Introduction

Interstitial lung diseases (ILDs) encompass a heterogeneous group of lung disorders, including connective tissue diseases (CTDs). These diseases vary widely in severity, progression rate, and clinical outcomes1. Some patients with ILD experience worsening respiratory distress, a decline in lung function, and progressive fibrosis on high resolution computed tomography (HRCT) despite conventional treatments such as corticosteroids and immunosuppressants2,3. This phenotype is termed as progressive fibrosing ILD (PF-ILD) and is characterized by poor prognosis and high mortality rate4,5.

CTDs such as rheumatoid arthritis (RA), dermatomyositis (DM), and systemic sclerosis (SSc) are often complicated by severe ILDs6. Some patients with CTD-associated ILD develop PF-ILD, similar to other ILDs, despite receiving immunosuppressive treatment7. Antifibrotic therapies such as nintedanib and pirfenidone suppress the progression of PF-ILD4,8. However, these medications face challenges in stopping the progression of fibrosis; thus, a comprehensive and effective treatment for PF-ILD is yet to be established.

PF-ILDs may share a common pathology regardless of the underlying disease; however, their pathogenesis and onset mechanisms remain unclear. Although several studies have identified biomarkers and clinical predictors of fibrosis development9,10,11, these initiatives have yet to produce reliable prognostic tools. The predictive factors are unclear, and the absence of appropriate biomarkers complicates the ability to forecast which patients are likely to experience fibrosis progression. Predicting PF-ILD at the time of ILD diagnosis would enable the selection of appropriate therapeutic options, potentially leading to improved patient outcomes.

Bronchoalveolar lavage fluid (BALF) is a pivotal research specimen for investigating the etiology of lung diseases, offering a less invasive procedure with a lower risk than lung tissue biopsies. BALF is instrumental for detecting malignancies, infections, and diagnosing ILDs and is particularly useful for patients exhibiting a probable usual interstitial pneumonia pattern on HRCT12,13,14. A recent study suggested that integrating BALF lymphocyte data with HRCT imaging findings could lead to prognostic predictions in patients with pulmonary fibrosis15. However, studies that use BALF to uncover the etiology of PF-ILDs are limited.

Single-cell RNA sequencing (scRNA-seq) is a recent technology that facilitates comprehensive examination of gene expression in individual cells16. This is useful for elucidating pathological conditions and identifying therapeutic targets. While this technology has been extensively applied in the study of various ILDs, including idiopathic pulmonary fibrosis (IPF), chronic obstructive pulmonary disease, and SSc-associated ILD17,18,19,20, single-cell analysis in the context of patients with PF-ILD remains underdeveloped.

The study aimed to analyze peripheral blood and BALF from ILD patients at the time of diagnosis using scRNA-seq to identify biomarkers that can predict PF-ILD. Furthermore, a detailed examination of immune cell function in patients with PF-ILD may provide valuable insights for identifying predictive biomarkers, optimizing treatment approaches, and contributing to improved patient outcomes. To the best of our knowledge, this is the first comprehensive single-cell analysis of blood and BALF samples from patients with PF-ILD. While the dataset analyzed in this study was also used in our previous research21, the objective and comparative focus of the present study are fundamentally different. In our previous research, we aimed to identify disease-specific immune cell features across various subtypes of CTD-ILDs, in order to elucidate the heterogeneity of their pathophysiology21. In contrast, the present study aims to identify immunological features and potential biomarkers that distinguish PF-ILD from non-PF-ILD, irrespective of the underlying CTD subtype. An important aspect of this study is that patients were categorized into PF-ILD and non-PF-ILD groups according to clinical follow-up data, rather than baseline diagnosis.

