Main

ILC3s have been shown to be part of discrete mucosal neuroimmune cell units2,3,4,5, raising the hypothesis that ILC3s may also integrate systemic neuroimmune circuits to regulate tissue integrity and organismic homeostasis. Circadian rhythms rely on local and systemic cues to coordinate mammalian physiology and are genetically encoded by molecular clocks that allow organisms to anticipate and adapt to extrinsic environmental changes6,7. The circadian clock machinery consists of an autoregulatory network of feedback loops primarily driven by the activators ARNTL and CLOCK and the repressors PER1–PER3, CRY1 and CRY2, amongst others6,7.

Analysis of subsets of intestinal ILCs and their bone marrow progenitors revealed that mature ILC3s express high levels of circadian clock genes (Fig. 1a–c, Extended Data Fig. 1a–d). Notably, ILC3s displayed a circadian pattern of Per1Venus expression (Fig. 1b) and transcriptional analysis of ILC3 revealed circadian expression of master clock regulators and ILC3-related transcription factors (Fig. 1c). To test whether ILC3s are regulated in a circadian manner, we investigated whether intestinal ILC3s require intrinsic clock signals. Thus, we interfered with the expression of the master circadian activator Arntl. Arntlfl mice were bred to Vav1Cre mice, allowing conditional deletion of Arntl in all haematopoietic cells (ArntlΔVav1 mice). Although ArntlΔVav1 mice displayed normal numbers of intestinal natural killer (NK) cells and enteric group 1 and 2 ILCs, gut ILC3s were severely and selectively reduced in these mice when compared to their wild-type littermate controls (Fig. 1d, e, Extended Data Fig. 2a, b). To more precisely define ILC3-intrinsic effects, we generated mixed bone marrow chimaeras by transferring Arntl-competent (Arntlfl) or Arntl-deficient (ArntlΔVav1) bone marrow against a third-party wild-type competitor into alymphoid hosts (Fig. 1f). Analysis of such chimaeras confirmed cell-autonomous circadian regulation of ILC3s, while their innate and adaptive counterparts were unperturbed (Fig. 1g, Extended Data Fig. 2c).

Fig. 1: Intestinal ILC3s are controlled in a circadian manner.
figure 1

a, Gene expression in CLPs, ILCPs and intestinal ILC3s. CLP and ILCP n = 4; ILC3 n = 6. b, PER1–VENUS mean fluorescence intensity (MFI). CLP and ILCP n = 6; ILC3 n = 4. c, Circadian gene expression in enteric ILC3s; n = 5. d, Intestinal ILC subsets in Arntlfl and ArntlΔVav1 mice; n = 4. e, Cell numbers of intestinal ILC3s and IL-17- and IL-22-producing ILC3 subsets in Arntlfl and ArntlΔVav1 mice; n = 4. f, Generation of mixed bone marrow chimaeras. g, Percentage of donor cells and cell numbers of ILC3s, IL-17 and IL-22-producing ILC3 subsets in the gut from mixed bone marrow chimaeras. Arntlfl n = 5, ArntlΔVav1 n = 7. b, c, White and grey represent light and dark periods, respectively. Data are representative of three independent experiments. n represents biologically independent samples (a, c) or animals (b, d–g). Data shown as mean ± s.e.m. a, Two-way ANOVA and Tukey’s test; b, c, cosinor analysis; d, e, g, Two-tailed Mann–Whitney U test. *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant.

Source Data.

Full size image

To investigate the functional effect of ILC3-intrinsic circadian signals, we deleted Arntl in RORγt-expressing cells by breeding RorgtCre mice (also known as RorcCre) to Arntlfl mice (ArntlΔRorgt mice). When compared to their wild-type littermate controls, ArntlΔRorgt mice showed a selective reduction of ILC3 subsets and IL-17- and IL-22-producing ILC3s (Fig. 2a, b, Extended Data Fig. 3a–j). Notably, independent deletion of Nr1d1 also perturbed subsets of enteric ILC3s, further supporting a role of the clock machinery in ILC3s (Extended Data Fig. 4a–e). ILC3s have been shown to regulate the expression of genes related to epithelial reactivity and microbial composition1. Analysis of Arntlfl and ArntlΔRorgt mice revealed a profound reduction in the expression of reactivity genes in the ArntlΔRorgt intestinal epithelium; notably, Reg3b, Reg3g, Muc3 and Muc13 were consistently reduced in Arntl-deficient mice (Fig. 2c). Furthermore, ArntlΔRorgt mice displayed altered diurnal patterns of Proteobacteria and Bacteroidetes (Fig. 2d, Extended Data Fig. 3j). To investigate whether disruption of ILC3-intrinsic ARNTL affected enteric defence, we tested how ArntlΔRorgt mice responded to intestinal infection. To this end, we bred ArntlΔRorgt mice to Rag1−/− mice to exclude putative T cell effects (Extended Data Fig. 3g–i). Rag1−/−ArntlΔRorgt mice were infected with the attaching and effacing bacteria Citrobacter rodentium2. When compared to their wild-type littermate controls, Rag1−/−ArntlΔRorgt mice had marked gut inflammation, fewer IL-22-producing ILC3s, increased C. rodentium infection and bacterial translocation, reduced expression of epithelial reactivity genes, increased weight loss and reduced survival (Fig. 2e–j, Extended Data Fig. 5a–j). These results indicate that cell-intrinsic circadian signals selectively control intestinal ILC3s and shape gut epithelial reactivity, microbial communities and enteric defence. Previous studies indicated that ILC3s regulate host lipid metabolism8. When compared to their wild-type littermate controls, the epithelium of ArntlΔRorgt mice revealed a marked increase in mRNA that codes for key lipid epithelial transporters, including Fabp1, Fabp2, Scd1, Cd36 and Apoe (Fig. 2k). Accordingly, these changes were associated with increased gonadal and subcutaneous accumulation of fat in ArntlΔRorgt mice when compared to their wild-type littermate controls (Fig. 2l, Extended Data Fig. 5k–n). Thus, ILC3-intrinsic circadian signals shape epithelial lipid transport and body fat composition.

