CD62L Is a Functional and Phenotypic Marker for Circulating Innate Lymphoid Cell Precursors

Innate lymphoid cells (ILCs) are important regulators of epithelial tissue integrity during homeostasis while they can additionally function as potent immune effector cells during inflammation. Based on their cytokine production and expression of subset-defining transcription factors, ILCs have been subdivided into three main groups (ILC1, 2, and 3), parallel to the subdivision of Th cell subsets (1, 2). ILC1 express the transcription factor T-BET and produce IFN-γ upon stimulation with IL-12 and/or IL-18 (3–5). ILC2 are defined by expression of the transcription factor GATA3, express the IL-33 receptor and, in humans, the chemokine receptor of Th2 cells (CRTH2), and produce type II cytokines (mainly IL-5 and IL-13) (6–15). ILC3 can be discriminated by the expression of the transcription factor retinoic acid orphan receptor (ROR) γt. ILC3 reside in epithelial tissues where they produce their main effector cytokines IL-17A and IL-22 (16–26).

The majority of studies specifically demonstrate the role of ILCs within healthy or diseased peripheral tissues, and a number of reports are available on ILCs in peripheral blood (PB) in patient cohorts (27–29). Recently, it was shown that circulating CD117⁺ ILCs in human PB include a population of ILC precursors (ILCP) that can differentiate into all three ILC subsets. This final differentiation step is proposed to occur within the tissues (30). To date, it is not clear in which tissue this differentiation takes place, and it is evident that homing receptors expressed by PB ILCP will determine their tissue migration. Adaptive lymphocytes are equipped to migrate to secondary lymphoid organs (SLO) when they are still naive; after which, they, upon activation within the SLO, migrate to a defined tissue (31). Interestingly, in mice, ILCs migrate to the intestine, and the skin upon programming in MALT and skin-draining lymphoid tissue, respectively (32, 33). Furthermore, ILCs present in PB of patients suffering from a range of immune-mediated diseases express different homing receptors when compared with ILCs in healthy individuals (3, 27–29, 34), which suggests that the migratory behavior of PB ILCs can be affected by ongoing immune reactions. Additionally, activated (i.e., NKp44⁺) ILC3 appear in the circulation and in the peripheral lymph nodes (LN) of patients with inflammatory disorders, whereas they are absent in homeostasis (27, 29, 35). Thus, to obtain their homing receptor profile to migrate as effector ILCs to a particular tissue, it is likely that ILCP need the LN microenvironment for further differentiation, similar to naive adaptive lymphocytes. Continuing the parallel to adaptive lymphocytes, it could mean that the differentiation step of ILCP to mature, effector ILCs also takes place in SLO.

In this study, we show that in homeostasis PB, ILCP express the naive T cell marker CD62L and that this population contains the majority of ILCs in PB, both in human and mouse. These cells use CD62L to enter peripheral LNs (pLN) in mice, whereas exit from pLN requires S1P. Furthermore, expression of CD62L is absent on murine LN–resident ILCs that express markers known to be associated with the described ILC subsets, implying that CD62L-expressing cells indeed belong to a precursor population, as is the case in humans. Most interestingly, in both Crohn disease (CD) patient blood and palatine tonsils, NKp44⁺ ILCs (activated ILC3) are uniformly negative for CD62L, suggesting that absence of CD62L associates with an activated phenotype. Accordingly, ILCs present in PB of CD patients have a strongly reduced expression of CD62L, further implicating CD62L as a functional marker for recirculating, nonactivated ILCs in both human and mouse and suggesting that CD62L expression on ILCP can potentially serve as a diagnostic marker in the clinic.

Ubq-GFP, Rag2^−/−, or wild type (wt) C57/Bl6 mice (10–16 wk of age, both male and female) were kept under specific pathogen-free conditions at the Academic Animal Research Center of the Vrije Universiteit (VU) University (Amsterdam, the Netherlands) and received water and food ad libitum. Mice were randomly assigned to treatment groups. Sample sizes of the different groups were determined based on published data of comparable experiments (36–39) and by subsequently applying a power analysis. No blinding was used during the experiments. All experiments have been approved by the VU University Scientific and Animal Ethics Committees according to Dutch law.

