CXCR6 Expression Is Important for Retention and Circulation of ILC Precursors

Innate lymphoid cells are present at mucosal sites and represent the first immune barrier against infections, but what contributes to their circulation and homing is still unclear. Using Rag2 −/− Cxcr6 Gfp/+ reporter mice, we assessed the expression and role of CXCR6 in the circulation of ILC precursors and their progeny. We identify CXCR6 expressing ILC precursors in the bone marrow and characterize their significant increase in CXCR6-deficient mice at steady state, indicating their partial retention in the bone marrow after CXCR6 ablation. Circulation was also impaired during embryonic life as fetal liver from CXCR6-deficient embryos displayed decreased numbers of ILC3 precursors. When injected, fetal CXCR6-deficient ILC3 precursors also fail to home and reconstitute ILC compartments in vivo. We show that adult intestinal ILC subsets have heterogeneous expression pattern of CXCR6, integrin α 4 β 7, CD62L, CD69, and CD44, with ILC1 and ILC3 being more likely tissue resident lymphocytes. Intestinal ILC subsets were unchanged in percentages and numbers in both mice. We demonstrate that the ILC frequency is maintained due to a significant increase of ILC peripheral proliferation, as well as an increased proliferation of the in situ ILC precursors to compensate their retention in the bone marrow.


Introduction
The family of innate lymphoid cells (ILCs) concerns cells devoid of rearranged antigen receptors. It comprises conventional EOMES expressing NK (cNK) cells and the helper subsets composed of three groups (ILC1, ILC2, and ILC3) in analogy with the T-helper cell nomenclature [1]. ILC1 designates the group of T-bet expressing cells that are producing the IFN whereas ILC2 includes ROR /GATA3 dependent cells producing type Th2 cytokines. ILC3 includes different subsets that are all ROR t dependent and produce IL-17 and/or IL-22 (for review, see [2]). In the intestine, the ILC3 populations constantly interact with the microbiota, dietary compounds, epithelial factors, and cytokines to preserve the epithelial barrier integrity. IL-22 producing cells from the ILC3 group mainly concerns the subset that expresses the NK cell receptor NKp46 (termed NCR + ILC3). In mice, this intestinal subset has been shown to protect against infection with the pathogen Citrobacter rodentium [3][4][5]. NCR + ILC3 cells are CCR6 − and do not produce IL-17A. They are rare in cryptopatches compared to CD4 + ILC3 (LTi-like) cells that express CCR6 and produce IL-17A. This LTi-like subset is the adult counterpart of the LTi cells present during fetal life and is crucial for the development of lymph nodes (LN) and Peyer's patches. A third subset of double negative (CD4 − NKp46 − ) ILC3 is also present in the lamina propria. It is usually denominated by NCR − ILC3 [6][7][8]. Intestinal ILC2 populations are crucial against parasitic infections [9][10][11] and may also be implicated in the regulation of gut homeostasis by limiting the inflammation [12].
The migration of lymphocytes to specific tissues is driven by the upregulation of adhesion and chemokine receptors. The T lymphocytes are imprinted with specific trafficking programs that are currently well known [13]. However, little is known concerning the circulation of ILC and their progenitors. All ILC groups derive from a bone marrow 2 Mediators of Inflammation (BM) precursor that expresses the ID2 transcription factor and the common chain of the interleukin 2 receptor [2]. Initial steps of ILC development start in the bone marrow. Intestinal lamina propria and spleen could also support the late stages of ILC differentiation, especially for the ILC3 lineage since precursors are unable to upregulate ROR t in the bone marrow [14]. Using mice bearing a targeted insertion of a GFP-reporter into the Cxcr6 locus (Cxcr6 Gfp/+ mice) [15], we have previously shown that a subset of ILC precursors both in the adult bone marrow and fetal liver expresses CXCR6 [14]. CXCR6 has also been described as expressed by most of the ILC intestinal subsets [16]. Homing and trafficking are achieved by a specific combination of adhesion molecules and chemokine receptors, and CXCR6 expression could be critical and correlated with differential homing and/or function of ILC subsets. It has been shown, for example, that hepatic NK cells rely on CXCR6 expression for the persistence of memory NK cells [17]. Hence, we decided to evaluate the role of CXCR6 in the circulation of both ILC progenitors and ILC subsets.
In this study, we assess the effect of CXCR6 deficiency on ILC precursors and ILC subsets from the bone marrow, intestinal lamina propria (LP), and mesenteric lymph nodes (mLN). The effect of CXCR6 loss could be analyzed using the Rag2 −/− Cxcr6 Gfp/Gfp mouse model since the GFP reporter insertion simultaneously inactivates the corresponding Cxcr6 allele [15], and Rag2 deletion ensures a proper ILC analysis devoid of T-cell contamination. We have observed an increased percentage of bone marrow CXCR6 + ILC precursors without any concomitant proliferation of those precursors, nor loss of differentiation potential, which suggests active bone marrow retention of ILC precursors after the CXCR6 loss. Similarly, CXCR6 contributes to circulation of fetal ILC3 precursors in the fetal liver (FL) as CXCR6deficient fetal livers were partially depleted of ILC3 precursors. Using injection and reconstitution analysis, we show that CXCR6 is required for circulation of ILC3 precursors to intestinal LP or liver.
We show that pattern expression of CXCR6 and integrin 4 7 and egress or circulation markers such as CD62L, CD69, and CD44 among ILC are heterogeneous and only intestinal ILC1 and ILC3 subsets were more likely to be tissue resident ILC populations. Finally, we show that CXCR6-deficiency does not alter the homeostatic balance of intestinal ILC subsets. However, those subsets were seeded and enriched in dividing and recently divided cells when CXCR6 deficiency occurs. The bone marrow derived ILC progenitor was similarly highly active and proliferative in CXCR6-deficient intestinal lamina propria, showing that this high proliferative state compensates retention of the bone marrow precursors.

