REAL-TIME MONITORING OF CELL AND BIOINK PROPERTIES USING DIELECTRIC IMPEDANCE SPECTROSCOPY FOR EXTRUSION BIOPRINTING A Dissertation Presented to the Faculty of the Graduate School of Cornell University in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy by Alicia Adina Matavosian August 2025 © 2025 Alicia Adina Matavosian ALL RIGHTS RESERVED REAL-TIME MONITORING OF CELL AND BIOINK PROPERTIES USING DIELECTRIC IMPEDANCE SPECTROSCOPY FOR EXTRUSION BIOPRINTING Alicia Adina Matavosian, Ph.D. Cornell University 2025 Bioprinting, a form of additive manufacturing, uses cell-laden hydrogels to pro- duce highly customizable constructs for applications in tissue engineering and regenerative medicine. Despite its potential, several challenges continue to im- pede the clinical translation of bioprinted constructs, particularly regarding re- producibility and quality assurance. Current quality control practices rely on the evaluation of critical quality attributes (CQAs) at various stages of the bio- printing workflow. These assessments are predominantly performed offline us- ing destructive analytical techniques, which necessitate the fabrication of multi- ple constructs. However, reproducing identical constructs remains problematic due to the variability in bioink properties, which are time-sensitive and require extensive optimization, as well as the complex and variable nature of cell collec- tion and processing procedures. Additionally, inconsistencies between printed constructs further complicate quality control efforts. As an alternative, the inte- gration of biosensors directly into the bioprinting process presents a promising strategy for enabling real-time, non-destructive monitoring of CQAs, thereby enhancing process control and construct reproducibility. The application of real-time monitoring technologies in bioprinting, along with their potential uses and future directions, is explored in Chapter 1. This evaluation of the current state of the field highlighted a significant gap: the lim- ited development of real-time, label-free methods for monitoring cellular be- havior during the bioprinting process. Such monitoring is essential for ensur- ing the consistency and clinical viability of bioprinted constructs. In response to this need, Chapter 2 demonstrates the feasibility of using dielectric impedance spectroscopy as an in-line method for assessing cell concentration and viability directly from a syringe. To further assess the utility of this approach within a bioprinting context, Chapter 3 examines how variations in alginate bioink properties affect the real-time detection of cells. Finally, to expand upon the applications for this technology, Chapter 4 presents modifications to the elec- trode housing and geometry, which resulted in improved manufacturability and heightened sensitivity to a wide range of cell concentrations. BIOGRAPHICAL SKETCH Alicia was born and raised in Knoxville, TN. Dreaming of being a veterinarian for 17 years, she shocked family and friends when she shifted gears and enrolled in biomedical engineering at the University of Tennessee- Knoxville. While at UTK, Alicia joined two research laboratories: the Tissue Regeneration Labora- tory under Dr. Madhu Dhar and the Upper Limb Assist Laboratory under Dr. Dustin Crouch. Alicia researched the gene expression of osteoblasts undergoing osteogenic differentiation and the muscle response to biomimetic scaffolds for sciatic nerve regeneration. She graduated from UTK during the COVID-19 pan- demic in 2020 as summa cum laude with a Bachelors of Science in Biomedical Engineering and a minor in Materials Science and Engineering. Alicia then went on to pursue her Ph.D. at Cornell University under Dr. Lawrence Bonassar. Her graduate research was centered on monitoring biologic cells in real-time during bioprinting. During her time in the Bonassar Lab, Alicia collaborated with a team of RD scientists at West Pharmaceutical Services and completed a clinical research internship under Dr. Jason Spector at Weill Cor- nell Medicine. Over the course of her career, Alicia has published 3 first-author papers, collaborated on 5 co-authored papers, and presented at several confer- ences. Alicia also investigated the business potential of her research through the regional I-Corps program and will continue developing the technology in a startup. iii This dissertation is dedicated to the individuals that inspired me to be the change I wanted to see in the world. iv ACKNOWLEDGEMENTS First and foremost, I would like to express my deepest gratitude to my advisor, Dr. Lawrence Bonassar, for his unwavering support, mentorship, and optimism throughout the course of my doctoral studies. Larry’s generous investment of time, guidance, and scientific insight has been instrumental in my development as a researcher and professional. I would also like to thank the other members of my special committee, Dr. Michelle Delco and Dr. Robert Shepherd, for their guidance and expertise. The decision to pursue a doctorate was shaped by the influence and encour- agement of many individuals throughout my life. In high school, Mrs. Heike challenged me to confront difficult truths in the world around me. During my undergraduate studies, Dr. Elizabeth Barker played a pivotal role in cultivat- ing my passion for research and connected me to my future research advisors. Within my family, my aunt and uncle, Ginger Atwood and Robert Matavosian, recognized and nurtured my engineering interests from a young age, helping to lay the foundation for my career. I must also acknowledge my brother, Alex Matavosian, whose sibling rivalry motivated me to reach the self-imposed goal of becoming “doctors” together. I am sincerely thankful to my closest friends—Samantha Bratcher, Lela Bones, Aria Yslas, Liz-Audrey, Aiyana Fortin, and Rigoberto Vazquez—for their friendship and camaraderie throughout my time in Ithaca, especially during the challenges of the COVID-19 pandemic. Your support, humor, and presence have profoundly shaped my graduate school experience, and I am grateful for all I have learned from each of you. Of course, my time in the Bonassar Lab would not have been the same with- out the support of its members. The senior lab members I’d like to thank the v most are Leigh Slyker, Sean Kim, Jongkil Kim, and Serafina Lopez for their patience and their critical feedback. The individuals in (Bio)printing team (Leigh Slyker, Julia Bellamy, India Dykes, Alexandra Griffin, Smriti Sridharan, and Mariana Rodriguez) reminded me of the joy of scientific innovation from weekly meetings with hot chocolate and pastries to intense scientific discussions in the break room. In particular, I acknowledge my undergraduate mentees, Alexandra Griffin and Mariana Rodriguez, for playing a meaningful role in my growth as a mentor, researcher, and teacher. Your enthusiasm for learn- ing was truly infectious, and your thoughtful questions continually inspired me to deepen my own understanding. Other notable lab members were (in no particular order) Aiyana Fortin, Alikhan Fidai, Carlos Urrea De La Puerta, Car- oline Thompson, Spencer Witt, Salman Matan, Ashley Cardenas, and Hamilton Young whose presence taught me that lab culture is shaped by both small, ev- eryday moments and meaningful, shared experiences. Last, I also want to thank the Schaffer-Nishimura lab for fostering an environment that encourages both casual and scientific discussion through the donation of their commercial-grade espresso machine to the break room. To my family and friends beyond Ithaca, thank you for your unwavering support throughout this journey. Your companionship and daily conversations helped bridge the distance from Tennessee and kept me grounded during the most difficult moments of graduate school. Lastly, I am deeply grateful to my partner, Josh Wilson, for his unwavering encouragement, understanding, and support. Before we met, my days often felt like a blur of deadlines, anxiety, and self-doubt. Your presence brought balance and perspective, reminding me to value life beyond the lab and to make space for meaningful moments each day. vi TABLE OF CONTENTS Biographical Sketch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii Dedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi 1 REIMAGINING BIOPRINTERS: REAL-TIME MONITORING FOR QUALITY CONTROL OF BIOPRINTED CONSTRUCTS AND FU- TURE VISION 1 1.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.3 In-situ Monitoring of Bioink Properties During Printing . . . . . . 7 1.3.1 Assessment of Dynamic Properties and Phase Transitions 9 1.3.2 Bioink Mixing/Homogeneity . . . . . . . . . . . . . . . . . 10 1.3.3 Bioink pH . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 1.3.4 Bioink Temperature and Bioprinter Environment . . . . . 15 1.3.5 Bioink Viscosity . . . . . . . . . . . . . . . . . . . . . . . . . 17 1.3.6 Print Accuracy . . . . . . . . . . . . . . . . . . . . . . . . . 18 1.3.7 Dynamic Bioink Properties: Summary and Future Direc- tions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 1.4 In-situ Monitoring of Cells During Bioprinting . . . . . . . . . . . 24 1.4.1 Cell Concentration . . . . . . . . . . . . . . . . . . . . . . . 27 1.4.2 Cell Viability . . . . . . . . . . . . . . . . . . . . . . . . . . 31 1.4.3 Detecting Multiple Cell Types . . . . . . . . . . . . . . . . . 34 1.4.4 Cell Monitoring: Summary and Future Directions . . . . . 37 1.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 1.6 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 2 REAL-TIME ASSESSMENT OF CELL CONCENTRATION AND VIA- BILITY ONBOARD A SYRINGE USING DIELECTRIC IMPEDANCE SPECTROSCOPY FOR EXTRUSION BIOPRINTING 43 2.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 2.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 2.3 Materials and Methods . . . . . . . . . . . . . . . . . . . . . . . . . 47 2.3.1 Smart Syringe creation and use . . . . . . . . . . . . . . . . 47 2.3.2 Smart Syringe DIS Parameters . . . . . . . . . . . . . . . . 49 2.3.3 Primary chondrocyte cell isolation . . . . . . . . . . . . . . 52 2.3.4 Cell concentration studies . . . . . . . . . . . . . . . . . . . 52 2.3.5 Flow Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 2.3.6 Cell viability studies . . . . . . . . . . . . . . . . . . . . . . 54 2.3.7 LIVE/DEAD images and image analysis . . . . . . . . . . 55 vii 2.3.8 Data analysis and sensitivity . . . . . . . . . . . . . . . . . 56 2.3.9 Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 2.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 2.4.1 Cell Concentration . . . . . . . . . . . . . . . . . . . . . . . 57 2.4.2 Cell Viability . . . . . . . . . . . . . . . . . . . . . . . . . . 63 2.4.3 Cell Concentration, Viability, and |Zcells| . . . . . . . . . . . 66 2.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 2.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 2.7 Acknowledgment . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 2.8 Conflicts of Interest . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 2.9 Supporting Information . . . . . . . . . . . . . . . . . . . . . . . . 77 3 ALGINATE BIOINK PROPERTIES INFLUENCE REAL-TIME IMPEDANCE MONITORING OF CELLS DURING EXTRUSION BIOPRINTING 82 3.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82 3.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 3.3 Materials and Methods . . . . . . . . . . . . . . . . . . . . . . . . . 87 3.3.1 Smart Syringe creation and DIS parameters . . . . . . . . . 87 3.3.2 Primary chondrocyte preparation . . . . . . . . . . . . . . 88 3.3.3 Alginate concentration studies . . . . . . . . . . . . . . . . 89 3.3.4 Sample pH studies . . . . . . . . . . . . . . . . . . . . . . . 89 3.3.5 Alginate crosslinking and bioprinting . . . . . . . . . . . . 90 3.3.6 Data analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 91 3.3.7 Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 3.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 3.4.1 Alginate concentration . . . . . . . . . . . . . . . . . . . . . 93 3.4.2 Sample pH . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96 3.4.3 Alginate crosslinking and bioprinting . . . . . . . . . . . . 99 3.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 3.7 Acknowledgment . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 3.8 Conflicts of Interest . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 3.9 Supporting Information . . . . . . . . . . . . . . . . . . . . . . . . 108 4 COMBINING ELECTRODES TYPES ENABLES HIGHER SENSITIV- ITY TOWARDS CELL CONCENTRATION 110 4.1 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 4.2 Materials and Methods . . . . . . . . . . . . . . . . . . . . . . . . . 112 4.2.1 Electrode housing and electrode fabrication . . . . . . . . . 112 4.2.2 Comparison of electrodes . . . . . . . . . . . . . . . . . . . 113 4.2.3 Combining multiple electrode types . . . . . . . . . . . . . 114 4.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 4.3.1 Comparison of electrodes for monitoring cell concentration 115 viii 4.3.2 Multi-electrode device . . . . . . . . . . . . . . . . . . . . . 116 5 CONCLUSIONS AND FUTURE DIRECTIONS 119 Bibliography 123 ix x LIST OF FIGURES 1.1 Reimagined bioprinter capable of both creating constructs and evaluating CQAs during printing. (a-d) Key stages for monitoring CQAs are 1.) bioinks and cells prior to printing, 2.) sample mixing, 3.) sample deposition, and 4.) construct printing. e.) To evaluate critical qyality attributes during these stages, various sensing technologies may be implemented for automatic, real-time data collection of dynamic printing variables, print accuracy, and cell health. This data can then be used to control a specified output variable through feedback during printing. f.) The collected data both optimizes the print and ensures that relevant CQAs are within target criteria. ...... 6 1.2 a.) This system represents a “smart” bioprinter, capable of monitoring the sample in real-time with the incorporation of in-line monitoring systems for bioink mixing, temperature, pH, and viscosity. b.) Various in-line methods that are compatible with real-time measurements of dynamic bioink properties. c.) Monitoring print accuracy through defect or error identification, prediction, and correction combines automated analysis with real-time imaging and control of printer parameters ........................................................................................................................................ 9 1.3 Adapted from reference (1) with permission. a.) Experimental setup with custom-printed syringe holder and in-line load cell. A test-frame produced a pre-set load on the syringe, mimicking a syringe pump. b.) Mechanical noise described the variation in force experienced by cell-laden alginate samples that were created with mixing cycles ranging from 8 to 200 ............................................................................ 12 1.4 Adapted from reference (1) with permission. a.) An ovine meniscus was imaged using CT (b) and bioprinted in alginate. c.) Constructs that failed from syringe tip clogging or showed poor shape fidelity due to bioink heterogeneity. d.) Homogeneous prints without defects .......................................................................... 19 1.5 Adapted from reference (2) with permission. a.) Overview of the closed-loop error detection and correction system using convolutional neural network-based ML to identify extrusion state as “good”, “over”, or “under” and take the appropriate action. b.) Prints of infill and line structures undergoing correction using red-stained high viscosity alginate (100 mg/mL). Poisoned G-code were areas where the extrusion rate was intentionally increased or decreased to create a defect ......................................... 21 1.6 Adapted from reference (3). a.) Experimental set-up with high-speed camera directed at the print bed and a temperature difference between the printhead and print bed. b.) (Top) Model of the 3-layer printed construct. (Bottom) Segmented and processed IR image from the thermal camera revealed only the most recently printed layer ............................................................................................................................. 22 1.7 a.) Representation of a remodeled bioprinter capable of monitoring cell properties (concentration, viability, and type) during bioprinting. b.) Recent advances xi in non-destructive methods for monitoring cell properties during printing are image- or spectroscopy-based. c-e.) Various in-line methods that are compatible with cell concentration, viability, and type ................................................................................ 26 1.8 Modified from reference (4). a.) Fluorescence monitoring system with a microscope directed at the print bed. b.) Heatmap showing the position of human embryonic kidney 293 cells printed at 2×106 cells/mL in Cellink Bioink RGD hydrogel. From top to bottom, layers 1, 2, 3 and 4 are displayed. c.) Plots of pixel intensity for the central region of each layer using samples at (top) 2×106 cells/mL and (bottom) 1×106 cells/mL .............................................................................................. 30 1.9 Modified from (5). a.) A custom in-line impedance sensor for DS was added between the syringe body and syringe tip. b.) The impedance |Z| of cell-laden samples was subtracted from acellular PBS to quantify |Zcells|. |Zcells| increased with cell concentration from 1 - 125×106 cells/mL at 25 kHz. Line represents the average of 5 independent trials, and bars show standard deviation. c.) |Zcells| changed accordingly as the cell concentration of the sample was altered under flow rate at 1 mL/min. Dark line showed the average of 3 independent trials ........................................................... 30 1.10 Adapted from reference (6). a.) Experimental method using light-sheet fluorescent microscopy to monitor cell-laden bioink. b.) Fluorescent cells were tracked over time in GelMA solution, pluronic, and gelled GelMA. The extracted cell tracks were color coded with mean velocity. c.) Plots of the tracked cell velocities and fitting to estimate the fluid velocity profile as a function of distance from the center of the capillary for GelMA solution, pluronic, and gelled GelMA ........................................ 33 1.11 Adapted from reference (7). a.) Multifunctional fibers were integrated into a syringe tip and sealed with epoxy for in-line impedance monitoring using dielectric impedance spectroscopy. b.) LIVE/DEAD assay of high or low viability samples of PC-12 cells at 5×105 cells/mL in alginate with nanofibrillated cellulose. c.) Impedance magnitude |Z| for high and low viability samples from 100 – 10,000 Hz. D.) Phase angle for both samples ................................................................................................. 34 1.12 Adapted from reference (8). a.) Schematic of acoustic printing platform with surface-enhanced Raman spectroscopy occurring immediately following printing for mouse red blood cells infused with gold nanorods (GNRs) and bacteria. b.) Mean spectra for different combinations of red blood cells (RBC), E. coli and S. epidermidis with 100 measurements each, taken from single droplets and mixed with GNRs. Blue, green, and red vertical lines represent wavenumbers displaying the biological peaks for each cell type .......................................................................................................... 36 2.1 DIS measured cellular response to electric currents. a.) As a cell suspension containing healthy and dead cells passed through the alternating current, a frequency- dependent response was measured as impedance |Z| and phase angle θ. In this diagram, bio-ink is flowing into the page. b.) A custom bioprinter loaded with a syringe and smart syringe. The smart syringe connected the syringe barrel to the syringe tip and xii was fitted with two antiparallel electrodes that generated an alternating current ........ 48 2.2 Primary chondrocyte concentration was controlled from 1×106 – 125×106 cells/mL and assessed from 1 – 25,000 kHz with n = 3 per concentration. Shaded areas represent sample standard deviation. a.) |Zcells| measurements of primary chondrocytes after subtracting the background PBS signal. Dark lines indicate sample mean. b.) θcells of primary chondrocytes after subtracting PBS background. c.) 15 measurements from each cell concentration collected from five independent studies were used to show the relationship between |Zcells| and cell concentration at 25 kHz. Each point represents a single replicate. d.) These 15 measurements were also used to determine the relationship between θcells and cell concentration at 25 kHz. e.) Device sensitivity described the degree to which |Zcells| changed with concentration across the frequency sweep and was generated by calculating the linear regression from Fig. 2c across all frequencies. Dark line represents sample mean. f.) Device sensitivity to changes in θcells was generated by calculating the linear regression from Fig. 2d across all frequencies .................................................................................................................... 59 2.3 Measuring cell concentration of primary chondrocytes at specified flow rate(s). Displayed data was from 25 kHz. a.) Diagram of sample flow using 3-way stopcock, smart syringe, and syringe pump. b.) Primary chondrocyte concentration was adjusted from 1×106 cells/mL to 100×106 cells/mL to 50×106 cells/mL with 120 s (2 min) intervals at a flow rate of 1 mL/min. Concentration transitions occurred over 30 s and 15 s. Dark line indicates the average of 3 independent trials. c.) Diagram of sample at 25×106 cells/mL with extrusion at a set flow rate for 240 s. d.) |Z| and θ were measured at 25 kHz across three flow rates relevant to bioprinting for 240 s (4 min) with n = 4 per flow rate and with cell concentration at 25×106 cells/mL. Data was normalized to t = 0 s for each flow rate, which occurred immediately after sample flow was initiated ................................................................................................................. 62 2.4 Primary chondrocyte samples suspended in PBS were measured with high (~94%), medium (~50%) and low (0%) viabilities with cell concentration remaining constant at 25×106 cells/mL. Frequency ranged from 0.1 kHz – 25,000 kHz, and n = 4 replicates per viability. Dark lines represent sample means, and shaded areas show standard deviation. a.) |Zcells| after subtracting the background PBS signal. Boxed region is expanded from 0.4 – 1,000 kHz. b.) θcells after subtracting the background PBS signal. c.) Linear correlation between |Zcells| and viability at 1 kHz for 3 independent experiments using 3 different smart syringes. The medium viability was not identical for all independent studies and was thus shown with multiple points. d.) Linear correlation between θcells and viability across 3 independent experiments using different devices. e.) Sensitivity to changes in |Zcells| as cell viability increases. f.) Device sensitivity to changes in θcells as cell viability increases ................................. 65 2.5 3-D representations of cell concentration, cell viability, and |Zcells| using 40 - 54 samples of primary chondrocytes suspended in PBS at controlled cell concentrations xiii and cell viabilities. a.) Best-fit plane with equation at 10 kHz with each point representing a sample average of 3 - 4 replicates and n = 54 samples. In the equation, x represents cell concentration, and y is viability. b.) Planes of best fit at (0.1, 10, and 10,000) kHz. The number of samples for each frequency are as follows: 0.1 kHz was n = 40, 10 kHz was n = 54, and 10,000 kHz was n = 54. c-e.) Best fit plane coefficients A, B, and C for frequencies (0.1, 1, 10, 100, 1,000, and 10,000) kHz .......................... 67 2.1S Assembly of a smart syringe. a.) Materials needed include breadboard wires, machined female luer lock, male luer locks, gold pin electrodes, wire connections, heat shrink, and epoxy (not pictured). b.) Strip the wire from one end of the breadboard wires then attach and strip a gold pin electrode. c.) Remove the plastic casing from the other end of the breadboard wire and add a wire connection. d/e.) Add heat shrink to both ends. f.) Insert the gold pin electrodes into the female luer lock and twist the two luer locks together. Add epoxy to the insertion point for the electrodes. g.) Top-view of assembled smart syringe ..................................................................... 78 2.2S Raw |Z| and θ measurements for primary chondrocytes suspended in PBS with concentration at 1×106, 10×106, 25×106, 50×106, 75×106, 100×106, and 125×106 cells/mL with n = 3 per concentration. PBS without cells was collected to measure background/media signal and was shown in blue. Samples replicates were randomly measured. Shaded regions represent the standard deviation with the dark lines showing the sample average. (a) (Left) Frequency-dependent measurements of |Z| from 1 kHz to 25,000 kHz with replicates and (right) zoomed-in view of 10 – 10,000 kHz. (b) (Left) Frequency-dependent measurements of θ from 1 - 25,000 kHz and (right) zoomed-in view of 10 kHz – 10,000 kHz .................................................................... 79 2.3S Primary chondrocytes suspended in PBS at 100×106 cells/mL were added to a syringe and allowed to rest without extrusion for 600 s (10 minutes) in a syringe pump. |Z| displayed at 25 kHz with n = 3 trials for the cell-based sample. Data was normalized to t = 0 s for each replicate. Dark line represents sample average with shaded area showing standard deviation ..................................................................... 80 2.4S Samples with cell concentrations at 1×106, 10×106, 25×106 and 50×106 cells/mL with n = 3 per concentration were assessed at 0, 30, 60, and 90 s at 2 mL/min for a.) |Z| normalized to t = 0 s and b.) θ normalized to t = 0 s ................................... 80 2.5S Raw |Z| and θ measurements of primary chondrocyte samples suspended in PBS with high (~95%), medium (~50%) and low (0%) viabilities. Cell concentration remained at 25×106 cells/mL, frequency varied from 0.1 – 25,000 kHz, and n = 4 - 5 replicates per viability. PBS without cells was measured before and each sample for comparison. Samples replicates were randomly measured. Shaded areas represent sample standard deviation. a.) |Z| for samples of varying viability across the frequency sweep. Boxed region is investigated in (b-d). b-d.) Zoomed-in view of 10 – 1,000 kHz with each viability and its corresponding PBS measurements displayed individually. e.) θ across the full frequency sweep. Boxed region is showcased in (f-h). f-h.) xiv Zoomed-in view of 10 kHz – 1,000 kHz for each viability and PBS pair.................... 81 3.1 Bioink properties influence |Z| measurements by altering the movement of ions in solution. In alginate bioink, changes to concentration, pH, or crosslinking lead to effects in counter-ion concentration and viscosity that have opposing influences on |Z|. Counter-ions (Na+) are present in solution due to the poly-anionic (-) nature of alginate. a.) As the concentration of alginate bioink increases, the number of alginate chains rises, increasing viscosity and counter-ion concentration. These changes in viscosity and ion balance have opposing effects on |Z|. b.) The anionic polymeric chains in alginate bioink become protonated as the sample becomes acidic, reducing counter-ion concentration. In addition, viscosity is decreased through the reduction of repulsive forces between negatively charged chains. c.) With the addition of CaCl2, alginate bioink is crosslinked into an egg-box structure, increasing viscosity and decreasing Na+ ions. This shift in conformational structure is anticipated to increase |Z| .................................................................................................................................. 86 3.2 Acellular and cellular (25×106 cells/mL) alginate bioinks with concentrations at 0.5, 1 and 3 w/v% were monitored using |Z| with frequency sweeps. Dark lines represented sample average and shaded regions showed standard deviation. a.) |Z| of acellular alginate was measured from 1 - 25,000 kHz with n = 4 per concentration. Boxed section depicted the region from 10 – 1,000 kHz with a dotted line representing 50 kHz. b.) At 50 kHz, |Z| was compared between acellular and cellular samples with n = 4 per alginate concentration. c.) To quantify the degree to which cells could be measured in each bioink concentration, |Zcells| was calculated from 1 - 25,000 kHz with n = 4. Brackets visualize regions with stable |Zcells|. d.) |Zcells| across multiple independently conducted experiments were amassed at 50 kHz with n = 11-16 per concentration. e,f.) Sensitivity to alginate concentration (|Z| Alg) and to cell detection (|Zcells| Alg) were determined as a rate of change and plotted as a function of frequency with n = 3. ..................................................................................................................... 95 3.3 |Z| was monitored across pH ranges in PBS (from 5.8 - 7.7) and alginate bio- ink (from 5.5 - 7.7) for acellular and cellular (at 25×106 cells/mL) samples. Dark lines represented sample average and shaded region showed standard deviation. a.) |Z| was monitored for acellular PBS samples using pH at 6.5, 7.0, and 7.4 from 1 - 25,000 kHz with n = 3 - 4 per pH. Boxed section depicts the region from 10 – 1,000 kHz with a dotted line representing 50 kHz. b.) Comparison of the |Z| values at 50 kHz for cellular and acellular PBS at different pHs with n = 3 - 4 per sample. c.) |Zcells| was calculated for PBS samples from 1 - 25,000 kHz with pH ranging from 6.5 - 7.4 and n = 4. Brackets indicated frequencies with greater signal stability and reduced variability. d.) Similar studies were performed for alginate samples. |Z| was measured for acellular alginate samples from 1 - 25,000 kHz with pH at 5.5, 6.7, and 7.7 and n = 4. Boxed section depicts the region from 10 – 1,000 kHz with a dotted line representing 50 kHz. e.) Comparison of the |Z| values for cellular and acellular xv alginate bio-ink at 50 kHz with n = 4. f.) |Zcells| was calculated for alginate bioink from 1 - 25,000 kHz with pH ranging from 5.5 - 7.7 and n = 4. Brackets visually depicted frequency regions with greater stability. g.) Comparison of |Zcells| for alginate bioink and PBS across multiple independent studies at 50 kHz with each point displaying the average of n = 3 - 8 measurements with standard deviation bars. Models were fit to the dataset using nonlinear 2nd order polynomials. The pH region for cell media is highlighted from 7.2 - 7.4. h.) Sensitivity of |Z| to sample pH (PBS or alginate) for acellular samples with n = 3. i.) Sensitivity of |Zcells| to sample pH (PBS or alginate) at a pH of 7.3, referencing the pH of typical cell media. ................................................ 98 3.4 |Z| of alginate bio-inks with CaCl2 crosslinker was measured immediately after mixing without flow (0 mL/min). Dark lines represented sample average, and the shaded region showed standard deviation. a.) To monitor changes in |Z| as a result of crosslinker, alginate with and without CaCl2 were compared across a frequency sweep from 1 - 25,000 kHz with PBS acting as a control group (n = 4). The boxed section depicts the region from 10 - 1,000 kHz with a dotted line representing 50 kHz. b.) Comparison of the |Z| values at 50 kHz with n = 4 per sample ................................ 100 3.5 Cell presence in alginate with CaCl2 was monitored in real-time using |Z| by adding or removing cells during extrusion from a syringe pump and bioprinter. Alginate bioink was at 1 w/v% with pH 7.3. a.) Experimental set-up for extruding acellular and cellular bioink using a syringe pump with the in-line smart syringe impedance sensor. b.) |Z| was measured as acellular and cellular bio-ink samples were transitioned in real-time with n = 5-6 per trial. Flow rate was set to 0.5 mL/min, and frequency was 1 kHz. PBS served as an acellular control. (b.i) Acellular alginate with CaCl2 was monitored using |Z| for 70 seconds, then was transitioned for cellular bioink (25×106 cells/mL or 100×106 cells/mL). Dotted lines represent the timed transition from acellular to cellular bioink during which samples mixed. (b.ii) The process was reversed with cellular samples (25×106 cells/mL or 100×106 cells/mL) transitioning to acellular bioink over 210 seconds. Dotted lines represent the mixing region. c.) Experimental set-up for printing acellular and cellular bioink from a single syringe with the mounted smart syringe. d.) Printed construct visually depicting the transition from acellular to cellular (25×106 cells/mL) bioink at 1 kHz. A heatmap depicting the shift in normalized |Z| was spatially mapped to the printed construct . 101 3.1S LIVE/DEAD staining of primary chondrocytes at a density of 25×106 cells/mL suspended in 0.5, 1, or 3 w/v% alginate. .................................................................... 108 3.2S LIVE/DEAD staining of primary chondrocytes at a density of 25×106 cells/mL suspended in (top row) 1 wt/v% alginate at pHs ranging from 6.4 - 7.6 or (bottom row) PBS at pHs ranging from 5.8 – 7.6 ............................................................................. 108 3.3S Sensitivity of |Zcells| to changes in pH in PBS and alginate bio-ink were determined from 1 – 25,000 kHz. a-d.) the rate of change for |Zcells| in acidic pHs from 5.8 – 6.7 was positive for primary chondrocytes in PBS and negative for cells in xvi alginate. e.) at a neutral pH of 7, cells suspended in either PBS or alginate were typically positive on average but displayed lower rate of change in comparison to acidic or basic pHs. f.) the rate of change for |Zcells| in a slightly basic solution at 7.6 pH was positive for primary chondrocytes in alginate and negative for cells in PBS, which is reversed from acidic solutions. ................................................................... 109 4.1 (Left) Zoomed top view of gold pin-shaped electrode nestled in the Smart Syringe housing. (Right) Distance between electrodes varied due to custom machining of the housing ............................................................................................................ 