Type 1 diabetes SIRPG risk variants increase SIRPγ expression and CD47 loss enhances CD8⁺ T cell cytotoxicity

Abstract

Type 1 diabetes (T1D) risk variants within SIRPG influence transcript splicing and alter the expression of signal regulatory protein gamma (SIRPγ) on human T cells; however, the functional consequences of the risk variants remain poorly defined. Here, we demonstrate that peripheral blood mononuclear cells from individuals carrying a T1D-associated risk haplotype exhibit increased membrane-bound SIRPγ and reduced expression of its ligand CD47 across CD4⁺ and CD8⁺ T cell subsets compared to a protective haplotype. To interrogate the impact of impaired CD47 signaling, we generated CRISPR/Cas9 knockout (KO) models in Jurkat and primary human CD8⁺ T cells, as well as a pancreatic β-cell line (βLox5). CD47-deficient T cells displayed heightened activation, reduced co-inhibitory receptor expression, and increased production of IL-2 and TNF following stimulation. Furthermore, CD47 KO CD8⁺ T cell avatars expressing an HLA-A*02:01-restricted, IGRP_265-273-reactive T cell receptor (TCR) exhibited enhanced cytotoxicity toward βLox5 targets, while CD47 KO βLox5 cells were more susceptible to apoptosis. These findings reveal that CD47 signaling constrains T cell effector function and β-cell susceptibility, and its disruption through altered SIRPγ-ligation may contribute to autoimmune pathogenesis in T1D.

Introduction

Adaptive immune cell signaling via co-stimulatory molecules and negative regulators represents a critical mechanism for maintaining peripheral tolerance¹. Several single-nucleotide polymorphisms (SNPs) conferring risk for autoimmune diseases, such as type 1 diabetes (T1D), are tagged to genes involved in T cell activation and function². Two T1D-risk SNPs within signal-regulatory protein gamma (SIRPG), rs2281808 (C>T, intronic, minor allele frequency (MAF): 0.27, odds ratio (OR): 1.11) and rs6043409 (G>A, Ala263Val, MAF: 0.20, OR 1.13), which are in tight linkage disequilibrium (LD; R² = 0.94, D’ = 0.98)^3–5 have been suggested to regulate T cell and natural killer (NK) cell activation^6–9.

SIRPγ binds the ubiquitously expressed transmembrane immunoglobulin superfamily receptor, integrin associated protein (IAP; CD47)⁷, which functions primarily as a marker of “self” on healthy cells to inhibit cell-mediated destruction following ligation with SIRPα expressed predominantly on antigen-presenting cells (APCs)¹⁰. A similar mechanism is hypothesized to occur when CD47 binds SIRPγ expressed on T and NK cell subsets⁷. However, in contrast with SIRPα, SIRPγ has a short cytoplasmic tail lacking a signaling motif and, thus, is thought to modulate CD47-dependent signaling indirectly by competing for CD47 binding, thereby influencing the availability of CD47 for signaling through other receptors^7,11,12. The interactions of SIRPs with CD47 are also hypothesized to regulate apoptosis, activation, and migration in diverse immune populations^7,13–20. However, the understanding of the impact of CD47–SIRPγ signaling on activation and effector function across T cell subsets remains incomplete.

Here, we tested the hypothesis that SIRPG risk alleles (rs2281808 [C]/rs6043409 [G]) and T1D disease status modulate SIRPγ and CD47 expression on peripheral blood CD4⁺ and CD8⁺ T cells, promoting effector phenotypes, activation, and proinflammatory cytokine production. We further hypothesized that complete loss of CD47 signaling would augment activation, effector molecule expression, and cytotoxic function of CD8⁺ T cells. Hence, we utilized CRISPR/Cas9 gene-editing to knockout (KO) CD47 in the βLox5 human pancreatic β-cell line and in human T cells, including the CD4⁺ Jurkat E6-1 cell line, polyclonal primary human CD8⁺ T cells, and primary human CD8⁺ T cells engineered to express an islet autoreactive T cell receptor (TCR)^21,22, to determine how this pathway influences β-cell killing in T1D.

Results

SIRPγ and CD47 are highly expressed on naïve T cells in peripheral blood

Using leukopak-derived human PBMCs, we assessed membrane-bound SIRPγ and CD47 expression on CD4⁺ and CD8⁺ T subsets by flow cytometry (Fig. 1A and B) for naïve (CD45RA⁺CD62L⁺), central memory (T_CM; CD45RA⁻CD62L⁺), effector memory (T_EM; CD45RA⁻CD62L⁻), and terminally differentiated effector memory re-expressing CD45RA (T_EMRA; CD45RA⁺CD62L⁻) subsets within CD4⁺ and CD8⁺ lineages (Figs. 1C and D, and S1A). Further, we measured SIRPγ and CD47 expression on regulatory T cell (Treg; CD25⁺CD127^lo/−) and T conventional cell (Tconv; CD25⁻CD127⁺) CD4⁺ T cell lineages with the latter subdivided to define T helper 1 (Th1; CD183⁺CD194⁻), T helper 2 (Th2; CD183⁻CD194⁺), T helper 17 (Th17; CD194⁺CCR6⁺), and T follicular helper (Tfh; PD-1⁺CD185⁺) cells (Figs. 1E and S1B).

