Taming Autoimmunity: Alpha-1 Antitrypsin Overexpressing Mesenchymal Stem/Stromal Cells Promote Regulatory T Cell Crosstalk to Reverse Diabetes

Abstract

Mesenchymal stem/stromal cell (MSC) therapy holds promise as a therapeutic option in diabetes treatment. The anti-inflammatory and immunomodulatory activities are enhanced when MSCs are engineered to overexpress alpha-1 antitrypsin (AAT-MSCs). Because a single infusion of AAT-MSCs reversed new-onset diabetes in over 50% of the female nonobese diabetic (NOD) mice, we used singlecell RNA sequencing, flow cytometry, and functional analyses to characterize how AAT-MSCs modulate CD4⁺ and CD8⁺ T cells in pancreatic lymph nodes (PLNs) and islets. AAT-MSC treatment increased T regulatory cells (Tregs) and more effectively suppressed T cell proliferation when stimulated with anti-CD3/CD28 antibodies. Treated mice exhibited reduced T helper 1 (Th1) cells and CD8⁺ cytotoxic T cells. In vitro studies confirmed the capacity of AAT-MSCs to promote Treg expansion in both mouse and human cells, drive CD8⁺ T cells toward an exhausted phenotype, and enhance mouse and human islet cell survival. Cellchat analysis showed that AAT-MSC therapy strengthens intercellular communication, especially signals originating from Tregs toward other PLN and islet cell populations. These findings clarify how AAT-MSCs modulate immune response and support their potential clinical application for type 1 diabetes and other autoimmune or inflammatory conditions.

Graphical Abstract

INTRODUCTION:

Type 1 diabetes (T1D) is characterized by autoimmune-mediated destruction of insulin-secreting pancreatic β cells that leads to hyperglycemia and lifelong insulin therapy. Currently, limited approaches are available to delay the onset or achieve a cure for T1D ¹. While insulin replacement therapy can effectively control glucose levels, it cannot restore pancreatic function and is associated with risks such as hypoglycemia and infection ². The ideal therapeutic approach for T1D would effectively restore immune system homeostasis and protect pancreatic β cells from further destruction. Currently, no single intervention can provide both benefits. Thus, there is a critical need to develop more effective strategies or technologies for treating T1D.

Mesenchymal stem/stromal cells (MSCs) are adult stem cells that can be isolated from various body tissues. They are increasingly recognized as a promising source for cell therapy due to their immunomodulatory, tissue repair, and regenerative functions ³. MSCs exert protective effects by secreting pro-mitotic, anti-apoptotic, anti-inflammatory, and immunomodulatory factors while mitigating metabolomic and oxidative stress imbalances and restoring homeostasis. The immunomodulatory function of MSCs has been observed in many disease models, including encephalomyelitis ⁴, T1D ⁵, multiple sclerosis ^6,7, graft-versus-host disease (GvHD) ⁸, and arthritis ⁹. The immunosuppressive properties are partly mediated by production of nitric oxide, indoleamine 2,3 dioxygenase (IDO), and transforming growth factor β (TGF-β), with complex interactions in other immune pathways ^10–14. Data from various studies, including ours, show that systemic infusion of MSCs in murine models of diabetes improved glycemic control, reduced pancreatic insulitis, prevented autoimmune destruction, and promoted pancreatic tissue repair ^15–18. Importantly, intravenously infused MSCs migrate to the injured pancreas or pancreatic islets in T1D mouse models ^19,20. Two recent clinical trials have demonstrated the preliminary safety and efficacy of MSCs in treating new-onset T1D in patients ^21,22. The U.S. Food and Drug Administration (FDA) has recently approved Remestemcel-L (Ryoncil, Mesoblast, Inc.), an allogeneic bone marrow-derived MSC therapy, for treating steroid-refractory GvHD in pediatric patients ²³.

In addition to their therapeutic effects, MSCs are also one of the most promising stem cell populations for use in gene therapy studies and trials, as they can be modified with a wide range of both viral and non-viral vector systems to produce therapeutic proteins, which further enhance their natural abilities to mediate repair within various tissues ²⁴. Alpha-1 antitrypsin (AAT) is an acute-phase reactant and serine protease inhibitor that suppresses multiple enzymes, including neutrophil elastase, cathepsin G, and others ²⁵. It also exerts anti-inflammatory and anti-apoptotic effects by suppressing cytokine production, complement activation, and immune cell infiltration ²⁶ while protecting pancreatic β cells from apoptosis ^27–30. In non-obese diabetic (NOD) mice, a single injection of AAT reduced the intensity of insulitis, increased β cell mass, promoted β cell regeneration, and prevented the onset of diabetes via modulating Tregs ^29,31. AAT has been shown to protect mouse and non-human primate islet grafts from failure by reducing islet cell death and promoting graft revascularization ^{30,32,33,25,34}. In our previous studies, human AAT-overexpressing MSCs (AAT-MSCs) demonstrated enhanced innate properties, including increased proliferative capacity and accelerated migration ³⁵. The infusion of AAT-MSCs showed better efficacy in preventing the onset of T1D in NOD mice than control MSCs ^35,36 , and provided significantly better protection against GvHD³⁷. This study aims to decipher the mechanistic basis of AAT-MSC action by examining their impact on immune cells, particularly CD4⁺ Tregs and exhausted CD8⁺ T cells, at single-cell resolution in both in vitro and in vivo models, providing mechanistic insights for potential clinical applications.

RESULTS:

AAT-MSC infusion reverses diabetes in NOD mice with newly onset T1D

We first evaluated the therapeutic efficacy of AAT-MSCs in reversing diabetes by administering a single dose of AAT-MSCs to female NOD mice with new onset T1D (Fig. 1a). Untreated control mice (n=11) exhibited a progressive rise in blood glucose levels, whereas AAT-MSC-treated mice (n=18) maintained significantly lower average blood glucose levels. By week 5 post-treatment, 11 out of 18 (61.1%) AAT-MSC-treated mice were diabetes-free, compared to none of the control mice (n=11). Five AAT-MSC-treated mice were sacrificed at week 5 for sample analysis. At week 10, 7 out of the remaining 13 (54%) AAT-MSC-treated mice continued to be diabetes free, with no mice in the control group achieving diabetes remission (n=11, Fig. 1c). Histological analysis (H&E staining) revealed a substantial reduction in immune cell infiltration in pancreatic islets of AAT-MSC-treated mice. Control mice exhibited only 28.2 ± 3.6% of islets with <5% immune cell infiltration, and 50.6 ± 4.3% with >75% infiltration. In contrast, AAT-MSC-treated mice displayed 51.1 ± 4.3 % and 26.2 ± 4.2%, respectively (Fig. 1, d–e). Among immune cell subsets, CD4⁺CD25⁺FoxP3⁺ Tregs play a pivotal role in regulating autoimmune responses, yet their numbers and function are often diminished in T1D ^38–41. AAT-MSC treatment increased both the abundance of CD4⁺CD25⁺FoxP3⁺ Tregs and the proportion of CD4⁺CD25⁺CTLA-4⁺ Tregs relative to controls (Fig. 1, f –i). Furthermore, PLNs from AAT-MSC-treated mice produced significantly lower levels of pro-inflammatory cytokines, including IFN-γ, IL-6, TNF-α, and IL-1β (Fig. 1, j–m). Collectively, these findings demonstrated that AAT-MSC infusion reverses hyperglycemia in new onset diabetic mice, likely by reducing immune cell infiltration to the islets, enhancing Tregs, and diminishing T cell activation within the PLNs.

