BATF acts as an essential regulator of IL-25–responsive migratory ILC2 cell fate and function

See allHide authors and affiliations

Science Immunology  10 Jan 2020:
Vol. 5, Issue 43, eaay3994
DOI: 10.1126/sciimmunol.aay3994

Early epithelial responder

Innate lymphoid cell 2 (ILC2) cells can be defined as two distinct cell subsets based on responsiveness to IL-25 (called iILC2) or IL-33 (called nILC2). Miller et al. identified the AP-1 superfamily transcription factor BATF as a modulator of ILC2 fate. BATF-deficient mice infected with Nippostrongylus brasiliensis lacked iILC2 cells, were unable to protect against infection, and displayed defective ILC2-cytokine responses. Further investigation revealed that IL-25–responsive BATF-dependent migratory iILC2 expressed low levels of arginase-1, making them distinct from tissue-resident nILC2. BATF-dependent iILC2 cells also produced IL-4 and IL-13 during helminth infection in the lung and intestine. These findings indicate that these BATF-dependent migratory iILC2s can respond early to epithelial damage and contribute to the restoration of the epithelium.


A transitory, interleukin-25 (IL-25)–responsive, group 2 innate lymphoid cell (ILC2) subset induced during type 2 inflammation was recently identified as iILC2s. This study focuses on understanding the significance of this population in relation to tissue-resident nILC2s in the lung and intestine. RNA-sequencing and pathway analysis revealed the AP-1 superfamily transcription factor BATF (basic leucine zipper transcription factor, activating transcription factor–like) as a potential modulator of ILC2 cell fate. Infection of BATF-deficient mice with Nippostrongylus brasiliensis showed a selective defect in IL-25–mediated helminth clearance and a corresponding loss of iILC2s in the lung characterized as IL-17RBhigh, KLRG1high, BATFhigh, and Arg1low. BATF deficiency selectively impaired iILC2s because it had no impact on tissue-resident nILC2 frequency or function. Pulmonary-associated iILC2s migrated to the lung after infection, where they represented an early source of IL-4 and IL-13. Although the composition of ILC2s in the small intestine was distinct from those in the lung, their frequency and IL-13 expression remained dependent on BATF, which was also required for optimal goblet and tuft cell hyperplasia. Findings support IL-25–responsive ILC2s as early sentinels of mucosal barrier integrity.


Disruption of mucosal barriers can occur in many ways and often initiates a type 2 immune response (1). Type 2 immunity is the process by which damage to an epithelial barrier is recognized, the pathologic agent is contained or eliminated in a way that minimizes prolonged inflammation, and damage to the underlying tissue is repaired to restore barrier integrity. This is orchestrated through the recruitment of both innate and adaptive immune cells to the site of barrier damage. In the case of helminth infection, these cells promote worm expulsion (weep and sweep response) and direct tissue repair (wound healing) (24). Innate and adaptive immune cells mediate both of these processes through the production of type 2 cytokines (5).

Tissue-resident group 2 innate lymphoid cells (ILC2s) respond to barrier disruption by sensing epithelial-derived alarmins and serve as early initiators of type 2 responses (68). Although enriched at mucosal sites, ILC2s reside in many tissues and respond to interleukin-25 (IL-25), IL-33, and thymic stromal lymphopoietin (TSLP) released by damaged mucosal epithelium (9, 10). Despite CD4+ T cells being the major producers of IL-4 and IL-13 at the peak of helminth infection, ILC2s expand in number in response to IL-25 and IL-33 and are prolific producers of IL-5 and IL-13 on a per-cell basis (6, 1113). In addition, ILC2s play an important role in intestinal homeostasis by IL-13–mediated goblet and tuft cell differentiation (1417) and help to mobilize eosinophils via IL-5 (11). Currently, ILC2 subsets can be separated into tissue-resident natural ILC2s (nILC2) or transient, circulating inflammatory ILC2 (iILC2) populations. Tissue-resident ILC2 cells proliferate locally at steady state and localize to adventitia (1821) and are highly motile within lung tissue after alarmin activation (22). Although tissue-resident ILC2 cells have been the subject of much study, less is known in regard to the significance of circulating ILC2 cells that mobilize in response to type 2 inflammatory cues in the bone marrow and intestine and arrive in the pulmonary vasculature (23, 24).

Until recently, ILC2s were thought to be a homogeneous population when compared with ILC1 and ILC3 subsets (25). However, recent studies have identified two distinct populations of ILC2s characterized by their differential responsiveness to IL-25 and IL-33 (26, 27). The IL-25–responsive subset, called iILC2, expresses high levels of the C-type lectin receptor, KLRG1, and low levels of CD90 and is transiently found in murine lungs early after Nippostrongylus brasiliensis infection or IL-25 administration. This differs from the IL-33–responsive nILC2 subset, which is tissue resident and displays intermediate levels of KLRG1 and high levels of CD90 (26). Although the relationship between IL-25– and IL-33–responsive ILC2s is not well understood, it has been suggested that tissue-resident nILC2s self-renew at steady state and proliferate locally, with some cells likely arriving from the bone marrow in response to IL-33, whereas migratory iILC2s that arrive in the lung have been suggested to originate from the intestine (19, 20, 23, 24). Whereas the role of tissue-resident nILC2s has been well defined, the significance of this migratory, IL-25–responsive iILC2 subset is less clear, and whether there are distinct lineage-determining factors involved in their selective development, fate, or function is not known.

Our previous studies identified the activator protein–1 (AP-1) superfamily transcription factor BATF (basic leucine zipper transcription factor, activating transcription factor–like) as an essential transcription factor involved in helminth clearance (28). BATF is required for the generation of follicular T helper (TFH) cells and T helper 2 (TH2) cells, as well as their production of type 2 cytokines (2831). As a result, the absence of BATF prevents both humoral and cell-mediated aspects of type 2 immunity. Despite its well-characterized role in adaptive immunity and type 2 cytokine production, a role for BATF in ILCs has not been described (32). This study reveals that BATF is required by pulmonary iILC2 cells in response to helminth infection and acts in a cell-intrinsic manner. Furthermore, we show that this BATF-dependent migratory iILC2 population is distinct from tissue-resident nILC2s via its low expression of the ILC2 marker arginase 1 (Arg1). Unexpectedly, it is the migratory iILC2 subset and not the tissue-resident subset that acts as the early source of IL-4 and IL-13 in the lung and intestine during infection. In addition, intestinal IL-25–responsive ILC2s serve as key early mediators of goblet and tuft cell hyperplasia, as helminths establish residence in the intestine. We propose that BATF-dependent migratory iILC2s act as early responders to damaged mucosal epithelium and orchestrate the initial steps in the reestablishment of barrier integrity at these sites.