Fig. 1. SIRPγ and CD47 expression are enriched on naïve human T cells.

Leukopak-derived PBMC were analyzed by flow cytometry. A Representative flow cytometry histogram showing membrane-bound SIRPγ and CD47 expression on CD4⁺ (pink) and CD8⁺ (blue) T cells relative to fluorescence-minus-one (FMO; black) controls. B Quantification of membrane-bound SIRPγ and CD47 stain index (SI; sample MFI divided by FMO MFI) on bulk CD4⁺ and CD8⁺ T cells. SIRPγ and CD47 SI across naïve (CD45RA⁺CD62L⁺), central memory (T_CM; CD45RA⁻CD62L⁺), effector memory (T_EM; CD45RA⁻CD62L⁻), and terminally differentiated effector memory RA (T_EMRA; CD45RA⁺CD62L⁻) subsets within CD4⁺ (C) and CD8⁺ (D) T cell lineages. E Membrane-bound SIRPγ and CD47 SI on CD4⁺ subsets, including Th1 (CD183⁺CD194⁻), Th2 (CD183⁻CD194⁺), Th17 (CD194⁺CCR6⁺), T follicular helper (Tfh; PD-1⁺CD185⁺), and regulatory T cells (Treg; CD25⁺CD127^lo/−). Each data point represents the mean of technical duplicates from each donor (n = 8). Significant p values are reported for B two-tailed paired t-test and C–E one-way repeated-measures ANOVA with Geisser–Greenhouse correction and Bonferroni correction for multiple comparisons.

Overall, we found that CD4⁺ T cells expressed greater levels of membrane-bound SIRPγ and CD47 compared to CD8⁺ T cells (Fig. 1B). Moreover, we determined that membrane-bound SIRPγ and CD47 expression was highest on naïve T cells in both the CD4⁺ and CD8⁺ compartments and declined sequentially across T cell memory subsets, from naïve to T_CM, T_EM, and T_EMRA. Within the CD4⁺ T cell compartment, Tregs exhibit greater membrane-bound SIRPγ expression compared to Th1. We also found that the Th2 subset exhibited increased CD47 expression compared to Th1 and Tfh. These data reveal that membrane-bound SIRPγ and CD47 expression are elevated on naïve T cells, Tregs, and Th2 T cells compared to other effector subsets.

CD47/SIRPγ co-expression is downregulated on T cells in subjects with T1D

After establishing the expression patterns of membrane-bound SIRPγ and CD47 across various T cell subsets in control donors, we evaluated potential differences between PBMCs from CTR (n = 15) and T1D (n = 16) subjects by flow cytometry (Table S1, Panels 1 and 2). We observed that T1D subjects exhibited increased membrane-bound SIRPγ expression across multiple CD4⁺ and CD8⁺ T cell subsets (excluding Th17, CD4⁺ T_EMRA, and memory CD8⁺ T cells) and a modest, although not significant, decrease in CD47 expression compared to CTR subjects (Fig. 2A).

Fig. 2. Increased SIRPγ and reduced CD47 expression across multiple CD4⁺ T cell subsets in T1D.

PBMC from individuals with type 1 diabetes (T1D, n = 16) and non-diabetic controls (CTR, n = 15) were analyzed by flow cytometry. A Log₂-fold change (Log₂FC) in stain index (SI; sample MFI divided by fluorescence-minus-one [FMO] MFI) for (left) membrane-bound SIRPγ and (right) CD47 was calculated for bulk CD4⁺ and CD8⁺ T cells, conventional CD4⁺ T cells (Tconv; CD4⁺ CD25^lo/−CD127⁺), regulatory T cells (Treg; CD4⁺CD25⁺CD127^lo/−), T helper subsets (Th1, Th2, Th17, and T follicular helper [Tfh]), and memory T cell subsets, including naïve (CD45RA⁺CD62L⁺), central memory (T_CM; CD45RA⁻CD62L⁺), effector memory (T_EM; CD45RA⁻CD62L⁻), and terminally differentiated effector memory RA (T_EMRA; CD45RA⁺CD62L⁻). Fold changes are plotted against −log₁₀ p values, with significant subsets labeled and colored in orange. B, C Log₂FC of the CD47/SIRPγ expression ratio across CD4⁺ and CD8⁺ T cell subsets in CTR (blue squares) and T1D (orange circles) subjects, with horizontal lines denoting group means. Each data point represents the mean of technical duplicates from an individual donor. Significant p values are reported for A–C from multivariable χ² modeling incorporating genotype, age, and T1D status.

