Arrest in the Progression of Type 1 Diabetes at the Mid-Stage of Insulitic Autoimmunity Using an Autoantigen-Decorated All-trans Retinoic Acid and Transforming Growth Factor Beta-1 Single Microparticle Formulation

Type 1 diabetes (T1D) is a disorder of impaired glucoregulation due to lymphocyte-driven pancreatic autoimmunity. Mobilizing dendritic cells (DC) in vivo to acquire tolerogenic activity is an attractive therapeutic approach as it results in multiple and overlapping immunosuppressive mechanisms. Delivery of agents that can achieve this, in the form of micro/nanoparticles, has successfully prevented a number of autoimmune conditions in vivo. Most of these formulations, however, do not establish multiple layers of immunoregulation. all-trans retinoic acid (RA) together with transforming growth factor beta 1 (TGFβ1), in contrast, has been shown to promote such mechanisms. When delivered in separate nanoparticle vehicles, they successfully prevent the progression of early-onset T1D autoimmunity in vivo. Herein, we show that the approach can be simplified into a single microparticle formulation of RA + TGFβ1 with surface decoration with the T1D-relevant insulin autoantigen. We show that the onset of hyperglycemia is prevented when administered into non-obese diabetic mice that are at the mid-stage of active islet-selective autoimmunity. Unexpectedly, the preventive effects do not seem to be mediated by increased numbers of regulatory T-lymphocytes inside the pancreatic lymph nodes, at least following acute administration of microparticles. Instead, we observed a mild increase in the frequency of regulatory B-lymphocytes inside the mesenteric lymph nodes. These data suggest additional and potentially-novel mechanisms that RA and TGFβ1 could be modulating to prevent progression of mid-stage autoimmunity to overt T1D. Our data further strengthen the rationale to develop RA+TGFβ1-based micro/nanoparticle “vaccines” as possible treatments of pre-symptomatic and new-onset T1D autoimmunity.


INTRODUCTION
Type 1 diabetes (T1D) is widely-viewed as a disorder of impaired glucoregulation primarily due to a pancreatic beta (β) cell-selective, largely lymphocyte-driven autoimmunity (1)(2)(3). Significant knowledge about pathogenesis of T1D derives from the spontaneously diabetic nonobese diabetic (NOD) mouse strain, believed to share etiopathogenesis with human T1D (4). TH1 and TH17 T-lymphocytes (T-cells) lead this process resulting in β cell destruction (5)(6)(7)(8). Dead and dying β cells are acquired by macrophages and DC in steady-state flux through islet structures which migrate into the pancreatic lymph nodes (PLN) where they amplify a vicious circle of T1D autoimmunity by triggering expansion of more β cell-autoreactive T-cells. Regulatory lymphocytes, especially T-cells that stably-express the Foxp3 transcription factor (Foxp3+ Tregs) prevent autoimmune diabetes in the NOD mouse (9) and there is strong evidence that they can regulate the pool of autoreactive effector CD4+ and CD8+ T-cells in humans (10)(11)(12). Indeed, autologous Tregs are now in different phases of clinical trials to preserve residual beta cell mass in new onset T1D patients (13)(14)(15)(16)(17). While ex vivo generation of Tregs has been realized, logistic hurdles stand in the way of expanding personalized Treg cell therapy at a population level (13)(14)(15)(16)(17). A simpler way to generate Tregs in vivo, and fortify other layers of immune tolerance using a "vaccine, " would be decisive in preventing and treating new onset disease.
We have demonstrated that microsphere formulations of antisense DNA oligonucleotides targeting the primary transcripts of CD40, CD80, and CD86 prevent and reverse T1D (18). We observed an increased prevalence of Foxp3+ Tregs together with a decrease in TH1 cytokine levels in successfully-treated, diabetes-free mice, especially in PLN, when injected into the abdominal region overlying the expected location of the pancreas. The effect was β-cell specific since T-cells from successfully-treated mice proliferated vigorously to alloantigen, but not to β-cell antigens in vitro. Protection offered by the microspheres was adoptively-transferable to immunodeficient recipients even in the presence of diabetogenic immune cells (18). In parallel, we had examined the effects of the antisense DNA oligonucleotides on dendritic cells (DC) and we discovered that they stimulated the DC to produce retinoic acid (RA) (19,20). These RA-producing DC were further shown to increase the frequency of IL-10+ regulatory B-lymphocytes (Bregs) as well as stimulate the proliferation of existing Bregs (19,20). We surmised that a microsphere formulation of the antisense DNA oligonucleotides could also confer RA-producing ability to DC. In fact, we subsequently discovered that DC isolated from the injection site of microsphere-formulated antisense DNA oligonucleotides acquired the ability to produce retinoic acid (RA) as they took up the microsphere formulations and migrated from this injection site to PLN. These DC, accumulated inside PLN, continued to produce RA. These microspheres were subsequently shown to mobilize DC and Tregs inside the PLN of mice exhibiting an ongoing ex vivo-induced inflammation (21).
While a number of factors, and different in vitro and in vivo environments can support the differentiation of Foxp3+ Tregs from precursor T-cells (22)(23)(24)(25), transforming growth factor beta (TGFβ) appears to be a common denominator (26)(27)(28)(29)(30)(31)(32). Evidence of the in vivo effect of TGFβ in Treg cell pool expansion is described in NOD mice treated with TGFβ. In these mice, TGFβ inhibits T1D development, and increases Treg frequency inside islets. Studies in human T-cells demonstrate TGFβ is necessary to induce Tregs. Treatment of human CD4+ cells with TGFβ increases the number of Tregs and expression of CD25 and intracellular CTLA-4. Expansion is due to increased proliferation and protection of cells from activation-induced apoptosis (33). TGFβ promotes induction of Tregs accompanied by an increase in Foxp3 expression. TGFβ converts CD4+ CD25-Foxp3-non-Tregs into CD4+ CD25+ Foxp3+ Tregs. On its own, however, TGFβ is unable to mediate Treg cell induction. RA appears to be an additional factor that licenses the induction process. In mucosal tissue, mature tolerogenic DC producing RA induce Foxp3+ Tregs via a TGFβ -dependent mechanism. RA enhances TGFβ signaling by increasing expression and phosphorylation of Smad3, and this results in increased Foxp3 expression, even in presence of proinflammatory IL-6 or IL-21 (34). Two studies addressed the role of retinoids in T1D using NOD mice. One examined the effect of high vitamin A concentrations on T1D development (35). RA was protective. The other demonstrated that RA inhibited disease development in multiple low dose streptozotocin and naturally-occurring T1D in NOD mice. Prevention was abrogated in Foxp3+ Treg-depleted mice. T1D hyperglycemia was reversible in new onset NOD mice so long as RA was available (36).
The conceptual feasibility of an RA-based immunoregulatory microparticle method to suppress inflammation has been demonstrated (37,38). PLGA-based particle RA formulation suppressed IL-17 production and RORγ(t) expression in Tcells polarized toward TH17 phenotype in vitro with similar potency to that of free drug. RA nanoparticles enhanced TGFβ -dependent Foxp3 expression and IL-10 production in T-cells in vitro with similar potency to free RA. T-cells polarized toward TH17 phenotype in presence of free and nanoparticulate RA similarly suppress ability to induce IL-6 production by fibroblasts. DC isolated from cervical lymph nodes and pulsed with PLGA nanoparticles efficiently induces proliferation of Foxp3+ Tregs in vitro. Further data demonstrate that nanoparticle formulations that contain either RA or TGFβ in combination with another drug (39,40), or in one instance administered together as separate particles (40) are effective in suppression of inflammation through regulatory cell networks.
Until now, no formulation that combined RA and TGFβ together with a T1D-relevant autoantigen, to induce antigenspecific immune hyporesponsiveness, was considered as a potential therapeutic vehicle. We present evidence, herein, that a novel and stable formulation of a RA and TGFβ-formulated single microparticle, decorated with a T1D-relevant autoantigen (Insulin B9-23 peptide) (41)(42)(43)(44) can prevent the onset of hyperglycemia when administered into NOD mice that are at the mid-stage of active islet-selective autoimmunity. Acute treatment of late stage autoimmune pre-diabetic NOD mice with the combined RA/TGFβ/T1D-relevant autoantigen microparticle formulation resulted in a mild increase in the frequency of regulatory B-lymphocytes (Bregs) inside the mesenteric lymph nodes (MLN), but not the PLN. These data suggest additional and potentially-novel mechanisms that RA and TGFβ could be modulating in the prevention of progression of mid-stage autoimmunity to hyperglycemic T1D.

