Ezh2 Controls Skin Tolerance through Distinct Mechanisms in Different Subsets of Skin Dendritic Cells.

Ezh2, a well-established epigenetic repressor, can down-regulate leukocyte inflammatory responses, but its role in cutaneous health remains elusive. Here we demonstrate that Ezh2 controls cutaneous tolerance by regulating Langerhans cell (LC) transmigration across the epidermal basement membrane directly via Talin1 methylation. Ezh2 deficiency impaired disassembly of adhesion structures in LCs, leading to their defective integrin-dependent emigration from the epidermis and failure in tolerance induction. Moreover, mobilization of Ezh2-deficient Langerin- dermal dendritic cells (dDCs) via high-dose treatment with a weak allergen restored tolerance, which is associated with an increased tolerogenic potential of Langerin- dDCs likely due to epigenetic de-repression of Aldh in the absence of Ezh2. Our data reveal novel roles for Ezh2 in governing LC- and dDC-mediated host protection against cutaneous allergen via distinct mechanisms.

Introduction

The skin is a complex organ that forms the primary barrier against the external environment. Keratinocytes, together with a diverse range of specialized immune cells, form a skin-associated lymphoid tissue that is critical for cutaneous immunosurveillance and host protection against repeated physical and pathogenic insults. Dysregulation of the skin immunity frequently results in various skin pathologies such as allergic contact dermatitis and psoriasis (Kaplan et al., 2005, Stratis et al., 2006).

Langerhans cells (LCs) are a group of professional antigen-presenting dendritic cells (DCs) that strategically populate the epidermis and actively survey the cornified layer for antigens via the extension and retraction of their dendrites (Merad et al., 2002). Despite considerable progress in our understanding of LC biology with the development of various LC ablation and transgenic mouse models, their role in the regulation of skin immunity remains controversial. LC migration to the skin-draining lymph nodes (sdLNs) is required not only for the initiation of adaptive immune responses (Igyarto et al., 2011) but also for the immune suppression during inflammation (Igyarto et al., 2009, Kaplan et al., 2005) and maintenance of steady-state tolerance to innocuous cutaneous antigens and commensal flora (Gomez de Aguero et al., 2012, van der Aar et al., 2013). Yet at the same time they have been demonstrated to be redundant to dermal DCs (dDCs) during skin immunization in different animal models (Fukunaga et al., 2008, Kissenpfennig et al., 2005). Such capacity of LCs to control the balance between immunogenicity and tolerance depends on multiple layers of gene regulation that permit dynamic changes in cellular functions including phagocytosis, migration, and cytokine production (Hacker et al., 2003, Sere et al., 2012). The most extensively studied of these LC regulatory mechanisms is transcriptional control of gene expression, which can potently modulate LC-induced immune responses (Medzhitov and Horng, 2009), but other factors influencing LC-mediated protection against skin injury and inflammation remain only poorly defined.

It is now well-established that leukocyte gene expression can be modified by epigenetic regulatory proteins including enhancer of Zeste homolog 2 (Ezh2), which represses gene expression by catalyzing the addition of methyl groups to histone 3 at lysine 27 (H3K27) (Cao et al., 2002, Czermin et al., 2002). Chromatin remodeling via histone modifications exerts major effects on leukocyte gene expression patterns (Jin et al., 2016, LaMere et al., 2016), but the ability of proteins such as Ezh2 to further modify immune cell function via non-epigenetic mechanisms is less well understood. Given its well-defined role as a transcriptional silencer, Ezh2 has previously been implicated in the epigenetic regulation of many immune genes including those associated with host cell defense such as interleukin (IL)-4 and IL-13 (Koyanagi et al., 2005). However, we recently reported an unexpected role for Ezh2 in the post-translational methylation of Talin1, which disrupts Talin1 binding to F-actin and thereby regulates dynamic changes in DC adhesion and migration programs upon encounter with integrin ligands (Gunawan et al., 2015, Loh and Su, 2016).

As Ezh2 has been implicated in the regulation of various immune cell functions including DCs (Gunawan et al., 2015, Su et al., 2003, Su et al., 2005), and its expression is frequently reduced in patients with cutaneous allergies (De Benedetto et al., 2011), we hypothesized that Ezh2 plays a major role in the control of skin dendritic-cell-mediated protection against skin inflammation. Here we report that Ezh2 critically regulates LC transmigration across the basement membrane via a non-epigenetic mechanism that is essential for the activation of regulatory T cells (Tregs) and host protection against hapten-induced contact hypersensitivity (CHS). Moreover, we identify a subset of skin-resident Ezh2-deficient Langerin– dDCs with enhanced aldehyde dehydrogenase (ALDH) activity, which can substitute for LCs during tolerance induction via high-dose treatment with a weak hapten. The increased tolerogenicity of Langerin– dDCs was likely due to epigenetic de-repression of Aldh in the absence of Ezh2. Collectively, our data reveal novel roles for Ezh2 in governing LC-migration- and DC-tolerogenicity-mediated host protection against cutaneous allergy.