For blocking lymphocyte homing into pLN, mice were i.v. injected with 250 μg blocking rat anti-mouse CD62L Ab (clone MEL-14) or with an isotype control (clone GL117) at 2-d intervals for the duration of the experiment. LN egress was blocked by daily i.p. injection with 0.3 mg/kg body weight FTY-720p (Echelon Biosciences, Salt Lake City, UT) or a DMSO control.

Upon sacrificing, pLN (axillary, brachial, and inguinal LNs were pooled together), mesenteric LN (mLN) and spleens were dissected. All organs were subsequently mechanically disrupted over a 70-μm nylon cell strainer (Falcon) and washed with PBS (B. Braun Melsungen) containing 2% (v/v) newborn calf serum. In spleen suspensions, erythrocytes were lysed using a buffer containing 0.15 M NH₄Cl, 10 mM KHCO₃, and 0.1 mM EDTA in PBS.

PB of inflammatory bowel disease patients and healthy controls was obtained upon receiving signed informed consent. All blood samples were collected according to the guidelines of the Medical Ethical Committee of the VU University Medical Center (Amsterdam, the Netherlands) in accordance with the Declaration of Helsinki and according to Dutch law. PBMCs were isolated by gradient centrifugation on Lymphoprep (d = 1.077; Fresenius Kabi Norge). The PBMCs were subsequently centrifuged and washed twice in cold PBS supplemented with 10% v/v donor plasma.

Palatine tonsils were obtained from the Department of Otolaryngology, Slotervaart Hospital, Amsterdam, the Netherlands. All tonsils were collected according to the guidelines of the Medical Ethical Committee of the VU University Medical Center (Amsterdam, the Netherlands) in accordance with the Declaration of Helsinki and according to Dutch law. Tonsil tissue was cut into small pieces and subsequently disrupted mechanically. Mononuclear cells were further isolated by filtering over a 70-μm nylon cell strainer (Falcon) and gradient centrifugation on Lymphoprep.

For flow cytometry, murine cells were stained with the following Abs: CD11b (M1/70), CD11c (N418), B220 (RA3-6B2), TCR-γ/δ/PE (eBioGL3), TCR-α/β/PE (H57-597, all PE), CD69/FITC (H1.2F3), NK1.1/FITC (PK136), CD90.2/FITC (53-2.1), Gata3-eF660 (TWAJ), EOMES/PE/eF610 (Dan11mag), CD62L/PE/Cy7 (MEL-14; all from Thermo Fisher Scientific, Waltham, MA), CD3/PE (145-2C11), integrin α₄β₇/APC (DATK32), NKp46/BV711 (29A1.4), CD127/BV421 (A7R34; both from BioLegend, San Diego, CA), and Rorγt/BV786 (Q31-378; BD Bioscience, Mountain View, CA). Anti-CD45 (clone MP33) was affinity-purified from hybridoma cell culture using Protein G Sepharose (Pharmacia, Uppsala, Sweden) and labeled in-house with Alexa Fluor 647 (Thermo Fisher Scientific). Detection of MEL-14 bound to cells isolated upon treatment was done using an Alexa Fluor 488–labeled Goat anti-Rat IgG (H+L) (Thermo Fisher Scientific), followed by staining with a fluorescently labeled anti-CD62L Ab (see above).

Human cells were stained with the following Abs: CD3 (UCHT1), TCR-α/β (IP26), TCR-γ/δ (B1.1), CD11c (3.9), CD19 (HIB19), CD14 (61D3), CD34 (4H11), CD94 (DX22; all FITC labeled), CD127/PE or APC (eBioRDR5), CD7/PerCP-eFluor700 (4H9), CD117/PeCy7 (104D2; all Thermo Fisher Scientific), CRTH2/PE/CF594 (BM16), CD56/BV786 (NCAM16.2), and CD62L/BV421 (DREG56; both BD Bioscience). The following reagent was obtained through the National Institutes of Health AIDS Reagent Program, Division of AIDS, National Institute of Allergy and Infectious Diseases, National Institutes of Health: α4-β7 mAb (Act-1, catalog no. 11718) from Dr. Aftab A. Ansari. The Ab was biotin-labeled in-house using EZ-Link NHS-Biotin (Thermo Fisher Scientific) according to manufacturers’ protocol. Biotin labeling was detected using Streptavidin-eF605NC or streptavidin, and Alexa Fluor 647 (both Thermo Fisher Scientific). Biotin-labeled recombinant human CCL19 used for staining of CCR7 was a kind gift from G. Graham (The Beatson Institute for Cancer Research, Glasgow, U.K.).

