Self‐healing encapsulation and controlled release of vaccine antigens from PLGA microparticles delivered by microneedle patches

Abstract There is an urgent need to reduce reliance on hypodermic injections for many vaccines to increase vaccination safety and coverage. Alternative approaches include controlled release formulations, which reduce dosing frequencies, and utilizing alternative delivery devices such as microneedle patches (MNPs). This work explores development of controlled release microparticles made of poly (lactic‐co‐glycolic acid) (PLGA) that stably encapsulate various antigens though aqueous active self‐healing encapsulation (ASE). These microparticles are incorporated into rapid‐dissolving MNPs for intradermal vaccination. PLGA microparticles containing Alhydrogel are loaded with antigens separate from microparticle fabrication using ASE. This avoids antigen expsoure to many stressors. The microparticles demonstrate bi‐phasic release, with initial burst of soluble antigen, followed by delayed release of Alhydrogel‐complexed antigen over approximately 2 months in vitro. For delivery, the microparticles are incorporated into MNPs designed with pedestals to extend functional microneedle length. These microneedles readily penetrate skin and rapidly dissolve to deposit microparticles intradermally. Microparticles remain in the tissue for extended residence, with MNP‐induced micropores resealing readily. In animal models, these patches generate robust immune responses that are comparable to conventional administration techniques. This lays the framework for a versatile vaccine delivery system that could be self‐applied with important logistical advantages over hypodermic injections.


| INTRODUCTION
While vaccines represent our strongest weapons against contagious disease, several obstacles still limit their maximum potential. For example, most vaccines require booster doses to induce protective levels of immunity. Besides, the large molecular size of vaccine antigens often prevents oral administration, and hypodermic injections are necessary to elicit the desired immune response. The reliance on repeated hypodermic injections creates many logistical challenges, such as difficulties with storage, disposal, and administration via healthcare professional. This not only increases costs, but also decreases availability, particularly in developing nations. If the full booster schedule is not administered, an individual may not develop protective immunity. Furthermore, hypodermic needles are designed for intramuscular (i.m.) delivery. However, the muscle has a low level of resident antigen-presenting cells (APCs), thus requiring higher doses than would be needed when compared to more APC-dense tissue such as the skin. [1][2][3][4] In an effort to increase the availability of vaccines and improve worldwide vaccination coverage, the next generation of vaccines should reduce reliance on hypodermic injections. This could be achieved through multiple approaches. One option is developing single-administration vaccines, which may offer protective immunity from a single dose. 5 A second concept, which could be accomplished separately or in tandem, is to utilize alternative delivery devices such as microneedle patches (MNPs) that avoid the logistical hurdles of hypodermic needles and are also more patient-friendly. 4,[6][7][8][9] A promising route to developing safe single-administration vaccines is through controlled antigen release. 5,10,11 This can be pulsatile to mimic current prime-boost paradigms, 12,13 or continuous to mimic a naturally developing infection. 14,15 In either case, the goal is to develop protective immunity via extended/delayed antigen exposure from a single administration. A common approach for controlled release is to encapsulate the active ingredient in a bio-erodible polymer such as a poly(lactic-co-glycolic acid) (PLGA). [16][17][18] While this approach has generated commercial success with various small molecules and peptides, it has not historically translated well to biomacromolecules such as protein antigens. 19 This is primarily due to the harsh stresses experienced during fabrication and sterilization of, and release from, the microparticles, which are known to damage sensitive proteins. [20][21][22][23][24] Newer approaches however allow for the separation of microparticle fabrication from the act of protein loading, thus allowing stable protein to be encapsulated. [24][25][26] This method, termed active selfhealing encapsulation (ASE), employs a protein-trapping agent inside the microparticles, which draws protein into the microparticles from an aqueous solution at high efficiency, followed by a dynamic selfhealing process of microparticles' surface pores that trap protein inside the microparticles (Figure 1a). 24,26,27 This method is well suited for protein antigens, and has great potential for the development of a single-administration controlled release vaccine delivery system. Whereas these PLGA controlled release systems have potential for reducing overall dosing requirements, they still rely on hypodermic needles for administration, which are often disliked by patients, require serious storage and disposal considerations, and generally must be administered by a healthcare professional. 1,6,8,28 MNPs are an attractive alternative, as they do not suffer from many of the obstacles mentioned above. In brief, MNPs are typically patches containing small sharp projections ($100-1,000 μm) that penetrate into superficial layers of the skin and deliver a therapeutic payload intradermally (i.d.). 4,6,7,28,29 Due to their small size, the patches cause little or no pain and generally no bleeding. 30,31 They also have reduced storage/disposal requirements, and may dissolve entirely after application, leaving behind no biohazardous sharps waste, which reduces risk of accidental stick or reuse. 32 Furthermore, MNPs are generally preferred by patients over traditional hypodermic injections, and can be successfully self-administered without a healthcare professional. 8 Lastly, by delivering the payload to the skin, they take advantage of the potent intradermal immune system, which can generate stronger responses than what is typical of the muscle, or can generate equivalent responses from lower doses. 1,4,33,34 Explored here is the combination of controlled protein antigen release from PLGA microparticles loaded via ASE with the logistical and immunological benefits of administration via microneedles. PLGA microparticles are first fabricated without antigen present, containing only the common vaccine adjuvant Alhydrogel, and trehalose as a stabilizing and pore-forming excipient. A variety of different vaccine antigens are then loaded into the same microparticle formulation using the ASE loading paradigm. These microparticles are then incorporated in a MNP, where the controlled antigen release behavior is evaluated FIGURE 1 (a) Schematic of aqueous ASE loading method. Porous microparticles containing trehalose-stabilized Alhydrogel are fabricated and freeze-dried. Microparticles are soaked in an antigen solution, antigen enters the pores and adsorbs to Alhydrogel. The solution is then mildly heated, healing the pores and entrapping the antigen. Microparticles can then be collected, washed, and utilized. (b) SEM images of porous ASE microparticles after fabrication and lyophilization (left), and after loading and partial self-healing (right). Scale = 20 μm in vitro. These patches readily penetrate skin and then rapidly dissolve to deliver the microparticles i.d. where they reside to release antigen.
This system has great potential as a self-applied and versatile controlled release vaccine delivery system.