To further understand the co-expression of SIRPγ and CD47 in T1D subjects, we compared the CD47/SIRPγ expression ratio in each T cell subset (Fig. 2B and C). The CD47/SIRPγ ratio reflects the relative surface abundance of the CD47 receptor and SIRPγ ligand rather than a direct biochemical measurement of ligand engagement. Accordingly, changes in this ratio are interpreted as shifts in the potential for CD47/SIRPγ interactions rather than evidence of receptor occupancy. Interestingly, CD4⁺ T cells, specifically Tconv, Treg, Th1, Tfh (Fig. 2B), and naïve/T_CM CD4⁺ subsets (Fig. 2C) from T1D subjects had significantly lower CD47/SIRPγ ratios compared to CTR subjects. However, there were no significant changes in CD8⁺ T cell memory subsets (Fig. 2C). These findings indicate that individuals with T1D have altered ratios of CD47/SIRPγ expression on CD4⁺ but not CD8⁺ T cells, compared to control subjects.

SIRPGT1D risk haplotype is associated with increased SIRPγ expression on CD4⁺subsets

Given that rs2281808 and rs6043409 are in strong LD^4,5,7, we stratified our subjects by combined homozygous risk CC/GG (C/G), heterozygous CT/GA (Het), or protective TT/AA (T/A) haplotypes to assess SIRPG risk allele-specific effects. Population frequencies did not differ among T cell subsets (Fig. S2A) or T cell memory subsets (Fig. S2B) between the risk or protected subjects. However, membrane-bound SIRPγ expression was significantly higher in risk haplotype and Het subjects across most CD4⁺ subsets and memory compartments (Fig. 3A and B). CD47 expression remained largely unchanged except for an increase on naïve CD4⁺ T cells in subjects with the protective haplotype (Fig. S3A and B).

Fig. 3. Increased SIRPγ expression across CD4+ and CD8+ T cell subsets in individuals homozygous for the T1D risk haplotype at rs2281808/rs6043409.

PBMC from individuals with type 1 diabetes (T1D; orange circles) and non-diabetic controls (CTR; blue squares) were analyzed by flow cytometry. Membrane-bound SIRPγ expression, quantified as stain index (SI; sample MFI divided by fluorescence-minus-one [FMO] MFI), on A bulk CD4⁺ and CD8⁺ T cells, conventional CD4⁺ T cells (Tconv; CD25^lo/−CD127⁺), regulatory T cells (Treg; CD25⁺CD127^lo/−), T helper subsets, including Th1, Th2, Th17, and T follicular helper (Tfh), and B naïve (CD45RA⁺CD62L⁺), central memory (T_CM; CD45RA⁻CD62L⁺), effector memory (T_EM; CD45RA⁻CD62L⁻), and terminally differentiated effector memory RA (T_EMRA; CD45RA⁺CD62L⁻) subsets within CD4⁺ and CD8⁺ T cell lineages stratified by rs2281808/rs6043409 haplotype: homozygous risk (C/G), heterozygous (Het), and protection (T/A). Data points represent individual donors, with bars indicating group means, and error bars representing SEM (C/G, n = 9; Het, n = 12; T/A, n = 10). Significant p values are reported for multivariable χ² modeling incorporating genotype, age, and T1D status.

Co-expression of CD47/SIRPγ on T cell subsets was not significantly altered between subjects with the risk and protection haplotypes (Fig. S4). Overall, these data indicate that SIRPG haplotypes primarily modulate membrane-bound SIRPγ abundance rather than T cell subset frequencies or CD47 expression.

SIRPGT1D risk alleles do not alter T cell activation or co-inhibitory markers, but influence cytokine production

To determine whether SIRPG T1D risk alleles affected T cell activation capacity, we evaluated activation (CD25 and CD69) and co-inhibitory (CTLA-4 and PD-1) marker expression following 24 h of anti-CD3/CD28 stimulation. Notably, CD25, CD69, CTLA-4, and PD-1 expression did not differ by SIRPG haplotype (Fig. S5A and B). In contrast, intracellular cytokine and effector molecule profiling for IL-2, IL-4, IL-10, IL-13, TNF, IFNγ, perforin, and granzyme B revealed selective differences, where subjects with the protection haplotype showed elevated IL-2 (CD4⁺) and IL-13 (CD8⁺) at 4 h, while risk haplotype subjects produced more TNF (CD4⁺) at 24 h (Fig. S6). These results suggest that although SIRPG variants do not markedly shift T cell activation thresholds, they may modulate cytokine responses. Importantly, we did not observe significant differences in T cell subset frequencies across haplotypes (Fig. S2A and B), indicating that the cytokine changes observed are unlikely to reflect alterations in the abundance of Th1, Th2, Th17, or Tfh populations.

