Epidermal γδ T cells originate from yolk sac hematopoiesis and clonally self-renew in the adult

Mammalian hematopoiesis is a multi-step and multi-site process. Early hematopoiesis occurs in the extra-embryonic yolk sac (YS) and is restricted to few hematopoietic lineages. Between embryonic day (E) 7 and E10 of mouse development, YS hematopoiesis generates erythrocytes, megakaryocytes, macrophages, and mast cells (Palis et al., 1999; Orkin and Zon, 2008; Ginhoux et al., 2010; Gomez Perdiguero et al., 2015; Hoeffel et al., 2015; Mass et al., 2016; Gentek et al., 2018). Starting around E10, multipotent hematopoietic stem cells (HSCs) are produced in the major arteries of the mouse embryo, including the aorta-gonado-mesonephros region (Boisset et al., 2010). In the adult, HSCs colonize the bone marrow (BM) and are thought to generate all hematopoietic lineages (Sawai et al., 2016). While most immune cells exclusively derive from HSCs in the mouse, some lineages such as macrophages and mast cells are sequentially generated from YS progenitors and HSCs (Gomez Perdiguero et al., 2015; Hoeffel et al., 2015; Mass et al., 2016; Gentek et al., 2018).

Myeloid cells exhibit marked tissue residency, whereas the majority of lymphocytes is circulating between lymphoid sites. However, populations of CD8⁺ memory T cells, innate-like lymphocytes such as innate lymphoid cells, and unconventional T cells, which include γδ T cells, also exhibit tissue residency primarily in barrier tissues, where they are thought to form a first line of defense against pathogens. One such tissue is the epidermis, the outermost layer of the skin that in mice harbors two resident immune cell populations: myeloid Langerhans cells (LCs) and lymphoid dendritic epidermal T cells (DETCs; Bergstresser et al., 1983). DETCs are intraepithelial γδ T cells that express a T cell receptor with the Vγ5 Vδ1 rearrangement and a restricted repertoire (Heilig and Tonegawa, 1986). DETCs modulate inflammation, protect against cutaneous neoplasia, and are implicated in skin wound healing (Girardi et al., 2002, 2003; Jameson et al., 2002; Sharp et al., 2005; Havran and Jameson, 2010; MacLeod et al., 2014). They develop from fetal thymic progenitors that seed the epidermis in low numbers before birth. Numbers of DETCs increase until adulthood is reached, after which they remain stable (Asarnow et al., 1988; Havran and Allison, 1988; Elbe et al., 1989; Havran et al., 1989; Payer et al., 1991; Jameson et al., 2004). However, the hematopoietic sources that generate these progenitors as well as the cellular mechanisms by which the DETC network is maintained once fully established are poorly characterized.

Here, we addressed the homeostatic maintenance mechanisms of DETCs and their hematopoietic origin using a combination of multicolor and hematopoietic fate mapping systems. We revealed that the adult DETC network undergoes slow homeostatic turnover mediated by tissue-resident proliferative units. The same process, albeit enhanced, replenishes adult DETCs after injury. Thus, DETCs maintain and replenish through the same cellular mechanisms as LCs (Ghigo et al., 2013). In the light of these similarities, we hypothesized that DETCs and LCs might also have common developmental origins. Hematopoietic lineage tracing demonstrated that DETCs are established from YS hematopoiesis independently of definitive HSCs, indeed reminiscent of LCs. Epidermal DETCs and LCs thus share cellular dynamics and hematopoietic features, characteristics that are likely at least partially controlled by their microenvironment.