111 4.2 Manufacturing of the curved parallel plate electrodes. a.) CAD model of the electrode housing split into two parts. b.) 3D printed housing with water-soluble support material. c.) Platinum sputter-coated Kapton tape measured prior to shaping. d.) Kapton tape was attached copy wire within the inner chamber of the housing unit using silver epoxy. e.) Two halves of the housing unit were combined. f.) Male and female luer locks were added to either end of the housing to support use with a syringe .................................................................................................................................... 113 4.3 Multi-electrode prototype, combining pin-shaped electrodes with curved parallel plate electrodes a.) into a single chamber or b.) in series .............................. 114 4.4 |Zcells| for pin electrodes and curved parallel plate electrodes across multiple cell concentrations displayed at 10 kHz. Shapes denote sample averages and bars represent the standard deviation. ............................................................................... 116 4.5 Fractional change of |Z| across multiple cell concentrations from 1 – 25,000 kHz. Black box indicates regions of high sensitivity to cell concentration from approximately 1 – 80 kHz ......................................................................................... 117 4.6 |Zcells| was calculated for measurements collected using the multi-electrode device. Measurements were plotted at 1, 10, and 100 kHz and fitted to a piecewise, weighted linear regression. a.) |Zcells| from the curved parallel plate electrodes increased with cell concentration, however, pin electrodes displayed a stagnation in |Zcells|. b.) At 10 kHz, the pin electrodes showed increasing |Zcells| with increasing cell concentration from 28.5 - 112 × 106 cells/mL. Curved parallel plate electrodes also showed increasing |Zcells|, particularly from 0.6 - 1.1 × 106 cells/mL. c.) Crosstalk between both electrodes greatly distorted the collected data, resulting in high noise. .................................................................................................................................... 118 xvii LIST OF TABLES 1.1 Key features and limitations of real-time monitoring methods for bioink mixing .......................................................................................................................... 13 1.2 Comparison of methods used for real-time monitoring of cell properties. Applications of these methods, their minimum requirements, and limitations are listed ...................................................................................................................................... 38 2.S1 Calculations for volume (v) passing through the electrodes during each frequency sweep with time (t) at various volumetric flow rates (Q) assessed in this study ............................................................................................................................. 77 2.S2 Calculations of cell residence time in the smart syringe at flow rates from 0.5 - 4 mL/min and with electrode distance at 0.05 or 0.1 cm. Cell residence time (tresidence) represents the approximate time that a cell spends in the path of the electrodes within the smart syringe .......................................................................................................... 77 2.S3 Fitting parameters for best-fit planes using Equation 3 by frequency. Color and asterisk indicate significance using R Critical Value tables with 0.05 significance for two-tailed tests ............................................................................................................. 80 xviii PREFACE The work presented in this dissertation first discusses the challenges associated with the clinical translation of bioprinted constructs with a focus on quality control and reproducibility. After establishing the complexities of measuring construct quality in cell-laden materials, the integration of sensors into the bioprinting process is proposed to monitor cell-laden bioinks in real-time. This work then details the application of in-line impedance sensors (referred to as the Smart Syringe (9)) onto a syringe for real-time measurement cell concentration and viability. This system is shown to be sensitive to changes in acellular and cellular alginate formulations and is compatible with bioprinting. To further refine this technology, changes in the sensor material and shape are shown to improve sensitivity to cell concentration. Overall, the research in this dissertation addresses a critical need for real-time monitoring of cells during bioprinting. Bioprinting uses printable hydrogels infused with cells to produce personalized constructs for tissue engineering and regenerative medicine. Although bioprinted constructs are produced for a variety of tissues, only 3 bioprinted products have entered clinical trials (10–13). Clinical translation for these constructs is limited by the extensive and costly regulatory requirements associated with quality control testing that occur at every step of production (14). Quality control currently relies upon the destructive, off-line testing of constructs post-printing. In addition to raising production costs due to manual and time-consuming handling, these tests render the construct unsuitable for implantation, necessitating the production of multiple identical constructs. Chapter 1 of this dissertation proposes an amelioration for this issue through the real-time assessment of construct quality during bioprinting (15). Automated, real-time measurement of bioink and cells expedites quality control xix testing, resulting in reduced production costs. In addition, real-time monitoring boosts the print quality of constructs by enabling the adjustment of print parameters, bioink formulation, or cell properties when combined with feedback. The collection and processing of cells greatly contributes to the overall cost of the bioprinted construct (16,17). Despite this cost, limited research has been dedicated to the real-time assessment of cell properties. Three methods show compatibility with monitoring cell properties in bioink. Optical imaging methods measure samples with high spatial resolution but are limited to low cell concentrations (<10×106 cells/mL) in narrow channels due to light scattering caused by viscous bioinks as well as cell clustering (18–21). Ultrasound bypasses the limitations of light scattering by using sound waves to measure cell concentrations in bioprinted constructs up to 100×106 cells/mL (22); however, this method is limited in its spatial resolution and depth penetration (23). In contrast to optical imaging and ultrasound, dielectric spectroscopy (DS or DIS) provides a rapid measurement of bulk cell properties. Upon exposure to an alternating current, the ions within the cell cytoplasm and in the surrounding substrate polarize, leading to the measurement of impedance and phase angle (24,25). Chapter 2 shows the real-time measurement of cell concentration and viability using the Smart Syringe in saline suspensions up to 125×106 cells/mL (26). Although measuring cell properties is possible using DIS, the influence of bioink on impedance is unknown. Notably, bioink formulations are adjusted to control the properties of the printed construct (27,28). Changes in bioink concentration, pH, or crosslinking may result in alterations to ion movement and interfere with impedance measurements of cell properties. This knowledge gap is addressed in Chapter 3, in which the effect of alginate concentration, pH, and crosslinking on impedance is determined for acellular and cellular samples (29). Following this discovery, acellular and cellular samples were monitored in real-time during bioprinting. xx Despite the capabilities of the Smart Syringe for monitoring cell properties, cell detection is limited to cell concentrations above 25×106 cells/mL. To expand upon the detectable range of cell concentrations, Chapter 4 details the optimization of the electrode material and shape. Thus, the research in this dissertation provides a method for monitoring cell- laden bioinks onboard a syringe. The specific aims for this dissertation are as follows: Specific Aim 1: Investigate the relationship between impedance and phase angle for cell concentration and viability onboard a syringe (Chapter 2). Specific Aim 2: Understand the effects of alginate concentration, pH, and crosslinking on the impedance of acellular and cellular samples. Demonstrate real- time cell detection during bioprinting (Chapter 3). Specific Aim 3: Improve device sensitivity to cell concentration through the optimization of electrode material and form (Chapter 4). 1 CHAPTER 1 REIMAGINING BIOPRINTERS: REAL-TIME MONITORING FOR QUALITY CONTROL OF BIOPRINTED CONSTRUCTS AND FUTURE VISION* 1.1. Abstract The use of bioprinters as depositional tools for bioinks and cells has expanded greatly over the past two decades. Bioprinting combines hydrogels with cells to produce customized constructs for personalized medicine. However, several challenges hinder the clinical use of these constructs. Quality control metrics for bioprinting involve the assessment of critical quality attributes (CQAs) at every stage of production. Currently, bioprinted constructs are manually assessed using destructive methods that occur post-production, requiring the creation of multiple products per patient. Reproducing printed constructs is difficult due to time-sensitive bioink properties that require lengthy optimization processes to print with accuracy. In addition, the collection, processing, and testing of cell bioactivity for each printed construct greatly increases production costs. To address these challenges, non- destructive, real-time monitoring can be integrated into the bioprinting process. The goal of this review paper is to reimagine the function of a bioprinter from a simple tool of production to one capable of evaluating constructs in real-time. This review features recent advances in the field for real-time monitoring with a focus on time-sensitive bioink properties, print accuracy, and cell health. Automated * Matavosian AA and Bonassar LJ. Reimagining Bioprinters: Real-Time Monitoring for Quality control of Bioprinted Constructs and Future Vision. Biofabrication. (In review) 2 assessment and quantification of time-sensitive bioink qualities such as mixing, pH, temperature, and viscosity will enhance construct quality by enabling the rapid optimization of printing parameters. Meanwhile, real-time monitoring of cell health through concentration, viability, and type serves as an indicator for bioactivity. Construct accuracy and reproducibility are also improved through the identification, prediction, and correction of defects during printing. Incorporating real-time monitoring into the bioprinting process using closed-loop feedback would improve the reproducibility, quality, and translation of constructs into the clinic. 1.2. Introduction Interest in personalized medicine has skyrocketed during the past decade with 286 FDA-approved personalized medicines available in the U.S. in 2020 compared to 36 in 2010 (30). As the demand for personalized medicine increases, bioprinting has taken the limelight as a fabrication method with the potential to produce customized, tissue-engineered implants with applications in organ transplantation, wound healing, tissue regeneration, and tissue reconstruction. Regardless of the type of bioprinting (extrusion, inkjet, or photopolymerization), a construct is printed using a phase transition from liquid to gel state by mixing bioink, living cells, and additives. Constructs are personalized by combining the 3-D fabrication of bioprinting with patient cells and models generated from computed tomography (CT) or magnetic resonance imaging (MRI). The scale of personalized implants that are produced using bioprinting vary from vascular networks (μm) to corneas (mm scale) and human hearts (cm scale) (31–35). 3 Although many types of bioprinted constructs are produced for research purposes, there currently are none approved for clinical use. In fact, as of writing this article, only 3 bioprinted products have entered clinical trials (10–13). A major hurdle for progressing through human clinical trials is the extensive regulatory requirements associated with quality control testing that occur at every step of production. Quality control is crucial for ensuring patient safety and is used to evaluate the reproducibility of the constructs, properties of the advanced materials used as bioinks, and documentation of cell processing. Because bioprinted constructs contain cells for a specific application, these products are designated as “combination” materials in the United States with the highest regulatory standards for quality control (36,37). These requirements include process controls for cell collection, transportation, and storage, as well as critical quality attributes (CQAs) for construct properties such as sterility, bioink synthesis, bioink properties, print accuracy, cell health, and cell bioactivity over time. For example, cells are obtained from a patient or donor, transported, and determined to be of sufficient quantity and viability. Simultaneously, bioink materials are assessed for safety, biocompatibility, and long-term storage (14). Next, patient models are created, converted to G-code, and manually assessed by manufacturers and surgeons while also taking extra measures to ensure that patient information is securely stored (14). The combination of assessing bioprinted products at every step of production and the manual testing methods used to complete regulatory requirements greatly increase the time and resources required for clinical trials. Assessment of CQAs is particularly costly for quality control of bioprinted constructs due to manually operated destructive testing methods. To quantify and 4 document CQAs, a typical medical product undergoes batch testing with a small number of products tested compared to products produced (38–40); however, bioprinted constructs are crafted for individual patients using a unique model as well as autologous or allogeneic cells and thus have a lot size of 1, lacking economy of scale (41,42). In fact, traditional methods for assessing CQAs involve destructive testing of samples pre- printing using rheological assessment of bioink and viability testing for cells as well as post-fabrication through histological sectioning and staining (43–45), fluorescent imaging (45–47), biochemical assays (44,47,48), and biomechanical testing (49–51). The destructive nature of traditional testing methods renders a construct unsuitable for implantation and thus requires that multiple constructs be produced per patient to fulfill regulatory requirements. In addition, the characterization of bioink and cell properties as well as handling of patient data involve manual, labor-intensive assessments by trained personnel that greatly expand the unit operations required per patient. The combination of lot sizing and manual assessment results in costly quality control for bioprinted constructs. To reduce the burden of quality control testing, we urge a shift away from measuring CQAs through manual, destructive testing methods. Instead, we encourage a reimagining of the current bioprinter from a tool of fabrication to a device that both produces and evaluates constructs (Fig. 1.1). This future bioprinter features an automated sensing platform using real-time monitoring to replace labor-intensive testing methods. Using sensors, electrodes, optics, or image-based methods CQAs for both bioink and cellular properties are collected for regulatory purposes, saving time and resources. In addition, real-time data collection enables closed-loop processes in 5 which crucial parameters are controlled through feedback and machine learning (ML). In contrast to our imagined printer, current bioprinters use open-loop processes, where a set of commands are executed through G-code to print a model. Notably, open-loop processes cannot be adjusted to ensure that target CQAs are met and are unable to adapt to changes in the sample or environment, resulting in costly print failures and manual optimization of print settings. By collecting CQAs during printing, these future bioprinters will produce higher quality constructs through real-time adjustments in printing metrics and will require fewer constructs per patient. Key stages of the printing process for monitoring CQAs in real-time include pre- processing the bioink and cells (Fig. 1.1a), mixing the bioink and cell solutions (Fig. 1.1b), extruding the sample using a deposition tool (Fig. 1.1c), and printing the construct (Fig. 1.1d). Sensing technology may be integrated into any stage to enable real-time monitoring of CQAs. Bioinks display dynamic process variables and properties during bioprinting, leading to the use of in-line sensors or electrodes, images, and videos to both acquire CQAs as well as monitor print accuracy (Fig. 1.1e). Meanwhile, cell-based monitoring methods emphasize the isolation of cell properties from the surrounding solution to measure indications of cell health such as concentration, viability, and type (Fig. 1.1e). Using this real-time data for dynamic bioink properties or cell monitoring can both inform the printing process and automatically optimize the construct using control systems (Fig. 1.1e). Lastly, real-time methods ensure that relevant CQAs are within target criteria, and stored data are submitted to regulatory agents post-printing (Fig. 1.1f). In this review, real-time monitoring methods for quality control are discussed in the 6 context of bioprinting. Relevant methods are proposed for monitoring CQAs, and emphasis is placed on dynamic process variables, bioink properties, and cell-properties with further analysis on feedback mechanisms that convert bioprinters into closed-loop systems. Future directions for each topic are proposed at the conclusion of each section. Figure 1.1: Reimagined bioprinter capable of both creating constructs and evaluating CQAs during printing. (a-d) Key stages for monitoring CQAs are 1.) bioinks and cells prior to printing, 2.) sample mixing, 3.) sample deposition, and 4.) construct printing. 7 e.) To evaluate critical qyality attributes during these stages, various sensing technologies may be implemented for automatic, real-time data collection of dynamic printing variables, print accuracy, and cell health. This data can then be used to control a specified output variable through feedback during printing. f.) The collected data both optimizes the print and ensures that relevant CQAs are within target criteria. 1.3. In-situ Monitoring of Bioink Properties During Printing Proper selection of bioink material is crucial to every aspect of the final product from shape fidelity to cell bioactivity. However, optimization of bioink material is difficult due to the dynamic printing system that results in changes in physical and rheological properties over time. For instance, hydrogels exhibit viscoelastic and shear-thinning behavior during extrusion (52–55), which influences: printing parameters through rheology (56,57); shape fidelity through crosslinking (58– 60); and cell viability through shear stresses (61–63). Bioink properties are currently assessed before and after printing, however, current testing methods fail to capture the dynamic changes that occur in the printing environment. For instance, rheological testing could involve exposing a sample of bioink to stress or strain while confined between two parallel plates. While this measurement is useful for characterizing material behavior and comparing batch-by-batch variations, it fails to recapitulate the forces experienced by bioinks during printing, which are influenced by extruder geometry and deposition rate (64). Because bioink properties dynamically change during printing, in-situ monitoring is crucial for creating reproducible and high-quality constructs (1). 8 Processing variables (mixing, temperature, and pH) as well as bioink properties (viscosity) affect print quality and are amendable to real-time monitoring. These four metrics can be monitored through in-line sensors or electrodes embedded into the printing process as well as image-based methods during printing (Fig. 1.2a). Several methods for monitoring bioink mixing, temperature, pH, and viscosity are listed in Fig. 1.2b. Notably, suboptimal bioink conditions result in print errors that compound with each layer, rapidly reducing the quality of the construct. Traditionally, constructs are manually assessed for print accuracy compared to the model (65), however, this process is time-consuming and does not recover resources expended for failed prints. Automatic identification of print errors in real-time provides the opportunity to both tune printer settings as well as correct these errors during the printing process. As a result, data collected from real-time methods can be used to identify errors during printing, and models can be created to predict and correct these errors (Fig. 1.2c). The following sub-sections discuss available methods for monitoring dynamic bioink properties as well as real-time prediction, identification, and correction of printing errors. 9 Figure 1.2: a.) This system represents a “smart” bioprinter, capable of monitoring the sample in real-time with the incorporation of in-line monitoring systems for bioink mixing, temperature, pH, and viscosity. b.) Various in-line methods that are compatible with real-time measurements of dynamic bioink properties. c.) Monitoring print accuracy through defect or error identification, prediction, and correction combines automated analysis with real-time imaging and control of printer parameters. 1.3.1. Assessment of Dynamic Properties and Phase Transitions Dynamic bioink properties have the potential to change over time during bioprinting and can influence the print accuracy of the construct. For instance, hydrogel crosslinking is tuned by bioink homogeneity, temperature, and pH (1,66–68). Crosslinking occurs upon adding external agents such as salts, chemicals, or light to reorganize the polymeric chains of the hydrogel into an interconnected network. This 10 process enables the layering of bioink into a 3-D construct, encapsulates cells, and allows for the diffusion of nutrients and water molecules (69). Due to the direct correlation between bioink crosslinking and print accuracy, the dynamic properties that affect bioink crosslinking must be closely monitored and optimized for quality control. Real-time quantification of these properties expedites optimization of bioink formulation, reduces printing errors, and leads to informed control of their values through feedback. Current methods for monitoring CQAs use off-line processes that involve manually testing samples. These labor-intensive methods fail to capture the time- sensitive qualities of bioink that influence its printability. In contrast, in-line sensors reside in the path of fluid flow and automatically monitor the sample. Because in-line sensors collect data on bioink during extrusion, they are compatible with multi-layered printing and can serve either as multi-use or single-use units. Integrating in-line sensors into the bioprinting process would expedite data collection for quality control as well as ensure that the printed construct is within target criteria. 1.3.2. Bio-ink Mixing/Homogeneity Bioink reproducibility relies on the homogeneity of the sample after mixing the bioink, crosslinker, and cells. Notably, undermixing samples leads to clumping and heterogenous distributions of bioinks and cells (70), and overmixing can result in over-gelled constructs with fractured morphology and reduced cell viability (71). Both byproducts of under and overmixing affect the mechanical characteristics of the printed construct by introducing regions of poor structural stability (72–74). Typically, 11 bioinks are mixed prior to printing using a tool such as a 3-way stopcock, static coupler, or automatic mixer (70,72,74,75), however, structurally complex constructs require additional processing. Due to the settling of bioink components over time, longer prints require additional mixing to prevent variations in print quality (72). The delay from mixing to assessment limits the ability to adjust the sample. Real-time monitoring of bioink homogeneity would enable the optimization of mixing procedures to produce high-quality constructs. Monitoring bioink mixing can occur through optical sensors, in-line load sensors, or ultrasound. Optical sensors are low-cost and quickly monitor for obstructions in fluid flow utilizing the reflectance or absorption of light to a transmitter (76). While optical sensors are commonly used due to their affordability, they are limited by the width of the imaging channel that light can pass through (< 600 μm) as well as the viscosity of the bioink solution due to light scattering (18–21). Thus, optical sensors are excellent for low-volume applications such as micro- extrusion bioprinting or inkjet printing. In contrast, high-volume or viscous bioinks are better monitored using in-line load sensors. These sensors measure the force exerted to deposit the sample at a set rate, and homogeneity is determined by the variation in force over time (1). Upon comparing alginate bioink samples mixed using 8 to 200 cycles, the sample created from 8 mixing cycles displayed greater compositional heterogeneity and produced higher variation in force compared to gels that were mixed more (Fig. 1.3) (1). Notably, in-line load sensors includes other factors such as noise generated by friction from the plunger on the syringe wall as well as the taper of the syringe barrel (1). 12 Lastly, bioinks may also be monitored using ultrasound. A transducer passes ultrasonic signals to the sample, and this signal is reflected based on the material differences at the bioink/air and bioink/print surface interfaces (23,77–79). This method monitors bioinks layer-by-layer to quantify internal flaws such as heterogeneity in bioink or crosslinking (77,80). Notably, sensitivity to these characteristics decrease as the thickness of the printed construct increases (80). Figure 1.3: Adapted from reference (1) with permission. a.) Experimental setup with custom-printed syringe holder and in-line load cell. A test-frame produced a pre-set load on the syringe, mimicking a syringe pump. b.) Mechanical noise described the variation in force experienced by cell-laden alginate samples that were created with mixing cycles ranging from 8 to 200. 13 Table 1.1: Key features and limitations of real-time monitoring methods for bioink mixing. Real-time monitoring of bioink mixing Optical sensors In-line load cell Ultrasound Key features Great accuracy for narrow channels (< 600 μm) (18) High resolution of low viscosity bioinks Continuous monitoring of viscous bioinks over time Layer-by-layer assessment (77) Limitation(s) Light scattering through high viscosity bioink and wider channels (19–21) Limited depth of view (< 1 mm) (81) Includes other factors such as frictional noise (1) Depth penetration (μm - mm scale) (23,77) 1.3.3. Bioink pH Several common hydrogels (collagen, gelatin, alginate, etc.) have charged amino acid side chains that influence their crosslinking behavior and are particularly affected by pH (82–87). Crosslinking time can be accelerated or slowed by pH, influencing both the rheological properties of the bioink and the print quality of the construct (58). In addition, cells are extremely sensitive to changes in pH with deviations from native pH resulting in alterations to cell bioactivity and viability (88,89). Before bioprinting, bioink pH is manually checked after mixing using a pH probe or pH strips, however, these methods occur off-line, taking precious time during which the sample may begin to crosslink. In addition, uneven mixing of bioink samples may create pockets of heterogeneous pH, which would lead to clumping and eventual nozzle clogging (1). As a result, real-time monitoring of pH can optimize bioink crosslinking and maintain 14 cell health. While monitoring pH in real-time is currently limited in bioprinting, pH is regularly measured for cell cultures over time, particularly in bioreactors. Commercially available pH sensors for cell-based applications are either electrochemically or optically based. Electrochemical probes are well-studied, easy to handle, sterilizable, multi-use, and measure sample pH within seconds. However, due to their bulky size, electrochemical probes are often used for off-line measurements prior to printing rather than real-time, in-line measurements (88). In addition, the accuracy of the pH measurements is limited by the homogeneity of the bioink solution. Optical sensors use absorbance or fluorescence to assess pH and are slim, portable, and compatible with in-line monitoring (88). Notably, optical sensors rely on colormetric assessment, typically using phenol red as an indicator of pH (90–92). While optical sensors provide great compatibility with bioprinting, the necessity of a pH indicator dye may limit their applications. Another option for real-time monitoring of pH are 3-D printed sensors, which can be low-cost, single-use, biocompatible, and flexible. These sensors monitor pH by measuring changes in resistance and can be incorporated either internally or externally to a syringe for in-line monitoring or attached to the bottom of a petri dish to monitor construct pH over time (93). Due to their compatibility with multiple steps during bioprinting, 3-D printed sensors are particularly useful for bridging the gap between measuring pH during bioprinting and monitoring it over time during cell culture. 15 1.3.4. Bioink Temperature and Bioprinter Environment Temperature is known to affect the viability and activity of encapsulated cells and controls gelation for a range of thermosensitive bioinks (e.g. collagen, gelatin, GelMA) (94–97). Temperature control during 3D printing is well-established and has a maximum error of 28% (98); however, this same error has a dramatic effect on bioink crosslinking and cell viability (99–101). Despite this temperature requirement, several parts in the bioprinting process are not thermally regulated. For instance, as bioink travels from the syringe barrel to the syringe needle, the temperature may vary from 22 - 35°C (102). In addition, the temperature of the ambient air between the deposition tip and the print bed ranges from 19.7 – 24.2°C (99). Last, as the layers of bioink increase in a printed construct, heat transfer from the print bed to the newest layer becomes less effective (103,104). As a result, there are four areas of note for temperature regulation in bioprinting: the bioink storage vessel (syringes or tanks), the printhead, the print bed, and the ambient air between the print head and bed. Controlling and preventing the formation of unregulated temperature gradients between these sites is crucial to reduce the effect of temperature shifts on bioink morphology and construct reproducibility. This challenge is addressed by enclosing the bioprinter in a single environment; however, temperature control is achieved by measuring at a single location (105). For instance, bioink lying along the outer perimeter of a syringe will maintain the set temperature of the enclosure, but heat may not transfer into the center of the syringe (94). As discussed previously, mixing plays a key role in the 16 homogeneity of the sample as well as the temperature distribution (Section 2.2 Bioink Mixing/Heterogeneity) (74,104,106). Combining in-line temperature sensors with a mixing method in an enclosure may improve the temperature consistency of bioink. In addition, optimization of feedback control systems such as proportional integral derivative (PID) can be leveraged to maintain a set temperature over time with reduced error. Integrating a PID system into the extruder nozzle of a 3-D printer improved temperature consistency over time for polylactic acid (PLA) and decreased the error to 7% (95). The stability of temperature in this system suggests that PID controllers can also be integrated into bioprinting systems. Thermosensitive bioinks rely on temperature change for gelation and require a different strategy. Rather than enclose the entire bioprinting system, a dual enclosure system with heat exchange is implemented. For instance, in the case of collagen bioink, a cooling system refrigerates the stored bioink, and a thermal insulator in the deposition tip maintains the temperature (102). The bioink is deposited through the ambient air onto the heated print bed where the temperature shift engages bioink crosslinking. As the construct grows thicker, less heat is transferred to new layers, affecting the crosslinking and quality of the printed construct (58,107,108). To better control the thermal environment for the printed construct, an enclosure separate from the cooling system may be created. Sensors such as platinum resistance thermometers, which measure temperature using changes in resistance, can be integrated into the walls of the enclosure for real-time monitoring and control (88,109). In addition, thermal cameras paired with temperature controllers can track and adjust temperature gradients while the bioink is extruded in real-time (110). Development of a controlled 17 hot and cold system for thermosensitive bioinks enables fine-tuning of the crosslinking process as well as improves the repeatability of the construct through proper heat transfer. 1.3.5. Bioink Viscosity Although bioink printability is greatly influenced by its rheological behavior during extrusion, viscosity is rarely measured in real-time with bioprinting. Such measurements are challenging due to the complex flow system formed by the dynamic shear-thinning and non-Newtonian behaviors of hydrogel bioinks in addition to the unique flow geometries formed during printing (111). To account for this intricacy, bioink viscosity is typically measured ex situ through a rheometer or viscometer using a portion of the total volume (83,112–114). Notably, this method of quantifying viscosity before printing does not account for the influence of printing process controls such as temperature gradients, mixing, pressure, or pH over time, and ex situ measurements fail to recapitulate strain and flow patterns from printing (111,114). Integrating real-time measurements of viscosity into the bioprinting process would assist with evaluation of bioink formulation and optimization of print quality by measuring bioink flow behavior. Real-time monitoring of viscosity for shear-thinning or non-Newtonian fluids presents a complex challenge. In straight channels of known dimensions, shear viscosity η is calculated as a function of the flow rate Q and pressure drop ΔP (115). Shear viscosity during laminar flow has been monitored in geometries that resemble a syringe or microtubing (capillary or slit-die geometries) by measuring the 18 pressure drop between two points (116,117) or by tracking the flow rate (118,119). In addition, for complex flows using agitators, effective viscosity and effective shear rate were calculated from the dissipated energy (111). While these methods are robust for analyzing complex fluids with continuous flow, bioinks experience discontinuity of flow while the sample exits the printhead. To account for these discontinuities, an image processing method or flow sensor may be used to compute the real-time flow rate. For instance, in a 3-D printing process, the rheological behavior of poly-(vinyl alcohol) (PVA) and PVA/chitosan droplets were monitored during extrusion through a combination of pressure sensors and video analysis of droplet volume (64). In addition, methods from previously discussed sections can also be utilized for real-time viscosity calculations. For example, bioink homogeneity was previously monitored by integrating an in-line load cell into the printing process (1). Combining measurements of the extrusional force alongside video analysis of flow rate would enable the real-time calculation of viscosity for bioink. 