52 lines
102 KiB
JSON
52 lines
102 KiB
JSON
[
|
||
{
|
||
"id": 1,
|
||
"chunk": "# High-Throughput Synthesis, Analysis, and Optimization of Injectable Hydrogels for Protein Delivery \n\nFei Xu, Brandon Corbett, Sydney Bell, Chiyan Zhang, Monika Budi Hartono, Zohreh Jomeh Farsangi, John MacGregor, and Todd Hoare\\* \n\nDepartment of Chemical Engineering, McMaster University, 1280 Main Street West, Hamilton, Ontario L8S 4L8, Canada \n\nSupporting Information \n\nABSTRACT: The development of in situ-gelling hydrogels that can enable prolonged protein release is increasingly important due to the emergence of a growing number of protein-based therapeutics. Herein, we describe a highthroughput strategy to fabricate, characterize, and subsequently optimize hydrazone-cross-linked in situ-gelling hydrogels for protein delivery. Hydrogels are fabricated using an automated high-throughput robot to mix a variety of thermoresponsive, nonthermoresponsive, charged, neutral, naturally sourced, and synthetic polymers functionalized with hydrazide or aldehyde groups, generating in situ-gelling \n\n \n\nhydrogels with well-defined compositions within a 96-well plate. High-throughput characterization strategies are subsequently developed to enable on-plate analysis of hydrogel swelling, mechanics, degradation, transparency, and protein (ovalbumin) release kinetics that yield results consistent with those collected using traditional bulk hydrogel analysis techniques. Dynamic regression and latent variable modeling are then applied to fit performance statistics to the collected data set; subsequently, numerical optimization is used to identify mixtures of precursor polymers that exhibit targeted combinations of minimal burst release, maximum total protein release, minimum release rate, and maximum transparency (the latter of particular relevance for ophthalmic protein delivery applications). Given the rapid throughput of the protocols developed (i.e., 126 hydrogels can be synthesized and screened in quadruplicate within hours), this approach offers particular promise for accelerating the identification of injectable hydrogel compositions relevant for both protein delivery as well as other biomedical applications for which clearly predefined materials properties are required.",
|
||
"category": " Abstract"
|
||
},
|
||
{
|
||
"id": 2,
|
||
"chunk": "# INTRODUCTION \n\nIn situ gelling hydrogels that can spontaneously gel in the body via physical1−3 and chemical reactions4,5 have attracted significant interest in the context of drug delivery and, more specifically, protein delivery. The internal gel porosity, typically on the same few nm length scale of most proteins, can be engineered by manipulating the cross-linker density6 and the structure of the gel building blocks7,8 to either irreversibly entrap the protein9,10 or control protein release at a rate proportional to the relative sizes of the protein and the pore network created.11 Furthermore, by introducing affinity groups into the hydrogel that can interact with proteins (e.g., charged moieties or hydrophobic domains), the affinity of the protein for the gel phase can be engineered to further tune release kinetics.12,13 If the in situ chemistry creates a degradable crosslink, further control over protein release kinetics can be achieved based on not only diffusion/affinity but also the rate of bulk matrix degradation.14−16 Leveraging these advantages, in situ-gelling hydrogels have been demonstrated to enable, for example, sustained release of vascular endothelium growth factor (VEGF) or bevacizumab for eye diseases.17,18 \n\nIn designing hydrogels specifically for protein delivery applications, the choice of the constituent polymer is particularly critical. The majority of effort has focused on one of two classes of polymers: (1) naturally occurring polymers that can be degraded via various oxidative or enzymatic pathways in vivo and have established records of biocompatibility (e.g., alginate, chitosan, dextran, hyaluronic acid, and various soluble cellulose derivatives),19,20 but can offer significant batch-to-batch variability and limited options for chemical functionalization, or (2) smart materials that can undergo physical changes in swelling upon the exposure of different environmental stimuli such as temperature, $\\mathrm{\\tt{pH}},$ or ionic strength. Temperature-responsive smart materials such as poly(N-isopropylacrylamide) (PNIPAM)21,22 or poly(ethylene oxide)/poly(propylene oxide) block copolymers (Pluronics23) have attracted particular interest given their potential to both facilitate in situ gelation upon heating from room temperature to physiological temperature as well as deswell in the body to entrap a protein cargo and thus prolong its release. Poly(oligo ethylene glycol methacrylate) (POEGMA) has more recently been explored as an alternative thermoresponsive polymer that exhibits similar properties to PNIPAM but avoids many of the regulatory issues that have limited the translation of PNIPAMbased materials to the clinic.4,24−27 In addition, the lower critical solution temperature (LCST) of POEGMA-based polymers can be tuned from 22 to ${\\tt>}90{\\ ^{\\circ}C}$ by changing the number of ethylene oxide (EO) repeat units in the side chain of the polymer.26−28 However, the successful translation of smart polymers requires solving key challenges around degradation29,30 and avoiding the convective burst release of the protein payload that is often observed as the gel deswells at physiological temperature.31,32 \n\nBased on these inherent limitations of both natural and synthetic polymer-based conventional hydrogels for protein delivery, coupled with the multiple potential variables (e.g., hydrogel cross-link density, degradability, porosity, chemical affinity) that must be optimized for each specific protein to be delivered, the development of hydrogel-based protein delivery vehicles remains a challenging and largely trial-and-error process for each protein type and dose targeted. Given the sharply increasing number of protein-based therapeutics now transitioning into the clinic,33 developing a rapid method to design a controlled delivery vehicle most suitable to deliver a given protein at the required dose over the required time is increasingly important. \n\nIn our previous work, we have extensively explored the use of hydrazide-aldehyde chemistry for creating functional in situgelling hydrogels based on PNIPAM,21 POEGMA,24 charged POEGMA derivatives,12,34 and natural polymers like carboxymethyl cellulose and dextran.35,36 Hydrazone chemistry is attractive, given its combination of rapid gelation (between seconds to minutes following mixing via coextrusion through a double barrel syringe) and hydrolytic degradability (tunable between weeks to months at neutral $\\mathrm{\\pH}$ ).12,24,31 In addition, by mixing different hydrazide and aldehyde precursor polymers with different properties (for example, natural degradable polymers, thermoresponsive polymers,30 and thermoresponsive polymers with different phase transition temperatures31), we have demonstrated the potential to precisely tune hydrogel properties by simple additive mixing, including the potential to prolong the release of model proteins through the formation of phase-separated domains within hydrogels.31 This mixingbased approach is attractive in that multiple hydrogels with significantly different properties can be fabricated by mixing a smaller set of precursor polymers in different ratios, reducing the required synthetic burden for hydrogel optimization. However, if a specific set of properties is desired, even this mixing approach becomes a slow trial-and-error process, particularly if the precursor polymers phase separate or interact in some way that makes the overall properties of the resulting hydrogel nonadditive (and thus less predictable) based on the precursor polymer content. \n\nGiven the clear potential of precursor polymer mixing to generate hydrogels with targeted properties, applying highthroughput fabrication techniques to produce such hydrogels via automated mixing protocols offers a potential solution to this challenge. High-throughput screening techniques have been widely used in the field of drug or cell-based assays;37,38 however, few reports have discussed the use of highthroughput robotics to fabricate and optimize hydrogels. This gap is likely due to the technical challenges inherent in both repeatedly fabricating gels in high throughput, as well as subsequently characterizing the properties of the resulting hydrogel arrays with sufficient speed that synthesizing hydrogels in high-throughput would be practically beneficial. However, if these challenges can be overcome, the high amount of data that can be generated quickly using highthroughput approaches is ideal for not only screening potential hydrogel compositions for particular end-uses but also for applying statistical optimization techniques that can fit multivariate mathematical models to the high-throughput data and (following model inversion) subsequently predict formulations that would offer better performance in terms of achieving a specific set of property targets. We have previously demonstrated the potential of latent variable methods to assist in designing multiresponsive polymers based on historical data;39 by using high-throughput data collection to rapidly collect this historical data, we expect that we can significantly accelerate the identification of hydrogel compositions suitable for protein delivery. \n\nHerein, we report high-throughput techniques to both fabricate and characterize hydrogel arrays based on the programmable mixing of 13 different precursor polymers (12 functionalized with hydrazide groups and 2 functionalized with aldehyde groups). In particular, precursor polymers with various charges (anionic, cationic, neutral), degradabilities (i.e., naturally sourced vs synthetic), and thermoresponsiveness (using both PNIPAM and POEGMA functional copolymers with various LCST values) were mixed in various ratios to fabricate 126 different hydrogels using a high-content robotics system in $<25~\\mathrm{\\min}$ . Subsequently, high-throughput test methods were developed and validated to probe the physical and pharmacokinetic properties of such hydrogels. Fitting a multivariate model to this high-throughput data set was demonstrated to enable a priori prediction of key hydrogel properties relevant to protein delivery; subsequent model inversion enabled the prediction of hydrogel recipes leading to improved protein release kinetics customized to the priorities of each drug release case.",
|
||
"category": " Introduction"
|
||
},
|
||
{
|
||
"id": 3,
|
||
"chunk": "# EXPERIMENTAL SECTION \n\nMaterials. Oligo(ethylene glycol) methyl ether methacrylate $(\\mathrm{OEGMA}_{475},$ $M_{\\mathrm{n}}~=~475~\\mathrm{g/mol})$ and di(ethylene glycol) methyl ether methacrylate $(\\mathbf{M}(\\mathrm{EO})_{2}\\mathbf{MA})$ were purchased from SigmaAldrich and were purified using an aluminum oxide (Sigma-Aldrich, type CG-20) column to remove the methyl ether hydroquinone (MEHQ) and butylated hydroxytoluene (BHT) inhibitors. N-(2,2- Dimethoxyethyl) methacrylamide (DMEMAm) was synthesized following our previous protocol.25 Dextran $\\left(M_{\\mathrm{r}}\\ =\\ 40000\\ \\mathrm{g/mol}\\right)$ chitosan (low molecular weight), sodium carboxymethyl cellulose (CMC, $M_{\\mathrm{w}}=90000\\mathrm{g/mol}$ , $\\mathrm{DS}\\ =\\ 0.9\\ \\cdot$ ), $N,N_{\\astrosun}$ -dimethylaminoethyl methacrylate (DMAEMA), $N$ -isopropylacrylamide (NIPAM, $99\\%$ ), acrylic acid (AA, $99\\%$ ), thioglycolic acid (TGA, $98\\%$ ), $N\\mathrm{.}$ - hydroxysuccinimide (NHS, $97\\%$ ), monochloroacetic acid (MACC), and $N,N.