RNA-sequencing and pathway analysis reveals BATF as a key transcription factor of pulmonary ILC2s

Because ILC2s are critical for helminth clearance, we sought to investigate transcriptional differences between KLRG1-positive ILC2s (which include all ILC2 subsets) and KLRG1-negative populations (non-ILC2s; enriched for ILC1 and ILC3 cells) in the lung after IL-33–induced ILC2 expansion. ILC2s were sorted as live, singlet, SSClowFSClowLinCD4CD90+KLRG1+ cells and compared with the KLRG1neg population (fig. S1, A to C). Consistent with previous RNA sequencing (RNA-seq) of ILC2s after IL-33 administration, KLRG1pos and KLRG1neg populations displayed distinct transcriptional profiles, as evidenced by principal components analysis (PCA) (fig. S1D) and pairwise hierarchical clustering (fig. S1E) (33, 34). Furthermore, the KLRG1pos cells displayed an enrichment in genes associated with phenotypic markers of ILC2 cells such as Il1rl1 (ST2, IL-33 receptor), Il17rb (IL-25 receptor), and Klrg1 (Fig. 1A), as well as increased expression of genes related to ILC2 function such as Arg1, Il4, Il13, and Il5 (Fig. 1B) (33, 3537). Predicting associated transcription factors from annotated affinities (PASTAA) analysis identified the AP-1 transcription factor Atf-1/Batf as among the factors predicted to be regulating significantly enriched genes within KLRG1pos compared with KLRG1neg populations (Fig. 1C). PASTAA is a bioinformatic tool used to predict transcription factors that likely regulate gene sets based on how strongly the transcription factor associates with eukaryotic promoters (38). Gene set enrichment analysis was used to calculate normalized enrichment scores for the transcription factors identified by PASTAA in KLRG1pos ILC2s (Fig. 1D).

Fig. 1 Batf is expressed by lung ILC2 cells and is predicted to regulate distinct gene sets in KLRG1pos populations.

WT C57BL/6 mice were injected intraperitoneally with IL-33 daily for 4 days, and pulmonary KLRG1neg (non-ILC2s) and KLRG1pos (ILC2s) were sorted on day 5 and used for RNA-seq. (A) Diagonal scatter plot comparing differentially expressed genes (Padj < 0.05) among KLRG1neg and KLRG1pos cells. (B) Heat map of ILC2-related genes and S1P receptor genes up-regulated in KLRG1pos ILCs. *P ≤ 0.05. (C) Transcription factors predicted by PASTAA analysis to be involved in regulating genes that were enriched among KLRG1pos, compared with KLRG1neg populations. (D) Normalized enrichment scores (NES) of the predicted transcription factors (TF) as assessed with gene set enrichment analysis of KLRG1pos ILC2s compared with the KLRG1neg population. Binding motifs as reported by TRANSFAC are also included in the table, where S = C or G, W = A or T, R = A or G, Y = C or T, M = A or C, and N = any base. RNA-seq analysis was compiled from three separate experiments with n = 3 to 6 mice per group.

BATF selectively promotes IL-25–responsive, KLRG1high iILC2s during helminth infection

ILC2s are identified as live, singlet, SSClowFSClowLinCD4CD90+KLRG1+ cells (fig. S2) and are required to clear N. brasiliensis within 5 days of infection after administration of IL-25 or IL-33 (6, 9). To investigate whether an impairment in ILC2 function exists in BATF-deficient mice, we assessed intestinal worm burden after administration of phosphate-buffered saline (PBS), IL-25, or IL-33 in wild-type (WT), Rag−/−, Batf−/−, and Rag−/−Batf−/− mice 5 days after N. brasiliensis infection. Although there was no difference in IL-33–dependent helminth expulsion between the groups of mice, Batf−/− and Rag−/-Batf−/− mice were significantly impaired in their ability to clear worms in response to IL-25 (Fig. 2, A and B). A closer investigation of pulmonary ILCs revealed that the CD90midKLRG1high IL-25–responsive iILC2 population was absent in Batf−/− mice (Fig. 2, C and D), whereas the IL-33–responsive CD90highKLRG1mid nILC2 (Fig. 2C and fig. S3A) and CD90highKLRG1neg (Fig. 2C and fig. S3B) populations remained intact. The N. brasiliensis–induced KLRG1high iILC2s were further phenotyped by low expression of ST2 (IL-33 receptor), high expression of IL-17RB (IL-25 receptor), positive for CD127, and high expression of BATF and GATA3 (Fig. 2E). Together, these data show that BATF deficiency more selectively impairs IL-25–responsive iILC2 cells than IL-33–responsive nILC2 cells.

Fig. 2 BATF is required by IL-25–responsive KLRG1high pulmonary ILC2s.

(A and B) Intestinal helminth burden after PBS, IL-25, or IL-33 administration to (A) Rag−/−, WT, or Batf−/− mice or (B) Rag−/−Batf−/− mice 5 days after infection with N. brasiliensis. Data combined from six (A) or two (B) independent experiments. (C to E) WT or Batf−/− mice were infected with N. brasiliensis, and pulmonary LinCD90+ ILC2s were assessed 5 days later by flow cytometry. (C) Flow cytometry plots of different CD90+ ILC subsets based on KLRG1 expression. (D) Number and frequency of KLRG1high ILC2s combined from seven independent experiments. (E) Representative histograms and compiled gMFI of various surface markers and transcription factors in the indicated ILC populations from infected WT mice. n = 6 to 18 mice pooled from two to seven separate experiments. Means ± SEM, *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001. Data were analyzed by two-tailed unpaired t test (D) and one-way ANOVA with Tukey post test (A, B, and E).

ILC2s are known to respond to epithelial alarmins such as IL-33, IL-25, and TSLP (6, 39). In the absence of helminth infection, intraperitoneal injections of IL-33 and IL-25, but not TSLP, led to an accumulation of CD90lowKLRG1high ILC2s in the lung (fig. S4A). Further analysis demonstrated that although IL-33 administration led to accumulation of KLRG1high ILC2s, these cells are not phenotypically distinct from the IL-33–induced KLRG1mid nILC2 subset in regard to ST2 and IL-17RB expression, whereas the KLRG1high population induced by IL-25 is unique based on alarmin receptor expression (fig. S4B). Therefore, IL-33 expands nILC2s, which can up-regulate KLRG1 expression in response to this alarmin. In contrast, IL-25 leads to the accumulation of iILC2s in the lung, which uniformly display high KLRG1 expression.