1.3.6. Print Accuracy The biggest challenge for construct reproducibility is the formation of defects. In general, defects include print path deviations, irregular edges, start/stop points, or thick/thin line width (Fig. 1.4). These defects are caused by multiple sources including bioink properties (rheological properties (120,121), batch variations (1,122), and heterogeneity from mixing (1)) and environmental factors (syringe nozzle, temperature, humidity, pH, suboptimal printing parameters (2,123–125)). Notably, 19 defects propagate with each printed layer, making the early identification of errors a top priority. However, there are currently no commercially available, closed-loop bioprinters capable of adjusting printing parameters (flow rate, pressure, velocity, position) in real-time to improve print accuracy. As a result, printing parameters are manually optimized through a guess-and-check process to minimize the formation of defects. Figure 1.4: Adapted from reference (1) with permission. a.) An ovine meniscus was imaged using CT (b) and bioprinted in alginate. c.) Constructs that failed from syringe tip clogging or showed poor shape fidelity due to bioink heterogeneity. d.) Homogeneous prints without defects. To reduce defects and rapidly optimize printing parameters, an automated analysis system capable of identifying, predicting and/or correcting defects is required. Optical imaging methods have dominated the bioprinting field for identifying printed defects due to their low-cost and rapid data collection. Defects are automatically identified by qualitatively and quantitatively comparing the print to the model, and error measurements that exceed a set threshold are flagged (126,127). These error 20 measurements are used to tune the construct’s G-code by adjusting extrusion rate and print location, leading to an iterative improvement in accuracy for the following printed constructs (128,129). This method of automatically identifying and correcting defects improves construct reproducibility and was shown to reduce the measured error to <15% (128). However, due to differences in bioink properties and rheological behavior, this optimization process must be repeated for new bioinks and for different models. To rapidly correct defects across a wide variety of bioinks, ML can be integrated into image processing. ML is trained across a wide range of scenarios from technical settings (camera zoom, image focus, and lighting) to construct qualities (bioink width, over/under extrusion, and layer height) (2,130). Upon predicting and identifying the defects, ML edits the G-code to improve accuracy on the following printed construct (131–134). In a more recent study, ML provided fully automated, on-the-fly correction of errors within 10 seconds of printing single-line prints using alginate and collagen bioinks (Fig 1.5) (2). These studies enforce the compatibility of ML with predicting and correcting defects caused by various factors. Notably, while ML has shown >90% accuracy in predicting defects, it requires training with multiple (30 – 4,000) images prior to use (2,130). To mitigate the resources required to generate this training dataset, a diverse collection of simulations may be generated (2). Once trained, ML is a powerful tool that automatically optimizes printing parameters to improve print accuracy and reduce the occurrence of defects. 21 Figure 1.5: Adapted from reference (2) with permission. a.) Overview of the closed- loop error detection and correction system using convolutional neural network-based ML to identify extrusion state as “good”, “over”, or “under” and take the appropriate action. b.) Prints of infill and line structures undergoing correction using red-stained high viscosity alginate (100 mg/mL). Poisoned G-code were areas where the extrusion rate was intentionally increased or decreased to create a defect. Although optical image methods are common for identifying defects, they are typically optimized for single-layered prints. Image-based methods struggle to segment layers of transparent bioink, leading to difficulty with calculating errors and flagging defects in multi-layer prints (128). New approaches for identifying defects in multi- layered prints utilize optical coherence tomography (OCT) or thermal imaging. OCT is a tomographic imaging system capable of penetrating up to 5 mm within bioprinted constructs to determine micron-resolution defects in layer thickness, layer fidelity, filament size, or pore size (123,135–138). Use of this method to iteratively correct 22 constructs led to an improvement in both construct accuracy and reproducibility, showing a decrease in the measured error from 40% to 7% and differences of only 9 μm in pore size between 6 different batches over 6 months (125). Early correction for multi- layered prints is particularly useful to conserve resources and prevent printing failures. To achieve this, OCT was used to identify defects such as start-stop points, straight-line deviations, and turnarounds, and these anomalies were corrected by adjusting the printing path, pressure, and velocity during a second pass over the located defect sites (123). As such, OCT enables the iterative or layer-by-layer correction of construct defects. In addition to OCT, thermal imaging can identify defects across multi-layer prints. As discussed in Section 2.4 Bioink Temperature and Bioprinter Environment, thermosensitive bioinks are stored at a different temperature in a syringe compared to the print bed. By exploiting the temperature gradient formed during printing, thermal imaging highlights only the newest printed layer for defect detection (3) (Fig 1.6). Combined with automated image analysis, thermal imaging captures defects layer-by- layer, providing an opportunity for on-the-fly error correction. Figure 1.6: Adapted from reference (3). a.) Experimental set-up with high-speed 23 camera directed at the print bed and a temperature difference between the printhead and print bed. b.) (Top) Model of the 3-layer printed construct. (Bottom) Segmented and processed IR image from the thermal camera revealed only the most recently printed layer. 1.3.7. Dynamic Bioink Properties: Summary and Future Directions Monitoring dynamic bioink properties and print accuracy enables the collection of data in real-time for quality control assessment while minimizing expended resources. In-line sensors integrated into the bioprinting process measure attributes that affect bioink crosslinking (mixing, pH, and temperature) as well as bioink viscosity in real-time. Meanwhile, optical imaging methods with integrated ML track the printed construct over time to identify defects and make corrections. To further re-define the capabilities of a bioprinter, we suggest that future studies use real-time monitoring in closed-loop bioprinters to advance the automated processing and optimization of bioinks as well as the structural complexity of constructs. Addressing these challenges will catalyze the ability to develop a smart bioprinter capable of both fabricating constructs and monitoring their quality in real-time. Processing and optimization of bioink currently relies on manual guess-and- check systems to obtain a high-quality construct. This manual method of optimizing print quality greatly consumes resources and limits the scalability of bioprinting to small batches. Real-time monitoring with feedback control enables the automation of both bioink processing and print parameter optimization. Combining imaging and in-line sensors into bioprinters would enable technological innovations for automatic mixing, 24 pH adjustment, and temperature control. In addition, automated bioink processing could be used to print complex architectures where current methods are not sufficient for recapitulating the multipart native structures. In all cases, print defects could be corrected on the fly, maximizing print accuracy and leading to high reproducibility of structurally complex constructs. Notably, technology for defect correction has been limited to single-layer constructs with iterative corrections on the following print (123,125,128–130). As such, on-the-fly defect correction has yet to be expanded into multi-layer prints. These rapid adjustments in dynamic bioink properties would improve the reproducibility and scalability of bioprinting. Several advancements in bioprinting are possible with the use of real-time monitoring in closed-loop bioprinters. Bioink properties could be tuned in real-time to emulate tissue structures with gradients, such as soft-to-hard interfaces or to print multiple tissue types for vascularization or innervation (139–142). Shifts in time- sensitive bioink properties could inform construct mechanics, leading to the formulation of mechanically robust bioinks (27,143,144). Control over the spatial patterning and properties of bioink would improve the print accuracy of multi-level and multi-scale structures, such as organs. Incorporating real-time monitoring into bioprinters would greatly expand the potential of printing by enhancing control over the printed product. 1.4. In-situ Monitoring of Cells During Bioprinting Monitoring cells over time is an indispensable measurement of bioactivity in an 25 implant. Although hydrogels are known for providing suitable environments for cell growth, bioprinting processes affect cell bioactivity through shear stress (62,71), temperature (94–97), pH (88,89), osmolarity (145,146), and nutrient deficiency (62,147,148), which may affect the reproducibility of the printed constructs or patient outcomes (62,82,149,150). Currently, constructs are subjected to destructive methods to characterize CQAs for bioactivity. The lack of real-time assessment of cells during printing results in an undesirable cycle of producing multiple identical constructs, culturing them to various time points, experimentally assessing them, then implanting a single construct into a patient. Notably, the cost of obtaining autologous cells and culturing them into a bioprinted construct is exorbitant. Two examples of FDA-approved personalized medicine treatments that culture and load patient cells onto scaffolds are autologous chondrocyte implantation (MACI) and cultured epidermal autografts (Epicel). MACI products are 3 x 5 cm sheets that are cultured for about a month to obtain a density of 500,000 chondrocytes per cm2 for cartilage repair. Epicel supplies 50 cm2 patches that are cultured with keratinocytes and murine fibroblasts for over two weeks to regenerate deep dermal burns. Collecting, transporting and culturing cells for MACI was estimated to cost $20,719 in 2014, and the overall cost of Epicel treatments is $79,000 on average (16,151). The high cost of production for both products includes cell processing, scaffold culturing, and product quality control testing. Despite the exorbitant costs related to cell processing, nearly all literature for in- line or real-time monitoring in bioprinting addresses dynamic bioink properties, 26 particularly print accuracy. Cell properties that are valuable for determining cell health include concentration, viability, and type (Fig. 1.7a). Notably, measurement of cell properties must be achieved using non-destructive methods such as optical-based imaging, spectroscopy, or ultrasound (Fig. 1.7b). Several methods for monitoring cell concentration, viability, and type are listed in Fig. 1.7c-e. Figure 1.7: a.) Representation of a remodeled bioprinter capable of monitoring cell properties (concentration, viability, and type) during bioprinting. b.) Recent advances in non-destructive methods for monitoring cell properties during printing are image- or spectroscopy-based. c-e.) Various in-line methods that are compatible with cell concentration, viability, and type. 27 1.4.1. Cell Concentration The quantity of cells in a bioprinted construct as well as their spatial distribution influences key biological activities for tissue regeneration such as cell viability, signaling, proliferation, and bioactivity through the production of ECM (152–156). However, cells suspended in viscous bioinks experience heterogeneous distributions caused by uneven mixing, cell settling, or hydrogel crosslinking that result in variability in cell expansion (157,158). Traditional methods for monitoring cell density in 3-D cell culture involve optical-based image analysis such as fluorescence microscopy or dissolved oxygen sensors post-printing (159–164). Image- based methods in particular are useful due to the spatial visualization of cells, however, they are limited by manual image collection/analysis and by the addition of fluorescent labels (165). Because image-based analyses are typically performed post- printing, cell concentration is passively measured but cannot be adjusted without printing a new construct. Removing the time-delay through automated imaging and analysis would enable rapid measurement of cell concentration and tuning of the deposited density. Notably, fluorescently labelled cells may not be compatible with applications in clinical trials (166). In addition to increasing the regulations and safety assessments needed for the printed product, fluorescent markers are difficult to image once implanted. To address these limitations, image-based methods with automated analysis have been integrated into bioprinting for real-time monitoring of cell concentration. Rapid assessment of cell quantity can occur either during bioink deposition or for 28 multi-layered constructs, immediately after each layer. During bioink deposition, monitoring the size and volume of bioink droplets provide insights into the cell count (165,166). Cell-laden bioink has previously shown a reduction in size compared to acellular bioink due to cellular contraction (47). As a result, printed lines, bioink strands, and droplets of cellular bioinks show decreased width and volume compared to acellular bioink during deposition (165). However, this change in size only indicates the presence of cells and does not provide their quantity. Previous studies using droplet analysis either rapidly measured the number of cells in droplets through fluorescent microscopy (<400 cells/droplet) (166) or predicted cell concentration based on bioink velocity profiles using ML (0 - 3×106 cells/mL) (165). Image-based methods can also monitor fluorescent cells on a layer-by-layer basis to track printed cell distribution (Fig. 1.8a-b) (4). Pixel intensity was shown to increase with additional cell-laden layers, which quantitatively and qualitatively measured cell distribution (using 0 - 4×106 cells/mL) (Fig 1.8c) (4). Image-based monitoring methods are particularly effective for cell concentrations <10×106 cells/mL but struggle to monitor higher cell concentrations due to light scattering (18–21). An alternative method for measuring cell concentration in real-time without labelling cells is dielectric spectroscopy (DS) or high-frequency ultrasound. DS measures the dielectric response of cells as they are exposed to an alternating current through sample impedance, phase angle, capacitance, or cell membrane conductivity or permittivity. Cell concentration and proliferation have previously been measured in bioprinted constructs post-fabrication (167,168), but only recently has this technique been used for real-time monitoring of cell properties using 29 in-line impedance sensors onboard a syringe with cell concentrations ranging from 0.1×106 - 125×106 cells/mL (Fig. 1.9a) (5,7). Regardless of whether the cell-laden bioink sample was monitored post printing or during printing, impedance and capacitance were shown to increase with cell concentration (Fig 1.9b) (5,167,168). In fact, sensitivity to changes in cell concentration was on the timescale of seconds (Fig. 1.9c) (5). In addition to DS, high-frequency ultrasound has shown great promise for label-free, real-time measurements of cell concentration in bioreactors (22,169). High frequency bursts are sent into the sample, and the generated acoustic backscatter is dependent on the quantity of cells in suspension. Acoustic backscatter was found to increase linearly with cell concentration using Chinese ovary hamster (CHO) concentrations from 0.1 - 100 ×106 cells/mL and mouse embryonic fibroblasts from 0.01 - 1 ×106 cells/mL (22,170). In addition to cell concentration, ultrasound revealed the distribution of human dermal fibroblast cells in GelMA hydrogels (23). High- frequency ultrasound provides measurements on both cell concentration and distribution without the use of fluorescent labels, but the spatial resolution is reduced compared to image-based methods (23). 30 Figure 1.8: Modified from reference (4). a.) Fluorescence monitoring system with a microscope directed at the print bed. b.) Heatmap showing the position of human embryonic kidney 293 cells printed at 2×106 cells/mL in Cellink Bioink RGD hydrogel. From top to bottom, layers 1, 2, 3 and 4 are displayed. c.) Plots of pixel intensity for the central region of each layer using samples at (top) 2×106 cells/mL and (bottom) 1×106 cells/mL. Figure 1.9: Modified from (5). a.) A custom in-line impedance sensor for DS was added between the syringe body and syringe tip. b.) The impedance |Z| of cell-laden samples was subtracted from acellular PBS to quantify |Zcells|. |Zcells| increased with cell concentration from 1 - 125×106 cells/mL at 25 kHz. Line represents the average of 5 independent trials, and bars show standard deviation. c.) |Zcells| changed accordingly 31 as the cell concentration of the sample was altered under flow rate at 1 mL/min. Dark line showed the average of 3 independent trials. 1.4.2. Cell Viability Maintaining high cell viability is crucial at all stages of the bioprinting process, and cell viability greatly affects the regenerative capability of a printed construct. Unfortunately, the bioprinting process and mechanical environment may negatively affect cell health (6,71,171–173). Cell viability is affected by environmental factors such as processing prior to printing (174), pH (88,89,175), and temperature (100) as well as effects from the printing process including mixing and shear stress (176). The greatest contributor for cell death is shear stress caused by the interaction between cells and the walls of a syringe, tube, or capillary in the presence of a velocity gradient (63,171,177). This velocity gradient reaches its peak when shear stresses are maximized as cells pass through the nozzle for printing (171,178). Although shear thinning hydrogels protect cells to some degree from shear stress, cell viability may decrease to as low as 40% (6,179–181). Cell viability is manually measured before printing using flow cytometry, trypan blue staining assays or automated cell counters and is measured after printing through fluorescent microscopy. However, these processes occur off-line and are time- and resource-consuming. In-line and real-time measurements of cell viability would identify regions of high and low cell health in the printed construct as well as provide an opportunity for correction. 32 Similar to real-time monitoring methods for cell concentration, cell viability is typically measured using optical imaging, spectroscopy, or ultrasound. As previously noted, image-based methods enable the continuous monitoring of samples during printing but require narrow imaging windows to reduce the effects of light scattering. To visualize the effects of shear stress on cells suspended in bioink, a 400 μm diameter capillary with cells suspended in GelMA at 1×106 cells/mL was imaged during bioink extrusion using light-sheet fluorescent microscopy (6). Through this real-time monitoring method, the unique velocity profile of cells in crosslinked and uncrosslinked bioink was visualized (6). Optical imaging provides high resolution monitoring of cell viability; however, the functionality of this method is limited by its scalability for image acquisition and analysis. DS and ultrasound are alternatives to optical imaging for applications requiring cell concentrations above 10×106 cells/mL. DS measurements rely upon the dielectric response of cells caused by polarization across the cell membrane in an alternating current. As a result, the way in which cell death occurs (apoptosis, temperature, pH, or shear) and the state of the cell membrane afterwards influences the impedance measurement (182). DS provided faster detection of cell death compared to off-line trypan blue assays (183). During printing, reduced viability (0% or 40%) was detected in chondrocytes at 25×106 cells/mL and in neuronal cells from rat adrenal glands (PC- 12) at 0.5×106 cells/mL (5,7). Similarly, high-frequency ultrasound measured increasing backscatter as a result of improving cell viability in a population of CHO cells at 5×106 cells/mL with viability ranging from 3 – 99% (22). 33 Figure 1.10: Adapted from reference (6). a.) Experimental method using light-sheet fluorescent microscopy to monitor cell-laden bioink. b.) Fluorescent cells were tracked over time in GelMA solution, pluronic, and gelled GelMA. The extracted cell tracks were color coded with mean velocity. c.) Plots of the tracked cell velocities and fitting to estimate the fluid velocity profile as a function of distance from the center of the capillary for GelMA solution, pluronic, and gelled GelMA. 34 Figure 1.11: Adapted from reference (7). a.) Multifunctional fibers were integrated into a syringe tip and sealed with epoxy for in-line impedance monitoring using dielectric impedance spectroscopy. b.) LIVE/DEAD assay of high or low viability samples of PC- 12 cells at 5×105 cells/mL in alginate with nanofibrillated cellulose. c.) Impedance magnitude |Z| for high and low viability samples from 100 – 10,000 Hz. D.) Phase angle for both samples. 1.4.3. Detecting Multiple Cell Types Recapitulating native tissues oftentimes requires the complex patterning of different cell types to create architecturally complex 3D co-cultures (35,184). Examples of spatially complex constructs that are printed with different cell types include full- scale constructs with vascular systems or epidermis and dermis layers for skin bioprinting. Interconnected 3D vascular systems improve nutrient diffusion in full-scale constructs. Microvessels are printed into multi-layered hollow channels using a blend 35 of bioink and cells (endothelial, smooth muscle, or mesenchymal stem cells) to provide structural support and promote cell communication (35,46,184). For skin bioprinting, current efforts to accelerate wound healing involve printing full thickness skin with epidermis and dermis layers. These layers are formed using different ratios of keratinocytes, fibroblasts, or stem cells to produce stratified epidermis and elastic dermis (12,185–191). In these examples, the spatial patterning of each cell type created specialized tissue layers. Despite the importance of cell spatial patterning, cell type is typically monitored post-printing and cannot be adjusted in real-time. Identifying cell type through in-line monitoring enables the tuning of cell spatial deposition and mixing ratios during printing. Monitoring cell type during construct fabrication may occur through DS or Raman spectroscopy. DS has previously identified cell types by size or morphology (192–194), biological activity (195–197), or differentiation (198–200). Using an in-line sensor for DS during bioprinting, the impedance signature of PC-12 cells was higher on average than mouse embryonic fibroblasts (NIH/3T3) cells (7). An additional study utilized DS to monitor cell-laden constructs post-printing, finding slight differences in relative permittivity between human adipose derived stem cells (hASC) and osteosarcoma cells (MG63) (168). Notably, both studies assessed two cell populations separately without the use of a co-culture. DS provided rapid identification of cell type and was suitable for in-line monitoring but would benefit from automated data analysis through ML when assessing co-cultures. Real-time identification of cell type could be expanded to detect unwanted pathogens in bioink. To identify sample contamination, pathogens such as bacteria are 36 cultured in vitro for hours to days (201,202). Notably, <2% of all bacteria can be cultured in this manner, showcasing both the time needed to test for contamination and the uncertainty associated with traditional methods (8,203–205). While DS has been used to monitor the dielectric response of bacteria such as E. coli (194), it has yet to be used for sterility assessments. Instead, a combination of surface-enhanced Raman spectroscopy and ML rapidly identified the presence of E. coli or S. epidermidis in blood droplets immediately following deposition from acoustic bioprinting (Fig. 1.12) (8). Surface-enhanced Raman spectroscopy quantifies the energy change from laser photons after interacting with particles in the sample, enabling the rapid and label-free identification of particles (greater than 2 μm) of differing shape or size (206). Notably, this method is most effective using small samples such as droplets combined with a metal such as gold or platinum. As a result, surface-enhanced Raman spectroscopy may prove useful for rapidly testing bioink batches for sterility. Figure 1.12: Adapted from reference (8). a.) Schematic of acoustic printing platform with surface-enhanced Raman spectroscopy occurring immediately following printing for mouse red blood cells infused with gold nanorods (GNRs) and bacteria. b.) Mean spectra for different combinations of red blood cells (RBC), E. coli and S. epidermidis with 100 measurements each, taken from single droplets and mixed with GNRs. Blue, green, and red vertical lines represent wavenumbers displaying the biological peaks for each cell type. 37 1.4.4. Cell Monitoring: Summary and Future Directions Quantification of cell bioactivity at multiple stages during the printing process is crucial for quality control, however, current methods are manual, destructive, and costly. This difficulty in measuring cell properties presents a major challenge for the field. Monitoring cell properties during printing enables the automated quantification of cell health and reduces costs associated with testing post-fabrication. To continue expanding upon real-time monitoring of cell health, we suggest that future studies address the need for automated data processing, the integration of feedback control into bioprinting, and the optimization of the printing process to reduce contamination risks. Investigation into solutions for these challenges will drastically alter the capabilities of bioprinting and facilitate the translation of printed constructs into clinical trials. Real-time measurement of cells in bioink was conducted using optical imaging, spectroscopy, ultrasound, or Raman spectroscopy (Table 2). Optical image-based methods showed great accuracy in tracking cell number or viability and were excellent for studies with 1-4 printed layers (4). However, optical imaging was limited to using fluorescent labels, low viscosity bioink, and low cell concentration (<10×106 cells/mL (207,208)). Meanwhile, DS was label-free and compatible with in-line measurements of high cell concentrations (5) and viability (5,7). Because DS measured the bulk sample response, it was not as accurate as image-based methods that counted individual cells. Additionally, DS was capable of monitoring multiple cell properties at once, but deconvolution of these properties requires complex data processing (5,209–211). In comparison, ultrasound-generated images of cell distribution and viability had 38 resolution between optical imaging and DS (23). Ultrasound was compatible with layer- by-layer monitoring, but required reference samples with the same speed of sound and temperature as the cell-laden bioink to calibrate the system (23). Lastly, Raman spectroscopy utilized the unique fingerprint of different cell types to evaluate the presence of bacteria in printed droplets (8). This method required the use of metallic additives to visualize samples and advanced data processing. While each method showed advantages and disadvantages, optical imaging, DS, and ultrasound could be improved through automatic data processing. To make full use of real-time monitoring, data processing must occur rapidly. Trained algorithms such as ML may reduce processing time for data analysis. Table 1.2: Comparison of methods used for real-time monitoring of cell properties. Applications of these methods, their minimum requirements, and limitations are listed. Applications Requirements Limitations Optical imaging • In-line or post-printing with high spatial resolution (~100-500 nm (212)) • Measures cell concentration < 10×106 cells/mL and viability • Single-cell measurements within the • Fluorescent labelling of cells • Light scattering limits imaging up to 4 layers (4) 39 focus of the objective Dielectric spectroscopy • In-line sensor • Measures cell concentration (0.1×106 - 125×106 cells/mL), viability, and type (5,7) • Measures cells that are between the electrodes (bulk sample) • Alternating current must pass through a conductive sample • Sensitive to multiple parameters based on frequency High- frequency ultrasound • Layer-by- layer measurements • Measures cell distribution and viability • Monitors regions based on transducer location (23,170) • Reference samples must have same speed of sound and temperature as samples (23) • Depth penetration from μm – mm scale • Sensitive to multiple parameters based on acoustic backscatter Surface- enhanced Raman spectroscopy • Droplet assessment • Measures cell type or sterility • Measures the energy released by excited molecules (206,213) • Requires metal additive such as gold to increase signal (8,213) • Sensitive to multiple parameters based on wavenumber • Depth penetration (μm) (8,213) 40 The previous sections showcased an array of exciting methods for monitoring cell concentration, viability, and type. Yet, these methods were integrated into open- loop bioprinters that are unable to respond to the collected data. Utilization of real-time monitoring in closed-loop bioprinters enables the control of cell properties using feedback. Directly tuning the concentration of cells to match target criteria will vastly improve construct reproducibility and lead to predictable construct bioactivity. Rapid identification of cell type could be used to adjust the ratio of cells in co-cultures during printing. In addition, cell viability and print accuracy could be optimized together by adjusting print parameters such as flow rate using feedback control (6,125). These real- time adjustments ensure that the printed construct is within regulatory criteria while minimizing material wasted from guess-and-check processes. Closed-loop bioprinters could also become self-contained units for fabrication, evaluation, and growth. Current bioprinters are tools for the fabrication of printed constructs. Introducing real-time feedback into the printing processes leads to the automated assessment of dynamic bioink properties and cell properties at multiple points during production. In addition to bioink and cell properties, sterility is a concern for every step of production. The fabrication of bioprinted constructs involves mixing bioinks prior to printing, loading the bioinks into the bioprinter, printing the construct, and incubating the sample. Contamination risks can be mitigated by reducing the number of steps involving sample handling. While many current bioprinters are enclosed in biosafety cabinets with temperature and humidity control, bioinks are processed before and after printing in different environments (105). Equipping bioprinters with the capability to automatically mix bioinks, print constructs, and 41 incubate the cell-laden constructs in an enclosed, sterile environment would reduce the risk for contamination. A hallmark of all tissues is their cellular complexity. To recapitulate native structures, cell concentration can be spatially patterned using closed-loop bioprinters. For instance, cartilage can be emulated using cellular gradients of chondrocytes and chondroblasts (168,214,215); full thickness skin could be printed using depth-dependent ratios of keratinocytes, fibroblasts, or stem cells (12,185–191); high-shape fidelity organoids could be formed using multiple cell types (216,217); whole organs could be produced with vascularization and innervation using accurate deposition of multiple cell types (35,142,155,218). As a result, closed-loop bioprinters can support the production of complex tissues. 1.5. Conclusion Through this article we have proposed the integration of real-time sensing technology into the bioprinting process to usher a new era of “smart” bioprinters. These new bioprinters would be capable of both producing and evaluating constructs, enabling the fine-tuning of crucial parameters as well as ensuring constructs meet regulatory criteria. The use of real-time monitoring during printing is anticipated to drastically lower production costs by reducing post-fabrication destructive testing and eliminating guess-and-check procedures. Several methods for monitoring bioink properties (homogeneity, pH, temperature, and viscosity), print accuracy, and cell properties (concentration, viability, type) were described in this article. By directly monitoring CQAs and standardizing construct quality, the path to clinical trials will 42 also become more accessible. Major avenues for improving real-time monitoring include the automation of data collection and analysis and the use of this data for feedback control mechanisms. 