Multiscale 3D genome organization underlies ILC2 ontogenesis and allergic airway inflammation

Abstract

Innate lymphoid cells (ILCs) are well-characterized immune cells that play key roles in host defense and tissue homeostasis. Yet how the 3-dimensional (3D) genome organization underlies the development and functions of ILCs is unknown. Herein, we carried out an integrative analysis of the 3D genome structure, chromatin accessibility, and gene expression in mature ILCs. Our results revealed that the local 3D configuration of the genome is rewired specifically at loci associated with ILC biology to promote their development and functional differentiation. Importantly, we demonstrated that the ontogenesis of ILC2s and the progression of allergic airway inflammation is determined by a unique local 3D configuration of the region containing the ILC-lineage defining factor Id2, which is characterized by multiple interactions between the Id2 promoter and distal regulatory elements bound by the transcription factors GATA-3 and RORα, unveiling the mechanism whereby the Id2 expression is specifically controlled in group 2 ILCs.

Innate lymphoid cells (ILCs) are effector innate cell populations that contribute to host defense against pathogens, metabolic homeostasis, and tissue repair but also play key roles in diseases such as cancer, asthma, or colitis¹. ILCs are known to be the innate counterpart of T lymphocytes and have been classified into three groups based on the transcription factors (TFs) and cytokines they express². Group 1 ILC is composed of natural killer (NK) cells and ILC1s, and are considered the innate counterparts of cytotoxic CD8⁺ T cells and CD4⁺ T helper 1 (Th1) cells, respectively. Group 2 ILC comprises ILC2s which, similarly to Th2 cells, produce IL-5 and IL-13 and express the TF GATA-3. Finally, group 3 ILC, similarly to Th17, requires the TF RORγt for their development and functions. Extensive work over the last decade has led to the identification of the TF networks that control the development and functions of each group of ILCs^2–4. However, how the three-dimensional (3D) organization of the genome contributes to the regulation of their development, cell identity, and functions remains unknown.

At large scales, chromosomes are segregated into regions of chromatin enriched in active genes and regions associated with repressive chromatin markers where genes tend to be silenced⁵. This higher level of chromatin compaction is referred to as compartment organization⁶. Topologically associating domains (TADs) represent the next layer of genome organization in which sub-megabase chromatin regions show a significant preference for internal interactions (intra-TAD) as compared to interactions with loci outside the given region (extra-TAD)⁷. In this manner, regulation of one or several genes can be coordinated by allowing different sets of promoter-enhancer interactions to occur. Finally, short- and long-range interactions between distal genomic regulatory elements (REs) such as loops and stripes, constitute the finest layer of spatial organization of the genome with critical roles in chromatin compaction, folding, and gene expression regulation. Despite our increased understanding of the 3D genome conformation, how large-scale 3D configurations of the genome or the structure of local interactions at specific genomic loci determine ILC development and functions is yet to be elucidated.

Results

The large-scale 3D genome structure of innate lymphoid cells

To establish how the 3D topology of the genome contributes to the development and functional specification of each ILC group, we coupled RNA-seq and ATAC-seq analysis to Hi-C. As high-quality Hi-C datasets require high numbers of cells with >90% viability, we selected the following representative cell types of each group of ILCs to perform ultra-deep Hi-C and RNA-seq analysis: mouse splenic NK cells (group 1), in vitro expanded bone marrow ILC2s (group 2), and the ILC3-like cell line MNK3 (group 3) since ILC3 numbers are limited and cannot be expanded in vitro⁸. Previous works demonstrated the high degree of similarity between primary ILC3 and MNK3 cells at both the transcriptional and the chromatin accessibility levels^9,10. Indeed, we estimated the similarity between MNK3 cells and primary ILC3s using principal component and differential accessibility analysis of ATAC-seq profiles and determined that 92% of accessible DNA regions overlap between primary ILC3s and MNK3 cells (Fig. 1a and Extended Data Fig. 1a–d). Moreover, MNK3 cells express classical functional markers associated with ILC3 functions in culture (Extended data Fig. 1e), indicating that these cells are a suitable experimental system to query the 3D genome architecture of ILC3s.

Fig. 1|. Large-scale 3D genome organization in ILCs.

a, Principal component analysis (PCA) of chromatin accessibility profiles of NK cells (group 1), bone marrow (BM) and lung ILC2s (group 2), small intestine (SI) CD4⁺ and NCR⁺ ILC3s, and MNK3 cells (group 3).

b, Genome-wide compartment distribution in CLPs and ILCs. Regions in the A and B compartments correspond to positive (red) and negative (blue) values in the PC1 bar. Black/Yellow bar indicates regions that overlap or not with genes.

c, A/B compartment distribution of genes known to play critical roles in CLPs and ILCs.

d, ‘Upset’ plot of the distribution of boundary elements. (bottom) Gray bars represent the total number of boundaries called in group 1, 2, and 3 ILCs. (middle) Number of boundaries shared by one, two or three ILC groups are represented by black bars. (top) Insulation scores of boundaries in each ILC group per intersection set as defined in the bottom panel. Box shows dataset quartiles and whiskers the distribution range. Dots represent outliers determined by the inter-quartile range.

e, TAD containing the ILC1 associated gene Eomes. TADs are delineated by boundaries conserved in all groups of ILCs (black rectangles) while non-conserved boundaries define sub-TADs (colored rectangles). Solid and dashed lines represent TADs and sub-TADs, respectively. Scale bar represents normalized contact frequency.

f, Scatterplot of the cross-boundary ratios of conserved TADs. Color gradient indicates the bias in the fold change of the TAD cross-boundary ratio in the corresponding comparison. Gray dots represent TADs with absolute log fold change <0.25. TADs containing genes involved in ILC biology are highlighted.

g, Heatmap of the top 25 differentially expressed genes in TADs with the highest comparative cross-boundary ratios in (left) group 1 ILCs, (middle) group 2 ILCs, and (right) group 3 ILCs. Genes with roles in each group of ILCs are denoted.

h, Transcription factor binding motif analysis over chromatin accessible regions located within TADs with the highest (log fold change >0.25) comparative cross-boundary ratios in (left) group 1 ILC, (middle) group 2 ILC, and (right) group 3 ILC. Numbers indicate the -log10(p-value) of denoted binding motifs (Two-sided binomial test).

At large scales, chromatin segregates into A compartments that often contain active gene-rich euchromatin, and B compartments associated with repressive histone marks mostly delineating inactive heterochromatin regions^6,11. Thus, we first assessed the distribution of A/B compartments in each ILC group and in common lymphoid progenitors (CLPs) to establish the extent to which the compartment structure of each ILC subset changes in relation to their earliest progenitor¹² (Extended Data Fig. 2a–b and Supplementary Table 1). Overall, ILCs and CLPs show similar distributions of compartment states as most of the genome (71.8%) do not change compartment status during the transition from CLPs to mature ILCs (Fig. 1b). Importantly, we observed that key TF driving the commitment to the ILC lineage (Id2, Nfil3, Tox, Tcf7, Gata3) already reside in conserved A compartment regions at the CLP stage, suggesting that they might be primed to be activated in these multipotent lymphoid precursors (Fig. 1c and Extended Data Fig.2c). Nevertheless, we also observed that around 28% of genomic regions flipped compartment status in at least one ILC subset and that these regions display gene expression patterns concordant with the compartment organization in each ILC group (Extended Data Fig. 2d–g and Supplementary Table 2). Altogether, these results indicate that most of the largescale 3D genome organization of ILCs is already pre-established at the CLP stage and thus, ILC specification and functional differentiation might be instead dependent on local chromatin remodeling at the mega and sub-megabase scale.

Topologically Associating Domains (TADs) are fundamental units of 3D genome organization that span up to a few megabases in size. TADs are demarcated by two genomic elements usually enriched in CTCF binding motifs that act as boundaries^7,13–15. The preference for local associations within TADs facilitates gene regulation by insulating chromatin regions containing promoter-enhancer interactions. Thus, to investigate the influence of the 3D genome organization at the sub-mega scale in ILCs, we next examined the TAD structure and its contribution to gene expression programs in each ILC group.

First, we looked at the distribution of boundary elements in ILCs and observed that approximately 36% of the boundaries were specific to one ILC subset. The remaining boundaries (64%) were, thus, shared by at least two ILC groups and more than half of these were conserved in all ILC groups (Fig. 1d). These observations indicate an important degree of conservation in the 3D genome organization at the megabase level across ILC groups. We then hypothesized that TADs containing genes associated with a specific ILC subset program should have higher frequencies of intra-TAD interactions when compared to the other two groups of ILCs. To perform a direct comparison, we sought to calculate a cross-boundary ratio which expresses the relative odds of loci within a TAD to interact with one another or with elements outside the TAD boundaries^12,16. To facilitate this comparison, we defined TADs as the genomic region between two conserved boundaries (Fig. 1e and Extended Data Fig. 3a–c). Interestingly, we found a shift towards higher cross-boundary ratios in TADs from group 2 ILCs (Extended Data Fig. 3b and Supplementary Table 3). More importantly, we observed that the TADs with comparatively higher cross-boundary ratios in each group of ILCs (log fold change > 0.25) contain genes with prominent biological roles in the given ILC program such as Eomes and Ikzf1 in group 1 ILCs, Il2ra and Rora in group 2 ILCs, or Kit in group 3 ILCs (Fig. 1f). Hereafter, we will refer to these TADs as ILC-subset associated TADs. In concordance, among the top 25 most differentially expressed genes within ILC-subset associated TADs, there are genes known to be critical for the development or functions of each ILC group (Fig. 1g). Moreover, each set of subset associated TADs also displays enrichment in motifs of TF with canonical roles in the development or functions of the corresponding ILC group (Fig. 1h). Altogether, these results indicate that, at the sub-mega base scale, the 3D chromatin organization contributes to the biology of each ILC group through selective enrichment of intra-TAD interactions.

The local 3D genome structure of innate lymphoid cells

Short- and long-range interactions between promoters and enhancers via looping represent the finest scale of spatial arrangement that directly relates to gene activity. As such, specific loop interactions and their combination determine cell identity and functions. We, therefore, examined how local interactions within ILC-subset associated TADs determine ILC-subset specific gene expression programs.

We first identified DNA loops that link or contain gene bodies (hereafter referred to as gene loops) within each set of ILC-subset associated TADs, and then we ranked those genes based on the number of loops they are involved in (Fig. 2a–b and Supplementary Table 4). Importantly, we observed that critical genes for the development or functions of each group of ILCs reside in highly DNA interacting neighborhoods as reflected by the number of loops around them (Fig. 2a–b and Supplementary Tables 5–7). Next, we sought to compare the average strength of each gene loop across the three groups of ILCs. To do so, we computed the fold change of its average contact frequency between every pair of cell types. Our comparisons show that genes specific to an ILC program have stronger (average) loop interactions in the corresponding ILC group (Fig. 2c) and that the strength of those loops correlates with their expression level, indicating that loop strength is critical for establishing gene expression programs in ILCs (Extended Data Fig. 3d). In addition, we also observed that regions that form stripes, which represent genomic loci that interact with entire contiguous regions^17,18, contain genes involved in ILC development and maintenance such as Id2, Gata3, Eomes, or Rorc (Fig. 2d, Extended Data Fig. 3e and Supplementary Table 8–10). Such stripes are thought to form via loop extrusion and are likely to constitute an additional mechanism of gene regulation at the local scale (Fig. 2e and Extended Data Fig. 3e). Taken together, these results indicate that the loop and stripe structure of key genes critical for ILC development and functions, correlate with their transcriptional activity and, in this way, local 3D interactions at the sub-TAD scale appear to contribute to ILC-subset specific programs.

