Single-cell analysis of recruited ILC2 cell fate during transition from circulation to establishment of tissue residency

TL;DR

This study explores how circulating ILC2 cells adapt and become lung-resident during infection, revealing their unique transition and functional contributions.

Contribution

The study reveals the gene expression changes and functional adaptations of circulating ILC2s as they transition to lung residency during infection.

Findings

Circulating ILC2s quickly adopt tissue-resident gene expression upon entering the lung.

Converted ILC2s retain some intestinal-related gene expression linked to enhanced functionality.

The transition to lung residency occurs mainly as cells enter the lung parenchyma from the vasculature.

Abstract

Circulating group 2 innate lymphoid cells (ILC2s) often serve as a first line of defense against infection prior to local expansion of lung-resident ILC2s. The fate of circulating ILC2s and their relationship with lung-resident ILC2s is not well understood. Using reporter mice in combination with single-cell RNA sequencing (scRNA-seq) and Cellular Indexing of Transcriptomes and Epitopes sequencing (CITE-seq), the fate of circulatory ILC2s was followed during the course of primary Nippostrongylus brasiliensis helminth infection. Circulating ILC2s rapidly acquire tissue-resident gene expression upon arrival to the lung. This transition occurs primarily as the cells enter the parenchyma from the vasculature. Despite acquiring a lung-resident phenotype by the peak of the immune response, these converted ILC2s retain some unique gene expression related to their intestinal origins which correlate with enhanced functionality. Findings provide insight into the tissue adaptation of circulatory ILC2s during recruitment to the lung and establish their contribution to the lung-resident ILC2 population.

Figures

Migratory iILC2s upregulate arginase as they enter the parenchyma and in a time-dependent fashion. (A) Experimental design. Mice were injected subcutaneously with Nb. On D5, D6, or D7 postinfection, intravenous labeling with anti-CD45-PE/Cy7 was performed. Lungs were harvested and stained for flow cytometry. (B) Representative plots and quantification of ILC2 populations (nILC2s and iILC2s) over time course. ILC2s are defined as CD45+ Lineage− KLRG1+. Mean ± SEM; n = 16 mice per time point, pooled from 4 independent experiments. One-way analysis of variance by Brown-Forsythe test with Dunnett’s T3 multiple comparisons test. (C) Representative plots and quantification of vascular labeling of nILC2s and iILC2s over time course. Mean ± SEM; n = 14 mice per time point pooled from 4 independent experiments. 2-way analysis of variance with Geisser-Greenhouse correction and uncorrected Fisher’s LSD, with individual variances computed for each comparison. (D) Representative plots and quantification of Arg1 expression in nILC2s, vascular iILC2s, and parenchymal iILC2s over time course. Mean ± SEM; n = 14 mice per time point pooled from 4 independent experiments. Two-way analysis of variance with uncorrected Fisher’s LSD using a single pooled variance. P values defined as follows: *P < 0.05; P < 0.01; *P < 0.005; P < 0.0001. Ab, antibody; ns, not significant.

Migratory iILC2s acquire a tissue-resident nILC2 transcriptome and phenotype as they enter the parenchyma and in a time-dependent fashion. (A) Experimental model. Mice were injected subcutaneously with Nb. Five or 7 d later, mice were given intravenous injection of anti-CD45-PE antibody and euthanized 5 min later. Lungs were harvested and stained for flow sorting. Total ILC2s were sorted and sequenced by scRNA-seq and CITE-seq. ILC2s are defined as CD45+ Lineage− KLRG1+ CD127+. Two mice per cohort were indexed by hashing and pooled for sequencing per experimental repeat (cohorts = D5, D7). Experimental protocol was repeated on 2 consecutive days. Total n = 4 mice per cohort pooled from 2 experiments. (B) UMAP of unsupervised clustering of cells by Seurat. (C) Clustering in panel B shown with metadata overlay for vascular labelling (green = parenchymal; red = vascular). (D) Expression of indicated ILC2 transcripts within D5 time point on UMAP. (E) Expression of indicated ILC2 transcripts within D7 time point on UMAP.

Migratory iILC2s acquire a tissue-resident nILC2 transcriptome and phenotype over the course of helminth infection. (A) Clustering shown with metadata overlay for both vascular labeling and time point (yellow = D5 vascular; orange = D5 parenchymal; purple = D7 vascular; blue = D7 parenchymal). (B) Heatmap of top 30 genes upregulated in each compartment, compared with all other compartments (compartments = D5 vascular, D5 parenchymal, D7 vascular, D7 parenchymal). Asterisks mark transcripts in panels C to E. (C) Dot plot of ILC2 transcript and protein expression (assayed by CITE-seq) across compartments. Blue indicates nILC2-associated genes, red indicates iILC2-associated genes. (D) Dot plot of cytokine transcript expression across compartments. (E) Dot plot of transcript expression for transcription factors whose transcripts are statistically significantly upregulated or downregulated across compartments, defined as false discovery rate–adjusted P value <0.05. Statistical testing was by Seurat v4.0.2 defaults for the FindMarkers() function (Wilcoxon rank sum test, followed by Bonferroni correction for multiple testing).

Migratory iILC2s stably upregulate arginase, whereas tissue-resident nILC2s maintain a steady phenotype. (A) Experimental model. Donor CD45.2 mice (Arg1YFP reporter positive) were injected subcutaneously with Nb. iILC2s and nILC2s were sorted from the lung 5 d later. Donor iILC2s or donor nILC2s were transferred into recipient CD45.1 mice at D3 of infection. Lungs were harvested either 3 d later or 10 d later and ILC2s characterized by flow cytometry. (B) Representative contour plots and quantification of Arg1/YFP expression of donor nILC2s and donor iILC2s. Mean ± SD, n = 3 mice per group per time point, pooled from 3 independent experiments with 1 biological replicate per cohort (nILC2 recipient, iILC2 recipient). Five total biological replicates for “before transfer” control cohort. One-way analysis of variance with Tukey’s multiple comparisons. P values defined as *P = 0.0009 and P < 0.0001. ns, not significant.

Migratory iILC2s acquire a tissue-resident nILC2 transcriptome and phenotype. (A) Experimental Model: Donor Arg1YFP CD45.2 mice were given rIL-25 intraperitoneally daily for 3 d. On the fourth day, iILC2s were sorted from the lungs and MLNs and transferred into congenic C57BL/6 5.1 recipient mice infected 3 d prior with Nb. Lungs of both recipient and untreated C57BL/6 5.1 mice were harvested 10 d later and ILC2s sorted and sequenced. Lung iILC2s and MLN iILC2s from IL-25–treated Arg1YFP CD45.2 (donor genotype) mice were also sorted and sequenced at the same time. (B) UMAP of unsupervised clustering of cells by Seurat. (C) Dot plot showing expression of nILC2- and iILC2-defining transcripts across Seurat clusters. Blue indicates nILC2-associated genes, red indicates iILC2-associated genes. (D) Expression of indicated nILC2- and iILC2-defining transcripts and proteins shown on UMAP. Blue indicates nILC2-associated genes, red indicates iILC2-associated genes. (E) Clustering of ILC2 populations by Seurat (same as panel B) encoded with known metadata of cell origin. (F) Trajectory analysis, with the starting point denoted with a star. Donor cells were pooled from 14 donor mice and split equally between 2 recipient mice per experiment. Two mice per cohort were indexed by hashing and pooled for sequencing per experiment (cohorts = MLN iILC2s, lung iILC2s, recipient, naïve). The experimental protocol was repeated on 2 consecutive days. Total n = 4 mice per cohort, pooled from 2 experiments.

iILC2-to-nILC2 conversion. Former circulating iILC2s resemble native, lung-resident nILC2s while retaining some distinct functional potential based on gene expression. (A) Breakout of UMAP by known metadata of cell origin. (B) Breakouts of UMAP by cell origin showing cells satisfying iILC2-qualifying combined marker (eigengene) expression and nILC2-qualifying combined marker (eigengene) expression. (C) Heatmap of top 30 genes upregulated in each cell type, compared with all other cell types. Asterisks mark transcripts in panels D to H. (D) Violin plots of gene expression and protein expression (assayed by CITE-seq) for iILC2- and nILC2-defining targets. No difference (nd) was defined as fold change (FC) (base 10) < 1.2, or false discovery rate–adjusted P value >0.05. Statistical testing was by Seurat v4.0.2 defaults for the FindMarkers() function (Wilcoxon rank sum test, followed by Bonferroni correction for multiple testing). (E) Dot plot of transcript expression for functional factors expressed by ILC2s across cell types in the congenic transfer model. (F) Dot plot of transcript expression of transcription factors responsible for regulating factors shown in panel E. (G) Dot plot of transcript expression for transcription factors whose transcripts are statistically significantly upregulated or downregulated across cell types, defined as false discovery rate–adjusted P value <0.05. Statistical testing was by Seurat v4.0.2 defaults for the FindMarkers() function (Wilcoxon rank sum test, followed by Bonferroni correction for multiple testing). (H) Dot plot of transcripts associated with migration and tissue residency across cell types.

