It has been reported that Temra and Tem cells expressed a lower level of SIRPG (20). To further define the expression of SIRPG in various subsets of peripheral T cells, we used CD45RO, CD27, CD45RA, and CD62L to define the phenotype of resting peripheral CD4+ T cells from healthy donors (Supplementary Figure S1A), and then examined the expression of SIRPG in each subset. We found that the “naïve” subset (CD45RO^-CD27⁺) already expressed a high level of SIRPG regardless of the status of CD45RA or CD62L (Figure 1A). The level remained high in “Tcm” subset (CD45RO⁺CD27⁺). The transition from SIRPG^hi to SIRPG^- started in “Tem” cells (CD45RO⁺CD27^-); a significant fraction of CD45RO⁺CD27^-CD45RA^-CD62L^- Tem cells were SIRPG^low or SIRPG^-. The level continued to drop through the CD45RO^-CD27^-CD45RA^-CD62L^- subset, whereas the classical Temra (CD45RO^-CD27^-CD45RA⁺) became SIRPG^-, a finding in agreement with the published data. A similar pattern of expression of SIRPG was observed in resting peripheral CD8+T cells (Supplementary Figures S1B, C).

Granzyme K (GzmK)-expressing CD8+T cells were recently found to be enriched in several autoimmune and inflammatory diseases (23), including rheumatoid arthritis. They are functionally distinct from GzmB-expressing cells and play a pathogenic role in autoimmune and inflammatory diseases. Thus far, there is no reliable surface marker to distinguish between GzmK⁺ and GzmB⁺ T cells. A recent RNA-seq analysis of peripheral T cells indicates that GzmB was highly expressed in SIRPG^low peripheral CD8+T cells compared to SIRPG^hi CD8+T cells (18). We therefore examined the expression of surface SIRPG, intracellular GzmK and GzmB in CD8+ T cells of PBMC from healthy donors (gating strategy in Supplementary Figure S1D). On average, 18%, 15%, 29%, and 38% of CD8+T cells were GzmB⁺GzmK^-, GzmB⁺GzmK⁺, GzmB^-GzmK⁺, and GzmB^-GzmK^- cells, respectively (Figure 1B). In agreement with the RNA-seq data, the great majority of GzmB⁺GzmK^- cells were SIRPG^- compared to the other subsets (Figure 1C). As naive cells are typically CD45RA⁺GzmB^-GzmK^-, we therefore used SIRPG and CD45RA to divide peripheral CD8+T cells into four populations: SIRPG^-CD45RA⁺ (12%), SIRPG^-CD45RA^- (3%), SIRPG⁺CD45RA⁺ (60%), and SIRPG⁺CD45RA^- (22%) (Figure 1D). On average, 67% of the SIRPG^-CD45RA⁺ cells (population a of Figure 1E) were GzmB single positive cells; 52% of the SIRPG⁺CD45RA⁺ cells (population c of Figure 1E) expressed neither GzmK nor GzmB; and 51% of the SIRPG⁺CD45RA^- cells (population d of Figure 1E) expressed only GzmK (Figures 1E, F). Thus, the combination of SIRPG and CD45RA can be used to enrich GzmB⁺GzmK^- (from 18% to 67%, population a) and GzmB^-GzmK⁺ (from 29% to 51%, population d) CD8+ T cells of peripheral blood. However, this distinction was blurred in CD8+ T cells of inflamed synovial fluid (gating strategy in Supplementary Figure S1E). There were much fewer GzmB⁺GzmK^- cells (2.4%) (Figure 1G), which still expressed less SIRPG compared to other subsets (Figure 1H), and a marked increase in the percentage of GzmB⁺GzmK⁺ cells (41%) and downregulation of CD45RA (Figures 1G, I). Nevertheless, the SIRPG⁺CD45RA⁺ gate (population c of Figure 1J) still partly excluded GzmB⁺GzmK⁺ cells (from 41% to 25%) and modestly enriched GzmB^-GzmK⁺ cells (from 36% to 55%) (Figures 1J, K).