Fig.1. AAT-MSC treatment reverses diabetes in mice with newly onset T1D.

a. Schematic experimental design. (b) Non-fasting blood glucose levels of new-onset T1D mice without treatment (CTR, n=11) or treated with AAT-MSC cells (1 × 10^⁶/mouse, i.v. infusion), once blood glucose level consistently exceeds 250 mg/dL (AAT-MSC, n=18). Among them, 5 mice from the AAT-MSC group were sacrificed at week 5 for analysis. c. Incidence of mice with T1D. d. Analysis of Treg population in PLNs of T1D mice 4 weeks post-treatment. Scatter plots (d, f) and the percentage of CD4⁺CD25⁺FOXP3⁺ cells and CD4⁺CD25⁺CTLA4⁺ cells in CD4⁺ (e, g) were significantly higher in AAT-MSC-treated mice. d-g: n=3 in CTR and n=5 in AAT-MSC. Each dot represents one mouse. h&i. Evaluation of insulitis in pancreatic tissue sections from control or AAT-MSC-treated T1D mice, stained with H.E. Individual histology scores were determined from 588 islets in the control group and 493 islets in the AAT-MSC-treated group. J-m. Measurement of IL-6, IFN-γ, TNF-α, IL-1β anti-CD3 stimulated PLN cells of CTR or AAT-treated mice. Data are presented as mean ± SEM. *P < 0.05, **P < 0.01, determined by unpaired two-tailed t-test with F test to compare variances.

ScRNA-seq analysis precisely identified the immune cell types affected by AAT-MSC infusion.

The activation of immune cells in T1D begins within the islets and is amplified in the PLNs, which serve as a hub for aberrant immune responses ⁴². To assess the impact of AAT-MSCs on the immune regulation of NOD mice, we injected AAT-MSCs into NOD mice at 8 weeks of age and analyzed immune cell profiles from PLNs and the islets 3 weeks later using scRNA-seq analysis (Fig. 2a). In the PLNs, a total of 8,878 cells from the control group and 8,144 cells from the AAT-MSC group were analyzed. Eight distinct cell clusters were identified based on an upregulated gene expression signature and the expression of defining marker genes. These include B cells-1 and 2 (determined by Ms4a1 expression), CD8⁺ T cells (Cd8a⁺), CD4⁺ T cells (Cd4⁺), CD4⁺ Treg (Cd4⁺ and FoxP3⁺), CD4⁺ interferon-stimulated gene-expressing immune cells (ISG-EIC, Cd4⁺ and Isg15⁺), NK cells (Ccl5⁺), and macrophages (Tyrobp⁺) (Fig. 2b–d). Portions of cell populations from control or AAT-MSC groups exhibited differences in each subcellar type (Fig. 2.e).

Fig. 2. Characterization of different cell types in PLNs and islets from CTR or AAT-MSC-treated mice using ScRNA-seq analysis.

a. Schematic of the experimental design for ScRNA-seq. b. t-SNE projection and graph-based clustering of PLN samples from each group. c. Gene set enrichment analysis (GSEA) summary of gene signatures for each population in PLN cells. d. Expression of canonical cell markers across PLN clusters. e. Portion of cells from AAT-MSC or control mice in each cell subtype in PLN cells. f. t-SNE projection and graph-based clustering of pooled islet samples from CTR or AAT-MSC-treated NOD mice. g: GSEA summary of gene signature for each population in islet cells. h. Expression of canonical cell markers across islet cell clusters. i. Distribution of cell subtypes in islets from AAT-MSC-treated or control mice.

In the islets, 6,814 cells were analyzed in the control group and 6,422 in the AAT-MSC-treated group. ScRNA-seq identified 17 clusters that highly express markers for immune cells and other pancreatic cell types, including B lymphocytes (Ms4a1⁺), CD4⁺ T cells (Cd4⁺), Treg (Cd4⁺Foxp3⁺), CD8⁺ T cells (Cd8a⁺), natural killer cells (Klrb1c⁺), acinar cells (Cpa1⁺), pancreatic alpha cells (Gcg⁺), β cells (Ins1⁺, Ins2⁺), delta cells (Sst⁺), ductal cells (Krt19⁺), and other cell types (Fig 2. f–h). Differences in cell portions between the groups were also observed (Fig. 2i).

3. Characterization of CD4⁺ T cells in the PLN and the islet

Next, we characterized CD4⁺ T cells, key players in the onset and maintenance of T1D, in the PLNs of AAT-MSC-treated and control mice. Using scRNA-seq, we identified six CD4⁺ T cell clusters: CD4 naïve 1 and 2 (Lef1⁺), Treg (Foxp3⁺), T effector (Cd69⁺), IFN-responsive cells (Stat 1⁺), and ISG-EIC cells (Isg15⁺) in the PLNs (Fig. 3a–b). Notably, portions of cell populations from control or AAT-MSC groups exhibited the most dramatic difference in IFN-responsive cells (20% in AAT-MSC-treated vs. 80% in control, Fig. 3c).

Fig. 3. Impact of AAT-MSC treatment on CD4⁺ T cells in the PLN.

a. t-SNE projection plots depicting CD4⁺ T cell subpopulation in the PLNs of AAT-MSC-treated and CTR mice. b. GSEA summary of gene signature associated with CD4⁺ T cell subset. c. Distribution of CD4⁺ T subtypes in the PLNs of CTR and AAT-MSC-treated mice. d &e. Differentially expressed genes in total CD4⁺ T cells (d) and Tregs (e) from AAT-MSC-treated or CTR mice. f. Gene ontology analysis of Treg biological processes in the PLNs. The frequencies of IFN-γ⁺ and IL-17⁺ in CD4⁺ T cells (g-i), CD4⁺CD25⁺ Foxp3⁺Helios⁺ (j&k), and CD4⁺CD25⁺CTLA4⁺ (l&m) were measured and quantified by flow cytometry. N=4 mice per group. *p < 0.05, **p < 0.01. Student’s t-test. n. Schematic diagram of the T cell immunosuppressive assay. Histogram (o) and percentage of proliferation (p) of mouse T cells cultured with or without anti-CD3/CD28 antibodies in the presence or absence of Tregs from control or AAT-MSC-treated mice (n=6 mice per group). CD3/CD28: anti-CD3 and anti-CD28 antibodies. *P < 0.05; **P < 0.01, ***P < 0.001analyzed by Student’s t-test.

The top upregulated genes in total CD4⁺ T cells from the AAT-MSC group compared to controls were Wdr89, Rpl35a, Tcf4, Rpl13a, and Rpl36a. In contrast, the most downregulated genes were Dnajb1, Ms4a4b, Hsp90aa1, Cotl1, and Ddit4 (Fig. 3d). Among the upregulated genes, T cell factor 4 (TCF4) plays a significant role in regulating immune responses within T cells. Dysregulation of TCF4 has been linked to autoimmune disease, particularly by disrupting the balance between pro-inflammatory and regulatory T cells and limiting the differentiation of pathological effector T cells⁴³. RPL13a, a component of the gamma interferon-activated inhibitor of translation (GAIT) complex, mediates IFN-γ-induced immune responses and suppresses inflammation. Among the downregulated genes, DNAJB1⁺ T cells play crucial roles in the development of autoimmune responses. High levels of DNAJB1 expression on T cells can be associated with increased T cell activation and proliferation, potentially enhancing the destructive immune response against pancreatic β cells in T1D ⁴⁴. MS4A4B, a membrane adapter protein, is related to antigen response, as naïve T cells transduced with MS4A4B can respond to lower antigen levels ⁴⁵. COTL1 modulates actin dynamics at the T cell immune synapse, which affects T cell motility and lamellipodial protrusion ⁴⁶ (Table S1). These transcriptional changes suggest that AAT-MSCs regulate the CD4⁺ T cell transcriptome by promoting genes that suppress inflammatory and autoimmune responses while inhibiting genes that enhance immune activation.

