Tolerant state T cells exhibit distinct gene expression profiles and TF distributions.
T cell tolerance could be induced by in vivo application of G-CSF in both mice and humans 4, 5, 6. We assessed the transcriptome, chromatin accessibility and 3D genome landscape of steady state CD4+ and CD8+ T cells (CD4+ Tss and CD8+ Tss, respectively) together with tolerant state CD4+ and CD8+ T cells (CD4+ Ttol and CD8+ Ttol, respectively) in human bone marrow (BM) (Fig. 1A and Fig. S1). Steady state and tolerant state T cells were distinct from each other both in CD4+ and CD8+ T cells, as shown in dendrograms generated using transcriptome data (Fig. 1B). Then we explored and validated genes that were differentially expressed during the induction of tolerance in human CD4+ and CD8+ T cells (CD4+ Ttol and CD8+ Ttol) by G-CSF. We found that CD8+ Ttol cells differentially expressed multiple genes, such as SOCS1, PRDM1 and SEMA7A (Fig. 1C-D), compared to CD8+ Tss cells. We observed changes in a core set of genes, including suppressor of cytokine signaling 1 (SOCS1), PR/SET domain 1 (PRDM1), in a similar manner in CD4+ Ttol and CD8+ Ttol cells (Fig. 1C-D, and S2A-B). However, both CD4+ Ttol and CD8+ Ttol cells also exhibited changes in specific sets of genes. Notably, genes with CD8+ T cell-based changes in expression included SEMA7A, whereas those with CD4+ T cell-based changes included nuclear factor-κB inhibitor alpha (NFKBIA) (Fig. 1C-D, and S2A-B). Thus, although there is a clear transcriptomic change shared by both CD4 and CD8 lineages, there were also unique differentially expressed genes and TFs in CD4+ and CD8+ T cells during tolerance induction in vivo. CD8 Ttol cells showed a significant downregulation of genes related to cell activation (Fig. 1E), and CD4 Ttol cells showed downregulation of cytokine signaling genes (Fig. S2C). These results indicated T cells from G-CSF administrated healthy donors exhibited tolerance phenotype. Furthermore, we validated the direct upregulation of SOCS1 expression level by G-CSF in highly purified CD3+ T cells from 7 independent heathy donor BM samples in vitro. The results showed that G-CSF stimulation led to a peak in SOCS1 mRNA production after 4 h of culture, followed by recovery after 72 h of culture (Fig. 1F, Fig. S3A-B). After cultured 72 h of CD3+ T cells with G-CSF stimulation in vitro, the G-CSFR expression level was significantly increased (Fig. S3C-D). Consistent with previous studies7, IL-2 was decreased in the G-CSF treatment group which indicated that G-CSF suppressed differentiation of T cells to the Th1 type (Fig. 1G, Fig. S3E). These results demonstrated that G-CSF directly upregulated SOCS1 expression levels via GCSFR.
G-CSF loses its protective role in a GVHD mouse model in the absence of Socs1.
To investigate the role of SOCS1 regulation on T cell function during G-CSF induction of tolerance in vivo, we established a T cell-specific Socs1 conditional knockout (cKO; LckCre-Socs1fl/fl) mouse model. Consistent with previous studies14, most cKO mice survived longer than 6 months. Occasionally, cKO mice developed dermatitis at 4 weeks (3 in 50 mice, Fig. S4A). There was slight splenomegaly in the cKO mice compared with Socs1fl/fl (WT) mice (Fig. S4B-C). Flow cytometry analysis showed that CD3+ T cell were decreased in cKO mice compared with WT mice (Fig. S4D). The transposition of the CD4/CD8 ratio in cKO mice represents a decrease in CD4+ T cells and an increase in CD8+ T cells in cKO mice compared with WT mice (Fig. S4E). The ratio of naïve CD4+ T cells was increased in the cKO mice compared with the WT mice (Fig. 2A-B). IFN-γ secretion in CD4+ and CD8+ T cells in cKO mice was increased compared with that in WT mice (Fig. 2C-D). These data suggested that loss Socs1 might induce severe GVHD in the HSCT mouse model.