Previous studies have shown that the level of surface SIRPG in bulk Th cells increased after they underwent two cycles of division (20). We further demonstrated that SIRPG expression in purified CD4+ T cells was enhanced by Fc-containing TNFα inhibitors adalimumab (ada) and etanercept (eta) (24), but not certolizumab (cer), which lacks Fc. In agreement with these observations, we found that the surface level of SIRPG in CD4+ and CD8+ T cells of anti-CD3 stimulated PMBC increased over a course of 6 day and was further induced by anti-CD3 re-stimulation (Figure 2A; Supplementary Figure S2A). The effect of ada when added at time 0 became obvious 72 hours after stimulation and was detected in blasting (high FSC, large) and non-blasting (low FSC, small) CD4+ T cells (Supplementary Figures S2B, 2B, C, Supplementary Figure S2C). Its effect was less obvious in CD8+ T cells (Figures 2B, C; Supplementary Figure S2C). Eta, but not cer or tocilizumab (toc, a humanized IgG1 against human IL-6R), also showed a strong trend of enhancing the expression of SIRPG in blasting CD4+ and CD8+ T cells (Figures 2B, C, and S2C). In addition, resting T cells once encountered ada during the initial stimulation with anti-CD3/anti-CD28 maintained a higher level of SIRPG up to day 6 and upon re-stimulation with anti-CD3 compared to those without encountering ada during the initial stimulation (Figure 2D). Ada still enhanced, albeit to a less degree, the expression of SIRPG when added 72 hours after the initial ant-CD3 stimulation (Figure 2E). This lesser effect was probably due to a high level of SIRPG at this time point even before adding ada. Interestingly, ada did not augment the expression of SIRPG in resting T cells from un-stimulated PBMC (Figure 2F). As resting peripheral T cells included antigen-experienced SIRPG^- Temra and/or Tem, the effect of ada on the expression of SIRPG is dependent on T cell activation regardless of the history of antigen exposure. By contrast, none of the drugs altered the level of CD4, CD8, or the CD4/CD8 ratio (Supplementary Figure S2D).

The function of SIRPG is still poorly understood. To investigate the function of SIRPG, we transfected anti-CD3-stimulated PBMC with control or two SIRPG sgRNAs. When examined 5 days after the transfection, more than 95% of the surviving cells were T cells, including CD4+ and CD8+ T cells (Figure 3A), and either SIRPG sgRNA reduced the expression of SIRPG in almost 80% of the CD4+ or CD8+ T cells to a level even lower than that observed in CD19+ cells of unmanipulated PBMC (Figure 3B), which expressed little or no SIRPG. The transfected T cells were maintained with IL-2 followed by re-stimulation with anti-CD3/IL-2. We found no difference in the expression of several activation markers, such as CD25, PD-1, and HLA-DR (Supplementary Figures S3A–C), and the production of IFNγ and TNFα (Supplementary Figures S3D, E) between control sgRNA-transfected (WT) and SIRPG sgRNA-transfected (KO) CD4+ or CD8+ T cells.

We subsequently mixed the WT and SIRPG KO cells from each donor (a total of 6 donors) at 1:1 ratio and stimulated the coculture with anti-CD3/IL-2 for 4 days, then sorted the stimulated cells into 4 populations: SIRPG⁺ CD4+, SIRPG^- CD4+, SIRPG⁺ CD8+ and SIRPG^- CD8+ (Figure 3C). The coculture enabled SIRPG⁺ and SIRPG^- cells to be equally stimulated. The sorted cells were then subjected to bulk RNA-seq. One SIRPG⁺ CD8 sample and its corresponding SIRPG^- CD8 sample were of poor quality and thus excluded from the analysis. After excluding low-expressing genes, we found a few differentially expressed genes (DEGs), including SIRPG, between SIRPG⁺ and SIRPG^- CD4 cells (Supplementary Figure S3F, Supplementary Table 1) and between SIRPG⁺ and SIRPG^- CD8 cells (Supplementary Figure S3G, Supplementary Table 2) based on the threshold of ≥1.5-fold change and FDR ≤0.05 in pair-wise comparison. To increase the sensitivity of detecting DEGs, we pooled SIRPG⁺ CD4 together with SIRPG⁺ CD8 and SIRPG^- CD4 together with SIRPG^- CD8. By comparing the pooled SIRPG⁺ and pooled SIRPG^- samples, we identified 46 genes, including SIRPG, that passed the threshold of downregulated genes in SIRPG^- cells compared to SIRPG⁺ cells (Figure 3D, Supplementary Table 3). STRING analysis of the down-regulated genes revealed that 20 the 46 genes, such as UBE2c, TOP2A, and CCNB1, fall into the “Cell Cycle” Reactome Pathway (FDR = 1.28x10^-14), whereas 7 of the 46 genes, including XCL1, XCL2, and CCL1, are in the “Cytokine-cytokine Receptor Interaction” KEGG Pathway (FDR = 0.00055) (Figure 3E). Interestingly, several surface markers, such as CRTAM, and TNFRSF18, were also downregulated. By contrast, there are only 15 genes, including TC2N and three MHC II molecules HLA-DPB1, HLA-DRB1, and HLA-DRB5, that are upregulated in pooled SIRPG^- cells compared to pooled SIRPG⁺ cells (Figure 3D). We then used qPCR to confirm the differential expression of several cell cycle genes, such as UBE2C, TOP2A, and CCNB1, and surface molecules, such as CRTAM and TNFRSF18, and XCL1 between SIRPG⁺ and SIRPG^- cells, which were similarly generated, stimulated, and sorted in independent experiments using a different set of donors (Figure 3F). The downregulation of CRTAM in SIRPG^- cells was also detected with FACS (Figures 3G-I). However, we were unable to detect a difference in TNRFSF18 MFI (Supplementary Figures S3H–J) probably because its differential expression is too modest for FACS, which is less sensitive than qPCR.

