A vasculature-resident innate lymphoid cell population in mouse lungs

Abstract

Tissue-resident immune cells such as innate lymphoid cells (ILC) are known to reside in the parenchymal compartments of tissues and modulate local immune protection. Here we use intravascular cell labeling, parabiosis and multiplex 3D imaging to identify a population of group 3 ILCs in mice that are present within the intravascular space of lung blood vessels (vILC3). vILC3s are distributed broadly in alveolar capillary beds from which inhaled pathogens enter the lung parenchyma. By contrast, conventional ILC3s in tissue parenchyma are enriched in lymphoid clusters in proximity to large veins. In a mouse model of pneumonia, Pseudomonas aeruginosa infection results in rapid vILC3 expansion and production of chemokines including CCL4. Blocking CCL4 in vivo attenuates neutrophil recruitment to the lung at the early stage of infection, resulting in prolonged inflammation and delayed bacterial clearance. Our findings thus define the intravascular space as a site of ILC residence in mice, and reveal a unique immune cell population that interfaces with tissue alarmins and the circulating immune system for timely host defense.

Innate lymphoid cells (ILC) are instrumental to immediate, local response to pathogens. Here the authors use parabiosis and multiplex 3D imaging to identify a mouse type 3 ILC population that resides in the intravascular space of lung, produces CCL4 for neutrophil recruitment, and protects from Pseudomonas aeruginosa infection in mice.

Introduction

Innate lymphoid cells (ILCs) are largely tissue-resident immune cells enriched in barrier sites where they contribute to host defense, tissue repair, metabolic homeostasis, and inflammatory diseases^1,2. During the perinatal period, ILCs develop from progenitors in the bone marrow or fetal liver and are seeded into their tissue niches^2–4. The maintenance of mature ILCs is mostly achieved through local self-renewal, although ILC replenishment by bone marrow- or lymphoid organ-derived precursors has also been suggested⁵. Signals derived from the microenvironment are chiefly responsible for ILC localization within their tissues of residence, as well as their adoption of tissue-specific functions^6–13. The sites ILCs occupy within tissues are critical for enabling them to mediate appropriate early immune responses. In the lungs, the adventitial cuff surrounding the bronchi and larger blood vessels is the predominate niche site occupied by lung-resident ILC2s and a population of fibroblast-like stromal cells in the same sites provide signals essential for ILC2 activation and expansion during helminth infection¹⁴. Additional work has shown that alveolar fibroblasts in the lungs facilitate the expansion and maturation of pulmonary ILC precursors into ILC3s by producing the niche signal insulin-like growth factor 1 (IGF-1)⁹.

ILC3s are characterized by expression of the transcription factor retinoic acid receptor-related orphan receptor γt (RORγt) and production of the canonical cytokines IL-17 and IL-22^15–18. In response to IL-1β and IL-23 stimulation, ILC3s are quickly activated and exert anti-bacterial function^19–23. In the gastrointestinal tract, two major groups of ILC3s are recognized, including the conventional ILC3s and lymphoid tissue inducer (LTi) ILC3s. Conventional ILC3s occupy the underlying lamina propria as well as form into tiny lymphoid aggregates termed cryptopatches²⁴. While LTi ILC3s are enriched in Peyer’s patches and isolated lymphoid follicles and contribute to the development of secondary lymphoid organs^25–28. Both ILC3 populations are essential for immune responses against various extracellular pathogens, particularly bacteria and fungi²⁹. Emerging evidence also shows that ILC3s significantly contribute to the protection against airway pathogens such as Klebsiella pneumoniae and Streptococcus pneumoniae^30–32. However, little is known about the sites of ILC3 localization and associated function within the lungs. It is also unknown whether mediators other than IL-17 and IL-22 are important for the function of these lung ILCs.

In this work, we seek to determine the sites of ILC3 localization within the lungs and whether this impacts effector functions. We identify an ILC3 population that resides in the intravascular space of the alveolar-capillary beds in the lungs of mice. We demonstrate that these vascular ILC3s produce a set of chemokines including CCL4 for neutrophil recruitment in the early stage of bacterial pneumonia in mice. Our study reveals that vasculature-resident ILC3s connect tissue damage signals to the circulatory system, and also expands our understanding of the repertoire of effector molecules derived from lung ILC3s, which can potentially serve as a therapeutic target during bacterial pneumonia.

Results

Identification of a vasculature-resident ILC3 population in the lung

To begin identifying potential sites occupied by ILC3s within the lungs, we administered fluorescently labeled CD45.2 antibodies intravenously (i.v.) by retro-orbital injection (Fig. 1a). Hematopoietic cells labeled with CD45.2⁺ were present within the blood vasculature (intravascular), while unlabeled CD45.2^– cells were present in alveolar or parenchymal tissue (extravascular) (Supplementary Fig. 1). The vast majority of CD4⁺ T cells in the lungs were located intravascularly and labeled with i.v. injected CD45.2 (ivCD45.2) antibody (Fig. 1b, c). In contrast, ILC2s consistent with their residence in the adventitial cuffs and lung parenchyma, were unlabeled, and thus well represent a tissue resident extravascular population. Unexpectedly, a substantial fraction of ILC3s (32% on average) showed antibody labeling consistent with their presence within the vasculature (Fig. 1b, c). We tested whether intravascular labeled ILC3s were in fact blood-borne, using ILC1s, conventional natural killer (cNK) and CD4⁺ T cells as controls due to their nature as circulatory cells. Indeed, the afore mentioned lymphocytes were readily detectable in the blood, whereas ILC3s were nearly undetectable (Fig. 1d). In addition, the numbers of CD45.2-labeled ILC3s were not affected by perfusion of the lung blood vessels (Fig. 1e), suggesting that the CD45.2-labeled ILC3s are highly adherent to the vascular endothelium rather than circulating freely in the blood. Together these data indicate two ILC3 populations in the lung, one located within the vasculature (vILC3s) and one associated with parenchymal tissue (tILC3s).

Fig. 1. Identification of intravascular ILC3s in mouse lung.

a Lung intravascular leukocytes were in vivo stained via intravenous (i.v.) injection of fluorescently labeled anti-CD45 antibody (ivCD45). Mice were euthanized 5 min later for flow cytometry analysis of lung leukocytes. b Representative FACS plots and c quantification of the percentage of intravascular-resident (ivCD45⁺) and tissue parenchyma-resident (ivCD45^–) Lin⁺ NK1.1^– CD4⁺ T cells, Lin^– Thy1⁺ RORγt^– NK1.1^– KLRG1^+/low ST2⁺ ILC2s, and Lin^– Thy1⁺ NK1.1^– ST2^– RORγt⁺ ILC3s in the lung of C57BL/6 (B6) mice as in (a). d FACS analysis of cell number and frequency of ILC3s, Lin^– RORγt^– NK1.1⁺ ILC1s, Lin⁺ CD4^– NK1.1⁺ conventional NK (cNK), and CD4⁺ T cells in the peripheral blood of B6 mice. e Cell numbers of vILC3s in the lungs of B6 mice that were perfused or not perfused following ivCD45 labeling. ns, not significant, P = 0.8793. f B6.SJL (CD45.1⁺) and B6 (CD45.2⁺) mice were surgically connected to generate parabiotic pairs. One month after surgery, both parabionts were given an i.v. injection of fluorescently labeled anti-Thy1 antibody for 5 min to stain intravascular lymphocytes in the lungs. The percentage of cells expressing CD45.1 or CD45.2 was determined by FACS for indicated populations, to determine exchange between each parabiont. g One month after surgery, both parabionts were treated by intraperitoneal (i.p.) injection with IL-1β and IL-23 daily for 7 days. ivThy1 labeling and FACS analysis of host-derived tILC3s and vILC3s were performed as in (f). h FACS quantification of vILC3 percentage of total lung ILC3s in B6 mice at different ages; ****P < 0.0001; **P = 0.0059. Data in (c–h) were shown as mean ± SD and represent findings from at least two independent experiments. n = 8 or 11 in (c), 5 in (d), 6 in (e), 4 in (f) and (g); in (h), n = 6 (3 weeks), 5 (6 weeks), 4 (9 weeks), 6 (16 weeks), or 9 (20 weeks). The statistics were obtained in (e) by unpaired two-tailed t test and in (h) by one-way analysis of variance with Tukey’s multiple comparisons test.

