Cbfβ2 deficiency preserves Langerhans cell precursors by lack of selective TGFβ receptor signaling

The skin is one of the body’s largest interfaces and is exposed to the outer environment, functioning as a physical barrier to protect against the invasion of pathogenic microorganisms. In addition to mechanical defense, two immune populations, namely dendritic epidermal T cells (DETCs) and Langerhans cells (LCs), reside specifically in the epidermis and participate in immunosurveillance. LCs are skin-specific dendritic cells that play an essential role in sensing pathogenic microorganisms and tissue damage to initiate immune responses and maintain skin homeostasis (Merad et al., 2008; Chopin and Nutt, 2015; Hieronymus et al., 2015; Collin and Milne, 2016). Consistent with such functions, the LC network is established immediately after birth when animals become exposed to the outside environment. Previous studies in mice showed that LC precursors, which arise from both yolk sac and fetal liver precursors (Hoeffel et al., 2012), migrate to the epidermis at 16.5 to 18.5 d postcoitus (dpc; Romani et al., 2010) and undergo sequential differentiation during neonatal periods to generate the adult LC network (Ginhoux and Merad, 2010; Perdiguero and Geissmann, 2016). During differentiation into mature LCs, precursors exhibit altered morphology, such as the protrusion of dendrites, and express the C-type lectin Langerin, MHC class II, and epithelial cell adhesion molecule (EpCAM; Chorro et al., 2009). Simultaneously, a proliferative burst in LC precursors begins at approximately postnatal day (P) 3, resulting in the establishment of a primary LC network in the epidermis within a week after birth in mice (Chorro et al., 2009; Ginhoux and Merad, 2010).

Adult LC steady-state homeostasis is maintained throughout life without replenishment by circulating precursors (Merad et al., 2008), whereas conventional DCs, which reside in other tissues, are continuously replaced by cells that differentiate from BM-derived DC precursors (Merad et al., 2008; Ginhoux and Merad, 2010; Chopin and Nutt, 2015; Schlitzer et al., 2015; Collin and Milne, 2016). In contrast, when the LC network is impaired by genetic treatment, such as in inducible Langerin-DTR mice (Bennett et al., 2005; Nagao et al., 2009), or via artificial or natural inflammation (Ginhoux et al., 2006; Seré et al., 2012), BM-derived Gr-1⁺ monocytes migrate to the epidermis to replenish the LC network.

The importance of the TGFβ superfamily in LC network formation has been studied in depth in both humans and mice. These studies highlight the role of the TGFβ superfamily as a major soluble environmental cue required for establishing the primary LC network (Merad et al., 2008; Collin and Milne, 2016). TGFβ superfamily signaling is triggered by binding to heterodimeric receptors, composed of a variable type I receptor that has distinct affinity to each TGFβ superfamily member and one common type II (TGFβR2) receptor. TGFβR2 is essential for the initiation of the intracellular signaling cascade, which activates several signal transducers including SMAD family proteins (Chen and Ten Dijke, 2016; Collin and Milne, 2016). Mice with genetic ablation of TGFβ1 or TGFβR2 lack the LC network in the epidermis (Borkowski et al., 1996; Kaplan et al., 2007). These findings confirm that TGFβ1 controls LC differentiation. However, more recent studies revealed another unexpected function of TGFβ1 signaling in the control of LC homeostasis. Ablation of TGFβ1 or TGFβR1 in mature LCs enhanced their egress from the epidermis (Kel et al., 2010; Bobr et al., 2012). Thus, TGFβ1 signaling through TGFβR1 is essential to preserve LCs in a quiescent state in addition to mediating their differentiation (Collin and Milne, 2016). Moreover, a recent study proposed that another member of the TGFβ superfamily, BMP7, plays a more prominent role in LC differentiation via binding to BMP receptor 1A (BMPR1A; Yasmin et al., 2013a). Thus, the current model of LC differentiation and homeostasis proposes that BMP7 and TGFβ1 are involved in distinct pathways, with BMP7 controlling cellular signaling that induces LC differentiation and TGFβ1 preserving the quiescent state of LCs (Collin and Milne, 2016). However, it remains unclear how cellular signaling, triggered by related but distinct TGFβ superfamily members, is regulated to control distinct cellular functions.

Several transcription factors are essential for LC differentiation. Both Id2 and PU.1 were shown to be required for LC differentiation from embryonic precursors (Seré et al., 2012; Chopin et al., 2013). However, LC differentiation from BM precursors depends only on PU.1 and not on Id2 (Chopin et al., 2013). Runx3 is another transcription factor that is essential for LC development (Fainaru et al., 2004). Runx3 belongs to the Runx protein family, which functions by forming heterodimers with the non-DNA binding β-subunit, Cbfβ protein (Wang et al., 1996). In mammals, two RNA splice variants, Cbfβ1 and Cbfβ2, are generated from a single Cbfb gene, and each variant has distinct amino acid sequences at the C terminus (Ogawa et al., 1993). However, whether Cbfβ1 and Cbfβ2 have nonredundant functions, particularly in regard to LC differentiation and homeostasis, has not been clarified in animal models.

In this study, we unraveled the essential function of Cbfβ2 in driving LC differentiation from embryonic precursors. Despite the complete inhibition of LC differentiation, embryonic-derived LC precursors were still detectable in the adult epidermis of Cbfβ2-deficient mice. Mechanistically, we showed that this was caused by the selective impairment of TGFβR signaling, which is essential only for promoting LC differentiation. Rescuing Cbfβ2 expression, using a novel transgenic model, resulted in the resumption of LC differentiation only during the neonatal stage, and not at the adult stage. Such defects in the developmental potential of LC precursors at the adult stage were caused by the intrinsic loss of BMP7/BMPR1A signaling, an essential driver of LC differentiation.

