Fabricating Degradable Thermoresponsive Hydrogels on Multiple Length Scales via Reactive Extrusion, Microfluidics, Self-assembly, and Electrospinning

We have successfully applied all these fabrication techniques to multiple polymer systems using only slight variations of the methods described in detail above for PNIPAM and POEGMA; however, users of these protocols must be cognizant of the potential issues that can arise when other polymers are substituted into these processes. In particular, increasing the viscosity of the precursor polymers can negatively impact both the processibility (especially in the microfluidic method) as well as the efficiency of mixing of the two precursor polymers. In addition, the gelation time of the polymers must be controlled at a rate dependent on the morphology targeted in order to avoid premature gelation that serves to inhibit flow or prevent interdiffusion of the reactive pre-polymers, essential to form the desired homogeneous gel structures. The specific limitations of each strategy, as well as approaches we have used to adapt these approaches to address such limitations at each fabrication length scale, are described below.

Bulk hydrogels via double barrel syringe co-extrusion
Gelation time is the key variable to control to ensure the efficacy of the double barrel syringe technique for forming bulk hydrogels. Polymers that gel too fast upon contact (<1-2 s) can clog the static mixer, either stopping flow entirely or resulting in non-stoichiometric amounts of the two precursor polymers being extruded from the syringe. We have found that gelation times >5 s are preferable (although not required) for the use of this technique; this is particularly important if replicate hydrogels are being cast for physical or mechanical analysis to ensure that each hydrogel cast has the same composition. Gelation time can be easily altered by changing the density of reactive functional groups on one or both precursor polymers (lower functional group density leading to slower gelation) or changing the concentration of the precursor polymers used to form the gel (lower concentrations leading to slower gelation)²¹. Alternately, replacing the (more reactive) aldehyde group with the (less reactive) ketone group as the electrophile in the gelling pair significantly reduces the gelation time without significantly changing the composition of the resulting hydrogel³⁵; polymers prepared with mixtures of aldehyde and ketone monomeric precursors can be used to tune the gelation time as desired without changing the concentration of precursor polymers used (and thus the mass percentage of solids in the resulting gel formed).

We would also note that the first hydrogel cast does not always have the same properties as subsequent hydrogels cast, an observation attributed to slight differences in the rate at which the contents of the two barrels actually reach the static mixer. As a result, we typically prime the double barrel syringe by extruding a small (<0.3 mL) fraction of gel prior to initiating the casting process to minimize such variability. Finally, while not typically problematic when using oligomeric synthetic pre-polymers, the viscosity of one or more precursor polymer solutions can pose a challenge in the context of this technique, both in terms of facilitating flow using simple thumb depression as well as promoting effective mixing within the static mixer. However, somewhat surprisingly, even precursor polymer solutions with sharply different viscosities still form relatively homogeneous hydrogels using the static mixer attachments described in the parts list (e.g., PNIPAM with a high molecular weight carbohydrate²⁶), suggesting that concerns about inefficient mixing as a result of mis-matched viscosities may not be significant at least on the bulk scale. If required, the use of a syringe pump (instead of the thumb) to drive flow and/or the use of a larger gauge needle at the outlet can help overcome issues associated with extrudability in these systems.

Microscale hydrogels via reactive microfluidics
The key step associated with the microfluidics approach for gel microparticle fabrication is the priming of the microfluidics chip with the two reactive polymers. If the polymers are delivered with different pressures or at different rates into the chip, the differential pressure can drive the backflow of one precursor polymer solution into the reservoir (or at least toward the reservoir) of the other precursor polymer. This results in gelation upstream from particle formation, effectively blocking flow and thus requiring chip disposal. The torturous path imprinted between each reservoir and the mixing point creates a significant resistance to backflow; however, even a trained operator will occasionally gel a chip before a stable flow regime is achieved. Based on our experience, between 1-2 min is typically required to stabilize the flows following the initiation of droplet formation (over which time relatively polydisperse gel microparticles are produced); if no problems are observed within the first 5-10 minutes of operation, it is likely that several hours of continuous monodisperse particle production can be achieved. The use of precursor polymers with relatively well-matched viscosities as well as non-instantaneous gelation times (at least >15 s preferable) greatly assists in avoiding such problems and promoting the formation of stable flows.

Note that various flow rates ranging from 0.01-0.1 mL/h in the aqueous phase and 1.1-5.5 mL/h in the oil phase have been tested using this chip design, leading to the fabrication of particles on the size range of ~25-100 µm according to the shear applied at the flow-focusing junction; faster flow rates equate to higher shear and thus smaller particles formed^31,32. Varying the oil flow rate while keeping the total aqueous flow rate low (~0.03 mL/h, as cited in the protocol) was found to be most efficient to control gel microparticle size without compromising either monodispersity or the lifetime of the device, both of which were observed to significantly decrease at the higher end of the cited total aqueous flow rates. Larger oil flow rates (>5.5 mL/h) to create smaller particles are possible, but increased the risk of chip delamination (a common limitation encountered with plasma-bonded PDMS microfluidic chips). Bonding the chips using another method may enable faster flow rates and thus smaller gel microparticle production, a strategy we are currently exploring. Decreasing the size of the nozzle may also help to reduce the size of the microparticles that could be produced, albeit at a heightened risk of premature gelation prior to particle formation. Slower flow rates tended to lead to flow instabilities and thus higher polydispersities and an increased risk of chip gelation; this limitation could be overcome by using a multichannel microfluidic flow control system that has higher stability and higher resolution than the standard syringe pumps used in this protocol.