A further comparison of the gene expression profiles of the two experimental groups revealed that the top upregulated genes in the IFN-responsive cells are Rps10, Tcf4, Uba52, Bcl2, and Ptpn20 (Figure S1a and Table S1). Among these, RPS10 is an IFN-responsive cellular component that acts as a negative regulator of the interferon response ⁴⁷. While Bcl-2 actively prevents cell death, it can trigger programmed cell death in activated CD4⁺ T cells in the presence of IFN-γ ⁴⁸. PTPN20 is associated with immune cell infiltration and tumor mutation burden in gastric cancer patients ⁴⁹. Conversely, the downregulated genes were Hsp90aa1 and Hsp90ab1, among others (Figure S1a).

The frequency of Tregs was comparable between the AAT-MSC and control groups. Related to gene expression change in the Treg subpopulation, the top upregulated genes in the AAT-MSC group relative to control were Uba52, Rps10, Zbtb20, and Ctla4 (Fig. 3e and Table S1). UBA52 and Rps10 are ubiquitin-ribosomal fusion proteins that play several essential roles in cellular function ⁵⁰. ZBTB20 plays a crucial role in Tregs, identifying and controlling a subpopulation that originates in the thymus and is critical for maintaining intestinal homeostasis ⁵¹. ZBTB20-expressing Tregs have distinct phenotypic and genetic characteristics from non-ZBTB20 Tregs by constitutively expressing IL-10 mRNA and producing high levels of IL-10 upon primary activation. They also express high levels of CD44, TIGIT, GITR, and ICOS compared with non-ZBTB20 Tregs and exert disease-protective effects ⁵¹. CTLA4 is critical for Treg function ^52,53. Top downregulated genes in Tregs include Rbm3, Dnaja1, P4ha1, Irf1, and CD69 (Fig. 3e). IRF1 has been shown to negatively regulate Tregs, as IRF1-deficient mice showed increased numbers of Tregs, and the absence of IRF1 leads to enhanced Treg development and function and has implications in autoimmune and inflammatory conditions ⁵⁴. CD69 regulates Treg differentiation and the secretion of IFN-γ, IL-17, and IL-22. It is also a common marker of precursor and mature resident memory T cells (TRMs) localized in peripheral tissues. Functional enrichment analysis of the most changed genes in Tregs revealed significant involvement in pathways related to the regulation of T cell activation, positive regulation of lymphocyte activation, cell-cell adhesion, and regulation of T cell differentiation (Fig. 3f and Figure S1b&c). Therefore, treatment with AAT-MSC regulates the expression of key genes that potentially improve Treg function.

We also characterized CD4⁺ T cells in islets harvested from AAT-MSC-treated and control mice. Three major types of CD4⁺ T cells were observed. These included naïve (Sell⁺), regulatory (Lag3⁺), and memory (Itgb1⁺) CD4⁺ T cells (Figure S2a and S2b). The AAT-MSCs group exhibited more Lag3⁺ Tregs and Itgb1⁺ memory CD4⁺ T cells than controls (Figure S2c). The top upregulated genes in the islet CD4⁺ Treg cells from the AAT-MSC group were Rpl38, Rps27, Rps21, Rpl37a, and Rps28 (Figure S2d). Among them, RPS27 encodes a protein that regulates genes involved in the immune response and inflammation ⁵⁵. RPS21 expression is related to immune checkpoints like CTLA4, LAG3, PDCD1, and TIGIT; RPS28 produces degraded nascent polypeptides (DRiPs) on MHC class I molecules, which are critical for T cell recognition ⁵⁶. The most downregulated genes include Cmss1, Cdk8, Rps6, Ppy, and Ins2 (Figure S3d). Notably, CDK8 typically promotes the differentiation of anti-inflammatory T regulatory cells (Tregs) while inhibiting the differentiation of Th1 and Th17 cells. RPS6 plays a crucial role in T cell development, as its deletion in mouse double-positive thymocytes completely blocks T cell maturation ⁵⁷.

Treatment with AAT-MSCs suppressed IFN-γ⁺ helper T cells while promoting the immunosuppressive effect of Tregs from the NOD mice

CD4⁺ T helper (Th1) cells are pivotal in immune responses and play a crucial role in T1D pathogenesis by promoting the differentiation, activation, and proliferation of CD8⁺ cytotoxic T cells ⁵⁸. These cells secrete IFN-γ or IL-17, which are associated with Th1 and Th17 responses, respectively ^59,60. To validate the scRNA-seq findings indicating a reduction in IFN-γ-responsive cells in the AAT-MSC-treated group, we isolated PLNs from NOD mice 3 weeks post-treatment with AAT-MSCs or controls and performed flow cytometry. The results showed that AAT-MSC treatment significantly inhibited the Th1 cell response, as evident by a lower percentage of CD4⁺IFN-γ ⁺ cells in the AAT-MSC-treated PLNs compared to controls (5.52 ± 0.44% in CTR vs. 4.30 ± 0.17% in AAT-MSC, p = 0.04, Fig. 3. g&h), potentially contributing to the reduced autoimmune response in NOD mice. In contrast, no significant difference was observed in the Th17 cell population between the two groups (1.41 ± 0.33% in CTR vs. 1.48 ± 0.11% in AAT-MSCs, p = 0.85, Fig. 3. g&h).

To evaluate the impact of AAT-MSC treatment on Treg numbers and function, we quantified the percentage of CD4⁺CD25⁺FoxP3⁺ Tregs in PLNs from AAT-MSC-treated and control NOD mice 3 weeks post-treatment. As shown in the scRNA-seq analysis, we observed no significant difference in the overall number of Tregs in PLNs between the two groups (6.94 ± 0.93% in CTR vs. 7.02 ± 0.48% in AAT-MSC, p = 0.88; Figure S1d&e). However, two Treg subpopulations associated with more immunosuppressive function ^52,53 were enhanced. AAT-MSC-treated mice exhibited a significant increase in Foxp3⁺Helios⁺ (25.27 ± 2.09% in CTR vs. 33.43 ± 0.63% in AAT-MSC, p = 0.02, Fig. 3. j&k) and CD4⁺CD25⁺CTLA4⁺cells (30.50 ± 1.60% in CTR vs. 49.17 ± 6.42% in AAT-MSC, p = 0.047, Fig. 3. l&m).

To determine whether AAT-MSCs functionally enhanced the immunosuppressive function of Tregs, Tregs were isolated from the PLNs of female Foxp3^EGFP NOD mice treated with AAT-MSCs or untreated and co-cultured with T cells from C57BL/6 mice using a standard T cell proliferation assay (Fig. 3n). Upon stimulation with anti-CD3 and anti-CD28 antibodies, T cells showed a rapid proliferation rate of 85.50 ± 2.38%. When co-cultured with Tregs from control NOD mice, T cell proliferation was reduced to 35.50 ± 2.38%. In contrast, T cells co-cultured with Tregs from AAT-MSC-treated mice exhibited a further reduction in proliferation to 25.70 ± 1.65% (n = 6 mice per group, p < 0.05 vs control Tregs, p < 0.001 vs. T cell with anti-CD3/CD28, ANOVA, Fig. 3. o&p). These results suggest that treatment with AAT-MSCs enhances the immunosuppressive function of Tregs, likely by the increased Helios and CTLA4 expression in these cells.