We first investigated the dynamics of DETC homeostasis in adulthood. In line with previous reports, we observed no contribution of transplanted BM to DETCs 3 mo after transplantation into mice that underwent full-body irradiation (Fig. 1 A and Fig. S1; Honjo et al., 1990; Sumaria et al., 2011). Similarly, DETCs remained exclusively of host origin in parabiotic mice (Fig. 1 B and Fig. S1), in which chimerism of circulating cells is achieved surgically through establishment of a joint circulation (Waskow, 2010; Kamran et al., 2013). Chimeric and parabiotic approaches therefore confirmed DETCs as tissue-resident cells that are radioresistant and maintain themselves independently of circulating progenitors under homeostatic conditions. These features were highly reminiscent of LCs and raised the possibility that DETCs do turn over in adult mice via local self-renewal, as shown for LCs (Merad et al., 2002; Hoeffel et al., 2012; Ghigo et al., 2013).

Tools that are now routinely used to track lymphocytes, such as intravital 2-photon microscopy, are inherently unsuited to resolve slow cellular dynamics that take days, weeks, or months, kinetics that likely apply to DETCs (Cahalan et al., 2002). Recombinase-driven multicolor fate mapping systems such as “Brainbow” mice, on the other hand, represent valuable tools to analyze cellular dynamics in situ (Livet et al., 2007; Hadjieconomou et al., 2011; Hampel et al., 2011; Ghigo et al., 2013; Pan et al., 2013; Mondor et al., 2016). These models enable clonal analysis in situ and thereby provide insights into cellular homeostasis and precursor-product relationships. The “Ubow” mouse is a derivative of the Brainbow model, in which the multicolor reporter cassette is under control of the human UbiquitinC (Ubc) promoter (Ghigo et al., 2013). This allows for combinatorial expression of three spectrally different fluorescent reporter proteins by Cre recombinase activity (Fig. 1 C). In the absence of Cre, Ubow cells ubiquitously express tdTomato. Upon Cre expression, cells perform a single and definitive recombination event that leads to the random acquisition of CFP and/or YFP, which is transmitted to their progeny. As DETCs express the fractalkine receptor CX3CR1 (Almeida et al., 2015), we intercrossed Ubow and Cx3cr1^Cre mice (Yona et al., 2013) to perform multicolor fate mapping of DETCs (Fig. 1, C–E). In Cx3cr1^Cre/wt Ubow^+/+ mice, DETCs (identified as CD45⁺ CD3⁺ TCR Vγ5⁺ MHCII⁻) indeed expressed CFP and/or YFP (Fig. 1 C), indicating that they have undergone recombination of the Ubow construct. Although Cre-mediated recombination occurs in a stochastic fashion in Ubow mice, we consistently observed a bias toward YFP expression, as reported before (Ghigo et al., 2013). We took advantage of this bias by tracking the rare CFP^+/+ DETC population (on average 2.5% of all DETCs) to exemplify the behavior of the entire population. We thus investigated the arrangement of colored DETCs in the epidermis of adult Cx3cr1^Cre/wt Ubow^+/+ mice by imaging. Confocal microscopy revealed that CFP^+/+ DETCs were, in fact, not randomly distributed in the epidermis but instead assembled in mono-colored foci (Fig. 1 D). This was confirmed by quantification of CFP^+/+ DETC clusters using “ClusterQuant” analysis, software that compares the experimentally observed assembly of colored cells in clusters with their computationally simulated random distribution (Ghigo et al., 2013). This approach indicated with high significance that CFP^+/+ DETCs form clusters (Fig. 1 E). Thus, like LCs, DETCs assemble in tissue-resident, proliferative units in the adult epidermis.

In Cx3cr1^Cre Ubow mice, recombination events are initiated as soon as DETCs express Cx3cr1 during their development. These data therefore cannot delineate whether DETCs turn over constantly or if cluster formation is the result of transient expansion confined to a specific developmental stage followed by minimal homeostatic turnover in the adult, as has been reported for LCs and suggested for DETCs (Havran and Allison, 1990; Havran et al., 1991; Payer et al., 1991; Matsue et al., 1993; Chorro et al., 2009). To estimate the extent of homeostatic DETC turnover in adult mice, we measured the incorporation of BrdU by DETCs over several weeks. LCs gradually incorporated BrdU over time, and after 5 wk, ∼20% of them were labeled (Fig. 2 A; Kamath et al., 2002). On the contrary, the cumulative fraction of BrdU-labeled DETCs reached only 2% over this period, confirming that adult DETCs are cycling very slowly at steady state (Fig. 2 A; Sumaria et al., 2011).

