Article Text
PDF
Abstract
Objectives Stratification approaches are vital to address clinical heterogeneity in Sjogren’s syndrome (SS). We previously described that the Newcastle Sjogren’s Stratification Tool (NSST) identified four distinct clinical subtypes of SS. We performed proteomic and network analysis to analyse the underlying pathobiology and highlight potential therapeutic targets for different SS subtypes.
Method We profiled serum proteins using O-link technology of 180 SS subjects. We used 5 O-link proteomics panels which included a total of 454 unique proteins. Network reconstruction was performed using the ARACNE algorithm, with differential expression estimates overlaid on these networks to reveal the key subnetworks of differential expression. Furthermore, data from a phase III trial of tocilizumab in SS were reanalysed by stratifying patients at baseline using NSST.
Results Our analysis highlights differential expression of chemokines, cytokines and the major autoantigen TRIM21 between the SS subtypes. Furthermore, we observe differential expression of several transcription factors associated with energy metabolism and redox balance namely APE1/Ref-1, FOXO1, TIGAR and BACH1. The differentially expressed proteins were inter-related in our network analysis, supporting the concept that distinct molecular networks underlie the clinical subtypes of SS. Stratification of patients at baseline using NSST revealed improvement of fatigue score only in the subtype expressing the highest levels of serum IL-6.
Conclusions Our data provide clues to the pathways contributing to the glandular and non-glandular manifestations of SS and to potential therapeutic targets for different SS subtypes. In addition, our analysis highlights the need for further exploration of altered metabolism and mitochondrial dysfunction in the context of SS subtypes.
- Sjogren's Syndrome
- Inflammation
- Patient Reported Outcome Measures
- Autoantibodies
- Autoimmune Diseases
Data availability statement
No data are available.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
Clinical heterogeneity is a significant barrier to understanding the pathobiology, identifying reliable biomarkers and developing efficacious therapeutics in Sjogren’s syndrome.
Stratification approaches have identified clinical subtypes of Sjogren’s syndrome.
WHAT THIS STUDY ADDS
This study demonstrates unique protein networks for clinical subtypes of Sjogren’s Syndrome with differences in expression of chemokines, cytokines, TRIM21 and master transcription factors.
Furthermore, these differences provide plausible mechanisms linking pathobiology with symptoms in Sjogren’s syndrome.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
Our data have relevance for trial design and therapeutic development as the networks provide clues to the biological pathways responsible for the glandular and non-glandular manifestations of Sjogren’s syndrome and to potentially effective therapeutic targets for Sjogren’s syndrome subtypes.
Furthermore, our data support the examination of differences in energy metabolism, mitochondrial function and redox status in the context of Sjogren’s syndrome subtypes.
Introduction
Sjogren’s syndrome (SS) is a chronic, immune-mediated inflammatory disease (IMID).
The clinical manifestations of SS are diverse and vary in severity greatly, with some patients reporting debilitating symptoms and a very poor quality of life and others reporting few symptoms. This clinical heterogeneity is responsible, at least in part, for difficulties in identifying reliable biomarkers and the variable responses to therapeutics in SS.
Stratification approaches are increasingly recognised as crucial to addressing heterogeneity.1–3 Using data from three European cohorts, the Newcastle Sjogren’s Stratification Tool (NSST) identified four clinical subtypes of SS: low symptom burden (LSB), high symptom burden (HSB), dryness dominant with fatigue (DDF) and pain dominant with fatigue (PDF).3 The subtypes have distinct molecular profiles, clinical manifestations and may respond differently to immunomodulatory therapies.4 Two independent groups have also identified similar symptom-based subtypes.1 2 Furthermore, longitudinal analysis demonstrated these subtypes to be stable over a 5-year follow-up period.2
Clinical heterogeneity in SS is likely underpinned by networks of dysregulated biological pathways, detailed knowledge of these pathways is vital to understanding clinical heterogeneity and has relevance for trial design and therapeutic development. In this study, we analyse the profile of serum protein expression of the four clinical subtypes to improve our understanding of the potential pathogenic mechanisms of the clinical heterogeneity and potential therapeutic targets for these different SS subtypes.