CD47 regulates activation and co-inhibitory signaling expression in CD4⁺ T cell-lineage Jurkat and primary CD8⁺ T cells

Given the observed alterations in membrane-bound SIRPγ and CD47 expression on CD4⁺ T cells, depending on T1D status and SIRPG risk haplotypes, we generated CD47 KO T cell models using CRISPR/Cas9 gene editing to isolate the functional role of CD47 in both CD4⁺ T cell-lineage Jurkat T cells²³ and primary CD8⁺ T cells. Following anti-CD3 stimulation, CD47 KO Jurkat cells showed significantly increased CD25 and CD69 expression and reduced CTLA-4 and PD-1 relative to mock controls (Figs. 4A, and S7A). Interestingly, SIRPγ expression was significantly increased in CD47 KO Jurkats (Figs. 4A and S7A). Intracellular cytokine and effector molecule analysis revealed increased IL-2, IL-10, IL-13, TNFα, and granzyme B and reduced IL-4, IFNγ, and perforin across timepoints in CD47 KO (Figs. 4B and S7B), supporting a broad shift toward an effector phenotype in CD4⁺ T cells with the loss of CD47 signaling.

Fig. 4. CD47-deficient Jurkat E6-1 T cells display reduced co-inhibitory and altered effector phenotypes following stimulation.

CD4⁺ T cell-lineage Jurkat E6-1 cells were subjected to CD47 knockout (KO; green squares) or mock transfection (gray diamonds) and stimulated in vitro with anti-CD3 before flow cytometric analysis. A After 24 h of stimulation, surface marker expression was quantified for CD25, CD69, CTLA-4, PD-1, and SIRPγ. Stain index (SI; sample MFI divided by fluorescence-minus-one [FMO] MFI) for each marker is shown for CD47 KO cells relative to paired mock-transfected controls. B After 4, 24, and 48 h of stimulation, intracellular effector molecule expression was quantified for IL-2, IL-4, IL-10, IL-13, TNF, IFNγ, perforin, and granzyme B. Data points represent the mean SI of technical duplicates from an individual Jurkat transfection (n = 6), with lines indicating paired measurements and error bars representing SEM. Significant p values are reported for A two-tailed paired t tests for parametric data or Wilcoxon signed-rank tests for nonparametric data, as appropriate and B two-way ANOVA with Bonferroni correction for multiple comparisons.

In primary CD8⁺ T cells, CD47 KO similarly decreased CTLA-4 and PD-1 and increased membrane-bound SIRPγ expression (Figs. 5A and S8A), with modest increases in IL-2 at 48 h and transient elevations in IL-4 and IL-13 at 4 h (Figs. 5B and S8B). Overall, CD47 KO had more pronounced effects in Jurkat cells, consistent with the higher baseline SIRPγ and CD47 expression observed in CD4⁺ T cells (Fig. 1).

Fig. 5. CD47-deficient primary human CD8⁺ T cells display reduced co-inhibitory marker expression and altered effector phenotypes following stimulation.

Primary human CD8⁺ T cells were subjected to CD47 knockout (KO; teal squares) or mock transfection (black diamonds) and stimulated in vitro with anti-CD3 before flow cytometric analysis. A After 24 h of stimulation, surface marker expression was quantified for CD25, CD69, CTLA-4, PD-1, and SIRPγ. The stain index (SI; sample MFI divided by fluorescence-minus-one [FMO] MFI) for each marker is shown for CD47 KO cells compared to paired mock-transfected controls. B After 4, 24, and 48 h of stimulation, intracellular effector molecule expression was quantified for IL-2, IL-4, IL-10, IL-13, TNF, IFNγ, perforin, and granzyme B. Data points represent the mean SI of technical duplicates from an individual donor (n = 8), with lines indicating paired measurements and error bars representing SEM. Significant p values are reported for A two-tailed paired t-tests for parametric data or a Wilcoxon signed-rank tests for nonparametric data, as appropriate and B two-way ANOVA with Bonferroni correction for multiple comparisons.

Loss of CD47 expression increases CD8⁺T cell cytotoxicity and β-cell susceptibility to cell-mediated killing

To determine how CD47 influences antigen-specific effector T cell function, we co-cultured CD47 KO or control IGRP-specific CD8⁺ T cell avatars with βLox5 (HLA-A*02:01) pancreatic cells^21,22,24,25. Following 16 h of co-culture, CD47 KO CD8⁺ avatars exhibited significantly enhanced cytotoxicity toward βLox5 targets (Figs. 6A and S9). We also generated CD47 KO βLox5 cells to assess the complementary role of CD47 in protecting pancreatic β-cells from cytotoxic killing^7,26,27. As hypothesized, these CD47 KO βLox5 cells showed greater susceptibility to apoptosis when co-cultured with IGRP-reactive CD8⁺ T cells (Fig. 6B). Together, these results demonstrate that loss of CD47 amplifies CD8⁺ T cell cytotoxicity and degrades β-cell resistance to cell-mediated killing, whereby the dysregulated CD47 signaling present in the context of T1D may augment the susceptibility of β-cells to CML by autoreactive CD8⁺ T cells.