Fig. 2|. Differential local 3D structure at key loci correlates with the gene expression programs of each ILC subset.

a, Rank plot of the number of loops that link or contain gene bodies in (left) group 1 ILC, (middle) group 2 ILC, and (right) group 3 ILC. Genes with known roles in each group of ILCs are denoted.

b, Examples of loop structures around ILC subset specific genes and Id2 in all ILC groups. Blue, green and red arcs refer to loops present in group 1, group 2, and group 3 ILC, respectively.

c, Pairwise comparisons of log fold changes of average gene loop strength in each group ILC vs the others. Dots represent genes expressed in (left) group 1 ILC, (middle) group 2 ILC, and (right) group 3 ILC. Color intensity indicates low-to-high gene expression levels in log(RPKM+1) units.

d, Rank plot of the stripiness of stripes detected in group 1 (left), group 2 (middle) and group 3 ILC (right). Genes with known roles in each group of ILCs are highlighted.

e, Example of the differences in stripe structure at the Id2 locus between ILCs. Pairwise visualization between ILCs of the contact heatmap of the 1 megabase region around the Id2 locus. Color boxes highlight the significant (p < 0.05) stripe imputed in the (blue) group 1, or (green) group 2, or (None) group 3 ILC contact frequency map. Scale bar in heatmap represents normalized contact frequency.

f, Visualization of the contact heatmap at the Id2 locus of bone marrow ILC2s overlapped with chromatin accessible regions determined by ATAC-seq, H3K27ac deposition (Gray and blue overlapped tracks, respectively) as well as GATA-3 and RORα binding determined by ChIP-seq. (Green and red overlapped tracks, respectively). Black arcs represent loop interactions detected in the heatmap. Shaded boxes delineate the locus control region 1 (LCR1 - Red shaded box) and the locus control region 2 (LCR2 - Blue shaded box). Dashed lines represent sub-TADs, solid lines represent TADs. Arrowheads indicate the position of conserved boundaries. Scale bar in heatmap represents normalized contact frequency.

3D organization of the Id2 locus underlies ILC2 ontogenesis

Commitment to the ILC lineage and maintenance of its identity is determined by the transcriptional repressor Id2¹⁹. ID2 binds to, and functionally inactivates a set of transcriptional activators such as E2A, E2–2, and HEB, which are important for adaptive lymphocyte development. As such, Id2 expression is now considered a hallmark of all ILC subsets in mice and humans.

Our integrative analysis revealed that the local spatial organization of the region containing the Id2 locus as well as its expression level is unique in each group of ILCs, indicating that Id2 expression is precisely controlled by the interplay between ILC-subset specific TFs and the remodeling of the 3D structure of the chromatin during development²⁰. Interestingly, Group 2 ILCs expressed higher levels of Id2 than other ILC subsets (Extended Data Fig. 3d). Moreover, the 3D architecture of the Id2 locus in group 2 ILCs shows multiple unique characteristics that correlate with the high expression of this transcriptional regulator in these cells: first, the cross-boundary ratio of the TAD containing Id2 was higher in ILC2s when compared to group 1 and 3 ILCs. Second, Id2 was among the top 25 most significantly expressed genes located only within TADs associated with ILC2s (Fig. 1g). Third, the mean strength of DNA loops that link or contain the Id2 gene body was stronger in ILC2s when compared to other ILC groups (Fig. 2c). Finally, although shorter than in group 1 ILC, the stripe at the Id2 locus was stronger in group 2 ILCs as reflected by its higher stripiness score (Fig. 2d). Indeed, Id2 was among the top 10 genes by stripiness in ILC2s (Fig. 2d). Altogether, these observations indicate that the frequency and strength of interactions between the Id2 gene body and downstream distal elements within the TAD, is higher in group 2 ILCs when compared to other ILC subsets. However, how the 3D folding of the Id2 locus underlies ILC2 ontogenesis by promoting high levels of Id2 expression remains unknown. Therefore, we hypothesized that a combination of long-distance DNA interactions between the Id2 promoter and cis-RE bound by TFs specifically required for ILC2 development control Id2 expression in these cells.

To establish which unique features of the 3D folding of the Id2 locus determine the expression of this lineage-defining factor in ILC2s, we examined chromatin accessibility, deposition of histone marks associated with active enhancers (H3K27ac), and loop interactions involving the Id2 locus and distal DNA elements in the three groups of ILCs (Extended Data Fig. 4a). Moreover, we used previously generated ChIP-seq datasets of ILC2s to identify DNA elements bound by RORα and GATA-3 two key TFs required for ILC2 development^4,21–26 (Fig. 2f). Using this approach, we identified a region located approximately 125 kb downstream of Id2 containing two segments highly enriched in H3K27ac, bound by RORα and GATA-3, which interacted with the Id2 promoter through DNA loops, specifically in group 2 ILCs (Hereafter referred to as locus control region 1 - LCR1) (Fig. 2f). We also identified a second region located 65kb downstream of Id2 (Hereafter, LCR2) characterized by an enhanced deposition of H3K27ac and a highly accessible region that interacted with Id2 in group 2, but not in other ILCs (Fig. 2f and Extended Data Fig. 4a). In addition, motif analysis showed that binding sites for key TFs controlling ILC2 development and functions, such as ETS1, GATA-3, RUNX, RAR:RXR, and RORα, are present within the LCR1 and LCR2 (Extended Data Fig. 5a and Supplementary Table 11). Altogether, these analyses suggest that distal DNA interactions between the Id2 promoter and cis-REs within the LCR1 or LCR2 might be required for the control of Id2 expression in ILC2s.

To establish whether the LCR1 or the LCR2 contain cis-REs necessary for the development, homeostasis or functions of ILC2s, we generated mice where we individually deleted these loci using the CRISPR/Cas9 system. We first determined the frequencies and numbers of each ILC group in their canonical tissue locations, in LCR1- and LCR2-deficient mice. Strikingly, ILC2s were dramatically reduced in the tissues examined such as lungs, visceral adipose tissue (VAT), skin, small intestine lamina propria (siLP) and bone marrow (BM) of LCR1^−/−, but not in LCR2^−/− mice (Fig. 3a–b). In contrast, group 1 ILCs (NK cells and ILC1s), ILC3s, or any other Id2 expressing myeloid or lymphoid cell remained unaltered in both mouse strains, indicating that the LCR1 might be specifically required for ILC2 development (Fig. 3c and Extended Data Fig. 5b–d). To confirm the ILC2-specific defect in LCR1-deficient mice, we crossed LCR1^−/− animals with Arg1-YFP reporter mice, which labels mature and developing ILC2s^27,28. As expected, ILC2s were absent in the lungs and BM of LCR1^−/−; Arg1-YFP mice (Extended Data Fig. 5e). In concordance with previous reports^29,30, this defect was associated with decreased eosinophil frequencies and numbers, as well as reduced levels of IL-5 in the serum (Extended Data Fig. 5f–g). Altogether, our results indicate that distal DNA interactions between the Id2 promoter and cis-REs within LCR1 are specifically required for the development or homeostasis of ILC2s.

Fig. 3 |. Multiple long-distance interactions between the Id2 promoter and the LCR1 specifically control ILC2 homeostasis or development.

a, Representative flow cytometry plots for lung and visceral adipose tissue (VAT) ILC2s.

b, Quantification of ILC2 numbers in the lung, VAT, skin, small intestine lamina propria (siLP), and bone marrow (BM) of wild-type, LCR1^−/−, and LCR2^−/− mice. siLP and skin are a pool of two independent experiments (skin, n=9 vs 9; siLP, n=8 vs 8). VAT and lungs are a pool of at least three independent experiments (VAT, n=11 vs 10; lungs, n= 15 vs 16 and 15 vs 17). BM is a pool of four experiments (BM, n= 15 vs 14). All experiments were repeated at least three times. p-values: ns = not significant, *** = p≤0.0005, **** = p≤0.0001 (Mann-Whitney U two-tailed test).

c, Quantification of ILC1s and NK cells in the liver, and siLP ILC3 numbers in wild-type, LCR1^−/−, and LCR2^−/− mice. Each histogram is a pool of two independent experiments, except for liver ILC numbers in LCR2^−/− mice which is one experiment. Experiments were repeated at least two times. (Liver, n=11 vs 11 and 4 vs 4; ILC3s, n=12 vs 13 and 8 vs 8). p-values: ns = not significant (ILC3, Mann-Whitney U two-tailed test; liver ILCs, two-way ANOVA with multiple comparisons and Bonferroni correction).

d, Schematic representation of the RORα and GATA-3 binding sites deletions in the mouse genome.

e, Quantification of BM ILC2 numbers in wild-type, LCR1^−/−, Gata3_BS^−/−, Rora_BS^−/−, and Rora_Gata3^dKO mice. Histogram is a pool of at least two independent experiments. Quantification was repeated at least two times (n= 18 wild-type; 3 LCR1^−/−; 8 Gata3_BS^−/−; 6 Rora_BS^−/− and 4 Rora_Gata3^dKO mice). p-values: ns = not significant, * = p≤0.05, ** = p≤0.01 (One-way ANOVA with multiple comparisons and Bonferroni correction).

f, Quantification of lung ILC2s in Rora_Gata3^WT and Rora_Gata3^dKO mice (n = 5 Rora_Gata3^WT and 4 Rora_Gata3^dKO mice). Data are a pool of two independent experiments. p-values: ns = not significant, * = p≤0.05 (Mann-Whitney U two-tailed test).

All error bars indicate Standard Error of the Mean (SEM).

The LCR1 contains two distinct cis-REs bound by the TFs RORα and GATA-3, which directly interact with the Id2 promoter in ILC2s. Thus, to determine the individual contribution of each cis-RE to the LCR1^−/− phenotype, we deleted the RORα and GATA-3 binding sites (BS) individually from the mouse genome and determined ILC2 numbers in the BM (Fig. 3d). Surprisingly, ILC2 development was unaffected in both mouse strains when compared to WT and LCR1^−/− mice (Fig. 3e). However, deletion of both the RORα and GATA-3 BS in cis from the mouse genome (referred to as Rora_Gata3^dKO) lead to a complete absence of ILC2s in the BM and lungs (Fig. 3f). Altogether, our results indicate that specific distal interactions between the Id2 promoter and unique cis-REs are required for group 2 ILC development.

cis-REs within LCR1 regulate the early development of ILC2s

ILC2s develop in the BM and fetal liver (FL) from CLPs, which subsequently differentiate into intermediate progenitors that eventually give rise to mature ILCs^31–33. More specifically, CLPs give rise to α₄β₇⁺ lymphoid progenitors (α-LPs) that differentiate into common helper-like ILC progenitors (ChILPs)^34–37. ChILPs are the precursors of a heterogeneous population of ILC precursors (ILCp) that includes ILC2 precursors (ILC2p), which is the latest developmental stage before mature ILC2s arise. To dissect when during development the LCR1 is required for ILC2 generation, we performed scRNA-sequencing on 2309 wild-type (WT) and 3729 LCR1^−/− Lin⁻, CD45⁺, CD127⁺ BM cells, which contain ILC2s and their progenitors. Using graph-based analyses we identified 15 different clusters expressing the Il7r (Fig. 4a and Extended Data Fig 6a). In contrast, only clusters 3–7, 9, 11, and 12 expressed Id2 and other genes critical for early ILC development and specification (Fig. 4b and Extended Data Fig. 6a). Next, we annotated each cluster based on a combination of genes that delineate early (Flt3, Notch1, and Bcl11a) and common ILC progenitors (Zbtb16, Tcf7, and Tox) (Fig. 4b). In addition, committed ILC subsets were assigned by the expression of TFs and surface molecules characteristic of each ILC subset such as Tbx21, Gata3, and Rorc for group 1, 2 and 3 ILC, respectively (Fig. 4b–c). Indeed, hierarchical clustering of this dataset revealed two major branches separating Id2-expressing clusters from clusters 1, 8, 13, 14 and 15 which consist of early lymphoid progenitors (Extended Data Fig. 6a–b).

