Single-Cell Analysis Reveals Novel Immune Perturbations in Fibrotic Hypersensitivity Pneumonitis

Abstract

Rationale

Fibrotic hypersensitivity pneumonitis (FHP) is a debilitating interstitial lung disease driven by incompletely understood immune mechanisms.

Objectives

To elucidate immune aberrations in FHP in single-cell resolution.

Methods

Single-cell 5′ RNA sequencing was conducted on peripheral blood mononuclear cells and BAL cells obtained from 45 patients with FHP, 63 patients with idiopathic pulmonary fibrosis (IPF), 4 patients with nonfibrotic hypersensitivity pneumonitis, and 36 healthy control subjects in the United States and Mexico. Analyses included differential gene expression (Seurat), TF (transcription factor) activity imputation (DoRothEA-VIPER), and trajectory analyses (Monocle3 and Velocyto-scVelo-CellRank).

Measurements and Main Results

Overall, 501,534 peripheral blood mononuclear cells from 110 patients and control subjects and 88,336 BAL cells from 19 patients were profiled. Compared with control samples, FHP has elevated classical monocytes (adjusted-P = 2.5 × 10⁻³) and is enriched in CCL3^hi/CCL4^hi and S100A^hi classical monocytes (adjusted-P < 2.2 × 10⁻¹⁶). Trajectory analyses demonstrate that S100A^hi classical monocytes differentiate into SPP1^hi lung macrophages associated with fibrosis. Compared with both control subjects and IPF, cells from patients with FHP are significantly enriched in GZM^hi cytotoxic T cells. These cells exhibit TF activities indicative of TGFβ and TNFα and NFκB pathways. These results are publicly available at http://ildimmunecellatlas.com.

Conclusions

Single-cell transcriptomics of patients with FHP uncovered novel immune perturbations, including previously undescribed increases in GZM^hi cytotoxic CD4⁺ and CD8⁺ T cells—reflecting this disease’s unique inflammatory T cell–driven nature—as well as increased S100A^hi and CCL3^hi/CCL4^hi classical monocytes also observed in IPF. Both cell populations may guide the development of new biomarkers and therapeutic interventions.

At a Glance Commentary

Scientific Knowledge on the Subject

Hypersensitivity pneumonitis is an interstitial lung disease caused by repeated exposure to inhaled antigens in genetically predisposed individuals. In fibrotic hypersensitivity pneumonitis (FHP), the disease persists and manifests similarly to other fibrotic lung diseases, such as idiopathic pulmonary fibrosis. Prior research suggests that FHP lungs exhibit increased activated T cells and that peripheral blood mononuclear cells can differentiate progressive from nonprogressive FHP, but a detailed understanding of the immune aberrations in FHP is still missing.

What This Study Adds to the Field

This study provides the first extensive single-cell catalog of the immune aberrations that occur in the peripheral blood and BAL of patients with FHP. Using samples from the United States and Mexico, we identify cytotoxic T-cell populations that are distinct to FHP, as well as monocyte populations common to FHP and idiopathic pulmonary fibrosis that are connected to fibrosis-related macrophages in BAL. These findings discover adaptive immune responses that may be specific to FHP, as well as innate immune responses that seem to be more universally characteristic of pulmonary fibrosis. These results, as well as the wide availability of our data at http://ildimmunecellatlas.com, should encourage the development of novel biomarkers, diagnostics, and immune-modulatory therapies tailored at patients’ immune profiles.

Hypersensitivity pneumonitis is a complex interstitial lung disease driven by repetitive exposure to inhaled antigens in genetically susceptible individuals (1–7). The acute stage involves activation of alveolar macrophages, dendritic cells, and T helper cell type 1 (Th1) cells, mediated by immune complexes (1, 2, 5). In fibrotic hypersensitivity pneumonitis (FHP), which is less understood, persistent inflammation, facilitated potentially by Th2 and Th17 cells, is assumed to possibly lead to fibroblast activation, extracellular matrix (ECM) deposition, and lung fibrosis (1, 7–10).

Inflammation plays a critical role in FHP pathophysiology, which makes understanding its immune alterations important for diagnosis and treatment. Previous studies demonstrated that the lungs of patients with FHP contain increased activated T cells, together with an increase in terminally differentiated CD4⁺ and CD8⁺ T cells and enriched T cell–mediated inflammatory pathways (11–18). Research using bulk RNA sequencing of peripheral blood mononuclear cells (PBMCs) identified a 74-gene signature distinguishing progressive from nonprogressive forms of FHP (15, 16). This study indicated that PBMCs may be informative of immune mechanisms in FHP, but the detailed depiction of specific immune cell types involved is limited.

In this study, we aimed to provide a comprehensive and high-resolution atlas of immune cell alteration in FHP by conducting single-cell RNA sequencing (scRNAseq) of both PBMC and BAL samples obtained from a cohort of patients with FHP from the United States and Mexico. Overall, this analysis allows us to identify both hallmarks of fibrosis in classical monocytes and etiology-specific immune signatures in cytotoxic T cells, which should enable targeted diagnostic and therapeutic interventions. An online data-sharing portal modeled after the IPF Cell Atlas (http://ildimmunecellatlas.com) complements this study (19, 20). Some of our results have been previously reported in abstracts (21–24) and in the form of a preprint (25).

