Immunomodulatory Dual-Sized Microparticle System Conditions Human Antigen Presenting Cells Into a Tolerogenic Phenotype In Vitro and Inhibits Type 1 Diabetes-Specific Autoreactive T Cell Responses

Current monotherapeutic agents fail to restore tolerance to self-antigens in autoimmune individuals without systemic immunosuppression. We hypothesized that a combinatorial drug formulation delivered by a poly-lactic-co-glycolic acid (PLGA) dual-sized microparticle (dMP) system would facilitate tunable drug delivery to elicit immune tolerance. Specifically, we utilized 30 µm MPs to provide local sustained release of granulocyte-macrophage colony-stimulating factor (GM-CSF) and transforming growth factor β1 (TGF-β1) along with 1 µm MPs to facilitate phagocytic uptake of encapsulated antigen and 1α,25(OH)2 Vitamin D3 (VD3) followed by tolerogenic antigen presentation. We previously demonstrated the dMP system ameliorated type 1 diabetes (T1D) and experimental autoimmune encephalomyelitis (EAE) in murine models. Here, we investigated the system’s capacity to impact human cell activity in vitro to advance clinical translation. dMP treatment directly reduced T cell proliferation and inflammatory cytokine production. dMP delivery to monocytes and monocyte-derived dendritic cells (DCs) increased their expression of surface and intracellular anti-inflammatory mediators. In co-culture, dMP-treated DCs (dMP-DCs) reduced allogeneic T cell receptor (TCR) signaling and proliferation, while increasing PD-1 expression, IL-10 production, and regulatory T cell (Treg) frequency. To model antigen-specific activation and downstream function, we co-cultured TCR-engineered autoreactive T cell “avatars,” with dMP-DCs or control DCs followed by β-cell line (ßlox5) target cells. For G6PC2-specific CD8+ avatars (clone 32), dMP-DC exposure reduced Granzyme B and dampened cytotoxicity. GAD65-reactive CD4+ avatars (clone 4.13) exhibited an anergic/exhausted phenotype with dMP-DC presence. Collectively, these data suggest this dMP formulation conditions human antigen presenting cells toward a tolerogenic phenotype, inducing regulatory and suppressive T cell responses.

Current monotherapeutic agents fail to restore tolerance to self-antigens in autoimmune individuals without systemic immunosuppression. We hypothesized that a combinatorial drug formulation delivered by a poly-lactic-co-glycolic acid (PLGA) dual-sized microparticle (dMP) system would facilitate tunable drug delivery to elicit immune tolerance. Specifically, we utilized 30 µm MPs to provide local sustained release of granulocyte-macrophage colony-stimulating factor (GM-CSF) and transforming growth factor b1 (TGF-b1) along with 1 µm MPs to facilitate phagocytic uptake of encapsulated antigen and 1a,25(OH) 2 Vitamin D 3 (VD3) followed by tolerogenic antigen presentation. We previously demonstrated the dMP system ameliorated type 1 diabetes (T1D) and experimental autoimmune encephalomyelitis (EAE) in murine models. Here, we investigated the system's capacity to impact human cell activity in vitro to advance clinical translation. dMP treatment directly reduced T cell proliferation and inflammatory cytokine production. dMP delivery to monocytes and monocyte-derived dendritic cells (DCs) increased their expression of surface and intracellular anti-inflammatory mediators. In co-culture, dMP-treated DCs (dMP-DCs) reduced allogeneic T cell receptor (TCR) signaling and proliferation, while increasing PD-1 expression, IL-10 production, and regulatory T cell (Treg) frequency. To model antigen-specific activation and downstream function, we co-cultured TCR-engineered autoreactive T cell "avatars," with dMP-DCs or control DCs followed by b-cell line (ßlox5) target cells. For G6PC2-specific CD8 + avatars (clone 32), dMP-DC exposure reduced Granzyme B and dampened cytotoxicity. GAD65reactive CD4 + avatars (clone 4.13) exhibited an anergic/exhausted phenotype with dMP-

