Survivin is an essential marker of the IFNg-producing cells phenotype. Flow-cytometry analysis of CD4+ cells from the blood of 22 patients with rheumatoid arthritis (RA) showed that the effector cells (TEFF, CD62LnegCD45RA+/-CD27neg) had higher levels of survivin than memory cells; 5.4–16.4% (mean 9.2%) of TEFF cells contained survivin and had highest amount of survivin per cell (Fig. 1a). The phenotype of survivin-producing TEFF cells was determined by RNA-seq analysis of CD4+ cells. Unsupervised clustering of the samples by the core genes characteristic of T-helper subsets22 identified accumulation of survivin/BIRC5 in the TEFF cluster marked by expression of Th1 signature genes (e.g., TBX21, EOMES, IL2RA, and IFNG) (Fig. 1b and Supplementary Fig. 1a) and cytokines (IFNg, IL9, IL10, and IL13) (Fig. 1d), which correlated with BIRC5 expression (Supplementary Fig. 1b). Comparison of BIRC5hi and BIRC5lo CD4+ cells revealed the complete Th1 signature enriched in BIRC5hi cells (Fig. 1d).
Availability and efficient metabolism of glucose are required for IFNg production and effector function of Th1 cells. Expression of the main glucose metabolism regulators HIF-1a and the catalytic subunit of AMPK-associated kinase (PRKAA1) differed between BIRC5hi and BIRC5lo CD4+ cells, but their expression of MYC and MTOR was similar (Fig. 1e). Since HIF1A expression is controlled by hypoxia, the selective enrichment in HIF1A prompted us to evaluate other genes of the hypoxia signature23. We found that BIRC5hi cells overexpress the canonical HIF-1a target genes, including lactate dehydrogenase (LDHA), enolase (ENO1), phosphoglycerate kinase 1 (PGK1), and aldolase A (ALDOA), corrrelated with glucose metabolism (Supplementary Fig. 1c). Specifically, BIRC5hi cells were deficient in the key regulator of glucose processing, PFKFB3 (Fig. 1e). As a result, glucose was shunted to the pentose phosphate pathway, as shown by increased expression of glucose-6-phosphate dehydrogenase (G6PD) and 6-phosphogluconolactonate (PGLS) and by increased expression of ATP citrate lyase (ACLY), indicating increased fatty acid metabolism. The correlation matrix of the core set of Th1 genes and glycolysis markers revealed clear distinctions in biological processes between BIRC5hi and BIRC5lo cells (Fig. 1f). The tight interactions in BIRC5hi cells suggested that survivin expression is functionally connected to these processes.
Survivin-bound chromatin is predicted to regulate carbohydrate metabolism
Since survivin has been reported to bind to genomic DNA elements that regulates gene transcription9-11, we investigate survivin-bound chromatin on the whole-genome level. Chromatin immunoprecipitation sequence (ChIP-seq) analysis of 12 CD4+ cell cultures pooled in 4 replicates (Fig. 2a) revealed 13,704 nonredundant survivin-ChIP peaks (enrichment against input, adjusted p < 10-5). The peaks were unevenly distributed across the genome and were specifically enriched in a region within 10–100 kb of the regulatory chromatin area occupied by promoters, enhancers, chromatin insulator regions, and CTCF binding sites (Fig. 2b, c).
To characterize the TF landscape of the survivin-ChIP peaks, we used the global ChIP-seq dataset for 1034 human transcriptional regulators in the ReMap database24 to annotate the set of nonredundant survivin-ChIP peaks. We identified 146 TF candidates that were significantly enriched across the survivin-ChIP peaks with 0 kb (minimal threshold for the overlapping peaks 10%) and 100-kb flanking regions (Fig. 2d) and compared them to regions within 1 Mb of the peaks. The TF candidates identified by this approach were tightly associated (Fig. 2d). The q significance of association with survivin was higher for TFs in the regions of 0 to 100 kb and lower for TFs within 1Mb.
To identify TFs in open chromatin of CD4+ cells, we used the ATAC-seq dataset (GSE13876724) to annotate nonredundant survivin-ChIP peaks. Survivin was tightly associated with a subset of TFs comparable to those identified by whole-genome analysis (Fig. 2d, inset). The q significance of the association did not change between the chromatin regions accessible at 2 and 4 h. The top TFs identified by both analyses were those regulating glucose and insulin metabolism, including CREBBP, KDM5B, DDX5, FOXK2, CTBP1, and IKZF1.
