Bivalent chromatin assimilates genome deposition of survivin

Annotation of chromatin bound to H3K4me3, H3K27me3, and H3K27ac, revealed a total of 6199 bivalent chromatin regions (BvCR) across the genome of CD4 + cells (Fig. 1A). The genome location of survivin peaks (n = 13,703, Fig. 1A) was found within 65% of BvCR (S + BvCR, 4068/6199 regions) (Fig. 1B, C). We found that, H3K4me3 mark dominated the BvCR (43%, H3K4me3-BvCR), followed by H3K27me3 (33%, H3K27me3-BvCR) and H3K27ac (24%, H3K27ac-BvCR) (Fig. 1C, Supporting Figure S2A). This frequency distribution of the dominant H3 mark was comparable for the BvCR and S + BvCR (Fig. 1C).

Fig. 1

Survivin accumulates in H3K4me3-dominant bivalent chromatin regions in primary CD4+ cells. A Heatmap of ChIP-seq peaks within and outside the bivalent chromatin regions (BvCR) defined by genomic overlap between the histone H3-marks. B BvCR and survivin deposition. C Frequency difference of BvCR without and with survivin. Chi-square test p-value is shown. Numbers within bars indicate percentage of BvCR dominant in individual H3 marks. D Box plots of survivin peak scores within BvCR dominant by H3K4me3, H3K27me3 and H3K27ac marks. Kolmogorov–Smirnov test p-values are shown. E Signal plot of fold change in mean H3 peak score in 2 kb window from peak center, in naïve and YM155-treated CD4+ cells within (black) and outside BvCR (colored). F Box plot of percentage tag change in H3 modifications after YM155 treatment, within BvCR dominant in H3K4me3 (green), H3K27me3 (red), and H3K27ac (violet) within and outside survivin colocalization. Wilcoxon unpaired p-values are indicated. G Frequency of changeable BvCR that shifts in dominant H3 after YM155 treatment. Kolmogorov–Smirnov test p-values are shown

Investigating strength of the survivin peak binding, we found that the average survivin peak score was higher in S + BvCR, compared to survivin peaks only. Besides, survivin peaks within BvCR were significantly larger in the H3K4me3-BvCR (Fig. 1D). Similarly, score of the individual H3-marks showed the highest peak scores within H3K4me3-BvCR and a significant difference in H3-marks deposition within S + BvCR compared to BvCR not confined to survivin (Supporting Figure S2B). Notably, presence of survivin within H3K4me3-BvCR resulted in lower tag deposition of H3K4me3, H3K27me3 and H3K27ac. Presence of survivin in H3K27me3-BvCR and H3K27ac-BvCR increased deposition of H3K27me3 in those regions. Together, these results demonstrated non-randomness of survivin binding across the genome being frequently annotated to BvCR, where deposition of survivin reciprocally adjusted H3-mark deposition, appreciably in H3K4me3-BvCR.

Binding of survivin to H3K4me3-BvCR regulates their functional status

Exploring survivin function within BvCR, we asked if survivin inhibition affected the deposition of individual histone H3 marks. To investigate this, we cultured CD4 + cells in presence of the survivin inhibitor YM155 and performed the chromatin sequencing analysis of H3K4me3, H3K27me3 and H3K27ac deposition. The adjusted average enrichment profile of BvCR showed that YM155-treated cells changed the deposition of all three H3 marks (Fig. 1E). Quantifying H3-marks deposition within BvCR of YM155-treated CD4 + cells, we observed a significant increase in deposition of all 3 modifications within H3K4me3-BvCR (Fig. 1F, the boxes are above the dotted line). This increase was more profound in the BvCR co-localized with survivin (Fig. 1F). In contrast, H3K27me3-BvCR and H3K27ac-BvCR responded to YM155 treatment by increasing deposition of H3K4me3 mark alone (Fig. 1F). Occasionally, the quantitative increase in H3 deposition caused a shift in the dominant H3 mark within BvCR (Fig. 1G). In total, such a shift occurred in 530 BvCR (8.55%) and was significantly less prevalent among the S + BvCR (325/4068 vs 205/2131, p-value = 0.03). We observed that the H3K4me3-BvCR or H3K27me3-BvCR shifted into each other in equal frequency, reflecting the functional bivalency of those chromatin regions (Fig. 1G). The H3K27ac-BvCR frequently lost their status and gained the dominance of H3K4me3 (32%) or H3K27me3 (23%). Therefore, the analysis of YM155-treated CD4+ cells showed that survivin inhibition increased the density of the lysine trimethylation on histone H3, largely increasing the proportion of H3K4me3-BvCR.