Co-culture with AAT-MSCs increases the number of mouse and human Tregs in vitro.

We next investigated whether co-culturing immune cells with AAT-MSCs could increase Treg numbers in vitro, both in mouse and human samples. First, we co-cultured mouse CD4⁺ T cells with AAT-MSCs, with or without anti-CD3 and anti-CD28 antibodies, for three days. After this period, we measured the proportion of Tregs by flow cytometry. Our data demonstrated that co-culturing with AAT-MSCs not only reduced the total number of CD4⁺ T cells but also significantly increased Treg numbers and percentages, regardless of anti-CD3/CD28 stimulation (Fig. 4. a–c). This increase in Tregs was likely driven by enhanced Treg proliferation, as indicated by cell proliferation analysis (Fig. 4d). Additionally, we observed significant increases in Foxp3⁺Ctla4⁺ and Foxp3⁺Helio⁺ subpopulations of Tregs in cells treated with AAT-MSCs (Fig. 4e&f), further confirming that AAT-MSCs not only promote Treg proliferation but also the expressions of immunosuppressive markers including Ctla4 and Helios, as observed in vivo. Next, we evaluated whether this effect could be replicated in human cells. We co-cultured peripheral blood mononuclear cells (PBMCs) from healthy donors with AAT-MSCs for three days and measured the percentages of the CD4⁺CD25⁺CD127^low Treg population (Fig. 4g). PBMCs cultured alone exhibited a Treg population of 4.09 ± 0.58%. In contrast, PBMCs co-cultured with AAT-MSCs at a 1:10 or 1:5 ratio showed significant increases in the Treg population, up to 6.21 ± 0.14% and 6.72 ± 0.22%, respectively (Fig. 4. h&i). These results prove that co-culture with AAT-MSCs increases mouse and human Treg numbers in vitro.

Fig. 4. Co-culture with AAT-MSCs increases Tregs among T cells from NOD mice and human PBMCs.

a. Schematic of AAT-MSCs and T cell co-culture. b&c: Scatter dot plots (b) and frequency (c) of mouse CD4⁺CD25⁺Foxp3⁺ Tregs, as well as CTLA4⁺ and Helio⁺ Treg (e and f) in cultures with or without anti-CD3/CD28 antibodies and/or AAT-MSCs. (d). The histogram shows Treg cell proliferation and the percentage of proliferation. g. Schematic of human Tregs induced by co-culture of PBMCs with AAT-MSCs. h, i. Analysis of CD4⁺CD25⁺CD127^lo Treg population in human PBMCs co-cultured with AAT-MSCs. Data are presented as mean ± SEM of at least three independent experiments, and the scatter dot plot displays individual data points. *p < 0.05, **p< 0.01, ANOVA test.

Treatment with AAT-MSCs favors an exhausted CD8⁺ T cell phenotype in islets.

CD8⁺ T cells are the primary immune cells that infiltrate the islets of NOD mice and T1D patients and play a significant role in the pathogenesis of T1D by directly killing β cells ^61–63. To investigate the impact of AAT-MSC treatment on CD8⁺ T cell profiles, we performed scRNA-seq analysis in islets of AAT-MSC-treated and control NOD mice at 3 weeks post-treatment. We identified three major sub-clusters of CD8⁺ T cells in the islets: proliferative (Mki67⁺), memory (CD28⁺), and exhausted (Tox⁺) CD8⁺ T cells. The islets of AAT-MSC-treated mice exhibited higher proportions of all three subpopulations (Fig. 5. a–c).

Fig. 5. AAT-MSCs increase exhausted CD8⁺ T cells in vivo and in vitro.

a. t-SNE projection plots of CD8⁺ T cells in the islet from the mice treated with AAT-MSC or CTR. b. GSEA summary of gene signature in subtypes of CD8⁺ T cells. c. Portions of the subtype population of CD8⁺ T cells in the islet from AAT-MSC-treated or CTR mice. d& e Differentially expressed genes in total and exhausted (Tox⁺) CD8⁺ T cells. f. Schematic measuring mouse exhausted CD8⁺ T cell induced by AAT-MSC co-culture in vitro. g, h. Scatter dot plot (g) and quantification (h) of exhausted CD8⁺ T cells induced by co-culture with AAT-MSCs. CD8⁺ PD-1⁺ cells were pre-gated, and representative flow cytometry data show the expression of exhausted CD8⁺ T cell markers, including TCF, Tim-3, Tox, and TIGIT, in the presence or absence of AAT-MSCs. Data are presented as mean ± SEM from at least three independent experiments, with individual data points shown in the scatter dot plot. *p < 0.05, **p< 0.01. One-way ANOVA with post-hoc correction.

The top upregulated genes in total CD8⁺ T cells from the AAT-MSC islets compared to control islets include Tomm7, Cox7c, Egr1, and Ndula2. In contrast, downregulated genes include Camk1d, gphn, and Cmss1, among others (Fig. 5d). Tomm7 or “let-7” is associated with CD8⁺ T cell activation, as a decrease in let-7 expression in activated T cells enhances clonal expansion and the acquisition of effector function ⁶⁴. Among the downregulated genes, CAMK1D is expressed in CD8⁺-activated T cells, not in resting cells. Its activity increases upon T cell activation, which plays a key role in CD8⁺ T cell proliferation, cytotoxicity, and responsiveness to stimulation ⁶⁵.

Next, we focused on analyzing the exhausted CD8⁺ T cell subpopulation because exhausted CD8⁺ T cells have been correlated with slow progression of and response to treatment in immunotherapy trials for T1D, such as anti-CD3 therapy (teplizumab) ^1,66,67,68. The upregulated genes in CD8⁺ exhausted T cells from the AAT-MSC group include Ctrb1, Rpl35, Rpl37a, and Rpl38, while the top downregulated genes are Camk1d and Gcg (Fig. 5e). Our data suggest that AAT-MSCs promote CD8⁺ T cells with a resting phenotype and low cytotoxicity, reflecting their exhausted phenotype.

In addition, the characterization of CD8⁺ T cells in the PLN showed there were three major CD8⁺ T cell populations: memory (Il7r⁺), cytotoxic (Nkg7⁺), and exhausted (Tox⁺) (Figure S3a and b). There were fewer Il7r⁺ memory CD8⁺ T cells in the AAT-MSCs compared to controls, while the numbers of the other three types of CD8⁺ T cells are comparable (Figure S3c). IL-7R⁺CD8⁺ T cells play a role in the autoimmune destruction of pancreatic β cells in T1D. These cells are involved in the disease–s progression, as they can become autoreactive and target insulin-producing β cells, contributing to the development of the condition ⁶⁹. AAT-MSC treatment likely suppresses CD8⁺ memory T cells, thus protecting β against cell death.

We compared gene expression in the Tox⁺ exhausted CD8⁺ T cell subpopulation (Figure S3d). Specifically, the expression of Ms4a4b and lef1, both markers for exhausted CD8⁺ T cells, was upregulated. In contrast, Stat1 and Klf2, critical factors for CD8⁺ T cell proliferation and function, were downregulated in exhausted CD8⁺ T cells from AAT-MSC-treated cells. These gene expression changes suggest that AAT-MSCs promote an exhausted CD8⁺ T cell phenotype.