To address the unique functions of Cbfb splice variants, we generated a mouse model lacking either Cbfβ1 or Cbfβ2 expression. Based on flow cytometric analyses (Fig. 1 A) of epidermal sheets from 2-mo-old Cbfβ2-deficient mice (Cbfb^2m/2m), only a few cells expressed MHC-II, whereas expression in the epidermal sheet of Cbfβ1-deficient mice (Cbfb^1m/1m) was similar to that in wild-type (Cbfb^+/+) mice. Immunohistochemical analyses indicated that MHC-II⁺ cells in the epidermis of Cbfb^2m/2m mice lacked Langerin expression. In addition, CD3⁺ DETCs, known as skin-specific γδT cells, were not present in the epidermis of Cbfb^2m/2m animals (Fig. 1, A and B). Thus, as with Runx3-deficient mice (Fainaru et al., 2004), Cbfβ2 deficiency resulted in a lack of LCs and DETCs in the skin, although a few MHC-II⁺ cells were present in the Cbfb^2m/2m epidermis. These MHC-II⁺ cells showed dendritic morphology, were larger than control Cbfb^+/+ MHC-II⁺ LCs (Fig. 1 B), and expressed F4/80 and CD11b (Fig. 1 C), which are LC-lineage markers known to be expressed from an early developmental stage (Chorro et al., 2009). Importantly, these MHC-II⁺ cells did not express mature LC markers (Fig. 1 C), such as CD11c, EpCAM, and CD24 (Chorro et al., 2009), but highly expressed the CX3 chemokine receptor 1 (CX3CR1; Fig. 1 D), a known marker expressed in recently seeded embryonic day (E) 18.5 LC precursors but not in differentiated LCs (Hoeffel et al., 2012). However, because these remaining MHC-II⁺CX3CR1-GFP⁺ cells in the adult Cbfb^2m/2m epidermis retained a marker signature reminiscent of LC precursors, we hypothesized that they represent LC precursors that are blocked in their differentiation, prompting us to next examine LC ontogeny in Cbfb^2m/2m mice.

LC differentiation from embryonic precursors begins around birth (P0; Merad et al., 2008). As previously reported (Chorro et al., 2009), embryonic LC precursors at P0 can be defined as CD45⁺CD3⁻CD11b⁺F4/80⁺ cells. No significant difference in the percentage of epidermal cells was detected in Cbfb^2m/2m newborns compared with Cbfb^+/+ controls (Fig. 2, A and B), indicating that in these cells, embryonic generation and skin-homing programs were unaffected by Cbfβ2 loss. Induction of MHC-II followed by EpCAM and Langerin expression concomitant with the loss of CX3CR1 expression represents normal LC differentiation after birth (Chorro et al., 2009; Nagao et al., 2009). At P0, 10–20% of CD45⁺CD3⁻CD11b⁺F4/80⁺ cells expressed MHC-II from both Cbfb^+/+ and Cbfb^2m/2m mice (Fig. 2, B and C). Consistent with this finding, Cbfb1 and Cbfb2 expression was increased on the P4 MHC-II⁺EpCAM⁺ population (Fig. S1). By P3, as expected, an MHC-II⁺EpCAM⁺ population lacking CX3CR1 expression emerged from Cbfb^+/+ mice; however, we could not detect any cells with this phenotype from Cbfb^2m/2m mice. By P7, nearly all LC-lineage CD45⁺CD3⁻CD11b⁺F4/80⁺ cells from Cbfb^+/+ mice expressed MHC-II, and 75% of these differentiated into EpCAM⁺ cells. In contrast, in the epidermis of P7 Cbfb^2m/2m mice, the MHC-II⁺EpCAM⁻ subset represented the largest population, and MHC-II⁻EpCAM⁻ precursors were still present. In 1-mo-old Cbfb^2m/2m mice, the accumulation of MHC-II⁺EpCAM⁻ cells was even more apparent, and MHC-II⁻EpCAM⁻ precursors were no longer detectable. These results indicated that Cbfβ2 ablation leads to developmental arrest during the transition from the MHC-II⁺EpCAM⁻ (precursor LC) to the MHC-II⁺EpCAM⁺ (mature LC) stage.

Despite the observed developmental block during the Cbfb^2m/2m postnatal period, we noted that some MHC-II⁺EpCAM⁺ cells emerged in the epidermis from 4-mo-old Cbfb^2m/2m mice (Fig. 2, B and C). Importantly, these MHC-II⁺EpCAM⁺ cells were negative for CX3CR1 expression (Fig. 2 B), suggesting that they were not LC precursors. Analyses of other relevant markers revealed that these MHC-II⁺EpCAM⁺ cells express Langerin, CD24, and CD11c (Fig. 3 A). In addition, they were smaller in size than the MHC-II⁺EpCAM⁻ subset (Fig. 3 A). Thus, we phenotypically and morphologically characterized these cells as mature LCs.