Fig. 4|. Multiple long-distance interactions between the Id2 promoter and the LCR1 determine early ILC2 development.

a, Uniform Manifold Approximation and Projection (UMAP) of the scRNA-seq data of 2309 LCR1^+/+ and 3728 LCR1^−/− live, CD45⁺, Lineage⁻, CD127⁺ bone marrow (BM) cells.

b, DotPlot visualization of genes that characterize early ILC, unspecified ILC progenitors (ILCp), or clusters with signatures associated with ILC1s, ILC2s, and ILC3s (denoted as Type 1, Type 2 and Type 3, respectively). Dot size represents the fraction of cells in a given cluster that express the listed gene. Color gradient represents the mean expression of the listed gene in a given cluster.

c, Heatmap displaying normalized-scaled expression levels of marker genes of the clusters identified in (a). Genes included correspond to the pool of top 10 markers expressed by each cluster. Genes with known roles in ILC biology are highlighted.

d, Pie chart of distribution and size and composition of clusters in (a) that express Id2 (C3–7, C9, C11–12). Inner ring represents relative cluster size. Outer ring represents the fraction of LCR1^+/+ (white) and LCR1^−/− (red) cells in each cluster.

e, Pseudotime plot illustrating the developmental trajectories of LCR1^+/+ and LCR1^−/− CD45⁺, Lineage⁻, CD127⁺ BM cells. The star (*) represents the selected starting point at the earliest progenitor identified – C8).

f, Quantification of CLP, α-LP, ChILP, ILCp, ILC2p, ILC3p and ILC2s from the BM of LCR1^+/+ and LCR1^−/− mice. Data represents the pool of at least three independent experiments, except for BM ILC3p numbers which is one experiment. (CLP, aLP, ILCp, ILC2p, n=15 vs 14; ChILP, n= 8 vs 8; ILC3p, n=5 vs 5). Error bars indicate Standard Error of the Mean (SEM); and p-values: ns = not significant, * = p≤0.05, ** = p≤0.01 (Mann-Whitney U two-tailed test).

g, ATAC-seq peaks at the Id2 locus during ILC2 development. Red box demarcates the LCR1, which contains the RORα and GATA-3 binding sites. CLP: common lymphoid progenitors; α-LP: alpha-lymphoid progenitors; rEILP: redefined early innate lymphoid progenitors; ChILP: common helper innates lymphoid progenitors; ILCp: innate lymphoid cell precursors; ILC2p: ILC2 precursors; ILC3p: ILC3 precursors.

Strikingly, by comparing the relative frequency of the different Id2-expressing clusters, we observed a drastic reduction in the proportions of cells composing ILC2s (C3), and immature ILC2s (C5) in LCR1-deficient mice when compared to WT animals (Fig. 4d). Moreover, transitioning ILC2s (C12) were only present in LCR1-deficient animals. Surprisingly, we also observed a 10-fold increase in the relative frequency of ILC3 precursors (C9) in LCR1^−/− mice (Fig. 4d and Extended Data Fig. 6c). In concordance, reconstruction of early ILC development in WT and LCR1-deficient mice through pseudo-time analysis confirmed the developmental trajectory from CLPs (C8) to ILC2s (C3) (Fig. 4e). Moreover, it demonstrated a developmental blockade at the ILC2p stage as well as a skewing towards cells with gene signatures associated with ILC3p (C9) and ILC1p (C11) in LCR1^−/− mice (Fig. 4a,e).

To corroborate our scRNA-seq results, we then quantified ILC progenitors and ILC2 populations in the BM of WT and LCR1-deficient mice using flow cytometry (Fig. 4f and Extended Data Fig. 6d). Indeed, we observed a significant reduction in the frequency and number of ILC2p and almost complete absence of BM ILC2s in LCR1^−/− mice, while ILCp population remained intact (Fig. 3c and Fig. 4f). Since ILC development starts during embryogenesis, we also quantified ILC2 progenitors in LCR1^+/+ and LCR1^−/− E15.5 embryos²⁸. Similarly, ILC2p were drastically reduced in fetal liver, fetal intestine, and fetal lungs of LCR1^−/− E15.5 embryos (Extended Data Fig. 6e). Moreover, the number of ILC3p was significantly increased in the BM of LCR1^−/− mice (Fig. 4f and Extended Data Fig. 6c). We then reasoned that the cis-REs bound by RORα and GATA-3 within the LCR1 should become accessible during early ILC development. Indeed, we observed that the RORα and GATA-3 BS only become fully accessible in ILCp (Fig. 4g). Altogether, our data strongly indicates that two long-range DNA interactions between the Id2 promoter and cis-REs within the LCR1 upon commitment to the ILC lineage are required for early ILC2 development.

Since ILC2s are absent in LCR1^−/− mice, we hypothesized that the cis-REs within the LCR1 might regulate Id2 expression during early ILC2 development. To address this possibility, we performed differential gene expression analysis in the Id2-expressing clusters. Indeed, Id2 was among the top downregulated genes in LCR1^−/− ILC precursors when compared to WT (Fig. 5a and Extended Data Fig. 6f). Interestingly, among all the TFs known to be required for ILC2 development (Id2, Bcl11b, Rora, Gata3, Zbtb16, Tox, Ets1, and Gfi1) only Id2 was downregulated throughout the ILC developmental trajectory in LCR1-deficient BM cells when compared to WT (Fig. 5b and Extended Data Fig. 6g). Moreover, gene ontology (GO) analysis showed that decreased levels of Id2 expression in LCR1-deficient developing ILCs were associated with pathways that positively regulate adaptive lymphocyte development and cell death (Fig. 5c). Of note, Id2 expression was slightly decreased in CLPs and throughout ILC2 development in LCR2-deficient mice (Extended Data Fig 6h). These results indicate that the activity of cis-REs within the LCR1 is necessary for adequate Id2 expression upon commitment to the ILC2 lineage, and that Id2 does not seem to control the expression of other key TFs involved in ILC2 development.

Fig. 5|. The LCR1 specifically controls group 2 ILC development through Id2 expression regulation.

a, Volcano plot displaying upregulated (red rectangle) and downregulated (blue rectangle) genes in LCR1^+/+ vs LCR1^−/− for clusters from Fig. 4a representing ILCp (C7), ILC2p (C3), and ILC2s (C5) populations.

b, Savitzky-Golay smoothing curve of the normalized expression levels of Id2, Gata3 and Rora along the pseudotime axis determined in Fig. 4e. P-value corresponds to Mann-Whitney U two-tailed test between smoothed signals from LCR1^+/+ (black curve) and LCR1^−/− (red curve) mice.

c, GO terms associated with differentially upregulated genes in cluster 7 (ILCp). For each term, the corresponding FDR is less than 0.05 (Two-sided binomial test). FDR: false discovery rate.

d, (top) Representative flow cytometry gating strategy of bone marrow (BM) ILC2s from lethally irradiated CD45.1⁺ wild-type mice (host) transplanted with CD45.2⁺ LCR1^−/− BM transduced with retroviral particles encoding Id2 and the GFP (Id2_RV) or an empty vector expressing only the GFP (Empty_RV). (bottom) Quantification of lung and BM ILC2s generated after BM transplantion of LCR1^−/− (CD45.2⁺) transduced cells with an empty retroviral vector (Empty_RV - green dots), or the Id2-encoding retroviral vector (Id2_RV - blue dots) into a CD45.1⁺ host. White dots represent BM ILC2 numbers in untransduced (CD45.2⁺) LCR1^+/+ animals. Data is a pool of two independent experiments that was repeated five times. (n= 5 LCR1^+/+; 8 LCR1^−/−Empty_RV; 12 LCR1^−/− Id2_RV). Error bars indicate Standard Error of the Mean (SEM). p-values: ns = not significant, ** = p≤0.01, *** = p≤0.0005, **** = p≤0.0001 (One-way ANOVA with multiple comparisons and Bonferroni correction).

To confirm that Id2 downregulation is the central defect in LCR1-deficient developing ILC2s, we transduced CD45.2⁺ LCR1^−/− BM progenitor cells with retroviral particles encoding Id2 and the green fluorescent protein (Id2-GFP), or an empty vector (Empty-GFP) which were then transferred to lethally irradiated congenic CD45.1⁺ mice. Eight weeks post engraftment we observed that mice transplanted with LCR1^−/− BM cells transduced with the Id2-GFP retroviral vector gave rise to ILC2s in the BM and lungs. However, mice that received LCR1-deficient BM cells transduced with the empty retroviral vector failed to rescue the ILC2 developmental blockade (Fig. 5d and Extended Data Fig. 6i). These results indicate that ILC2s deficiency in LCR1-deficient mice occurs as a consequence of insufficient Id2 expression in ILC progenitors.

cis-REs within the LCR1 promote allergic airway inflammation

Allergic airway inflammation (AAI) is characterized by airway hyperactivation induced by eosinophilic infiltration, type 2 cytokines, and mucus hyperproduction^38–40. Early responses during AAI are driven by local activation of ILC2s upon allergen exposure which in turn promotes CD4⁺ T cell recruitment and Th2 polarization. Indeed, recent novel genetic tools have shown a non-redundant role for ILC2s during AAI^26,41,42. In addition, it has been reported that lung ILC2s are composed of cells that seed tissues at steady-state and cells that expand or are de novo generated during inflammatory responses⁴³. However, whether the development and functions of ILC2s during inflammation are determined by a similar or different 3D architecture of the Id2 locus is unknown.

To address the role of LCR1 during AAI, we challenged LCR1^+/+ and LCR1^−/− mice with House Dust Mite (HDM) for 5 consecutive days after an initial challenge six days before, and quantified immune cell infiltration to the lungs, type 2 cytokine production, and lung histopathology (Fig. 6a). Importantly, except for ILC2s, we did not detect significant changes in the frequencies or total numbers of major lymphoid populations, indicating that the selective ablation of ILC2s does not impair the colonization or homeostasis of these cells at steady-state (Extended Data Fig. 7a–c). Interestingly, 16 days post-HDM challenge, lung ILC2s were still drastically reduced in LCR1-deficient mice (Fig. 6b), a phenomenon that we also observed in the context of acute allergic challenges with IL-33 and papain, indicating a non-redundant role for the LCR1 in ILC2 development at steady-state and during AAI (Extended Data Fig. 8a–e).

Fig. 6|. The LCR1 determines the progression of allergic airway inflammation.

a, Schematic representation of the house dust mite (HDM) challenge.

b, (left) Quantification of lung parenchyma ILC2s at day 16 after the initial HDM challenge. Data is a pool of two age and sex-matched independent experiments (n=13 vs 13). The experiment was repeated three times. p-values: **** = p≤0.0001 (Mann-Whitney U two-tailed test). (right) Representative flow cytometry plots of ILC2s in PBS and HDM-treated mice.

c, Quantification of CD45⁺ cells at day 16 after the initial HDM challenge in the lung parenchyma and bronchoalveolar lavage fluid (BALF) of LCR1^+/+ and LCR1^−/− mice. Data is a pool of two age and sex-matched independent experiments for the lung parenchyma and two age-matched pooled experiments for the BALF. (Lung parenchyma, n=13 vs 13; BALF, n=15 vs 14). The experiments were repeated three times. p-values: ns = not significant, * = p≤0.05, ** = p≤0.01 (Mann-Whitney U two-tailed test).

d-e, Quantification of total T cells, CD4⁺ and CD8⁺ T cells, activated CD4⁺ and CD8⁺ T cells, CD4⁺ Th2 cells, Tregs, and eosinophils in the (d) lung parenchyma and (e) BALF of LCR1^+/+ and LCR1^−/− mice 16 days after the initial HDM challenge. Data is a pool of two age and sex-matched independent experiments for the lung parenchyma and two age-matched pooled independent experiments for the BALF. (Lung parenchyma, n=13 vs 13; BALF, n=15 vs 14). The experiments were repeated three times. p-values: ns = not significant, * = p≤0.05, ** = p≤0.01 (Mann-Whitney U two-tailed test).

f, Quantification of serum IL-5 in LCR1^+/+ and LCR1^−/− mice 16 days after the initial HDM challenge. Data is a pool of two age-matched independent experiments (n= 8 vs 9). p-values: * = p≤0.05 (Mann-Whitney U two-tailed test).

g, Representative histological sections of lungs stained with Periodic Shift Acid (PAS-left) and hematoxylin and eosin (H&E-right) of 2 LCR1^+/+ and 2 LCR1^−/− mice at day 16. Black arrows indicate areas of mucus production; star indicates cell infiltration.