Arg1 expression differentially defines nILC2s and iILC2s

Arg1 has been described as a marker for tissue-resident ILC2s and ILC2 precursors (20, 35, 36), yet its expression among IL-25–responsive relative to IL-33–responsive ILC2s remained uncharacterized. Arg1YFP reporter mice revealed a separate and distinct KLRG1highArglow population that was absent in both uninfected Arg1YFP and N. brasiliensis–infected Batf−/−Arg1YFP mice (Fig. 3A) (40). Most of the KLRG1mid tissue-resident nILC2s showed evidence of high Arg1 [yellow fluorescent protein (YFP)] expression regardless of infection or Batf status. In contrast, KLRG1high iILC2s from infected WT Arg1YFP mice expressed little Arg1 (YFP) (Fig. 3, B and C). Note that whereas Arg1 levels in KLRG1high iILC2s are low, expression is not absent as evidenced by the geometric mean fluorescence intensity (gMFI) of YFP (arginase) when compared with a WT C57BL/6 reporter negative mouse (represented by the dashed line in Fig. 3D). Thus, Arg1 expression is a unique identifier of ILC2 subsets in vivo.

Fig. 3 KLRG1mid, but not KLRG1high, pulmonary ILC2s express high levels of Arg1.

(A to D) Arg1YFP and Batf−/−Arg1YFP mice were infected with N. brasiliensis, and pulmonary ILC2s were assessed 5 days later by flow cytometry. (A) Contour plots and combined data of the percentage of KLRG1high iILC2s. (B) Histograms displaying the percentage of YFP+ cells in the gated populations depicted in the contour plot. (C) Compiled data showing the percentage of cells that are YFP+ in (B). (D) Compiled data of the gMFI of YFP of each population in (B). The dashed line represents the gMFI of the reporter negative mouse. Data in (A) to (D) are from a single experiment with n = 2 to 4 mice per group, representative of four individual experiments. (E) Arg1YFP mice were given consecutive daily HDM (10 μg) by oropharyngeal aspiration for 4 days and rested on day 5, and pulmonary ILC2s were assessed by flow cytometry on day 6. Representative contour plots display the frequency of KLRG1high iILC2s in the lung, and the bar graph shows combined data from n = 6 to 7 mice per group compiled from two separate experiments. **P ≤ 0.01, ***P ≤ 0.001, as determined by one-way ANOVA and Tukey post hoc analysis (A, C, and D) or two-tailed unpaired t test (E).

House dust mite administration does not induce pulmonary iILC2s

We next wanted to understand whether iILC2s could be detected in other forms of acute type 2 pulmonary inflammation. To do this, we treated Arg1YFP mice for four consecutive days with house dust mite (HDM) extract via oropharyngeal aspiration. Unlike N. brasiliensis infection, which induced robust iILC2 recruitment to the lung, administration of HDM did not lead to recruitment of iILC2s at the time point tested (Fig. 3E). These results indicate that the early induction or recruitment of iILC2 cells to the lung can vary among different settings of type 2 inflammation. The unique events involved in iILC2 induction during helminth infection that are absent in allergen models are likely to reveal important insight into iILC2 biology and function.

BATF works in a cell-intrinsic manner to modulate iILC2 cell fate during helminth infection

To determine whether the decreased frequency of pulmonary iILC2s in Batf−/− mice was cell intrinsic, we generated mixed bone marrow chimeric mice. Irradiated WT CD45.1/CD45.2 hosts were reconstituted with equal numbers of WT CD45.2 and Batf−/− CD45.1 bone marrow and infected with N. brasiliensis 8 weeks later (Fig. 4A and fig. S5A). Of the KLRG1high cells, the majority originated from WT bone marrow, whereas significantly fewer originated from either residual host cells or Batf−/− bone marrow (Fig. 4, B and C). Looking at the total LinCD90+ pulmonary ILC compartment, ~23% of the cells were residual CD45.1/CD45.2 host cells, indicating their radio-resistant nature, a phenomenon that has already been observed in ILC3s (fig. S5B) (41). Furthermore, there was a major defect in the ability of Batf−/− bone marrow to repopulate the total ILC compartment because WT bone marrow gave rise to most LinCD90+ cells. When gating on ILCs of each donor origin first and then assessing KLRG1 expression, the KLRG1highCD90+ iILC2 subset could only be generated from the WT donor and not Batf−/− bone marrow (fig. S5, B and C), further confirming the cell-intrinsic requirement for BATF. Surprisingly, we also observed the expansion of pulmonary KLRG1high ILC2s derived from WT bone marrow in uninfected mice (fig. S5, D and E). Although this expansion required BATF, it does raise the question of whether homeostatic expansion alone can result in increased KLRG1 expression among ILC2s. In support, the apparent lack of nILC2s in the KLRG1mid gate after irradiation indicates that, similar to IL-33 administration (fig. S4A), KLRG1 expression may increase on nILC2s under certain conditions.

Fig. 4 Cell-intrinsic requirement of BATF for pulmonary KLRG1high ILC2s.

(A) Schematic depicting the generation and infection of bone marrow chimeric mice. Cells were analyzed 5 days after N. brasiliensis infection. (B) Contour plots showing congenic markers delineating LinCD90+KLRG1high lung cells that arose from CD45.1 Batf−/− or CD45.2 WT donors or CD45.1/CD45.2 residual host cells. (C) Combined results from two separate experiments; n = 7. *P ≤ 0.05, **P ≤ 0.01, as determined by one-way ANOVA and Tukey post hoc analysis.

ILC2s in the small intestine lamina propria (SILP) were also assessed in the bone marrow chimeric mice. In contrast to what was observed in the lung, residual ILC2s from the recipient mice were absent, indicating potential tissue-specific differences in irradiation efficiency. Moreover, although there was a reduction in the frequency of SILP ILC2s in a few Batf−/− mice, this result was not statistically significant across all mice (fig. S5, F and G).

KLRG1high iILC2s are the major ILC2 source of IL-4 and IL-13 5 days after helminth infection

To further investigate the function of pulmonary ILC2s, we identified immunologically relevant pathways and genes enriched in ILC2s through RNA-seq. Using genes that were at least twofold enriched in KLRG1pos ILC2s, we identified cytokine activity and cytokine-cytokine receptor interactions as significantly associated pathways by both Gene Ontology molecular function and KEGG (Kyoto Encyclopedia of Genes and Genomes) analysis (Fig. 5, A and B). Furthermore, expression of genes for type 2 cytokines such as Il4, Il5, and Il13 increased in KLRG1pos ILC2s relative to the KLRG1neg population (Fig. 5C). The emphasis on pathways involved in type 2 cytokine expression and ILC2 function was consistent with a key role for BATF in TH2 and TFH cell function (28, 30, 31).

Fig. 5 KLRG1high iILC2s are responsible for type 2 cytokine production in the lung 5 days after N. brasiliensis infection.

