Abstract

   There is a significant knowledge gap in how T cells promote emphysema
   in smokers with chronic obstructive pulmonary disease (COPD).
   Single-cell RNA sequencing (scRNA seq) analysis of human samples and
   relevant clinical data can provide new mechanistic insights into
   disease pathogenesis. We generated a human lung scRNA seq dataset with
   extensive disease characteristic annotation and analyzed a second
   independent scRNA seq dataset to examine the pathophysiological role of
   T cells in emphysema. Comparisons of pulmonary immune landscapes in
   emphysematous (E)-COPD, non-emphysematous (NE)-COPD, and control showed
   positive enrichment of T cells in E-COPD. Pathway analyses identified
   upregulated inflammatory states in CD4 T cells as a distinguishing
   feature of E-COPD.  Compared to controls, glucocorticoid receptor NR3C1
   CD4 T cells were enriched in NE-COPD but were reduced in E-COPD.
   Interactions between macrophages and NR3C1^+ CD4 T cell subsets via
   CXCL signaling were strongly predicted in E-COPD but were absent in
   NE-COPD and control. The relative abundance of CD4 CXCR6^high effector
   memory T cells positively correlated with preserved lung function in
   E-COPD but not in NE-COPD. These findings suggest that NR3C1^+ and
   CXCR6^high effector memory subsets of CD4 T cells distinguish the
   immune-pathophysiological features of emphysema in human lungs.
   Targeting relevant T cell subsets in emphysema might provide new
   therapeutic opportunities.

   graphic file with name 42003_2025_8698_Figa_HTML.jpg

   Subject terms: Chronic inflammation, Predictive markers
     __________________________________________________________________

   scRNAseq analysis of human lung T cells complemented with clinical data
   provides new mechanistic insights into the pathophysiological role of
   adaptive immune cells in emphysema.

Introduction

   Long-term cigarette smoking profoundly changes the transcriptomic
   landscapes in the lungs^[42]1–[43]3 and is directly linked to the
   development of chronic obstructive lung diseases (COPD)^[44]4,[45]5.
   Studies in human smokers and experimental animal models of
   smoke-induced emphysema suggest that both the adaptive^[46]6–[47]10 and
   innate^[48]11,[49]12 immune systems contribute to the tissue
   destruction and severity of lung disease in COPD. Many human studies,
   however, classify COPD based on the degree of airway flow limitation,
   but the extent of lung tissue destruction, emphysema, which is the most
   clinically consequential outcome in smokers^[50]13–[51]17, is not
   reported. Under-reporting and excluding distinct subphenotypes have
   culminated in a poor understanding of the immune-mediated
   pathophysiological changes in human emphysema.

   Genome-wide association studies (GWAS)^[52]18–[53]20,
   proteomic^[54]21,[55]22, and transcriptomic^[56]23,[57]24 studies
   recognize emphysema as an independent disease phenotype in COPD.
   Notably, radiographic detection of emphysema can be discordant with the
   physiological manifestation of airflow obstruction^[58]25–[59]27. These
   observations call for a better characterization of emphysema and COPD
   endotypes in preclinical and clinical studies.

   The molecular mechanisms underlying the diverse clinical presentations
   of COPD remain less clear. Inflammation is strongly associated with the
   development of COPD among tobacco smokers, and the emphysema variant of
   COPD is linked to the activation of the adaptive immune system^[60]28.
   Specifically, identification of oligoclonal autoreactive T cells in
   humans and experimental emphysema models^[61]7,[62]8,[63]29 suggests a
   pathogenic role for antigens that can activate and expand immune cells
   in the lungs. Although the contribution of adaptive immunity in the
   pathophysiology of emphysema is widely acknowledged, how different
   types and states of T cells and their interactions with innate immune
   cells in the lungs emerge and promote disease development in emphysema
   remain unclear.