To validate this hypothesis, we examined the effect of losing Socs1 in T cells on the protective role of G-CSF in a well-established murine GVHD model (C57BL/6 to BALB/c). Donor WT or cKO mice received 5 daily subcutaneous injections of either PBS or 5 µg human G-CSF, and spleens were harvested on day 6. BALB/c recipient mice received 8 Gy total body irradiation (TBI), and 3×106 T cells from the spleen were transplanted intravenously from the respective donors the following day. We transplanted 5×106 T cell-depleted bone marrow cells (TCD-BM) as protective cells from the WT PBS group donor mice to all groups of recipient mice. As shown in Fig. 2E, G-CSF prolonged the survival of WT mice compared with the PBS group (green vs. blue); however, G-CSF accelerated the death of GVHD mice in the cKO group (yellow vs. red). A validation experiment in which 5×106 TCD-BM and 2×106 T cells were transplanted from the spleen also indicated that G-CSF exacerbated GVHD and shortened the life span of cKO mice compared with WT mice (Fig. 2F). Flow cytometry analysis also showed that G-CSF inhibited CD62L expression levels in both WT and cKO mice (Fig. S4F-G), which is consistent with our previous studies in clinical samples.
We further investigated the G-CSF-administered donor-derived T cell phenotype in recipient mice in GVHD models. Compared with the WT mice, the naïve populations of both CD4+ and CD8+ T cells were increased in the cKO group (Fig. 2G-H). The proliferation ability of CD4+ T cells from cKO mice was significantly increased compared with that of CD4+ T cells from WT mice (Fig. 2I). Moreover, we transplanted 1×106 MLL-AF9-induced AML cells into nonirradiated WT or cKO mice to investigate T cell function in the context of leukemia. The results showed that the loss of Socs1 in T cells prolonged survival in leukemic mice and delayed leukemia progression (Fig. 2J). This result indicated that losing Socs1 might activate T cell function in the tumor environment. Taken together, these results demonstrated that Socs1 is the key mediator in G-CSF-induced T cell tolerance.
High expression levels of SOCS1 impairs T cell proliferation and decreases aGVHD occurrence after HSCT.
To further investigate the role of SOCS1 in maintaining T cell tolerance, we used lentivirus to overexpress SOCS1 in steady-state T cells and found that the SOCS1 expression level was increased approximately 30-fold in the SOCS1 overexpression group (SOCS1 OE) compared with the control group (CT) (Fig. 3A). High expression of SOCS1 inhibited T cell proliferation, and more T cells were blocked in the G0 stage in the SOCS1 OE group compared to the CT or noninfection control group (Fig. 3B). The proliferative ability of CD4+ T cells was decreased in the SOCS1 OE group compared to the CT or noninfection group, while CD8+ T cells showed no change (Fig. 3C-D, Fig. S5A). This result is consistent with the in vivo study in which the proliferation ability of CD4+ T cells from Socs1 cKO mice was increased compared with that of CD4+ T cells from WT mice (Fig. 2I). Moreover, high SOCS1 expression in T cells also promoted TIGIT expression (Fig. 3E, Fig. S5B-C). There were no significant differences in the secretion of cytokines, such as IFN-γ, IL-2, IL-17, IL-4, and IL-10, by CD4+ T and CD8+ T cells between the SOCS1 OE group and CT group (Fig. 3F, Fig. S6A-C).
We further validated the relationship between the expression level of SOCS1 in T cells and aGVHD occurrence in patients after allo-HSCT. The results showed that there was a lower expression level of SOCS1 in the patients with aGVHD than in the patients without aGVHD at the same timepoint after allo-HSCT (Fig. 3G). This result indicated that a low expression level of SOCS1 might induce aGVHD after allo-HSCT.