Cell Preparation.
Bone marrows (BM) were flushed out of femurs and tibias; mesenteric lymph nodes (mLN), fetal livers (FL), and fetal spleens (FS) were mechanically dissociated to obtain cell suspensions. Additionally, BM red blood cells were lysed by incubation with ammonium chloridepotassium bicarbonate solution. Cells from BM, mLN, and FL were magnetically depleted of Lin + cells via staining with biotin labeled antibodies to lineage markers, followed by the use of Streptavidin MicroBeads (Miltenyi).
Small intestine was washed from its contents by PBS injection, and Peyer's patches, if present, were removed. After being cut open longitudinally, small intestine was cut in 1 cm fragments. Fragments of small intestine were incubated 30 min at 37 ∘ and 5% CO 2 in RPMI medium (Gibco) plus 20% fetal calf serum (FCS), HEPES buffer (10 M, Sigma-Aldrich), and then vortexed thoroughly for 4 minutes for removal of epithelial cells and intraepithelial lymphocytes. Remaining fragments of small intestine were incubated 30 min at 37 ∘ and 5% CO 2 in RPMI medium plus 20% FCS, HEPES Buffer (10 M), and collagenase type VIII (250 g/mL, Sigma) for the isolation of LP lymphocytes. Cell suspension was centrifuged, resuspended in a 40% solution of Percoll (GE Healthcare), and underlaid with a 75% solution of Percoll. After centrifugation 20 min at 600 g, cells were collected at the 40-75% interface.
All cells were then collected in cold HBSS (Gibco) plus 1% FCS. Cell suspensions were counted on Malassez cell, and dead cells were excluded using Trypan Blue.

Flow Cytometry.
After antibody staining, cells were washed and dead cells were excluded by Propidium Iodide staining (250 ng/mL, Sigma-Aldrich). Cells were stained intracellularly using Foxp3 Permeabilzation/Fixation Kit according to manufacturers notice (eBiosciences). Cells were then incubated for 5 min with DAPI (4,6-diamidino-2phenylindole, 2 g/mL, Life Technologies), and washed prior to cell acquisition. FACSCanto II and LSR Fortessa (BD Biosciences) were used for flow cytometry acquisition, with Diva6 software (BD Biosciences), and analyzed with FlowJo v8 software (TreeStar). For visual purposes, only 8000 events maximum are shown in each FACS dot plot.
Cells were purified with a FACSAria III sorter (BD Biosciences). Cells were recovered in PBS for cell injection, or OPTIMEM (Gibco) plus 10% FCS for cell culture. All antibodies were purchased from BD Biosciences, eBiosicences, or Biolegend.

