Background

Despite a vast array of pharmacologic agents available to treat rheumatoid arthritis (RA), management is often complicated by insufficient treatment response, drug toxicity and contraindications, poor access to care and/or medications, and/or damage that predates medical intervention. These barriers lead to or are accompanied by systemic manifestations, disease-associated co-morbidities, chronic pain, physical inactivity, dysmobility, and poor physical function. Thus, further advances in RA care require identification of factors contributing to persistent deficiencies in quality of life and physical function, despite access to excellent anti-rheumatic medications.

Importantly, inactivity and muscle wasting are two important contributors to RA-related morbidity and mortality. Approximately half of patients with RA do not perform even a single bout of weekly physical exercise [1]. The sedentary lifestyle common to patients with RA gives rise to physical deconditioning and muscle atrophy, both of which are associated with osteoporosis, impaired immune function, glucose intolerance, insulin resistance, loss of independence, and increased mortality [2].

In addition to physical inactivity, other factors that likewise promote muscle loss and disability in patients with RA include inadequate protein ingestion, glucocorticoid treatment, and pro-inflammatory cytokines, all resulting in reduced myocyte protein synthesis and increased protein degradation [2, 3]. Inflammation can impact normal muscle turnover and responses to injury, both of which require an exquisitely coordinated remodeling process involving activation, proliferation and differentiation of muscle stem cells—also known as satellite cells. These processes are mediated largely by signals from intramuscular immune cells: neutrophils, regulatory T cells, pro-inflammatory M1 macrophages, and anti-inflammatory M2 macrophages.

The established roles of inflammation in both skeletal muscle remodeling and RA pathophysiology raise obvious questions regarding the potential interplay between muscle dysfunction and RA morbidity. Whereas the link between pro-inflammatory cytokines and muscle dysfunction has been investigated intensely in the context of diseases such as diabetes and cancer cachexia, this topic has remained surprisingly unexplored in RA. In the current study we sought to identify molecular perturbations in muscle specimens from individuals with RA, and to test the hypothesis that skeletal muscle inflammatory markers and derangements in tissue remodeling might contribute to metabolic decline and disability in these patients. Herein, we report that disease-activity-related muscle inflammatory markers are related to physical inactivity, and moreover, that disrupted skeletal muscle repair processes are associated with greater disability. These findings support a model in which skeletal muscle deterioration contributes to a vicious cycle of disease activity, muscle inflammatory signaling and disrupted remodeling, physical inactivity, and disability in patients with RA.

Methods

Design and participants

This was a cross-sectional investigation of individuals with RA and matched controls collected from the Durham, NC area. The RA group met the following criteria: (1) RA diagnosis meeting American College of Rheumatology 1987 criteria [4]; (2) seropositive disease (positive rheumatoid factor or anti-cyclic citrullinated peptide) or evidence of erosions on hand or foot imaging; (3) no medication changes within the three months prior to study enrollment; and (4) daily prednisone use ≤5 mg. Healthy participants without a diagnosis of RA, without joint pain, and without joint swelling lasting more than a week were matched to individual participants with RA by gender, race, age within 3 years, and body mass index (BMI) within 3 kg/m2. Exclusions included current pregnancy, type 2 diabetes mellitus, and known coronary artery disease. Further specific details on questionnaires and measurement protocols have previously been described [5]. This study was in compliance with the Helsinki Declaration and was approved by the Duke University Institutional Review Board.

Assessments of both groups included questionnaires, physical exams for disease status, fasting blood collection, intravenous glucose tolerance tests for insulin sensitivity, 7 days of accelerometer-measured physical activity, computed tomography (CT) imaging of abdomen and thigh, and vastus lateralis muscle biopsies [5]. Disability (health assessment questionnaire-disability index (HAQ-DI) and co-morbidities (co-morbidity index) were assessed by previously published questionnaires [6, 7]. Disease activity assessed by the disease activity score in 28 joints (DAS-28) was determined from a patient-completed visual analog scale, physician-determined numbers of tender and swollen joints, and erythrocyte sedimentation rate [8]. Plasma concentrations of inflammatory markers and cytokines were determined by immunoassay [5] and nuclear magnetic resonance (NMR) spectroscopy (GlycA) [9]. Insulin sensitivity was determined using Bergman’s minimal model [10] and concentrations of glucose and insulin (glucose: Beckman-CoulterDXC600; insulin: electrochemiluminscent assay from Meso Scale Discovery) at each of 29 time points during the intravenous glucose tolerance test.