Fig. 2: ILC3-intrinsic Arntl regulates gut homeostasis and defence.
figure 2

a, Enteric ILC3s and subtypes in Arntlfl and ArntlΔRorgt mice; n = 4. b, Gut T helper cells in Arntlfl and ArntlΔRorgt mice; n = 5. c, Expression of epithelial reactivity genes in ArntlΔRorgt mice compared with Arntlfl mice; n = 5. d, qPCR analysis of Proteobacteria in stools from Arntlfl and ArntlΔRorgt mice (see Methods). Arntlfl n = 5; ArntlΔRorgt n = 6. e–j, Data from C. rodentium-infected Rag1−/−Arntlfl and Rag1−/−ArntlΔRorgt mice. e, Histopathology of colon sections; n = 5. f, Colitis score; n = 5. g, Colon length; n = 5. h, Infection burden; Rag1−/−Arntlfl n = 6, Rag1−/−ArntlΔRorgt n = 7. i, Bacterial translocation to the spleen; Rag1−/−Arntlfl n = 6, Rag1−/−ArntlΔRorgt n = 7. j, Survival; n = 5. k, Expression of epithelial lipid transporter genes in Arntlfl (n = 4) and ArntlΔRorgt (n = 5) mice. l, Gonadal and subcutaneous adipose tissue in Arntlfl and ArntlΔRorgt mice; n = 5. d, White and grey represent light and dark periods, respectively. Scale bars, 250 μm. Data are representative of at least three independent experiments; n represents biologically independent animals. Data shown as mean ± s.e.m. a, b, f, g, i, k, Two-tailed Mann–Whitney U test; d, cosinor analysis; h, two-way ANOVA and Sidak’s test; j, log-rank test; l, two-tailed unpaired Student’s t-test. *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant.

Source Data.

Full size image

To further investigate how cell-intrinsic Arntl controls intestinal ILC3 homeostasis, initially we studied the diurnal oscillations of the ILC3 clock machinery. When compared to their wild-type littermate controls, ArntlΔRorgt ILC3s displayed a disrupted diurnal pattern of activator and repressor circadian genes (Fig. 3a). Sequentially, we used genome-wide transcriptional profiling of Arntl-sufficient and -deficient ILC3s to interrogate the effect of a deregulated circadian machinery. Diurnal analysis of the genetic signature associated with ILC3 identity1 demonstrated that the vast majority of those genes were unperturbed in Arntl-deficient ILC3s, suggesting that ARNTL is dispensable to ILC3 lineage commitment (Fig. 3b, Extended Data Fig. 6a–c). To test this hypothesis, we first studied the effect of ablation of Arntl in ILC3 progenitors. ArntlΔVav1 mice had unperturbed numbers of common lymphoid progenitors (CLPs) and innate lymphoid cell progenitors (ILCPs; Fig. 3c, Extended Data Fig. 6d). Sequentially, we analysed the effects of Arntl ablation in ILC3s in other organs. Compared to their littermate controls, ArntlΔRorgt mice had normal numbers of ILC3s in the spleen, lungs and blood, in contrast to their pronounced reduction in the intestine (Figs. 2a, 3d, e, Extended Data Fig. 6e). Notably, enteric ArntlΔRorgt ILC3s showed unperturbed proliferation and apoptosis-related genetic signatures (Extended Data Fig. 6b, c), suggesting that ArntlΔRorgt ILC3s may show altered migration to the intestinal mucosa9. When compared to their wild-type littermate controls, ILC3s in ArntlΔRorgt mice showed a marked reduction in gut postcode molecules—which are essential receptors for intestinal lamina propria homing—and accumulated in mesenteric lymph nodes9 (Extended Data Fig. 6f). Notably, the expression of the integrin and chemokine receptors CCR9, α4β7 and CXCR4 was selectively and hierarchically reduced in ArntlΔRorgt ILC3s (Fig. 3f–h, Extended Data Fig. 6g–m). To investigate whether ARNTL could directly regulate expression of Ccr9, we performed chromatin immunoprecipitation (ChIP). Binding of ARNTL to the Ccr9 locus in ILC3s followed a diurnal pattern, with increased binding at Zeitgeber time (ZT) 5 (Fig. 3i). Thus, ARNTL can contribute directly to the expression of Ccr9 in ILC3s, although additional factors may also regulate this gene. In conclusion, while a fully operational ILC3-intrinsic circadian machinery is not required for lineage commitment and development of ILC3s, cell-intrinsic clock signals are required for a functional ILC3 gut receptor postcode.