Methods

Patient Cohort

To study FHP’s immune landscape at patients’ initial presentation, PBMC and BAL samples were procured from healthy control subjects and patients within months of diagnosis. PBMCs were obtained from patients with FHP (n = 39), patients with IPF (n = 53), and healthy control subjects (n = 36) from the United States (New Haven, CT and Chicago, IL) and Mexico (Mexico City); BAL samples were from different patients with FHP (5), IPF (10), and nonfibrotic hypersensitivity pneumonitis (NFHP) (4) from Mexico (Tables 1 and 2). A multidisciplinary team confirmed diagnoses following established international guidelines (4, 6).

Table 1.

Summary of Demographic and Clinical Characteristics across All Cohorts for Peripheral Blood Samples, with P Values for Comparisons of Variables across Specified Groups

Patient Variable	CTRL (n = 27)	FHP (n = 37)	IPF (n = 45)	P Value
				FHP vs. CTRL	FHP vs. IPF	IPF vs. CTRL
Age, mean ± SD	69.44 ± 7.59	72.97 ± 11.98	68.95 ± 11.00	0.98	0.81	0.80
Sex
Male	17 (63)	12 (32)	31 (69)	0.022	1.2 × 10−3	0.62
Female	10 (37)	25 (68)	14 (31)
Race
White	26 (97)	36 (97)	44 (98)	1	1	1
Non-White	1 (3)	1 (3)	1 (2)
Ethnicity
Hispanic	9 (33)	10 (27)	7 (16)	0.59	0.09	0.28
Non-Hispanic	18 (67)	27 (73)	38 (84)
Smoking
Never	7 (26)	7 (19)	12 (27)	0.16	0.10	1
Ever	15 (56)	5 (14)	27 (60)
No data	5 (18)	25 (67)	6 (13)
% FVC, mean ± SD	92.11 ± 18.60	65.50 ± 16.74	72.97 ± 11.98	0.0037	0.17	0.015
% DlCO, mean ± SD	104.67 ± 14.29	42.58 ± 14.18	42.72 ± 12.84	1.5 × 10−8	0.97	8.8 × 10−8
GAP score
0	2 (7)	0 (0)	1 (2)	0.12	0.11	0.02
1	1 (4)	0 (0)	2 (4)
2	2 (7)	5 (14)	2 (4)
3	2 (7)	1 (3)	5 (11)
4	0 (0)	3 (8)	9 (2)
5	0 (0)	2 (5)	11 (24)
6	0 (0)	1 (3)	6 (13)
No data	20 (75)	25 (67)	9 (20)
Origin city
New Haven	18 (67)	8 (22)	34 (76)	N/A	N/A	N/A
Chicago	0 (0)	25 (68)	6 (13)
Mexico City	9 (33)	4 (10)	5 (11)

Definition of abbreviations: CTRL = control; FHP = fibrotic hypersensitivity pneumonitis; GAP = gender, age, and physiology index; IPF = idiopathic pulmonary fibrosis; N/A = not applicable.

Data are given as n (%) unless otherwise noted.

Table 2.

Summary of Demographic and Clinical Characteristics for BAL Samples from Mexico, with P Values for Comparisons of Variables across Specified Groups

Patient Variable	NFHP (n = 4)	FHP (n = 5)	IPF (n = 10)	P Value
				FHP vs. NFHP	FHP vs. IPF	IPF vs. NFHP
Age, mean ± SD	47.75 ± 8.85	53.83 ± 18.70	64.10 ± 8.86	0.53	0.24	8.9 × 10−3
Sex
Male	1 (25)	1 (20)	8 (80)	1	0.08	0.09
Female	3 (75)	4 (80)	2 (20)
Race
White	4 (100)	5 (100)	10 (100)	1	1	1
Ethnicity
Hispanic	4 (100)	5 (100)	10 (100)	1	1	1
Smoking
Never	2 (50)	3 (60)	5 (50)	1	1	1
Ever	2 (50)	2 (40)	5 (50)
% FVC, mean ± SD	58.67 ± 14.57	48.67 ± 16.44	45.00 ± 7.00	0.51	0.16	0.55
% DlCO, mean ± SD	45.00 ± 7.00	43.80 ± 16.45	52.57 ± 28.29	0.66	0.82	0.67
GAP score
0	0 (0)	0 (0)	0 (0)	0.44	0.36	0.37
1	0 (0)	0 (0)	0 (0)
2	1 (25)	1 (20)	0 (0)
3	2 (50)	2 (40)	3 (30)
4	0 (0)	1 (20)	1 (10)
5	0 (0)	0 (0)	2 (20)
6	0 (0)	0 (0)	1 (10)
No data	1 (25)	1 (20)	3 (30)
Origin city
Mexico City	4 (100)	5 (100)	10 (100)	N/A	N/A	N/A

Definition of abbreviations: CTRL = control; FHP = fibrotic hypersensitivity pneumonitis; GAP = gender, age, and physiology index; IPF = idiopathic pulmonary fibrosis; N/A = not applicable; NFHP = nonfibrotic hypersensitivity pneumonitis.

Data are given as n (%) unless otherwise noted.

