Blood samples from participants were obtained after receiving written informed consent and following approval of the Institutional Review Board at the Veterans Affairs (VA) Nebraska-Western Iowa Health Care System and University of Nebraska Medical Center (UNMC). Blood was collected between January 2022 and January 2023. Details of the enrollment criteria have been previously published (9). Briefly, all patients with RA fulfilled the 1987 American College of Rheumatology classification criteria (22). The presence of RA-ILD was confirmed by a board-certified subspecialist (pulmonologist or rheumatologist) and the presence of supportive chest computed tomography (CT) findings. Patients were defined as having RA without ILD in the absence of a clinical diagnosis of ILD or past chest imaging findings suggestive of ILD. All patients with IPF without clinically apparent autoimmune disease were diagnosed by a board-certified pulmonologist and had confirmatory findings of UIP by chest CT imaging. However, these IPF patients were not systematically tested for circulating autoantibodies. A convenience sample of “control” comparator patients was recruited from the VA allergy clinic and included patients with allergic rhinitis on specific allergy desensitization and in the absence of clinically apparent lung disease and/or autoimmunity. Venous blood samples were collected in EDTA tubes and processed within 2 hours of collection.

Participants completed the University of California, San Diego (UCSD) Shortness of Breath Questionnaire (SOBQ), a 24-item dyspnea questionnaire with scores ranging from 0 to 120, with higher scores indicating greater dyspnea (23). For patients with RA, clinical data, disease severity measures, and medication use were recorded by the treating rheumatologists. Lung function indices for RA-ILD and IPF were abstracted from patients’ medical records.

Peripheral blood mononuclear cells (PBMCs) were first isolated from peripheral blood by Histopaque-1077 (Sigma-Aldrich, Burlington, MA) density gradient centrifugation. PBMCs were washed and CD14⁺ monocytes were selected using MACS magnetic microbead positive selection kit (Miltenyi Biotec, Germany). Post-selection monocyte purity was determined using Diffquick staining and enumeration of cytospin preparations. Monocyte purity ranged from 85.6% to 90.6%. Next, CD14⁺ selected monocytes were washed and lysed with RLT buffer containing β-mercaptoethanol for RNA isolation as per manufacturer’s instructions with Qiagen RNAeasy Micro Kit (Qiagen, Germantown, MD).

Quality and quantity of total RNA was evaluated using a Fragment Analyzer (Agilent, Santa Clara, CA) and Nanodrop (ThermoFisher), respectively. Total RNA (>4 ng/μl) was hybridized and processed per the manufacturer’s suggested protocol with capture and reporter probes to prepare target-probe complexes using reagents from the Human Innate Immune Myeloid (770 genes) and Human Fibrosis (770 genes) profiling panel (NanoString, Seattle, WA). There were 175 overlapping genes between panels and gene data from first panel (human innate immune myeloid panel) were utilized for combined analyses, noting congruency between panels. Sequencing libraries were generated by the UNMC Next Generation Sequencing Core beginning with 500 ng of total RNA from each sample using the NuGEN Universal Plus mRNA-Seq library kit from TECAN (Redwood City, CA) following the manufacturer’s recommended procedure. Resulting libraries were assessed for size of insert by analysis of an aliquot of each library on a Bioanalyzer instrument (Agilent Technologies, Santa Clara, CA). Each library had a unique indexing identifier barcode allowing the individual libraries to be multiplexed together for efficient sequencing. Multiplexed libraries were sequenced on a single 150 cycle mid-output flow cell of the NextSeq550 Sequencer (Illumina) using a 2 x 75 bp paired end protocol to generate a total of approximately 50 million pairs of reads for each sample.