Fig. 3: ILC3-intrinsic circadian signals regulate an enteric receptor postcode.
figure 3

a, Relative expression of circadian genes in enteric ILC3s from Arntlfl and ArntlΔRorgt mice; n = 3. b, RNA sequencing (RNA-seq) analysis of gut ILC3s from Arntlfl and ArntlΔRorgt mice at ZT5 and ZT23; n = 3. c, Numbers of CLPs and ILCPs in Arntlfl and ArntlΔVav1 mice; n = 4. d, e, ILC3s in spleen (d, n = 3) and lung (e, n = 6) of Arntlfl and ArntlΔRorgt mice. f, g, Expression of α4β7 (f) and CCR9 (g) by gut ILC3s in Arntlfl and ArntlΔRorgt mice; n = 4. h, Circadian variation in expression of CCR9 by intestinal ILC3s in Arntlfl and ArntlΔRorgt mice; n = 4. i, ChIP analysis of binding of ARNTL to the Ccr9 locus in enteric ILC3s; n = 3. A–J denote putative ARNTL DNA-binding sites. Data are representative of three independent experiments. n represents biologically independent animals (a, c–h) or samples (b, i). a, h, White and grey represent light and dark periods, respectively. Data shown as mean ± s.e.m. a, Two-way ANOVA; c–g, two-tailed Mann–Whitney U test; h, cosinor analysis; i, two-tailed unpaired Student’s t-test. *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant.

Source Data.

Full size image

Circadian rhythms allow organisms to adapt to extrinsic environmental changes. Microbial cues can alter the rhythms of intestinal cells10,11, and feeding regimens are major circadian entraining cues for peripheral organs, such as the liver12. In order to define the environmental cues that entrain circadian oscillations of ILC3, we initially investigated whether microbial signals affect the oscillations of ILC3s. Treatment of Per1Venus reporter mice with antibiotics did not alter the amplitude of circadian oscillations, but did induce a minute shift in the acrophase (timing of the peak of the cycle; Fig. 4a). We then tested whether feeding regimens, which are major entraining cues of oscillations in the liver, pancreas, kidney, and heart12, could alter ILC3 rhythms. To this end, we restricted food access to a 12-h interval and compared Per1Venus oscillations to those observed in mice with inverted feeding regimens12. Inverted feeding had a small effect on the amplitude of ILC3 oscillations but did not invert the acrophase of ILC3s (Fig. 4b, Extended Data Fig. 7a), in contrast to the full inversion of the acrophase of hepatocytes12 (Extended Data Fig. 7b). As these local intestinal cues could not invert the acrophase of ILC3s, we hypothesize that light–dark cycles are major regulators of enteric ILC3 oscillations6. To test this hypothesis, we placed Per1Venus mice in light-tight cabinets on two opposing 12-h light–dark cycles. Inversion of light–dark cycles had a profound effect on the circadian oscillations of ILC3s (Fig. 4c). Notably, and in contrast to microbiota and feeding regimens, light cycles fully inverted the acrophase of Per1Venus oscillations in ILC3s (Fig. 4c, Extended Data Fig. 7c). Furthermore, light–dark cycles entrained ILC3 oscillations, as revealed by their maintenance upon removal of light (constant darkness; Fig. 4d, Extended Data Fig. 7d), confirming that light is a major environmental entraining signal for ILC3 intrinsic oscillations. Together, these data indicate that ILC3s integrate systemic and local cues hierarchically; while microbiota and feeding regimens locally adjust the ILC3 clock, light–dark cycles are major entraining cues of ILC3s, fully setting and entraining their intrinsic oscillatory clock.

Fig. 4: Light-entrained and brain-tuned cues shape intestinal ILC3s.
figure 4

a–d, PER1–VENUS MFI in gut ILC3s from mice treated with or without antibiotics (Ab) (a; n = 3); with restricted or inverted feeding (b; n = 3); with opposing light–dark cycles (c; n = 3); and with opposing light–dark cycles followed by constant darkness (d; n = 3). e, Magnetic resonance imaging of sham- and SCN-ablated (xSCN) mice; n = 11. White arrows indicate location of lesion. f, PER1–VENUS MFI in enteric ILC3s from sham- or SCN-ablated mice; n = 3. g, Expression of circadian genes in enteric ILC3s from Arntlfl and ArntlΔCamk2a mice; n = 3. h, CCR9 expression in gut ILC3s from Arntlfl and ArntlΔCamk2a mice; n = 3. i, Expression of epithelial reactivity genes in the small intestine from Arntlfl and ArntlΔCamk2a mice; n = 3. j, qPCR analysis of Proteobacteria in stools from Arntlfl and ArntlΔCamk2a mice; n = 4. k, Expression of lipid transporter genes in the epithelium of the small intestine in Arntlfl and ArntlΔCamk2a mice; n = 3. l, Gonadal and subcutaneous adipose tissue in Arntlfl mice (n = 5) and ArntlΔCamk2a mice (n = 4). a, b, White and grey represent light and dark periods, respectively. Data shown as mean ± s.e.m. n represents biologically independent animals. a–d, f–k, Cosinor analysis; f–k, cosine fitted curves; amplitude (Amp) and acrophase (Acro) were extracted from the cosinor model. l, Two-tailed unpaired Student’s t-test. *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant.

Source Data.

Full size image

The suprachiasmatic nuclei (SCN) in the hypothalamus are main integrators of light signals6, suggesting that brain cues may regulate ILC3s. To assess the influence of the master circadian pacemaker on ILC3s, while excluding confounding light-induced, SCN-independent effects13,14, we performed SCN ablation by electrolytic lesion in Per1Venus mice using stereotaxic brain surgery15. Strikingly, whereas sham-operated mice displayed circadian Per1Venus oscillations in ILC3s, ILC3s in SCN-ablated mice lost the circadian rhythmicity of Per1Venus and other circadian genes (Fig. 4e, f, Extended Data Fig. 8a–d). Because electrolytic lesions of the SCN may cause scission of afferent and efferent fibres in the SCN, we further confirmed that brain SCN-derived cues control ILC3s by genetic ablation of Arntl in the SCN14. Arntlfl mice were bred to Camk2aCre mice to allow forebrain- and SCN-specific deletion of Arntl (ArntlΔCamk2a)14. When compared to their control counterparts, ILC3s from ArntlΔCamk2a mice showed severe arrhythmicity of circadian regulatory genes and of the enteric postcode molecule CCR9 (Fig. 4g, h, Extended Data Fig. 9a–f). In addition, ArntlΔCamk2a mice showed alterations in epithelial reactivity genes and microbial communities, particularly Proteobacteria and Bacteroidetes (Fig. 4i, j, Extended Data Fig. 9g–i). Finally, the intestinal epithelium of ArntlΔCamk2a mice showed disrupted circadian expression of lipid epithelial transporters, and these changes were associated with increased gonadal and subcutaneous fat accumulation (Fig. 4k, l). Together, these data indicate that light-entrained and brain-tuned circuits regulate enteric ILC3s, controlling microbial communities, lipid metabolism and body composition.