Experimental Animals
Female NOD/LtJ mice were purchased from the Jackson Laboratories (Bar Harbor, ME) at 6-8 weeks of age and housed up to 34 weeks. Prior to randomization into the treatment arms, NOD female mice between 9 and 11 weeks of age were pre-screened to insure absence of overt hyperglycemia. Blood glucose was assessed using the One Touch Ultra Blood Glucose Meter (Lifescan, Malvern, PA). Animals were maintained in a specific pathogen-free environment in the Animal Facility of the Allegheny Health Network Research Institute. All procedures utilized were in full compliance with and approved by the Institutional Animal Care and Use Research Committee of the Allegheny Health Network Research Institute.

Synthesis and Characterization of RA and TGFβ-Formulated Microparticles
We have previously described surface-functionalized poly lacticco-glycolic acid (PLGA) particles as drug carriers (45)(46)(47)(48)(49). In particular, surface nickel (Ni)-formulated PLGA microparticles (PLGA-Ni) serves as the backbone of our formulations. PLGA-Ni particles were prepared by incorporating the metal chelating lipid 18:1 DOGS-NTA-Ni into the PLGA matrix using a doubleemulsion solvent evaporation method. Briefly, 90 mg of PLGA (50:50) dissolved in 2.4 ml dichloromethane (DCM) was admixed with 0.6 ml of 10 mg/ml DOGS-NTA-Ni dissolved in the same solvent. The organic mixture was slowly added to 20 ml 1% polyvinyl alcohol aqueous solution and homogenized at 25,000 rpm for 5 min on ice. These microparticles are designated as PLGA-Ni. Additional PLGA-Ni formulations include all-trans retinoic acid (RA) (Sigma-Aldrich, St. Louis, MO) or Nile Red dye (Thermo Fisher Scientific, Waltham, MA). 5.0 mg of RA or Nile red dye were loaded in the organic phase (DCM) with PLGA and the lipid. Using a solvent extraction method, loading of RA was determined to be 4.8 µg per mg of dry powder, with the efficiency determined to be 61% by BCA protein assay (data not shown). Microparticles containing Nile Red dye are designated as PLGA-Red in this manuscript. Particles were precipitated by evaporating DCM and collected by centrifugation, before lyophilization in 2% trehalose for longterm storage at 4 • C in a desiccator until use. His-tagged proteins were attached after washing and resuspension in low protein binding microcentrifuge tubes (Thermo Fisher Scientific). Five microgram of the following His-tagged proteins were incubated per 10 mg of microparticles for 1 h: G protein (pG), GFP (Thermo Fisher Scientific), TGFβ1 (Prospec, Rehovot, Israel), or human insulin B9-23 peptide (8 µg) (R&D Biosystems, Minneapoplis, MN). Using the BCA assay, 45.5% of the His-tagged TGF β1protein was associated with the microparticles (data not shown). RA-formulated, TGFβ1 and insulin B9-23 peptidemodified microparticles (all three components constituting the microparticle) were designated Ins-RT-NP.