All error bars indicate Standard Error of the Mean (SEM). Intranasal (i.n). Green-bordered dots denote males; orange-bordered dots denote females.

Notably, ILC2 deficiency was associated with lower CD45⁺ cell infiltration in the lung parenchyma and the bronchoalveolar lavage fluid (BALF) in LCR1-deficient mice when compared to WT animals (Fig. 6c). More specifically, activated CD4⁺ T cells, Th2 cells, CD8⁺ T cells, and Tregs in the lungs as well as IL-5 levels in serum were significantly decreased in HDM-challenged LCR1^−/− animals (Fig 6d–f). Moreover, we observed a non-statistically significant decrease of eosinophils numbers in ILC2-deficient mice, while other CD45⁺ immune cell populations remained unchanged (Fig. 6d–e and Extended Data Fig. 9a). Importantly, histological analysis also revealed a drastic decrease in overall cell infiltration and mucus production in the lungs of LCR1^−/− mice (Fig.6g and Extended Data Fig. 9b). Of note, in concordance with the well-described role of ILC2s in the context of acute AAI models^44–46, eosinophils were also significantly decreased in the lung parenchyma of LCR1^−/− mice when compared to WT animals in response to intranasal administration of IL-33 and papain (Extended Data Fig. 8c,e).

To exclude cell-intrinsic effects on CD4⁺ Th2 cell differentiation or functions as a consequence of LCR1 ablation, we assessed type 2 cytokine production in Th2 cells 16 days after the initial HDM challenge as well as in in vitro polarized CD4⁺ Th2 cells. We did not observe any differences in the activation, proliferation, or type 2 cytokine production in ex vivo restimulated lung infiltrating Th2 or in vitro polarized Th2 cells (Extended Data Fig. 9c–d), indicating that a CD4⁺ T cell-intrinsic defect in LCR1^−/− mice is not likely responsible for decreased Th2 numbers in the lungs of these animals during HDM-induced AAI. Altogether, while we cannot fully exclude the possibility that the LCR1 has additional and yet-to-be-discovered roles during AAI, these results strongly indicate that disruption of the 3D architecture of the Id2 locus impairs ILC2 numbers irreversibly and consequently HDM-induced AAI.

Discussion

Herein, we report that the local spatial configuration of the genome is significantly rewired specifically at loci associated with ILC biology to promote their development and functional differentiation. Importantly, we demonstrated that the ontogenesis of ILC2s and the progression of allergic airway inflammation is controlled by a unique local 3D configuration characterized by multiple long-distance interactions between the Id2 gene body and distal cis-RE bound by the ILC2-associated TF GATA-3 and RORα, unveiling the mechanism whereby the Id2 expression is specifically controlled in group 2 ILCs through a dynamic remodeling of the 3D architecture of its locus early in development.

We observed that approximately one-third of the genome compartments are repositioned in the transition from CLP to mature ILC. Furthermore, we found that compartment flipping correlated with the gene expression patterns detected in ILCs at these flipping regions. These results are in agreement with studies in developing B and T cells showing that transcriptional activity is affected by compartment status positioning. Interestingly, TFs required for the priming of the T and B lineages (Bcl11b and Ebf1, respectively) are located within B compartments in early progenitors and repositioned to A compartments upon lineage commitment^12,47,48. Surprisingly, our results show that the ILC lineage-defining transcriptional regulator ID2 as well as other key TFs required for early ILC development are already within A compartments at the CLP stage. These findings suggest that the default 3D chromatin conformation during lymphocyte development might be that of innate lymphocytes, and that commitment to the adaptive lymphocyte lineage requires an active process of compartment repositioning at loci containing lineage defining TFs.

At the megabase scale, we observed higher yields of intra-TAD interactions in regions containing key genes with specific roles in each group of ILCs. Interestingly, these regions were also enriched in motifs of TFs required for the commitment and differentiation of each ILC subset (Eomes, GATA-3, RORα, or RORγt), and of architectural proteins such as CTCF and ETS1⁴⁹. It has been previously suggested that architectural proteins play active roles in cell fate choices. Thus, it is possible that differential chromatin connectivity at the TAD scale may contribute to ILC development through the interplay of ILC-subset specific TFs and subclasses of architectural proteins. Future work should be aimed at establishing how the expression of TFs such as Eomes, GATA-3, RORα, or RORγt contribute to the 3D architecture of early ILC precursors.

Examination of the TAD containing Id2 revealed that its 3D organization in group 2 ILCs has multiple unique characteristics: first, the intra-TAD interactions and the mean strength of DNA loops around the Id2 gene body are stronger in ILC2 when compared to other ILCs. Second, Id2 is the anchor of a stripe only in group 1 and group 2 ILC. However, the stripiness score was much higher in the latter. Third, we identified the LCR1 located 125 kb downstream of Id2 containing two chromatin accessible elements highly enriched in H3K27ac, bound by GATA-3 and RORα, which interact with the Id2 promoter specifically in group 2 ILC. Importantly, our analyses revealed that both loop interactions promote Id2 expression in group 2 ILC, whereas the LCR2 and the Rroid locus are dispensable for this process. These results, along with the previous report showing that Id2 expression in group 1 ILC is regulated by cis-REs within the lncRNA Rroid locus²⁰, indicate that the activity of multiple promoter-enhancer interactions bound by different TFs is required for the tight regulation of this lineage defining TF in the different ILC subsets. How GATA-3, RORα, and other architectural proteins work in concert to establish a unique 3D configuration of the Id2 locus remains to be elucidated. In addition, whether these TFs bind directly to DNA or are recruited through protein-protein interactions to their corresponding target regions and the signals that induce their expression in group 2 ILCs are key questions to be addressed in future work.

Using scRNA-seq and flow cytometry we established that ablation of the LCR1 blocks ILC2 development at the ILC2p stage in the BM and FL, which was accompanied by a skewing towards a developmental trajectory that gives rise to ILC3p and ILC1p-like cells. Moreover, using ATAC-seq we showed that the RORα and GATA-3 binding sites within the LCR1 are closed in CLPs and Flt3⁺ α-LPs, and become fully accessible at the ILCp stage. These results suggest that precursors for each group of ILCs might acquire unique 3D configurations at the Id2 locus characterized by multiple distal promoter-enhancer interactions that are required for their development and maturation. Hence, similarly to group 1 and 2 ILC, it is plausible to infer that another set of distal cis-REs that directly interact with the Id2 promoter, is necessary for group 3 ILC development. Addressing this question in future work will shed new light on early ILC development and the mechanisms that determine commitment to the ILC3 lineage.

We showed that disruption of unique promoter-enhancer interactions specifically blocks group 2 ILC development and consequently, progression of AAI is blunted. These results highlight the importance of understanding the precise contribution of the 3D genome configuration to the transcriptional programs of specific immune cell subsets. The knowledge obtained from these studies can be harnessed for developing genetic tools to probe the functions of specific cell subsets in the context of an otherwise intact immune system⁵⁰. This can be a particularly valuable tool in the study of CD4⁺ T helper cells and ILCs since the core components of their transcriptional programs are shared. More importantly, how alterations in the spatial organization of the genome contribute to the development of inflammatory disorders remains poorly understood. Hence, we propose that an integrative analysis of chromatin accessibility patterns, 3D genome architecture, and single nucleotide polymorphisms associated with immunological diseases in specific immune cells, will shed new light on the etiology of autoimmune and inflammatory pathologies.

Methods

Mice.

The LCR1^−/−, LCR2^−/−, Rora_BS^−/− and Gata3_BS^−/−, Rora_Gata3^dKO mice were generated using the CRISPR/Cas9 system. The sgRNA sequences used for the generation of knockout mice are listed in the Supplementary Table 13. Genomic deletion of each locus was confirmed by PCR and Sanger sequencing. All strains were backcrossed onto the C57BL/6 background for at least 3 generations to control potential off-target effects. B6.SJL-Ptprc^a Pepc^b/Boy (CD45.1⁺ -- Strain #:002014), and B6.129S4-Arg1tm1.1Lky/J (Arg1-YFP -- Strain #:015857) mice were purchased from The Jackson Laboratory. All mice were bred and maintained under pathogen-free conditions at an American Association for the Accreditation of Laboratory Animal Care accredited animal facility at the University of Pennsylvania. Mice were housed in accordance with the procedures outlined in the Guide for the Care and Use of Laboratory Animals under the animal study protocol 805188 approved by the institutional Animal Care and Use Committee at the University of Pennsylvania. All experiments were performed using 6- to 19-week-old mice that were sex and age matched mice, and all experiments were performed in both males and females for this study. Rora_Gata3^dKO were backcrossed for 2 generations to control potential off-target effects and were used at an age of 4- to 5-week-old. Mice were housed at 21°C +/− 2°C, 55% humidity (+/− 10%) with 12 h light dark/ cycle in 7–7 IVC caging with environmental enrichment of plastic houses plus paper bedding.

Cell isolation.

Cells from spleen and lymph node were isolated by physical dissociation and filtered through a 70μm cell strainer. For bone marrow cell isolation, femurs were collected and crushed using a mortar and pestle. All red blood cells were lysed using ACK lysis buffer (GIBCO).

Lungs cell isolation.

Lungs were isolated, minced with scissors, and digested in PBS containing FBS (2%), Collagenase D (1mg/ml), DNase I (0.2mg/ml) for 35 minutes at 37°C with shaking at 200 RPM using a MaxQ4450 (Thermo Scientific). The digested lungs were then passed through a 70μm cell strainer.

Liver cell isolation.

Mice were perfused with 10ml PBS and then transferred to DMEM on ice. The liver was then removed from media and mechanically dissociated using a tissue grinder, then filtered through a 100μm cell strainer. To pellet hepatocytes, the digested livers were centrifuged at 20g for 5min at 4°C. Leukocytes were then purified over an 80/40% Percoll gradient.

Small intestine lamina propria digestion.

small intestines were first flushed with PBS to remove feces, then PP were removed using scissors. The small intestines were then open lengthwise and tissues were shaken in a petri dish containing cold PBS to remove any remaining feces. Mucus and epithelial cells were removed by first incubating 2×15min at 180 RPM and 37°C in PBS containing FBS (2%), HEPES (20mM, pH7.2–7.5) and EDTA (10mM, pH 8) using a MaxQ4450 (Thermo Scientific). Intestines were thoroughly washed 2x with ice cold PBS and minced into 1 cm pieces using scissors. The minced small intestines were then digested PBS containing FBS (2%), HEPES (20mM, pH7.2–7.5), Collagenase D (1mg/ml), DNAse I (0.2mg/mL), and Dispase (0.1 U/ml). The digested intestines were filtered through a 100μm cell strainer and purified over an 80/40% Percoll gradient.

Adipose tissue digestion.