PBMC and BAL Sample Processing

To facilitate long-term quality and transfer to New Haven for downstream processing, PBMC and BAL samples were cryopreserved in 10% DMSO in heat-inactivated fetal bovine serum and stored in liquid nitrogen. In New Haven, single-cell libraries were constructed from these samples (10X Genomics [CG00039; CG000331]). Chicago cohort samples were multiplexed to reduce batch effects and demultiplexed using their whole exome (26).

Sequencing, Quality Control, and Annotation

Single-cell complementary DNA libraries were sequenced to a depth of 150 million reads per sample. Standard quality control and filtering were conducted in Seurat. Cells were annotated using manual and automated methods (27, 28).

Differential Cell Type Composition

To evaluate cell shifts across disease groups, cell type proportions between FHP and control, FHP and IPF, and FHP and NFHP (BAL) were compared using the Wilcoxon rank-sum test, using Benjamini-Hochberg multiple hypothesis testing correction (false discovery rate).

Differential Gene Expression and Pathway Analysis

FHP-enriched cell clusters across different experimental groups were identified using the chi-square test. Differential gene expression analysis was conducted on these cell clusters using Seurat’s FindMarkers. EnrichR was used to ascertain differentially enriched pathways in these FHP-predominant cells (29).

Transcription Factor Activity Imputation

To study TF (transcription factor) roles in different cell types, TF activity scores were imputed from target genes’ expression (30, 31).

Trajectory Analyses

To relate peripheral blood findings to the lung, PBMC and BAL myeloid cells were integrated. Pseudotime and RNA velocity analyses were used to infer cellular trajectories and identify differentially expressed genes over time (32–35).

Data Availability

Data are deposited in GEO (GSE271789). A data portal is available (http://ildimmunecellatlas.com) based on the IPF cell atlas (19, 20, 35). The online supplement contains additional details.

Results

Single-Cell Transcriptomics Reveal Qualitative Shifts in Immune Cell Populations

PBMC samples from 39 patients with FHP, 53 patients with IPF, and 36 healthy control subjects from both the United States (Chicago, IL and New Haven, CT) and Mexico (Mexico City) were processed for 5′ scRNAseq (Figure 1A and see Table E1 in the online supplement). These samples were well matched for age, ethnicity, race, and smoking history across patients with FHP, patients with IPF, and control subjects. Pulmonary function tests and gender, age, and physiology index scores differed between control subjects and patients with FHP or IPF but not significantly between patients with FHP and patients with IPF. After quality control, 501,534 cells from 110 samples (37 FHP, 45 IPF, and 27 control) remained (Table 1), including 215,495 from FHP, 176,236 from IPF, and 109,804 from control samples. Major PBMC cell types were identified (Figures 1B and E1). Analysis for U.S. samples was combined, as they shared similar overall cell proportional shifts (adjusted-P = 0.237) and top variable genes (adjusted-P < 2.2 × 10⁻¹⁶) (Figure E2). Nineteen BAL samples from distinct Mexican patients (5 FHP, 10 IPF, and 4 NFHP) were well matched for demographic and clinical variables (Table 2). After quality control, 88,336 cells remained: 25,857 from FHP samples, 47,781 from IPF samples, and 14,698 from NFHP samples (Figure 1D). Qualitative cell distribution shifts by disease are observed in PBMCs and BAL (Figures 1C and 1E).

Figure 1.

Data overview. (A) Geographic overview of the study samples, with Uniform Manifold Approximation and Projection (UMAP) representations of cells from patients from each city (New Haven, CT; Chicago, IL; Mexico City, Mexico). UMAP representation of 501,534 PBMCs, labeled by (B) all major cell types and (C) conditions, and of 88,336 BAL cells, also labeled by (D) all major cell types and (E) conditions. BAL = bronchoalveolar lavage cells; CTRL = control; FHP = fibrotic hypersensitivity pneumonitis; IPF = idiopathic pulmonary fibrosis; mDC = myeloid dendritic cell; NFHP = nonfibrotic hypersensitivity pneumonitis; NK = natural killer; pDC = plasmacytoid dendritic cell; PBMCs = peripheral blood mononuclear cells.

FHP Harbors Quantitative Immune Cell Shifts Compared with Healthy Controls

scRNAseq can quantify cell type proportion shifts between FHP and healthy controls (Figures 2A and 2B). Specifically, FHP has decreased proportions of memory B cells (FHP: 1.6%; control: 3.0%; adjusted-P = 2.2 × 10⁻⁴), CD4⁺ regulatory T cells (Tregs) (FHP: 0.86%; control: 1.6%; adjusted-P = 4.5 × 10⁻³), mucosal-associated invariant T (MAIT) cells (FHP: 0.42%; control: 1.6%; adjusted-P = 4.9 × 10⁻⁴), and CD56^hi natural killer (NK) cells (FHP: 0.70%; control: 1.4%; adjusted-P = 1.2 × 10⁻⁴). Conversely, FHP has increased classical monocytes (FHP: 25%; control: 15%; adjusted-P = 2.5 × 10⁻³), platelets (FHP: 1.5%; control: 1.4%; adjusted-P = 6.4 × 10⁻³), and overall monocytes (FHP: 31%; control: 21%; adjusted-P = 0.011) (Figures 2C and E3).

Figure 2.