Microparticle Microscopy and Size Determination
Fluorescence microscopy was performed with a IX53 inverted microscope (Olympus, Shinjuku, Japan) with a 20x objective. His-tagged GFP was admixed with particle suspensions, fixed on slide cover and mounted with Slowfade Diamond solution. Scanning electron microscopy (SEM) of the microparticle formulations was conducted after placing the sample solution (200 µL) directly onto sample stubs. Stubs were then covered in foil and left in a desiccator overnight before testing. Samples were examined using a S-3400N Scanning Electron Microscope 14 (Hitatchi, Tokyo, Japan) equipped with an AXS XFlash Detector 5010 (Bruker, Billerica, MA) and a BrukerNano e-Flash 1000+ (Bruker, Billerica, MA). Hydrodynamic size of the microparticle formulations was measured using a Zetasizer Nano-S; particles were washed to remove trehalose and resuspended in ultrapure water (pH 7.4); measurements were made with dilutions at 5, 2.5, 0.25, 0.0125, 0.00025, and 0.0000025 mg per ml to obtain consensus.

Formulation Release Kinetics
His-tagged GFP was used as a TGFβ1 surrogate to determine release kinetics in vitro. Briefly, 5 mg of the PLGA-Ni microparticles were washed and resuspend in 100 µl of pH 8 Tris buffer with 0.1% BSA and loaded with 200 µl of 0.025 mg/ml His-GFP to 0.2 ml pH 8 Tris buffer (0.5 ml Tris buffer in total) and incubated for 1 h at room temperature. Particles were then washed to remove unbound proteins before re-suspended in 1 ml of release medium (1% BSA) and transferred into a 1 ml syringe fitted with a 0.22 µm syringe. Samples (100 µl) were dispensed from the syringe kept at 37 • C at the specific time points. Equal volume of the release medium was replaced in the syringe after each sampling. The concentration of His-GFP collected at each time point was determined by measuring the fluorescence intensity (λ = 508 nm) using a Tecan M1000 spectrophotometer (Tecan, Mannedorf, Switzerland) with an established standard curve. Similarly PLGA-Ni microparticles bound with TGFβ1 were placed in a complete elution buffer. A small sample (5 µl) of the elution was loaded into individual wells of a 6-12% polyacrylamide gradient gel, together with molecular weight marker proteins (Thermo Fisher Scientific) and separated by SDS-PAGE in single dimension. The separated proteins were visualized in the gel in situ using silver staining. RA was found to release gradually from the particles at 37 • C in PBS. In a preparation of 10 mg of PLGA-Ni RA containing microparticles was dissolved in 1 ml of PBS, and the rate of release was monitored over 18 days in vitro by UV absorbance.

Ex vivo Microparticle Uptake
Spleens were harvested from NOD female mice euthanized at 10 weeks of age. A single cells suspension was made by first physically dissociating the spleen followed by a 10 min incubation in RBC lysis buffer (eBioscience) to remove red blood cells.
The cell suspension was then strained through a 100 µm filter and plated at 1 × 10 6 cells/well in complete RPMI cell culture media (Sigma Aldrich) plus 10% fetal bovine serum (Gibco). Cells were plated in a 96-well tissue culture plate in a final volume of 200 µl and allowed to settle prior to addition of microparticles. PLGA-Red microparticles were resuspended in tissue culture media. 0.25 mg of reconstituted particles were then added to the cell wells and incubated together at 37 • C for 18 h. Cells were then collected and washed three times to remove extra-cellular particles. The uptake of particles by cells was measured by flow cytometry (BD Influx) and the data were analyzed by FlowJo software version 10.7.1.

In vivo Microparticle Administration
Microparticles were prepared at a concentration of 10 mg per 200 µL vehicle volume for administration to mice. PLGA-Ni and PLGA-Red microparticles were resuspended in sterile saline in low protein binding microcentrifuge tubes. PLGA (RA)-Ni were incubated with His-tagged recombinant human TGFβ1 at a concentration of 25 µg/mL, or 5 µg total weight and His-tagged human insulin B9-23 peptide (3-letter amino acid sequence: H-Ser-His-Leu-Val-Glu-Ala-Leu-Tyr-Leu-Val-Cys-Gly-Glu-Arg-Gly-OH) was added at a concentration of 40 µg/mL, or 8 µg total weight. The microparticle suspension was maintained dispersed by pipet mixing and incubated at room temperature for 30 min. After incubation, the suspension was again mixed by pipetting and loaded into 1 ml syringes with 27 gauge needles. Eleven week-old female NODLtJ/Shi mice were administered the microparticle formulations by subcutaneous (s.c.) injection into the abdomen. None of the mice were hyperglycemic or exhibited dysglycemia at the time of microparticle administration.

