Fibroblast-like synoviocytes orchestrate daily rhythmic inflammation in arthritis

Abstract

Rheumatoid arthritis is a chronic inflammatory disease that shows characteristic diurnal variation in symptom severity, where joint resident fibroblast-like synoviocytes (FLS) act as important mediators of arthritis pathology. We investigate the role of FLS circadian clock function in directing rhythmic joint inflammation in a murine model of inflammatory arthritis. We demonstrate FLS time-of-day-dependent gene expression is attenuated in arthritic joints, except for a subset of disease-modifying genes. The deletion of essential clock gene Bmal1 in FLS reduced susceptibility to collagen-induced arthritis but did not impact symptomatic severity in affected mice. Notably, FLS Bmal1 deletion resulted in loss of diurnal expression of disease-modulating genes across the joint, and elevated production of MMP3, a prognostic marker of joint damage in inflammatory arthritis. This work identifies the FLS circadian clock as an influential driver of daily oscillations in joint inflammation, and a potential regulator of destructive pathology in chronic inflammatory arthritis.

Keywords: circadian clock; fibroblast-like synoviocytes; inflammation; matrix metalloprotease; rheumatoid arthritis.

Figure 1.

FLS exhibit altered rhythmic activity during CIA. (a) Schematic representation of CIA experimental plan. (b) Total paw score and paw thickness (day of sample collection) measurements of mice used for FLS isolation. Mean ± s.e.m., Welch’s t‐test, n = 10–11 per condition. (c) FLS from CIA or naive mice were isolated at ZT4 or ZT16 and analysed by RNAseq (n = 5–6 per time per condition). DE analysis identified transcripts which are significantly up- or downregulated with disease and/or time. (d) Peak expression level (left) and scaled variance (centre) were compared for the top 1000 significantly differentially expressed genes in naive or CIA FLS, and the magnitude of time-of-day difference calculated (right; difference in mean expression level at ZT4 and ZT16 after z-scoring for top 1000 genes rhythmic in naive FLS). Error bars show median ± interquartile range, distributions compared by Welch’s t‐test. (e) Nos2 and Mmp3 show time-of-day-dependent expression in FLS from inflamed joints. Mean ± s.e.m., two-way ANOVA treatment effect and time-dependent differences are indicated. n = 5–6. (f) FLS core clock gene expression in cells isolated from inflamed paws. Mean ± s.e.m., two-way ANOVA. (g) Four-way Venn comparison of time-of-day-dependent gene expression between ZT4 and ZT16 timepoints in FLS RNAseq and whole joint RNAseq datasets [11]. (h) Comparison of cytokine/chemokine expression changes with CIA in joints and isolated FLS. Of 92 transcripts considered (CCL, CXCL, interleukin, interferon, TNF and TGF families), 81 were detected and 58 were differentially expressed with CIA in at least one sample type. Fold changes (FC) in expression were calculated from RNAseq data at ZT4; bubble size represents the significance of DE between CIA and naive FLS.

Figure 2.

Loss of molecular clock function alters FLS transcriptional profile. (a,b) Quantification of BMAL1 ablation at transcript (a) and protein (b) levels for Cre− and Cre+ mice. Mean ± s.e.m., Welch’s t‐test, n = 3–7 per genotype. (c) FLS sorted into CD90.2− and CD90.2+ subpopulations were analysed to confirm LL versus SLL identity (Prg4 and Thy1 marker gene expression, respectively; left, centre) and comparable expression of Cre recombinase (right). Gene expression is presented as mean ± s.e.m., RM two-way ANOVA, post hoc subpopulation comparison, n = 4 independent replicates. (d) Gene expression of FLS isolated from joints was analysed using split pool ligation-based transcriptome sequencing (SPLIT-seq). Cells were clustered (left) and subpopulation identity determined based on cluster-specific expression of highly variable genes, identifying six FLS subpopulations (as well as chondrocytes, osteogenic cells and minor populations of other cell types), with no significant difference in population distribution between genotypes (right); two-way ANOVA, n = 3 per genotype. (e) Volcano plot showing differential gene expression between naive Cre− (grey) and Cre+ (pink) samples. Comparative expression analysis using edgeR considered pseudo-bulk populations of cells from clusters 1–8. (f) Analysis of genes showing significant DE between genotypes identified enriched KEGG pathways (adj. p < 0.05, dashed line). Fraction of pathway indicates the proportion of genes associated with a pathway which showed significant DE between genotypes. (g) Circadian (left) and collagen/cell matrix-associated genes (right) are differentially expressed between genotypes (FDR < 0.05, dashed line). DE is presented as the absolute value of log₂ fold change, where grey indicates increased expression in Cre− mice and pink indicates increased expression in Cre+ mice. HVG, highly variable gene.

Figure 3.

Mice lacking Bmal1 expression in FLS show altered CIA development. (a) Development of symptoms in Cre+ mice was significantly reduced compared with littermate Cre− controls. Fisher’s exact test; error bars indicate 95% confidence interval. (b) Quantification of IgG1 and IgG2a anticollagen antibody response in plasma from naive, asymptomatic and symptomatic CIA mice. Two-way ANOVA, post hoc treatment comparison, n = 4–18 per condition. (c) Daily paw score (top) and thickness measurement (bottom) found no difference in disease severity between mice that develop symptoms. Two-way ANOVA, mean ± s.e.m., n = 17–26 mice/genotype. (d,e,f) Flow cytometry analysis of immune cell infiltration into joints with disease, comparing total immune cell proportion (d) and cell identity (e) between genotypes. Flow cytometry analysis of FLS identity (f) comparing LL versus SLL between disease and time-of-day. Open circles indicate ZT8 sample collection, filled circles indicate ZT20 sample collection. Two-way ANOVA, post hoc treatment (d, e) or genotype (f) comparison, n = 3–8 per time per condition.

Figure 4.

Mice lacking Bmal1 expression in FLS lose rhythmicity of disease effectors in joints. (a) Inflammatory gene expression in naive and CIA joints from Cre− and Cre+ mice. Three-way ANOVA results and time-dependent differences in gene expression are indicated, n = 4–7 per condition. (b) Administration of an MMP activity-dependent fluorescent probe identified localised enzymatic activity in inflamed (but not asymptomatic) CIA paws. (c) MMP3 level in plasma from arthritic mice across time. RM two-way ANOVA with post hoc time comparison, n = 7–10 per genotype.