   Despite significant efforts in identifying biomarkers that may predict
   disease outcomes in tobacco smokers, immune profiles in the peripheral
   blood have not always mirrored tissue immunity in the
   lungs^[64]18,[65]30. This knowledge gap suggests the need to examine
   lung tissue cellular profiles to decipher how immune cells promote lung
   tissue destruction. Single-cell (sc)RNA sequencing has been
   revolutionary in characterizing the cellular landscape at a
   high-throughput scale in single-cell resolution^[66]31, but large-scale
   studies that include exhaustively annotated clinical information are
   rare and have not focused on the role of T cells in human emphysema.

   In this study, we analyzed scRNA sequencing data using a total of 108
   human lung tissues. We used sixty-two human lung samples from
   well-characterized groups of smokers and non-smokers with extensive
   disease characteristic annotation and radiographic quantification of
   emphysema using percent low attenuation area to separate cases into
   three strata: emphysema predominant (E-COPD), non-emphysematous COPD
   (NE-COPD), and controls. We took advantage of the transcriptomic
   information of different immune cell types and cell states in the lungs
   to investigate the factors that may precipitate the emergence of
   pathogenic cell types across different COPD endotypes. A published
   scRNA sequencing dataset from 46 samples was used to validate the main
   results.

Results

Distinct human lung immune cell landscape in emphysema

   We acquired human lung tissue samples from 62 individuals and performed
   scRNA sequencing (Fig. [67]1a). Patients were first stratified to COPD
   (n = 37) and Control (n = 25) based on airflow obstruction
   (FEV[1]/FVC). COPD patients were further categorized based on computed
   tomography (CT)-based measurement of low attenuation area percentages
   (LAA%) below 950 Hounsfield units for each subject (Supplementary
   Fig. [68]1a–g, Table [69]1). We used the 5% LAA cutoff as the accepted
   threshold for the presence of emphysema^[70]32. Percent LAA separated
   the COPD cohort into non-emphysematous COPD (NE-COPD) (n = 21) and
   predominant emphysema (E-COPD) phenotypes (n = 16).

Fig. 1. Positive enrichment of T cells and negative enrichment of macrophages
in Emphysematous COPD (E-COPD).

   [71]Fig. 1
   [72]Open in a new tab

   a Study design for in-house discovery dataset (b–f), and (g) for
   validation dataset (h–j). UMAP embedding of the global cellular
   landscape for the in-house discovery dataset (b) and the published
   validation dataset (h). Collective cellular proportions across disease
   groups in the in-house discovery dataset (c) and published validation
   dataset (i). Pairwise comparison of cell type differential abundance
   using scCODA in in-house discovery dataset between E-COPD vs. Control
   (d), E-COPD vs. NE-COPD (e), and NE-COPD vs Control (f). Pairwise
   comparison of cell type differential abundance using scCODA in the
   published validation dataset between COPD and Control (j). The final
   parameter shown on the horizontal axis indicates degrees of enrichment.
   Positive values suggest positive enrichments, and negative values
   suggest negative enrichments. The false discovery rate (FDR) for scCODA
   differential abundance analyses was set at a threshold of 0.25. Artwork
   was generated from Bioicons ([73]https://bioicons.com/) and NIH NIAID
   BioArt Source ([74]https://bioart.niaid.nih.gov/). Modifications of the
   artwork were performed in Inkscape ([75]https://inkscape.org/).

Table 1.

   Patient characteristics, in-house dataset one
   Variable Control N = 25 No Emphysema COPD N = 21 Emphysema COPD
   N = 16 p-value^a
   Age, Median (IQR) 62 (52–72) 72 (67–75) 68 (66–68) 0.034
   Sex, n (%) 0.062
    F 14 (56) 9 (43) 3 (19)
    M 11 (44) 12 (57) 13 (81)
   Smoke, n (%) <0.001
    Current 5 (20) 8 (38) 0 (0)
    Former 9 (36) 12 (57) 16 (100)
    Never 11 (44) 1 (4.8) 0 (0)

   PackYear,

   Median (IQR)
   15 (0–25) 28 (16–54) 39 (20–52) 0.001
   GOLD, n (%) <0.001
    Control 25 (100) 0 (0) 0 (0)
    GOLD 0 0 (0) 7 (33) 0 (0)
    GOLD 1 0 (0) 6 (29) 2 (13)
    GOLD 2 0 (0) 8 (38) 2 (13)
    GOLD 4 0 (0) 0 (0) 12 (75)