Results

Baseline characteristics of study participants

This study included 43 patients with ILD, 12 of whom fulfilled the criteria for PF-ILD. The baseline patient characteristics are presented in Table 1 and Supplementary Table S1 online. The patients with PF-ILD were older than those without PF-ILD. Additionally, a higher proportion of neutrophils was observed in the BALF of patients with PF-ILD. In the BALF, an increased percentage of neutrophils (> 3%) was observed in 12 (3 with RA, 3 with SJS, 2 with DM, 2 with AAV, 1 with IPF, and 1 with IIP) out of 43 patients, of whom 7 were included in the PF-ILD group. HRCT imaging revealed that the non-specific interstitial pneumonia pattern was predominant in both patient groups. However, there was an observed trend of increased usual interstitial pneumonia patterns in the PF-ILD group compared with the non-PF-ILD group. The organizing pneumonia pattern was not observed in the PF-ILD group. Furthermore, no significant differences were observed between the groups regarding underlying diseases, pulmonary function test results, blood test parameters, or smoking history. In the non-PF-ILD group, 1 patient diagnosed with RA and another with SSc received immunosuppressive treatment for symptoms other than pulmonary lesions at the time of enrolment. Excluding these 2 patients, the remaining individuals did not receive immunosuppressive treatment.

Table 1 Clinical characteristics of patients.
Full size table

Differences in immune cells proportions

Peripheral blood and BALF were procured from the 43 enrolled patients with ILD at the time of enrolment, and scRNA-seq was performed. To analyze gene expression within the immune cell populations of both blood and BALF in patients with ILD, Seq-Well, a portable platform for scRNA-seq, was utilized22. Major cell types were clustered and annotated based on marker gene expression. Blood cells were annotated as monocytes, dendritic cells (DCs), eosinophils, T/natural killer (NK) cells, and B/plasma cells using uniform manifold approximation and projection, according to previous methods18,21. The blood immune cell profiles of patients with PF-ILD exhibited a decrease in DCs compared to those of non-PF-ILD patients (Fig. 1a). Additionally, multicolor flow cytometry (MCFC) revealed that DCs were substantially fewer in the PF-ILD group compared to the non-PF-ILD group (Supplementary Fig. S1 online). Following our methodology used for blood analysis, we evaluated the distribution and gene expression profiles of immune cells in BALF and annotated them into mononuclear myeloid cells, neutrophils, eosinophils, mast cells, T/NKcells, B/plasma cells, and proliferating cells18. Amongst the major cell types, no significant differences were observed between the two patient groups (Fig. 1b). Furthermore, MCFC analysis revealed no differences in major cell types between the two groups (Supplementary Fig. S2 online).

Fig. 1
figure 1

Single-cell RNA-sequencing analysis of blood samples and bronchoalveolar lavage fluid (BALF) samples from 43 patients with interstitial lung disease (ILD). a) Comparative analysis of the proportions of immune cells of blood cells in patients with non-progressive fibrosing (PF) ILD and patients with PF-ILD using the Wilcoxon rank-sum test. b) Comparative analysis of the proportions of immune cells of BALF cells in patients with non-PF ILD and patients with PF-ILD using the Wilcoxon rank-sum test.

Full size image

This study detailed the subclassification of cell groups by analyzing distinct gene expression patterns, as outlined in previous methods18. We conducted a comparative analysis of the cell subtypes between the PF-ILD and non-PF-ILD groups (Fig. 2, Fig. S3 and S4 in the Supporting Information). Mononuclear myeloid cells, including monocytes and macrophages, were further classified into multiple subtypes based on their marker gene expression profiles21. Within the PF-ILD group, there was an observable trend towards an increased prevalence of macrophages expressing various chemokines, including C-X-C motif chemokine ligand (CXCL) 10, C-C motif chemokine ligand (CCL) 4, and CCL13. In contrast, nuclear enriched abundant transcript 1 + macrophages were less prevalent in the PF-ILD group than in the non-PF-ILD group (Fig. 2a). A significant finding in the comparison of B/plasma cell subtype classifications in BALF was the increased presence of plasma cells in the PF-ILD group compared to the non-PF-ILD group (P = 0.0054) (Fig. 2b and c, and 2d).