| Fabrication and evaluation of ASE-loaded PLGA microparticles
The formulation parameters of the ASE PLGA microparticles were selected to produce spherical, porous microparticles within the desired size range (10-60 μm) that demonstrated self-healing when incubated in solution above the hydrated PLGA glass-transition temperature (T g ). 27,35 The T g of the dry microparticles was 46.5 C, while after hydration this value dropped to 32.6 C (Supporting Information). The observed T g depression of the hydrated microparticles is expected because of the well-known plasticization effect of water on polymers. 36 The microparticles were well formed and highly porous as observed via scanning electron microscopy (SEM; Figure 1b). The hydrated microparticles had a volume-weighted mean diameter of 35.0 μm; larger than the limit up to which phagocytic cells can internalize a particle. 37 Thus, encapsulated antigen will likely be hidden from the immune system until it is released from the microparticles as soluble or adjuvant-bound protein.
The major advantage of the ASE loading strategy is it allows formulation optimization of the preformed microspheres in the absence of protein. This reduces the amount of potentially expensive protein wasted during pilot formulation studies. Any microparticles larger than the desired size could be excluded from the final product with particle sieves without wasting antigen.
To evaluate the microparticles' ability to load different antigens, dry and unloaded microparticles were co-incubated with various antigen solutions during a loading gamut that included 48 hr at 42 C.
After this period, the pores on the microparticle surface had partially or fully healed ( Figure 1b). This serves to close off some diffusion pathways for soluble antigen, and slows the inital burst release. While many different loading conditions, including varying antigen concentration, volume, or maximum temperature, successfully produced antigen-loaded microparticles, it was found using ovalbumin (OVA) as the model antigen that 0.5 ml of a 1 mg/ml OVA solution incubated with 20 mg of microparticles for 2 days at 4 C, followed by 1 day at room temperature, and 2 days at 42 C produced the best combination of wt/wt antigen loading and encapsulation efficiency (EE%) for this formulation (Supporting Information). Also, a variety of antigens, both model and clinically relevant, were successfully encapsulated into the exact same formulation of microparticles-that is, alterations to the microparticles were not needed to accommodate different antigens. The changes in wt/wt loading generally correlated with the antigens' affinity for the Alhydrogel that was included in the formulation (Table 1). It should be noted that the time used for self-healing encapsulation in the above protocol is undesirably long. Means to accelerate the loading via the use of plasticizers in the polymer, for example, are currently under investigation.
Alhydrogel is a common vaccine adjuvant currently included in many different vaccines. 38,39 It was loaded into the microparticles at 3.5% (theoretical wt/wt). The adjuvant binds to antigens to create colloidal particles, thus extending their residence time and increasing phagocytosis. [38][39][40] Here, Alhydrogel also acts as an agent to preferentially sequester antigen inside the microparticles. That is, during incubation antigen diffuses into the microparticle pores where it binds Alhydrogel to become trapped inside the microparticles (i.e., loading) before the surface pores heal under elevated temperature. Because Alhydrogel can bind to most proteins at pH above the protein's pI, it offers a versatile system to work with many different antigens without the need to change the microparticle formulation. The primary additional consideration is for thermoliable antigens-in this case the anthrax antigen rPA and the plague antigen F1-V. In order for these antigens to remain stable during the loading conditions, an appropriate stabilizer, such as 20% wt/wt trehalose added to the antigen solution was necessary, as has been previously reported. 41 While the addition of trehalose as an excipient stabilized the antigen, it also interferes with loading. When 20% trehalose was added to OVA controls, loading was reduced by 45% (data not shown). Thus, future studies focused on formulations with OVA and Hepatitis B surface antigen (rHBsAg).