To track this slow homeostatic turnover, we turned to a model in which Ubow recombination can be induced in adult DETCs. We generated Cx3cr1^CreERT2/wt Ubow^+/+ mice (Chorro et al., 2009; Yona et al., 2013). In these mice, injection of a low dose of tamoxifen induces recombination in some DETCs, as a consequence of which they stochastically acquire expression of CFP and/or YFP, hence generating an initial mosaic of colored DETCs. We reasoned that longitudinal analysis of tamoxifen-treated Cx3cr1^CreERT2/wt Ubow^+/+ mice should unravel any slow homeostatic DETC turnover occurring in the adult. As BM-derived cells do not contribute to adult DETC homeostasis, we reasoned that every dying DETC should be replaced by a tissue-resident one. If DETCs were capable of replacing their dying neighbors by local proliferation, mono-colored foci composed of dividing DETCs and their progeny should develop over time. 1 wk after tamoxifen administration to adult mice, ∼6–7% of DETCs had undergone recombination and were randomly distributed (Fig. 2, B–D). Analysis of DETC cluster formation in adult Cx3cr1^CreERT2/wt Ubow^+/+ mice up to 6 mo after tamoxifen treatment revealed that the initial mosaic of CFP^+/+, YFP^+/+, and CFP⁺YFP⁺ DETCs was indeed progressively replaced by rare foci of mono-colored DETCs similar to the ones observed in constitutive Cx3cr1^Cre/wt Ubow^+/+ mice (Fig. 1 and Fig. 2, C and D).

Collectively, these data uncovered that the adult DETC network does slowly turn over at steady state. This turnover is achieved through local self-renewal, resulting in clonal units comprised of proliferating DETCs and their progeny. Such cellular dynamics are highly reminiscent of LCs, whose homeostasis also relies on clonal proliferative units (Ghigo et al., 2013). However, this process is considerably slower for DETCs, taking several months to be detectable in the Ubow system. These distinct turnover rates probably reflect differences in the frequency with which these cell types need to be replaced. Indeed, LCs, but not DETCs, continuously migrate from the epidermis to skin draining lymph nodes, both under inflammatory conditions and at steady state (Merad et al., 2008). Such constant need for partial replenishment of the LC network likely explains their higher turnover (Merad et al., 2002; Chorro et al., 2009). The low-grade clonal self-renewal of DETCs, on the other hand, might serve to replace the few DETCs dying over time.

The magnitude of and mechanisms underlying DETC turnover might be different, however, when homeostasis is perturbed. To investigate whether local self-renewal is also involved in the maintenance and replenishment of the DETC network under conditions of tissue injury, we used tape stripping (TS), a model in which the repeated application of adhesive tape to the ear induces epidermal abrasion and an inflammatory response (Holzmann et al., 2004; Strid et al., 2011). We first subjected WT mice to TS and measured the proliferation of DETCs at the peak of inflammation (10 d) and when the epidermis had been repaired (30 d) using 5-ethynyl-2′-deoxyuridine (EdU) incorporation (Fig. 3 A). TS induced enhanced LC and DETC proliferation at the peak of inflammation, with 6% of DETCs in S-phase compared with 1% in untreated control epidermis. After 30 d, DETC proliferation had returned to baseline. To address whether DETC restoration following injury was also supported by the mobilization of BM-derived progenitors, we performed TS on BM chimeric mice (Fig. 3 B). As previously observed (Ghigo et al., 2013), BM-derived cells engrafted in the LC network of the repaired epidermis at low frequency. DETCs, on the other hand, did not receive any BM contribution by 30 d after TS. Together, these results demonstrate that DETCs locally respond to injury with a transient increase in proliferation.