Cell
Culture. OP9-DL4 stromal cells were plated one day prior to culture experiment in culture medium: OPTIMEM, 10% FCS, -Mercaptoethanol (500 M, Gibco), penicillin (5 U/mL, Gibco), and streptomycin (5 g/mL, Gibco). Bone marrow precursors were FACS sorted in 400 L of culture medium and plated on OP9-DL4 cells at 20 cells per well. Medium was complemented with in-lab produced cytokines (IL-7, c-KitL and Flt3-L). Cells were cultured for 10 days at 37 ∘ C and 5% CO 2 and then harvested and stained for analysis.
To confirm the ILC potential of the two different ILCP subsets (CXCR6 + and CXCR6 − ), we cultured them on OP9-DL4 stromal cell lines with cytokines to promote cell survival and differentiation (IL-7, c-KitL, and Flt3-L) and analyzed their progeny after 10 days of culture. We confirmed that both CXCR6 + and CXCR6 − ILCP cells were able to give rise to ILC1 (ID2 + NKp46 Figure 1(b)). Interestingly, all ILCs that were obtained expressed variable levels of CXCR6. ILC1 progeny cells were distributed between CXCR6 − and CXCR6 hi levels, with a majority of CXCR6 int cells, regardless of their progenitor (CXCR6 + or CXCR6 − ILCP), showing that ILC1 could acquire or loose CXCR6 expression. Amongst remaining non-ILC1 cells, all cells are ID2 + and are enriched in CXCR6 hi cells (up to 77%), but for a smaller fraction that is CXCR6 − . Because in our previous work, we showed that all ROR t + cells (which represent between 30% and 50% in NK1.1 − NKp46 − fraction after culture of ILCP, Figure 1(b)) were comprised of CXCR6 hi cells, and the remaining ID2 + IL-7R + ILC2 cells are both CXCR6 hi and CXCR6 − , recapitulating the phenotype of ex vivo bone marrow ILC2P.
We analyzed then the contribution of CXCR6 to different lymphoid precursor compartments, using Rag2 −/− Cxcr6 +/+ (wt), Cxcr6 Gfp/+ (HZ), or Cxcr6 Gfp/Gfp (KO) mice. As expected, CLP compartment that does not express CXCR6 was not affected by Cxcr6 deletion either in frequency or in numbers. No difference was observed for total numbers of ILCP and ILC2P in CXCR6-deficient bone marrow (Figure 2(a)). Levels of CXCR6-GFP between HZ and KO mice were not different (Figure 2(b)). However, when looking at the enrichment in CXCR6 expressing cells, a significant increase of CXCR6 + precursors is detected in both ILCP and ILC2P subsets (Figure 2(c)). ILCP from both HZ and KO mice were able to give rise to cNK/ILC1 cells (NK1. showing that their differentiation potential was not affected by CXCR6-deficiency ( Figure 2(d)). This increase in CXCR6 + precursors was not due to a particular proliferative behavior of bone marrow precursors in CXCR6-deficient conditions. The lymphoid precursor subsets display similar frequency of both Ki67 + and DAPI + cells in CXCR6-deficient and competent bone marrow (Figures 2(e) and 2(f)).
Overall, our results show that, in CXCR6-deficient mice, a fraction of ILC precursors that still expresses GFP but no functional CXCR6 protein under the control of Cxcr6 promoter is retained in the bone marrow. This confinement to the bone marrow concerns both ILCP and committed ILC2P and indicates that CXCR6 contributes to cell egress of all ILC subsets from the bone marrow.

E15.5 FS Lin
In conclusion, CXCR6 highly contributes to fetal ILC3 precursors and ILC3 circulation, especially towards the intestine and the liver.

Intestinal ILC Compartments Have Heterogeneous Expression of CXCR6 and Egress
Markers. CXCR6 is heterogeneously expressed by ILC subtypes. By analyzing Rag2 −/− Cxcr6 Gfp/+ intestinal LP, we show that ILC1 are all CXCR6 hi and the diverse ILC3 subsets express different levels of CXCR6, with LTi-like cells expressing lower levels than NCR + and NCR − ILC3 subsets. In contrast, cNK are all CXCR6 − , and intestinal ILC2 subsets were separated into a large CXCR6 − and a smaller CXCR6 + fraction (Figure 4(a)).
We further analyzed expression of egress or retention markers such as CD62L, CD69, and CD44, as well as CXCR6 and integrin 4 7 in intestinal LP and also mesenteric lymph nodes (mLN) ILC populations (Figures 4(b) and 4(c)). All ILC subsets expressed CD44 in all organs. Most of the CD62L + mLN cells belong to the ILC1-cNK subset. In the LP, these cells largely become CD62L − as their ILC2 and ILC3 intestinal counterparts. Only intestinal ILC2 subsets were CD69 lo , whereas intestinal ILC1-cNK and all ILC3 subsets were CD69 + , suggesting less retention of ILC2 cells in the tissue [21,22]. In contrast, all mLN ILC subsets were CD69 − . CXCR6-GFP expression pattern in mLN ILC subsets is similar to what could be observed in LP ILC subsets (Figure 4(a)). Finally, integrin 4 7 was only detected in ILC2 subset but not in other ILC subsets in both LP and mLN (Figure 4(c)).
In conclusion, only intestinal ILC1 and ILC3 subsets that express high levels of both CXCR6 and CD69 are more likely to be tissue resident ILC populations.