The choice of oil was critical to the success of this protocol, as heavier oils (favorable in terms of preventing gel microparticle agglomeration after collection) led to much less consistent particle formation at the nozzle than the light silicone oil reported in the protocol. We hypothesize this reduced reproducibility is a result of lower consistency of syringe pumping of heavier oils, leading to more variable shear at the mixing point. Avoiding gel microparticle aggregation in the collection flask was also a challenge, particularly immediately at the exit from the microfluidic device at which point in situ gelation was not complete and large numbers of available reactive functional groups were available to form bridges between colliding particles in the collection bath. This challenge is addressed by: increasing the length of the exit channel on the microfluidic chip itself, maintaining the gel microparticles in laminar flow for a longer period of time to promote more complete gelation; adding the side channels after the nozzle to feed more oil into the chip and thus better separate the gel microparticles in this post-mixing channel without affecting the shear fields at the nozzle itself or the particle production rate; and adding a magnetic mixer to the collection flask to avoid gel microparticle sedimentation and maintain a larger average separation between adjacent particles. While very slow gelling polymers would likely improve the device stability and minimize issues with priming, such systems also were observed to significantly increase the risk of gel microparticle aggregation, as a larger number of reactive functional groups remains unreacted (and thus able to form inter-particle bridges) over a longer period of time. As such, gelation times on the order of 15-60 s appear to be optimal for this technique: slow enough to enable priming but fast enough to ensure most reactive functional groups are consumed prior to the gel microparticles exiting the laminar flow channel into the collection flask.

Finally, removal of the templating oil is essential to ensure that the resulting particles maintain the smart properties anticipated based on the composition of the pre-polymers added and enable use of these particles in a biomedical context. The pentane washing procedure described was highly effective in this regard for general gel microparticle production. However, the application of this technique in a direct biomedical context (e.g., on-chip cell encapsulation) would require re-evaluation of this protocol. We have also explored the use of olive oil, suggested to be a more inert oil in the context of contacting cells³⁶, as the dispersant. While particle formation was possible, the gel microparticle populations were significantly more polydisperse than could be achieved with mineral oil, at least with the current chip design. Thus, while the chip appears to be adaptable to both synthetic polymer and natural polymer gel microparticle formation³¹, a modified design may be required to exploit this technique more broadly across all possible material combinations.

Nanoscale hydrogels via reactive self-assembly
Nanogels have been formed using a very wide range of processing conditions, including different concentrations of seed polymer (0.5-2 wt%), different ratios of crosslinking:seed polymer (0.05-0.2), different temperatures (40-80 °C), different mixing speeds (200-800 rpm), and different heating times following the addition of the crosslinker polymer (2-60 min)²⁸. In terms of concentrations, the trends observed are generally as would be predicted, as higher concentrations of seed polymer lead to larger nanogels and higher ratios of crosslinker:seed polymer lead to nanogels with higher crosslink densities and thus lower thermoresponsivities. It should be emphasized that increasing the seed polymer concentration too high ultimately leads to bulk aggregation as opposed to nanoaggregation, consistent with what is observed in the conventional free radical precipitation process for forming thermoresponsive nanogels³. Shorter heating times were also found to be favorable for forming smaller and more monodisperse particles. We hypothesize that holding the nanoaggregate at longer times at a temperature above the LCST one or both of the precursor polymers increases the probability of aggregation upon nanogel collision, with the increased hydrophobicity of the hydrazone bond relative to either the precursor aldehyde or hydrazide functional groups making this aggregation more likely as the degree of crosslinking achieved is increased. Ultimately, shorter heating times are favorable from a process perspective, as a monodisperse nanogel population can be formed in as little as 2 min after crosslinker polymer addition; 10 min was found to be the longest time that could consistently produce monodisperse nanogels while also allowing for the production of more highly crosslinked nanogels. Interestingly, the method is remarkably insensitive to mixing, with nearly identical particle sizes and particle size distributions resulting from mixing at different speeds or even scaling the process to larger volumes. While initially surprised by this result, it likely speaks to the primary role of thermodynamics in regulating nanogel production.