Previous studies showed that circulating Gr-1⁺ monocytes, which are derived from BM progenitors, migrate to the epidermis and give rise to LCs in LC-depleted or -deficient epidermis (Ginhoux et al., 2006; Seré et al., 2012). Given the arrest of LC differentiation and the low numbers of LC-lineage cells in the epidermis of Cbfb^2m/2m mice, we speculated that an influx of Gr-1⁺ monocytes into the epidermis of adult Cbfb^2m/2m mice occurs, even during steady states. Indeed, Gr-1⁺MHC-II⁻ cells were detected in CD45⁺CD3⁻CD11b⁺ epidermal populations as soon as 1 mo after birth in Cbfb^2m/2m mice (Fig. 3 B). These epidermal Gr-1⁺ cells expressed lower levels of F4/80 than MHC-II⁺ cells (Fig. 3 B). The percentage of Gr-1⁺ cells in the CD45⁺CD3⁻CD11b⁺ epidermal population increased by up to 20% from 1 to 4 mo after birth, whereas these cells were almost undetectable in the Cbfb^+/+ epidermis (Fig. 3 B). These results not only suggested that mature LCs could emerge from Gr-1⁺ monocytes, but also raised questions regarding the origin of MHC-II⁺EpCAM⁻ LC precursor-like cells in the epidermis of adult Cbfb^2m/2m mice.

Radioresistance of LCs is a hallmark of their embryonic ontogeny (Merad et al., 2002). To further characterize the MHC-II⁺EpCAM⁻ cells in adult Cbfβ^2m/2m mice, we performed a BM chimera experiment using CD45.1⁺ wild-type BM cells injected into sublethally irradiated CD45.2⁺ Cbfb^+/+ or CD45.2⁺ Cbfb^2m/2m recipient mice. 7 mo later, nearly all CD11c⁺ DCs in the spleen of both Cbfβ^+/+ and Cbfβ^2m/2m recipients expressed the donor marker CD45.1 (Fig. 4 A). In contrast, the majority of LCs were of host origin (CD45.2) in Cbfβ^+/+ recipient animals (Fig. 4 A). However, in the epidermis of Cbfb^2m/2m recipient mice, 25% of the MHC-II⁺ population was of donor origin, suggesting that these donor cells were derived from donor Gr-1⁺ blood monocytes. Importantly, the majority of the MHC-II⁺ population with immature characteristics such as low Langerin expression remained of host origin (CD45.2) and consisted of LC precursors that exhibited radioresistance and tissue-residency properties of bona fide skin-residential LCs.

To examine tissue residency in noninflammatory conditions, we performed parabiosis experiments, in which parabionts shared the same blood circulation, allowing exchange of hematopoietic cells between them (Merad et al., 2002). After 7 mo, the splenic CD11c⁺ population of each parabiont consisted equally of CD45.1⁺ and CD45.2⁺ cells in both CD45.1⁺ Cbfb^+/+:CD45.2⁺ Cbfb^+/+ and CD45.1⁺ Cbfb^+/+:CD45.2⁺ Cbfb^2m/2m pairs (Fig. 4 B). In contrast, the epidermal MHC-II⁺ population was hardly exchanged in the CD45.1⁺ Cbfβ^+/+:CD45.2⁺ Cbfb^+/+ pair, and CD45.1- and CD45.2-expressing cells were detected only in the epidermis of CD45.1 and CD45.2 parabionts, respectively (Fig. 4 B). In sharp contrast, a significant number of CD45.1⁺ cells showing characteristics of mature LCs were detected in the epidermis of CD45.2⁺ Cbfb^2m/2m parabionts (Fig. 4 B), further supporting the contention that recruitment and differentiation from circulating Gr-1⁺ monocytes into mature LCs takes place in the Cbfb^2m/2m epidermis. However, we were still able to identify a significant population of host (CD45.2⁺) MHC-II⁺ cells that did not express any markers of maturation, such as EpCAM, CD24, and Langerin (Fig. 4 B).

Altogether, our observations indicated that Cbfβ2-deficient LC precursor-like cells retain properties of tissue residency, and homeostasis is maintained in an epidermal-specific manner, as was the case for mature LCs.

We next examined whether and to what extent embryonic-derived precursors survive throughout life and contribute to the maintenance of the MHC-II⁺EpCAM⁻ population in adult Cbfβ2-deficient mice. For this, we performed a fate-mapping experiment using a transgenic mouse line expressing the tamoxifen-inducible Mer-iCre fusion protein driven by the Csf1r promoter (Csf1r-iCre transgenic mice; Schulz et al., 2012). Injection of 4-hydroxytamoxifen (4-OHT) into pregnant animals at 8.5 dpc resulted in the specific marking of embryonic macrophages and their progeny with YFP expression from the Rosa26-STOP-YFP allele (Rosa26^YFP; Schulz et al., 2012). We crossed Cbfb^+/2m mice harboring both Rosa26^YFP and Csf1r-iCre transgenes, selected offspring according to the desired genotype, and examined them at 4 mo of age when both MHC-II⁺EpCAM⁻ and MHC-II⁺EpCAM⁺ populations should be present in the epidermis of Cbfb^2m/2m mice (Fig. 2). In double transgenic mice, 1% of EpCAM⁺ cells expressed YFP, a tagging efficiency similar to that reported previously (Schulz et al., 2012). In Cbfb^2m/2m offspring, a similar percentage of YFP⁺ cells was detected in MHC-II⁺EpCAM⁻ LC precursor-like cells, whereas no YFP-expressing cells were detected in epidermal MHC-II⁺EpCAM⁺ mature LCs (Fig. 4 C). These results strongly support that the MHC-II⁺EpCAM⁻ population is of embryonic origin and is likely to survive in the epidermis until adulthood.