To directly test whether this seemingly autonomous DETC replenishment following injury was mediated by local clonal proliferation, we analyzed cluster formation in adult tamoxifen-treated Cx3cr1^CreERT2/wt Ubow^+/+ mice that were then subjected to TS of one ear (Fig. 3, C and D). Confocal imaging revealed that DETCs assembled in clusters already 1 mo later in TS but not untreated control ears. These findings were in line with the enhanced proliferative response observed by EdU incorporation and indicate that clonal self-renewal replenishes the DETC network upon perturbation. Hence, DETC homeostasis and replenishment seem to depend on similar cellular strategies.

At steady state, DETCs are polarized, immotile cells that possess anchored projections at keratinocyte tight junctions (Chodaczek et al., 2012). However, DETCs are known to round up following wounding and could thus reacquire motility, migrate, and proliferate to replenish the damaged epithelial network in a distant location (Havran and Jameson, 2010; Strid et al., 2011). Therefore, an alternative, albeit not mutually exclusive, mechanism by which the damaged DETC network could be replenished independently from peripheral progenitors is through relatively long-range migration and subsequent proliferation of DETCs (progenitors) from distant unperturbed locations, a possibility that our experiments do not formally exclude.

To assess migration of DETCs in epidermal wounding, we established skin grafts, in which a piece of WT skin was grafted onto the back of Cx3cr1^gfp/wt Ubc^tdt mice (Fig. 3 E). In this model, host DETCs can be identified by virtue of their GFP expression, while the graft border is visible from the mismatch between ubiquitously tdTomato-expressing host cells and the nonfluorescent graft. Whole mount confocal imaging 4 mo after grafting revealed that host DETCs were only found in the graft within 600 µm of the border and constituted a sizeable fraction of 10–20% only within 300 µm. Taking into account the total graft size of ∼10 mm × 15 mm, host DETCs were thus mostly restricted to the immediate edge of the wound. These data therefore argue against long-range migration of DETCs and favor a model in which the DETC network in damaged epidermis is primarily restored through proliferation of neighboring DETCs or their progenitors.

Overall, DETCs and LCs display striking similarities: both are tissue-resident cells occupying the epidermis, where they form comparably dense networks of radioresistant cells that self-maintain independently of input from the BM and circulating progenitors. At steady state, both networks are dynamic structures composed of adjacent proliferative units. Clonal proliferation is also the major mechanism by which DETCs and LCs are replenished upon perturbation, albeit at different frequencies. Intriguingly, DETCs and LCs also share some developmental characteristics: both lineages depend on the transcription factor Runx3, and, as outlined above, although they continue to mature in the perinatal period in terms of their phenotype and numbers, both networks are established prenatally (Havran and Allison, 1990; Payer et al., 1991; Elbe et al., 1992; Fainaru et al., 2004, 2005; Woolf et al., 2007; Hoeffel et al., 2012). Indeed, neither LCs nor DETCs showed any contribution of transplanted BM in the epidermis of adult mice that had been shield-irradiated as newborns (Fig. 4 A), a model that allows BM engraftment under relatively unperturbed conditions in this developmental window (Scott et al., 2016; Baratin et al., 2017). Considering these similarities, we hypothesized that DETCs and LCs might also share hematopoietic characteristics. The majority of LCs derives from fetal liver monocytes recruited to the skin starting around E14.5, while the remainder originates from progenitors produced by early YS hematopoiesis (Hoeffel et al., 2012). DETCs are established from fetal thymic progenitors that settle in the epidermis at mid- to late gestation, similar to LC precursors (Elbe et al., 1989, 1992; Havran and Allison, 1990; Payer et al., 1991).