Full text

Introduction

Group 2 innate lymphoid cells (ILC2s) act as early orchestrators of type 2 inflammation in the context of mucosal barrier immunity.1–3 While ILC2s are found in many tissues, these innate immune cells are enriched in mucosal sites.3^,^4 ILC2s express GATA3 and rapidly produce type 2 cytokines IL-4, IL-5, and IL-13 in response to tissue alarmins IL-25, IL-33, and TSLP.2^,^5–11 Depending on the tissue in which they reside, ILC2s can exhibit preferential responsiveness to different alarmins.12 For example, intestinal-resident ILC2s are poised to respond to IL-25 released by epithelial tuft cells during homeostasis, while lung-resident ILC2s preferentially respond to stromal-derived IL-33.13–16 Both IL-25–responsive ILC2s (termed “inflammatory” ILC2s [iILC2s]) and IL-33–responsive ILC2s (termed “natural” ILC2s [nILC2s]) work together to establish early type 2 immune responses in the lung and intestine after infection by intestinal helminths.12

In the context of Nippostrongylus brasiliensis (Nb) helminth infection, circulatory iILC2s and lung-resident nILC2s respond in 2 distinct waves.17 Prior work has shown that circulatory iILC2s represent the first wave arriving from the intestine within 4 to 5 d of Nb infection.12^,^18 These studies reveal that circulating ILC2s from the intestine are the first ILC2s to generate type 2 cytokines in the lung and comprise 30% to 40% of the total pulmonary ILC2 pool at the time of their arrival at 4 to 5 d of infection.12^,^18^,^19 A careful accounting of iILC2 and nILC2 numbers in the lung and other tissues between day 0 (D0) and D5 of Nb infection and blocking of cell egress and resultant immigration to the lung through the use of FTY720 confirmed that the appearance of this first wave of iILC2s is via circulation and not a result of an early alteration in phenotype or expansion of lung-resident nILC2s.18^,^19 In contrast, the second wave of responding ILC2s comprises almost entirely of cells with a lung-resident, nILC2 phenotype.12 Lung-resident nILC2 expansion begins a week after initial infection, peaking around 2 wk.12 This second wave has largely been attributed to the expansion of lung-resident nILC2 populations, with some de novo seeding by non–lung-resident ILC2s and their progenitors.4^,^12^,^20 With respect to the dynamics of these 2 waves, circulatory iILC2 numbers peak prior to nILC2 activation and decline as lung-resident nILC2 expansion takes hold.12

By the peak of the lung-resident nILC2 response to Nb infection, circulating iILC2s are absent from the lung.4^,^12 While studies using parabiosis, fate labeling, in vivo photoactivation, and FTY720 blockade have cemented the circulatory nature of IL-25–responsive iILC2s, their origin, and their differential responsiveness to tissue alarmins, the ultimate fate of circulatory iILC2s and their relationship to tissue-resident nILC2s after arrival to the lung has remained less clear.18^,^19^,^21^,^22 Specifically, what contributes to the absence of circulatory iILC2s at the peak of the immune response in the lung? Do these cells die, exit, or instead alter their phenotype to blend in with the established lung-resident nILC2 pool as the immune response proceeds? In support of this latter possibility, in vitro studies have shown that, when cultured with IL-2, IL-7, and tissue alarmins, iILC2s lose expression of the IL-25 receptor IL17RB and acquire expression of the IL-33 receptor ST2.12 Transition from an IL-25–responsive population to an IL-33–responsive population was also observed in vivo after transfer of iILC2s into Nb-infected Rag2^−/−^Il2rγ^−/−^ mice, which lack both adaptive immune cells and ILCs.12 Together, these findings support the concept that the transient nature of circulating iILC2s could result from an alteration in phenotype as opposed to their exit or death, at least under certain conditions. However, a systematic investigation surrounding the transition of circulatory ILC2s into lung-resident ILC2s is lacking.

The present study had 3 objectives. First, to understand the transient nature of circulating iILC2s in the lung, it is important to determine the extent that these cells undergo a phenotypic conversion toward lung-resident nILC2s. Second, if conversion is occurring, when and where is it taking place in the lung? Last, to fully understand the process of conversion toward a lung-resident cell, we needed to explore the gene programs that define this phenotypic transition at single-cell resolution over time. To meet these objectives, we leveraged single-cell sequencing in conjunction with intravenous labeling and cell transfer models to identify and characterize the phenotypic conversion in place and time. Specifically, the dynamics and phenotype of the pulmonary iILC2 compartment was assessed from the start of the first wave of circulatory iILC2s reaching the lung, through the second wave of expansion by lung-resident nILC2s. The data show that circulatory iILC2s begin to acquire a lung-resident nILC2 phenotype shortly after arriving in the lungs. Adaptation to a tissue-resident phenotype continues with time. Importantly, some features of their original intestinal origin persist in the converted iILC2 population. Such features are consistent with certain subsets of memory ILC2s.23 In sum, the data help to define an important process whereby circulatory iILC2s from the intestine contribute to the lung-resident nILC2 niche post–primary helminth infection.

Methods

Mice

C57BL/6 CD45.2 (Stock #000664), C57BL/6 CD45.1 (Stock #002014), and YARG Arg1^YFP^ (Stock #015857) were purchased from the Jackson Laboratory and bred in-house.24 YARG mice were maintained on both CD45.1 and CD45.2 backgrounds. Mice were maintained in accordance with guidelines established by the Institutional Animal Care and Use Committee at National Jewish Health. Mice were 6 to 12 wk old at the start of experiments. Infected mice were housed under ASBL-2 conditions. Mice of both sexes were used in all experiments.

Nb infection

Nb was maintained according to previously published protocol.25 Mice were injected subcutaneously in the lower back with ∼500 L3 larvae in 200 µL sterile-filtered 0.9% saline.

IL-25 treatment

Donor mice were administered 300 ng recombinant IL-25 (rIL-25) (R&D Biosystems 1399-IL/CF) in 100 µL phosphate-buffered saline (PBS) by intraperitoneal injection, daily for 3 d (D0–D2). On the fourth day (D3), lungs and mesenteric lymph nodes (MLNs) were harvested to isolate ILC2s.

Intravenous labeling

Mice were administered 2 µg of anti-CD45 (clone 30F11; BioLegend) in 200 µL PBS by intravenous injection into the tail vein. Five minutes later, mice were euthanized and lungs were harvested for flow cytometry (fluorophore: PE/Cy7) or scRNA-seq and CITE-seq (fluorophore: PE).

Tissue harvest and preparation of single-cell suspension

Upon harvest, lungs were placed in RPMI 1640 (Corning 15-040-CV) on ice. Lungs were then digested in digestion media comprising of RPMI 1640 with DNAse I (200 µg/mL, Gold Bio D-300-100), Liberase TH (50 µg/mL; MilliporeSigma 5401135001), collagenase (250 µg/mL; Sigma-Aldrich C7657), and hyaluronidase (1 mg/mL; Sigma-Aldrich H3506). Lungs in digestion media were dissociated on the GentleMacs Octo Dissociator by Miltenyi Biotec using the 37C_m_LDK_1 program (37 °C for 31 min). Resultant single-cell suspension was filtered through a 70 µm filter into PBS. Next, erythrocyte load was reduced by resuspending in 3 mL ACK lysing buffer for 3 min, followed by quench with 10 mL PBS. Cells were then resuspended in 2% fetal calf serum (FCS) for further downstream processing.

Upon harvest, MLNs were placed in 2% FCS on ice. Lymph nodes were then collected into PBS via manual dissociation and filtration through a 70 µm filter using the flat end of a 3 mL syringe plunger. If necessary, erythrocyte load was reduced using the same method as described for lungs (resuspend in 3 mL ACK lysing buffer for 3 min followed by quench with 10 mL PBS). Cells were then resuspended in 2% FCS for further downstream processing.

Flow staining

Cells were blocked for 10 min on ice with TruStain FcX (BioLegend anti-mouse CD16/32 clone 93) and then stained with appropriate antibodies in 2% FCS. Lineage markers were as follows: CD3ε (145-2C11), CD4 (RM4-5), CD8a (53-6.7), CD11b (M1/70), CD11c (N418), FcεR1α (MAR-1), GR-1 (RB6-8C5), NK-1.1 (PK136), TCRβ (H57-597), TCRγ/δ (GL3), and Ter119 (TER-119). Additional antibodies were CD45 (30-F11), CD45.1 (A20), CD45.2 (104), KLRG1 (2F1/KLRG1), CD90.1 (OX-7), CD90.2 (30-H12), CD127 (A7R34), and biotinylated IL-17RB (752101). All antibodies were acquired from BioLegend except biotinylated IL-17RB, which was from Novus Biologics. For live/dead discrimination, cells were resuspended in DAPI at final concentration of either 150 nM or 300 nM. ILC2s were defined as CD45+ Lineage− KLRG1+, in combination with either CD90 expression or YFP expression (in YARG Arg1^YFP^) mice. For the intravenous labeling sequencing experiment, ILC2s were additionally defined as CD127+.

