Transcription factor Tox2 is required for metabolic adaptation and tissue residency of ILC3 in the gut

SUMMARY

ILC3 are the major subset of gut-resident ILC with essential roles in infections and tissue repair, but how they adapt to the gut environment to maintain tissue residency is unclear. We report that Tox2 is critical for gut ILC3 maintenance and function. Gut ILC3 highly expressed Tox2, and depletion of Tox2 markedly decreased ILC3 in gut but not at central sites, resulting in defective control of Citrobacter rodentium infection. Single-cell transcriptional profiling revealed decreased expression of Hexokinase-2 in Tox2-deficient gut ILC3. Consistent with the requirement for hexokinases in glycolysis, Tox2^−/− ILC3 displayed decreased ability to utilize glycolysis for protein translation. Ectopic expression of Hexokinase-2 rescued Tox2^−/− gut ILC3 defects. Hypoxia and IL-17A each induced Tox2 expression in ILC3, suggesting a mechanism by which ILC3 adjust to fluctuating environments by programming glycolytic metabolism. Our results reveal the requirement for Tox2 to support the metabolic adaptation of ILC3 within the gastrointestinal tract.

eTOC Blurb

Group 3 innate lymphoid cells (ILC3) play critical roles in gut homeostasis. Das et al. show transcription factor Tox2 is required by ILC3 to metabolically adapt to the gut environment. They identify IL-17 as an important signal that induces Tox2 expression in gut ILC3, but not in other immune cells.

Graphical Abstract

INTRODUCTION

Innate lymphoid cells (ILC) are immune cells that lack specific antigen receptors but possess similar effector functions as T cells^1,2. Group 3 ILCs (ILC3) are innate counterparts of type 17 helper T (Th17) cells that express transcription factor ROR-γt and produce interleukin (IL)-17 and IL-22. ILC3 are heterogenous, comprising NKp46 receptor-expressing ILC3 (NCR⁺ ILC3) and CCR6-expressing ILC3 (CCR6⁺ ILC3), some of which are CD4-expressing lymphoid tissue inducer (LTi) cells. ILC3 are present in primary and secondary lymphoid organs, including the spleen, lymph nodes and thymus³. They are enriched in the small intestinal lamina propria, intestinal crypts, and Peyer’s patches, where they are exposed to local signals such as IL-17A which is abundant in the gut microenvironment^4–6. ILC3 have essential roles in lymphoid organogenesis and tissue repair⁷. Gut-resident ILC3 protect the intestinal mucosa against extracellular bacteria and maintain the intestinal barrier via production of IL-22, IL-17, IFN-γ, GM-CSF and TNF-α^8,9. They also enhance antitumor immunity^10,11.

The development and function of ILC3 require transcription factors, including ROR-γt, GATA-3 and Ahr^12–14. However, other transcription factors implicated in generation and function of ILC3 remain to be identified. Residence in tissues imposes distinct metabolic requirements^15,16, but the nature of these requirements and the transcriptional mechanisms that control them are poorly understood. Metabolic rewiring allows ILC3 to mount a more robust immune response than naïve ILC3 after Citrobacter rodentium¹⁷. mTOR and the hypoxia-inducible factor (HIF-1α) are implicated in controlling intestinal ILC3 function, but it is unclear whether this is tissue-specific^18–21. HIF-1α is a broadly expressed transcription factor that is stabilized under hypoxic conditions²², but the sites of ILC3 residence within the gut are not hypoxic^23–25. Nevertheless, the gut lumen and epithelial layers are hypoxic, and gut injury and infection generate hypoxic environments^26–29. Thus, gut ILC3 may undergo metabolic adaptations to persist in challenging environments generated by injury and infection. However, our knowledge of the transcriptional controllers orchestrating their metabolic adjustments remains incomplete.

Here, we aimed to understand the role of the transcription factor Tox2 (thymocyte selection-associated high mobility group box protein family member 2) in ILC3. We report that the persistence of ILC3 in gut required Tox2. Tox2 was highly expressed in gut ILC3 as compared to mesenteric lymph node (mLN) ILC3, and Tox2 was required for persistence of gut ILC3. IL-17A and hypoxia were each sufficient to induce Tox2 expression in ILC3 to drive metabolic rewiring necessary for residence in the gut. The results suggest that ILC3 possess a tailored mechanism that allows them to sense and metabolically adapt to fluctuating gut environments, thus enabling ILC3 to persist in challenging conditions created by infection and injury.

RESULTS

Tox2 is specifically required in gut ILC3

Tox2 is a sequence non-specific HMG-box transcription factor in the same family as Tox³⁰. Tox2 contributes to the CD8⁺ T-cell exhaustion program³¹. Tox2 is also important for T follicular helper (Tfh) cell development and for the generation and maintenance of memory Tfh cells^32,33.

Tox is first expressed at the Early Innate Lymphoid Progenitor (EILP) stage of ILC development, and its expression is maintained in Innate Lymphoid Progenitors (ILCP) and mature ILC^34,35. Tox2 is expressed at the next stage of ILC development after EILP, in ILCP³⁴. We found Tox2 was highly expressed in small intestinal lamina propria (gut) ILC3 (hereafter termed gILC3) in contrast to mLN ILC3 (hereafter termed mLN-ILC3) (Fig 1A). For measuring Tox2 mRNA, ILC3 were identified using surface markers. We observed co-expression of Kit and ROR-γt in Lin⁻EpCAM⁻CD45⁺ cells from WT and Tox2^−/− gut (S1A), confirming past work establishing Kit as a suitable marker for identification of ROR-γt⁺ ILC3^36–38. Moreover, these Kit⁺ROR-γt⁺ cells expressed CD127 (S1A). Tox2 was expressed in lung ILC2, whereas liver-resident group 1 ILC (ILC1 and NK cells) showed low expression of Tox2 (Fig 1A). To determine the importance of Tox2, we generated germline Tox2^−/− mice³¹. Unlike in Tox-deficient mice, there was no defect in numbers of EILP and ILCP in bone marrow (BM) of adult Tox2^−/− mice (S1B). However, ILC3 were reduced in gut of Tox2^−/− mice (Fig 1B). The number of ILC3 was unchanged at other sites such as mLN, spleen, peripheral lymph nodes and thymus (Fig 1C). The reduction in numbers was evident among NCR⁺, CCR6⁺ and NCR⁻CCR6⁻ subsets of gILC3 (Fig 1D, Fig 1E and S1C). We observed a similar defect in ILC3 number in caecum, colon and Peyer’s patches (S1D). We did not observe defects in cellularity of peripheral lymph nodes such as inguinal, axillary, brachial and mandibular nodes (S1E), and these lymph nodes were present in all mice examined. However, the number of Peyer’s patches in Tox2^−/− mice was slightly but consistently reduced (S1F).

Figure 1. Tox2 is specifically required for gut ILC3 (gILC3) and not ILC3 at central sites.

(A) Expression of Tox2 relative to β-actin was measured in sorted All Lymphoid Progenitors (ALP), NK, ILC1, ILC2 and ILC3 subsets from indicated organs by quantitative reverse-transcription PCR (qPCR). n=4–5 mice.

(B) Representative fluorescence-activated cell sorting (FACS) plot and cellularity of ILC3 defined as Lin⁻EpCAM⁻CD45⁺CD127⁺ROR-γt⁺ in gut of WT and Tox2^−/− mice (left). n=7 mice.

(C) Representative FACS plot and cellularity of ILC3 in indicated organs of WT and Tox2^−/− mice (n=6–9 mice).

(D-E) Bar graph showing absolute number of NCR⁺, CCR6⁺ and NCR⁻CCR6⁻ ILC3 subsets from gut (D) and mLN (E) of WT and Tox2^−/− mice (n=9 mice).

(F) Reconstitution of ILC3 from gut and mLN along with B and NK cells from spleen were examined at 12 weeks post reconstitution. Bar graph shows percentage of chimerism. (n=6 mice).

(G) Schematic overview of experimental strategy(top). Tox2 expression was analyzed in indicated sorted cell populations by qPCR (bottom). n=4 mice.

(H) Representative FACS plot of ILC3 in gut and quantification of ILC3 in gut and mLN from Rag^−/− and Rag^−/−Tox2^−/− mice. Data are representative of at least three independent experiments. Error bars are SEM. See also Figure S1.

To determine when Tox2 is required during ontogeny, we analyzed Tox2 expression and numbers of ILC3 in mice during development. ILCP from adult BM and E15.5 fetal gut expressed Tox2, as did group 3 ILC isolated from both adult and fetal E15.5 embryos (S1G). We found that group 3 ILC were reduced in Tox2^−/− E15.5 intestines (S1H). Furthermore, we found both NCR⁺ and CCR6⁺ ILC3 subsets were reduced in Tox2^−/− mice at all ages analyzed (S1I).

The ILC3 defect in the gut was cell-intrinsic, evident by mixed stem cell chimeras (Fig 1F). To assess whether residence in the gut induced Tox2 expression, we adoptively transferred mLN-ILC3 into immunodeficient NOD scid gamma (NSG) mice and assessed Tox2 expression in donor ILC3 sorted from gut and spleen one-week post transfer (Fig 1G). We found that ILC3 recovered from spleen maintained low expression of Tox2 whereas ILC3 recovered from gut increased Tox2 expression to levels comparable to gILC3 (Fig 1G). We generated Rag1^−/−Tox2^−/− mice and observed that ILC3 number and frequency were reduced in gut but not mLN when compared to Rag1^−/−control mice (Fig 1H), confirming the ILC3 defect in gut of Tox2^−/− mice is independent of T cells.

We assessed other ILC populations in small intestine and observed group 1 ILC remained unaltered between WT and Tox2^−/−mice, whereas ILC2 were reduced in Tox2^−/− mice (S1J). Group 1 ILC numbers at other sites such as in liver, colon and spleen were comparable between WT and Tox2^−/− mice (S1K). Consistent with the defect in gut, ILC2 numbers were reduced by approximately three-fold in lung and colon of Tox2^−/− mice (S1L). B cell, myeloid and T cell numbers were unchanged (S1M).

Together, these results established that Tox2 is required in ILC2 at all sites we assessed, and only in gILC3 but not by ILC3 at central sites. We focused our further inquiries on the gut-specific requirement for Tox2 in ILC3.

Tox2 ablation leads to reduced ILC3 response upon Citrobacter rodentium infection

We explored the functional consequences of Tox2 ablation using Citrobacter rodentium infection, in which production of IL-22 by ILC3 in gut limits infection^39–46. At steady state, we did not observe differences in the frequency of IL-22⁺ and IL-17A⁺ cells in NCR⁺ and CCR6⁺ ILC3 between WT and Tox2^−/− mice (S2A). We infected WT and Tox2^−/− mice with C. rodentium and assessed ILC3 numbers and responses (Fig 2A). Tox2^−/− mice lost more body weight upon C. rodentium infection than WT controls (Fig 2B). The bacterial load on day 6 p.i. was higher in feces, spleen and liver from Tox2^−/− mice compared to WT mice (Fig 2C). However, colon length at day 6 and day 10 p.i. did not differ between WT and Tox2^−/− mice (S2B). On day 6 p.i., there were fewer ILC3 in colonic lamina propria (cLP) of Tox2^−/− mice compared to WT (Fig 2D). The remaining ILC3 in gut of Tox2^−/− infected mice showed lower frequencies of IL-22 producers as compared to WT controls (Fig 2E). T cell frequencies were unaltered between WT and Tox2^−/− mice on day 6 p.i. (S2C). We also infected Rag1^−/−Tox2^−/− mice to rule out any contribution of T cells in modulating ILC3 responses. Consistent with the infected Tox2^−/− mice data, Rag1^−/−Tox2^−/− mice lost more body weight (Fig 2F), displayed higher bacterial burden (Fig 2G), possessed fewer ILC3 (Fig 2H) and fewer IL-22 producers (Fig 2I) compared to control Rag1^−/− mice after day 10 p.i.. IFN-γ production by group 1 ILC was similar in Tox2^−/− and Rag1^−/−Tox2^−/− mice (S2D). Furthermore, IL-22 producers were reduced in NCR⁺, CCR6⁺ and NCR⁻CCR6⁻ gILC3, from both Tox2^−/− (S2E) and Rag^−/−Tox2^−/− (S2F) mice compared to their respective Tox2^+/+ controls.