Unlike for LCs, however, the hematopoietic source of these progenitors, and by extension of adult DETCs, has not been identified yet. We sought to address the ontogeny of DETCs using genetic fate mapping. We recently established that Cdh5-CreERT2 mice can be used to efficiently and precisely label either early YS- or HSC-dependent definitive hematopoiesis in a temporally defined manner (Gentek et al., 2018). This model exploits the fact that Cdh5 (encoding VE-cadherin) is expressed by endothelial cells, but down-regulated immediately upon endothelial-to-hematopoietic transition (Breier et al., 1996; Zovein et al., 2008). In Cdh5-CreERT2 Rosa^tdt mice, administration of a single dose of 4-hydroxytamoxifen (4OHT) at defined embryonic time points therefore labels hemogenically active endothelial cells and fate maps their hematopoietic output. Specifically, 4OHT treatment at E7.5 labels the hemogenic endothelium of the YS and its early output, including microglia. Administration of 4OHT at E10.5, on the other hand, labels the hemogenic endothelium of the aorta-gonado-mesonephros and consequently, definitive HSCs and their progeny such as adult circulating leukocytes, but not microglia (Gentek et al., 2018; Fig. 4 B). Using this model, we found that the epidermis of adult Cdh5-CreERT2^+/wt Rosa^tdt/wt mice contains a minor but sizeable population of E7.5-labeled DETCs generated by early YS hematopoiesis (Fig. 4 B). Unlike LCs, however, adult DETCs showed only marginal E10.5 labeling, indicating that they do not originate from adult-type, HSC-dependent definitive hematopoiesis. These findings were confirmed in Runx1^CreERT2/wt Rosa^yfp/wt mice, in which 4OHT administration at E7.5 fate maps early YS hematopoiesis, while pulsing at E9.5 primarily fate maps HSC-dependent definitive hematopoiesis (Ginhoux et al., 2010; Hoeffel et al., 2012, 2015; Fig. 4 C). Thus, while early YS hematopoiesis marginally contributes to adult DETCs, the majority of DETCs originates from neither early YS progenitors nor definitive HSC-dependent hematopoiesis.

While seemingly surprising, these findings are in line with the recruitment of progenitors with T cell potential to the thymus anlagen as early as E11 (Jotereau et al., 1987; Masuda et al., 2005; Bell and Bhandoola, 2008). Irrespective of whether they represent the extreme of a continuum or a truly distinct wave, the majority of DETCs could be produced by late YS hematopoiesis. In support of this hypothesis, DETCs were fate mapped to a similar degree as LCs in Runx1^CreERT2/wt Rosa^yfp/wt mice treated with 4OHT at E8.5, an approach that primarily labels the output of late YS, but not early YS and definitive hematopoiesis (Hoeffel et al., 2015; Fig. 4 C). Unlike LCs, however, DETCs never showed any contribution of cells labeled with 4OHT in Csf1r^CreERT2 Rosa^yfp/wt mice, a model known to label erythro-myeloid progenitors generated by HSC-independent YS hematopoiesis, which contribute to the first wave of LCs (Gomez Perdiguero et al., 2015; Hoeffel et al., 2015; Fig. 4 D). Thus, although they share developmental kinetics, DETCs and LCs are established from distinct progenitors.

Several candidates exist for such DETC progenitors. A recent study identified a population of embryonic “developmentally restricted HSCs” that exhibits marked differentiation bias toward tissue resident innate-like lymphocytes, including TCR Vγ5⁺ T cells. Furthermore, progenitors with lymphoid and lymphomyeloid potential have been described in the YS (Palacios and Imhof, 1993; Yoshimoto et al., 2012; Böiers et al., 2013). Such progenitors are not endowed with long-term reconstitution capacity. However, once seeded in their tissue of residency, their progeny could be maintained by local self-renewal throughout life. This would be in agreement with the mode of homeostatic maintenance we unraveled here for adult DETCs. Based on these considerations and our fate mapping data, we conclude that DETCs are produced by YS hematopoiesis. To our knowledge, this makes DETCs the only lymphoid lineage with identified YS origin. Much like their adult maintenance and replenishment mechanisms, their developmental origin closely resembles that of LCs, with which they co-occupy the epidermis. This suggests that in addition to lineage-specific factors like IL-34 for LCs and IL-2 and IL-15 for DETCs, other factors produced by, e.g., keratinocytes might regulate local development and maintenance of both DETCs and LCs (Ye et al., 2001; Greter et al., 2012; Wang et al., 2012).