We then used a parabiosis model to further assess whether vILC3s are vasculature-resident cells or are circulating in the bloodstream at steady state or following cytokine treatment. Parabiotic mice were generated via surgically joining a CD45.2-expressing C57BL/6 mouse and a CD45.1-expressing B6.SJL mouse. These phenotypically distinct markers allow determination of cell origin following migration between the parabionts using flow cytometry. Four weeks post parabiosis surgery, parabiotic mice were i.v. injected with PE-Thy1.2 antibody (Fig. 1f). CD4⁺ T cells, which continually traffic between the parabionts showed a 50:50 distribution of CD45.1⁺ and CD45.2⁺ cells within the paired animals (Fig. 1f). Whereas ILC2s, which are tissue resident in the lungs and were used as a control for non-migratory cells, underwent negligible exchange (Fig. 1f). No discernable transfer of tILC3s or vILC3s was detected between parabiotic mice (Fig. 1f), suggesting that both populations are tissue-resident and sessile within that organ in the steady state. Furthermore, treatment of parabiotic mice with IL-1β plus IL-23, which induce ILC3s to proliferate and produce effector molecules³³, did not lead to exchange of vILC3s or tILC3s between parabiotic mice (Fig. 1g).

We also characterized vILC3s in mice at different ages. vILC3s were detected at three weeks of age as a low proportion of the total lung ILC3s. The fractional representation of vILC3 increased with age, and the number doubled at six weeks of age and then remained stable until a decline at twenty weeks (Fig. 1h). Taken together these results indicate that vILC3s populate the intravascular niches in the early life and are self-maintained in the adult lung blood vessels without going into circulation, even upon cytokine stimulation.

We also addressed the possibility of whether vILC3s are perivascular and their staining with ivCD45 is due to potential ‘leakage’ of i.v. injected antibodies into the perivascular region. For this purpose, we performed a control stain targeting pericytes, which wrap blood capillaries in the lung and express CD140b (platelet-derived growth factor receptor-β, PDGFR-β)³⁴. A mix of anti-CD140b and anti-CD45.2 antibodies were i.v. injected into mice 5 minutes before euthanasia, and a control bulk staining was performed in parallel to validate CD140b antibody (Supplementary Fig. 2a). We found that lung pericytes were not stained by i.v. injected CD140b antibody and, consistently, 99% of ILC2s were not stained by ivCD45.2, whereas 95% of CD4⁺ T cells were labeled by ivCD45.2 (Supplementary Fig. 2b, c). Our results are in accordance with previously published work³⁵, suggesting that there is no detectable leakage of i.v. injected antibodies into the perivascular region or parenchyma tissue of the lung.

vILC3s reside in the intravascular space and are widely distributed in the capillary beds

We next sought to determine the exact anatomical location of vILC3s and tILC3s. We performed clearing-enhanced 3D (Ce3D) imaging³⁶, using RORγt-GFP reporter mice in conjunction with ivCD45 antibody labeling. We scanned a lung lobe with a Z stack of 300 μm thickness to obtain a high-resolution 3D view of the macro-scale tissue architecture, identifying bronchioles and large blood vessels. vILC3s (CD3⁻ GFP⁺ ivCD45⁺) and tILC3s (CD3⁻ GFP⁺ ivCD45⁻) presented distinct distribution patterns with vILC3s broadly distributed throughout the distal lung, in contrast to tILC3s that were largely clustered, with many in proximity to large αSMA-labeled veins (Fig. 2a and Supplementary Movie 1), which are coated by the irregularly formed αSMA⁺ smooth muscle cells³⁷. High magnification imaging showed the presence of vILC3s within Lyve-1⁺ vasculature in alveoli (Fig. 2b). Given that Lyve-1 labels both lymphatic and vascular endothelial cells, we used PECAM1 (CD31) to mark the vascular endothelium³⁸. XYZ 3-D rotation of tissue sections confirmed that vILC3s were located inside CD31⁺ capillaries (Fig. 2c). To highlight further the intravascular localization of vILC3s, we used a rotating 3D image, viewing the cells at different angles of confocal scan (Supplementary Movie 2). When viewing a z-stack image of a CD31-marked capillary from the top or the bottom, vILC3s were not observed; but when focusing on the middle layers within the capillary vILC3s were seen (Supplementary Movie 2), emphasizing that vILC3s are intravascular resident. In contrast, the majority of tILC3s were located in lymphoid clusters that also contained CD3⁺ T cells and were in proximity to veins (Fig. 2d and Supplementary Movie 3). Quantification of distances within each subset, for cell-to-cell, cell-to-airway, and cell-to-large αSMA⁺ vessels, confirmed that tILC3s are clustered together and located in proximity to the large veins, whereas vILC3s are widely distributed in the capillary beds surrounding the airways (Fig. 2e). Together these results revealed that the blood vasculature is a previously unappreciated niche site for the residency of ILC3s.

Fig. 2. vILC3s are widely distributed in the capillary beds of the lung.

a Representative clearing-enhanced 3D (Ce3D) imaging of the lung from RORγt-GFP reporter mice that received ivCD45 labeling. tILC3s and vILC3s were identified based on indicated marker expression. The vessels that are coated by the irregularly formed αSMA⁺ smooth muscle cells are veins. b Three representative images of ivCD45⁺ CD3^– RORγt-GFP⁺ vILC3s (white arrows) in capillary beds. Lyve-1 labels blood vessel and lymphatic endothelial cells. c x-y (top left), y-z (top right), and x-z (bottom) views of vILC3s that were localized within the intravascular space of capillary vessels. CD31 is blood vessel endothelial cell marker. d Three representative images of ivCD45^– CD3^– RORγt-GFP⁺ tILC3s (white arrows) in the lymphoid clusters in tissue and in proximity to the veins. e Quantification of the shortest distance among cells within each group, and shortest distance to the airways or αSMA⁺ vessels from vILC3s or tILC3s. n = 386 for tILC3 and 579 for vILC3 in (e). Data in (e) shown as mean ± SD and two-tailed non-parametric Mann-Whitney U test was used to calculate P-values. Data in (a–e) were representative of two independent experiments.

vILC3s are phenotypically distinct from tILC3s

We next surveyed a group of mouse strains that lack various immune cells to confirm the lineage identity of vILC3s. We examined Rag1^–/– mice (no T and B lymphocytes), Klrg1^CreGata3^fl/fl mice (no ILC2s)³⁹, Tcrb^–/– mice (no αβT cells), muMt ^–/– mice (no B cells), as well as Rorc^–/– mice (that lack both ILC3s and their adaptive counterparts Th17 cells⁴⁰). Both vILC3s and tILC3s were completely depleted in Rorc^–/– mice; in contrast, the numbers of vILC3s were not altered in the lungs of Tcrb^–/–, muMt^–/– and Klrg1^CreGata3^fl/fl mice (Fig. 3a and Supplementary Fig. 3a–c). Notably, Rag1^–/– mice harbored significantly higher numbers of vILC3s, tILC3s and ILC1s but not ILC2s (Fig. 3a and Supplementary Fig. 3c–e), presumably due to the lack of competition with adaptive lymphocytes for IL-7, a niche cytokine that maintains ILC and T cell homeostasis⁴¹. Flow cytometry analysis showed that vILC3s and tILC3s had similar expression levels of CD127 (IL-7Rα) (Fig. 3b and Supplementary Fig. 4a). vILC3s expressed a higher level of CCR7 and Sca-1 than tILC3s, whereas tILC3s contained a higher proportion of MHC-II⁺ and CD4⁺ cells (Fig. 3b and Supplementary Fig. 4a). When stained with NKp46 and CCR6, vILC3s were found to consist of three subsets, NKp46⁺, CCR6⁺ and NKp46^–CCR6^–, similar to gut ILC3s (Fig. 3c), although the proportion of NKp46⁺ subset in the lung is much lower than in the gut. tILC3s consisted of two subsets, CCR6⁺ and NKp46^– CCR6^–, lacking an NKp46⁺ subset. Consistent with NKp46 expression, both vILC3s and tILC3s expressed lower levels of T-bet than gut NKp46⁺ ILC3s (Fig. 3d). Collectively, these results indicated that vILC3s share the same lineage identity with tILC3s, both requiring RORγt for their development and homeostasis, although the two seem to differ slightly in terms of cell surface marker expression.