$ -dimethylformamide (DMF, $99\\%$ ) were purchased from Sigma-Aldrich and used as received. Adipic acid dihydrazide (ADH, Alfa Aesar, $98\\%$ ), $N^{\\prime}$ -ethyl-N-(3-(dimethylamino)propyl)-carbodiimide (EDC, Carbosynth, Compton $\\mathrm{CA},$ commercial grade), 2,2- azobisisobutyric acid dimethyl ester (AIBMe, Wako Chemicals, $98.5\\%$ ), and dioxane (Caledon Laboratories, $99\\%$ ) were used as received. Sodium hydroxide (NaOH) and hydrogen chloride (HCl) were purchased from LabChem Inc. and used as received. NIH 3T3 mouse fibroblast cells were purchased from ATCC (Cedarlane Laboratories, Burlington, ON, Canada). Dulbecco’s Modified Eagle’s Medium (DMEM, Life Technologies), fetal bovine serum (FBS, ThermoFisher), penicillin-streptomycin (PS, ThermoFisher), PrestoBlue cell viability reagent (ThermoFisher), a Bradford reagent kit (Sigma-Aldrich), and albumin from chicken egg white (ovalbumin, Sigma-Aldrich) were used as received. Milli- $\\mathrm{\\DeltaQ}$ grade distilled deionized water (DIW) and $10~\\mathrm{mM}$ PBS was used for experiments. \n\nTable 1. Synthetic Recipes for POEGMA Polymer Precursors \n\n\n<html><body><table><tr><td></td><td>OEGMA475 (mmol)</td><td>M(EO)MA (mmol)</td><td>AA (mmol)</td><td>DMEMAm (mmol)</td><td>TGA (mmol)</td><td>AIBMe (mmol)</td><td>DMAEMA (mmol)</td></tr><tr><td>POH30</td><td>0.0</td><td>21.3</td><td>9.0</td><td></td><td>0.1</td><td>0.16</td><td></td></tr><tr><td>PO10H30</td><td>1.8</td><td>18.0</td><td>8.5</td><td></td><td>0.1</td><td>0.16</td><td rowspan=\"6\">3.5</td></tr><tr><td>PO50H30</td><td>4.2</td><td>4.2</td><td>3.6</td><td></td><td>0.1</td><td>0.16</td></tr><tr><td>PO100H30</td><td>8.4</td><td>0.0</td><td>3.6</td><td></td><td>0.02</td><td>0.16</td></tr><tr><td>PO100H30C20</td><td>8.4</td><td>0.0</td><td>5.1</td><td></td><td>0.02</td><td>0.16</td></tr><tr><td>PO100H30A20</td><td>8.4</td><td>0.0</td><td>8.6</td><td></td><td>0.02</td><td>0.16</td></tr><tr><td>PO10A30</td><td>1.8</td><td>18.0</td><td></td><td>7.5</td><td>0.1</td><td>0.22</td></tr><tr><td>PO100A30</td><td>8.4</td><td>0.0</td><td></td><td>3.5</td><td>0.02</td><td>0.22</td></tr></table></body></html> \n\nSynthesis and Characterization of Polymer Precursors. Chemical Characterization. $^{1}\\mathrm{H}$ NMR was performed using a Bruker ${600~\\mathrm{MHz}}$ spectrometer using deuterated chloroform as the solvent. The carboxylic acid content of relevant polymer precursors was determined by base-into-acid titration (ManTech Associates) using $100\\ \\mathrm{mM\\NaOH}$ as the base. The lower critical solution temperatures (LCST) of the polymer precursors were determined from a $10~\\mathrm{mg/}$ mL polymer solution in $10\\ \\mathrm{mM}$ PBS using a Variant Cary Bio 100 UV−vis spectrophotometer ramped over a temperature range of $20-$ $80~^{\\circ}\\mathrm{C}$ at a rate of $0.5~\\mathrm{{}^{\\circ}C/m i n}$ . The molecular weight of polymer precursors was determined using gel permeation chromatography (GPC). Samples soluble in water were analyzed using a system consisting of a Waters 515 HPLC pump, a Waters 717 plus Autosampler and three columns (Waters Ultrahydrogel-120, $-250,$ $-500$ ; $7.8\\times300\\ \\mathrm{mm}$ ; $6~\\mu\\mathrm{m}$ particles), with a continuous phase consisting of $0.5\\mathrm{~M~}$ sodium nitrate and $25\\ \\mathrm{mM}\\ 2$ -(cyclohexylamino) ethanesulfonic acid maintained at $\\mathrm{pH}10.0$ at $30~^{\\circ}\\mathrm{C}$ . Samples soluble in DMF were analyzed using a Polymer Laboratories PL-50 GPC system consisting of three Phenomenex PhenogelTM columns ( $300\\times$ $4.6\\ \\mathrm{mm}$ , ${\\mathfrak{s}}\\mu\\mathrm{{m}};$ pore size: 100, 500, 104 Å), with the continuous phase consisting of DMF containing $50\\ \\mathrm{\\mM}$ LiBr maintained at room temperature. All samples for GPC were filtered using a $0.2\\ \\mu\\mathrm{m}$ filter. Hydrazide-Functionalized POEGMA. The synthetic recipes and resulting chemical properties of POEGMA hydrazide precursor polymers produced are shown in Table 1. All hydrazide functionalized POEGMA polymers were synthesized following recipes previously reported.40 In brief, diethylene glycol methacrylate $(\\mathbf{M}(\\mathrm{EO})_{2}\\mathbf{MA},$ , $n=$ 2), oligo ethylene glycol methacrylate $\\mathrm{'OEGMA_{475}},$ $n=8,9$ ), acrylic acid (AA), 2,2-azobisisobutryic acid dimethyl ester (AIBMe), thioglycolic acid (TGA), and $20~\\mathrm{mL}$ of dioxane were added in a 50 mL Schlenk three-neck flask and purged with nitrogen for $30\\ \\mathrm{min}$ before being placed into a $75~^{\\circ}\\mathrm{C}$ oil bath. The polymerization was completed in $^{4\\mathrm{h}}$ under magnetic stirring to obtain poly(OEGMA-coAA) copolymer, followed by rotary evaporation to remove extra dioxane. Subsequently, poly(OEGMA-co-AA) was redissolved in 100 mL of DIW, and the acrylic acid residues were functionalized with hydrazide groups via carbodiimide chemistry using a 5-fold excess of adipic acid dihydrazide and an 2.5-fold excess of EDC $\\mathrm{(pH=4.75)}$ . The resulting POEGMA precursor polymers are labeled as $\\mathrm{PO}_{\\boldsymbol{x}}\\mathrm{H}_{\\boldsymbol{y}},$ where $x$ represents the mole percentage of $\\mathrm{OEGMA_{475}}$ relative to the sum of the $\\mathrm{OEGMA_{475}}$ (long chain, high transition temperature) and $\\mathbf{M}(\\mathrm{EO})_{2}\\mathbf{MA}$ (short chain, low transition temperature) OEGMA monomers added and $y$ is the overall mol $\\%$ of monomer residues bearing hydrazide groups. By changing the ratio of $\\mathrm{OEGMA_{475}/}$ $\\begin{array}{r}{\\mathbf{M}(\\mathrm{EO})_{2}\\mathbf{M}\\mathbf{A},}\\end{array}$ the LCST of POEGMA precursors can be adjusted from low $\\mathrm{(PO_{0}H_{30}}$ and $\\mathrm{PO}_{10}\\mathrm{H}_{30}\\big)$ to medium $\\mathrm{(PO}_{50}\\mathrm{H}_{30}\\mathrm{)}$ to high $\\mathrm{(PO}_{100}\\mathrm{H}_{30}\\right)$ .28,31 The hydrazide content was determined by the difference in the titrated $-\\mathrm{COOH}$ content of the polymers before and after ADH functionalization. Charged POEGMA precursors $\\mathrm{PO}_{100}\\mathrm{H}_{30}$ -cationic and $\\mathrm{PO}_{100}\\mathrm{H}_{30^{-}}$ anionic were synthesized using the same protocols but adding $20~\\mathrm{mol}$ $\\%$ of $N,N$ -dimethylaminoethyl methacrylate (cationic) or $2\\Bar{0}\\mathrm{~mol~}\\%$ extra acrylic acid (anionic), respectively (Table 1).12,34 All precursors were dialyzed for purification ( $^{6+}$ hours for 6 cycles using dialysis tubing with molecular weight cutoff (MWCO) of $3.5~\\mathrm{\\kDa})$ , lyophilized, and stored in a 15 wt $\\%$ solution of $10\\ \\mathrm{mM}\\ \\mathrm{PBS}$ at 4 $^{\\circ}\\mathrm{C}$ . Polymers were characterized as described above, with the additional step of base-into-acid conductometric titration (as previously described) to determine the net cationic and anionic charge content. \n\nHydrazide-Functionalized PNIPAM (PNIPAM-Hzd). PNIPAM-Hzd precursor was synthesized following previous work.21,41 In brief, 4.00 $\\mathsf{g}$ $\\textsl{g}(0.035\\ \\mathrm{mol})$ of NIPAM, $\\boldsymbol{1.00\\mathrm{~g}}$ $\\mathrm{\\bar{\\langle}0.014~m o l\\rangle}$ of acrylic acid, $87\\ \\mu\\mathrm{L}$ $\\left(1.25\\mathrm{\\mmol}\\right)$ of TGA and $55.5~\\mathrm{\\mg}$ $\\mathrm{(0.24~mmol)}$ of AIBMe were dissolved in $20~\\mathrm{mL}$ of ethanol and polymerized for overnight at $56^{\\circ}\\mathrm{C}$ under nitrogen. The solution was then lyophilized, redissolved in DIW, and functionalized with hydrazide groups using ADH/EDC chemistry as described for the POEGMA precursors. The resulting PNIPAM-Hzd polymer was then dialyzed $^{\\circ+}$ hours for 6 cycles, $\\mathrm{MWCO}~=~3.5~\\mathrm{\\kDa};$ ) and lyophilized for storage. The hydrazide content was determined by the difference in the titrated $-\\mathrm{COOH}$ contents of the polymers before and after ADH functionalization. \n\nHydrazide-Functionalized Dextran (Dextran-Hzd). Hydrazidefunctionalized dextran was prepared following previous work.42−44 A total of $\\textsf{S g}$ ( $\\left(0.13\\mathrm{\\mmol}\\right)^{\\cdot}$ ) of dextran from Leuconostoc spp (SigmaAldrich, $M_{\\mathrm{r}}\\sim40000)$ was dissolved in $42~\\mathrm{mL}$ of a $3~\\mathrm{M}$ solution of ${\\mathrm{NaOH}}_{;}$ , after which $7.29\\ \\mathrm{g}$ $\\mathrm{77~mmol)}$ of chloroacetic acid was added and stirred at room temperature until dissolution. Subsequently, the solution was heated at $70~^{\\circ}\\mathrm{C}$ for $90~\\mathrm{{min}}$ , cooled back to room temperature, and neutralized to $\\mathrm{pH}~7.0$ by adding acetic acid. The product was precipitated with methanol and collected through vacuum filtration to acquire a raw product that was stirred in acetone overnight to fully precipitate. Subsequently, the resulting product was washed with acetone and dried in an oven at $60~^{\\circ}\\mathrm{C}$ . The resulting carboxymethylated dextran was then functionalized with hydrazide groups by adding 2.5-fold excess EDC and 5-fold excess ADH in water $\\left(\\mathsf{p H4.75}\\right)$ . The final product was obtained by dialysis ( $^{6+}$ hours for 6 cycles, $\\mathrm{MWCO}\\ =\\ 12\\ \\mathrm{kDa}$ ) and lyophilized for storage. The hydrazide content was determined by the difference in the titrated −COOH content between the carboxymethylated dextrans before and after ADH functionalization. \n\nHydrazide-Functionalized Chitosan (Chitosan-Hzd). ChitosanHzd was prepared following previously reported protocols.45,46 A total of $_{\\textrm{1g}}$ of chitosan was slowly added to a solution of sodium hydroxide (NaOH, $43.75~\\mathrm{\\mmol}$ dissolved in $2{\\mathrm{~mL~}}$ of Milli-Q DIW) under magnetic stirring to swell (but not dissolve) the chitosan powder, followed by the addition of $8{\\mathrm{~mL~}}$ of 2-propanol. The mixture was stirred for $^{\\textrm{1h,}}$ after which $1.75\\mathrm{g}\\left(0.02\\mathrm{mol}\\right)$ of monochloroacetic acid predissolved in $2\\mathrm{mL}$ of isopropanol was added. After $^{4\\mathrm{h},}$ the reaction was stopped by adding $50~\\mathrm{\\mL}$ of $70\\%$ ethanol. The resulting carboxymethylated chitosan was then purified by dialysis ( $^{6+}$ hours for 6 cycles, $\\mathrm{MWCO}=12\\ \\mathrm{kDa}$ ) and functionalized with hydrazide groups using the same EDC/ADH chemistry outlined for dextran. The final product was dialyzed and lyophilized for storage. The hydrazide content was determined by the difference in the titrated −COOH content between the carboxymethylated chitosan before and after ADH functionalization. \n\nHydrazide-Functionalized Sodium Carboxymethyl Cellulose (CMC-Hzd). Synthesis of CMC-Hzd was reported in previous work.44 A total of $1.0\\mathrm{~g~}$ of CMC $\\left(0.01\\mathrm{\\mmol}\\right)^{\\cdot}$ ) and $_{3\\mathrm{~g~}}$ ( $\\cdot0.02\\ \\mathrm{mol},$ ) \n\nTable 2. Chemical Characterization of Polymer Precursors \n\n\n<html><body><table><tr><td>category</td><td>polymers</td><td> functional group</td><td>functionality mol %</td><td>M, 10° g mol-1</td><td> PDI</td><td>charge</td><td>LCST °C</td><td>concn mg/mL</td></tr><tr><td>group T</td><td>POH30</td><td>NHNH</td><td>30.5a</td><td>16.5b</td><td>2.95</td><td>neutral</td><td>36.6</td><td>150</td></tr><tr><td rowspan=\"12\">group C</td><td>PO10H30</td><td>NHNH</td><td>31.9a</td><td>19.8h</td><td>2.18</td><td>neutral</td><td>52.6</td><td>150</td></tr><tr><td>PO50H30</td><td>NHNH</td><td>29.6a</td><td>31.1b</td><td>1.76</td><td>neutral</td><td>>80</td><td>150</td></tr><tr><td>PO100H30</td><td>NHNH</td><td>29.5a</td><td>29.7℃</td><td>2.82</td><td>neutral</td><td>>80</td><td>150</td></tr><tr><td>PNIPAM-Hzd</td><td>NHNH</td><td>26.5a</td><td>25.3b</td><td>1.43</td><td>neutral</td><td>32.0</td><td>60</td></tr><tr><td>CMC40-Hzd</td><td>NHNH</td><td>23.3a</td><td>90d</td><td>N/A</td><td>neutral</td><td>>80</td><td>40</td></tr><tr><td>PO10A30</td><td>CHO</td><td>31.6d</td><td>17.4b</td><td>2.24</td><td>neutral</td><td>44.5</td><td>150</td></tr><tr><td>PO100H30</td><td>NHNH</td><td>29.5a</td><td>29.7℃</td><td>2.82</td><td>neutral</td><td>>80</td><td>150</td></tr><tr><td>PO100H30C20</td><td>NHNH</td><td>33.3a</td><td>31.2</td><td>3.75</td><td>cationic</td><td>>80</td><td>150</td></tr><tr><td>PO100H30A20</td><td>NHNH</td><td>27.8a</td><td>29.8</td><td>2.82</td><td>anionic</td><td>>80</td><td>150</td></tr><tr><td>CMC20-Hzd</td><td>NHNH</td><td>23.3a</td><td>90℃</td><td>N/A</td><td>neutral</td><td>>80</td><td>20</td></tr><tr><td>dextran-Hzd</td><td>NHNH</td><td>25.7a</td><td>40°</td><td>N/A</td><td>neutral</td><td>>80</td><td>15</td></tr><tr><td>chitosan-Hzd</td><td>NHNH</td><td>26.7a</td><td>lowe</td><td>N/A</td><td>cationic</td><td>>80</td><td>20</td></tr><tr><td>PO100A30</td><td>CHO</td><td>28.2d</td><td>24.1℃</td><td>3.25</td><td>neutral</td><td>>80</td><td>150</td></tr></table></body></html>\n\naDetermined by base-into-acid titration. bDetermined by DMF GPC. cDetermined by aqueous GPC. dDetermined by $\\mathrm{^{1}H}$ NMR. eInformation provided by supplier. $\\mathrm{LCST}=$ lower critical solution temperature \n\nof ADH were dissolved in $200~\\mathrm{mL}$ of DIW. Following this, premade solutions of $\\boldsymbol{0.07\\mathrm{g}}$ $\\mathrm{0.6~mmol})$ of NHS (suspended in $4~\\mathrm{mL}$ of a 1:1 $\\mathrm{DMSO/H_{2}O}$ solution) and $_{0.3\\mathrm{~g~}}$ ( $1.6\\ \\mathrm{mmol})$ of EDC (dissolved in 1 mL of a 1:1 $\\mathrm{DMSO/H_{2}O}$ solution) were added to the flask sequentially. The $\\mathrm{\\pH}$ of CMC solution was adjusted to $\\mathsf{p H}6.8$ using $\\mathrm{\\DeltaNaOH}$ and HCl solutions until no longer changed $({\\sim}4\\ \\mathrm{h})$ . The resulting hydrazide-functionalized polymer was dialyzed dialysis $^{'}6+$ hours for 6 cycles, $\\mathrm{MWCO}=12\\mathrm{\\kDa}$ ) and lyophilized for storage. The hydrazide content was determined by the difference in the titrated −COOH content before and after ADH functionalization. Two different concentrations of CMC-Hzd $\\mathrm{\\CMC_{40}–H z d}$ and $\\mathrm{CMC}_{20^{-}}$ Hzd) were used, with the subscript in each case corresponding to 40 and $20~\\mathrm{mg/mL}$ . \n\nAldehyde-Functionalized POEGMA Polymer Precursors. Aldehyde-functionalized POEGMA precursors were synthesized as previously described.4,25 Briefly, diethylene glycol methacrylate $\\mathbf{\\left(M(EO)}_{2}\\mathbf{MA}$ , $\\begin{array}{l c l}{n}&{=}&{2,}\\end{array}$ ), oligo ethylene glycol methacrylate $(\\mathrm{OEGMA}_{475},$ $\\begin{array}{r l r}{n}&{{}=}&{8,9}\\end{array}$ ), $N\\mathrm{.