Physical activity was measured with 7 days of accelerometry. After completing assessments, accelerometers (RT3, Stayhealthy, Inc., Monrovia, CA, USA) were provided to participants. Participants also received a pre-addressed and postage-applied box for return and directions for wearing on the waist above the right knee during waking hours for 7 days. Accelerometer data were evaluated for validity and non-wear time, and categorized into metabolic equivalents (METs) as previously described [11]. After data cleaning, valid data were available for 41 persons with RA and 31 controls. Time spent exercising was defined as the sum of time spent performing activity at METs equal to or greater than 3. CT scan analyses were performed using OsiriX (Pixmeo) to determine adipose and muscle tissue size and muscle tissue density (greater tissue density is indicative of less inter-muscular adipose tissue) [5]. Standard Bergstrom needle muscle biopsies were performed on the vastus lateralis in the fasting state; participants consumed only water during the 12 hours overnight prior to the biopsy [12]. Tissue was flash frozen in liquid nitrogen and stored at −80 ° C.

Skeletal muscle inflammatory marker measurements

Flash frozen muscle samples (5–10 mg) were homogenized in a buffer consisting of 1% Nonidet-P40, 1 mM EDTA, 150 mM NaCl, and 20 mM Tris-Cl for ELISA-based measures of muscle (m) interleukin (IL)-1β, mIL-6, mIL-8, m-tumor necrosis factor (TNF)-α (MSD 4-plex; K15053D-1) and m-toll like receptor (TLR)-4 (Abnova; KA1238). Assays were performed according to the manufacturers’ directions except for the addition of a 30-minute, room temperature, blocking step with 5% BSA followed by three PBS-T washes. Concentrations were normalized to starting masses. Spike-and-recovery assays for all analytes achieved 80–100% recovery confirming lack of assay interference by muscle homogenates. For each cytokine, the mean intra-assay and inter-assay coefficients of variation were: mIL-1β 8.5%, 13.2%; mIL-6 3.5%, 1.5%; mIL-8 4.0%, 4.0%; mTNF-α 8.4%, 10.4%; and mTLR-4 1.7% (only one plate was used for analyses).

Gene expression analyses

Muscle samples were selected for gene expression analyses in an effort to span the range of RA disease activity seen in the larger sample; these corresponded to the following DAS-28 categories: remission (n = 7), low (n = 4), moderate (n = 6), and high activity (n = 3). For each RA muscle sample, the corresponding sample from a control matched by age, gender, and BMI was included.

For RNA preparation, flash frozen muscle samples (20–30 mg) were homogenized in 1 mL TRIzol® (Thermo Fisher Scientific, Inc, Waltham, MA, USA). Biotinylated total RNA was generated using the Illumina TotalPrep RNA amplification kit (Life Technologies, Grand Island, NY, USA); 200 nanograms of RNA were used for the kit. The quality of the RNA was determined using the Bioanalyzer RNA Nano chip assay (Agilent, Santa Clara, CA, USA). Quantification of the RNA was determined using the Quant-iT RiboGreen RNA Assay Kit. The Human HT-12v3 Expression BeadChip (Illumina, San Diego, CA, USA) was used for quantitative whole genome RNA profiling. Biotinylated RNA (750 ng) was hybridized to the BeadChip and washed; Cy3-SA was then introduced to the hybridized samples and the BeadChips scanned on the Illumina iScan system according to manufacturer’s protocol. Quality control was performed using the Illumina GenomeStudio tools.

Gene expression fold-differences between groups were compared in Partek Genomics Suite (Partek, Inc.; St. Louis, MO, USA). For pathway analyses, differentially expressed genes (p < 0.02) were evaluated using the Ingenuity Pathway Analysis software (IPA, www.ingenuity.com). IPA identified the canonical pathways containing the greatest number of significant, differentially expressed genes in the dataset. IPA also generated novel networks of related genes and molecules based on the relationships present in the current literature.

