Abstract

   Rheumatoid arthritis (RA) is an autoimmune disease that exhibits a high
   degree of heterogeneity, marked by unpredictable disease flares and
   significant variations in the response to available treatments. The
   lack of optimal stratification for RA patients may be a contributing
   factor to the poor efficacy of current treatment options. The objective
   of this study is to elucidate the molecular characteristics of RA
   through the utilization of mitochondrial genes and subsequently
   construct and authenticate a diagnostic framework for RA. Mitochondrial
   proteins were obtained from the MitoCarta database, and the R package
   limma was employed to filter for differentially expressed mitochondrial
   genes (MDEGs). Metascape was utilized to perform enrichment analysis,
   followed by an unsupervised clustering algorithm using the
   ConsensuClusterPlus package to identify distinct subtypes based on
   MDEGs. The immune microenvironment, biological pathways, and drug
   response were further explored in these subtypes. Finally, a
   multi-biomarker-based diagnostic model was constructed using machine
   learning algorithms. Utilizing 88 MDEGs present in transcript profiles,
   it was possible to classify RA patients into three distinct subtypes,
   each characterized by unique molecular and cellular signatures. Subtype
   A exhibited a marked activation of inflammatory cells and pathways,
   while subtype C was characterized by the presence of specific innate
   lymphocytes. Inflammatory and immune cells in subtype B displayed a
   more modest level of activation (Wilcoxon test P < 0.05). Notably,
   subtype C demonstrated a stronger correlation with a superior response
   to biologics such as infliximab, anti-TNF, rituximab, and
   methotrexate/abatacept (P = 0.001) using the fisher test. Furthermore,
   the mitochondrial diagnosis SVM model demonstrated a high degree of
   discriminatory ability in distinguishing RA in both training
   (AUC = 100%) and validation sets (AUC = 80.1%). This study presents a
   pioneering analysis of mitochondrial modifications in RA, offering a
   novel framework for patient stratification and potentially enhancing
   therapeutic decision-making.

Supplementary Information

   The online version contains supplementary material available at
   10.1186/s12967-023-04426-7.

   Keywords: Rheumatoid arthritis, Mitochondrial proteins, Immune
   microenvironment, Unsupervised machine learning, Stratification

Introduction

   Rheumatoid arthritis (RA) is a heterogeneous and prevalent autoimmune
   inflammatory arthritis [[35]1], leading to a rise in the number of
   disabled life years attributed to RA worldwide. However, these trends
   exhibit regional and national variations. Furthermore, RA is an
   autoimmune disease with an unknown etiology, and past risk factors
   include respiratory exposure, genetics, intestinal health, oral health,
   gender, lifestyle, and habits [[36]2].

   Currently, Nonsteroidal anti-inflammatory drugs [[37]3],
   Glucocorticoids [[38]4, [39]5], and Disease-Modifying Anti-Rheumatic
   Drugs (methotrexate, sulfasalazine, minocycline, hydroxychloroquine,
   and azathioprine) are commonly utilized as the primary pharmacological
   interventions for managing patients with RA These drugs exert their
   therapeutic effects through immunosuppressive and anti-inflammatory
   mechanisms [[40]6–[41]9]. Despite the wide range of treatment options
   available to RA patients, the current standard treatment regimen is
   associated with a multitude of adverse effects [[42]10]. In the realm
   of research, the utilization of big data has the potential to unveil
   novel (sub-) phenotypes in unsupervised analyses, thereby enhancing
   precision in medical interventions through the facilitation of
   innovative targeted therapeutic strategies. Dana E Orange et al.
   conducted comprehensive analyses of patient samples, leading to the
   identification of three distinct subtypes of rheumatoid arthritis, with
   strong associations observed between these subtypes and disease
   activity [[43]11]. Additionally, Rodrigo Cánovas et al. [[44]12]
   discovered different subtypes of juvenile idiopathic arthritis, which
   has significantly contributed to the advancement of our understanding
   of this disease. Therefore, it is essential to understand RA subtypes
   and their molecular characterizations in order to better select
   patients and develop individualized therapy based on phenotypes and
   molecular signatures.