Figure 2. Tox2 ablation leads to reduced ILC3 response upon Citrobacter rodentium infection.

(A) Schematic overview of experimental strategy.

(B) Weight loss of WT and Tox2^−/− mice after infection with C. rodentium (n=8–10 mice).

(C) Bacterial load measured in feces and indicated organs on day 6 post infection (p.i.) in WT and Tox2^−/− mice (n=4–5 mice).

(D) Representative FACS plot and cellularity of cLP ILC3 from WT and Tox2^−/− mice on day 6 p.i. (n=4–5 mice)

(E) Representative FACS plot and frequency of IL-22⁺ and IL-17A⁺ cLP ILC3 on day 6 p.i. in WT and Tox2^−/− mice (n=4–5 mice).

(F) Weight loss of Rag^−/− and Rag^−/−Tox2^−/− mice after infection with C. rodentium (n= 8–10 mice).

(G) Bacterial load measured in feces and indicated organs on day 7 p.i. in Rag^−/− and Rag^−/−Tox2^−/− mice. n=4–5 mice.

(H) Representative FACS plot and cellularity of cLP ILC3 from Rag^−/− and Rag^−/−Tox2^−/− mice on day 7 p.i. n=4–5 mice.

(I) Representative FACS plot and frequencies of IL-22⁺ and IL-17A⁺ cLP ILC3 on day 7 p.i. in Rag^−/− and Rag^−/−Tox2^−/− mice. n=4–5 mice. Data are representative of at least three experiments. Error bars are SEM. See also Figure S2.

These results confirmed the specific requirement of Tox2 in generating functional gILC3 responses against C. rodentium.

Tox2 is required by gut ILC3 for their persistence in the gut

We generated mice carrying a conditional allele of Tox2 (Tox2^flox, S3A). We crossed Tox2^flox mice with IL7r^Cre47 mice, thus deleting Tox2 in all lymphoid lineage cells (IL7r^CreTox2^fl/fl). This mouse recapitulates the phenotype of the Tox2 germline deficient mouse, thus establishing that the Tox2^flox mouse strain efficiently enables conditional deletion of Tox2 (S3B). To assess the role of Tox2 specifically in ILC3, we crossed Tox2^flox mice with Rorc^Cre mice. As expected, the frequency and number of gILC3, but not mLN-ILC3, were reduced upon Tox2 ablation in Rorc⁺ cells (Fig 3A and S3C for ILC3 subsets).

Figure 3. Tox2 is required by gILC3 for their persistence in the gut.

(A) Representative FACS plot and cellularity of ILC3 from mLN (mLN-ILC3) and gut (gILC3) of Rorc^CreTox2^+/+ and Rorc^CreTox2^fl/fl mice (n=5 mice).

(B) Schematic overview of experimental strategy.

(C) Tox2 expression in sorted mLN-ILC3 and gILC3 from ER^T2CreTox2^+/+ and ER^T2CreTox2^fl/fl mice by qPCR (n=5 mice).

(D) Representative FACS plot and cellularity of mLN-ILC3 and gILC3 in ER^T2CreTox2^+/+ and ER^T2CreTox2^fl/fl mice after tamoxifen treatment (n=5 mice).

(E) Representative histogram (top) showing Rorc expression history in gut NK cells, ILC1 and ILC3 from Rorc^CreRosa-YFPTox2^+/+ and Rorc^CreRosa-YFPTox2^fl/fl mice. Bar graph (bottom) shows the quantification of Rorc-YFP⁺ ILC1 (ROR-γt⁻) from mLN and gut (n=4 mice).

(F) Visualization (top) and quantification (bottom) of cryptopatches (CP) and isolated lymphoid follicles (ILF) in duodenum of Rorc^CreRosa-YFP Tox2^+/+ and Rorc^CreRosa-YFPTox2^fl/fl mice (n=3 mice). Data are representative of three independent experiments. Error bars are SEM. See also Figure S3.

We assessed whether Tox2 ablation affected homing by altering CCR7 expression in ILC3 and ILC3 subsets from mLN and gut. Tox2 deficiency did not impact CCR7 expression on mLN-ILC3 and gILC3 (S3D and S3E for ILC3 subsets). We observed no defect in proliferation of mLN-ILC3 and gILC3 as measured by BrdU incorporation (S3F). However, gILC3 from Tox2^−/− mice showed more death as compared to WT controls as measured by cleaved caspase 3 staining whereas mLN-ILC3 showed comparable frequency of cells with cleaved caspase 3 between WT and Tox2^−/− mice (S3G).

To assess whether Tox2 is required for persistence of mature ILC3 in gut, we crossed Tox2^flox mice with ER^T2Cre mice (Fig 3B), leading to decreased Tox2 expression in mature ILC3 (Fig 3C). Conditional deletion of Tox2 reduced gILC3 but not mLN-ILC3 number, indicating Tox2 is required for persistence of mature ILC3 in gut (Fig 3D).

ILC3-to-ILC1 plasticity is observed in mucosal tissues including gut^48–50. To determine whether Tox2 deficiency affected the ILC3-to-ILC1 transition, we generated Rorc^CreRosa-YFPTox2^fl/fl mice, which allowed tracking Rorc expression history in ILC3-ILC1 populations. We observed approximately 30% of ILC1 showed a history of Rorc expression in gut (Fig 3E) and the number of ROR-γt⁻Rorc-YFP⁺ ILC1 (“ex-ILC3”) remained unaltered in the absence of Tox2 in mLN and gut (Fig 3E). We also evaluated solitary intestinal lymphoid tissue structures using the same mouse model to assess if reduced gILC3 in Tox2^−/− affected the formation of cryptopatches (CP) and isolated lymphoid follicles (ILF) in the duodenum, jejunum and ileum. There were fewer CP in duodenum compared to other sections in Rorc^CreRosa-YFPTox2^fl/fl mice but the number of ILF remained comparable across all sections analyzed (S3H). Hence, we focused on duodenum for quantification of CP and ILF. Deficiency of Tox2 in Rorc-expressing cells resulted in smaller CP and ILF (Fig 3F). The number of CP was reduced, however the number of ILF was only minimally affected in Rorc^CreRosa-YFPTox2^fl/fl mice compared to Rorc^CreRosa-YFPTox2^+/+ (Fig 3F).

Together these results establish that Tox2 is required for the survival and thus persistence of mature ILC3 in gut.

Gut ILC3 from Tox2^−/− mice are transcriptionally distinct from their WT counterparts

To examine whether Tox2 deficiency leads to transcriptional alterations in ILC3, we profiled mLN-ILC3 and gILC3 from WT and Tox2^−/− mice using single-cell RNA-sequencing (scRNAseq). We identified ILC3 clusters in mLN (S4A) and gut (S4B). To rule out the possibility of other contaminants, we integrated our mLN-ILC3 data with a published mLN-ILC3 dataset which used RORγt-eGFP BAC-transgenic reporter mice to sort ILC3⁵¹. We observed the mLN-ILC3 subsets overlapped with the published dataset (S4A). Two broad subsets of ILC3: NCR⁺ ILC3 (Ncr1⁺Tbx21⁺Rorc^lowEomes⁻) and CCR6⁺ ILC3 (Ccr6⁺Cd4⁺Rorc⁺) were identified based on gene expression in mLN (S4A) and gut (S4B). Rorc transcripts in RORγt-eGFP BAC-transgenic mice were considerably lower in NCR⁺ ILC3 as compared to CCR6⁺ ILC3 (S4A). This can explain why we detected lower Rorc transcripts in wild-type (non-transgenic) NCR⁺ ILC3 (S4A). While Rorc expression was low in NCR⁺ mLN-ILC3, it was appreciably higher than in other cells such as NK cells (S4A). Visualization of data using UMAP showed mLN-ILC3 from WT and Tox2^−/− mice overlapped with each other suggesting they are transcriptionally similar, whereas gILC3 from WT and Tox2^−/− mice occupied different positions in the manifold, indicating gILC3 from Tox2^−/− and WT mice are transcriptionally distinct (Fig 4A and Fig 4B). We observed 97 differentially expressed genes (DEG) in CCR6⁺ ILC3 and 95 DEG in NCR⁺ ILC3 in the gut as compared to only 14 DEG and 11 DEG in CCR6⁺ ILC3 and NCR⁺ ILC3 in mLN-ILC3, respectively, between WT and Tox2^−/− mice (S4C).

Figure 4. gILC3 from Tox2^−/− mice are transcriptionally distinct from their WT counterparts.

scRNA sequencing analysis performed on sorted as Lin⁻EpCAM⁻CD5⁻kit⁺ cells from mLN and gut isolated from WT and Tox2^−/− mice (A-E)

(A) UMAP representation of mLN-ILC3 (top) and gILC3 (bottom) isolated from WT and Tox2^−/− mice, colored and clustered by original identity.

(B) UMAP representation of two broad ILC3 subsets, NCR⁺ (Ncr1) and CCR6⁺ (Ccr6) ILC3 in mLN-ILC3 (top) and gILC3 (bottom) isolated from WT and Tox2^−/− mice, colored and clustered by original identity.

(C) Heatmap of selected enriched and reduced genes in gILC3 and mLN-ILC3 as controls from WT and Tox2^−/− mice. Hexokinase-2 (Hk2) expression is indicated in red box.

(D) Scheme of glycolysis pathway including genes and metabolic products (left). Heatmap of glycolytic enzymes in mLN-ILC3 and gILC3 from WT and Tox2^−/−mice (right).

(E) Heatmap of selected oxidative phosphorylation (OXPHOS) pathway genes in mLN-ILC3 and gILC3 from WT and Tox2^−/− mice.

(F) Hk2 expression relative to β-actin in mLN-ILC3 and gILC3 sorted from WT and Tox2^−/− mice (left) and from ER^T2CreTox2^+/+ and ER^T2CreTox2^fl/fl mice (right) by qPCR after tamoxifen treatment (n=4–5 mice).

(G-H) Hk2 expression was analyzed by flow cytometry in ILC3 and CD4 T cells as control from mLN (G) and gut (H) of WT or Tox2^−/− mice (n=5 mice).

(I-J) Violin plot and histogram showing the expression of T-bet and Bcl6 in gILC3 from WT and Tox2^−/− mice (n=5 mice). Data are representative of three independent experiments. Error bars are SEM. See also Figure S4.

We assessed the top enriched and reduced genes in Tox2-deficient gILC3 (Fig 4C, full gene list in Supplementary Table 1). Genes that were enriched in the absence of Tox2 were associated with effector function (Il17f, Fgl2) and antigen presentation and processing (H2-Q4, Cd74) whereas the genes that were reduced were associated with migration (Cxcr4, Cxcr5) and metabolism (Arg1, Ldhb, Hk2) (Fig 4C and Supplementary Table 2).

