To define the expression landscape of the CHASERR–CHD2 axis across human immune cells, we analyzed single-cell RNA sequencing (scRNA-seq) data from the Asian Immune Diversity Atlas (AIDA), a comprehensive dataset of 1,265,624 peripheral blood mononuclear cells (PBMCs) (10). We found that CHASERR expression was significantly higher in T cells than in B or NK cells (adjusted p < 0.05; Supplementary File 2). Among T cell subsets, the highest expression levels were detected in double-negative regulatory T cells, naïve T cells, and regulatory T cells (Treg), while the lowest expression was observed in gamma-delta T cells and mucosal-associated invariant T (MAIT) cells (Figures 1A–C). Further stratification revealed that CHASERR expression was higher in central memory (TCM) compared to effector memory (TEM) T cells in both CD4+ and CD8+ lineages (Figure 1D). Among monocytes, CHASERR was more highly expressed in CD16+ than in CD14+ subsets.

As expected from its known repressive role, CHD2 expression often exhibited an inverse relationship with CHASERR—most notably in Tregs and double-negative T cells, where CHD2 levels were significantly lower (Supplementary File 1: Supplementary Figures S1A–D). However, in many other cell types, including naïve CD4+ and CD8+ T cells, CHD2 expression mirrored that of CHASERR, suggesting both repressive and co-regulated modes of interaction depending on cellular context. Indeed, although CHASERR and CHD2 were positively correlated across most cell types, their divergent behavior in Tregs and monocytes implies complex, cell-type-specific regulation.

We next sought to validate these findings experimentally. Using fluorescence-activated cell sorting (FACS), we isolated populations of primary human CD4+ T cell subsets—naïve, TCM, TEM, TEMRA, Th1, Th1-17, Th17, and Tregs—and quantified CHASERR and CHD2 expression via qPCR (Figure 2A). Consistent with the AIDA data, both genes were most highly expressed in naïve T cells. We also confirmed divergent expression patterns: CHASERR was elevated in TCM relative to TEM cells, while CHD2 showed the opposite trend. Among helper T subsets, CHASERR was highest in Th17 and Tregs, whereas CHD2 was highly expressed in Th1 and Th17 but reduced in Tregs.

To ensure robustness, we turned to two independent public datasets. Analysis of the DICE database (11), which contains bulk RNA-seq from 13 immune cell types, confirmed that CHASERR is most highly expressed in naïve CD4+ T cells and naïve Tregs, with the lowest levels in activated T cells (Figures 2B, C). Notably, CHD2 showed less heterogeneity across T cell subsets. We also analyzed scRNA-seq data from CD4+ T cells in healthy and autoimmune donors (12), which again revealed elevated CHASERR expression in naïve and Treg populations, with the highest levels in activated Tregs (Figure 2D, Supplementary File 1: Supplementary Figures S2C, D).

Together, these results firmly establish that CHASERR is most abundant in naïve, regulatory, and central memory T cells, and reveal both concordant and antagonistic expression patterns with its target, CHD2, across immune cell subtypes.

To further investigate CHASERR expression patterns across CD4+ T cell subtypes, we analyzed bulk RNA-seq data from naïve and memory CD4+ T cells subjected to five distinct polarization conditions using different cytokine combinations (13). After 16 hours of polarization, CHASERR expression decreased in both naïve and memory T cells across all cytokine conditions. However, following five days of polarization, naïve T cells showed increased CHASERR expression in Treg and TH17 cells compared to both resting cells and cytokine-free controls (Figure 3A). CHD2 expression exhibited a similar pattern in polarized naïve T cells at day 5, though in iTregs and Th17 cells, CHD2 levels did not exceed those observed in resting cells. The marked variability in CHASERR and CHD2 expression across polarized naïve CD4+ T cell subtypes, coupled with consistently low expression in memory cells, suggests that their transcription may be repressed in memory T cells while remaining plastic in naïve populations.