H3K4me3-BvCR control the DNA damage response

To connect BvCR and long-distance gene regulation, we exploited the GeneHancer database [53] of experimentally confirmed connections between cis-RE and genes. We found that 59–65% of BvCR were located within cis-RE (Fig. 2A). Focusing on the transcriptome of CD4+ cells, we identified 4212 protein-coding genes connected to the cis-RE containing BvCR (Fig. 2A). Consistent with the highest frequency, the H3K4me3-BvCR had the largest number of connected genes transcribed in CD4+ cells (Fig. 2A).

Fig. 2

BvCR dominant in H3K4me3 together with survivin regulate transcription of DNA damage response genes. A Analysis strategy. BvCR within genomic regulatory elements (cis-RE, grey boxes) connected to genes, filtered on the protein-coding genes expressed in CD4+ cells, by RNA-seq. Transcription difference in CD4+ cells treated with IFNg or IFNγ + YM155 compared to sham cultures was calculated by DESeq2. Differentially expressed genes (DEG) were defined by a nominal p-value < 0.05. B Radar plot of Spearman’s rho correlations between H3K4me3 and H3K27me3 tag deposition change in all and survivin-positive BvCR and transcription change in CD4+ cells treated with IFNγ or IFNγ + YM155. Arrows indicate direction of transcription change. C Bubble plot of enrichment in biological processes among CD4.+ expressed genes connected to all and survivin-positive BvCR. Bubble size indicates protein number in the process. Color intensity shows false discovery range (FDR). D DNA damage response (DDR) network. Nodes are colored by dominant H3 mark in BvCR connected to genes within nodes (top map) and by transcription change after IFNγ or YM155 treatment (bottom map). Size of bubble corresponds to percentage of BvCR-connected genes within each node. DDR, DNA damage response. MMR, mismatch repair. RFC, replicator factor C. SSB, single strand break. DSB, double-strand break. HR, homologous recombination. MRN, MRE11-RAD50-NBS1. E Heatmap of normalized tag deposition of H3 marks, by ChIP-seq, in BvCR connected to DEG treated with IFNγ + YM155. Shaded squares indicate survivin-positive BvCR. Genes connected to multiple BvCR are marked in bold. F Heatmap of RNA-seq transcription difference in genes annotated to DNA repair and stress response categories. Transcription difference was calculated by DESeq2 statistics, p-values * < 0.05, ** < 0.01, *** < 0.001

To decipher if functional changes in the BvCR affected transcription of the connected genes, we investigated response of these genes to survivin inhibition after IFNγ stimulation. Analyzing the transcriptome in CD4+ cells, we found that 40% of the BvCR-connected genes were differentially expressed after IFNγ and/or YM155 treatment, i.e., were IFNγ- and/or survivin-sensitive (Fig. 2A). The changeable BvCR were less frequent among S + BvCR, while those connected to both the IFNγ-sensitive and survivin-sensitive genes were significantly predominant among S + BvCR (Supporting Figure S3A) reiterating our previous report [25] that genomic localization of survivin mediated IFNγ-dependent transcription.

To validate internal relation between the YM155-induced changes in H3 tag deposition in BvCR and the connected gene transcription, we built a linear regression model between these parameters in the IFNγ- and YM155-treated CD4+ cells compared to mock. Upregulation of IFNγ-sensitive and survivin-sensitive genes connected to the survivin-positive H3K4me3-BvCR had a strong direct correlation (Spearman r > 0.5) to the change in H3K4me3 and H3K27me3 tag deposition (Fig. 2B, Supporting Figure S3B), suggesting that survivin contributed to the dynamics of H3 tail deposition. Within H3K27me3-BvCR, the correlation between transcription and the survivin-dependent H3 tag deposition change was weaker (Supporting Figure S3C). These findings clearly demonstrated that 1) H3K4me3 was the epigenetic mark transducing survivin deposition in cis-RE into IFNγ-sensitive and survivin-sensitive regulation of transcription, and 2) transcription of survivin-sensitive genes was dependent on an interplay between H3K4me3 and H3K27me3 deposition in H3K4me3-BvCR.