To identify biological processes regulated by chromatin-bound survivin, we analyzed the functions of the 146 TFs that co-localized with survivin peaks (Fig. 2e) and 2749 protein-coding genes expressed in CD4+ cells (RNA-seq, normalized raw counts >0.5) and used Gene Ontology terms to annotate them to the survivn peaks (Supplementary Fig. 2a). This approach identified functional groups that regulate chromatin, protein modification, and metabolism (Fig. 2e). Other functional groups regulated the response to hypoxia and organic substances, including glucose (Fig. 2e and Supplementary Table 1). In agreement with the functional annotation of TFs, gene set enrichment analysis of the protein-coding genes expressed in CD4+ cells and annotated to the survivin peaks revealed significant enrichment in processes regulating cellular biogenesis, carbohydrate metabolism, and hydrolase activity (Supplementary Fig. 2b).
Thus, survivin is frequently located near cis-regulatory elements (REs) and is functionally linked to regulation of protein and carbohydrate metabolism.
Survivin restricts PFKFB3 expression and changes the metabolic requirements of CD4+ cells. To investigate the role of survivin in the predicted biological processes, we used YM155 to inhibit survivin function10,25 in freshly isolated CD4+T cells using YM155. Cells were polarized with IFNg for the final 2 h. Comparison of differentially expressed genes (DEGs) identified by RNA-seq analysis of YM155-treated (0 and 10 nM) CD4+ cells (nominal p < 0.05, DESeq2) with those annotated to survivin peaks showed that 11.8% (24 h) and 4.5% (72 h) of the protein-coding genes expressed in CD4+ cells were sensitive to survivin inhibition (Fig. 3a). Using the curated TRRUST database of gene regulatory relationships, we identified the central metabolism regulators HIF-1a, c-MYC, and SP1 as the upstream transcriptional supervisors of the DEGs after 24 h and 72 h of survivin inhibition. Other effects were attributed to the activity of SMAD4, JUN, NF-kB, RELA, ETS1 TFs at 24 h and to interferon regulatory factor 1 (IRF1) and the MHC class II transactivator at 72 h (Fig. 3b).
To study in detail the enzymes involved in cellular glucose utilization (Fig. 3c), we analyzed YM155-treated and IFNg-polarized CD4+ cells by RNA-seq. We found that mRNA levels of PFKFB3 and LDHA increased rapidly, promoting conversion of pyruvate into lactate, and that PGLS and ACLY mRNA levels decreased, indicating downregulation of the pentose phosphate pathway and fatty acid metabolism (Fig. 3d). These effects of survivin inhibition blocked the alterations in carbohydrate metabolism seen in the BIRC5hi CD4+ cells from RA patients (Fig. 1f) but did not alter the mRNA levels of HIF1A or its metabolic targets HK2, ALDOA, ENO1, and GAPDH.
To assess the role of survivin in regulating glucose uptake by CD4+ cells, we measured the accumulation of glucose labeled with the fluorescent dye D-glucose derivate 2-N-nitrobenz-2-oxa-1,3diazol-4-amino]-2 deoxy-D-glucose (2NBD) in CD4+ cells activated with aCD3/IFNg. YM155 reduced uptake of 2NBD-glucose (Fig. 3e), resulting in decreased expression of the HIF-1a-controlled sugar transporters GLUT1 (encoded by SLC2A1), glucose-6-phosphate translocase (SLC37A4), and proton-associated sugar transporter A (SLC45A1) and the hypoxia-sensing proteins SESN2, PYHIN1, and NLRX1 (Fig. 4a). Survivin inhibition also activated expression of (1) the sugar sensors CDK5R1, KLF10, IL21R, and RXRA, (2) the transporters of neutral amino acids SLC7A5, and (3) the transporter of glutamate SLC1A4 controlled by c-Myc (Fig. 3g).
Survivin inhibition resets TGF-beta/SMAD signaling and promotes phenotype transition in CD4+T cells. The reset of PFKFB3-dependent glucose metabolism reduced IFNg production by CD4+ cells treated with YM155 for 24 h and 72 h (Fig. 4a) and inhibited IFN-dependent processes (Supplementary Fig. 2c). After 24 h, canonical IFN-sensitive genes were repressed, including cytotoxic PRF1 and GNL1, the proinflammatory cytokines CXCL8 and IL1b, and receptors that promote clonal T-cell expansion (IL2RA, SLAMF7, IL10RA) and joint homing (CX3CR1, ITGB3, ICAM2, TREM25) (Fig. 4b). The downregulation of IFN-sensitive genes was even more pronounced after 72 h and involved multiple IRF1-dependent genes (e.g., SOCS1 and HLA family genes). Importantly, the IFN-sensitive genes included in autoimmunity signatures of RA26, systemic lupus erythematosus 27 and Sjögren’s syndrome28 (e.g., IRF7, GAS6, IFI35, IFITM2, ISG15, ISG20, ODF3B) were also downregulated. (Fig. 4c).