}$ -(2,2-dimethoxyethyl) methacrylamide (DMEMAm), AIBMe initiator, and TGA chain transfer agent were dissolved in $20~\\mathrm{mL}$ of dioxane and reacted for $^\\textrm{\\scriptsize4h}$ at $75~^{\\circ}\\mathrm{C}$ under nitrogen. The resulting poly(OEGMA-co-DMEMAm) polymer precursors $(\\sim4\\textrm{g})$ were subsequently dissolved in $75~\\mathrm{mL}$ of DIW, followed by adding $25~\\mathrm{\\mL}$ of $1.0\\mathrm{~M~HCl}$ and reacted at room temperature for $24\\mathrm{~h~}$ to convert the acetal groups in the DMEMAm residues to aldehyde groups to form $\\mathrm{PO}_{x}\\mathrm{A}_{y},$ where $x$ represents the $\\mathrm{OEGMA}_{475}/\\mathrm{M}(\\mathrm{EO})_{2}\\mathrm{\\bar{M}A}$ ratio (as with the hydrazide polymers) and y represents the overall mole fraction of monomers bearing an aldehyde group (Table 1). The resulting precursors were dialyzed $^{\\prime}6+$ hours for 6 cycles, using dialysis tubing with MWCO of $3.5\\mathrm{\\kDa}$ ), lyophilized, and stored in $15\\%$ solutions in $10~\\mathrm{mM}$ PBS at $4~^{\\circ}\\mathrm{C}$ . The aldehyde content of the polymers was determined by $^{1}\\mathrm{H}$ NMR comparing the aldehyde proton $(\\delta\\sim9)$ with the methoxy proton ( $\\overset{\\cdot}{\\delta}$ $\\sim3.5)$ . \n\nHydrogel Preparation. Hydrogels were prepared using an automated material screening high-throughput robotics system (Tecan Evo 200). Polymer precursor solutions were first loaded in a 24 well-plate ( $2\\ \\mathrm{\\mL}$ /well for each precursor polymer). Subsequently, 8 available robotic arms were used to aspirate preprogrammed volumes of each hydrazide precursor polymer into different wells of a 96 well-plate (Greiner, VWR) using a factorial design strategy, in which each possible equal volume combination of the six hydrazide precursor polymers within each series (Table 2) was pipetted sequentially into each well (Figure 1; see Tables S1 and S2 for detailed recipes of polymers dispensed into each well). The pipetting parameters were optimized for different ranges of liquid types; for example, the aspiration speed used for low-viscosity POEGMA polymer precursors was $800\\ \\mu\\mathrm{L}/s,$ while an aspiration speed of $100\\ \\mu\\mathrm{L}/\\mathrm{s}$ was used for the relatively viscous chitosan precursor. After dispensing all the hydrazide precursor polymers, the plate was automatically moved to a shaker platform and mixed at 1700 rpm for $10\\mathrm{~s~}$ to promote mixing of all added hydrazide components. The plate was then moved back to the pipetting platform, and the relevant aldehyde precursor polymer was added column by column to each well, with the plate moved back to the shaker for $\\textit{\\textbf{5s}}$ after each column of aldehyde polymer was added. To avoid bubbles and promote mixing when adding the aldehyde polymer precursors, (1) $70~\\mu\\mathrm{L}$ of POA was aspirated, but only $60~\\mu\\mathrm{L}$ was dispensed in each well (ensuring no air is injected), and (2) the pipet tips were fully immersed in the hydrazide polymer solutions. \n\n \nFigure 1. Schematic of compositions of hydrazide polymer precursors in a 96-well plate. \n\nHigh-Throughput Hydrogel Characterization. Hydrogel Swelling. A total of $120~\\mu\\mathrm{L}$ of $10\\ \\mathrm{mM}$ PBS was added to each well containing a hydrogel and incubated for $^{48\\mathrm{~h~}}$ at $22\\ ^{\\circ}\\mathrm{C},$ a time confirmed to achieve equilibrium swelling in previous experiments12,24 and a temperature below the LCST values of each polymer precursor. The “find contact” function of a Mach-1 micromechanical tester fitted with a $1\\ \\mathrm{mm}$ diameter rounded tip indenter (Biomomentum, Laval, Canada) was used to track the change in the height of the hydrogel in each well before and after swelling according to the height of hydrogel in each well, as per eq 1. \n\n \nFigure 2. Model form for partial least-squares analysis of high-throughput hydrogel data. \n\n$$\n{\\mathrm{normalizedvolume}}={\\frac{V_{\\mathrm{final}}}{V_{\\mathrm{initial}}}}={\\frac{A\\times H_{\\mathrm{gel}}}{120~{\\upmu\\mathrm{L}}}}\n$$ \n\nHere, $V_{\\mathrm{final}}$ and $V_{\\mathrm{initial}}$ are the volumes of gel before $\\left(t=0\\mathrm{h}\\right)$ and after swelling $t=48\\ \\mathrm{h},$ , respectively, $A$ is the cross-sectional area of each well in 96 well-plate $(0.{\\bar{3}}4\\thinspace\\mathrm{cm}^{2})$ , and $H_{\\mathrm{gel}}$ is the height of gel measured by Mach-1 tester (i.e., the height of the gel at the test time point subtracted by the premeasured height of the empty well). Error bars represent the standard deviation of four independent measurements $\\left(n=4\\right)$ . \n\nHydrogel Mechanics. The compressive modulus of the hydrogels was measured inside the 96-well plates using a $1~\\mathrm{mm}$ diameter rounded tip indenter and the Mach-1 micromechanical tester with a multiwell plate attachment. The modulus was measured by finding the contact in each well and performing a $10\\%$ compression, with the modulus corresponding to the slope of the resulting stress versus strain curve. Error bars represent the standard deviation of four independent measurements $\\left(n=4\\right)$ . \n\nHydrogel Degradation. Hydrogel degradation was determined by tracking the change in the compressive modulus of the hydrogels in acid-accelerated conditions to allow for comparisons between the degradation rates of different hydrogels under practical-to-measure timeframes. Following an equilibrium swelling step ( $\\ln\\mathrm{\\Delta}$ of PBS/ well, $22\\ ^{\\circ}\\mathrm{C},$ after $^{72\\mathrm{~h~}}$ ), the PBS was removed and a compressive modulus measurement was done in each well using the Mach-1 micromechanical tester as described above. Subsequently, $100\\ \\mathrm{mM}$ HCl ( $120~\\mu\\mathrm{L}$ per well, $22\\ ^{\\circ}\\mathrm{C})$ was added to each well, and the compressive modulus measurement was repeated at predetermined time intervals until complete gel degradation (considered to be the point at which the modulus of the residual hydrogel was below the detection threshold of the find contact measurement). Error bars represent the standard deviation of four independent measurements $\\overset{\\mathcal{-}}{\\left(n\\right.}=4\\overset{\\mathcal{-}}{\\left)}$ . \n\nTransparency. The transmittance of each hydrogel was determined using a VICTOR 3 multilabel microplate reader operating at a wavelength of $595~\\mathrm{{\\nm}}$ . The transmittance was scanned over a temperature range of 25 to $40\\ ^{\\circ}\\mathrm{C},$ using an equilibration time of 10 min at each fixed temperature measurement. Error bars represent the standard deviation of four independent measurements $\\left(n=4\\right)$ . \n\nDrug Release Kinetics. Hydrogels were prepared as described above using the high-throughput robotics system but also dissolving $25~\\mathrm{mg/mL}$ ovalbumin in the POA component (i.e., the component added consistently to each well), resulting in a total of $1.5~\\mathrm{mg}$ protein encapsulated in each gel. For drug release experiment, hydrogels were first fabricated inside a 96-well MultiScreen-Mesh filter plate and then submerged in a 96-well receiver with $10\\ \\mathrm{mM}$ PBS (EMD Millipore, see Supporting Information, Figure S1). A total of $100\\mu\\mathrm{L}$ of PBS was added on the top of each hydrogel, while $250\\mu\\mathrm{L}$ of PBS was added in the bottom (i.e., in the receiving chamber) of each insert, after which the samples were incubated at $37~^{\\circ}\\mathrm{C}$ over a one-month period. At predetermined intervals, $20~\\mu\\mathrm{L}$ of medium was taken from the reserved plate for each sample and assayed for protein concentration using a Bradford assay, with the concentration calculated based on a calibration of standard ovalbumin concentrations $\\displaystyle{{'R}^{2}=0.99},$ ). Error bars represent the standard deviation of four independent measurements $\\left(n=4\\right)$ . All volumes of PBS was replaced with fresh PBS after measurement every time. \n\nDynamic and Latent Variable Modeling and Analysis. Dynamic Modeling of Drug Release Kinetics. Data from the drug release kinetics measurements were first modeled using dynamic regression modeling to convert the kinetic curves into fitting parameters that could be incorporated into the multivariate statistical model. The release kinetics profiles were fit to a modified first-order (i.e., diffusion-governed) model (eq 2): \n\n$$\n\\hat{y}(t)=y_{f}+(y_{0}-y_{f})e^{-t/\\tau}\n$$ \n\nwhere $\\hat{y}(t)$ represents the predicted protein concentration at time $t,$ and $y_{0},y_{\\beta}$ and $\\tau$ are modeling parameters representing the initial drug concentration, final drug concentration (at infinite time), and first order rate constant (reflecting the release rate), respectively. Note that the $y_{0}$ term was included to compensate for burst release early in the release process, while the $y_{f}$ term was incorporated to compensate for potential protein entrapment inside the gels, particularly relevant for temperature-responsive hydrogels that significantly dehydrate over time. Parameters that minimized the sum of squared error between the experimental and predicted data were fit by explicitly finding the zero of the gradient of the quadratic cost function for each given kinetic profile using Matlab’s fsolve function. Model fits were evaluated both qualitatively (by visual inspection) and quantitatively (using the squared prediction error statistic). \n\nIngredient Modeling. To assess the impact of different hydrogel chemistries (i.e., mixtures of predefined precursor polymers) on hydrogel performance, a multivariate modeling technique developed by Muteki and MacGregor47 was used. The basis of this method is to combine raw ingredient properties (here, the different precursor polymer compositions) with ratios using linear mixing rules to calculate “pseudo” mixture properties. These mixture properties can then be used in the input space of a regressive (typically latent variable) model to predict product performance. Here, for each hydrogel fabricated, the concentration (wt $\\%$ ), molecular weight, and degree of functionalization $\\mathrm{(mmol/g)}$ of each precursor polymer used to form the mixed hydrogels were used to calculate the pseudo mixture values of each of those properties based on linear mixture rules (i.e., by weighting the properties of each ingredient by the weight percentage of that ingredient used in the formulation). The resulting pseudo mixture properties were subsequently used in a partial least-squares regression to relate the mixture formulations to the performance parameters of the resulting hydrogels, as per the model form shown in Figure 2. The model was fit using the NIPALS algorithm and Aspen Technology’s ProMV software package. To ensure a good model fit, only experiments with a “good” dynamic model fits (as assessed by the sum of squared error) were used in the training data of the model. This approach would not be statistically appropriate if the objective of this analysis was to establish confidence in model coefficients for gaining mechanistic insight; however, given that the objective of the modeling in this case was instead to advise a direction for further experimentation toward achieving better hydrogel properties, limiting the input data to only the highest quality samples available is justifiable. This choice is particularly justifiable in this work in light of the relatively high noise associated with the moderate-to-high uncertainties observed with many (but not all) of the high-throughput characterization techniques developed, uncertainties that complicate explicit interpretation of the trends from each data set analyzed unless a data filtering approach was used. \n\nOptimization. To make informed decisions about what formulations would achieve target hydrogel performance characteristics based on predefined criteria, the identified partial least-squares (PLS) model was inverted to allow for the prediction of polymer mixtures (bounded by what range of concentrations is physically possible given the rheological properties of each mixture