Deciphering the mechanisms by which neuroimmune circuits operate to integrate extrinsic and systemic signals is essential for understanding tissue and organ homeostasis. We found that light cues are major extrinsic entraining cues of ILC3 circadian rhythms, and surgically or genetically induced deregulation of brain rhythmicity resulted in altered ILC3 regulation. In turn, the ILC3-intrinsic circadian machinery controlled the gut receptor postcode of ILC3s, shaping enteric ILC3s and host homeostasis.

Our data reveal that ILC3s display diurnal oscillations that are genetically encoded, cell-autonomous and entrained by light cues. While microbiota and feeding regimens could locally induce small adjustments to ILC3 oscillations, light–dark cycles were major entraining cues of the ILC3 circadian clock. Whether the effects of photonic signals on ILC3s are immediate or rely on other peripheral clocks remains to be elucidated16,17. Nevertheless, cell-intrinsic ablation of important endocrine and peripheral neural signals in ILC3s did not affect gut ILC3 numbers (Extended Data Fig. 10a-i). Our work indicates that ILC3s integrate local and systemic entraining cues in a distinct hierarchical manner, establishing an organismal circuitry that is an essential link between the extrinsic environment, enteric ILC3s, gut defence, lipid metabolism and host homeostasis (Extended Data Fig. 10j).

Previous studies demonstrated that ILCs integrate tissue microenvironmental signals, including cytokines, micronutrients and neuroregulators3,4,18,19. Here we show that ILC3s have a cell-intrinsic circadian clock that integrates extrinsic light-entrained and brain-tuned signals. Coupling light cues to ILC3 circadian regulation may have ensured efficient and integrated multi-system anticipatory responses to environmental changes. Notably, the regulation of ILC3 activity by systemic circadian circuits may have evolved to maximize metabolic homeostasis, gut defence and efficient symbiosis with commensal organisms that have been evolutionary partners of mammals. Finally, our current data may also contribute to a better understanding of how circadian disruptions in humans are associated with metabolic diseases, bowel inflammatory conditions and cancer20.

Methods

Mice

Nod scid gamma (NSG) mice were purchased from Jackson Laboratories. C57BL/6J Ly5.1 mice were purchased from Jackson Laboratories and bred with C57BL/6J mice to obtain C57BL/6 Ly5.1/Ly5.2 (CD45.1/CD45.2). Mouse lines used were: Rag1−/− (ref. 21), Rag2−/−Il2rg−/− (ref. 22,23), Vav1Cre (ref. 24), RorgtCre (ref. 25), Camk2aCre (ref. 26), Il7raCre (ref. 27), Per1Venus (ref. 28), RetGFP (ref. 29), Rosa26RFP (ref. 30), Nr1d1−/− (ref. 31), Arntlfl (ref. 32), Nr3c1fl (ref. 33) and Adrb2fl (ref. 34). All mouse lines were on a full C57BL/6J background. All lines were bred and maintained at Champalimaud Centre for the Unknown (CCU) animal facility under specific pathogen-free conditions. Male and female mice were used at 8–14 weeks old, unless stated otherwise. Sex- and age-matched mice were used for analysis of small intestine epithelium lipid transporters and quantification of white adipose tissue. Mice were maintained in 12-h light–dark cycles, with ad libitum access to food and water, if not specified otherwise. For light inversion experiments mice were housed in ventilated, light-tight cabinets on defined 12-h light–dark cycles (Ternox). Camk2aCreArntlfl (ArntlΔCamk2a) mice and wild-type littermate controls were maintained in constant darkness as previously described14. Mice were systematically compared with co-housed littermate controls unless stated otherwise. Power analysis was performed to estimate the number of experimental mice required. All animal experiments were approved by national and local institutional review boards (IRBs), Direção Geral de Veterinária and CCU ethical committees. Randomization and blinding were not used unless stated otherwise.

Cell isolation

Isolation of small intestine and colonic lamina propria cells was as previously described2. In brief, intestines and colons were thoroughly rinsed with cold PBS1×, Peyer patches were removed from the small intestine, and intestines and colons were cut into 1-cm pieces and shaken for 30 min in PBS containing 2% FBS, 1% HEPES and 5 mM EDTA to remove intraepithelial and epithelial cells. Intestines and colons were then digested with collagenase D (0.5 mg/ml; Roche) and DNase I (20 U/ml; Roche) in complete RPMI for 30 min at 37 °C, under gentle agitation. Cells were passed through a 100-μm cell strainer and purified by centrifugation for 30 min at 2,400 rpm in a 40/80 Percoll (GE Healthcare) gradient. Lungs were finely minced and digested in complete RPMI supplemented with collagenase D (0.1 mg/ml; Roche) and DNase I (20 U/ml; Roche) for 1 h at 37 °C under gentle agitation. Cells were passed through a 100-μm cell strainer and purified by centrifugation for 30 min at 2,400 rpm in a 40/80 Percoll (GE Healthcare) gradient. Spleen and mesenteric lymph node cell suspensions were obtained using 70-μm strainers. Bone marrow cells were collected by either flushing or crushing bones and filtered using 70-μm strainers. Erythrocytes from small intestine, colon, lung, spleen and bone marrow preparations were lysed with RBC lysis buffer (eBioscience). Leukocytes from blood were isolated by treatment with Ficoll (GE Healthcare).