INTRODUCTION
Ideal immunotherapy approaches in antigen-specific autoimmune disease must abrogate autoimmunity without the need for broad and sustained systemic immunosuppression. In the case of type 1 diabetes (T1D), insulin is a primary target antigen during disease development and thus, also for tolerance induction to prevent disease onset (1,2). Historically, attempts to tolerize against insulin have demonstrated exceptional safety profiles, yet failed to meet clinical endpoints in major T1D prevention trials (3)(4)(5)(6). In the case of established T1D, therapeutic success will likely hinge on elimination of the autoreactive T cells presumably responsible for the destruction of pancreatic b-cells or external measures to control T cell behavior (7). Indeed, non-antigen-specific strategies targeting T cells have shown success in subjects with or at-risk for T1D, temporarily maintaining C-peptide production or delaying disease onset, but the decline in C-peptide and T1D progression eventually resumes, suggesting treated subjects do not develop lasting tolerance to islet antigens (8)(9)(10)(11)(12)(13)(14). The field has called for the use of combination therapies as a potentially more effective strategy to augment T cell targeted agents (15)(16)(17)(18)(19). To address this, we developed a novel biomaterial therapy to deliver immunomodulatory agents along with autoantigen as a means to recruit and tolerize dendritic cells (DCs) for robust antigen-specific T cell tolerance (20,21). Here, we extensively characterized human immune cell responses in vitro as an important bridge to clinical translation for this novel dual sized microparticle (dMP) formulation.
DC-based therapies have gained interest in both cancer and autoimmunity due to the unique ability of DCs to direct inflammatory or anti-inflammatory T cell effector responses to their presented antigen (22)(23)(24)(25)(26). Many approaches center around the generation of antigen-specific presenting DCs ex vivo; however, high cost, impaired cell migration, and poor survival upon delivery complicate clinical translation (27,28). Thus, we have pursued strategies to direct DC function in vivo. The intrinsic phagocytic capacity of DCs and other antigen presenting cells (APCs), as well as their abundant presence as immune sentinels in the epidermis (29), make them an attractive target for subcutaneously delivered payloads encapsulated in a particulate biomaterial (30,31). Antigen-loaded particles are an area of particular interest in various disease models (28,32,33). Poly-lactic-co-glycolic acid (PLGA) has been well characterized as a delivery vehicle to DCs and is a component in multiple FDA-approved products (e.g., dissolvable sutures) (34). Previous work from our group and others demonstrated that PLGA microparticles (MPs) of approximately 1 mm in diameter are efficiently endocytosed for directed endosomal delivery, while particles 30 mm in diameter, too large to be taken up by APCs, provide controlled local release of encapsulated factors extracellularly to generate a tolerogenic milieu (35,36). Our strategy involves combining a disease-relevant autoantigen with immunomodulatory agents selected for their ability to recruit DCs, create a suppressive APC phenotype, and induce durable antigen-specific T cell tolerance. We previously screened immunomodulatory agents of interest encapsulated in PLGA for their abilities to effect tolerogenic activity by murine bone-marrow derived DCs in mixed lymphocyte reactions. The resultant dMP was comprised of large MPs (30 mm) encapsulating transforming growth factor b1 (TGF-b1) and granulocytemacrophage colony-stimulating factor (GM-CSF) for extracellular conditioning, along with small MPs (1 mm) containing 1a,25(OH) 2 Vitamin D 3 (VD3) and denatured insulin antigen for phagocytic engulfment (20,36).
The tolerogenic capacity of the individual agents has been previously characterized in several settings. TGF-b1 is a potent immunoregulatory cytokine capable of suppressing effector function and cytokine production by both innate and adaptive immune cells (37). TGF-b1 treated DCs demonstrate reduced expression of MHC-II, co-stimulatory molecules, and inflammatory cytokines; increased production of the tolerogenic enzyme, indoleamine 2,3-dioxygenase (IDO) (38,39); increased capacity for the induction of antigen-specific regulatory T cells (Tregs); and deletion of antigen-specific effector T cells (40). Additionally, as a critical mediator in differentiation and development of myeloid DCs, GM-CSF has been shown to promote DC recruitment in multiple disease applications (41), including the non-obese diabetic (NOD) mouse model of T1D where adoptive transfer of GM-CSF exposed DCs promoted the expansion of Foxp3 + Tregs and delayed diabetes onset (42). VD3 is well-known for its ability to steer DCs to a tolerogenic phenotype by inhibiting their maturation and promoting anti-inflammatory cytokine production, thus reducing T cell proliferation and effector response (43)(44)(45)(46)(47). Additionally, deficiencies in vitamin D, its receptor, and binding proteins have been found in multiple autoimmune and autoinflammatory conditions, including T1D, multiple sclerosis, rheumatoid arthritis, systemic lupus erythematosus, and Crohn's disease (48)(49)(50). Thus, we designed our dMP formulation to impact multiple tolerogenic pathways active in innate and adaptive immune subsets for the induction of antigen-specific immune regulation.
This carefully selected combination of tolerogenic agents and disease-relevant autoantigen, delivered via PLGA MP encapsulation for subcutaneous injection, has been tested in two murine models of antigen-specific autoimmunity. This therapy successfully prevented diabetes in NOD mice and reduced disease severity in an early treatment model of experimental autoimmune encephalomyelitis (EAE) (21,51). Often, efficacy in mouse models does not scale to trials in human subjects, highlighting the need for in vitro preclinical assays to test dose-response in target cells, as well as off-target or unexpected effects (52). Hence, we modeled biomaterial therapeutic responses to the immunomodulatory dMP agents in human subjects via in vitro culture and phenotyping of primary human monocytes, monocyte-derived DCs (hereafter referred to as DCs), primary T cells, and autoreactive T cell avatars engineered via T cell receptor (TCR) gene transfer (53) as a step toward supporting clinical translation.