component) that would achieve hydrogel compositions with targeted properties. Optimization was performed by solving the following quadratic programming problem using Matlab (eqs 3−8): \n\n$$\n\\underset{r,z}{\\mathrm{min}}(\\hat{y}-y_{t})^{T}P_{y}(\\hat{y}-y_{t})+\\left(\\frac{t}{s_{a}}\\right)^{T}P_{H T}{}^{2}\\Bigg(\\frac{t}{s_{a}}\\Bigg)+(x-P t)^{T}P_{S P E_{x}}{}(x-P t)\n$$ \n\n$$\n{\\begin{array}{r l}&{s\\cdot t\\cdot\\sum r_{i}=1.0}\\\\ &{r_{i}\\in[0.1]\\ \\forall\\ i}\\\\ &{x=[r^{T}X_{p}r^{T}]^{T}}\\\\ &{{\\widehat{\\boldsymbol{y}}}=Q t}\\\\ &{t=x^{T}W^{*}}\\end{array}}\n$$ \n\ncontrolling protein release kinetics from hydrogels. Table 2 shows the chemical properties of each of the precursor polymers. \n\nwhere $\\hat{\\pmb{y}}$ represents the predicted product qualities (here, the drug release kinetic parameters and transparency); $y_{t}$ represents the target properties; $P_{y},P_{H T^{2}},$ and $P_{S P E_{x}}$ are symmetric penalty matrices that can be adjusted to weight the relative importance of meeting the target hydrogel performance, minimizing extrapolation within the latent variable space, and minimizing extrapolation away from the latent space respectively; $\\boldsymbol{\\mathrm{\\ell}}_{Q}$ is the PLS coefficient matrix relating outcomes to the vector of latent space scores $\\mathbf{\\Delta}_{t,\\tiny{\\begin{array}{r l}\\end{array}}}$ and $W^{*}$ is the matrix of PLS coefficients relating inputs $\\pmb{x}$ to scores; the interested reader is referred to reference 64 for more complete details on the meaning of these scores.47 The input, $\\mathbf{\\delta}_{\\pmb{x},}$ is comprised of the ratios of the pseudo mixture properties $\\left(r^{T}X_{p}\\right)$ and the precursor polymer ratios used to form hydrogels in each well $(r)$ . The first two constraints (eqs 4 and 5) enforce that the sum of the ratios should be 1 and all ingredient ratios must be between 0 and 1. The second term in eq 2 prevents extrapolation of the model within the latent space beyond the statistical limits of the high throughput input data, while the third term of the objective function enforces a soft constraint on the squared prediction error (SPE) of the input space; the latter is essential in this case to reflect the covariance structure of the inputs since not all of the polymer recipe variables modeled are completely independent (e.g., wt $\\%$ polymer is confounded by the significant higher viscosities and thus lower concentrations of CMC-Hzd and Chitosan-Hzd that can be used to prepare hydrogels).",
|
||
"category": " Materials and methods"
|
||
},
|
||
{
|
||
"id": 4,
|
||
"chunk": "# RESULTS AND DISCUSSION \n\nSynthesis of Hydrogel Precursors. Six hydrazidefunctionalized hydrogel precursor polymers were synthesized for each series of materials studied (i.e., thermoresponsive vs nonthermoresponsive and charged vs neutral). This number was chosen given that all possible combinations of each of the six precursor polymers could be prepared on a single 96-well plate (63 total hydrogels/series), enabling comprehensive high-throughput characterization of hydrogel swelling, degradation, transparency, and drug release kinetics all in a single synthetic and analysis step. The two series of polymers were selected based on the integral role of hydrogel porosity (regulated by thermoresponsiveness) and protein−hydrogel interactions (regulated by both hydrophobicity and charge) on \n\nAll synthetic polymers (POEGMA, PNIPAM) were synthesized via free radical copolymerization using a chain transfer agent to limit the molecular weight of polymers. The measured number-average molecular weight $\\bar{(\\boldsymbol{M_{\\mathrm{n}}})}$ of both hydrazide and aldehyde-functionalized POEGMA or PNIPAM polymers was measured to lie between 16 and $22\\times10^{3}\\mathrm{g/mol},$ below the $\\ensuremath{M_{\\mathrm{n}}}\\sim40\\times10^{3}\\mathrm{g/mol}$ associated with in vivo renal clearance (Table 2).30,48 The degree of hydrazide or aldehyde functionalization between the different POEGMA precursor polymers was also designed to be similar (Table 2), such that the degree of cross-linking is likely to be comparable between different mixed combinations. The lower critical solution temperatures (LCSTs) of the $\\mathrm{PO}_{x}$ POEGMA-based hydrazidefunctionalized precursor polymers varied between ${\\sim}37\\ ^{\\circ}\\mathrm{C}$ for $\\mathrm{PO}_{0}\\mathrm{H}_{30}$ to ${>}80{}^{\\circ}\\mathrm{C}$ for $\\mathrm{PO}_{50}\\mathrm{H}_{30}$ and $\\mathrm{PO}_{100}\\mathrm{H}_{30},$ consistent with our previous reports.31,48 Note that the resulting volume phase transition temperatures (VPTTs) of the single-component hydrogels were significantly lower than the precursor polymer LCST values due to the consumption of the more polar hydrazide groups upon hydrazone bond formation, resulting in hydrogel VPTT values ranging from ${\\sim}26\\ ^{\\circ}\\mathrm{C}$ for ${\\mathrm{PO}}_{0}$ (collapsed at $37\\ ^{\\circ}\\mathrm{C}\\$ ), ${\\sim}32{\\mathrm{-}}33^{\\circ}\\mathrm{C}$ for $\\mathrm{PO}_{10}$ (slightly collapsed at $37{}^{\\circ}\\mathrm{C}\\mathrm{)}$ ), ${>}80{}^{\\circ}\\mathrm{C}$ for $\\mathrm{PO}_{50}$ (swollen at $37\\ {}^{\\circ}{\\bf C}{\\bf\\dot{\\Psi}},$ ), and ${>}80{}^{\\circ}\\bar{\\mathrm{C}}$ for $\\mathrm{PO}_{100}$ (swollen at $37\\ ^{\\circ}\\mathrm{C}$ ).31,48 As such, the thermoresponsive POEGMA precursor polymers selected for screening span the full range of the phase transition temperatures relevant to physiological protein release. CMC-Hzd was included in the thermoresponsive series as a nonthermoresponsive carbohydrate component for comparison with the high transition temperature $\\mathrm{PO}_{50}$ and $\\mathrm{PO}_{100}$ components to assess how using a temperature-independent, but more highly viscous, precursor component would affect the protein release properties. Note that $595\\%$ of the −COOH groups on all of the PNIPAM, POEGMA, and CMC precursor polymers were converted to hydrazide groups as per conductometric titration, such that all precursor polymers in this first series have similar (essentially neutral) net charges. \n\nCharge was introduced into the precursor polymers by (1) copolymerizing the cationic comonomer DMAEMA (cationic, $\\mathrm{P}\\bar{\\mathrm{O}}_{100}\\mathrm{\\dot{H}}_{30}\\mathrm{C}_{20}\\big)$ or the anionic comonomer AA (anionic, $\\mathrm{PO}_{100}\\mathrm{H}_{30}\\mathrm{A}_{20}\\right)$ to form $20\\mathrm{~\\mol~\\}\\%$ functional monomer POEGMA-based synthetic copolymers or (2) selecting natively cationic (chitosan) or anionic (carboxymethyl cellulose) naturally sourced polymers. As such, different cationic, anionic, and amphoteric hydrogels with different charge contents can be prepared by mixing different combinations of cationic, anionic, and neutral $\\mathrm{PO}_{100}$ -based hydrazide precursors together with the nonthermoresponsive neutral aldehyde polymer $\\mathrm{PO}_{100}\\mathrm{A}_{30}$ . \n\nThe concentrations of each polymer used were selected based on (1) the viscosity of the precursor polymers (ensuring that robotic pipetting is feasible), (2) the concentration at which the single component hydrazide and aldehyde gelling pair gels within $2{-}30\\ \\operatorname*{min}$ (ensuring that mixing is possible prior to gelation), and (3) published recipes of single component hydrazide and aldehyde gelling pairs that have been demonstrated to yield mechanically robust hydrogels with relevant biomedical properties.21,34,36,40,43,44,48 As such, while the mass concentrations and functional group densities of polymer in each combination gel do not match, the resulting hydrogels all have similar gelation times and compressive moduli on the same order of magnitude (tens of $\\mathbf{kPa}_{,}$ , see Supporting Information, Figure S2). In addition, the latent variable statistical model applied to the resulting data takes these different concentrations into account in the optimization protocol while also allowing them to vary in the optimization step in order to achieve target hydrogel properties. \n\n \nFigure 3. Schematic of the high-throughput robotic fabrication approach and the structures of the hydrazide and aldehyde-functionalized polymer precursors used for hydrogel preparation. \n\nCytotoxicity measurements using the Presto Blue assay in conjunction with NIH 3T3 mouse fibroblast cells indicated that high cell viability was maintained after $24\\mathrm{~h~}$ of incubation with each hydrogel precursor up to concentrations of at least 2 $\\mathrm{mg/mL},$ , a relatively high concentration for in vitro cytotoxicity screening (Supporting Information, Figure S3). As such, coupled with the degradability of the hydrogel networks (enabled by the presence of the naturally sourced polymers and the hydrolytically labile hydrazone cross-links), both sets of combinatorial hydrogels tested have potential for in vivo use. \n\nPreparation of Multicomponent Hydrogels Using High-Throughput Robotics. Multicomponent hydrogels were prepared using a Tecan Evo 200 robot to mix preformed solutions of hydrogel precursor solutions in $10\\mathrm{\\mM}$ PBS at preprogrammed ratios, as shown visually in Figure 1. Two categories of materials were separately assayed: thermoresponsive versus nonthermoresponsive hydrogels (group T) and charged versus neutral hydrogels (group C; Figure 3). Each possible combination of the six hydrazide precursor polymers in each set was pipetted by the robot into separate wells of a 96-well plate according to combinatorial theory (eq 9). \n\n$$\nC_{6}^{1}+C_{6}^{2}+C_{6}^{3}+C_{6}^{4}+C_{6}^{5}+C_{6}^{6}=63\n$$ \n\nFor example, $C_{6}^{2}$ refers to each possible combination of two premixed hydrazide polymers were cross-linked with one aldehyde polymer (15 combinations total). A total of $60~\\mu\\mathrm{L}$ of hydrazide precursor polymer was dispensed into each well for each combination tested, corresponding to $30~\\mu\\mathrm{L}$ of each of two precursor polymers dispensed, $20~\\mu\\mathrm{L}$ of each of three precursor polymers dispensed, and so on; this design approach facilitates the creation of each of the mixed hydrogel combinations listed above without changing the total gel volume. The result of this protocol was that 126 different hydrogels could be fabricated in quadruplicate in ${\\sim}25~\\mathrm{min}$ . \n\nPhysiochemical Properties of Combinatorial Hydrogels. To address the key challenge of high-throughput materials screening (i.e., the characterization of the properties of the fabricated materials), we developed a series of analytical techniques for assessing the typically reported properties of hydrogels (mechanics, swelling, degradation, and transparency) relevant for biomedical applications that could be performed reliably at reasonably high speeds ( $^{\\cdot}<2\\mathrm{~h~}$ per plate of 63 hydrogels) without requiring removal of the hydrogels from the wells. Mechanics, swelling, and degradation measurements were all performed using a Mach-1 micromechanical tester fitted with a multiwell plate adapter that allows the instrument to individually address each well of a 96-well plate. \n\nSwelling. Hydrogel swelling was measured by comparing the point at which the microindenter contacted the hydrogel interface (i.e., normal force $>0.005\\mathrm{~N~}$ ) before and after a $^{48\\mathrm{~h~}}$ swelling period in $10\\mathrm{mM}\\mathrm{PBS}$ at $22{}^{\\circ}\\mathrm{C};$ the room temperature test condition was chosen to minimize the effect of the phase transition of thermoresponsive hydrogels on the measured swelling results. Figure 4 shows the swelling responses of single-component hydrogels in each series, while Supporting Information, Figures S4 (thermoresponsive series) and S5 (charged series) show the results for each combinatorial hydrogel fabricated. As anticipated, the thermoresponsive $\\mathrm{PO}_{0}\\mathrm{H}_{30}$ and PNIPAM-Hzd single component hydrogels deswelled over the test period, yielding equilibrium swelling ratios of $0.91\\pm0.09$ $\\left(\\mathrm{PO}_{0}\\mathrm{H}_{30}\\right)$ and $0.82\\pm0.11$ (PNIPAMHzd), respectively. While the absolute swelling ratios are lower for the constrained (in-plate) and unconstrained (free disk) swelling measurements, the swelling data in Figure 4 correlates well $\\ ^{\\prime}R^{2}\\ =\\ 0.84)$ with the unconstrained swelling measurements performed using the conventional disk method (Supporting Information, Figure S6), confirming the predictive ability of the multiwell plate assay for measuring hydrogel swelling. Similarly, mixing $\\mathrm{PO}_{0}\\mathrm{H}_{30}$ or PNIPAM-Hzd with any of $\\mathrm{PO}_{10}\\mathrm{H}_{30},$ $\\mathrm{PO}_{50}\\mathrm{H}_{30},$ $\\mathrm{PO}_{100}\\mathrm{H}_{30},$ or $\\mathrm{CMC}_{40}–\\mathrm{Hzd}$ resulted