Fig. 3. vILC3s and tILC3s are phenotypically and transcriptionally distinct.

a Lung intravascular leukocytes were in vivo stained via i.v. injection of fluorescently labeled anti-CD45 antibody (ivCD45). Mice were euthanized 5 min later for flow cytometry analysis of lung leukocytes. Cell numbers of vILC3s in the lungs of B6 (WT), Rag1^–/–, Gata3^fl/fl Klrg1^Cre, Tcrβ^–/–, muMt^–/– and Rorc^–/– mouse lines determined by flow cytometry; ****P < 0.0001. b Histograms of CD127, CCR7, CD4, MHC-II and Sca-1 expression in vILC3s, tILC3s and ILC2s from the lungs of B6 mice determined by flow cytometry. c FACS analysis of NKp46 and CCR6 expression in gut ILC3s, lung tILC3s, vILC3s and ILC1s. d Histograms of T-bet expression in gut ILC3s, lung tILC3s and vILC3s. e vILC3s and tILC3s were FACS sorted from RORγt-GFP reporter mice and subjected to a bulk RNA-Seq analysis. f Volcano plot of differentially expressed genes in tILC3s (blue) and vILC3s (red). X axis represents log2 transformed fold change. Y axis represents negative log10 fold of adjusted P-value. In the volcano plots colored dots represented significant genes with adjusted P-value < 0.01. g Heatmaps of the expression of secreted factors (left) and cell adhesion molecules (right) between tILC3s and vILC3s. h Metascape analysis of signaling networks enriched in vILC3s compared to tILC3s. Each node represents a functional term. The size of the node is proportional to the number of genes that fall into the corresponding term and the color reflects its cluster identity. n = 3, 4, 5 or 11 in (a). The data are shown as mean ± SD. Data in (a–d) were represent results from at least two independent experiments. The statistics in (a) were obtained by one-way analysis of variance with Dunnett’s multiple comparisons test.

To gain greater insights into the functionality of vILC3s, we performed bulk RNA-sequencing analysis of sorted vILC3s and tILC3s from the lungs of unchallenged RORγt-GFP mice (Fig. 3e). Approximately 100 genes were differentially expressed between the two subsets (Fig. 3f), highlighting their shared lineage identity. vILC3s expressed high levels of the chemokines Ccl4, Ccl5 and Ccl6 (Fig. 3f, g), suggesting that they are involved in immune cell recruitment. Steady-state vILC3s and tILC3s expressed similar levels of Il17a and Il17f, while Il22 was higher in vILC3s (Fig. 3g). Additionally, compared to tILC3s, vILC3s expressed higher levels of Cd7, Klrc1, transcription factors including Tbx21, Stat3 and Stat4, signaling receptors including Notch1, Notch2 and Nr4a1 (Fig. 3f, g and Supplementary Fig. 4b), consistent with their potential roles in mediating inflammatory signaling. Notably, vILC3s expressed a variety of adhesion molecules including Icam1, Itgb2 and Itgb7 (Fig. 3g), which potentially facilitate the interaction between vILC3s and endothelial cells and sustain vILC3 residency in the vasculature. In addition, vILC3s expressed S1pr1 (Fig. 3g), a receptor for sphingosine 1-phosphate (S1P) that plays a pivotal role in regulating immune cell trafficking as well as in supporting T cell survival⁴². In contrast, tILC3s highly expressed insulin-like growth factor-binding proteins Igfbp6 and Igfbp7 (Fig. 3f), the ligand of which (IGF1) has been shown to promote ILC3 development in newborn lungs⁹. tILC3s also expressed lymphotoxin α (Lta) (Fig. 3g), which plays a critical role in the development and maintenance of secondary lymphoid organs⁴³. Since tILC3s were located mainly in lymphoid clusters, it is possible that tILC3s are LTi-like cells and contribute to the formation and maintenance of the lymphoid cluster structures in the lung. Notably, Metascape analysis showed that vILC3s were enriched with transcripts related to signaling networks including vasculature development, regulation of cell motility, response to wounding, and cell-cell adhesion (Fig. 3h). These findings indicate that the transcriptomic features of vILC3s are consistent with their vascular localization and potential immunological functions.

vILC3s produce a set of chemokines including CCL4 in the early stage of bacterial infection

ILC3s primarily respond to the alarmin cytokines IL-23 and IL-1β during inflammation or infection^44–46. To gain insights into the functionality of vILC3s and tILC3s, we injected IL-23 plus IL-1β intraperitoneally into C57BL/6 mice and monitored their response dynamics. The total number of RORγt⁺ ILC3s in the lung increased steadily over the course of cytokine treatment (Fig. 4a). On day 7 post first treatment, both vILC3s and tILC3s increased ~4 folds in number (Fig. 4b). Interestingly, the majority of vILC3s (70% vILC3s versus 30% tILC3) expressed Ki67 on day 3 following cytokine stimulation (Fig. 4c), indicating greater proliferation of vILC3s than tILC3s. Moreover, vILC3s produced a far greater amount of IL-17A than tILC3s following ex vivo stimulation with phorbol myristate acetate (PMA) and ionomycin, although both subsets produced similar amounts of IL-22 upon stimulation (Fig. 4d, e).

Fig. 4. vILC3s rapidly respond to cytokine stimulation.

a Lung intravascular leukocytes were in vivo stained via i.v. injection of fluorescently labeled anti-CD45 antibody. Mice were euthanized 5 min later for flow cytometry analysis of lung leukocytes. B6 mice were i.p. treated with IL-1β and IL-23 daily for up to 7 days, and the numbers of total lung ILC3s were analyzed by flow cytometry; ns, not significant, P = 0.6526; **P = 0.007; ****P < 0.0001. b Numbers of tILC3s and vILC3s in the lungs of B6 mice that were i.p. treated with IL-1β and IL-23 daily for 7 days; ****P < 0.0001 (left); **P = 0.0041 (right). c Percentage of Ki67⁺ tILC3s and vILC3s as in (a); ***P = 0.0006. d Representative FACS plots and (e) quantification of the percentage of IL-17A⁺ and IL-22⁺ tILC3s and vILC3s from Rag1^–/– mice stimulated with PMA and ionomycin (in the presence of IL-2, IL-6, IL-1β and IL-23) ex vivo for 4 hours; ****P < 0.0001 (left); ns, P = 0.0546 (right). n = 4 or 6 in (a–c), 7 in (e). The data are shown as mean ± SD. Data in (a–c) and (e) are representative results from at least two independent experiments. The statistics for comparison of two variables in (b) and (e) were obtained by unpaired two-tailed t test; for more than two variables statistics in (a) and (c) were obtained by one-way analysis of variance with Dunnett’s multiple comparisons test or two-way ANOVA with Bonferroni’s multiple comparisons test.

We then investigated the physiological function of vILC3s using a pneumonia mouse model infected with Pseudomonas aeruginosa, a leading cause of nosocomial infections and pneumonia⁴⁷. We used the P. aeruginosa Schroeter Migula strain (ATCC 10145), which is genetically close to moderately virulent isolates PA103 and PAO1, and relatively distant from highly virulent isolate PA14 (Supplementary Fig. 5a). Through a virulence test, we established a working dose of this strain (8 × 10⁵ CFU), which induced notable bodyweight loss in C57BL/6 mice without causing substantial mortality (Supplementary Fig. 5b, c). ILC3s, as well as other groups of ILCs, are known to provide critical immune protection during the early stage of infection⁴⁸. We therefore focused on the acute phase of P. aeruginosa infection. Similar to IL-23 plus IL-1β stimulation, P. aeruginosa infection induced the expansion of lung ILC3s (Fig. 5a). On day 3 post infection, 45% of vILC3s compared to 20% of tILC3s expressed Ki67 (Fig. 5b), indicating that vILC3s underwent more rapid expansion than tILC3s during the early stage of P. aeruginosa infection. Interestingly, the number of both vILC3s and tILC3s plateaued on day 7 post infection, yet remained significantly increased even after the infection was resolved on day 30 (Fig. 5c, d). However, for both subsets the rate of proliferation had returned to steady state, based on Ki67 expression (Fig. 5b).

Fig. 5. vILC3s rapidly respond to P. aeruginosa infection and produce a unique set of chemokines.

a B6 mice were given a single dose (8 × 10⁵ CFU) of P. aeruginosa via intranasal administration, and total ILC3s in the lungs were determined by flow cytometry on day 3, 5, 7, and 30 post infection; *P = 0.0423; ***P = 0.0004; **P = 0.0029. b Percentage of Ki67⁺ tILC3s and vILC3s as in (a); day 3, ***P = 0.0001; day 7, **P = 0.009. c, d Cell numbers of tILC3s and vILC3s on day 7 (in c: ***P = 0.001 (left); **P = 0.0092 (right)) or day 30 (in d: ***P = 0.0002 (left); ***P = 0.0004 (right)) post infection as in (a). e vILC3s and tILC3s were sorted from the lungs of RORγt-GFP mice on day 3 post P.a. infection, and subjected to bulk RNA-seq analysis. Volcano plot of differentially expressed genes in tILC3s (blue) and vILC3s (red). X axis represents log2 transformed fold change. Y axis represents negative log10 fold of adjusted P-value. In the volcano plots colored dots represented significant genes with adjusted P-value < 0.01. f Real-time qPCR analysis of Ccl4 mRNA expression in tILC3s and vILC3s sorted from the RORγt-GFP mice on day 3 post P.a. infection. Ccl4 level was relative to Gapdh mRNA; **P = 0.0077. g The read counts of scRNA-seq were extracted from the published datasets (GSE192890) and applied together with bulk RNA-seq in I to a modified R script based on the DotPlot function in Seurat. Both the size and color of the dots indicate the average expression scale of the gene. n = 5, 6 or 9 in (a–d), 5 in (f). The data in (a–d) and (f) were shown as mean ± SD and representative data from at least two independent experiments. The statistics for comparison of two variables (c), (d) and (f) were obtained by unpaired two-tailed t test; for more than two variables (a) and (b) statistics were obtained by one-way analysis of variance with Dunnett’s multiple comparisons test or two-way ANOVA with Bonferroni’s multiple comparisons test.