(A to C) RNA-seq was performed on KLRG1pos and KLRG1neg LinCD90+ lung cells from mice treated with IL-33 as described in Fig. 1. (A and B) Genes that are significantly different and increased at least twofold in KLRG1pos relative to KLRG1neg cells were used in (A) Gene Ontology molecular function analysis and (B) KEGG pathway analysis. Results are shown as −log10 P values to denote pathways that are significantly associated with genes up-regulated in KLRG1high ILC2s. Activity groups or pathways that relate to cytokine signaling are highlighted in red. (C) Quantification of type 2 cytokine expression in KLRG1pos and KLRG1neg populations. (D to G) IL44get and Batf−/−IL44get and (H to K) IL13yetcre13 and Batf−/−IL13yetcre13 mice were infected with N. brasiliensis, and pulmonary LinCD90+ ILC2s were assessed 5 days later by flow cytometry. (D and H) Contour plot showing IL-4 (GFP) (D) or IL-13 (YFP) (H) cytokine reporter expression of KLRG1mid (blue) and KLRG1high (red) ILC2s. Gates were set according to a reporter negative control. (E and I) Bar graphs comparing the percent of KLRG1mid and KLRG1high ILC2s that express either IL-4 (GFP) with n = 8 mice combined from three separate experiments (E) or IL-13 (YFP) with n = 9 to 10 mice combined from four experiments (I). (F and J) Representative contour plots of IL-4 (GFP) (F) or IL-13 (YFP) (J) and KLRG1 expression on LinCD90+ lung ILC2s. (G and K) Bar graphs comparing the percent of KLRG1+ ILCs expressing IL-4 (GFP) with n = 6 to 8 mice combined from three separate experiments (G) or IL-13 (YFP) with n = 6 to 7 mice combined from three separate experiments (K). *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001, as determined by two-tailed unpaired t test.

To expand on these findings and extend the analysis of type 2 cytokine production to nILC2 and iILC2 subsets, C57BL/6 IL44get-GFP reporter mice were infected with N. brasiliensis, and 5 days later, IL-4 mRNA transcription, as detected by green fluorescent protein (GFP) expression, was assessed by flow cytometry in the KLRG1high and KLRG1mid ILC2 subsets. IL-4 expression during the early response to helminths is restricted to KLRG1high iILC2s with very little contribution from KLRG1mid nILC2s (Fig. 5, D and E). The role of BATF in this process is clear because there is very little IL-4 mRNA being transcribed in the total LinCD90+ ILC compartment in Batf−/− IL44get-GFP mice (Fig. 5, F and G). A similar pattern is seen in IL13yetcre13-YFP reporter mice where the KLRG1high iILC2 subset is responsible for significantly more IL-13, as detected by YFP expression, than the KLRG1mid nILC2s (Fig. 5, H and I). Again, BATF is critical for IL-13 transcription by ILCs because there is essentially no reporter expression in Batf−/− IL13yetcre13-YFP mice (Fig. 5, J and K). Defects in IL-4 and IL-13 production by Batf−/− ILC2s were confirmed using IL-4 and IL-13 protein reporter mice (fig. S6) (12, 42). These data highlight the importance of KLRG1high iILC2s in initiating the type 2 cytokine response in the lung during helminth infection and identify BATF as a key modulator of iILC2 function in type 2 immunity.

Expression of IL-13 5 days after helminth infection is confined to migratory KLRG1high iILC2s

Because pulmonary KLRG1high iILC2s are a transient population that may migrate to the lung from peripheral sites (26), we investigated the chemokine gene signature of ILC2s from our initial RNA-seq study. Using genes that were enriched at least twofold in KLRG1pos ILC2s, Gene Ontology molecular function and KEGG pathway analysis indicated that overall chemokine receptor family and C-C chemokine receptor activity were significantly more likely to be involved in the recruitment of KLRG1pos ILC2s to the lung than expected by chance (Fig. 6, A and B). In support, an array of chemokine receptors was shown to be differentially regulated between KLRG1pos relative to KLRG1neg ILCs (Fig. 6C). Specifically, enhanced expression of Ccr2, Ccr3, and Ccr1, all of which have been shown to mediate leukocyte migration to the lung during allergic inflammation [reviewed in (43)], as well as Cxcr6, which directs ILC2s to the lung (44), was detected in KLRG1pos ILC2s relative to the KLRG1neg population (Fig. 6C). To investigate whether the IL-13–producing KLRG1high iILC2 population is migratory, we administered the sphingosine-1-phosphate receptor agonist FTY720 to IL13yetcre13-YFP mice upon infection with N. brasiliensis and found that mice treated with FTY720 exhibited a significantly reduced frequency and number of KLRG1high iILC2s in the lung compared with mice given saline (Fig. 6, D and E). Similarly, blockade of lymphocyte migration led to a near-complete absence of IL-13–producing LinCD90+KLRG1+ ILC2s in the lung (Fig. 6, F and G), indicating that at this early stage of helminth infection, most IL-13 is produced by migratory KLRG1high iILC2s rather than tissue-resident nILC2.

Fig. 6 Appearance of KLRG1high pulmonary iILC2s in the lung is blocked by FTY720.

(A to C) RNA-seq was performed on KLRG1pos and KLRG1neg LinCD90+ lung cells from mice treated with IL-33 as described in Fig. 1. (A) Gene Ontology molecular function and (B) KEGG pathway analyses highlighting gene groups and pathways that are significantly associated with genes that increased at least twofold in KLRG1pos relative to KLRG1neg populations, as performed in Fig. 5. Gene groups and pathways associated with cell trafficking are highlighted in red. (C) Heatmap of chemokine receptor expression in indicated populations from three separate RNA-seq experiments. Significant differential expression is denoted by an asterisk. (D to G) IL13yetcre13 mice were infected with N. brasiliensis on day 0 and given saline control or FTY720 on days 0, 2, and 4 after infection. Pulmonary LinCD90+ ILC2s were assessed on day 5 by flow cytometry. (D) Representative contour plots of the frequency of KLRG1mid and KLRG1high lung ILC2s. (E) Combined frequency and number of KLRG1high cells; n = 3 from a single experiment representative of three independent experiments. (F) Contour plots of YFP (IL-13 mRNA) expression within the KLRG1+ subset of LinCD90+ ILC2s. (G) Combined frequency and number of YFP+ LinCD90+KLRG1high cells; n = 3 from a single experiment representative of three independent experiments. *P ≤ 0.05, **P ≤ 0.01, as determined by two-tailed unpaired t test.

Because intraperitoneal injection of IL-33 led to an expansion of KLRG1high ILC2s that resembled IL-25–induced KLRG1high iILC2s, we investigated whether any of these KLRG1high IL-33–responsive ILC2s might be migrating from the periphery. Mice given IL-33 and FTY720 displayed similar frequencies and numbers of CD90+ KLRG1high ILC2s to those given only IL-33 (fig. S7, A and B). Thus, unlike IL-25, which promotes the appearance of circulating iILC2 cells in the lung, IL-33 increased KLRG1 expression on nILC2 but did not lead to the mobilization of iILC2 cells.