The TGFb/SMAD pathway often counteracts the pro-inflammatory properties of IFNg, and SMAD4 is a predicted upstream regulator of the DEGs (Fig. 2b). We therefore investigated the effects of survivin inhibition on this pathway (Fig. 4d). Among the top DEGs (nominal p < 0.005; Supplementary Fig. 3), we found upregulation of (1) the E3 ubiquitin ligases SMURF2, SPSB1, SIAH3, LDLRAD4, and PMEPA1, which facilitate proteolysis required for T-cell reprogramming; (2) SMAD7 and its co-repressors SKI and SKIL, which physically interact with the receptor-activated SMADs; and (3) the chromatin-binding SMAD3 co-factors JUN, FOXO1, and BACH1 (Fig. 4e).
In agreement with the increased glycolytic activity of PFKFB3 and LDHA, which control the NOTCH1 and FOXO1 pathways29-31, survivin inhibition increased mRNA levels of FOXO1 and NOTCH1 (Fig. 4e). Consequently, CD4+ cells expressed higher levels of the surface receptors CD44, IL21R, ITGA5, and CXCR3 downstream of NOTCH1 and the FOXO1 target genes IL2RB, CCR5, CCR7, and CXCR4 (Fig. 4b). These transcriptional changes enable the phenotype transition of CD4+ cells.
Survivin colocalizes with IRF1 and SMAD3 on chromatin. To characterize chromatin bound by survivin, we used the JASPAR database of human TF binding sites to analyze motifs in genomic regions covered by the survivin peaks. This analysis revealed enrichment in IRF-binding motifs in all 4 independent ChIP-seq replicates. Predominant among the IRF motifs were IRF1 and IRF8, both containing the conserved GAAA repeat (Fig. 5a). The survivin peaks were also enriched in the composite motifs AP1:IRF (AICE motif, GAAAnnnTGAc/gTCA) and SPI1:IRF (EICE motif, GGAAnnGAAA). Multiple binding sites for each motif were frequently present in a single survivin peak. The ISRE motif (GRAASTGAAAST), which bound two IRFs, was also enriched compared to the whole genome, yet infrequent within the survivin peaks (Fig. 5a).
To connect survivin peaks with transcription, we annotated the whole set of survivin-ChIP peaks to open chromatin in aCD3/aCD28 activated CD4+ cells, using ATAC-seq data (GSE13876724). We found that 12.3% (2 h) and 21.5% (4 h of cell stimulation) of the peaks were located within 0–10 kb of open chromatin regions. An independent de novo motif search in those survivin peaks revealed up to 4.88-fold enrichment in the binding motifs of IRF1 and the SMAD3/SMAD4 complex, against the randomized background of all open chromatin (Fig. 5B and Supplementary Fig. 4a). No enrichment in JUND and JUN motifs was found. These finding confirmed the functional specificity of survivin binding.
Next, we looked for evidence of physical interaction between survivin and the predicted TF partners, we immunoprecipitated survivin from lysates of THP1 cells. After affinity isolation, protein denaturation, and separation by electrophoresis, western blotting with specific antibodies showed that IRF1 and SMAD3 co-precipitated with survivin (Fig. 5c). Neither IRF8 nor c-MYC, JUN, or JUND were identified in the precipitated material from two independent experiments.
Thus, survivin is recruited to open chromatin containing sequence-specific motifs through its binding to IRF1 and SMAD3. This finding provides molecular evidence that the IRF/survivin/SMAD3 complex helps coordinate the survivin-dependent transcriptional control we observed in the functional experiments (Figs. 3 and 4).
IRF1 and SMAD3 partner with survivin to regulate gene transcription. Since survivin-ChIP peaks accumulated in regulatory chromatin occupied by enhancers (Fig. 2c), we analyzed their presence in the cis-REs of the top protein-coding DEGs (Supplementary Fig. 3). Using the likelihood score for the enhancer–gene pairing32, we identified 117 REs that were paired to DEGs and associated with survivin peaks within 0–10 kb and 852 REs with no survivin peaks (Fig. 6a, b). These two groups of REs were similar in GeneHancer (GH) score, length/size of REs, and distance to the transcription start site (TSS) (Supplementary Fig. 5a).