Flow cytometry

Cell counts were performed by adding 20 to 25 µL of count beads (Bangs Laboratories 580) per sample tube and scaling cell counts by fraction of loaded beads that were detected and fraction of organ in the sample tube. Flow cytometry was performed on BD Fortessa and BD LSR II (BD Biosciences). Flow analysis was performed with FlowJo v10 (TreeStar).

CITE-seq and hashtag staining for sequencing

For sequencing experiments, biological replicates were split across 2 consecutive days, with 2 biological samples per cohort processed each day. On each processing day, the biological replicates within each cohort were labeled using hashtag antibodies and pooled for downstream processing after sorting.

Sequencing antibodies were the following (all Biolegend TotalSeq-A chemistry set): anti-mouse Hashtag 1 (TotalSeq-A0301; catalog# 155801; clone: M1/42; 30F11), anti-mouse Hashtag 2 (TotalSeq-A0302; catalog# 155803; clone: M1/42; 30F11), anti-PE (TotalSeq-A0911; catalog# 2; clone: PE001), anti-mouse IL-33Rα (TotalSeq-A0837; catalog# 145317; clone: DIH9), streptavidin (TotalSeq-A0975; catalog# 405365; barcode sequence: TGCCAGCCCTTTGTA), and anti-human/mouse integrin β7 (ITGβ7) (TotalSeq-A0214; catalog# 321227; clone: FIB504). Per the manufacturer’s recommendations, CITE-seq and hashing antibodies were prepared at a concentration of 1.0 µg per 10^6^ cells in 100 µL, and streptavidin was prepared at a concentration of 0.06 µg per 10^6^ cells in 100 µL.

After staining for flow cytometry, samples were stained with CITE-seq and hashing antibodies per the manufacturer’s recommendations. Briefly, after flow staining, samples were washed and resuspended in master mix for sequencing: anti-PE, anti-human/mouse ITGβ7, anti-mouse IL-33Rα, and one of anti-mouse Hashtag 1 or Hashtag 2. For the congenic transfer experiment, master mix for sequencing also included streptavidin, with corresponding biotinylated anti-IL-17RB included in the flow staining master mix. Samples were incubated for 30 min at 4 °C in the dark, followed by washing (add 1 mL PBS, centrifuge) and resuspension in DAPI.

Cell sorting

When processing for cell transfers, MLNs were pooled directly upon harvest and lungs were pooled after dissociation, up to 8 mice per pooled tissue. When processing for sequencing, biological samples were kept separate throughout the entire staining and sorting process. To increase downstream ILC2 recovery, single-cell suspensions were enriched using the EasySep Mouse Hematopoietic Progenitor Cell Isolation Kit (Stemcell 19856), following the manufacturer’s instructions. Samples were then stained and sorted into complete RPMI 1640 media using the FACSAria III (BD Biosciences). Sorted cells were washed once (add PBS, centrifuge) and resuspended in appropriate buffers for downstream applications.

Adoptive cell transfer

Recipient mice received 200 µL of cell suspension via intravenous injection into the tail vein. Sorter counts were used to estimate the number of cells transferred. For transferring nILC2s versus iILC2s into different recipient mice, recipient mice received between 30,000 and 80,000 iILC2s and between 20,000 and 130,000 nILC2s, which were obtained from 3 to 4 donor mice (lungs) pooled. For congenic transfer in the sequencing experiment, 14 donor mice pooled (both lungs and MLNs) yielded ∼800,000 cells, which were split equally between two recipients, resulting in ∼400,000 donor iILC2s per recipient mouse.

Sequencing preparation

After ensuring sort quality on postsorts, individual samples were pooled equally by cohort based on sorter counts (50/50 split of Hashtag 1 and Hashtag 2 sample). For the congenic transfer experiment, recipient samples were sorted into a tube for cells identified as donor cells and a tube for cells identified as host endogenous cells. Biological replicates were then pooled such that all donor cells were retained, and endogenous cells were supplemented equally up to 20,000 cells total in the final pooled sample. These pooled-by-cohort samples were then spun down and resuspended in 1× PBS + 0.4% bovine serum albumin, counted, and loaded to construct emulsion droplets.

Cell capture, gene expression and ADT/HTO library preparation, and sequencing

Following manufacturer’s recommendations, the emulsion droplets were constructed using a 10x Genomics IX controller. Gene Expression and ADT/HTO library preparation was performed by combining the instructions in the TotalSeq A Dual Index Protocol and either 10x Genomics Chromium Single Cell 3′ Reagent Kits V3.1 (intravenous labeling experiments) or 10x Genomics Chromium Single Cell 3′ Reagent Kits V4 (congenic transfer experiments). cDNA was generated for the gene expression and ADT/HTO libraries from the same captured cells. For the gene expression, quality control on the generated cDNA was performed on an Agilent 2100 Bioanalyzer using a High Sensitivity DNA Kit (Agilent Technologies). One quarter of the cDNA was used to build the gene expression libraries, and indexed using the 10x, Set TT A index plate. Gene expression libraries were pooled based on the projected number of cells captured and sequenced using an Illumina NextSeq 2000, aiming for 20,000 reads per cell. For the ADT/HTO libraries, one-quarter of the cDNA generated was used to build the ADT/HTO libraries and indexed using the TotalSeq-A indexes listed in the protocol. ADT/HTO libraries were pooled based on the number of cells captured and sequenced using an Illumina NextSeq 2000, aiming for 5,500 reads/cell. All libraries for a single experiment (both experimental days) were sequenced together in a single sequencing run.

Bioinformatic analysis

Reads were aligned to the refdata-gex-GRCm39-2024-A reference transcriptome using Cell Ranger v9.0.0 (10x Genomics). Downstream analysis was done using Seurat v4.0.2.

For the vascular labeling experiment, cells were filtered to retain those with at least 3,000 total transcripts (nCount_RNA > 3,000) and at least 2,000 unique transcripts (nFeature_RNA > 2,000). Cells with either metric above the 98th percentile for that metric were also filtered out. Further, cells with more than 1,000 reads on PE were also filtered out. Raw PE-CITE-seq reads were then used to determine vascular versus parenchymal classification. Cells with >100 PE reads were classified as vascularly labelled, while cells with 100 PE reads or less were classified as parenchymal.

For the congenic transfer experiment, cells were filtered to retain those with at least 2,500 total transcripts (nCount_RNA >2,500) and at least 1,500 unique transcripts (nFeature_RNA >1,500). Cells with either metric above the 98th percentile for that metric were also filtered out. Further, cells with more than 150 reads on PE were also filtered out. Raw PE-CITE-seq reads were then used to determine donor/endogenous cell classification for recipient samples. Cells with greater than 15 PE reads were classified as donor cells, while cells with 15 PE reads or less were classified as endogenous host cells.

In sum, individual samples were assigned metadata for the following: vascular/parenchymal (vascular labeling experiment), donor/endogenous (congenic transfer experiment), time point, replicate, organ, and treatment. Initial principal component analysis (PCA) and Uniform Manifold Approximation and Projection (UMAP) based clustering was examined to determine that biological replicates had good overlap and there was no batch effect, providing justification for merging data for downstream analysis.

Quality-controlled and metadata-assigned samples were merged across all biological samples for all conditions in an experiment using the Seurat merge function. This combined Seurat object was then log-normalized for RNA, following which we identified variable features (FindVariableFeatures) and scaled the data for all genes. Principal component analysis was performed on this postprocessed RNA data and dimensions for final clustering determined by titration along the inflection points of the resultant elbow plot. UMAP and clustering were then calculated using default Seurat settings (UMAP using normalized Laplacian + noise and nearest neighbor graph using SNN, Modularity Optimizer version 1.3.0, and Louvain algorithm). ADT data were normalized using center log ratio and normalizing per feature across cells [NormalizeData with assay = “ADTsubset” (excluding PE), method = “CLR”, margin = “2”]. These data were then scaled, to enable visualization on heatmaps. Anti-PE, already used to determine vascular/nonvascular or donor/endogenous, was excluded from this postprocessing.

For trajectory analysis, Monocle3 was used to visualize the lineage trajectory of ILC2 populations. To be compatible with Monocle3, the Seurat object was converted into a cell_data_set and dimensional reduction was performed by UMAP. The trajectory map was visualized using the order_cells function with the top of the MLN iILC2 cluster as the root node.

For eigengenes (combined expression of canonical markers), iILC2 and nILC2 expression was defined as follows: for a marker expressed in the subset, expression must be greater than or equal to the 10th percentile of samples comprised only of that subset (e.g. 10th percentile of MLN and lung iILC2s combined for KLRG1); for a marker that is not expressed is that subset, expression must be less than or equal to the 90th percentile of samples comprising only of that subset (e.g. 90th percentile of lung naïve nILC2s for KLRG1). Specifically, qualification requirements were as follows. To qualify as a cell of iILC2 identity, the cell must express—relative to baseline defined by MLN and lung iILC2s—above the 10th percentile for Klrg1 and Il17RB (protein), and below the 90th percentile for Arg1, Il1rl1, Thy1, and IL33Rα (protein). To qualify as a cell of nILC2 identity, the cell must express—relative to baseline defined by uninfected naïve lung nILC2s—above the 10th percentile for Arg1, Il1rl1, Thy1, and IL33Rα (protein), and below the 90th percentile for Klrg1 and IL17RB (protein).