Fig. 2
figure 2

Subtypes of mononuclear myeloid cells and B/plasma cells in bronchoalveolar lavage fluid (BALF). a) Comparative analysis of the proportions of mononuclear myeloid cells in BALF between patients with non-progressive fibrosing (PF) interstitial lung disease (ILD) and patients with PF-ILD. (b) Uniform manifold approximation and projection (UMAP) visualization of B/plasma cells in BALF. Five clusters were identified. Each cluster is represented by a unique color, number, and cell type. (c) Dot plots show the top marker genes per cluster. (d) Comparative analysis of the proportions of B/plasma cells between non-PF-ILD patients and PF-ILD patients. Abbreviations: PF: progressive fibrosing; ILD: interstitial lung disease; CXCL: C-X-C motif chemokine ligand; CCL: C-C motif chemokine ligand; NEAT: nuclear enriched abundant transcript; FABP: fatty acid binding protein; APOC: apolipoprotein C; SPP: secretory pathway peptidase; PPARG: peroxisome proliferator-activated receptor gamma; DC: dendritic cell

Full size image

Analysis of gene ontology enrichment

Following this, we investigated differences in gene expression between the PF-ILD and non-PF-ILD groups (Fig. 3). Variance in gene expression was analyzed using DESeq2 (Supplementary Tables S2 and S3 online). In the BALF of the PF-ILD group, compared to that of the non-PF-ILD group, the expression of genes related to cytokines and chemokines, such as interleukin (IL)−1, TNFAIP6, CXCL8, and CCL20, was increased in monocytes and macrophages. Moreover, the expression of XBP1 and ELL2, which are related to plasma cell differentiation, increased in B/plasma cells21,23.

Fig. 3
figure 3

Gene ontology (GO) enrichment analysis of differentially expressed genes in mononuclear myeloid cells and B/plasma cells in bronchoalveolar lavage fluid (BALF). The p-value cutoff for genes was set at 0.01. (a) Dot plot, (b) Gene-Concept Network visualization of significantly enriched GO terms of mononuclear myeloid cells in BALF. (c) Dot plot, (d) Gene-Concept Network visualization of significantly enriched GO terms of B/plasma cells in BALF. Abbreviations: PF: progressive fibrosing; ILD: interstitial lung disease.

Full size image

To gain deeper insight into these differential gene expression patterns, we conducted a gene ontology (GO) enrichment analysis. This analysis of blood cells did not identify any terms uniquely associated with the PF-ILD group among the various cell groups examined. In contrast, the GO enrichment analysis of BALF identified notable findings in the PF-ILD group compared to the non-PF-ILD group: mononuclear myeloid cells exhibited enrichment terms related to various aspects of the immune response (Fig. 3a and b), predominantly related to innate immunity, adaptive immunity, and complement activation; and B/plasma cells were substantially enriched for terms related to the endoplasmic reticulum (Fig. 3c and d).