   LAA %,

   Median (IQR)
   0 (0–1) 1 (0–2) 19 (16–30) <0.001

   FEV1%

   predicted,

   Median (IQR)
   100 (94–100) 77 (65–80) 23 (19–34) <0.001
   FEV1/FVC, Median (IQR) 80 (74–80) 67 (63–73) 27 (22–58) <0.001
   Steroids, n (%) 1 (4.0) 4 (19) 10 (63) <0.001
   [76]Open in a new tab

   IQR interquartile range.

   ^aKruskal–Wallis rank sum test; Pearson’s Chi-squared test; Fisher’s
   exact test.

   A total of 48,569 cells from the control, 45,692 cells from NE-COPD,
   and 34,172 cells from the E-COPD group passed quality control.
   Twenty-four cell types were identified (Fig. [77]1b and Supplementary
   Fig. [78]1h). Immune cells represented over 90% of recovered cells,
   with macrophages (~48%) and αβT lymphocytes (~33%) together comprising
   over 80% of total cells (Fig. [79]1c). Other top immune populations
   include monocytes (~8%), NK (~10%), and cDCs (~4%). Top recovered
   non-immune components include alveolar epithelial cells (~2%) and
   vascular and lymphatic endothelial cells (~2%).

   Because of the disparity of total cell counts and sample numbers in
   both COPD and control groups, we opted for the proportion-based
   Bayesian modeling approach, scCODA, for compositional analyses between
   different groups^[80]33. We compared cell-type enrichment distribution
   in E-COPD vs Control (Fig. [81]1d) and E-COPD vs NE-COPD (Fig. [82]1e).
   CD14 monocytes, B cells, and αβ CD4 T cells were positively enriched,
   whereas macrophages and epithelial cells showed negative enrichments in
   E-COPD compared to controls. In addition to the same cell populations,
   CD8 T cells were also enriched in E-COPD compared to NE-COPD, but this
   enrichment was not detected when compared to controls. In NE-COPD,
   macrophages showed positive enrichment when compared to controls
   (Fig. [83]1f). These findings indicate that the E-COPD phenotype has a
   distinctive immune profile from NE-COPD, suggesting that while adaptive
   immunity is highly associated with E-COPD, positive enrichment of
   macrophages is a characteristic feature in NE-COPD.

   Because active smoking can strongly affect systemic and lung tissue
   immunity, we next examined its effect on immune cells. In the Control
   group, 20% of cases were current, 36% former, while 44% were never
   smokers (Table [84]1). In the NE-COPD group, 38% of cases were current,
   57% former, while 5% were never smokers (Table [85]1). Because all
   patients in E-COPD were former smokers, we assessed the effect of
   smoking in the Control and NE-COPD groups. In the NE-COPD group, when
   comparing former smokers with never-smokers, macrophages were
   positively enriched, whereas CD8 and NK cells were negatively enriched
   (Supplementary Fig. [86]2a–c). In the same group, current smokers
   exhibited positive enrichment of pulmonary macrophages, but CD4 and CD8
   T cells were negatively enriched (Supplementary Fig. [87]2).
   Consistently, comparing current smokers with former smokers, CD4 T
   cells were negatively enriched (Supplementary Fig. [88]2e). In the
   Control group, when comparing former smokers with never smokers, CD4 T
   cells were positively enriched, but macrophages were negatively
   enriched (Supplementary Fig. [89]2f–h). Current smokers, compared to
   former smokers, showed positive enrichment of CD8 T cells and
   macrophages but negative enrichment of CD16 monocytes (Supplementary
   Fig. [90]2i). Comparison between current and never smokers in the
   Control group did not yield any statistically significant results that
   passed the false discovery rate threshold.