Flow cytometry analysis and cell sorting

For cytokine analysis ex vivo, cells were incubated with PMA (phorbol 12-myristate 13-acetate; 50 ng/ml) and ionomycin (500 ng/ml) (Sigma-Aldrich) in the presence of brefeldin A (eBioscience) for 4 h before intracellular staining. Intracellular staining for cytokines and transcription factors analysis was performed using IC fixation and Staining Buffer Set (eBioscience). Cell sorting was performed using FACSFusion (BD Biosciences). Sorted populations were >95% pure. Flow cytometry analysis was performed on LSRFortessa X-20 (BD Biosciences). Data were analysed using FlowJo 8.8.7 software (Tree Star). Cell populations were gated in live cells, both for sorting and flow cytometry analysis.

Cell populations

Cell populations were defined as: bone marrow (BM) common lymphoid progenitor (CLP): Lin−CD127+Flt3+Sca1intc-Kitint; BM innate lymphoid cell progenitor (ILCP): Lin−CD127+Flt3−CD25−c-Kit+α4β7high; BM ILC2 progenitor (ILC2P): Lin−CD127+Flt3−Sca1+CD25+; small intestine (SI) NK: CD45+Lin−NK1.1+NKp46+CD27+CD49b+CD127− EOMES+ or CD45+Lin−NK1.1+NKp46+CD27+CD49b+CD127−; small intestine ILC1: CD45+Lin−NK1.1+NKp46+CD27+CD49b−CD127+Tbet+ or CD45+Lin−NK1.1+NKp46+CD27+CD49b−CD127+; small intestine ILC2: CD45+Lin−Thy1.2+KLRG1+GATA3+ or CD45+Lin−Thy1.2+KLRG1+Sca-1+CD25+; lamina propria, spleen, mesenteric lymph node and lung ILC3: CD45+Lin−Thy1.2highRORγt+ or CD45+Lin−Thy1.2highKLRG1−; ILC3-IL-17+: CD45+Lin−Thy1.2highRORγt+IL-17+; ILC3-IL-22+: CD45+Lin−Thy1.2highRORγt+IL-22+; for ILC3 subsets additional markers were used: ILC3-NCR−CD4−: NKp46−CD4−; ILC3-LTi CD4+: NKp46−CD4+; ILC3-CCR6−NCR−: CCR6−NKp46−; ILC3-LTi-like: CCR6+NKp46−; ILC3-NCR+: NKp46+; SI Th17 cells: CD45+Lin+Thy1.2+CD4+RORγt+; colon Tregs: CD45+CD3+Thy1.2+CD4+CD25+FOXP3+; colon Tregs RORγt+: CD45+CD3+Thy1.2+CD4+CD25+FOXP3+RORγt+. The lineage cocktail for BM, lung, small intestine lamina propria, spleen and mesenteric lymph nodes included CD3ɛ, CD8α, CD19, B220, CD11c, CD11b, Ter119, Gr1, TCRβ, TCRγδ and NK1.1. For NK and ILC1 staining in the small intestine, NK1.1 and CD11b were not added to the lineage cocktail.

Antibody list

Cell suspensions were stained with: anti-CD45 (30-F11); anti-CD45.1 (A20); anti-CD45.2 (104); anti-CD11c (N418); anti-CD11b (Mi/70); anti-CD127 (IL7Rα; A7R34); anti-CD27(LG.7F9); anti-CD8α (53-6.7); anti-CD19 (eBio1D3); anti-CXCR4(L276F12); anti-NK1.1 (PK136); anti-CD3ɛ (eBio500A2); anti-TER119 (TER-119); anti-Gr1 (RB6-8C5); anti-CD4 (RM4-5); anti-CD25 (PC61); anti-CD117 (c-Kit; 2B8); anti-CD90.2 (Thy1.2; 53-2.1); anti-TCRβ (H57-595); anti-TCRγδ (GL3); anti-B220 (RA3-6B2); anti-KLRG1 (2F1/KLRG1); anti-Ly-6A/E (Sca1; D7); anti-CCR9 (CW-1.2); anti-IL-17 (TC11-18H10.1); anti-rat IgG1k isotype control (RTK2071); anti-streptavidin fluorochrome conjugates from Biolegend; anti-α4β7 (DATK32); anti-Flt3 (A2F10); anti-NKp46 (29A1.4); anti-CD49b (DX5); anti-Ki67 (SolA15); anti-rat IgG2ak isotype control (eBR2a); anti-IL-22 (1H8PWSR); anti-rat IgG1k isotype control (eBRG1); anti-EOMES (Dan11mag); anti-Tbet (eBio4B10); anti-FOPX3 (FJK-16s); anti-GATA3 (TWAJ); anti-CD16/CD32 (93); 7AAD viability dye from eBiosciences; anti-CD196 (CCR6; 140706) from BD Biosciences; anti-RORγt (Q31-378) and anti-mouse IgG2ak isotype control (G155-178) from BD Pharmingen. LIVE/DEAD Fixable Aqua Dead Cell Stain Kit was purchased from Invitrogen.