Tissue Single Cell Collection and Processing
PLN, MLN and skin (a 1 × 1 cm 2 patch at the injection site) were resected from randomly-selected diabetes-free mice from all the treatment arms. Lymph nodes were physically dissociated and the tissue was strained through 100 µm filters (Fisher Scientific) to produce single cell suspensions. Viability of cells was assessed by Trypan Blue staining and counted in a Countess II FL device (Thermo Fisher). Skin was isolated and processed into single cells as described previously (50). Briefly, hair around the injection site on the mouse abdomen was removed with hair trimmers. After euthanasia a 1 × 1 cm 2 patch of skin was resected. Excess fat and connective tissue were scraped off the sample. Skin was incubated in a HBSS solution containing 5 mM EDTA, 10 mM HEPES, and 10% FBS for 30 min at 37 • C. Skin was then placed in HBSS media supplemented with 0.7 mg/mL collagenase (Sigma Aldrich) and cut into small pieces with surgical scissors. The tissue was incubated for 30 min at 37 • C, vigorously vortexed, and single cells were isolated through a 70 µm filter (Thermo Fisher Scientific). Isolated cells were further purified by density gradient centrifugation on Ficoll (GE Healthcare) prior to further processing.

Flow Cytometry
Single cells from lymph nodes, spleen and skin from mice randomly-selected for euthanasia in all treatment groups were treated with Mouse BD Fc block (BD Pharmigen) for 5 min to reduce nonspecific antibody binding. Cells were incubated with fluorochrome-labeled antibodies at the manufacturer's recommended dilutions for 30 min. Cells were then permeabilized for internal staining with FoxP3/Transcription Factor Staining Buffer Set (eBioscience) for nuclear proteins according to manufacturer's protocols. All antibodies were obtained from BD Pharmigen unless specified otherwise. DC were characterized as CD11c+ (clone REA754; Miltenyi Biotech) and CD45+ (clone 30-F11). Tregs were characterized as CD4+ (clone RM4-5), CD25+ (clone PC61), and FoxP3+ (Internal nuclear, clone FJK-16s) (eBioscience). Bregs were characterized as B220+ (clone RA3-6B2), CD19+ (clone 1D3), CD1d+ (clone 1B1), and CD5+ (clone 53-7.3). Elsewhere, we (20) and others (51)(52)(53)(54)(55)(56) have shown that the precursor B-cells to IL-10-producing Bregs as well as >50% of IL-10 actively-producing Bregs lie inside this CD1d+ CD5+ B-cell population. Appropriate fluorochrome and antibody type matched isotypes were used, as well as single stain controls for cytometer compensation. Cytometry was conducted in the BD Influx cell sorter with 50,000 events recorded per sample. The data were analyzed with FlowJo software version 10.7.1. Data presented exclude cellular debris and are displayed as a percentage of target cells within the total intact cell population.

Incidence of Hyperglycemia in Mice in vivo and Insulitis Scoring
Female NOD mice were randomly distributed into saline, PLGA-Ni, or Ins-RT-NP treatment groups with an n = 12 animals per group. Microparticles or vehicle were administered on Day 0, 3, 7, 14, and 21. Blood glucose was monitored twice a week. An animal was considered diabetic if two consecutive readings, spaced 2 days apart were ≥300 mg/dL. Animals were euthanized within 4 days of diabetes confirmation. Tissues collected from euthanized mice included PLN, MLN, pancreas, and spleen. To assess the grade of insulitis in randomly-selected sections of pancreata from diabetic mice as well as RT-NP recipients who were diabetesfree at 33 weeks of age, we visualized hematoxylin/eosin-stained sections by light microscopy at 10X magnification and assigned scores as follows: 1 = up to 25% of the islet mass infiltrated; 2 = between 25 and 50% of the islet infiltrated; 3 = between 50 and 75% of the islet infiltrated; and 4 = Between 75 and 100% of the islet infiltrated. Scoring was conducted in a blinded manner (i.e., pathologist was not aware of the treatment assigned to the mouse from which the sections of the pancreas were derived).

Statistical Analyses of the Data
Statistical analysis was performed with GraphPad Prism software version 7.0c. Student's t-tests, ANOVA, and Kruskal-Wallis tests were performed as appropriate to compare the statistical relevance of differences in outcomes between and among treatment groups. A p-value of ≤0.05 was considered significant. Survival curves (Log-Rank test, Mantel-Cox) were analyzed for significance between two treatments at a time, where a p < 0.0167 was deemed significant (Bonferroni correction for three groups). Graphs are displayed as mean and standard deviation. A Dixon q test was performed to remove any data point outliers.