   We next used an independent and publicly available scRNA-seq
   dataset^[91]34,[92]35 that included end-stage COPD lung explant and
   rejected donor lung samples as controls to validate our findings
   (Fig. [93]1g, Supplementary [94]3a–c, Table [95]2). Because information
   regarding %LAA was not available for the second COPD cohort, we could
   not stratify radiographically confirmed emphysema and were only able to
   compare end-stage COPD to controls. We found similar cellular
   enrichment patterns in the validation dataset with positive enrichments
   of the αβ CD4 and CD8 T lymphocytes, B cells, and negative enrichment
   of macrophages in end-stage COPD when compared to controls
   (Fig. [96]1h–j; Supplementary Fig. [97]3d).

Table 2.

   Patient characteristics, validation dataset two
       Variable      Control N = 28 COPD N = 18 p-value^a
   Gender, n (%)                                0.64
    Female           12 (43)        9 (50)
    Male             16 (57)        9 (50)
   Age, Median (IQR) 47 (31–63)     62 (58–66)  0.002
   Race, n (%)                                  >0.99
    Asian            1 (3.6)        0 (0)
    Black            1 (3.6)        0 (0)
    Latino           1 (3.6)        0 (0)
    White            25 (89)        18 (100)
   Smoking, n (%)    6 (21)         17 (94)     <0.001
   [98]Open in a new tab

   IQR interquartile range.

   ^bPearson’s Chi-squared test; Wilcoxon rank sum test; Fisher’s exact
   test.

   Together, these two independent datasets provided evidence that the
   emphysema variant of COPD is characterized by the positive enrichment
   of αβ CD4 T cells and negative enrichment of pulmonary macrophages.

Unique transcriptomic signatures of αβ CD4 T cells in emphysema

   Because CD4 T cells were consistently positively enriched in E-COPD and
   end-stage COPD, we then set out to assess functional alternations in
   CD4 T cells using transcription factor activity analyses (Fig. [99]2a),
   gene set over representation analyses (ORA) (Fig. [100]2b), and gene
   set variation analysis (GSVA) (Fig. [101]2c–f). The transcription
   factor activity inference data showed increased STAT3, JUN, NFKB, and
   RELA activities in E-COPD compared to the NE-COPD and controls
   (Fig. [102]2a). STAT3 is a critical transcription factor downstream of
   IL6 signaling and can induce its expression^[103]36–[104]38. Consistent
   with these findings, we found CD4 T cells in E-COPD exhibit upregulated
   IL2-STAT5 and IL6-JAK-STAT3 signaling by both ORA and GSVA analyses and
   E-COPD showed increased IL-17 production pathway in GSVA analysis
   compared to controls (Fig. [105]2b, e–g). Responses to chemokines and
   cytokines were also upregulated at an individual patient level in
   E-COPD compared to control or NE-COPD cohorts (Fig. [106]2c, d). IL6
   induces expression of IL21 and IL23R in CD4 T cells, which are upstream
   of transcription factor, RORC, and IL17 expression^[107]39. STAT3 is
   also indispensable in the development of Th17 cells^[108]40,[109]41.
   Notably, IL6-STAT3 pathways and IL-17 upregulation in CD4 in E-COPD
   further corroborate the pathogenic role of IL17/Th17 in human
   emphysema^[110]42,[111]43. Furthermore, CD4 T cells in E-COPD also
   showed upregulation of inflammatory pathways, including
   interferon-gamma (IFN-γ), TNF, and IFN-α responses (Fig. [112]2b).

Fig. 2. Unique transcriptional signatures of CD4 T cells in Emphysematous
COPD (E-COPD).