ILC2s are present at a reduced frequency in the intestine of BATF-deficient mice but fail to produce IL-13

Intestinal homeostasis and helminth expulsion are modulated by ILC2 responsiveness to IL-25 (1416). This prompted us to investigate whether BATF influenced ILC2s in the SILP. Like those in the lung, SILP ILC2s were gated on live, singlet, SSC-AlowFSC-AlowLinCD90+CD4 cells (fig. S2). Unlike the lung where intermediate and high levels of KLRG1 identify unique subsets of nILC2 and iILC2s, only KLRG1high ILC2s were found in the intestine of naïve or N. brasiliensis–infected mice (Fig. 7A). This KLRG1-expressing population represents the only ILC2 population in this tissue. For this reason, we refer to SILP ILC2s as “KLRG1pos.” Moreover, this KLRG1pos population was also observed in the SILP regardless of Batf status, although infected Batf−/− mice displayed a significantly reduced frequency compared with infected WT mice (Fig. 7, A and B). To determine whether BATF controlled type 2 cytokine expression in SILP ILC2s similar to what was observed for lung ILC2s, IL-13 production was assessed after helminth challenge in IL13yetcre13-YFP or Batf−/−IL13yetcre13-YFP mice. The results again indicate a critical role for BATF in IL-13 transcription by ILC2 because significantly reduced frequencies of YFP+ cells were detected in the SILP of Batf−/− mice compared with their Batf-sufficient counterparts (Fig. 7, C and D).

Fig. 7 BATF is required for ILC2 function in the small intestine after helminth infection.

(A and B) WT or Batf−/− mice were infected with N. brasiliensis or remained uninfected. LinCD90+KLRG1+ ILC2s in the SILP were assessed 5 days after infection by flow cytometry. (A) Contour plots of SILP ILC2s in indicated mice before and after infection. (B) Compiled data comparing the frequency of LinCD90+KLRG1+ cells in the SILP between infected WT (blue) and Batf−/− (red) mice; n = 7 to 9 mice, combined from three separate experiments. (C and D) IL13yetcre13 and Batf−/−IL13yetcre13 were infected with N. brasiliensis or remained uninfected. LinCD90+KLRG1+ ILC2s in the SILP were assessed 5 days later by flow cytometry. (C) Contour plots of YFP (IL-13) expression versus KLRG1 expression. (D) Compiled data comparing the frequency of LinCD90+KLRG1+ cells that express YFP (IL-13) in the SILP between infected WT (blue) and Batf−/− (red) mice; n = 5 mice, combined from two separate experiments. (E) Representative histograms of various surface markers and transcription factors in SILP ILC2s from infected WT mice. (F and G) WT and Batf−/− mice were infected with N. brasiliensis or remained uninfected, and LinCD90+KLRG1+ ILC2s in the SILP were assessed 5 days later by flow cytometry. (F) Representative contour plots showing the percentage of LinCD90+KLRG1+ ILC2s expressing BATF. Gates were set according to Batf−/− mice. (G) Bar graph showing the percentage of LinCD90+KLRG1+ ILC2s expressing BATF; n = 6 to 8 mice, combined from three separate experiments. (H and I) Arg1YFP and Batf−/−Arg1YFP mice were infected with N. brasiliensis, and SILP ILC2s were assessed 5 days later by flow cytometry. (H) Representative contour plots showing the percentage of LinCD90+KLRG1+ ILC2s that express YFP (Arg1 mRNA). Gates were set according to reporter negative mice. (I) Bar graph showing the percentage of LinCD90+KLRG1+ ILC2s that express YFP (Arg1) between infected WT (blue) and Batf−/− (red) mice; n = 6 mice, combined from two separate experiments. **P ≤ 0.01, ***P ≤ 0.001, as determined by two-tailed unpaired t test.

These results show that ILC2 populations may differ between the lung and SILP. This is interesting because pulmonary iILC2s have been reported to originate from the SILP (24). To explore this further, we phenotypically profiled the KLRG1pos SILP ILC2s and found that they resemble KLRG1high pulmonary iILC2s in their high expression of IL-17RB, CD127, and GATA3 (Fig. 7E). Also consistent with the iILC2 phenotype, KLRG1+ SILP ILC2s lacked significant ST2 expression (Fig. 7E). The lack of ST2 expression by SILP ILC2s reveals differential alarmin responsiveness among ILC2 cells in different mucosal tissues. Despite these similarities, several major differences were evident between lung KLRG1high and SILP ILC2 populations. First, although BATF expression was detected among all KLRG1high iILC2s in the lung, only a subset of ILC2s in the SILP expressed this transcription factor (Fig. 7, F and G). Helminth infection expands the BATF-expressing SILP ILC2s relative to uninfected mice. Second, Arg1 expression was elevated among all KLRG1pos ILC2s in the SILP but not KLRG1high iILC2s in the lung (Fig. 7, H and I). Moreover, the presence of ILC2s in the SILP is not significantly affected by FTY720 administration, indicating that most KLRG1pos SILP ILC2s are resident to the intestine (fig. S7, C and D).

BATF-deficient mice have impaired tuft and goblet cell hyperplasia after helminth infection

In the SILP, ILC2-derived IL-13 is critical for worm expulsion via goblet cell hyperplasia, smooth muscle contractility, and increased tuft cell generation. Thus, we investigated whether BATF deficiency would ultimately affect tuft cells and mucin-producing epithelial cells after N. brasiliensis infection. Immunohistochemistry of IL-25–producing tuft cells, marked by expression of doublecortin-like kinase 1 (DCLK-1), revealed a significant decrease in cell number per millimeter of villus in Batf−/− mice compared with WT mice 8 days after N. brasiliensis infection (Fig. 8, A and B). Furthermore, BATF deficiency led to a significant reduction of mucin-producing (MUC2) epithelial cells (Fig. 8, C and D). These results are consistent with Batf−/− mice having decreased frequencies of IL-13–producing ILC2s relative to WT mice in their small intestine during N. brasiliensis infection (Fig. 7, C and D).

Fig. 8 BATF-deficient mice have reduced tuft cells and mucin-producing epithelial cells after helminth infection.

WT and Batf−/− mice were infected with N. brasiliensis, and immunohistochemistry was performed on the small intestine 8 days later. (A) Representative image of DAPI (blue) and DCLK (red) staining. (B) Bar graphs depicting the number of DCLK+ cells per millimeter of villus; n = 28 to 77 individual villi counted from four individual mice per group, combined from two separate experiments. (C) Representative image of DAPI (blue) and MUC2 (red) staining. (D) Bar graphs depicting the number of MUC2+ foci per millimeter of villus; n = 22 to 23 individual villi from two mice per group from one experiment. Scale bar, 100 μm. ****P ≤ 0.0001, as determined by two-tailed unpaired t test.