Methods
Patients
The UK Primary Sjogren’s Syndrome Registry (UKPSSR) is a nationwide observational cohort of clinically well-characterised patients with primary SS who fulfil the 2002 American European Consensus Group classification criteria. The UKPSSR holds detailed clinical and laboratory data that is collected prospectively, including patient-reported symptoms collected using standardised questionnaires.
Newcastle Sjogren’s Stratification Tool
Patients were classified into four symptom-based subgroups using the NSST.3 The tool uses a logistic regression model to classify patients using their responses for ESSPRI pain, ESSPRI fatigue, ESSPRI dryness and Hospital Anxiety and Depression scales (HADs).
Proteomic analysis
We profiled serum proteins using O-link technology of 180 SS subjects. Subjects were chosen to include 45 individuals from each symptoms-based subtype. The clinical and demographic data of the 180 subjects is shown in online supplemental table 1. We used 5 O-link proteomics panels: ‘inflammation’, ‘immune response’, ‘organ damage’, ‘cardiovascular’ and ‘metabolism’, with 454 unique proteins measured.
Supplemental material
ELISA
We performed a combined technical and independent validation of IL-6 expression in serum. We measured the serum IL-6 concentration of 193 SS patients and 18 healthy controls selected from the UKPSSR. Samples were analysed on 96-well ELISA plates (Invitrogen), read at 450 nm. Out of the 193 SS patients 132 had been used in the O-link analysis and 61 were independent. The samples were selected to be equally distributed between the NSST subtypes.
Statistical analysis
Differential expression estimates are derived from a generalised linear model using the LSB group as a reference, after adjustment for baseline age, sex and plate batch. As we anticipate that these serum proteins may act directly or indirectly through interactions with other proteins, we used a network analysis approach—the ARACNE algorithm.5 ARACNE analysis calculates degrees of mutual information between the variables of a multidimensional dataset and reflects shared information and correlation between variables within a network. Nodes within the clinical network represent individual variables and the edges between them represent the relationship between those variables. Edges are filtered to those having significant (p<1−10) mutual information as determined by the ARACNE algorithm. Therefore, where nodes are not connected by an edge there is a much weaker level of association. Differential expression estimates were overlaid on these networks to reveal key subnetworks that are differentially expressed between the subtypes. We applied node degree, betweenness and eigenvector centrality to identify the influential ‘hub’ proteins within the network. The ELISA data were analysed using Kruskal-Wallis test.
We reanalysed the tocilizumab trial using linear regression including terms for baseline subgroup, treatment and the subgroup×treatment interaction. Covariates, age, sex, body mass index, C reactive protein, IgG titre, C4 levels, lymphocyte count, were also included for adjustment of the model. Individual contrasts of change in FACIT-F score were assessed using Wilcoxon rank-sums tests. Statistical analysis was performed in R V.4.1.3 and JMP Pro V.15 and network visualisation in Cytoscape V.3.8.2.
Results
ARACNE protein network
ARACNE networks for HSB, PDF and DDF subtypes are shown in figure 1. Overlaying differential expression estimates for serum proteins on the networks, we identified that the differentially expressed proteins between SS subtypes were highly related to one another and localise to subnetworks. The two key subnetworks for differential expression are highlighted by ellipses.
Whole O-link ARACNE network in which nodes represent individual proteins and edges represent the connections between them based on mutual information. Node colour scale represents differential expression of proteins in the (A) DDF subtype, (B) HSB subtype and (C) PDF subtype in each case using the LSB subtype as a comparator. Superior and inferior black ellipses highlight subnetworks with substantial differences in relative expression of serum protein between the three groups. DDF, dryness dominant with fatigue; HSB, high symptom burden; PDF, pain dominant with fatigue.
The superior most subnetwork (circled), with higher expression in HSB and PDF subtypes, contains proteins involved in innate immunity and inflammation, redox balance, as well as several pleiotropic transcription factors associated with energy metabolism. The inferior subnetwork (ovoid), with higher expression in the DDF subtype, consisted of chemokines, cytokines and proteins associated with lymphocyte activation.