Skeletal muscle metabolic intermediate measurements

Metabolites were measured in muscle from all participants (n = 102). Flash frozen muscle biopsies weighing approximately 25 mg were diluted 20 times (wt:vol) in ice-cold 50% acetonitrile containing 0.3% formate and homogenized for 120 sec in a TissueLyser II (Qiagen) at 30 Hz. Amino acids, organic acids, and acylcarnitines were analyzed using stable isotope dilution techniques in the Duke Molecular Physiology Metabolomics Core. Amino acids and acylcarnitine measurements were made by flow injection tandem mass spectrometry (MS) as previously described [13, 14]. The data were acquired using a Micromass Quattro Micro liquid chromatography (LC)-MS system running MassLynx 4.0 software (Waters Corporation, Milford, MA, USA). Organic acids were quantified using methods described previously [15] employing Trace Ultra GC coupled to ISQ MS operating under Xcalibur 2.2 (Thermo Fisher Scientific, Austin, TX, USA). All data are expressed as picomoles/mg tissue.

Statistical analyses

Accounting for the repeated measures in matched participants, patients with RA and controls were compared using mixed models. Muscle inflammatory molecules and metabolic intermediates were logarithmically transformed prior to group comparisons. Bivariate associations were assessed with Spearman correlation. Gene expression fold-changes were compared in Partek using analysis of variance (ANOVA). All other statistical analyses were performed using SAS 9.4 (SAS, Cary, NC). All data are available from the corresponding author upon reasonable request.

Results

Clinical measures and skeletal muscle inflammatory markers

As shown in Table 1, persons with RA were well-matched to controls by age, gender, and BMI. Patients with RA were recruited based on the inclusion criteria described and without respect to physical activity levels, body mass or body composition; similarly controls were included upon matching a patient with RA by age, gender, and BMI. Despite this, patients with RA and controls were similar with respect to physical activity levels, abdominal and thigh adipose depot size, muscle area, and muscle density [5, 11]. In those with RA, there was more co-morbidity, disability, and systemic inflammation; specifically, greater serum concentrations of high sensitivity C-reactive protein (hs-CRP), IL-6, and TNF-α (p < 0.05 for all) [5]. When skeletal muscle inflammatory markers were compared, there was approximately two times greater concentrations of the muscle cytokines, mIL-6 (p = 0.006) and mIL-8 in RA (p = 0.059) (Table 1).

Table 1 Participant characteristics
Full size table

Akin to disease activity, RA muscle inflammatory markers exhibited variation across a broad range (Table 1). Muscle inflammatory marker concentrations were positively associated with disease activity (mIL-1β, mIL-8), disability (mIL-1β, mIL-6), and pain (mIL-1β, mTNF-α, mTLR-4) (p < 0.05 for all) (Table 2). Muscle cytokines, mIL-1β and mIL-8, were negatively correlated with use of biological agents; mTNF-α was negatively correlated with use of non-biological disease-modifying therapy (p < 0.05 for all) (Table 2). Importantly, there were no correlation between muscle inflammatory marker concentrations and prednisone treatment.

Table 2 Skeletal muscle inflammatory marker correlations in patients with rheumatoid arthritis
Full size table

In addition to disease-related factors, muscle cytokine concentrations (mIL-1β, mIL-6, and mIL-8) were negatively associated with exercise minutes (p < 0.05 for all) (Table 2). Higher mIL-1β and mIL-6 concentrations were associated with less total physical activity (total METs; p < 0.05 for both) (Table 2). Although, both mIL-6 and mIL-8 were positively correlated with the systemic inflammatory marker, GlycA (p < 0.05 for both) (Table 2), muscle inflammatory marker concentrations were not related to insulin sensitivity or systemic cytokine concentrations.

Skeletal muscle gene expression

To better understand the etiology of RA muscle inflammatory markers, we compared RA (n = 20) and control (n = 20) skeletal muscle gene expression: 1939 genes were significantly upregulated or downregulated in RA samples (p < 0.05); 445 genes were identified when using a more stringent definition of significance (p < 0.02).

To identify other relationships between differentially expressed RA muscle genes, pathway analyses were performed using IPA, which has thousands of canonical pathways onto which our experimental gene expression differences were overlaid. Of those canonical pathways, IPA identified several pathways impacted by differential gene expression in muscle in RA (p < 0.05) (Table 3). Except for glycolysis and methionine degradation, these canonical pathways were identified because of reduced RA muscle gene expression for nuclear factor (NF)-kβ2, both nuclear factor of activated T cells (NFAT)5 and NFATC4, or all three. Also, none of the canonical pathways was predicted to be activated or inhibited by gene expression differences in muscle in RA (Z-scores < |2|) (Table 3) [16].