in hydrogels with significantly suppressed swelling over the $^{48\\mathrm{~h~}}$ incubation period for each binary or ternary combinations tested (Supporting Information, Figure S4A,B). The binary combination of $\\mathrm{PO}_{0}\\mathrm{H}_{30}$ and $\\mathrm{CMC}_{40}–\\mathrm{Hzd}$ (T2−12) exhibited a particularly noteworthy deswelling ratio of $0.70\\pm0.06,$ despite the fact that the $\\mathrm{CMC}_{40}$ -Hzd single component gels swell over the same time period (Figure 4A); similarly, the ternary combination of $\\mathrm{PO_{10}H_{30}}+\\mathrm{PO_{100}H_{30}}+$ PNIPAM-Hzd (T3−3) \n\n \nFigure 4. Volume-based swelling ratios of (A) thermoresponsive vs nonthermoresponsive (T series) single-component hydrogels and (B) charged vs neutral (C series) single-component hydrogels before (blue) and after (red) swelling for $^{48\\mathrm{~h~}}$ in $10~\\mathrm{mM}$ PBS at room temperature $22\\ {}^{\\circ}\\mathrm{C};$ ; $n=$ 4). See Supporting Information, Figures S4 and S5, for the corresponding results for the combinatorial hydrogels. \n\n \nFigure 5. Compressive modulus ( $\\dot{\\boldsymbol{t}}=0$ h data points) and degradation kinetics (tracked by changes in the measured compressive modulus over time of exposure to $0.1\\mathrm{~M~}$ HCl at $22\\ ^{\\circ}\\mathrm{C}$ ) for (A) thermoresponsive vs nonthermoresponsive $\\mathrm{\\bar{T}}$ series) single-component hydrogels and (B) charged vs neutral (C series) single-component hydrogels $\\left(n=4\\right)$ ; (C) Comparison of the compressive modulus of hydrogels in the $\\mathrm{\\DeltaT}$ series before and after acid degradation over $72\\ \\mathrm{h};$ (D) Comparison of the compressive modulus of hydrogels in the C series before and after acid degradation over $24\\mathrm{~h~}$ . See Supporting Information, Figures S7 and S8, for the corresponding results for the combinatorial hydrogels. \n\nexhibited a swelling ratio of $0.70\\pm0.07_{.}$ , despite the presence of just one precursor polymer (PNIPAM-Hzd) that deswelled as a single component gel (Figure 4A). In contrast, each fivecomponent hydrogel in which $\\mathrm{PO}_{0}\\mathrm{H}_{30}$ and PNIPAM-Hzd represented two of the five components (i.e., comprise $40\\%$ of the total volume and ${>}50\\%$ of the total polymer mass, Supporting Information, Figure S4D) exhibited substantial swelling, with the minimum swelling ratio among the four such hydrogels tested being $1.1\\pm\\:0.3$ and the others reaching as high as $1.6\\pm0.4$ . As such, combinations of precursor polymers result in nonadditive swelling properties, suggesting the potential utility of this mixing approach to generate new and optimized protein release kinetics. We hypothesize that these observed nonadditive effects on hydrogel properties to probable phase separation within these combination gels, creating segregated mass distributions between pro-swelling and antiswelling precursor polymer domains. \n\nSwelling ratios for the charged hydrogel series also showed nonlinear effects, although the general trends were much more consistent. Comparing Figure 4A and 4B, the swelling ratios in Figure 4B were typically higher, consistent with the higher hydrophilicity of the cross-linking polymer $\\mathrm{PO}_{100}\\mathrm{A}_{30}$ compared to $\\mathrm{PO}_{10}\\mathrm{A}_{30}\\big)$ , as well as the charged nature of many of the hydrazide precursor polymers that can drive Donnan equilibrium-related swelling at the physiological test $\\mathrm{\\tt{pH}}$ . Combinations of two, five, or six precursor polymers (Supporting Information, Figure $S5\\mathrm{A,D}$ ) all resulted in hydrogels with generally similar swelling ratios to the single component hydrogels; in contrast, combinations of three or four precursor polymers (Supporting Information, Figure S5B,C) resulted in hydrogels with significantly higher swelling ratios. The amphoteric hydrogel combinations of $\\mathrm{PO}_{100}\\mathrm{H}_{30}\\mathrm{A}_{20}$ $^+$ dextran-Hzd $^+$ chitosan-Hzd (C3−19) and $\\mathrm{PO}_{100}\\mathrm{H}_{30}\\mathrm{A}_{20}+$ CMC-Hzd $^+$ dextran-Hzd $^+$ chitosan-Hzd (C4−15) exhibited particularly notable swelling ratios of $1.9\\pm0.4$ and $2.2\\pm0.1$ , respectively, suggesting a benefit to creating mixed charge hydrogels to promote high swelling. In this context, combinatorial mixing can again enable access to a broader potential range of hydrogel swelling responses than possible with the single-component hydrogels. \n\nMechanics. A simple $10\\%$ compression protocol from the point of contact was used in combination with the high throughput accessory of the Mach-1 micromechanical tester to measure the compressive modulus at $22~^{\\circ}\\mathrm{C}$ both before and after the incubation of the hydrogels in $10~\\mathrm{mM}$ PBS at room temperature $\\left(22^{\\circ}\\mathrm{C}\\right)$ for $^{72\\mathrm{h},}$ with the time chosen to ensure all gels reach equilibrium swelling before measurement. The compressive modulus of single-component POEGMA hydrogels after swelling decreased from $\\mathrm{PO}_{0}\\mathrm{H}_{30}$ $(59\\pm34\\mathrm{\\kPa})$ to $\\mathrm{PO}_{100}\\mathrm{H}_{30}$ ( $\\langle33\\pm5\\mathrm{\\kPa}\\rangle$ (Figure 5A, time $t=0\\mathrm{~h~}$ ), a trend consistent with previous observations on bulk gels as the proportion of the long-chain $\\mathrm{OEGMA}_{475}$ monomer is reduced.48 Furthermore, charged hydrogels prepared with POEGMA-based precursor polymers exhibited substantially higher moduli than those prepared based on the higher molecular weight naturally sourced precursors (Figure 5B, time $t=0\\mathrm{h}$ ), again consistent with previous reports.34 As such, the high-throughput mechanics assay yields modulus trends mirroring those achieved with the conventional technique. \n\nMixing different precursor polymers again demonstrates nonlinear effects (Supporting Information, Figures S7 and S8), although the relatively large error bars observed particularly with some of the higher-modulus mixtures limit the scope of conclusions that can be drawn. For example, among the binary thermoresponsive combinations (Figure S7A), the combination of $\\mathrm{PO}_{0}\\mathrm{H}_{30}+\\mathrm{PO}_{50}\\mathrm{H}_{30}$ (T2−10, in which one component is thermoresponsive and the other is not at physiological temperature) exhibited a compressive modulus of $89\\pm21\\mathrm{{kPa}}_{;}$ equivalent to a $\\mathrm{PO}_{0}\\mathrm{H}_{30}$ single-component gel and somewhat higher than a $\\mathrm{PO}_{50}\\mathrm{H}_{30}$ single-component gel $\\left(p\\ <\\ 0.1\\right)$ ; in contrast, the combination of $\\mathrm{PO}_{0}\\mathrm{H}_{30}+\\mathrm{PO}_{10}\\mathrm{H}_{30}$ $(\\mathrm{T}2-2,$ in which both components are thermoresponsive but to different degrees at physiological temperature) resulted in a hydrogel with a much lower compressive modulus of $15~\\pm~7~\\mathrm{{kPa}},$ , substantially weaker than either single-component gel $(p\\ <$ 0.01 for both comparisons). We hypothesize this difference is again related to differences in phase separation, with the double thermoresponsive binary gel $(\\mathrm{T}2-2)$ more likely to generate bulk phase-segregated domains and the single thermoresponsive binary gel (T2−10) more likely to form a continuous nonresponsive phase with collapsed thermoresponsive domains that may mechanically reinforce the hydrogel. Such differences are suppressed as more components are added to the gels and, thus, the probability of some form of macroscopic phase separation is increased, with all five- and six-component hydrogels exhibiting similar moduli within experimental error. Similar general trends are observed with the charged precursor data (Supporting Information, Figure S8). Binary and ternary combinations (Figures S7A and S8B) exhibited substantially higher variability in modulus values than the five and six-component mixtures (Figure S8D), with amphoteric hydrogels such as $\\mathrm{PO_{100}H_{30}~+~\\mathrm{PO_{100}H_{30}C_{20}~+~}}$ $\\mathrm{PO}_{100}\\mathrm{H}_{30}\\mathrm{A}_{20}$ (C3−1), exhibiting particularly high moduli, consistent with the demonstrated high capacity of amphoteric hydrogels for retaining water in the presence of salt (i.e., PBS) and facilitating dual ionic/covalent cross-linking at physiological $\\mathrm{\\ttpH}$ .34 \n\nDegradation. Degradation kinetics were assessed by tracking the decrease in the compressive modulus of the hydrogels over time upon exposure to acidic degradation conditions $\\langle0.1\\mathrm{M}\\mathrm{HCl},\\bar{2}2^{\\circ}\\mathrm{C}\\rangle$ ; note that this acidic condition was chosen to accelerate the degradation of the hydrazone bond to enable comparisons between the degradation potential of various hydrogel compositions on a shorter time scale. Upon exposure to 0.1 M HCl, the compressive modulus of most hydrogels decreased to ${\\sim}50\\%$ or less of the initial modulus very quickly (comparing $t=0$ and $^{2\\mathrm{~h~}}$ in Figure 5A and B), with the modulus of most of the single-component charged series hydrogels decreasing to nearly zero after $24\\ensuremath{\\mathrm{~h~}}$ of incubation consistent with visually observed gel degradation at this time point (Figure 5D). In contrast, thermoresponsive hydrogels based on POEGMA or, in particular, PNIPAM (which deswells the most relative to its preparation state at room temperature, Figure 4A) retained at least ${\\sim}20{\\mathrm{-}}50\\%$ of their initial modulus after $24\\mathrm{h}$ and persisted for at least $^{72\\mathrm{{h}}}$ in the presence of 0.1 M HCl (Figure 5C). This notably slower degradation of the thermoresponsive hydrogels and, in particular, the PNIPAM-Hzd hydrogel, is consistent with previous observations,21,30 as well as theory, in that introducing charge promotes swelling (Figure 4B) and thus faster hydrolytic degradation (Figure 5B), while introducing thermoresponsive components suppresses swelling (Figure 4A) and thus also degradation (Figure 5A). On this basis, the suitability of the high-throughput measurement protocol used for probing degradation rates is confirmed. \n\n \nFigure 6. Transmittance of (A) thermoresponsive vs nonthermoresponsive (T series) single-component hydrogels and (B) charged vs neutral (C series) single-component hydrogels as a function of temperature ${\\mathit{\\check{n}}}=4{\\mathit{\\check{\\Psi}}},$ ). See Supporting Information, Figures S9−S16, for the corresponding results for the combinatorial hydrogels. \n\nNonadditive effects are again observed in the degradation performance of the combinatorial hydrogels, particularly in the thermoresponsive binary and ternary mixture hydrogels (Supporting Information, Figures S7A,B). Of particular note, $\\mathrm{PO}_{0}\\mathrm{H}_{30}+$ PNIPAM-Hzd (T2−11) and $\\mathrm{PO}_{10}\\mathrm{H}_{30}+\\mathrm{PO}_{0}\\mathrm{H}_{30}+$ PNIPAM-Hzd $\\left(\\mathrm{T}3-6\\right)$ , both of which contain mixtures of only thermoresponsive polymers, exhibited no significant change in their compressive modulus after $^{72\\mathrm{~h~}}$ of acid degradation $\\left(p>$ 0.1 for both pairwise comparisons between the modulus measurements at the 0 and $^{72\\mathrm{~h~}}$ time points); this represents a substantially slower degradation rate than any of the single component thermoresponsive hydrogels (Figure 5A). Combining the nonthermoresponsive but more viscous CMC-Hzd with $\\mathrm{PO}_{0}\\mathrm{H}_{30}$ and PNIPAM-Hzd (T3−19) similarly suppresses degradation over the $^{72\\mathrm{~h~}}$ test period, consistent with the slower diffusion of water expected into this hydrogel. While the mixture modulus results are less dramatically different within the charged series, more neutral binary and ternary combinations (e.g., $\\mathrm{PO}_{100}\\mathrm{H}_{30}$ + Dextran − C2−4 or $\\mathrm{\\Delta}\\mathrm{\\supset_{100}H_{30}}+\\mathrm{PO_{100}H_{30}}\\mathrm{-cat}+\\mathrm{Dextran}-\\mathrm{C}3-\\mathrm{\\B{:}}$ ) show somewhat reduced degradation rates, consistent with the lower observed swelling in those hydrogels (Figure 4B). \n\nTransparency. Transparency measurements give insight into the homogeneity of the gels, allowing us to correlate, at least in part, the nonlinear changes in swelling, mechanics, and degradation noted in the previous sections to potential phase separation within these materials. Optical transparency is also an important parameter in some applications of injectable hydrogels (e.g., ophthalmic delivery) that are particular targets for protein-based therapies.49 Transmittance as a function of temperature was measured by ramping the temperature of a microplate reader from 25 to $40~^{\\circ}\\mathrm{C}$ and tracking the resulting transmittance at ${595}\\ \\mathrm{nm}$ , far outside the window in which any of the components of any of the combinatorial hydrogels would absorb due to chemical bonding; as such, the transmittance measurement corresponds to the light scattered by each hydrogel and, by extension, the number and size of phase-separated domains present in each hydrogel sample. As expected, low VPTT hydrogels such as $\\mathrm{PO}_{0}\\mathrm{H}_{30}$ and PNIPAMHzd showed low transmittance $48\\%$ for $\\mathrm{PO}_{0}\\mathrm{H}_{30}$ and $45\\%$ for PNIPAM-Hzd) even at $25~^{\\circ}\\mathrm{C}_{\\mathrm{\\ell}}$ , consistent with the deswelling response observed for both these hydrogels at room temperature (Figures 4A and 6A); further decreased transmittance values were observed as the temperature was increased to 40 $^{\\circ}\\mathrm{C}$ ( $29\\%$ for $\\mathrm{PO}_{0}\\mathrm{H}_{30},$ $30\\%$ for PNIPAM-Hzd, Figure 6A). In contrast, only a slight reduction in transmittance was observed for the $\\mathrm{PO}_{10}\\mathrm{H}_{30}$ single-component hydrogel (consistent with the reported ${\\sim}33\\ ^{\\circ}\\mathrm{C}$ onset transition temperature of this hydrogel48), and no change in transmittance was observed for higher transition temperature or nonthermoresponsive singlecomponent gels over the full temperature range probed (Figure 6A). None of the charged gels exhibited a thermal phase transition (Figure 6B), with all gels showing transmittance values of ${\\sim}95\\%$ , as expected within the probed temperature range. \n\nUpon mixing different components, a substantially broader range of transmittances was achieved, with the thermoresponsive (T series) hydrogels showing transmittances spanning from ${\\sim}20\\%$ (or less) to $595\\%$ (Supporting Information, Figures S8, S10, S12, and S14). For example, the ${\\mathrm{T}}3{-}3$ hydrogel noted to have a particularly large deswelling response $\\left(\\mathrm{PO_{10}H_{30}}+\\mathrm{PO_{100}H_{30}}+\\right.