B-cell hyperactivity in the DDF proteome
The DDF subtype has the highest salivary gland (SG) dysfunction, anti-SSA/SSB antibody positivity and prevalence of B-cell lymphoma.3 Within the DDF proteome, the inferior subnetwork of proteins was upregulated compared with other subtypes. This subnetwork contains chemokines, cytokines and cell surface markers associated with lymphocyte migration, antigen presentation, germinal centre activity and plasma cell maturation (figure 2A).
Bar charts showing the estimated difference between each phenotype using the LSB phenotype as a comparator group. Estimates are derived from a generalised linear model (GLM) testing for differences in the phenotypes for each protein after adjustment for baseline age, sex and batch. (A) Selected proteins from the inferior subnetwork showing differences in inflammatory cytokines and chemokines associated with the DDF, HSB and PDF subtypes (B) Selected proteins from the superior subnetwork showing the major differences associated with the DDF, HSB and PDF subtypes. DDF, dryness dominant with fatigue; HSB, high symptom burden; LSB, low symptom burden; PDF, pain dominant with fatigue.
The DDF proteome showed the highest levels of chemokines associated with lymphocyte migration to the SGs including IFN-γ-induced chemokines CXCL9, CXCL10 and CXCL11. These chemokines exert their effects via CXCR3 and are strongly associated with T-cell infiltration of the SG.6
CXCR3 ligands are elevated in the ductal epithelium, tear film and peripheral blood of SS patients.5 Subtypes reporting minimal glandular dysfunction showed lower levels of CXCR3 ligands in addition to their established lower IFN module activity and decreased anti-SSA and anti-SSB positivity.3
Chemokines associated with ectopic lymphoid structures (CXCL13, CCL19 and CX3CL1) and lymphoma risk (CXCL13) in SS were also increased in the DDF proteome. Elevated CXCL13 and CCL19 levels have been demonstrated in the saliva and serum of SS patients and correlate with increased lymphoid organisation.7 CXCL13 is of particular importance in pathobiology of SS, with expression correlated with T-follicular helper cell levels, lymphocytic focus scores, B-cell hyperactivity and lymphoma development.8 9 Another IFN-γ-induced chemokine, CX3CL1 functions as T-cell chemoattractant and colocalises with ectopic lymphoid structures in SG biopsies. CX3CL1 receptor, CX3CR1, has been shown to be transcriptionally upregulated in B-cells from SS patients and tightly correlates with increased focus scores.10 11
Serum autoantibodies, raised free light chains, B-cell infiltrates in the SG and patients' increased risk of B-cell lymphoma highlight the influence of B-cells in the pathogenesis of SS. The DDF subtype exhibited the highest expression of cytokines associated with B-cell stimulation including IL-2, IL-4, IL-10, IL-13 and lymphotoxin-alpha (TNF-β). SERPINAA9, whose expression is restricted to germinal centre B-cells and lymphoid malignancies, was also found to be increased. Expression of BAFF in the DDF proteome was not significantly elevated above that of the PDF and HSB subtypes. However, this may reflect an elevated level across all SS subtypes. The expression of costimulatory molecules, including CD28 and 4-IBB, was also highest within the DDF subtype. Furthermore, high expression of signalling lymphocyte activation molecule 1 (SLAMF1) and its downstream adaptor SLAM-associated protein (SAP; encoded by SH2D1A) also points towards robust lymphocyte activation in this subtype.9
We also examined the expression of other important proteins linked to the pathogenesis of SS. TRIM21 (Ro52, SSA) is a major autoantigen in SS. It functions as an E3 ligase and as an intracellular receptor recognising cytoplasmic immune complexes that escape endosomal degradation.12 Furthermore, altered TRIM21 expression in SS may have roles in B-cell function and the type-1 interferon (IFN) regulation.12 The DDF subtype has the highest anti-SSA antibody positivity and high IFN module activity, interestingly our analysis showed it also had the lowest TRIM21 protein expression.