   The disruption of mitochondrial homeostasis has been implicated in the
   development of RA [[45]13–[46]15]. The imbalance of the endostatin
   environment resulting from mitochondrial impairment plays a crucial
   role in the pathology of RA [[47]16]. In the context of RA, METTL3 is
   responsible for mediating inflammatory responses by activating the
   NF-κB pathway and facilitating FLS activation [[48]17]. The heightened
   expression of SIRT4 promotes the secretion of TNF-a and IL-6, thereby
   expediting the process of bone destruction in individuals with
   osteoarthritis [[49]18, [50]19]. Moreover, PTEN Methylation has been
   found to promote inflammation and the activation of fibroblast-like
   synoviocytes in Rheumatoid Arthritis [[51]20]. Both mitochondrial
   metabolism and immune-inflammation are significant pathogeneses of RA.
   However, their interplay in RA remains unexplored and necessitates
   further investigation.

   This study employed unsupervised clustering methods to identify
   different subtypes in patients with RA based on mitochondrial gene
   expression profiles from whole blood. The subtypes were thoroughly
   characterized using cellular, molecular, and clinical features to gain
   a deeper understanding of the underlying biological mechanisms. The
   identified characteristic genes were then applied to independent groups
   of RA patients to evaluate the therapeutic outcomes of conventional
   triple Infliximab and anti-TNF. Additionally, machine learning was
   utilized to develop a diagnostic tool based on the identified features.
   This study aims to provide a reference for clinical precision treatment
   and early diagnosis of RA patients.

Materials and methods

Processing of RA gene expression data

   The Gene Expression Omnibus (GEO) database furnished microarray gene
   expression data for rheumatoid arthritis samples, with a comprehensive
   account of the study design, data preprocessing, and data
   interpretation for the six microarray datasets ([52]GSE110169,
   [53]GSE93272, [54]GSE58795, [55]GSE15258, [56]GSE37107, and
   [57]GSE68215 in Additional file [58]1: Table S1). Several biologic
   agents were included, namely: Infliximab ([59]GSE58795), anti-TNF
   ([60]GSE15258), rituximab ([61]GSE37107), methotrexate/abatacept
   ([62]GSE68215). Drug information was extracted from medical records.
   Additionally, microarray datasets [63]GSE110169 and [64]GSE93272 were
   segregated into training and test datasets. To mitigate background
   noise and normalize quantiles for microarray data, we retrieved raw
   files in ‘CEL’ format and employed the Affy and Simpleaffy packages for
   robust multiarray averaging.

Differentially expressed mitochondrial genes: screening and function and
pathway enrichment analysis

   The present study employed the Mitocarta 3.01 database to identify gene
   sets that are associated with mitochondria, with a specific focus on
   the 1,136 unique human mitochondrial genes [[65]21]. The R package
   limma was utilized to filter differentially expressed genes that are
   linked to mitochondria between samples of individuals with RA and
   healthy control samples. False-positive outcomes were corrected using
   the false-discovery rate (FDR). The criteria for identifying
   Mitochondria-associated differentially expressed genes (MDEGs) were an
   adjusted p-value of less than 0.05 and a log fold change (logFC) of
   greater than 0.32. To ascertain the enrichment of pathways, a Metascape
   analysis was executed for GO and KEGG pathways, wherein functional
   pathways with a p < 0.05 were deemed significantly enriched [[66]22].
   Pearson correlation coefficients were employed to scrutinize gene
   expression correlations.

Clustering of Mitochondria-related expression-driven subgroups in RA

   To gain further insights into molecular subtype heterogeneity within
   MDEGs profiles associated with RA, the R package ConsensuClusterPlus
   was utilized to perform hierarchical agglomerative clustering using the
   ‘km’ method, which is based on Euclidean distance, The parameter
   settings were as follows: maxK = 6, reps = 1000, pItem = 0.8,
   pFeature = 1, clusterAlg="km”, distance="euclidean”. The “km” option
   performs kmeans clustering directly on a data matrix, with items and
   features resampled. This process was repeated 1000 times to ensure
   clustering stability [[67]23]. The optimal cluster allocation was
   determined through the utilization of a cumulative distribution
   function (CDF). Principal component analysis (PCA) was employed to
   visualize the differences between subtypes. The identification of
   differentially expressed genes (MDEGs) was conducted across the three
   subtypes.

Characterization of RA subtypes based on cellular, molecular, and clinical
characteristics

   The present study assessed immune cell infiltration in patients with RA
   through the utilization of the ‘Xcell’ R package, which facilitated the
   computation of the enrichment of 64 immune genes [[68]24].
   Additionally, the immune function of three subgroups of participants
   was determined via single-sample gene set enrichment analysis (ssGSEA)
   [[69]25]. Pathways linked to RA were curated based on literature
   references and GSEA outcomes, and gene sets were sourced from the KEGG