Quantification and statistical analysis

Statistical calculations for the sequencing analyses were obtained from Seurat v4.0.2 as part of the standard output from the functions used to generate the corresponding plots. For flow cytometry experiments, statistical calculations were performed with Prism 10.1.0 (GraphPad Software). Data are presented as the mean ± SD from the specified number of mice unless otherwise noted. Comparison of data between two groups was analyzed using a 2-tailed unpaired t test to determine statistical significance. Significance across multiple groups was determined by 1- or 2-way analysis of variance with multiple correction, depending on experimental design. Statistically nonsignificant differences were defined relative to alpha = 0.05 and denoted with “ns.” If exact P value is not noted, P values were defined as follows: **P *< 0.05; ***P *< 0.01; ****P *< 0.005; *****P *< 0.0001.

Results

Circulatory ILC2s begin to acquire tissue-resident features shortly after arriving in the lung

While it has previously been shown that circulatory iILC2s arrive in the lung from the small intestine by D4 postinfection with Nb, we were interested in what happens to iILC2s in subsequent days.18 Using Arg1^YFP^ reporter mice, in which tissue-resident nILC2s are Arg1/YFP^high^ and circulating iILC2s are Arg1/YFP^low^, we characterized the dynamics of pulmonary ILC2s from D5 to D7.18^,^19^,^24^,^26^,^27 To identify cells in the vasculature versus in the parenchyma, we performed intravenous labeling with anti-CD45 antibody 5 min prior to euthanasia (Fig. 1A).28 At D5, iILC2s (red box: CD90^low^, KLRG1^high^) comprise on average ∼40% of the total ILC2 population, decreasing in proportion thereafter (Fig. 1B;  Fig. S1A), similar to other findings showing numbers and relative percentage of iILC2s over a comparable timeframe.12^,^18

*Migratory iILC2s upregulate arginase as they enter the parenchyma and in a time-dependent fashion. (A) Experimental design. Mice were injected subcutaneously with Nb. On D5, D6, or D7 postinfection, intravenous labeling with anti-CD45-PE/Cy7 was performed. Lungs were harvested and stained for flow cytometry. (B) Representative plots and quantification of ILC2 populations (nILC2s and iILC2s) over time course. ILC2s are defined as CD45+ Lineage− KLRG1+. Mean ± SEM; n = 16 mice per time point, pooled from 4 independent experiments. One-way analysis of variance by Brown-Forsythe test with Dunnett’s T3 multiple comparisons test. (C) Representative plots and quantification of vascular labeling of nILC2s and iILC2s over time course. Mean ± SEM; n = 14 mice per time point pooled from 4 independent experiments. 2-way analysis of variance with Geisser-Greenhouse correction and uncorrected Fisher’s LSD, with individual variances computed for each comparison. (D) Representative plots and quantification of Arg1 expression in nILC2s, vascular iILC2s, and parenchymal iILC2s over time course. Mean ± SEM; n = 14 mice per time point pooled from 4 independent experiments. Two-way analysis of variance with uncorrected Fisher’s LSD using a single pooled variance. P values defined as follows: *P < 0.05; **P < 0.01; ***P < 0.005; ***P < 0.0001. Ab, antibody; ns, not significant.

Taking intravenous labeling into consideration, nILC2s (blue box: CD90^high^, KLRG1^int^) were found exclusively within the parenchyma, which was consistent with tissue-residency, and circulatory iILC2s were present in both the vasculature and parenchyma across all time points. However, the percentage of iILC2s in each compartment changed between D5 and D7. While primarily in the vasculature at D5, iILC2s were primarily in the parenchyma by D7 (Fig. 1C).

Next, we wanted to investigate whether and where iILC2s begin to acquire a lung-resident phenotype. Using Arg1/YFP expression as a marker of a tissue-resident phenotype, 80% to 85% of lung-resident nILC2s were Arg1/YFP^high^ throughout the time course, consistent with their lung-residency (Fig. 1D, blue, top). In contrast, vascular and parenchymal iILC2s exhibited variable Arg1/YFP expression. With respect to vascular iILC2s, only 10% to 15% were Arg1/YFP^high^ at D5, increasing in percentage at D6 before plateauing at D7 (Fig. 1D, pink, middle). In contrast, ∼25% of parenchymal iILC2s were already Arg1/YFP^high^ at D5, with this percentage reaching ∼70% of the parenchymal iILC2 population by D7 (Fig. 1D, green, bottom). While 70% of parenchymal iILC2s acquiring a lung-resident phenotype is significant, this likely underestimates the full conversion rate, as we later show that conversion is also associated with an increase in CD90 expression, making converted iILC2s indistinguishable from nILC2s in this gating scheme. In sum, circulating iILC2s rapidly acquire tissue-resident attributes after arriving in the lung, with onset of this phenotype coinciding with entry into the lung parenchyma.

Circulatory ILC2s rapidly acquire a tissue-resident transcriptome as they enter the lung parenchyma

The reduction in CD90^low^ KLRG1^high^ iILC2s in the lung between D5 and D7 coincided with an increase in Arg1 expression in the same population. This indicated that circulating iILC2s begin to acquire markers of tissue-resident nILC2s. However, the extent to which this conversion occurs was unclear. To more rigorously assess the acquisition of a tissue-resident identity, we performed single-cell RNA sequencing (scRNA-seq) along with CITE-seq on pulmonary ILC2s. ILC2s (CD45+, Lineage−, CD127+, CD90+, KLRG1+) were sorted from the lung and sequenced at D5 and D7 of Nb infection (Fig. S2).29 As before, intravenous injection with anti-CD45 antibody (PE-conjugated) 5 min prior to sacrifice was used to differentiate vascular and parenchymal populations (Fig. 2A). Instead of using flow cytometry to differentiate vascular and parenchymal populations, we now leveraged CITE-seq. Specifically, CITE-seq enabled the bioinformatic identification of vascularly labelled cells via the binding of DNA-barcoded anti-PE secondary antibody (Fig. S3A). Using CITE-seq to differentiate vascular from parenchymal cells in combination with standard scRNA-seq, transcriptomes of these populations could be distinguished. CITE-seq was also used to assay protein expression of surface markers ITGβ7 and the IL-33 receptor IL33Rα/ST2.

Migratory iILC2s acquire a tissue-resident nILC2 transcriptome and phenotype as they enter the parenchyma and in a time-dependent fashion. (A) Experimental model. Mice were injected subcutaneously with Nb. Five or 7 d later, mice were given intravenous injection of anti-CD45-PE antibody and euthanized 5 min later. Lungs were harvested and stained for flow sorting. Total ILC2s were sorted and sequenced by scRNA-seq and CITE-seq. ILC2s are defined as CD45+ Lineage− KLRG1+ CD127+. Two mice per cohort were indexed by hashing and pooled for sequencing per experimental repeat (cohorts = D5, D7). Experimental protocol was repeated on 2 consecutive days. Total n = 4 mice per cohort pooled from 2 experiments. (B) UMAP of unsupervised clustering of cells by Seurat. (C) Clustering in panel B shown with metadata overlay for vascular labelling (green = parenchymal; red = vascular). (D) Expression of indicated ILC2 transcripts within D5 time point on UMAP. (E) Expression of indicated ILC2 transcripts within D7 time point on UMAP.

Unsupervised clustering revealed 5 clusters (Fig. 2B), with clusters clearly differentiating based on intravascular labeling (Fig. 2C). Considering the D5 time point alone, we see that the vascular (red) and parenchymal (green) compartments have distinct transcriptomes, as indicated by their separation on the UMAP (Fig. 2D). On a UMAP, transcriptionally similar cells are observed closer together while those more dissimilar from one another are separated by greater distances. At D5, the vascular compartment was enriched for iILC2 markers (Il17rb, Klrg1, Itgb7, ITGβ7), with concordance between gene (scRNA-seq) and protein expression (CITE-seq) for ITGβ7. In contrast, the parenchymal compartment was enriched for nILC2 markers (Thy1, Arg1, Il1rl1, IL33Rα), with concordance between gene expression and protein expression for the IL-33 receptor (Fig. 2D). While both parenchymal and vascular compartments expressed Mki67, a marker linked to cellular proliferation, expression was enriched in the vascular compartment, indicating that circulatory iILC2s are more proliferative than tissue-resident nILC2s.

By D7, we see that the vascular compartment has shifted to be adjacent to the parenchymal compartment, indicating that the circulatory iILC2s are acquiring a tissue-resident nILC2-like transcriptome over time (Fig. 2E). Consistent with this overall shift in signature, we now see that the parenchymal compartment has gained expression of iILC2 markers (Il17rb, Klrg1, Itgb7, ITGβ7), though most markers are still expressed more strongly in the vascular compartment. In contrast, the parenchymal compartment still retains expression of nILC2 markers (Thy1, Arg1, Il1rl1, IL33Rα). Finally, proliferative capacity as assessed by Mki67 expression has shifted from the vascular compartment at D5 into the parenchyma at D7. Important to these findings, the tissue-resident population has expanded in relative size while the vascular compartment has reduced. These data are consistent with circulatory iILC2s entering the parenchyma.