Next, we examined the transcriptomes of both ILC3 subsets during infection. vILC3s and tILC3s were sorted from the lungs of RORγt-GFP reporter mice on day 3 post P. aeruginosa infection and were then subjected to bulk RNA-seq. Comparison of differentially expressed genes revealed that 56 and 40 transcripts were significantly upregulated and downregulated between tILC3s and vILC3s, respectively (Fig. 5e). Notably, vILC3s expressed high levels of chemokine compared to tILC3s, including Ccl4, Ccl5, Ccl3, Ccl6 and Ccl9 (Fig. 5e and Supplementary Fig. 6a). Consistently, Metascape analysis revealed that signaling networks associated with cell migration, cell division and IL-17 signals were highly enriched in vILC3s compared to tILC3s (Supplementary Fig. 6b). vILC3s expressed the IL-12 receptors Il12rb1 and Il12rb2, whereas tILC3s expressed the IL-18 receptor Il18r1 (Supplementary Fig. 6a), suggesting that vILC3s and tILC3s respond to different cytokine signals for their activation. vILC3s also expressed colony stimulating factor 1 (Csf1) (Supplementary Fig. 6a, d), a primary growth factor required for the survival, proliferation and differentiation of monocytes and macrophages⁴⁹. In response to P. aeruginosa infection, vILC3s expressed higher levels of Il17f mRNA (Supplementary Fig. 6a) and had upregulated IL-17A signaling compared with tILC3s (Supplementary Fig. 6b), consistent with protein production upon PMA/ionomycin stimulation (Fig. 4d, e). In contrast, activated tILC3s expressed high levels of Il22, which modulates epithelial cell homeostasis and regeneration^50–52, and Ltb (encoding Lymphotoxin-β), a cytokine required for the development of secondary lymphoid structures⁵³ (Supplementary Fig. 6a, c). These results suggest that vILC3s regulate immune cell recruitment and monocyte activation during infection and that tILC3s are associated with epithelial remodeling and lymphoid organ development.

Given that vILC3s expressed Il12rb1 and Il12rb2 in response to P. aeruginosa infection (Supplementary Fig. 6a) and that IL-12 has been shown to regulate plasticity of ILC3-ILC1 transitions in the intestine⁵⁴, we tested whether IL-12 had any effects on the phenotype and function of vILC3s and tILC3s in vitro. In response to IL-12, vILC3s produced a moderate amount of IFN-γ, while tILC3s did not produce any (Supplementary Fig. 6e, f). When cultured with IL-12 plus IL-1β and IL-23, both subsets produced a similar amount of IL-22, and again, vILC3s produced IFN-γ to a similar level as culture with IL-12 alone (Supplementary Fig. 6e, f). These results further suggest that these two subsets of lung ILC3s respond to different inflammatory signals and vILC3s have an intrinsic plasticity to produce IFN-γ.

Comparison between vILC3 and tILC3 revealed that Ccl4 was the most highly expressed immune-related gene in vILC3s during infection (Fig. 5e). In line with this finding, Ccl4 transcripts were also the most prominent in P. aeruginosa-activated vILC3 versus steady-state vILC3 (Supplementary Fig. 6d), which was further confirmed by real-time qPCR (Fig. 5f). We integrated our bulk RNA-seq data with a published single-cell (sc) RNA-seq dataset that profiled the majority of lung CD45⁺ immune cells during P. aeruginosa infection⁵⁵, using a previously described bioinformatic pipeline⁵⁶. These clusters were annotated manually based on the curated lineage marker genes (Fig. 5g, rows from neutrophils to CD8 T cells). Since there is no standard way to infer the percentage of cells expressing a specific gene from bulk RNA-seq, a modified dot plot was used to show the expression level of selected genes in clusters from scRNA-seq (Fig. 5g, rows from neutrophils to CD8 T cells) and cells from bulk RNA-seq (Fig. 5g, row vILC3s and tILC3s). We observed that some chemokines were preferentially produced by certain immune cell types, for example, Ccl3 and Ccl6 by monocytes, macrophages and vILC3s, Ccl5 by NK/ILC1s, NKT and CD8⁺ T cells. Strikingly, we observed that vILC3s, as well as monocytes and NK/ILC1s, were among the highest expressers of Ccl4 (Fig. 5g). CCL4, also known as macrophage inflammatory proteins (MIP)-1β, is a chemotactic cytokine that orchestrates the recruitment of many immune cells during infection and inflammation^57,58. Thus, we next investigated the functional role of vILC3-derived CCL4 during P. aeruginosa infection.

vILC3-produced CCL4 recruits neutrophils for bacteria clearance

We used anti-CCL4 neutralizing antibodies to block CCL4 function at the early stage of P. aeruginosa infection (Fig. 6a). C57BL/6 mice receiving CCL4 neutralizing antibodies showed significantly more body weight loss than the control group that received IgG (Supplementary Fig. 7a). Given that Ccl5 was also found to be highly expressed by vILC3s at the steady state and during infection, we tested the impact of CCL5 neutralization. CCL5 blockade also resulted in a greater body weight loss than control IgG group (Supplementary Fig. 7b), although the effect was less pronounced than with CCL4 neutralization. We next performed CCL4 neutralization in Rag1^–/– mice, which have the advantage of containing a greater number of ILC3s within the lungs than WT mice (Fig. 3a), and also lack adaptive lymphocytes, including Th17 cells, removing them as a variable. CCL4 blockage again resulted in significantly more body weight loss than the control IgG group on day 2 and day 3 post P. aeruginosa infection (Fig. 6b). Furthermore, Rag1^–/– mice receiving two doses of CCL4 neutralizing antibodies failed to regain their starting bodyweight on day 7 post P. aeruginosa infection (Supplementary Fig. 7c, d). We also tested CCL4 neutralization in Rorc^–/– mice, which lack ILC3s and Th17 cells, and did not observe exacerbated loss of body weight as compared to control (Fig. 6c). Moreover, body weight restoration in Rorc^–/– mice was unaffected by CCL4 neutralization post P. aeruginosa infection (Supplementary Fig. 7e). These results indicate the important role of vILC3-derived CCL4 in the protection of the lung during bacterial infection.

Fig. 6. vILC3-derived CCL4 promotes the recruitment of neutrophils for host defense against bacterial infection.