Exclusion of dead cells was done using staining with Fixable Viability Dye eFluor 780 (Thermo Fisher Scientific). Fixation and permeabilization of all cells for intracellular staining was performed using the Foxp3/Transcription Factor Fixation/Permeabilization Kit (Thermo Fisher Scientific).

Analysis was performed using either LSRFortessa ×20 (BD Bioscience) or CyAn ADP (Beckman Coulter) flow cytometer. Further analysis was done using FlowJo Software version 10 for Microsoft (Tree Star, San Carlos, CA). Gating was done based on fluorescence minus one controls.

Isolated mononuclear cells were enriched for CD117⁺ cells using the CD117 MicroBead Kit (Miltenyi Biotec, Bergisch Gladbach, Germany). The cells were subsequently stained using the same Abs as listed above, with the exception of CD117/PE/eVio770 (A3C6E2; Miltenyi Biotec). Cells were then sorted using a MoFlo cell sorter (Beckman Coulter) or a FACS Fusion (BD Bioscience), and the sorted cells were collected in DMEM (Life Technologies; Thermo Fisher Scientific) supplemented with 10% v/v FCS and penicillin, streptomycin, and glutamine. Sorted ILCs were cultured in 10 ng/ml recombinant human IL-7 (rhIL-7) (PeproTech, Rocky Hill, NJ), 10 ng/ml recombinant human stem cell factor (SCF; R&D Systems, Minneapolis, MN), and either or not in presence of 10 ng/ml rhIL-1β and 50 ng/ml rhIL-23 (both PeproTech). When cocultured with the mouse stromal cell line OP9, cells were cultured for 7 d with addition of rhIL-7 as above.

For clonal expansion, human PB ILCs were sorted as one cell per well in 96-well plates and cultured on OP9 cells in DMEM supplemented with 10% v/v FCS and penicillin, streptomycin, and glutamine in the cytokine conditions mentioned below. Sorted ILCs were cultured in 10 ng/ml rhIL-7 (PeproTech), 100 U/ml rhIL-2 (Chiron), 20 ng/ml rhIL-1β, and either 50 ng/ml rhIL-23 (PeproTech), 20 ng/ml rhIL-12 (PeproTech), and 20 ng/ml IL-18 (R&D Systems) or 20 ng/ml IL-25 (Miltenyi Biotec) and 20 ng/ml IL-33 for 18 d prior to restimulation with PMA and ionomycin (see below). Medium was refreshed every week.

Bulk-sorted mouse ILCs from PB, spleen, and LNs were cultured for 10 d on OP9 cells in DMEM supplemented with 10% v/v FCS and penicillin, streptomycin, and glutamine in the cytokine conditions mentioned below. Sorted ILCs were cultured in 25 ng/ml recombinant mouse IL-7 and 20 ng/ml recombinant mouse SCF (both PeproTech), 20 ng/ml recombinant mouse IL-12 (PeproTech), and 20 ng/ml IL-18 (R&D Systems) or 20 ng/ml IL-25 (Miltenyi Biotec) and 20 ng/ml IL-33 (PeproTech).

For intracellular cytokine stainings, cells were restimulated using PMA and ionomycin (both Sigma-Aldrich, St. Louis, MO), in the presence of GolgiPlug Protein Transport Inhibitor (BD Bioscience). Intracellular FACS staining for the detection of cytokines was done using the same procedures as listed above. Detection of human IL-22, IL-13, and IFN-γ was done using anti-human IL-22/PE (22URTI; Thermo Fisher Scientific) and IL-13/APC (JES105A2) and IFN-γ/BV510 (4S.B3; both BioLegend). Detection of mouse IL-22, IL-13, IL-17A, and IFN-γ was done using anti-mouse IL-22/PerCP-eF710 (1H8PWSR), IL-13/eF450 (eBIO13A; both Thermo Fisher Scientific) IL-17A/PE (TC11-18H10; BD Bioscience), and IFNγ/FITC (XMG1.2; BioLegend).