We next examined whether MHC-II⁺EpCAM⁻ cells from adult Cbfb^2m/2m mice retain the developmental potential to differentiate into mature LCs. Because of the lack of established methods available to induce LC differentiation from epidermal precursors in vitro, we tested whether restoration of Cbfβ2 expression in vivo could induce LC differentiation by establishing another mouse strain in which Cbfβ2 expression could be induced from the Rosa26 locus after Cre-mediated excision of STOP signals (a Rosa26^Cbfb2 locus). The presence of internal ribosome entry site GFP sequences downstream of Cbfb2 cDNA in the Rosa26^Cbfb2 allele allowed us to identify cells expressing transgenic Cbfβ2 by tracing reporter GFP expression. The Rosa26^Cbfb2 mouse strain was crossed with another transgenic strain expressing the Mer-iCre fusion protein from the Rosa26 locus (a Rosa26^ERT2 locus) and was sequentially crossed with Cbfb^2m/2m mice. To obtain newborn mice harboring the Cbfb^2m/2m:Rosa26^{ERT2Cre/Cbfb2} genotype, we crossed Cbfb^+/2m:Rosa26^{ERT2Cre/Cbfb2} mice (Fig. 5 A).

We performed two sets of experiments, one with restored Cbfβ2 expression at birth representing “normal” development and one with restoration at a later stage in adult animals. First, we administered 4-OHT for 5 d from P0 and analyzed the epidermal content after 2 wks. GFP⁺ cells were detected from ∼10% of CD45⁺MHC-II⁺cells (Fig. 5 B) as well as from 0.4% of CD45⁻ cells (Fig. S2 A). In this experimental setting, there were no GFP⁺ hematopoietic cells detected in CD11b⁺ blood monocytes (Fig. S2 B), confirming the skin-specific induction of Cbfβ2 by topical 4-OHT administration. Importantly, all GFP⁺ cells in the CD45⁺MHC-II⁺ population became EpCAM⁺ cells (Fig. 5 B), indicating that the induction of Cbfβ2 when LC differentiation normally occurs is enough to restore LC development in Cbfb^2m/2m mice. Second, we topically administered 4-OHT to 4-wk-old Cbfb^2m/2m mice for 5 d. In contrast to administration at P0, transgenic Cbfβ2 expression at 4 wks, which was confirmed by reporter GFP expression in 10% of CD45⁺ cells, failed to induce EpCAM⁺ expression in CD45⁺MHC-II⁺ cells (Fig. 5 B). This result indicated that, at least in terms of EpCAM induction, epidermal resident Cbfb^2m/2m MHC-II⁺EpCAM⁻ cells lost their developmental ability to mature after 1 mo.

To gain insights into the molecular basis of the diminishing developmental potency of epidermal-residual LC precursors in Cbfb^2m/2m mice, we prepared epidermal CD45⁺CD3⁻MHC-II⁺ cells from P5 and 1.5-mo-old animals (referred hereafter to as postnatal and adult, respectively) and compared gene expression profiles between Cbfb^+/+ and Cbfb^2m/2m mice using barcode-based digital RNA sequences (Shiroguchi et al., 2012). We found more than 400 differentially expressed genes among four cell subsets (Fig. S3 A), and primary component analyses showed that these four subsets clustered separately (Fig. 6 A). According to PC1 values, postnatal cells were relatively close to each other. Some genes including Smad7, encoding inhibitory Smad, and Runx3 are commonly up-regulated in control LCs and adult Cbfb^2m/2m LC precursors (Fig. S3 B). In addition, comparing combinatorial gene expression changes between postnatal and adult Cbfb^2m/2m cells with regard to cytokine–receptor pairs showed that the expression of both Tgfbr1 and Tgfbr2 was elevated in adult Cbfb^2m/2m cells (Figs. 6 B and S3, C and D). In adult Cbfb^2m/2m cells, the expression of Tgfb1 was maintained at the same level as in postnatal cells (Fig. 6 B). In addition, adult Cbfb^2m/2m cells expressed higher levels of Lamtor2, which encodes a component of LAMTOR complexes that transmit TGFβ1 signals (Sparber et al., 2015), and Axl, which transduces signals that result in the preservation of a quiescent state (Bauer et al., 2012; Fig. S3 D). This suggested that the TGFβR1 pathway, which is required to maintain a quiescent state, is functionally enhanced in these cells. In contrast, the Bmp7-Bmpr1a transcript pair was reduced in adult Cbfb^2m/2m cells compared with postnatal Cbfb^2m/2m cells (Fig. 6 B). In addition, levels of the Ctnnb1 transcript were reduced in adult Cbfb^2m/2m cells (Fig. S3 D). This gene encodes β-catenin and is induced by TGFβ1 to promote LC differentiation synergistically with vitamin D receptor (VDR; Yasmin et al., 2013b). Interestingly, the expression of Vdr was also reduced in adult Cbfb^2m/2m cells (Fig. S3 D). These results indicated that postnatal and adult Cbfb^2m/2m cells have district gene expression signatures, particularly related to TGFβ superfamily signaling, with high and low expression of the Tgfb-Tgfbr1 and Bmp7-Bmpr1a axes, respectively, in adult cells.