FOXO1 expression also differed across subtypes, with the lowest expression of FOXO1 in the DDF proteome. This pleiotropic transcription factor has important roles in lymphocyte responses including pivotal roles in regulatory T-cell (Treg) function.13 14 Inactivation of FOXO1 is essential for optimal B-cell proliferation and long-lived plasma cell survival via its inactivation of caspases 3 and 7.15 In keeping with robust B-cell responses in the DDF subtype, the DDF proteome showed reduced FOXO1 and caspase 3 expression.12 Ubiquitin-specific protease USP8 was also decreased in the DDF subtype. Reduced USP8 is linked with altered FOXO1 activity via its ubiquitination of 14-3-3, the molecule responsible for sequestering FOXO1 within the cytoplasm.14 Alterations in USP8 expression are also linked to aberrant immune regulation including impaired Treg function and altered PD-L1 expression.16
NF-κβ signalling, oxidative stress and altered metabolism in the HSB and PDF proteomes
The HSB and PDF subtypes report increased burden from non-glandular manifestations of SS such as pain and mood disturbances. Within the HSB and PDF proteomes, there was increased expression of proteins in the superior subnetwork (figure 2B). Network metrics identified apurinic/apyrimidinic endonuclease 1/redox factor-1 (APE1/Ref-1) as an influential ‘hub-protein’ in the subnetwork (figure 3). APE1/Ref-1 is the main endonuclease in the base excision repair (BER) pathway, the main mechanism of oxidative DNA damage repair.17 In addition, APE1/Ref-1 functions as a reductive activator of various transcription factors including c-Jun, p53, nuclear factor kappa B (NF-κβ), activator protein-1 (AP-1), hypoxia-inducible factor 1α (HIF-1α) and nuclear erythroid factor-like 2 (NRF2).18 19 These transcription factors coordinate a diverse range of cellular and mitochondrial processes including redox status, biomolecule metabolism, apoptosis and inflammation.
Subnetwork of the first neighbour nodes of APEX1, the node within the network with the highest degree centrality. APEX1 shares an edge with every other node within the subnetwork and differential expression of the neighbour nodes of APEX1 is similar to APEX1 differential expression. We hypothesise that small changes in ‘hub’ nodes and their neighbours can have a greater impact on the network. The colour scaling shows the estimated difference between each phenotype using the LSB phenotype as a comparator group. The red colour indicates a higher expression compared with LSB and the blue colour indicates a lower expression level compared with LSB. Estimates are derived from a generalised linear model (GLM) testing for differences in the phenotypes for each protein after adjustment for baseline age, sex and batch. Differential expression of the nodes within this subnetwork differs markedly between the subgroups (A) DDF, (B) HSB and (C) PDF. DDF, dryness dominant with fatigue; HSB, high symptom burden; LSB, low symptom burden; PDF, pain dominant with fatigue.
Increased expression of APE1/Ref-1 was demonstrated in HSB and PDF proteome and potentially reflects increased activity of the BER pathway in response to oxidative stress. DNA damage and the type 1 IFN response are recognised features in the pathogenesis of SS. Consistent with this, expression of TP53-inducible glycolysis and apoptosis regulator (TIGAR) was also increased in these subtypes and influential in the superior subnetwork. TIGAR activity is initiated by p53 after a cell has experienced a low level of DNA damage or stress.20 Several proteins involved in the antioxidant response were in the superior subnetwork and upregulated in the HSB and PDF subtypes, including Glutarexodin1 (GLRX) and NAD+Kinase (NADK). GLRX is an essential thioltransferase which primary role is to reverse glutathionylation. Glutathionylation is a major mechanism of redox regulation affecting DNA repair-related enzymes as well as ratio of reduced glutathione, the main intracellular antioxidant. Similarly, NADK influences the redox balance through the synthesis of crucial cofactor NADPH. Homocysteine, the vital amino acid in glutathione production, may also vary between subtypes with expression of adenosylhomocysteinase (AHCY) the enzyme responsible for homocysteine synthesis altered between the subtypes. The HSB cohort showed the highest expression of AHCY, with the protein closely associated with GLRX in our network analysis.