Fig. 6. Enhanced βLox5 target cell killing by CD47-deficient IGRP-specific CD8⁺ T cell avatars and susceptibility of CD47-deficient βLox5 cells.

IGRP_265–273-reactive CD8⁺ T cell avatars (clone 32) were assessed for cytotoxic activity against βLox5 target cells by co-culture before flow cytometric analysis. A Mock-transfected (black diamonds) or CD47 knockout (KO; teal squares) CD8⁺ T cell avatars were co-cultured for 16 h with unedited βLox5 cells at a 1:1 effector-to-target ratio. B Unedited IGRP-reactive CD8⁺ T cell avatars were co-cultured for 16 h with mock-transfected (black diamonds) or CD47 KO (pink triangles) βLox5 cells at a 1:1 ratio. βLox5 cell death was quantified by flow cytometry and reported as total apoptotic frequency, defined as the combined percentage of early apoptotic (PI⁻Annexin V⁺) and late apoptotic (PI⁺Annexin V⁺) cells. Data points represent the mean of technical duplicates from an individual experiment (n = 6), with error bars representing SEM. Significant p values are reported for A two-tailed paired t tests and B Wilcoxon signed-rank tests for nonparametric data.

Discussion

Understanding immune checkpoints that regulate autoreactive T cell activation and cell-mediated killing is critical for identifying targets for intervention in T1D pathogenesis. Toward this goal, our study provides insight into the immunomodulatory role of the CD47–SIRPγ axis by assessing SIRPG T1D-risk versus protection alleles and determining how complete loss of CD47 signaling affects T cell cytotoxicity. Furthermore, the investigation of this pathway in human cells is crucial, as SIRPγ is absent on murine T cells⁷. Consistent with prior studies, SIRPG risk allele SNPs rs2281808/rs6043409 have been associated with genetic risk of autoimmune diseases such as T1D, rheumatoid arthritis (RA), systemic lupus erythematosus (SLE), and ulcerative colitis (UC)^3,4,28,29, although the mechanisms by which CD47–SIRPγ signaling regulates T cell biology remain debated^6,7,9,28,30.

Prior reports suggest that SIRPγ may either enhance or inhibit activation^{6,7,9,12,28,30}. In our study, we determined that although SIRPG risk alleles did not significantly alter T cell subset frequencies or activation/co-inhibitory marker expression, they were associated with clear CD4⁺ subset-specific shifts in CD47/SIRPγ expression and cytokine production. We also discovered that subjects with T1D exhibited lower CD47 expression but significantly higher membrane-bound SIRPγ expression across several CD4⁺ T cell subsets. Furthermore, complete loss of CD47 expression altered activation and downstream effector phenotypes in both CD4⁺ lineage Jurkat cells and primary CD8⁺ T cells.

SIRPγ and CD47 are highly expressed on CD4⁺ and CD8⁺ T cells^6,11,12,28, with enrichment on CD4⁺ T cells and particularly on Tregs, suggesting potential immunoregulatory importance^31,32. Thus, the contribution of the CD47–SIRPγ axis to Treg functionality represents an important direction for future investigation^33–35.

Across all subsets, in both CTR and T1D subjects and in our CD47 KO models, we observed increased membrane-bound SIRPγ expression when CD47 was reduced or absent, suggesting a potential negative feedback regulatory mechanism or receptor modulation upon ligation. The mechanisms underlying these observations remain unclear, yet other traditional negative feedback loop receptors, such as the Notch:Delta signaling in T cell regulation^36,37, have been shown to employ alternative splicing, receptor endocytosis, or shedding.

SIRPG has been implicated in autoimmune disease risk across various GWAS studies^{3,4,6,7,28,29,38}, yet the specific contributions of the SNPs rs2281808/rs6043409 to these diseases remain unclear. Some groups, like Sinha et al.^6,8,28, suggest that rs2281808 can alter SIRPγ expression and effector functionality, whereas rs6043409 influences SIRPγ isoform distribution⁹. Given the strong LD between these variants^{3–5,7,29,38}, we evaluated their combined effects on SIRPγ and CD47 expression and T cell regulation. Our findings corroborate prior work^6,8,9,28, showing decreased membrane-bound SIRPγ expression in subjects with the protective haplotype. Based on the Smith et al. studies⁹, differences in isoform 1 and 2 usage may be contributing to lower membrane-bound SIRPγ in subjects with the protective haplotype, possibly through increased secretion⁷.