To examine the function of Cbfβ2 in TGFβR signaling during LC differentiation, BM progenitors were cultured with TGFβ1 and GM-CSF, a cytokine combination that induces LC-like cell development (Chopin et al., 2013). After 3 d of culture, CD205⁺EpCAM⁺ LC-like cells were differentiated from Cbfb^+/+ BM progenitors. However, the differentiation of CD205⁺EpCAM⁺ cells from Cbfβ^2m/2m progenitors was markedly inefficient (Fig. 6 C). Furthermore, the differentiation of CD205⁺EpCAM⁺ cells was impaired by dorsomorphin, an inhibitor of BMP type I receptors including BMPR1A, but not by SB431542, an inhibitor of TGFβR1. These observations confirmed that signaling through BMPR1A is more important for LC differentiation.

Consistent with the retention of Cbfβ^2m/2m LC precursors in the adult epidermis, the expression of CD40, CD86, and CCR7, which is induced by loss of TGFβR1 signaling and involved in LC egress from the epidermis (Kel et al., 2010), was comparable between Cbfb^+/+ and Cbfb^2m/2m cells (Fig. 6 D). Altogether, these observations suggested that reduced expression of Bmpr1a diminishes the developmental potential of embryonic Cbfb^2m/2m LC precursors, whereas the migratory capacity required for epidermal egress is inhibited by enhanced TGFβR1 signaling (Fig. 6 E).

A previous study reported the absence of MHC-II⁺ cells in the Runx3-deficient epidermis (Fainaru et al., 2004). Because Runx3 must interact with Cbfβ to exert its function, a discrepancy in presence or absence of epidermal MHC-II⁺ cells between Runx3- and Cbfβ2-deficient mice could be explained by the functional redundancy of Cbfβ1. We therefore examined the effect of total ablation of Cbfβ function in DC-lineage cells using Cbfb^F/F:CD11c-Cre mice. Although the CD45⁺MHC-II⁺ population was significantly reduced in the epidermis of Cbfb^F/F:CD11c-Cre mice, one third of these cells were present as MHC-II⁺EpCAM⁻ LCs (Fig. 7 A). These MHC-II⁺EpCAM⁻ cells did not up-regulate CD86 or CCR7 expression (Fig. 7 B), indicating that the maintenance of a quiescent state by TGFβR1 signaling is likely independent of Cbfβ. This observation not only suggested that MHC-II⁺ LC precursors could be present in Runx3-deficient mice, but also indicated that Runx3/Cbfβ2 complexes have a unique role in transmitting selective signals through BMPR1A during stimulation by the TGFβ superfamily.

Our observations provide several novel insights into the underlying mechanisms through which the LC network is established and maintained. In this study, we first unraveled a unique role for Cbfβ2 in establishing the epidermal LC network. We showed that in Cbfβ2-deficient mice, embryonic-derived LC precursors can persist in the epidermis until adulthood through impaired selective TGFβ superfamily signaling, which is essential for LC differentiation. However, the conditional rescue of Cbfβ2 expression also revealed that such epidermal-resident Cbfβ2-deficient LC precursors do not maintain developmental potency.

It is widely accepted that inhibition of the TGFβ1/TGFβR1 pathway leads to the depletion of LCs in the adult epidermis. Initially, the simplest interpretation was that TGFβ1 is essential for LC differentiation, specifically to drive embryonic LC precursors to differentiate into the EpCAM⁺/Langerin⁺ stage (Borkowski et al., 1996; Kaplan et al., 2007; Chorro et al., 2009; Nagao et al., 2009). However, other studies subsequently revealed that signaling through TGFβR1 has another role in the maintenance of a quiescent state in LCs (Kel et al., 2010; Bobr et al., 2012), specifically, promoting their retention in the epidermis. It was proposed that these two pathways, LC differentiation and maintenance in a quiescent state, are likely to be regulated by different TGFβ superfamily receptor pairs. Whereas BMP7-BMPR1A plays a prominent role in LC differentiation, TGFβ1-TGFβR1 signaling is crucial for the inhibition of LC activation and the retention of these cells in the epidermis (Kel et al., 2010; Yasmin et al., 2013a). Here we found that the expression of Lamtor2 and Axl, which are involved in maintaining a quiescent state downstream of TGFβR1 signaling, was increased in skin-resident LC precursors of adult Cbfb^2m/2m mice. In contrast, expression of Bmpr1a, which is involved in LC differentiation, was reduced. Therefore, our results suggest that the selective inhibition of BMPR1A-dependent signaling, through the loss of Cbfβ2, inhibited LC differentiation when intact TGFβR1-mediated signals were present to preserve a quiescent state. Consequently, Cbfβ2-deficient LC precursors persisted in the epidermis.

EpCAM was previously shown to increase the motility of LCs in vivo (Gaiser et al., 2012). In this regard, it is possible that defective EpCAM expression in Cbfb^2m/2m LC precursors is involved in an additional mechanistic level of epidermal retention. However, given that TGFβR1-deficient LC precursors lacking EpCAM expression migrated out from the epidermis (Kel et al., 2010), it seems that TGFβR1 signaling rather than EpCAM expression drives LC retention in the epidermis. Although this model for distinct roles among TGFβ superfamily members predicts that BMP7- or BMPR1A-deficient mice should retain MHC-II⁺ LC precursors in the adult epidermis, similar to what was observed in this study using Cbfβ2-deficient mice, neonatal or embryonic lethality of these mice (Mishina et al., 1995; Zouvelou et al., 2009) limited such analysis.