| Development of microneedle patches containing ASE microparticles
Microparticles were then incorporated into microneedles composed of highly water-soluble materials (i.e., polyvinyl alcohol [PVA] and sucrose). In this way, the microneedles dissolve quickly in the skin (thereby allowing the MNP to be removed from the skin within a few minutes), leaving the microparticles deposited as a depot within the skin. MNPs were prepared using "standard" microneedles ( Figure 2a) and using "pedestal" microneedles that were mounted atop a pedastal to improve microneedle insertion into deformable skin (Figure 2c).
The PVA/sucrose composition was selected as it maximized solid content of the filling solution, while also providing an acceptable viscosity.
Previous work has explored incorporating nanoparticles into microneedles, [42][43][44] but incorporation of microparticles into microneedles has received limited attention. 29 In this study, the microparticles are large and thus remain extracellular during release. Furthermore, this is the first time microparticles loaded via the ASE loading  [24], which utilized similar microparticle formulations and an identical loading approach. AESEM.
technique have been utilized in a MNP, which is expected to improve antigen stability. 24 When making standard patches, microparticles could be readily observed in the microneedles, with few particles in the backing ( Figure 2a,b). The process was also easily adapted to include a pedestal design that increased the functional length of the microneedles while keeping the microparticles localized to the microneedle portion ( Figure 2c,d).
The standard and pedestal MN patches contained approximately 244 and 208 μg of microparticles, respectively ( Table 2). The difference was likely due to the extra manipulation required of the pedestal patches. Using the model antigen OVA, which loads into the microparticles at 1.6% (wt/wt), this corresponded to a final antigen dose of 4.0 and 3.4 μg/patch for standard and pedestal patches, respectively (Table 2). Also, each patch is expected to contain less than 10 μg of alhydrogel-well below the FDA limit of 0.85 mg/dose, even if multiple patches were administered. The antigen loading would be expected to change when using different antigens. To adjust dosage, several options are possible, such as changing the number of microneedles in the array, using multiple patches, or diluting the microparticles with a packing excipient. It may be challenging to incorporate additional microparticles in this size range into a microneedle without changing the overall geometry.
Pedestal-based microneedles are helpful for overcoming the elasticity of the skin and ensuring more full penetration/insertion of the microneedles into the tissue. Using a standard pyramidal/conical microneedle design, it is common for only 25% of the total microneedle volume to be dissolved or deposited in the tissue. [45][46][47] The pedestal design utilized here was crafted using three-dimensional (3D)printed master parts that were re-cast using soluble materials. While 3D printing lacks the micron-scale precision and accuracy of photolithograpy, presice dimensions and smooth surfaces are not generally required of the pedestal part, so 3D printing was an effective means of reducing fabrication costs and time. In addition, by creating a pedestal patch that is fully soluble, it eliminates considerations for disposal of biohazardous waste versus other two-part systems. 47,48 While the standard microneedles had a height of 600 μm, and the pedestal part was 800 μm tall, the final tip-to-base height of the pedestal patches was 1,183 AE 6 μm, suggesting roughly 200 μm of overlap between the pedestal and the microneedle, as confirmed by confocal imaging (Figure 2d).    51 Thus, a lower percentage desorbs and more remains inside the microparticles as a particulate complexed to Alhydrogel. 49,52 It is noteworthy that in vivo, the dissolution and clearance of the PVA/sucrose binding material would be anticipated to take additional time. Thus, the early stage of antigen release is expected to occur slower in vivo than under the in vitro test described above.