We next deciphered the impact of AAT-MSCs on exhausted CD8⁺ T cells in vitro. Mouse splenocytes were co-cultured ex vivo with AAT-MSCs in the presence or absence of anti-CD3 and anti-CD28 antibodies, and the frequency of exhausted CD8⁺ T cells was assessed (Fig. 5f). Co-culturing splenocytes with AAT-MSCs increased the frequency of both total CD8⁺ T cells and PD-1⁺ exhausted CD8⁺ T cells (Fig. 5g&h). The frequency of progenitor-exhausted CD8⁺ T cells (characterized by the expression of PD-1, TCF, and Tim3) and terminal exhausted CD8⁺ T cells (identified by expression of PD-1, Tox, and/or TIGIT) were significantly higher in the AAT-MSCs-co-culture group compared to those from controls (Fig. 5h). These findings suggest that AAT-MSC promotes the transition of T cells into a CD8⁺ T cell exhaustion phenotype.

Tregs from AAT-MSC-treated islets established cell-to-cell interactions with other cells in islets, likely via the enhanced IL-2 signaling

Using the cellchat application, we analyzed communications in PLN and islet cells using scRNA-seq data. Cells in control PLNs exhibited fewer interactions (75 interactions) than those from AAT-MSC PLNs (104 interactions, Fig. 6a&b, Figure S4a). Among the incoming signaling pathways in PLN cells, TGF-β, a key Treg differentiation factor, exhibits the highest relative strength in Tregs from AAT-MSC-treated mice compared to controls, but not much difference in the outgoing pathways (Figure S4b–d).

Fig. 6. Enhanced cell-cell communications in PLNs and islets from AAT-MSC-treated mice compared to controls as analyzed by Cellchat.

a. Netvisual aggregate of cells from PLNs of CTR (a) or AAT-MSC (b) mice. c. Incoming signals in immune cell subtypes from PLNs of CTR or AAT-MSC mice. d. Netvisual aggregate of cells from islets of CTR (d) or AAT-MSC (e) mice. f. Outgoing signals in immune cell subtypes from islets of CTR or AAT-MSC mice. g-h: Dot plots showing ligand-receptor prediction analysis between CD4⁺ Tregs and other immune cells in islets from CTR (g) or AAT-MSC (h)-treated mice.

In islets, control islet cells showed 2,211 interactions, while AAT-MSC-treated islet cells had 2,593 interactions (Fig. 6d &e and Figure S4e). Notably, AAT-MSC islets showed enhanced IL-2 outgoing signaling (Fig. 6f) and incoming signals (Figure S4f). Furthermore, compared with controls, Tregs from AAT-MSC-treated islets are highly interactive with the microenvironment and establish a more complex cell-to-cell communication network with other cell types via multiple pathways, as shown by increased outgoing signaling and ligand-receptor prediction analysis using Cellchat (Fig. 6g&h). Our results reveal that treatment with AAT-MSCs enhances cell communication, particularly from Tregs to other cell types in the PLN and islet.

8. Tregs educated by AAT-MSCs promote the conversion of effector CD8⁺ T cells into an exhausted phenotype

We investigated whether AAT-MSC-educated Tregs can directly promote the generation of an exhausted CD8⁺ T cell phenotype by enhancing communication with other cells. We separated CD4⁺GFP⁺ Tregs and CD4⁺GFP⁻ non-Tregs from spleen and PLN cells from the Foxp3^EGFP NOD mice and co-cultured them with or without AAT-MSCs for 72 hours. Subsequently, all four groups of cells were co-cultured with mouse CD8⁺ T cells for an additional 72 hours, and the percentages of exhausted CD8⁺ T cells were measured by flow cytometry. In some experiments, these cells were co-cultured with freshly isolated mouse islets to assess their impact on islet cell death (Fig. 7a).

Fig. 7. AAT-MSC-educated Tregs induce exhausted CD8⁺ T cell phenotypes and promote human islet survival ex vivo.

a. Schematic illustration of Treg education with AAT-MSCs, its impact on CD8⁺ T cell exhaustion, and islet survival (b and c) Scatter plots (b) and percentages (c) of exhausted CD8⁺ T cells expressing PD-1⁺, Tim-3, and TIGIT in the presence or absence of AAT-MSC co-culture for 72 hours were analyzed by flow cytometry. d-e: Microscope (d) images and quantification (e) of human islet cell death (measured by TUNEL⁺/total islet cells) and insulin⁺ cells/total islet cells in the presence or absence of CD8⁺ T cells and/or AAT-MSCs co-culture, quantified by TUNEL assay and insulin staining. Red: insulin⁺ cells, green: TUNEL⁺ cells, blue: DAPI. The bar graph represents the mean ± standard error of the mean (SEM), and the scatter dot plot represents an individual data point. * p< 0.05; **, p < 0.01, ***, P < 0.001. NS: not significant, ANOVA test.

Our data revealed that expression of the exhaustion markers PD-1, Tim3, and TIGIT was elevated to varying degrees in CD8⁺ T cells in groups co-cultured with GFP⁺ Tregs or AAT-MSC-educated Tregs, with the latter showing more pronounced effects. In contrast, GFP⁻non-Tregs alone had little, if any, effect on CD8⁺ T cell exhaustion. However, non-Tregs educated by AAT-MSCs showed an increase in exhausted CD8⁺ T cells compared to non-Tregs cultured without AAT-MSCs (Fig. 7b&c). These findings suggest that Tregs, but not non-Tregs, can directly promote CD8⁺ T cell exhaustion, and that AAT-MSC education enhances this effect in both Tregs and non-Tregs, particularly in Tregs. Furthermore, TUNEL staining of islets (Fig. 7. d&e) and a live and dead cell analysis (Figure S5) confirmed a positive correlation between islet cell survival and the percentage of exhausted CD8⁺ T cells: higher levels of islet cell death were observed when islets were co-cultured with CD8⁺ T cells or non-Treg cells, whereas islet survival was significantly better in islets co-cultured with AAT-MSC-educated Tregs and CD8⁺ T cells. These findings suggest that AAT-MSCs primarily protect islet survival through Tregs and their ability to convert CD8⁺ T cells into an exhausted phenotype that supports islet survival.

DISCUSSION:

The therapeutic effects of naïve or AAT-overexpressing MSCs in T1D and GvHD have been described in mouse models and in clinical trials ^21,35,37,70. However, the mechanisms by which these cells suppress autoimmunity and re-establish homeostasis in vivo to halt the destruction of pancreatic β cells remain incompletely explored. In this study, we demonstrated that a single infusion of AAT-MSCs reverses diabetes via modulating subtypes of immune cells critical for T1D. AAT-MSC infusion increased Treg immunosuppressive function by enhancing Helios and CTLA4 expression and suppressing IFN-responsive (Stat1⁺) cells. Meanwhile, AAT-MSC infusion is correlated with more exhausted CD8⁺ T cells and contributes to β cell survival. In vitro, co-culturing splenocytes or PBMCs promoted Treg generation and led to CD8⁺ T cell exhaustion. Our comprehensive data demonstrated that AAT-MSCs can be used for T1D therapy.