Tox2 is required for Hk2 induction in gut ILC3

Our scRNAseq data analysis identified Hexokinase-2 (Hk2) as a putative target of Tox2. We focused on Hk2 because it is the first enzyme for glycolysis and is rate-limiting^52,53. Other Hexokinases such as Hk1, Hk3 were not differentially expressed in our scRNAseq data (Fig 4D and S4D). Based on the decreased expression of Hk2, we examined the expression of genes involved in glycolysis (Fig 4D) and oxidative phosphorylation (OXPHOS) (Fig 4E). mLN-ILC3 from both WT and Tox2^−/− mice exhibited higher expression of genes associated with OXPHOS as compared to gILC3 from WT and Tox2^−/− mice (Fig 4E and S4E for ILC3 subsets), suggesting a differential reliance of mLN-ILC3 and gILC3 on OXPHOS and glycolysis, respectively.

We performed qPCR for Hk2 in mLN-ILC3 and gILC3 and confirmed that Hk2 expression was increased in gILC3 as compared to mLN-ILC3 from WT mice (Fig 4F). Hk2 expression was reduced in gILC3 from germline Tox2^−/− mice as well as from tamoxifen-treated ER^T2CreTox2^fl/fl mice as compared to WT mice (Fig 4F and S4F for ILC3 subsets ). We assessed Hk2 protein expression and found that while Hk2 expression in mLN-ILC3 and CD4 T-cells from WT and Tox2-deficient mice were similar (Fig 4G), Hk2 was markedly elevated in gILC3 as compared to mLN-ILC3 (Fig 4H). Moreover, Hk2 protein expression was reduced specifically in gILC3 from Tox2^−/− mice (Fig 4H). To assess whether residence in the gut induced Hk2 expression by ILC3, we adoptively transferred sorted mLN-ILC3 and gILC3 from WT and Tox2^−/− mice into NSG mice. We observed mLN-ILC3 that homed to gut increased Hk2 expression. As expected, donor mLN-ILC3 from Tox2^−/− mice that homed to gut did not induce Hk2 expression (S4G).

We also determined whether Tox2 regulated Hk2 expression in other cells that needed Tox2 such as ILC2 and Tfh cells³². We observed that Tox2 deficiency did not alter Hk2 expression in ILC2 (S4H) or in Tfh cells (S4I). We sought to determine whether Hk2 might be regulated by Tox in addition to Tox2. Tox-deficient mice lack nearly all ILCs except gILC3³⁵. We found Hk2 expression remained similar between WT and Tox^−/− NCR⁺ and CCR6⁺ gILC3 subsets (S4J), indicating Tox and Tox2 play distinct roles in gILC3.

In agreement with our earlier data examining cytokine production and proliferation in gILC3 at steady state, the expression of key ILC3 cytokine genes and proliferation genes were minimally altered between WT and Tox2^−/− ILC3 subsets from mLN and gut (S4K). Furthermore, T-bet and Bcl6, which are regulated by Tox2 in other lineages^32,54 were not altered at mRNA or protein levels between WT and Tox2^−/− gILC3 (Fig 4I–J).

To investigate the impact of Tox2 deficiency on chromatin landscapes, we performed scATAC sequencing using gILC3 from WT and Tox2^−/−, with WT mLN-ILC3 serving as an additional control. Global chromatin accessibility in total gILC3 and gILC3 subsets between WT and Tox2^−/− mice were similar (S4L). The single exception was one region at the Tox2 locus in gILC3 that was deleted in Tox2^−/− mice (S4M). We looked at the promoter and enhancer regions of Hk2 locus and did not observe any differentially accessible peaks between WT and Tox2^−/− gILC3 subsets (S4M) suggesting Tox2 does not alter chromatin landscapes in ILC3 lineage cells.

Together these results indicate that ILC3 induce Hk2 expression in the gut, in a Tox2-dependent manner, and Tox2 does not alter chromatin accessibility in ILC3.

Tox2 deficiency leads to attenuated protein translation in gILC3

The data presented above suggest a dependence on distinct metabolic programs in tissue-resident gILC3 and mLN-ILC3. As protein translation is tightly coupled with ATP production^55,56, we monitored incorporation of puromycin into nascent proteins as a surrogate of energy generation in mLN-ILC3 and gILC3. To quantify the specific contributions of glycolysis and OXPHOS to protein synthesis, we monitored puromycin incorporation in WT mLN-ILC3 and gILC3 following treatment with the inhibitors 2-deoxyglucose (2DG) and oligomycin, respectively⁵⁷. Oligomycin treatment almost completely abrogated puromycin incorporation in mLN-ILC3, even moreso than 2DG (Fig 5A), indicating a strong dependence of mLN-ILC3 on mitochondrial energetics and a low glycolytic capacity (S5A). However, the inverse situation was detected in WT gILC3. 2DG resulted in a greater reduction in puromycin MFI than oligomycin-treated gILC3 (Fig 5A), demonstrating that WT gILC3 have higher dependence on glycolysis and a lower mitochondrial dependence than WT mLN-ILC3 (Fig 5A and S5A).

Figure 5. Tox2 deficiency leads to attenuated protein translation in gILC3 with decreased dependence on glycolysis.

(A) Representative histograms of puromycin antibody staining (solid lines) on mLN-ILC3, gILC3 overlaid against a without puromycin control (black dotted line) in indicated groups. Control puromycin MFI is represented in black line in each plot. Bar graph represents the MFI of puromycin gated on mLN-ILC3 and gILC3 from WT mice (n=5 mice).

(B-C) Representative histograms and quantification of puromycin incorporation by ILC3 and CD4 T cells as control from mLN (B) and gut (C) of WT and Tox2^−/− mice treated with either puromycin (solid lines) or PBS (dotted lines). n=5–6 mice.

(D) Quantification of MFI of puromycin incorporation in indicated groups from WT and Tox2^−/− mice gated on mLN-ILC3 (n= 5 mice).

(E) Representative histograms of puromycin antibody staining (solid lines) and quantification on gILC3 from WT and Tox2^−/− mice in indicated groups. Control histograms showing fluorescence in the absence of puromycin staining are presented (dotted lines). n= 5 mice.

(F-G) Bar graph represents the concentration of lactate (F) and basal extracellular acidification rate (ECAR) and oxygen consumption rate (OCR) (G) measured in supernatant from 3-day culture of sorted GFP⁺ transduced MNK-3 cells with empty virus (GFP), Tox2.GFP virus and Hk2.GFP virus along with untraduced MNK-3 cells (UT) control (n>9). Data are representative of at least two independent experiments. Error bars are SEM. See also Figure S5.

Next, we evaluated in vivo translation by injecting mice with puromycin⁵⁶. There was no difference in puromycin incorporation in mLN-ILC3 or mLN CD4⁺ T cells from WT and Tox2^−/− mice (Fig 5B). There was a reduced incorporation of puromycin in gILC3 from Tox2^−/− mice as compared to WT controls whereas gut CD4⁺ T cells exhibited similar incorporation (Fig 5C and S5B for gILC3 subsets). Thus, Tox2 is specifically required to maintain global protein translation in gILC3.

We then assessed the contributions of glycolysis and OXPHOS to protein translation in WT and Tox2^−/− ILC3. Puromycin incorporation was equivalent in all tested conditions between WT and Tox2^−/− mLN-ILC3 (Fig 5D and S5C), consistent with low Tox2 expression in mLN-ILC3 and the relative lack of transcriptional alterations in mLN-ILC3. However, puromycin incorporation was reduced in Tox2^−/− gILC3 compared to WT gILC3 (Fig 5E). As described above, 2DG treatment reduced puromycin incorporation in WT, but not Tox2^−/− gILC3 (Fig 5A and 5E), highlighting defective glycolysis in the latter (Fig 5E). These results indicate that WT gILC3 are characterized by a Tox2-dependent glycolytic capacity (S5D). These trends were also observed in NCR⁺ and CCR6⁺ ILC3 subsets (S5E). Furthermore, WT gILC3 showed substantial puromycin MFI in presence of oligomycin but not with 2DG+Oligomycin, indicating lower dependence on OXPHOS and higher glycolytic capacity (Fig 5E and S5D). These results are consistent with the hypothesis that higher expression of Tox2 in gILC3 induces Hk2 expression, and increased Hk2 in turn confer enhanced glycolytic capacity on WT gILC3.

To further support these protein translation data, we directly monitored glycolysis and oxygen consumption in gILC3. Due to low gILC3 cell numbers obtained from Tox2^−/− mice, we instead utilized the ILC3 cell line MNK-3⁵⁸. We transduced MNK-3 cells with Tox2 and Hk2 (S5F) as parental MNK-3 cells do not express Tox2 (S5G). Enforced expression of Tox2 in MNK-3 cells resulted in increased Hk2 mRNA (S5G) and protein (S5H) expression while Hk1 protein expression remained unaltered (S5I). Furthermore, enforced expression of either Tox2 or Hk2 resulted in increased lactate production (Fig 5F) and higher basal ECAR (Fig 5G), a measure of glycolysis, as compared to MNK-3 cells transduced with empty virus. In contrast, Tox2 or Hk2 expression minimally altered the oxygen consumption rate (OCR) (S5J) and basal OCR in Tox2 or Hk2-overexpressed MNK-3 cells compared to control cells (Fig 5G).

Together, these results show that mLN-ILC3 and gILC3 exhibit distinct metabolic programs with a bias towards glycolysis in gILC3 which is Tox2-dependent. These data further indicate that Tox2 supports glycolysis in ILC3 lineage cells.

Ectopic Hk2 expression rescues Tox2^−/− ILC3 gut residency

The results above suggest that Tox2 promotes glycolysis in gILC3, likely through the increased expression of Hk2. To assess the importance of Hk2 regulation by Tox2 in vivo, we ectopically expressed Hk2 in Tox2^−/− BM progenitors using lentivirus, and adoptively transferred the transduced BM progenitors into NSG mice (Fig 6A). Ectopic expression of Hk2 in Tox2^−/− progenitors rescued the number of Tox2^−/− ILC3 in the gut (S6A, S6B, Fig 6B, Fig 6C and S6C). Both NCR⁺ and CCR6⁺ gILC3 subsets were rescued upon Hk2 overexpression in Tox2^−/− progenitors (Fig 6D). Furthermore, this rescue was specific for ILC3 lineage cells since Hk2 overexpression in Tox2^−/− progenitors did not affect NK cell numbers (Fig 6C) or rescue the ILC2 defect (S6D). It is likely that the Tox2 requirement in ILC2 reflects its control of genes required in ILC2 development from ILCP; these gene targets remain to be discovered. These results implicate Hk2 as a critical target of Tox2 which is essential for persistence of ILC3 in gut.

Figure 6. Ectopic Hexokinase-2 expression rescues Tox2^−/− ILC3 gut residency.

(A) Schematic overview of experimental strategy.

(B) Representative FACS plot showing gut ILC3 recovered from gut after 12 weeks reconstitution of GFP and Hk2.GFP transduced WT and Tox2^−/− BM progenitors.

(C-D) Quantification of recovered ILC3 and NK cells (C) and ILC3 subsets (D) from NSG gut in indicated groups. n=4–9 mice.

(E) Schematic overview of in vitro ILC3 culture experiment.