| In vitro controlled release
The stability of the antigen released during this early phase was Release from microparticles not incorporated into MNPs was generally similar, but with a slightly larger burst release and higher total percent soluble release (Supporting Information).
To confirm that the remaining fraction of antigen (antigen that did not desorb from Alhydrogel and release from the microparticles as soluble antigen) was still inside the microparticles and had not released as a soluble aggregate or degradation product, microparticles were subjected to total nitrogen analysis after 35 days of in vitro release (Table 3). After day 35, the standalone microparticles had released $70% of encapsulated OVA. As the polymer and other excipients are nitrogen-free, the total nitrogen content can be correlated back to protein content. Roughly 27% of encapsulated protein was recovered (97% total recovery). This strongly suggests the fraction of antigen that is not released during the soluble release phase is remaining inside the microparticles as a ligand-bound particulate complexed with Alhydrogel, although protein aggregation could not be ruled out.
To evaluate the release characteristics of this remaining antigen fraction, a fluorescently-labeled OVA (fOVA) was encapsulated into microparticles using the ASE technique. Figure 4a shows the colloidal particles formed by adsorbing fOVA onto Alhydrogel-small particulates no more than a few microns in diameter. This was more apparent at week 4, where the complex was now more visible, and heavy microparticle degradation was obvious. By week 6, the microparticles were fully degraded and the remaining fraction of Alhydrogel-complexed antigen was released and available for presentation to the immune system. Again, because of the larger size of these microparticles, antigen still encapsulated inside the microparticles is hidden from the immune system until release.

| Skin penetration and microparticle delivery
To evaluate skin penetration, excised porcine inner ear tissue was used. Standard and pedestal patches were pressed into taut skin with the thumb. Standard patches produced a full 100 clearly identifiable microchannels, while pedestal patches produced an average of 98 AE 2 (n = 5, AESEM) microchannels (Figure 5a,b). This suggests the patches possess the mechanical integrity necessary to penetrate skin tissue.  To verify that after the microneedles penetrate skin they dissolve i.d. to deliver microparticles, pedestal MNPs were fabricated with microparticles loaded with fOVA. The resulting MNPs were applied as above, but the patches were allowed to remain in the tissue for   After removing the partially dissolved patches from the tissue, it was apparent that some microparticles had not been deposited and remained on the patch after administration (Supporting Infomation).
The fraction left in the patch could not be determined gravimetrically, as the patches picked up a considerable amount of tissue and hair.
Rather, a GPC method was developed to quantify the mass of polymer left on the patch after administration. As shown in Table 2, the standard patches only delivered 25% of the microparticles (consistent with previous results, when presented), 45,46 while the addition of the pedestal improved this significantly, to 55%. In addition to the elasticity of the skin, insertion is likely limited by rapid dissolution of the microneedle tip, which could quickly become dull after insertion and prevent further tissue penetration. To further improve delivery, slower dissolving and materials could be investigated, possibly coupled with more advanced microneedle-fabrication methods.