The role of tissue-resident ILC2s during helminth infection and allergic disease is well defined (45). These cells are prominent producers of IL-5 and IL-13 and likely work with canonical TH2 cells to promote eosinophil mobilization, wound healing, and worm clearance. In addition to their role in anti-helminth immunity and allergic disease, tissue-resident ILC2s modulate mucosal barrier homeostasis by responding to tuft cell–derived IL-25. Together, their positioning in tissues and their responsiveness to tissue-derived signals support the long-held idea that these tissue-resident nILC2 cells act as early sensors of mucosal barrier integrity. The data provided here show that migratory iILC2s also contribute to barrier immunity and serve as an important, early source of IL-4 and IL-13 in the lung after helminth infection (fig. S8). Thus, although tissue-resident nILC2s are prominent type 2 cytokine producers in the lung at the peak of the response, circulating iILC2s appear to be transient, early initiators of type 2 immune hallmarks at mucosal sites.

The selective dependence on the AP-1 superfamily transcription factor BATF in the fate and function of iILC2s relative to tissue-resident nILC2s provides further support for the unique nature of these populations in settings of type 2 immunity and helps to explain some of the heterogeneity observed among ILC2 populations in response to different tissue alarmins. Previous single-cell RNA-seq revealed that both IL-25– and IL-33–responsive ILC2s expressed BATF, but no significance of its expression in ILCs was assigned (32, 33). Here, the data indicate that BATF, although expressed in nILC2s, is highly expressed in iILC2s. BATF works in a cell-intrinsic manner among iILC2s and appears critical in both the mobilization of iILC2s into circulation and the expression of IL-4 and IL-13. Although we show that expression of GATA3 and IL-13 is common in both iILC2s and tissue-resident nILC2s, IL-4 expression is not commonly observed in mature nILC2s (6, 8, 9). Hence, the observation that migratory iILC2s express both IL-4 and IL-13 is intriguing because this is reminiscent of TH2 cells, which make both cytokines. IL-4 production has been reported among some intestinal ILC2s in response to colonization by the helminth Heligmosomoides polygyrus (46). The factors involved in migratory iILC2 expression of IL-4 are not clear, but the high expression of BATF in this population indicates one potential mechanism, as we and others have shown that BATF plays an important role in IL-4 production by T cells (2830). Future studies specifically investigating the regulome of migratory iILC2 cells and their relationship to other ILC and type 2 cytokine-producing immune cell populations will be important to better understand these functional differences (47).

Another unexpected difference between migratory and tissue-resident ILC2s is their expression of Arg1, which has been used to define ILC2s in mice and humans (20, 35, 36). Arg1 plays a key role in ILC2 metabolism and allergic inflammation. Here, we show that although tissue-resident nILC2s in the lung and intestine all express Arg1, most of the migratory lung iILC2 cells express very little, if any, of this enzyme. This may indicate either that iILC2s are a unique Arg1low ILC2 subset independent from nILC2 or ILC2 precursors (18, 20, 35, 36) or that Arg1 expression is tied to tissue residency. In this context, circulating iILC2s may be Arg1 negative, while they transit to the lung via the blood. The lack of Arg1 in iILC2 cells is revealing when placed in the context of previous publications, suggesting that these cells arise from intestinal emigrants (24). Given that all ILC2s in the intestine express high levels of Arg1 and pulmonary iILC2s lack significant expression of this enzyme, if pulmonary iILC2s originate in the intestine, then these intestinal emigrants would have to extinguish expression of Arg1 before arrival in the lung. If this is the case, then the time it takes for an intestinal ILC2 to make its way to the lung as an iILC2 would have to be longer than 16 to 24 hours given the half-life of YFP in the Arg1YFP reporter mouse. This may indicate that Arg1 expression is tied to tissue-specific signals received by ILC2s, as these cells establish a more permanent tissue residency. Recent evidence that TH2 cells become “licensed” to become tissue-resident effectors only after they enter the lung and sense alarmins supports this idea (48). The tissue in which an ILC resides can impart unique identities (49).

Alternatively, migratory iILC2 cells may be replaced by IL-33–responsive populations emigrating from the bone marrow or are outcompeted by proliferating tissue-resident populations already established at mucosal sites (23, 50). Although it was recently argued that migratory iILC2s do not contribute to the expansion of pulmonary ILC2 numbers after N. brasiliensis infection, a clear decrease (10%) of neonatally derived tissue-resident ILC2 cells was observed in the lung after helminth infection compared with controls, indicating that de novo replacement of the local tissue-resident population does occur (20). Further studies are required to delineate between these models. In any event, Arg1 expression serves as a unique identifier for iILC2 from nILC2 cells and helps to stratify tissue-resident and circulating ILC2 populations.

One limitation of this study is that it remains unclear whether human ILC2s are similarly composed of distinct ILC2 subsets that differentially respond to tissue alarmins. Human ILC2s that lack IL-33 receptor expression also lack Arg1 expression (36). This is phenotypically consistent with the migratory, IL-25–responsive population described here in mice, and whether this represents a BATF-dependent migratory iILC2 equivalent in humans is intriguing. Another consideration lies within the model of ILC2 induction. The differences in response to parasitic helminth infection, which mobilizes iILC2s to the lung, compared with allergen exposure, which does not, suggest a more nuanced role for ILC2 cells at mucosal barriers than previously appreciated. How these different ILC2 subsets are influenced by environmental factors, disease states, and lifestyle choices will be an important next step in our understanding of ILC2 biology and their orchestration of type 2 immunity.


Study design

When possible, a sample size of four to six mice was chosen for mouse studies involving N. brasiliensis infection and cytokine reporters. This sample size was chosen to maximize significance and minimize use of animals (based on previous publications). All such studies were repeated at least twice. All studies involving N. brasiliensis infection and analysis in the lung had a defined endpoint of 5 days (Figs. 2 to 7) given that pulmonary iILC2s peak at day 5 in this setting of type 2 inflammation. Worm clearance was assessed on day 5 when alarmins were provided (Fig. 2). Tuft cells and mucus production were assessed on day 8 corresponding to when worm expulsion occurs in the N. brasiliensis model of helminth infection (Fig. 8). All data including outliers were included in the data analysis and figures with the exception of one mouse that did not receive a productive infection determined by a lack of worms in the intestine at day 5 and lung eosinophilia similar to uninfected controls. The specific number of mice used and experimental repeats performed are included in each figure legend.

The major research objective outlined at the start of the study was to identify whether BATF played a role in ILC2 development and/or function. The initial hypothesis that BATF would influence ILC2s was based on the requirement of this transcription factor for the function of TH2 and TFH cells in helminth clearance. Once a role for BATF was confirmed specifically in IL-25–mediated helminth clearance in both WT and Rag−/− mice, we hypothesized that BATF played a specific role in IL-25–responsive ILC2 subsets as compared with tissue-resident IL-33–responsive ILC2 populations.

In these studies, mouse tissues were used as research subjects. For experimental design, the models of type 2 immunity used were helminth infection and HDM administration. The helminth N. brasiliensis was used, and all infections and methods were performed under controlled laboratory settings. The study was performed so that Batf−/− mice were compared with WT counterparts. For scoring of tuft cells and mucus-producing epithelial cells in intestinal sections, slides were blinded before analysis and scoring.