a P.a.-infected mice were intravenously treated with normal IgG or an anti-CCL4 neutralizing antibody at 24 and 48 hours post infection. b Bodyweight loss of Rag1^–/– mice as in (a); ns, not significant, P = 0.8514; *P = 0.0101; ***P = 0.0001. c Bodyweight loss of Rorc^–/– mice as in (a); ns. d The numbers of neutrophils (CD11c^– CD11b⁺ Ly6G⁺ Siglec-F⁺); ***P = 0.0001 (NT vs. Pa + αIgG); *P = 0.0188 (NT vs. Pa + αCCL4); *P = 0.0117 (Pa + αIgG vs. Pa + αCCL4), macrophages (CD11c⁺ F4/80⁺); **P = 0.0042 (NT vs. Pa + αIgG); *P = 0.0374 (NT vs. Pa + αCCL4); ns, P = 0.5294 (Pa + αIgG vs. Pa + αCCL4), and cNK cells; ns, P = 0.2948 (NT vs. Pa + αIgG); ns, P = 0.3733 (NT vs. Pa + αCCL4); ns, P = 0.8688 (Pa + αIgG vs. Pa + αCCL4), in the lungs of Rag1^–/– mice at 72 hours post infection. e The numbers of neutrophils; *P = 0.0242 (NT vs. Pa + αIgG); **P = 0.0032 (NT vs. Pa + αCCL4); ns, P = 0.336 (Pa + αIgG vs. Pa + αCCL4), macrophages; ns, P = 0.7645 (NT vs. Pa + αIgG); ns, P = 0.5164 (NT vs. Pa + αCCL4); ns, P = 0.9094 (Pa + αIgG vs. Pa + αCCL4), and CD4⁺ T cells; ns, P = 0.4859 (NT vs. Pa + αIgG); ns, P = 0.969 (NT vs. Pa + αCCL4); ns, P = 0.466 (Pa + αIgG vs. Pa + αCCL4) in the lungs of Rorc^–/– mice at 72 hours post infection. f Leukocytes isolated from bone marrow or blood of B6 mice were subjected to a Transwell system with added recombinant CCL4 in the lower chamber. The percentage of transmigrated neutrophils was determined by flow cytometry; ***P = 0.0008 (left); ****P < 0.0001 (right). g Sorted vILC3s from P.a.-infected mice were added (5 × 10³ cells/chamber) into the lower chamber with αCCL4 neutralizing antibody or isotype control (both at 100 ng/ml), and leukocytes isolated from blood were subjected to the insert of Transwell plate. Control had no vILC3s present in the lower chamber. The percentage of transmigrated neutrophils was determined by flow cytometry; **P = 0.0034. h Bacterial burden of P.a. in the lungs of Rag1^–/– mice at 72 h post infection; ****P < 0.0001. i Three representative H&E staining of paraffin-embedded lung sections in (h). j Lung injury score (LIS) in (i); *, P = 0.036. k Static compliance of the lungs of Rag1^–/– mice measured by flexiVent system at 72 hours post infection; *P = 0.0306. n = 9 in (b), 6 in (c), 6, 7, 8 or 11 in (d), 4, 6 or 8 in (e), 3, 6 or 9 in (f), 4 in (g), 7 or 8 in (h), 6 in (j), and 10 in (k). Data in (b–h), (j) and (k) was shown as mean ± SD and represent data from at least two independent experiments. The statistics for comparison of two variables in (f–h), (j) and (k) were obtained by unpaired two-tailed t test; for more than two variables in (b–e) the statistics were obtained by one-way analysis of variance with Tukey’s multiple comparisons test.

Recruitment of neutrophils from the circulation is crucial for elimination of P. aeruginosa bacteria from the lung⁵⁹. Neutrophils highly express the CCL4 receptor CCR5⁶⁰, but the role of the CCL4-CCR5 axis in neutrophil recruitment during infection is unknown. Flow cytometry analysis revealed that the recruitment of neutrophils to the lung was significantly impaired in Rag1^–/– mice receiving CCL4 neutralizing antibodies, with lower numbers of neutrophils present on day 3 post P. aeruginosa infection (Fig. 6d). Interestingly, CCL4 blockage led to only a modest decrease in the percentage of extravascular neutrophils among the total neutrophils in the lung (Supplementary Fig. 7f), suggesting that that vILC3-derived CCL4 plays an important role in recruiting circulatory neutrophils and sustaining them in the infected lungs, rather than promoting trans-endothelial migration from the blood vessels into the parenchyma and alveoli.

The numbers of macrophages or NK cells in the lung were not impacted by CCL4 neutralization (Fig. 6d). Furthermore, a significantly higher number of neutrophils remained in the lungs of Rag1^–/– mice receiving two doses of CCL4 neutralizing antibodies but not in mice receiving IgG at day 7 post infection (Supplementary Fig. 7g), likely due to delayed immune responses in accordance with the delayed body weight restoration (Supplementary Fig. 7d). However, CCL4 neutralization did not affect the recruitment of neutrophils, CD4⁺ T cells and macrophages to the lung in Rorc^–/– mice (Fig. 6e and Supplementary Fig. 7h).

We then asked whether CCL4 directly modulates neutrophil migration as a chemoattractant. We subjected neutrophils from bone marrow or peripheral blood to in vitro chemotaxis assays, in which recombinant CCL4 was added in the lower chamber of a Transwell system. The presence of CCL4 led to an enhancement of neutrophil migration in vitro (Fig. 6f), suggesting that CCL4 has a role in directly regulating neutrophil migration and recruitment. We then tested the effect of vILC3-derived CCL4 on neutrophil migration. Sorted vILC3s from P. aeruginosa-infected mice were loaded into the lower chamber with either anti-CCL4 neutralizing antibody or isotype control antibody. Neutrophils within the peripheral blood were loaded into the Transwell insert. We found that vILC3s significantly prompted neutrophil transmigration, and CCL4 neutralization diminished such an effect (Fig. 6g). These results demonstrated that vILC3-derived CCL4 can directly induce neutrophil transmigration in vitro.

Neutrophil recruitment is critical to clearance of P. aeruginosa from the lungs during infection. In accordance with reduced neutrophil recruitment, CCL4 neutralization led to increased bacterial burdens in the lung (Fig. 6h). Hematoxylin and eosin (H&E) staining and tissue injury analysis showed that, compared to the control IgG group, the mice receiving CCL4 neutralizing antibodies exhibited elevated tissue disruption and injury scores in the lung (Fig. 6i, j). Further suggesting that CCL4-mediated neutrophil recruitment provides critical protection against tissue damage at the early stage of infection. Finally, we tested the impact of CCL4 blockade on lung compliance, which reflects the intrinsic elastic properties of the respiratory system. Previous studies have shown that bacterial infection results in decreased compliance due to exudate and inflammation⁶¹. Indeed, lung compliance was significantly decreased in mice receiving CCL4 blocking antibody when compared to infection alone (Fig. 6k), indicating exacerbated pneumonia and compromised lung physiological function in these mice. Taken together these findings highlight the important role of vILC3-derived CCL4 in host defense by recruiting neutrophils at the early stage of infection.

Discussion

Here, we employed intravascular cell labeling, parabiosis mice, multiplex 3D imaging and transcriptome profiling to identify and characterize an ILC3 population that resides within the vascular space of capillaries in the lung. These vasculature-resident ILC3s respond at the early stage of bacterial infection, producing cytokines such as IL-17 and chemokines including CCL4 which facilitates the recruitment of neutrophils to the lung tissue for bacteria clearance.

Similar to other tissue-resident cells, ILCs are generally found to localize in the parenchymal compartments of barrier tissues, for instance, ILC2s and ILC3s in the lamina propria of the intestine⁶²; ILC3s close to the sebaceous glands of the skin⁶³; and ILC2s in adventitial cuffs of the lung¹⁴. The identification of vILC3s extends this paradigm, revealing that the vasculature of the lung is an underappreciated site for immune cell residency, or at a minimum for ILC3 residency. vILC3s appeared in early life (Fig. 1h), suggesting that vILC3s originate directly from bone marrow progenitors distributed to the lung during early development rather than induced later in life. vILC3s expressed high levels of adhesion molecules, which likely facilitates their retention in the lung vasculature. Although alveolar fibroblast-produced IGF-1 is critical for lung ILC3 development in newborn mice⁹, vILC3s might require additional niche signals for their maintenance considering the spatial separation from alveolar fibroblasts by endothelial walls. Along this line, vILC3s expressed high levels of S1pr1 and Notch1/2 (Fig. 3g and Supplementary Fig. 4b). S1P is an important survival signal for circulatory T cells and the blood contains a high level of S1P compared to parenchymal tissues⁶⁴. Notch signaling plays an important role in the differentiation and maintenance of intestinal ILC3s^65,66. Thus, it is possible that S1P and/or Notch signals serve as additional niche factors contributing to vILC3 maintenance.

Although our parabiosis experiments suggest that vILC3s do not circulate widely in the blood stream, there is a possibility that vILC3s actively patrol capillary networks in the alveoli without leaving the lung. Patrolling behavior by a Nr4a1^hi Cx3cr1^hi monocyte subpopulation is critical for the maintenance of lung homeostasis^67–71. Significantly, vILC3s but not tILC3s expressed high levels of Nr4a1 and Cx3cr1 (Fig. 3g). It will be interesting in the future to determine whether at homeostasis vILC3s patrol lung tissue by crawling on the apical side of vascular endothelium to mediate immuno-surveillance in a manner similar to the monocyte subpopulation.

It is intriguing that tILC3s are primarily localized in the lymphoid clusters in close proximity to veins. The expression of Lta and Ltb suggests that they are potentially LTi-like cells, prompting lymphoid organogenesis. Lung ILC3s have been shown to be dispensable for the formation of the inducible bronchus-associated lymphoid tissue (iBALT)⁷². However, it is possible that tILC3s are required for the formation and/or maintenance of naturally existing lymphoid cluster structures. Unlike intestinal ILC3s that express MHC-II^73,74, only a very small proportion of tILC3s express MHC-II (Fig. 3b). tILC3s are localized with T cells in lymphoid clusters, raising the possibility of tILC3 interactions with T cells to prompt adaptive immune responses.