Cell type compositional comparisons of fibrotic hypersensitivity pneumonitis (FHP) versus control samples. (A) Uniform Manifold Approximation and Projection (UMAP) of PBMC samples from only patients with FHP and control subjects. (B) Cell type composition by FHP or healthy controls. (C) Percentage of cell types for each sample, grouped by disease subtype. Only significant cell type comparisons are shown. *P < 0.05, **P < 0.01, and ***P < 0.001. CTRL = control; PBMCs = peripheral blood mononuclear cells; CTL = cytotoxic lymphocyte; gdT = gamma delta T; Inter. = intermediate; ILC = innate lymphoid cell; mDC = myeloid dendritic cell; MPC = monocyte-platelet complex; MAIT = mucosal-associated invariant T cell; NK = natural killer; pDC = plasmacytoid dendritic cell; TEM = T effector memory; TReg = regulatory T cell.

Although not a proxy for healthy controls, NFHP BAL samples were compared with FHP samples to investigate fibrosis-related cell type compositional changes. In BAL cells, FHP demonstrates comparatively increased total macrophages (FHP: 59%; NFHP: 20%; adjusted-P = 0.032), proliferating macrophages (FHP: 5.5%; NFHP: 1.1%; adjusted-P = 0.032), and decreased Tregs (FHP: 1.5%; NFHP: 3.6%; adjusted-P = 0.032; Figure E4).

FHP Is Enriched in S100A^hi and CCL3/CCL4^hi Classical Monocytes versus Control Subjects

Classical monocytes are increased in FHP, but so far, they have not been analyzed at the single-cell resolution. Thus, 150,817 monocytes from FHP and control samples were subsetted and reclustered (Figures 3A–3C). Two FHP-enriched cell subpopulations were identified in FHP (adjusted-P < 2.2 × 10⁻¹⁶; Figures 3D and 3E): S100 calcium binding protein A (S100A^hi) (Figure 3F) and chemokine ligand 3/4 (CCL3^hi/CCL4^hi) (Figures 3G and E5 and Tables E2 and E3). These subtypes are also enriched in IPF (adjusted-P < 2.2 × 10⁻¹⁶; Figure E6). Pathway analysis revealed that S100A^hi monocytes are highly enriched in TGFβ regulation of ECM (representative genes: ITGAM, VCAN, NAIP, CCR2, CKLF, GM2A, ALOX5AP; adjusted-P = 0.045) and neutrophil degranulation and activation (S100A8, S100A9, CR1, FPR1, FOLR3, MNDA, SELL, CLEC4D, CXCR2, CTSD, ELANE, RETN, HEPB2, MCEMP1, ITGAM; adjusted-P = 2.5 × 10⁻¹⁸). CCL3^hi/CCL4^hi monocytes are enriched in lung fibrosis (CSF3, EDN1, IL6, PLAU, CXCL8, CCL3, CCL4, CXCL2, TNF, MMP9; adjusted-P = 3.6 × 10⁻⁷), TNFα-NFκB signaling (DUPS2, IER3, CXCL2, CD83, CXCL3, G0S2, TNFAIP3, PPP1R15A, TNF, IL1B, CCL4, NFKBIA, ICAM1, PDE4B; adjusted-P = 2.8 × 10⁻⁸¹), and IL-10 signaling (IL10, TNF, CSF3, ILR1RN, CXCL8, CCL3L1, CCL20, CD80, IL1R2, CXCL1, PTGS2, CXCL2, ICAM1, CXCL10, IL1A, IL6, IL1B, CCL3, CCL4; adjusted-P = 1.2 × 10⁻¹⁷). These results indicate the convergence of TGFβ and NFκB in fibrosis-associated monocytes and their potential role in regulating neutrophil activity in FHP.

Figure 3.

S100A^hi and CCL3^hi/CCL4^hi monocyte subpopulations enriched in fibrotic hypersensitivity pneumonitis (FHP) versus control. (A) Monocytes (circled) were subset from the overall data and reclustered. A Uniform Manifold Approximation and Projection (UMAP) representation of these reclustered cells is shown by (B) monocyte subtypes and (C) condition. Density graphs of monocytes for (D) controls and (E) FHP. (F) Gene expression of S100A^hi classical monocytes: S100A8, S100A9, S100A12, and CTSD. (G) Gene expression of CCL3^hi/CCL4^hi classical monocytes: CCL3, CCL4, CXCL3, NFKBIA, ICAM1, and IL1B. (H) UMAP of monocytes clustered by DoRothEA-VIPER transcription factor (TF) activity scores. (I) Breakdown of the DoRothEA-based clusters by percentage represented in control subjects and FHP. Significantly different clusters in FHP are marked in orange. (J) Heatmap demonstrating TFs with top variable imputed activities across monocyte DoRothEA clusters, with hierarchically clustered rows and columns. CTRL = control.

Increased Proinflammatory TF Activities in Classical Monocytes in FHP

To explore gene regulation in FHP, we imputed TF activities in classical monocytes and reclustered them based on their TF profiles (Figure 3H). Six clusters (0, 3, 4, 5, 7, 9) are expanded in FHP versus control samples (adjusted-P < 2.2 × 10⁻¹⁶; Figures 3I and 3J), with most (0, 4, 5, 9) also enriched in IPF (P < 2.2 × 10⁻¹⁶; Figure E6), suggesting shared transcriptional regulatory mechanisms. The clusters’ key regulators include NFKB1 (adjusted-P = 0.0042), SPI1 (adjusted-P = 0.05), and CEBPB1 (adjusted-P = 0.027) (Figure 3K). ATF3 and BHLHE40 are among the most significantly enriched TFs in FHP compared with control subjects across all clusters (Figure E7). Together, these results suggest a coordinated network of transcriptional regulation driving monocyte inflammation and differentiation in FHP.