Table 3 Canonical pathways implicated in gene expression in muscle in rheumatoid arthritis
Full size table

In addition to canonical networks, pathway analyses generate novel networks connecting differentially regulated molecules based on published associations. The IPA-generated novel network with the highest connection score depicted significant differences in expression of genes associated with connective tissue, dental, and dermatological diseases (Fig. 1; Table 4). The prominent pathway connections in muscle in RA were centered on regulation of the NF-kB complex, specifically NF-kB2. These were in the setting of differential regulation of genes in muscle repair and glycolysis.

Fig. 1
figure 1

Novel network identified by muscle gene expression in rheumatoid arthritis (RA): gene expression was determined in muscle from 20 persons with RA and 20 controls matched by age, gender, and body mass index. The network shows connections between genes with differential expression in RA relative to control muscle. Genes in red were upregulated and genes in green were downregulated in muscle in RA

Full size image
Table 4 Novel network genes
Full size table

To augment traditional pathway analyses, we evaluated the 20 genes with the largest muscle expression differences in RA and control samples (Table 5) and examined gene members of well-established skeletal muscle anabolic, catabolic, and inflammatory pathways (Table 6). The top 20 upregulated and downregulated genes by fold difference were associated with muscle remodeling, satellite cell maturation, exercise intolerance, and/or energy metabolism; for these genes, the range of differences in expression was 20–50% (Table 5). Except for NF-kB2, there was no differential expression of canonical genes involved in skeletal muscle anabolic, catabolic, or inflammatory pathways (Table 6).

Table 5 Genes with the greatest differences in expression between patients with rheumatoid arthritis (RA) and controls
Full size table
Table 6 Genes involved in skeletal muscle anabolic, catabolic, and inflammatory pathways
Full size table

Skeletal muscle metabolic intermediates

When concentrations of skeletal muscle metabolic intermediates were compared between RA (n = 51) and controls (n = 51), muscle pyruvate concentrations were 46% greater in muscle in RA than in controls (p < 0.001) (Table 7). There were no significant differences in the concentrations of muscle amino acids, other organic acids, or acylcarnitines in RA compared to controls (Table 7). However, several muscle amino acids and acylcarnitines were related to RA disease activity and disability. For instance, greater concentrations of glycine, serine, aspartate/asparagine, and ornithine and lower muscle concentrations of alanine and fumarate were related to greater disease activity (p < 0.05) (Table 8). Greater muscle concentrations of glycine, proline, ornithine, arginine, and aspartate/asparagine were related to greater disability (p < 0.05) (Fig. 2); in contrast, lower concentrations of several long-chain unsaturated acylcarnitines were related to greater disease activity and disability (p < 0.05) (Table 8).

Table 7 Skeletal muscle metabolic intermediate concentrations
Full size table
Table 8 Relationships between rheumatoid arthritis clinical features and muscle metabolic intermediates
Full size table
Fig. 2
figure 2

Schematic depiction of muscle injury repair showing potential impact of cytokine, gene expression, and amino acids on satellite cell activation, macrophage function, and fibrosis in muscle from patients with rheumatoid arthritis (RA). Boxes show gene IDs for genes differentially regulated in patients with RA and in controls. See Table 3 for gene descriptions

Full size image

Discussion

Here, we report that in RA, skeletal muscle exhibits molecular alterations in inflammatory markers, transcriptional profiles, and metabolic signatures. Both at protein and transcriptional levels, muscle had a pro-inflammatory phenotype in RA. Additionally, differential gene expression in muscle in RA was indicative of dysregulation of muscle repair, promotion of glycolysis, and poor mitochondrial function. Upregulated glycolysis and mitochondrial inefficiency were supported by greater muscle concentrations of the glycolytic end-product pyruvate in RA. Further, disease activity and disability were related to lesser concentrations of long-chain acylcarnitines and greater concentrations of amino acid precursors for muscle fibrosis. Taken together, these alterations in proteins, gene expression, and metabolic intermediates were indicative of muscle in RA in a state of chronically activated, yet dysregulated remodeling with increased glycolysis, mitochondrial inefficiency, and fibrotic material (Fig. 2).