We next examined whether CHASERR and CHD2 might be regulated by FOXP3, a master transcriptional regulator of Treg cell differentiation (14). In human Treg cells, FOXP3 ChIP-seq analysis revealed a binding peak within 2,000 base pairs of the CHASERR transcription start site (TSS) (Supplementary File 1: Supplementary Figure S3A). In mouse Treg cells, Foxp3 bound directly to both the Chd2 and Chaserr promoter regions (Supplementary File 1: Supplementary Figure S5B). Additionally, Foxp1, another Foxp family member essential for T cell quiescence and differentiation, was found to occupy the promoter regions of both Chaserr and Chd2 in mouse Treg cells and spleen CD8+ T cells (Supplementary File 1: Supplementary Figure S3B).

The elevated expression of both CHASERR and CHD2 in naïve compared to effector T cells prompted us to investigate their temporal dynamics during T cell activation. We analyzed six bulk RNA-seq datasets from a meta-analysis by Rade et al. (15), in which T cells were activated via anti-CD3/anti-CD28 antibodies.

Notably, CHASERR expression began to decline immediately following T cell activation (Figure 3B, Supplementary File 1: Supplementary Figure S4). In contrast, CHD2 expression exhibited a transient increase, peaking approximately one hour post-activation before gradually declining (Supplementary File 1: Supplementary Figure S5). This sequential pattern—CHASERR downregulation followed by transient CHD2 upregulation—suggests that the CHASERR-CHD2 axis may serve as an early regulatory module in T cell activation.

To identify potential functional targets of this axis, we examined whether genes altered upon CHASERR knockdown in fibroblasts (FANTOM6 project (16)) were also differentially expressed during T cell activation (15). Several genes downregulated after CHASERR knockdown — including RICTOR, RBL2, and USP30 — were also suppressed during T cell activation. Conversely, genes upregulated upon knockdown, such as HN1 and OSBPL3, showed increased expression during activation (Supplementary File 1: Supplementary Figure S6). These overlapping signatures suggest that CHASERR may regulate a conserved set of targets across cell types, including T cells.

The specific functions of these genes further support their role in promoting T cell activation: upregulation of HN1, a regulator of cell proliferation and microtubule stability (17, 18), alongside downregulation of the cell cycle inhibitor RBL2 (19), likely facilitates exit from quiescence, proliferation, and cytoskeletal remodeling essential for T cell function.

To explore potential functional relationships involving CHASERR in T cells, we performed co-expression analysis using metacells—aggregated cell states—generated from the AIDA dataset to address single-cell data sparsity. The CHASERR expression profile across metacells appeared generally consistent with patterns observed at the single-cell level (Figures 4A, B, Supplementary File 1: Supplementary Figure S7).

We identified genes showing correlation with CHASERR (Spearman’s |ρ| > 0.5, adjusted p < 0.05) across T cell subtypes (Supplementary File 3). Genes positively correlated with CHASERR included several regulators associated with quiescence, such as BTG1, BTG2 (20), TOB1 (21), FOXP1 (22), and ZFP36L2 (23). Genes negatively correlated with CHASERR in memory and effector T cells showed some enrichment in pathways related to effector functions, including proliferation, migration, and cytotoxicity (Figure 4C, Supplementary File 4). Some of these pathways included CORO1A (coronin-1A), an actin-binding protein that has been reported to facilitate Ca²⁺ mobilization (24) and may contribute to T cell survival (25).

We also computed Gene Ontology (GO) Biological Process activity scores for each metacell (Supplementary File 5). The T cell activation pathway showed a negative correlation with CHASERR expression (Figure 4D), which appears consistent with the pattern of higher CHASERR levels in quiescent naïve T cells and their reduction upon activation.

Along with the observed downregulation of CHASERR following T cell activation and the presence of cell cycle regulators among CHASERR-sensitive genes, these results may suggest a potential role for CHASERR in maintaining T cell quiescence. Its downregulation appears to coincide with the acquisition of effector functions, though further investigation would be needed to establish causal relationships.

To explore potential functional consequences of CHASERR loss across T cell subtypes, we examined enrichment patterns for genes differentially expressed in FANTOM6 CHASERR knockdown experiments (FDR ¡ 0.05) within metacells from the AIDA dataset (Figure 4E).