The strong cell cycle signal suggests that SIRPG regulates the proliferation of T cells. The transcript counts of MKI67, which encodes the proliferation marker Ki-67, were also statistically lower in SIRPG KO cells in the bulk RNA-seq analysis but the fold reduction did not reach the threshold of 1.5-fold (Figure 4A). We therefore repeated CRISPR, set up the coculture, restimulated the coculture with anti-CD3/IL-2, and stained the cells for intracellular Ki-67. Indeed, the MFI of Ki-67 was already lower in SIRPG KO cells prior to restimulation compared to WT cells (Figure 4B). This difference was observed in both CD4+ and CD8+ cells and not due to less activation of SIRPG^- cells because the MFI of CD25, PD-1, and HLA-DR, three activation markers, was not lower in SIRPG^- population compared to SIRPG⁺ population (Supplementary Figures S3A–C). In agreement with the difference in the level of intracellular Ki-67, the number of WT cells also increased more robustly than SIRPG KO cells when separately stimulated for 4 days (Figure 4C) and the ratio of SIRPG⁺/SIRPG^- in the restimulated coculture increased from day 0 to day 4 (Figure 4D). This increase was comparably observed in CD4+ cells and CD8+ cells, resulting in no alteration in the CD4/CD8 ratio. Restimulation of the coculture with anti-CD3+/- IL-2 on day 4 further increased the SIRPG⁺/SIRPG^- ratio (Figure 4E). Thus, providing SIRPG in trans did not rescue the proliferation of SIRPG^- cells. The preferential expansion of SIRPG⁺ cells was not due to attenuated apoptosis because the percentage of AnnexinV⁺/7AAD⁺ cells was comparable between SIRPG⁺ and SIRPG^- cells (Supplementary Figure S4). In addition, the preferential expansion of SIRPG⁺ cells required anti-CD3 stimulation because no such preferential expansion was observed when the coculture was maintained with only IL-2 for up to 2 weeks (Figure 4F).

The only known ligand of SIRPG is CD47, which is expressed ubiquitously, including in T cells (Figure 5A). If the SIRPG-mediated proliferation and expression of CRTAM requires its interaction with CD47, then deficiency of CD47 should also lead to impaired proliferation and CRTAM expression. To examine the dependence on CD47 of SIRPG-mediated proliferation and CRTAM expression, we also generated CD47^- and control CD47⁺ primary T cells with CRISPR followed by sorting (Figures 5A, B). Interestingly, ablation of CD47 led to a subtle but reproducible increase in surface SIRPG MFI (Figure 5C). Despite the slight increase in SIRPG MFI, CD47^- cells expanded very comparably to CD47⁺ cells when stimulated separately for 4 days (Figure 5D). The subtle difference in SIRPG MFI between CD47⁺ and CD47^- cells was also observed when they were cocultured at 1:1 ratio and stimulated together for 4 days (Figure 5E). Unexpectedly, the CD47+/CD47- ratio in the coculture decreased over the 4-day course (Figure 5F), a trend consistent with the observation that murine CD47^- cells proliferate more robustly than CD47⁺ cells in response to anti-CD3/anti-CD28 stimulation in vitro presumably due to the disruption of the interaction between CD47 and thrombospondin 1 (25). In addition, The CRTAM MFI of CD47^- cells was also comparable to that of CD47⁺ counterparts in the coculture (Figure 5G). These features are either different or opposite to those of SIRPG^- cells, suggesting that the SIRPG-mediated proliferation and expression of CRTAM is independent of interaction with CD47.