To explore the cellular functions regulated by the BvCR, we searched for biological processes engaging 4212 BvCR-connected genes in CD4+ cells. We discovered that the DNA damage response (DDR, GO:0006974) was the principal pathway regulated by the BvCR (Fig. 2C, Supporting Table T1) followed by the nucleosome-modifying processes including cell cycle process (GO:0022402), mRNA metabolism (GO:0016071), RNA polymerase-dependent transcription (GO:0045944), DNA metabolism (GO:0006259), and chromatin organization (GO:0006325) (Fig. 2C). Consistent with a strong correlation between transcription and H3K4me3 deposition change, the genes active in these six processes were connected to the survivin-positive H3K4me3-BvCR. Such a connection was neither found in the H3K4me3-BvCR lacking survivin deposition (Fig. 2C) nor in the H3K27me3-BvCR. On the contrary, the genes connected to H3K27me3-BvCR represented immunologically relevant processes of T cell differentiation (GO:0030217), RNA splicing (GO:0043484), and RNA processing (GO:0006396) (Fig. 2C). Further, 65% of the genes connected to H3K4me3-BvCR were annotated to any of the six pathways above (Supporting Figure S4A), including a subset of genes annotated to more than two of the pathways. These results connected H3K4me3-BvCR to regulation of the DDR pathway in CD4+ cells.

H3K4me3-BvCR regulate the functional DDR network

To discriminate between specific tasks within the DDR regulated by the H3K4me3-BvCR, we utilized the recently proposed DDR interaction network [15] which employed a systems biology approach to catalogue protein–protein interactions and assign them into functional DDR assemblies (Fig. 2D). Annotation of the BvCR-connected genes to the DDR network identified 197 genes which were distributed between 89% of the network nodes (Fig. 2D, top. Supporting Table T2). Notably, 63% of the nodes were dominated by H3K4me3-BvCR connected genes, including the core nodes of DNA repair, chromatin regulators, stress response and ribosome, and contained the genes of the BRG1/SWI complex including SMARCA4, SMARCE1, SMARCB1, and the Replicator Factor C complex including RFC1, MSH2, and MSH6. Gene Ontology enrichment analysis showed that more than 60% of H3K4me3-BvCR connected genes in the DDR pathway were multifunctional (Supporting Figure S4A). For example, the genes of the BRG1/SWI complex subunits were represented in > 4 pathways, which pointed at their central role in the nucleosome-modifying processes supervised by H3K4me3-BvCR. Further, we noticed that the H3K4me3-BvCR-connected genes organized the nodes of stress response, single-strand, and double-strand break repair in the DDR network, while H3K27me3-BvCR controlled the nodes of homologous recombination through the MRE11-RAD50-NBS1 complex and p53. Ribosome and stress response nodes contained only a minor fraction of the BvCR-connected genes (Fig. 2D, top, Supporting Table T2).

Among the BvCR-controlled nodes were SSB/DSB repair including specific branches of mismatch repair genes ATAD5, RFC1, CHTF18, MSH2, MSH6; DNA replication genes ATAD5, RFC1, FANCI, CHTF18, MSH2, MSH6, CTPS1, MCM8; base excision repair MSH2, and nucleotide excision repair USP7, XAB2; Fanconi anemia complex genes FANCI, RMI2, FANCA, MCM8; homologous recombination genes MRE11, XRCC3, RMI2, UIMC1; the G1 cell cycle arrest category genes STK11, CAB39; and multiple stress response genes.

Survivin inhibition counteracted IFNγ effects and triggered DNA damage recognition and repair (Fig. 2D, bottom). To underpin molecular mechanisms connecting the H3K4me3 deposition with the transcriptional response to IFNγ, we retrieved the IFNγ-sensitive and survivin-sensitive genes of the DDR pathway connected to H3K4me3-BvCR (Supporting Figure S4A). We found that the majority (63–68%) of the survivin-sensitive genes (n = 28, Fig. 2E), and the IFNγ-sensitive genes (n = 45, Supporting Figure S4B) were connected to the survivin-positive H3K4me3-BvCR. Furthermore, several of the survivin-sensitive genes (FANCI, STK11, BCL2, FOXP1, CEBPG) and the IFNγ-sensitive genes (VRK1, RTEL1, DOT1L, MYC, BACH1, MDM4, AXIN2, XRCC3, HIPK2, HMGN1, BRD4, PYHIN1) were connected to > 1 H3K4me3-BvCR, which multiplied survivin control.