G-CSF Regulates Target Gene Expression by Chromatin structure alteration.
Previous results identified SOCS1 as a key immune checkpoint for T cell tolerance, thus the regulatory network of SOCS1 needs to be further investigated. To explore the regulatory mechanism of SOCS1 during T cell tolerance induction, we performed transcription factor enrichment analysis in chromatin regions with differences in accessibility between CD8+ Ttol and CD8+ Tss cells using ATAC-seq data 15. The regions with high chromatin accessibility were located in promoter and distal intergenic and distinct in steady state and tolerant state T cells (Fig. S7A-B). We observed higher chromatin accessibility at the promoter and upstream elements of cell-type specific genes than at other regions in CD4+ and CD8+ Tss (Fig. S7C-E), consistent with the characteristics of these two cell lineages 16. These results highlight the feasibility and reliability of ATAC-seq in investigating the genome landscape and chromatin accessibility of human T cells. We found that TFs such as STAT3 were specifically activated in CD8+ Ttol compared to CD8+ Tss cells (Fig. 4A). T cell activation related TFs such as JunB, AP-1 were suppressed in CD8+ Ttol compared to CD8+ Tss cells (Fig. 4B). Differential expression of genes and differential activation of TFs including TXNIP and RUNX1 between CD4+ Ttol and CD4+ Tss were observed (Fig. S7E-F). According to the relationship between TFs and target genes 17, we plotted the network of upregulated genes and enhanced regulatory transcription factors in CD8 cells (Fig. 4C), which shows that SOCS1 is highly expressed and that the TF most strongly regulating SOCS1 is STAT3. We also identified a regulatory network of highly expressed genes and enhanced TFs in CD4+ cells (Fig. S7G).
Next, we analyzed the spatial regulatory relationship between transcription factors and target genes which is 3D genome interaction maps of CD4+ and CD8+ T cells from steady state to tolerant state. We obtained high-resolution maps (5 kb) of the 3D genome structure, including A/B compartments, TAD structure and loop structure, of all the samples (Fig. 4D). The HiCRep SCC scores 18 of the Hi-C matrices showed that cells of different lineages have different 3D genome structures (Fig. 4E). The loop length of CD4+ Tss was significantly longer than that of CD8+ Tss (median length: CD4 Tss, 210 kb; CD8 Tss, 190 kb; Fig. 4F-G). This finding indicates that the chromatin structure of CD4+ T cells is more variable than that of CD8+ T cells at the level of loops during the induction of tolerance.
The switch of A/B compartments from steady-state T cells to tolerant cells is shown for chromosome 10 in Fig. S8A. Approximately 7.4% and 7.3% of the genome regions in CD4+ and CD8+ T cells, respectively, switched from compartment A to B and were associated with downregulated gene expression after tolerance induction. Meanwhile, 6.6% and 4.5% of the genome regions in CD4+ and CD8+ T cells, respectively, switched from compartment B to A and were associated with upregulated gene expression (Fig. S8B-C). The TAD boundaries were more accessible and were enriched with CTCF and H3K27ac signals 13 (Fig. S8D). We found that most of the TAD boundaries (> 90%) are conserved and that the length of the TADs decreased slightly in the tolerant CD4+ and CD8+ T cells compared to the steady-state cells (Fig. S8E-F).
We demonstrated that more than 53% of the genes upregulated in CD8 Ttol cells compared with CD8+ Tss cells, including SOCS1, PRDM1, and KLF9, are located in the loop anchor regions (Fig. 4H, P < 1e-16). These observations showed that most of the differentially expressed genes may contact distal cis-regulatory elements via chromatin loops during the in vivo induction of human T cell tolerance. We found 7,481 chromatin loops in CD8+ Tss and 6,186 chromatin loops in CD8+ Ttol cells, with 4,786 chromatin loops appearing in both (Fig. 4I). Compared with those in CD8+ Tss cells, STAT3 and PRDM1 motifs were enriched in the gained loop anchors and were highly expressed in CD8+ Ttol cells, which is consistent with the ATAC-seq results (Fig. 4A, 4J). This suggests that STAT3 may be a structural protein that mediates the gain of chromatin loops. The lost loop anchor-enriched TFs included ZNF416 and TCF4 in CD8+ Ttol cells (Fig. 4J, Fig. S8G).