| Skin resealing and in vivo microparticle tracking
A potential concern for advancing MNP technologies is the submillimeter pores introduced in the skin by application of the patch. If these micropores do not close quickly the potential for infection may exist, although prior reserarch suggests this risk is small. 53 While several studies have investigated the kinetics of skin resealing, the existing literature focuses on solid, nondissolving-type MNPs that do not deposit any material in or otherwise occlude the micropores. [54][55][56] Thus, it was necessary to explore the skin resealing kinetics after

| Immunizations via ASE microparticles and microneedles
To determine if the microparticles delivered hypodermically or via a MNP stimulate an immune response, mice were dosed with On day 42, blood was drawn and analyzed for anti-OVA or anti-rHBsAg total IgG serum levels, as well as IgG1 and IgG2c, which are indicators of Th2 and Th1-type immunity, respectively. 61 For both antigens, all microparticle and/or MNP-dosed groups showed high antigen-specific total IgG levels compared with the sham and soluble OVA groups, and were as-good-as or better than the conventional vaccine group with Alhydrogel-adsorbed antigen (Figure 8a,b) Blood was also drawn and analyzed on day 20, 1 day before booster doses). In this case, microparticles, but not MNPs, showed significantly higher antigen-specific IgG levels compared to controls. The the Alhydrogel-adsorbed group already showed a response. As was shown in Figure 3, the rHBsAg released more slowly (less released in the soluble phase), and thus the response takes longer to develop.
However, research suggests that slower releasing antigens generate stronger final immune responses than quick releasing antigen. 15 These results support further scientific development of hypodermic needle-free vaccination via ASE microparticle-containing MNPs, and that the ASE microparticles used in the MNPs may be a useful method for sustained exposure of antigen to the immune system. The MNPs produced responses that were generally equivalent to i.d. injection of microparticles, but did not rely on a hypodermic needle for injection. This seemingly minor detail actually has enormous consequences for improving vaccination coverage for reasons mentioned above, including higher patient acceptability, self-application, and easier storage/disposal. It is unexpected that i.m. injection of microparticles produced equivalent or occasionally stronger responses than i.d./ MN administration, as this trend is typically reversed in the existing literature. 33,42 The controlled release potential of these polymer-based delivery systems was also apparent. For example, rHBsAg was shown to release more slowly than OVA from the microneedles in vitro ( Figure 3). Before booster doses, it appeared that Alhydrogeladsorbed rHBsAg was producing a more robust IgG response, whereas after boost the responses were nearly equivalent. The faster releasing OVA, on the other hand, generated a response more quickly and surpassed the conventional formulation by day 42. This trend may suggest that the slower release of antigen could delay the development of the immune response, but may also lead to the production of a stronger response once release is complete. This is also true of the comparison between the MNPs and i.d. microparticles, as the patches were shown to release antigen more slowly than free microparticles. Lastly, by incorporating the antigen-loaded microparticles into a soluble MNP, multiple logistical, and scientific advantages become apparent. These patches are smaller, can be self-administered, and will dissolve completely to avoid creating biohazardous sharps waste. They are also more likely to be preferred by patients, as they generate minimal or negligible pain and bleeding and are unlikely to induce needlephobia. MNPs also utilize the powerful intradermal immune system, which may be more advantageous than traditional i.m. delivery.