Analyzing cis-RE of the survivin and IFNγ-sensitive DDR genes connected to BvCR (n = 28 + 45), we found that survivin inhibition caused a significant change in H3K4me3 tag deposition in the corresponding BvCR (Fig. 2E, Supporting Figure S4B, S4D). The infrequency of H3K27me3-BvCR and H3K27ac-BvCR can be appreciated from the sparsity of tag deposition in these H3 modifications (Fig. 2E, Supporting Figure S4B). The functional bivalency in transcription of DNA repair genes is reflected by the change in tag deposition of H3K4me3-BvCR and H3K27me3-BvCR in cis-RE connected to these genes (Supporting Figure S5). In contrast to the whole BvCR-connected transcriptome, only a minor part of changeable H3K4me3-BvCR were connected to the DDR, indicating a stability of the DDR-related H3K4me3 mark after survivin inhibition.

Analyzing the DDR network mediated through H3K4me3-BvCR, we found that numerous IFNγ-sensitive (Supporting Figure S4C) and survivin-sensitive genes (Fig. 2F) were annotated to one or more of the DDR processes. IFNγ suppressed the SSB/DSB repair and ubiquitin response genes FANCI, MSH6, MSH2, ATAD5 and activated the genes involved in stress response and cell cycle control CDKN1A, STK11, RMI2, TP53, PIDD1 (Fig. 2F, Supporting Figure S4C). Inhibition of survivin by YM155 counteracted these effects of IFNγ and upregulated the DNA repair genes (Fig. 2F), suggesting that a bimodal response to IFNγ was survivin dependent. Thus, interrogation of the DDR network controlled by H3K4me3-BvCR revealed that survivin acted as a critical mediator of IFNγ effects regulating the SSB/DSB repair, stress response and cell cycle control.

Binding of survivin to BvCR is sequence-specific and assembles TF complexes

To detail survivin binding to BvCR in CD4+ cells, we investigated if TF assisted the reading of the histone marks. Previous studies reported that survivin controlled transcription by association with IRF1, SMAD3, and PRC2 complex [25, 31, 60]. Hence, we performed a motif enrichment analysis of the DNA sequences in the S + BvCR (n = 4068), survivin-negative BvCR (n = 2131), and the survivin peaks outside BvCR (n = 7823) (Fig. 3A). Analyzing 400–700 bp sequences for motif presence, we discovered that the S + BvCR and survivin peaks were frequently enriched (E-value > 100) with identical complex motifs (Fig. 3A), which were estimated as potential binding sites of 331 TF within these genome regions (Supporting Table T3). Notably, the TF motif enrichment was found solely in the survivin binding regions (S + BvCR and Survivin) and was largely absent in other BvCRs (Fig. 3A), which implied that survivin, not histone H3 modifications, accounted for the sequence specificity of the binding.

Fig. 3

Survivin binding within H3K4me3-BvCR is sequence-specific and mobilizes BRG1/SWI complex. A Bar plot of motif enrichment in survivin peaks within and outside of BvCR. MEME motifs enriched within BvCR. Venn diagram of TF identified by DNA sequence motif (MEME suite) and location (ChIP-seq) and mass spectrometry (MS). Group 1 had two unique motifs with different abundances indicated by E-value. B1 Dot plot of enrichment for human proteins/TF within BvCR, by ChIP-seq based ReMap2022 database. B2 Box plot of enrichment for BRG1/SWI complex proteins in ReMap2022 database. Counts underneath protein names indicates the number of overlaps with BvCR. C Coomassie-stained electrophoresis gel depicts survivin-bound proteins precipitated from THP1 cell lysate, separated by molecular weight (MW). Lanes represent two independent experiments (1 and 2, respectively). Bands with BRG1/SWI complex proteins identified by mass spectrometry are indicated by boxes. D Table of BRG1/SWI subunits identified in nuclear material of THP1 cells precipitated by survivin using liquid chromatography mass spectrometry. E Ribbon diagram of the canonical BRG1/SWI complex (PDB ID: 6LTJ) depicts the regions binding survivin (red). Ribbon color corresponds to the normalized fluorescence intensity of survivin binding in the peptide binding array. F Ribbon diagram of protein–protein interaction between the human survivin-H3 tail complex (PDB ID: 3UEF) and ARID1A residues 1722-ARG, 1726-GLU, 1732-LYS, 1738-ASP, 1833-ARG, 1853-GLU, 1855-ILE, and 1862-LYS (F1) and SMARCC2 residues 600-GLU, 643-PRO, 646-ASP, 647-PRO, 650-GLU, 651-ASP, 656-LEU, 682-SER, 683-VAL, 700-PHE, 701-SER, 702-LYS, and 703-MET (F2) in the canonical BRG1/SWI complex (PDB ID: 6LTJ). Survivin, arctic blue; ARID1A, magenta; SMARCC2, rose; interaction residues, yellow