Vascular and parenchymal ILC2 transcriptomes change over the course of infection

Using metadata, unsupervised clusters were readily delineated by location (vascular and parenchymal) and day of infection (Fig. 3A). The UMAP shows that there is considerably more overlap between vascular and parenchymal clusters at D7 than at D5. This is also borne out when looking at a heatmap of the top 30 defining transcripts in each population (Fig. 3B). The heatmap shows that some cells in the vascular compartment at D7 share extensive similarities with gene expression of cells in the parenchymal compartment (Fig. 3B). A portion of cells sharing similar gene expression across vascular and parenchymal compartments is not observed at day 5 of infection. Together, these data are consistent with recent vascular immigrants making up part of the parenchymal compartment, with a small number acquiring a parenchymal phenotype while still in the vasculature.

Migratory iILC2s acquire a tissue-resident nILC2 transcriptome and phenotype over the course of helminth infection. (A) Clustering shown with metadata overlay for both vascular labeling and time point (yellow = D5 vascular; orange = D5 parenchymal; purple = D7 vascular; blue = D7 parenchymal). (B) Heatmap of top 30 genes upregulated in each compartment, compared with all other compartments (compartments = D5 vascular, D5 parenchymal, D7 vascular, D7 parenchymal). Asterisks mark transcripts in panels C to E. (C) Dot plot of ILC2 transcript and protein expression (assayed by CITE-seq) across compartments. Blue indicates nILC2-associated genes, red indicates iILC2-associated genes. (D) Dot plot of cytokine transcript expression across compartments. (E) Dot plot of transcript expression for transcription factors whose transcripts are statistically significantly upregulated or downregulated across compartments, defined as false discovery rate–adjusted P value <0.05. Statistical testing was by Seurat v4.0.2 defaults for the FindMarkers() function (Wilcoxon rank sum test, followed by Bonferroni correction for multiple testing).

Considering specific nILC2- and iILC2-defining factors again, we see nILC2-associated Arg1, Il1rl1, and Thy1 increasing in expression across time and between vascular and parenchymal compartments (Fig. 3C). Conversely, iILC2-associated Klrg1 and Itgb7 reduce in expression over time and between lung vasculature and parenchyma. These patterns in gene expression are corroborated by protein expression of IL33Rα and ITGβ7 (Fig. 3C). Interestingly, iILC2-associated Il17rb expression is also increased among parenchymal cells between D5 and D7. In context, the data may indicate an enrichment in the parenchyma of recent vascular immigrants that can maintain responsiveness to IL-25 (Fig. 3C).

Transcriptomic changes suggest that iILC2s adopting an nILC2 phenotype may also gain tissue-resident nILC2 functions. If this were the case, it would be predicted that circulating iILC2s would lose IL-4 production and gain IL-5 production as they convert in phenotype.2^,^18^,^19^,^30 Indeed, Il4 and Il13 are most highly expressed in the D5 vasculature and are downregulated across time and compartment, being lowest in ILC2s of D7 parenchyma. On the other hand, Il5 is most highly expressed in the D5 parenchyma, with a slight decrease in expression between D5 and D7, and a corresponding increase in expression over the same timeframe in the vasculature (Fig. 3D). This is consistent with the parenchymal lung-resident nILC2 population expressing Il5 but being diluted as more circulatory iILC2s convert and immigrate into the parenchyma.

To understand what may be playing a role in this apparent iILC2-to-nILC2 conversion, we examined transcription factors that were significantly upregulated or downregulated across metadata-defined clusters (Fig. 3E). Bach2 and Dach2 appear to be restricted to tissue-resident ILC2s as they exhibit selective and high parenchymal expression, albeit slightly downregulated over time, again possibly due to dilution by vascular immigrants. Zbtb20, Nr4a3, and Fosb are distinctly upregulated across both time and space, indicating that these factors are increased as iILC2s enter the parenchyma. Conversely, Batf is highly expressed only in ILC2s found in the vasculature at D5. This pattern is consistent with circulating iILC2s rapidly turning off Batf expression upon arriving in the lung, a finding consistent with BATF playing a dominant role in circulatory iILC2 function.19

Transferred iILC2s express arginase as they establish lung residence

To directly assess whether circulatory iILC2s convert into lung-resident nILC2s, we performed congenic transfer experiments (Fig. 4A). Specifically, donor Arg1^YFP^CD45.2 mice were infected with Nb, and lung ILC2 subsets were sorted and adoptively transferred into congenic CD45.1 mice. To maintain the correct inflammatory state in the lung, recipient mice were also infected with Nb 3 d prior to transfer. Transferred ILC2 subsets were assessed for Arg1/YFP expression either 3 d later (D6 for host) or 10 d later (D13 for host) (Fig. S4A–C). While transferred nILC2s maintained high Arg1/YFP expression throughout the time course, transferred iILC2s gradually increased arginase expression (Fig. 4B). Consistent with the rapid acquisition of tissue residency observed in our vascular labeling experiments, the majority of the transferred iILC2 population (∼70%) had already converted to an Arg1/YFP^high^ phenotype 3 d after transfer. By 10 d after transfer, all transferred iILC2s expressed an Arg1/YFP^high^ phenotype resembling transferred nILC2s. This is consistent with intravenous labeling experiments showing that vascular ILC2s quickly acquire Arg1 after entering the parenchyma (Figs. 1D, 2D, 2E, and 3C). These findings directly confirm that circulating iILC2s quickly acquire an identity consistent with lung-resident nILC2s during the course of pulmonary helminth infection.

*Migratory iILC2s stably upregulate arginase, whereas tissue-resident nILC2s maintain a steady phenotype. (A) Experimental model. Donor CD45.2 mice (Arg1YFP reporter positive) were injected subcutaneously with Nb. iILC2s and nILC2s were sorted from the lung 5 d later. Donor iILC2s or donor nILC2s were transferred into recipient CD45.1 mice at D3 of infection. Lungs were harvested either 3 d later or 10 d later and ILC2s characterized by flow cytometry. (B) Representative contour plots and quantification of Arg1/YFP expression of donor nILC2s and donor iILC2s. Mean ± SD, n = 3 mice per group per time point, pooled from 3 independent experiments with 1 biological replicate per cohort (nILC2 recipient, iILC2 recipient). Five total biological replicates for “before transfer” control cohort. One-way analysis of variance with Tukey’s multiple comparisons. P values defined as ***P = 0.0009 and ***P < 0.0001. ns, not significant.

Transferred IL-25–responsive iILC2s acquire a tissue-resident identity in the lung

To more comprehensively characterize iILC2-to-nILC2 conversion, we leveraged single-cell sequencing in an ILC2 congenic transfer model (Fig. 5A). Similar to our vascular labeling experiments, we used a DNA-barcoded antibody to bioinformatically discriminate transferred ILC2s from recipient cells. In this case, anti-CD45.2-PE was recognized by an anti-PE CITE-seq antibody, thus differentiating donor (transferred) and recipient (endogenous) populations (Fig. S3B). Further, as IL-25 administration has been shown to drive iILC2 mobilization similar to Nb infection but gives rise to increased numbers of circulatory iILC2s compared with Nb infection, we turned to recombinant IL-25 to maximize our numbers for transfer and downstream sequencing.12^,^21 Overall experimental design was as follows (Fig. 5A): Arg1/YFP^low^ iILC2s were sorted from the lungs and MLNs of Arg1^YFP^CD45.2 donor mice treated with IL-25 (Fig. S1B, C). Sorted Arg1^YFP^CD45.2 donor iILC2s were injected into B6 CD45.1 recipients that had been infected with Nb 3 d prior. Ten days after transfer, ILC2 populations were sorted and sequenced. Sequenced ILC2 populations were (1) lung ILC2s from naïve CD45.1 mice (naïve lung nILC2s), (2) lung iILC2s from rIL-25–treated Arg1^YFP^CD45.2 mice (D3 lung iILC2s), (3) MLN iILC2s from rIL-25–treated Arg1^YFP^CD45.2 mice (D3 MLN iILC2s), (4) endogenous lung ILC2s from day 13 Nb-infected and iILC2-transferred mice (D13 lung ILC2^endog-5.1^), and (5) transferred lung ILC2s from day 13 Nb-infected and iILC2-transferred mice (D13 lung ILC2^transf-5.2^) (Fig. S5A–C). The rIL-25–treated populations represent the starting transcriptome for iILC2s, while naïve lung ILC2s represent the baseline tissue-resident nILC2 transcriptome. In addition to differentiating transferred from endogenous cells, CITE-seq was also used here to assay protein expression of surface markers ITGβ7, the IL-33 receptor IL33Rα/ST2, and the IL-25 receptor IL17RB.