S100A^hi and CCL3/CCL4^hi Classical Monocytes Are Enriched in Fibrosis and Develop into SPP1^hi Lung Macrophages

To understand how blood S100A^hi and CCL3^hi/CCL4^hi classical monocytes relate to lung macrophages in FHP, we analyzed PBMC and BAL cells, combining 122,232 PBMC and 59,944 BAL cells (Figures 4A–4C). S100A^hi and CCL3^hi/CCL4^hi classical monocyte subtypes and previously described secreted phosphoprotein 1 (SPP1^hi) profibrotic macrophages (SPP1, MERTK, LGMN, RNASE1, SEPP1, FOLR2; adjusted-P = 1.1 × 10⁻²⁸; Figures 4D and E8) were identified (20, 37–42). Monocle3 trajectory analysis, using gene expression profiles, demonstrated that S100A^hi peripheral monocytes differentiate into SPP1^hi macrophages (Figure 4E). Early markers like CCR2, SPI1, CD14, and S100A8 decrease over pseudotime, whereas CCL18, CD68, MERTK, PLA2G7, CCL3, and ICAM1 increase continuously. The transition to macrophages also includes rises and falls in CCR5, CEBPB, CCL2, and CSF1R expression, then MMP9, CSF1, and SPP1 (Figure 4F). RNA velocity analysis infers cellular fates from spliced to unspliced transcripts, not gene expression profiles. The transcriptional activity of monocytes (65–67% unspliced reads) exceeds that of macrophages (57–61%), suggesting more active monocyte cell turnover (Figure E9) (34). RNA velocity trajectories demonstrated that FHP monocytes differentiate into SPP1^hi macrophages, not alveolar macrophages (Figure 4G). CellRank identified five terminal states, mainly classical monocytes or alveolar macrophages, with fibrosis-associated monocytes in FHP overwhelmingly converging to SPP1^hi macrophages (Figure E10). These findings suggest that FHP monocytes can transition into profibrotic macrophages and should be further evaluated as potential biomarkers and easily accessible therapeutic targets in blood.

Figure 4.

Fibrosis-associated monocytes transition into profibrotic SPP1^hi macrophages. (A) Peripheral blood mononuclear cell (PBMC) monocytes and (B) BAL macrophages were integrated from the original data. (C) Uniform Manifold Approximation and Projection (UMAP) representation of integrated myeloid cells by cell type. (D) Localization of CCL3^hi/CCL4^hi monocytes (CCL3, CCL4), S100A^hi monocytes (S100A8, S100A9, TLR4), and profibrotic SPP1^hi macrophages (SPP1, MERTK, LGMN, PLA2G7). (E) Pseudotime trajectories across all integrated PBMC and BAL myeloid cells. (F) Gene expression changes over pseudotime between S100A^hi classical monocytes to profibrotic SPP1^hi macrophages, colored by pseudotime. (G) RNA velocity vector fields projected onto the UMAP embedding, demonstrating velocity vectors pointing from S100A^hi monocytes to SPP1^hi macrophages. CTRL = control.

FHP Exhibits Increased BAL T Lymphocytes versus IPF

The preceding sections compared patients with FHP to control subjects and patients with NFHP. Because FHP and IPF can present similarly clinically, comparing them can elucidate FHP-specific immune perturbations. We assessed cell proportional shifts in FHP and IPF, finding no significant differences in PBMCs (Figures 5A, 5B, and E11). However, for FHP, BAL analysis indicated fewer alveolar macrophages (FHP: 37%; IPF: 68%; adjusted-P = 0.0027) and total macrophages (FHP: 60%; IPF: 88%; adjusted-P = 0.0047), but increased CD4⁺ T cells (FHP: 23%; IPF: 5.6%; adjusted-P = 0.0027) and total T cells (FHP: 37%; IPF: 9.4%; adjusted-P = 0.0047; Figure E4), in agreement with previous reports (6, 43).

Figure 5.

Cell type compositional comparisons of fibrotic hypersensitivity pneumonitis (FHP) and idiopathic pulmonary fibrosis (IPF). (A) Uniform Manifold Approximation and Projection (UMAP) of peripheral blood mononuclear cell samples from only FHP and IPF. (B) Cell type composition by FHP or IPF. CTL = cytotoxic lymphocyte; gdT = gamma delta T; Inter. = intermediate; ILC = innate lymphoid cell; mDC = myeloid dendritic cell; MPC = monocyte-platelet complex; MAIT = mucosal-associated invariant T cell; NK = natural killer; pDC = plasmacytoid dendritic cell; TEM = T effector memory; TReg = regulatory T cell.