Bone marrow transplantation

Bone marrow CD3− cells were FACS sorted from Arntlfl, Vav1CreArntlfl, Rag1−/−Arntlfl, Rag1−/−RorgtCreArntlfl, Nr1d1+/+, Nr1d1−/− and C57BL/6 Ly5.1/Ly5.2 mice. Sorted cells (2 × 105) from Arntl- or Nr1d1-deficient and -competent wild-type littermate controls were intravenously injected in direct competition with a third-party wild-type competitor (CD45.1/CD45.2), in a 1:1 ratio, into non-lethally irradiated NSG (150cGy) or Rag2−/−Il2rg−/− (500cGy) mice (CD45.1). Recipients were analysed 8 weeks after transplantation.

Quantitative RT–PCR

RNA from sorted cells was extracted using RNeasy micro kit (Qiagen) according to the manufacturer’s protocol. Liver, small intestine (ileum) and colon epithelium was collected for RNA extraction using Trizol (Invitrogen) and zirconia/silica beads (BioSpec) in a bead beater (MIDSCI). RNA concentration was determined using Nanodrop Spectrophotometer (Nanodrop Technologies). For TaqMan assays (Applied Biosystems) RNA was retro-transcribed using a High Capacity RNA-to-cDNA Kit (Applied Biosystems), followed by a pre-amplification PCR using TaqMan PreAmp Master Mix (Applied Biosystems). TaqMan Gene Expression Master Mix (Applied Biosystems) was used in real-time PCR. Real-time PCR analysis was performed using StepOne and QuantStudio 5 Real-Time PCR systems (Applied Biosystems). Hprt, Gapdh and Eef1a1 were used as housekeeping genes. When multiple endogenous controls were used, these were treated as a single population and the reference value calculated by arithmetic mean of their CT values. The mRNA analysis was performed as previously described35. In brief, we used the comparative CT method (2−ΔCT), in which ΔCT(gene of interest) = CT(gene of interest) − CT(housekeeping reference value). When fold change comparison between samples was required, the comparative ΔCT method (2−ΔΔCT) was applied.

TaqMan gene expression assays

TaqMan Gene Expression Assays (Applied Biosystems) were the following: Hprt Mm00446968_m1; Gapdh Mm99999915_g1; Eef1a1 Mm01973893_g1; Arntl Mm00500223_m1; Clock Mm00455950_m1; Nr1d1 Mm00520708_m1; Nr1d2 Mm01310356_g1; Per1 Mm00501813_m1; Per2 00478113_m1; Cry1 Mm00500223_m1; Cry2 Mm01331539_m1; Runx1 Mm01213404_m1; Tox Mm00455231_m1; Rorgt Mm01261022_m1; Ahr Mm00478932_m1; Rora Mm01173766_m1; Ccr9 Mm02528165_s1; Reg3a Mm01181787_m1; Reg3b Mm00440616_g1; Reg3g Mm00441127_m1; Muc1 Mm00449604_m1; Muc2 Mm01276696_m1; Muc3 Mm01207064_m1; Muc13 Mm00495397_m1; S100a8 Mm01276696_m1; S100a9 Mm00656925_m1; Epcam Mm00493214_m1; Apoe Mm01307193_g1; Cd36 Mm01307193_g1; Fabp1 Mm00444340_m1; Fabp2 Mm00433188_m1; and Scd1 Mm00772290_m1.

Quantitative PCR analysis of bacteria in stools at the phylum level

DNA from faecal pellets of female mice was isolated with ZR Fecal DNA MicroPrep (Zymo Research). Quantification of bacteria was determined from standard curves established by qPCR as previously described2. qPCRs were performed with NZY qPCR Green Master Mix (Nzytech) and different primer sets using a QuantStudio 5 Real-Time PCR System (Applied Biosystems) thermocycler. Samples were normalized to 16S rDNA and reported according to the 2−ΔCT method. Primer sequences were: 16S rDNA, F-ACTCCTACGGGAGGCAGCAGT and R-ATTACCGCGGCTGCTGGC; Bacteroidetes, F-GAGAGGAAGGTCCCCCAC and R-CGCTACTTGGCTGGTTCAG; Proteobacteria, F-GGTTCTGAGAGGAGGTCCC and R-GCTGGCTCCCGTAGGAGT; Firmicutes, F-GGAGCATGTGGTTTAATTCGAAGCA and R-AGCTGACGACAACCATGCAC.

C. rodentium infection

Infection with C. rodentium ICC180 (derived from DBS100 strain)36 was performed at ZT6 by gavage inoculation with 109 colony-forming units (CFUs)36,37. Acquisition and quantification of luciferase signal was performed in an IVIS Lumina III System (Perkin Elmer). Throughout infection, weight loss, diarrhoea and bloody stools were monitored daily.

CFU measurement

Bacterial translocation was determined in the spleen, liver, and mesenteric lymph nodes, taking in account total bacteria and luciferase-positive C. rodentium. Organs were removed, weighed and brought into suspension. Bacterial CFUs from organ samples were determined via serial dilutions on Luria broth (LB) agar (Invitrogen) and MacConkey agar (Sigma-Aldrich). Colonies were counted after 2 days of culture at 37 °C. Luciferase-positive C. rodentium was quantified on MacConkey agar plates using an IVIS Lumina III System (Perkin Elmer). CFUs were determined per volume (ml) for each organ.

Antibiotic and dexamethasone treatment

Pregnant females and newborn mice were treated with streptomycin (5 g/l), ampicillin (1 g/l) and colistin (1 g/l) (Sigma-Aldrich) in drinking water with 3% sucrose. Control mice were given 3% sucrose in drinking water as previously described38. Dexamethasone 21-phosphate disodium salt (200 μg) (Sigma) or PBS was injected intraperitoneally at ZT0. After 4, 8, 12 and 23 h (ZT 4, 8, 12 and 23) mice were killed and analysed.

