Introduction
Rheumatoid arthritis (RA) is a chronic inflammatory disease affecting approximately 1% of adults [
1]. RA is associated with significant comorbidities most notably premature cardiovascular disease and insulin resistance [
2,
3] and is classically rhythmic, with hallmark symptoms such as pain and stiffness typically being worse in the early morning. This has been attributed, at least in part, to underlying circadian patterns in cytokine production, including IL-6 and IL-1β [
4,
5].
All human physiology is circadian; the external environment is sampled by light input through the retina, and neural transmission to the central clock in the suprachiasmatic nucleus (SCN) [
6]. The core cellular clock comprises a transcription-translation feedback loop in which heterodimers of BMAL1 and CLOCK drive expression of PER, and CRY proteins in one negative feedback arm, and the two REV-ERB paralogues in the other feedback arm [
7]. The period of the clock is affected by post-translational modifications of the feedback proteins, notably by phosphorylation [
8].
Perturbation of the circadian clock has a significant detrimental effect on immune response, whether by genetic targeting of core clock components, or imposition of phase shifts in light-dark [
9]. Indeed, shift work has been associated with risk of RA in women [
10]. Although numerous lines of evidence support a role for the immune system in affecting circadian clock function [
11], most observations have been made either in healthy animals or in response to acute inflammatory challenges, and there is a notable lack of studies in human subjects with prevalent, chronic inflammatory disease, such as RA.
Here, we use a systems biology approach to analyse circadian patterns of gene regulation, and protein phosphorylation in circulating immune cells from patients with rheumatoid arthritis. We selected these cells as they are accessible for serial sampling and include some of the effector cell types responsible for rheumatoid arthritis disease expression. Surprisingly, we found greater time-of-day variation in gene expression profile in RA patients than controls, a gain of rhythmic function, driven by changes in the phosphoproteome. Such an increase was also seen in serum lipid changes, particularly of the ceramide class; a gain in circadian rhythmicity was also seen in mouse experimental arthritis, driven by the acquisition of circadian rhythmicity of the ceramide synthases in the liver. Therefore, chronic joint inflammation serves as a circadian organiser, differentially coupling lipid-metabolic pathways to the core clock.
Discussion
A characteristic feature of RA is the change in disease activity over the day, measured by symptoms, and circulating markers of the inflammatory process such as IL-6 [
31,
32]. Indeed, there have been attempts to add circadian logic to therapy, with modified release prednisolone tablets entering the clinic [
33]. Embedding circadian logic into drug development, and biomarker studies of RA, requires an understanding of the circadian contribution to disease pathogenesis and expression, as an oscillating drug target phase may not be obvious from patient-reported symptoms.
We identified preserved behavioural, endocrine and immune cell circadian rhythmicity in patients with active RA. RA patients showed far more differences in gene expression and phosphopeptide abundance by time of day than controls. This difference was essentially lost in cells cultured ex vivo, implying the presence of a strong endogenous circadian entraining factor. Such molecules have been proposed to circulate in serum, but currently their identity remains unknown [
34,
35].
In the RA group, we identified major gene ontology groups involved in immune response as being regulated by time of day within the circulating immune cell compartment. Among these, there was strong enrichment for TLR4 signalling components. Analysing the genes differentially regulated at dawn between RA and control, we identified the enrichment for transcription factor binding motifs including STAT3, which is an important signalling mediator of IL-6 action. Further analysis of the immune cell phosphoproteome identified the separation of profiles by time of day only in the RA group, and among the phosphopeptides with a time-of-day signature, we inferred proline-directly serine phosphorylation, which marks MAP kinase action, and arginine-associated serine phosphorylation, which indicates PKA action. Further analysis of the phosphopeptides themselves identified changes in protein kinase phosphorylation, including multiple members of the MAP kinase family, and an interesting reciprocal phosphorylation site on protein kinase A (PRKACA T196/198). Involvement of the MAP kinases in time-of-day responses within the immune cell population may explain the enrichment of STAT3-binding sites within time-of-day-regulated genes, as MAP kinases phosphorylate STAT3. These data identify a MAP kinase-STAT3 circuit as time of day regulated in patients with active RA.
To extend our time-of-day observations, we analysed cellular responses to ex vivo activation. Here we found that the time-of-day effects were largely lost, but disease-specific differences remained. The inferred signalling networks activated included those served by multiple members of the MAP kinase family, again, and both STAT3 and STAT5. As the time-of-day differences were not seen in the ex vivo cell studies, we confined the analysis of disease-specific changes. Here, we identified a major change in the MAP kinase phosphosite on STAT3 (S727), again finding a disease-specific role for this transcription factor, a target of the IL-6-directed therapy in RA.
Ceramides play a critical role in regulating inflammation and may be produced locally within foci of inflammation, or elsewhere (e.g. the liver). In our CIA mouse model, we found similar patterns of rhythm serum ceramides as emerged in human RA. The CIA mouse model offers the advantage of no drug exposures, and controlled whole-life environmental exposures, thereby greatly increasing confidence that the ceramides result from the inflammatory process. Conservation of the newly rhythmic ceramides in RA would suggest functional importance in response to chronic inflammation. Ceramides have emerged as important drivers of insulin resistance in type II diabetes [
36,
37]. Their emergence here as coupled to the circadian clock in RA suggests a possible role mediating the adverse metabolic profile associated with accelerated atherosclerosis in RA [
38,
39].
We also found oscillating ceramide synthases surprisingly in the liver, and not in the joint, suggesting the liver, or other organs such as the gut [
40], as the origin for the rhythmic ceramides and thereby identifying a systemic coupling of ceramide metabolism to joint inflammation. Further analysis showed that IL-1β, a rhythmically expressed cytokine in CIA [
21], with peak serum concentration at ZT6, inhibits hepatic
CERS4, resulting in a near coincident nadir in serum
CERS4 product ceramide concentration. A similar circadian organising effect of disease was reported for lung cancer, mediated by IL-6 [
34], or in response to changes in the gut microbiome, mediated by polyamine metabolites [
41]. Therefore, our data show that in human chronic inflammatory disease, circadian mechanisms are co-opted to establish a new homeostatic oscillatory state, but whether this is a disease-sustaining step remains to be determined.