MP Fabrication and Characterization
PLGA MPs were manufactured as previously described (51) with some noted modifications. Briefly, a 50:50 polymer composition of PLGA (molecular weight (MW) 44,000 g/mol; Corbion Purac), was used in a standard water-oil-water double solvent evaporation technique. Emulsions were formed with the aqueous phase comprised of Ultrapure H 2 O (Barnstead GenPure, Thermo Fisher Scientific). Poly-vinyl alcohol (PVA; MW approximately 15,000 g/mol; Fisher Scientific) was used as an emulsion stabilizer. To incorporate the desired protein(s), 100mg PLGA polymer was dissolved in methylene chloride (Fisher Scientific) at 5% w/v ratio. Protein solution containing either TGF-b1, GM-CSF, or VD3) was added to 5% PLGA solution and homogenized to form a primary emulsion. This emulsion was added to 2 mL of 5% PVA solution and homogenized to form the secondary emulsion. After transfer to a beaker containing 30 mL 1% PVA, resultant MPs were agitated using a magnetic stirrer for 4-6 h to evaporate residual methylene chloride. The remaining solution was centrifuged at 10,000xg for 10 min to collect MPs and washed 3x with Ultrapure H 2 O. MPs were then flash-frozen in liquid nitrogen, lyophilized, and stored at −20°C or −80°C until use. MP size distributions were measured using the Microtrac Nanotrac Dynamic Light Scattering Particle Analyser (Microtrac). Loading efficiency in MPs was measured using solvent extraction in DMSO followed by spectrophotometric analysis of protein content (51).

Peripheral Blood Sample Collection and Processing
Following the provision of written informed consent, deidentified blood samples were collected from subjects without autoimmunity by venipuncture into sodium heparin coated Vacutainer tubes (BD) in accordance with University of Florida IRB201400709 and processed for leukocyte subsets via negative selection and Ficoll density gradient separation within 12 h of collection.

Monocyte/Macrophage Culture
Peripheral blood mononuclear cells (PBMC) were isolated by density gradient centrifugation over Ficoll separation medium (GE) using established protocols. For monocyte/macrophage cultures, PBMC were incubated in 24-well tissue culture plates at 5x10 6 /mL in complete RPMI with 10% fetal bovine serum for 24 h, after which nonadherent cells were washed away to leave adherent monocytes. After MP incubation steps, cells were removed from tissue culture plastic via plate incubation on ice for 10 min and gentle scraping to avoid loss of surface marker expression potentially associated with protease treatment.

Monocyte-Derived DC Generation and Culture
Monocytes were isolated from heparinized peripheral blood via negative selection (RosetteSep, StemCell) followed by density gradient centrifugation over Ficoll separation medium (GE). Monocytes were maintained at 10 6 /mL in DMEM (Gibco) with 10% heat-inactivated human serum and 50 ng/mL each of recombinant human GM-CSF and IL-4 (Peprotech) in ultralow attachment plates (Corning) for 7-10 days. Small MPs (containing VD3 or equivalent mass unloaded PLGA, 5 µg) were incubated in wells with harvested DC in ultra-low attachment plates to allow for antigen/PLGA uptake. Large MPs (containing TGF-b1 and GM-CSF or equivalent mass unloaded PLGA) were added at 5 mg per 10 6 DC in 0.4 mM pore size hanging well inserts (Miltenyi) to prevent cell overcrowding, for two days prior to testing for phenotype or stimulus response.