ChIP assay

Enteric ILC3s from adult C57BL/6J mice were isolated by flow cytometry. Cells were fixed, cross-linked and lysed, and chromosomal DNA–protein complexes were sonicated to generate DNA fragments ranging from 200 to 400 base pairs as previously described2. DNA–protein complexes were immunoprecipitated using LowCell# ChIP kit (Diagenode), with 1 μg of antibody against ARNTL (Abcam) and IgG isotype control (Abcam). Immunoprecipitates were uncrosslinked and analysed by qPCR using primer pairs flanking ARNTL putative sites (E-boxes) in the Ccr9 locus (determined by computational analysis using TFBS tools and Jaspar 2018). Results were normalized to input intensity and control IgG. Primer sequences were: A: F-CATTTCATAGCTTAGGCTGGCATGG; R-CTAGCTAACTGGTCTCAAAGTCCTC; B: F-GCCTCCCTTGTACTACCTG AAGC; R-TCCCAACACCAGGCCGAGTA; C: F-AGGGTCAATTTCTT AGGGCGACA; R-GCCAAGTGTTCGGTCCCAC; D: F-TCTGGCTTCT CACCATGACCACT; R-TCTAAGGCGTCACCACTGTTCTC, E: F-TTTGG GGAATCATCTTACAGC AGAG; R-ATTCATCCTGGCCCTTTCCTTCTTA; F: F-GCTCCACCTCATAGTTGTCTGG; R-CCATGAGCACGTGGAGAGAAAG; G: F-GGTCGAATACCGCGTGGGTT; R-CCCGGTAGAGGCTGCAAGAAA; H: F-AGGCAAATCTGGGCCTATCC; R-GGCCCAGTACAGAGGGGTCT; I: F-GGCTCAGGCTAGCAGGTCTC; R-TGTTTGGCCAGCATCCTCCA; J: F-ACTCAGAGGTGCTGTGACTCC; R-AGCTTTAGGACCACAATGGGCA.

Food restriction (inverted feeding)

Per1Venus mice fed during the night received food from 21:00 to 9:00 (control group), whereas mice fed during the day had access to food from 9:00 to 21:00 (inverted group). Food restriction was performed during nine consecutive days as previously described12. For food restriction in constant darkness, Per1Venus mice were housed in constant darkness with ad libitum access to food and water for 2 weeks. Then, access to food was restricted to the subjective day or night, for 12 days, in constant darkness.

Inverted light–dark cycles

To induce changes in light regime, Per1Venus mice were placed in ventilated, light-tight cabinets on a 12-h light–dark cycle (Ternox). After acclimation, light cycles were changed for mice in the inverted group for 3 weeks to completely establish an inverse light cycle, while they remained the same for mice in the control group, as previously described39. For inverted light–dark cycle experiments followed by constant darkness, after establishing an inverse light–dark cycle, mice were transferred into constant darkness for 3 weeks.

SCN lesions

Bilateral ablation of the SCN was performed in 9–12-week-old Per1Venus males by electrolytic lesion using stereotaxic brain surgery, as described previously15. Mice were kept under deep anaesthesia using a mixture of isoflurane and oxygen (1–3% isoflurane at 1 l/min). Surgeries were performed using a stereotaxic device (Kopf). After identification of the bregma, a hole was drilled through which the lesion electrode was inserted into the brain. Electrodes were made by isolating a 0.25-mm stainless steel insect pin with a heat shrink polyester tubing, except for 0.2 mm at the tip. The electrode tip was aimed at the SCN, 0.3 mm anterior to bregma, 0.20 mm lateral to the midline, and 5.8 mm ventral to the surface of the cortex, according to the Paxinos Mouse Brain Atlas, 2001. Bilateral SCN lesions were made by passing a 1-mA current through the electrode for 6 s, in the left and right SCN separately. Sham-lesioned mice underwent the same procedure, but no current was passed through the electrode. After surgery animals were housed individually under constant dark conditions with ad libitum food and water and were allowed to recover for 1 week before behavioural analysis. Successfully SCN-lesioned mice were selected by magnetic resonance imaging (MRI), arrhythmic behaviour and histopathology analysis.

Magnetic resonance imaging

Screening of SCN ablated mice was performed using a Bruker ICON scanner (Bruker, Karlsruhe, Germany). RARE (Rapid Acquisition with Refocused Echoes) sequence was used to acquire coronal, sagittal and axial slices (five slices in each orientation) with the following parameters: RARE factor = 8, TE = 85 ms, TR = 2,500 ms, resolution = 156 × 156 × 500 µm3 (30 averages). For high-quality images, a 9.4-T BioSpec scanner (Bruker, Karlsruhe, Germany) was used. This operates with Paravision 6.0.1 software and is interfaced with an Avance IIIHD console. Anatomical images (16 axial and 13 sagittal slices) were acquired using a RARE sequence with RARE factor = 8, TE = 36 ms, TR = 2,200 ms and resolution of 80 × 80 × 500 µm3 (12 averages).

Behavioural analysis

Sham-operated and SCN-ablated mice were individually housed and after a 24-h acclimation period their movement was recorded for 72 h, in constant darkness, using the automated animal behaviour CleverSys system. Data were auto scored by the CleverSys software. Videos and scoring were visually validated. Circadian rhythmicity was evaluated using the cosinor regression model40,41.