1.6. Acknowledgements We acknowledge the National Science Foundation I-Corps program. 43 CHAPTER 2 REAL-TIME ASSESSMENT OF CELL CONCENTRATION AND VIABILITY ONBOARD A SYRINGE USING DIELECTRIC IMPEDANCE SPECTROSCOPY FOR EXTRUSION BIOPRINTING* 2.1. Abstract Bioprinting produces personalized, cell-laden constructs for tissue regeneration through the additive layering of bio-ink, an injectable hydrogel infused with cells. Currently, bioprinted constructs are assessed for quality by measuring cellular properties post-production using destructive techniques, necessitating the creation of multiple constructs and increasing the production costs of bioprinting. To reduce this burden, cell properties in bio-ink can be monitored in real-time during printing. We incorporated dielectric impedance spectroscopy (DIS) onto a syringe for real-time measurement of primary chondrocytes suspended in phosphate buffered saline (PBS) using impedance (|Z|) and phase angle (θ) from 0.1 – 25,000 kHz. Cell concentration and viability ranged from 0.1×106 cells/mL - 125×106 cells/mL and from 0% - 94%, respectively. Samples with constant or with changing cell concentration were exposed to various flow conditions from 0.5 - 4mL/min. The background PBS signal was subtracted from the sample, allowing for comparisons across devices and providing insight into the dielectric properties of the cells, and was labelled as |Zcells| and θcells. |Zcells| shared a linear correlation with cell concentration and viability. Flow rate had *Matavosian AA, Griffin AC, Bhuiyan DB, Lyness AM, Bhatnagar V, Bonassar LJ. Real-time assessment of cell concentration and viability onboard a syringe using dielectric impedance spectroscopy for extrusion bioprinting. Biofabrication. 2025 Feb;17(2):025018. 44 minimal effect on our results, and |Zcells| responded on the order of seconds as cell concentration was altered over time. Notably, sensitivity to cell concentration and viability were dependent on frequency and were highest for |Zcells| when θcells was minimized. Cell concentration and viability showed an additive effect on |Zcells| that was modelled across multiple frequencies, and deconvolution of these signals could result in real-time predictions of cell properties in the future. Overall, DIS was found to be a suitable technique for real-time sensing of cell concentration and viability during bioprinting. 2.2. Introduction With the increasing demand for regenerative medicine and personalized implants, bioprinting has emerged as a promising technique. Bioprinting uses bio-ink, a printable hydrogel infused with autologous or allogeneic cells, to build 3-D constructs layer-by-layer with the spatial resolution needed to mimic the structure of native tissue (219–223). Notably, 3-D bioprinted constructs have already entered Phase I/IIa combined clinical trials for ear replacements (10,11) and Phase I/II for trachea transplants (13). However, quality control for cell-laden printed constructs is critical, due to the effect of extrusion bioprinting conditions on cell bioactivity (62,149,150). For instance, cell proliferation and viability decrease as a result of shear stress (62,150), temperature (62,149), pH (62,82), or osmolarity (145,224) resulting in reduced cell activity and remodeling. Quality control processes measure the cell bioactivity of the construct and rely on destructive techniques such as histological sectioning and staining (43–45), 45 fluorescent imagining (45–47), and biochemical analysis (44,47,48). To assess cell bioactivity, several constructs are bioprinted per patient and destroyed to confirm key characteristics (168,225). As a result, expenses for bioprinting are inflated from the cost of cell production for each construct. Currently, implants on the market that require autologous cell harvest that are cultured and loaded onto scaffolds include autologous chondrocyte implantation (MACI, e.g.) and cultured epidermal autografts (Epicel, e.g.), both of which are supplied by Vericel. MACI infuses autologous chondrocytes onto a collagen scaffold for cartilage knee repair, and Epicel incorporates autologous keratinocytes and murine fibroblasts into epidermal sheets as a skin replacement for burn victims. Culturing cells for these products costs over $15,900 for MACI and an estimated $5,292 per patch for Epicel as of 2014 with 100 patches used on average per patient (16,17,226,227). These costs hinder the growth and scaling up of bioprinting as well as limit its affordability for patients. In-line monitoring of cell bioactivity during printing would provide cost effective measurements of construct quality that would reduce the need for destructive testing (168,196).Several detection techniques are available for non-destructive measurement of cell bioactivity in other cell-dispensing systems such as microfluidic devices and syringe pumps. For instance, optical techniques such as optical spectroscopy, flow cytometry, or light-sheet fluorescence microscopy have previously estimated sample viability and differentiation during sample extrusion by quantifying the absorption, emission, or fluorescence of the sample when exposed to light (6,19,228). However, optical techniques are limited by the visibility of the sample through light scattering from viscous bio-ink and cell clustering from densities greater 46 than 10×106 cells/mL (19–21). Because of these challenges, optical techniques are well- suited for small-scale applications, but are not practical for printing large constructs on an industrial or clinical level with cell concentrations up to 100×106 cells/mL (47,229). Dielectric impedance spectroscopy (DIS) circumvents the limitations of light scattering and cell clustering by measuring bulk sample dielectric response to an electrical stimulus. When cells suspended in a conductive solution are exposed to an alternating current, they become polarized as ions within the cell cytoplasm and the surrounding media align. Known as the Maxwell Wagner Effect, the degree to which cells polarize depends on the frequencies used to generate the alternating current, the viability of the cell, and the dielectric properties of the cell, which are influenced by cell type, size, density, and media (25). DIS measures the impedance magnitude (|Z|) and phase angle (θ) of the sample, which together indicate cell resistance, reactance, and behavior and are used to determine cell properties (25,198). This technique has previously been used to collect real-time data relating to cell concentration (24,167,168), viability (182,230), type (7,197), and state of differentiation (199,200) using cell densities up to 25×106 cells/mL (24,199) in 2-D and 3-D cell cultures as well as monitoring cell proliferation and viability in bioprinted alginate constructs post- fabrication (167,168). More recently, DIS was integrated into a syringe tip as a proof- of-concept to monitor cells in suspension (7). However, DIS has yet to be validated in a dynamic sample extrusion method on a scale relevant to bioprinting. The primary goal of our study was to assess the compatibility of DIS for in-line monitoring of cell concentrations, cell viabilities, and flow rates that are relevant to tissue bioprinting while onboard a syringe. We previously designed a device (referred 47 to as the “smart syringe,” (9)) that attaches to commercial syringes for real-time measurement of cellular properties using DIS. However, the capabilities of our device for monitoring cells in conditions relevant to bioprinting and quality control are unknown. Our objective was to use the smart syringe to measure the |Z| and θ of primary chondrocytes for 1.) concentrations ranging from 1×106 - 125×106 cells/mL, 2.) samples with changing cell concentration over time, 3.) flow rates from 0.5 - 4 mL/min, and 4.) viability from 0 - 94%. Since bio-inks are oftentimes saline-based (231), we suspended primary chondrocytes in phosphate buffered saline (PBS) to assess the compatibility of this device with monitoring samples over 100×106 cells/mL with flow rate in a simplified environment. To determine the dielectric behavior of samples within a wide range of cell concentrations and viabilities, we created calibration curves using frequencies up to 0.1 - 25,000 kHz. Cell-laden solutions were extruded at flow rates relevant to bioprinting, and cell concentration varied with time to determine the real- time capabilities of DIS for bioprinting. In addition, we assessed the combined effects of cell concentration and viability on |Zcells| to determine the potential of monitoring both cell properties simultaneously. 2.3. Materials and Methods 2.3.1. Smart Syringe creation and use 48 Figure 2.1: DIS measured cellular response to electric currents. a.) As a cell suspension containing healthy and dead cells passed through the alternating current, a frequency- dependent response was measured as impedance |Z| and phase angle θ. In this diagram, bio-ink is flowing into the page. b.) A custom bioprinter loaded with a syringe and smart syringe. The smart syringe connected the syringe barrel to the syringe tip and was fitted with two antiparallel electrodes that generated an alternating current. DIS was integrated into a syringe for in-line monitoring of cell properties. By transmitting an alternating current between two electrodes, cells were temporarily polarized, allowing for the detection of cell properties using |Z| and θ (Fig 2.1a). The smart syringe was engineered to house the electrodes, and this device was reproducible using male and female luer locks, wiring, heat shrink, and a sealant such as epoxy (Fig 2.1Sa). AWG breadboard wires (Amazon, Seattle, WA) were stripped to reveal the copper wires, and the gold electrodes (RadioShack Tech Plus) were crimped onto one end of the wires (Fig. 2.1Sb). The plastic casing on the other end of the wires was removed, and a wire connection was added (Fig. 2.1Sc). Heat shrink (Digikey, Thief River Falls, MN) reinforced the binding between the electrodes and wires (Fig. 2.1Sd- e). The electrodes were press-fit into a machined female luer lock (Qosina, Ronkonkoma, NY) with a distance of 1-1.5 mm measured from each electrode’s midline. Distance between electrodes varied due to the difficulty of drilling into a 49 circular, uneven surface. The female luer lock and electrodes assembly was connected to male luer lock (Qosina) (Fig. 2.1Sf-g). To prevent air leaks, gorilla glue epoxy (Amazon) was applied to both ends of the heat shrink as well as across the point of insertion between the electrodes and female luer lock. After drying for 24 hours, the smart syringe device was attached to a commercial syringe and could be mounted on a bioprinter (Fig. 2.1b). The smart syringe was cleaned after each use by washing with 70% ethanol followed by two washes of a PBS solution composed of 1X PBS with 100 U/ml penicillin, 100 μg/ml streptomycin, and 250 ng/ml amphotericin B (VWR, Radnor, PA). Each replicate was tested individually and the order was randomized for each experiment. In total, 6 smart syringe devices were used in this study. On average |Z| measurements varied by 13% across different devices. Multiple devices were required as a result of electrode fouling caused by a build-up of salt and cellular debris which directly affected |Z| measurements after several experiments. 2.3.2. Smart Syringe DIS Parameters An alternating current was sent to the electrodes in the smart syringe by a Digilent Analog Discovery 2 device (Digilent, Pullman, WA). All studies used an AC voltage of 50 mV with zero DC offset and averaging occurring every 50-100 ms. One electrode served as the working electrode, and the other received the signal. An impedance analyzer attachment (Digilent) combined with the Waveforms software (Digilent) converted the signal into |Z| and θ. |Z| and θ were exported to Microsoft Excel for data analysis. Cell concentration studies used a logarithmic frequency sweep from 1 50 – 25,000 kHz with 18.5 points per decade and 101 steps from start to end frequency. The range of frequencies was expanded for cell viability studies to 0.1 – 25,000 kHz because viable and dead cells have previously shown greater differences in θ at low frequencies (7). Additionally, the frequency range was reduced to 0.1 – 1,000 kHz for flow rate studies after determining that sensitivity to cell properties declines at higher frequencies and to hasten data measurement by reducing frequency sweep collection time. Cell-laden solutions for bioprinting applications may suffer from heterogeneously distributed cell content due to uneven mixing, cell settling, or pre- mature bio-ink crosslinking (157,158). To monitor heterogeneity and changes in concentration, cell suspensions could be assessed periodically. The smart syringe monitored a small portion of the sample as it passed between the electrodes by sampling frequencies over a set time. This volume (vsmart syringe) was approximated using the volumetric flow rate (Q) and the time (t) for a frequency sweep from 0.1 – 1,000 kHz, similar to the flow rate studies, and is shown in Equation 1. However, a gap exists on either side of the electrodes where cells may evade detection (Figure 2.1b). The volume for this gap in addition to the volume of the electrodes (vgap) were approximated and were subtracted from vsmart syringe to calculate vcurrent (Equation 2). These calculations were displayed in Supplemental Table S1. 𝑣𝑠𝑚𝑎𝑟𝑡 𝑠𝑦𝑟𝑖𝑛𝑔𝑒 = 𝑄𝑡 (1) Q = volumetric flow rate (mL/min) t = time for single frequency sweep (s) vsmart syringe = volume passing through device (mL) 51 𝑣𝑐𝑢𝑟𝑟𝑒𝑛𝑡 = 𝑣𝑠𝑚𝑎𝑟𝑡 𝑠𝑦𝑟𝑖𝑛𝑔𝑒 − 𝑣𝑔𝑎𝑝 (2) vgap = volume avoiding electrodes (mL) vcurrent = volume passing between electrodes for measurements (mL) In addition to volume passing the electrodes, the amount of time that the cells spend in the alternating current was also calculated. To determine the cell residence time, velocity between the electrodes was calculated using Equation 3, where Q was the volumetric flow rate, A was the area of the electric field, and V was the velocity. A was calculated by multiplying the distance between the electrodes with the electrode thickness, and V was calculated from electrode thickness divided by residence time. Equation 3 was rewritten as Equation 4 and used to solve for the residence time of the cell, t. These values were displayed in Supplemental Table S2 for various flow conditions and electrode distances. 𝑽 = 𝑸 𝑨 (3) A = area of the electric field (mm2) V = velocity (mm/s) 𝒕𝒓𝒆𝒔𝒊𝒅𝒆𝒏𝒄𝒆 = 𝒅𝑻𝟐 𝑸 (4) d = distance between electrodes, ranges from 0.5 – 1.0 mm T = thickness of electrodes = 0.9 mm tresidence = residence time(s) 52 2.3.3. Primary chondrocyte cell isolation From a total of 44 neonatal bovine condyles, (Copper City Meats, Rome, NY) primary chondrocytes were isolated as described previously (229,232). Chondrocytes were selected for this study due to their wide applications in bioprinting and their high cell yield from isolation techniques. Condyles were dissected, suspended in PBS (VWR), chopped, and left to digest overnight in a 0.2 - 0.3% wt/v collagenase solution (DMEM 4.5 g/L glucose with L-glutamine and without sodium pyruvate (VWR), 100 U/ml penicillin (VWR), 100 μg/ml streptomycin (VWR), and 250 ng/ml amphotericin B (VWR), and 2 - 3 mg/mL collagenase type 2 (Worthington Biochemical, Lakewood, NJ)). After digesting for 16 - 20 hours, the solution was pipetted through a 40 μm cell strainer (VWR) and centrifuged. To wash the collagenase from the cells, the sample was aspirated, replaced with PBS, and centrifuged. This step was repeated before counting the cells using a hemacytometer (Hausser Scientific, Horsham, PA) with trypan blue staining (VWR), and light microscopy (Nikon Eclipse TS100). 2.3.4. Cell concentration studies Primary chondrocytes were divided into subgroups and diluted to concentrations at (1×106, 10×106, 25×106, 50×106, 75×106, 100×106, and 125×106) cells/mL using PBS. There was a total of 15 measurements per cell concentration that were collected from 5 independent experiments with n = 3 replicates each. Although bioprinting uses cell concentrations up to 100×106 cells/mL, a higher concentration at 125×106 cells/mL was included to better understand the behavior of the dielectric properties at high densities (47). Sample concentration and viability were assessed using trypan blue and 53 cell counting. Samples were stored on ice until immediately before use to preserve cell viability and to simulate preprocessing of typical bio-inks before printing, such as collagen gel. Chondrocytes have previously been shown to maintain viability in saline at 4°C for 24 hours with only a 25% decrease in viability after 3 days (233). Each sample was individually loaded into a syringe, and measurements were collected when the sample was in full contact with the smart syringe’s electrodes. Cell concentration studies were performed without flow rate. 2.3.5. Flow Rate To assess real-time changes in impedance, samples of primary chondrocytes were created at 1×106 cells/mL and 100×106 cells/mL. The samples were loaded into 10 mL BD plastic syringes (VWR) and connected to a smart syringe using a 3-way stopcock (MoHawk Medical, Utica, NY) and Tygon® tubing (McMaster-Carr, Elmhurst, IL). The flow rate was set to 1 mL/min (equivalent to 16 mm3/s) using a dual channel syringe pump (Harvard Apparatus), and 1×106 cells/mL was allowed to flow in isolation for 120 seconds. During a brief transition period lasting 30 seconds, several events occurred. First, flow was paused, and the 1×106 cells/mL sample was moved out of the syringe pump. Next, the syringe containing 100×106 cells/mL was added to the syringe pump, and flow was resumed. By the end of the 30 second transition period, the 100×106 cells/mL sample had passed through the smart syringe device. The 100×106 cells/mL sample then flowed in isolation for 120 seconds. Lastly, both 1×106 cells/mL and 100×106 cells/mL samples flowed at the same time for 120 seconds after a brief 15 second transition period. During this transition period, the 1×106 cells/mL sample was 54 returned to the syringe pump, and flow was not paused. This dual flow of both samples created a new concentration of 50×106 cells/mL. Three independent studies were conducted using this method, creating a total of n = 3. Cells in suspension are known to settle over time due to gravity (157,158). To isolate the potential confounding factor of cell settling, a sample of primary chondrocytes were suspended in PBS at a concentration of 100×106 cells/mL. The sample was added to the smart syringe device and mounted to a syringe pump without flow rate. |Z| was collected from 0.1 kHz – 1,000 kHz for 600 s. The sample was vortexed and the study was repeated with n = 3. Similar to previous experiments, the sample rested on ice between each trial to preserve cell viability. Since bioprinting oftentimes uses a range of flow rates to print various materials, the effect of flow rate on data collection was investigated. Samples at 25×106 cells/mL were extruded using a syringe pump at (0.5, 2, and 4) mL/min (equivalent to 8.3, 33, 67 mm3/s, respectively) with n = 4 per flow rate. Data collection began immediately after sample flow was initiated. |Z| and θ were normalized to t = 0 s for each flow rate to highlight potential changes in measurement quality. This study assessed various flow rates using samples at a single cell concentration. To further explore the effect of flow rate on data collection, an additional study was performed at a single flow rate of 2 mL/min using samples at (1×106, 10×106, 25×106, and 50×106) cells/mL with n = 3 per cell concentration. 2.3.6. Cell viability studies 55 Primary chondrocytes were divided into two equal groups. One group was left on ice until needed. Meanwhile, the other group was placed in a 50 ml centrifuge tube (Falcon, Corning, NY), allowed to rest in a liquid nitrogen bath for 10 min, then was rapidly thawed in a water bath at 37°C. This process of rapidly freezing and thawing the cells induced cell death through the generation of ice crystals (234). Cells in both the healthy and dead groups were counted using trypan blue and diluted to 25×106 cells/mL in PBS with target viabilities of 95% and 0%. Portions of the heathy cell sample and the dead cell sample were combined in a 1:1 ratio to create a third sample at 25×106 cells/mL with target viability of 45%. There was a total of 12 measurements per cell viability collected over 3 independent experiments with n = 4 replicates each. Viability and concentration were confirmed using trypan blue exclusion assay and a hemacytometer. Cell viability studies were conducted without flow rate. 2.3.7. LIVE/DEAD images and image analysis Primary chondrocytes suspended in PBS were collected pre- and post- testing with the smart syringe using LIVE/DEAD® assays (Thermofisher, Waltham, MA). Each sample was stained using calcein AM and ethidium homodimer-1 according to the manufacturer’s instructions. Fluorescence images were taken at 100X using a Keyence BZ-X810 microscope with n = 3 images per sample across 3 independent experiments. ImageJ was used to merge image channels, and LIVE/DEAD® image analysis was quantified using a custom MATLAB code. This code separated cells into live and dead categories using Otsu’s thresholding method (235). LIVE/DEAD images were 56 subjected to a Gaussian Blur then processed with a watershed separation code to distinguish cells in close proximity to one another. 2.3.8. Data Analysis and Sensitivity |Z| and θ measurements were collected from primary chondrocytes suspended in PBS and from PBS samples without cells as a background measurement. Taking the difference between these measurements (Equations 5, 6) isolated the dielectric properties of primary chondrocytes from the PBS and allowed for standardization between independent studies and different smart syringe devices. In Equation 5, |Zsample| indicates the magnitude of impedance from cells and PBS, |ZPBS| describes the magnitude of impedance for PBS without cells, and |Zcells| represents the adjusted magnitude of impedance for primary chondrocytes. Equation 6 followed a similar nomenclature for phase angle. Device sensitivity was calculated as the slope of the dielectric measurement over the cell property at each frequency. |𝒁𝒄𝒆𝒍𝒍𝒔| = |𝒁𝒔𝒂𝒎𝒑𝒍𝒆| − |𝒁𝑷𝑩𝑺| (5) 𝜽𝒄𝒆𝒍𝒍𝒔 = 𝜽𝒔𝒂𝒎𝒑𝒍𝒆 − 𝜽𝑷𝑩𝑺 (6) 2.3.9. Statistics All statistical tests were performed in collaboration with the Cornell Statistical Consulting Unit. |Zcells| and θcells for cell concentration studies and viability studies were analyzed with a linear regression at 25 kHz and 1 kHz, respectively. These frequencies were selected due to increased sensitivity to cell concentration and viability at these values and fall within similar frequency ranges as previously reported literature (7,24). 57 The first measurement from each cell viability experiment was omitted due to instrumental error. Flow rate studies used a mixed linear model between flow rates as well as over time within replicates. All linear models were created in GraphPad Prism. Quantified LIVE/DEAD images used a student’s t-test on Microsoft EXCEL. MATLAB was used to generate planes of best fit at (0.1, 1, 10, 100, 1,000, and 10,000) kHz for 40-54 samples from 0.1×106 to 125×106 with viability from 0 - 94%. Fewer samples were collected from 0.1 - 1 kHz compared to 10 - 10,000 kHz. Best-fit planes were fitted to Equation 7, where A represented the intercept of the plane, B shows the sensitivity to cell concentration, x signified cell concentration (106 cells/mL), C was sensitivity to cell viability, and y was viability as a ratio of living cells and total cells. |𝒁𝒄𝒆𝒍𝒍𝒔| = 𝑨 + 𝑩𝒙 + 𝑪𝒚 (7) Planar models of cell concentration, viability, and |Zcells| were assessed using the goodness of fit parameters SSE, R, and R2 generated from MATLAB, and significance was determined using two-tailed critical R value tables from GraphPad. 2.4. Results 2.4.1. Cell Concentration We measured |Z| and θ using cell concentrations from 1×106 cells/mL to 125×106 cells/mL across a frequency sweep from 1 - 25,000 kHz. |Z| decreased with frequency (Fig. 2.2Sa), and θ was negative from 1-1,000 kHz and shifted to positive values from 5,000-25,000 kHz (Fig. 2.2Sb). To compare measurements across independent experiments and isolate the cellular response, Equations 5 and 6 were used to calculate 58 |Zcells| and θcells, respectively. |Zcells| ranged from -3 to 31 Ω and showed monotonic decrease with frequency and 3 major regions: from approximately 1 - 4 kHz |Zcells| drastically decreased; from 4 - 1,000 kHz |Zcells| remained constant; and from 1,000- 25,000 kHz |Zcells| decreased (Fig 2.2a). Notably, |Zcells| increased with cell concentration at all frequencies (Fig. 2.2a). Similar to |Zcells|, θcells depended on frequency as a function of concentration with 3 regions of behavior: from approximately 1 - 4 kHz θcells increased with concentration; from 4 - 1,000 kHz θcells experienced noise; and from 1,000-25,000 kHz θcells decreased with concentration (Fig. 2.2b). Notably, θcells was relatively small, ranging from –2.5 to 1 degrees. This indicated that θcells was influenced by both cell concentration and frequency, however, these changes were small in comparison to θPBS. 59 Figure 2.2: Primary chondrocyte concentration was controlled from 1×106 – 125×106 cells/mL and assessed from 1 – 25,000 kHz with n = 3 per concentration. Shaded areas represent sample standard deviation. a.) |Zcells| measurements of primary chondrocytes after subtracting the background PBS signal. Dark lines indicate sample mean. b.) θcells of primary chondrocytes after subtracting PBS background. c.) 15 measurements from each cell concentration collected from five independent studies were used to show the relationship between |Zcells| and cell concentration at 25 kHz. Each point represents a single replicate. d.) These 15 measurements were also used to determine the relationship between θcells and cell concentration at 25 kHz. e.) Device sensitivity described the degree to which |Zcells| changed with concentration across the frequency sweep and was generated by calculating the linear regression from Fig. 2.2c across all frequencies. Dark line represents sample mean. f.) Device sensitivity to changes in θcells was generated by calculating the linear regression from Fig. 2.2d across all frequencies. To determine the correlation between dielectric properties and cell concentration, this study was repeated with 5 independent experiments and was shown 60 at 25 kHz (Fig. 2.2c-d). |Zcells| increased linearly with concentration across all studies and ranged in value from -3 to 21 Ω (R2 = 0.90, Fig 2.2c). θcells also showed a linear relationship with cell concentration, but the intercept of the correlation was dependent on the experiment (R2 = 0.78, Fig 2.2d). To further assess the frequency-dependence of |Zcells| and θcells, we calculated the device sensitivity, defined as the slope 𝑑|𝑍𝑐𝑒𝑙𝑙𝑠| 𝑑𝑥 and 𝑑𝜃𝑐𝑒𝑙𝑙𝑠 𝑑𝑥 where x represented cell concentration, and these derivatives were assessed across the full range of concentrations measured. The smart syringe showed the greatest sensitivity to cell concentration from 10 – 1000 kHz for |Zcells| (Fig. 2.2e) and from 1 - 10 kHz or 10,000 - 25,000 kHz for θcells (Fig. 2.2f). In fact, 25 kHz was selected for Fig. 2.2c and Fig. 2.2d because it offered rapid data collection and laid within this range for both values. These studies revealed the repeatability of our measurements as well as the correlations between |Zcells|, θcells, and cell concentration in a syringe-based device. To simulate conditions during bioprinting, we assessed |Zcells| as cell concentration was altered in real-time during extrusion. Two syringes loaded with either 1×106 cells/mL or 100×106 cells/mL were connected to a smart syringe using a 3-way stopcock and attached to a syringe pump (Fig 2.3a). By adjusting which samples flowed and whether they flowed in isolation or unison, the cell concentration entering the smart syringe was altered from 1×106 cells/mL to 100×106 cells/mL to 50×106 cells/mL. At 25 kHz |Zcells| ranged on average from -4 to 7.3 Ω for 1×106 cells/mL, 13 to 22.2 Ω for 100×106 cells/mL and remained constant at 14 Ω for 50×106 cells/mL (Fig. 2.3b). As cell concentration changed, |Zcells| measurements reflected the new sample on the order of seconds. Notably, cells typically settle to the bottom of the syringe when suspended 61 in solution. To determine if this phenomenon influenced the |Z| measurements of primary chondrocytes suspended in PBS over time, a sample at 100×106 cells/mL was allowed to rest in a syringe pump without extrusion for 600 s (10 min) (Supplemental Fig. 2.3S). Nearly all samples, including PBS without cells, displayed a downward trend in |Z| when normalized to t = 0 s, suggesting cell settling did not contribute to changes in |Z| over time in our device. Combined, these data highlight the capability of the smart syringe to monitor high cell concentrations above 1×106 cells/mL with and without flow rate. The influence of flow rate on |Zcells| and θcells was investigated by extruding samples at a single cell concentration with different flow conditions. Samples at 25×106 cells/mL were extruded at (0.5, 2, and 4) mL/min for 240 s. After normalizing the results to t = 0 s, neither |Zcells| nor θcells changed over time with flow rate (p > 0.05 using a mixed linear model, Fig. 2.3c). These results were confirmed for additional cell concentrations at (1×106, 10×106, 25×106, and 50×106) cells/mL at 2 mL/min (Supplemental Fig. 2.4Sa-b). As a result, flow rates from 0.5 - 4 mL/min showed minimal effect on the collection of |Zcells| and θcells for cells in solution, showcasing the potential for in-line monitoring of cells in conditions that simulate bioprinting. 62 Figure 2.3: Measuring cell concentration of primary chondrocytes at specified flow rate(s). Displayed data was from 25 kHz. a.) Diagram of sample flow using 3-way stopcock, smart syringe, and syringe pump. b.) Primary chondrocyte concentration was adjusted from 1×106 cells/mL to 100×106 cells/mL to 50×106 cells/mL with 120 s 63 (2 min) intervals at a flow rate of 1 mL/min. Concentration transitions occurred over 30 s and 15 s. Dark line indicates the average of 3 independent trials. c.) Diagram of sample at 25×106 cells/mL with extrusion at a set flow rate for 240 s. d.) |Z| and θ were measured at 25 kHz across three flow rates relevant to bioprinting for 240 s (4 min) with n = 4 per flow rate and with cell concentration at 25×106 cells/mL. Data was normalized to t = 0 s for each flow rate, which occurred immediately after sample flow was initiated. 2.4.2. Cell Viability Since bioprinting conditions can negatively affect bioactivity and cell viability, understanding the effect of the smart syringe on viability as well as determining the dielectric properties of samples with low viability are crucial. Samples with high (94%), medium (~50%) and low (0%) viability were created at 25×106 cells/mL, and these samples were stained via LIVE/DEAD assay, imaged, and quantified. Qualitatively, LIVE/DEAD images confirmed the intended ratio of viable and dead cells (Fig. 2.4a i-iii). Next, we assessed for changes in cell viability caused by the smart syringe device. High, medium, and low viability samples were collected pre- and immediately post-testing with the smart syringe. Pre-testing and post-testing groups showed similar quantities of viable and non-viable cells (Fig 2.4a i-vi), and this finding was confirmed through quantification of the images (Fig. 2.4b, p > 0.05 using student’s t-test). As such, sample viability was not altered by testing or handling with the smart syringe device. Next, we measured |Z| and θ for samples with high, medium, and low viability at 25×106 cells/mL. Raw values of |Z| showed a sharp decrease with frequency from 0.1 – 4 kHz (Supplemental Fig 2.5Sa), and viability influenced whether |Z| was higher than or lower than PBS (Supplemental Fig. 2.5Sb-d). In contrast, θ displayed similar values 64 between samples and PBS (Supplemental Fig. 2.5Se-h). After removing the background of the samples, |Zcells| ranged from -27.3 to 24.2 Ω at high viability, -41.2 to 6.31 Ω at medium viability, and -94.2 to 0.52 Ω at low viability (Fig. 2.5a). These values indicate that |Zcells| increased with viability, similar to previous literature, in a syringe environment (182). Across the tested frequency sweep, |Zcells| displayed behavior that could be separated into two regions: from 0.1 – 4 kHz |Zcells| increased and showed the greatest range in values; and from 4 - 25,000 kHz |Zcells| remained constant with a comparably smaller range (Fig. 2.5a). In contrast to |Zcells|, θcells ranged from -1.2 to 2.3 degrees for all samples and showed minimal dependence on frequency (Fig. 2.5b). From 0.1 – 4 kHz, θcells decreased from 2 to 0 degrees, then θcells was afflicted by noise centered about 0 degrees from 4 – 25,000 kHz (Fig. 2.5b). By repeating this study with 3 independent experiments, correlations were determined between |Zcells|, θcells, and cell viability. At 1 kHz, |Zcells| increased with cell viability (Fig. 2.5c, R2 = 0.70). However, no correlation was found between θcells and cell viability (Fig. 2.5d, R2 = 0.001). Device sensitivity to cell viability was calculated using the slope 𝑑|𝑍𝑐𝑒𝑙𝑙𝑠| 𝑑𝑦 and 𝑑𝜃𝑐𝑒𝑙𝑙𝑠 𝑑𝑦 with y representing cell viability and indicated the rate at which |Zcells| and θcells changed with respect to cell viability. |Zcells| showed high sensitivity from 0.1 – 4 kHz with measurements ranging from -29.4 to 50.5 Ω/106 cells/mL, but sensitivity remained relatively constant from 3 to 6 Ω/106 cells/mL across 4 – 25,000 kHz (Fig. 2.5e). In contrast, sensitivity using θcells ranged from -0.83 to 0.87 deg/%viability from 0.1 – 4 kHz and experienced noise centered about 0 deg/%viability from 4 – 25,000 kHz (Fig. 2.5f). Notably, 1 kHz was selected for Fig. 2.5c and Fig. 2.5d because it was within the sensitive range for both values and minimized the noise 65 experienced at lower frequencies. These results show the dependence of |Zcells| on cell viability and frequency as well as the lack of correlation between θcells and viability. Figure 2.4: Primary chondrocyte samples suspended in PBS were measured with high (~94%), medium (~50%) and low (0%) viabilities with cell concentration remaining constant at 25×106 cells/mL. Frequency ranged from 0.1 kHz – 25,000 kHz, and n = 4 replicates per viability. Dark lines represent sample means, and shaded areas show standard deviation. a.) |Zcells| after subtracting the background PBS signal. Boxed region is expanded from 0.4 – 1,000 kHz. b.) θcells after subtracting the background PBS signal. c.) Linear correlation between |Zcells| and viability at 1 kHz for 3 independent experiments using 3 different smart syringes. The medium viability was not identical for all independent studies and was thus shown with multiple points. d.) Linear correlation between θcells and viability across 3 independent experiments using different devices. e.) Sensitivity to changes in |Zcells| as cell viability increases. f.) Device sensitivity to changes in θcells as cell viability increases. 