TGFβ1 and Human Insulin B9-23 Peptide Bind Efficiently to PLGA-Ni Microparticles
The ability of PGLA-Ni microparticles to adsorb His-tagged proteins was confirmed by fluorescent microscopy. Figure 1A shows His-tagged GFP conjugated PLGA-Ni microparticles. PLGA-Ni microparticles were also imaged in scanning electron microscopy where they exhibited spherical morphology (Figure 1B), and self-association when formulated as Histagged TGFβ1 microparticles ( Figure 1C). In the absence of TGFβ1, the particles are found in the micron size range with a hydrodynamic diameter (D h ) of 867.5 ± 41.6 nm (PDI = 0.18; Figure 1D). Adding TGFβ1 to the particles increased the D h to 1,088.6 ± 15.6 nm (PDI = 0.15). The kinetics of RA release was subsequently characterized in vitro. RA was found to release gradually from PLGA-Ni microparticles at 37 • C in PBS with an average release rate of 1 µg/day from 10 mg of microparticles over 10 days in vitro, resulting in a total release of approximately 15 µg by day 18 at 37 • C. This represents 31% of the total RA contained in 10 mg of microparticles (Figure 2A). PLGA-Ni microparticles incubated with His-tagged GFP (as a protein surrogate for TGFβ1) showed a more rapid and complete release of GFP ( Figure 2B). This is indicative of the His-tag more readily detaching from the Ni ion compared to RA released from the intact or deformulating particle. GFP concentration was maintained at 150 ng/mL, yielding a concentration above the anticipated physiologically-active concentration of TGFβ1 (57). Additional preparations of PLGA-Ni microparticles were incubated with His-tagged TGFβ1 (a 17.3 kDa non-glycosylated fragment fused to a 4.5 kDa amino terminal hexa-histidine tag) and subsequently eluted. Figure 2C shows a representative result following SDS-PAGE of eluates in vitro from the TGFβ1formulated microparticles indicating that upon elution, the TGFβ1 protein exists predominately as dimers and monomers, with some higher order oligomers.

DC Take up PLGA-Ni Particles in vitro as Well as in vivo, Following Subcutaneous Administration
We first sought to confirm that the PLGA-Ni backbone would be taken up by DC in vitro. CD45+ CD11c+ DC were identified by flow cytometry (Supplementary Figure 1) in freshly-obtained splenocytes after a 18 h exposure to PLGA-Ni or PLGA-Red microparticles. In Figures 3A-C we show that Nile Red dye was detected in up to 85% of CD45+ CD11c+ DC incubated with PLGA-Red microparticles, with little detection in cell unexposed to the PLGA-Red microparticles or cells incubated with PLGA-Ni microparticles that were not formulated with Nile Red. We then sought to determine the uptake of PLGA-Red by DC in the skin following subcutaneous (s.c.) injection into 8-10 week-old female NOD mice. In Figure 3D, we show Having confirmed that PLGA-Ni delivery resulted in substantial accumulation inside DC, with the expectation that these DC would be converted into tolerogenic cells using RA and TGFβ1formulated microparticles, we then sought to determine if a PLGA-Ni formulation of RA and TGFβ1 additionally-engineered with adsorbed insulin peptide B9-23 (Ins-RT-NP) could alter the progression of the underlying autoimmunity in NOD female mice toward overt diabetic hyperglycemia. In Figure 4A, we show the timeline of microparticle injections, where each of the mice were administered five s.c. injections of either vehicle saline, PLGA-Ni, or Ins-RT-NP starting at 11 weeks of age. This is an age where insulitis is evident inside the pancreas of NOD mice. Prior to randomization into the treatment arms, NOD female mice between 9 and 11 weeks of age were prescreened to insure absence of hyperglycemia. Fasting blood glucose, immediately prior to treatment, did not exceed 131 mg/dL (Figure 4B). In Figure 4C, we show that Ins-RT-NP administration beginning at 11 weeks of age when substantial insulitis is known to exist, prevented the onset of hyperglycemia in a significant proportion of recipients compared to mice treated with microparticles that were devoid of RA, TGFβ1 and insulin B9-23 peptide, or mice treated with vehicle alone. It is noteworthy that the diabetes-free state was maintained for a substantial amount of time (33 week-old mice; 22 weeks diabetes-free following treatment) without additional injections of the microparticles. Insulitis grade was significantlylower in pancreas sections of 33 week-old diabetes-free RT-NP recipients compared to all the diabetic mice in the control arms (Figures 4D,E).

Increased Frequency of Bregs but Not Foxp3+ Tregs in Ins-RT-NP-Treated NOD Mice
Data by others indicate that nanoparticle formulations of RA and TGFβ can convert T-cells into Tregs (37), and when administered (nanoparticles or hydrogels) into NOD female mice, they foster an increase in the number of Foxp3+ Tregs inside the PLN and/or the pancreata of diabetes-free mice (58,59,62,63). Since there were no surviving mice in the control treatment arms at 22 weeks following test article administration, it proved challenging for us to interpret the differences in Tregs in PLN and MLN obtained from vehicle or PLGA-Ni-treated mice to those in the diabetes-free Ins-RT-NP recipients. Therefore, to ensure matched populations and matched times at which PLN and MLN were collected in order to obtain interpretable data, we used an acute model of test article administration. For this, 11 week-old NOD mice in all treatment arms (n = 7-9) were euthanized 1 or 3 days following microparticle or vehicle administration. We then measured regulatory immune cell population numbers by flow cytometry (Supplementary Figures 2, 3) in PLN and MLN from these mice. Much to our surprise, and in contrast to our expectations, the frequency of Foxp3+ Tregs inside the PLN among all treatment groups was not different (Figure 5). There was a surprising lower, albeit statistically-not significant, Treg content in the MLN of mice treated with Ins-RT-NP compared to the controls (Figure 6). Considering the possibility that the viability of the cells inside the pancreatic and/or mesenteric lymph nodes could be affected by microparticles drained into the tissue (PLN and/or MLN), we measured the viability of single cells from dispersed lymph node tissue. We did not see any significant differences in viability of single cells from PLN and MLN of diabetes-free mice (Supplementary Figure 4). We then looked for possible differences in Breg frequency. Although we could not discern any differences inside the PLN of mice among the three treatment arms (Figure 5), we confirmed a small but statistically-relevant difference inside the MLN of mice treated with the Ins-RT-NP (Figure 6) at 3 days following administration. The ratio of Tregs to Bregs in PLN at 1 or 3 days following test article administration was statisticallyindistinguishable among the treatment arms ( Figure 5). In the MLN, however, the ratio was statistically-different in the Ins-RT-NP-treatment arm at 3 days post-administration compared to the other treatments (Figure 6).