Analysis of cytokine and chemokine levels and complement activation

RNA-seq analysis revealed disparities in immune cell fractionation and gene expression between patients with and without PF-ILD. To further elucidate these differences, we quantified levels of cytokines and chemokines in the plasma and BALF supernatants using an enzyme-linked immunosorbent assay (ELISA), allowing for a comparative analysis between the two groups (Fig. 4, Supplementary Fig. S5 online). Within the BALF of patients with PF-ILD, a notable abundance of several cytokines was observed. Specifically, the level of IL-6 was significantly elevated (P = 0.0039), while CXCL10 and arginase were markedly increased (P = 0.012, P = 0.049). In contrast, cytokine levels in the plasma samples showed no significant differences between the two groups. Considering the pronounced expression of genes related to the complement system in monocytes and macrophages in BALF, along with recent insights into the association between lung injury and the complement system24, we also analyzed complement components. ELISA results indicated elevated levels of complement components C3a, C4a, and C5a in the BALF of patients with PF-ILD compared to those of patients without PF-ILD (P = 0.052, P = 0.0093, P = 0.0028). However, in the plasma samples, complement levels did not differ significantly between the groups. To explore the relationship between systemic and pulmonary inflammation, we analyzed the correlations between plasma and BALF concentrations of biomarkers measured by ELISA in all 43 participants, using Spearman’s correlation coefficients (Supplementary Tables 4 and 5). Several cytokines and chemokines, including IP-10, TARC, IL-23, TNF-α, IL-6, and IFN-γ, showed significant positive correlations, suggesting that these inflammatory mediators tend to increase in parallel in both plasma and BALF compartments. In contrast, C3a showed a significant negative correlation, indicating compartmentalized complement activation in the lung. The other 10 biomarkers showed no significant correlations.

Fig. 4
figure 4

Measurement of cytokines and chemokines, and complement components in plasma and bronchoalveolar lavage fluid (BALF) supernatants using the enzyme-linked immunosorbent assay method. a) Comparative analysis of IL-6, CXCL10 and Arginase between patients with non-progressive fibrosing (PF) interstitial lung disease (ILD) and patients with PF-ILD. b) Comparative analysis of C3a, C4a, C5a, and C1q between patients with non-PF-ILD and patients with PF-ILD patients. Abbreviations: PF: progressive fibrosing; ILD: interstitial lung disease; IL: interleukin; CXCL: C-X-C motif chemokine ligand.

Full size image

As a further consideration, to account for potential bias, we reanalyzed the data excluding the 3 IPF cases, which showed similar results and indicated that their inclusion did not substantially influence the study’s main findings (Supplementary Table S6 online, Supplementary Fig. S6 online).

Discussion

In this study, we conducted a comprehensive analysis of blood and BALF from patients with ILDs using scRNA-seq. Mononuclear myeloid cells in the BALF were classified into 15 distinct subtypes. We compared the immune cell phenotypes and gene expression patterns in patients who fulfilled the PF-ILD criteria with those in patients who did not fulfil them. In the BALF from the PF-ILD patient group, a higher prevalence of macrophages expressing chemokines and cytokines, including CXCL10, CCL4, and CCL13, was observed than that in the non-PF-ILD group. CXCL10, also referred to as interferon gamma-induced protein 10, is a chemokine secreted by various cell types in response to interferon gamma. It belongs to the CXC chemokine family and plays a pivotal role in the pathogenesis of autoimmune diseases such as RA, SLE, and SSc25. High levels of CXCL10 have been observed in the serum of anti-melanoma differentiation-associated gene 5 antibody-positive patients with rapidly progressive ILD26, in the serum of patients with RA, and in both the serum and BALF of patients with SSc27,28,29. The present study confirmed an increase in the expression of CXCL10 + macrophages within the BALF and elevated levels of CXCL10, as detected by ELISA, suggesting their significant involvement in the pathogenesis of PF-ILD. These findings suggest that in patients with PF-ILD, macrophages may induce a more active inflammatory response by recruiting immune cells, such as neutrophils and lymphocytes, through the production of cytokines and chemokines. The activation of innate and adaptive immunity, as indicated by the GO analysis of mononuclear myeloid cells, is consistent with the presence of an enhanced inflammatory response in PF-ILD.