$ PNIPAM-Hzd, Figure 4A) also exhibited a particularly low transmittance ( $\\cdot<20\\%$ over the full temperature range), consistent with phase separation among the different transition temperature thermoresponsive components of this hydrogel. While the inclusion of CMC-Hzd preserved high deswelling, it also significantly increased the transmittance of the resulting hydrogels; the T2−12 hydrogel $\\left(\\mathrm{PO}_{0}\\mathrm{H}_{30}+\\mathrm{CMC}\\mathrm{-}\\mathrm{Hzd}\\right)$ that exhibited similar deswelling to the $_{\\mathrm{T}3-3}$ hydrogel maintained a transmittance of ${\\sim}50\\%$ at $25~^{\\circ}\\mathrm{C}$ . The PNIPAM-Hzd component particularly appears to play a key role in creating hydrogels with lower transmittances; for example, in the five-component thermoresponsive combinatorial gels (Figure S3D), the one hydrogel prepared without the PNIPAM-Hzd component $(\\mathrm{T}5-2)$ still maintained $>60\\%$ transmittance, while all formulations containing PNIPAM-Hzd had transmittances $<40\\%$ . \n\nInterestingly, some charged (C series) hydrogels yielded transmittances as low as $\\sim60\\%$ (Supporting Information, Figures S9, S11, S13, and S14), despite the fact that each single-component hydrogel was highly transparent (Figure 6B). In particular, binary mixtures of different carbohydrates (e.g., $\\mathrm{CMC}_{20}–\\mathrm{Hzd}+$ chitosan-Hzd − C2−14 or $\\mathrm{CMC}_{20}–\\mathrm{Hzd}+$ dextran-Hzd − C2−15) exhibited significantly lower transmittances than achieved by mixing different POEGMA precursor polymers, a result consistent with the different base chemistries and higher viscosities of the carbohydrate starting materials that could result in more thermodynamic phase separation and less effective mixing during hydrogel fabrication. In addition, higher-order mixtures, including chitosan-Hzd (e.g., C4−10, C4−12, C4−15, or any of the five-component gels aside from $C5{-}1$ , which excludes chitosan-Hzd), all exhibit lower transmittances than other combinations, potentially attributable to the reduced solubility of the carboxymethylated chitosan at physiological pH following the consumption of a portion of those carboxyl groups during the hydrazide functionalization process. As such, minimizing the use of PNIPAM-Hzd and chitosan-Hzd results in hydrogels with enhanced transparency without changing the swelling response. \n\n \nFigure 7. Cumulative ovalbumin release kinetics from (A) thermoresponsive vs nonthermoresponsive (T series) single component hydrogels and (B) charged vs neutral (C series) single component hydrogels in $10~\\mathrm{mM}$ PBS at $37^{\\circ}\\mathrm{C}$ $\\left(n=4\\right)$ ). See Supporting Information, Figures S9−S15, for the corresponding results for the combinatorial hydrogels. \n\nDrug Release Kinetics. To assess the kinetics of protein release from the combinatorial hydrogels, ovalbumin was dissolved in the aldehyde precursor solution used to prepare the high-throughput hydrogels, resulting in a uniform loading of $1.5\\mathrm{\\mg}$ ovalbumin/well. Ovalbumin $\\left(M_{\\mathrm{w}}\\:=\\:45\\:\\mathrm{\\kDa}\\right)$ was chosen as our model protein based on its reported role as an effective analogue of ranibizuman $(M_{\\mathrm{w}}=46~\\mathrm{kDa})$ ), a key antiVEGF antibody for treatment of eye disease such as age-related macular degeneration (AMD).50 By using a multiwell filter plate with individual well collectors, the cumulative protein release from each sample can be individually tracked over a one month release period. Figure 7 shows the cumulative ovalbumin release profiles (measured via the Bradford assay) for each single-component gel. Thermoresponsive hydrogels showed higher $\\%$ drug release than nonthermoresponsive hydrogels (Figure 7A), with the PNIPAM-Hzd gels that deswell the most exhibiting the maximum total drug release $(89\\%)$ and the highest day one burst release $(70\\%)$ , consistent with convective transport of the protein out of the gel as the thermal collapse occurs. Nonthermoresponsive gels such as $\\mathrm{CMC}_{40}–\\mathrm{Hzd}$ and $\\mathrm{PO}_{100}\\mathrm{H}_{30}$ exhibited substantially lower burst releases ( ${\\sim}32\\%$ day one release) but also significant retention of protein, with only ${\\sim}44\\%$ overall release achieved after one month. In comparison, the intermediate transition temperature $\\mathrm{PO}_{50}\\mathrm{H}_{30}$ single component hydrogel exhibited comparable burst release to the nonthermoresponsive gels ( ${\\sim}42\\%$ day one release) but significantly higher total release over the test period $\\left({>}60\\%\\right)$ , suggesting that moderate thermoresponsiveness may be beneficial for achieving prolonged protein release. The charged single component hydrogels (Figure 7B) showed significantly less burst release (between ${\\sim}10\\%$ to $28\\%$ after 1 day compared to ${\\sim}28\\%$ for the neutral $\\mathrm{PO}_{100}\\mathrm{H}_{30}$ control) but also significantly higher protein retention, particularly for the carbohydrate-based hydrogels, which retained between 58 and $68\\%$ of their cargo, even after one month (at least in the absence of oxidative degradation that is known to primarily degrade carbohydrate precursor polymers in vivo). This general result is consistent with the increased affinity of the charged gels for proteins,51,52 a result further demonstrated by the longer time frame of controlled release observed in the charged gels (up to 6−7 days) relative to the thermoresponsive gels (up to 2−3 days) despite the significantly higher water contents and, thus, lower diffusion resistances of the charged hydrogels (Figure 4B). \n\nThe combinatorial hydrogels again show interesting results in terms of manipulating the burst release, total duration of significant protein release, and the entrapped protein fraction of each hydrogel (Supporting Information, Figures S9−S15). While the charged combinations all trended similarly (i.e., samples with less burst release also released less protein, Figures S10, S12, and S14), the thermoresponsive hydrogels showed some independence among these variables. For example, the $\\mathrm{PO_{10}H_{30}\\ +\\ P O_{100}H_{30}\\ +\\ P O_{0}H_{30}}$ hydrogel containing a mixture of POEGMA-based polymers with different phase transition temperatures $\\left(\\mathrm{T}3\\mathrm{-}1\\right)$ exhibited a high burst release of ${\\sim}60\\%$ and a high cumulative release of ${>}90\\%$ , but effectively sustained ovalbumin release effectively over ${\\sim}10$ days, while the ${\\mathrm{T}}3{-}3$ gel in which the $\\mathrm{PO}_{0}\\mathrm{H}_{30}$ component is replaced with PNIPAM-Hzd exhibited similar burst and total ovalbumin release, but a substantially shorter release time of ${\\sim}3$ days. The binary combinations exhibiting the highest degree of deswelling (T2−2 and T2−10, Figure S4A) also exhibited the fastest and highest total ovalbumin release consistent with higher convective water transport from these gels, while binary hydrogels formed by combining CMCHzd with an intermediate transition temperature POEGMA precursor (i.e., $\\mathrm{PO}_{10}\\mathrm{H}_{30}\\ -\\ \\mathrm{T}2{-}5$ or $\\mathrm{PO}_{50}\\mathrm{H}_{30}\\ -\\ \\mathrm{T}2\\mathrm{-}14)$ exhibited the lowest total release. These results suggest that mixing different precursor polymers may be an effective approach for tuning protein release from hydrogels. Note that hydrazone chemistry has previously been shown to maintain good protein activity, suggesting its relevance for designing in situ-gelling protein delivery vehicles. \n\nLatent Variable Analysis and Optimization. Given the large amount of data generated by the high-throughput fabrication and characterization approaches developed, latent variable methods are ideally suited to fit the high-throughput data to a mathematical model and subsequently invert the model to identify optimal mixtures of precursor polymers that will provide desirable properties. While the thermoresponsive and charged polymer series were fabricated separately based on the experimental limitations of the high-throughput robotics system, a single latent variable model was built that combines data from both series by assigning a value of zero to any precursor polymer not included in a given hydrogel. For quantifying drug release kinetics, the kinetic curves in Figure 7 and Supporting Information, Figures S9−S15, were fit to the model $\\hat{\\boldsymbol{y}}^{}(t)=\\dot{\\boldsymbol{y}}_{f}+(\\boldsymbol{y}_{0}-\\boldsymbol{y}_{f})e^{-t/\\tau},$ in which $y_{0}$ approximated the burst release of protein from each gel, $y_{f}$ compensated for potential entrapment of protein within the gel, and $\\tau$ was the time constant related to the rate of protein release. A representative model fit is shown in Supporting Information, \n\n \nFigure 8. Coefficient plots relating the key output variables to the recipe variables (i.e., the type of polymer used, concentration of polymer used, degree of polymer functionalization, and polymer molecular weight) for (A) $y_{0},$ (B) $y_{\\beta}$ (C) $\\tau,$ and (D) hydrogel transparency. \n\nFigure S16, while Supporting Information, Figure S17 shows the root mean squared error for all drug release kinetic points evaluated. Overall, the modified first-order model can accurately fit the release profiles measured, with an average root mean squared error of only $1.7\\%$ drug released. As such, the model parameters extracted from these fits can reliably describe the overall drug release kinetics achieved with most of the 126 hydrogels analyzed via three fitted parameters that can be incorporated directly into predictive models. \n\nA global model was first attempted to be built that included all the parameters measured via high-throughput (i.e., swelling ratio, compressive modulus, degradation rate constant, transparency, and drug release kinetics). However, only the transparency $(R^{2}\\ =\\ 61.2\\%$ , $Q^{2}\\ =\\ 58.1\\%$ ) and drug release kinetics ${\\left(R^{2}=[y_{0};85.0\\%,y_{f};61.2\\%,\\tau;59.1\\%],\\right.