Although SIRPG haplotypes did not significantly alter T cell subset frequencies or CD25, CD69, CTLA-4, and PD-1 expression before or after stimulation, haplotype-specific differences in effector molecules were observed. Protective haplotype subjects showed increased IL-2, IL-13, and granzyme B production, whereas TNF levels were higher in risk haplotype subjects. Given TNF’s connection to various autoimmune diseases and its capacity to increase CD47 expression via TNFR1³⁹, interactions between TNF and CD47–SIRPγ signaling merit further investigation.

To isolate the functional contribution of CD47 independent of genotype, we generated CD47 KO T cell models. CD47 KO Jurkat cells displayed increased CD25, a later marker of T cell activation, and CD69, an earlier marker of T cell activation, after 24 h of stimulation^40,41. CD47 KO CD8⁺ T cells showed similar trends of greater activation. Both models exhibited markedly reduced CTLA-4 and PD-1 expression, suggesting that CD47 regulates both co-inhibitory and co-stimulatory signaling. We also observed shifts in intracellular effector molecule production, including TNF⁴², IFNγ⁴², perforin⁴³, granzyme B⁴³, IL-2⁴⁴, IL-4⁴⁵, IL-10⁴⁶, IL-13⁴⁷, and notably higher expression of IL-2, which has a dichotomous role in promoting inflammation⁴⁸ and Treg functionality⁴⁹. Comparatively, CD47 KO Jurkats demonstrated more pronounced changes in surface and intracellular effector molecule expression than CD47 KO CD8⁺ T cells, consistent with the higher baseline SIRPγ/CD47 expression in CD4⁺ T cells.

Finally, we evaluated how CD47 loss influences antigen-specific cytotoxicity toward pancreatic β cells, where CD8⁺ T cell infiltration and HLA class I hyperexpression are prominent during T1D pathogenesis^50–52. CD47 KO IGRP-specific CD8⁺ T cell avatars showed enhanced killing of βLox5 cells, and CD47 KO βLox5 targets were significantly more susceptible to cell-mediated killing. Our findings support prior reports in the streptozotocin-induced C57BL/6J mouse model⁵³ and the human EndoC-βH1 cell line⁵⁴, demonstrating that the loss of CD47 on pancreatic β-cells enhances their vulnerability towards cell-mediated destruction. Notably, the expression of CD47 on stem cell-derived β-cells (sBC) and induced pluripotent stem cell (iPSC)-derived β-cells has been described as an immunomodulatory approach for conferring graft protection^55,56, highlighting the translational relevance of this pathway.

In summary, SIRPG risk haplotypes are associated with increased membrane-bound SIRPγ and decreased CD47 expression on T cells. Further, a complete loss of CD47 signaling enhances CD4⁺ T cell activation and CD8⁺ T cell cytotoxicity while augmenting pancreatic β-cell susceptibility to cell-mediated killing. Our findings identify the CD47/SIRPγ checkpoint as a possible therapeutic target for T1D, with translational potential both for modulating the CD47–SIRPγ axis in T cells and augmenting CD47 expression on β cells to enhance survival.

Methods

Human samples

Experiments using de-identified banked human peripheral blood mononuclear cell (PBMC) samples were approved by the University of Florida Institutional Review Board (IRB, Approval: IRB201400709) as non-human exempt research and performed in accordance with the Declaration of Helsinki. We selected cryopreserved PBMCs (N = 31) from the University of Florida Diabetes Institute (UFDI) DiaBank (see Table S2) based on their type 1 diabetes (T1D) status (n = 15 control [CTR], n = 16 T1D) and SIRPG genotypes at rs2281808 and rs6043409 (CC/GG risk, n = 9; CT/GA heterozygous [Het], n = 12; TT/AA protective, n = 10), determined using our custom UFDIchip⁵⁷. Both alleles were combined into a haplotype, given the strong LD (R² ≥ 0.95; D’ ≥ 0.95) observed^4,5,7. The study cohort was age-balanced across study groups, with additional demographic information provided in Table S2. For flow cytometric immunophenotyping, 2 × 10⁶ PBMCs per sample were thawed at 37 °C and washed three times into warm complete RPMI 1640 (cRPMI) as previously described⁵⁸. PBMC viability and cell count were established using the Cellometer Auto 2000 Cell Viability Counter (Nexcelom Bioscience) with the ViaStain acridine orange/propidium iodide (AO/PI) staining solution (Nexcelom Bioscience, Lawrence, MA, USA).