Migratory iILC2s acquire a tissue-resident nILC2 transcriptome and phenotype. (A) Experimental Model: Donor Arg1YFP CD45.2 mice were given rIL-25 intraperitoneally daily for 3 d. On the fourth day, iILC2s were sorted from the lungs and MLNs and transferred into congenic C57BL/6 5.1 recipient mice infected 3 d prior with Nb. Lungs of both recipient and untreated C57BL/6 5.1 mice were harvested 10 d later and ILC2s sorted and sequenced. Lung iILC2s and MLN iILC2s from IL-25–treated Arg1YFP CD45.2 (donor genotype) mice were also sorted and sequenced at the same time. (B) UMAP of unsupervised clustering of cells by Seurat. (C) Dot plot showing expression of nILC2- and iILC2-defining transcripts across Seurat clusters. Blue indicates nILC2-associated genes, red indicates iILC2-associated genes. (D) Expression of indicated nILC2- and iILC2-defining transcripts and proteins shown on UMAP. Blue indicates nILC2-associated genes, red indicates iILC2-associated genes. (E) Clustering of ILC2 populations by Seurat (same as panel B) encoded with known metadata of cell origin. (F) Trajectory analysis, with the starting point denoted with a star. Donor cells were pooled from 14 donor mice and split equally between 2 recipient mice per experiment. Two mice per cohort were indexed by hashing and pooled for sequencing per experiment (cohorts = MLN iILC2s, lung iILC2s, recipient, naïve). The experimental protocol was repeated on 2 consecutive days. Total n = 4 mice per cohort, pooled from 2 experiments.

Unsupervised clustering reveals 4 clusters, with the major clusters being 1 to 3 (Fig. 5B). Clusters 1 and 2 have higher expression of iILC2-specific transcripts (Il17rb, Klrg1, Itgb7) and show evidence of recent proliferation (Mki67) (Fig. 5C). Furthermore, the iILC2-associated genes Klrg1 and Itgb7 and the IL-25 receptor (protein) are broadly expressed within clusters 1 and 2, confirming iILC2 identity (Fig. 5D). Cluster 3 exhibits higher expression of nILC2-specific transcripts (Thy1, Arg1, Il1rl1) and is not proliferative (Fig. 5C). Similarly, nILC2 markers Thy1, Arg1, and* Il1rl1* are broadly expressed throughout cluster 3 (Fig. 5D). This suggests that clusters 1 and 2 comprise iILC2s and cluster 3 comprises lung-resident nILC2s.

However, important differences in transcript and protein expression within cluster 3 provide evidence supporting iILC2s converting to nILC2s. While Il1rl1 transcripts are expressed throughout cluster 3, protein expression is enriched only in the upper area of the cluster. The lower region of cluster 3, with a relative dearth of IL-33Rα protein, corresponds to the same region that is enriched for the iILC2 transcripts Il17rb and Klrg1 as well as the iILC2 integrin ITGβ7. These observations are consistent with a subpopulation of cells that are enriched for iILC2 transcripts/protein embedded within cluster 3 that largely segregates as nILC2s.

Metadata and trajectory analysis support conversion of transferred iILC2s into lung-resident nILC2s during helminth infection

Overlaying sample identity (cell type) on the UMAP reveals that clusters cleanly align with known sequenced populations (Fig. 5E). Consistent with the specific marker expression examined thus far, cluster 1 comprises D3 MLN iILC2s, cluster 2 comprises D3 lung iILC2s, and cluster 3 comprises naïve lung nILC2s, D13 lung ILC2^endog-5.1^, and D13 lung ILC2^transf-5.2^. Of note, the area of cluster 3 corresponding to cells expressing Il17rb, Klrg1, and ITGβ7, but low for IL33Rα, contains the densest population of D13 lung ILC2^transf-5.2^ cells (Fig. 5D, E). This corresponds to transferred iILC2s retaining some iILC2 characteristics despite their transcriptional profiles being essentially identical to that of lung-resident nILC2s.

The fact that unsupervised clustering grouped D13 lung ILC2^transf-5.2^ with naïve lung nILC2s and endogenous D13 lung ILC2^endog-5.1^ (cluster 3) and that this cluster is distinct from the baseline donor iILC2 populations (clusters 1 and 2) indicates that iILC2s undergo a significant transformation toward acquiring a transcriptome consistent with lung-resident nILC2s at this time point. These former iILC2s no longer reflect their origins and instead acquire transcriptomes that more closely resemble those of the pre-existing tissue-resident nILC2 population. However, increased Il17rb and ITGβ7, and low IL33Rα expression on these converted cells indicate that there may be nuanced differences between these newly converted ILC2s in the lung and lung-resident nILC2s that are seeded at birth.

Finally, trajectory analysis with Monocle3 further supports the model of iILC2s converting into tissue-resident nILC2s. When given a starting point, trajectory analysis finds the shortest path of transcriptomic changes that connect populations through putative developmental time (pseudotime). In this case, the D3 MLN iILC2 population (marked with a star) is a logical place to begin the trajectory path because iILC2s are known to originate in the small intestine and first travel to the MLNs.18^,^21^,^31 Here, the trajectory path proceeds from the baseline D3 MLN iILC2s to the baseline D3 lung iILC2s, to the combined D13 lung ILC2^endog-5.1^/D13 lung ILC2^transf-5.2^/naïve lung nILC2 populations. Of particular relevance to conversion, the trajectory path first passes through the portion of cluster 3 that corresponds with the densest transferred ILC2 cell population (Fig. 5F). The final destination is naïve lung nILC2s representing the furthest fate according to pseudotime. This fits with the idea that as the immune response resolves, lung ILC2s—both the originally seeding population and converted iILC2s—return to a less activated and more resting lung-resident phenotype.

Transcriptomes of converting ILC2s resemble lung-resident ILC2s but maintain some prior iILC2 identity

In further support of migratory iILC2s converting into lung-resident nILC2s, we see a combination of canonical iILC2 markers localizing to clusters comprising iILC2s from IL-25–treated MLNs (D3 MLN iILC2s) and lung (D3 lung iILC2s) but absent in clusters comprising naïve nILC2s (naïve lung nILC2s) or ILC2 populations from Nb-infected lungs (D13 lung ILC2^transf-5.2^ and D13 lung ILC2^endog-5.1^) (Fig. 6A, B). In contrast, combined expression of canonical nILC2 markers appears only in naïve lung nILC2s, D13 lung ILC2^endog-5.1^, and D13 lung ILC2^transf-5.2^ populations (Fig. 6A, B). This is significant with respect to the D13 lung ILC2^transf-5.2^ population, given that it represents cells originating from IL-25–treated MLNs and lung iILC2 populations. Thus, using a stringent system to define iILC2s and nILC2s, the data show that, after arrival in the lung, former iILC2s acquire transcriptomes and protein expression consistent with canonical nILC2 populations.

iILC2-to-nILC2 conversion. Former circulating iILC2s resemble native, lung-resident nILC2s while retaining some distinct functional potential based on gene expression. (A) Breakout of UMAP by known metadata of cell origin. (B) Breakouts of UMAP by cell origin showing cells satisfying iILC2-qualifying combined marker (eigengene) expression and nILC2-qualifying combined marker (eigengene) expression. (C) Heatmap of top 30 genes upregulated in each cell type, compared with all other cell types. Asterisks mark transcripts in panels D to H. (D) Violin plots of gene expression and protein expression (assayed by CITE-seq) for iILC2- and nILC2-defining targets. No difference (nd) was defined as fold change (FC) (base 10) < 1.2, or false discovery rate–adjusted P value >0.05. Statistical testing was by Seurat v4.0.2 defaults for the FindMarkers() function (Wilcoxon rank sum test, followed by Bonferroni correction for multiple testing). (E) Dot plot of transcript expression for functional factors expressed by ILC2s across cell types in the congenic transfer model. (F) Dot plot of transcript expression of transcription factors responsible for regulating factors shown in panel E. (G) Dot plot of transcript expression for transcription factors whose transcripts are statistically significantly upregulated or downregulated across cell types, defined as false discovery rate–adjusted P value <0.05. Statistical testing was by Seurat v4.0.2 defaults for the FindMarkers() function (Wilcoxon rank sum test, followed by Bonferroni correction for multiple testing). (H) Dot plot of transcripts associated with migration and tissue residency across cell types.

The conversion of the transferred population is illustrated on a broader scale by a heatmap of the top 20 most differentially expressed genes in each subset (Fig. 6C). While most similar to their endogenous D13 counterparts, transferred cells also share significant similarity with naïve nILC2s. This is in stark contrast to the lack of similarity between these transferred cells at D13 (D13 lung ILC2^transf-5.2^) and their starting transcriptome illustrated by iILC2s derived post–rIL-25 from the MLNs and lung (D3 MLN iILC2s; D3 lung iILC2s). Significantly, many genes associated with canonical tissue-resident phenotype and function, such as Arg1, Areg, and Il1rl1, as well as some canonically associated with circulatory iILC2s (Il4) are identified in this unbiased approach (asterisks on heatmap).