66 2.4.3. Cell concentration, viability, and |Zcells| The smart syringe showed strong linear correlations between |Zcells| and two key cell properties: concentration and viability. However, the combined effects of these properties on |Zcells| are unknown. To evaluate these combined effects, measurements were collected for 40 - 54 samples ranging from 0.1×106 – 125×106 cells/mL and 0 - 94% viability across a frequency sweep from 0.1 - 25,000 kHz. To simplify the large dataset, samples were plotted at 10 kHz with a best-fit plane using Equation 7 (Fig. 2.6a, R2 = 0.35, p < 0.0001 using R critical value tables). The best-fit plane displayed a moderate correlation between |Zcells| and the cell properties at 10 kHz. |Zcells| ranged from -6.38 to 20.3 Ω and increased as a result of both cell concentration and viability. In fact, the effects of these cell properties on |Zcells| were additive with high concentration and viability providing the largest |Zcells| values (Fig. 2.6a). Most samples were well- represented by the fitted model, however, points at the extremes, such as 0% viability with high concentration, deviated from the plane (Fig. 2.6a). To further assess the influence of frequency on these results, the best fit planes for (0.1, 10, and 10,000) kHz were generated with Equation 7 (Fig. 2.6b). In comparison to 10 kHz, 0.1 kHz showed a sharp decrease in |Zcells| values ranging from -543 to 171 Ω and displayed a negative correlation. In addition, 10,000 kHz showed increased |Zcells| values ranging from -9.1 to 16 Ω with a flatter slope than 10 kHz (Fig. 2.6b). To quantitatively assess the best-fit planes across frequency, the equation coefficients A, B, and C were compared at (0.1, 1, 10, 100, 1,000, and 10,000) kHz (Fig. 2.6 c-e). Coefficient A, representing the intercept of the planes, increased with frequency (Fig. 67 2.6c). Sensitivity to cell concentration was characterized by coefficient B, and sensitivity to viability was depicted by coefficient C. Both sensitivities showed low negative values at 0.1 kHz, then greatly increased and remained high from 1 – 1,000 kHz until decreasing at 10,000 kHz (Fig. 2.6d-e). This change in sensitivity also follows an improvement in goodness of fit values for best-fit planes from 1-10,000 kHz (Supplemental Table 2.S3). However, the goodness of fit for 0.1 kHz was low, suggesting that a different modeling technique is needed for low frequencies. By modeling samples with known concentration and viability on a plane, we have shown that concentration and viability have an additive effect on |Zcells| and that the degree to which this occurs is influenced by the frequency. Figure 2.5: 3-D representations of cell concentration, cell viability, and |Zcells| using 40 - 54 samples of primary chondrocytes suspended in PBS at controlled cell 68 concentrations and cell viabilities. a.) Best-fit plane with equation at 10 kHz with each point representing a sample average of 3 - 4 replicates and n = 54 samples. In the equation, x represents cell concentration, and y is viability. b.) Planes of best fit at (0.1, 10, and 10,000) kHz. The number of samples for each frequency are as follows: 0.1 kHz was n = 40, 10 kHz was n = 54, and 10,000 kHz was n = 54. c-e.) Best fit plane coefficients A, B, and C for frequencies (0.1, 1, 10, 100, 1,000, and 10,000) kHz. 2.5. Discussion Bioprinting is limited by the use of destructive techniques to determine construct quality. This greatly increases production costs and limits the scaling potential of bioprinting. Since the printing process may negatively affect cell bioactivity (62,149,150), non-destructive, in-line sensing of bio-ink would provide valuable assessment of cell properties that influence bioactivity such as cell concentration and viability. The objective of this study was to monitor cell concentration and viability in real-time onboard a syringe. To achieve this, we used the smart syringe, a device that incorporates DIS onto a syringe to measure the dielectric properties of |Z| and θ, and we monitored samples using frequencies from 0.1 - 25,000 kHz. Since bio-inks are typically saline-based, we assessed bulk cell suspensions in PBS to measure dielectric properties in a simplified environment (231). This study used concentrations ranging from 0.1×106 - 125×106 cells/mL, flow rates from 0.5 – 4 mL/min, and cell viabilities from 0 - 94%. Assessment of these parameters will determine the practicality of using DIS for real- time monitoring of cell concentration and viability during bioprinting. In addition, this study determined the combined effects of cell concentration and viability on |Zcells| using a planar model at several frequencies. Understanding how cell properties influence the dielectric measurements can shape how DIS is utilized for in-line monitoring in the 69 future. Bioprinting tissues can use a wide range of cell concentrations, depending on the application and shape of the construct (47,229). We demonstrated that large concentrations up to 125×106 cells/mL follow predictable and repeatable dielectric patterns in |Zcells| and θcells based on frequency. Previously, |Z| and other related dielectric properties such as permittivity have shown a linear, frequency-dependent relationships with cell concentration in 2-D cell culture and 3-D bioreactors using cell concentrations up to 10×106 cells/mL (167). Our findings expand upon this relationship, showing that the linear trend continues beyond 100×106 cells/mL onboard a syringe. In fact, the linearity of the relationship suggests that greater concentrations may follow the same trend. Notably, all measurements were collected using primary chondrocytes suspended in PBS rather than hydrogels that are typically used in bioprinting. While this created a simplified environment for monitoring cell properties in a conductive solution, |Z| measurements may change as a result of suspension medium. Measurement of cell concentration was dependent on the frequency. In an alternating current, the movement of ions within cells and in the surrounding solution have been previously described as the α-dispersion (<1 kHz) and β-dispersion (1 – 25,000 kHz) (20,25,236,237). The α-dispersion measures ionic diffusion, whereas the β-dispersion leads to cell polarization (25,236). In this study, sensitivity indicated frequency regions with the greatest potential for sensing cell concentration. Although the α-dispersion showed the highest |Zcells| values, low frequencies from 1 – 1,000 kHz within the β-dispersion held the highest sensitivity to changes in cell concentration for samples from 1×106 to 125×106 cells/mL. Interestingly, this range of frequencies 70 follows previous literature for monitoring cells despite occurring in a 3-D environment while onboard a syringe (25,238). Notably, some previous studies have found that frequencies greater than 10,000 kHz are effective for measuring cell properties such as viability and cell type for concentrations up to 0.2×106 cells/mL (195,239). However, frequencies from 10,000 - 25,000 kHz provided reduced sensitivity in this study using the smart syringe device. This difference in sensitivity could be linked to the concentration of cells or the design of our device. Data from the current study suggests that for cell concentrations greater than 1×106 cells/mL in a suspension, measuring |Zcells| and θcells from 1 - 1,000 kHz is recommended. In addition, to our knowledge, this is the first study to report changes in θcells with cell concentration. θ indicates cell response to the alternating current with negative values showing capacitive behavior and positive values describing inductive behavior. In their natural environment within cartilage, chondrocytes are sensitive to fluxes in local ion concentration caused by compression and relaxation of the joint (195). This responsiveness to ions in the local environment is related to θ and cell membrane capacitance (240). As frequency increased, θcells showed opposite effects for high and low concentration samples during the α-dispersion and β-dispersion. For example, high concentration samples ≥ 100×106 cells/mL showed inductive θcells at low frequencies, whereas low concentration samples < 75×106 cells/mL displayed capacitive θcells. This suggests that as cell concentration increases, the behavior of the cells becomes more pronounced in the α-dispersion and β-dispersion when compared to the surrounding solution. Extrusion of cell suspensions is crucial for bioprinting tissues, however DIS has not 71 previously been evaluated for in-line monitoring using flow rates relevant to tissue printing. To assess the practicality of monitoring cell concentrations with flow rate, two cell-laden solutions were extruded through the smart syringe at set time intervals using a 3-way stopcock. As cell concentration varied with time, |Zcells| changed on the order of seconds to reflect the concentration. Real-time changes in |Zcells| with cell concentration indicate the compatibility of DIS onboard a syringe for measurement of cell bioactivity during extrusion. In-line monitoring of samples during extrusion could benefit future bioprinting applications such as printing zonally complex tissues. For example, menisci require regions of high and low cell density for load transfer and to recapitulate native extracellular matrix (214,241), however, previous efforts to bioprint cell gradients rely on post-fabrication destructive testing to confirm the distribution of cells (242,243). Incorporation of the smart syringe device for bioprinting complex tissues would enable real-time monitoring of cell placement in the construct. Additionally, the rate of data collection could be hastened in future applications by reducing the frequency range. Previously, the influence of flow rate on |Z| and θ measurement has only been investigated for microfluidic devices using cell concentrations at 1×106 cells/mL with speeds up to 40 μL/min (193). In these microfluidic devices, increased flow rate led to increased |Z| and decreased θ (244). However, flow rates have not been investigated in macro-scale systems with speeds or durations relevant to bioprinting. This study found that flow rates from 0.5 – 4 mL/min while onboard a syringe exhibited no influence on |Zcells| or θcells over 240 s for cell concentrations ranging from 1×106 - 50×106 cells/mL. Because measurement quality was not affected by the flow rate, the duration of 72 extrusion, or cell settling, the smart syringe will likely support data collection throughout the bioprinting process, which can range from seconds to hours, using frequencies from 1 – 1,000 kHz. Cell viability is a crucial marker of cell bioactivity and plays a key role in the dielectric properties of the sample. DIS relies on the presence of a plasma membrane, such that dead cells with punctured or otherwise damaged membranes will not store charge as well as viable cells (7,168,183). Comparing samples pre- and post-testing revealed that exposure to DIS onboard a syringe did not alter sample viability. Upon examining the dielectric properties, |Zcells| was found to be dependent on cell viability and frequency; however, θcells was independent of cell viability. In theory, |Zcells| would be minimized for dead cells without a functioning plasma membrane with expected values near 0 Ω, indicating that dead cells have similar |Z| as the background media. However, low viability samples had the widest range of negative |Zcells|, indicating that dead cells showed greatly reduced |Z| compared to PBS. This response from dead cells has been reported previously and may be explained by the release of ions from the cell cytoplasm after death. Increasing ions within a solution would improve conductivity and reduce impedance (7,230). Simultaneously, medium and low viability samples showed increased cell clustering compared to high viability samples, which may have increased the impedance and contributed to the range in |Zcells| values. Though sample viability and |Zcells| shared an increasing relationship, there was no correlation with viability and θcells using the smart syringe. This suggests that viability does not change the capacitive or inductive behavior of the primary chondrocytes exposed to an alternating current in the smart syringe. In contrast, previous literature found that dead 73 cells showed higher θ than viable cells (7). This difference in θ behavior likely stems from the smart syringe’s comparatively lower sensitivity to this variable, suggesting that |Zcells| is a stronger predictor for cell concentration and cell viability than θcells. We found that both concentration and viability have a linear relationship with |Zcells| that is dependent on the frequency. Understanding of the combined effects of these cellular properties on |Zcells| would greatly improve the applications of DIS for real-time monitoring of bioprinting. To visualize the combined effect on |Zcells|, 54 samples were fitted to a planar model and displayed at 10 kHz. Interestingly, an additive effect was shown with both cell concentration and cell viability contributing to increased |Zcells|. Notably, samples at extreme points with low concentration (0.1 ×106 - 10×106 cells/mL) or low viability (0%) produced |Zcells| that were vastly lower than predicted by the planar model. This finding indicates that optimal samples for bioprinting with high cell concentration and viability will produce high |Zcells|, whereas sub-optimal samples will produce low |Zcells|. Using the best-fit plane as a real-time metric for optimal and sub-optimal samples may lead to a reduction of destructive testing used in bioprinting, however, this model could be improved further by accounting for samples at extreme points. Investigation of multiple best-fit planes revealed that |Zcells| increased with frequency, however higher |Zcells| values did not lead to improved sensitivity. Rather, sensitivity to cell concentration and viability were enhanced from 1 – 1,000 kHz. In fact, planar models ≥ 1 kHz were found to be reliable representations of the data, with the plane at 10 kHz showing the highest correlation. This finding highlights the importance of frequency on sample behavior and enforces that optimal sensing of both cell concentration and viability can occur at 74 10 kHz. Several studies have investigated cell dielectric properties at low frequencies during the α-dispersion. This region is plagued by noise caused by ions in the substrate coating the electrodes in a double layer that increases impedance (245–247). Known as electrode polarization, several techniques have been employed to reduce its influence on data quality such as adding electrode coatings, increasing electrode spacing, and using high frequencies (≥ 1 kHz) to avoid its effects. For this reason, studies often use 4 electrodes with 2 for imposing current and 2 for measuring impedance (248–250). While we only used 2 electrodes, this study subtracted the background PBS from the sample to reduce the effects of electrode polarization and to compare measurements across multiple devices with slight variations in electrode distance. While electrode distance ranged from 1 - 1.5 mm across electrode midpoints for the devices used in this study, no differences were noted when comparing |Zcells| and θcells. Interestingly, electrode polarization persisted at 0.1 kHz, resulting in an unreliable planar fit and suggesting that cells in suspension increased this effect, possibly resulting from ionic diffusion during the α-dispersion. The outcomes of these studies showcase the compatibility of DIS onboard a syringe for measurement of cell concentration and viability. Our device, the smart syringe, was sensitive to changes in cell properties with and without flow rate. These results indicate that DIS can provide in-line sensing of cellular properties while bio- ink is extruded during bioprinting. While samples of known concentration and viability were used in this study, future studies could use the planar models created in this study to predict the cellular properties of unknown samples using |Zcells|. We 75 anticipate that matching the |Zcells| values to planes at multiple frequencies from 1 - 1,000 kHz will result in a more accurate prediction. While our device is capable of measuring |Z| and θ in real-time, all analyses shown in this study were performed off- line after the experiments were conducted. Furthermore, a reference measurement of the background solution was required prior to monitoring cellular suspensions. These limitations highlight the need for an integrated system capable of generating waveforms across a frequency sweep then rapidly collecting, analyzing, and displaying dielectric measurements on the order of seconds. Integrating a system capable of rapid assessment of cell dielectric properties at multiple frequencies would bridge this time gap. Additionally, the models used in this study could be further improved by accounting for additional cell properties, such as cell type. Previous studies have found that cell types have a unique impedance signature across a frequency sweep (199,251). Machine learning algorithms could be trained to analyze cell behavior across multiple frequencies for rapid monitoring of bio-ink during printing. In addition to bioprinting, the smart syringe could be adapted to other fields that would benefit from in-line sensing of cell properties during extrusion or transportation, such as cell therapy or cell manufacturing. Optimization of DIS onboard a syringe for these fields would also benefit from an enhanced model that incorporates dielectric behavior across several frequencies. 2.6. Conclusion Overall, this study showed that DIS incorporated into a syringe is a suitable technique for real-time, in-line sensing of cell concentration and cell viability at flow 76 rates from 0.5 - 4 mL/min. This study used a custom device called the smart syringe to measure |Z| and θ onboard a syringe for primary chondrocytes suspended in PBS. |Zcells| showed a linear correlation with for cell concentrations from 1×106 - 125×106 cells/mL and with viabilities from 0 - 94%. In fact, these cellular properties had an additive effect on |Zcells| that could be modeled as a plane with high reliability at (1, 10, 100, 1,000, and 10,000) kHz. Sensitivity to cell concentration and viability using |Zcells| was greatest when θcells was centered about 0 degees from 1 - 1,000 kHz. When cell concentration was varied at specific time points with flow rate, |Zcells| changed on the order of seconds to reflect the sample. Although |Zcells| adjusted in real-time with cell concentration, varying sample flow rate from 0.5 – 4 mL/min did not affect data collection over 4 min. This study supports the use of DIS as an alternative technique to destructive testing for bioprinting. Additionally, this work suggests that cell concentration and viability of unknown samples can be predicted using a planar model for |Zcells| at different frequencies. 2.7. Acknowledgement We acknowledge funding from West Pharmaceutical Services. We also acknowledge the contributions of the Cornell Statistical Consulting Unit for assistance with statistical testing of the results. Some figures were created using BioRender.com. 2.8. Conflicts of Interest Didarul Bhuiyan, Alex Lyness, and Vivek Bhatnagar were employees of West 77 Pharmaceutical Services during the production of this manuscript. Dr. Bonassar is a co-founder of and holds equity in 3DBio Corp and is a consultant for Fidia Farmaceutici, SpA and Histogenics, Inc. 2.9. Supporting Information Q (mL/min) 0.5 1.0 2.0 4.0 0.5 1.0 2.0 4.0 t (s) 7.0 7.0 7.0 7.0 7.0 7.0 7.0 7.0 vsmart syringe (mL) 0.06 0.12 0.23 0.47 0.06 0.12 0.23 0.47 vgap (mL) 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 vcurrent (mL) 0.05 0.12 0.23 0.47 0.05 0.12 0.23 0.47 Table 2.S1: Calculations for volume (v) passing through the electrodes during each frequency sweep with time (t) at various volumetric flow rates (Q) assessed in this study. Q (mL/min) 0.5 1 2 4 0.5 1 2 4 d (cm) 0.10 0.10 0.10 0.10 0.05 0.05 0.05 0.05 T (cm) 0.09 0.09 0.09 0.09 0.09 0.09 0.09 0.09 A (cm2) 0.009 0.009 0.009 0.009 0.005 0.005 0.005 0.005 V (cm/s) 0.93 1.85 3.70 7.41 1.85 3.70 7.401 14.8 t residence (s) 0.097 0.049 0.024 0.012 0.049 0.024 0.012 0.006 Table 2.S2: Calculations of cell residence time in the smart syringe at flow rates from 0.5 - 4 mL/min and with electrode distance at 0.05 or 0.1 cm. Cell residence time (tresidence) represents the approximate time that a cell spends in the path of the electrodes within the smart syringe. 78 Figure 2.1S: Assembly of a smart syringe. a.) Materials needed include breadboard wires, machined female luer lock, male luer locks, gold pin electrodes, wire connections, heat shrink, and epoxy (not pictured). b.) Strip the wire from one end of the breadboard wires then attach and strip a gold pin electrode. c.) Remove the plastic casing from the other end of the breadboard wire and add a wire connection. d/e.) Add heat shrink to both ends. f.) Insert the gold pin electrodes into the female luer lock and twist the two luer locks together. Add epoxy to the insertion point for the electrodes. g.) Top-view of assembled smart syringe. (a) 79 Figure 2.2S: Raw |Z| and θ measurements for primary chondrocytes suspended in PBS with concentration at 1×106, 10×106, 25×106, 50×106, 75×106, 100×106, and 125×106 cells/mL with n = 3 per concentration. PBS without cells was collected to measure background/media signal and was shown in blue. Samples replicates were randomly measured. Shaded regions represent the standard deviation with the dark lines showing the sample average. (a) (Left) Frequency-dependent measurements of |Z| from 1 kHz to 25,000 kHz with replicates and (right) zoomed-in view of 10 – 10,000 kHz. (b) (Left) Frequency-dependent measurements of θ from 1 - 25,000 kHz and (right) zoomed-in view of 10 kHz – 10,000 kHz. 80 Figure 2.3S: Primary chondrocytes suspended in PBS at 100×106 cells/mL were added to a syringe and allowed to rest without extrusion for 600 s (10 minutes) in a syringe pump. |Z| displayed at 25 kHz with n = 3 trials for the cell-based sample. Data was normalized to t = 0 s for each replicate. Dark line represents sample average with shaded area showing standard deviation. Figure 2.4S: Samples with cell concentrations at 1×106, 10×106, 25×106 and 50×106 cells/mL with n = 3 per concentration were assessed at 0, 30, 60, and 90 s at 2 mL/min for a.) |Z| normalized to t = 0 s and b.) θ normalized to t = 0 s. 81 Figure 2.5S: Raw |Z| and θ measurements of primary chondrocyte samples suspended in PBS with high (~95%), medium (~50%) and low (0%) viabilities. Cell concentration remained at 25×106 cells/mL, frequency varied from 0.1 – 25,000 kHz, and n = 4 - 5 replicates per viability. PBS without cells was measured before and each sample for comparison. Samples replicates were randomly measured. Shaded areas represent sample standard deviation. a.) |Z| for samples of varying viability across the frequency sweep. Boxed region is investigated in (b-d). b-d.) Zoomed-in view of 10 – 1,000 kHz with each viability and its corresponding PBS measurements displayed individually. e.) θ across the full frequency sweep. Boxed region is showcased in (f-h). f-h.) Zoomed-in view of 10 kHz – 1,000 kHz for each viability and PBS pair. Table 2.S3: Fitting parameters for best-fit planes using Equation 3 by frequency. Color and asterisk indicate significance using R Critical Value tables with 0.05 significance for two-tailed tests. *p < 0.05, **p < 0.001, ***p < 0.0001 f (kHz) A (Ω) B (Ω mL / 10 6 cells) C (Ω / % viability) SSE R2 R 0.1 -16.70 -17.4 -0.388 8.40E+05 0.008 0.092 1 -4.90 12.2 0.080 6700 0.136  0.304* 10  -5.68  9.81 0.078 146 0.347 0.589*** 100 -4.58 9.39  0.056  1640  0.260 0.512*** 1,000  -4.61  8.84  0.058 1580 0.265 0.515*** 10,000  0.596  5.03  -0.070 877 0.282 0.495** Coefficients Goodness of Fit Fitting Parameters for 3-D Surface Plots 82 CHAPTER 3 ALGINATE BIOINK PROPERTIES INFLUENCE REAL-TIME IMPEDANCE MONITORING OF CELLS DURING EXTRUSION BIOPRINTING* 3.1. Abstract Bioprinting processes have greatly advanced in recent years through improvements in print accuracy and bioink optimization. Despite this progress, optimizing cell bioactivity still relies on guess-and-check processes with destructive testing post- printing. Measuring cell bioactivity during printing would improve print quality and inform complex printing processes, such as cell gradients or bioink transitions. Real- time monitoring using dielectric impedance spectroscopy alleviates this burden by correlating impedance |Z| to cell properties. However, the influence of bioink properties on these measurements is unknown. Using an in-line impedance sensor, we assessed the effect of alginate bioink concentration, pH, and crosslinking on impedance from 1 – 25,000 kHz and determined how these properties influenced the detection of primary chondrocytes. Increasing the alginate concentration, decreasing the pH, or crosslinking with CaCl2 resulted in an increase in impedance. In nearly all samples, the addition of cells resulted in an increase in impedance compared to acellular samples, and this difference in impedance was used to quantify cell presence, termed |Zcells|. Higher alginate concentrations at 1 w/v% and 3 w/v% showed greater *Matavosian AA, Griffin AC, Bonassar LJ. Alginate bioink properties influence real-time impedance monitoring of cells during extrusion bioprinting. Biofabrication. (Submitted). 83 |Zcells|, indicating reliable cell detection. Although |Zcells| varied greatly with alginate or PBS pH, similar measurements were found in pH resembling cell media. Optimal frequency ranges for monitoring acellular and cellular samples were from 10 – 100 kHz and 1,000 – 25,000 kHz. Furthermore, cells were detected in real-time as acellular and cellular alginate bioinks were transitioned during bioprinting. This transition in cell concentration was spatially mapped to deposited bioink, providing a visual display of bioink transition using impedance. In summary, DIS was capable of detecting cells suspended in alginate bioink and showed potential for real-time mapping of cell deposition. 3.2. Introduction Recent advancements in bioprinting technology have led to improved print accuracy through support baths (107,252,253), optimized bioink formulations through machine learning (254–256), correcting print defects (2,131,133), and printing full-scale organs (10,12,107). Despite these major advancements, bioprinting is still limited to guess-and- check processes to assess print quality. Print quality is affected by multiple factors, including process parameters (bioink batch, bioink formulation), printing parameters (extrusion rate, temperature, layer height), and cell bioactivity (cell density, viability, type, proliferation). Of particular note is the influence of bioprinting processes on cell bioactivity, which may affect the regenerative properties of the construct. For instance, bioprinting may decrease cell viability through shear stress (62,63,177) or influence ECM production and cell signaling through heterogeneous cell distribution caused by bioink mixing and cell settling (152,153,155,157). As a result, assessment and 84 optimization of cell bioactivity is a high priority in printed constructs. Real-time monitoring of cell-laden bioink enables the rapid quantification of crucial cell parameters during the printing process. Dielectric impedance spectroscopy (DIS) is a real-time monitoring method that uses an alternating current to non-destructively polarize cells (24,25,198). Cell polarization is caused by ion reorganization both inside cells and in the surrounding substrate and is quantified as impedance |Z|. Using |Z|, cell properties such as cell concentration (5,24,167,168), viability (5,7,182), type (7,197), and state of differentiation (199,200) were tracked over time. DIS has previously been utilized to monitor bioprinted constructs post-fabrication (167,168,196,211) and has more recently been incorporated into a syringe tip as an in-line sensor (5,7). Although cells have been assessed using DIS previously, the influence of bioink on |Z| measurements have not been explored. Bioink properties are adjusted to control the material properties of the printed construct (27,28). For instance, sodium alginate is commonly used in printing due to its biocompatibility and tunability (58,69). Changing the concentration, pH, or crosslinking of the alginate alters the viscosity, gelation, construct mechanics, and cell viability of the printed construct (68,69,257). In addition to adjusting the construct, changes in alginate concentration, pH, or crosslinking may inadvertently result in alterations to ion movement that would influence |Z| measurements (Fig 3.1). Alginate is an anionic polyelectrolyte that is negatively charged in solution due to the availability of carboxyl groups (69,257,258). As such, increasing alginate concentration would increase the quantity of excess Na+ and lower |Z|. Conversely, higher concentrations of alginate are more viscous and would increase |Z| (259–261) (Fig. 3.1a). Similarly, as alginate pH decreases, protonation of 85 carboxyl groups decreases counter-ion concentration and increases viscosity (68,262) (Fig. 3.1b). In this case, counter-ion concentration may increase |Z|, whereas viscosity would decrease it (Fig. 3.1b). In contrast, alginate crosslinking results in a physical change to the conformation of the chains to create an egg-crate model, increasing |Z| by decreasing counter-ion concentration and increasing viscosity (263,264) (Fig. 3.1c). Elucidating the effect of these parameters that are commonly used to tune alginate bioinks on |Z| would enable the use of real-time impedance monitoring of cells suspended in a variety of bioink formulations. We previously designed a device with in-line impedance sensors that attach to commercial syringes, referred to as the smart syringe (5,9). This previous work demonstrated the ability to monitor cell deposition in saline. However, the influence of bioink physical and biochemical attributes on |Z| are unknown. The goals of this study are to (1) understand the effects of a.) alginate concentration, b.) pH, and c.) crosslinking on the |Z| of acellular and cellular samples and (2) demonstrate real-time cell detection during printing. 86 Figure 3.1: Bioink properties influence |Z| measurements by altering the movement of ions in solution. In alginate bioink, changes to concentration, pH, or crosslinking lead to effects in counter-ion concentration and viscosity that have opposing influences on |Z|. Counter-ions (Na+) are present in solution due to the poly-anionic (-) nature of alginate. a.) As the concentration of alginate bioink increases, the number of alginate chains rises, increasing viscosity and counter-ion concentration. These changes in viscosity and ion balance have opposing effects on |Z|. b.) The anionic polymeric chains in alginate bioink become protonated as the sample becomes acidic, reducing counter-ion concentration. In addition, viscosity is decreased through the reduction of repulsive forces between negatively charged chains. c.) With the addition of CaCl2, alginate bioink is crosslinked into an egg-box structure, increasing viscosity and decreasing Na+ ions. This shift in conformational structure is anticipated to increase |Z|. 87 3.3. Materials and Methods 3.3.1. Smart Syringe creation and DIS parameters Smart syringes were produced using previously established methods (5). To summarize, custom machined female luer locks (Qosina, Ronkonkoma, NY) were press-fit with two antiparallel gold electrodes (RadioShack Tech Plus) with 1.0 ± 0.1 mm spacing between the electrodes. The ends of the gold electrodes were crimped onto stripped AWG copper wires (Amazon, Seattle, WA), and heat shrink (Digikey, Thief River Falls, MN) was used as insulation. The female luer lock with electrodes was screwed onto a male luer lock (Qosina), and all connections were sealed using gorilla glue epoxy (Amazon). All devices were assessed for functionality using 1X phosphate- buffered saline (PBS) (VWR, Radnor, PA) prior to use. Smart syringes measure |Z| of alginate bioinks in real-time. An alternating current was sent to the smart syringe devices using a Digilent Analog Discovery 2 device (Digilent, Pullman, WA). |Z| was monitored in alginate concentration, pH, and crosslinking studies using frequency sweeps from 1 – 25,000 kHz with 18.5 points per decade and 101 steps from the initial to the final frequency. To increase the sampling rate for extruded bioink, the frequency sweep was shortened to 1 – 1,000 kHz for syringe pump studies and 10 kHz for bioprinting studies with 8.5 measurements per second. All experiments used a voltage of 50 mV with zero DC offset, 1 kΩ resistance, and 100 ms averaging. Data collection involved loading each sample, measuring |Z|, and cleaning the device. Samples were individually loaded into the smart syringe device, and |Z| was 88 collected when there was full contact between the sample and electrodes. To reduce electrode fouling, devices were cleaned after each measurement by leaving the electrodes in contact with 1% acetic acid for 3 minutes or more, followed by washes with PBS, 70% ethanol, and PBS again. 3.3.2. Primary chondrocyte preparation Primary chondrocytes were isolated from a total of 37 neonatal bovine stifle joints (Copper City Meats, Rome, NY), as described previously (229,232). Chondrocytes were selected in this study due to their previously investigated dielectric properties as well as their widespread applications in bioprinting (5,195,265,266). Condyles were dissected and chopped into approximately 3 mm3 cubes. These cartilage cubes were digested overnight in 0.2-0.3% w/v collagenase solution (DMEM 4.5 g/L glucose with L- glutamine (VWR), 100 U/ml penicillin (VWR), 100 µg/mL streptomycin (VWR), and 250 ng/mL amphotericin B, and 2-3 mg/mL collagenase type 2 (Worthington Biochemical, Lakewood, NJ)). After digesting for 16 - 20 hours, the cell and collagenase solution was pipetted through a 40 μm cell strainer (VWR) and centrifuged at 600 g- force for 12 minutes. Next, the supernatant was aspirated, and the cell pellet was washed with PBS and centrifuged at 400 g-force for 9 minutes. The supernatant was aspirated again, and the wash step was repeated with PBS. Cells were counted using a hemacytometer (Hausser Scientific, Horsham, PA) with trypan blue (VWR) and light microscopy (Nikon Eclipse TS100). Before testing, all samples remained on ice until 20 minutes before testing, during which they were allowed to thaw to room temperature. Notably, chondrocytes preserve viability when suspended in saline at 4°C and have 89 shown only a 25% decrease in viability after 3 days (233). 