Allogeneic T Cell Response Assay
CD4 + and CD8 + T cells were isolated from peripheral blood by negative selection (RosetteSep, StemCell) followed by density gradient separation. T cells were labeled with fluorescent proliferation dye [CellTrace Violet (CTV), Invitrogen] as per manufacturer's instructions and cocultured with MP treated DC at a 1:1 ratio for up to 7 days to assess proliferation and effector function. In some experiments, memory (CD45RO + ) and naïve (CD45RA + ) T cells were FACS sorted (BD FACSAria) prior to DC coculture. Proliferation was quantified by gating the frequency of dividing cells, or by calculating the proliferation index (average number of divisions of responding cells; calculated in FlowJo as proliferation index = sum ( i * N(i)/2 i )/ sum (N(i)/2 i )where i = division number (undivided = 0) and N(i) = number of events in division i (54).

Lentiviral Transduction and Generation of T Cell Avatars
Isolated CD4 + T cells were transduced with a multicistronic lentiviral TCR clone 4.13 as previously described (53). This clone reacts to GAD 555-567 in the context of HLA-DRB1*04:01 and expresses an eGFP reporter on a pCNFW lentiviral vector backbone, driven by a cytomegalovirus (CMV) promoter. Isolated CD8 + T cells were transduced with a multicistronic lentiviral TCR expression construct encoding TCR clone 32, which recognizes the autoantigen glucose-6-phosphatase catalytic subunit 2 (G6PC2, formerly known as IGRP) in the context of HLA-A*02:01 (55,56). In brief, T cells were resuspended in 1 mL of complete RPMI media (cRPMI) and cultured in a 24-well plate (250,000 cells/well). Cells were bead activated using anti-CD3 and anti-CD28 loaded Dynabeads (Life Technologies) at a 1:1 cell to bead ratio. After 48 h, protamine sulfate (Sigma) was added to a final concentration of 8 mg/mL. Lentivirus stock was added dropwise [multiplicity of infection (MOI) = 3]. Spinoculation was then performed by centrifuging the plate at 1000xg for 30 min at 32°C, followed by addition of IL-2 (200 U/mL) on days 2, 5, and 7.

Supernatant Cytokine Analysis
Plates were centrifuged to remove cell debris, supernatants collected, and stored at -20°C until batch analysis of cytokines via Luminex multiplex bead assay (Millipore) or ELISA (BD OptEIA, BD Biosciences) as per manufacturer's instructions.

Antigen-Specific CD4 + T Cell Avatar Proliferation Assay
Seven days post-transduction, CD4 + T cell avatars (clone 4.13) were magnetically depleted of stimulation beads and flow-sorted (BD FACSAria) into GFP + (de novo TCR + ) and GFP -(mocktransduced) populations prior to equilibration in IL-2 (50 U/mL) and IL-7 (10 ng/mL) for 5 days. Semi-quiescent cells were labeled with CTV proliferation dye for 1:1 co-culture with MP-treated HLA-DR4+ genotype selected (57) DCs (dye-labeled where indicated to ease gating strategies) in the presence of GAD 555-567 peptide. Proliferation and expression of intracellular transcription factors, FOXP3 and Eomes, were assessed via flow cytometry at day 5.

Antigen-Specific CD8 + T Cell Avatar Killing Assay
Previously cryopreserved CD8 + G6PC2-reactive TCR avatars (clone 32) were thawed and rested for 24 h with the homeostatic cytokine IL-7 (10 ng/ml) prior to coculture with MP-treated HLA-A2+ genotype selected (57, 58) DCs for 24 h in tissue culture treated plates at a 1:1 DC : CD8 + T cell ratio. Nonadherent CD8 + T cell avatars were removed from the plate, washed, and resuspended in fresh media for the killing assay. The human b-cell line, blox5 (59,60), was maintained under standard culture conditions, labeled with CTV fluorescent dye, and plated 18 h prior to the killing assay to achieve 80-90% confluency. At time of assay, media was removed and CD8 + T cell avatars were seeded at an effector to blox5 target cell ratio of 5:1. Cell death was assessed at 18 h via flow cytometric analysis of Annexin V (BD Biosciences) and viability dye (Life Technologies) staining via an assay established in the lab (61).

Statistical Analysis
Data were visualized using GraphPad Prism v.8 and analyzed by t tests or one-way ANOVA with multiple comparison testing as indicated in figure legends, with p < 0.05 considered significant.

MP Characterization
PLGA MPs were manufactured in a single large batch to provide standardization across in vitro experiments. MP size distribution, assessed by dynamic light scattering, showed that MPs exhibited the desired size characteristics ( Figure 1A). Release kinetics in aqueous solution were determined at 1, 3, and 7 days to model payload release in vitro ( Figure 1B). Encapsulation efficiencies were quantified via solvent extraction and spectrophotometric analysis, with TGF-b1 at 63.1 ± 6.6%, GM-CSF at 53.2 ± 7.3%, and VD3 at 73.4 ± 8.2% encapsulation efficiency. To control for biomaterial-induced cellular responses, in vitro assays were conducted with identical masses of drug-loaded PLGA MPs versus PLGA-only controls in each size class. These mass values, in combination with the release kinetics, were used to estimate in vitro concentrations of each agent at defined assay time points of 24 and 48 h ( Table 1) and thereby, assess the effects of MP treatment on cellular phenotype and function.