Association of STAT3 with SOCS1 Expression in Tolerant T Cells
We next explored the regions of spatial interaction with the promoter of SOCS1 and determined which transcription factors bind to these regions. On chromosome 16, the SOCS1 gene is located within one TAD in CD8+ Tss cells (Fig. 5). Then, we investigated the chromatin spatial structure, histone modification and TF binding sites around the SOCS1 gene. The genome-browser view of CTCF and STAT3 binding sites suggests that the CTCF protein mediates the interaction between the SOCS1 locus and the downstream chromatin region, and STAT3 proteins mediate the interaction between SOCS1 and upstream super enhancers. From Hi-C data, the interaction between the SOCS1 locus and downstream heterochromatin is weakened, and the interaction between SOCS1 and upstream super enhancers is enhanced in CD8+ Ttol cells compared to CD8+ Tss cells (Fig. 5). These results suggest that a new association of STAT3 with SOCS1 expression emerged during the in vivo induction of human T cell tolerance. In support of this hypothesis, genome-wide statistics showed that genes with long-range interactions with heterochromatin tended to be expressed at low levels, while genes with long-range interactions with enhancers tended to be highly expressed (Fig. S8H).
STAT3 mediates the spatial interaction between enhancers and promoters in the whole genome.
To investigate whether STAT3 competes with CTCF in regulating target genes, we performed ChIP-seq and CUT&Tag experiments to detect the colocalization of STAT3 and CTCF. For example, many of the binding sites of STAT3 and CTCF are colocalized in and around SOCS1 (Fig. 6A) and TXNIP (Fig. 6B), which are upregulated after G-CSF mobilization. Furthermore, STAT3 and CTCF colocalized in the whole genome analysis of human CD8 T cells and GM12878 cell lines (Fig. 6C-E and Fig. S9A-D). There was a significant overlap between the CTCF peaks and the STAT3 peaks in CD8 T cells, as shown by Venn diagram (P < 1e-10, Fig. 6F). The peaks of STAT3 binding are enriched in promoter and enhancer regions (Fig. 6G-H). Then, we classified the peaks of STAT3 binding into promoter regions or enhancer regions (Fig. 6H and Fig. S9E-G), and there is a significant spatial interaction between the promoter regions and enhancer regions (Fig. 6I-J). These results strongly suggest that the STAT3 complex is involved in enhancer and promoter interactions. Consistent with our observation, previous studies showed STAT3 could regulate chromatin topology and mediate transcription during T cell differentiation 19, 20, 21. STAT4 binding in the genome contributes to the specification of the nuclear architecture around Ifng during Th1 differentiation22. Furthermore, we observed both CTCF and STAT3 foci in the nuclei of Jurkat cells by immunofluorescence staining (Fig. S9H). Collectively, these results suggest that a STAT3-mediated enhancer-promoter interaction induces SOCS1 expression during the in vivo induction of T cell tolerance (Fig. S9I).
Furthermore, we detected STAT3 protein levels in SOCS1 overexpressed Jurkat T cell line. The results showed that a high SOCS1 expression level inhibited the phosphorylation of STAT3 (Fig. S10A-B). Moreover, the Western blotting results in the spleen cells from mice with T cell-specific Socs1 knock out showed that the phosphorylation of STAT3 was upregulated when Socs1 expression was inhibited (Fig. S10C-D). These results indicated that SOCS1 regulated the activation of STAT3 through a negative feedback loop.