AAT-MSCs enhanced the immunosuppressive effects of Tregs. Treg insufficiency and dysfunction were observed in T1D patients and NOD mice, and increased Tregs are associated with a delay in T1D progression in children ⁷¹. There are persistent efforts to stimulate Tregs and suppress T effector cells to protect β cells, and direct infusion of Tregs has been shown to benefit T1D therapy ⁷². Tregs use several mechanisms for immune regulation, including the production of the anti-inflammatory cytokine IL-10, depletion of IL-2 needed for the proliferation of effector T cells, and transmission of inhibitory signals to both APCs and T cells via CTLA-4 ⁷³. We found that AAT-MSC infusion enhances Treg immunosuppressive function by promoting the expression of Helios and CTLA4, two critical transcriptional factors for Treg function. Helios expression can ensure stable suppressive and anergic phenotype expression under intense inflammatory responses ⁷⁴. This aligns with research conducted by other groups, which demonstrated that MSCs can stimulate the growth of Tregs by releasing certain factors and restore the disrupted Th17/Treg balance in autoimmune disease ^75,76. Furthermore, AAT-MSCs, either through cell-cell interactions, IL-10 secretion, or AAT, increase Treg populations and/or function, both in vivo in NOD mice and in vitro when co-cultured with mouse splenocytes or human PBMCs. This is likely achieved by promoting the expression of key genes involved in Treg stability and/or function, by converting T effector cells into Tregs, or by protecting Tregs from cell death. This function could provide insights into the increased immunosuppressive function of Tregs observed in vivo, potentially contributing to the therapeutic effects of AAT-MSCs in modulating immune responses.

Additionally, treatment with AAT-MSCs suppressed IFN-responsive T cells. Previous studies have shown that human umbilical cord-derived MSCs inhibit STAT1/3 signaling in T cells by secreting chitinase-3-like protein 1 (CHI3L1) and upregulating peroxisome proliferator-activated receptor δ (PPARδ). MSCs interfere with the phosphorylation of STAT1 and its binding to IFN-stimulated response elements, thereby controlling the expression of interferon-stimulated genes ⁷⁷. This mechanism is consistent with our finding that AAT-MSCs reduce Th1 cells in the PLN. Our study demonstrates that a significant portion of the therapeutic effects in T1D is attributed to the inhibitory effect of AAT-MSCs on Stat1⁺ IFN-responsive cells. The emergence of CD4⁺ T cell subsets detrimental to pancreatic β cell survival plays a crucial role in T1D. The canonical Th1 cytokine, IFN-γ, contributes to Th1 cell development, as evidenced by studies showing that mice deficient in IFN-γ (IFN-γ^−/−) exhibit aberrant Th2 cell development when challenged with pathogens that typically induce Th1 responses ⁷⁸. IFN-γ influences Th1 development by regulating IL-12Rβ2 chain expression and promoting IL-12 secretion from macrophages directly or through activated NK cells ⁷⁹. In addition, IFN-γ produced by Th1 cells can suppress the growth of Th2 cells ⁸⁰. In our study, AAT-MSC therapy suppressed IFN-responsive Th1 cells in the PLNs of mice, resulting in a significant decrease in IFN-responsive cells in PLNs treated with AAT-MSCs compared to controls. This subpopulation of cells is characterized by elevated expression of the transcription factors Stat1 and Lef1, as well as the membrane-bound type II C-lectin receptor CD69. Stat1 is a transcriptional factor that mediates signaling downstream of IFN-α, IFN-β, and IFN-γ. CD69 is known for its rapid appearance on lymphocyte surfaces following activation and is an early marker of lymphocyte activation ⁸¹; it associates with a predisposition to autoimmune conditions and affects Th1/Treg balance and the suppressive activity of Tregs ⁸².

Another novel finding is that AAT-MSC treatment favors a CD8⁺ exhausted T cell phenotype, characterized by increased PD-1, Tim3, Tox, and TIGIT expression in islet CD8⁺ T cells, as identified by scRNA-seq and confirmed in vitro. In humans, exhausted cells are hyperproliferative, exhibit expanded, rearranged T cell receptor junctions, and express exhaustion-associated markers, including TIGIT and KLRG1⁸³. In both NOD mice and humans, CD8⁺ T cells are a major component of immune infiltration in islets. Islet antigen-reactive CD8⁺ T cells can also be reproducibly detected in the blood and pancreas of people with T1D ⁶⁹. In recent years, exhausted CD8⁺ T cell signatures have been linked to slower progression of T1D after diagnosis and clinical response to immunotherapy ⁶⁷. For example, two distinct CD8⁺ T cell signatures exist at different stages of T1D: a proinflammatory signature in children with newly diagnosed T1D and a co-inhibitory signature in autoantibody-positive children who later progressed to T1D, suggesting that CD8⁺ T cell signatures could be used as biomarkers to evaluate T1D progression⁸³. Thus, our data show for the first time that AAT-MSCs promote the exhaustion of CD8⁺ T cells, which might have contributed to islet survival in NOD mice.

Our results also provided novel insights into the role of AAT-MSC-educated Tregs in promoting CD8⁺ T cell exhaustion and their subsequent protective effect on islet survival. The data suggest that Tregs, particularly those educated by AAT-MSCs, upregulate exhaustion markers in CD8⁺ T cells, indicating that Tregs, but not non-Tregs, can directly induce CD8⁺ T cell exhaustion. Additionally, our data showed a correlation between the exhaustion of CD8⁺ T cells and enhanced islet cell survival in the in vitro cell culture system. Therefore, AAT-MSCs confer a protective effect on islet survival, mainly by promoting Treg-mediated CD8⁺ T cell exhaustion and reducing islet cell death.

The in vitro studies suggest that AAT-MSCs can enhance the number of Tregs, possibly through multiple mechanisms, including the direct conversion of T effector cells into Tregs and the protection of Tregs from cell death. These functions could provide insight into the increased immunosuppressive function of Tregs observed in vivo, potentially contributing to the therapeutic effects of AAT-MSCs in modulating immune responses. Taken together, under autoimmune diabetes conditions, the function of Tregs decreases while the activity of T effector cells increases, leading to islet cell death. However, when AAT-MSCs are given, they boost Treg function and reduce effector T cell activity. Tregs enhanced by AAT-MSCs can also transform CD8⁺ T cells from an effector phenotype to an exhausted phenotype. These exhausted CD8⁺ T cells may interact with other impaired effector T cells, favoring the survival of pancreatic islet cells.

Our studies have limitations. Consistent with prior reports, systemically delivered MSCs, including AAT-MSCs, exhibit limited persistence in vivo, typically localizing transiently to the lungs, liver, and spleen before clearance within days to weeks⁸⁴. Although AAT levels were not directly measured in vivo, our in vitro data confirm stable AAT secretion by transduced MSCs during their viable period³⁵. We are planning follow-up studies to track AAT-MSC distribution, quantify in vivo AAT levels over time, and assess the durability of the effect through repeated dosing. We pooled PLN or islet cells from multiple mice for scRNAseq, as RNA sample pooling strategies can optimize both the cost of data generation and statistical power for differential gene expression analysis ⁸⁵. PLNs and islets were analyzed 3 weeks post-cell infusion, a time point we anticipated would reveal the impact of MSCs, making it an optimal time to compare the cellular landscape in the PLNs and islets between control and treated NOD mice. However, additional studies may reveal more pronounced effects. In addition, this study was conducted in female NOD mice that have a higher incidence of diabetes onset; therefore, potential sex-specific differences in therapeutic response were not addressed and should be evaluated in future studies.