Results are given as the mean ± SEM or SD. Statistical analysis was done using GraphPad Prism 4 Software (La Jolla, CA). Because of small sample size, we could not assume normal distribution and/or equal variance and thus used either one- or two-tailed Mann–Whitney U tests for comparisons between two groups or two-way ANOVA with Bonferonni correction for comparisons between multiple groups, as described in the figure legends. Significance is indicated by *p ≤ 0.05, **p < 0.01, or ***p < 0.001. We did not use statistical methods to predetermine sample size of human samples, nor were the investigators blinded to sample identity or results. Samples were not randomized, except for the in vivo experiments in mice, as described above.

Naive T cells migrate through lymphoid organs in search of an activation stimulus. To do so, circulating T cells need to express high levels of CD62L, which allows for migration into pLNs (36). To investigate whether human PB ILCP have the same migratory potential, we investigated expression of CD62L on ex vivo isolated ILCs from PB. PB ILCP were defined as Lin⁻CD127⁺cKit⁺CRTH2⁻CD7⁺CD45RA⁺CD56⁻ cells (30) (for gating, see Fig. 1A) and were confirmed to be genuine ILCs by their intermediate to high expression levels of the transcription factor GATA3, reported to be expressed by all ILCs (40–42) (Fig. 1A). Human PB ILCP almost uniformly expressed CD62L (Fig. 1A), and expression levels of CD62L on PB ILCP were comparable to these on PB T cells (Fig. 1B). CD62L expression levels on PB ILCP were also similar to circulating Lin⁻CD127⁺CRTH2⁺ ILC2, whereas they were slightly higher compared with CD56⁺ NK cells (Fig. 1C). Additional FACS stainings including Abs targeting TCR-αβ and TCR-γδ did not affect our results (Supplemental Fig. 1A), confirming the identity of CD62L⁺ PB ILCP as true ILC. In line with previous reports on PB ILCP (30) and tonsil-derived CD62L⁺ ILC3 (43), and despite being phenotypically similar to tonsil ILC3 (i.e., Lin⁻CD127⁺cKit⁺ CRTH2⁻), PB CD62L⁺ ILCP produced no IL-22 following ex vivo culture in IL-1β and IL-23 for 3 d and subsequent restimulation with PMA and ionomycin (Fig. 1D). Only when CD62L⁺ cells were cultured on OP9 stromal cells (7 d), some production of IL-22 was observed, and the cells appeared to be potent producers of IL-13 while producing little to no IFN-γ (Fig. 1E, data not shown). To determine whether these CD62L⁺ PB ILCP indeed constitute a population of true ILC progenitors, clonal outgrowth assays on OP9 stromal cells were set up. Single cell per well–sorted CD62L⁺ ILCP showed potential to grow out clonally with the ability to produce IL-22, IL-13, and IFN-γ (Fig. 1F). Interestingly, the majority of clones produced IL-13, regardless of the culture condition. Also, we did not see any differentiation bias between CD62L⁺ and CD62L⁻ clones (Fig. 1F) or in bulk cultures (Supplemental Fig. 1B), furthermore confirming that both PB subsets are precursor populations. It should be noted though, that CD62L marks the vast majority of circulating ILCP (∼95%; Fig. 1A).

Although prevalent on PB ILCP, expression of CD62L was nearly absent on phenotypically similar ILC3s derived from tonsils (CD62L is expressed on 5.9 ± 3.6% of Lin⁻CD127⁺cKit⁺ tonsil ILC3s; Fig. 1G). In agreement with CD62L marking ILCP, CD62L⁺ ILCP in tonsils expressed lower levels of RORγt as compared with the CD62L⁻ ILC3 (Lin⁻CD127⁺cKit⁺CRTH2⁻CD62L⁻; Fig. 1H). Interestingly, tonsil-derived CD62L⁻ ILC3 expressed higher levels of RORγt when compared with CD62L⁻ ILCP from PB (Supplemental Fig. 1C), suggesting that loss pf CD62L in tonsils marks mature ILCs, although this is not necessarily true in PB. These data suggest the CD62L⁺ cells in tonsils and PB might belong to the same, circulating subset in line with the reported presence of ILCP within tissues (30).