3.3.3. Alginate concentration studies Alginate concentration tunes the printability of the bioink and the mechanics of the printed construct. Alginate was prepared with final concentrations at 0.5, 1, and 3 w/v% to determine the effect of concentrations commonly used in bioprinting on |Z|. Pronova UP LVG sodium alginate powder (Novamatrix, Berlin, Germany) was reconstituted in 1X PBS to reach the final concentration. Alginate was sterilized using an 8 μm syringe filter (VWR) at low concentration, then lyophilized and reconstituted at the desired stock concentration. Primary chondrocytes were suspended into alginate at 25 × 106 cells/mL. This concentration of cells was selected because it was well-representative of concentrations used for tissue engineering and showed reproducible results in prior studies (5,47,170). There was a total of 11 - 16 measurements per alginate concentration for acellular and cellular bioinks that were collected across four independent experiments. Independent experiments were conducted on different days using different cell batches and smart syringe devices. To reduce the creation of air bubbles in 1 and 3 w/v% alginate bioinks, samples were mixed using a combination of low-speed vortexing and manual stirring. After monitoring |Z|, the cellular bioinks were stained using LIVE/DEAD® assays (Thermofisher, Waltham, MA), and imaged at 200X magnification using a Keyence BZ-X810 microscope with 3 - 4 images collected per sample. Scale bars were added to these images using ImageJ. 3.3.4. Sample pH studies Sample pH is crucial for cell viability and bioink printability. 1 w/v% alginate-based 90 and PBS-based samples were prepared with pH ranging from 5.5 – 7.7 to determine the effect of pH on |Z| measurements. pH was confirmed for all samples using pH test strips and a double-junction pH probe (EW-55510-18 Cole-Parmer, Vernon Hills, IL). Cell- laden samples used primary chondrocytes suspended at 25×106 cells/mL in either alginate or PBS. There was a total of 3 - 8 measurements per pH for acellular and cellular samples collected across independent experiments. After measuring |Z|, the cellular bioinks were stained using LIVE/DEAD® assays (Thermofisher, Waltham, MA), and imaged at 200X magnification with 3 - 4 images collected per sample. Scale bars were added using ImageJ. 3.3.5. Alginate crosslinking and bioprinting Crosslinking enables bioink to form layers in a printed construct and is vital to the printing process. To measure the influence of crosslinking on |Z|, acellular alginate with and without crosslinker were compared to PBS (n = 4). Alginate was crosslinked by mixing with 30 mM CaCl2 in a 7:3 volume ratio, which has previously shown optimal printability (58). Bioprinting complex geometries may involve alternations or transitions between acellular and cellular bioinks (124,181,267). To measure the presence or absence of cells during extrusion, crosslinked alginate samples were monitored during transitions with cell concentrations at (0, 25, and 100) ×106 cells/mL. Five bioink transitions were tested: (1) 0×106 cells/mL to 25×106 cells/mL, (2) 0×106 cells/mL to 100×106 cells/mL, (3) 25×106 cells/mL to 0×106 cells/mL, (4) 100×106 cells/mL to 0×106 cells/mL, and (5) PBS to PBS (n = 3 - 5). Samples were loaded into 10 mL BD plastic syringes (VWR), 91 passed through Tygon® tubing (McMaster-Carr) into a three-way stopcock (MoHawk Medical), and flowed through the Smart Syringe device, as previously shown (5). Extrusion was set to 0.5 mL/min (8.3 mm3/s) using a dual-channel syringe pump (Harvard Apparatus). The first sample flowed for 70 seconds, mixed with the second sample, then was replaced by the second sample with a total time of 210 seconds. Data collection began immediately after the first droplet exited the Smart Syringe. Real-time |Z| measurements were used to spatially map the deposition of transitioning bioinks. A line construct was printed using a syringe loaded with samples at 0×106 cells/mL and 25×106 cells/mL with n = 3. To load two samples into a single syringe, the acellular bioink was front-filled into a syringe. Then the 25×106 cells/mL bioink was back-loaded into the syringe by removing the plunger and adding the sample (124). A thin wire was placed into the syringe while the plunger was reinserted, allowing air to escape. To improve visibility, the 25×106 cells/mL bioink included fluorescein isothiocyanate (FITC), which was previously shown not to affect cell viability (47). Images were collected of each print using an iPhone 15 Pro. 3.3.6. Data analysis |Z| measurements were collected for acellular and cellular samples. As previously described, the change in |Z| between these samples was calculated as |Zcells| using Equation 5 (5). This value quantified the change in |Z| caused by the presence of cells as well as interactions between the cells and substrate. |Zcells| also enabled the standardization of measurements between independent experiments and different smart 92 syringe devices. |𝑍𝑐𝑒𝑙𝑙𝑠| = |𝑍𝑐𝑒𝑙𝑙𝑢𝑙𝑎𝑟 𝑠𝑎𝑚𝑝𝑙𝑒| − |𝑍𝑎𝑐𝑒𝑙𝑙𝑢𝑙𝑎𝑟 𝑠𝑎𝑚𝑝𝑙𝑒| (5) To determine the sensitivity of |Z| to critical process parameters, |Z| was fit to a linear function with alginate concentration and sample pH. Meanwhile, the sensitivity of |Zcells| was fit to a linear function with alginate concentration and a quadratic function for sample pH. Sensitivity was defined as the change in |Z| or |Zcells| as a result of alginate concentration or sample pH. Using acellular samples, sensitivity to alginate concentration was calculated through the derivative of the linear correlation between concentration and |Z| at each point from 1 – 25,000 kHz. Meanwhile, the sensitivity to sample pH was calculated using the derivative of the second-order polynomial at each frequency from 1 – 25,000 kHz for alginate-based and PBS-based acellular samples. Bioprinting and syringe pump studies involving flow rate were analyzed by normalizing |Z| to t = 0 seconds to highlight changes over time. For bioprinted constructs, the deposition of bioink was correlated to real-time measurements of |Z| based on the flow rate. A heatmap showing this change in |Z| was overlaid onto the image of the construct using a custom MATLAB code. This code averaged 4 |Z| measurements per 3 mm2 of imaged construct to visualize changes in |Z| during bioprinting. 3.3.7. Statistics Assessments at 50 kHz used a 2-way ANOVA with Tukey post-hoc for comparing between acellular samples and Šídák's multiple comparisons test for comparing between acellular and cellular samples. Comparisons between PBS, alginate, and crosslinked 93 alginate were conducted using a one-way ANOVA with Tukey post-hoc. Data was displayed at 50 kHz because this frequency displayed stable values with high sensitivity and low standard deviation. In all experiments, replicates that were greater than two standard deviations above the sample mean were excluded from analysis as outliers. 3.4. Results 3.4.1. Alginate concentration |Z| was dependent on both alginate concentration and frequency. Samples at 0.5, 1, and 3 w/v% alginate were monitored from 1 – 25,000 kHz to determine the effect of alginate concentration on |Z| (Fig. 3.2a). For all samples, |Z| decreased with frequency until reaching a nearly constant value at approximately 20 kHz. In addition, at 10 – 25,000 kHz, |Z| decreased on average with alginate concentration (Fig. 3.2a). At 50 kHz, 0.5 w/v% and 1 w/v% alginate bioinks showed higher |Z| than 3 w/v% (p < 0.0001) (Fig. 3.2b). Comparing acellular and cellular samples, on average |Z| was higher for cell-laden bioinks at 1 w/v% and 3 w/v% alginate (p > 0.1) (Fig 3.2b). Using Equation 5, |Zcells| was calculated as the difference between cellular and acellular samples across the frequency sweep (Fig. 3.2c). |Zcells| revealed the concentrations of alginate bioink that were amenable to detecting cells as well as their optimal frequency ranges. Particularly from 10 – 25,000 kHz, |Zcells| for 1 w/v% and 3 w/v% alginate showed positive values on average, indicating a higher capacity for monitoring cells at these concentrations (Fig. 3.2c). At 50 kHz, independent repetitions of this study supported the use of 1 w/v% and 3 w/v% alginate for monitoring cells, displaying on average |Zcells| at 6.22 Ω and 2.48 Ω, respectively (Fig. 3.2d). Notably, |Zcells| at 0.5 w/v% alginate showed wide variation with an average value of -0.442 Ω, 94 showing low compatibility with detecting cells. LIVE/DEAD staining revealed similar cell concentrations and viabilities between all alginate concentrations, indicating that the |Z| measurements were based on alginate concentration rather than these cell properties (Fig. 3.1S). Sensitivity was calculated at each frequency as the derivative of |Z| or |Zcells| as a function of alginate concentration. Frequency regions with narrow variations in standard deviation and greater distance from the minima indicated increased change in |Z| or |Zcells| as a result of alginate concentration. Additionally, regions with stable sensitivity provided reliable |Z| measurements with greater repeatability. Taking these factors into account, sensitivity for |Z| and |Zcells| both showed stable frequency ranges from 10 – 25,000 kHz (Fig. 3.2e and 3.2f). 95 Figure 3.2: Acellular and cellular (25×106 cells/mL) alginate bioinks with concentrations at 0.5, 1 and 3 w/v% were monitored using |Z| with frequency sweeps. Dark lines represented sample average and shaded regions showed standard deviation. a.) |Z| of acellular alginate was measured from 1 - 25,000 kHz with n = 4 per concentration. Boxed section depicted the region from 10 – 1,000 kHz with a dotted line representing 50 kHz. b.) At 50 kHz, |Z| was compared between acellular and cellular samples with n = 4 per alginate concentration. c.) To quantify the degree to which cells could be measured in each bioink concentration, |Zcells| was calculated from 1 - 25,000 kHz with n = 4. Brackets visualize regions with stable |Zcells|. d.) |Zcells| across multiple independently conducted experiments were amassed at 50 kHz with n = 11-16 per concentration. e,f.) Sensitivity to alginate concentration (|Z| Alg) and to cell detection (|Zcells| Alg) were determined as a rate of change and plotted as a function of frequency with n = 3. (** indicated p<0.01 and **** indicated p < 0.0001) 96 3.4.2. Sample pH pH of both alginate-based and PBS-based samples greatly influenced |Z| and |Zcells|. Acellular and cellular samples suspended in either PBS or alginate were monitored from 1 – 25,000 kHz to isolate the effect of pH on |Z| for both solutions. |Z| sharply decreased below 60 kHz and plateaued at higher frequencies (Fig. 3.3a). On average |Z| for PBS samples decreased as pH increased (Fig. 3.3a). At 50 kHz, acellular PBS at 6.5 pH showed higher |Z| than samples with pH at 7.0 or 7.4 (p < 0.01) (Fig. 3.3b). In addition, all cell-laden samples showed an increase in |Z| (6.5 pH p = 0.02, 7.0 pH p = 0.0007, and 7.4 pH p = 0.002) (Fig. 3.3b). To determine if sample pH affected cell detection, |Zcells| was determined using Equation 5. For all frequencies tested, |Zcells| for PBS samples were positive on average, indicating compatibility for monitoring cells; however, |Zcells| experienced high levels of noise from 100 – 1,200 kHz (Fig. 3.3c). Similar to PBS, alginate pH was correlated to |Z| with frequency dependence. |Z| increased as alginate pH decreased (Fig. 3.3d). This trend was further supported at 50 kHz using acellular samples (p < 0.01 between acellular samples) (Fig. 3.3e). Adding cells increased |Z| on average in all alginate pHs, however this difference was reduced for bioink samples at 6.7 pH in comparison to 5.5 and 7.7 (5.5 pH p < 0.0001, 6.7 pH p = 0.2, and 7.4 pH p = 0.0001) (Fig. 3.3e). In fact, |Zcells| for alginate at 6.7 pH continued to show lower values on average compared to 5.5 and 7.7 from 4 – 25,000 kHz (Fig. 3.3f). Notably, nearly all |Zcells| values for alginate bioink at all tested pHs were positive, showcasing the ability to monitor for cell presence (Fig. 3.3f). In both PBS and alginate, differences in sample pH influenced the ability to monitor cells. To elucidate the relationship between sample pH and |Zcells|, independent 97 trials were performed across 5.5 - 7.7 pH. Interestingly, |Zcells| for PBS and alginate samples displayed the same average value of 2.0 Ω at 7.3 pH, which represents typical cell culturing media (268); however, as pH deviated from this point, |Zcells| depended on solution (Fig. 3.3g). PBS-based samples showed a negative second order function with |Zcells| reaching their peak at pH 7.0 (Fig. 3.3g). Meanwhile, alginate-based samples were fitted to a positive second order function with |Zcells| increasing as pH deviated from 7 (Fig. 3.3g). LIVE/DEAD staining revealed that cell viability drastically decreased when suspended in PBS with low pH when compared to alginate (Fig. 3.2S). Sensitivity was calculated at each frequency as the derivative of |Z| or |Zcells| with respect to pH. This value identified ranges of frequency that provided consistent and reliable measurements. Sensitivity of |Z| to changes in acellular alginate pH or PBS pH showed similar behavior overall with consistent measurements from 10 – 25,000 kHz (Fig. 3h). Because |Zcells| as a function of pH followed a quadratic function, sensitivity for |Zcells| was determined at specified pHs. At 7.3 pH, alginate and PBS showed the greatest sensitivity from 1 – 10 kHz, however measurements were more stable from 10 – 100 kHz and 1,000 – 25,000 kHz (Fig. 3.3i). In comparison, at pHs ranging from 5.8 – 7.6, sensitivity was shown to increase as pH deviated from 7 (Fig 3.3S). 98 Figure 3.3: |Z| was monitored across pH ranges in PBS (from 5.8 - 7.7) and alginate bio-ink (from 5.5 - 7.7) for acellular and cellular (at 25×106 cells/mL) samples. Dark lines represented sample average and shaded region showed standard deviation. a.) |Z| was monitored for acellular PBS samples using pH at 6.5, 7.0, and 7.4 from 1 - 25,000 99 kHz with n = 3 - 4 per pH. Boxed section depicts the region from 10 – 1,000 kHz with a dotted line representing 50 kHz. b.) Comparison of the |Z| values at 50 kHz for cellular and acellular PBS at different pHs with n = 3 - 4 per sample. c.) |Zcells| was calculated for PBS samples from 1 - 25,000 kHz with pH ranging from 6.5 - 7.4 and n = 4. Brackets indicated frequencies with greater signal stability and reduced variability. d.) Similar studies were performed for alginate samples. |Z| was measured for acellular alginate samples from 1 - 25,000 kHz with pH at 5.5, 6.7, and 7.7 and n = 4. Boxed section depicts the region from 10 – 1,000 kHz with a dotted line representing 50 kHz. e.) Comparison of the |Z| values for cellular and acellular alginate bio-ink at 50 kHz with n = 4. f.) |Zcells| was calculated for alginate bioink from 1 - 25,000 kHz with pH ranging from 5.5 - 7.7 and n = 4. Brackets visually depicted frequency regions with greater stability. g.) Comparison of |Zcells| for alginate bioink and PBS across multiple independent studies at 50 kHz with each point displaying the average of n = 3 - 8 measurements with standard deviation bars. Models were fit to the dataset using nonlinear 2nd order polynomials. The pH region for cell media is highlighted from 7.2 - 7.4. h.) Sensitivity of |Z| to sample pH (PBS or alginate) for acellular samples with n = 3. i.) Sensitivity of |Zcells| to sample pH (PBS or alginate) at a pH of 7.3, referencing the pH of typical cell media. (* indicated p<0.05, ** indicated p<0.01, *** indicated p < 0.001, and **** indicated p <0.0001) 3.4.3. Alginate crosslinking and bioprinting The previous experiments were performed in the absence of a crosslinker. Acellular alginate with and without CaCl2 crosslinker and PBS were compared to monitor the effect of crosslinker on |Z| (Fig. 3.4a). Initially, |Z| sharply decreased and leveled out by 60 kHz, remaining at a nearly constant value until 25,000 kHz (Fig. 3.4a). At 50 kHz, crosslinked alginate showed higher |Z| compared to uncrosslinked alginate (p = 0.03) (Fig. 3.4b). 100 Figure 3.4: |Z| of alginate bio-inks with CaCl2 crosslinker was measured immediately after mixing without flow (0 mL/min). Dark lines represented sample average, and the shaded region showed standard deviation. a.) To monitor changes in |Z| as a result of crosslinker, alginate with and without CaCl2 were compared across a frequency sweep from 1 - 25,000 kHz with PBS acting as a control group (n = 4). The boxed section depicts the region from 10 - 1,000 kHz with a dotted line representing 50 kHz. b.) Comparison of the |Z| values at 50 kHz with n = 4 per sample. After assessing the influence of alginate concentration, pH, and crosslinking on |Z|, these parameters were optimized to monitor cells in real-time during bioprinting. Using 1 w/v% alginate at 7.3 pH, acellular and cellular bioinks were transitioned in real-time on a syringe pump and a bioprinter. Cell transitions were first tested in a syringe pump 0.5 mL/min (Fig. 3.5a). As acellular bioink transitioned to cellular bioinks at either 25×106 cells/mL or 100×106 cells/mL, normalized |Z| increased (p = 0.1 and p = 0.001, respectively comparing start and end points) (Fig. 3.5bi). In contrast, when the process was reversed, with cellular bioink at 25×106 cells/mL or 100×106 cells/mL transitioning to acellular bioink, normalized |Z| decreased on average (p = 0.02 and p = 0.01, respectively comparing start and end points) (Fig. 3.5bii). The PBS control group was shown to slightly decrease in normalized |Z| over time, highlighting the stark difference 101 in behavior when cells were added or removed from solution (Fig. 3.5b). Bioprinting relies on the accurate spatial deposition of printed material. Adding or removing cells from a solution may be a necessary part of the printing process based on printing technique, equipment limitations, or construct application. However, due to bioink mixing, the deposition of cell-laden material is difficult to predict. To gain spatial resolution on the deposition of acellular and cellular bioinks, two samples were loaded into a syringe and printed into a line construct (Fig. 3.5c). A heatmap of normalized |Z| showed the increase in values as the bioink transitioned from 0×106 cells/mL to 25×106 cells/mL (Fig. 3.5d). In addition, the mixing region in the center of the construct is clearly depicted through normalized |Z| (Fig. 3.5d). Figure 3.5: Cell presence in alginate with CaCl2 was monitored in real-time using |Z| by adding or removing cells during extrusion from a syringe pump and bioprinter. 102 Alginate bioink was at 1 w/v% with pH 7.3. a.) Experimental set-up for extruding acellular and cellular bioink using a syringe pump with the in-line smart syringe impedance sensor. b.) |Z| was measured as acellular and cellular bio-ink samples were transitioned in real-time with n = 5-6 per trial. Flow rate was set to 0.5 mL/min, and frequency was 1 kHz. PBS served as an acellular control. (b.i) Acellular alginate with CaCl2 was monitored using |Z| for 70 seconds, then was transitioned for cellular bioink (25×106 cells/mL or 100×106 cells/mL). Dotted lines represent the timed transition from acellular to cellular bioink during which samples mixed. (b.ii) The process was reversed with cellular samples (25×106 cells/mL or 100×106 cells/mL) transitioning to acellular bioink over 210 seconds. Dotted lines represent the mixing region. c.) Experimental set-up for printing acellular and cellular bioink from a single syringe with the mounted smart syringe. d.) Printed construct visually depicting the transition from acellular to cellular (25×106 cells/mL) bioink at 1 kHz. A heatmap depicting the shift in normalized |Z| was spatially mapped to the printed construct. 3.5. Discussion Bioprinting has seen many improvements over recent years, but assessment of cell bioactivity still relies on guess-and-check processes that occur post-printing. In-line sensors enable real-time measurements of cell properties during bioprinting. Although DIS has previously shown correlations between impedance and cell properties such as density, viability, and type, the influence of bioink properties on these measurements was unknown (5,7,167,168). The objective of this study was to understand the effects of alginate concentration, pH, and crosslinking on the impedance of acellular and cellular samples as well as to demonstrate real-time monitoring of cells during bioprinting. These studies show that changes in alginate formulation influence impedance and that cells are detectable across a variety of such formulations. For instance, increasing alginate concentration, decreasing pH, or crosslinking with CaCl2 led to an increase in impedance. In nearly all cases, adding cells to a sample further increased impedance, and measuring this increase enabled the detection of cells. These 103 results showcase the potential of monitoring cell-laden bioinks during printing. Many studies demonstrate the tuning of alginate concentration to enhance construct mechanics and print accuracy (58,257,260). Monitoring cells suspended in a variety of alginate concentrations requires an understanding of how bioink concentration influences impedance. This study found that as alginate concentration increased, impedance decreased. This change in impedance indicated that counter-ion concentration was the dominating influence over impedance rather than viscosity, suggesting that bioink viscosity can be adjusted without affecting cell detection. Cell- laden samples were shown to increase impedance compared to acellular samples, which aligned with previous studies (5,24,167). Interestingly, |Zcells| was higher at greater alginate concentrations. This difference in |Zcells| may have resulted from the chondrocytes responding to changes in osmolarity caused by excess Na+ ions from increasing alginate concentration. Because chondrocytes originate from negatively charged cartilage, they express a wide array of ion channels to maintain homeostasis rapidly (269,270). Osmolarity has been shown to either stimulate chondrocyte bioactivity or cause cell death (145,271,272). As a result, chondrocyte ion channel activity in different alginate concentrations may have led to changes in polarization. Sample pH is crucial for cell bioactivity and viability (88,89). Lowering the pH in either PBS-based or alginate-based solutions led to an increase in impedance, showing that counter-ion concentration was the main contributor to impedance in either solution. Our previous study monitored a wide range of chondrocyte concentrations suspended in PBS and showed that |Zcells| at 25×106 cells/mL ranged from 1.1 – 6.9 Ω at 25 kHz with a pH of 7.3 (5). In comparison, at 50 kHz, |Zcells| at 104 7.3 pH showed slightly lower values in alginate ranging from -0.63 – 5.0 Ω and reduced variability in PBS ranging from 2.6 – 3.0 Ω. Notably, cells were detectable in both solutions on average. Similar to alginate concentration, impedance increased at most pHs with the addition of cells. However, the degree to which impedance increased depended on pH, frequency, and solution. Alginate and PBS showed opposite behavior in |Zcells| as the pH deviated from physiological pH (6.9 - 7.2) and from typical cell media (7.2 - 7.4) (175,268). This difference in behavior was likely caused by chondrocyte ion channel activity and viability. LIVE/DEAD staining revealed that at 5.8 pH, PBS showed drastically lower cell viability compared to acidic cell-laden alginate samples. As previously reported, cell death released ions into the solution, reducing impedance and resulting in negative |Zcells| (5,7,230). This change in |Zcells| indicates that cells suspended in PBS can be reliably detected in solutions from 7.0 - 7.4 pH, but not in solutions <6.5 pH. In contrast, cells encapsulated in alginate may have preserved viability on the timescale of hours, potentially due to buffering by the carboxylic groups on alginate (172,273). Nevertheless, acidic conditions may have damaged chondrocyte ion channels, resulting in an increase in |Zcells| as pH deviated from media pH (175). Collectively, these findings emphasize tuning the detection system to the bioink to optimize cell detection. Changes in sensitivity across the dielectric spectrum may provide an opportunity to differentially monitor bioink and cells. The optimal regions for detecting cells were from 10 – 100 kHz and 1,000 – 25,000 kHz. These regions correspond to the β-dispersion in which cells experience polarization through 105 interactions between the cytoplasm and surrounding substrate in an alternating current (25). Impedance spectra also contain an α-dispersion at frequencies <10 kHz, which monitors the diffusion of ionic species such as counter-ions (25,274). As a result, polyanionic solutions experience heightened sensitivity in the α-dispersion. Combining α- and β-dispersions enables the measurement of changes in bioink as well as the detection of cells. Using parameters that were optimized from the experiments on alginate concentration and sample pH, cells were detected in real-time during printing. In addition to introducing a conformational change in crosslinked alginate that increased viscosity, the quantity of counter-ions was decreased. While it is unknown whether viscosity or counter-ions had greater influence over impedance, combined, both factors resulted in increased impedance for crosslinked alginate. One of the goals for real-time monitoring is to assess changes in bioink cell concentration. In monitoring transitions between acellular and cellular samples, impedance increased when cell-seeded bioink moved through the sensor. Additionally, this increase was heighted for samples at higher cell concentrations. When extruding cell-laden bioink that transitioned into acellular bioink, impedance decreased as cells were removed from the sensor. In both cases, real-time changes in impedance reflected changes in cell concentrations over time. This technology enables heightened control over cell deposition such that complex constructs that require cellular gradients such as cartilage (214,215) or skin (48,185,190) may be printed using controlled cell concentrations from real-time monitoring and feedback. This potential is supported by the spatial mapping of normalized impedance over the course of a printed construct. 106 By correlating impedance measurements to cell concentration in real-time, on-the-fly corrections can be made to the printed construct, improving print reproducibility and quality (2). A limitation in this study was electrode fouling caused by alginate. Fouling can result in inaccurate impedance measurements due to alginate coating the electrodes in a layer of ions (275). To limit the effects of fouling in this study, the electrodes were cleaned between each measurement, and new smart syringe devices were used for independent trials that were conducted on different days. To compare measurements across devices and trials, normalization (|Zcells| or impedance / |Zt=0|) was required. Addressing the challenge of electrode fouling would increase device longevity and improve data sensitivity over time. The outcome of these studies showed the vast potential of monitoring acellular and cellular alginate using DIS. Our device, the smart syringe, monitored changes in impedance in a variety of alginate bioink formulations, and it showed real-time detection of cells during bioprinting. While samples of known cell concentration were used in this study, future endeavors may determine cell density through impedance and use these values to control the output. Notably, although impedance was measured in real-time, analyses shown in this study were performed off-line. Combining an in- line impedance sensor with a PID system capable of generating waveforms, rapidly analyzing data, and tuning deposition based on a target concentration would address this limitation and enable greater control of printed constructs. 107 3.6. Conclusion Cell bioactivity is a crucial aspect of construct quality, yet monitoring cell properties during bioprinting is limited. This study used an in-line impedance sensor to examine the influence of alginate concentration, pH, and crosslinking on impedance and to demonstrate real-time monitoring of cell-laden bioink during bioprinting. Impedance increased as a result of increasing alginate concentration, decreasing pH, or crosslinking with CaCl2. During these studies, cell-laden samples exhibited increased impedance compared to acellular samples, demonstrating the detectability of cells in various alginate bioink formulations. The application of this technology was showcased through real-time monitoring of cell-laden and acellular bioinks as they were transitioned during bioprinting. This work shows that real-time monitoring of bioinks can provide insight into cell properties during deposition, suggesting that these properties can be controlled with the integration of feedback. 3.7. Acknowledgement We wish to acknowledge Cornell Engineering Learning Initiatives for partially funding this research. In addition, we acknowledge the efforts of Mariana D. Rodriguez for her contributions towards creating experimental methods for the alginate concentration study as well as generating preliminary data. 3.8. Conflicts of Interest Alicia A. Matavosian and Lawrence J. Bonassar are lead inventors on a patent for the smart syringe device. 108 3.9. Supporting Information Figure 3.1S: LIVE/DEAD staining of primary chondrocytes at a density of 25×106 cells/mL suspended in 0.5, 1, or 3 w/v% alginate. Figure 3.2S: LIVE/DEAD staining of primary chondrocytes at a density of 25×106 cells/mL suspended in (top row) 1 wt/v% alginate at pHs ranging from 6.4 - 7.6 or (bottom row) PBS at pHs ranging from 5.8 – 7.6. 109 Figure 3.3S: Sensitivity of |Zcells| to changes in pH in PBS and alginate bio-ink were determined from 1 – 25,000 kHz. a-d.) the rate of change for |Zcells| in acidic pHs from 5.8 – 6.7 was positive for primary chondrocytes in PBS and negative for cells in alginate. e.) at a neutral pH of 7, cells suspended in either PBS or alginate were typically positive on average but displayed lower rate of change in comparison to acidic or basic pHs. f.) the rate of change for |Zcells| in a slightly basic solution at 7.6 pH was positive for primary chondrocytes in alginate and negative for cells in PBS, which is reversed from acidic solutions. 110 CHAPTER 4 COMBINING ELECTRODES TYPES ENABLES HIGHER SENSITIVITY TOWARDS CELL CONCENTRATION* 4.1. Problem Statement The preceding chapters established the critical need for real-time monitoring during bioprinting to evaluate construct quality and demonstrated the feasibility of using an in-line impedance-based device—the Smart Syringe—for real-time measurement of cellular properties during bioink deposition. The Smart Syringe exhibited sensitivity to both cell concentration and viability in saline solutions and further demonstrated the capability to distinguish between acellular and cellular alginate formulations during continuous flow through the electrodes. Although this device showed excellent compatibility with real-time monitoring of cells, the design of the Smart Syringe presented several limitations that constrained both its sensitivity to low cell concentrations and the reproducibility of device fabrication. As detailed in Chapter 2, impedance magnitude was positively correlated with cell concentration (see Fig. 2.2); however, the device consistently achieved reliable detection only at concentrations ≥ 25×10⁶ cells/mL. At concentrations below this threshold, impedance signals from cellular samples were insufficiently distinguishable from background measurements, limiting the lower bound of quantifiable cell concentrations. * Matavosian A, Bonassar L, Bhuiyan D, Bhatnagar V. Electrode device with expanded sensitive range for real-time, non-destructive measurement of cell concentration. 10789-01-US, 2024. p. 1–36. 111 Manufacturing limitations further contributed to variability in device performance. Specifically, the geometry of the Smart Syringe housing posed challenges during electrode integration. Drilling into the curved surface of the syringe body introduced inconsistencies in electrode placement, resulting in inter-electrode distances ranging from 1.0 to 1.5 mm (Fig. 4.1). These spatial inconsistencies led to variability in baseline impedance values across devices, necessitating signal normalization to enable cross-device comparisons. To address these limitations, improvements to the Smart Syringe were designed with manufacturability and cell sensitivity in mind. The goals for this enhancement were to design a chamber with a reproducible distance between the electrodes and to improve detection of cell-laden solutions. The redesigned device aimed to improve manufacturability, reduce inter-device variability, and extend the sensitivity range to detect lower cell concentrations, thereby broadening its applicability for real-time monitoring in diverse bioprinting contexts. Figure 4.1: (Left) Zoomed top view of gold pin-shaped electrode nestled in the Smart Syringe housing. (Right) Distance between electrodes varied due to custom machining of the housing. 112 4.2. Materials and Methods 4.2.1. Electrode housing and electrode fabrication Electrode distance was set by redesigning the housing unit on CAD (SolidWorks). The interior shape of the original Smart Syringe was maintained with tapered clearance holes running from the outside to the inside of the chamber (Fig. 4.2a). A chamber was inset into the inner wall of the housing to hold the electrodes. To maintain resolution in the clearance holes, the housing units were split into two halves and 3-D printed using VeroClear (StrataSys, Minnetonka, MN) with water- soluble support material (Fig. 4.2b). To improve detection of cell-laden samples, the reduction of noise was prioritized. To achieve this, the shape of the electrode was exchanged from pins coated in gold to parallel sheets made of platinum. Flattening the shape of the electrode increases the surface area and reduces sources of noise from electrode resistance, increasing signal to noise ratio (276). Although gold and platinum show similar electric performance on a millimeter scale, platinum is more ductile and scratch- resistant (277). Electrodes were crafted from Kapton tape sputter-coated with platinum. The Kapton tape was shaped into the electrodes, folded in half such that platinum was on either side, and adhered to the housing unit using silver epoxy (EP77MFMED, MasterBond, Hackensack, NJ) (Fig. 4.2c, 4.2d). 