(F-G) Representative FACS plot showing recovery of CD45⁺ sorted mLN-ILC3 (F) and gILC3 (G) from WT and Tox2^−/− mice post 24h culture in indicated groups (n= 4–5 mice).

(H) Quantification of the recovered frequencies of mLN-ILC3 (left) and gILC3 (right) from (F) and (G). n=5 mice. Data are representative of three independent experiments. Error bars are SEM. See also Figure S6.

To further test the importance of Hk2 regulation by Tox2, we evaluated whether bypassing Hk2 via supplementation of downstream metabolites would improve the survival of gILC3 in vitro. First, we assessed whether high glucose (25mM), pyruvate (10mM) or acetate (5mM) would enhance the survival of WT and Tox2^−/− mLN-ILC3 cultured for 24 hours on irradiated OP9 cells (Fig 6E). The frequency of mLN-ILC3 recovered after culture was minimally altered by the absence of Tox2 and survival was also not altered by any of the nutrient supplements we tested (Fig 6F and Fig 6H). The frequency of recovered gILC3 was lower than that of mLN-ILC3 in standard glucose media (Fig 6F and Fig 6G). Consistent with a high dependence of gILC3 on glucose, high glucose media increased the recovery of WT gILC3 (Fig 6G and Fig 6H). However, this was not the case for Tox2^−/− gILC3. The addition of pyruvate or acetate, bypassing Hk2, increased the recovery of Tox2^−/− gILC3 (Fig 6G and Fig 6H). Thus, even in the absence of Tox2-induced Hk2, gILC3 survival can be rescued in vitro by metabolites generated downstream of the enzymatic activity of Hk2.

Our results indicate that Hk2 is a critical target of Tox2 which is essential for persistence of gILC3 by increasing their glycolytic metabolism.

ILC3 Tox2 expression is induced by hypoxia and IL-17A

Our experiments highlighted differences in metabolic programming of ILC3 at different sites as a function of Tox2 and Hk2 expression. Nevertheless, these experiments did not determine why Tox2 expression is higher in gILC3 as compared to mLN-ILC3 (Fig 1A). ILC3 reside in the gut lamina propria which is not hypoxic, but gut injury and infection generate hypoxic environments^25–29. HIF-1α is a broadly expressed transcription factor that is stabilized under hypoxic conditions²². Thus, we examined whether HIF-1α was differentially expressed between mLN-ILC3 and gILC3. We detected increased HIF-1α expression in gILC3 as compared to mLN-ILC3, irrespective of the genotype of the mice, in scRNAseq data and by antibody measurement of protein levels (S7A and S7B), whereas HIF-1α protein was similar between mLN and gut CD4⁺ T cells in WT and Tox2^−/− mice (S7C). Flow cytometry analyses revealed increases in HIF-1α in both NCR⁺ and CCR6⁺ gILC3 (Fig 7A), and we identified 3 putative binding sites of HIF-1α near the Tox2 transcription start site (S7D).

Figure 7. ILC3 Tox2 expression is induced by hypoxia and IL-17A.

(A) Representative histogram (left) and quantification (right) of HIF-1α expression on NCR⁺ and CCR6⁺ ILC3 from mLN and gut of WT and Tox2^−/− mice (n=5 mice).

(B-C) Expression of Tox2 and Hk2 by qPCR in mLN-ILC3 and gILC3 (B) and Tox2 expression in ILC3 subsets (C) from WT and Tox2^−/− mice following a 6h incubation in atmospheric normoxia (21% O₂): Nor and hypoxia (5% O₂): Hyp conditions (n=5–7 mice).

(D-E) Quantification of ILC3 (D) and ILC3 subsets (E) from mLN and gut of Rorc^CreHif1a^+/+ and Rorc^CreHif1a^fl/fl mice (n= 4–7 mice).

(F) Expression of Tox2 and Hk2 by qPCR in ILC3 subsets from Rorc^CreHif1a^+/+ and Rorc^CreHif1a^fl/fl mice (n=3 mice).

(G-H) Tox2 expression in mLN-ILC3 subsets after 6h of indicated cytokine treatment from WT mice (G) and Rorc^CreHif1a^+/+ and Rorc^CreHif1a^fl/fl mice (H) by qPCR (n≥3).

(I-J) Representative FACS plot and quantification of ILC3 (I) and ILC3 subsets (J) from mLN (top) and gut (bottom) of WT and Il17a^−/− mice (n= 6–8 mice).

(K) Expression of Tox2 (top) and Hk2 (bottom) in indicated groups by qPCR (n= 6–8 mice).

(L-M) HIF-1α antibody expression (left) and quantification (right) on mLN-ILC3 and gILC3 (L) and ILC3 subsets (M) of WT and Il17a^−/− mice (n= 4 mice). Data are representative of at least two independent experiments. Error bars are SEM. See also Figure S7.

We examined whether hypoxia would induce HIF-1α, Tox2 and Hk2 expression in ILC3. HIF-1α expression was increased upon exposing mLN-ILC3 to hypoxia (5% O₂) for 6 hours; however, HIF-1α expression in gILC3 did not show further increase after incubating gILC3 under hypoxia (S7E). In both mLN-ILC3 and gILC3, hypoxia resulted in induction of Tox2 expression by qPCR (Fig 7B). Indeed, a 6-hour exposure of mLN-ILC3 to hypoxia induced Tox2 expression to levels that were similar to those detected in gILC3 (Fig 7B, Fig 7C for ILC3 subsets). Exposure of WT mLN-ILC3 to hypoxia also resulted in induction of Hk2, to levels similar to those detected in gILC3 (Fig 7B). However, Hk2 expression was not increased in either Tox2^−/− mLN-ILC3 or gILC3 in response to hypoxic conditions (Fig 7B). Hence, acute exposure to hypoxic conditions in vitro resulted in the induction of Tox2 and Hk2 expression in mLN-ILC3, and Hk2 induction by hypoxia required Tox2.

To determine whether HIF-1α is required for Tox2 and Hk2 induction in both NCR⁺ and CCR6⁺ gILC3, we assessed Rorc^CreHif1a^fl/fl mice. Consistent with previous work^18,19, we saw a 2-fold reduction in gILC3 and no defect in NCR⁺ ILC3 in gut (S7F, Fig 7D and 7E). We observed a reduction in CCR6⁺ gILC3 (Fig 7E) but mLN-ILC3 numbers remained unaffected (Fig 7D and 7E). Tox2 and Hk2 expression was reduced in CCR6⁺ gILC3 but not in NCR⁺ gILC3 isolated from Rorc^CreHif1a^fl/fl mice (S7G and Fig 7F), suggesting that HIF-1α-mediated Tox2 and Hk2 induction is restricted to CCR6⁺ ILC3 in vivo. Thus, the results with HIF-1α deficiency can only partially explain the defects seen in Tox2^−/− mice where both NCR⁺ and CCR6⁺ gILC3 are affected.

Next, we determined the role of another gut-specific factor, IL-17 which is abundant in small intestinal lamina propria and plays an important role in gut barrier function at steady state and after injury and infection^4,5,59,60. Synergy between hypoxia and IL-17 was reported in keratinocytes in a wound-repair model⁶¹. We therefore explored the role of IL-17A in regulating Tox2 expression in ILC3. We observed that IL-17A was sufficient to induce Tox2 in both ILC3 subsets (Fig 7G) as well as in HIF-1α-deficient ILC3 subsets (Fig 7H), suggesting IL-17A regulates Tox2 expression in both ILC3 subsets in a HIF-1α-independent manner. We analyzed Il17a-germline deficient mice and found that unlike the results with HIF-1α deficiency, IL17A deficiency resulted in reduced numbers of both NCR⁺ and CCR6⁺ gILC3 subsets, while mLN-ILC3 numbers remained unaltered (Fig 7I and 7J for ILC3 subsets). Furthermore, Tox2 and Hk2 expression were reduced in Il17a^−/− gILC3 (S7H and Fig 7K for ILC3 subsets) compared to the WT control, indicating that IL-17A is required for induction of Tox2 expression in both ILC3 subsets.

IL-17 stabilizes HIF-1α in keratinocytes and fibroblasts^61,62. To delineate the relationship between HIF-1α and IL-17A in gILC3, we assessed expression of HIF-1α in Il17a^−/− mice and detected a reduction of HIF-1α expression in gILC3 (Fig 7L and Fig 7M for ILC3 subsets).

Our results thus establish that HIF-1α and the cytokine IL-17A can each induce Tox2 expression in gILC3. The results are consistent with the notion that ILC3 sense higher levels of IL-17 in the gut microenvironment at steady state, and respond by increasing expression of Tox2 and Hk2, which in turn supports glycolysis in gILC3 (S7I). Together, our data identify Tox2 as a critical regulator of metabolic programming and survival of ILC3 in the gut environment.

DISCUSSION

We have identified an essential role for Tox2 in the persistence of ILC3 in gut, but not at central sites. Tox2 expression was higher in gILC3 compared to mLN-ILC3 counterparts. The numbers of NCR⁺ and CCR6⁺ ILC3 were greatly reduced in gut but not central sites in mice lacking Tox2 in ILC3. Consistently, Tox2^−/− mice failed to effectively control C. rodentium infection.

We observed increased expression of Hk2 in gILC3 as compared to mLN-ILC3; this induction was Tox2-dependent and ILC3-specific as it was not observed in other immune cell types in gut. Because hexokinases are rate-limiting enzymes for glycolysis, we assessed the metabolic dependence of both central and gut ILC3 on glycolysis and mitochondrial respiration. gILC3 showed reduced dependence on oxidative phosphorylation but increased glycolytic ability compared to mLN-ILC3. The increased glycolytic ability of gILC3 required Tox2, as it was impaired in Tox2^−/− gILC3. Forced Tox2 expression in the ILC3 cell line MNK-3 increased basal ECAR and lactate production, further highlighting the role of Tox2 in glycolytic metabolism of ILC3. Ectopic expression of Hk2 in Tox2^−/− BM progenitors rescued the ability of Tox2^−/− ILC3 to persist in the gut, supporting the conclusion that Hk2 is a critical target of Tox2 which is essential for persistence of ILC3 in gut. Indeed, Tox2^−/− gILC3 exhibited defective survival ex-vivo, even in the presence of high concentrations of glucose. The survival could be at least partially restored by supplementing Tox2^−/− gILC3 cultures with pyruvate or acetate, that are metabolic intermediates downstream of Hk2.

We further explored the rationale underlying this metabolic adaptation of gut resident ILC3. Gut injury and inflammation generate hypoxic environments within which immune cells must execute their protective functions^18,28,61. We observed higher HIF-1α expression in gILC3 compared to mLN counterparts. Furthermore, we found that hypoxia was sufficient to induce expression of Tox2 and Hk2 in ILC3. We found that HIF-1α is required in CCR6⁺ ILC3 but not NCR⁺ ILC3 in vivo; as in the absence of HIF-1α, only CCR6⁺ ILC3 showed lower Tox2 and Hk2 expression. Whether these differences are intrinsic to ILC3 subsets or reflect their occupancy of distinct anatomical locations²⁵ requires further investigation.

Because HIF-1α appeared dispensable in NCR⁺ gILC3 for Tox2 induction in vivo, and the effects of HIF-1α ablation in CCR6⁺ gILC3 for Tox2 induction were modest, we sought other mechanisms which could induce Tox2 expression distinct from HIF-1α. IL-17 is abundant in small intestinal lamina propria at steady state^4,5. We found that IL-17A was required for appropriate induction of HIF-1α, Tox2 and Hk2 in both NCR⁺ and CCR6⁺ ILC3. Additionally, IL-17A could increase Tox2 expression in ILC3 independently of HIF-1α in vitro.