The SIRPG-mediated proliferation was also recapitulated in Jurkat cells, which normally express a high level of SIRPG (Figure 6A). We transfected Jurkat cells with SIRPG sgRNA. The efficiency of SIRPG sgRNA was approximately 75% when examined 3 days after the transfection (Figure 6A), resulting in de facto coculture of SIRPG⁺ and SIRPG^- cells. This coculture allowed us to compare their expansion without any stimulation. In agreement with the data obtained from primary T cells, the SIRPG⁺/SIRPG^- ratio in the SIRPG sgRNA-transfected samples increased over a course of 2–3 weeks (Figures 6A, B). We further sorted SIRPG⁺ and SIRPG^- Jurkat cells from the SIRPG sgRNA-transfected samples to more than 90% purity (Figure 6C) and co-cultured the cells at 1:1 ratio (Figure 6D). Again, the SIRPG⁺/SIRPG^- ratio increased from approximately 1 to nearly 3 over a course of 2 weeks (Figure 6D) and the percentage of Ki67⁺ cells and the MFI of Ki67 were lower in SIRPG^- cells than those in SIRPG+ cells (Figure 6E). Jurkat cells also express a high level of CD47 (Figure 6F). The efficiency of CD47 sgRNA was approximately 50%, also resulting in a de facto coculture of CD47⁺ and CD47^- cells. However, the ratio of CD47⁺/CD47^- did not increase over a course of more than 2 weeks (Figure 6F).

To confirm the role of SIRPG in promoting proliferation, we stably expressed in KO Jurkat cells an exogenous Flag-tagged SIRPG (Flag-WT) driven by a CMV promoter (Figure 7A). While we were able to obtain nearly 100% SIRPG⁺ population after selection and sorting (Supplementary Figure S5A), the percentage of SIPRG⁺ fluctuated over a wide range (20% to 98%) (Supplementary Figure S5B). The cause of this fluctuation is still unclear and is probably due to the instability of the CMV promoter after integration into genome (26, 27). However, we found that the ectopic Flag-WT increased the expression of Ki-67 compared to its empty vector (Figures 7B, C) and that the number of Flag-WT Jurkat cells expanded more robustly than control cells over a course of one week (Figure 7D). We also used the dilution of CFSE to quantify the proliferation of Jurkat cells and found that the CFSE dilution is modestly higher in Flag-WT Jurkat cells compared to control SIRPG KO Jurkat cells (Supplementary Figure S5C).

We subsequently generated a series of deletion mutants of Flag-SIRPG (Figure 7A). The mutants were separately introduced into SIRPG KO Jurkat cells. While deletion of the D1 domain (dD1) or the cytoplasmic domain (dCD) did not affect the surface expression of SIRPG, deletion of the D2 (dD2) or D3 (dD3) domain rendered the Flag-SIRPG barely detectable on the surface (Supplementary Figure S5A), suggesting that D2 and D3 are critical for maintaining the structure and/or stability of SIRPG protein. Again, the percentage of SIRPG⁺ cells driven by each mutant fluctuated even after selection and repeated sorting. Flag-dD1 also enhanced the expression of Ki-67, the proliferation and the expansion of Jurkat cells (Figures 7B-D, Supplementary Figure S5C). By contrast, Flag-dCD was unable to restore the expression of Ki-67 or cell proliferation/expansion. The V263A SNP was predicted to cause a conformational change altering the stability, structure, or function of SIRPG protein and was very likely a causal SNP of type 1 diabetes. We also introduced the SNP into the Flag-SIRPG (Flag-VA). It was as readily detected with FACS as Flag-WT 3 days after the transfection (Supplementary Figure S5A), indicating that the Flag-VA mutant is structurally stable enough to be expressed on the plasma membrane. It was also as efficient as Flag-WT in restoring the expression of Ki-67 and proliferation/expansion of SIRPG KO Jurkat cells (Figures 7B-D, Supplementary Figure S5C). Taken together, our structural and functional analyses indicate that SIRPG-mediated proliferation depends on its CD but not its D1 and that the V263A conversion minimally influences this function.

Health donor PBMCs were purified from leukoreduction collars obtained from the Crimson Biomaterials Collection Core Facility, which prospectively collects discarded clinical materials matching investigator-defined criteria against available information on clinical samples. Synovial fluid samples are discarded samples collected exclusively as part of routine clinical care at the Brigham and Women’s Hospital Arthritis Center and are anonymous to the research team. This study has been approved by Partners Human Research Committee (PHRC), Boston, MA.