Together, the data show that, similar to naive T cells during homeostasis, human PB ILCP express homing molecules that allow preferential homing to SLOs.

To address whether the expression of CD62L is functional and required for ILCP migration to LNs, we addressed whether a population of CD62L⁺ ILCs could be found in adult mice. Hereto, we investigated CD62L expression on cells derived from pLN, mLN, spleen, and PB. In these tissues, ILCs (defined as Lin⁻CD127⁺; for gating, see Fig. 2A) were found to express CD62L (Fig. 2B, 2C). Again, addition of Abs against TCR-αβ and TCR-γδ did not influence the results, and T cell contamination of our CD62L⁺ ILC gate was thus excluded (Supplemental Fig. 1D). The percentage of CD62L⁺ ILCs of all ILCs was lower in LNs when compared with spleen and PB (Fig. 2C). Further characterization of CD62L⁺ ILCs learned that CD62L expression was mutually exclusive with CD69, an early T cell activation marker as well as tissue resident memory T marker (44) in all tissues we investigated in line with human ILCP (30, 45) (Fig. 2B). Confirming their identity as ILCs, CD62L⁺ ILCs in murine pLN, mLN, spleen, and PB expressed CD90 as well as GATA3 at intermediate levels (Supplemental Fig. 1E, 1F) (40–42). Furthermore, the expression level of CD62L was comparable to the levels on T cells (Supplemental Fig. 1G), similar to what we observed for human ILCP (Fig. 1C). Intriguingly, CD62L-expressing ILCs almost uniformly lacked subset-defining markers in all the tissues analyzed (ILC1: NKp46⁺ST2⁻Rorγt⁻, ILC2: NKp46⁻ST2⁺Rorγt⁻, and ILC3: NKp46^+/−ST2⁻Rorγt⁺) (Fig. 2D–G). Only a small proportion of CD62L-expressing cells were found to coexpress RORγt in mouse PB (Fig. 2D, Supplemental Fig. 1H). To confirm that CD62L-expressing cells are genuine ILCP, Lin⁻CD90⁺CD127⁺NKp46⁻ST2^-α4β7⁻CD62L⁺ cells were sorted from PB, pLN, and spleen of Rag2^−/− mice and cultured in different cytokine conditions. During sorting, α4β7 was used as a surrogate marker for ILC3, as intracellular staining for Rorγt requires fixation and permeabilization of the cells. After 10 d of culture on OP9 mouse stromal cells and subsequent restimulation in PMA and ionomycin, these cells produced ILC-associated cytokines upon culture (mainly IL-17A, IL-22 and IFN-γ, and, to a lesser extent, IL-13), proving that these cells indeed have the potential to differentiate into cytokine-producing ILCs in culture (Fig. 2H).

Recently, it was shown in parabiotic mice, based on a gating strategy different from ours (46), potentially leaving out this CD62L⁺ subset, that ILCs are tissue-resident, nonmigratory cells. For this reason, we stained cells isolated from pLN, mLN, and spleens and compared the gating strategies (Supplemental Fig. 2A). These analyses revealed that the (lymphoid) tissue resident populations as described by Gasteiger et al. (46) do not include the population of Lin⁻CD127⁺CD90⁺Gata3^intCD62L⁺ we describe in this study in both pLN and mLN. For spleen, only a part of the tissue-resident ILCs expressed CD62L (Supplemental Fig. 2B).

Altogether, our data show that a subset of Lin⁻CD127⁺ ILCs present in PB and lymphoid tissues express CD62L in both mouse and human.