This represents the first report of significant markers of inflammation in muscle in RA. The clinical importance of these molecules is demonstrated by the significant association of several muscle cytokines with RA disease activity, disability, pain, and physical inactivity. The IPA-generated novel network centered on downregulation of NF-kB2, a protein that promotes non-canonical NF-kB signaling and opposes inflammatory signaling [17]. Downregulation of NF-kB2 would be predicted to favor coordinated upregulation of pro-inflammatory NF-kB signaling in muscle in RA. We were unable to determine if the muscle cytokines and pro-inflammatory transcripts in RA were derived from myocytes, inflammatory cells, or other cellular sources. Interestingly, muscle cytokine concentrations did not reflect those measured in circulation, suggesting these disease-associated inflammatory markers stem from local rather than systemic events.

Based on the strong relationships between muscle inflammatory markers and disability, pain and physical inactivity, we suspected that increased intramuscular cytokines may be indicative of a disrupted muscle remodeling process. In fact, muscle gene expression alterations in RA were consistent with promotion of satellite cell differentiation and upregulation of several facets of the normally well-coordinated process of muscle remodeling. For instance, muscle in RA was characterized by downregulation of ZFP36, the gene that encodes tristetraprolin (TTP), which reduces inflammation by destabilizing pro-inflammatory cytokine transcripts [18, 19] and prevents satellite cell activation by destabilizing myogenic regulatory factor, MyoD, mRNA [20]. Thus, the reduction in ZFP36 expression in muscle in RA would be expected to promote pro-inflammatory cytokine production and satellite cell activation.

Other gene expression changes also suggest both chronic activation and temporal dysregulation of muscle remodeling. For instance, downregulation of ZBTB16 would promote inflammation and proliferation of autoreactive T cells [21]. In contrast to the reduced ZFP36 expected to promote satellite cell activation, the increased CITED2 would be expected to reduce satellite cell activation [22, 23]. Increased expression of ABRA, RCAN1, VGLL2, MYF6 and decreased expression of RRAD would promote differentiation of satellite cells [24–30]. More descriptions of differentially expressed genes are provided in Additional file 1.

Gene expression alterations indicative of glycolysis promotion and poor mitochondrial function were supported by greater muscle concentrations of the glycolytic end-product pyruvate in RA. Further, disease activity and disability were related to lower concentrations of fatty-acid-derived long-chain acylcarnitines. One plausible explanation for this relationship is that fewer long-chain acylcarnitines indicate less oxidative metabolism and fewer mitochondria, consistent with a glycolytic phenotype. RA disease activity and disability were also related to higher concentrations of amino acid precursors for muscle fibrosis. M2-type macrophages contain arginase, which metabolizes arginine to ornithine [31]. Ornithine is converted to proline, which provides a substrate for resident fibroblasts to generate collagen. In addition to proline, collagen formation also requires glycine; glycine and proline each account for a third of the collagen amino acids. While collagen is critical for extracellular matrix production, in the setting of a chronically activated remodeling process, excess collagen production leads to fibrosis [31]. Thus, the relationships between these amino acids and disease activity and disability may indicate a fibrotic process in muscle associated with active disease that contributes to RA-associated disability.

There were several limitations to this study. RA medication regimens were not uniform among participants, and effects of these medications on skeletal muscle are unclear. Twenty-five percent of patients with RA used prednisone at low doses, which is not expected to have significant myopathic effects; despite this, they had significant alterations in muscle inflammatory markers and systemic inflammation relative to controls. Without histopathologic assessment or single cell isolations, we were unable to determine the cellular source of muscle cytokines, transcripts, or metabolites. Our findings indicate that either RA medication regimens or the RA disease process itself alters skeletal muscle inflammatory molecules, transcriptional profiles, and metabolic pathways.

Conclusions

Taken together, these alterations in pro-inflammatory cytokines, gene expression, and metabolic intermediates are indicative of RA muscle in a state of chronically activated, yet dysregulated remodeling, with increased glycolysis, mitochondrial inefficiency, and fibrosis. It is very likely these changes contribute to the ongoing issues of exercise intolerance and disability in persons with RA. Future work should be directed at understanding whether these deficits may be mitigated by combining pharmacologic treatment with physical activity, to reduce inflammatory signaling and/or fibrosis while promoting skeletal muscle efficiency. Therefore, to improve the lives of patients with RA, future work should be directed toward understanding the role of skeletal muscle in RA, and interactions between treatment regimens, physical activity, and influences of skeletal muscle on the clinical status in RA.