Cellular redox status regulates many important metabolic enzymes and transcription factors. This relationship is bidirectional with metabolic pathways simultaneously affecting reactive oxygen species (ROS) and antioxidant consumption. The effects of APE1/Ref-1 and TIGAR on cellular metabolism are well studied in cancer cells. Furthermore, through its redox control, APE1/Ref-1 is directly linked with HIF1α-dependent and p53-dependent metabolic changes. TIGAR also effects cellular energy production, notably glycolysis, in response to cellular ROS levels.20 EGLN1, encoding prolyl hydroxylase domain 2 (PHD2) was also influential in the superior subnetwork. PHD2 is the critical PHD isomer mediating HIF-1a protein degradation under normoxia and may exert its influence on the broader network via HIF-1a. It may also affect the network through HIF-independent activity and is postulated as regulator of NF-κB, in the context of inflammation. Transcription factors FOXO1 and BTB Domain And CNC Homolog 1 (BACH1) were also closely associated with APE1/Ref-1 in our network and showed increased expression in the HSB and PDF subtypes (figure 4). These pleiotropic transcription factors are well-recognised regulators of metabolic processes, and our analysis confirmed their role in influencing the surrounding divergent proteins. FOXO1 is involved in mediating gluconeogenesis, glycogenolysis and lipid metabolism.21 Notably, it is phosphorylated and degraded by insulin activation of PI3K/PKB signalling and is linked with metabolic diseases via insulin resistance and hypertriglyceridaemia.21 BACH1 has been reported to influence glycolysis, oxidative phosphorylation and the TCA cycle.22 Thus, small changes in expression of these influential transcription factors may have broad reaching consequences. Proteins involved in lipid metabolism were also altered between subtypes, with insulin-like growth factor binding proteins 1 and 2 (IGFBP1, IFBP2), adipokine fatty acid binding protein 4 (FABP4) and perilipin (PLIN1) differentially expressed.
Selected proteins from the superior subnetwork showing the major differences between the LSB and DDF (A), HSB (B) and PDF (C) subtypes. The network is derived from ARACNE analysis of protein expression data where nodes represent individual proteins and edges represent the mutual information and therefore the strong association between nodes. The colour scaling shows the estimated difference between each phenotype using the LSB phenotype as a comparator group. The red colour indicates a higher expression compared with LSB and the blue colour indicates a lower expression level compared with LSB. Estimates are derived from a generalised linear model (GLM) testing for differences in the phenotypes for each protein after adjustment for baseline age, sex and batch. DDF, dryness dominant with fatigue; HSB, high symptom burden; LSB, low symptom burden; PDF, pain dominant with fatigue.
DNA damage and redox imbalance are intrinsically linked to inflammation and many of the proteins in the superior subnetwork were associated with innate immune response, NF-κβ and proinflammatory pathways. IL-1a, IL-6, MCP-3, CCL3, OSM, AREG, TREM1, ITGB2, RNASE 3, VCAN, ROR1, MPO, PRTN3, AZU1, CLEC4A and CLEC4D were all upregulated in comparison to LSB and DDF. Ligands for RAGE (receptor for advanced glycation endproducts), S100A12 and S100p, were also found in the superior subnetwork and had increased expression in HSB and PDF subtypes. RAGE has a range of ligands that are released in response to cellular stress and is implicated as a driver of the inflammatory response via its NF-κβ response promoter.23 RAGE both promotes DNA and RNA uptake into endosomes and decreases the immune recognition threshold for activation of the primary DNA recognising transmembrane receptor TLR9.
HSB proteome
The HSB subtype has the highest patient-reported scores for anxiety and depression (HADS).3 Proteome analysis identified differential expression in several neuroimmunoendocrine pathways that distinguish the HSB proteome from PDF and other subtypes. These pathways propose a mechanistic link between immune dysfunction and the cognitive and non-glandular manifestations of SS. The HSB proteome had the lowest expression of Kynurenine aminotransferase 1 (KYAT1), Catechol-O-methyltransferase (COMT), fibroblast growth factor 2 (FGF2) with distinct high expression of cytokines IL-6 and IL-1a.