}$ , $Q^{2}=\\bar{[{y_{0}};{83.5\\%}},$ yf: $58.1\\%,\\ \\tau\\colon55.9\\%]\\big)$ , where $R^{2}$ represents the percentage of overall variance explained and $Q^{2}$ represents the percentage of variance explained during cross-validation, could be fit with reasonable predictive confidence. The obvious nonlinear effects outlined in the discussion for the swelling, degradation, and mechanics measurements, coupled with the relatively large uncertainty inherent in some of the raw compressive modulus measurements, likely account for these relatively poor fits. As such, an alternative model was built that included all the gels tested from both series but only transparency and the drug release parameters as $y$ -variables, resulting in a 12 component PLS model which explained $67\\%$ of the variance in the output space (Supporting Information, Figure S18). The observed versus predicted plots for each of the drug release kinetic parameters (Supporting Information, Figures S19−S21) and the transparency measurements (Supporting Information, Figure S22) confirm the utility of the model for predicting hydrogel transparency and drug release kinetics. \n\nCorrespondingly, the coefficient plots associated with the model fit, summarized in Figure 8, as well as the loading biplot (Supporting Information, Figure S23) reflect many of the key qualitative trends identified in the experimental data analysis. In the coefficient plots in Figure 8, positive coefficients indicate that the variable in question is positively correlated with the hydrogel property considered in each panel, while negative coefficients indicate a negative correlation; the larger the absolute magnitude of the coefficient and the smaller the error bar (particularly if the error range does not cross zero), the larger the effect of that particular variable on a given hydrogel property. CMC suppresses drug burst, while PNIPAM-Hzd and other thermoresponsive POEGMA polymers ( $\\mathrm{\\DeltaPO_{0}H_{30}}$ and $\\mathrm{PO}_{10}\\mathrm{H}_{30}\\big)$ promote drug burst $(y_{0},$ Figure 8A) as well as more complete release of the protein from the gel $\\left(y_{\\hat{f}}\\right.$ Figure 8B). The incorporation of charged POEGMA precursor polymers and chitosan both result in larger $\\tau$ values (i.e., slower release) due to electrostatic interactions with the protein cargo (Figures 7B and 8C), while hydrogels prepared with the more hydrophilic $\\mathrm{PO}_{100}\\mathrm{H}_{30}$ precursor polymer result in lower $\\tau$ values (i.e., faster release) due to their high swelling (Figures 4A and 8C). The thermoresponsive PNIPAM-Hzd and $\\mathrm{PO}_{0}\\mathrm{H}_{30}$ precursor polymers negatively contribute to transparency, while the $\\mathrm{PO}_{100}$ precursor polymers enhance transparency (Figure 8D). Note that none of the pseudo variables representing weightings of different polymer compositions (i.e., molecular weight, polymer concentration, and $\\mathrm{{mmol}/\\mathrm{{g}}}$ functionalization) show obvious trends, given that these variables are highly correlated with the physicochemical properties of the polymers; for example, hydrogels prepared with Chitosan-Hzd or CMC-Hzd will inherently have lower weighted polymer concentrations and degrees of functionalization based on their higher viscosities and lower numbers of derivatizable functional groups. However, the match between the model predictions of the effects of each polymer component and both qualitative observations and theory clearly suggest the potential of the model for the prediction of key gel properties. \n\nTable 3. Model-Predicted Optimal Recipes for Each Optimization Criteriona \n\n\n<html><body><table><tr><td rowspan=\"2\">polymer precursors</td><td colspan=\"4\">M1-</td><td colspan=\"4\">M2-</td><td colspan=\"4\"> M3-</td></tr><tr><td>1</td><td>2</td><td>3</td><td>4</td><td>1</td><td>2</td><td>3</td><td>4</td><td>1</td><td>2</td><td>3</td><td>4</td></tr><tr><td>POH30</td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td></tr><tr><td>PO10H30</td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td>1</td><td></td><td></td><td></td></tr><tr><td>PO50H30</td><td>31</td><td>50</td><td>14</td><td>7</td><td>15</td><td>30</td><td></td><td>50</td><td></td><td></td><td></td><td></td></tr><tr><td>PO100H30</td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td>7</td><td></td><td></td><td></td></tr><tr><td>PNIPAM-Hzd</td><td></td><td></td><td></td><td></td><td>35</td><td>20</td><td>50</td><td></td><td></td><td></td><td></td><td>18</td></tr><tr><td>CMC40-Hzd</td><td>19</td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td></tr><tr><td> PO100H30C20</td><td></td><td></td><td>36</td><td>43</td><td></td><td></td><td></td><td></td><td>14</td><td>21</td><td>39</td><td>20</td></tr><tr><td>PO100H30A20</td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td>10</td><td>13</td><td>10</td><td>12</td></tr><tr><td>CMC20-Hzd</td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td>10</td><td>13</td><td>1</td><td></td></tr><tr><td>dextran-Hzd</td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td>8</td><td>3</td><td></td><td></td></tr><tr><td>chitosan-Hzd</td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td></tr><tr><td>PO10A30</td><td></td><td></td><td></td><td></td><td></td><td></td><td>39</td><td>7</td><td></td><td></td><td></td><td>4</td></tr><tr><td>PO100A30</td><td>50</td><td>50</td><td>50</td><td>50</td><td>50</td><td>50</td><td>11</td><td>43</td><td>50</td><td>50</td><td>50</td><td>46</td></tr></table></body></html>\n\nNumbers correspond to $\\mu\\mathrm{L}$ of each precursor component solution defined in Table 2 that were added to each well. \n\n \nFigure 9. Properties of optimized hydrogel formulations: $(\\mathbf{A},\\mathbf{B})$ minimize burst $\\left(y_{0}\\right)$ , maximize total protein release $\\left(y_{f}\\right)$ , maximize transparency (M1) optimization: (A) Cumulative ovalbumin release kinetics as a function of time ( $10\\ \\mathrm{mM}$ PBS, $37\\ ^{\\circ}\\mathrm{C})$ ); (B) transmittance measured at a wavelength of $595\\mathrm{nm}$ as a function of temperature. (C, D) Minimize burst $\\left(y_{0}\\right)_{\\cdot}$ , minimize release rate (maximize $\\tau$ ), and maximize total release $\\left(y_{\\mathscr{f}}\\right)$ M2) optimization: (C) cumulative ovalbumin release kinetics as a function of time ( $10\\ \\mathrm{mM}$ PBS, $37\\ ^{\\circ}\\mathrm{C})$ ); (D) transmittance measured at a wavelength of ${\\mathfrak{s o s}}_{\\mathrm{nm}}$ as a function of temperature. (E, F) Minimize burst $\\left(y_{0}\\right)$ , minimize release rate (maximize $\\tau$ ), maximize total release $\\left(y_{f}\\right)$ , and maximize transparency (M3) optimization: (E) cumulative ovalbumin release kinetics as a function of time ( $10\\mathrm{\\mM}$ PBS, $37~^{\\circ}\\mathrm{C}$ ); (F) transmittance measured at a wavelength of $595\\ \\mathrm{nm}$ as a function of temperature. Error bars represent the standard deviation of four replicate measurements $\\left(n=4\\right)$ ). \n\nThe model was next inverted to optimize the hydrogel compositions for protein delivery based on one of three criteria, each of which are relevant to different protein delivery applications: (M1) minimize burst $\\left(y_{0}\\right)$ , maximize total protein release $(y_{f})$ , maximize transparency (ideal for ophthalmic drug delivery in which minimizing the burst release is more important than maximizing the overall release period); (M2) minimize burst $\\left(y_{0}\\right)$ , minimize release rate (maximize $\\tau$ ), maximize total release $(y_{f})$ (ideal for other drug release applications in which transparency is not important); and (M3) minimize burst $\\left(y_{0}\\right)$ , minimize release rate (maximize $\\tau$ ), maximize total release $\\left(y_{f}\\right)$ , and maximize transparency (ideal for ophthalmic drug delivery applications in which slower release is as important as minimal burst). Each maximization/ minimization objective was equally weighted, although different weightings could be incorporated if desired to emphasize the importance of one or more parameters versus others. Four improved compositions $\\left(\\mathrm{A-D}\\right)^{-}$ were subsequently predicted for each optimization case, allowing any of the 12 hydrazide precursor polymers or 2 aldehyde precursor polymers available to be mixed at any mass ratio to create a new hydrogel. Table 3 shows the optimized recipes selected, while Supporting Information, Figure S23, shows the loading biplot displaying the initial high-throughput data in reduced two-dimensional space together with the relative locations of the new recipes identified by the model. \n\nOf note, the use of $\\mathrm{PO}_{50}\\mathrm{H}_{30}$ is suggested by multiple optimized recipes for criteria M1 and M2, consistent with the qualitative observations around the potential utility of this precursor polymer for prolonging release while avoiding convective burst. $\\mathrm{PO}_{100}\\mathrm{A}_{30}$ was recommended as the aldehyde polymer for most formulations, although combinations of the dual thermoresponsive $\\mathrm{PO}_{10}\\mathrm{A}_{30}/$ PNIPAM-Hzd precursor polymers were recommended for the M2 optimization in which transparency was not a targeted property. M3 also recommends the use of amphoteric hydrogels in each predicted optimized recipe, consistent with the model objectives to achieve both transparency and longer release periods (i.e., larger $\\tau$ values). Neither chitosan-Hzd nor $\\mathrm{PO}_{0}\\mathrm{H}_{30}$ is recommended for any formulation, likely based on the issues around transparency with chitosan-containing hydrogels (Figure 6) and the lower internal phase separation and thus potential for protein partitioning observed with $\\mathrm{PO}_{0}\\mathrm{H}_{30}$ relative to PNIPAM-Hzd (which otherwise exhibits similar properties). \n\nThe predicted optimized recipes were then fabricated and characterized for protein release kinetics and transmittance using the same high-throughput analysis techniques, the results of which are shown in Figure 9. Substantial improvements in hydrogel properties are achieved based on the optimization trials. For optimization M1, much higher transparencies are observed ( $597\\%$ in cases in which maximum transparency is a specific target) without suppressing the total release ( ${>}80\\%$ for $\\mathbf{M}1{-}2$ and $>70\\%$ for all other tests conducted) or promoting substantial burst release; this represents a combination of properties not achieved with any of the initial high-throughput screened recipes. For optimization M2, low burst release and high total release are similarly observed, although no obvious improvement was achieved with release duration despite it being a target parameter of the optimization; of note, one formulation had low transparency consistent with transparency not being a target parameter in this optimization. For optimization M3, low burst release, high transparency, and prolonged release periods ${\\bf>}10$ days) are all achieved with higher total release amounts than observed for any of the initial high-throughput samples with even comparable release durations, again showing the benefits of this optimization approach for designing functional hydrogels. \n\nFurther improvements in gel properties may be achievable upon additional iterations as the results from the first cycle of optimization are added back into the model, enabling the model to become more informed over a broader sample space. While the volume of high-throughput data provided herein as training data is not necessarily required for pursuing such a model-based optimization strategy, the rapid nature of data collection using the high-throughput protocols developed (i.e., successful synthesis of quadruplicate samples of 126 hydrogels within $<25~\\mathrm{\\min}$ and subsequent characterization of those hydrogels in $<30~\\mathrm{min}$ per plate for swelling, $<2\\mathrm{~h~}$ per plate for mechanics/degradation, and $<2~\\mathrm{\\min}$ per plate for transparency), coupled with the facile mixing-based synthetic protocol for preparing hydrazone cross-linked hydrogels that is ideally suited for automated liquid handling systems, make this combination of high-throughput synthesis and statistical modeling both practical and effective. To our knowledge, the approaches described herein for injectable hydrogel fabrication and characterization allow by far the fastest screening of gel properties reported to date. Such high-speed analysis is beneficial to identify functional hydrogel compositions for applications such as drug delivery in which the interplay between different hydrogel properties make explicit prediction of effective compositions challenging.",
|
||
"category": " Results and discussion"
|
||
},
|
||
{
|
||
"id": 5,
|
||
"chunk": "# CONCLUSIONS \n\nWe demonstrate the potential of automated high-throughput synthesis and characterization strategies to prepare and screen in situ-gelling hydrogel compositions for targeted applications. In particular, we demonstrate how mixtures of charged and thermoresponsive precursor polymers can reduce burst release, maximize total release, and slow the overall release kinetics of a model protein (here, ovalbumin). Hydrogel compositions optimization is significantly aided by latent variable statistical modeling strategies that provide predictive potential to identify new compositions with improved target properties based on previously collected data. The combination of a highthroughput strategy to rapidly collect large amounts of data (particularly enabled by the development of the suite of highthroughput hydrogel characterization protocols developed and optimized in this paper, the major bottleneck of most materials high-throughput screening applications) with a “big data” latent variable statistical approach that can quantitatively interpret this data thus represents a promising approach to rapidly identifying new injectable hydrogels for protein delivery and/or other applications in which suites of required target properties can be precisely defined (and, thus, optimized for) at the start of the development process.",