To impose on this finding, we integrated the genome location of BvCR, S + BvCR, and survivin peaks with the location occupied by human TFs identified via chromatin sequencing in a catalogue of human cells [52]. The integrated TF enrichment analysis revealed that 215 TFs had a prevalent binding within BvCR (Fig. 3B, Supporting Table T3), which was stronger in the survivin binding regions (S + BvCR and Survivin) (Fig. 3B1). The targeted search for genomic occupancy by the DDR pathway controlling BRG1/SWI complex subunits demonstrated significant overlaps with BvCR (Fig. 3B2), presenting a likelihood that survivin binding to BvCR defined genome localization of the BRG1/SWI complex. Remarkably, the core subunits of BRG1/SWI complex overlapped frequently with H3K4me3-BvCR (Supporting Figure S6A), while the subunits specific for canonical (ARID1B, DPF2, SMARCA2) and polybromo-complexes (ARID2, and BRD7) were less enriched (Fig. 3B2).

Survivin anchors BRG1/SWI complex to BvCR

To experimentally challenge the survivin-dependent control of the DDR function, we performed protein analysis of nuclear content of THP1 cells immunoprecipitated with survivin. Survivin-IP material was separated by molecular weight using electrophoresis in two independent experiments (Supporting Figure S6B). Proteomic analysis of individual electrophoretic bands by nano LC–ESI–MS mass spectrometry revealed numerous peptides unique for the BRG1/SWI complex subunits including SMARCA2/4, SMARCC1, SMARCC2, SMARCD1, SMARCD2, SMARCE1, DPF2, PBRM1 (Fig. 3C, D) signifying a physical association of survivin with the BRG1/SWI complex. Particularly, we observed high and reproducible sequence coverage of the base module subunits SMARCC1, SMARCD1, SMARCE1, and SMARCD2 that anchored the complex to chromatin (Lane 2, Fig. 3C, D).

To detail the interaction between survivin and the BRG1/SWI complex, we applied the peptide model based on the functional group composition of each protein of the complex [31, 57]. We identified that SMARCC2, SMARCD1, SMARCA2, SMARCA4, and BRD7 exhibit comparable ratios between the predicted survivin-binding peptides to the total number of generated peptides (Rbind-values range 0.39–0.43) and harbored a high fraction of regions favorable to survivin binding (Supporting Table T5). Introduction of mutations in the binding positions of the peptides demonstrated robustness of the predicted binding between survivin and those proteins. A value representing the ratio between the number of mutations that support binding to the total number of mutations, was mapped onto the 3D structure of canonical BRG1/SWI complex, thereby exposing stability of the survivin binding regions (Supporting Figure S6C). DPF2 and PHF10 appeared enriched in survivin-binding regions (Rbind-values 0.46 and 0.51, respectively). Conversely, ARID1A and ARID2 exhibited low survivin-binding ratio comparable to the median of the proteome (Supporting Table T5).

Motivated by the peptide-based prediction, we performed a direct interaction experiment between survivin and BRG1/SWI complex subunits through a peptide-binding array covering the complete sequence of BRG1, ACTL6A, ARID1A, SMARCC2, SMARCD1, SMARCB1, SMARCE1, DPF2, and PHF10 proteins. The fluorescence signal generated by labeled survivin was strongest with the peptides of core subunits BRG1, SMARCC2, SMARCE1, and subunits of the canonical BRG1/SWI complex ARID1A, and DPF2 proteins (Fig. 3E, Supporting Figure S7A). The peptides of SMARCD1, SMARCB1, and PHF10 proteins had somewhat lower binding intensity although all above 30,000. The signal was observed from several peptides and formed continuous confluent regions of survivin interaction depicted in three-dimensional structure (Fig. 3E).