   [113]Fig. 2
   [114]Open in a new tab

   a Transcription factor activity estimation with a univariate linear
   model of CD4 T cells using decoupleR in Controls, Nonemphysematous COPD
   (NE-COPD), and Emphysematous COPD (E-COPD) in the in-house discovery
   dataset. b Functional enrichment of biological terms by
   over-representation analyses (ORA) of CD4 T cells across Controls,
   NE-COPD, and E-COPD using Hallmark pathway gene set in the in-house
   discovery dataset. Gene set variation analyses (GSVA) of Leukocyte
   response to cytokine (Gene Ontology) (c), Leukocyte response to
   chemokine (Gene Ontology) (d), IL2 mediated signaling (KEGG) (e), and
   IL6-JAK-STAT3 Signaling (KEGG) (f) and IL17 production (g) across three
   disease groups in the in-house discovery dataset. Statistical
   significance between the 3 groups was determined using the
   Kruskal–Wallis’ test. Pairwise comparisons were performed using
   Wilcoxon rank-sum test with Holm’s correction. Significance code:
   p < 0.05(*), p < 0.01(**), p < 0.001 (***), ns: non-significant.

   Together, these results suggest that augmented inflammatory pathways
   are a unique feature of CD4 T cells in the lungs of E-COPD and that the
   positive enrichments of CD4 T cells might be a result of increased
   responsiveness to chemokine and cytokine-mediated recruitment signals.

Identification of distinct CD4 T cell subsets in human lungs

   We next examined the heterogeneities of CD4 T cell subsets in the lung.
   Broadly, we identified 5 distinct subsets of CD4 T cells
   (Fig. [115]3a–d). Effector memory CD4 T cells were identified by the
   expression of S100A4 and IL32^[116]44,[117]45. Naïve and central memory
   T cells were identified by the expression of CCR7, TCF7, CD62L (SELL),
   and LEF1, while naïve T cells were further distinguished by the absence
   of memory markers S100A4 and IL32^[118]45. Regulatory T (Treg) cells
   were identified by the concurrent expression of FOXP3, IL2RA, TIGIT,
   CTLA4, and IKZF2^[119]46–[120]49. A subset of CD4 T cells expressing
   glucocorticoid receptor NR3C1 showed distinct gene expression profiles
   (Fig. [121]3b, c) and transcription factor activity (Fig. [122]3d),
   which we termed the NR3C1^+ subset. To confirm the CD4 T cell subsets,
   we performed pseudotemporal analysis of the different populations using
   the Palantir modeling method^[123]50. As expected, naïve and central
   memory CD4 T cells exhibited the lowest levels of differentiation,
   whereas Tregs showed the highest level of differentiation
   (Fig. [124]3e–g). Notably, effector memory CD4 T cells exhibited an
   intermediate level of differentiation, whereas NR3C1-expressing CD4 T
   cells showed the second-highest levels of differentiation
   (Fig. [125]3e–g). We next confirmed the presence of distinct CD4 T cell
   subsets using an independent scRNA seq dataset (Supplementary
   Fig. [126]4a–c).

Fig. 3. Identified CD4 subsets in the human lungs.

   [127]Fig. 3
   [128]Open in a new tab

   a UMAP embedding of identified CD4 subsets in the in-house discovery
   dataset. b UMAP embedding of major subset marker expression: Central
   memory CD4 (TCF7, KLF2), Naïve CD4 (KLF12, BACH2), NR3C1 (NR3C1), Treg
   (FOXP3), Effector memory (S100A4, IL32). c Heatmap of gene signature
   expression in the identified 5 subsets of CD4 T cells in the in-house
   discovery dataset. d Transcription factor activity was estimated with
   the univariate linear model using decoupleR for each CD4 subset. e UMAP
   embedding of Palantir pseudotime. f Differentiation trajectory of CD4
   subsets estimated by cellrank using Palantir pseudotime. g Violin plots
   of palantir pseudotime values for identified CD4 subsets.

Reduced relative abundance of NR3C1^+ CD4 T subset in E-COPD

   We next examined the relative abundance of glucocorticoid-receptor
   NR3C1-expressing T cells in the study cohorts. We found that the
   relative abundance of the NR3C1^+ subset of CD4 was significantly
   reduced in E-COPD but was increased in NE-COPD relative to control
   (Fig. [129]4a). Notably, alterations in the relative abundance of the
   NR3C1 subset were independent of individual patients’ glucocorticoid
   therapy and the types of steroids used (Fig. [130]4b, c). Mapping of
   NR3C1 targets in Omnipath^[131]51 showed that pro-inflammatory IL6,
   RELA, and JUN are among the inhibited targets, indicating the possible
   role of this subset of T cells in inflammation control (Fig. [132]4d).
   Mapping the signature genes of the NR3C1 CD4 subset to pathway gene
   sets showed that TGFβ signaling is among the top upregulated pathways,
   suggesting that TGFβ signaling might be critical for either the
   function or maintenance of this T cell subset (Fig. [133]4e).

