Comprehensive analysis of miRNA expression in T-cell subsets of rheumatoid arthritis patients reveals defined signatures of naive and memory Tregs

Abstract

Disturbed expression of microRNAs (miRNAs) in regulatory T cells (Tregs) leads to development of autoimmunity in experimental mouse models. However, the miRNA expression signature characterizing Tregs of autoimmune diseases, such as rheumatoid arthritis (RA) has not been determined yet. In this study, we have used a microarray approach to comprehensively analyze miRNA expression signatures of both naive Tregs (CD4+CD45RO-CD25++) and memory Tregs (CD4+CD45RO+CD25+++), as well as conventional naive (CD4+CD45RO-CD25-) and memory (CD4+CD45RO+CD25-) T cells (Tconvs) derived from peripheral blood of RA patients and matched healthy controls. Differential expression of selected miRNAs was validated by TaqMan-based quantitative reverse transcription-PCR. We found a positive correlation between increased expression of miR-451 in T cells of RA patients and disease activity score (DAS28), erythrocyte sedimentation rate levels and serum levels of interleukin-6. Moreover, we found characteristic, disease- and treatment-independent, global miRNA expression signatures defining naive Tregs, memory Tregs, naive Tconvs and memory Tconvs. The analysis allowed us to define miRNAs characteristic for a general naive phenotype (for example, miR-92a) and a general memory phenotype (for example, miR-21, miR-155). Importantly, the analysis allowed us to define miRNAs that are specifically expressed in both naive and memory Tregs, defining as such miRNA signature characterizing the Treg phenotype (that is, miR-146a, miR-3162, miR-1202, miR-1246 and miR-4281).

Figure 1

Purification of naive and memory Tconv and Treg populations. (a) Representative fluorescence-activated cell sorting (FACS) dot plot depicting the gating strategy for purification of CD4+, naive and memory Treg and Tconv subsets, based on the expression levels of CD45RO and CD25: I naive Tregs (CD45RO−CD25++), II memory Tregs (CD45RO+CD25+++), III memory Tconvs (CD45RO+CD25−) and IV naive Tconvs (CD45RO−CD25−). (b) Representative histogram depicting FACS analysis of Foxp3 protein expression in four CD4+ T-cell subsets. (c) Naive Tregs, memory Tregs, memory Tconvs and naive Tconvs were FACS purified from peripheral blood mononuclear cell (PBMC) of healthy donors, and Fosxp3 transcript levels were analyzed by qRT-PCR. Relative expression, normalized to the average of TATA box binding protein and RPII reference genes is shown. Lines indicate median expression values. Each dot represents a separate donor (n=4). Statistical analysis: repeated measures analysis of variance test with Bonferroni's multiple comparison post-test. ***P<0.0001.

Figure 2

Unsupervised hierarchical clustering of globally expressed miRNAs. (a) MiRNA expression signatures characterizing Tconvs and Tregs were assessed in 32 samples derived from six RA patients and two HC (Table 1, array) using miRNA microarray. A heatmap of the 121 human miRNAs detected in at least 50% of samples, ranked according to their expression in naive Tconvs is shown. Top legend (red to blue) depicts range of log expression values. Gray coloring indicates expression below the detection limit. Similarities between samples were assessed by one-way unsupervised hierarchical clustering (Pearson's correlation, complete linkage) using the Genesis software. (b) Enlargement of the hierarchical tree from a. Color coding corresponds to the four T-cell phenotypes. HC, healthy control; RA new, RA patients before the start of treatment with disease modifying anti-rheumatic drugs; RA MTX, RA patients treated for 3 months with MTX alone; RA MTX anti-TNF-α, RA patients treated with MTX and anti-TNFα monoclonal antibody; N, naive; M, memory.

Figure 3

Expression of miR-451 correlates with DAS28. (a) MiR-451 expression was assessed in Tconvs and Tregs, of six RA patients receiving different treatments, and two HC by miRNA microarray (Table 1, array). MiR-451 levels are higher in all T-cell subsets of patients treated with MTX. The horizontal lines indicate the median values. Statistical analysis: analysis of variance test with the Benjamini–Hochberg correction for multiple testing. (b) MiR-451 expression was assessed in Tconvs and Tregs in a new cohort of ten RA patients receiving different treatment and four HC by qRT-PCR (Table 1, qRT–PCR validation). Expression was normalized to the average of RNU44 and RNU48 reference genes and to a calibrator sample (RA-29 naive Tconv), which was set to 1. The horizontal lines indicate the median values. Statistical analysis: the Kruskal–Wallis test with Dunn's multiple comparison post-test. HC, healthy control; RA new, RA patients before the start of treatment with disease modifying anti-rheumatic drugs; RA MTX, RA patients treated for 3 months with MTX alone; RA MTX anti-TNFα, RA patients treated with MTX and anti-TNFα monoclonal antibody. (c–e) MiR-451 expression was assessed in naive Tconvs of 27 RA patients (Table 1, all patients) by qRT–PCR. Expression was normalized to the average of RNU44 and RNU48 reference genes and a calibrator sample (RA-29 naive Tconv), which was set to 1. Each dot represents one RA patient. (c) MiR-451 expression correlates positively with DAS28 in RA patients. (d) MiR-451 expression correlates positively with the erythrocytes sedimentation rate (ESR) in RA patients. (e) MiR-451 expression correlates positively with IL-6 levels detected in the serum of RA patients. *P<0.05, **P<0.01, ***P<0.0001.

Figure 4

MiRNA signatures characterizing T-cell subsets. (a) Heatmap (Pearson's correlation, median centering miRNAs, complete linkage) showing the miRNA expression signatures that characterize Tconvs and Tregs in 32 samples derived from six RA patients and two healthy controls, using miRNA microarray. Statistical analysis of 121 detected miRNAs revealed 42 miRNAs significantly differently expressed between T-cell subsets of eight donors (n=32 samples, Friedman's test, P<0.005, fold difference ⩾2 in at least one of six comparisons). Top legend (red to blue) depicts range of expression values (log values). Each T-cell subset is characterized by specific miRNA signature. (b) Eight miRNAs, differentially expressed between the four clusters as identified in a, were selected for qRT–PCR validation in a new cohort of donors (n=14, Table 1, qRT–PCR validation). Relative expression, normalized to the average of RNU44 and RNU48 reference genes is shown. Data are depicted as median values with interquartile range. Statistical analysis: Friedman's test with Dunn's multiple comparison post-test. *P<0.05, **P<0.01, ***P<0.0001.

Figure 5

Venn diagram depicting subset specificity of miRNAs. MiRNAs specific for each T-cell subset were selected based on an expression level of ⩾70% as depicted in Table 2.