|
||
"category": " Conclusions"
|
||
},
|
||
{
|
||
"id": 6,
|
||
"chunk": "# ASSOCIATED CONTENT",
|
||
"category": " References"
|
||
},
|
||
{
|
||
"id": 7,
|
||
"chunk": "# $\\otimes$ Supporting Information \n\nThe Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.biomac.9b01132. \n\nTables describing all hydrogel compositions fabricated, schematics of the high-throughput drug release apparatus, high throughput data associated with all combinatorial hydrogel mechanics, swelling, degradation, transparency, and ovalbumin release kinetics, cytocomptability tests on the precursor polymers, sample model fits and squared error plots associated with the fits to the drug release kinetics data, goodness of fit data for the latent variable model constructed, observed versus actual plots for hydrogel transparency and drug release parameters, and a loading biplot showing the relative positions of the hydrogels tested in latent variable space are provided (PDF)",
|
||
"category": " Results and discussion"
|
||
},
|
||
{
|
||
"id": 8,
|
||
"chunk": "# AUTHOR INFORMATION \n\nCorresponding Author \n$^{*}\\mathrm{E}$ -mail: hoaretr@mcmaster.ca. \nORCID $\\circledcirc$ \nTodd Hoare: 0000-0002-5698-8463 \nNotes \nThe authors declare no competing financial interest.",
|
||
"category": " Abstract"
|
||
},
|
||
{
|
||
"id": 9,
|
||
"chunk": "# ACKNOWLEDGMENTS \n\nFunding from the Natural Sciences and Engineering Research Council of Canada (Strategic Project Grant # STPGP447372- 13) is gratefully acknowledged. Funding of Corbett’s postdoctoral fellowship by ProSensus Inc. and Mitacs (Accelerate Grant #IT08155) is also acknowledged. ProMV software for performing the latent variable analysis was provided free of charge by ProSensus Inc.",
|
||
"category": " Acknowledgments"
|
||
},
|
||
{
|
||
"id": 10,
|
||
"chunk": "# REFERENCES \n\n(1) Loh, X. J.; Peh, P.; Liao, S.; Sng, C.; Li, J. Controlled Drug Release from Biodegradable Thermoresponsive Physical Hydrogel Nanofibers. J. Controlled Release 2010, 143 (2), 175−182. \n(2) Zhang, J.; Muirhead, B.; Dodd, M.; Liu, L.; Xu, F.; Mangiacotte, N.; Hoare, T.; Sheardown, H. An Injectable Hydrogel Prepared Using a PEG/Vitamin E Copolymer Facilitating Aqueous-Driven Gelation. Biomacromolecules 2016, 17 (11), 3648−3658. \n(3) Hoffman, A. S. Hydrogels for Biomedical Applications. Adv. Drug Delivery Rev. 2002, 54 (1), 3−12. \n(4) Bakaic, E.; Smeets, N. M. B.; Hoare, T. Injectable Hydrogels Based on Poly(Ethylene Glycol) and Derivatives as Functional Biomaterials. RSC Adv. 2015, 5 (45), 35469−35486. \n(5) Nair, D. P.; Podgórski, M.; Chatani, S.; Gong, T.; Xi, W.; Fenoli, C. R.; Bowman, C. N. The Thiol-Michael Addition Click Reaction: A Powerful and Widely Used Tool in Materials Chemistry. Chem. Mater. 2014, 26 (1), 724−744. \n(6) Takeoka, Y.; Watanabe, M. Tuning Structural Color Changes of Porous Thermosensitive Gels through Quantitative Adjustment of the Cross-Linker in Pre-Gel Solutions. Langmuir 2003, 19 (22), 9104− 9106. \n(7) Breulmann, M.; Davis, S. A.; Mann, S.; Hentze, H.-P.; Antonietti, M. Polymer−Gel Templating of Porous Inorganic Macro-Structures Using Nanoparticle Building Blocks. Adv. Mater. 2000, 12 (7), 502− 507. \n(8) Griffin, D. R.; Weaver, W. M.; Scumpia, P. O.; Di Carlo, D.; Segura, T. Accelerated Wound Healing by Injectable Microporous Gel Scaffolds Assembled from Annealed Building Blocks. Nat. Mater. 2015, 14 (7), 737−744. \n(9) Lin, C.-C.; Anseth, K. S. PEG Hydrogels for the Controlled Release of Biomolecules in Regenerative Medicine. Pharm. Res. 2009, 26 (3), 631−643. \n(10) Mateen, R.; Ali, M. M.; Hoare, T. A Printable Hydrogel Microarray for Drug Screening Avoids False Positives Associated with Promiscuous Aggregating Inhibitors. Nat. Commun. 2018, 9 (1), 602. (11) Yu, Y.; Chau, Y. Formulation of In Situ Chemically CrossLinked Hydrogel Depots for Protein Release: From the Blob Model Perspective. Biomacromolecules 2015, 16 (1), 56−65. \n(12) Bakaic, E.; Smeets, N. M. B.; Badv, M.; Dodd, M.; Barrigar, O.; Siebers, E.; Lawlor, M.; Sheardown, H.; Hoare, T. Injectable and Degradable Poly(Oligoethylene Glycol Methacrylate) Hydrogels with Tunable Charge Densities as Adhesive Peptide-Free Cell Scaffolds. ACS Biomater. Sci. Eng. 2018, 4 (11), 3713−3725. \n(13) Smeets, N. M. B.; Patenaude, M.; Kinio, D.; Yavitt, F. M.; Bakaic, E.; Yang, F.-C.; Rheinstadter, M.; Hoare, T. Injectable Hydrogels with in Situ-Forming Hydrophobic Domains: Oligo(D,LLactide) Modified Poly(Oligoethylene Glycol Methacrylate) Hydrogels. Polym. Chem. 2014, 5 (23), 6811−6823. \n(14) Koehler, K. C.; Anseth, K. S.; Bowman, C. N. Diels−Alder Mediated Controlled Release from a Poly(Ethylene Glycol) Based Hydrogel. Biomacromolecules 2013, 14 (2), 538−547. \n(15) Buwalda, S. J.; Boere, K. W. M.; Dijkstra, P. J.; Feijen, J.; Vermonden, T.; Hennink, W. E. Hydrogels in a Historical Perspective: From Simple Networks to Smart Materials. J. Controlled Release 2014, 190, 254−273. \n(16) Hennink, W. E.; van Nostrum, C. F. Novel Crosslinking Methods to Design Hydrogels. Adv. Drug Delivery Rev. 2012, 64, 223−236. \n(17) Yu, Y.; Lau, L. C. M.; Lo, A. C.; Chau, Y. Injectable Chemically Crosslinked Hydrogel for the Controlled Release of Bevacizumab in Vitreous: A 6-Month In Vivo Study. Transl. Vis. Sci. Technol. 2015, 4 (2), 5. \n(18) Lau, C. M. L.; Yu, Y.; Jahanmir, G.; Chau, Y. Controlled Release Technology for Anti-Angiogenesis Treatment of Posterior Eye Diseases: Current Status and Challenges. Adv. Drug Delivery Rev. 2018, 126, 145−161. \n(19) Dang, J. M.; Leong, K. W. Natural Polymers for Gene Delivery and Tissue Engineering. Adv. Drug Delivery Rev. 2006, 58 (4), 487− 499. \n(20) Shelke, N. B.; James, R.; Laurencin, C. T.; Kumbar, S. G. Polysaccharide Biomaterials for Drug Delivery and Regenerative Engineering. Polym. Adv. Technol. 2014, 25 (5), 448−460. \n(21) Patenaude, M.; Hoare, T. Injectable, Degradable Thermoresponsive Poly(N-Isopropylacrylamide) Hydrogels. ACS Macro Lett. 2012, 1 (3), 409−413. \n(22) Sivakumaran, D.; Bakaic, E.; Campbell, S. B.; Xu, F.; Mueller, E.; Hoare, T. Fabricating Degradable Thermoresponsive Hydrogels on Multiple Length Scales via Reactive Extrusion, Microfluidics, SelfAssembly, and Electrospinning. J. Vis. Exp. 2018, 2018 (134), 1−12. (23) Yoon, J. J.; Chung, H. J.; Park, T. G. Photo-Crosslinkable and Biodegradable Pluronic/Heparin Hydrogels for Local and Sustained Delivery of Angiogenic Growth Factor. J. Biomed. Mater. Res., Part A 2007, 83 (3), 597−605. \n(24) Smeets, N. M. B.; Bakaic, E.; Patenaude, M.; Hoare, T. Injectable and Tunable Poly(Ethylene Glycol) Analogue Hydrogels Based on Poly(Oligoethylene Glycol Methacrylate). Chem. Commun. 2014, 50 (25), 3306−3309. \n(25) Smeets, N. M. B.; Bakaic, E.; Patenaude, M.; Hoare, T. Injectable and Tunable Poly(Ethylene Glycol) Analogue Hydrogels Based on Poly(Oligoethylene Glycol Methacrylate). Chem. Commun. 2014, 50 (25), 3306−3309. \n(26) Lutz, J.-F. Thermo-Switchable Materials Prepared Using the OEGMA-Platform. Adv. Mater. 2011, 23 (19), 2237−2243. \n(27) Lutz, J.-F.; Akdemir, Ö .; Hoth, A. Point by Point Comparison of Two Thermosensitive Polymers Exhibiting a Similar LCST: Is the Age of Poly(NIPAM) Over? J. Am. Chem. Soc. 2006, 128 (40), 13046−13047. \n(28) Lutz, J.-F.; Hoth, A. Preparation of Ideal PEG Analogues with a Tunable Thermosensitivity by Controlled Radical Copolymerization of 2-(2-Methoxyethoxy)Ethyl Methacrylate and Oligo(Ethylene Glycol) Methacrylate. Macromolecules 2006, 39 (2), 893−896. (29) Zhang, Y. S.; Khademhosseini, A. Advances in Engineering Hydrogels. Science 2017, 356, 3627. \n(30) Patenaude, M.; Hoare, T. Injectable, Mixed Natural-Synthetic Polymer Hydrogels with Modular Properties. Biomacromolecules 2012, 13 (2), 369−378. \n(31) Bakaic, E.; Smeets, N. M. B.; Dorrington, H.; Hoare, T. Offthe-Shelf” Thermoresponsive Hydrogel Design: Tuning Hydrogel Properties by Mixing Precursor Polymers with Different LowerCritical Solution Temperatures. RSC Adv. 2015, 5 (42), 33364− 33376. \n(32) Ward, M. A.; Georgiou, T. K. Thermoresponsive Polymers for Biomedical Applications. Polymers 2011, 3 (3), 1215−1242. \n(33) Caló, E.; Khutoryanskiy, V. V. Biomedical Applications of Hydrogels: A Review of Patents and Commercial Products. Eur. Polym. J. 2015, 65, 252−267. \n(34) Bakaic, E.; Smeets, N. M. B.; Barrigar, O.; Alsop, R.; Rheinstädter, M. C.; Hoare, T. PH-Ionizable in Situ Gelling Poly(Oligo Ethylene Glycol Methacrylate)-Based Hydrogels: The Role of Internal Network Structures in Controlling Macroscopic Properties. Macromolecules 2017, 50 (19), 7687−7698. \n(35) Campbell, S. B.; Patenaude, M.; Hoare, T. Injectable Superparamagnets: Highly Elastic and Degradable Poly(N-Isopropylacrylamide)−Superparamagnetic Iron Oxide Nanoparticle (SPION) Composite Hydrogels. Biomacromolecules 2013, 14 (3), 644−653. (36) Sivakumaran, D.; Maitland, D.; Hoare, T. Injectable MicrogelHydrogel Composites for Prolonged Small-Molecule Drug Delivery. Biomacromolecules 2011, 12 (11), 4112−4120. \n(37) Fernandes, T. G.; Diogo, M. M.; Clark, D. S.; Dordick, J. S.; Cabral, J. M. S. High-Throughput Cellular Microarray Platforms: Applications in Drug Discovery, Toxicology and Stem Cell Research. Trends Biotechnol. 2009, 27 (6), 342−349. \n(38) Xu, F.; Wu, J.; Wang, S.; Durmus, N. G.; Gurkan, U. A.; Demirci, U. Microengineering Methods for Cell-Based Microarrays and High-Throughput Drug-Screening Applications. Biofabrication 2011, 3 (3), 34101. \n(39) Tzoc Torres, J. M. G.; Nichols, E.; MacGregor, J. F.; Hoare, T. Designing Multi-Responsive Polymers Using Latent Variable Methods. Polymer 2014, 55 (2), 505−516. \n(40) Bakaic, E.; Smeets, N. M. B.; Dorrington, H.; Hoare, T. Offthe-Shelf” Thermoresponsive Hydrogel Design: Tuning Hydrogel Properties by Mixing Precursor Polymers with Different LowerCritical Solution Temperatures. RSC Adv. 2015, 5 (42), 33364− 33376. \n(41) Campbell, S.; Maitland, D.; Hoare, T. Enhanced Pulsatile Drug Release from Injectable Magnetic Hydrogels with Embedded Thermosensitive Microgels. ACS Macro Lett. 2015, 4 (3), 312−316. (42) Mateen, R.; Hoare, T. Carboxymethyl and Hydrazide Functionalized $\\beta$ -Cyclodextrin Derivatives: A Systematic Investigation of Complexation Behaviours with the Model Hydrophobic Drug Dexamethasone. Int. J. Pharm. 2014, 472 (1), 315−326. \n(43) Mateen, R.; Hoare, T. Injectable, in Situ Gelling, CyclodextrinDextran Hydrogels for the Partitioning-Driven Release of Hydrophobic Drugs. J. Mater. Chem. B 2014, 2 (32), 5157−5167. \n(44) Yang, X.; Bakaic, E.; Hoare, T.; Cranston, E. D. Injectable Polysaccharide Hydrogels Reinforced with Cellulose Nanocrystals: Morphology, Rheology, Degradation, and Cytotoxicity. Biomacromolecules 2013, 14 (12), 4447−4455. \n(45) Anitha, A.; Maya, S.; Deepa, N.; Chennazhi, K. P.; Nair, S. V.; Tamura, H.; Jayakumar, R. Efficient Water Soluble O-Carboxymethyl Chitosan Nanocarrier for the Delivery of Curcumin to Cancer Cells. Carbohydr. Polym. 2011, 83 (2), 452−461. (46) Fei Liu, X.; Lin Guan, Y.; Zhi Yang, D.; Li, Z.; De Yao, K. Antibacterial Action of Chitosan and Carboxymethylated Chitosan. J. Appl. Polym. Sci. 2001, 79 (7), 1324−1335. \n(47) Muteki, K.; MacGregor, J. F.; Ueda, T. Rapid Development of New Polymer Blends: The Optimal Selection of Materials and Blend Ratios. Ind. Eng. Chem. Res. 2006, 45 (13), 4653−4660. \n(48) Smeets, N. M. B.; Bakaic, E.; Patenaude, M.; Hoare, T. Injectable Poly(Oligoethylene Glycol Methacrylate)-Based Hydrogels with Tunable Phase Transition Behaviours: Physicochemical and Biological Responses. Acta Biomater. 2014, 10 (10), 4143−4155. (49) Ng, E. W. M.; Shima, D. T.; Calias, P.; Cunningham, E. T., Jr.; Guyer, D. R.; Adamis, A. P. Pegaptanib, a Targeted Anti-VEGF Aptamer for Ocular Vascular Disease. Nat. Rev. Drug Discovery 2006, 5, 123−132. \n(50) Osswald, C. R.; Kang-Mieler, J. J. Controlled and Extended Release of a Model Protein from a Microsphere-Hydrogel Drug Delivery System. Ann. Biomed. Eng. 2015, 43 (11), 2609−2617. (51) Schmaljohann, D. Thermo- and PH-Responsive Polymers in Drug Delivery. Adv. Drug Delivery Rev. 2006, 58 (15), 1655−1670. (52) Mura, S.; Nicolas, J.; Couvreur, P. Stimuli-Responsive Nanocarriers for Drug Delivery. Nat. Mater. 2013, 12, 991−1003.",
|
||
"category": " References"
|
||
}
|
||
] |