To achieve low polydispersities, the colloidal stability and the degree of hydration of the nanoaggregate appear to be the key variables. For example, nanoaggregates prepared using the more hydrophilic hydrazide-functionalized polymers as the seed as opposed to the less hydrophilic aldehyde-functionalized polymers lead to nanogels with significantly lower polydispersities. The difference between the experimental assembly temperature and the LCST of the seed polymer is also critical. Operating at a temperature just above the seed polymer LCST ((T – LCST) < 5 °C) offers the highest probability of monodisperse nanogel formation; operating well above the LCST creates more hydrophobic and collapsed nanoaggregates that are more likely to aggregate and less likely to crosslink, while operating below the LCST results in a relatively non-compact seed polymer that cannot be effectively or reproducibly crosslinked. For the best prediction of particle monodispersity, we recommend first performing a UV/vis scan to measure the onset LCST of the seed polymer and subsequently performing the self-assembly process at a temperature 1-2 °C above that LCST.

Note that nanogels produced using this method could be lyophilized and redispersed without any change in colloidal stability, often not possible for self-assembled structures and in our view attributable to our crosslinking stabilization method. We also anticipate that only the seed polymer needs to be thermoresponsive for this method to work; use of cross-linking polymers that are either non-responsive or responsive to other stimuli may further broaden the ultimate applicability of this technique. Finally, since the mixing of the two reactive precursor polymers is in this case passive as opposed to active, gelation time is much less important in terms of process control relative to the other fabrication strategies described. However, even in this technique, keeping the total crosslinking time <30 min is desirable to minimize the risk of particle aggregation.

Nanofibrous hydrogels via reactive electrospinning
Controlling the gelation time of the reactive pre-polymers is again essential to the success of gel nanofiber production. In particular, approximately matching the residence time of the precursor polymers in the static mixer (controlled by changing the flow rate of solution from the double-barrel syringe as well as the length and tortuosity of the static mixer) with the bulk gelation time of the precursor polymers is essential both to preserve spinnability as well as ensure effective crosslinking of the spun fibers between the needle and the collector. Faster gelation leads to ineffective Taylor cone development and thus poor spinnability, while slower gelation results in an aqueous solution instead of a gel hitting the collector, resulting in spreading and the ultimate formation of a thin film gel instead of nanofibers. Working at residence times slightly below the bulk gelation time has also been found to be effective (and indeed preferable to reduce the risk of needle clogging) since water evaporation as the solution is spun effectively concentrates the precursor polymers in the stream and thus accelerates gelation kinetics during the spinning process. In this same vein, operating at higher needle-to-collector distances (>10 cm) is generally favorable in this process, as shorter distances reduce the time available for water evaporation and thus require more stringent control over the relationship between residence time and gelation time in order to preserve a nanofibrous product.

Note that the use of PEO (or another high molecular weight and easily electrospun polymer) is essential in this protocol to promote nanofiber formation, as the short and highly branched POEGMA oligomers cannot alone reach an adequate degree of entanglement to induce electrospinning; instead, electrospray results at all process conditions tested for POEGMA-only formulations (although this may also have applications for making degradable gel particles using this same chemistry). A minimum PEO concentration of 1 wt% (1 MDa molecular weight) is required to maintain a fully nanofibrous morphology. Note that the PEO can be removed from the fibers following a simple soaking procedure (deionized water, 24 h) without disrupting the integrity of the nanofibrous network; in this way, PEO acts more as a transient electrospinning aid than an essential component of the final nanofibrous product. Note also that various types of collectors, including simple aluminum foil (to create thin layer hydrogels that can delaminate from the collector upon soaking) as well as a rotating aluminum disk (to create thicker scaffolds) can be used in conjunction with this same technique, provided the other process variables controlling the rate of gelation, the rate of electrospinning, and the rate of water evaporation during electrospinning remain unaltered.

Interestingly, depending on the method used to prepare the different morphologies, significant differences have been observed in the degradation times of hydrogels prepared from the same hydrogel precursors. For example, POEGMA nanofibrous hydrogels degrade slower than bulk POEGMA hydrogels with the same composition despite their significantly higher surface area and thus access to water to hydrolyze the hydrazone bonds. We relate these differences to the inherent contrasts between the described protocols in terms of the geometry of mixing the precursor polymers, which may lead to internal gel homogeneities and/or morphologies that are significantly different and/or the in situ concentration of polymer precursors on the same time scale as gelation, particularly relevant in electrospinning due to the simultaneous water evaporation and crosslinking observed in this process. While this may somewhat complicate the choice of the precursor polymers if one polymer is targeted for use in each protocol, it may also offer a technical opportunity in terms of making hydrogels with one chemical composition but very different physical properties.

Overall, the methods described provide a strategy for fabricating degradable (or at least renally clearable) analogues of thermoresponsive polymers on multiple length scales (bulk, micro, and nano) and with multiple types of internal structures (particles or fibers). Such protocols address the key barriers to the successful translation of conventionally-prepared synthetic thermoresponsive materials to the biomedical field: injectability and degradability. We are continuing to explore the application of such materials in both drug delivery and tissue engineering applications ranging from the physical targeting of cancers, the transport of drugs across the blood-brain barrier, the therapeutic delivery of proteins at the back of the eye, the directional growth of tissues, and the thermoreversible adhesion and differentiation of cells, among other applications.