Because glycolysis is often required for the maintenance of cellular energetics under hypoxic conditions^63,64, our results suggest that Tox2 promotes lineage-specific metabolic adaptation that endows ILC3 with the tailored ability to survive and function at IL-17 rich sites, in anticipation of hypoxic conditions that are created by injury and infection. Indeed, alterations in the metabolic status of gILC3 were found to be associated with their function during infection^17,65, although these studies did not compare gILC3 with ILC3 at other sites. Our results provide a mechanism for how metabolic changes might occur, as they predict gILC3 to become more glycolytic in presence of high levels of IL-17 and possibly other gut signals present during infection.

ILC3 can play important roles in immune responses to tumors. ILC3 were the first cells to infiltrate tumors after cisplatin-based chemotherapy, subsequently recruiting T cells and so improving responses to immune checkpoint blockade¹¹. Tumors can be IL-17 rich⁶², and are frequently hypoxic⁶⁶, resulting in a shift in ATP production from OXPHOS to glycolysis⁶⁷. Thus, it is tempting to speculate that tumor-infiltrating ILC3 exhibit metabolic adaptations that are similar to gILC3, and further, that the ability of ILC3 to program glycolytic metabolism in response to IL-17 and hypoxia allows them to pioneer colonization of hypoxic sites such as tumors.

Collectively, our present findings demonstrate that metabolic adaptation of gILC3 is under lineage-specific transcriptional control. Tox2 controls persistence of gILC3 by regulating expression of Hk2 and promoting glycolysis. We speculate that enhanced glycolysis in gILC3 is a tissue-specific and lineage-specific adaptation that enables their rapid function under conditions of hypoxia generated by infection or other injury. Thus, ILC3 possess a tailored mechanism involving Tox2 and Hk2, that allows gILC3 to respond in a manner that differs from the response of other lymphocytes residing in the intestinal niche.

Limitations of the study

Our study implicates regulation of Hk2 and glycolysis as the mechanism by which Tox2 controls the persistence of gILC3 at steady-state. We speculate that enhanced glycolysis is a tissue-specific and lineage-specific adaptation in gILC3 that enables their function in hypoxic environments and at IL-17 rich sites, such as occur in infections and tumors. However, we did not test the Tox2-Hk2 axis in these settings. Although ectopic expression of Hk2 was sufficient to rescue persistence of Tox2^−/− gILC3, Tox2 may have additional gene targets in ILC3 that require further investigation. Signals distinct from IL-17 and hypoxia may also affect Tox2 expression and the consequent metabolic adaptation in ILC3; these are appropriate subjects for future studies.

STAR METHODS

Resource Availability

Lead contact

Further information and requests for resources and reagents should be directed to the Lead Contact, Avinash Bhandoola (avinash.bhandoola@nih.gov).

Materials availability

All requests for resources and reagents generated in this study will be fulfilled by the lead contact.

Data and code availability

scRNA-seq and scATAC-seq data are deposited to GEO and are publicly available with citation accession PRJNA870085.

This paper does not report original code.

Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Experimental model and study participant details

Mice

B6 (CD45.2), B6-Ly5.1 (CD45.1), B6.129-Hif1a^tm3Rsjo/J (strain#:007561), B6.129X1-Gt(ROSA)26Sor^tm1(EYFP)Cos/J (strain#:006148) and Rag1^−/− mice were purchased from Jackson Laboratory. Tox2^−/− mice have previously been described. Tox2 flox mice were generated by using Cas9 protein, two synthetically modified guide RNAs (TCAAGGGCCTCCTGACTAGT and TGGAGATGTGTCCAAAATAG, Synthego) to generate 2 cuts, one on each side of Exon 5, and a single-stranded DNA donor template containing this region, with loxP sites on each side, along with ~500 bp of homology. Guide RNA sequences were identified using sgRNA Scorer 2.0⁶⁸ and single-stranded DNA was generated using the Guide-it Long ssDNA production system (Takara). Microinjections to generate knockout mice by CRISPR-Cas9 were performed by Transgenics/Cryopreservation-Laboratory Animal Science Program National Cancer Institute core on a C57BL/6NCr background. ER^T2Cre mouse⁶⁹ (Taconic line 10471), IL7r^Cre47, Tox^−/−70, Tcf7^EGFP71, Il17a^−/−72 and Rorc^Cre73 mice were previously been described. Tox2^−/−.Tcf7^GFP mice were generated for EILP quantification in bone marrow. NOD-scid IL2Rgamma^null (NSG) mice were obtained from National Cancer Institute. All mice except NSG mice are maintained on C57/BL6 background. Littermates were used as WT control mice for all experiments. Mice were 6–12 weeks of age and of either sex, except for assessment of thymic ILC3 when 7-day old mice were used. Animal procedures were approved by relevant National Institutes of Health Animal Care and Use Committees.

Plasmids, cell culture and transduction

Full length mouse Tox2 cDNA (Cat# MR212216; Origene) was cloned into MSCV-IRES-GFP (MIGR1) and retroviral supernatant preparation was performed as described⁷⁴.

Hk2 cDNA (Cat# MC202280; Origene) along with a synthetic gene fragment containing 2A-GFP (Twist Biosciences) was cloned together into BamHI/EcoRI digested FUGW plasmid (Gift from David Baltimore, 14883, Addgene). Generation of infective lentiviral particles was performed in HEK293T (American Type Culture Collection) cells as described⁷⁵.

MNK-3 cell line was kindly provided by Juan Carlos Zuniga-Pflucker, Jorge Henao-Mejia, Michael Michieletto and the late James Carlyle.

Method details

Isolation of hematopoietic cells

For Lung: Mice were euthanized with CO₂, perfused through the left ventricle of the heart by injecting 10 ml PBS. Lungs were cut into small pieces and digested in Hank’s balanced salt solution (HBSS) containing 0.025mg/ml Liberase TL (Roche Diagnostics) and 10U/ml DNase I (Roche Diagnostics) for 30 min at 37°C on a shaker (220 rpm). Cells were filtered by using a cell strainer. Leukocytes were enriched by centrifugation (10 min, 650 g) on a 40%/80 % Percoll gradient (GE Healthcare), and RBCs were lysed by ammonium chloride solution.

For Lamina propria: Peyer’s patches were removed from small intestines. Intestinal tissues were washed, cut into 1–1.5 cm sections, incubated with Hanks’ Balanced Salt Solution (HBSS) containing 5mM EDTA and 1mM DTT at 37°C for 20 mins to remove epithelial cells. For cell isolation from colon, tissue sections were incubated with RPMI media containing Pen/Strep, L-Glutamine, NEAA, Sodium Pyruvate and HEPES (GIBCO). The digested tissue was minced and dissociated in RPMI containing 10% FBS, liberase TL (1 mg/ml; Roche) and DNase1 (100 μg/ml; Roche) with constant stirring at 37°C for 30 min.

Fetal intestines from E15.5 were processed as described earlier³⁸.

For BM cells: tibia, femur, and pelvic bones were flushed with 10 ml RPMI-containing 2% FBS and penicillin-streptomycin (100 U/ml) using 27.5-gauge needle, the single-cell suspension was filtered through a 70 μm cell strainer and washed once with 5 ml of the same media.

For lymph nodes, spleen and thymus: Tissues were mechanically disrupted. Single-cell suspension was filtered through a 70 μm cell strainer and washed once with 5 ml of the same media. This was followed by RBC lysis by ACK lysis buffer (Quality Biological).

Liver tissues were mashed through 70 μm cell strainer and leukocytes were collected at the interface of a 40%/80% Percoll gradient (GE Healthcare).

Antibodies and Flow Cytometry

Bone marrow cell suspensions were incubated with a mix of purified rat, mouse and hamster immunoglobulin G (IgG) before the addition of specific antibodies. Antibodies specific for Ly-6D (49H4), B220 (RA3–6B3), CD19 (1D3), Mac-1 (M1/70), Gr-1 (8C5), CD11c (N418), Ter119 (TER119), NK1.1 (PK136), CD5 (53–7.3), CD3ε (2C11), CD8α (53–6.72), CD8β (H35–17.2), CD4 (GK1.5), TCRβ (H57), TCRγδ (GL-3), Kit (2B8), CCR7 (4B12), NKp46 (29A1.4), Sca-1 (D7), Thy1.2 (53–2.1), α₄β₇ (DATK32), IL-7Rα (A7R34), CD25 (PC61.5), CD45.1 (A20), CD45.2 (104), EpCAM (G8.8), α₄β₇ (DATK32), Flt3 (A2F10.1), CCR6 (G034E3), CD49a (HMα1), CD49b (DX5), CD5 (53–7.3), KLRG1 (2F1) were purchased from ThermoFisher Scientific or BD Biosciences or BioLegend. Antibodies specific for transcription factors and cytokines: RORγt (Q31–378), GATA-3 (TWAJ); T-bet (eBio4B10(4B10)), EOMES (Dan11mag), HIF-1α (D1S7W), Bcl6 (BCL-DWN), IFN-γ (XMG1.2), IL-22 (IL22JOP) and IL-17 (eBio17b7) from ThermoFisher Scientific, Cell signaling and BD biosciences. HK-I (ab150423) and HK-II (ab209847) were purchased from Abcam.

ILC lineage ‘cocktail’ used to sort ALP (LinILC) for bone marrow is a mix of the following antibodies: anti-Ly-6D, B220, CD19, Mac-1, Gr-1, CD11c, Ter119, NK1.1, CD5, CD3ε, CD8α, CD8β, CD4, TCRβ and TCRγδ. LinILC2: B220, CD19, NK1.1, Gr-1, CD11b, CD11c, Ter119, CD5, CD3ε, TCRβ and TCRγδ. ILC3 lineage cocktail ‘Lin’: B220, CD19, NK1.1, Gr-1, CD11b, Ter119, CD5, CD3ε, TCRβ and TCRγδ. ILC3 lineage cocktail for Fig 1B and S1A ‘Lin’: B220, CD19, NK1.1, KLRG1, Gr-1, CD11b, Ter119, CD5, CD3ε, TCRβ and TCRγδ. ‘Lin’ for fate mapping experiment: B220, CD19, Gr-1, Ter119, CD5, CD3ε, TCRβ and TCRγδ. All analyses were done on singlet live cells. For intracellular staining to detect cytokines and transcription factors, cells were first stained for cell surface molecules and permeabilized using the eBioscience’s transcription factor staining buffer set (cat: 00–5523-00) according to the manufacturer’s instructions. For fixation of fate-map YFP signal, cells were fixed 20 min at room temperature in 2% paraformaldehyde and stained in BD Bioscience perm buffer⁴⁸. FITC Active Caspase-3 Apoptosis kit was used to measure cleaved Caspase-3 (BD Biosciences, cat: 550480). APC BrdU Flow kit (552598; BD Bioscience) was used to measure proliferation. Live/dead discrimination was performed by staining with DAPI or LIVE/DEAD Fixable Blue Dead Cell stain kit (L-34962). Samples were acquired using a flow cytometer (LSRFortessa; BD) and analyzed using FlowJo software (Tree Star). Cell populations were sorted using an Aria flow cytometer (BD). All cell sort purities were >95%. Absolute cell numbers were obtained using an Accuri C6 PLUS flow cytometer (BD).