| CONCLUSIONS
While additional modifications to the system could further improve its utility, this work lays a foundation for a self-administered vaccine system that is applicable to a variety of vaccines and thus disease states.
The inner-water phase was prepared by concentrating Alhydrogel (2%, Invivogen) to 6.35% via centrifugation and removal of excess solution, then 8% wt/vol trehalose was added and the slurry was

| Total nitrogen analysis
Total protein content was extrapolated from total nitrogen content using a modified automated Dumas technique. 62 Microparticle pellets were washed 3× with ddH 2 O, then freeze-dried. About 1-4 mg of microparticles were massed into tin pans, which were crimped to remove excess air. Samples were run on a Leco TrueSpec Micro CHN.
The instrument was first blanked without samples to establish atmospheric baselines. Carbon, hydrogen, and nitrogen standards were then set in the anticipated range of nitrogen mass using USP-grade EDTA. Lyophilized antigen standards were run to verify the percent nitrogen in the protein and set a Protein Factor. Microparticle samples were then dropped into the combustion chamber at 1,050 C, which converts all nitrogen to nitrogen gas, which is then quantified by a thermal conductivity cell. Protein content was determined by multiplying the nitrogen mass by the protein factor after first subtracting the nitrogen mass from negative controls (unloaded microparticles).
Percent protein could then be determined by dividing protein mass by total sample mass.

| Confocal microscopy
To visualize the distribution of antigen inside the microparticles after encapsulation, microparticles were loaded using an Ovalbumin-Alexa Fluor 647 conjugate (fOVA) similar to as described above. After washing, the microparticles were resuspended in ddH 2 O and placed on a glass slide with a coverslip and cross-sectional Z-stacked images were taken on a Nikon A-1 spectral confocal laser scanning microscope (CLSM) operating with a Cy5 filter and NIS Elements viewing and analysis software.
To evaluate the particulate release fraction, fOVA-loaded microparticles were resuspended in PBST at 37 C. At predetermined time points, a sample of the suspension was removed and washed with ddH 2 O before similarly imaging as above via CLSM. Images were compared against Alhydrogel that had similarly been loaded with fOVA and washed of unbound antigen.

| Microneedle insertion
For ex vivo evaluation of mechanical integrity, excised porcine ear tissue was used. The shaved inner skin with cartilage attached was separated from the outer skin and subcutaneous fat, and pinned taut.
MNPs were gently placed tip-down onto the skin, and pressed in firmly with the thumb for 10 s. The patch was then removed and Gentian Violet (Ricca Chemical Co.) was applied to the application site for 1 min before being wiped away with an alcohol pad. The application site was then cut away and imaged on a stereomicroscope (n = 5 for each patch type).

| In vivo microparticle tracking
The treatment of all experimental animals in these procedures were in accordance with University committee on use and care of animals (University of Michigan UCUCA), and all NIH guidelines for the care and use of laboratory animals. Pedestal MNPs were made loaded with fOVA and applied to male albino C57BL/6J mice as described above.
Two patches were applied per mouse, to the left anterior and right posterior dorsal flank. At predetermined time-points, the whole animal was anesthetized and imaged using a PerkinElmer IVIS Spectrum imaging system. Fluorescence data was processed using a region-ofinterest analysis with background subtraction using Living Image 4.5 software. Other study groups included mice given an i.d. injection to the same locations of an equivalent delivered dose of fOVA-loaded microparticles or soluble fOVA. Mice were kept on an alfalfa-free diet to reduce autofluorescence. Depilatory cream was not reapplied during the study, but hair was kept trimmed using electric razors (n = 4 mice/group, two applications per mouse). taken per application site, per animal, at each timepoint, and the TEWL chamber was allowed to re-equilibrate to environmental conditions before each measurement. To measure TEWL, the VapoMeter was gently pressed against the application site without manual tension applied to the skin. Data are presented as percent increase over an application control using ANOVA with Fisher's LSD. The application control consisted of a flat PVA/sucrose mock patch that did not contain any microneedles, but was applied similarly to other groups.

| Immunizations
C57Bl/6 (for OVA groups) or BALB/c (for rHBsAg groups) mice, 5-6 weeks old, five mice/group, were purchased from Jackson Laboratories. The choice of mouse strain was reliant on reagents available for the different antigens. One day prior to priming and booster immu- is an inventor of patents that have been or may be licensed to companies developing MNP-based products, a paid advisor to companies developing MNP-based products, and is a founder/shareholder of companies developing MNP-based products (e.g., Micro Biomedical).
The resulting potential conflict of interest has been disclosed and is managed by the Georgia Institute of Technology and Emory University.