To further illustrate this point, the expression of key iILC2 and nILC2 markers (genes and proteins) are depicted in violin plots by cell population (Fig. 6D). Consistent with a transition to a lung-resident ILC2 identity, donor-derived lung ILC2s on D13 of Nb infection show a loss of canonical iILC2 gene (Itgb7 and Klrg1) and protein (ITGβ7) expression. Also consistent was an observed gain in canonical lung-resident nILC2 genes (Arg1, Thy1, and Il1rl1) and protein (IL33Rα). The exception to this pattern is the IL-25 receptor, for which transcript is strongly expressed in transferred ILC2s by D13 (D13 lung ILC2^transf-5.2^). Interestingly, IL-25 receptor protein expression was reduced to levels similar to that in endogenous D13 ILC2s (D13 lung ILC2^endog-5.1^) and naïve nILC2s (naïve lung nILC2s). This loss of active receptor at the cell surface but maintenance of gene expression correlates with a poised state that can rapidly upregulate IL-25 responsiveness in the future. Collectively, this acquired expression profile renders these former iILC2s as similar to lung-resident ILC2 populations with the exception that converted iILC2s may exhibit increased IL-25 responsiveness.

Transferred iILC2s convert toward cytokine profiles consistent with nILC2s while retaining some iILC2 characteristics

Given the observed conversion of transferred iILC2s into lung-resident nILC2s, we were interested in understanding if these cells also changed their gene expression to reflect acquisition of cytokine profiles consistent with nILC2 identity. Specifically, iILC2s express IL-4 and IL-13 while nILC2s express IL-5 and IL-13.12^,^19 In support of conversion to nILC2-like cytokine capacity, both D13 lung ILC2^transf-5.2^ and D13 lung ILC2^endog-5.1^ express Il5 and Il13 at similar levels and higher than naïve nILC2s (naïve lung nILC2s) (Fig. 6E). However, only the transferred D13 lung ILC2^transf-5.2^ expressed significant Il4. This is indicative of circulatory iILC2s gaining nILC2-associated Il5 production capacity but retaining iILC2-associated Il4 production capacity after conversion in the lung.

Gained Il5 expression and retained Il4 expression among D13 lung ILC2^transf-5.2^ indicated that converted iILC2s may express a unique type 2 transcription factor profile relative to stereotypical nILC2s and their original iILC2 state. IL-5 and IL-13 production in nILC2s is linked to the canonical Th2 transcription factors GATA-3 (Gata3) and Stat6 (Stat6), which are required to maintain canonical type 2 cytokine expression.32–34 In contrast, c-Maf (Maf) has been linked to noncanonical IL-4 expression in the absence of GATA-3.9^,^35–39 Consistent with IL-4 expression in the absence of IL-5 in iILC2s, Maf expression was enriched in both D3 MLN iILC2s and D3 lung iILC2s (Fig. 6F). In contrast, converted iILC2s (D13 lung ILC2^transf-5.2^) and endogenous nILC2s (D13 lung ILC2^endog-5.1^; naïve lung nILC2s) are enriched for Gata3 and Stat6 expression (Fig. 6F). Transcription factor expression indicates that D13 lung ILC2^transf-5.2^ cells maintain a low level of noncanonical Maf-driven IL-4 expression, but also acquire canonical Gata3-driven cytokine production, which may be enhanced by Stat6 signaling. When this transcription factor expression is paired with cytokine expression, this could help explain why converted iILC2s continue to produce IL-4 (noncanonical c-Maf) in addition to producing IL-5 and IL-13 (canonical GATA3 and Stat6). Together the data suggest that while converted iILC2s become considerably more tissue resident–like in identity, they do retain some unique iILC2 characteristics.

The retention of limited iILC2 characteristics in converted iILC2s (D13 lung ILC2^transf-5.2^) is also evident when investigating Il17a and Ahr gene expression. iILC2s arriving in the lung early after immune insult or alarmin administration are known to express IL17A.12 Further, both c-Maf and the receptor Ahr are linked to IL-17 production in ILC3.40^,^41 Our data indicate that Il17a expression is not only maintained in D13 lung ILC2^transf-5.2^ cells, but it is also upregulated from baseline when comparing expression to D3 MLN and lung iILC2s. Furthermore, Ahr expression is also upregulated and maintained in D13 lung ILC2^transf-5.2^ cells, in contrast to negligible Il17a and very low Ahr in pulmonary nILC2s (naïve lung nILC2s and D13 lung ILC2^endog-5.1^) (Fig. 6E).

Further consistent with converting iILC2s acquiring gene expression reflective of tissue-resident nILC2s in the lung, D13 lung ILC2^transf-5.2^ upregulated Areg and Csf2 (GMCSF) relative to iILC2 baseline (D3 MLN iILC2s and D3 lung iILC2s). In the case of Areg, transferred cells (D13 lung ILC2^transf-5.2^) attained similar expression levels to that in the endogenous population (D13 lung ILC2^endog-5.1^), which in turn also reflected an upregulation from baseline naïve lung nILC2 expression levels (Fig. 6E). This dynamic indicates that both tissue-resident nILC2s and converting iILC2s upregulate Areg expression. In the case of GMCSF, which is known to be expressed by nILC2s in the lung, D13 lung ILC2^transf-5.2^ exhibit increased expression of Csf2 relative to their origins (D3 MLN iILC2s and D3 lung iILC2s), but do not reach comparable levels as naïve lung nILC2s and D13 lung ILC2^endog-5.1^ (Fig. 6E).42

To reveal potential drivers of iILC2-to-nILC2 conversion, we looked at transcription factors that were significantly upregulated or downregulated across populations (Fig. 6G). Notably, the same transcription factors we identified in the vascular labeling experiments were also (but not exclusively) found in these congenic transfer experiments (Bach2, Dach2, Zbtb20, N4ra3, Fosb, and Batf) and their patterns of regulation were also analogous (Figs. 3E and 6G). Of note, transferred D13 lung ILC2^transf-5.2^ showed similar transcription factor profiles as their naïve lung nILC2s and D13 lung ILC2^endog-5.1^ counterparts, further highlighting the potential involvement that these transcription factors may have in iILC2s acquiring a lung-resident, nILC2 identity.

The above data are consistent with circulating iILC2s acquiring properties of lung-resident nILC2s after entering the lung. However, to be considered true tissue-resident nILC2s, converted iILC2s would need to be maintained in the lung and not re-enter circulation. The tissue-resident nature of nILC2s is consistent with the definition used for conventional tissue-resident memory T cells (TRMs).43^,^44 TRMs lose their ability to egress from tissues by shutting off responsiveness to lymphatic sphingosine-1-phosphate (S1P) gradients. S1P receptor (S1PR) expression is mediated by Kruppel-like factor 2 and TRM lack expression of both Klf2 and S1PRs.45^,^46 Because circulating ILC2s, similarly to T cells, express S1PR1 and S1PR4 and require S1P-mediated signaling to enter circulation, we theorized that iILC2s, like TRM, would also lose S1PR expression after entering the lung.19^,^21^,^22 Consistent with this hypothesis, circulating iILC2s from IL-25–treated MLNs (D3 MLN iILC2s) and lung (D3 lung iILC2s) strongly expressed transcripts of S1pr1, S1pr4, and Klf2, while transferred iILC2s (D13 lung ILC2^transf-5.2^) had lost S1PR and Klf2 expression by day 13 of Nb infection (Fig. 6H). The lack of S1PRs and Klf2 was mirrored in both endogenous lung ILC2s (D13 lung ILC2^endog-5.1^) and naïve nILC2s (naïve lung nILC2s) cells, consistent with a transition to tissue-residency by iILC2s (Fig. 6H).

If lung residency was achieved, we also theorized that these converted iILC2s would lose their original intestinal-associated chemokine receptor expression (CCR9) and increase lung- and TRM-associated chemokine receptors (CCR8, CXCR4, CXCR6).47–51 Indeed, D13 lung ILC2^transf-5.2^ cells lost Ccr9 expression and gained Ccr8, Cxcr4, and Cxcr6 expression. Furthermore, transferred iILC2s also acquired gene expression for PD1, CD69, and CD103 (Itgae), 3 established TRM markers (Fig. 6H).52 Of note, CD69 inhibits S1PR expression and cell egress.53 Together, the data show that circulating iILC2s adapt an nILC2-like identity and acquire markers consistent with the ability to establish tissue-residency.