Progenitor population and mature ILC definition

ALP: LinILC⁻Kit⁺Flt3^hiCD127⁺α₄β₇⁻; EILP: LinILC⁻Kit⁺α₄β₇⁺Tcf7^GFP+Thy1⁻; ILCP: LinILC⁻Kit⁺α₄β₇⁺Tcf7^GFP+Thy1⁺CD127⁺ (all in bone marrow); and ILCP: CD45⁺EpCAM⁻CD5⁻CD19⁻CD11b⁻CD127⁺Kit⁺α₄β₇⁺CD4⁻NK1.1⁻ST2⁻ and LTi: CD45⁺CD5⁻CD19⁻CD11b⁻NK1.1⁻CD127⁺Kit⁺CD4⁺ (in fetal gut) were described earlier^34,38.

For Spleen and Liver, NK cells: Lin⁻ (Lin for NK and ILC1: CD3ε, CD5, CD19, B220) NK1.1⁺NKp46⁺DX5⁺CD49a⁻; ILC1: Lin⁻NK1.1⁺NKp46⁺DX5⁻CD49a⁺; for Gut and Colon, NK: Lin⁻EpCAM⁻CD45⁺NK1.1⁺T-bet⁺Eomes⁺ ILC1: Lin⁻EpCAM⁻CD45⁺NK1.1⁺T-bet⁺Eomes⁻. For the fate mapping experiment in mLN and gut, NK cells: Lin⁻EpCAM⁻CD45⁺NK1.1⁺Eomes⁺ and ILC1⁴⁸: Lin⁻EpCAM⁻CD45⁺NK1.1⁺T-bet⁺Eomes⁻ROR-γt⁻.

Lung ILC2 (surface stain): LinILC2⁻CD45⁺CD127⁺Thy1⁺ST2⁺CD25⁺; Lung ILC2 (intracellular stain): LinILC2⁻CD45⁺Thy1⁺CD127⁺GATA-3⁺; Gut and Colon ILC2 (intracellular stain): Lin⁻ (Lin: CD3ε, CD5, CD19, B220) EpCAM⁻CD45⁺NK1.1⁻ROR-γt⁻GATA-3⁺.

mLN and gut ILC3 using surface stain: Lin⁻CD45⁺EpCAM⁻Kit⁺CD127⁺; ILC3 in d7 thymus: LinCD45⁺CD127⁺ROR-γt⁺; ILC3 in gut, mLN, caecum, colon, Peyer’s patch, pLN, spleen and for fate mapping experiment were defined by intracellular staining: Lin⁻CD45⁺EpCAM⁻CD127⁺ROR-γt⁺.

Bone Marrow Transplantation

For competitive bone marrow chimeras⁷⁶, 18,000 sorted LSK cells from donor mice (CD45.2) were mixed with 2,000 competitor WT LSK cells (CD45.1 or CD45.1⁺CD45.2⁺) and intravenously transferred into lethally irradiated (900 rads) WT recipient mice (CD45.1).

Cell culture and transduction

MNK-3 cells were cultured as described⁵⁸. MNK-3 cells were transduced using control GFP virus, Tox-2 virus and Hk2 virus with 4 μg/ml of polybrene (Cat# 7711/10; R&D SYSTEMS) at 900 g for 2 hours, following which viral supernatant was removed after 16 hours and GFP+ cells were sorted after 72 hours culture with 10 ng/ml of IL-7 and IL-15 (Peprotech). Sorted GFP+ cells were rested for additional 2 days before performing Seahorse assay.

Bone marrow LSK and ALP were transduced using Retronectin (Takara). 24-well plates were coated with 50 μg/ml Retronectin according to the manufacturer’s instructions. High-titre GFP or Hk2-GFP lentiviral supernatants along with sorted bone marrow LSK and ALP in IL-3, IL-6 (10 ng/ml), SCF (100 ng/ml), Flt3-ligand, IL-7 (30 ng/ml) and polybrene (6 μg/ml) were added into wells and centrifuged at 25°C for 2h, following which viral supernatants were removed after 16h incubation. Fresh medium with cytokines (IL-7, SCF and Flt3-ligand (30 ng/ml)) was added and transduced cells were sorted 24 hours post-infection.

Seahorse assay

Oxygen consumption rate (OCR) and Extracellular acidification rate (ECAR) were measured using the XFe96 Extracellular Flux Analyzer (Seahorse Biosciences, Agilent). Agilent XFe96 Extracellular Flux Assay cartridges were hydrated overnight in pure water, then water was replaced by 200 μL of Seahorse Calibrant and incubated for at least 1h at 37°C (non-CO₂). MNK-3 cells (10⁵/well) were harvested 2 days post-sort after resting in media. Murine T-cells from C57Bl6 mice were selected from mice spleens (EasyStep mouse T-cell isolation kit #1951A) and cultured with IL2 only (1e5/well) for 3 days. Cells were resuspended in 80 μL Agilent XF media (buffered RPMI pH7.4) in the presence of glucose (11 mM) and L-glutamine (2 mM), and pyruvate (1 mM) and seeded in 96-well Seahorse XF96 cell culture microplates (Seahorse Biosciences, Agilent) coated with CellTak (Corning 354240) 20 min at room temperature. Cells were incubated at 37°C (non-CO₂) for 30 minutes, then the total volume of each well was set to 180 μL by adding 100 μL of XF media. Oligomycin (Sigma-Aldrich #04876, 2 μM), FCCP (Sigma-Aldrich #C2920, 1 μM), Rotenone (Sigma-Aldrich R8875, 1 μM), Antimycin A (Sigma-Aldrich A8674, 1 μM) were injected as indicated.

Adoptive transfer of ILC3 and transduced progenitors

Sorted 5000 gILC3 or mLN-ILC3 were adoptively transferred into NSG mice by intravenous injection. The donor population was sorted from spleen and gut one week post-transfer.

20,000–50,000 transduced BM LSK and ALP were adoptively transferred into NSG mice by intravenous injection. Mice were analyzed for donor-derived ILC3 (defined as CD45.2⁺GFP⁺CD11b⁻B220⁻CD19⁻NK1.1⁻CD5⁻CD3ε⁻Kit⁺CD127⁺ using surface staining and CD45.2⁺CD11b⁻B220⁻CD19⁻NK1.1⁻CD5⁻CD3ε⁻CD127⁺ROR-γt⁺ using intracellular staining); NK cells (defined as CD45.2⁺GFP⁺B220⁻CD19⁻CD5⁻CD3ε⁻NK1.1⁺) and ILC2 [Lin⁻(CD11b, B220, CD19, NK1.1, CD5, CD3ε)CD45.2⁺CD127⁺GATA-3⁺] in gut were assessed 12 week post-reconstitution.

ILC3 in vitro cultures

ILC3 culture with metabolites: Sorted 2000 gILC3 or 5000 mLN-ILC3 were cultured on irradiated OP9 stromal layers in α-MEM supplemented with 20% FBS, glutamine, penicillin, streptomycin, IL-7, IL-2 (30ng/ml) were added without or with glucose (25 mM; G8270), acetate (5 mM; S5636) from Sigma-Aldrich and pyruvate (10 mM; 11360–070) from ThermoFisher Scientific.

ILC3 culture with metabolic inhibitors: mLN and gut lymphocytes were treated for 30 min with Control, 2-Deoxy-D-Glucose (2DG, final concentration 100mM), Oligomycin (final concentration 1μM) or both. Puromycin (10 μg/ml) was added for another 30 min. After incubation cells were stained for puromycin as described below.

For PMA/Ionomycin stimulation: mLN and gut lymphocytes were stimulated with 50 ng/ml PMA and 500 ng/ml Ionomycin from Sigma-Aldrich for 2.5 hours and Brefeldin A (ThermoFisher Scientific) was added in the last 2 hours of stimulation.

For hypoxia experiments: Sorted gILC3 (2000 cells) and mLN-ILC3 (5000 cells) were incubated for 6h in 5% O₂ to mimic hypoxic condition in the gut¹⁸ or maintained in control atmospheric oxygen conditions (21% O₂). Tox2 and Hk2 expression was then evaluated by qPCR.

Puromycin staining

The antibiotic puromycin (PMY) is an aminoacyl-tRNA analog that binds to the ribosome A-site and is covalently incorporated onto the 3′end of the nascent polypeptide resulting in chain termination. After fixation and permeabilization, the puromycylated polypeptides can be detected by an intracellular stain with an antibody specific to puromycin. The intracellular RPM signal from PMY treated mice minus the RPM signal from PBS treatment generates a measure of the total amount of translation⁷⁷. Mice were injected intravenously with 100 μl of puromycin (PMY) (10 mg/ml) in PBS (Life Technologies) that was warmed to 37 °C. Mice were sacrificed 30 min post PMY injection and mLN and gut were processed as stated above. Samples were stained with LIVE/DEAD Fixable Blue Dead Cell Stain Kit (ThermoFisher Scientific) on ice for 10 min, then fixed and permeabilized in fixation/permeabilization buffer (1% paraformaldehyde, ThermoFisher Scientific; 0.0075% digitonin, Wako) in PBS for 20 min at 4 °C. Cells were then labeled with anti-PMY-AF488 antibody (J.Yewdell in house generated) for 1 hour.

BrdU staining

Mice were injected intraperitoneally (i.p) with BrdU (1.5 mg) 18 hours before sacrifice, and mLN and gut were processed as stated above. Cell suspensions were intracellularly stained for BrdU using the BD bioscience APC BrdU flow kit (552598), according to the manufacturer’s instructions.

Tamoxifen treatment

Tamoxifen (Sigma-Aldrich) was dissolved in corn oil (Sigma-Aldrich C8267–500ml) at a concentration of 25 mg/ml by shaking 6 h at 37 °C. Adult mice were administered with 200 μl tamoxifen (5 mg per mouse) via intraperitoneal injection daily, 5 times on alternate days. Mice were sacrificed 3 weeks post treatment⁷⁸.

SILT quantification

Small intestines were transversally cut into three equal-sized segments (duodenum, jejunum and ileum). After removal of contents by using forceps and performing washes with PBS, they were cut longitudinally. The opened segments were flattened and incubated in 4% paraformaldehyde/PBS for 30 min. Then, tissue sections were swiss-rolled⁷⁹ and incubated in paraformaldehyde/PBS at 4 °C overnight. Fixed intestines were first washed with PBS and dehydrated performing an incubation with 15% sucrose for 24 hours, followed by another incubation in 30% sucrose for 24 hours. Then, swiss-rolls were embedded in Tissue-Tek O.C.T. compound (Sakura Finetek USA). After discarding approximately the first 800μm of the embedded tissue by using cryostat HM525 (Thermo Scientific), five sections of 12 μm in thickness were cut, each section was 100 μm apart and stained using antibodies B220 (RA3–6B2, 1:50, eBioscience), CD3ε (500a2, 1:50, eBioscience). YFP+ cells represented RORγt expression. Sections were counterstained with DAPI and mounted using Prolong Glass Antifade Mountant (Invitrogen)⁸⁰. Whole stained tissue sections were imaged at 20X magnification by using SoRa CSU-W1 spinning disk microscope (Nikon). NSI-Element software (Nikon) was used to stitch approximately 150–175 image frames by applying the blending method to form the whole tissue section. To quantify CP and ILF, we focused on duodenum because in our pilot experiment, the difference in CP number was prominent in duodenum than in other sections analyzed however the number of ILF was comparable across the sections analyzed between WT and Tox2^−/− mice.CPs and ILFs were counted in stitched images in 5 sections taken at 100μm intervals through the depth of the duodenum.