DISCUSSION
Although insulin therapy can adequately manage the day-to-day glucose control in T1D, the lifestyle changes, the considerable risk of insulin-induced hypoglycemia, and the progression of many T1D patients to cardiovascular, renal, neural, and ophthalmic complications, indicate that pharamcologic insulin is not a cure. The once widely-held dogma that insulinrequiring patients were devoid of a substantial mass of beta cells inside the pancreas due to autoimmune destruction, has now yielded to a more balanced view. Based on accumulating evidence, a respectable mass of beta cells remains in T1D patients whose disease duration can be long-standing (64-69) that is functionally-impaired, although potentially "reactivatable" if the underlying autoimmunity as well as nonautoimmune pancreas-selective innate inflammation can be attenuated. A number of approaches have been implemented over the past decade to depress the underlying autoimmunity including single or combination agents from the categories of pharmacologic immunosuppressives, cytokine neutralizing antibodies, lymphocyte depletion, and T1D autoantigens (70)(71)(72)(73)(74)(75)(76)(77)(78)(79)(80). Where animal studies demonstrated sufficient efficacy, when implemented as clinical trials, each of the referenced approaches could not achieve long-term insulin independence (70)(71)(72). In fact, systemic as well as parenteral administration of most of these agents resulted in significant toxicity and adverse effects.
Micro/nanoparticle-based delivery of immunomodulatory agents, with or without autoantigen co-delivery to treat autoimmune disease, offers many attractive advantages over systemic immunomodulators. Site-specific delivery minimizes off-target effects, licenses leukocytes that can be "reprogrammed" to build regulatory leukocyte networks that can become antigen-specific, even in the absence of exogenous supply of autoantigen(s). We have demonstrated the latter using a specific abdominal site in mice and non-human primates, whose epidermal space is served by lymphatic drainage that accumulates or transits, in part, inside the PLN (18,21). Taking advantage of this intriguing anatomical feature, we demonstrated that autoantigen-free microsphere formulations of antisense DNA targeting the primary transcripts of CD40, CD80, and CD86, FIGURE 4 | Prevention of diabetes onset following Ins-RT-NP administration. Non-diabetic female NOD mice at 11 weeks of age were randomly assigned to three treatment groups (n = 12 per group). Each mouse received five subcutaneous abdominal injections of saline vehicle, 10 mg of PLGA-Ni in solution, or 10 mg of Ins-RT-NP in solution as shown in the dosing schedule timeline (A). The timeline also shows the tissue harvest time points ("*") from an additional cohort of animals ("acute study cohort"; Figure 5). (B) No significant difference was observed in the starting blood glucose values (prior to microparticle injection; p = 0.981, Kruskal-Wallis test). (C) Mice were assessed for diabetes status by measuring tail vein blood glucose, where two consecutive readings over 300 mg/dL indicated diabetes onset. Shown in the graph are the survival curves representing diabetes-free mice. Diabetes incidence in Ins-RT-NP-treated mice was significantly lower when compared to saline (p = 0.0119) and PLGA-Ni (p = 0.0001) treatment. A log-rank test (Mantel-Cox) confirms significance in the differences among the treatment groups (p < 0.0167; p < 0.05 Bonferonni-corrected for three groups). (D) Insulitis grade was significantly-lower in pancreas sections of 33 week-old diabetes-free RT-NP recipients compared to all the diabetic mice in the control arms. Shown are images at 10X magnification representative of >5 islets/field visualized per treatment arm. (E) Summary of insulitis scores between all diabetic mice and the Ins-RT-NP recipients who were diabetes-free at 33 weeks of age.
preferentially accumulated inside PLN, generated Foxp3+ Tregs therein, and in NOD mice, prevented the progression of latestage dysglycemia to overt clinical diabetes and also achieved reversal of new-onset hyperglycemia and long-term stability of a normoglycemic state (18,21).
Our technology (18,21) was developed to exploit the phagocytic nature of epidermis-residing as well as migratory DC to take up exogenously-administered particles loaded with molecules that are able to direct them toward a functionallytolerogenic state in vivo. Since then, other groups have developed conceptually similar architecture with different approaches to stabilize functional tolerance in DC (40,58,59,(81)(82)(83)(84)(85)(86)(87)(88)(89)(90)(91)(92). These approaches have resulted in successful treatment of autoimmune disease in animal models of T1D (58,59) and multiple sclerosis (81)(82)(83)85). As we continue to decipher the mechanisms that our earlier-reported microspheres use to confer tolerogenic functional properties to DC beyond the knockdown of the primary transcripts of CD40, CD80, and CD86, we have uncovered the involvement of RA (19). The combination of the DNA antisense oligonucleotides stimulates the production of RA by DC in a manner that involves cell-intrinsic retinaldehyde dehydrogenase 1 (RALDH1) (19).