Cytotoxic GZM^hi T Lymphocytes Are Increased in FHP Compared with Controls and IPF

Given the lymphocytosis seen in FHP BAL and prior work implicating T cells in FHP, lymphocytes were reprocessed, yielding 223,782 cells and 11 annotated subtypes (Figures 6A–6C). Overall, T lymphocytes highly expressing granzymes (GZM^hi) levels—serine proteases indicating their cytotoxicity (44)—are significantly more prevalent in FHP than in controls or IPF. FHP samples from the United States and Mexico are enriched in GZM^hi CD4⁺ and CD8⁺ T cells (adjusted-P < 1.1 × 10⁻⁶), with both types highly expressing GZMA and GZMB (Figures 6D–6F and Tables E4 and E5) and CD8⁺ T cells exhibiting a GZMH^hi phenotype previously described in autoimmunity (45) (Figure E12). Cytotoxic CD4⁺ T cells enriched in patients with FHP (adjusted-P < 2.2 × 10⁻¹⁶) expressed genes involved in IL-2 signaling pathway (adjusted-P = 4.2 × 10⁻²³), allograft rejection (TGBF1, ISRGN, GZMA, ITGAL, IL2RG, CCL4/5; adjusted-P = 1.0 × 10⁻⁸), and IFN-γ response (TGBF1, IFITM2/3, STAT4, CCL5, IRF1, HLA-DRB1, LY6E; adjusted-P = 5.9 × 10⁻¹¹) (Figures 6G and E13 and Table E4).

Figure 6.

Fibrotic hypersensitivity pneumonitis (FHP) harbors distinct CD8⁺ T cell phenotype compared with idiopathic pulmonary fibrosis (IPF) and controls. (A) T cells (circled) were subset from the overall data and reclustered. A Uniform Manifold Approximation and Projection (UMAP) representation of these cells is shown by (B) T cell subtypes and (C) condition. Density graphs of T cells for (D) controls, (E) FHP, and (F) IPF. (G) Gene expression of cytotoxic CD4 T cells enriched in FHP: HLA-DRB1, KLRB1, TGFB1, S100A10, PLEK, and VAMP. (H) Gene expression of GZMH^hi CD8⁺ T cells enriched in FHP: FGFBP2, GZMH, CCL5, CTSW, KLRD1, and IL1B. (I) UMAP of CD8⁺ T cells clustered by DoRothEA-VIPER TF (transcription factor) activity scores. (J) Breakdown of the DoRothEA-based clusters by control, FHP, and IPF. Significantly enriched clusters in FHP are marked in orange. (K) Heatmap demonstrating TFs with top variable imputed activities across CD8⁺ T cell DoRothEA clusters, with hierarchically clustered rows and columns. CTRL = control.

Similarly, FHP demonstrates significant enrichment of GZM^hi CD4⁺ and CD8⁺ T cells (adjusted-P < 2.2 × 10⁻¹⁶) than IPF. A GZMH^hi subpopulation (cluster 2) is increased significantly in FHP versus IPF and is enriched in IL-2 pathway (IFITM3, PRF1, IKZF3, LY6E, NKG7, CTSD, STAT4, KLF6; adjusted-P = 1.4 × 10⁻²⁸), IL-12 signaling (LCK, TBX21, STAT4, B2M, HLA-DRB1, GZMA/B; adjusted-P = 9.8 × 10⁻³), and TNFα-NFκB signaling (GADD45B, KLF6, IRF1, CCL4/5, IER2, BHLHE40, ZFP36; adjusted-P = 0.040) (Figures 6H and E14; Table E5).

Finally, CD8⁺ T cell populations displayed geographical differences. Beyond the globally enriched GZMH^hi CD8⁺ T cells, a GZMK^hi CD8⁺ T cell subpopulation was enriched only in patients with FHP from Mexico (adjusted-P < 4.3 × 10⁻²⁸⁸; Figure E15). GZMK^hi cells are enriched in TNFα-NFκB signaling (CCL4/5, NFKBIA, CD68, JUN, GADD45B, ID2 DUSP1/2, TNFAIP3, ZFP36; adjusted-P = 3.4 × 10⁻¹⁴), IL-2 signaling pathway (ZFP36, JUN, DUSP1, CCL4/5, NKG7, CXCR4, GZMA/K, DUSP6, CTSW, CD69; adjusted-P = 7.7 × 10⁻⁹), IL-1 regulation of ECM (ZFP36, JUN, TNFAIP3, NFKBIA, CCL4/5, DUSP6; adjusted-P = 3.2 × 10⁻⁸), and IL-12 signaling (GZMA, GADD45B, CCL4L2, CD8A/B, HLA-DRB1; adjusted-P = 3.2 × 10⁻⁸). Activated BAL CD8⁺ T lymphocytes enriched in patients with FHP had high HLA type II expression (HLA-DRB1, HLA-DPB1, HLA-DPA1, HLA-DQA1, HLA-DQA2, HLA-DQB1; adjusted-P < 2.2 × 10⁻¹⁶; Figure E16). The same populations emerge significantly in a female-patient subanalysis, ruling out potential sex-related confounding (Figure E17).