Fig. 4. Reduced NR3C1 CD4 T cells and increased myeloid cell-NR3C1 CD4
interactions in Emphysematous COPD (E-COPD) compared to nonemphysematous COPD
(NE-COPD).

   [134]Fig. 4
   [135]Open in a new tab

   a Percentages of NR3C1 in CD4 T cells in the in-house discovery
   dataset. b Percentages of NR3C1 in CD4 T cells across 3 disease groups
   stratified by steroids usage. c Percentages of NR3C1 in CD4 T cells
   across 3 disease groups stratified by types of steroids used. d Top
   downstream targets of NR3C1 genes mapped using the Omnipath database,
   blue arrows indicate inhibition by NR3C1, and red arrows indicate
   induction by NR3C1. e EnrichR analyses results of NR3C1 CD4 subset
   signature genes using BioPlanet 2019 and WikiPathway 2019 Human gene
   sets. Interactions via CXCL signaling between identified major cell
   types and CD4 subsets in Control (f), NE-COPD (g), and E-COPD (h).
   Horizontal bar plots represent incoming signal strength estimated by
   Cellchat. A vertical bar plot represents outgoing signal strength.
   Tiles in the heatmap represent the summation of incoming and outgoing
   signal strengths. Statistical significance between the three groups was
   determined using the Kruskal–Wallis test. Pairwise comparisons were
   performed using the Wilcoxon rank-sum test with Holm’s correction.
   Significance code: p < 0.05(*), p < 0.01(**), p < 0.001 (***). Artwork
   was generated from Bioicons ([136]https://bioicons.com/) and NIH NIAID
   BioArt Source ([137]https://bioart.niaid.nih.gov/). Modifications of
   the artwork were performed in Inkscape ([138]https://inkscape.org/).

   Because CD4 T cells exhibited heightened responses to chemokines and
   cytokines in E-COPD (Fig. [139]2c–fd), we next performed interactome
   modeling to identify cellular interactions between different immune
   cells in this cohort. Interactome modeling estimated that CXCL signals
   mediate interactions between myeloid and T cells (Supplementary
   Fig. [140]5a–c). CXCL signaling activities were largely absent between
   NR3C1^+ CD4 T cells and myeloid in either the control or NE-COPD groups
   (Fig. [141]4f, g, Supplementary Fig. [142]5a, b). In contrast,
   interactions between myeloid cells, especially pulmonary macrophages
   and NR3C1^+ CD4 T cell subset via CXCL signaling, were strongly
   predicted in E-COPD (Fig. [143]4h, Supplementary Fig. [144]5c). CCL
   signals were also predicted to mediate interactions between T cell
   subsets and myeloid cells in all three groups (Supplementary
   Fig. [145]5d–f). However, in E-COPD, CCL signals were primarily
   predicted to mediate interactions between T cell subsets (Supplementary
   Fig. [146]5f), whereas in NE-COPD and controls, CCL signals were
   largely predicted to mediate myeloid-T cell interaction (Supplementary
   Fig. [147]5d, e).

   Together, these findings suggest that, independent of steroid usage,
   reduced relative abundance of the NR3C1^+ subset of CD4 T cells in the
   E-COPD cohort might be a distinguishing factor that separates them from
   the NE-COPD and controls. These findings also suggest a specific
   potential interaction between lung myeloid compartments, including
   macrophages with NR3C1^+ CD4 T cells, that may contribute to their
   reduced relative abundance in E-COPD.