To visualize the most likely binding site(s) of survivin on the protein surfaces of the BRG1/SWI complex, we conducted protein–protein docking between survivin and the mammalian BRG1/SWI complexes based on the known 3D structures (cBAF, PDB ID: 6LTJ; PBAF, PDB ID: 7VDV). These docking analyses confirmed that both cBAF and PBAF use SMARCA4, SMARCC1/C2, and SMARCD1 subunits for binding with survivin. The absence of interaction between SMARCE1 and survivin in our predicted complexes may be attributed to the potential incompleteness of the EM structures of the cBAF and PBAF complexes. Additionally, cBAF-specific subunits ARID1A and DPF2 and PBAF-specific subunits PHF10, ARID2, PBRM1, and BRD7 contribute to the interactions with survivin (Supporting Figure S7A and S7B). Together, this highlighted the distinctive roles of specific BRG1/SWI subunits in chromatin accessibility regulation [61, 62]. The cBAF subunits exhibit a larger interaction area with survivin compared to the PBAF subunits (34 residues from cBAF vs. 21 residues from PBAF). Accordingly, the docking energy score for survivin binding to canonical complexes was somewhat lower compared to polybromo BRG1/SWI complexes (Supporting Figure S7A and S7B). Structural modelling of the binding interphase between the complexes and survivin demonstrated maximal binding contacts with amino acid residues of SMARCC2, SMARCD1, and ARID1A (experiment 1), SMARCB1, SMARCE1, and ARID1A (experiment 2), and ACTL6A, SMARCB1 and ARID1A (experiment 3), which supported the hypothesis that these subunits provided a probable interaction platform for survivin. Notably, the amino acid residues of SMARCC2, SMARCD1, SMARCE1 and ARID1A interacting with survivin in the docking experiments were also localized in the peptides with strongest binding fluorescent intensity to survivin in the peptide array. Based on the protein–protein docking analysis, we propose specific residues involved in the interactions between survivin and the ARID1A and SMARCC2 subunits, respectively (Fig. 3F1, F2. Supporting Figure S7A and S7B). Essentially, survivin binds the SANT-domain of SMARCC2 responsible for assembly, and stability of BRG1/SWI complexes [61, 62]. Canonical subunit ARID1A, and polybromo-specific subunit PHF10, contributed differently to the compositionally predicted and physically confirmed interactions with survivin. Interaction between ARID1A and survivin occurred within the region containing the LXXLL nuclear receptor recognition motif important for its role in gene regulation, cell biology and disease [63]. The amino acid residues of DPF2, ARID2, PBRM1, and BRD7 subunits had predicted interaction with survivin in the docking simulations, despite that their composition compatibility with survivin was low. Together, the combination of biomolecular interaction experiments through mass spectrometry, compositional analysis, peptide binding, and structural modelling, advocates in favor of survivin binding with BRG1/SWI complex through the DNA anchoring module. The differences in binding residues observed between protein docking experiments and peptide array experiments could be attributed to the structural rigidity of the BRG1/SWI complex structures which limit their ability to undergo conformational changes of the interaction regions. As a result, survivin is only able to interact with residues on the surface of these complexes.

Inhibition of survivin and JAK-STAT signaling caused cell cycle arrest and enhanced the DNA damage recognition

To examine colocalization of survivin and the BRG1/SWI complex, we performed immuno-histochemistry targeting BRG1 and survivin. We used BRG1 as a representative for the whole complex as BRG1 is the catalytic subunit and therefore a fixed component of all type of BRG1/SWI complexes. The convergence of survivin and BRG1 in nuclear and peri-nuclear area was visualized in THP1 cells known for abundant survivin expression. The analysis was performed on a pixel-by-pixel basis and presented 33–88% overlap in fluorescence produced by survivin and BRG1 (Fig. 4A, Supporting Figure S8).