Quantification of glucose and mitochondrial dependence

Glucose and Mitochondrial dependencies along with glycolytic and FAO and AAO capacities are calculated as previously described⁵⁵.

Glucose dependence (%) = 100*((Puromycin MFI of Control group - Puromycin MFI of 2-DG treated group)/(Puromycin MFI of Control group - Puromycin MFI of both inhibitors treated group))

Mitochondrial dependence (%) = 100*((Puromycin MFI of Control group - Puromycin MFI of Oligomycin treated group)/ (Puromycin MFI of Control group - Puromycin MFI of both inhibitors treated group))

Glycolytic capacity (%) = 100-Mitochondrial dependence

FAO and AAO capacity (%) = 100-Glucose dependence

Puromycin MFI for each group is calculated by subtracting values of puromycin MFI of each group from the control MFI (without puromycin).

Citrobacter infection

Citrobacter rodentium (strain ICC169) was cultured overnight in Luria-Bertani (LB) broth at 37°C as previously described⁸¹. Mice were infected orally by gavage with 10⁸ colony-forming units (CFU) resuspended in sterile PBS. The mice were assessed on day 6 for ILC3 function. To determine the bacterial load in spleen, liver and feces, tissues were homogenized, serially diluted and plated onto LB-agar plates containing nalidixic acid, and the colonies were counted after incubation overnight at 37°C.

Lactate secretion assay

MNK-3 cells transduced with GFP, Tox2.GFP and Hk2. GFP were sorted for GFP+ cells. 10⁵ sorted GFP+ cells along with untransduced MNK-3 cells were cultured for 24 hours in high glucose media similar as MNK-3 cell media described above. Culture supernatant was harvested and lactate was measured using the colorimetric L-Lactate Assay Kit (Abcam).

Quantitative RT-PCR

A total of 5000 sorted mLN-ILC3 (Lin⁻EpCAM⁻CD45⁺CD127⁺Kit⁺), gILC3 (Lin⁻EpCAM⁻CD45⁺CD127⁺Kit⁺), along with ILC3 subsets (pregated on ILC3 and further gated on NKp46 for NCR⁺ ILC3 and CCR6 for CCR6⁺ ILC3), ALP (LinILC⁻Kit⁺Flt3^hiCD127⁺α₄β₇⁻), ILCP (LinILC⁻Kit⁺α₄β₇⁺Tcf7^GFP+Thy1⁺CD127⁺) from bone marrow, ILCP (CD45⁺EpCAM⁻CD5⁻CD19⁻CD11b⁻CD127⁺Kit⁺α₄β₇⁺CD4⁻NK1.1⁻ST2⁻) and LTi (CD45⁺CD5⁻CD19⁻CD11b⁻NK1.1⁻CD127⁺Kit⁺CD4⁺) from fetal gut, ILC2 from lung (LinILC2⁻CD45⁺CD127⁺Thy1⁺ST2⁺CD25⁺), ILC1 (Lin⁻NK1.1⁺NKp46⁺DX5⁻CD49a⁺) and NK cells (Lin⁻NK1.1⁺NKp46⁺DX5⁺CD49a⁻) from liver were lysed in RLT buffer, and total RNA was isolated using the RNeasy Micro Kit (Qiagen) according to the manufacturer’s instructions. Reverse transcription was performed using the SuperScript VILO cDNA Synthesis Kit (Invitrogen). Quantitative PCR was performed on a StepOne Plus Real-Time PCR System (Applied Biosystems) using TaqMan probes designed against Tox2 (Mm01241014; ThermoFisher scientific) and Gapdh (Mm9999915; ThermoFisher scientific), Hk2 (Mm00443385). Results were analyzed using the ΔΔ cycle threshold method and plotted as relative expression normalized to β-actin.

scRNA-Seq and analysis

Cells from mLN and gut were isolated by sorting from WT or Tox2^−/− mice by gating for Lin⁻CD5⁻CD45⁺kit⁺ to mainly enrich for ILC3. Sorted cells were loaded on the 10X Chromium platform (10X genomics). Libraries were constructed using the Chromium Single Cell 3’kits according to the manufacturer’s instructions (v3 chemistry, 10x Genomics). Libraries were sequenced with a NextSeq (v3 chemistry, Illumina). A primary analysis was performed with the Cell Ranger (version 6.0.0) software using the default parameters. Data set in Fig. 4: WT mLN (2585 cells), Tox2^−/− mLN (2816 cells), WT gut (4134 cells), Tox2^−/− gut (3985 cells), expressing a median of 1500–4000 genes per sample and all captured at the same time. The median number of UMI counts ranged between 4000 and 9500 per cell.

Single-cell analysis was performed using Seurat (version 4.1.0)⁸², applying default settings unless otherwise stated. We excluded cells with >10% mitochondrial gene content and >5000 genes per cell. Mitochondrial content and cell-cycle related genes were regressed out. mLN and gut samples were processed separately. We applied Seurat’s NormalizeData function to normalize read counts. We performed UMAP dimensional reduction using 20 principal components. Cells were clustered using Seurat’s FindClusters function with a resolution of 0.5. Contaminants were identified based on gene expression signatures determined using the function FindAllMarkers with a minimum Log2 fold change threshold of 0.25 and the Wilcoxon Rank Sum test. Based on the DEG analysis, clusters marked by significant expression of the following genes were removed: Cd19 for B cells, CD3e for T-cells, Epcam for epithelial cells, Adgre1 for monocyte-macrophage, Eomes and high levels of Klrb1c for NK and ILC1, and high expression of Gata3 and Il17rb for ILC2⁵¹. Cells were then clustered using Seurat’s FindClusters function at a resolution of 0.1. Clusters were annotated as NCR⁺ ILC3 by expression of Ncr1 and Tbx21 and as CCR6⁺ ILC3 by expression of Ccr6 and Cd4. For both mLN and gut samples separately, DEG comparing WT versus Tox2^−/− ILC3 was generated using FindMarkers function, with min.pct set to 0.12 and significant genes having an adjusted p-value<0.05, reported as Supplementary Table 1. Heatmap shows the gene expression of averaged RNA normalized Z scores.

To confirm the identity of mLN-ILC3, we further integrated our mLN dataset with a published mLN-ILC3 dataset⁵¹ which used RORγt-eGFP BAC-transgenic mice (see S4A). We normalized read counts and regressed out mitochondrial and cell-cycle related genes for this analysis. The integration was performed with the FindIntegrationAnchors and IntegrateData functions using the canonical correlation analyses (cca) approach, with all default variables. We performed UMAP dimensional reduction using 20 principal components. The list of genes identifying NCR⁺ and CCR6⁺ ILC3 subsets was generated using the FindMarkers function with default parameters. The ILC3 subset gene list from mLN was cross-referenced with mLN-ILC3 clusters in published work⁵¹, and the shared set of genes was retained. The lists of genes defining ILC3 subsets from mLN and gut are given in Supplementary Table 3. Violin plots shown in the figures were made using Seurat’s normalized data. Average gene expression from cells of the same cluster was generated using the Seurat’s Average Expression function. Gene ontology pathway analysis of DEGs was carried out with the ClusterProfiler R package (version 4.0.5).

scATAC-Seq and analysis

The libraries obtained from different experimental conditions were aggregated using the cellranger-atac aggr pipeline. Subsequently, the fragment files were loaded into R using Signac⁸³. Only high-quality cells were retained, based on the following parameters: cells needed to have more than 3000 and less than 30000 number of peaks, the percentage of reads in peaks needed to be higher than 15, the transcription start site enrichment needed to be higher than 15 and having a nucleosome signal less than 4 was required. The data was analyzed using a standard Signac workflow, using a q0 cutoff for variable features and excluding the first latent semantic indexing dimension (LSI). Gene activity was predicted and leveraged to isolate clusters containing ILC3s. The ILC3s were reanalyzed using the same workflow, and annotations were transferred from the scRNAseq data using the FindTransferAnchors, TransferData and AddMetaData functions. Specifically, canonical correlation analysis was used to identify “anchors” and the first LSI dimension was excluded from the TransferData function. Annotated UMAPs were generated using the DimPlot function and the coverage plots were generated using the CoveragePlot function. Differentially abundant regions were assessed using the FindMarkers function, using a logistic regression test and providing “nCount_peaks” as latent variable.

Quantification and statistical analysis

Statistical significance was performed with GraphPad Prism. Differences between groups were determined by a two-tailed unpaired Student’s t test or by one-way ANOVA with multiple comparisons. For statistical significance, asterisks indicate the level of significance, * p<0.05, ** p<0.01, *** p<0.001, ****p<0.0001.

Supplementary Material

1

Supplementary Table 1. List of differentially expressed genes (DEGs) in mLN-ILC3 and gILC3 subsets isolated from WT and Tox2^−/− mice (Related to Figure 4).

2

Supplementary Table 2. GO analysis of the DEGs in gILC3 from WT and Tox2^−/− mice (Related to Figure 4).

4

Supplementary Table 3. List of genes defining NCR⁺ and CCR6⁺ ILC3 subsets from mLN and gut (Related to STAR Methods).