For NanoString analysis, the nSolver Analysis Software was utilized to manage and analyze data. After importing raw data, experimental properties included background subtraction parameters: 30, normalized mean type: geometric mean, positive control normalization, threshold minimum: 0.3 and threshold maximum: 3.0, code set content normalization threshold minimum: 0.1 and maximum: 10, ratio data parameters: all pairwise ratios. Analysis properties included normalization of data to housekeeping genes, utilization of gene-specific z-score genes, exclusion of controls, and preset exclusion of genes with less than 10% of samples outside of the interval specified by the upper and lower bound values (myeloid innate panel lower bound 1.0 to upper bound 323,700 and fibrosis panel lower bound 1.0 to upper bound 64,364) In total, this included 736 for myeloid innate panel and 301 for fibrosis panel. Using unsupervised, machine-based, hierarchical agglomerative clustering with settings of linkage type of average and metric type of Pearson correlation, clustering by participant was performed with corresponding data visualization by heat map. Next, using the same settings for individual participant clustering, gene transcript average counts were normalized to housekeeping genes by participant group. Subsequent hierarchical agglomerative clustering was conducted for the entirety of the gene set as well as for the 30 genes with the highest average gene transcripts, which is referred to as “top differentially expressed genes (DEGs)”. Next, ratio data to include log₂ ratio with p-value determined from NSolver software analysis was exported for subsequent supervised analyses. Volcano plots and heatmaps of statistically significant (p value < 0.05 or -log₁₀ p value > 1.3) and log₂-fold changes were generated using GraphPad Prism software, version 10.2.2 (GraphPad, San Diego, CA, USA). Gene enrichment analyses were performed using Ingenuity Pathway Analysis (IPA; Qiagen Inc., https://www.qiagenbioinformatics.com/products/ingenuity-pathway-analysis). The R package GOPlot was utilized to visualize the relationship between genes and enriched pathways.

Characteristics of the 19 study participants (5 with RA-ILD, 5 with RA in absence of ILD, 5 with IPF, and 4 controls) are summarized in Table 1. Participants with RA-ILD, RA, and IPF were older than controls, though age was overlapping among the disease groups. Participants across groups were predominately male and self-reported White race, reflecting characteristics of the larger VA population. A history of ever smoking was higher among participants with RA and RA-ILD followed by those with IPF and lowest in controls. Shortness of breath, as assessed by the UCSD-SOBQ, although generally low, demonstrated increased symptoms (total score) in patients with RA-ILD, RA, and IPF vs. controls with the highest symptomatic dyspnea observed in those with IPF. Based on CT findings, RA-ILD patients were categorized as having 1) typical UIP pattern with honeycombing, reticular pattern with peripheral traction, and/or fibrosis (n=4, 80%) or 2) probable UIP pattern (indeterminate phenotype) with bibasilar interstitial thickening and subpleural reticulation (n=1, 20%) (24, 25). Further characteristics of patients with RA and RA-ILD are listed in Supplemental Table 1.

Gene expression profiles of circulating CD14⁺ monocytes from patients with RA-ILD were compared to those with RA without lung disease, IPF without autoimmunity, and controls for 770 myeloid innate genes and 770 fibrosis genes. Heatmap analyses with corresponding hierarchical dendrograms demonstrate the unsupervised clustering of the NanoString Human Myeloid Innate (Figure 1A) and Fibrosis (Figure 1B) profiling panels of the gene set of each subject as well as the top 30 Differentially Expressed Genes (DEGs) by myeloid innate (Figure 1C) and fibrosis (Figure 1D) panels. Note, this hierarchical dendrogram clustering by participant group was the same for the full gene set (data not shown) and top 30 DEGs shown here. Unsupervised myeloid innate gene analysis of monocytes demonstrated a clear stratification of patients with RA-ILD clustering together with IPF as opposed to RA in the absence of ILD (RA alone) or controls whereas RA alone aligned more closely with controls. In comparison, the fibrosis genes of monocytes from each patient with RA-ILD clustered together without clear associations with any of the other comparator patient groups. Controls and patients with RA alone demonstrated visual stratification together whereas patients with IPF did not demonstrate observable clustering of monocytes by the fibrosis genes interrogated.

By these machine-based analyses, the relative expression of inflammatory innate immune genes (DUSP1, NAMPT, CASP1, ANXA1, LTA3) and fibrosis genes (LYN, IQGAP1, VIM, FCGR3A/B, PECAM1, STAT1 and HLA-A/G/BE) were increased in the monocytes of RA-ILD patients. Relative to the comparator groups, there was reduced DEG expression in RA-ILD monocytes for a large number of genes.