Normal Intestinal ILC1 and ILC3 Compartments in CXCR6-Deficient Mice Are Compensated by Higher Proliferative Capacities of Respective ILC Compartments and of an In Situ Progenitor Cell.
We confirmed that, in CXCR6deficient mice, there was no impact on the repartition of the different ILC subsets at homeostasis, except for cNK-ILC1 that are significantly (but slightly only) increased in CXCR6deficient LP (Figure 5(a)), in line with previous publications [16]. Overall, the balance of intestinal ILC subsets was not majorly affected, as we observed similar repartitions between wt, HZ, and KO mice. Intestinal ILC are mainly composed of
Similar to HZ mice, we tested CD69, CD62L, CD44, CXCR6-GFP, and integrin 4 7 expression among ILC subsets in LP and mLN from KO mice. We show that the expression pattern of those markers was not significantly perturbed by CXCR6-deficiency ( Figure 5(b)).
Then, we tested whether each of the ILC subsets was similarly distributed between the different stages of proliferation ( Figure 5(c)). We defined that there was an important increase (up to two fold increase) in Ki67 + cells amongst cNK-ILC1 and all ILC3 subsets between HZ or wt control and CXCR6-deficient conditions (Figures 5(c) and 5(d)). In contrast, the Ki67 + frequency of ILC2 cells remained unchanged. Additionally, DAPI staining showed that all ILC subsets were enriched in cells that were in S-G2-M stages (i.e., DAPI + ) in the CXCR6-deficient condition ( Figure 5(e)). These results show that intestinal ILC1 and ILC3 compartments, and to a certain extent ILC2, were seeded with dividing DAPI + and active recently divided Ki67 + cells, after CXCR6 ablation.
We analyzed the putative in situ intestinal ILC precursor fraction that is isolated as Lin − NKp46 − NK1.1 − GATA3 − ROR t − IL-7R + cells ( Figure 5(f)). This fraction, gated using surrogate markers as described to identify ILC subsets in Cxcr6 Gfp/+ Id2 Yfp/+ mice (as in Figure 4(a)), expresses CXCR6 and ID2 ( Figure 5(f)). No difference in the frequency of those cells is observed, but they were significantly enriched in both Ki67 + (active G1-S-G2 cells) and DAPI + (dividing S-G2-M cells) in CXCR6-deficient intestines ( Figure 5(g)). This result determines that this compartment is highly active in the CXCR6-deficient mice and contributes to proliferation, thus differentiation to seed the respective mature ILC compartments.
In conclusion, these results show that in situ intestinal ILC precursor cells compensate a defect in homing to the intestine by homeostatically proliferating and result in a proper balance of diverse ILC compartments.

Discussion
ILCs share numerous characteristics with the T-helper cell subsets and we considered that they might have similarities in their trafficking features even if they do not have to encounter antigen. For naïve T lymphocytes, it was demonstrated that the intestinal homing marker 4 7 is upregulated after their activation in the mLN [23]. For the ILC lineage, this intestinal homing marker is already expressed by their precursors in the bone marrow. Moreover, CXCR6 was also shown to be expressed by the most mature fraction of 4 7 + medular ILC precursors. Hence, CXCR6 may represent one of the chemokine receptors important for the egress and homing of ILC precursors. We decided to study the role of CXCR6 by comparing diverse features of ILC populations between Cxcr6 Gfp/+ and Cxcr6 Gfp/Gfp mice in steady state conditions. We crossed our mouse models to obtain Rag2 −/− Cxcr6 Gfp/+ and Rag2 −/− Cxcr6 Gfp/Gfp mice since most studies of ILC function count on Rag2 −/− mice to avoid an important contamination of T cells.
The determination of CXCR6 expression fractions among the ID2 + ILC precursors was performed using Cxcr6 Gfp/+