C57BL/6 Batf−/− (51), Rag1−/−, Arg1YFP (40), and WT CD45.2 mice were originally purchased from The Jackson Laboratory. C57BL/6 IL44get/KN2 and IL13yetcre13/yetcre13 and BALB/c IL44get and IL13Smart were provided by R. Locksley (University of California, San Francisco). Batf−/− mice were crossed onto the following mRNA and/or protein reporter mice: IL-4 (IL44get or IL4KN2), IL-13 (IL13yetcre13), and Arg1 (Arg1YFP). C57BL/6 Batf−/− mice were crossed to Rag1−/− mice to generate Rag−/-Batf−/− mice. In addition, C57BL/6 Batf−/− mice were bred to BALB/c mice for >10 generations and crossed onto the following reporter mice: IL-4 mRNA (IL44get) and IL-13 protein (IL13Smart13). Both male and female mice were used in all experiments except for the bone marrow chimera studies in which only female donors were used to avoid potential rejection. All mice were maintained in specific pathogen–free conditions and in accordance with guidelines established by the Institutional Animal Care and Use Committee at National Jewish Health. Mice used in these experiments were aged 6 to 12 weeks.

Preparation of single-cell suspensions for flow cytometry and sorting

Murine lungs were harvested, rinsed with PBS, finely minced using a razor blade, and digested with collagenase (250 μg/ml), liberase (250 μg/ml), hyaluronidase (1 mg/ml), and deoxyribonuclease (DNase; 200 μg/ml) in RPMI 1640 for 30 min. In some experiments, murine lungs were dissociated using the 37c_m_LDK_1 program on a gentleMACS Octo Dissociator with Heaters (Miltenyi Biotec) with the enzyme cocktail listed above. Cells were then subjected to erythrocyte lysis and filtered before staining.

Murine small intestine was prepared according to a previous report (14). Briefly, 10 cm of small intestine distal to the stomach was isolated, flushed with PBS, and opened longitudinally, and the Peyer’s patches were removed. Tissue was incubated at room temperature, rocking in Hank's balanced salt solution (HBSS) with 5% fetal bovine serum (FBS), 10 mM Hepes, 10 mM dithiothreitol, and 5 mM EDTA for 15 min. This was repeated in fresh solution after vigorous vortexing. Next, intestines were incubated with rocking in HBSS with 5% FBS and 10 mM Hepes for 20 min and then vortexed vigorously. Each intestine was then cut into 3- to 5-mm pieces and incubated in HBSS with 5% FBS, 10 mM Hepes, DNase (200 μg/ml), and liberase (40 μg/ml). Tissue was further dissociated in GentleMACS C tubes, quenched in 10% FBS, and filtered through 80-μm nylon mesh, and then lymphocytes were separated by centrifugation with lymphocyte separation medium (Corning, 25-072).

Staining for flow cytometry

Single-cell suspensions of mouse lung and small intestine were incubated with Fc block (TruStain FcX, BioLegend). Cells were then stained with antibodies to mouse: CD3 (145-2C11), CD4 (RM4-5), CD8 (53-6.7), CD11b (M1/70), CD11c (N418), CD19 (6D5), CD45.1 (A20), CD45.2 (104), CD49b (DX5), CD90.2 (30-H12), CD127 (SB/199), B220 (RA3-6B2), FcεR1α (MAR-1), Gr-1 (RB6-8C5), IL-17RB (9B10), KLRG1 (2F1), NK1.1 (PK136), ST2 (DIH9), TCRγδ (GL3), TER119 or human CD2 (S5.5), or CD4 (S3.5). Cells were resuspended in 2% fetal calf serum (FCS) containing 4′,6-diamidino-2-phenylindole (DAPI). For intracellular transcription factor staining, Fc receptors were blocked and cells were stained with live/dead fixable violet dye (Thermo Fisher Scientific) before fixation and permeabilization (Foxp3 Transcription Factor Staining Kit, eBioscience) according to the manufacturer’s instructions. Cells were then incubated with anti-mouse/human BATF (MBM7C7) or anti-mouse GATA3 (TWAJ). Data were collected on an LSRII (BD) or an LSRFortessa (BD) and analyzed with FlowJo software v10 (Treestar). For analysis of ILCs, cells were gated as lineage negative, which excludes cells labeled with the following antibodies: CD3, CD4, CD8, TCRγδ, CD11b, CD11c, CD19, B220, CD49b, FcεR1α, Gr-1, NK1.1, and Ter119.

RNA-seq and bioinformatics analysis

After IL-33 treatment, Lin, CD90+ KLRG1neg, or KLRG1pos populations from WT Il44get mice were sorted to >85% purity into a solution of 50% FCS, 50% PBS, and 20 U of ribonuclease inhibitor on a BD FACSAria Fusion. RNA was isolated (RNeasy Micro Kit, Qiagen), and libraries were constructed by SMARTer first-strand complementary DNA (cDNA) synthesis (SMARTer Ultra Low Input RNA Kit v4, Clontech) and full-length double-stranded cDNA (dscDNA) amplification by long-distance polymerase chain reaction (LD-PCR), followed by purification and validation. Samples were then fragmented, tagged, and indexed (Nextera XT) and sequenced on a HiSeq 2500 as routinely performed by the National Jewish Health Genomics Facility. FASTQ files were generated using the Illumina bcl2fastq converter (version 2.17), and read quality was assessed using FastQC (version 0.11.5). Nextera TruSight adapters were trimmed with skewer (version 0.2.2) (52) and mapped with STAR aligner (version 2.4.1d) (53) to the GRCm38 assembly of the mouse genome using gene = annotations from Ensembl version 90 ( Gene reads were counted with featureCounts from Subread software package (v1.5.2) (54). Pairwise comparisons were conducted using the Wald test in the DESeq2 package (version 1.81) (55) for the R statistical software (version 3.2.0). P values were adjusted for multiple testing using the method by Benjamini and Hochberg (56). For PCA, heatmaps, and expression plots, raw read counts per sample and gene were normalized to transcripts per million (TPM). PCA was performed using the prcomp function in R version 3.3.2. Heatmaps were generated using the pheatmap function (version 1.0.8) ( in R version 3.3.2. Gene Ontology (molecular function), KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway, and prediction of transcription factors were conducted using DAVID (Database for Annotation, Visualization and Integrated Discovery) version 6.8 functional annotation tool ( PASTAA (predicting associated transcription factors from annotated affinities) was performed using the webtool at and validated by using gene sets within collection 3-Transcription factor targets in MSigDB. Gene set enrichment analysis was performed with the WebGestalt 2019 webtool assessing transcription factor targets within the network functional database ( Transcription factor binding motifs were reported by TRANSFAC v7.4. All pathway analyses used expression data from the KLRG1pos ILCs that were significantly different from KLRG1neg population and at least twofold enriched.