FHP CD8⁺ T Cells Demonstrate Increased TF Activities Involved in TGFβ Signaling

Reclustering PBMC CD8⁺ T lymphocytes by TF activity scores identified 12 clusters (Figure 6I). FHP cluster compositions differ significantly versus control subjects (adjusted-P = 3.6 × 10⁻⁸) and IPF (adjusted-P = 5.1 × 10⁻⁹). Five clusters (0, 3, 4, 8, 10) are increased versus IPF and controls (adjusted-P < 1 × 10⁻⁹), showing enrichment of SMAD2/3/4 (FOXA1, BLHLE40, ATF3, GATA4, MXI1; adjusted-P = 3.6 × 10⁻⁵) TF activities (Figures 6J and 6K). TFs enriched in these clusters also demonstrate TGFβ (adjusted-P = 0.0015) and TNFα-NFκB signaling (adjusted-P = 0.0020). The TFs with significantly increased activities in FHP include LHX2, GLI2, and SOX11 (Figure E18). These results again highlight TGFβ and NFκB signaling convergences in FHP and altered lymphocyte differentiation.

Discussion

Our study presents the most extensive single-cell analysis yet of peripheral blood and BAL samples from patients with FHP. Patients with FHP exhibit enriched GZM^hi cytotoxic CD4⁺ and CD8⁺ T cells in PBMCs versus control subjects and patients with IPF, with geographically restricted GZMK^hi CD8⁺ T cells in Mexican patients with FHP. Within myeloid cells, we identify novel S100A^hi and CCL3^hi/CCL4^hi classical monocyte subtypes in FHP and IPF. We link these peripheral classical monocytes to previously described profibrotic SPP1^hi tissue macrophages (20, 37–42). We uncover previously undescribed adaptive and innate immune cell populations in FHP and provide a data-sharing, mining, and dissemination portal (https://www.ildimmunecellatlas.com) to facilitate community use.

One of our key findings is identifying FHP-specific cytotoxic T cells expressing high granzyme levels (GZMA, GZMB, GZMH, GZMK), including cytotoxic CD4⁺, GZMH^hi CD8⁺, and GZMK^hi CD8⁺ T cells. IPF and FHP have long been suggested to differ in immune etiology, with lymphocytes predominantly driving FHP (1, 5, 6, 11–14, 18). Yet, past research has focused on Th2 and Th17 cells, with the role of cytotoxic T cells underexplored. We show significant enrichment of cytotoxic CD8⁺ and CD4⁺ T cells in FHP PBMCs. T cell activation–related pathways—IL-2 and IFN-γ in CD4⁺ T cells; IL-2, IL-12, and TNFα and NFκB in GZMH^hi CD8⁺; and IL-1, IL-2, IL-12, and TNFα/NFκB in GZMK^hi CD8⁺ T cells—and active TFs in profibrotic TGFβ pathways suggest these cells’ potential to sustain inflammation and propagate fibrosis in FHP (43, 44). These cells also highly express CCL3, CCL5, CXCR3, GZMA, TNFa, and IL2R (Tables E4 and E5), indicating their likely contribution to disease-specific genes discovered by Selman and colleagues (17). These activation signatures also emerged in regions of mild lung fibrosis (11). Of particular interest are the GZMH^hi and GZMK^hi CD8⁺ T cells associated with autoimmunity in FHP. A repertoire restricted GZMH^hi CD8⁺ T cell population was originally identified in patients with systemic lupus erythematosus (SLE) (45). We find GZMH^hi CD8⁺ T cells in FHP without the high exhaustion signature seen in SLE, suggesting FHP-enriched cells resemble, rather than mirror, an autoimmune-associated adaptive immune response and should be further studied in detail. GZMK^hi cells have also been tied with autoimmune and inflammatory conditions, including rheumatoid arthritis, primary Sjogren’s syndrome, and coronavirus disease (COVID-19) (46, 47). Only our Mexican cohort displayed them, suggesting a distinct genetic or environmental influence on adaptive immune responses in FHP. This result illustrates our study’s high resolution and underscores the need for future studies to include populations from multiple geographical or genetic backgrounds.

Peripheral blood monocytes have been associated with disease severity and outcome in IPF (37, 48). Recently, an IPF PBMC single-cell study confirmed that the validated outcome-predictive 52-gene signature in progressive IPF reflected genes mostly expressed in monocytes (36, 49, 50). Although bulk RNA studies of the lungs have identified some myeloid-related genes, specific blood monocyte subpopulations were not previously characterized at the single-cell resolution in FHP (11–18). We find that monocytes in FHP exhibit cellular phenotypes akin to IPF but not control samples, highlighting a conserved innate immune response across fibrotic diseases. Our single-cell analysis uncovered two distinct classical monocyte subpopulations—S100A^hi and CCL3^hi/CCL4^hi—which are markedly enriched in FHP. S100A^hi monocytes exhibit gene enrichments linked with TGFβ-mediated ECM regulation and neutrophil degranulation, pathways related to fibrosis and neutrophilic inflammation. Damage-associated molecular patterns S100A8/9/12 activate the TLR4 and MyD88 pathway, enhancing TGFβ signaling and contributing to poor outcomes in diseases like IPF and severe COVID-19 outcomes (12, 51–56). Although neutrophils are not detected in PBMCs, S100A^hi monocytes may recruit and activate neutrophils through high FPR-1 and IL-8 expression, leading to neutrophil degranulation and subsequent lung damage (57, 58). Furthermore, neutrophilia is associated with lung function decline and early mortality in patients with IPF (59). The CCL3^hi/CCL4^hi monocytes may be a potential source of these chemokines found in the lung by Selman and colleagues (17). CCL3^hi/CCL4^hi monocytes are enriched in TNFα and NFκB and IL-10 signaling pathways, which are instrumental in driving Th1, Th2, and Th17 responses, promoting fibrosis, and recruiting neutrophils (60–64). Samples from patients with IPF also demonstrate increased CCL3, CCL4, TNFa, IL-10, CXCL8, CXCL2, and IL-18 further underscoring these monocytes as shared cell types in fibrotic disease (61, 65–78). The discovery of specific monocyte subsets—S100A^hi and CCL3^hi/CCL4^hi monocytes in both FHP and IPF—may identify distinct innate cell populations that both represent and contribute to pulmonary fibrosis, thus representing targets for developing novel diagnostics and therapeutics.

