Genes and biological processes associated with increasing (worsening) QuickDASH score across tissue types in shoulder OA
Complement activation fragments in cartilage and synovium occur without the need for parent molecules from the blood stream [16] indicating complement activity is a local phenomenon. The injurious sequelae of complement activation in shoulder OA is likely compounded by complement pathways present in bone, capsule, synovium, fat and muscle; although, downregulated in end stage disease and associated with worsening shoulder function as shown in our study. Further complement pathway analysis revealed expression of immunoglobulins in bone may explain the creation of C4 and C2 activators. Alternatively, contamination of bone with blood could explain this, however given equivalent findings in fat, a relatively avascular tissue, this is thought less likely.
Other pathways strongly negatively correlated with worsening shoulder disability in all tissue types included Fc gamma receptors (FcγRs), which have been implicated in osteoclastogenesis and may be critical regulators of inflammatory arthritis but not demonstrated in shoulder OA to date [17].
Notable and novel genes associated with increasing (worsening) QuickDASH score in OA across tissue types
Synovium and capsule
Upregulation of NR4A3 (NOR-1) was observed in our study and is consistent with NR4A receptors upregulation in synovium and cartilage of patients with synovial hyperplasia, a key mechanistic piece in the pathogenesis of OA [18]. NR4A2 transcriptional activity was shown in synovium of osteoarthritic knees contributing to synovial hyperplasia, a precursor to cartilage destruction; however, NR4A3 mRNA activity remained the same in osteoarthritic and normal knees [19]. NR4A3 is a transcription factor linked to cellular proliferation, differentiation and oxidative metabolism and its upregulation could be a biomarker specific to shoulder OA.
SIK1 activity is robust in normal articular cartilage but decreased in knee OA [20]. Upregulation in capsule and synovium in our series is representative of an inflammatory state in shoulder OA given the observation that SIK inhibition induces an anti-inflammatory phenotype in macrophages [21]. Genes responsible for OA and osteoporosis may demonstrate pleiotropy given significant overlap in association studies [22]. Because SIK inhibition is osteoclastogenic [23], it would be worthwhile to determine potential relationships with the ‘bone former’ OA phenotype, and upregulation with subchondral cyst formation, or ‘bone resorber’ and in OA.
Activation of other inflammatory genes in shoulder OA included CXCL5, a chemokine within injured tissues. CXCL5 neutralisation reduces joint inflammation and bone destruction in murine Rheumatoid Arthritis models [24]. This gene also acts to reduce sensitivity to sunburn pain [25] and our study details an association with worsening shoulder function and pain as per QuickDASH responses.
Fat
Zinc dependant transcription signalling causes OA progression after biomechanical injury in vivo via inflammatory mediators [26] and downregulation of FEZF2 in shoulder OA is a novel finding of our study.
In murine models, TBC1D3 expression in macrophages delayed wound healing by altering extracellular vesicle delivery between stromal and immune cells. Diagnostic targeting of extracellular vesicles or secretory machinery that mediates tissue repair may be possible [27] and downregulation in our osteoarthritis cohort points to utility as a potential biomarker of worsening disease.
Subchondral bone
The gene products of FRZB regulate chondrocyte differentiation and survival [28]. Downregulation of FRZB has been demonstrated in murine osteoarthritic knee models [29] and inhibits the Wnt/β-catenin pathway. In human chondrocytes in-vitro, FRZB expression associates with increased MMP13 expression [30], which breaks down extracellular matrix by cleavage of collagen II [31]. Our findings therefore support literature citing downregulation of FRZB in knee OA models and confirm its relevance in human shoulder OA.
Biological processes downregulated in OA, compared with instability: pathways linked to RORα, RAC1, NOTCH2, CREB1, and PI3K
RORα, RAC1, NOTCH2 and CREB1 is upregulated in knee OA [32–34]. We also found upregulation of these genes in bone in our study although, this was not statistically significant (p > 0.05), perhaps owing to a small sample size. Physiological chondroprotection, temporal or tissue dependant mechanisms may explain downregulation of these implicated genes in capsular tissue biopsies in our study.
Activation of the PI3K/AKT/mTOR signalling promotes chondrocyte proliferation, reduces apoptosis [35] and is inhibited in knee OA [36]. We observed PI3K and KAT6A downregulation in capsular tissue of osteoarthritic shoulders in keeping with disease mechanisms described for other joints [35, 37, 38].
Significantly differentially expressed genes in shoulder OA compared with instability
Targeting mitochondrial dysfunction has been proposed for the treatment of OA [39]. In knee OA, cartilage degeneration can result from diminished ATP production [40], increased oxidative stress [41, 42], and calcium dysregulation [43]. Downregulation of the following mitochondrial genes in association with worsening shoulder function in our OA cohort is novel: MT-TE, MT-TY and Mt-ts1. The tricarboxylic acid pathway is generally upregulated in OA, probably to increase ATP production for cellular repair, which is more pronounced in late disease [44]. Of the top 20 expressed genes, ATP5MD was singularly upregulated in our study.
Limitations
Despite a wide range of baseline QuickDASH scores (38.6–86.4) within our cohort, transcriptomic inferences made according to worsening shoulder disability, cannot necessarily be applied across the spectrum of disease. Tissue samples obtained at different stages of disease is practically challenging but would enable identification of unique transcriptomic patterns occurring over time.
Low patient numbers in the OA group (N = 6) yielded a low number of differentially expressed genes across the 5 different periarticular tissues. Both parametric and non-parametric analyses were undertaken yielding similar results, suggesting a likely negligible batch effect; further, non-parametric analyses were reported upon given the low sample size.
Shoulders with instability demonstrate an anabolic; rather than inflammatory or catabolic phenotype [9]. Thereby, after a time of quiescence following the traumatic event, comparison of genes and gene pathways in an inert environment is possible. Despite significant differences in age and sex between our two groups, previous transcriptomic work comparing adhesive capsulitis to instability demonstrated no significant relationship between gene expression, age, or sex further justifying utility as a control group [12]. Recurrent shoulder instability is a risk factor for OA decades later; what is less clear is if gene expression changes because of the initial traumatic event, recurrent instability episodes, or some other mechanism. By excluding patients with a history of shoulder instability in the OA group we assumed that genetic changes in shoulder instability, whatsoever, are unrelated to gene changes associated with osteoarthritis.
Metabolic syndrome associated OA is a recognised entity, but pathophysiological mechanisms remain unproven [45]. There were clear between-group differences for each constituent risk factor of metabolic syndrome, namely BMI, HbA1c, hypercholesterolaemia, and hypertension, when comparing OA and instability. Whole transcriptome next generation sequencing of blood has demonstrated enrichment of genes associated with inflammation, protein synthesis and mitochondrial dysfunction [46]. Further studies investigating transcriptomic expression changes in patients with metabolic syndrome, with and without OA are required before our observed gene and pathway differences pertaining to inflammation (RAC1, CREB1), protein synthesis (NOTCH2, PI3K/AKT/mTOR pathway) and mitochondrial genes (MT-TE, MT-TY, Mt-ts1, MT-ATP8) can be accurately interpreted as integral to the disease process rather than associated with metabolic syndrome itself. To do this, larger sample sizes are required than investigated here.