The combination of RA and TGFβ1 is critical in peripheral generation of immunosuppressive T-cells characterized by the expression of the Foxp3 transcription factor (Foxp3+ Tregs) (93)(94)(95)(96)(97)(98). Tregs prevent emergence of autoimmune disease (99,  Figure 4A) were euthanized at 1 or 3 days following test article administration. Single cell suspensions from resected PLN were then assessed by flow cytometry for Treg and Breg frequency. There were no discernible differences in the Treg or Breg frequency in the PLN among the mice of the three treatment arms at 1 or 3 days following microparticle administration. Similarly the ratios between Treg and Breg cells were unaltered. Data are combined from the outcomes in two independent animal cohorts examined at different times in the study (n = 7-9 per arm). 100). Subsets of Foxp3+ Tregs are found in mice and humans: natural Treg and induced Treg (iTreg) (101,102). In the periphery, iTreg differentiate from naive T-cells under subimmunogenic conditions of antigen presentation and in the presence of RA and TGFβ (103). That RA is a cofactor in the generation of iTregs stems from in vitro findings that mesenteric lymph node and lamina propria DC potently induced iTreg differentiation in the presence of TGFβ (98,104,105). Addition of RA to co-cultures with splenic DC and TGFβ enhances iTreg induction (94,98,105). Exogenous RA sustains iTreg generation in conditions that typically oppose it, such as presence of inflammatory cytokines (IL-6, IL-21) and high co-stimulatory environments (34,94,98). These observations made for a very compelling case to combine RA with TGFβ1 with a T1D-relevant autoantigen in a microparticle formulation. One could argue that non-formulated RA and TGFβ1 coadministered could achieve similar if not identical outcomes thereby questioning the need to engineer micro-/nanoparticle formulations of the agents. To our knowledge, and also in our experience, administration of RA or TGFβ1 alone or together as an injectable suspension cannot prevent T1D (unpublished observations) even though iTreg frequency is increased inside the site of the injection in the skin but not the draining lymph nodes. Also, the iTregs are not functionally-stable in suppression assays (unpublished observations). As the studies reported herein were underway, a number of supportive findings were published using particle designs different from ours. Keselowsky et al., using a mixture of two separate microparticle populations (population 1 consisting of TGFβ or IL-10 and population 2 consisting of either rapamaycin or RA), conferred a tolerogenic-reminiscent phenotype in DC in vitro (40). They also developed a more complex variation where one of the microparticle populations contained either GM-CSF or TGFβ and the other contained vitamin D3 and denatured insulin, to prevent progression of early-stage autoimmunity to diabetic hyperglycemia and to reverse, to some degree, new-onset hyperglycemia (59). This approach resulted in protective outcomes similar to those achieved by Miller et al. in mouse models of multiple sclerosis (81,83,85). Mechanistically, the involvement of DC as mediators of a tolerogenic state, mainly via the increase in Treg  Figure 4A) were euthanized at 1 or 3 days following test article administration. Single cell suspensions from resected MLN were then assessed by flow cytometry for Treg and Breg frequency. A non-significant downward trend is seen in Treg frequency in MLN of mice treated with Ins-RT-NP vs. saline. Breg frequency exhibited an upward trend in mice treated with Ins-RT-NP, becoming statistically-significant significant compared to control test articles by the time of the third injection (n = 9, ANOVA p = 0.0262). Tukey's multiple comparisons determined p = 0.0215 between the saline and Ins-RT-NP treatment group. The Treg/Breg ratio was similarly significantly-different with the three-injection treatment (ANOVA p = 0.0035, post test p = 0.0025 for saline vs. Ins-RT-NP). Statistical significance is displayed as * for p < 0.05 and ** for p < 0.01 as assessed by ANOVA. Data are combined from the outcomes in two independent animal cohorts examined at different times in the study.
In addition to the increase in the frequency of Tregs, we have shown that one mechanism by which RA participates in the induction of immune hyporesponsiveness is through the differentiation of B-lymphocytes into IL-10+ Bregs as well as the proliferation of existing Bregs (19,20). Although initial studies indicated that suppression ability was specific for IL-10+ CD1d+ CD5+ B-cells, it was later discovered that the progenitors of as well as the majority of IL-10producing Bregs resided inside CD1d+ CD5+ population and that additionally, IL-10 expression was not a condition sine qua non for CD1d+ CD5+ B-cell suppressive ability (51)(52)(53)(54)(55)(56). To date, none of the previous studies in the delivery of potentially-tolerogenic micro/nanoparticles into animal models of autoimmunity have studied their effects on Bregs as possible unique or Treg-overlapping cellular mediators of the action of the micro/nanoparticles. Our report here is the first to demonstrate the possible involvement of Bregs in the absence of increases in Treg frequency in NOD mice acutely-treated with single microparticle-formulated RA, TGFβ1 and insulin autoantigen peptide. What is, at the moment, unclear, is why Bregs are elevated inside the MLN when compared to the PLN (Figures 5, 6). Although NOD mice treated with Ins-RT-NP exhibit Foxp3 positivity at 22 weeks from the time of the last microparticle injection (data not shown), in the absence of age-matched diabetes-free and age-matched newonset diabetic mice treated with the control test articles, it is currently not known if the frequency of Tregs or Bregs changes longitudinally resulting in a greater population of these cells inside the PLN and/or the MLN of Ins-RT-NP-treated mice. If so, the observations we present following acute administration will not reflect a mechanistically-interpretable outcome, when measured longitudinally. These studies are currently underway with large animal cohorts to capture age-matched diabetic and non-diabetic control microparticle recipients. Nevertheless, the microparticle formulation, when delivered into NOD mice is protective against the progression of mid-stage autoimmunity to overt clinical hyperglycemia.