In this study, we only focused on T cells, although other cell types in the innate immune system, such as APCs, may also contribute to the outcome of AAT-MSC protection. As a T cell-driven autoimmune disease that targets β cells, a key aspect of treatment is to dampen autoimmune attacks, particularly among individuals in the early stages of T1D. Both MSC and AAT showed profound protection in autoimmune diseases. Therefore, the observed effects may be due to a combination of MSCs and AAT. AAT-MSCs may prolong survival, enhance the function of MSCs, and induce sustained suppression of autoimmunity and β cell protection. MSC and AAT target different aspects of T1D, and AAT-engineered MSCs exhibit dramatically improved functionality, with consistent secretion of AAT at high levels. Furthermore, although donor- and passage-matched naïve MSC controls could modulate early inflammatory responses, they are unlikely to account for the coordinated transcriptional and functional immune reprogramming observed across multiple orthogonal datasets. Additional studies are needed in some aspects of our work to determine the relative contribution of the AAT gene edits compared to AAT or MSC treatment alone. Another limitation of this study is that the proposed mechanism primarily focused on the interaction between AAT-MSCs and T cells, while the potential direct effects of AAT-MSCs on pancreatic islets were not directly investigated. The higher abundance of α-, β-, and δ-cells in control islets may reflect the potential inclusion of diabetes-resistant NOD mice (~10-20% never develop diabetes), which could contribute to variability in cellular composition.

These data suggest the potential therapeutic applications of AAT-MSCs in autoimmune and other diseases. As an engineered cell therapy, any AAT-MSCs would require compliance with the FDA’s regulatory guidelines for advanced molecular and cellular therapeutics. These guidelines emphasize comprehensive chemistry, manufacturing, and control information, measurement of potency, and safety evaluations to ensure product consistency. Assessment of molecular modification stability, off-target effects, and long-term safety would be critical prior to clinical translation. While these considerations fall outside the scope of the present mechanistic study, they represent essential steps for future therapeutic development.

In summary, through comprehensive in vitro and in vivo analyses, we identified major T cell subpopulations and intracellular signaling pathways underlying AAT-MSC-mediated effects on diabetes remission. AAT-MSCs achieve therapeutic effects in T1D by suppressing autoreactive T cell infiltration and activation, enhancing Treg function, and promoting CD8⁺ T cell exhaustion. These mechanisms collectively contribute to the protection against T1D progression.

MATERIALS AND METHODS:

Preparation of AAT-MSCs

Human bone marrow-derived MSCs were isolated from bone marrow and infected with lentivirus to overexpress AAT, as described previously ³⁵. AAT-MSCs were incubated at 37°C in 5% CO₂ in α-MEM supplemented with 10% fetal bovine serum (FBS), 2mM L-glutamine, 100U/mL penicillin, and 100mg/mL streptomycin. Once the cultures reached 90% confluency, cells were subcultured or stored in liquid nitrogen. Cells at passage 6~8 were used in this study. All reagents were from Thermo Fisher Scientific (Waltham, MA) unless otherwise specified.

Mice and AAT-MSC infusion

Female NOD mice purchased from the Jackson Laboratory (Bar Harbor, ME) were used in the study. Mice with new onset diabetes (two consecutive non-fasting blood glucose readings > 250 mg/dl) were administered a single dose of 1 × 10⁶ AAT-MSCs intravenously or no treatment (controls). Non-fasting blood glucose levels of mice were measured by tail vein prick using the AlphaTRAK glucose meter twice weekly. Mice were followed for 10 weeks, and those with two consecutive high blood glucose levels (> 500mg/dL) were sacrificed. Mice with random blood glucose levels < 200 mg/dl were considered diabetes free. The Institutional Animal Care and Use Committee at the Medical University of South Carolina approved all mouse studies.

Hematoxylin and eosin (H&E) staining of the pancreas

The whole pancreas was isolated, fixed, and embedded in paraffin. At least three sections, spanning every 100 μm, were selected from serially sectioned slides and stained with H&E according to standard protocol. Insulitis scores were graded as follows: grade 0, a normal islet with <5% of immune cell infiltration; grade 1, 5-25% of the islet were infiltrated by immune cells; grade 2, 25–50% of the islet were infiltrated; grade 3, 25-50% of the islet were infiltrated; grade 4, >75% of islet was infiltrated. Each islet was evaluated by at least three people independently. Data were pooled from sections obtained from different mice in each group.

RNA isolation and RT-PCR analysis

Total RNA was isolated from cells using TRIzol reagent (Invitrogen, Waltham, MA) and then reverse-transcribed with M-MLV reverse transcriptase and oligo(dT) 18 primers. Real-time PCR was performed using SYBR Green I on a CFX96 Real-Time PCR Detection System (Bio-Rad, Hercules, CA). The thermal profile for qPCR was 95 °C for 10 minutes, followed by 40 cycles of 95 °C for 15 seconds and 60 °C for 1 minute. For a relative quantification, the results were normalized to 18S rRNA. Each sample was run in duplicate. The sequence of each primer pair is listed in Table S2.

ScRNA-seq, library preparation, alignment, and analysis

Female NOD mice (8 weeks old) received a single dose of AAT-MSCs (1x10⁶ cells per mice). Age-matched untreated mice served as controls. Three weeks after treatment, PLNs were harvested, minced, and enzymatically digested into a single-cell suspension with collagenase, then filtered through a cell strainer to remove debris. Islets were isolated from the pancreas using the standard procedure and dissociated into single cells with Accutase (StemCell Technology) for 10 min at 37 °C. Samples from 3- 8 mice were pooled for analysis. Single-cell suspensions were loaded onto the Chromium Controller (10X Genomics) and processed with the Chromium Single Cell 3′ Library & Gel Bead Kit (10X Genomics, v2) according to the manufacturer’s protocol. Libraries were sequenced on Illumina NovaSeq6000.

Raw base call (BCL) files were analyzed using CellRanger (v7.0.0) ⁸⁶. The “fast” command was used to generate FASTQ files, and the “count” command was used to generate raw gene-cell expression matrices. Ambient RNA contamination was inferred and removed using CellBender (v0.2.0) with standard parameters. The mm10 mouse genome was used for alignment, and gencode. vM25 was used for gene annotation and coordinates⁸³. Data from PLNs and islets were analyzed individually. Samples were combined in R using the “Read10X” function from the Seurat package (v4.3.0) ⁸⁷, and an integrated Seurat object was generated. Filtering was conducted by retaining cells with fewer than 30,000 unique molecular identifiers (UMIs), more than 500 genes, and mitochondrial content of less than 20 percent. Cell cycle analysis was conducted using CellCycleScoring with a list of cell cycle markers ⁸⁸. Doublets were removed using scDblFinder (v1.12.0) ⁸⁹. The SCTransform workflow was used for count normalization and initial integration and to identify highly variable genes ⁹⁰ using 30 principal components with a resolution of 0.3 for Louvain clustering and UMAP. Cluster marker genes were identified using FindAllMarkers with the Wilcoxon Rank Sum test and standard parameters. Cell annotation was performed using two approaches: 1) scType, an ultrafast unsupervised method for cell type annotations ⁹¹, and 2) Manual curation by gene markers to reflect the prediction results.

Identification of differentially expressed genes (DEGs)

Genes differentially expressed between the Control and AAT-MSC groups were identified. The R package LIBRA (v1.0.0) was used to perform zero-inflated regression analysis ⁹². Genes were defined as significantly differentially expressed at Benjamini–Hochberg correction FDR < 0.05 and abs (log2(Fold Change)) > 0.3.