Performing scRNAseq of PBMCs and BAL cells from patients with FHP has provided us valuable insights into the origins of SPP1^hi macrophages, which are associated with fibrosis. Recent single-cell studies of human pulmonary fibrosis lungs have identified these macrophages (20, 36–42), yet it remains unclear whether they originate from tissue-resident cells or are derived from monocytes. Our trajectory analyses linked these peripheral monocytes to the lung, offering an in silico model of how fibrosis-associated monocytes differentiate into SPP1^hi lung macrophages. Prior single-cell lung studies identified SPP1^hi macrophages in fibrotic lung disease that are enriched in CHI3L1, MARCKS, IL1RN, PLA2G7, MERTK, LGMN, CCL18, and MMP9 (20, 37–42). The debate regarding the origin of these profibrotic macrophages continues, with some evidence suggesting they are derived from self-renewing interstitial macrophages (39, 79), whereas other studies, including ours, support a monocyte-derived hypothesis. Notably, Joshi and colleagues have demonstrated in a mouse model that selective depletion of monocyte-derived macrophages expressing profibrotic markers can attenuate asbestos-induced pulmonary fibrosis (80). Our findings link S100A^hi monocytes to SPP1^hi profibrotic macrophages, detailing a dynamic gene expression transition involving genes like S100A9, CTSD, MMP9, SPP1, CD68, CCL18, MERTK, and PLA2G7 that are not expressed contemporaneously. These data may inspire research into targeting monocytes expressing “early peak” genes to prevent their differentiation into fibrogenic macrophages, offering a potential novel therapeutic strategy.

Beyond gene expression, scRNAseq reveals cell-type composition changes that bulk methods cannot and allowed us to uncover unique cellular shifts in FHP compared with both controls and IPF. FHP is characterized by increased peripheral total monocytes, classical monocytes, and platelets, consistent with patterns seen in other fibrotic and autoimmune conditions, such as SLE and rheumatoid arthritis (48, 81–86). Our group has also discovered increased monocytes in more severe IPF (35). In addition, FHP demonstrated significant decreases in peripheral memory B cells, MAIT cells, Tregs, and CD56^hi NK cells, aligning with decreased memory B cells seen in rheumatoid arthritis–interstitial lung disease (87, 88) and fewer circulating MAIT cells in severe COVID-19 (89). Reduced CD56^hi NK cells are consistent with systemic inflammation–associated diseases (90–93), whereas decreased Tregs fall in line with global Treg dysfunction seen in systemic inflammation (94). In BAL, increased macrophages and decreased Tregs suggest a proinflammatory state driven by myeloid cells, with insufficient regulatory cells to attenuate the inflammation. Together, these cell population shifts in FHP mirror those that occur in both autoimmune and fibrotic diseases, demonstrating the contribution of both processes (2, 95). Although it is not immediately clear how this information will impact our understanding of FHP, the availability of these detailed profiles and the gene expression patterns associated with them should ignite interest and enthusiasm for better understanding and targeting immune mechanisms in FHP.

Our study has several limitations. First, our data’s heterogeneity and limited cohort size may affect the generalizability of our findings. This heterogeneity could arise from technical variability or demographic and regional differences among the cohorts. To mitigate technical variability, we used a randomized block design and standardized protocols across all samples that were processed in the same laboratory. In addition, to control for potential sex-driven signals, we conducted subanalyses restricted to female samples, which confirmed our findings. We also controlled for regional differences by analyzing patients from New Haven and Chicago together and patients from Mexico separately. This approach enabled us to identify both common and distinct signatures. Overall, the significance and reproducibility of our results suggest that our strategies addressed these limitations. Naturally, other limitations, such as replication in large, prospective, and independent cohorts and assessment of treatment effects, are beyond the scope of this first large-scale study of single-cell immune profiles in FHP, but we hope our results generate enthusiasm for performing such studies in the future.

In conclusion, our study represents the first extensive single-cell analysis of peripheral blood and BAL from patients with FHP. We identified novel monocyte populations common to IPF and FHP, as well as distinct cytotoxic T cell populations enriched in FHP but not IPF. These findings underscore the importance of conducting prospective comparative studies on immune aberrations in fibrotic lung diseases of different etiologies and their response to state-of-the-art therapies. Furthermore, the consistency of certain findings across U.S. and Mexican cohorts is reassuring, yet differences highlight the need for further research involving diverse populations from diverse geographic, ethnic, and exposure backgrounds. Our results’ broad availability will likely generate significant interest in FHP, encouraging studies that could facilitate the development of targeted diagnostics, biomarkers, and immune-modulatory therapies tailored to specific patients.