Given the plasticity of ILCs, a potential question is whether circulating NK/ILC1s can acquire a vILC3-like phenotype during adaptation to the intravascular niche. Although only a small portion of vILC3s express NKp46 (Fig. 3c), it is possible that NK/ILC1s lose the expression of NKp46 and T-bet but upregulate RORγt, becoming vILC3-like cells. Since the vILC3 population contains three phenotypic subsets based on CCR6 and NKp46 expression, it presents another hypothesis: when ILC3 progenitors populate from the bone marrow to peripheral tissue sites during early development, a fraction of progenitor cells may establish in the vasculature of alveolar-capillary beds, surviving and developing to vILC3s. We attempted to identify immature vILC3s or progenitor-like cells in the lung vasculature at the neonatal stage, but it is technically challenging given that neonatal mice are too small for i.v. injection. An interesting future area of exploration will be illuminating the developmental origin of vILC3s as well as tILC3s.

Another interesting question is whether there is a dynamic transition between vILC3s and tILC3s. Tissue microenvironment has a significant impact on the phenotype and functions of innate lymphoid cells⁷⁵. Cytokine milieu can lead ILC3s to phenotypically switch from NKp46⁻ ILC3s into NKp46⁺ ILC3s in response to IL-2, IL-1β, IL-23, and IL-12, and this transition has significant implications for their effector functions^54,76,77. Upregulation of T-bet in CCR6⁻ RORγt⁺ ILCs in the intestine has been shown to be essential for protection against Salmonella enterica infection and signals for this upregulation are derived from the commensal microbiota and IL-23⁷⁸. Rapid activation and proliferation of vILC3 during infections could suggest a possibility that they may enter and populate into the niches associated with tILC3s, acquiring tILC3 phenotypic traits.

vILC3s are widely distributed across capillary beds in proximity to the airways, from which inhaled pathogens invade the lung. Therefore, vILC3s are ideally positioned at the interface of the circulatory system and alveoli, acting as a first responder to fight against pathogenic bacteria. CCL4 neutralization results suggest that vILC3s are critical for arresting circulatory neutrophils in the lung at the early stage of infection. It is notable that vILC3s at steady state or in response to P. aeruginosa do not express typical chemokines associated with neutrophil chemotaxis, such as CXCL1 (KC), CXCL2 (MIP-2) or CXCL5⁷⁹, and this likely accounts for the modest effects we found in vitro. However, our finding expands the repertoire of ILC3 effector molecules and it is possible that vILC3-derived CCL4 works synergistically with other effectors, to enhance neutrophil recruitment and responses. CCL4 blockage led to a slight decrease in the percentage of extravascular (tissue) neutrophils among the total neutrophils present in the lung (Supplementary Fig. 7f), thus it is possible that vILC3-derived CCL4 has little role in promoting trans-endothelial migration of neutrophils from the capillaries into the parenchyma; instead, it recruits circulatory neutrophils and sustain them in the infected lungs. Other chemoattractant signals from cells residing in the tissue, alveolar macrophages in particular, likely promote neutrophil trans-endothelial migration into the alveoli through their production of chemokines such as CXCL1^80,81. Therefore, vILC3s may act in conjunction with other immune cells that are resident in the alveolar-capillary beds, forming part of a recruitment network facilitating pathogen clearance.

Lymphocytes are critical factors in host defense against P. aeruginosa^82,83. Among these, ILC3s have been demonstrated to have an important role in protection against P. aeruginosa, as well as other airway pathogens³², through the production of IL-17 and IL-22. The majority of IL-17A produced during P. aeruginosa infection comes from ILC3s and prevents the establishment of chronic infection⁸⁴. IL-17 plays a critical role in orchestrating neutrophil expansion and recruitment, and inducing a variety of antimicrobial peptides and defensins^85–88. IL-17 receptor deficiency leads to a reduction in blood neutrophil numbers but no effect on bronchoalveolar lavage neutrophils⁸⁴, suggesting that IL-17 has less involvement in neutrophil transendothelial migration. This echoes our findings that vILC3-derived CCL4 enhances neutrophil recruitment from the circulation rather than promoting transendothelial migration.

ILC3s infiltrate the mouse lung during the first week of life⁸⁹. These newly arrived ILC3s have been shown to be indispensable for defense against bacterial pneumonia in newborn mice⁹⁰. We observed that the frequency of vILC3s was reduced in older mice, at 20 weeks of age. In humans, aging is a major factor associated with an increase in the prevalence and risk of pulmonary infections⁹¹. Whether there is a correlation between a decrease in vILC3s with aging and an increased risk of infections will be an interesting future area of investigation.

It is critical to prevent the transition to chronic persistent infection and the establishment of biofilms by P. aeruginosa. This change from acute to chronic infection is highly significant for patients with cystic fibrosis (CF)⁹². Individuals with CF are highly susceptible to P. aeruginosa infection and this leads to an accelerated decline in lung function. In the context of CF, excessive neutrophil influx in the lungs without the capacity to clear infection is a pathological hallmark of this disease^93–95. IL-22 is associated with reducing neutrophil influx during infection, and IL-22 neutralization led to an increase in neutrophil recruitment and tissue damage⁹⁶. Therefore, although vILC3s prompt neutrophil recruitment at the early stage of acute P. aeruginosa infection, IL-22-expressing vILC3s and tILC3s could be potential therapeutic targets to limit pathologic neutrophilia in CF.

A limitation of the work presented here is the current lack of an ILC3-specific knockout mouse line. Recently a RORγ^STOP/CD4 mouse has been developed and, in combination with bone marrow chimera, this line can perverse Th17 cells and lymph node structures⁹⁷, but still lacks many other RORγt-expressing populations besides ILC3s. We considered adoptive transfer experiments, however these would require sorting large numbers of vILC3s from the lungs of mice, which is technically challenging given the low numbers of vILC3s. It is also a concern whether the transferred vILC3s can be re-constituted in the same intravascular space, given that ILC3s are populated into tissues in a specific niche window during early development. In the future, generating vILC3 knockout and vILC3-specific CCL4 knockout genetic mouse lines will be helpful for further understanding the physiological roles of this cell population.

Methods

Mice

C57BL/6 (B6), B6.SJL and Rag1^–/– mice were obtained from Taconic. Tcrb^–/– and muMt^–/– mice were obtained from Jackson Laboratory. Gata3^fl/fl mice⁹⁸ provided by J. Zhu at NIAID and Klrg1^Cre mice⁹⁹ provided by R. Flavell at Yale University were bred in the lab to generate Gata3^fl/flKlrg1^Cre/+(ILC2-deficient). RORγt-GFP reporter mice¹⁰⁰ were provided by I. Ivanov at Columbia University. The homozygous of RORγt-GFP mice were RORγt-deficient (Rorc^–/–).

Both male and female mice at the age of 8–16 weeks old were used. We did not observe any differences between males and females. Mice were bred and maintained in designated Columbia University animal facilities in specific pathogen-free conditions. Control and experimental animals were co-housed under a 12-h light-dark cycle, with ambient temperature of 20–26 degree and humidity of 30–70%. All animal experiments were performed under the approval by the Institute Animal Care and Use Committee (IACUC) of Columbia University.

In vivo labeling of intravascular leukocytes

2 µg PE-conjugated anti-CD45.2 (clone 104) or anti-Thy1.2 (clone 30-H12) antibody was administered into mice under anesthesia via retro-orbital injection. Five minutes later, mice were euthanized by cervical dislocation and the lung leukocytes were isolated for further flow cytometry analysis.

Isolation of leukocytes

Lung tissues were collected after perfusion, disrupted into small pieces and digested for 20 min at 37 °C with Liberase (Merck, Cat #5401119001) plus DNase I (Merck, Cat #4536282001). Tissue pieces were strained into single cell suspension followed by treatment with ACK lysis buffer (Life Technologies, Cat #A1049201). Single cells were kept in PBS (Corning, Cat #46-013-CM) solution with 3% fetal bovine serum (FBS) (ThermoFisher, Cat #A5670701) and used for further analysis.

To isolate leukocytes in the blood, the blood was collected into a tube containing 50 μl 500 mM EDTA (ThermoFisher, Cat #AM9262). Blood-EDTA was added to 5 mL ACK lysis buffer and incubated at room temperature for 5 min, then centrifuged and the supernatant was removed. Cell pellets were resuspended in PBS with 3% FBS for further analysis.