Gene ontology analyses

The functional annotation of the identified DEGs was performed using enrichGO from the clusterProfiler R package ⁹³.

Cell-cell interaction analysis

Intercellular communication network analysis was performed using the standard workflow of the R package CellChat (v1.4.0) ⁹⁴. Cellchat analysis can predict ligand-receptor-mediated interactions from single-cell transcriptomic data, enabling comparison of signaling pathways between experimental groups.

Multiplex analysis of cytokine production

Cells from PLNs were cultured in 96-well plates coated with 10 μg/ml anti-CD3 antibody for 96 hours at 37 °C at 5% CO₂. The supernatants were then collected and stored at −80 °C. Cytokine levels were measured by analyzing the supernatant using a Premixed Analyte Kit (Mouse Custom 6-Plex, AssayGenie, Ireland), which included TNF-α, IL-1β, IL-17A, IFN-γ, IL-6, and IL-10. The assay was performed following the manufacturer’s instructions, and fluorescence signals were acquired using a CytoFLEX LX flow cytometer (Beckman Coulter Life Sciences, IN, USA).

Flow cytometry analysis

Single cells were fixed with Fixation/Permeabilization Concentrate and Diluent buffer set (Invitrogen) for 30 minutes on ice. Fixed cells were washed with a flow cytometry staining buffer (FACS buffer, Invitrogen). For Treg detection, fixed cells were incubated at room temperature (RT) for 30 mins in permeabilization buffer (Invitrogen) with the following antibodies: Brilliant^™ violet 605 (BV 605) anti-mouse CD4, Phycoerythrin-cyanine7 (PE-Cy7) anti-mouse CD25, PE anti-mouse Foxp3, Alexa fluor 647 (AF647) anti-mouse Helios and Phycoerythrin-cyanine5 (PE-Cy5) anti-mouse CD152 (CTLA4). For intracellular cytokine detection, cells were stimulated in RPMI 1640 (10% FBS, 200 μg/ml penicillin, and 50 μM 2-mercaptoethanol) containing phorbol 12-myristate 13-acetate (PMA; 50 ng/ml, Sigma-Aldrich), ionomycin (1 μg/ml; Sigma-Aldrich), and 1x protein transport inhibitor (BD Biosciences) for 4 h to facilitate intracellular cytokine accumulation. Cells were washed, fixed, and stained in permeability buffer at RT for 30 mins with BV 605 anti-mouse CD4, PE-Cy7 anti-mouse CD25, AF647 anti-mouse IFN-γ, and Peridinin chlorophyll-cyanine5.5 (PerCP-Cy5.5) anti-mouse IL-17A. For detection of exhausted CD8⁺ T cells, fixed cells were stained in permeabilization buffer at RT for 30 min with APC-Cy7 anti-mouse CD8, BV 421 anti-mouse CD279 (PD-1), PE anti-mouse TCF7/TCF1, PerCP-Cy5.5 anti-mouse CD366 (Tim-3), AF647 anti-mouse Tox, and PE-Cy7 anti-mouse TIGIT. Flow cytometry analysis was performed on BD LSRFortessa Cell Analyzer (BD Biosciences) and analyzed using FlowJo software. A list of antibodies used and dilutions is provided in Table S3.

T cell immunosuppressive assay

PLNs were isolated from AAT-MSC-treated or untreated Foxp3^GFPNOD mice 3 weeks after treatment. CD4⁺CD25⁺GFP⁺ Treg cells were sorted from the PLNs. Responder T cells were obtained from splenocytes of C57BL/6 mice by flow cytometry sorting. Sorted responder T cells were labeled with CellTrace Violet (CTV) for 10 minutes at 37°C. Treg cells (3×10⁵) were co-cultured with CTV-labeled responder T cells at a 1:1 ratio in RPMI 1640 medium supplemented with 10% FBS, 2-mercaptoethanol, and glutamine (Thermo Fisher Scientific) in 96-well plates. Cells were stimulated with anti-CD3 (2 μg/mL; BioXCell) and anti-CD28 (2 μg/mL; BioXCell) antibodies for 3 days. T-cell proliferation was assessed by flow cytometry using the standard method ⁹⁵.

Co-culture of AAT-MSCs with human PBMCs

AAT-MSCs (3×10⁵) were seeded into 96-well plates and allowed to adhere for 2 h at 37°C in a 5% CO₂ incubator. Human PBMCs (3×10⁵ per well) were then added for co-culture. After 4 days, PBMCs were harvested and stained with APC anti-human CD4, PerCP-Cy5.5 anti-human CD25, and Alexa Fluor 700 (AF700) anti-human CD127 antibodies. The frequency of CD4⁺CD25^highCD127^lo cells was quantified by flow cytometry.

Co-culture of Tregs, AAT-MSCs, with human CD8⁺ T cells and islets

CD4⁺CD25⁺GFP⁺ Tregs and CD4⁺CD25⁺GFP⁻ non-Tregs cells were isolated from FoxP3^EGFP NOD mice and stimulated with anti-CD3 (5 μg/ml) and anti-CD28 (5 μg/ml) antibodies. Cells were then co-cultured (or “educated”) with AAT-MSCs at a 1:1 ratio for 72 hours. Following education, Tregs or non-Tregs were co-cultured with CD8⁺ T cells for an additional 72 h, and the frequency of exhausted CD8⁺ T cells was quantified by flow cytometry using canonical T cell exhaustion markers, including PD-1, TIGIT, Tox, and Tim3. CD8⁺ T cells were subsequently isolated using the MagniSort^™ Mouse CD8⁺ T cells Enrichment Kit (Invitrogen, Waltham, MA) and co-cultured with mouse islets harvested from C57BL/6 mice for 3 days before assessing islet cell death.

Islet live/dead staining and terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) Assay

For live/dead staining, islets were incubated with 0.1 μM SYTO-13 and 10μg/ml ethidium bromide (EB) in PBS for 15 min in the dark. Stained islets were imaged with a fluorescent microscope, and the percentages of live and dead cells per islet were quantified and compared between groups. For apoptosis detection using the TUNEL assay, cryopreserved islet sections were fixed with cold acetone and washed with PBS, and incubated with the fluorescein labeling reagent from the In Situ Cell Death Detection Kit (Roche) for 1 h at RT. Following PBS rinsing, sections were permeabilized with 0.1% Triton X-100 for 15 min at RT, then incubated overnight at 4°C with an anti-insulin antibody. The following day, sections were washed and incubated with the appropriate secondary antibody for 1 h at RT. Fluorescence images were acquired using an EVOS M5000 fluorescence microscope (Thermo Fisher).

Statistical Analysis

Data are presented as mean ± standard deviation (SD) or standard error of the mean (SEM). Z scores were calculated for selected variables. Group differences were assessed using one-way ANOVA with post hoc tests. Comparisons between two groups were performed using Student’s t-test. A value of p < 0.05 was considered statistically significant.

AAT-overexpressing MSCs reverse new-onset diabetes in NOD mice by enhancing Treg expansion, suppressing pathogenic CD4⁺ and CD8⁺ T cells, and improving immune communication in pancreatic tissues. This immunomodulation supports islet survival and highlights AAT-MSCs as a promising therapy for type 1 diabetes and other inflammatory diseases.

Data Availability Statement:

The datasets generated and/or analyzed in this study are available from the corresponding author upon written request.