In summary, our study demonstrated that adult DETCs undergo slow homeostatic turnover and identified clonal self-renewal as the underlying cellular mechanism. The same mechanism, albeit accelerated, is responsible for replenishing the DETC network following injury. DETCs therefore bear striking resemblance to LCs that reside in the same environment. Moreover, we revealed that DETCs and LCs also share an origin from HSC-independent YS hematopoiesis that, within lymphocytes, is unique to DETCs. Collectively, these data suggest that the microenvironment imposes a dominant degree of regulation on its resident immune cells, independently of their lineage identity. Whether such environmental imprinting also applies to other tissues should be subject to future studies.

Mice housed at the Centre d’Immunologie de Marseille-Luminy (Marseille, France) were maintained under specific pathogen–free conditions in a 12-h light/dark cycle at an ambient temperature of 22°C. Animals were kept in individually ventilated cages equipped with sterile quarter-inch corncob bedding and cotton pads for environmental enrichment. They were fed with irradiated standard pellet chow and reverse osmosis water. Experiments were performed in accordance with French and European guidelines for animal care under the permission no. 5–01022012 following review and approval by the local animal ethics committee in Marseille, France. Mice were on a C57BL/6J background (minimally eight backcrosses) and used between 6 and 12 wk of age, unless stated otherwise. Cx3cr1^Cre (B6J.B6N(Cg)-Cx3cr1^{tm1.1(cre)Jung}/J, strain 25524; Jung et al., 2000), Cx3cr1^CreERT2 (B6.129P2(C)-Cx3cr1^{tm2.1(cre/ERT2)Jung}/J, strain 20940; Yona et al., 2013), Ubc^gfp (UBC-GFP 30Scha/J, strain 4353), and Rosa^tdt (also known as Ai14, strain 7914; Madisen et al., 2010) were purchased from The Jackson Laboratory. C57BL/6J (WT), CD45.1 (CD45.1 WT), and RjOrl:SWISS (CD1) mice were from Janvier Labs. Ubc^tdt (Ubow) mice were generated at the Service des Animaux Transgéniques (Orléans, France) as described previously (Ghigo et al., 2013). Cdh5-CreERT2 mice (Sörensen et al., 2009) were a gift from Ralf Adams (Max Planck Institute for Molecular Biomedicine, Muenster, Germany). Experiments using Runx1^CreERT2 (Samokhvalov et al., 2007) and Csf1r^CreRT2 mice (The Jackson Laboratory) were performed at the Singapore Immunology Network (Singapore) following approval by the local ethics committee.

10-wk-old WT CD45.1 and Cx3cr1^gfp/wt CD45.2 mice were surgically joined as described previously (Waskow, 2010). Parabionts were analyzed 2 mo later.

To generate regular BM chimeras, mice were subjected to full-body irradiation using an x-ray source and a total dose of 7 Gy. Mice were then reconstituted with at least 5 × 10⁶ BM cells prepared using standard procedures. For shielded newborn chimeras, newborn mice were placed in a 6-mm-thick lead shield that only left their hind legs exposed while protecting the rest of the body from irradiation. Mice were reconstituted intravenously with 3 × 10⁷ BM cells depleted of CD11b⁺ cells through negative magnetic selection.