Among the TF ChIP-seq peaks that co-localized with survivin peaks (10% overlap, 0-kb flanks) in ChIP-seq datasets (Fig. 2d), 58 TFs were expressed in CD4+ cells and were more abundant in survivin-containing REs than in the whole genome and or in the remaining REs (all p<10-3) (Fig. 6c). IRF1 and SMAD3 were among the most frequent and abundant survivin partners in REs paired to DEGs, as shown by density distribution analysis (Fig. 6d). Principal component analysis of the distribution of the enriched TFs across the REs, followed by unsupervised clustering of the components (Fig. 6e) revealed that REs clustered by total density of TFs (TF-poor and TF-rich) (Supplementary Fig. 5b) rather than by gene association and further by association of TFs with IRF1 or SMAD3 (Fig. 6e). Thus, survivin is present in TF complexes with distinct functions and diverse protein compositions.
Using the BioGrid database to analyze protein–protein interactions, we identified histone acetyltransferase EP300 and glycogen synthetase kinase 3B as the only common interactors of IRF1 and SMAD3 (Fig. 6f). EP300, a polyvalent protein that recruits TFs to distant enhancers, was enriched in survivin-containing REs and physically interacted with several other enriched TFs (Fig. 6e, f), providing a broad platform for building multiprotein complexes. Some of the IRF1 and SMAD3 interactors were differentially expressed in BIRC5hiCD4+ cells of RA patients (Fig. 6g, Supplementary Fig. 4b, c).
Survivin has a specific pattern of transcriptional regulation. To explore the mode of survivin-specific transcriptional regulation, we analyzed chromatin regions containing genes highly sensitive to survivin inhibition. Several common features emerged, including (1) long-range interactions between survivin-containing REs and the promoters of target genes, (2) the location of survivin-containing REs among REs clustered into regulatory modules, and (3) the location of survivin-containing REs on repressed/poised chromatin. These features are clearly seen in three genes critical for survivin-dependent metabolism in CD4+ cells: PFKFB3, BIRC2, and SMURF2 (Fig. 7a–c).
PFKFB3 was the main target of the survivin-dependent metabolic effects in CD4+ cells. We identified 4 survivin-ChIP peaks associated with 5 high-scored REs paired to PFKFB3 (Fig. 7a). These REs covered a region extending from ~20 kb upstream to 100 kb downstream of PFKFB3. Both the upstream and the downstream REs contained ChIP peaks for IRF1 and SMAD3 grouped together with the survivin peaks (Fig. 6e). Publicly available data (e.g.,, HiC, eQTL) indicated internal connection between these REs and the PFKFB3 promoter. Functional segmentation in CD4+ cells annotate the upstream REs to the active promoter and the downstream REs to the repressed region. Thus, activation of those areas after survivin inhibition would enhance PFKFB3 expression. Near the upstream REs were single nucleotide polymorphisms (SNPs) identified by genome-wide association studies (GWAS) provide additional weight to those loci and implicated in T-cell-mediated disorders, including latent autoimmune diabetes, and thyroiditis at the upstream REs and type 1 diabetes, RA, thyroiditis, and celiac disease at the downstream REs (Fig. 7a). Thus, the binding of survivin to those REs may be important for PFKFB3 expression and glucose utilization in T cells.
BIRC2 was significantly upregulated by survivin inhibition. Fig. 7b shows an extended region of ~550 kb containing 26 REs paired to and surrounding BIRC2. Two of these REs, located ~100 kb and ~400 kb downstream of the TSS, were associated with 3 survivin-ChIP peaks. Despite their distant location, both REs were strongly linked to BIRC2 (GH scores of 1.56 and 10.95, respectively). Both REs contained multiple IRF1 and SMAD3 ChIP-seq peaks and were located within the repressed/poised chromatin according to the functional chromatin segmentation in CD4+ cells. The BIRC2 locus contains few immunologically relevant SNPs.
Survivin activated transcription of SMURF2. Regulatory chromatin around SMURF2 forms a dense cluster of 25 adjacent REs that span the region of ~100 kb upstream of the TSS and cover the gene start (Fig. 7c). Four survivin peaks are annotated to 12 of those REs, 11 of which were in poised/inactive chromatin. An additional survivin peak was in SMURF2, outside of any RE. REs paired to SMURF2 differed in length, TF density, and presence of survivin partners IRF1 and SMAD3 (Fig. 6e). Simultaneous activation of the poised REs triggered by survivin inhibition was predicted by te RoadMap data that reveal a higher-order regulatory unit at this site. Therefore, cooperative activation of the clustered REs is a plausible mechanism for the pronounced upregulation of SMURF2 expression.