As CD62L⁺ T cells have been shown to recirculate between LNs and blood during steady state, and ILCP express this marker in human (Fig. 1A) and mouse (Fig. 2B), we addressed whether ILCs show similar recirculation patterns. As such, we inhibited pLN entry by in vivo blockade of CD62L using an inhibitory Ab (clone MEL-14). The 20 × 10⁶ total Ubq-GFP splenocytes were injected in combination with either MEL-14 or a control Ab (cntrl Ab) in wt hosts. Twenty-four hours after injection, spleens, pLN, and mLN were analyzed. Adequate blockade of CD62L was confirmed, as MEL-14 injection led to an increase of total cells in the spleens, whereas the opposite was true for pLN (data not shown). In agreement, both the percentage and the absolute count of the GFP⁺ cells were decreased in pLN upon CD62L blockade (Supplemental Fig. 3A). No difference was seen in mLN (data not shown), which also allows α₄β₇-dependent cell entry via MAdCAM-1 expression on high endothelial venules in these LNs (38, 39). CD62L blockade specifically blocked the entry of host-derived CD62L⁺ ILCs in pLNs, whereas no effect of the treatment was observed when the entire population of Lin⁻CD127⁺ ILCs was analyzed (Fig. 3A, 3B). In addition, ILCs were accumulating in the spleen (Fig. 3A). To further address the long-term effect of blocking LN entry of circulating CD62L⁺ ILCs, we treated mice with either MEL-14 or cntrl Ab for the duration of 1 wk. Analysis on day 7 revealed an increase in spleen cellularity and a concomitant decrease in LN cellularity, again confirming successful CD62L blockade (Supplemental Fig. 3B). CD62L blockade resulted in a significant increase in both the relative and absolute size of the ILC population in the spleens, as expected for recirculating cells (36) (Fig. 3C). Unexpectedly, the ILC population in LNs showed a relative increase but did not change in absolute numbers (Fig. 3C). When we focused, however, on CD62L⁺ ILCs, these cells were significantly increased in spleen and PB, whereas they were reduced in pLN, both in relative and in absolute numbers (Fig. 3D), supporting a model of continuous CD62L-dependent migration of CD62L⁺ ILCs between LNs and the bloodstream.

Because the absolute numbers of ILCs in pLN did not change upon treatment with MEL-14, whereas the CD62L⁺ population was diminished (Fig. 3C, 3D), we questioned whether this seeming discrepancy was a result of ILC proliferation. Increased proliferation of intestinal ILCs has been reported in mice lacking adaptive lymphocytes (Rag^−/−), which was lost upon adoptive transfer of CD4⁺ T cells (47). Similar to in Rag^−/− intestines, CD62L blockade results in a strong decrease in T cell cellularity within LNs (36, 48), potentially freeing up essential growth factors for ILCs. We, therefore, analyzed ILCs from spleen, pLN, mLN, and PB for the proliferation marker Ki67 upon treatment with MEL-14. MEL-14 treatment resulted in an increased percentage of proliferating ILCs in the spleen, mLN, and pLN, whereas proliferation was most pronounced in the pLN (Fig. 4A). Although a slight increase in Ki67 staining was apparent in CD62L⁺ ILCs in pLN upon treatment with MEL-14, tissue-resident CD62L⁻ population showed the most prominent expression of this proliferation marker (Fig. 4B, 4C). Moreover, upon MEL-14 treatment, Ki67⁻ ILCs accumulated in the spleen, whereas this was not seen in pLN (Fig. 4D). These observations further support the notion that ILC accumulation in the spleen of MEL-14–treated mice is due to altered migratory behavior of CD62L⁺ ILCs, whereas in pLN, local proliferation of tissue resident ILCs compensates for the loss of the CD62L⁺ population.

Together, our data show that CD62L⁺ ILCs migrate to LNs in a CD62L-dependent fashion and accumulate in the spleen once CD62L is blocked.

To further address the molecular requirements for CD62L⁺ ILCs migration through SLOs, LN egress was blocked for the duration of 1 wk by treating mice with FTY-720. This causes the internalization of S1P receptors, among which is S1PR1, which is essential for lymphocyte egress from LN (49). In human SLOs, S1PR1 RNA is expressed by human cKit⁺CRTH2⁻NKp44⁻ ILCs, which include CD62L⁺ ILCP (45). Similar to mouse SLO CD62L⁺ ILCs (Fig. 2B), this population lacks expression of CD69, allowing the expression of S1PR1, which is mutually exclusive with CD69 expression (50). In accordance with arrest of LN exit, spleens of FTY-720–treated mice contained fewer cells than those of DMSO-treated (control) animals, whereas the opposite was seen in pLN and mLN (Supplemental Fig. 3C). When focusing on total Lin⁻CD127⁺ ILCs, an absolute increase was observed in pLN upon treatment with FTY-720 (Fig. 5A). More specifically, this increase was caused by a specific increase in CD62L⁺ ILCs and not CD62L⁻ ILCs (Fig. 5B, 5C), implying that the exit of CD62L⁺ ILCs, which had entered during the time of FTY-720 treatment, was blocked, whereas tissue-resident CD62L⁻ ILCs were not affected.