Cdh5-CreERT2 fate mapping was performed as in Gentek et al. (2018). Briefly, timed matings were set up between WT females aged 6–10 wk and Cdh5-CreERT2^+/wt Rosa^tdt/tdt or Cdh5-CreERT2^+/+ Rosa^tdt/tdt males. The presence of vaginal plugs was assessed the morning after, which was considered 0.5 d after conception. To induce reporter recombination in the offspring, a single dose of 4-hydroxytamoxifen (4OHT) was delivered by i.p. injections to pregnant females at E7.5 or E10.5. 4OHT solutions were supplemented with progesterone to counteract adverse effects of 4OHT on pregnancies. Females received a dose of 1.2 mg 4OHT and 0.6 mg progesterone. In cases when females could not give birth naturally, pups were delivered by C-section and cross fostered with lactating CD1 females. Runx1^CreERT2 and Csf1r^CreRT2 fate mapping experiments were performed at the Singapore Immunology Network (Singapore) as described previously (Hoeffel et al., 2015). Fate mapped Cdh5-CreERT2^+/wt Rosa^tdT/wt and Runx1^CreERT2/wt Rosa^yfp/wt mice were analyzed in the adult, and an analysis time course was performed for Csf1r^CreRT2 mice. Statistical significance was assessed by an unpaired, nonparametric t test (Mann-Whitney).

Adult Cx3cr1^CreERT2/wt Ubow^+/+ mice received a single dose of tamoxifen (1 mg) dissolved in corn oil (both Sigma-Aldrich) by i.p. injection.

For longitudinal analysis of BrdU incorporation, mice were administered BrdU in their drinking water for up to 4 wk, the maximal possible duration due to BrdU toxicity, similar to what has previously been published for LCs (Merad et al., 2002). BrdU incorporation was detected using a BrdU Flow cytometry kit (BD Biosciences) following the manufacturer’s protocol. To quantify cells currently in S-phase, EdU incorporation assays were performed. Mice were injected i.p. with 1 mg of EdU and sacrificed 14 h later. Epidermal DETCs and LCs were enriched using CD45 microbeads (Miltenyi Biotec), and EdU incorporation was detected using the Click it EdU flow cytometry kit (Life Technologies) according to the manufacturer’s instructions.

Mice were anesthetized with ketamine/xylazine. The dorsal side of the left ear was stripped by repeated application (15 times) of tape (3M comply indicator tape, 18 mm).

Recipient mice were anesthetized with ketamine/xylazine. A single piece of back skin (∼1 cm × 1.5 cm) was grafted in the center of the dorsal region of the recipient. Mice were treated with ibuprofen for 1 mo following surgery. Bandages were removed 7 d after surgery, and graft integrity was observed weekly.

Cell suspensions were obtained through mechanical dissociation and enzymatic digest. Briefly, brain and epidermis samples were harvested, finely minced with scissors, and digested in RPMI medium (Invitrogen) containing 0.1 mg/ml DNase I (Sigma-Aldrich) and 400 U/ml Collagenase I (Gibco). Epidermis was harvested from the ear skin. Briefly, dorsal and ventral sheets were separated using forceps, and the epidermis was removed by incubation in Dispase in PBS for 1 h at 37°C before digestion. Digests were performed for 30–45 min at 37°C under continuous agitation (900 rpm), and samples were regularly pipetted up and down to support mechanical dissociation and digest. Blood was drawn from the retro-orbital vein, and red blood cell lysis was performed before staining. Digested tissue samples were filtered, and single-cell suspensions were stained with fixable live dead stains (BioLegend), blocked in a suspension containing normal rat (Jackson ImmunoResearch) and mouse serum (Thermo Fisher Scientific) and anti CD16/32 (“TruStain Fx”; BioLegend). Surface staining was performed in FACS buffer supplemented with Brilliant Stain Buffer (BD Biosciences) for 20–30 min at 4°C. Data were acquired on LSRII and LSR Fortessa X-20 instruments (BD Biosciences) and analyzed using FlowJo (Treestar) and Prism (GraphPad) software.