For experiments using fresh human PBMCs, peripheral blood leukopaks (N = 38) were purchased from LifeSouth Blood Centers (Gainesville, FL, USA), with a median age of 21 years (range 18–24 years) and biological sex distribution of 50% male and 50% female. PBMCs from leukopaks were enriched for CD8⁺ T cells with the RosetteSep Human CD8⁺ T cell Enrichment Cocktail (StemCell Technologies, Vancouver, BC, Canada) and further isolated using density gradient centrifugation with Ficoll-Paque PLUS (Cytiva, Marlborough, MA, USA). Per manufacturer instructions, pre-enriched CD8⁺ T cells were further isolated using the Human Naïve CD8⁺ T cell Isolation Kit (Miltenyi Biotec, Bergisch Gladbach, NW, Germany). For in vitro experiments described below, primary CD8⁺ T cells were cultured at 37 °C with 5% CO₂ in cRPMI supplemented with recombinant human IL-2 (rh-IL2; 100 IU/mL; Roche, Basel, CH-BS, Switzerland) and recombinant human IL-7 (rh-IL7; 10 ng/mL; R&D Systems, Minneapolis, MN, USA).

Cell lines

Jurkat T cells (Clone E6-1 cells; TIB-152) were purchased from the American Type Culture Collection (ATCC, Manassas, VA, USA) and cultured in cRPMI at 37 °C with 5% CO₂.

The human pancreatic β-cell line, βLox5, was kindly provided by Dr. Fred Levine (Sanford Children’s Health Research Center, La Jolla, CA, USA), as previously described^21,59,60. βLox5 cells were cultured in complete Dulbecco’s modification of Eagle’s medium (cDMEM) formulated as described⁶⁰, at 37 °C with 5% CO₂.

Flow cytometry

Flow cytometry was performed in duplicate. In brief, 4 × 10⁵ cells were incubated with Live/Dead Fixable viability dye (Invitrogen, Waltham, MA, USA) per the manufacturer’s instructions, then washed twice. Cells were surface-stained with fluorescently labeled antibodies (Table S1, Panels 1–4), then washed, resuspended in stain buffer, and acquired immediately.

For intracellular staining, cells were permeabilized using the eBioscience Fix/Perm Kit (Invitrogen) according to the manufacturer’s instructions. Permeabilized cells were stained with intracellular antibodies (Table S1, Panel 4), then washed twice with Perm Buffer, resuspended in stain buffer, and incubated at 4 °C overnight in the dark.

Data were acquired using a 3-laser (16V-14B-8R) Aurora spectral flow cytometer and SpectroFlo software (Cytek, Fremont, CA, USA). Spectral unmixing was accomplished using UltraComp eBeads Compensation Beads and the ArC Amine Reactive Compensation Bead Kit (Invitrogen) per kit instructions. Data were analyzed using FlowJo v10.8 Software (BD Biosciences, Ashland, OR, USA). Gating strategies are visualized in Fig. S1. Staining index for each marker was calculated as: mean fluorescence intensity (MFI) of the sample/MFI of fluorescence minus one (FMO) control.

CD47 KO

CRISPR/Cas9 gene targeting methods were adapted from the Amaxa 4D-Nucleofector protocol (Lonza, Basel, CH-BS, Switzerland). Single guide RNA (sgRNA) sequences targeting CD47 (Fig. S10A and Table S3) were designed using Synthego bioinformatics⁶¹. Jurkat E6-1, βLox5, or leukopak-derived primary human naïve CD8⁺ T cells were prepared at 1 × 10⁶ cells per electroporation condition (electroporation-only [mock-transfected] and CD47 KO). Ribonucleoprotein (RNP) complexes were created using a 6:1 molar ratio of 120 pmol of sgRNA (Synthego, Menlo Park, CA, USA) to 20 pmol of TrueCut^™ recombinant Streptococcus pyogenes Cas9 Protein v2 (Thermo Scientific), along with Nucleofector SE solution (for Jurkat cells and βLox5 cells; Lonza) or Nucleofector P3 solution (for CD8⁺ T cells; Lonza) in Nucleofector Supplement. Cell pellets were re-suspended in 20 µL of negative control (no sgRNA and no Cas9) or CD47 KO RNP complex solution, placed into a 16-well nucleocuvette strip (Lonza), and electroporated using the EH100 program in the Lonza 4D Nucleofector X Unit for electroporation. After electroporation, cells were recovered with 37 °C cRPMI (supplemented with IL-2 and IL-7 for primary human CD8⁺ T cells) and incubated at 37 °C with 5% CO₂ for three days. Viable cells were washed and utilized in downstream assays, outlined below. Confirmation of KO was done using PCR primers (Table S4) and a touchdown PCR amplification program (Table S5), where amplified CD47 DNA was created at the cut section and 250 bp upstream and downstream (~525 bp product) and sent to Genewiz© for sequencing (Fig. S10A).