In this study, we discovered that the BALF of the PF-ILD group contained a significantly higher number of plasma cells than that of the non-PF-ILD group. In the PF-ILD group, GO enrichment analysis of B cells revealed significant enrichment of terms related to the endoplasmic reticulum. Considering that the endoplasmic reticulum is abundant in plasma cells30, it is possible that the increased plasma cell count in patients with PF-ILD contributes to this difference in gene expression. Furthermore, gene expression analysis of the entire B cell population revealed increased expression of XBP1, a key factor in plasma cell differentiation and antibody production21. This implies that the B cells in the lungs of patients with PF-ILD may differentiate into plasma cells. Previous reports have shown that in patients with IPF, B cells circulating in the bloodstream not only exhibit a higher degree of antigen differentiation and greater plasmablast proportions than those in controls, but the extent of B cell differentiation is also correlated with the patients’ lung volumes31. Additionally, previous research has indicated that the suppression of plasma cells can effectively reduce pulmonary fibrosis32, implying that B cells, particularly plasma cells, play a crucial role in the pathogenesis of PF-ILD.

We also observed a significant increase in the levels of IL-6 and complement factors C3a, C4a, and C5a in the BALF of patients with PF-ILD. IL-6 is a critical cytokine involved in various diseases33. IL-6 may be linked to pulmonary fibrosis in patients with IPF, and in IL-6-deficient mice, it has been shown that pulmonary fibrosis induced by bleomycin is suppressed, highlighting the vital role of IL-6 in lung fibrosis34,35. Taken together, these results strengthen the hypothesis that IL-6 plays a critical role in the development of pulmonary fibrosis. Increased concentrations of these complement components suggest activation of the complement system in the lungs. In contrast, no differences were observed in complement activation within the plasma, suggesting that complement activation primarily occurs locally in the lungs. Previous studies have indicated that the activation of the complement system is linked to the activation of tissue growth factor β, which is important in the pathogenesis of fibrosis36. Higher levels of C3 and C5a in both lung tissue homogenates and plasma derived from patients with IPF suggest local and systemic complement activation26. Furthermore, other research has similarly shown the activation of the complement system in patients with sarcoidosis and SSc complicated by ILD37,38. Systemic complement activation was not observed in the present study. This may be attributable to the collection of samples from patients in the initial stages of the disease, when inflammation and complement activation were limited to the lungs. In this study, comparative analysis of blood cells, unlike that of BALF, showed little difference between the PF-ILD and non-PF-ILD groups. These results suggest that changes occurring in the early stages of the disease may be localized in the lungs.

The lack of early stage treatment options for PF-ILD represents a significant challenge in the development of therapeutic strategies. Targeting inflammation at an early stage may prevent or mitigate subsequent fibrotic progression. Our findings indicate that changes in immune cells are observed at the time of diagnosis in patients with PF-ILD. Initially, there is an increase in inflammatory cytokines, such as IL-6, in the pulmonary environment of patients with PF-ILD, along with an increase in complement components and plasma cells. Therefore, therapeutic strategies that include anti-IL-6 medications, which are commonly used in the treatment of RA and other autoimmune disorders39, may be viable options for future treatment of PF-ILD. Targeting the complement system and plasma cells may also be effective treatment options.

Our study had some limitations. First, the study was limited to only two medical facilities and involved a small study population. The diversity of underlying diseases among the participants suggests the possibility of specific immunological responses based on disease background, which may have influenced the results of the study. However, due to the limited number of samples, a detailed examination of the differences between these diseases could not be conducted. The presence of patients with PF-ILD, regardless of their underlying diseases, suggests a common pathology independent of the underlying disease. To adequately assess the unique risks of PF-ILD that are independent of underlying diseases and to understand its pathology, more comprehensive and large-scale research is necessary. Second, the study only included patients eligible for BALF, potentially excluding those with critical systemic conditions. Third, this study focused on the initial diagnosis without tracking longitudinal changes, preventing the assessment of fibrosis progression over time. Fourth, the study did not include a healthy or disease control group, which limits our ability to determine whether the observed features are specific to PF-ILD. Future studies should include appropriate controls to clarify disease-specific signatures.