To prepare epidermal sheets, the dorsal and ventral halves of ear skin were pulled apart as for flow cytometry and incubated in Dispase for 40 min. Epidermal sheets were separated from the dermis, fixed in AntigenFix (Microm Microtech) for 20 min at room temperature, and washed in phosphate buffer. DETCs were identified by immunostaining TCR Vγ5 expression. Immunofluorescence confocal microscopy was performed with a Zeiss LSM 780 confocal microscope. Final image processing was performed with Imaris software (Bitplane) and Adobe Photoshop.

In addition to critical reagents listed throughout this section, the following antibodies were used for flow cytometry and confocal imaging: anti–mouse CD45 (30-F11; BD Biosciences), CD3 (17A2; BioLegend), TCR Vγ5 (536; BioLegend), MHCII (M5/114.15.2; eBioscience), F4/80 (BM8; BioLegend), CD64 (X54-5/7.1; BioLegend), CD11b (M1/70; BioLegend), and Ly6C (AL-21; BD Biosciences).

Quantification and statistical analysis of the spatial distribution of DETCs consist of three modules combined in the ClusterQuant software, which we developed with the CognitionMaster image analysis platform based on C#/.NET. In the first module, DETC classes are defined using the Region-of-Interest plugin (ROImanager) of ClusterQuant/CognitionMaster. The respective cells are labeled manually to obtain the cell center coordinate information for the subsequent processing step. The second module computes a two-dimensional Voronoi mesh based on the manually labeled cell centers. The resulting tessellation establishes the neighborhood relationship between the cells, which is used to evaluate the number of same-type cells assembled in clusters as well as the respective cluster size. This step provides a robust and objective means of cell cluster quantification. To assess the degree of randomness of the observed cell distribution/cluster formation, Monte Carlo simulations are used. These are based on the experimentally determined cell numbers. For each experimental dataset, 10,000 simulations are performed, in which a number of cells corresponding to the experimentally found cells is randomly distributed over the same image areas. For each simulation run, the spatial statistics (Voronoi mesh, neighborhood, and cell cluster features) are computed in same way as for the experimental data. The results of all simulations are then compared with the “original,” experimentally obtained data. To this end, the ratio is formed between the number of simulations in which the average cluster size is larger than in the experimental data and the total number of Monte Carlo simulations. This yields P values that indicate the probability that experimentally observed clusters could have been achieved by random cell distribution. For more details, see Ghigo et al. (2013).

Fig. S1, related to Fig. 1, shows the gating strategies exploited to identify epidermal DETCs and LCs as well as representative flow cytometry data for BM chimeric and parabiotic mice.

We thank the Centre d'Immunologie de Marseille-Luminy core facilities for animal housing, flow cytometry, and imaging (Plateforme d’Imagerie Commune du Site de Luminy, ImagImm, member of the national infrastructure France-BioImaging, supported by the French National Research Agency, ANR-10-INBS-04).

This study was funded by grants from the Agence National pour la Recherche (ANR-10-INBS-04-01 France-BioImaging, ANR-08-JCJC-0134). Additional support was provided by institutional grants from Institut National de la Santé et de la Recherche Médicale, Centre National de la Recherche Scientifique, and Aix-Marseille University to the Centre d'Immunologie de Marseille-Luminy. F. Ginhoux is a European Molecular Biology Organization Young Investigators Programme awardee and is supported by Singapore Immunology Network core funding as well as a Singapore National Research Foundation Senior Investigatorship (NRF2016NRF-NRFI001-02).

The authors declare no competing financial interests.

Author contributions: M. Bajénoff, R. Gentek, F. Ginhoux, and C. Ghigo designed the study. R. Gentek and M. Bajénoff wrote the manuscript. R. Gentek, C. Ghigo, G. Hoeffel, A. Jorquera, R. Msallam, S, Wienert, F. Klauschen, and M. Bajénoff performed and analyzed experiments. F. Ginhoux and G. Hoeffel edited the manuscript.