Small intestines were collected by emptying the fecal contents. Peyer’s patches were removed, and small intestines were opened longitudinally, cut into small pieces and then shaken at 37 °C for 40 min in PBS media containing 10% FBS, 5 mM EDTA and 1 mM dithiothreitol (ThermoFisher, Cat #R0861) to dissociate intraepithelial leukocytes. The remaining fragments were washed twice with PBS and digested at 37 °C for 45 min in HBSS media containing Liberase and DNase I. The digested tissues were strained to yield a single-cell suspension, which was followed by treatment with ACK solution.

Parabiosis

CD45.1-expressing B6.SJL mice and CD45.2-expressing C57BL/6 mice of the same sex and age were co-housed for two weeks prior to the procedure to establish harmonious cohabitation. At the time of procedure, mice were anesthetized to full relaxation. Mice were then placed back to back, one partner to have the procedure done on its left side and the other on the right side. The surgical areas of the two mice were shaved and hair removal cream was applied to remove all hair. The surgical area was sterilized by alternating iodine ethanol wipes. The mice had matching skin incisions made from the elbow to the knee and the subcutaneous fascia was separated from the skin of each mouse, creating approximately 0.5 cm of free skin along both sides of the whole incision. The forelimbs, and rear limbs, of the parabionts were joined together at the elbow using a non-absorbable 3-0 silk suture, respectively. The skin of the two mice were then joined by stitching with an absorbable 4-0 polyglactin suture ventrally from the knee to the elbow. Analgesic was administered right after surgery for pain control and repeated post operation as needed. Parabiotic mice were maintained under antibiotic-containing water (10⁶ U/liter polymyxin B sulfate and 1.1 g/L neomycin sulfate) for two weeks after surgery and then returned to regular drinking water. Cytokine injections were performed on parabiotic pairs one month after the surgery. Prior to euthanasia, 2 µg PE-conjugated anti-Thy1.2 (clone 30-H12) antibody was administered into mice via retro-orbital injection. Five min later, mice were euthanized and the lung leukocytes were isolated for further flow cytometry analysis. Intravascular cells were identified as ivThy1.2-positive.

Pseudomonas aeruginosa culture and infection

Pseudomonas aeruginosa (P.a.) Schroeter Migula (ATCC #10145) master stocks were prepared following manufacturer’s instructions and aliquoted for storage at −80 °C. For infection, bacteria from frozen aliquots were grown in nutrient broth (Sigma-Aldrich, Cat. #N7519-250G) overnight at 37 °C with shaking and harvested in log phase at an optical density at 600 nm (OD600) of 0.2 (OD600 = 0.2; 2 × 10⁷ CFU/mL). Bacteria were centrifuged for 10 min, rinsed and resuspended in sterile PBS containing 20% glycerol and stored at −80 °C.

Prior to infection, P.a. aliquots were thawed at 37 °C and the infection was performed within 1 hour of thawing. Under anesthesia, mice were given 40 µl P.a. (8 × 10⁵ CFU) via intranasal (i.n.) administration. Bodyweight and general condition of mice was monitored daily over a 7-day period.

Bacterial burden in the lung

Lungs were collected and homogenized in 1 ml of sterile PBS and then used for quantitative culture on Nutrient agar (Sigma-Aldrich, Cat. #Y1500-250G) at 37 °C for 24 h. Serial dilutions were performed and viable counts after 24 h of incubation were expressed as the mean ± SD Colony Forming Unit (CFU)/ml.

In vivo cytokine treatment and chemokine neutralization

Recombinant IL-1β and IL-23 were from Peprotech and were injected intraperitoneally (i.p.) into mice daily for up to 7 days. To block CCL4 or CCL5, 40 μg anti-CCL4 (ThermoFisher, Cat. #46907) or 30 μg anti-CCL5 (ThermoFisher, Cat. #21418) antibody was retro-orbitally injected into each mouse at 24 h and 48 hours post P.a. infection. The same amount of Rat IgG2a (ThermoFisher, Cat. #02-9688, isotype control for anti-CCL4) or Mouse IgG1 (ThermoFisher, Cat. #02-6100, isotype control for anti-CCL5) was administered into control groups.

Analysis of ex vivo cytokine production

Cells from the lungs were cultured in a CO₂ incubator at 37 °C in RPMI with 10% FBS containing IL-2, IL-6, IL-1β, IL-23, and IL-12 (Perprotech) for 1 h. PMA (50 ng/ml; Sigma, Cat # P1585-1MG), ionomycin (500 ng/ml; Sigma, Cat # I3909-1ML) and Golgi Stop (BD Biosciences, Cat #554724) were then added into the culture medium for an additional 4 h incubation. For an unstimulated control group, only Golgi Stop was added for an additional 4 h incubation. Cells were subjected to intracellular staining of IFN-γ, IL-17A, and IL-22 and subsequent flow cytometry analysis.

Flow cytometry and cell sorting

Isolated leukocytes in PBS with 3% FBS were blocked with anti-CD16/CD32 (2.4G2, Harlan Laboratories) and incubated for cell surface staining with fluorochrome-conjugated antibodies with LIVE/DEAD fixable dead cell stain. Cell surface staining was performed at room temperature with shaking and washing performed at 4 °C. For transcription factor and cytokine staining, cells were fixed, permeabilized and stained using Foxp3/Transcription Factor Staining Buffer Kit (eBioscience, Cat #00-5523-00), and cells were stained overnight for intracellular markers in permeabilization buffer at 4 °C.

Fluorochrome-conjugated antibodies used for intracellular and cell surface staining can be found in Supplementary Table 1. The LIVE/DEAD fixable dead cell stain (dilution 1:1000) was from Life Technologies. CountBright Absolute Counting Beads (Life Technologies) were added into cell suspensions before FACS analysis. Cells were analyzed on an LSRII or LSR Fortessa flow cytometer (BD Biosciences), and data were analyzed with FlowJo software. Lineage antibody cocktail included CD3, CD5, CD19, B220, CD11b, CD11c, TCRγδ, Gr-1, FcεR1, CD8α, DX5, and TER119.

For cell sorting, cells were stained and washed in PBS solution with 10% FBS at 4 °C. vILC3s were sorted as live Lin⁻ Thy1⁺ NK1.1⁻ RORγt-GFP⁺ ivCD45.2⁺ and tILC3s were sorted as live Lin⁻ Thy1⁺ NK1.1⁻ RORγt-GFP⁺ ivCD45.2⁻. Cells were sorted on a FACS Aria cell sorter (BD Biosciences).

Ce3D imaging

Fluorochrome-conjugated antibodies used for Ce3D imaging can be found in Supplementary Table 2. For Ce3D imaging, RORγt-GFP reporter mice were i.v. injected with CD45 antibody and then sacrificed 5 min later. The lungs were collected and fixed with BD CytoFix/CytoPerm (BD Biosciences, Cat. #554714) diluted in PBS solution (1:4) for 24 h at 4 °C. Following fixation, all tissues were washed briefly (5 min per wash) in PBS and incubated in 30% sucrose for 24 hours at 4 °C before embedding in OCT compound (Sakura Finetek, Cat. #4583). Frozen samples were sectioned at 200 μm thickness with a CM1950 cryostat (Leica Biosystems). The samples were hydrated and washed with PBS to remove OCT. Samples were incubated for at least 12 h in BD Perm/Wash Buffer (BD Bioscience, Cat #554723) containing 1% mouse Fc block (BD Bioscience, Cat. #553142) and stained with titrated antibodies in BD Perm/Wash Buffer (BD Bioscience, Cat. #554723) containing 1% Fc block for 24 hours at room temperature on a shaker. Stained samples were washed with BD Perm/Wash Buffer three times for at least 20 min at room temperature on a shaker, fixed with 1% PFA for 10 min, and further washed with PBS for 20 min. Samples were transferred on a slide with two silicon isolators (Grace BioLabs, Cat. #664407) and treated with 200 μl of Ce3D medium^36,101 with modified recipe (1.82 g Histodenz (Millipore Sigma, Cat. #D2158-100G) per 1 ml 40% N-methylacetamide (Millipore Sigma, Cat. #M26305–500G) in PBS) inside a chemical fume hood and sealed with a coverslip (Electron Microscopy Sciences, Cat. #63766-01) and incubated at room temperature on a shaker overnight. After removing the old Ce3D medium, cleared samples were mounted with 40 μl of new Ce3D medium and sealed with a coverslip with two SecureSeal Imaging Spacers (Grace Bio-Labs, Cat. #654002). Image acquisition was performed at a voxel size of 0.285 × 0.285 × 1.5 µm, with a total imaging volume of 4225 × 3950 × 150 µm (XYZ). Images were acquired using an inverted Leica SP8 confocal microscope equipped with a 20× (NA 0.75) or 40× (NA 1.3) oil objective with 4 HyD and 1 PMT detectors and a white light laser that produces a continuous spectral output between 470 to 670 nm as well as 405, 685, and 730 nm lasers.