Key resources table.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Ly-6D Antibody, anti-mouse, Biotin	Miltenyi Biotec	130-115-310
CD45R (B220) Monoclonal Antibody (RA3-6B2), Biotin	ThermoFisher Scientific	13-0452-86
CD19 Monoclonal Antibody (eBio1D3 (1D3)), Biotin	ThermoFisher Scientific	13-0193-85
CD11b Monoclonal Antibody (M1/70), Biotin	ThermoFisher Scientific	13-0112-86
Ly-6G/Ly-6C Monoclonal Antibody (RB6-8C5), Biotin	ThermoFisher Scientific	13-5931-86
CD11c Monoclonal Antibody (N418), Biotin	ThermoFisher Scientific	13-0114-85
TER-119 Monoclonal Antibody (TER-119), Biotin	ThermoFisher Scientific	13-5921-85
NK1.1 Monoclonal Antibody (PK136),Biotin	ThermoFisher Scientific	13-5941-85
CD5 Monoclonal Antibody (53-7.3), Biotin	ThermoFisher Scientific	13-0051-85
CD3e Monoclonal Antibody (eBio500A2 (500A2)), Biotin	ThermoFisher Scientific	13-0033-86
CD8a Monoclonal Antibody (53-6.7), Biotin	ThermoFisher Scientific	13-0081-85
CD8 beta Monoclonal Antibody (CT-CD8b), Biotin	ThermoFisher Scientific	MA5-17846
CD4 Monoclonal Antibody (GK1.5), Biotin,	ThermoFisher Scientific	13-0043-85
TCR gamma/delta Monoclonal Antibody (eBioGL3 (GL-3, GL3), Biotin	ThermoFisher Scientific	13-5711-85
TCR beta Monoclonal Antibody (H57-597), Biotin	ThermoFisher Scientific	13-5961-85
KLRG1 Monoclonal Antibody (2F1), APC-eFluorTM 780	ThermoFisher Scientific	47-5893-82
Streptavidin eFluorTM 450 Conjugate	ThermoFisher Scientific	48-4317-82
Streptavidin PE Conjugate	ThermoFisher Scientific	12-4317-87
Streptavidin FITC Conjugate	ThermoFisher Scientific	11-4317-87
Streptavidin APC-eFluor™ 780 Conjugate	ThermoFisher Scientific	47-4317-82
CD117 (c-Kit) Monoclonal Antibody (ACK2), Alexa Fluor 700	ThermoFisher Scientific	56-1172-82
CD117 (c-Kit) Monoclonal Antibody (2B8), PE-Cyanine7	ThermoFisher Scientific	25-1171-82
CD127 Monoclonal Antibody (A7R34), APC	ThermoFisher Scientific	17-1271-82
APC/Cyanine7 anti-mouse CD326 (Ep-CAM) Antibody	BioLegend	118218
Brilliant Violet 785™ anti-mouse/human KLRG1 (MAFA) Antibody	BioLegend	138429
FITC anti-mouse CD326 (Ep-CAM) Antibody	BioLegend	118207
PerCP/Cyanine5.5 anti-mouse CD326 (Ep-CAM) Antibody	BioLegend	118220
CD45.2 Monoclonal Antibody (104) eFluorTM 450	ThermoFisher Scientific	48-0454-82
CD45.1 Monoclonal Antibody (A20) PE-eFluorTM 610	ThermoFisher Scientific	61-0453-82
PerCP/Cyanine5.5 anti-mouse CD45.2 Antibody	BioLegend	109828
CD45.2 Monoclonal Antibody (104), PE	ThermoFisher Scientific	12-0454-83
CD45.1 Monoclonal Antibody (A20), PerCP-Cyanine5.5	ThermoFisher Scientific	45-0453-82
CD197 (CCR7) Monoclonal Antibody (3D12), APC	ThermoFisher Scientific	17-1979-42
CD335 (NKp46), PerCP-eFluor™ 710, clone: 29A1.4	ThermoFisher Scientific	46-3351-82
CD335 (NKp46) Monoclonal Antibody (29A1.4), PE, eBioscience™	ThermoFisher Scientific	12-3351-82
APC/Cyanine7 anti-mouse CD335 (NKp46) Antibody	BioLegend	137646
NK1.1 Monoclonal Antibody (PK136), Super Bright 780	ThermoFisher Scientific	78-5941-82
NK1.1 Monoclonal Antibody (PK136), APC-eFluor 780	ThermoFisher Scientific	47-5941-82
Brilliant Violet 605™ anti-mouse CD90.2 (Thy-1.2) Antibody	BioLegend	140318
Alexa Fluor® 700 anti-mouse CD90.2 (Thy-1.2) Antibody	Biolegend	140324
NK1.1 Monoclonal Antibody (PK136), PE-Cyanine7	ThermoFisher Scientific	25-5941-82
Ly-6A/E (Sca-1) Monoclonal Antibody (D7), PerCP-Cyanine5.5	ThermoFisher Scientific	45-5981-82
PE-CF594 Rat Anti-Mouse CD135	BD bioscience	562537
Anti-Mouse Integrin alpha 4 beta 7 (LPAM-1) PerCP-eFluor® 710	ThermoFisher Scientific	46-5887-82
Integrin alpha 4 beta 7 (LPAM-1) Monoclonal Antibody (DATK32 (DATK-32)), PE	ThermoFisher Scientific	12-5887-83
FITC anti-mouse CD25 Antibody	Biolegend	101908
PE/Cy7 anti-mouse CD196 (CCR6) Antibody	Biolegend	129816
APC anti-mouse CD49a Antibody	Biolegend	142606
CD49b (Integrin alpha 2) Monoclonal Antibody (DX5), APC-eFluor 780	ThermoFisher Scientific	47-5971-82
BV605 Rat anti-mouse CD5	BD biosciences	563194
Pacific Blue™ anti-mouse CD5 Antibody	Biolegend	100642
Brilliant Violet 605™ anti-mouse CD3 Antibody	Biolegend	100237
CD3e Monoclonal Antibody (145-2C11), eFluor 450	ThermoFisher Scientific	48-0031-82
Brilliant Violet 605™ anti-mouse CD19 Antibody	Biolegend	115540
APC/Cy7 anti-mouse/human CD45R/B220 Antibody	Biolegend	103224
PE-CF594 Mouse Anti-Mouse RORγt	BD Biosciences	562684
T-bet Monoclonal Antibody (eBio4B10 (4B10)), eFluor 450	ThermoFisher Scientific	48-5825-82
Gata-3 Monoclonal Antibody (TWAJ), PE, eBioscience™	ThermoFisher Scientific	12-9966-42
Gata-3 Monoclonal Antibody (TWAJ), eFluor 660	ThermoFisher Scientific	50-9966-42
EOMES Monoclonal Antibody (Dan11mag), Alexa Fluor 488	ThermoFisher Scientific	53-4875-82
HIF-1α (D1S7W) XP® Rabbit mAb (PE Conjugate) #59370	Cell Signaling Technology	59370S
PE Mouse anti-Bcl-6	BD Biosciences	561522
Alexa Fluor® 488 Mouse anti-Bcl-6	BD Biosciences	561524
IFN gamma Monoclonal Antibody (XMG1.2), PE-Cyanine7	ThermoFisher Scientific	25-7311-82
IL-22 Monoclonal Antibody (1H8PWSR), PE, eBioscience	ThermoFisher Scientific	12-7221-82
IL-17A Monoclonal Antibody (eBio17B7), PE-Cyanine7	ThermoFisher Scientific	25-7177-82
Anti-puromycin -AF488	J. Yewdell	Generated in house
CD45R (B220) Monoclonal Antibody (RA3-6B2), APC	ThermoFisher Scientific	17-0452-82
Alexa Fluor® 594 Streptavidin	Biolegend	405240
Recombinant Anti-Hexokinase 1 antibody [EPR10134(B)]	abcam	ab150423
Recombinant Anti-Hexokinase II antibody [EPR20839] (ab209847)	abcam	ab209847
Bacterial and virus strains
Citrobacter rodentium (strain ICC169)	Y.Belkaid	Bouladoux, Harrison, and Belkaid 201781
Chemicals, peptides, and recombinant proteins
Mouse serum	Millipore Sigma	M5905-5ML
Rat serum	Millipore Sigma	R9759-5ML
normal Armenian hamster IgG	Santa Cruz	sc-3886
16% Paraformaldehyde, methanol-free	Fisher scientific	AA433689M
Liberase™ TL Research Grade	Millipore Sigma	05 401 020 001
DNase 1	Millipore Sigma	10104159001
Percoll®	Millipore Sigma	17-0891-01
HBSS calcium Magnesium	ThermoFisher scientific	24020-117
DTT (dithiothreitol)	ThermoFisher scientific	R0861
EDTA (0.5 M), pH 8.0	ThermoFisher scientific	AM9260G
DAPI	ThermoFisher scientific	62248
Digitonin (5%)	ThermoFisher scientific	BN2006
polybrene	R&D SYSTEMS	7711/10
RetroNectin® Recombinant Human Fibronectin Fragment	Takara	T100B
Recombinant Murine IL-15	peprotech	210-15
Recombinant Murine IL-7	peprotech	217-17
Recombinant Murine SCF size B	peprotech	250-03
Recombinant Murine IL-6	peprotech	216-16
Recombinant Murine IL-1β	peprotech	211-11B
Recombinant Murine Flt3-Ligand, size B	peprotech	250-31L
Recombinant Murine IL-3 size B	peprotech	213-13
Sucrose	Millipore Sigma	S0389
Tissue-Plus™ O.C.T Compound	ThermoFisher scientific	23-730-571
Prolong Glass antifade mountant	Fisher scientific	P36980
PMA	Millipore Sigma	P1585
Ionomycin	MilliporeSigma	I9657
Brefeldin A	ThermoFisher scientific	B7651
Tamoxifen	Millipore Sigma	T5648
Puromycin	Millipore Sigma	540222
Oligomycin	Millipore Sigma	04876
FCCP	Millipore Sigma	C2920
Rotenone	Millipore Sigma	R8875
Antimycin A	Millipore Sigma	A8674
D-(+)-Glucose	Millipore Sigma	G8270
sodium acetate cell culture grade	Millipore Sigma	S5636
Sodium pyruvate	ThermoFisher scientific	11360-070
2-Deoxy-D-glucose	Millipore Sigma	D6134
Critical commercial assays
Foxp3 / Transcription Factor Staining Buffer Set	ThermoFisher scientific	00-5523-00
BD Cytofix/Cytoperm™ Fixation/Permeablization Kit	BD Biosciences	554714
FITC Active Caspase-3 Apoptosis Kit	BD Biosciences	550480
LIVE/DEAD™ Fixable Blue Dead Cell Stain Kit, for UV excitation	ThermoFisher scientific	L23105
APC BrdU Flow Kit	BD Biosciences	552598
L-Lactate Assay Kit (Colorimetric)	abcam	ab65331
SuperScript™ VILO™ cDNA Synthesis Kit	ThermoFisher scientific	11754050
Deposited data
scRNA-seq and scATAC-seq data	This study	PRJNA870085
Experimental models: Cell lines
MNK-3 cells	N/A	Allan et al. 201558
OP9		Weber et al. 201174
HEK293T	N/A	Ruiz et al. 201175
Experimental models: Organisms/strains
Tox2−/−	Our lab	Seo et al. 201931
Tox2flox/flox	This study	N/A
B6.129-Hif1atm3Rsjo/J	Jackson Laboratory	Strain# 007561
B6.129X1-Gt(ROSA)26Sortm1(EYFP)Cos/J	Jackson Laboratory	Strain# 006148
Tox−/−	N/A	Aliahmad and Kaye 200870
ERT2Cre	Taconic Biosciences	Strain# 10471
IL7rCre	N/A	Schlenner et al. 201047
Tcf7EGF	N/A	Yang et al. 201571
Il17a−/−	N/A	Zhang et al. 201972
Rorc Cre	N/A	Eberl et al. 200473
Tox2−/−.Tcf7GFP	This study	N/A
NOD-scid IL2Rgammanull (NSG)	National Cancer Institute	N/A
B6 CD45.1/CD45.2	N/A	Generated in house
Oligonucleotides
Tox2-Floxxed-Left Forward	This paper	5’-GAATGGACATATATGTGGGTC-3’
Tox2-Floxxed-Left Reverse	This paper	5’-CCAGCCATAACTTCGTATAATG-3’
Tox2-Floxxed-Full Forward	This paper	5’-GAATGGACATATATGTGGGTC-3’
Tox2-Floxxed-Full reverse	This paper	5’-CTCAATTAATGATCAAGGGTGG-3’
Recombinant DNA
MSCV-IRES-GFP (MIGR1) plamid	N/A	Weber et al. 201174
Tox2 cDNA	Origene	MR212216
Hk2 cDNA	Origene	MC202280
FUGW plasmid	Addgene	14883
Software and algorithms
Flowjo 10.0	N/A	https://www.flowjo.com/
Prism 10	GraphPad	https://www.graphpad.com/scientific-software/prism/
NSI-Element software	Nikon	https://www.microscope.healthcare.nikon.com/products/software/nis-elements
Seurat v4.1.0	Rahul Satija Lab (NYU)	Hao et al. 202182
R v4.0.5	N/A	https://www.r-project.org/
Signac 1.10.0	Cell ranger-atac-2.1.0	Stuart et al. 202183

HIGHLIGHTS.

• Tox2 is required for gut ILC3 persistence and response to C. rodentium

• ILC3 in the gut are metabolically distinct from ILC3 in central lymphoid sites

• Tox2 regulates Hexokinase-2 (Hk2) expression and glycolysis in gut ILC3

• Hypoxia and IL-17A each induce Tox2 expression in gut ILC3