Genes upregulated following CHASERR knockdown tended to show higher enrichment scores in cytotoxic T cell populations compared to naïve T cells. Meanwhile, genes downregulated after CHASERR knockdown (particularly in the ASO G0272888 AD 07 experiment) appeared most enriched in CD8+ naïve T cells. This pattern suggests a potential inverse relationship between CHASERR expression and the enrichment scores of genes responsive to its knockdown.

These observations raise the possibility that genes affected by CHASERR knockdown might represent functional targets in T cells, and that loss of CHASERR could potentially contribute to transcriptional states associated with cytotoxic and effector T cell function. However, further validation in T cell models would be needed to confirm this relationship.

To identify potential therapeutic compounds that could counteract the effects of CHASERR loss, we screened the LINCS database using the transcriptional signature from FANTOM6 CHASERR knockdown experiments (ASO G0272888 AD 07; Supplementary File 6). Among the top-ranked candidates, we selected cyclosporine A (CsA) — a known immunosuppressant that forms a complex with cyclophilin to inhibit calcineurin, thereby preventing nuclear translocation of NFAT and T cell activation (26) - for experimental validation.

We first tested whether CsA could modulate CHD2 expression in the context of CHASERR deficiency. Following CHASERR knockdown in primary human fibroblasts, we observed the expected increase in CHD2 expression by qPCR (Figure 5A), consistent with previous reports of CHASERR-mediated repression (5). While CsA treatment did not restore CHASERR expression itself — likely due to persistent ASO-mediated knockdown — it significantly reduced CHD2 expression in CHASERR-deficient cells. This provides only a preliminary indication that CsA may partially compensate for CHASERR loss by normalizing CHD2 levels through a mechanism independent of CHASERR re-expression in human.

To validate these findings in mouse model, we analyzed scRNA-seq data from lymph nodes of mice with experimental autoimmune uveitis (EAU) treated with CsA (27). Consistent with human data, Chaserr expression was highest in CD4+ naïve and regulatory T cells and lowest in CD8+ cytotoxic T cells (Figures 5B–D). We also confirmed conserved positive correlations between Chaserr and key quiescence factors Foxp1 and Zfp36l2 (adjusted p < 0.05; Supplementary File 7), and Gene Ontology analysis revealed enrichment of Chaserr-correlated genes in T cell activation pathways, including Satb1—a regulator of chromatin remodeling in T cells (28) (Figure 5E, Supplementary File 8).

In the EAU model, Chaserr expression was reduced across T cell subsets, and CsA treatment further decreased its expression (Figure 5F). In contrast, Chd2 levels were elevated in uveitis but reduced following CsA treatment (Figure 5G). These results align with our in vitro data and suggest that CsA can counteract Chd2 overexpression in settings of CHASERR deficiency, even in an active autoimmune context.

While CsA does not restore CHASERR expression, our findings indicate that it may mitigate key transcriptional consequences of CHASERR loss, particularly dysregulation of CHD2. This supports further investigation of calcineurin-NFAT pathway inhibitors as potential therapeutic strategies for conditions linked to CHASERR deficiency.

To investigate the expression dynamics of CHASERR and CHD2 across immune cell populations, we analyzed 10 datasets:

To identify and sort naive and memory CD4+ T cell subsets, PBMCs (510⁶ cells in 100) were stained with a cocktail of fluorescently labeled antibodies: CD3-BB700 (clone OKT3, 1:20 dilution), CD4-BV786 (clone OKT4, 1:20), CD45RA-BV480 (clone HI100, 1:20), and CD197 (CCR7, clone G043H7, 1:20). Naïve CD4+ T cells were defined as CD45RACCR7, central memory (CM) as CD45RA-CCR7-, effector memory (EM) as CD45RA-CCR7-, and terminally differentiated effector memory (TEMRA) as CD45RA+CCR7-. A representative gating strategy is provided in Supplementary File 1: Supplementary Figures S8A1–A6.