Taken together, our data confirm that CD62L⁺ ILCs are able to recirculate via LNs, similar to naive T cells.

As CD62L⁺ ILCP have the capacity to recirculate between SLO and PB, they could become activated when they encounter an immune-activated LN, similar to naive adaptive lymphocytes. To address this, we analyzed ILCs in PB of CD patients undergoing a control colonoscopy (Table I) for expression of NKp44, reported as activation marker for ILC3 (51) in combination with the expression of CD62L. Five of the seven patients were suffering from a relapse when blood was drawn for analysis (active disease in Table I). Lin⁻CD127⁺cKit⁺CRTH2⁻ ILCs present in PB of two out of seven patients suffering from CD showed an increased presence of NKp44⁺ ILCs (Fig. 6A). Remarkably, all seven patients had a relatively reduced population of CD62L⁺ ILCP in PB as compared with the mean of the healthy donors (Fig. 6B). This was caused by an increase in total numbers of cKit⁺CRTH2⁻ CD62L⁻ ILCs, as the total numbers of CD62L⁺ ILCP did not change (Fig. 6C). No changes were observed in the composition of other PB ILC population (i.e., relative frequencies Lin⁻CD127⁺, Lin⁻CD127⁺cKit⁻CRTH2⁻, Lin⁻CD127⁺cKit⁺CRTH2⁻ and Lin⁻CD127⁺CRTH2⁺ ILCs; Supplemental Fig. 4C). This was specific for Lin⁻CD127⁺cKit⁺CRTH2⁺ ILCs, as similar proportions of lineage⁺ lymphocytes (i.e., B and T cells; for gating, see Supplemental Fig. 4A) expressed CD62L in patients and healthy controls (Supplemental Fig. 4B). As seen in mouse for CD69, and in line with previous observations (45), the expression of NKp44 was mutually exclusive with CD62L on PB ILCs (Fig. 6D), further implying that CD62L marks nonactivated cells. Similarly, and as previously shown (43, 45), NKp44⁺ ILCs derived from human tonsils were uniformly negative for CD62L expression (Fig. 6E). In addition, α4β7⁺ and CXCR3⁺ ILCs from all CD patients lacked CD62L expression, whereas this was not the case in healthy controls (Fig. 6F), showing that activated, gut-homing Lin⁻CD127⁺cKit⁺CRTH2⁻ ILCs do not express CD62L in these patients. Overall, our data support a model in which CD62L is a functional marker for recirculating ILCP, which are reduced in PB of CD patients.

In this study, we demonstrate for the first time, to our knowledge, that CD62L⁺ ILCP have the capacity to circulate between PB and LNs. Their migration from blood to pLN is dependent on CD62L, whereas LN exit requires S1P receptors, similar to naive T cells. CD62L expression was mutually exclusive with expression of activation markers (e.g., NKp44 and CD69), and CD62L⁺ ILCP are reduced in numbers in PB of CD patients. These results show that CD62L functionally marks ILCP and that expression of CD62L on ILCP has potential to serve as a clinical marker in inflammatory diseases.

In a number of recent reports, CD62L was shown to mark a population of less-differentiated ILC3 and ILC progenitors in both humans (43, 52) and mice (53). In line with these reports, we show in this study for the first time, to our knowledge, that the recently reported PB ILCP (Lin⁻CD127⁺CRTH2⁻cKit⁺CD7⁺CD45RA⁺) (30) express CD62L and accordingly lower levels of the lineage defining transcription factor RORγt in comparison with tonsil-derived Lin⁻CD127⁺CRTH2⁻cKit⁺CD62L⁻ cells (i.e., mature ILC3). In addition, CD62L⁺ ILCP in both tonsil and PB expressed similar levels of this transcription factor, further implying that these cells belong to the same (circulating) subset. As expression of CD62L is furthermore mutually exclusive with expression of activation markers [e.g., NKp44 on human ILC3 and CD69 on mouse SLO ILCs (Figs. 6D, 2B, respectively)] and present on cells that express lower levels of lineage-defining markers (Figs. 1F, 2D–G), we conclude that CD62L is a functional marker for recirculating ILCP.