Jurkat E6-1 activation

1 × 10⁵ CD47 KO and mock-transfected Jurkat T cells were incubated with 20 µg/mL of AffiniPure F(ab’)₂ Fragment Goat Anti-Mouse IgG (H + L) (Jackson ImmunoResearch, West Grove, PA, USA) for crosslinking with 5 µg/mL of soluble anti-human CD3 in cRPMI for 24 h at 37 °C with 5% CO₂. Samples were stained (Table S1, Panel 3) and analyzed on a Cytek Aurora spectral flow cytometer as described above.

Primary CD8⁺ T cell activation

CD47 KO and mock-transfected primary human CD8⁺ T cells were plated at 1 × 10⁵ cells/well on non-tissue culture-treated 12-well plates and activated with Human T-Activator CD3/CD28 Dynabeads at a 1:1 bead:cell ratio in cRPMI with 100 IU/mL IL-2 and 10 ng/mL IL-7 for 24 h. Cells were then stained (Table S1, Panel 3) and analyzed by flow cytometry.

CD47 KO T cell activation with intracellular staining

CD47 KO and mock-transfected Jurkat and CD8⁺ T cells were seeded at 1 × 10⁵ cells per well on 48-well plates coated with 5 µg/mL anti-CD3 plus 5 µg/mL soluble anti-CD28. GolgiStop (BD Biosciences) was added at a concentration of 1.33 µL/mL, 4 h prior to collection at 4-, 24-, and 48-h time points. Negative control (media alone) and positive control conditions (10 ng/mL phorbol myristate acetate [PMA; Millipore Sigma, Burlington, MA, USA] plus 0.5 µL/mL 500 nM ionomycin [Millipore Sigma]) for 4 h were processed in parallel. Cells were stained (Table S1, Panel 4) and assessed by flow cytometry.

T cell avatar generation

CD47 KO and mock-transfected primary CD8⁺ T cells were transduced with lentiviral vectors containing a multicistronic TCR (clone 32) and enhanced green fluorescent protein (eGFP) reporter expression construct to generate CD8⁺ T cell “avatars” specific for islet-specific glucose-6-phosphatase catalytic subunit-related protein (IGRP_265–273), as previously described^21,22,24. This TCR has been reported⁶² to recognize IGRP in the context of HLA-A*02:01 when presented by the βLox5 pancreatic cell line^21,24.

Assessing cell-mediated cytotoxicity

Cell-mediated lympholysis (CML) assays were performed in duplicate per condition as previously described^21,24 using CD47 KO and mock-transfected CD8⁺ T cell avatars co-cultured with CD47 KO and mock-transfected βLox5 target cells. In brief, βLox5 cells were labeled with CellTrace Violet (CTV) cell proliferation dye (Invitrogen) according to the manufacturer’s instructions, then resuspended in cDMEM and plated at 30,000 cells/well in a 48-well plate. CD8⁺ T cell avatars were added at 0:1 and 1:1T cell-to-target cell ratios, and cultures were incubated for 16 h at 37 °C with 5% CO₂. Supernatants containing T cells and non-adherent βLox5 cells were removed and reserved for flow cytometric analysis. Adherent βLox5 cells were dissociated from the well by incubating with 2.5% trypsin (Gibco) solution at 37 °C with 5% CO₂ for 5 min, then added to the reserved supernatants. Samples were stained with viability master mix and assessed by flow cytometry as described above using Propidium Iodide (PI)/Annexin V (Table S1, Panel 5). CTV⁺ βLox5 cells from CML co-cultures were gated to assess live, early apoptosis, late apoptosis, and necrotic cell frequencies (Fig. S9), and these values were normalized against no T cell controls.

Statistical analysis

All assays were performed in duplicate per condition. Data were tested for normality and accordingly, analyzed by parametric or non-parametric tests as defined in the figure legends, using GraphPad Prism Software version 9 with a post-hoc corrected p-value of 0.05 set as the threshold for significance. For multivariable modeling concerning controls vs. T1D subjects and risk alleles vs. protection alleles, Chi-square testing was done to account for age, T1D status, and genotyping to determine p-values.

Author contributions

Conceptualization: R.C.S. and T.M.B.; Methodology: R.C.S., R.B., M.E.B., L.D.P., and T.M.B.; Investigation: R.C.S., M.E.B., and L.D.P.; Formal analysis: R.C.S. and R.B.; Visualization: R.C.S., M.E.B., and R.B.; Funding acquisition: R.C.S., L.D.P., M.E.B., and T.M.B.; Project administration: T.M.B.; Resources: L.M.J. and T.M.B.; Supervision: T.M.B.; Writing—original draft: R.C.S. and M.E.B.; Writing—review and editing: R.C.S., M.E.B., L.D.P., R.B., L.M.J., and T.M.B.

Data availability

The datasets generated and/or analyzed during the current study are not publicly available due to institutional and regulatory data-sharing guidelines, but are available from the corresponding author on reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s44483-026-00002-5.