Kynurenine aminotransferases are involved in processing kynurenine. Altered tryptophan metabolism, via serotonin-kynurenine imbalance or via an imbalance in kynurenine metabolites, may cause immune system dysregulation. Furthermore, it has associations with depression and the CNS manifestations of IMIDs including SS.24 Similarly, COMT function can be influenced by the immune system and is intrinsically linked to depression and anxiety via catecholamine degradation. Indeed, polymorphisms related to decreased COMT activity have been linked to higher circulating basal and stress-induced levels of catecholamines and increased reported depression.25 Reduced FGF2 expression also has links to depression and anxiety disorders.26 This warrants further investigation in the context of SS, given the role of SGs in growth factor production. There was also altered protein expression of FKBP4 across subtypes. FKBP4 is involved in the glucocorticoid receptor-ligand complex with FKBP5 and HSP90 and may lead to altered glucocorticoid response between the subtypes. This resembles tryptophan metabolism, whereby in mediating immune responses as well as metabolic and endocrine functions, small changes in glucocorticoid response may significantly alter proteome and clinical phenotype.
Another mechanism by which the immune system may drive non-glandular and cognitive manifestations of SS is via proinflammatory cytokines IL-6 and IL-1.27 IL-1a, IL-1Ra and IL-6 all showed increased expression in HSB proteome. We also used ELISA to perform a combined technical and independent validation of IL-6 expression of 211 UKPSSR serum samples (193 SS patients, 18 healthy controls, online supplemental figure 3). IL-6 has established links with depression in preclinical and clinical studies. Increased levels of IL-6 have been demonstrated in patients with depression, depressive symptomology and patients at risk of developing depression.27 Increased circulating IL-6 has also been implicated in contributing to fatigue. A randomised control trial of IL-6 receptor monoclonal antibody, tocilizumab, in 110 SS patients did not meet primary endpoints, however, there was no patient stratification used in the analysis and our reanalysis of this trial shows a reduction in the FACIT-Fatigue score for the tocilizumab arm in the HSB subtype (online supplemental figure 2).28 IL-1 has been linked with fatigue, with raised cerebrospinal fluid levels of IL-1Ra, a naturally occurring IL-1 receptor antagonist, found to correlate with fatigue in SS patients.29 Given the pleiotropy and redundancy characteristic of the cytokine response the relationship between cytokines and cognitive symptoms is likely complex.
Discussion
The four clinical subtypes of SS have distinct clinical manifestations and may respond to immunomodulatory therapies differently.3 Proteomic analysis showed differential expression of chemokines, cytokines and master transcription factors between the SS subtypes. Furthermore, the differentially expressed proteins were interrelated on our network analysis. Our data support the concept that distinct molecular networks underlie the clinical subtypes of SS.
Our analysis highlights pathways that potentially shape the clinical manifestations of each subtype (table 1). The clinical phenotype of the DDF subtype is characterised by glandular dysfunction. This subtype has the highest objectively measured SG dysfunction and patient-reported dryness scores, along with the highest markers of B-cell activation and risk of developing B-cell lymphoma.3 Consistently, proteomic analysis demonstrated higher expression of chemokines orchestrating lymphocyte migration and ELS formation in the SGs and increased expression of cytokines promoting B-cell hyperactivity and autoantibody production. This suggests therapeutics targeting B-cells and costimulation would show optimal performance within this subtype. Reanalysis of the initially disappointing TRACTISS trial has shown rituximab significantly improved unstimulated and stimulated salivary flow at week 48 in the DDF subtype.3
Proteome summary
The altered expression of major autoantigen TRIM21 between subtypes warrants further investigation. TRIM21 expression was lowest in the DDF subtype, which also have the highest anti-TRIM21 antibody seropositivity and high IFN modular activity.3 TRIM21 expression is enhanced by type-1 IFNs and TRIM21 also acts as negative regulator of IFN signalling, via its ubiquitylation of interferon regulating factors and their subsequent proteasomal degradation. Anti-TRIM21 antibody has been shown to limit TRIM21 E3 ligase activity in vitro, however, the mechanisms by which anti-TRIM21 antibody, intracellular and serum TRIM21 and IFN signalling interact requires further investigation.