Two recent reports show that SLOs play an important role in determining the migration patterns of ILCs in mice as imprinting in skin or gut-draining LN was required for migration of ILCs to the skin or intestine, respectively (32, 33). We showed that whereas the CD62L⁺ population in pLN was influenced in vivo by treatment with MEL-14 and FTY-720, blocking LN entry and egress of this population, respectively, no changes were seen in the CD62L⁻ population (Figs. 3–5). This latter observation is in agreement with reports showing that mature (i.e., CD62L⁻) ILC are tissue-resident cells (46).

Our model predicts that similar to what is known in T cell biology, CD62L⁺ ILCP will differentiate into CD62L⁻ ILCs, which express subset-defining markers. This differentiation in LN will most likely occur under the influence of proinflammatory stimuli derived from Ag-presenting cells and, most likely, additional factors. Upon acquiring such activating signals, ILCs will start expressing homing receptors, allowing migration to inflamed tissues upon their re-entry into the circulation. Indeed, an influx of blood-derived ILC2s was found in the lungs following local inflammation in a parabiotic mouse model (46). When they arrive at the inflamed site, the ILCs will provide help to innate immune cells by their secretion of T cell cytokines at a time that adaptive lymphocytes are just starting to divide within the LNs. Supporting this model, in several settings of human immune-mediated disease, the phenotype of ILCs in PB has been shown to vary in patients when compared with healthy individuals (27, 29, 54). Most importantly, patients lacking ILCs in PB were shown to lack ILCs in tissues as well (55). The latter provides further evidence that circulating ILCP are not a separate population and are a source of tissue-residing, fully matured ILCs. Although this does not prove that the cells mature in SLO, the fact that the vast majority of ILCP express CD62L, whereas activated and matured ILCs do not, supports our premise. In support of receiving activation signals and subsequently homing cues for inflamed tissues within SLOs, ILCs within the circulation express homing receptors for the diseased organs [e.g., a₄β₇ in CD patients (Fig. 6E) and cutaneous lymphocyte Ag in psoriasis patients (27, 29, 34)]. Importantly, however, these data do not exclude that, under certain conditions, ILCP could also differentiate in organs other than SLOs. The fact that circulating CD62L⁻ ILCP, although constituting a very small minority of the cells, can differentiate into all ILC subsets shows that these CD62L⁻ cells are indeed ILCP. The absence of CD62L on these ILCP predicts that their migratory capacity is different and that these cells likely will differentiate into fully matured ILCs at other locations. In line with this, an alternative explanation for the increase in numbers of CD62L⁻ ILCP in PB of CD patients can be direct recruitment of ILCP from the bone marrow to the affected organ (in this case, the intestine), bypassing the need for LN activation. Further characterization of this CD62L⁻ population should show whether these cells are activated cells or ILCP directly recruited from the bone marrow.

Taken together, we describe in this article for the first time, to our knowledge, that the majority of PB ILCP express CD62L and migrate through the mammalian body in much the same way as has been demonstrated for naive T cells. We propose a model in which circulating ILCP are activated in LNs when an inflammation is ongoing and subsequently obtain the homing receptors for the inflamed tissues. In the future, it will be interesting to address whether the presence or absence of CD62L⁺ ILCP can serve as a sensitive biomarker for CD as well as for other inflammatory disorders and whether this correlates with disease activity.

We thank Dr. T. Cupedo and Prof. Dr. G. Kraal for carefully reading the manuscript. We also thank Dr. G. Graham for the kind gift of biotin-labeled recombinant human CCL21 for FACS staining of human CCR7 and Prof. Dr. F. Koning for the helpful discussion.

The authors have no financial conflicts of interest.