MP Responses in Isolated Cell Subsets
Previous work in the NOD mouse revealed a diverse immune cell composition surrounding the subcutaneous MP injection site, with neutrophils predominating in immediate response to PLGA material, followed by accumulation of the desired target of myeloid cells, as well as a significant number of T cells in response to drug-loaded dMP (20,21). We therefore sought to investigate direct effects of dMPs on human cells belonging to those subsets specifically attracted by the dMP over PLGAnamely, monocytes/macrophages, DCs, and T cells-to assess therapeutic modulation.

Monocytes
The capacity of dMPs to induce phenotypic alterations in isolated human monocytes was investigated ( Figure 2). Monocytes were incubated for 48 h under unstimulated conditions with media alone (UN) or both large and small empty MPs (PLGA) versus with MPs encapsulating dMP agents GM-CSF, VD3, and TGF-b1(dMP)) in the presence or absence of LPS for an additional 24 h of culture time (72 hr total). Tolerogenic markers of interest were assessed by flow cytometry of live CD14 + monocytes (representative gating in Figure 2A). dMP treated monocytes exhibited a significant increase (approximately 30%) in frequency of cells expressing the scavenger receptor CD163, a marker associated with antiinflammatory M2 macrophages (62) ( Figure 2B, dMP vs. UN and PLGA conditions, p < 0.01). Expression of the suppressive tryptophan-catabolizing enzyme, IDO, increased in response to dMP in the absence of LPS induction ( Figure 2C, UN vs. dMP, p < 0.05), whereas PD-L1 was robustly induced in dMP treated cells above PLGA effects in in the presence of LPS stimulus ( Figure 2D, dMP vs. UN/PLGA, p < 0.001). As expected, monocytes cultured with dMPs exhibited enhanced surface and intracellular expression of anti-inflammatory mediators.

T Cells
Total CD3 + T cells were isolated from healthy control subjects by negative selection, proliferation dye-labeled, and treated for 48 h with single-component MP-encapsulated agents (GM-CSF, VD3, or TGF-b1) or the complete dMP formulation (GM-CSF, VD3, and TGF-b1) as compared to UN and PLGA controls. T cells were subsequently harvested from MPs via densitygradient centrifugation and cultured with plate-bound anti-CD3/anti-CD28 TCR stimulus for five days, whereupon cell proliferation and phenotype of both CD4 + and CD8 + T cells were assessed by flow cytometry (Figure 4). potentially underlying the observed reduced proliferation revealed that TGF-b1 MPs increased the frequency of PD1 + CD4 + T cells compared to UN and VD3 MP treated T cells (p < 0.01), although PD1 + CD8 + T cell frequencies were not significantly altered ( Figure  4C). These findings demonstrate the interactions of individual MP components in the complete dMP in direct suppression of T cell activation and effector responses. Moreover, TGF-b1 MPs bolstered the frequency of Tregs among total CD4 + T cells ( Figure 4D) as compared to UN (p < 0.0001), VD3 MP (p < 0.01), and dMP treated T cells (p < 0.01).

TCR Activation With dMP-DC
With the observation of a tolerogenic phenotype in dMP-DCs, we next sought to confirm functional tolerogenic capacity via allogeneic human T cell co-culture. DCs (n = 3 donors) were proliferation dye labeled, then left untreated (UN-DCs), or conditioned with empty MPs (PLGA-DCs) or dMP (dMP-DCs) for 48 h, followed by MP removal via density gradient centrifugation. DCs from each donor were next co-cultured with proliferation dye-labeled T cells from two allogeneic donors (total experimental n = 6) at a 1:1 DC:T cell ratio. Cocultures (experimental n = 6) were incubated in media alone or with soluble CD3/CD28 TCR stimulus for T cell activation for 24 h, and examined by flow cytometry for both DC phenotype and T cell activation ( Figure 5). dMP-DC response to T cell activation mirrored that seen with LPS stimulation (Figure 3). Namely, reduced HLA-DR expression was observed in response to dMP ( Figure 5A, black bars) as compared to both UN-DCs and PLGA-DCs (p < 0.0001). PD-L1 expression was concomitantly increased in dMP-DCs as compared to UN-DCs (p < 0.01) and PLGA-DCs ( Figure 5B, p < 0.05). Accordingly, both CD4 + and CD8 + T cells had significantly reduced frequency of cells expressing the TCR-activation marker, CD69, at 24 h post-stimulation in the presence of dMP-DCs ( Figure 5C) compared to PLGA-DCs (p < 0.01) or UN-DCs (p < 0.01). CD4 T cell CD69 MFI was also significantly reduced with dMP-DC compared to UN-DC ( Figure 5D, p < 0.01). These findings support the notion that dMP treatment blunts immediate TCR activation, giving the potential to influence downstream effector responses.