Fig. 4

Survivin inhibition and JAKi treatment disrupts cell cycle progression and enhances DNA damage repair. A Colocalization of survivin (red) and BRG1 (yellow) in nucleus (blue) of THP1 cells, visualized by confocal microscopy at resolution 40X. Nuclear area is identified by Hoechst stain. Colocalization is calculated by overlap of fluorescence pixels, in ImageJ JACoB plugin. B Representative histogram of cell proliferation in THP1 cells treated with YM155 (0 nM and 10 nM). (right) Bar plot of THP1 cells proliferation cultured with survivin inhibitor YM155 (0–25 nM) for 72 h. Dilution of CellTrace Violet (CTV) proliferation dye was used to monitor generations of proliferating cells. D, E THP1 cells were treated with YM155 (20 nM), JAKi (50 µM), or sham (DMSO) for 24 h, fixed and stained for DNA damage using antibodies against BRG1 (red), γH2AX (green) and nuclear stain (blue). F γH2AX foci were counted in each nucleus of 90–200 cells per treatment using PepSlide Analyzer in ImageJ. P-values are obtained by Mann–Whitney U test. G Representative histogram of 7AAD+CD4+ cell distribution by phases of the cell cycle created by FlowJo software. Colored areas indicate G1 (blue), S (yellow) and G2 (green) phases. H Frequency distribution of 7AAD+ cells by cell cycle phases in CD4+ cells treated with IFNγ (50 ng/ml), and JAKi (10 µM) compared to sham (DMSO). P-values are obtained by Wilcoxon paired test. I Frequency distribution of 7AAD+ cells by cell cycle phases in CD4+ cells treated with YM155. P-values are obtained by Friedman test. J Heatmap of qPCR gene expression change by median log2FC in CD4+ cells treated with IFNγ (50 ng/ml), YM155 (10 nM) or JAKi (10 mM) for 48 h, compared to control (DMSO), measured by qPCR. Asterisk indicates p-values calculated by Wilcoxon paired test. * < 0.05. K A model of intervention in IFNγ signaling, IFN, survivin-BRG1 complex and BvCR dependent transcription of DNA damage response (DDR) genes by treatment with JAK-inhibitor (JAKi) and survivin inhibitor YM155. cis-RE, regulatory element

Based on the acquired results, we hypothesized that the DDR network in CD4+ cells was controlled by S + BvCR located within cis-RE of the human genome. To aid this function, survivin anchored the BRG1/SWI complex to the BvCR via transduction of the activating effects of IFNγ through JAK-STAT signaling.

To gather experimental verification of this model, we investigated the role of survivin in cell proliferation and cell cycle progress. For this, we cultured THP1 cells in presence of survivin inhibitor YM155 for 72 h. The CellTrace Violet dye was added to the culture medium to monitor new generations of proliferating cells by dye dilution. We found that survivin inhibition resulted in a decreased frequency of new generations of THP1 cells clearly seen in YM155 concentrations above 5 nM, while the part of maternal undivided cells increased (Fig. 4B, C). Additionally, survivin inhibition by YM155 promoted the accumulation of phosphorylated histone γH2AX+ foci which recognized the double-strand breaks in THP1 cells (Fig. 4D, F). A similar accumulation of γH2AX foci was observed after JAK/STAT inhibition (Fig. 4 E, F).

Next, we studied the cell cycle progress in CD4+ cells using the DNA binding fluorescent dye in flow cytometry (Fig. 4G). We observed that the progression of cell cycle was opposed by inhibition of survivin using YM155 causing a significant and dose dependent retention of cells in G1 phase and depleting cells in S and G2 phases, consistent with cell cycle arrest (Fig. 4I). Similar alterations in cell cycle progress were induced by inhibition of JAK/STAT signal downstream of IFNγ receptor (Fig. 4H). Therefore, both survivin and JAK/STAT inhibition disrupted proper cell cycle progression in CD4+ cells.

The IFNγ stimulation activated the functional link between survivin and energy supply in CD4+ cells by upregulating BIRC5 and repressing PFKFB3 (Fig. 4J) [25], the gene important for efficient homologous recombination [64]. Notably, the IFNγ-treated cells reduced mRNA levels of SMARCA4, SMARCC1, SMARCE1, which was associated with low transcription of the central DNA repair genes MRE11, MSH6, and FANCI connected to the H3K4me3-BvCR in CD4+ cells (Fig. 4J). This effect of IFNγ on DNA repair genes was abrogated by inhibition of survivin or JAK-STAT signaling (Fig. 4J).

Together, the in vitro studies validated the proposed hypothesis in which survivin exploited bivalency in aiding IFNγ signaling and BRG1/SWI complex in the DDR activity. Inhibition of IFNγ signaling and survivin was associated with cell cycle arrest and accumulation of damaged DNA (Fig. 4K). This triggered the subsequent activation of DNA repair genes, including transcription of the BRG1/SWI complex subunits.

BRG1 expression defined a specific phenotype of CD4+ cells in patients with rheumatoid arthritis

To investigate the impact of the BRG1/SWI complex in the survivin-dependent DDR control in autoimmune cells, we used transcriptome datasets of CD4+ cells of 24 RA patients (Supporting Fig. 1A). Guided by BRG1/SMARCA4 transcription, we found that both survivin and IFNγ were highly co-expressed in the BRG1hiCD4+ cells (Fig. 5A). Mapping of the differentially expressed genes to the DDR network revealed that the nodes of DNA repair, replication, and G1 arrest were upregulated in BRG1hi cells (Supporting Figure S9), pointing at unbalanced DDR control in these cells.