113 Figure 4.2: Manufacturing of the curved parallel plate electrode. a.) CAD model of the electrode housing split into two parts. b.) 3D printed housing with water-soluble support material. c.) Platinum sputter-coated Kapton tape measured prior to shaping. d.) Kapton tape was attached copy wire within the inner chamber of the housing unit using silver epoxy. e.) Two halves of the housing unit were combined. f.) Male and female luer locks were added to either end of the housing to support use with a syringe. 4.2.2. Comparison of electrodes Cell detection was compared between the two electrode types. Primary chondrocytes isolated from neonatal bovine stifles were diluted to concentrations ranging from 0.1 × 106 cells/mL – 65 × 106 cells/mL in PBS. The order of tested samples was randomized, and each concentration had n = 3 replicates. Both electrodes were flushed between measurements using PBS. |Zcells| was calculated for both a.) e.) d.) c.) b.) f.) 114 electrodes using equation 5, and this measurement indicated each electrode’s capability to detect cells compared to the background. 4.2.3. Combining multiple electrode types Both electrodes were combined in series to monitor cells across a wide range of concentrations (Fig. 4.3). Primary chondrocytes suspended in PBS were diluted to concentrations ranging from 0.6×106 cells/mL - 112×106 cells/mL with n = 3 per concentration. This multi-electrode device was flushed between measurements using PBS. To ascertain the capabilities of each electrode in this unit, fractional change was calculated using Equation 6. |𝑍| 𝑓𝑟𝑎𝑐𝑡𝑖𝑜𝑛𝑎𝑙 𝑐ℎ𝑎𝑛𝑔𝑒 = |𝑍𝑐𝑒𝑙𝑙𝑢𝑙𝑎𝑟 |−|𝑍𝑃𝐵𝑆| |𝑍𝑐𝑒𝑙𝑙𝑢𝑙𝑎𝑟| ∗ 100 (6) Figure 4.3: Multi-electrode prototype, combining pin-shaped electrodes with curved parallel plate electrodes a.) into a single chamber or b.) in series. 115 4.3. Results 4.3.1. Comparison of electrodes for monitoring cell concentration Pin electrodes displayed an increasing trend in |Zcells| as cell concentration increased, however this trend was more pronounced from 4.5 – 50 × 106 cells/mL (Fig. 4.4). In contrast, curved parallel plate electrodes displayed an increasing trend from 0.1 – 25 × 106 cells/mL and showed comparatively lower standard deviation at each concentration (Fig. 4.4). Interestingly, |Zcells| for curved parallel plate electrodes suggested a high compatibility with detecting cells with values up to 40 times higher than the pin electrodes (Fig. 4.4). Overall, pin electrodes were found to be more sensitive to higher cell concentrations, whereas curved parallel plate electrodes were capable of detecting cells at lower cell concentrations. This revelation suggested that cells could be monitored across an expanded range of cell concentrations by collecting measurements with both electrodes. 116 Figure 4.4: |Zcells| for pin electrodes and curved parallel plate electrodes across multiple cell concentrations displayed at 10 kHz. Shapes denote sample averages and bars represent the standard deviation. 4.3.2. Multi-electrode device Using a single multi-electrode device, pin electrodes and curved parallel plate electrodes displayed overall similar trends in behavior with 1 – 100 kHz showing the most reliable measurements for the detection of cells (Fig. 4.5). As previously shown, curved parallel plate electrodes showed an increase in |Z| fractional change with cell concentration, however, this trend is most highlighted in lower cell concentrations from 0.6 – 11.6 × 106 cells/mL (Fig. 4.5). 1 2 3 4 5 -5 0 5 10 15 -200 0 200 400 600 20 40 60 Cell Concentration (M/mL) G o ld Z c e ll s (Ω ) P la tin u m Z c e lls (Ω ) Gold Electrodes Platinum Electrodes P in e le ct r o d e | Z ce ll s| ( Ω ) C u rv e d p a r a ll e l p la te e le ct ro d e |Z ce ll s| ( Ω ) Pin electrodes Curved parallel plate electrodes 117 Figure 4.5: Fractional change of |Z| across multiple cell concentrations from 1 – 25,000 kHz. Black box indicates regions of high sensitivity to cell concentration from approximately 1 – 80 kHz. Multi-electrode sensitivity to cells was further assessed by calculating |Zcells| across frequencies with high reliability. At 10 kHz, the pin electrodes displayed typical |Zcells| measurements with values increasing alongside cell concentration (Fig. 4.6). For curved parallel plate electrodes, |Zcells| displayed a sharp increase from 0.7 – 1.1 × 106 cells/mL followed by a gradual increase at concentrations greater than 11.6 × Curved parallel plate electrodes Pin electrodes 118 106 cells/mL at both 1 kHz and 10 kHz (Fig. 4.6). However, |Zcells| values were sharply reduced compared to Fig. 4.4, suggesting limitations with reproducibility due to the manufacturing method of the electrodes. At 100 kHz, electrode cross-talk greatly affected |Z|, leading to unreliable calculations of |Zcells|. Figure 4.6: |Zcells| was calculated for measurements collected using the multi-electrode device. Measurements were plotted at 1, 10, and 100 kHz and fitted to a piecewise, weighted linear regression. a.) |Zcells| from the curved parallel plate electrodes increased with cell concentration, however, pin electrodes displayed a stagnation in |Zcells|. b.) At 10 kHz, the pin electrodes showed increasing |Zcells| with increasing cell concentration from 28.5 - 112 × 106 cells/mL. Curved parallel plate electrodes also showed increasing |Zcells|, particularly from 0.6 - 1.1 × 106 cells/mL. c.) Crosstalk between both electrodes greatly distorted the collected data, resulting in high noise. Pin electrodes Curved parallel plate electrodes a.) b.) c.) 119 CHAPTER 5 CONCLUSIONS AND FUTURE DIRECTIONS The goals of this dissertation were to establish the need for real-time monitoring during bioprinting, to measure cell concentration and in real-time using in-line sensors, to determine the influence of alginate bioink on impedance, and to improve the sensitivity of the Smart Syringe device to low cell concentrations. Current quality assurance practices for bioprinted constructs rely heavily on destructive testing, necessitating the fabrication of multiple constructs per patient case, thereby significantly escalating production costs and limiting clinical scalability. Chapter 1 established the foundational argument for integrating real-time sensing technologies into the bioprinting pipeline. Through the continuous monitoring bioink homogeneity, pH, temperature, and viscosity, the optimization of both formulation and print accuracy was achieved in a significantly reduced timeframe compared to standard methods. The real-time detection and correction of defects during printing further improved the reproducibility of constructs and minimized the consumption of valuable bioink and cellular material. Moreover, by capturing real-time data on key cellular parameters, including concentration, viability, and cell type, it became possible to fine-tune printing parameters to optimize viability and to engineer gradients that more accurately reflect the heterogeneous microenvironments of native tissues. Despite these promising developments, current bioprinting platforms lack the integrated infrastructure necessary for seamless real-time monitoring. Existing systems 120 capable of such measurements typically rely on customized configurations with limited implementation of feedback control loops capable of converting sensor outputs into actionable printer responses. The absence of standardized interfaces and protocols across different bioprinter platforms and construct production workflows presents a substantial barrier to the widespread adoption of smart bioprinting technologies. To address these limitations, Chapter 2 introduced an in-line impedance sensor for the real-time measurement of cell concentration and viability during sample deposition. Experimental results demonstrated a linear correlation between impedance and both cell concentration and viability, with the sensitivity of detection modulated by the frequency of the applied signal. Notably, conditions of high concentration and high viability produced additive increases in impedance, whereas the presence of non- viable cells at high concentrations led to a marked reduction in impedance. Navigating the relationship between these variables would enable predictive modeling of cell concentration and viability with potential utility extending beyond bioprinting to other applications that are reliant upon viable cell preparations, such as bioreactors or cell therapies. Although the Smart Syringe was shown to successfully detect cells in saline, the influence of bioink on impedance was not understood. Specifically, variations in alginate concentration, pH or crosslinking were hypothesized to affect impedance through ion movement caused by changes in viscosity and counter-ion concentration. Chapter 3 addressed this knowledge gap through a systematic investigation aimed at decoupling the effects of these individual parameters on the impedance response. Experimental results revealed that impedance increased with higher alginate 121 concentration, lower pH, and crosslinking with CaCl2. Furthermore, across all tested conditions, cell-laden alginate samples consistently exhibited higher impedance compared to their acellular counterparts, confirming the capability of the impedance- based system to detect cells within a range of bioink formulations. These findings underscore the versatility of the Smart Syringe for in-line characterization of cell- laden bioink. In addition, this chapter demonstrated the feasibility of using impedance measurements to track spatial variations in cell distribution. This capability lays the foundation for integrating impedance-based feedback mechanisms into bioprinting workflows, thereby facilitating precise control over cellular deposition patterns essential for replicating the structural and functional complexity of native tissues. To expand the operational capabilities of the Smart Syringe, a redesign of the electrode configuration was undertaken with the objective of enhancing sensitivity to cell concentrations below 25×10⁶ cells/mL. The implementation of curved parallel plates improved detection sensitivity at these lower cell concentrations. Furthermore, a multi-electrode configuration that combined the original gold-pin electrodes in series with the newly developed curved platinum parallel plate electrodes demonstrated promising performance, particularly in the context of complex biological samples. Despite these advances, several limitations were identified that warrant further development. Notably, the reproducibility of results measured by the curved parallel plate electrodes was limited. This variability is attributed to the manual fabrication process, which involved hand-shaping and positioning the electrodes within the syringe housing. Future iterations of the device would benefit from standardized manufacturing protocols to ensure consistent electrode dimensions, placement, and 122 electrical performance, thereby improving dataset reproducibility and facilitating broader deployment. Additionally, biofouling emerged as a significant factor affecting long-term device sensitivity. Both pin and curved parallel plate electrodes experienced performance degradation over time, attributed to the accumulation of ions, proteins, alginate, or cellular debris on the electrode surfaces (275). A decline in sensor functionality was observed after approximately 12 - 84 sample measurements, with variation dependent on the duration of testing and the tested solution. To address this, future design improvements should explore the integration of anti-fouling strategies, such as the application of thin-film coatings or the use of specialized cleaning solvents, to extend device longevity and maintain sensitivity (278,279). 123 BIBLIOGRAPHY 1. Cohen DL, Lo W, Tsavaris A, Peng D, Lipson H, Bonassar LJ. Increased Mixing Improves Hydrogel Homogeneity and Quality of Three-Dimensional Printed Constructs. Tissue Engineering Part C: Methods. 2011 Feb;17(2):239–48. 2. Kelly D, Sergis V, Ventura i Blanco L, Mason K, Daly AC. Autonomous Control of Extrusion Bioprinting Using Convolutional Neural Networks. Advanced Functional Materials. n/a(n/a):2424553. 3. Gugliandolo SG, Prioglio E, Moscatelli D, Colosimo BM. A new solution for in situ monitoring of shape fidelity in extrusion-based bioprinting via thermal imaging. International Journal of Bioprinting. 2024 Mar 22;10(3):2021. 4. Margarita A, Gugliandolo SG, Santoni S, Moscatelli D, Colosimo BM. A novel solution for real-time in-situ cell distribution monitoring in 3D bioprinting via fluorescence imaging. Biofabrication. 2025 Mar;17(2):021001. 5. Matavosian AA, Griffin AC, Bhuiyan DB, Lyness AM, Bhatnagar V, Bonassar LJ. Real-time assessment of cell concentration and viability onboard a syringe using dielectric impedance spectroscopy for extrusion bioprinting. Biofabrication. 2025 Feb;17(2):025018. 6. Poologasundarampillai G, Haweet A, Jayash SN, Morgan G, Moore JE, Candeo A. Real-time imaging and analysis of cell-hydrogel interplay within an extrusion- bioprinting capillary. Bioprinting. 2021 Aug 1;23:e00144. 7. Haring AP, Jiang S, Barron C, Thompson EG, Sontheimer H, He JQ, et al. 3D bioprinting using hollow multifunctional fiber impedimetric sensors. Biofabrication. 2020 Jul 1;12(3):035026. 8. Safir F, Vu N, Tadesse LF, Firouzi K, Banaei N, Jeffrey SS, et al. Combining Acoustic Bioprinting with AI-Assisted Raman Spectroscopy for High-Throughput Identification of Bacteria in Blood. Nano Lett. 2023 Mar 22;23(6):2065–73. 9. Matthews J, Bonassar L. System and devices for monitoring cell-containing materials and methods of their use. US12019040B2, 2021. p. 1–23. 10. Auregen Biotherapeutics, SA. A Phase 1/2, Multi-Center, Open-Label, Single- Arm Safety, Tolerability, and Efficacy Study of an Auricle and Wedge Subcutaneous Implant Combination Product (AUR-201) in Patients With Unilateral Microtia [Internet]. clinicaltrials.gov; 2024 Feb [cited 2023 Dec 31]. Report No.: NCT06072040. Available from: https://clinicaltrials.gov/study/NCT06072040 11. Auregen Biotherapeutics, SA. Long-Term Follow-Up Study of Unilateral Microtia Patients Who Participated in Study AUR 201 05 and Had the AUR-201 124 Implant in Place at Final Visit (24 Weeks After the First Implantation Surgery) [Internet]. clinicaltrials.gov; 2024 Feb [cited 2023 Dec 31]. Report No.: NCT06078566. Available from: https://clinicaltrials.gov/study/NCT06078566 12. Principal Investigator: Dominique Casanova. A DERMO-EPIDERMAL AUTOLOGOUS SKIN SUBSTITUTE FOR FURTHER THERAPEUTIC USE (BIOPSKIN) [Internet]. clinicaltrials.gov; 2021 Jun [cited 2025 Jan 26]. Report No.: NCT04925323. Available from: https://clinicaltrials.gov/study/NCT04925323 13. Bae JS. Development of the Practical Usage Based Technology Using the Patient Customized Bioprinting Trachea for the Regeneration of Respiratory Tract (Trachea) [Internet]. clinicaltrials.gov; 2023 Sep [cited 2023 Dec 31]. Report No.: NCT06051747. Available from: https://clinicaltrials.gov/study/NCT06051747 14. Li P, Faulkner A, Medcalf N. 3D bioprinting in a 2D regulatory landscape: gaps, uncertainties, and problems. Law, Innovation and Technology. 2020 Jan 2;12(1):1–29. 15. Matavosian AA, Bonassar LJ. Reimagining Bioprinters: Real-Time Monitoring for Quality Control of Bioprinted Constructs and Future Vision. Manuscript submitted for publication. 16. Mistry H, Connock M, Pink J, Shyangdan D, Clar C, Royle P, et al. The cost- effectiveness of autologous chondrocyte implantation. In: Autologous chondrocyte implantation in the knee: systematic review and economic evaluation [Internet]. NIHR Journals Library; 2017 [cited 2024 Jul 16]. Available from: https://www.ncbi.nlm.nih.gov/books/NBK424065/ 17. Phillips TJ. Tissue-Engineered Skin: An Alternative to Split-Thickness Skin Grafts? Archives of Dermatology. 1999 Aug 1;135(8):977–8. 18. Munson MS, Yager P. Simple quantitative optical method for monitoring the extent of mixing applied to a novel microfluidic mixer. Analytica Chimica Acta. 2004 Apr 1;507(1):63–71. 19. Downes A, Mouras R, Elfick A. Optical Spectroscopy for Noninvasive Monitoring of Stem Cell Differentiation. Journal of Biomedicine and Biotechnology. 2010;2010:1–10. 20. B. Kell D, Markx GH, Davey CL, Todd RW. Real-time monitoring of cellular biomass: Methods and applications. TrAC Trends in Analytical Chemistry. 1990 Jun 1;9(6):190–4. 21. Beć KB, Grabska J, Huck CW. Biomolecular and bioanalytical applications of infrared spectroscopy – A review. Analytica Chimica Acta. 2020 Oct;1133:150–77. 22. Akbari S, Anderson P, Zang H, Ganjian A, Balke R, Kwon T, et al. Non- 125 invasive real-time monitoring of cell concentration and viability using Doppler ultrasound. SLAS Technology. 2022 Dec 1;27(6):368–75. 23. Ruland A, Gilmore KJ, Daikuara LY, Fay CD, Yue Z, Wallace GG. Quantitative ultrasound imaging of cell-laden hydrogels and printed constructs. Acta Biomaterialia. 2019 Jun 1;91:173–85. 24. Ducommun P, Kadouri A, Von Stockar U, Marison IW. On‐line determination of animal cell concentration in two industrial high‐density culture processes by dielectric spectroscopy. Biotech & Bioengineering. 2002 Feb 5;77(3):316–23. 25. Heileman K, Daoud J, Tabrizian M. Dielectric spectroscopy as a viable biosensing tool for cell and tissue characterization and analysis. Biosensors and Bioelectronics. 2013 Nov;49:348–59. 26. Matavosian AA, Griffin AC, Bhuiyan DB, Lyness AM, Bhatnagar V, Bonassar LJ. Real-time assessment of cell concentration and viability onboard a syringe using dielectric impedance spectroscopy for extrusion bioprinting. Biofabrication. 2025 Feb 25;17(2). 27. Forget A, Blaeser A, Miessmer F, Köpf M, Campos DFD, Voelcker NH, et al. Mechanically Tunable Bioink for 3D Bioprinting of Human Cells. Advanced Healthcare Materials. 2017;6(20):1700255. 28. Wang J, Cui Z, Maniruzzaman M. Bioprinting: A focus on improving bioink printability and cell performance based on different process parameters. International Journal of Pharmaceutics. 2023 Jun 10;640:123020. 29. Matavosian AA, Griffin AC, Bonassar LJ. Alginate bioink properties influence real-time impedance monitoring of cells during extrusion bioprinting. Manuscript submitted for publication. 30. Statista [Internet]. [cited 2025 Mar 11]. Personalized medicine numbers U.S. 2008-2020. Available from: https://www.statista.com/statistics/687992/number-of- personalized-medicines/ 31. Hinton TJ, Lee A, Feinberg AW. 3D bioprinting from the micrometer to millimeter length scales: Size does matter. Current Opinion in Biomedical Engineering. 2017 Mar 1;1:31–7. 32. Lee A, Hudson AR, Shiwarski DJ, Tashman JW, Hinton TJ, Yerneni S, et al. 3D bioprinting of collagen to rebuild components of the human heart. Science. 2019 Aug 2;365(6452):482–7. 33. Isaacson A, Swioklo S, Connon CJ. 3D bioprinting of a corneal stroma equivalent. Exp Eye Res. 2018 Aug;173:188–93. 126 34. Kutlehria S, Dinh TC, Bagde A, Patel N, Gebeyehu A, Singh M. High- throughput 3D bioprinting of corneal stromal equivalents. Journal of Biomedical Materials Research Part B: Applied Biomaterials. 2020;108(7):2981–94. 35. Stankey PP, Kroll KT, Ainscough AJ, Reynolds DS, Elamine A, Fichtenkort BT, et al. Embedding Biomimetic Vascular Networks via Coaxial Sacrificial Writing into Functional Tissue. Advanced Materials. 2024;36(36):2401528. 36. Nielsen J, Kaldor ,Jenny, Irwin ,Adam, Stewart ,Cameron, and Nicol D. Bespoke Regulation for Bespoke Medicine? A Comparative Analysis of Bioprinting Regulation in Europe, the USA and Australia. Journal of 3D Printing in Medicine. 2021 Sep 30;5(3):155–67. 37. Hourd P, Medcalf ,Nicholas, Segal ,Joel, and Williams DJ. A 3D Bioprinting Exemplar of the Consequences of the Regulatory Requirements on Customized Processes. Regenerative Medicine. 2015 Oct 1;10(7):863–83. 38. Bate R, Hess K. The role of pre-shipment batch testing in ensuring good medicine quality. Malariaworld J. 2012 Dec 5;3:12. 39. Xiong H, Yu LX, Qu H. Batch-to-Batch Quality Consistency Evaluation of Botanical Drug Products Using Multivariate Statistical Analysis of the Chromatographic Fingerprint. AAPS PharmSciTech. 2013 Jun 1;14(2):802–10. 40. Harer J. Process Validation for Medical Device Manufacturing. In: Medical Devices and In Vitro Diagnostics [Internet]. Springer, Cham; 2023 [cited 2025 Jun 1]. p. 1–30. Available from: https://link.springer.com/referenceworkentry/10.1007/978-3-030-98743-5_19-1 41. Rowley J, Abraham E, Campbell A, Brandwein H, Oh S. Meeting Lot-Size Challenges of Manufacturing Adherent Cells for Therapy. BioProcess International. 2012;10(3). 42. Suigyama H, Shiokaramatsu M, Kino-oka M. A distribution-based approach for determining lot sizes in the filling of human-induced pluripotent stem cells. Regenerative Therapy. 2019 Dec 15;12:94–101. 43. Jorgensen AM, Varkey M, Gorkun A, Clouse C, Xu L, Chou Z, et al. Bioprinted Skin Recapitulates Normal Collagen Remodeling in Full-Thickness Wounds. Tissue Engineering Part A. 2020 May;26(9–10):512–26. 44. Pitacco P, Sadowska JM, O’Brien FJ, Kelly DJ. 3D bioprinting of cartilaginous templates for large bone defect healing. Acta Biomaterialia. 2023 Jan 15;156:61–74. 45. Costa JB, Park J, Jorgensen AM, Silva-Correia J, Reis RL, Oliveira JM, et al. 3D Bioprinted Highly Elastic Hybrid Constructs for Advanced Fibrocartilaginous 127 Tissue Regeneration. Chem Mater. 2020 Oct 13;32(19):8733–46. 46. Muthusamy S, Kannan S, Lee M, Sanjairaj V, Lu WF, Fuh JYH, et al. 3D bioprinting and microscale organization of vascularized tissue constructs using collagen-based bioink. Biotechnology and Bioengineering. 2021;118(8):3150–63. 47. Diamantides N, Dugopolski C, Blahut E, Kennedy S, Bonassar LJ. High density cell seeding affects the rheology and printability of collagen bioinks. Biofabrication. 2019 Aug;11(4):045016. 48. Skardal A, Devarasetty M, Kang HW, Mead I, Bishop C, Shupe T, et al. A hydrogel bioink toolkit for mimicking native tissue biochemical and mechanical properties in bioprinted tissue constructs. Acta Biomaterialia. 2015 Oct 1;25:24–34. 49. Van der Valk DC, Van der Ven CFT, Blaser MC, Grolman JM, Wu PJ, Fenton OS, et al. Engineering a 3D-Bioprinted Model of Human Heart Valve Disease Using Nanoindentation-Based Biomechanics. Nanomaterials. 2018 May;8(5):296. 50. Choe R, Devoy E, Jabari E, Packer JD, Fisher JP. Biomechanical Aspects of Osteochondral Regeneration: Implications and Strategies for Three-Dimensional Bioprinting. Tissue Engineering Part B: Reviews. 2022 Aug;28(4):766–88. 51. Zhang T, Yan KC, Ouyang L, Sun W. Mechanical characterization of bioprinted in vitro soft tissue models. Biofabrication. 2013 Nov;5(4):045010. 52. Ahearne M, Yang Y, El Haj AJ, Then KY, Liu KK. Characterizing the viscoelastic properties of thin hydrogel-based constructs for tissue engineering applications. Journal of The Royal Society Interface. 2005 Jul 28;2(5):455–63. 53. Webber RE, Shull KR. Strain Dependence of the Viscoelastic Properties of Alginate Hydrogels. Macromolecules. 2004 Aug 1;37(16):6153–60. 54. Krishnan L, Weiss JA, Wessman MD, Hoying JB. Design and Application of a Test System for Viscoelastic Characterization of Collagen Gels. Tissue Engineering. 2004 Jan;10(1–2):241–52. 55. Guvendiren M, D. Lu H, A. Burdick J. Shear-thinning hydrogels for biomedical applications. Soft Matter. 2012;8(2):260–72. 56. Herrada-Manchón H, Fernández MA, Aguilar E. Essential Guide to Hydrogel Rheology in Extrusion 3D Printing: How to Measure It and Why It Matters? Gels. 2023 Jul;9(7):517. 57. Bom S, Ribeiro R, Ribeiro HM, Santos C, Marto J. On the progress of hydrogel-based 3D printing: Correlating rheological properties with printing behaviour. International Journal of Pharmaceutics. 2022 Mar 5;615:121506. 128 58. Freeman FE, Kelly DJ. Tuning Alginate Bioink Stiffness and Composition for Controlled Growth Factor Delivery and to Spatially Direct MSC Fate within Bioprinted Tissues. Sci Rep. 2017 Dec 6;7(1):17042. 59. Malda J, Visser J, Melchels FP, Jüngst T, Hennink WE, Dhert WJA, et al. 25th Anniversary Article: Engineering Hydrogels for Biofabrication. Advanced Materials. 2013;25(36):5011–28. 60. Schwab A, Levato R, D’Este M, Piluso S, Eglin D, Malda J. Printability and Shape Fidelity of Bioinks in 3D Bioprinting. Chem Rev. 2020 Oct 14;120(19):11028–55. 61. Sánchez-Sánchez R, Rodríguez-Rego JM, Macías-García A, Mendoza-Cerezo L, Díaz-Parralejo A. Relationship between shear-thinning rheological properties of bioinks and bioprinting parameters. Int J Bioprint. 2023 Feb 16;9(2):687. 62. Xu HQ, Liu JC, Zhang ZY, Xu CX. A review on cell damage, viability, and functionality during 3D bioprinting. Military Med Res. 2022 Dec 16;9(1):70. 63. Nair K, Gandhi M, Khalil S, Yan KC, Marcolongo M, Barbee K, et al. Characterization of cell viability during bioprinting processes. Biotechnology Journal. 2009;4(8):1168–77. 64. Yokpradit A, Tongloy T, Kaewpirom S, Boonsang S. A Real-time Rheological Measurement for Biopolymer 3D Printing Process. Sensors and Materials. 2018 Oct 12;30(10):2199. 65. Cohen DL, Lipton JI, Bonassar LJ, Lipson H. Additive manufacturing for in situ repair of osteochondral defects. Biofabrication. 2010 Sep;2(3):035004. 66. Zhang XZ, Wu DQ, Chu CC. Effect of the crosslinking level on the properties of temperature-sensitive poly(N-isopropylacrylamide) hydrogels. Journal of Polymer Science Part B: Polymer Physics. 2003;41(6):582–93. 67. Ishida K, Uno T, Itoh T, Kubo M. Synthesis and Property of Temperature- Responsive Hydrogel with Movable Cross-Linking Points. Macromolecules. 2012 Aug 14;45(15):6136–42. 68. Deol PK, Kaur A, Kooner JK, Gill AS, Singh M, Kaur IP. Alginate Based Polyelectrolyte Complexes for Drug Delivery and Biomedical Applications. In: Jana S, Jana S, editors. Alginate Biomaterial: Drug Delivery Strategies and Biomedical Engineering [Internet]. Singapore: Springer Nature; 2023 [cited 2025 May 29]. p. 179–205. Available from: https://doi.org/10.1007/978-981-19-6937- 9_8 69. Axpe E, Oyen ML. Applications of Alginate-Based Bioinks in 3D Bioprinting. Int J Mol Sci. 2016 Nov 25;17(12):1976. 129 70. Wu D, Pang S, Röhrs V, Berg J, Ali ASM, Mei Y, et al. Man vs. machine: Automated bioink mixing device improves reliability and reproducibility of bioprinting results compared to human operators. IJB. 2024 Feb 12;0(0):1974. 71. Ouyang L, Yao R, Zhao Y, Sun W. Effect of bioink properties on printability and cell viability for 3D bioplotting of embryonic stem cells. Biofabrication. 2016 Sep;8(3):035020. 72. Bhattacharyya A, Janarthanan G, Tran HN, Ham HJ, Yoon J, Noh I. Bioink homogeneity control during 3D bioprinting of multicomponent micro/nanocomposite hydrogel for even tissue regeneration using novel twin screw extrusion system. Chemical Engineering Journal. 2021 Jul 1;415:128971. 73. Zhang H, Tan Y, Hao L, Zhang S, Xiao J, Poon CS. Intelligent real-time quality control for 3D-printed concrete with near-nozzle secondary mixing. Automation in Construction. 2024 Apr 1;160:105325. 74. Dani S, Ahlfeld T, Albrecht F, Duin S, Kluger P, Lode A, et al. Homogeneous and Reproducible Mixing of Highly Viscous Biomaterial Inks and Cell Suspensions to Create Bioinks. Gels. 2021 Dec;7(4):227. 75. Diamantides N, Slyker L, Martin S, Rodriguez MR, Bonassar LJ. Pre‐ glycation impairs gelation of high concentration collagen solutions. J Biomedical Materials Res. 2022 Dec;110(12):1953–63. 76. Li Z, Zhang B, Dang D, Yang X, Yang W, Liang W. A review of microfluidic- based mixing methods. Sensors and Actuators A: Physical. 2022 Sep 1;344:113757. 77. Yang T, Jin Y, Smith LM, Dahotre NB, Neogi A. Real-time in-situ ultrasound monitoring of soft hydrogel 3D printing with subwavelength resolution. Commun Eng. 2024 Nov 9;3(1):1–13. 78. Han E, Peters IR, Jaeger HM. High-speed ultrasound imaging in dense suspensions reveals impact-activated solidification due to dynamic shear jamming. Nat Commun. 2016 Jul 20;7(1):12243. 79. Nam SY, Ricles LM, Suggs LJ, Emelianov SY. Imaging Strategies for Tissue Engineering Applications. Tissue Engineering Part B: Reviews. 2015 Feb;21(1):88–102. 80. Mercado KP, Helguera M, Hocking DC, Dalecki D. Noninvasive Quantitative Imaging of Collagen Microstructure in Three-Dimensional Hydrogels Using High- Frequency Ultrasound. Tissue Engineering Part C: Methods. 2015 Jul;21(7):671– 82. 81. Georgakoudi I, Rice WL, Hronik-Tupaj M, Kaplan DL. Optical Spectroscopy and Imaging for the Noninvasive Evaluation of Engineered Tissues. Tissue 130 Engineering Part B: Reviews. 2008 Dec;14(4):321–40. 82. Stepanovska J, Otahal M, Hanzalek K, Supova M, Matejka R. pH Modification of High-Concentrated Collagen Bioinks as a Factor Affecting Cell Viability, Mechanical Properties, and Printability. Gels. 2021 Dec;7(4):252. 83. Diamantides N, Wang L, Pruiksma T, Siemiatkoski J, Dugopolski C, Shortkroff S, et al. Correlating rheological properties and printability of collagen bioinks: the effects of riboflavin photocrosslinking and pH. Biofabrication. 2017 Jul;9(3):034102. 84. Garcia-Villen F, Guembe A, José MR, Zúñiga T, Ruiz-Alonso S, Saenz-del- Burgo L, et al. Characterization and assessment of new fibrillar collagen inks and bioinks for 3D printing and bioprinting. Int J Bioprint. 2023 Mar 16;9(3):712. 85. Sæther HV, Holme HK, Maurstad G, Smidsrød O, Stokke BT. Polyelectrolyte complex formation using alginate and chitosan. Carbohydrate Polymers. 2008 Nov 21;74(4):813–21. 86. Rosenblatt J, Devereux B, Wallace DG. Injectable collagen as a pH-sensitive hydrogel. Biomaterials. 1994 Oct 1;15(12):985–95. 87. Goudie KJ, McCreath SJ, Parkinson JA, Davidson CM, Liggat JJ. Investigation of the influence of pH on the properties and morphology of gelatin hydrogels. Journal of Polymer Science. 2023;61(19):2316–32. 88. O’Mara P, Farrell A, Bones J, Twomey K. Staying alive! Sensors used for monitoring cell health in bioreactors. Talanta. 2018 Jan 1;176:130–9. 89. Souza A, Parnell M, Rodriguez BJ, Reynaud EG. Role of pH and Crosslinking Ions on Cell Viability and Metabolic Activity in Alginate–Gelatin 3D Prints. Gels. 2023 Nov;9(11):853. 90. Udomsom S, Budwong A, Wongsa C, Sangngam P, Baipaywad P, Manaspon C, et al. Automatic Programmable Bioreactor with pH Monitoring System for Tissue Engineering Application. Bioengineering. 2022 May;9(5):187. 91. Jeevarajan AS, Vani S, Taylor TD, Anderson MM. Continuous pH monitoring in a perfused bioreactor system using an optical pH sensor. Biotechnology and Bioengineering. 2002;78(4):467–72. 92. Xu X, Smith S, Urban J, Cui Z. An in line non-invasive optical system to monitor pH in cell and tissue culture. Med Eng Phys. 2006 Jun;28(5):468–74. 93. Goh GL, Agarwala S, Tan YJ, Yeong WY. A low cost and flexible carbon nanotube pH sensor fabricated using aerosol jet technology for live cell applications. Sensors and Actuators B: Chemical. 2018 May 1;260:227–35. 131 94. Suthithanakom S, Juemjutitam S, Arunwattanamongkol N, Wongpakham T, Pimpin A, Chancharoen R. Temperature Control for Hydrogel Bio-Printing. In: 2019 IEEE International Conference on Cybernetics and Intelligent Systems (CIS) and IEEE Conference on Robotics, Automation and Mechatronics (RAM) [Internet]. 2019 [cited 2025 May 14]. p. 310–5. Available from: https://ieeexplore.ieee.org/document/9095808 95. Hakme da Silva AM, Silva ES, Magalhães DSF, Fraga MA, Correa de Paula e Silva SG, Ferreira PHD, et al. Enhancing Temperature Control in 3D Bioprinting: A Comprehensive Study on PID Control Strategies. Journal of Biomedical Nanotechnology. 2024 Nov 1;20(11):1747–55. 96. Abdollahi A, Malek-Khatabi A, Razavi MS, Sheikhi M, Abbaspour K, Rezagholi Z, et al. The recent advancement in the chitosan-based thermosensitive hydrogel for tissue regeneration. Journal of Drug Delivery Science and Technology. 2023 Sep 1;86:104627. 97. Iberite F, Badiola-Mateos M, Loggini S, Paci C, Ruspi J, Iachetta D, et al. 3D bioprinting of thermosensitive inks based on gelatin, hyaluronic acid, and fibrinogen: reproducibility and role of printing parameters. Bioprinting. 2024 Jun 1;39:e00338. 98. Zhang P, Gao Q, Yu K, Yao Y, Lu L. Investigation on the Temperature Control Accuracy of a Print Head for Extrusion 3D Printing and Its Improved Design. Biomedicines. 2022 Jun;10(6):1233. 99. Yu K, Gao Q, Lin Z, Wang F, Zhang P, Liu Y, et al. Investigation of the temperature gradient control in the printing space for the material extrusion of medical biodegradable hydrogel. Computers in Biology and Medicine. 2024 Jan 1;168:107722. 100. Gómez-Blanco JC, Mancha-Sánchez E, Marcos AC, Matamoros M, Díaz- Parralejo A, Pagador JB. Bioink Temperature Influence on Shear Stress, Pressure and Velocity Using Computational Simulation. Processes. 2020 Jul;8(7):865. 101. Li Q, Zhang B, Xue Q, Zhao C, Luo Y, Zhou H, et al. A Systematic Thermal Analysis for Accurately Predicting the Extrusion Printability of Alginate–Gelatin- Based Hydrogel Bioinks. Int J Bioprint. 2021 Jun 22;7(3):394. 102. Guo R, Tang W. Optimizing Printhead Design for Enhanced Temperature Control in Extrusion-Based Bioprinting. Micromachines. 2024 Aug;15(8):943. 103. Kholil A, Syaefudin EA, Juniar A, Rohim MK, Dwiyati ST. Simulation of temperature distribution in 3D printing heated chamber on orientation and temperature variations with ABS material. J Phys: Conf Ser. 2023 Sep;2596(1):012007. 132 104. Roehm KD, Madihally SV. Bioprinted chitosan-gelatin thermosensitive hydrogels using an inexpensive 3D printer. Biofabrication. 2017 Nov;10(1):015002. 105. Matamoros M, Gómez-Blanco JC, Sánchez ÁJ, Mancha E, Marcos AC, Carrasco-Amador JP, et al. Temperature and Humidity PID Controller for a Bioprinter Atmospheric Enclosure System. Micromachines (Basel). 2020 Nov 12;11(11):999. 106. Janmaleki M, Liu J, Kamkar M, Azarmanesh M, Sundararaj U, Nezhad AS. Role of temperature on bio-printability of gelatin methacryloyl bioink in two-step cross-linking strategy for tissue engineering applications. Biomed Mater. 2020 Dec;16(1):015021. 107. Rahimnejad M, Adoungotchodo A, Demarquette NR, Lerouge S. FRESH bioprinting of biodegradable chitosan thermosensitive hydrogels. Bioprinting. 2022 Aug 1;27:e00209. 108. Betsch M, Cristian C, Lin YY, Blaeser A, Schöneberg J, Vogt M, et al. Incorporating 4D into Bioprinting: Real-Time Magnetically Directed Collagen Fiber Alignment for Generating Complex Multilayered Tissues. Advanced Healthcare Materials. 2018;7(21):1800894. 