T Cell Phenotype and Function Post-Activation With dMP-DCs
Downstream CD4 + effector T cell response to dMP-DCs was determined via co-culture of proliferation dye-labeled T cells with VD3-TGF-, PLGA-, or UN-pretreated allogeneic DCs at a DC:T cell ratio of 1:5 for 7 days. Proliferation and cellular phenotype were assessed by flow cytometry (Figure 6). dMP-DCs did not support alloantigen-induced proliferation of both CD45RO + memory T cells ( Figure 6A, dMP vs. UN and PLGA p < 0.01) and CD45RA + naïve T cells at day 5 ( Figure 6B  DCs (p < 0.001) ( Figure 6C). Both FOXP3 + Helios + Treg frequency among CD4 + T cells ( Figure 6D, dMP vs. UN/ PLGA, p < 0.05) and PD1 expression on CD4 + T cells were increased in response to dMP-DCs at day 7 ( Figure 6E, dMP vs. UN and PLGA, p < 0.0001). Despite reduced proliferation and therefore, lower total T cell number in all dMP-DC co-cultures, IL-10 production in culture supernatant was increased in cultures assessed at day 5 compared to PLGA-or UN-DC cocultures ( Figure 6F, dMP vs. UN, p < 0.05; dMP vs. PLGA, p < 0.01). Multiplex measurement of supernatant cytokines on day 7 ( Figure 6G) revealed a trend toward decreased proinflammatory cytokines (IL-6, TNFa, IFNg, GM-CSF, and IL7), a significant decrease in IL-8 (p < 0.01) and IL-12p70 (p < 0.05) as compared to UN-DC cocultures, and a significant decrease in IL-5 (p < 0.001) from both UN-and PLGA-DC cocultures. This decrease in T H 1 and proinflammatory cytokines, with a concomitant significant increase in IL-4 compared to both control conditions (p < 0.001) and maintenance of IL-10 levels compared to untreated cultures, is suggestive of skewing to a T H 2 phenotype. These alterations in T cell skewing were specific to the dMP, as PLGA-DC stimulus induced no significant changes in cytokine levels as compared to UN-DC. Given the reduced T cell proliferation, cytokine profiles were generally suppressed in presence of dMP, although production of IL-2 and IL-4 were enhanced in these cultures (p < 0.05). IL-10 production was elevated in response to dMP-DCs on day 5 when measured via ELISA ( Figure 6F, p < 0.05), yet dMP-DC induced IL-10 was near UN-DC levels on day 7 with multiplex assay ( Figure 6G), consistent with declining IL-10 levels over culture time for in vitro T cell stimulations (67). (F) IL10 production in supernatant at day 5 was elevated (n = 9) with dMP-DCs. (G) A subset of T cell proliferation assays (displayed in (C)) were assessed for cytokines by multiplex bead assay and showed a general trend for decrease in cytokine production with T cell-dMP DC coculture (scaled for display and comparison by log transformation of raw values). % proliferation was included as a scale reference. Significance compared to dMP condition is indicated for individual outcome measures. (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 by ANOVA with Tukey's multiple comparison testing). Modeling of Antigen-Specific Effector T Cell Response to dMP CD4 + and CD8 + T cell avatars expressing TCRs reactive to the T1D autoantigens GAD (clone 4.13) and G6PC2 (clone 32), respectively, were generated from naïve T cells using lentiviral vectors as previously described (53). As TCR transduction requires T cell preactivation and expansion, sorted T cell avatars were equilibrated in homeostatic cytokines IL-2 and IL-7 for two days. This was followed by incubation for 24 h with nonadherent lymphocytedepleted APCs from HLA-compatible donors, pretreated with PLGA or dMPs for 48 h ( Figure 7A). CD8-32 avatars were assessed for phenotype and CTL function in killing assays using the lox5 human pancreatic b-cell line as target cells (55). dMP-APC conditioned CD8-32 cells were moderately suppressed in killinglox5 cells ( Figure 7B, p < 0.05) and accordingly, exhibited significantly decreased expression of the serine protease Granzyme B in response to dMP-APCs versus PLGA-APCs, indicative of reduced cytotoxic potential ( Figure 7C, 30% decrease in gMFI, p < 0.01). While untransduced GFP -CD4 + T cells (mock transduced) showed somewhat reduced proliferation in response to dMP-APC with allogeneic stimulus, CD4-GAD4.13 avatars showed dramatically reduced proliferation in response to dMP-APCs with exogenous GAD 555-567 peptide stimulus (GFP + , 75% decrease in proliferating cells; p < 0.0001, Figure 7D). Modest increases in the frequency of FOXP3 + cells and FOXP3 gMFI were observed in dMP-APC activated CD4 + T cell avatars following exposure to autoantigen, whereas dMP-APC activation significantly increased frequency (p < 0.01) and expression intensity (4.5-fold increase, p < 0.0001) of the exhaustionassociated transcription factor Eomes ( Figures 7E, F) (68,69). Altogether, these data indicate dMP-APCs are capable of strongly directing T1D antigen-specific CD4 + T cell responses, with more modest control over CD8 + T cells.