Visceral adipose tissue (VAT) was finely dissected with scissors and digested in PBS containing BSA (2%), Collagenase D (0.5mg/ml), and DNAse I (0.2 mg/ml) at 37°C for 30 min with gentle agitation. Digests were filtered through 70-μm cell strainer and centrifuged at 450g for 5 min at 4°C to pellet cells.

Skin digestion.

Mice were first shaved using a hair trimmer and then 2cm² of skin was used for digestion. Skin was first minced into small pieces prior digestion. Skin pieces were digested in PBS containing FBS (2%), Collagenase IX (1mg/ml), DNAse I (0.2mg/ml) and Hyaluronidase (0.5mg/ml) for 90 min at 37°C with gentle shaking. Digested tissues were then filtered through a 70uM cell strainer and purified over a 67/44% Percoll gradient at 970g for 20 min at 25°C. Cells were collected at the 67/44% interphase.

IL-5 serum dosage.

For serum quantification, mice were anesthetized and blood was harvested by intracardial puncture. Blood was then transferred to a BD Microtainer Serum Separator Tube and left at RT for at least 30 minutes followed by a 2 mins centrifugation at 11000 rcf. IL-5 from serum of naïve or HDM-challenged mice was measured using an Enhanced Sensitivity Flex Set with Enhanced Sensitivity Cytometric Bead Array kit (BD) following the manufacturer’s recommendations.

HDM-induced allergic asthma.

HDM extracts (Dermatophagoides pteronyssinus extracts; Greer Laboratories lot #361863, 385930) were used to induce allergic airway inflammation⁵¹. Briefly, mice were sensitized intranasally with 20 μg HDM extracts on day 0 and subsequently challenged with 10 μg/mouse per day on days 7–13. Three days after the last challenge, mice were anesthetized and used either for immune cell quantification in the lung parenchyma or BALF, as well as for lung histology and measurement of serum IL-5. For BALF was collected by administration of two times 1ml of FACS buffer (PBS + 2% FBS + 2mM EDTA, pH 8). For characterization of cell infiltration in the lung parenchyma. For lung parenchyma experiments, lungs were isolated, minced with scissors, and digested in PBS containing FBS (2%), Collagenase D (1mg/ml), DNase I (0.2mg/ml) for 35 minutes at 37°C with shaking at 200 RPM using a MaxQ4450 (Thermo Scientific). The digested lungs were then passed through a 70μm cell strainer. Cells were then stained and analyzed by flow cytometry.

Eight HDM challenge experiments containing a total of 58 and 35 HDM-challenged males and females were performed as ILC2s are absent in both sexes. Out of these eight experiments, three experiments were used to measure immune cell infiltration in the lung parenchyma, three experiments were used to measure immune cell infiltration in the BALF and for histopathological analysis, and two experiments were used to measure IL-5 levels in serum. For the three experiments performed for measuring immune infiltration to the lung parenchyma, we challenged a total of 20 male and 16 female age and sex-matched mice (6.9–17.2-week-old), which were part of two independent cohorts of males and one independent cohort of females. Two representative experimental cohorts containing 10 male and 16 female mice (6.9–12.9-week-old) were pooled for statistical analysis and data representation. For the three experiments performed for measuring immune infiltration to the BALF, we challenged a total of 21 male and 15 female age-matched mice (6.9–19.5-week-old), which were part of three independent cohorts that contained both males and females. Two representative experimental cohorts containing 14 male and 15 female mice (6.9–8.0-week-old) were pooled for statistical analysis and data representation. For the two experiments performed for measuring IL-5 levels in serum, we challenged a total of 13 males and 4 females age-matched mice (7.5–8.5-week-old mice), which were part of one independent cohort of males and one independent cohort that contained both males and females. These two experiments were pooled for statistical analysis and data representation.

Lung histology.

Lung tissues from 5 males (3 LCR1^+/+, 2 LCR1^−/−) were fixed in 10% formalin for 24 hours and then placed in ethanol 70% before embedding in paraffin. Lungs sections were performed by the molecular and pathology and imaging core at the Center for Molecular Studies in Digestive and Liver Diseases (P30DK050306) and were stained with hematoxylin and eosin (H&E) or periodic acid schiff (PAS) staining.

IL-33 injection.

Mice were anesthetized with Isoflurane and treated with 200ng of recombinant murine IL-33 (Peprotech – 210–33) diluted in 20μL of PBS every 24 hours for 4 days. Mice were sacrificed 24 hours after the last administration and lungs were collected to assess lung ILC2 and eosinophil numbers. Two rIL-33 challenge experiments containing a total of 18 age-matched 9.4–9.7-week-old males were performed.

Papain challenge.

Mice were anesthetized with isoflurane followed by intranasal administration of 30μg of papain (Sigma-Aldrich – 5125–50GM) diluted in 20 μL of PBS once per day for 4 days. Mice were sacrificed 24 hours after the last administration and lungs were collected to assess lung ILC2 and eosinophil numbers. One experiment containing a total of 12 age matched 11.4–12.2-week-old males was performed.

Retroviral particles production.

To generate retroviral particles for Id2-overexpression, HEK-293T cells were purchased from (ATCC). Briefly, HEK-293T cells were maintained in high glucose DMEM medium 1X with L-Glutamine (Invitrogen), supplemented with 100 U/mL penicillin and 100 mg/mL streptomycin (Gibco) with 10% FBS. Retroviral vectors were packaged in HEK-293T cells. Briefly, 4×10⁶ HEK 293T cells were plated in 5 ml media in 10 cm dishes on the day prior to transfection. During transfection, 15 μg of MSCV-IRES-GFP or MSCV-GFP (control) plasmid was co-transfected with packaging plasmid, 15 μg of pCL-Eco, using Lipofectamine 3000 (Invitrogen). The cells were then returned to the incubator for 6 hours. Subsequently, the medium was changed to fresh media. Virions were collected 24 and 48 hrs after transfection, snap-frozen, and stored at −80°C for future use. Viral particles were titrated in NIH3T3 cells that were obtained from ATCC using serial dilution of HEK-derived supernatant.

Bone marrow chimera and retroviral BM transduction.

Mice were first treated 4 days before take down with 5mg of 5-Fluorouracil. Bone marrow from CD45.2⁺ LCR1^−/− mice was isolated by crushing bones with a mortar and pestle. Following red blood cell lysis, cells were counted and plated at 2.5×10⁶ cells/ml in a 6-well plate in IMDM containing FBS (15%) + Penicillin/Streptomycin (1X) and 10ng/ml of IL-3, 5ng/ml of IL-6 and 100ug/mL of SCF (Biolegend) and left in a 5% CO₂ incubator at 37°C overnight. The next day, cells were counted and readjusted to 5×10⁶ cells/mL containing the previous cytokine mix and retroviral supernatants supplemented with 8 μg/mL of Polybrene (EMD Millipore) was added to the appropriate wells. Plates were then spinfected at 1800g 32°C for 90 min and incubated 4 hours at 32°C prior an overnight incubation at 37°C. Transduced BM was then washed two times with PBS and transferred to lethally irradiated CD45.1⁺ hosts via retro orbital injection (down to 0.2×10⁶ cells per mouse). Congenic recipients were irradiated with 1100 rad and cells were injected later the same day. Mice were bled in order to check BM reconstitution after 4 weeks and mice were sacrificed and analyzed after 8 weeks post engraftment.

Antibodies, Flow cytometry, and cell sorting.

All antibodies used are listed in the Nature Portfolio and were diluted in FACS buffer (PBS + 2% FBS + 2mM EDTA, pH 8) and used to stain single cell suspensions for 30 minutes at 4°C. Cells were washed with FACS buffer and either fixed for intracellular staining using the Foxp3 staining buffer (eBioscience) or were fixed with 2% PFA then subjected to flow cytometry. Dead cells were eliminated by incubation of cell suspension in Viability Dye (eFluor780 or eFluor506) diluted in PBS for 10mins at 4°C. Cell counting was performed using 123count eBeads (ThermoFischer Scientific, ref:01–1234-42) following manufacturer’s recommendations.

Cell sorting.

For sorts, splenic NK cells (Live, CD45⁺, CD3e⁻, NK1.1⁺, NKp46⁺) and BM ILC2s (Live, CD45⁺, Lin⁻, CD90.2⁺, ST2⁺) were sorted using a BD FACS Aria II SORP. Lineage (Lin) markers included CD3, CD5, CD11b, CD11c, Gr1, TCRγ/δ, NK1.1, B220, CD19 and Ter119 if not stated otherwise. For single-cell RNA-seq, BM cells were first enriched using the biotinylated Lineage antibodies. Remaining cells (Progenitors) were then stained and sorted as follows: Live, CD45⁺, Lin⁻, NK1.1⁻CD127⁺.

MNK3 in vitro culture.

MNK-3 cells were cultured in DMEM containing FBS (10%), Glutamine (1X), sodium pyruvate (1X), 2-mercaptoethanol (1X), HEPES (10mM, pH7.2–7.5), Penicillin/Streptomycin (1X). MNK-3 cells were maintained in 10 ng/mL of mouse IL-7 (Biolegend) and 10 ng/mL of mouse IL-15 (Biolegend). To maintain high cell viability, the medium was changed every 3 days, and cells were passaged when reaching confluency. For characterization of MNK3 cell markers, cultured cells were stimulated with 10ng/ml of IL-2, IL-23 and IL-1β and expression of defined functional markers was addressed by flow cytometry.

ILC2 in vitro culture.

50,000 ILC2s were sorted from enriched BM of Arg1-YFP mice by sorting Live CD45⁺ Lin⁻ ST2⁺ YFP⁺ cells and expanded in vitro in complete RPMI supplemented with IL-33 (10ng/ml), IL-25 (10ng/ml), IL-7 (10ng/ml) and IL-2 (10ng/ml) with medium replacement every two days. Viability and proliferation were calculated using a hemocytometer and trypan blue incorporation.

Hi-C analysis.

Hi-C libraries were generated on up to 10⁶ cells using Arima-Hi-C kit (Arima Genomics) and Accel-NGS Plus DNA Library kit (21024 Swift Biosciences), according to the manufacturer’s recommendations. Libraries were validated for quality and size distribution using Qubit dsDNA HS Assay Kit (Invitrogen, Q32851) and TapeStation 2200 (Agilent). Libraries were paired-end sequenced (43bp+43bp) on NextSeq 550 (Illumina). The common lymphoid progenitor Hi-C dataset comes from previously published study under the GEO accession number GSE79422¹². All samples were aligned using the HiC-Pro pipeline as recommended by the authors⁵². Next, Hi-C matrices were converted into the cooler format and normalized by matrix balancing with default parameters⁵³.

Compartment analysis.

Compartments were called using the cooltools command line interface (CLI) on each Hi-C dataset at 50kb resolution. We used the GC content file generated by cooltools as the reference track. A weighted compartment score was assigned to genes that span multiple regions with different compartment scores (weighting factor = Number of bps overlapping each region).

TAD analysis.

TADs were called using the HiCExplorer tools^54–56. The function hicFindTADs was applied to group 1, group 2 and group 3 ILCs Hi-C datasets at 25kb resolution, with default parameters. Conserved boundaries were determined by pooling all the TAD boundaries initially reported by hicFindTADs and selecting the ones that appeared in all datasets of interest. Two boundaries were considered the same if they differed by no more than one bin (25kb). Elements within a TAD can interact with other elements that are also within the TAD (intra-TAD interaction) or outside the TAD (extra-TAD interaction). For each TAD, we divided the sum of all intra-TAD interactions by the sum of all extra-TAD interactions in each group ILC. This quantity is referred to as the cross-boundary ratio¹⁶. Next, the cross-boundary ratios of a given TAD, were compared across cell types by computing corresponding pairwise log fold changes between ILCs. Fig. 1f was constructed by selecting, for each cell type, TADs with cross-boundary ratios whose log fold change was greater than 0.25 when compared to the remaining cell types (termed type 1, type 2, type 3 TADs correspondingly). Next, we selected genes within these TADs, whose RPKM log fold change was greater than 0.25 with respect to the remaining cell types. Motif analysis of these regions was carried out by intersecting the coordinates of each TAD with the ATAC-seq peaks of the corresponding cell type, and then, calling the findMotifsGenome.pl function of the HOMER package with default parameters.