In contrast to DDF, the clinical phenotype of HSB and PDF subtypes focuses on non-glandular manifestations. Proteome analysis showed increased expression of proteins involved in DNA repair, inflammation and the antioxidant response in these subtypes. In addition, there was altered expression across the subtypes of several transcription factors involved in cellular metabolism. The differentially expressed proteins were tightly associated in our network with the metabolic transcription factors demonstrating significant influence within the subnetwork. Increasing evidence highlights the importance of metabolic processes and mitochondrial dysfunction in rheumatic disease.30 Inflammation, mitochondrial function and redox imbalance are intrinsically linked to each other and the DNA damage response and altered energy metabolism. Mitochondrial and DNA damage can drive inflammatory responses. Likewise, mitochondrial dysfunction and DNA damage can be caused by chronic exposure to inflammation, including type I IFNs.30 Constitutive immune system activation can shift balance of glycolysis and oxidative phosphorylation, altering ATP return, the cellular redox environment (via ROS production and antioxidant consumption) and biomolecule availability. Mitochondrial dysfunction and altered metabolic processes have already been linked to proinflammatory manifestations in SLE and RA.30 Altered energy metabolism, mitochondrial dysfunction and redox imbalance potentially play important roles in the fatigue and pain experienced by the HSB and PDF subtypes (figure 5). Despite this, these processes are poorly understood in SS and remain unaddressed by current therapeutics. Our data support the examination of differences in metabolism and mitochondrial function in the context of SS subtypes.
Proposed factors contributing to the symptomology of HSB and PDF subtypes. Altered metabolism, mitochondrial dysfunction, chronic inflammation and oxidative stress all represent potential pathways that may contribute to the non-glandular manifestations. It is possible that it is the protective counter-response to features like oxidative stress, chronic inflammation or DNA damage that is responsible for the symptomatology. HSB, high symptom burden; PDF, pain dominant with fatigue.
Coexisting anxiety and depression are common in SS and may share similar underlying molecular networks. Our data show altered expression of proinflammatory cytokines as well as proteins involved in kynurenine, glucocorticoid and catecholamine metabolism between the subtypes. Whether these pathways represent a causal link with depression and anxiety or are simply correlated with ongoing inflammation requires further investigation. Therapeutics directed at cytokines are of intense interest in mood disorders and fatigue, and the increased expression of IL-6 and IL-1a in the subtype reporting the highest depression, anxiety and fatigue, alongside tocilizumab trial reanalysis showing a reduction in the fatigue score within this subtype, suggest potential merit in revisiting trials of monoclonal antibodies against these cytokines with stratification tools and revised endpoints. Peripheral IL-6 antagonism must, however, be approached with caution as it may lead to increased IL-6 in the CNS and a worsening of depressive symptoms.31 In a small cohort trial of IL-1 receptor antagonist, Anakinra, the Fatigue Severity Scale did not differ from the control group at 4 weeks, though significantly more patients in the Anakinra group had a fatigue reduction of more than 50% when using VAS.29 Administration of IL-1 receptor antagonists reduced fatigue in RA patients and suitably powered, stratified trials are required in SS. Work revisiting the JOQUER trial showed Hydroxychloroquine had some efficacy in improving ESSPRI scores for the HSB subtype.4 Hydroxychloroquine has several proposed mechanisms including altering lysosomal activity, inhibiting TLR-7 and 9 signalling pathways and interfering with cyclic GMP-AMP synthase signalling. Long-term treatment of hydroxychloroquine in RA has been shown to reduce the circulating levels of IL-1 and IL-6.32 Also of note is hydroxychloroquine’s beneficial metabolic effects in several studies.32 Whether hydroxychloroquine improves ESSPRI scores in the HSB subtype by one or a combination of these mechanisms requires further investigation.