For the sorting of Treg and Th1/Th17 subsets, PBMCs (510⁶ cells in 100) were stained with a separate antibody panel: CD3-BB700 (clone OKT3, 1:20), CD4-BV786 (clone OKT4, 1:20), CD183-BV421 (CXCR3, clone G025H7, 1:20), CD196-BB515 (CCR6, clone 11A9, 1:20), CD127-AF647 (clone HIL-7R M21, 1:20), and CD25-PE (clone M-A251, 1:20). The gating strategy is shown in Supplementary File 1: Supplementary Figures S9A1–A5.

All antibodies were obtained from BD Biosciences (USA), except CD196 and CD183, which were sourced from BioLegend (USA). Cell sorting was performed on a BD FACS Aria III sorter equipped with 405, 488, 561, and 633 nm lasers. Compensation was carried out using anti-mouse Ig compensation beads (BD Biosciences) stained with respective antibodies, and compensation matrices were automatically calculated using BD FACSDiva software (v9.0.1). Cells were sorted in purity mode into 5 mL tubes. Postsort validation was performed by reanalyzing 100 µL of sorted cells mixed with 100 PBS (Supplementary File 1: Supplementary Figures S8B1–B4, S9B1–B4). Final sorted cell counts were as follows: 336,500 naive cells, 276,000 CM cells, 170,500 EM cells, 33,590 TEMRA cells, approximately 50,000 Tregs, 177,000 Th1 cells, 57,500 Th17 cells, and 101,780 Th1–17 cells.

RNA was extracted by direct cell lysis with ExtractRNA reagent (Evrogen). 5 ng of total RNA was Total RNA was extracted from sorted CD4+ T cell populations by direct lysis using ExtractRNA reagent (Evrogen). For cDNA synthesis, 5 ng of total RNA was reverse transcribed using the MMLV RT kit with dT20 primer (Evrogen). Quantitative real-time PCR was performed using 5X SYBR Green master mix (Evrogen). Gene expression levels were normalized to the housekeeping gene B2M and analyzed using the ΔΔCt method (36). All primer sequences are provided in Supplementary File 9.

All scRNA-seq data were processed using Seurat v4.3.3 (37). For the cyclosporine dataset, reads were aligned to the GRCm39 reference genome using CellRanger v8.0.1 (https://support.10xgenomics.com) with Ensembl v111 annotation. Quality control filtering excluded cells with <200 or >2500 detected genes, total UMI counts <500 or >10,000, or mitochondrial gene content >15%. Genes detected in <1% of cells were removed. Count normalization was performed using the LogNormalize function in Seurat. Following normalization and scaling, we performed nearest-neighbor analysis, Louvain clustering, and batch correction via Harmony integration (38). T cell clusters were identified by expression of Cd3e, Cd3d, and Cd3g, with further subclustering using established marker genes (Supplementary File 1: Supplementary Figure S10) (39).

Differential expression analysis for the AIDA dataset was performed by comparing each T cell type against all other cell types from the layer 1 annotation using the FindMarkers function in Seurat. We applied the Wilcoxon rank-sum test with return.thresh = 1 and calculated adjusted p-values using Bonferroni correction.

For subtype-specific analyses within T cells in both the AIDA and CD4+ T datasets, we identified differentially expressed genes between each T cell subtype and all other T cell subtypes using the FindAllMarkers function in Seurat with Wilcoxon rank-sum tests (return.thresh = 1) and Bonferroni-adjusted p-values. Additionally, we performed differential expression analysis on pseudobulk expression profiles aggregated by donor and cell type using the same statistical approach.

Bulk RNA-seq differential expression results for the DICE database were obtained from the DICE Portal (https://dice-database.org/) (11). These analyses were performed using DESeq2 (40), and we retained only results with an adjusted p-value < 0.05 (Benjamini Hochberg method).

Differential expression analysis was conducted on T cell activation datasets (GSE90569, GSE96538, GSE94396, GSE52260, GSE140244). Gene expression at the 1-hour and 2-hour time points was compared against the 0-hour control. The analysis was performed using DESeq2 (40), and p-values were adjusted for multiple testing using the Benjamini-Hochberg method.