Cross-boundary ratio formula.

Let $M_{i,j}^{IL{C}_k}$ be the contact frequency between loci i and j in $IL{C}_k$ as determined by Hi-C, and let T be a TAD detected $IL{C}_k$ in. Thus, the cross-boundary ratio of T is given by,

$$c{b}_T=\frac{\sum _{i,j\in T}{M}_{i,j}^{IL{C}_k}}{\sum _{i\in T,j\in T}{M}_{i,j}^{IL{C}_k}}$$

The numerator represents the total count of interactions between elements belonging to the TAD, while the denominator represents the sum of interactions between the loci in the TAD with loci outside the TAD.

Loop analysis.

We called loops at three different resolutions (5, 10, 25kb) using the Mustache software with parameters; -pt 0.1⁵⁷. Next, for each dataset, we pooled together loops called at different resolutions by examining their anchors (overlapping of both anchors defined equivalent loops) using the bedtools suite⁵⁸.

Stripe analysis.

Stripes were detected using the software Stripenn (https://github.com/ysora/stripenn) with default parameters¹⁸. Stripenn uses a Canny edge detector algorithm borrowed from image processing analysis, to find vertical and horizontal patterns in the HiC maps that correspond to highly interactive regions of chromatin. These regions are then classified and scored in order to identify significant stripes.

Library preparation of bulk RNA-sequencing.

NK cell, in vitro expanded ILC2s and MNK3 cell line were washed once with 1x PBS before resuspending pellet in 350 μL Buffer RLT Plus (Qiagen) with 1% 2-Mercaptoethanol (Sigma), vortexed briefly, and stored at −80°C. Subsequently, total RNA was isolated using the RNeasy Plus Micro Kit (Qiagen). RNA integrity numbers were determined using a TapeStation 2200 (Agilent), and all samples used for RNA-seq library preparation had RIN numbers greater than 9. Libraries were prepared using the SMARTer^® Stranded Total RNA-seq Kit v2-Pico Input Mammalian kit (Takara). Two technical replicates were generated for each experiment. Libraries were validated for quality and size distribution using a TapeStation 2200 (Agilent). Libraries were paired-end sequenced (38bp+38bp) on a NextSeq 550 (Illumina). Bulk RNA data was aligned to the mm10 reference genome using STAR_2.7.8a with --outFilterMultimapNmax 1 --outFilterScoreMinOverLread 0 --outFilterMatchNminOverLread 0 --outSAMtype BAM SortedByCoordinate --alignEndsType Local --outReadsUnmapped Fastx. Additionally, we counted reads using HTseq-count with -f bam -r name -s no -t exon -i gene_id -m intersection-nonempty parameters⁵⁹. Next, we transformed read counts into RPKM for subsequent analyses.

Preparation of ATAC-sequencing samples and library.

ATAC-seq was performed as previously described with minor modifications^60,61. Fifty thousand cells were pelleted at 550 × g and washed with 50 μL ice-cold 1X PBS, followed by treatment with 50 μL of lysis buffer (10 mM Tris-HCl [pH 7.4], 10 mM NaCl, 3 mM MgCl2,0.1% IGEPAL CA-630). Nuclei pellets were resuspended in a 50 μL transposition reaction containing 2.5μL Tn5 transposase (FC-121–1030; Illumina). The reaction was incubated in a 37°C heat block for 45min. Tagmented DNA was purified using a MinElute Reaction Cleanup Kit (Qiagen) and amplified with varying cycles, depending on the side reaction results. Libraries were purified using a QIAQuick PCR Purification Kit (Qiagen). Libraries were validated for quality and size distribution using a TapeStation2200 (Agilent). Libraries were paired-end sequenced (38bp+38bp) on a NextSeq 550 (Illumina). ATAC-seq fastq files were aligned to the mm10 reference genome using the mem function of the software bwa-0.7.17 with additional parameters; -M. ATAC-seq peaks were subsequently called using MACS2 software with additional parameters; -nomodel -f BAM -B --keep-dup all --broad --broad-cutoff 0.1 -q 0.1. ATAC-seq profiles between primary ILC3s and the MNK3 cell line were compared by first creating a reference peak catalog^62,63. The reference peak catalog consisted of the pool of peaks called in each cell type with the corresponding raw counts in each sample. Next, we performed differential analysis using DESeq2⁶⁴. Additionally, we extracted the normalized counts computed by DESeq2 to perform the Principal Component Analysis between different ATAC-seq profiles shown in Fig. 1a. Motif analysis over the Id2 region containing LCR1, LCR2 was performed using HOMER.

Intracellular cytokine staining:

Indicated cells were plated in 96-well round bottom plates in RPMI (Gibco) containing FBS (10%) HEPES (20mM, pH7.2–7.5), sodium pyruvate (1X), Glutamine (1X) Penicillin/Streptomycin (1X), of 2-mercaptoethanol (1X) and held at 37°C for 4 hours. Media containing PMA (Sigma; final concentration 100ng/mL) and Ionomycin (Sigma; final concentration 10ng/mL) was then added to the appropriate wells in addition with GolgiPlug (1X - BD Bioscience). Cytokines and transcription factor expression were measured by intracellular staining using the “Foxp3 staining buffer” (Ebioscience).

CD4⁺ T Cell Purification, Stimulation, and Flow Cytometry:

Naive CD62L⁺CD4⁺ T cells were obtained by negative selection using the Naïve CD4⁺ T cell purification kit from STEMCELL Technologies (19765). Naive CD4⁺ T cells were stimulated with plate bound anti-CD3 antibody (Biolegend) with soluble anti-CD28 (Biolegend) under Th2 cell-polarizing medium containing 10ug/mL of anti-IFN-γ, 50ng/ml of IL-4, and 1ng/mL of IL-2. Cytokines and transcription factor expression were measured by intracellular staining using the “Foxp3 staining buffer” (Ebioscience). Antibodies were all purchased from Ebioscience, BD Bioscience or Biolegend. All samples were acquired and analyzed with the LSR II flow cytometer (Becton Dickinson) and analyzed using FlowJo software (TreeStar).

Single cell RNA-sequencing.

BM progenitors (Live, CD45⁺, Lin⁻, NK1.1⁻, CD127⁺) were obtained by pooling LCR1^+/+ or LCR1^−/− animals (equal number of males and females; between 8- and 12-week-old). After cell sorting cells were washed two times with PBS containing 0.01% of Bovine Serum Albumin (BSA, Sigma #A8806) cell viability and cell counts were assessed using an hemacytometer and a trypan blue staining. Libraries were constructed using Chromium^™ Controller and Chromium^™ Single Cell 3’ Library & Gel Bead Kit v3.1 (10x Genomics) according to the manufacturer’s protocol for 10000 cells recovery. Briefly, cellular suspension was added to the master mix containing RT Reagent B, Template Switch Oligo, Reducing Agent B and RT Enzyme C. 70mL of master mix with cell suspension was then loaded to the Row 1, while Row 2 and Row 3 were filled with 50mL of Gel Beads 45mL of Partitioning Oil respectively. After run completion, 100mL of GEM mix was used for reverse transcription and was performed using a C1000 Touch Thermal cycler (Bio-Rad) at the following conditions: 53°C for 45 min, 85°C for 5 min, held at 4°C. Post GEM-RT cleanup was performed with DynaBeads MyOne Silane Beads (Thermo Fisher Scientific). cDNA was amplified using C1000 Touch Thermal cycler at the following conditions: 98°C for 3 min, 12 cycles of (90°C for 15 s, 67°C for 20 s and 72°C for 1 min), 72°C for 1 min, held 4°C. Amplified cDNA was cleaned with the SPRIselect Reagent Kit (Beckman Coulter) and quality was assessed using a TapeStation (Agilent). Libraries were constructed following the manufacturer’s protocol and sequenced with a targeted sequencing depth of 20.000 read pairs per cell on pair-end mode on either a NextSeq 500/550 or a NovaSeq with recommended loading concentration (1.8pM and 300pM for NextSeq and NovaSeq respectively).

scRNA-sequencing normalization and clustering.

First, we filtered out low quality or dead cells by excluding any cell that had less than 1100 UMI and more than 8% of mitochondrial genes, also, we excluded genes expressed in less than 10 cells. Next, we processed the data following the steps recommended in the SCTransform tutorial⁶⁵. Briefly, we computed the percentage of mitochondrial genes (percent.mt) in addition to the cell cycling score (S.score, g2m.score) of each cell, and then called the SCTransform function of the Seurat package⁶⁶. For the SCTransform function we selected the following parameters:

method=‘GlmGamPoi’, var.to.regress=(‘percent.mt’,’S.score’,’g2m.score’)

Subsequently, we applied dimensional reduction using the RunPCA, RunUMAP, RunTSNE functions and created a Shared Nearest Neighbors (SNN) graph using the FindNeighbors function with default parameters. Finally, we called the function FindClusters with parameters: resolution=0.8, algorithm=4, method=‘igraph’, to obtain the clusters reported in this work.

Differential Expression Analysis.

We created a catalog of differentially expressed genes for several comparisons between clusters and between conditions. To do so, we found marker genes of each cluster using the FindAllMarkers command in Seurat with parameters; logfc.threshold=log(2), only.pos=TRUE. For differentially expressed genes both within clusters (between treat and control condition) and between clusters, we used the FindMarkers function with parameters; logfc.threshold=log(2), only.pos=TRUE. Genes with adjusted p-values smaller than 10⁻⁵ and average log₂ fold change less than 0.25 were considered as differentially expressed in subsequent analysis. Pseudotime trajectories were calculated using the R package Monocle3 with default parameters⁶⁷. We computed pseudotime trajectories independently for LCR1^+/+ and LCR1^−/− scRNA-seq data. For this analysis, we selected a random cell in cluster 8 (a cluster enriched in markers of early progenitor cell populations) as the starting point of the trajectory. Additionally, to create the curves in Fig. 5b and Extended Data Fig. 6g we used the scipy.signals package. We divided the pseudotime axis in 30 bins, computed mean expression levels per bin for each gene, and applied a Savitzky-Golay smoothing over the resulting signal (window_lenght=81, polyorder=3).

Gene Ontology.

GO Enrichment Analysis was performed using the online tool PANTHER⁶⁸. For this analysis, we compared LCR1^+/+ vs LCR1^−/− expression profiles in cluster 7 (Fig. 5c). To do so, we divided the list of differentially expressed genes into up- and down-regulated genes based on the sign of their fold change. Next, we passed each list separately to PANTHER (http://geneontology.org/).

Statistical analysis.

No statistical methods were used to pre-determine sample sizes but our sample sizes are similar to those reported in our previous publications and data collection was appropriately blocked. For all experiments, data distribution was assumed to be normal but this was not formally tested. Two-tailed Mann and Whitney unpaired t-test was used for comparisons of two conditions within one group. Two-tailed one-way ANOVA was used to compare more than two conditions for one group. Two-tailed two-way ANOVA was used to compare two conditions across multiple groups. All graphs show the mean and the standard error of the mean (SEM). One- and Two-way ANOVA were corrected for multiple comparison using Bonferroni correction. Prism 10 GraphPad Software (version 9.4.0 or earlier versions) was used to calculate p-values which represent: ns = not significant, * = p ≤ 0.05, ** = p ≤ 0.01, *** = p ≤ 0.0005, **** = p ≤ 0.0001.

Data collection and sample exclusion.