Our results also revealed that Cbfβ1 and Cbfβ2 have distinct roles in TGFβR signaling. First, Cbfβ1 could not compensate for loss of Cbfβ2 function, which is necessary for LC differentiation. The presence of MHC-II⁺EpCAM⁻ LC precursors in Cbfb^f/f:CD11c-Cre mice, in which the function of Cbfβ is completely abrogated, suggests that such LC precursors could similarly be present in the epidermis of adult Runx3-deficient mice. These observations tempted us to speculate that Runx3/Cbfβ complexes are not essential for the transmission of signals that are essential for the maintenance of LCs in a quiescent state during TGFβR1 signaling, although the Cbfβ-independent function of Runx3 in this regulation was not formally excluded. Given that Runx proteins interact with Smad proteins (Ito and Miyazono, 2003), Smad family proteins potentially interact with Runx3/Cbfβ2 complexes to regulate LC differentiation during BMP7-BMPR1A signaling. Although Smad2, Smad3, and Smad4 are not required for LC differentiation in vivo (Xu et al., 2012; Li et al., 2016; Zhang et al., 2016), it has been reported that BMP7 induces phosphorylation of Smad1/5/8 proteins in vitro (Yasmin et al., 2013a). Thus, these Smad1/5/8 proteins might be involved in transmitting signaling from BMPR1A to induce LC differentiation. Alternatively, it is possible that Runx3/Cbfβ2 complexes regulate LC differentiation during BMP7 stimulation in a Smad-independent manner. Understanding the molecular basis of Cbfβ2-dependent LC differentiation will require the identification of molecules that interact specifically with Cbfβ2.

We showed that Cbfb^2m/2m LC precursors can reside in the epidermis but fail to retain maturation potential upon rescue by Cbfβ2 expression. Our transcriptome analysis showed that expression of the TGFβ1-TGFβR1 axis was higher in adult Cbfb^2m/2m LC precursors. Although it remains unclear whether negative feedback regulation could restrain this axis, it is possible that the loss of Cbfβ2 might interfere with this process, resulting in enhanced TGFβR1 expression. In contrast, the expression of Ctnnb1, which is also induced by TGFβ1 stimulation in vitro in human CD34⁺ hematopoietic progenitor cells (Yasmin et al., 2013b), was reduced in adult Cbfb^2m/2m LC precursors. Thus, responsiveness to TGFβ1 might differ among putative TGFβ target genes in Cbfβ2-deficient LC precursors. Alternatively, the induction of Ctnnb1 by TGFβ1 is mediated by receptors other than TGFβR1, such as BMPR1A. It is also possible that continuous exposure to TGFβ1 stimulation decreases responsiveness to TGFβ1 stimulation at certain gene loci such as Ctnnb1. As such, it was previously reported that forced expression of β-catenin in CD34⁺ hematopoietic progenitor cells accelerates LC differentiation, and this effect was enhanced by VDR activity (Yasmin et al., 2013b). Interestingly, Cbfb^2m/2m LC precursors in the adult epidermis exhibited reduced expression of Ctnnb1 and Vdr, suggesting that the loss of these molecules, combined with low BMPR1A expression, is involved in the diminished maturation potential of Cbfb^2m/2m LC precursors. It will be of interest to determine whether the down-regulation of these molecules arises from continuous TGFβR1 stimulation or whether Cbfβ2 plays a direct role in maintaining their expression. Nevertheless, during the postnatal period, the expression of Vdr and Bmpr1a was higher in Cbfb^2m/2m cells than in Cbfb^+/+ cells, indicating that Cbfβ2 is dispensable for the induction of these genes.

Results of this study and those of previous studies have established different roles for distinct members of the TGFβR superfamily in establishing and maintaining the primary LC network (Collin and Milne, 2016), inducing LC differentiation, and maintaining LC-lineage cells in a quiescent state. Regarding the physiological relevance of two TGFβ superfamily regulatory pathways, our results showed that loss of selective signaling that is involved in LC differentiation allows LC precursors to reside in the epidermis of Cbfb^2m/2m mice; however, these cells lose developmental potential with time. It is unclear whether changes in developmental potency are a common feature of embryonic LC precursors that reside in the epidermis beyond the postnatal period, because no other models showing the retention of such cells have been reported. Thus, the exact mechanisms resulting in diminished developmental potential remain elusive. However, it is conceivable that LC precursors that cannot differentiate into LCs are useless and therefore should be eliminated from the epidermis. In this regard, two pathways composed of different TGFβR superfamily members might function to survey responsiveness to the TGFβ superfamily. A pathway that mediates the epidermal egress of undifferentiated LC precursors, which could result from either insufficient levels of BMP7/TGFβ1 in the skin environment or intrinsic defects in integrating those environmental cues, might be beneficial to vacate the space needed for the recruitment of BM-derived monocytes to restore the functional LC network. Thus, surveying the responsiveness to the TGFβ superfamily might increase a probability for meditating replenishment of the LC network by other resources. Given that LC differentiation from BM precursors was shown to be independent of TGFβR signaling (Borkowski et al., 1997), which is consistent with emergence of LC cells in older Cbfb^2m/2m mice, the second adult LC developmental program would restore the LC network even after the primary LC network fails to differentiate from embryonic LC precursors. This hypothesis might explain why undifferentiated LC precursors are hardly detected in the adult epidermis. Contrary to TGFβR-independent LC differentiation from BM precursors in vivo, LC differentiation from BM cells in vitro depends on TGFβ/BMPR1A signaling. It is important to understand how BMPR1A signaling supports LC differentiation from BM cells in vitro.