Fig. 5

Immunomodulating treatment affects DNA damage response in CD4+ cells of patients with rheumatoid arthritis. A Dot correlation plot of normalized mean expression of BRG1, BIRC5 and IFNG genes in CD4+ cells of patients with rheumatoid arthritis. Spearman’s rho values are indicated. B Venn diagram of differentially expressed genes (DEG) in BRG1hiCD4+ cells connected to BvCR and in IFNγ-treated CD4+ cells. Heatmap of Spearman’s rho correlation values of genes connected to BvCR in BRG1hi and BRG1lo cells identified by weighted correlation network analysis (WGCNA). C Heatmap of expression difference in T cell specific markers identified by RNA-seq in BRG1hiCD4+ cells and in CD4+ cells before and after treatment with abatacept (ABAT, n = 14), tocilizumab (TOCI, n = 6) and methotrexate (MTX, n = 28) and in CD4+ cells of JAKi-treated (n = 23) and untreated (n = 9) RA patients. Expression difference in CD4+ cells of treated and untreated patients was calculated by DESeq2. Nominal p-values are indicated. * < 0.05, ** < 0.01, *** < 0.001. D Venn diagram of DEG changed with treatment in the DNA damage response (DDR) network. DDR network map of DEG changed with treatment. Node size indicates the percentage of BRG1hi DEG. Node color indicates the percentage of DEG in the node. E Heatmap of expression difference in BRG1/SWI complex proteins in BRG1hiCD4+ cells and in CD4+ cells after treatment, by RNA-seq. Expression difference was calculated by DESeq2. Nominal p-values are indicated. * < 0.05, ** < 0.01, *** < 0.001. F Heatmap of expression difference in DDR network genes in BRG1hiCD4+ cells and in CD4+ cells after treatment

Next, we investigated if BvCR-connected genes abnormally regulated in BRG1hi cells were associated with the pathogenic phenotype in RA CD4+ cells. A total of 63% (846/1336) of the genes involved in at least one of the main pathways controlled by BvCR were differentially expressed in BRG1hiCD4+ cells, including the DDR pathway (Fig. 5B). To identify the functional modules of the genes co-expressed with BRG1, we applied the weighted gene correlation network analysis (WGCNA), asking if the hub genes that showed high co-expression in the BRG1hi and BRG1lo cells were functional in the BvCR-controlled biological processes. The WGCNA approach identified two modules, where the hub genes were positively associated with BRG1hi cells (n = 498) and with BRG1lo cells (n = 348) (Fig. 5B, Supporting Table T4). Several core subunits of the BRG1/SWI complex including BRG1/SMARCA4, SMARCC1, SMARCD1, SMARCE1 and canonical subunit ARID1A were accumulated in the module of BRG1hi cells (Fig. 5E). Additionally, this module showed the enrichment for the DNA repair pathway (GO:0006281, FDR = 2.8e-34) and cell cycle pathway (R-HSA-1640170, FDR = 1.3e-36) (Fig. 5B, F). Common cell cycle genes TP53 and CDKN1A controlled by BvCR were also highly upregulated in the BRG1hi cells while the ATM gene was repressed. The module of BRG1hi cells included the IFNγ-sensitive genes of PIAS4, MSH6, FANCI, MRE11, TELO2, and SMC3 (Fig. 5F) that changed H3K4me3 tag deposition after survivin inhibition (Fig. 2F, Supporting Figure S5). Several regulators of transcription including FOXP1, CBX4, PBX1, LEF1 were repressed and present in the module of BRG1lo cells (FDR = 1.35e-70).

To translate the profile of BRG1hiCD4+ cells into joint pathology in rheumatoid arthritis, we utilized characteristics of the inflammatory cells identified in RA synovia with a single cell resolution [65]. With focus on CD4+ cells, we investigated if the key markers of the synovial clusters were differentially expressed in the BRG1hiCD4+ cells and connected to BvCR. We found that transcriptome of BRG1hiCD4+ cells was enriched in the synovial cytotoxic GNLYhiHOPXhi cells expressing IFNG, TNF and GZMA, and the peripheral T-helper cells abundant in the immune check-point receptors PD1/PDCD1, CTLA4, and LAG3 (Fig. 5C) and TNF-superfamily receptors OX40/TNFRSF4, GITR/TNFRSF18, LAIR2, known for the ability to infiltrate inflamed tissue and to drive autoimmunity in RA [66,67,