Data collection and analysis were not performed blind to the conditions of the experiments. We excluded 2 mice that did not show an increase of CD45⁺ cells in the lung parenchyma of HDM-treated mice when compared to the PBS-treated animals. We corroborated our exclusion criteria by performing an Iglewics and Hoagling’s outlier test with modified z-scores, which confirmed our exclusion rationale (modified z-score > 2, p < 0.05).

Extended Data

Extended Data Fig. 1|. Comparison of accessibility between the MNK3 cell line and primary ILC3.

a, Venn diagram representation of the differential chromatin accessibility between ATAC-seq peaks of MNK3 and small intestine (SI) CD4⁺ ILC3, and SI NCR⁺ ILC3 cells. Absolute numbers of peaks differentially open (red) or closed (blue) in MNK3, or with similar accessibility (gray) in all samples, are represented.

b, View of the chromatin accessibility (ATAC-seq signal) at the Rorc, Il17a, Il17f, Il22 and Il23r loci in small intestine (SI) CD4⁺ ILC3, SI NCR⁺ ILC3, and MNK3 cells.

c, Venn diagram representation of the differential accessibility analysis results in the region displayed in the right panel.

d, View of the chromatin accessibility (ATAC-seq signal) at the Id2 locus in SI CD4⁺ ILC3, SI NCR⁺ ILC3, and MNK3 cells. Absolute numbers of peaks differentially open (red) or closed (blue) in MNK3, or with similar accessibility (gray) in all samples, are shown.

e, Representative flow cytometry plots showing cell viability and the expression of ILC3-specific markers (IL-22, GM-CSF, CD90.2, RORγt, IL-17 and ID2) by MNK3 cells after stimulation with IL-2, IL-23, and IL-1β. The red line represents activated MNK3 cells and the black line represents unstained MNK3 cells.

Extended Data Fig. 2|. Changes in compartment distribution in ILCs are accompanied by changes in expression patterns concordant with euchromatin/heterochromatin state.

a, Fraction of regions in the A or B compartments in common lymphoid progenitors (CLP), group 1 (splenic NK cells), group 2 (bone marrow ILC2s), and group 3 (MNK3) ILC.

b, Proportion of regions in CLPs that flipped compartment status in at least one ILC group (brown) or that remained unchanged in all ILC groups (gray). Sidebar indicates the proportion of flipped regions that were in the A (red) or B (blue) compartment at the CLP stage.

c, Examples of compartment state distribution in the vicinity of genes with known roles in ILC biology such as Id2, Tcf7, Nfil3, Gata3, Lifr and Tox. Regions in the A and B compartment correspond to positive (red) and negative (blue) values in the PC1 bar, respectively. Each square represents 25kb.

d, Percentage of expressed genes among genes within compartments that underwent similar flipping from CLP. The Y-axis represents the different combinations of final compartment states in ILC for flipping regions originally in the status indicated on the left. Parenthetical numbers indicate the total number of genes overlapping the corresponding group of regions. Expressed genes correspond to genes with average RPKM > 1. RPKM: Reads per kilobase of transcripts per million mapped reads. Letter triplets represent the compartment status in group 1, group 2 and group 3 ILC, respectively.

e, (left) Heatmap of the compartment status in ILCs of flipping regions that were in the B compartment in CLPs. Regions in the A and B compartment correspond to positive (red) and negative (blue) values in the PC1 bar. (right) Comparison of gene expression levels between group 1 (blue boxes), group 2 (green boxes) and group 3 (red boxes) ILCs of groups of genes undergoing similar compartment flipping from CLP. The Y-axis represents the different combinations of final compartment states in ILC. RPKM: Reads per kilobase of transcripts per million mapped reads. Letter triplets represent the compartment status in group 1, group 2 and group 3 ILC, respectively. Box shows dataset quartiles and whiskers the distribution range. Dots represent outliers as determined by the inter-quartile range.

f, (left) Heatmap of the compartment status in ILCs of flipping regions that were in the A compartment in CLP. Regions in the A and B compartment correspond to positive (red) and negative (blue) values in the PC1 bar. (right) Comparison of gene expression levels between group 1 (blue boxes), group 2 (green boxes) and group 3 (red boxes) ILCs of groups of genes undergoing similar compartment flipping from CLP. The Y-axis represents the different combinations of final compartment states in ILC. RPKM: Reads per kilobase of transcripts per million mapped reads. Letter triplets represent the compartment status in group 1, group 2 and group 3 ILC, respectively. Box shows dataset quartiles and whiskers the distribution range. Dots represent outliers as determined by the inter-quartile range.

g, Table with the p-values (Mann-Whitney U two-tailed test) for Fig.1 e,f of the pairwise comparison between ILC of expression levels of genes that underwent similar compartment flippings from the A or B compartment in CLP. Letter triplets represent the compartment status in group 1, group 2 and group 3 ILC, respectively.

Extended Data Fig. 3|. Local structures of the 3D genome architecture are comparable at the sub-megabase scale and correlate with gene expression in each ILC group.

a, Comparison of the pileup plots computed over conserved vs ILC-specific (left) boundary elements and (right) Topological Associated Domains (TADs) in each group of ILCs. Numbers in the corner indicate average signal strength.

b, Boxplot of cross-boundary ratios of TADs for group 1 (blue), group 2 (green) and group 3 (red) ILCs. Box shows dataset quartiles and whiskers the distribution range. Dots represent outliers as determined by the inter-quartile range.

c, Boxplot of the size distribution of TADs defined by boundaries detected in each ILC (blue; group 1 ILC, green; group 2 ILC, red; group 3 ILC) and by boundaries conserved across ILCs (white). Box shows dataset quartiles and whiskers the distribution range. Dots represent outliers as determined by the inter-quartile range.

d, Expression levels of genes with critical roles in ILC biology in group 1 (blue bars), group 2 (green bars) and group 3 (red bars) ILCs obtained by RNA-sequencing. RPKM: Reads per kilobase of transcripts per million mapped reads.

e, Pileup plots of stripes detected in group 1 (left), group 2 (middle) and group 3 ILC (right). Axes in units of mean feature (stripes) size as computed for each group ILC.

Extended Data Fig. 4|. 3D architecture, chromatin accessibility, and enhancer activity patterns at the Id2 locus in each ILC group.

a, Visualization of the contact heatmap at the Id2 locus in (top) group 1, (middle) group 2, and (bottom) group 3 ILCs. Chromatin accessible regions determined by ATAC-seq (gray track), and H3K27ac deposition (blue track) determined by ChIP-seq (group1 and group 2) or Cut & Run (group 3) in each group of ILCs. Colored arcs represent loops detected in the corresponding contact map (blue; group 1 ILC, green; group 2 ILC, red; group 3 ILC). The location of the Id2 promoter is indicated with an extended vertical dashed line. Dotted and solid lines represent sub-TAD and TAD boundaries, respectively. Arrowheads indicate the position of boundaries conserved across all ILCs. Scale bar in heatmap represents normalized contact frequency.

Extended Data Fig. 5|. The locus control region 1 (LCR1), but not LCR2, is specifically required for the development of all tissue resident ILC2s.

a, Representative transcription factor binding motifs enriched in ATAC-seq peaks (peaks from all ILCs pooled) within LCR1 (red section) and LCR2 (blue section). Numbers represent log odds detection scores as computed by HOMER.

b, (top) Quantification of absolute numbers of splenic CD4⁺ T cells, CD8⁺ T cells, and NK cells in their different maturation states in LCR1^+/+ and LCR1^−/− mice. (bottom) Quantification of absolute numbers of splenic naive (CD62L⁺, CD44⁻) and effector (CD62L⁻, CD44⁺) CD4⁺ and CD8⁺ T cell populations, as well as CD8⁺ and CD11b⁺ dendritic cells. Splenic T cell numbers are one representative experiment that was repeated six times (n = 8 LCR1^+/+ and 8 LCR1^−/−). Splenic NK cell numbers are a pool of two independent experiments. This experiment was repeated four (n = 9 LCR1^+/+ and 9 LCR1^−/−). Tregs represents a pool of two independent experiments (n = 9 LCR1^+/+ and 9 LCR1^−/−) and was repeated three times. Myeloid cell numbers represent one experiment (n = 3 LCR1^+/+ and 3 LCR1^−/−). Error bars = SEM; and p-values: ns = not significant (Treg panel: Mann-Whitney U two-tailed test; Splenic T cells NK cell and Myeloid cell: Two-way ANOVA with multiple comparison and Bonferroni correction).

c, Quantification of absolute numbers of splenic CD4⁺ T cells, CD8⁺ T cells, and NK cells in their different maturation state in LCR2^+/+ and LCR2^−/− mice (n = 4 LCR2^+/+ and 4 LCR2^−/−). Data represents one experiment that was repeated two times. Error bars = SEM; and p-values: ns = not significant, * = p≤0.05, ** = p≤0.01, *** = p≤0.0005 (Two-way ANOVA with multiple comparison and Bonferroni correction).

d, Representative flow cytometry gating strategy used for the identification of group 1 (NK cell), group 2 (ILC2s), and group 3 (ILC3s) ILC in small intestine lamina propria (siLP) of LCR1^+/+ and LCR1^−/− animals.

e, Quantification of lungs and bone marrow (BM) ILC2s numbers of Arginase1 reporter mouse (Arg1^YFP) and Arg1^YFP; LCR1^−/− mice (Live, CD45⁺, Lineage⁻, CD127⁺, CD90.2⁺, ST2⁺, YFP⁺) at steady state. Data is a pool of two independent experiments and were repeated two times. Each dot represents an individual mouse (n = 7 Arg1^YFP; LCR1^+/+ and 7 Arg1^YFP; LCR1^−/−). Error bars = SEM; and p-values: ns = not significant, * = p≤0.05, ** = p≤0.01, *** = p≤0.0005 (Mann-Whitney U two-tailed test).

f, Quantification of IL-5 in the serum of LCR1^+/+ and LCR1^−/− mice at steady state by cytometry bead assay (CBA). Data is representative of two experiments. Each dot represents an individual mouse, (n = 7 LCR1^+/+ and 6 LCR1^−/−). Error bars = SEM; and p-values: ns = not significant, * = p≤0.05 (Mann-Whitney U two-tailed test).

g, (top) Representative flow cytometry gating strategy used for the identification eosinophils in the blood, spleen, and lungs of LCR1^+/+ and LCR1^−/− mice at steady state and (bottom) their subsequent quantification. Data represents one experiment. Each dot represents an individual mouse, (n = 4 LCR1^+/+ and 4 LCR1^−/−). Error bars = SEM; and p-values: ns = not significant, * = p≤0.05 (Mann-Whitney U two-tailed test).