For the cyclosporine treatment dataset, we compared gene expression between the control and uveitis groups, as well as between the uveitis and cyclosporine-treated groups, within each annotated cell type using the FindMarkers function in Seurat (Wilcoxon rank-sum test, adjusted p-value < 0.05).

Metacells were constructed from T cells in the AIDA dataset using hdWGCNA v0.4.0 (41). Libraries with mean CHASERR expression in the lowest 5% were excluded from analysis. Metacells were generated by mean aggregation of normalized counts using k = 75 nearest neighbors, with a maximum of 10 shared cells between any two metacells. Harmony integration was applied using the library_uuid covariate. The final analysis excluded generalized ‘CD4-positive, alpha-beta T cell’ and ‘CD8-positive, alpha-beta T cell’ annotations to focus on specific T cell subtypes.

Pairwise gene expression correlations were computed using Spearman’s rank correlation on aggregated, normalized expression vectors. Correlations with |ρ| > 0.5 and Bonferroni-adjusted p-values < 0.05 were retained for further analysis. Genes showing significant correlations underwent Gene Ontology enrichment analysis (Biological Process category) using clusterProfiler v4.14.3 (42), with all detected genes serving as the background set and a significance threshold of q-value < 0.05.

Module scores for genes upregulated and downregulated in FANTOM6 CHASERR knockdown experiments were calculated using the AddModuleScore function in Seurat. Gene Set Variation Analysis (GSVA) v2.0.5 (43) was used to compute GO Biological Process pathway scores for each metacell. Spearman correlations between GSVA pathway scores and CHASERR expression levels were calculated, retaining only results with Bonferroni-adjusted p-values < 0.05.

We processed Foxp1 ChIP-seq data from mouse spleen CD8+ T cells from raw FASTQ files using the nf-core/chipseq pipeline (44). Reads were aligned to the mm10 reference genome using BWA v0.7.17-r1188 (45). Peak calling was performed with MACS3 v3.0.3 (46) (q-value < 0.05) for each biological replicate, and replicates were merged using the Irreproducible Discovery Rate (IDR) framework v2.0.4.2 (47). FOXP3 ChIP-seq data from human Treg cells (GSE43119) were processed using the same nf-core/chipseq pipeline parameters, with alignment to the hg38 genome. ChIP-seq peaks were visualized using the IGV browser (https://igv.org/app/).

To identify compounds capable of counteracting the transcriptional signature of CHASERR knockdown, we submitted differentially expressed genes (FDR < 0.05) from the FANTOM6 CHASERR knockdown experiment (ASO_G0272888_AD_07) to the L1000FWD platform (https://maayanlab.cloud/l1000fwd/) (48). Candidate drugs were prioritized based on combined scores, with emphasis on compounds whose gene expression signatures were inversely correlated with the CHASERR knockdown phenotype.

Human dermal fibroblasts (HDFb d75, adult female) were obtained from the Cell Culture Collection (IDB RAS, Russia). Cells between passages 5–10 were maintained at 37 °C with 5% CO 2 in DMEM/F12 medium (Gibco) supplemented with 1×Penicillin-Streptomycin (Paneco) and 10% v/v FBS (BioWest). For lncRNA knockdown, cells at 75–80% confluency were transfected with ASOs targeting CHASERR using FectoMEM transfection medium (Bioinnlabs) and GenJect40 transfection reagent (Molecta) at a final concentration of 10 pmol ASO per cm² growth surface. Total RNA was isolated by direct cell lysis with ExtractRNA reagent (Evrogen). Reverse transcription was performed on 50 ng of total RNA using MMLV RT kit with dT 20 primer (Evrogen). Quantitative PCR was carried out using 5× SYBR Green master mix (Evrogen). Three biological replicates were analyzed to assess cyclosporine effects—expression was measured at 20 hours post-treatment in the first experiment and at 24 hours in subsequent replicates. All experiments were analyzed independently using the ΔΔCt method (36) with normalization to PPIA. ASO sequences and qPCR primers are provided in Supplementary File 9.