In conclusion, our study aimed to elucidate the pathophysiology of PF-ILD by analyzing blood and BALF samples using scRNA-seq and ELISA. Notably, differences were observed in the subsets and gene expression of mononuclear myeloid and B cells in the BALF of the PF-ILD group. Furthermore, the ELISA revealed elevated levels of specific cytokines and complement activation in the BALF of the PF-ILD group. These findings suggest that active inflammation in the lungs may be associated with the progression of pulmonary fibrosis and that analyzing BALF may be useful for assessing immune responses in the lungs. These findings are valuable for understanding PF-ILD pathogenesis, which is essential for developing effective treatment strategies and improving patient outcomes.

Materials and methods

Overview and study focus

In this study, we utilized the same dataset previously reported in our earlier publication21. All samples and experimental procedures—including BALF collection, sample processing, scRNA-seq, and downstream analyses—were performed exactly as described in the prior study. Although the dataset and experimental procedures remain unchanged, the current analysis was conducted with a different research objective. In particular, this study aims to identify immune features associated with fibrotic progression by classifying patients into PF-ILD and non-PF-ILD groups, irrespective of the underlying CTD subtype.

Experimental procedures and analytical methods

The procedure for collecting BALF samples, as well as the protocols for MCFC, scRNA-seq, ELISA, and the associated analytical procedures, were identical to those described in our previous study21, except for detection of differentially expressed genes, which utilized only the DESeq2 strategy. For cell-type annotation, the cell-type labels queried from the dataset18 were used for annotation. In addition, we validated these cell-type annotations using marker genes. Marker genes per cluster were defined as the most significant DE genes between identified clusters using a Wilcoxon rank sum test for differential gene expression implemented in Seurat. For a detailed characterization of the cells in the dataset, the cells of the identified major cell types were isolated, and scaling, dimensionality reduction using PCA, and data integration were performed. To annotate the subclusters in the major cell types, we used marker genes in combination with a priori knowledge from the public domain. For more details, please refer to that publication21.

Flowcytometry gating strategy and antibody panels are shown in Supplementary Fig. S7 and Supplementary Table S7 online.

Human specimens

Studies involving human participants were approved by the Ethics Committee of the Kyoto Prefectural University of Medicine (approval number: ERB-C-1471) and complied with the principles of the Declaration of Helsinki. Written informed consent was obtained from each patient prior to specimen collection. This study included 43 patients newly diagnosed with ILD who underwent bronchoscopy and BALF collection at two facilities: Graduate School of Medical Science, Kyoto Prefectural University of Medicine and Japanese Red Cross Kyoto Daiichi Hospital, between September 2019 and October 2021.

Study population and diagnostic criteria

This population included patient 31, who had been excluded from the previous study21, but otherwise remained consistent with previous research participants. Patients diagnosed RA, SSc, Sjögren syndrome (SJS), DM, antineutrophil cytoplasmic antibody-associated vasculitis (AAV), IPF, idiopathic interstitial pneumonia (IIP), and hypersensitivity pneumonitis followed specific classification criteria40,41,42,43,44,45,46,47,48,49,50, and the radiological diagnosis followed the methods from our previous study21.

Definition of PF-ILD

Our study focused on patients newly diagnosed with ILD to investigate the characteristics of PF-ILD. The follow-up period for this study was 24 months. Patients with PF-ILD were classified based on the INBUILD trial criteria4. as meeting at least one of the following criteria within the 24 months, despite standard treatment: a relative decline in the forced vital capacity of at least 10% of the predicted value; a relative decline in the forced vital capacity of 5% to less than 10% of the predicted value and worsening of respiratory symptoms or an increased extent of fibrosis on HRCT; or worsening of respiratory symptoms and an increased extent of fibrosis.

Statistical analysis for comparison

To compare categorical variables between the PF-ILD and non-PF-ILD groups, Fisher’s exact test was conducted. For continuous variables, normality was first assessed using the Shapiro-Wilk test. If the data followed a normal distribution, a t-test was performed, otherwise, the Wilcoxon rank-sum test was used. Statistical significance was defined as a p-value < 0.01.