DISCUSSION
In development and testing this dMP platform, we implemented a sequential iterative strategy of in vitro screening with formulation testing in mouse immune cells and validation of in vivo efficacy in two murine models of autoimmunity (21,51), followed by both replicative in vitro experiments and discovery efforts in human cells as reported herein. We validated target immune effects using standard immune function assays, then modeled autoimmune responses to evaluate indirect immunomodulation of engineered autoreactive T cells exposed to dMP-APCs. These data support the clinical translatability of this dMP formulation, as therapeutic goals defined and met in animal models of autoimmunity (21,51) were recapitulated in vitro in human cells.
Notably, dMP treatment of both monocytes and monocytederived DCs robustly induced a suppressive cell phenotype and dampened responses to inflammatory stimulus. Human dMP-DC results were analogous to those seen in murine bone marrow derived DCs (20,36), with marked resistance to LPS-induced maturation after brief incubations with soluble GM-CSF plus MP encapsulated TGF-b1 and VD3. dMP-DCs displayed a  comprehensive suppressive phenotype, with high expression of multiple immunoregulatory markers (PD-L1, ILT3, Galectin 9) and demonstrated ability to restrain T cell responses. Expression of ILT3, induced on DCs in vitro by IL-10 in combination with IFN or VD3, influences unprimed T cells to become suppressive (70). Induction of ILT3 and PD-L1 on DCs has been shown with individual agents in the dMP formulation, namely VD3 (45,47,71,72). PD-L1 + DCs are known to restrain T cell activation, differentiation, and proliferation during priming (73,74). Administration of Galectin 9 at supraphysiological levels has shown promise in suppressing T H 1 inflammation, (66), promoting Treg (75), and preventing T1D development in the NOD mouse via islet expression (76). Additionally, Galectin 9 expression on DCs assists their migration toward chemokines (77), thus suggesting that dMP treatment could enhance DC migratory capacity from the injection site to draining lymph nodes. Importantly, dMP-DCs maintained their desirable tolerogenic surface marker profile following both LPS stimulus and T cell crosstalk post-anti-CD3/CD28 activation. T cells responding to dMP-DCs exhibited reduced immediate TCR activation and downstream proinflammatory responses. Notably, these data indicate that dMP exposure influenced T cell skewing, with a major shift toward a T H 2 phenotype characterized by high levels of IL-4 production. The skewing of cytokine production after exposure to dMP-DCs versus PLGA-DCs or UN-DCs is consistent with lower effective TCR signal strength (78, 79), biasing T cell differentiation toward a T FH /T H2 phenotype. This work included assessment of dMP influence on human monocytes/macrophages, which were abundantly present at the MP injection site in vivo in the NOD mouse (21) and could provide additional local immunoregulation beyond the tolerogenic DC and cytokine milieu promoted by dMP. The potential for dMPmonocytes to induce robust T cell suppression was demonstrated, whereas HLA-matched donor monocytes were not differentiated to DCs prior to incubation with dMP and subsequent T cell co-culture. In addition to DCs, monocytes and macrophages could provide additional antigen-specific immunoregulation of T cells encountered (80) and deletion of autoreactive T cells in peripheral and inflamed sites through the observed increased IDO expression (81). Indeed, M2 macrophages have been shown to promote tolerance of transplant tissue and suppression of xenoimmune response via IDO expression and production of TGF-b1 (82,83). Additionally, human monocytes have a unique ability to activate latent TGF-b1 on surrounding cells, providing a potential feedforward loop for this potent suppressive mediator (84).
We measured T cell responses following direct culture with the dMP formulation. In agreement with the prior literature, TGF-b1 MPs alone drove induction of FOXP3 and PD-1 (85), whereas only the full dMP provided both significant reduction in TCR-induced cellular proliferation and inflammatory cytokine production. Our data indicate that in addition to indirect suppression by interactions with dMP-tolerized DCs in lymph nodes, T cell phenotype and activity could be directly influenced at the injection site through mechanisms of infectious tolerance demonstrated in TGF-b1 producing T H 3 peripheral Tregs (86). Notably, while the specific dMP formulation is tunable for disease-specific applications, the combination of agents in the dMP tested here synergized to influence T cell effector function. This may provide additional benefit in the case of antigenspecific T cell localization to a peripheral antigen depot, as recently reported in an antigen-defined T1D mouse model (87).