Extended Data Fig. 6|. The LCR1 impairs ILC2 development through reduction of Id2 expression and skewing toward ILC1- and ILC3-like ILC progenitors.

a, (left) Expression levels of Il7r and Id2 over the Uniform Manifold Approximation and Projection of the scRNA-seq data set described in Fig. 4a. (right), Violin plot representation of the expression levels of Id2 and Il7r across all identified clusters in Fig.4a. For each cluster (horizontal axis), the expression levels in LCR1^+/+ (gray) and LCR1^−/− (red) mice is shown

b, Hierarchical representation of similarity between the clusters identified in Fig. 4a. Similarity was measured using complete linkage over the Pearson correlation matrix between clusters.

c, Gating strategy used to identify putative ILC3p in the bone marrow of LCR1^+/+ and LCR1^−/− animals. Lin⁻, CD127⁺ cells were pre-gated on Singlets, Live, CD45⁺ cells

d, Gating strategy for the identification of CLP: Common Lymphoid Progenitors (Live, CD45⁺, Lin⁻, CD127⁺, CD25⁻, ICOS⁻, CD135⁺, α4β7⁻); α-LP: alpha-Lymphoid Progenitors (Live, CD45⁺, Lin⁻, CD127⁺, ICOS⁻, CD135⁺, α4β7⁺, Id2⁺, CD25⁻, CD117⁺); ChILP: Common helper Innate Lymphoid Progenitors (Live, CD45⁺, Lin⁻, CD127⁺, CD135⁻, α4β7⁺, CD25⁻); ILCp: Innate Lymphoid Cell precursors (Live, CD45⁺, Lin⁻, CD127⁺, ICOS⁻, CD25⁻, CD117⁺, Sca1⁻, α4β7⁺, ID2⁺); ILC2p: group 2 Innate Lymphoid Cell precursors (Live, CD45⁺, Lin⁻, CD127⁺, CD90.2⁺, ST2⁺ or CD25⁺ , ICOS⁺ ).

e, Quantification of ILC2 progenitors in the fetal liver (FL), fetal intestine (FI) or fetal lungs (FLu)of LCR1^+/+ and LCR1^−/− E15.5 embryos. Data represents the pool of two independent experiments. Each dot represents an individual embryo (numbers: FL: =21 LCR1^+/+ and 20 LCR1^−/− ; FI and FLu: 11 LCR1^+/+ and 10 LCR1^−/−). Error bars = SEM; and p-values: ns = not significant, * = p≤0.05, ** = p≤0.01, *** = p≤0.0005, **** = p≤0.0001 (Mann-Whitney U two-tailed test).

f, Volcano plot displaying upregulated (red rectangle) and downregulated (blue rectangle) genes in LCR1^+/+ when compared to LCR1^−/− cells for clusters in Fig. 5a representing ILCp populations.

g, Savitzky-Golay smoothing curve of the normalized expression levels of Tox, Bcl11b, Ets1, Tcf7, and Zbtb16 along the pseudotime axis determined in Fig. 5e. P-value corresponds to Mann-Whitney U test between smoothed signals from LCR1^+/+ (black curve) and LCR1^−/− (red curve) mice.

h, (left) Quantification of ID2 mean fluorescence intensity (MFI) in LCR2^+/+ and LCR2^−/− bone marrow (BM) CLP, a4b7⁺ CLP, aLP, ChILP, ILCp, ILC2p gated as shown in Extended Data Fig. 6d. (right) Representative histogram of ID2 expression in LCR2^+/+ and LCR2^−/− BM. Each dot represents an individual mouse (n = 3 LCR2^+/+ and 4 LCR2^−/−) and was observed in 2 independent experiments. Error bars = SEM; and p-values: ns = not significant, * = p≤0.05, ** = p≤0.01 (Two-way ANOVA with multiple comparison and Bonferroni correction).

i, Representative FACS plots and histograms of GFP expression in wild-type (CD45.1⁺) or LCR1^−/− CD45.2 BM previously transduced with retroviral particles encoding an empty vector expressing the GFP (Empty_RV - green) or retroviral particles encoding Id2 and the GFP (Id2_RV – blue), at least eight weeks post engraftment in CD45.1⁺ hosts. Each dot represents an individual mouse (n= 8 LCR1^−/− Empty_RV; 12 LCR1^−/− Id2_RV). Error bars = SEM; and p-values: ns = not significant, * = p≤0.05, ** = p≤0.01 (Two-way ANOVA with multiple comparison and Bonferroni correction).

Extended Data Fig. 7|. Disruption of the 3D architecture of the Id2 locus reduces ILC2 numbers in the lung but does not drastically change other immune cell populations at steady state.

a, Representative flow cytometry gating strategy used for the identification of immune cell populations in digested lungs or in bronchoalveolar lavage fluid (BALF) at steady state or after 16 days of house dust mite (HDM) challenge in LCR1^+/+ and LCR1^−/− animals.

b, Quantification of frequencies among CD45⁺ cells and absolute numbers of lung parenchyma ILC2s in LCR1^+/+ and LCR1^−/− mice at steady state, (n = 8 LCR1^+/+ and 8 LCR1^−/−). Data are representative of one experiment that was repeated five times. Error bars = SEM; and p-values: ns = not significant, * = p≤0.05, ** = p≤0.01, *** = p≤0.0005 (Mann-Whitney U two-tailed test).

c, Quantification of total number of cells of the indicated immune cell populations in Extended Data Fig. 7a in the lung parenchyma of LCR1^+/+ and LCR1^−/− mice at steady state, (n = 12 LCR1^+/+ and 12 LCR1^−/−). Data are a pool of two independent experiments, except CD4⁺ and CD8⁺ TRM which is a representative experiment. (n= 8 LCR1^+/+ and 8 LCR1^−/−). Error bars = SEM; and p-values: ns = not significant, * = p≤0.05 (Mann-Whitney U two-tailed test or Two-way ANOVA with multiple comparison and Bonferroni correction).

Extended Data Fig. 8|. Acute intranasal papain and IL-33 challenges in LCR1-deficient mice.

a, Schematic representation of (left) the experimental acute papain or (right) recombinant IL-33 (rIL-33) challenges. Mice were challenged intranasally with either 30mg of papain (in 20mL of PBS) or with 200ng of rIL-33 (in 20ml of PBS) for four consecutive days. Euthanasia was carried out 24 hours after the last challenge. Lung parenchyma ILC2s and eosinophil numbers were assessed by flow cytometry.

b, Flow cytometry gating strategy used for the identification of lung parenchyma ILC2s and eosinophils in LCR1^+/+ and LCR1^−/− mice challenged with papain or LCR1^+/+ mice treated with PBS (20mL). Cells were pre-gated on SSC/FSC and Singlets.

c, Quantification of ILC2 frequency among CD45⁺ cells, ILC2 and eosinophil numbers in LCR1^+/+ and LCR1^−/− mice challenged with papain, or LCR1^+/+ mice treated with PBS (20mL). Data represents one experiment with males (green dots; n= 4 LCR1^+/+ PBS, 4 LCR1^+/+, and 4 LCR1^−/− papain – 11.4–12.2-week-old). Error bars = SEM; and p-values: ns = not significant, * = p≤0.05, ** = p≤0.01, *** = p≤0.0005, **** = p≤0.0001 (One-way ANOVA with multiple comparisons and Bonferroni correction).

d, Flow cytometry gating strategy used for the identification of lung parenchyma ILC2s and eosinophils in LCR1^+/+ and LCR1^−/− mice challenged with rIL-33 (200ng) or LCR1^+/+ mice treated with PBS (20mL). Cells were pre-gated on SSC/FSC and Singlets.

e, Quantification of ILC2 frequency among CD45⁺ cells, ILC2 and Eosinophil numbers in LCR1^+/+ and LCR1^−/− mice challenged with rIL-33 (200ng), or LCR1^+/+ mice treated with PBS (20mL). Data is representative of two independent experiments (green dots; n= 3 LCR1^+/+ PBS, 3 LCR1^+/+ rIL-33, and 4 LCR1^−/− rIL-33 – 9.4–9.9-week-old) and was repeated two times. Error bars = SEM; and p-values: ns = not significant, * = p≤0.05, ** = p≤0.01, *** = p≤0.0005, **** = p≤0.0001 (One-way ANOVA with multiple comparisons and Bonferroni correction).

Extended Data Fig. 9|. Disruption of the 3D architecture of the Id2 locus in ILC2 reduces HDM-induced allergic airway inflammation progression.

a, Quantification of NK and NKT cells numbers in the lungs of LCR1^+/+ and LCR1^−/− mice 16 days after the initial HDM challenge. Data is a pool of two age and sex-matched independent experiments containing males (green dots; n= 5 LCR1^+/+ and 5 LCR1^−/− – age 12.9-week-old) and females (orange dots, n= 8 LCR1^+/+ and 8 LCR1^−/− – age 6.9-week-old). The experiments were repeated three times. Error bars = SEM; and p-values: ns = not significant, * = p≤0.05, ** = p≤0.01 (Mann-Whitney U two-tailed test).

b, Representative histological sections of lungs stained with hematoxylin (H&E - top) or Periodic Shift Acid (PAS - bottom) of 1 LCR1^+/+ mice treated with PBS. The scale represents a length of 500mM and the magnification used is 40x.

c, Representative flow cytometry plot (left) and quantification (right) of IL-5⁺ and IL-13⁺ producing CD4⁺ Th2 cells (CD4⁺, GATA3⁺) from lung parenchyma. Cells were ex vivo stimulated with PMA and Ionomycin for 4 hours in presence of Golgi inhibitor and identified as Live, CD45⁺, TCRβ⁺, CD4⁺, GATA-3⁺ cells Data is a pool of two age and sex-matched independent experiments containing males (green dots; n= 5 LCR1^+/+ and 5 LCR1^−/− – age 12.9-week-old) and females (orange dots, n=8 LCR1^+/+ and 8 LCR1^−/− – age 6.9 week-old). The experiment was repeated three times. Error bars = SEM; and p-values: ns = not significant (Mann-Whitney U two-tailed test).

d, (top left) Representative flow cytometry plots of IL-13⁺ and IL-5⁺ producing naive CD4⁺ T cells from LCR1^+/+ and LCR1^−/− mice polarized under Th2 conditions for 6 days. (Bottom left) Proportion of IL-13⁺, IL-5⁺ and IL-13⁺/IL-5⁺ producing Th2 cells. (Top right), Quantification of the geometric mean fluorescence intensity (gMFI) of GATA-3, and frequency of activated (CD44⁺) CD4⁺ Th2 cells at day 6 after polarization. Data is representative of three independent experiments. Each dot represents an individual mouse, (Females; n = 4 LCR1^+/+ and 4 LCR1^−/−). Error bars = SEM; and p-values: ns = not significant (Mann-Whitney U two-tailed test).

Supplementary Material

Supplementary Tables 1-13

Supplementary Table 1. Compartment scores of common lymphoid progenitor, group1, group 2 and group 3 ILCs. Each region is annotated with the corresponding set of overlapping genes. Related to Fig. 1 and Extended Data Fig. 1.

Supplementary Table 2. RPKM values of group1, group 2 and group 3 ILCs bulk RNA-sequencing data. Related to Fig. 1, 2 and Extended Data Fig. 1–3.

Supplementary Table 3. Cross boundary ratio of conserved TADs in group1, group 2 and group 3 ILCs. Related to Fig. 2.

Supplementary Table 4. Coordinates of loops called in group 1, group 2 and group 3 ILCs. Loops pooled together. Seventh column contains unique identifier for further reference in Supplementary Tables 5–7 and Extended Data Fig. 7. Related to Fig. 3 and Extended Data Fig. 3.

Supplementary Table 5–7. Unique identifiers of loops detected in group 1 (Table 5), group 2 (Table 6) and group 3 (Table 7) ILCs. Related to Fig. 3 and Extended Data Fig. 3.

Supplementary Table 8–10. Output of the stripenn algorithm for group 1 (Table 8), group 2 (Table 9) and group 3 (Table 10) ILCs. Related to Extended Data Fig. 3.

Supplementary Table 11. List of motifs identified in LCR1, LCR2 and Rroid regions found by HOMER (knownMotifs.txt file). Related to Extended Data Fig. 5.

Supplementary Table 12. List of marker genes of the clusters identified by scRNA-sequencing.

Supplementary Table 13. sgRNA used to generate LCR1^−/−, LCR2^−/−, Gata3_BS^−/−, Rora_BS^−/− and Rora_Gata3^dKO mice.

Data availability

All bulk RNA-seq, ATAC-seq, Hi-C and scRNA-seq datasets have been uploaded to the Gene Expression Omnibus repository (accession n° GSE191312). The following published datasets were used: GEO: GSE111871²³, GEO: GSE77695⁴, GEO: GSE130775⁹, GEO: GSE169542⁴¹, GEO: GSE146745²², GEO: GSE79422¹².