In the present study, we conducted transcriptome analysis of B cell subpopulations. First, we found that expression of LINC00487 was upregulated in B cell subsets derived from patients with pSS. Further, LINC00487 expression significantly correlated with disease activity and was induced by IFNα stimulation. Second, using WGCNA, we identified several key networks and hub genes in the B cells of patients with pSS.
IFN signalling is a central component of the pathogenesis of pSS [28]. Type 1 IFN signalling plays a crucial role in the development of autoreactive B cells in a mouse model of autoimmune disease [5]. Principal component analysis using gene sets for WGCNA demonstrated that Bm1/naïve B cells of pSS were clearly distinguished from those of HC and showed a trend of expression patterns closer to the more highly differentiated subset. These results suggest that molecular dysfunction, such as disruption of peripheral tolerance, might start at an early stage in the maturation of B cells. Indeed, the frequencies of naïve B cells expressing autoreactive antibodies are significantly increased in patients with pSS [29].
Interestingly, expression of the HLA class II gene was upregulated. GWAS studies of pSS found associations of HLA-DQA1 and HLA-DQB1 loci [6,7]. Further, IFNα induces the expression of HLA-DQA1 [30]. In patients with pSS, HLA-DQA1 and HLA-DQB1 alleles are associated with higher concentrations of anti-SSA and SSB antibodies [31]. These findings suggest that aberrant interactions among IFN signalling and HLA class II genes may trigger the breakdown of B cell tolerance, leading to the development of pSS.
LINC00487 was upregulated in all B cell subsets of pSS and its expression significantly correlated with the disease activity scores of pSS and ISGs, although ESSDAI score of our patients skewed toward the low end. Moreover, we found that IFNα was an upstream regulator of LINC00487 in B cells. The sequence of LINC00487, which belongs to the class of long intergenic non-coding RNAs and resides on human chromosome 2, is atypically long (> 40,000 bases). Although LINC00487 is one of the hub genes in the normal development of human B cells [32], many of its properties, including its function, are unexplained. Further, there is no ortholog or paralog of this gene in species other than humans that may provide clues to its function in human cells. However, according to AceView, one of three transcriptional variants derived from LINC00487 has potential to encode a protein in silico prediction [33].
Four genomic locations are considered candidate enhancers of LINC00487 transcription. The targets of the regulators of LINC00487 overlap with those of other ISGs [34]. Moreover, referring to the public transcriptome database of microarray analyses of healthy humans, expression of LINC00487 is higher specifically in centroblasts and centrocytes of the germinal centre [35]. IFNα promotes the autoreactivity of B cells via germinal centre pathways [5]. Further, LINC00487 expression is upregulated in the subgroup of diffuse large B-cell lymphoma with molecular characteristics of germinal centre B cells, compared with other subgroups, and is associated with the efficacy of B cell depletion therapy [36]. Therefore, our study suggests that upregulation of LINC00487 expression in all B cell subsets may reflect or regulate the enrichment of a germinal centre-like reaction by IFNα from an early stage of B cell development, leading to B cell autoreactivity in patients with pSS.
Gene co-expression analysis revealed an aberrant network in B cell subpopulations. The top significant upstream regulator of the grey module of pSS, which was associated with clinical disease activity score and enriched in the early stage of B cell development in pSS, was the gene encoding T-cell acute lymphocytic leukemia protein 1 (TAL1). B cell development is stringently controlled by stage-specific transcription factors. The transcription factor TAL1 regulates genes such as IKAROS family zinc finger 3 (IKZF3), which is one of hub genes in the grey module of pSS (Figure 5C). IKZF3 is a lineage-specific transcription factor that is important in the regulation of B cell proliferation and development [37].
In the pre-GC B cells-associated module of pSS, SOX4 was identified as an upstream regulator and a hub gene. In mice, SOX4 regulates the differentiation of early-stage B cells by activating the expression of Rag1 and Rag2 [38]. Further, SOX4 contributes to the formation of ectopic lymphoid-like structures via promoting CXCL13, which is ligand of CXCR5 on naïve B cells and critical for migration into light zone of germinal centre undergoing somatic hypermutation, production from PD-1hiCXCR5–CD4+ T cells [39]. Additionally, in proteome analysis, CXCL13 positively correlates with the disease activity score and serum IgG levels of patients with pSS [19]. Further study is needed to explore the role of SOX4 in mature B cells.
Further, we identified miR-21 as an upstream regulator of the yellow module of pSS. Although miR-21 is upregulated in peripheral blood mononuclear cells of pSS [40], the present study is the first to propose its involvement in pSS via B cell dysregulation. Interestingly, miR-21 regulates the immune response of memory T cells via induction of transcription networks, such as SOX4 [41], implying that the interaction between miR-21 and SOX4 may also affect function of B cells.
pSS is enormously heterogeneous disease from the molecular point of view. In WGCNA, despite filtering out genes with high variation in each cell subpopulation, module expression showed the strong variation among samples in the same group, in particular Bm1 subset (Figure 4B). One of the reasons for the high variability seen in Bm1 subset may be related to the fact that some cells with the CD38-IgD+ express CD27 [20]. Namely, the CD38-IgD+ B cell subpopulation includes both naive Bm1 cells and IgD+ memory B cells. However, the finding in the current study suggested that there might be different regulatory mechanisms within each subgroup, although it was information that couldn’t be analyzed due to the limited sample size.
The current study suffers from several limitations. First, patients in our cohort have low disease activity/severity. Because there were few patients with high disease activity before immunosuppressive treatment, it was difficult to include such patients in this study. Therefore, correlation between expression of LINC00487 and disease activity score is needed to be validated in other cohort including patients with high ESSDAI scores. Second, healthy controls were younger than the pSS patients, and this could be a confounding variable. Third, sample size was limited. To overcome limitation about small sample size, we validated by qPCR using another cohort, supporting results derived from microarray analysis. However, regarding WGCNA, we couldn’t include subjects enough to replicate by qPCR.