Finally, we assessed the impact of the dMP on antigen-specific T cell responses to two b-cell autoantigens in T1D. We utilized our established platform for generating T cell avatars by multi-cistronic TCR expression to assess dMP-mediated suppression of islet antigenspecific T cell activity (53). With this platform, we previously demonstrated that autoreactive T cells may play a direct role in bcell killing, as G6P2C-specific CD8 + cytotoxic T lymphocyte avatars are able to directly lyse both b-like cell lines and primary human bcells (55). Moreover, our studies of human tissues from the Network for Pancreatic Organ donors with Diabetes (nPOD) have demonstrated that clone 4.13, originally identified in T1D PBMC (88), can be found within the islets of a human T1D pancreas (89). Hence, this system represents an excellent platform for preclinical translation. dMP-DCs were able to potently suppress the proliferation and drive anergy in CD4 + GAD4.13 T cell avatars, and though dMP-DCs reduced Granzyme B expression by G6PC2 avatars, they only moderately reduced cytotoxic effects. These results suggest that our dMP formulation, despite decreasing effector molecule expression and inducing markers of T cell exhaustion, is not capable of fully reversing the cytotoxic activity of an activated CTL with high affinity TCR. For optimal reversal of T1D in human subjects, these results suggest potential opportunity for pretreatment with a depleting agent, such as anti-CD3 (teplizumab) or anti-thymocyte globulin (ATG, thymoglobulin) prior to reconditioning the host with this tolerogenic dMP vaccine (90).
The experiments conducted herein further support the development of antigen-specific tolerogenic vaccine strategies that leverage the many benefits of controlled release from biocompatible MPs. Experimental conditions described herein were meant to model the effect on cells encountering the persistent injection site microparticle depot induced by dMP injection, wherein the local concentration of agents released by microparticles is high, followed by trafficking to draining LN to present ingested antigen.
Here, we expand the scope of the dMP from preventing and reversing T1D in the NOD mouse (20,21) and ameliorating EAE in a mouse model of multiple sclerosis (51) to illustrate efficacy in human cells, notably using autoreactive T cell clones specific for two T1D-associated antigens. Importantly, the dMP system is modular in nature and could be adapted to include additional antigens and epitopes that continue to emerge as targets of autoreactive T cells in T1D (91,92). In summary, the use of bioengineering approaches, including the dMP formulation reported here, offers the capacity to safely deliver effective vaccines to intervene in the disease process of chronic autoimmune conditions.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by University of Florida Institutional Review Board. Written informed consent to participate in this study was provided by the participant or in the case of minors, the participants' legal guardian/next of kin.

AUTHOR CONTRIBUTIONS
MAB conceived the study, designed and performed the experiments, analyzed the data, and wrote the manuscript. JMS designed the experiments, manufactured the microparticles, and wrote the manuscript. ALP wrote the manuscript. CHW, MAA, and TMB contributed intellectually and wrote the manuscript. BGK conceived and directed the study, and wrote the manuscript. All authors contributed to the article and approved the submitted version. As guarantor of this work, BGK assumes full responsibility for the ethical collection, analysis, and reporting of the data and the decision to publish.

FUNDING
This work was supported by the National Institutes of Health (R01 DK091658, R01 DK098589, R01 DE027301, and R01 AI133623 to BGK; T32 DK108736 to JMS; and P01 AI042288 to MAA and TMB), the University of Florida Clinical and Translational Science Institute Network Science Pilot award, and the McJunkin Family Charitable Foundation, Inc.