DEGs of circulating CD14⁺ monocytes of RA-ILD, RA, and IPF participants were compared with controls with manual curation of the significant (p<0.05) top upregulated and top downregulated genes identified by myeloid innate and fibrosis profile panels (Figure 2A). By fold-change (log₂-fold) of participant group to controls, 27, 35, and 37 myeloid innate (panel) and 77, 15, and 32 fibrosis (panel) up-regulated genes with 50, 9, and 11 myeloid innate and 17, 5, and 14 fibrosis downregulated genes were identified in monocytes of RA-ILD, RA, and IPF (respectively) (Supplementary Tables 2–7). Correspondingly, comparisons were also conducted for RA-ILD vs. RA and RA vs. IPF by myeloid innate and fibrosis profile panels (Figure 2B, Supplementary Tables 8–11). For RA-ILD vs. RA, there were 15 and 25 upregulated and 28 and 12 downregulated genes (myeloid innate and fibrosis panels, respectively). For RA-ILD vs. IPF, there were 3 and 25 upregulated and 19 and 20 downregulated (myeloid innate and fibrosis panels, respectively).

The top 10 DEGs from each panel identified in RA-ILD patients (vs. controls) are summarized in Table 2 (and shown in Figure 2) with name and/or main biologic processes. Genes identified include those involved in regulating inflammation and fibrosis (SOCS3, CECAM1, LTB4R2, CLEC7A, IRF7, PHYKPL, GBP5, RAPGEF2, HIVEP1), epigenetic modification (KDM5D, KMT2D, OGT), and M1-like monocyte/macrophage activation (GBP5, RAPGEF2 (Ras-activators), FCGR1A, SOCS3)

Next, the DEGs for patients (RA-ILD, RA, and IPF) vs. controls for the myeloid innate and fibrosis profile panel were combined for a total of 1365 unique genes. In total, there were 103 upregulated and 66 downregulated genes for RA-ILD, 66 upregulated and 14 downregulated for RA, and 64 upregulated and 25 downregulated for IPF, with shared genes across patient groups represented by Venn diagram (Figure 3). There were 12 upregulated genes shared by RA-ILD, RA, and IPF (FCGR1A, CEACAM1, ANOS1, NOS2, HDAC4, CFLAR, GCNT1, STATE3, TLR2, CHUK, PTPN1, and NUMB), 3 upregulated genes shared by RA-ILD and RA (SERPINE 3, NECTIN2, KDM3B) and 12 upregulated genes shared by RA-ILD and IPF (IRF7, PTPRC, JAK2, JAK3, IL4R, ACAD10, RICTOR, MAPK14, EP400, CASP4, CDKLN2C, CSNK1D). There were fewer shared downregulated genes, with 2 downregulated genes shared by RA-ILD, RA, and IPF (CD1C, CD4), 1 gene (COL4A2) shared by RA-ILD and RA, and 8 genes (CD34, LGALS3, NFYC, SERPINF1, ALAD, GNG2, RPLP0, HADH) shared by RA-ILD and IPF. Thus, there were numerically more DEGs shared between patients with RA-ILD and IPF as compared to RA.

To achieve further insights into the functional consequences of differential gene expression in CD14⁺ monocytes of RA-ILD participants as well as that observed in RA and IPF, a pathway enrichment analysis using all the combined DEGs from the myeloid innate immunity and fibrosis profile panels was performed. For RA-ILD participants, this analysis confirmed the differential regulation of genes involved in macrophage differentiation and activation in RA with the following signaling pathways implicated: IL-12, glucocorticoid receptor, and IL-27 signaling pathways; neuroinflammatory pathways were also implicated (Figure 4). Pathways involved in regulating epithelial mesenchymal transition by growth factors was also implicated, potentially suggesting a role for macrophages in orchestrating wound repair, regulation, and fibrosis important in ILD development. For RA patients without lung disease, perturbation of the following signaling pathways was demonstrated: pathogen induced cytokine, immunogenic cell death, Th1 and Th2, macrophage classical activation, IL-27, and interferon regulatory pathways (Supplementary Figure 1, top panel). For IPF patients, neuroinflammation, production of nitric oxide and reactive oxygen species and iNOS, IL-12, IL-27, and macrophage alternative activation signaling pathways were significantly affected (Supplementary Figure 1, bottom panel). Thus, alteration of the IL-12, IL-27 and neuroinflammation signaling pathways was shared between RA-ILD and IPF.