Image processing and analysis

Images were converted to.ims files using Imaris File Converter (Oxford Instruments). The Channel Arithmetics function of Imaris was used to remove spectral spillover and autofluorescence. Gaussian filters were applied to denoise the images. Segmentation for ILC3s was performed in Imaris (version 10.0.0) using the Surface module to define individual cells. Segmentation artifacts were excluded by volume thresholds and manual correction. Intensity filters for CD3 and ivCD45, based on the mean intensity within the segmented surfaces, were used to classify tILC3s and vILC3s. The Spot module was used to place seed points on the αSMA image for use in the shortest distance calculation between each ILC3 population and airways or vessels. Airways and vessels were determined by the morphology of the smooth muscle cells, and the spots were manually partitioned into airway smooth muscle cells and vascular smooth muscle cells³⁷. The x, y, and z coordinates of the centroid of segmented tILC3s, vILC3s, airway smooth muscle cells, and vascular smooth muscle cells were exported and used for downstream distance analysis. All nearest neighbor distances were calculated in 3D space using numpy.min in Python 3.11.5.

Lung physiology measurements

To assess lung physiology, mice were anesthetized under isoflurane (1.5–2% v/v). Mice were i.p. injected with Propoflo 28 (75 mg/kg, Zoetis, Cat# PRO-00312R2) for cannulation of the trachea with blunted 18-G cannula. Mice were then connected to a SciReq flexiVent FX small rodent ventilator, setup to deliver oxygen and isoflurane (1–1.5% v/v). The ventilator was programmed to maintain a positive end expiratory pressure (PEEP) of 3 cmH₂O and ventilation was initiated at 150 breaths/min, with tidal volume 10 mL/kg. Mice were paralyzed with i.p. injection of pancuronium bromide (0.8 mg/kg, Merck, Cat #P1918-10MG). The animals were stabilized for 10 min of regular ventilation and a standard lung volume history established by delivering two total lung capacity (TLC) maneuvers to a pressure limit of 27 cmH₂O and holding for three seconds. Next, two sets of baseline measurements were collected with Snapshot and Quickprime perturbations in addition to PV-loops. Data were fit to the single-compartment model to provide values for compliance.

Histology

Hematoxylin and eosin (H&E) images were obtained with Leica AT2 digital scanning system. At least 10 regions of interest (ROI) were randomly generated using ImageJ software with a published custom script¹⁰². The ROI images were evaluated for Lung injury score (LIS) according to the official ATS Workshop Report on Features and Measurements of Experimental Acute Lung Injury in Animals¹⁰³.

Transwell assay

Bone marrow cells were obtained by flushing tibias and femurs and straining cells into single cell suspension followed by treatment with ACK solution. Blood cells were obtained by bleeding mice and collecting blood into EDTA-containing tubes followed by treatment ACK solution. Single cell suspensions were collected into transwell media (HBSS containing Ca²⁺ and Mg²⁺ (ThermoFisher, Cat #24020117), 0.25% fatty acid-free low endotoxin BSA (Merck, Cat #A8806), and 15 mM HEPES–NaOH at pH 7.4 (37 °C)), and resuspended at 5 × 10⁶ cells/ml. Of these cell suspensions, 200 μl were added to the top of a transwell filter (polycarbonate, 3 μm pore; Millipore) inserted into a 24-well plate (Ultra-low attachment; Costar, Corning), which contained 300 μl of the above buffer with 50 ng/ml CCL4 (Perprotech). The plates were incubated in a cell incubator (37 °C, atmospheric CO₂, humidified). Two hours later, the filter inserts were removed, the medium in the lower chamber was collected and made up to 1 ml with HBSS. The cells recovered from each of the wells were analyzed by FACS. Neutrophils were gated as Ly-6G⁺ CD11b⁺ CD11c⁻ Siglec-F⁻.

For transwell assays testing the effects of vILC3-derived CCL4 on neutrophil chemotaxis, the above outlined procedure was followed with the exception that: sorted vILC3s (5 × 10³ cells/chamber) were added to the lower chamber in transwell media containing either 100 ng/ml anti-CCL4 (ThermoFisher, Cat. #46907) or 100 ng/ml isotype control for anti-CCL4 (Rat IgG2a: ThermoFisher, Cat. #02-9688).

RNA isolation and RT-qPCR

Sorted tILC3s and vILC3s were disrupted in TRIZol reagent (ThermoFisher, Cat #15596026), and total RNA was purified according to the manufacturer’s protocol. Reverse transcription was performed by using a ProtoScript II First-Strand cDNA synthesis kit (New England Biolabs, Cat #E6560S) and oligo(dT)20 primers. qPCR was performed using PowerUp SYBR Green Master Mix on a QuantStudio 3 system (Applied Biosystems). Gapdh was used to normalize the RNA content of the samples. Primer sequences for Gapdh and Ccl4 are as follows (5’-3’): Gapdh forward: GGGGTCCCAGCTTAGGTTC, Gapdh reverse: TTCACACCGACCTTCACCATT, Ccl4 forward: CCCACTTCCTGCTGTTTCTCCGCG, and Ccl4 reverse: GAGGAGGCCTCTCCTGAAGT.

Bulk RNA sequencing

Total RNA was isolated from sorted vILC3s and tILC3s from RORγt-GFP mice by using an miRNeasy Micro kit (Qiagen). RNAs were reverse transcribed, amplified and indexed according to the protocol of NEBNext Single Cell/Low Input RNA Library Prep kit for Illumina (Cat. # E6420S).

The reads of bulk RNA-seq were aligned to mm10 mouse genome with GENCODE M25 gene annotation by STAR¹⁰⁴ to get the read count of each gene. These read counts were normalized to “count per 10000”. DESeq2 was used to evaluate the significance of gene expression difference between groups. Differential genes were defined as >2-fold increase/decrease and adjusted p-value < 0.05. Metascape¹⁰⁵ analysis was done on differentially expressed genes and generate the GO enrichment networks. Top 17 clusters were shown in the network figure. The read counts for scRNA-seq were extracted from GSE192890¹⁰⁶. Genes with less than 50 total read counts in every sample were filtered out. Cells that had more than 10% reads mapped to mitochondrial genes were removed. The ribosomal and red cell genes were removed. The resulting read counts were processed by Seurat¹⁰⁷ and followed by principal component analysis (30 PCs) and Harmony¹⁰⁸ projection. Uniform manifold approximation and projection (UMAP) dimension reduction were performed on the 30 Harmony dimensions. The Louvain algorithm was used to detect cell clusters (resolution=0.1). These clusters were subsequently annotated based on canonical marker genes. The read counts of scRNA-seq were normalized to “count per 10000” which is consistent with the normalization of read counts of bulk RNA-seq for comparison. A typical dot plot made by Seurat package shows the percent of cells expressing a specific gene with dot size and the average expression level with color gradients. However, there is no standard way to infer the percentage of cells expressing a specific gene in bulk RNA-seq. To compare gene expressions from single-cell and bulk RNAseq, a modified R script based on the DotPlot function in Seurat was generated to make the dot plot where both size and color of the dot indicate the average expression level of selected genes in each cluster from scRNA-seq and in each sample from bulk RNA-seq.

Statistical analysis

GraphPad Prism 9 software was used for statistical analysis. Statistical difference was determined via an unpaired two-tailed Student’s t-test or Mann-Whitney U test for pairwise comparisons. One-way analysis of variance followed by Dunnett’s multiple comparisons test or Tukey’s multiple comparison test, and two-way ANOVA with Bonferroni’s multiple comparisons test were used for comparison between more than two groups. P-values of ≤0.05 were considered to represent means with a statistically significant difference.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

S.S. designed, performed, interpreted the majority of the experiments and drafted the manuscript. H.I. performed and interpreted 3D imaging experiments. V.D., C.W., R.Z., Y.Z., M.C., and S.P. assisted with experiments. J.J., W.Z., Y.F., and U.B. performed RNA-seq data analysis and interpretation. J.Q. interpreted the data and drafted the manuscript. Y.H. designed the experiments, interpreted the data and finalized the manuscript. All authors contributed to the discussion of experimental findings and preparation of the manuscript.

Peer review

Peer review information

Nature Communications thanks Marco Colonna and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

RNA-Seq data are available in the Gene Expression Omnibus database (accession number GSE274097. All data are included in the manuscript and supplementary information or available from the authors, as are unique reagents used in this article. The raw numbers for charts and graphs are available in the Source data file. Other data are available upon request. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-58982-1.