Histopathology analysis

Mice infected with C. rodentium were killed by CO2 narcosis, the gastrointestinal tract was isolated, and the full length of caecum and colon was collected and fixed in 10% neutral buffered formalin. Colon was trimmed in multiple transverse and cross-sections and caecum in one cross-section42, and all were processed for paraffin embedding. Sections (3–4 μm) were stained with haematoxylin and eosin and lesions were scored by a pathologist blinded to experimental groups, according to previously published criteria43,44,45. In brief, lesions were individually scored (0–4 increasing severity) for: mucosal loss; mucosal epithelial hyperplasia; degree of inflammation; extent of the section affected in any manner; and extent of the section affected in the most severe manner, as previously described45. The score was derived by summing the individual lesion and extent scores. Mesenteric (mesocolic) inflammation was noted but not scored. Liver, gonadal and subcutaneous fat from Arntl∆Rorgt mice was collected, fixed in 10% neutral buffered formalin, processed for paraffin embedding, sectioned into 3-μm-thick sections and stained with haematoxylin and eosin. The presence of inflammatory infiltrates was analysed by a pathologist blinded to experimental groups. For the SCN lesions experiment, mice were killed with CO2 narcosis, necropsy was performed and brain was harvested and fixed in 4% PFA. Coronal sections of 50-µm thickness were prepared with a vibratome (Leica VT1000 S), from 0.6 to –1.3 relative to the bregma, collected on Superfrost Plus slides (Menzel-Gläser) and allowed to dry overnight before Nissl staining. Stained slides were hydrated in distilled water for a few seconds and incubated in Cresyl Violet stain solution (Sigma-Aldrich) for 30 min. Slides were dehydrated in graded ethanol and mounted with CV Mount (Leica). Coronal sections were analysed for the presence or absence of an SCN lesion (partial versus total ablation, unilateral versus bilateral) in a Leica DM200 microscope coupled to a Leica MC170HD camera (Leica Microsystems, Wetzlar, Germany).

Microscopy

Adult intestines from RetGFP mice were flushed with cold PBS (Gibco) and opened longitudinally. Mucus and epithelium were removed, and intestines were fixed in 4% PFA (Sigma-Aldrich) at room temperature for 10 min and incubated in blocking/permeabilizing buffer solution (PBS containing 2% BSA, 2% goat serum, 0.6% Triton X-100). Samples were cleared with benzyl alcohol-benzyl benzoate (Sigma-Aldrich) prior to dehydration in methanol18,46. Whole-mount samples were incubated overnight or for 2 days at 4 °C using the following antibodies: anti-tyrosine hydroxylase (TH) (Pel-Freez Biologicals) and anti-GFP (Aves Labs). Alexa Fluor 488 goat anti-chicken and Alexa Fluor 568 goat anti-rabbit (Invitrogen) were used as secondary antibodies overnight at room temperature. For SCN imaging, RFPΔCamk2a and RFPΔRorgt mice were anaesthetized and perfused intracardially with PBS followed by 4% paraformaldehyde (pH 7.4, Sigma-Aldrich). The brains were removed and post-fixed for 24 h in 4% paraformaldehyde and transferred to phosphate buffer. Coronal sections (50 µm) were collected through the entire SCN using a Leica vibratome (VT1000s) into phosphate buffer and processed free-floating. Sections were incubated with neurotrace 500/525 (Invitrogen, N21480) diluted 1/200 and mounted using Mowiol. Samples were acquired on a Zeiss LSM710 confocal microscope using EC Plan-Neofluar 10×/0.30 M27, Plan Apochromat 20×/0.8 M27 and EC Plan-Neofluar 40×/1.30 objectives.

RNA sequencing and data analysis

RNA was extracted and purified from sorted small intestinal lamina propria cells isolated at ZT5 and ZT23. RNA quality was assessed using an Agilent 2100 Bioanalyzer. SMART-SeqII (ultra-low input RNA) libraries were prepared using Nextera XT DNA sample preparation kit (Illumina). Sequencing was performed on an Illumina HiSeq4000 platform, PE100. Global quality of FASTQ files with raw RNA-seq reads was analysed using fastqc (ver 0.11.5) (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Vast-tools47 (version 2.0.0) aligning and read processing software was used for quantification of gene expression in read counts from FASTQ files using VASTD-DB47 transcript annotation for mouse genome assembly mm9. Only the 8,443 genes with read count information in all 12 samples and an average greater than 1.25 reads per sample were considered informative enough for subsequent analyses. Preprocessing of read count data, namely transforming them to log2(counts per million) (logCPM), was performed with voom48, included in the Bioconductor49 package limma50 (version 3.38.3) for the statistical software environment R (version 3.5.1). Linear models and empirical Bayes statistics were used for differential gene expression analysis, using limma. For heat maps, normalized RNA-seq data were plotted using the pheatmap (v1.0.10) R package (http://www.R-project.org/). Heat-map genes were clustered using Euclidean distance as metric.

Statistics

Results are shown as mean ± s.e.m. Statistical analysis was performed using GraphPad Prism software (version 6.01). Comparisons between two samples were performed using Mann–Whitney U test or unpaired Student’s t-test. Two-way ANOVA analysis was used for multiple group comparisons, followed by Tukey’s post hoc test or Sidak’s multiple comparisons test. Circadian rhythmicity was evaluated using the cosinor regression model40,41,51, using the cosinor (v1.1) R package. A single-component cosinor fits one cosine curve by least squares to the data. The circadian period was assumed to be 24 h for all analysis and the significance of the circadian fit was assessed by a zero-amplitude test with 95% confidence. A single-component cosinor yields estimates and defines standard errors with 95% confidence limits for amplitude and acrophase using Taylor’s series expansion51. The latter were compared using two-tailed Student’s t-test where indicated. Results were considered significant at *P < 0.05, **P < 0.01, ***P < 0.001.

Reporting summary

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