109. Vojinović V, Cabral JMS, Fonseca LP. Real-time bioprocess monitoring: Part I: In situ sensors. Sensors and Actuators B: Chemical. 2006 Apr 26;114(2):1083– 91. 110. Moncal KK, Ozbolat V, Datta P, Heo DN, Ozbolat IT. Thermally-controlled extrusion-based bioprinting of collagen. J Mater Sci: Mater Med. 2019 Apr 30;30(5):55. 111. Jang HK, Hong SO, Lee SB, Kim JM, Hwang WR. Viscosity measurement of non-Newtonian fluids in pressure-driven flows of general geometries based on energy dissipation rate. Journal of Non-Newtonian Fluid Mechanics. 2019 Dec 1;274:104204. 112. Slyker L, Diamantides N, Kim J, Bonassar LJ. Mechanical performance of collagen gels is dependent on purity, α1/α2 ratio, and telopeptides. J Biomedical Materials Res. 2022 Jan;110(1):11–20. 113. O’Connell C, Ren J, Pope L, Zhang Y, Mohandas A, Blanchard R, et al. Characterizing Bioinks for Extrusion Bioprinting: Printability and Rheology. In: Crook JM, editor. 3D Bioprinting: Principles and Protocols [Internet]. New York, NY: Springer US; 2020 [cited 2025 Jun 1]. p. 111–33. Available from: https://doi.org/10.1007/978-1-0716-0520-2_7 114. Amorim PA, d’Ávila MA, Anand R, Moldenaers P, Van Puyvelde P, Bloemen V. Insights on shear rheology of inks for extrusion-based 3D bioprinting. 133 Bioprinting. 2021 Jun 1;22:e00129. 115. White FM. Viscous fluid flow. 2nd ed. New York: McGraw-Hill; 1991. 614 p. (McGraw-Hill series in mechanical engineering). 116. Laun HM. Polymer melt rheology with a slit die. Rheol Acta. 1983 Mar 1;22(2):171–85. 117. Brunn PO, Vorwerk J. Determination of the steady-state shear viscosity from measurements of the apparent viscosity for some common types of viscometers. Rheologica Acta. 1993;32(4):380–97. 118. Pipe CJ, McKinley GH. Microfluidic rheometry. Mechanics Research Communications. 2009 Jan 1;36(1):110–20. 119. Pipe CJ, Majmudar TS, McKinley GH. High shear rate viscometry. Rheol Acta. 2008 Jul 1;47(5):621–42. 120. Gillispie GJ, Copus J, Uzun-Per M, Yoo JJ, Atala A, Niazi MKK, et al. The correlation between rheological properties and extrusion-based printability in bioink artifact quantification. Materials & Design. 2023 Sep 1;233:112237. 121. Maciel BR, Baki K, Oelschlaeger C, Willenbacher N. The Influence of Rheological and Wetting Properties of Hydrogel-based Bio-Inks on Extrusion- based Bioprinting. Chemie Ingenieur Technik. 2022;94(3):393–401. 122. Murphy SV, De Coppi P, Atala A. Opportunities and challenges of translational 3D bioprinting. Nat Biomed Eng. 2020 Apr;4(4):370–80. 123. Yang S, Chen Q, Wang L, Xu M. In situ defect detection and feedback control with three-dimensional extrusion-based bioprinter-associated optical coherence tomography. Int J Bioprint. 2022 Oct 27;9(1):624. 124. Wenger L, Strauß S, Hubbuch J. Automated and dynamic extrusion pressure adjustment based on real-time flow rate measurements for precise ink dispensing in 3D bioprinting. Bioprinting. 2022 Dec 1;28:e00229. 125. Wang L, Xu M en, Luo L, Zhou Y, Si P. Iterative feedback bio-printing- derived cell-laden hydrogel scaffolds with optimal geometrical fidelity and cellular controllability. Sci Rep. 2018 Feb 12;8(1):2802. 126. Uzun-Per M, Gillispie GJ, Tavolara TE, Yoo JJ, Atala A, Gurcan MN, et al. Automated Image Analysis Methodologies to Compute Bioink Printability. Advanced Engineering Materials. 2021;23(4):2000900. 127. Strauß S, Meutelet R, Radosevic L, Gretzinger S, Hubbuch J. Image analysis as PAT-Tool for use in extrusion-based bioprinting. Bioprinting. 2021 Mar 134 1;21:e00112. 128. Armstrong AA, Alleyne AG, Wagoner Johnson AJ. 1D and 2D error assessment and correction for extrusion-based bioprinting using process sensing and control strategies. Biofabrication. 2020 Aug;12(4):045023. 129. Armstrong AA, Norato J, Alleyne AG, Wagoner Johnson AJ. Direct process feedback in extrusion-based 3D bioprinting. Biofabrication. 2019 Dec;12(1):015017. 130. Bonatti AF, Vozzi G, Chua CK, Maria CD. A Deep Learning Quality Control Loop of the Extrusion-based Bioprinting Process. Int J Bioprint. 2022;8(4):620. 131. Arduengo J, Hascoet N, Chinesta F, Hascoet JY. Open-Loop Control System for High Precision Extrusion-Based Bioprinting Through Machine Learning Modeling. Journal of Machine Engineering. 2024 Apr 2;24(1):103–17. 132. Sun J, Jing L, Fan X, Gao X, Liang YC. Electrohydrodynamic printing process monitoring by microscopic image identification. International Journal of Bioprinting. 2018 Dec 14;5(1):164. 133. Jin Z, Zhang Z, Shao X, Gu GX. Monitoring Anomalies in 3D Bioprinting with Deep Neural Networks. ACS Biomater Sci Eng. 2023 Jul 10;9(7):3945–52. 134. Ogunsanya M, Isichei J, Parupelli SK, Desai S, Cai Y. In-situ Droplet Monitoring of Inkjet 3D Printing Process using Image Analysis and Machine Learning Models. Procedia Manufacturing. 2021 Jan 1;53:427–34. 135. Yang S, Wang L, Chen Q, Xu M. In situ process monitoring and automated multi-parameter evaluation using optical coherence tomography during extrusion- based bioprinting. Additive Manufacturing. 2021 Nov 1;47:102251. 136. Gardner MR, Lewis A, Park J, McElroy AB, Estrada AD, Fish S, et al. In situ process monitoring in selective laser sintering using optical coherence tomography. Opt Eng. 2018 Apr;57(4):041407. 137. DePond PJ, Guss G, Ly S, Calta NP, Deane D, Khairallah S, et al. In situ measurements of layer roughness during laser powder bed fusion additive manufacturing using low coherence scanning interferometry. Materials & Design. 2018 Sep 15;154:347–59. 138. Wang L, Xu M, Zhang L, Zhou Q, Luo L. Automated quantitative assessment of three-dimensional bioprinted hydrogel scaffolds using optical coherence tomography. Biomed Opt Express. 2016 Feb 19;7(3):894–910. 139. Sant S, Hancock MJ, Donnelly JP, Iyer D, Khademhosseini A. Biomimetic gradient hydrogels for tissue engineering. The Canadian Journal of Chemical 135 Engineering. 2010 Dec;88(6):899–911. 140. Seidi A, Ramalingam M, Elloumi-Hannachi I, Ostrovidov S, Khademhosseini A. Gradient biomaterials for soft-to-hard interface tissue engineering. Acta Biomaterialia. 2011 Apr 1;7(4):1441–51. 141. Richards D, Jia J, Yost M, Markwald R, Mei Y. 3D Bioprinting for Vascularized Tissue Fabrication. Ann Biomed Eng. 2017 Jan 1;45(1):132–47. 142. Das S, Gordián-Vélez WJ, Ledebur HC, Mourkioti F, Rompolas P, Chen HI, et al. Innervation: the missing link for biofabricated tissues and organs. npj Regen Med. 2020 Jun 5;5(1):1–19. 143. Clark CC, Aleman J, Mutkus L, Skardal A. A mechanically robust thixotropic collagen and hyaluronic acid bioink supplemented with gelatin nanoparticles. Bioprinting. 2019 Dec 1;16:e00058. 144. Ng WL, Yeow CHE, Huang X, Lee S, Yeong WY. Physically crosslinked gelatin bio-inks with enhanced printability, degradation and mechanical robustness for multi-modal bioprinting. Interdisciplinary Medicine. n/a(n/a):e20250058. 145. Amin AK, Huntley JS, Bush PG, Simpson AHRW, Hall AC. Osmolarity Influences Chondrocyte Death in Wounded Articular Cartilage. JBJS. 2008 Jul 1;90(7):1531. 146. Luh EH, Shackford SR, Shatos MA, Pietropaoli JA. The Effects of Hyperosmolarity on the Viability and Function of Endothelial Cells. Journal of Surgical Research. 1996 Jan 1;60(1):122–8. 147. Saito Y, Yoshida Y, Akazawa T, Takahashi K, Niki E. Cell Death Caused by Selenium Deficiency and Protective Effect of Antioxidants*. Journal of Biological Chemistry. 2003 Oct 10;278(41):39428–34. 148. Gudapati H, Yan J, Huang Y, Chrisey DB. Alginate gelation-induced cell death during laser-assisted cell printing. Biofabrication. 2014 Aug;6(3):035022. 149. Adhikari J, Roy A, Das A, Ghosh M, Thomas S, Sinha A, et al. Effects of Processing Parameters of 3D Bioprinting on the Cellular Activity of Bioinks. Macromolecular Bioscience. 2021;21(1):2000179. 150. Boularaoui S, Al Hussein G, Khan KA, Christoforou N, Stefanini C. An overview of extrusion-based bioprinting with a focus on induced shear stress and its effect on cell viability. Bioprinting. 2020 Dec 1;20:e00093. 151. Phillips TJ. Tissue-Engineered Skin: An Alternative to Split-Thickness Skin Grafts? Archives of Dermatology. 1999 Aug 1;135(8):977–8. 136 152. Ng WL, Yeong WY, Naing MW. Polyvinylpyrrolidone-Based Bio-Ink Improves Cell Viability and Homogeneity during Drop-On-Demand Printing. Materials (Basel). 2017 Feb 16;10(2):190. 153. O’Brien FJ. Biomaterials & scaffolds for tissue engineering. Materials Today. 2011 Mar 1;14(3):88–95. 154. Tran V, Wen X. 4 - Rapid prototyping technologies for tissue regeneration. In: Narayan R, editor. Rapid Prototyping of Biomaterials [Internet]. Woodhead Publishing; 2014 [cited 2025 May 1]. p. 97–155. Available from: https://www.sciencedirect.com/science/article/pii/B9780857095992500040 155. Patel DB, Gray KM, Santharam Y, Lamichhane TN, Stroka KM, Jay SM. Impact of cell culture parameters on production and vascularization bioactivity of mesenchymal stem cell-derived extracellular vesicles. Bioengineering & Translational Medicine. 2017;2(2):170–9. 156. Zhou H, Weir MD, Xu HHK. Effect of Cell Seeding Density on Proliferation and Osteodifferentiation of Umbilical Cord Stem Cells on Calcium Phosphate Cement-Fiber Scaffold. Tissue Engineering Part A. 2011 Nov;17(21–22):2603–13. 157. Xu H, Martinez Salazar DM, Xu C. Investigation of Cell Concentration Change and Cell Aggregation Due to Cell Sedimentation during Inkjet-Based Bioprinting of Cell-Laden Bioink. Machines. 2022 May;10(5):315. 158. Launiere CA, Czaplewski GJ, Myung JH, Hong S, Eddington DT. Rheologically biomimetic cell suspensions for decreased cell settling in microfluidic devices. Biomed Microdevices. 2011 Jun 1;13(3):549–57. 159. Lamping SR, Zhang H, Allen B, Ayazi Shamlou P. Design of a prototype miniature bioreactor for high throughput automated bioprocessing. Chemical Engineering Science. 2003 Feb 1;58(3):747–58. 160. Trampe E, Koren K, Akkineni AR, Senwitz C, Krujatz F, Lode A, et al. Functionalized Bioink with Optical Sensor Nanoparticles for O2 Imaging in 3D- Bioprinted Constructs. Advanced Functional Materials. 2018;28(45):1804411. 161. Figueiredo L, Le Visage C, Weiss P, Yang J. Quantifying Oxygen Levels in 3D Bioprinted Cell-Laden Thick Constructs with Perfusable Microchannel Networks. Polymers. 2020 Jun;12(6):1260. 162. Kim SH, Kwon JS, Cho JG, Park KG, Lim TH, Kim MS, et al. Non-invasive in vivo monitoring of transplanted stem cells in 3D-bioprinted constructs using near-infrared fluorescent imaging. Bioengineering & Translational Medicine. 2021;6(2):e10216. 163. Zhao L, Lee VK, Yoo SS, Dai G, Intes X. The integration of 3-D cell printing 137 and mesoscopic fluorescence molecular tomography of vascular constructs within thick hydrogel scaffolds. Biomaterials. 2012 Jul 1;33(21):5325–32. 164. Hong S, Samson AAS, Song JM. Application of fluorescence resonance energy transfer to bioprinting. TrAC Trends in Analytical Chemistry. 2020 Jan 1;122:115749. 165. Huang X, Ng WL, Yeong WY. Predicting the number of printed cells during inkjet-based bioprinting process based on droplet velocity profile using machine learning approaches. J Intell Manuf. 2024 Jun 1;35(5):2349–64. 166. Chen X, O’Mahony AP, Barber T. The assessment of average cell number inside in-flight 3D printed droplets in microvalve-based bioprinting | Journal of Applied Physics | AIP Publishing. Journal of Applied Physics [Internet]. 2022 Jun 10 [cited 2025 May 7];131(224701). Available from: https://pubs.aip.org/aip/jap/article/131/22/224701/2837051/The-assessment-of- average-cell-number-inside-in 167. Lee SM, Han N, Lee R, Choi IH, Park YB, Shin JS, et al. Real-time monitoring of 3D cell culture using a 3D capacitance biosensor. Biosensors and Bioelectronics. 2016 Mar;77:56–61. 168. Narayanan LK, Thompson TL, Shirwaiker RA, Starly B. Label free process monitoring of 3D bioprinted engineered constructs via dielectric impedance spectroscopy. Biofabrication. 2018 Jun 28;10(3):035012. 169. Africk SA, Colton CK. System and method for ultrasonic measuring of particle properties - Google Patents [Internet]. US7543480B2, 2009 [cited 2025 May 13]. Available from: https://patents.google.com/patent/US7543480B2/en 170. Mercado KP, Helguera M, Hocking DC, Dalecki D. Estimating Cell Concentration in Three-Dimensional Engineered Tissues Using High Frequency Quantitative Ultrasound. Ann Biomed Eng. 2014 Jun 1;42(6):1292–304. 171. Boularaoui S, Al Hussein G, Khan KA, Christoforou N, Stefanini C. An overview of extrusion-based bioprinting with a focus on induced shear stress and its effect on cell viability. Bioprinting. 2020 Dec 1;20:e00093. 172. Aguado BA, Mulyasasmita W, Su J, Lampe KJ, Heilshorn SC. Improving viability of stem cells during syringe needle flow through the design of hydrogel cell carriers. Tissue Eng Part A. 2012 Apr;18(7–8):806–15. 173. Chang R, Nam J, Sun W. Effects of dispensing pressure and nozzle diameter on cell survival from solid freeform fabrication-based direct cell writing. Tissue Eng Part A. 2008 Jan;14(1):41–8. 174. Watts MJ, Linch DC. Optimisation and quality control of cell processing for 138 autologous stem cell transplantation. British Journal of Haematology. 2016;175(5):771–83. 175. Wu MH, Urban JPG, Cui ZF, Cui Z, Xu X. Effect of extracellular ph on matrix synthesis by chondrocytes in 3D agarose gel. Biotechnol Prog. 2007;23(2):430–4. 176. Xu HQ, Liu JC, Zhang ZY, Xu CX. A review on cell damage, viability, and functionality during 3D bioprinting. Military Medical Research. 2022 Dec 16;9(1):70. 177. Shi J, Wu B, Li S, Song J, Song B, Lu WF. Shear stress analysis and its effects on cell viability and cell proliferation in drop-on-demand bioprinting. Biomed Phys Eng Express. 2018 Jun;4(4):045028. 178. Li M, Tian X, Kozinski JA, Chen X, Hwang DK. Modeling mechanical cell damage in the bioprinting process employing a conical needle. J Mech Med Biol. 2015 Oct;15(05):1550073. 179. Aguado BA, Mulyasasmita W, Su J, Lampe KJ, Heilshorn SC. Improving viability of stem cells during syringe needle flow through the design of hydrogel cell carriers. Tissue Eng Part A. 2012 Apr;18(7–8):806–15. 180. Singh S, Choudhury D, Yu F, Mironov V, Naing MW. In situ bioprinting – Bioprinting from benchside to bedside? Acta Biomaterialia. 2020 Jan 1;101:14–25. 181. Ravanbakhsh H, Karamzadeh V, Bao G, Mongeau L, Juncker D, Zhang YS. Emerging Technologies in Multi-Material Bioprinting. Advanced Materials. 2021;33(49):2104730. 182. Patel P, Markx GH. Dielectric measurement of cell death. Enzyme and Microbial Technology. 2008 Dec;43(7):463–70. 183. Sharma R, Blackburn T, Hu W, Wiltberger K, Velev OD. On-chip microelectrode impedance analysis of mammalian cell viability during biomanufacturing. Biomicrofluidics. 2014 Sep 11;8(5):054108. 184. Jia W, Gungor-Ozkerim PS, Zhang YS, Yue K, Zhu K, Liu W, et al. Direct 3D bioprinting of perfusable vascular constructs using a blend bioink. Biomaterials. 2016 Nov 1;106:58–68. 185. Andrade TAM, da Silva VA, Scheck K, Garay T, Sharma R, Willerth SM. 3D Bioprinting a Novel Skin Co-Culture Model Using Human Keratinocytes and Fibroblasts. Journal of Biomedical Materials Research Part A. 2025 Jan;113(1):e37831. 186. Skardal A, Mack D, Kapetanovic E, Atala A, Jackson JD, Yoo J, et al. Bioprinted Amniotic Fluid-Derived Stem Cells Accelerate Healing of Large Skin 139 Wounds. Stem Cells Translational Medicine. 2012 Nov 1;1(11):792–802. 187. Min D, Lee W, Bae IH, Lee TR, Croce P, Yoo SS. Bioprinting of biomimetic skin containing melanocytes. Experimental Dermatology. 2018;27(5):453–9. 188. Michael S, Sorg H, Peck CT, Koch L, Deiwick A, Chichkov B, et al. Tissue Engineered Skin Substitutes Created by Laser-Assisted Bioprinting Form Skin-Like Structures in the Dorsal Skin Fold Chamber in Mice. PLOS ONE. 2013 Mar 4;8(3):e57741. 189. Koch L, Deiwick A, Schlie S, Michael S, Gruene M, Coger V, et al. Skin tissue generation by laser cell printing. Biotechnology and Bioengineering. 2012;109(7):1855–63. 190. Tarassoli SP, Jessop ZM, Al-Sabah A, Gao N, Whitaker S, Doak S, et al. Skin tissue engineering using 3D bioprinting: An evolving research field. Journal of Plastic, Reconstructive & Aesthetic Surgery. 2018 May 1;71(5):615–23. 191. Pourchet LJ, Thepot A, Albouy M, Courtial EJ, Boher A, Blum LJ, et al. Human Skin 3D Bioprinting Using Scaffold-Free Approach. Advanced Healthcare Materials. 2017;6(4):1601101. 192. Asami K. Dielectric spectroscopy reveals nanoholes in erythrocyte ghosts. Soft Matter. 2012 Feb 22;8(11):3250–7. 193. Spencer D, Morgan H. High-Speed Single-Cell Dielectric Spectroscopy. ACS Sens. 2020 Feb 28;5(2):423–30. 194. Sarró E, Lecina M, Fontova A, Gòdia F, Bragós R, Cairó JJ. Real-time and on- line monitoring of morphological cell parameters using electrical impedance spectroscopy measurements. Journal of Chemical Technology & Biotechnology. 2016;91(6):1755–62. 195. Stacey MW, Sabuncu AC, Beskok A. Dielectric characterization of costal cartilage chondrocytes. Biochimica et Biophysica Acta (BBA) - General Subjects. 2014 Jan;1840(1):146–52. 196. Shohan S, Zeng Y, Chen X, Jin R, Shirwaiker R. Investigating dielectric spectroscopy and soft sensing for nondestructive quality assessment of engineered tissues. Biosensors and Bioelectronics. 2022 Nov;216:114286. 197. Sabuncu AC, Asmar AJ, Stacey MW, Beskok A. Differential dielectric responses of chondrocyte and Jurkat cells in electromanipulation buffers. Electrophoresis. 2015 Jul;36(13):1499–506. 198. Daoud J, Asami K, Rosenberg L, Tabrizian M. Dielectric spectroscopy for non-invasive monitoring of epithelial cell differentiation within three-dimensional 140 scaffolds. Phys Med Biol. 2012;57. 199. Bagnaninchi PO, Drummond N. Real-time label-free monitoring of adipose- derived stem cell differentiation with electric cell-substrate impedance sensing. Proc Natl Acad Sci USA. 2011 Apr 19;108(16):6462–7. 200. Nordberg RC, Zhang J, Griffith EH, Frank MW, Starly B, Loboa EG. Electrical Cell-Substrate Impedance Spectroscopy Can Monitor Age-Grouped Human Adipose Stem Cell Variability During Osteogenic Differentiation. Stem Cells Translational Medicine. 2017 Feb 1;6(2):502–11. 201. Lincoln CK, Gabridge MG. Chapter 4 Cell Culture Contamination: Sources, Consequences, Prevention, and Elimination. In: Mather JP, Barnes D, editors. Methods in Cell Biology [Internet]. Academic Press; 1998 [cited 2025 Jun 1]. p. 49–65. (Animal Cell Culture Methods; vol. 57). Available from: https://www.sciencedirect.com/science/article/pii/S0091679X0861571X 202. Myers DK, Goldberg AM, Poth A, Wolf MF, Carraway J, McKim J, et al. From in vivo to in vitro: The medical device testing paradigm shift. ALTEX - Alternatives to animal experimentation. 2017 Nov 1;34(4):479–500. 203. Wade W. Unculturable bacteria—the uncharacterized organisms that cause oral infections. J R Soc Med. 2002 Feb;95(2):81–3. 204. Bodor A, Bounedjoum N, Vincze GE, Erdeiné Kis Á, Laczi K, Bende G, et al. Challenges of unculturable bacteria: environmental perspectives. Rev Environ Sci Biotechnol. 2020 Mar 1;19(1):1–22. 205. Hahn MW, Koll U, Schmidt J. Isolation and Cultivation of Bacteria. In: Hurst CJ, editor. The Structure and Function of Aquatic Microbial Communities [Internet]. Cham: Springer International Publishing; 2019 [cited 2025 May 10]. p. 313–51. Available from: https://doi.org/10.1007/978-3-030-16775-2_10 206. Han XX, Rodriguez RS, Haynes CL, Ozaki Y, Zhao B. Surface-enhanced Raman spectroscopy. Nat Rev Methods Primers. 2022 Jan 6;1(1):1–17. 207. Benavides OR, Gibbs HC, White BP, Kaunas R, Gregory CA, Walsh AJ, et al. Volumetric imaging of human mesenchymal stem cells (hMSCs) for non- destructive quantification of 3D cell culture growth. PLOS ONE. 2023 Mar 28;18(3):e0282298. 208. Cortesi M, Giordano E. Non-destructive monitoring of 3D cell cultures: new technologies and applications. PeerJ. 2022 May 12;10:e13338. 209. Albakri B, Diniz ATS, Benner P, Muth T, Nakajima S, Favaro M, et al. Machine learning-assisted equivalent circuit identification for dielectric spectroscopy of polymers. Electrochimica Acta. 2024 Aug 20;496:144474. 141 210. Wang S, Bingke Y, Zhou Y, Qu Y, Zhang D, Chen J, et al. Dielectric spectroscopy with machine learning integrated approach to analysis the influence of N:P ratio on algae growth modeling for forecasting. Microchemical Journal. 2024 May 1;200:110254. 211. Shohan S, Harm J, Hasan M, Starly B, Shirwaiker R. Non-destructive quality monitoring of 3D printed tissue scaffolds via dielectric impedance spectroscopy and supervised machine learning. Procedia Manufacturing. 2021;53:636–43. 212. Fouquet C, Gilles JF, Heck N, Santos MD, Schwartzmann R, Cannaya V, et al. Improving Axial Resolution in Confocal Microscopy with New High Refractive Index Mounting Media. PLOS ONE. 2015 Mar 30;10(3):e0121096. 213. Ember KJI, Hoeve MA, McAughtrie SL, Bergholt MS, Dwyer BJ, Stevens MM, et al. Raman spectroscopy and regenerative medicine: a review. npj Regen Med. 2017 May 15;2(1):1–10. 214. Buma P, Meel M van, Tienen TG van, Veth RPH. Tissue Engineering of the Meniscus. In: Pietrzak WS, editor. Musculoskeletal Tissue Regeneration: Biological Materials and Methods [Internet]. Totowa, NJ: Humana Press; 2008 [cited 2024 Dec 19]. p. 327–46. Available from: https://doi.org/10.1007/978-1- 59745-239-7_16 215. Iannucci LE, Boys AJ, McCorry MC, Estroff LA, Bonassar LJ. Cellular and Chemical Gradients to Engineer the Meniscus-to-Bone Insertion. Advanced Healthcare Materials. 2019;8(7):1800806. 216. Lawlor KT, Vanslambrouck JM, Higgins JW, Chambon A, Bishard K, Arndt D, et al. Cellular extrusion bioprinting improves kidney organoid reproducibility and conformation. Nat Mater. 2021 Feb;20(2):260–71. 217. Wang D, Guo Y, Zhu J, Liu F, Xue Y, Huang Y, et al. Hyaluronic acid methacrylate/pancreatic extracellular matrix as a potential 3D printing bioink for constructing islet organoids. Acta Biomaterialia. 2023 Jul 15;165:86–101. 218. Jorgensen AM, Yoo JJ, Atala A. Solid Organ Bioprinting: Strategies to Achieve Organ Function. Chem Rev. 2020 Oct 14;120(19):11093–127. 219. Cohen DL, Lipton JI, Bonassar LJ, Lipson H. Additive manufacturing for in situ repair of osteochondral defects. Biofabrication. 2010 Sep 1;2(3):035004. 220. Mirdamadi E, Tashman JW, Shiwarski DJ, Palchesko RN, Feinberg AW. FRESH 3D Bioprinting a Full-Size Model of the Human Heart. ACS Biomater Sci Eng. 2020 Nov 9;6(11):6453–9. 221. You S, Xiang Y, Hwang HH, Berry DB, Kiratitanaporn W, Guan J, et al. High cell density and high-resolution 3D bioprinting for fabricating vascularized tissues. 142 Science Advances. 2023 Feb 22;9(8):eade7923. 222. Zimmermann R, Hentschel C, Schrön F, Moedder D, Büttner T, Atallah P, et al. High resolution bioprinting of multi-component hydrogels. Biofabrication. 2019 Jul;11(4):045008. 223. Altunbek M, Afghah F, Caliskan OS, Yoo JJ, Koc B. Design and bioprinting for tissue interfaces. Biofabrication. 2023 Feb;15(2):022002. 224. Krouwels A, Melchels FPW, van Rijen MHP, Öner FC, Dhert WJA, Tryfonidou MA, et al. Comparing Hydrogels for Human Nucleus Pulposus Regeneration: Role of Osmolarity During Expansion. Tissue Engineering Part C: Methods. 2018 Apr;24(4):222–32. 225. Narayanan LK, Thompson TL, Bhat A, Starly B, Shirwaiker RA. Investigating Dielectric Impedance Spectroscopy As a Non-Destructive Quality Assessment Tool for 3D Cellular Constructs. In: Volume 4: Bio and Sustainable Manufacturing [Internet]. Los Angeles, California, USA: American Society of Mechanical Engineers; 2017 [cited 2024 Apr 23]. p. V004T05A013. Available from: https://asmedigitalcollection.asme.org/MSEC/proceedings/MSEC2017/50756/Los %20Angeles,%20California,%20USA/268644 226. Niemeyer P, Schubert T, Grebe M, Hoburg A. Treatment Costs of Matrix- Associated Autologous Chondrocyte Implantation Compared With Microfracture: Results of a Matched-Pair Claims Data Analysis on the Treatment of Cartilage Knee Defects in Germany. Orthop J Sports Med. 2019 Dec 5;7(12):2325967119886583. 227. Kleintjes W, Kleintjes W, Thomas G, Salence B, Thaele S, Stevenson N, et al. Clinics in Surgery A Novel Technique for Composite Cultured Epithelial Autograft in a Patient with Extensive Burn Wounds: A Case Report OPEN ACCESS Citation: A Novel Technique for Composite Cultured Epithelial Autograft in a Patient with Extensive Burn Wounds: A Case. Clin Surg. 2017 Aug 1;2(1579). 228. Wilson BC, Jermyn M, Leblond F. Challenges and opportunities in clinical translation of biomedical optical spectroscopy and imaging. J Biomed Opt. 2018 Mar 6;23(03):1. 229. Chang SCN, Rowley JA, Tobias G, Genes NG, Roy AK, Mooney DJ, et al. Injection molding of chondrocyte/alginate constructs in the shape of facial implants. J Biomed Mater Res. 2001 Jun 15;55(4):503–11. 230. Opel CF, Li J, Amanullah A. Quantitative modeling of viable cell density, cell size, intracellular conductivity, and membrane capacitance in batch and fed‐batch CHO processes using dielectric spectroscopy. Biotechnology Progress. 2010 Jul;26(4):1187–99. 143 231. Immohr MB, Adrego FDS, Teichert HL, Schmidt V, Sugimura Y, Bauer S, et al. 3D-bioprinting of aortic valve interstitial cells: impact of hydrogel and printing parameters on cell viability. Biomed Mater. 2022 Nov;18(1):015004. 232. Genes NG, Rowley JA, Mooney DJ, Bonassar LJ. Effect of substrate mechanics on chondrocyte adhesion to modified alginate surfaces. Archives of Biochemistry and Biophysics. 2004 Feb;422(2):161–7. 233. Voss JR, Lu Y, Edwards RB, Bogdanske JJ, Markel MD. Effects of thermal energy on chondrocyte viability. American Journal of Veterinary Research. 2006 Oct 1;67(10):1708–12. 234. Li Y, Wang H, Tingrui P. Intracellular ice formation (IIF) during freeze–thaw repetitions. International Journal of Heat and Mass Transfer. 2013 Sep 1;64:436– 43. 235. Otsu N. A Threshold Selection Method from Gray-Level Histograms. IEEE Transactions on Systems, Man, and Cybernetics. 1979 Jan;9(1):62–6. 236. Prada J, Vega-Castillo P, Krautschneider W. Design of a Wide Tuning-Range CMOS 130-nm Quadrature VCO for Cell Impedance Spectroscopy. In Hamburg, Germany: IEEE conference; 2015. 237. Davey CL, Davey HM, Kell DB, Todd RW. Introduction to the dielectric estimation of cellular biomass in real time, with special emphasis on measurements at high volume fractions. Analytica Chimica Acta. 1993 Jul 1;279(1):155–61. 238. Canali C, Heiskanen A, Muhammad HB, Høyum P, Pettersen FJ, Hemmingsen M, et al. Bioimpedance monitoring of 3D cell culturing—Complementary electrode configurations for enhanced spatial sensitivity. Biosensors and Bioelectronics. 2015 Jan 15;63:72–9. 239. Afshar S, Fazelkhah A, Braasch K, Salimi E, Butler M, Thomson DJ, et al. Full Beta-Dispersion Region Dielectric Spectra and Dielectric Models of Viable and Non-Viable CHO Cells. IEEE J Electromagn RF Microw Med Biol. 2021 Mar;5(1):70–7. 240. Bao JZ, Davis CC, Schmukler RE. Frequency domain impedance measurements of erythrocytes. Constant phase angle impedance characteristics and a phase transition. Biophys J. 1992 May;61(5):1427–34. 241. Abbadessa A, Crecente-Campo J, Alonso MJ. Engineering Anisotropic Meniscus: Zonal Functionality and Spatiotemporal Drug Delivery. Tissue Engineering Part B: Reviews. 2021 Apr;27(2):133–54. 242. Lv H, Deng G, Lai J, Yu Y, Chen F, Yao J. Advances in 3D Bioprinting of Biomimetic and Engineered Meniscal Grafts. Macromolecular Bioscience. 144 2023;23(12):2300199. 243. Lee CH, Rodeo SA, Fortier LA, Lu C, Erisken C, Mao JJ. Protein-releasing polymeric scaffolds induce fibrochondrocytic differentiation of endogenous cells for knee meniscus regeneration in sheep. Science Translational Medicine. 2014 Dec 10;6(266):266ra171-266ra171. 244. Feng Y, Huang L, Zhao P, Liang F, Wang W. A Microfluidic Device Integrating Impedance Flow Cytometry and Electric Impedance Spectroscopy for High-Efficiency Single-Cell Electrical Property Measurement. Anal Chem. 2019 Dec 3;91(23):15204–12. 245. Aljabal Z. Dielectric spectroscopy of organic solvents of varying polarity [Internet] [M.Sc.]. [Department of Physics and Physical Oceanography]: Memorial University of Newfoundland; 2016 [cited 2024 May 20]. Available from: https://www.semanticscholar.org/paper/Dielectric-spectroscopy-of-organic- solvents-of-Poduska-Cheng/742308d5f05aa01c1adac5fd4be26664a58a1fb6 246. Di Biasio A, Ambrosone L, Cametti C. The Dielectric Behavior of Nonspherical Biological Cell Suspensions: An Analytic Approach. Biophysical Journal. 2010 Jul;99(1):163–74. 247. Davey CL, Kell DB. The influence of electrode polarisation on dielectric spectra, with special reference to capacitive biomass measurements. II: Reduction in the contribution of electrode polarisation to dielectric spectra using a two- frequency method. Bioelectrochemistry and Bioenergetics. 1998 Aug;46(1):105– 14. 248. Ishai PB, Talary MS, Caduff A, Levy E, Feldman Y. Electrode polarization in dielectric measurements: a review. Meas Sci Technol. 2013 Aug;24(10):102001. 249. Mazzeo BrianA, Flewitt AJ. Two- and four-electrode, wide-bandwidth, dielectric spectrometer for conductive liquids: Theory, limitations, and experiment. Journal of Applied Physics. 2007 Nov 19;102(10):104106. 250. Schwan HP. Electrode Polarization Impedance and Measurements in Biological Materials. Annals of the New York Academy of Sciences. 1968;148(1):191–209. 251. Paivana G, Mavrikou S, Kaltsas G, Kintzios S. Bioelectrical Analysis of Various Cancer Cell Types Immobilized in 3D Matrix and Cultured in 3D-Printed Well. Biosensors (Basel). 2019 Nov 14;9(4):136. 252. Corbett DC, Olszewski E, Stevens K. A FRESH Take on Resolution in 3D Bioprinting. Trends in Biotechnology. 2019 Nov 1;37(11):1153–5. 253. Brunel LG, Christakopoulos F, Kilian D, Cai B, Hull SM, Myung D, et al. 145 Embedded 3D Bioprinting of Collagen Inks into Microgel Baths to Control Hydrogel Microstructure and Cell Spreading. Advanced Healthcare Materials. 2024;13(25):2303325. 254. Freeman S, Calabro S, Williams R, Jin S, Ye K. Bioink Formulation and Machine Learning-Empowered Bioprinting Optimization. Front Bioeng Biotechnol [Internet]. 2022 Jun 13 [cited 2025 Jun 18];10. Available from: https://www.frontiersin.org/journals/bioengineering-and- biotechnology/articles/10.3389/fbioe.2022.913579/full 255. Lee J, Oh SJ, An SH, Kim WD, Kim SH. Machine learning-based design strategy for 3D printable bioink: elastic modulus and yield stress determine printability. Biofabrication. 2020 May;12(3):035018. 256. Mukherjee A, Sarker S, Kumar R, Sahani A, Das B. Chapter 9 - Optimization of bio-ink using machine learning. In: Sharma CP, Chandy T, Thomas V, editors. Artificial Intelligence in Tissue and Organ Regeneration [Internet]. Academic Press; 2023 [cited 2025 Jun 18]. p. 155–74. Available from: https://www.sciencedirect.com/science/article/pii/B9780443184987000168 257. Deniz I, Imamoglu E, Keskin-Gundogdu T, Deniz I, Imamoglu E, Keskin- Gundogdu T. Properties and Applications of Alginates [Internet]. InTechOpen; 2022 [cited 2025 Jun 9]. Available from: https://www.intechopen.com/books/10670 258. Smidsrød O. Solution properties of alginate. Carbohydrate Research. 1970 Jun 1;13(3):359–72. 259. Duggirala S, Deluca PP. Rheological Characterization of Cellulosic and Alginate Polymers. PDA Journal of Pharmaceutical Science and Technology. 1996 Sep 1;50(5):290–6. 260. Gao T, Gillispie GJ, Copus JS, PR AK, Seol YJ, Atala A, et al. Optimization of gelatin–alginate composite bioink printability using rheological parameters: a systematic approach. Biofabrication. 2018 Jun;10(3):034106. 261. Jia J, Richards DJ, Pollard S, Tan Y, Rodriguez J, Visconti RP, et al. Engineering alginate as bioink for bioprinting. Acta Biomaterialia. 2014 Oct 1;10(10):4323–31. 262. Yang J, Chen S, Fang Y. Viscosity study of interactions between sodium alginate and CTAB in dilute solutions at different pH values. Carbohydrate Polymers. 2009 Jan 22;75(2):333–7. 263. Fernandes PCG, Filgueiras VF, Matte BF, Lopes JH. A comprehensive rheological study on the influence of ion charge density and valence in ionotropically crosslinked alginate hydrogels for bioprinting. International Journal of Biological Macromolecules. 2025 Jun 1;316:144720. 146 264. Smidsrod O, Skjakbrk G. Alginate as immobilization matrix for cells. Trends in Biotechnology. 1990;8:71–8. 265. Apelgren P, Amoroso M, Lindahl A, Brantsing C, Rotter N, Gatenholm P, et al. Chondrocytes and stem cells in 3D-bioprinted structures create human cartilage in vivo. PLOS ONE. 2017 Dec 13;12(12):e0189428. 266. Lan X, Boluk Y, Adesida AB. 3D Bioprinting of Hyaline Cartilage Using Nasal Chondrocytes. Ann Biomed Eng. 2024 Jul 1;52(7):1816–34. 267. Kuzucu M, Vera G, Beaumont M, Fischer S, Wei P, Shastri VP, et al. Extrusion-Based 3D Bioprinting of Gradients of Stiffness, Cell Density, and Immobilized Peptide Using Thermogelling Hydrogels. ACS Biomater Sci Eng. 2021 Jun 14;7(6):2192–7. 268. Das R h. j., van Osch G j. v. m., Kreukniet M, Oostra J, Weinans H, Jahr H. Effects of individual control of pH and hypoxia in chondrocyte culture. Journal of Orthopaedic Research. 2010;28(4):537–45. 269. Hall A, Horwitz E, Wilkins R. The cellular physiology of articular cartilage. Experimental Physiology. 1996;81(3):535–45. 270. Wu Q qian, Chen Q. Mechanoregulation of Chondrocyte Proliferation, Maturation, and Hypertrophy: Ion-Channel Dependent Transduction of Matrix Deformation Signals. Experimental Cell Research. 2000 May 1;256(2):383–91. 271. Huang Y, Zhang Y, Ding X, Liu S, Sun T. Osmolarity influences chondrocyte repair after injury in human articular cartilage. J Orthop Surg Res. 2015 Jan 28;10(1):19. 272. Mang T, Lindemann S, Gigout A. Increasing the Medium Osmolarity Reduces the Inflammatory Status of Human OA Chondrocytes and Increases Their Responsiveness to GDF-5. International Journal of Molecular Sciences. 2020 Jan;21(2):531. 273. Lopez Hernandez H, Grosskopf AK, Stapleton LM, Agmon G, Appel EA. Non-Newtonian Polymer–Nanoparticle Hydrogels Enhance Cell Viability during Injection. Macromolecular Bioscience. 2019;19(1):1800275. 274. Asami K. Characterization of biological cells by dielectric spectroscopy. Journal of Non-Crystalline Solids. 2002 Jul;305(1–3):268–77. 275. Szunerits S, Pagneux Q, M’Barek YB, Vassal S, Boukherroub R. Do not let electrode fouling be the enemy of bioanalysis. Bioelectrochemistry. 2023 Oct 1;153:108479. 276. Ahmed R, Reifsnider K. Study of Influence of Electrode Geometry on 147 Impedance Spectroscopy. In American Society of Mechanical Engineers Digital Collection; 2010 [cited 2025 Jun 25]. p. 167–75. Available from: https://dx.doi.org/10.1115/FuelCell2010-33209 277. Fan B, Wolfrum B, Robinson JT. Impedance scaling for gold and platinum microelectrodes. J Neural Eng. 2021 Sep;18(5):056025. 278. Geise RJ, Adams JM, Barone NJ, Yacynych AM. Electropolymerized films to prevent interferences and electrode fouling in biosensors. Biosensors and Bioelectronics. 1991 Jan 1;6(2):151–60. 279. Hanssen BL, Siraj S, Wong DKY. Recent strategies to minimise fouling in electrochemical detection systems. Reviews in Analytical Chemistry. 2016 Apr 1;35(1):1–28.