Altogether, our results revealed an essential role for Cbfβ2 in the establishment of the primary LC network. Two major immune cell types, DETCs and LCs, are absent from the epidermis of Runx3- and Cbfβ2-deficient mice. Our comparative genomics approach showed that the splicing event that generates Cbfβ2 likely emerged as early as in bony fish (M. Tenno and I. Taniuchi, unpublished data), despite Runx3 being already present in lamprey (Nah et al., 2014). A recent study showed that LC-like cells are present in catfish (Kordon et al., 2016), prompting us to speculate on an evolutionarily conserved role for Cbfβ2 in LC differentiation. Interestingly, the interaction between Runx2 and VDR was proposed to be involved in shaping the genetic program of osteoblastogenesis (Marcellini et al., 2010). Further understanding of the Cbfβ2-dependent establishment of LC networks in several species will improve understanding of the evolution of the skin immune system.

In order to generate the Cbfb^1m and Cbfb^2m mutant allele, the splicing donor signal in exon 5 of murine Cbfb gene was mutated by homologous recombination in embryonic stem cells. To generate the Rosa26^Cbfb2 allele, cDNA encoding Cbfβ2 was amplified by PCR to add AscI sites at both ends, and this was ligated into an AscI-cleaved pCTV vector (15912l Addgene). C57/BL6 mice congenic for the CD45 locus were purchased from Sankyo Labo Service Coro. CSFIR-iCre transgenic mice (JR019098) and Rosa26^ERT2Cre mice were purchased from the Jackson Laboratory and Artemis Pharmaceuticals, respectively. Cx3cr1^gfp reporter mice, Rosa26^YFP reporter mice, and CD11c-Cre transgenic mice were a gift from S. Jung (Weizmann Institute of Science, Rehovot, Israel), F. Costantino (Columbia University, New York, NY), and B. Reizis (Columbia University), respectively. All animal experiments were performed in the animal facility at the RIKEN Center for Integrative Medical Sciences according to Institutional Animal Care Guidelines. Mice of different groups were cohoused and randomly assigned for analyses.

To prepare single-cell suspensions from epidermis, ears were split into dorsal and ventral halves. After removal of cartilage, ears were incubated with 0.5% trypsin and 1 mM EDTA for 1 h at 37°C. The epidermis was peeled from the dermis and was dissociated into single cells by mashing through a 70-µm cell strainer (BD Biosciences). The resulting single-cell suspensions were stained with the following antibodies purchased from BD Biosciences or eBiosciences: CD4 (RM4-5), CD8 (53-6.7), B220 (RA3-6B2), CD11b (M1/70), CD45 (30-F11), CD45.1 (A20), CD45.2 (104), CD3 (145-2C11), MHC-II (M5/114.15.2), CD207 (eBioL31), CD11c (HL3), EpCAM (G8.8), CD24 (M1/69), F4/80 (BM8), CCR7 (4B12), and CD86 (GL1). For intracellular staining, cells were permeabilized before staining with antibodies. Multicolor flow cytometry analysis was performed using a BD FACScanto II (BD Biosciences), and data were analyzed using FlowJo (Tree Star) software. Cell subsets were sorted using a FACSAria II (BD Biosciences).

Ears were split into dorsal and ventral halves. After removal of cartilage and incubation in PBS containing 20 mM EDTA for 1 h at 37°C, the epidermis was peeled from dermis and fixed with 2% PFA in PBS for 30 min on ice. Fixed epidermal sheets were sequentially incubated with PBS, 25% MtOH/PBS, 50% MtOH/PBS, 75% MtOH/PBS, and 100% MtOH for 15 min on ice at each incubation. The epidermis was incubated with 100% MtOH at −20°C overnight. The next day, the epidermis was sequentially incubated with 100% MtOH, 75% MtOH/PBS, 50% MtOH/PBS, 25% MtOH/PBS, and PBS for 15 min on ice at each incubation. The epidermis was first blocked with 5% skim milk, 2% BSA, and 0.1% Triton X-100 in PBS for 1 h on ice then stained with anti-CD3, anti–MHC-II, anti-EpCAM, and anti-Langerin antibodies in blocking solution for 2 h, followed by five washes with PBS/0.05% Tween on ice. After being mounted on MAS-coated slides (Matsunami Glass) using Fluoromount (Diagnostic BioSystems), the stained epidermis sheet was analyzed using TCS SP2 AOBS fluorescent microscopy (Leica).

RNAs were isolated using TRIzol (Thermo Fisher Scientific) according to the manufacturer’s instruction. cDNAs were synthesized from total RNAs with oligo dT primers using Superscript III reverse transcription (Invitrogen). Amplification was performed with SYBR Green Real-Time PCR Master Mix (Thermo Fisher Scientific). Primer sequences are listed in Table S1.

Surgical procedures to generate parabiotic pairs between CD45.2 Cbfb^+/+ or Cbfb^2m/2m and CD45.1 Cbfb^+/+ mice were performed as described previously (Rossi et al., 2005). After 7 mo, tissues of each mouse were dissected and analyzed by flow cytometry.

Total BM cells (5 × 10⁵ cells) from 8-wk-old CD45.1 Cbfb^+/+ mice were intravenously injected into sublethally irradiated (9.5 Gy) 6-wk-old CD45.2 Cbfb^+/+ or Cbfb^2m/2m mice. 7 mo after injection, mice were killed, and flow cytometry analysis was performed.