This study is not without limitations, as the clinical phenotype in SS is likely to be underpinned by networks of dysregulated biological pathways rather than one or a few pathways. We used a high-throughput proteome analysis which allowed us to profile a wide array of proteins covering many biological pathways and processes, however, expression and ARACNE analysis of the entire proteome was beyond the scope of this study. Consequently, certain pathways and the expression of potentially important mediators of SS may not have been examined. Furthermore, our work gives a static picture of the interconnected pathways. The sensitivity of the network to change is of vital importance and future work should focus on validating the proteome in additional cohorts, as well as examining it longitudinally and in response to specific treatments. Additionally, we acknowledge that examining blood proteome may not reflect the proteome in target organs, however, because the target organ for many of the debilitating symptoms (such as pain and fatigue) is unknown, blood is a reasonable starting point in the search for differences between symptom-based subtypes.33 Finally, this is study of SS subtypes, and while differential expression between SS patients and healthy controls have been documented for several of the proteins discussed, how the protein network varies between SS patients and healthy individuals or non autoimmune sicca patients requires further study.
In summary, distinct clinical subtypes have now been established by several independent groups.1 3 In this article, we demonstrate each subtype has a unique protein network with differences in chemokines, cytokines and master transcription factors. Furthermore, these differences provide plausible mechanisms to explain the distinct clinical phenotypes. These findings have relevance for trial design and therapeutic development, as the molecular networks provide clues to the pathways driving the glandular and non-glandular manifestations of SS and to potentially effective therapeutic targets for different SS subtypes.
Data availability statement
No data are available.
Ethics statements
Patient consent for publication
Ethics approval
This study involves human participants and research ethics approval was granted by the UK North-West Research Ethics Committee. All participants provided informed consent. Participants gave informed consent to participate in the study before taking part.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Handling editor Josef S Smolen
Twitter @DocFelten
Contributors This work was drafted and designed by JT, JC and W-FN. Experimentation and analysis were completed by JSB, JC, JT, LS and KW. JSB and JT wrote the original manuscript. JC, J-EG and W-FN provided revisions. All authors read, contributed and approved the final version. W-FN is the author acting as guarantor.
Funding This work was supported in part by FOREUM, MRC (grant reference: 0800629) NECESSITY, NIHR Biomedical Research Centre at Newcastle University and the Newcastle upon Tyne Hospitals NHS Trust. The reanalysed Tocilizumab trial was sponsored by Hôpitaux Universitaires de Strasbourg. Roche Chugai provided Tocilizumab and the placebo and a grant to fund the study but had no role in the study design, data collection, analysis, interpretation or manuscript preparation, revision or approval of the manuscript. The French patient’s association (Association Française du Gougerot-Sjogren et des Syndromes Secs, AFGS) gave a grant to fund the study.
Competing interests VD-P has undertaken clinical trials and provided consultancy or expert advice in the area of Sjogren’s disease to the following companies: MedImmune, UCB, Abbvie, Sanofi, Novartis and BMS. BF has undertaken consultancy for Novartis, BMS, Servier, Galapagos, Roche, Sanofi and Janssen and received research funding from Janssen, Galapagos, Servier and Celgene. SJB has undertaken consultancy for Novartis, BMS, Iqvia and Janssen and accepted hospitality from Novartis to attend a working dinner at the BSR conference regarding secukinumab in inflammatory arthritis in the last year. MB has received consultancy and/or advisory board fees and/or grant support from GSK, Janssen, Ono Pharmaceutical, Horizon Therapeutics in the last two years. W-FN has undertaken clinical trials and provided consultancy or expert advice in the area of Sjogren’s syndrome to the following companies: GlaxoSmithKline, MedImmune, UCB, Abbvie, Roche, Eli Lilly, Takeda, Resolves Therapeutics, Sanofi, Novartis and BMS. No other potential conflict of interest relevant to this article was reported.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.