N. brasiliensis infection

N. brasiliensis were prepared as previously described (57), and 500 L3 larvae in 0.9% saline were injected subcutaneously in the center rear flank of mice. Mice were euthanized 5 or 8 days after infection.

Administration of epithelial alarmins

Mice were given intraperitoneal recombinant IL-33 (300 ng; BioLegend), IL-25 (300 ng; R&D Systems), or TSLP (800 ng; R&D Systems) daily for 3 days and rested 1 day, and then lung ILC2 populations were assessed by flow cytometry.

Epithelial alarmin–induced helminth expulsion

Mice were infected with 500 L3 N. brasiliensis larvae. On the day of infection and for an additional 3 days, mice were administered sterile PBS or 500 ng of recombinant IL-33 (BioLegend) or IL-25 (R&D Systems) intraperitoneally. Mice were euthanized 5 days after infection, the small intestine was opened longitudinally and placed in HBSS at 37°C for 1 hour, and worms were manually enumerated.

Administration of HDM

Mice were given 10 μg of HDM extract in 50 μl of PBS daily for 4 days by oropharyngeal aspiration, rested 1 day, and euthanized the following day for analysis of pulmonary ILC2 populations by flow cytometry as described.

Generation of bone marrow chimeras

CD45.1/CD45.2 WT mice were irradiated with one dose of 9 Gy and reconstituted with a 1:1 ratio of 2 × 106 CD45.2 WT and 2 × 106 CD45.1 Batf−/− bone marrow cells by tail vein injection. Mice were maintained on a diet containing trimethoprim and sulfamethoxazole (SEPTRA) for 4 weeks before being switched to normal chow and were rested for a total of 8 weeks before N. brasiliensis infection.

Treatment with FTY720

FTY720 was purchased from Sigma-Aldrich and administered intraperitoneally at ~1 μg/g mouse weight in sterile 0.9% saline. FTY720 was given on the day of N. brasiliensis infection or IL-33 administration and re-administered 2 and 4 days after infection/treatment.

Immunohistochemistry and quantification of DCLK+ and MUC2+ cells

Mice were euthanized 8 days after infection, and the first 10 cm of the duodenum and jejunum was flushed with PBS, fixed in 4% paraformaldehyde for 4 hours, washed in PBS overnight, and placed in 30% sucrose for 6 hours before embedding in optimal cutting temperature (OCT) medium. Cryosections (μm) for both DCLK and MUC2 staining were cut and placed on slides for histology. For detection of DCLK+ cells, slides were incubated with 1% H2O2 in PBS for 45 min, TNB (2-nitro-5-thiobenzoic acid) and Fc block for 30 min, and biotin/avidin for 30 min each. Tissues were stained with 1:1000 rabbit anti-DCLK (Abcam, 31704) in TNB followed by 1:500 biotinylated donkey anti-rabbit antibody (Jackson ImmunoResearch), streptavidin horseradish peroxidase (PerkinElmer), 1:200 555-tyramide (PerkinElmer), and DAPI and mounted with a coverslip (Vectashield). For detection of MUC2+ cells, tissues were blocked with 5% bovine serum albumin and Fc block and stained with 1:200 anti-mouse MUC2 (Santa Cruz Biotechnology, H-300), followed by goat anti-rabbit 647 (Life Technologies), and DAPI. Images were obtained on a Marianas microscope. Images were blinded, and the number of DCLK+ cells per millimeter of villus was quantified using Fiji (ImageJ), where only intact villi were assessed. A similar blinded quantification was used for MUC2 staining, but because mucin is released from the cell, the number of MUC2+ foci per millimeter of villus was assessed by enumerating only the regions where discrete MUC2 staining originates.

Statistical analyses

Data are mean ± SEM from the specified number of mice combined from multiple experiments unless otherwise noted. Statistical calculations were performed with Prism 7.0 software (GraphPad). Comparison of three or more groups was performed using a one-way analysis of variance (ANOVA) test, followed by Tukey post hoc analysis. Comparison of data between two groups was analyzed using a two-tailed unpaired t test to determine statistical significance. In the figures indicated, N.S. designates nonsignificant statistical differences. In all figures, only statistical differences between WT and Batf−/− mice in either naïve, infected, or treated groups are displayed. Statistically significant differences are indicated by *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

RNA-seq data availability

The data and software code used to generate and support this study will be available from corresponding authors upon request. RNA-seq data are freely available in the NCBI Gene Expression Omnibus ( under the accession number GSE125816.


Fig. S1. Gating strategy for sorting cells for RNA-seq.

Fig. S2. Gating strategy for murine pulmonary and SILP ILCs.

Fig. S3. Quantification of KLRG1neg and KLRG1mid pulmonary ILC2s.

Fig. S4. Induction of pulmonary ILC2s by alarmins.

Fig. S5. Bone marrow chimeric mice reveal that KLRG1high ILC2s require BATF and expand even in the absence of infection.

Fig. S6. Quantification of type 2 cytokine protein reporter production by lung ILC2s.

Fig. S7. FTY720 treatment does not affect nILC2s in the lung or any ILC2 population in the SILP.

Fig. S8. Graphical summary.

Table S1. Raw data spreadsheet.


Acknowledgments: We thank R. Locksley for mouse reporter strains; S. Sobus and J. Loomis for assistance with flow cytometry, sorting, and microscopy; M. Chapman and K. Walton for assistance and generation of RNA-seq libraries; E. Gelfand for access to gentleMACS dissociator; C. Jakubzick, S. Gibbings, M. Dell’Aringa, I. Brown, M. Schoenbach, A. Miller, and J. Barerra for technical and laboratory support; and M. Rosenbaum, S. Crow, M. Crumrine, and the NJH Biological Resource Center for excellent animal care and support. Funding: This work was supported by NIH grant AI119004 (to R.L.R.) and Windsweep Farm Fellowship (M.M.M.). Author contributions: Conceptualization: M.M.M., K.B., and R.L.R.; validation: M.M.M. and P.S.P.; formal analysis: B.P.O., T.D., and P.S.P.; investigation: M.M.M., P.S.P., and K.B.; statistical analysis: M.M.M., P.S.P., and T.D.; writing the original draft: M.M.M. and R.L.R.; writing, reviewing, and editing: all authors. Competing interests: The authors declare that they have no competing interests. Data and materials availability: The data used to generate and support this study will be available from corresponding authors upon request. RNA-seq data are freely available in the NCBI Gene Expression Omnibus ( under the accession number GSE125816. All mice are either commercially available or available under a material transfer agreement (MTA). For IL44get C57BL/6 mice that are under a current MTA, please contact the corresponding authors to initiate a request for access to these mice.

Stay Connected to Science Immunology

Navigate This Article