Divergent abundance of lung PPARG^+ macrophages associates with disease
phenotypes

   Because we found a potential interaction between lung macrophages and
   the NR3C1^+ T-cell subset, we next examined the heterogeneity of human
   pulmonary macrophages in the lungs of the same cohort. We classified
   human lung macrophages into four distinct subsets: (1) a proliferative
   subset marked by the expression of MKI67, (2) PPARG macrophages, (3)
   Monocytic macrophages expressing CD14 and IL1β, and (4) C1Q macrophages
   (Fig. [148]5a, b). These four subsets of macrophages exhibited distinct
   gene expression and transcription factor activity profiles
   (Fig. [149]5c, d). All subsets of pulmonary macrophages, including
   PPARG, monocytic, and C1Q pulmonary macrophages in E-COPD, showed
   upregulated inflammatory pathways such as IFN-γ signaling
   (Supplementary Fig. [150]6a–c), which has been associated with the
   inhibition of glucocorticoid signaling^[151]52. In NE-COPD, PPARγ
   macrophages were relatively increased (Fig. [152]5e) suggesting
   increased renewal or persistence of this subset of macrophages. In
   contrast, the decreased relative abundance of PPARG macrophages in
   E-COPD mirrored the reduced proportion of NR3C1^+ CD4 T cells in E-COPD
   (Figs. [153]4a and [154]5e). Further, in NE-COPD, where PPARγ
   macrophages were relatively increased, pairwise comparisons of
   ligand-receptor activities predicted increased interactions between
   PPARγ macrophages with NR3C1^+ CD4 T cells through integrins, cell
   adhesion molecules and chemokine/cytokine receptors such as ITGB2,
   ALCAM, and CXCR4 (Supplementary Fig. [155]7a–c). In E-COPD, increased
   interaction through IL-1β and ADRB2 was estimated to be upregulated
   between PPARγ and NR3C1^+ CD4 T cells (Supplementary Fig. [156]7b)
   compared to NE-COPD. Notably, however, in NE-COPD, interactions through
   SIGLEC1-SPN(CD43) between PPARγ macrophages and NR3C1^+ CD4 T cells
   were estimated to be increased (Supplementary Fig.  [157]7e). CD43 is
   critical for CD4 T cell trafficking^[158]53 and is a known
   counter-receptor for SIGLEC1^[159]54.

Fig. 5. Reduced tissue resident macrophages in emphysematous COPD (E-COPD)
compared to nonemphysematous COPD (NE-COPD).

   [160]Fig. 5
   [161]Open in a new tab

   a UMAP embedding of identified pulmonary macrophage subsets. b UMAP
   embedding of major lineage markers’ expressions for identified
   macrophage subsets. PPARG ITGB8 macrophage (PPARG), C1Q macrophages
   (C1QB, HLA-DQB1), Monocytic macrophages (CD14, TREM2) and proliferative
   macrophages (MKI67). c Heatmap of gene signature expression of
   identified macrophage subsets. d Heatmap of DecoupleR estimation of
   transcription factor activity in identified macrophage subsets. e
   Percentage of PPARG ITGB8 macrophages in pulmonary macrophages.
   Statistical significance between the 3 groups was determined using the
   Kruskal–Wallis test. Pairwise comparisons were performed using the
   Wilcoxon rank-sum test with Holm’s correction. Significance code:
   p < 0.05(*), p < 0.01(**), p < 0.001 (***).

NR3C1^+ CD4 subset correlates with lung inflammation

   Given the limitation of scRNA seq studies in identifying tissue-based
   associations between immune cells, we next performed spatial
   transcriptomics to establish the correlation between NR3C1^+ CD4 T
   cells and inflammatory cells in the lungs. We employed the Nanostring
   GeoMx platform to measure bulk RNA transcript levels in microscopic
   regions of interest (ROI), including the lung parenchyma, using samples
   from the same cohorts as the in-house single-cell RNA sequencing
   dataset. This spatial transcriptomics data of the lung parenchyma from
   the GeoMx platform were then deconvoluted for the relative abundance of
   cell types with CibersortX using the in-house single-cell
   RNA-sequencing defined cell type signatures as references. ^[162]55 Our