Discussion

The stability of the tissue-resident nILC2 population has been well characterized.54 While self-renewal is the major mechanism of maintenance during homeostasis, prenatal and postnatal nILC2 populations are replaced by bone marrow and local tissue progenitors over time.4^,^20 Expansion of lung-resident ILC2s also occurs as a result of pulmonary insults and inflammation.3^,^12^,^55^,^56 While increased ILC2 numbers after pulmonary insult has been attributed to local proliferation of lung-resident nILC2s, a significant contribution from outside sources has been observed. After successful clearance of Nb, 10% to 15% of the lung ILC2 pool is derived from non–lung-resident ILC2 origins.4^,^22 The cellular source or sources of the de novo contribution to the lung-resident ILC2 pool after pulmonary insult have remained incompletely defined. While some contribution likely reflects further seeding and differentiation by bone marrow–derived progenitors,18^,^20^,^51 recent work highlights a potential contribution by circulatory ILC2s from the intestine.18^,^22 In support, data here show that circulatory iILC2s make upward of 40% of the total pulmonary ILC2 population early during Nb infection. This early wave of circulating iILC2s from the intestine being mobilized to the lung has also been observed with other intestinal-dwelling helminths and protozoa.57^,^58 Nevertheless, phenotypically, this circulatory iILC2 population was absent from the pulmonary ILC2 pool by D12 of Nb infection and the fate of these circulatory iILC2s throughout the pulmonary response has remained unclear.4^,^12^,^21^,^22

Experiments herein explore the fate of circulating iILC2s over the course of primary Nb infection. The key finding is that these former circulating cells acquire a tissue-resident-like phenotype, consistent with a potentially important contribution to the lung-resident ILC2 pool. This is likely similar to what is observed during early ontogeny, in which progenitors from the bone marrow and yolk sac and local progenitors adopt a tissue-specific phenotype.4^,^20^,^59 Findings herein show that this phenotypic switch is rapid and begins as the circulating ILC2s first arrive in the lung, which occurs between D4 and D5 post Nb infection. While initially much of the conversion occurs in the lung parenchyma, transcriptomic analysis shows that a small percentage of circulating cells begin to adopt a tissue-resident gene program while still in the vasculature. The percentage of vascular conversion appears to increase with time, indicated by increased expression of tissue-resident genes by ILC2s in the vascular compartment. Tissue-resident gene expression is supported by the observed increase in Arg1, a tissue-resident marker.26^,^27 Together, this may reflect vascular iILC2s gaining access to tissue-specific signals as a result of damage to capillaries and leakage from surrounding tissue which occurs during helminth larval entry and migration through the lungs between 1 and 3 d postinfection.17^,^60

While the tissue-specific signals that drive this conversion remain unknown, the epithelial alarmin IL-33 could play a role similar as that observed in the neonatal period in which it appears critical in the maturation of early-life ILC2s into long-lasting, IL-33–responsive lung-resident populations.61 However, tissue alarmins are likely not sufficient to achieve full conversion as ILC2s in mice lacking receptors for IL-33, IL-25, and TSLP still exhibit lung-specific transcriptomes.59 Other factors that could be contributing are leukotrienes, neuropeptides, and prostaglandins, which all have been shown to activate IL-33–responsive ILC2s or work synergistically with IL-33.62–68 Potentially of equal importance in the acquisition of a lung-resident phenotype may be the loss of non–lung-derived signals. For example, expression of intestinal homing molecules α4β7 and CCR9, molecules that are lost from iILC2s as they convert toward an nILC2 phenotype in the lung, has been linked to the gut-specific tissue factor retinoic acid.47^,^69^,^70 Important to this possibility, retinoic acid–driven α4β7 and CCR9 expression on T cells requires the AP-1 factor BATF.70 BATF is a key transcription factor that differentiates lung iILC2s from nILC2s, and data herein show that BATF expression is lost as iILC2s transition from their intestinal identity into a more lung-resident phenotype.19

As these former circulating iILC2s spend time in the lung, they become increasingly similar to the lung-resident nILC2 population seeded in early life. Specifically, the vast majority of genes expressed by naïve nILC2s in the lung and endogenous lung-resident ILC2s responding to Nb infection are paralleled in expression by circulating iILC2s that are retained in the lung during the response. However, these converted ILC2s do maintain some of their original properties as iILC2s, such as Il17rb, Il4, and Il17a gene expression. The maintained IL-25 receptor gene expression may indicate that circulating ILC2s that acquire lung-residency remain more poised to regain IL-25 responsiveness relative to conventional nILC2s. However, their reduced expression of the receptor at the cell surface within 10 to 13 d postinfection suggests that they are more limited in response to IL-25 than when they first appeared in the lung as early iILC2 immigrants. Whether these converted ILC2s remain poised for IL-25 receptor expression overtime or what signals are required to maintain receptor expression at the cell surface is of interest given that prolonged, high-dose recombinant IL-25 administration has been linked to the generation of IL-25–responsive memory ILC2s.23 The relationship between these converted iILC2s and IL-25–responsive, memory ILC2s remains unclear, but they both originate in the intestine, which could indicate that IL-25 availability in the lung is a key factor for maintaining optimal IL-25 responsiveness over time.

Similarly, enhanced cytokine gene expression by converted ILC2s may suggest a heightened ability to produce IL-4 and IL-17A compared with other lung-resident ILC2 populations. In support of this idea, converted ILC2s expressed transcript for Ahr, a receptor known to be important in modulating IL-17A in ILC2s.71 The maintenance of IL-17 in circulating and converted ILC2s may also indicate why circulating iILC2s are associated with steroid-resistant, neutrophilic asthma.72–74 As such, blocking conversion of circulating ILC2s may be desirable in treatment of severe asthma.

To this effect, we have also identified transcription factors that are differentially regulated over the course of conversion from iILC2 identity to nILC2 identity. In particular, Bach2, Fos, Fosb, Junb, Nr4a2, and Nr4a3 are increased and stably maintained during conversion. Many of these transcription factors are part of the AP-1 family and have been shown to regulate both CD4 and CD8 T cell differentiation through interaction with other AP-1 family members.75^,^76 In addition to AP-1 factors, the Nr4a family of transcription factors are also involved in cell fate determination, notably playing roles in memory T cell differentiation.77 In many ways these former circulating cells appear to be adopting a memory-like program similar to what has been described in T cells after allergen challenge.78  Batf, on the other hand, which is highly expressed by circulating ILC2s and required for these cells to reach the lung, was found to be stably downregulated as circulating ILC2s adapt a more lung-resident identity.19 Loss of BATF expression in converting ILC2s also tracked with loss of intestinal homing ITGβ7 (of α4β7) and chemokine receptor CCR9. This is consistent with what has been observed in T cells, where the loss of α4β7 and CCR9 on intestinal CD4^+^ T cells was linked to the absence of BATF.70

The data herein also show that circulating iILC2s likely lose the ability to emigrate from the lung through the loss of S1PR expression. Thus, these cells appear distinct from ILC2s extruding from the lung late during Nb infection or which are observed to enter circulation upon IL-33 administration.22^,^79 Furthermore, during conversion, former iILC2s gain both tissue residency markers such as PD1, CD69, and CD103, and tissue-resident and lung-associated chemokine receptors such as CXCR6 and CCR8.44^,^50^,^52 These shifts clearly indicate a transition from a circulatory cell arriving to the lung from the intestine to a cell establishing lung residency. These findings indicate that iILC2s follow a similar path as circulating effector T cells when differentiating into tissue-resident memory cells.43 Here, recruited iILC2s (ILC2_rec_) adopt a tissue-resident phenotype (ILC2_convt_) consistent with allowing them to join the tissue-resident nILC2 pool (ILC2_res_).

The precise contribution of ILC2_convt_ to the long-term lung-resident nILC2 pool relative to other contributors such as ILC2_res_ undergoing self-renewal or those being reseeded by bone marrow or local progenitors still needs to be worked out. However, data here strongly suggest an important contribution of ILC2_convt_ to the tissue-resident population. Whether this will become more prevalent after each immune insult or whether such contributions by circulatory ILC2s to the lung-resident populations will increase naturally with age remain important areas of investigation.17 Specifically, a redistribution of mature ILC2s from one mucosal tissue to another after infection could have evolved as a way to maintain mucosal-resident ILC2 numbers, which are rapidly replaced in some tissues after seeding in early life and also naturally decline with age.4^,^80 However, the possibility that the lung ILC2 niche eventually reaches “capacity,” preventing subsequent entry and contribution by further waves of circulatory iILC2s after subsequent infections and pulmonary insults or with age, remains equally plausible.

While likely advantageous with respect to protection against infection and maintaining barrier integrity over time, ILC2_rec_ and ILC2_convt_ may also play a detrimental role. Specifically, patients with severe asthma and chronic rhinosinusitis with nasal polyps have elevated circulating ILC2 populations.73 This might indicate that ILC2_convt_ are contributing to the prolonged and excessive type 2 inflammation observed in the lungs and upper airways of these patients. If true, preventing the accumulation of ILC2_rec_ and subsequent conversion and retention of ILC2_convt_ in the tissue-resident ILC2 pool may limit disease pathobiology and future exacerbations. Inhibiting functionality or preventing conversion of circulating ILC2s also may be of particular benefit in treatment of other chronic lung and upper airways diseases associated with a presence of increased circulating iILC2-like cells such as chronic rhinosinusitis with nasal polyps and chronic obstructive pulmonary disease.73^,^81–83 Certainly, the receptors, transcription factors, and pathways identified here represent potential therapeutic targets to minimize the conversion of circulating ILC2s into lung-resident ILC2s. Such an idea has been proposed for blocking the contribution of circulating T cells in asthma responses.84 Together, these findings warrant a more in-depth look at how circulating ILC2s and how their conversion into tissue-resident populations might impact health and disease.

Supplementary Material

vkaf352_Supplementary_Data