Our Ins-RT-NP delivery system has several advantages over the other reported approaches, referenced above. First is in the design complexity. Our system consists of a single formulation that brings together only two bioactive agents (RA and TGFβ1) with the addition of a unique T1D-relevant autoantigen. Second, from a biologic aspect, DC need acquire only one population of particles without the potential of competition for uptake and/or selectivity in action imposed by a second population. Third, from a potential future regulatory aspect, a single population of easilycharacterizable particles that are manufactured to predictable physicochemical properties facilitates a target product profile that would be human-specific compared to more complex microparticle chemistry and the requirement for more than one particle population. Finally, only a short course of microsphere treatment (five injections) was required to achieve a long-term diabetes-free state (Figure 4). It is possible that even fewer injections could achieve comparable outcomes, something that is under investigation currently. This suggests that, even if the induced Bregs are the only layer of autoimmunity regulation, their effect is potentially long lasting (e.g., due to a longitudinal expansion and/or a long half-life of these Bregs in vivo). Although this does not rule out the possibility that other regulatory cell changes occur in different organs or time points of the treatment progression.
Recently, an orally-delivered formulation of TGFβ and RA was reported to result in decreased potentially-autoreactive effector T-cells inside the pancreas of mice rendered diabetic by multiple low-dose streptozotocin (MLDS) (106). The reported data suggest suppressed insulitis and a statistically-greater proportion of diabetes-free mice in the RA/TGFβ microparticle treatment arm compared to controls. The conceptual novelty and achievement of an orally-available microparticle formulation is very noteworthy and certainly commendable, however, the findings raise some important questions that need to be addressed. First is the model. All studies, including ours, using micro/nanoparticle approaches to treat T1D, use the NOD/LtJ strain mouse strain, widely regarded as the best in vivo test system for T1D immunotherapeutics, so long as one is aware of the model's limitations (72,(107)(108)(109)(110). The authors, instead used a model (MLDS in C57BL/6 mice) that does not reflect the genetic component as well as the chronic development of an underlying autoimmune state. Second, diabetes incidence was only monitored up to 28 days following the MLDS treatment. In our view, this time frame is insufficient to draw any conclusions about the stability of any efficacy. Third, most of the measured differences in the frequency of leukocyte populations recovered from mice in the different treatment arms in that report are not readily discernible. The differences appear to be quite small. Nevertheless, that report can be considered a first in terms of demonstrating proof-in-concept, at least in terms of oral delivery of a potentially-tolerogenic microparticle system that has a broad impact on treating autoimmunity.
Our findings support the ongoing trajectory to develop TGFβ/RA-based tolerogenic micro/nanoparticle "vaccines" for the treatment of T1D (38-40, 58, 81). The strengths of our system and approach include a simple one-population microparticle with the efficacy to prevent the progression of mid-stage autoimmunity characterized by fulminant insulitis to overt diabetic hyperglycemia achieved by only five s.c. injections of an amount of particles that would translate to a welltolerated human-relevant target product profile. Furthermore, we have achieved long term stability of the normoglycemic state (33 weeks) that may in part be due to the regulation of the progressive autoimmunty via Bregs, at least in the MLN, even though the increase in Bregs we present herein (Figure 6) was an acute response specifically to the Ins-RT-NP. It remains to be determined if Breg frequency in diabetesfree mice remains increased in Ins-RT-NP compared to agematched diabetic control mice (untreated and control test articletreated recipients). The differences in the Treg:Breg ratio we observe in the MLN among the treatment arms at 3 days postadministration of the Ins-RT-NP are intriguing and remain to be mechanistically-understood. It is possible that our particles act by modifying the biology of other T-cells by a number of overlapping and complementary mechanisms of action. For example, DC that take up these particles could modify the numbers and/or action of Tr1 regulatory T-cells and not Foxp3+ Tregs. Or, the surface-bound TGFβ1 on our particles could bind to TGF receptors on DC and /or other regulatory leukocytes to license their suppressive abilities that are expressed only inside secondary lymphoid organs. These possibilities, among a number of others, are currently under investigation so that we can better understand the mechanism of action of our particle system.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.

ETHICS STATEMENT
The animal study was reviewed and approved by AHN IACUC; Allegheny Health Network IACUC.

AUTHOR CONTRIBUTIONS
BP, CE, and YG administered the microparticles into the mice, collected the tissues and organs, conducted the immunophenotyping, conducted the statistical analyses, and wrote the initial version of the manuscript. WM, YW, and WL manufactured the microparticle formulations and characterized those formulations, wrote the sections in the materials and methods part of the manuscript that describe the manufacture and characterization of the particles, and verified the statistical analyses. RL and BP conducted the flow cytometry and analysis of the outcomes and participated in the statistical analysis of those data. MT and NG edited the manuscript from the first draft onwards, discussed the findings and the significance with the co-authors, and verified the statistical outcomes. NG as principal investigator of the study was responsible for the oversight of the study at all levels and ensured that the outcomes were reproducible, that the experiments were rigorously conducted, and that the data presented in this manuscript reflect the raw data obtained faithfully.