The marking of yolk sac–derived cells using Rosa26^YFP and CSF1R-iCre transgenic mouse strains was performed as previously described (Schulz et al., 2012). In brief, 4-OHT (2 mg/mouse; Sigma-Aldrich) and progesterone (4 mg/mouse; Sigma-Aldrich) were injected once at 8.5 dpc intraperitoneally into pregnant Cbfb^+/2m:Rosa26^+/YFP:CSF1R-iCre females, which were crossed with Cbfb^+/2m:Rosa26^+/YFP:CSF1R-iCre transgenic male mice. Epidermal tissues were prepared for flow cytometric analyses from offspring harboring the Cbfb^+/+:Rosa26^+/YFP:CSF1R-iCre or Cbfb^2m/2m:Rosa26^+/YFP:CSF1R-iCre genotype at 4 mo of age.

To induce Cbfβ2 expression from the Rosa26^Cbfb2 allele in Cbfb^2m/2m mice at different ages, we first set up intercrossing between Cbfb^+/2m:Rosa26^Cbfb2/ERT2 mice. Topical administration of 4-OHT (1 mg/ml in EtOH) on the ears of all newborn pups was initiated at P0 and applied for an additional 4 d. Genotyping was performed at 1 wk of age, and pups harboring Cbfb^2m/2m:Rosa26^Cbfb2/ERT2 were selected for further analyses at 2 wks of age. Cbfb^2m/2m:Rosa26^Cbfb2/ERT2 mice at 4 wks of age were treated similarly, receiving topical administration of 4-OHT on the ears for 5 d; animals were analyzed 2 wks after the first 4-OHT treatment.

100 CD45⁺CD3⁻MHC-II⁺ cells were sorted from epidermis of P5 and 1.5-mo-old Cbfb^+/+ and Cbfb^2m/2m mice into a tube using a FACSAria III (BD Biosciences). To normalize results using detection efficiency, the same amount of spike-in RNAs (a known number of External RNA Controls Consortium RNA molecules) was added into every tube, and the cells were lysed. After fragmentation of RNA molecules by temperature elevation, cDNAs were synthesized by reverse transcription using a reverse transcription primer in which the first common primer sequence molecular barcode, 14 T bases, and 1 V (A or G or C) base were tandemly arranged from the 5′ end. A second common primer sequence was also attached to the 3′ end of each generated cDNA molecule by template switching during reverse transcription. Subsequently, cDNAs were amplified using the primer that contains an Illumina adapter sequence, sample index, and the first common primer sequence with the other primer that contains another Illumina adapter sequence, sample index, and the second common primer sequence. Three independent libraries were generated from each subset with different sample indexes, and in total six libraries were sequenced together in a single MiSeq run (150 cycles; Illumina). Sequencing data were mapped against the mouse genome (mm10 assembly) using TopHat2 (Kim et al., 2013). This strategy, called digital RNA-seq (Shiroguchi et al., 2012), provides the mean number of mRNA molecules per cell for each gene and each indexed sample by counting the number of unique molecular barcodes instead of the number of sequenced cDNA molecules. Differential gene expression of digital RNA-seq was determined using DESeq2. To identify genes differentially expressed among groups, a moderated one-way ANOVA was performed. Genes were considered differentially expressed when they had false discovery rates of <0.01 and log2 fold-changes of >1 or <−1. To evaluate the combinatorial gene expression changes of cytokine–receptor pairs, we first picked up all cytokines and receptors according to the annotations in UniProt database. Cytokine–receptor interactions were reconstructed based on the protein–protein interaction network in the STRING database. All sequencing datasets have been deposited in the Genome Expression Omnibus database under accession no. GSE95487.

For the differentiation of LC-like cells from BM progenitors (Chopin et al., 2013), lineage (CD4, CD8, B220, MHC-II, CD11b, CD11c, NK1.1, Ter119, and Gr-1)-maker negative fractions were prepared from the BM cells of Cbfb^+/+ and Cbfb^2m/2m mice using the MACS system (Miltenyi Biotec). Cells (5 × 10⁵) were cultured in RPMI 1640 (Nacalai Tesque) supplemented with 10% FCS, 20 ng/ml GM-CSF (R&D Systems), and 10 ng/ml TGFβ1 (R&D Systems) for 3 d in a 24-well plate in the absence or presence of 5 µM dorsomorphin (041-3376; Wako Chemicals) or 5 µM SB431542 (Cay-13031; Cayman Chemical). For inhibition assays, BM progenitors were pretreated with dorsomorphin or SB431542 1 h before adding cytokines.

Fig. S1 shows increased Cbfb1 and Cbfb2 expression during LC differentiation. Fig. S2 shows conditional and skin-specific induction of Cbfb2 expression from the Rosa26^Cbfb2 allele. Fig. S3 shows a summary of the analysis of cytokine–receptor pairs and expression levels of selected genes. Table S1 shows sequence information for primers used for real-time quantitative PCR.

We thank M. Ikegaya for genotyping, T. Ishikura for embryonic stem cell aggregation, and K. Fukuhara for help in performing digital RNA sequencing.

This work was supported by Japan Society for the Promotion of Science KAKENHI grants 14F04214 and JP 19390118 (I. Taniuchi) and JP 23790552 (M. Tenno), the RIKEN Center for Integrative Medical Sciences YCI program (K. Shiroguchi), and Japan Science and Technology Agency/PRESTO grant JPMJPR15F3 (K. Shiroguchi).

The authors declare no competing financial interests.

Author contributions: M. Tenno performed phenotypic analyses. K. Shiroguchi, E. Kawakami, K. Koseki, K. Kryukov, and T. Imanishi performed digital RNA-seq and analyzed data. S. Muroi generated mouse strains. F. Ginhoux provided essential advice to design experiments and helped to write the manuscript. I. Taniuchi designed experiments and wrote the manuscript.