Results

Ezh2 Controls LC-Mediated Tolerance Induction

Ezh2 expression is frequently reduced in patients with cutaneous allergies (De Benedetto et al., 2011). To determine if Ezh2 expression in skin DCs plays a protective role against cutaneous allergies, we induced CHS, which mimics allergic contact dermatitis in humans, in control mice (Ezh2) for comparison with animals in which the Ezh2 gene was deleted only in CD11c+ cells (CD11c-cre; Ezh2). However, when control animals and CD11c-cre; Ezh2 mice were painted with two consecutive doses of strong hapten 2,4- dinitrofluorobenzene (DNFB) to induce CHS, mice with Ezh2-deficient DCs did not exhibit greater extent of ear swelling than control mice (Figure 1A). We next investigated whether Ezh2 expression in DCs plays a role in the induction of cutaneous tolerance. Here, the mice were tolerized with the weak hapten 2,4-dinitrothiocyanobenzene (DNTB) followed by CHS induction using DNFB (Figure 1B). We observed that tolerization with DNTB was able to suppress the development of DNFB-induced ear swelling in control mice (Figure 1C). In contrast, CD11c-cre; Ezh2 mice developed a robust CHS response that peaked at day 2–3 in spite of prior DNTB treatment, implying an important function of Ezh2-expressing DCs in the induction of cutaneous tolerance (Figure 1C). Consistent with the enhanced ear swelling, analysis of whole-mount ear tissues revealed an accumulation of enlarged DC clusters (Figure 1D), which are indicative of DC activation and enhanced antigen-presenting capacity (Natsuaki et al., 2014). Accordingly, the sdLNs from these animals contained increased numbers of interferon-γ-producing CD8+ T cells with a CD44+ CD62Llo effector or CD44+ CD62L+ central memory phenotype (Figures S1A and S1B). Although we were unable to detect any gross difference in the total numbers of Tregs present in sdLNs between control and CD11c-cre; Ezh2 mice (Figure S1C), a significantly lower proportion of activated Foxp3+ Tregs was found in the sdLNs of DNTB-painted CD11c-cre; Ezh2 mice, as determined by their surface expression of co-stimulatory molecule ICOS (Figures 1E and 1F).

Figure 1

Ezh2-Expressing LCs Induce Cutaneous Tolerance

(A) Ezh2 (Ezh2f/f) or CD11c-cre; Ezh2 (Ezh2Δ/Δ) mice were sensitized with 0.5% DNFB on shaved back skin before ear challenge with 0.1% DNFB 5 days later. Ear swelling was determined daily over a period of 5 days post-challenge (mean ± SEM). Data shown are from 2 independent experiments (n = 8).

(B and C) (B) Mice were tolerized with 0.1% DNTB on shaved abdomen skin before treated as described in (A). (C) Ear swelling response was determined daily over a period of 5 days post-challenge (mean ± SEM). Data shown are from 3 independent experiments (n = 9). From left to right *p = 0.01; **p = 0.0002, 0.0005, 0.0006, 0.0004.

(D) Ezh2; CD11c-YFP and CD11c-cre; Ezh2; CD11c-YFP mice were tolerized as described in (B). CD11c-YFP+ dermal DC clustering was visualized in whole-mount ear tissues harvested from untreated steady-state mice or from mice during the elicitation phase of CHS (48 hr after ear challenge with DNFB). Data shown are representative of 3 independent experiments (n = 3). Scale bars, 50 μm.

(E) Mice were tolerized with 0.1% DNTB on shaved abdomen skin, and Treg activation status in sdLNs was determined by flow cytometry on day 5. Dot plots were pre-gated on singlet, live cells, and CD3ɛ+ CD4+ Foxp3+ Tregs.

(F) Scatterplot indicates absolute number of ICOS+ Tregs in sdLNs of untreated or tolerized mice. Data shown are from 3 independent experiments (n ≥ 5) and were analyzed by two-tailed Student's t test (from left to right; **p = 0.003, 0.01).

(G and H) BM chimeric mice were treated as described in (B), and ear swelling was assessed (mean ± SEM). Ezh2 and CD11c-cre; Ezh2 mice are CD45.2. Data were analyzed by two-tailed Student's t test (n > 3) (from left to right *p = 0.02, 0.04, 0.05).

See also Figures S1 and S2.

The skin is a complex tissue consisting of multiple phenotypically and functionally heterogeneous CD11c+ DC subsets. To determine which DC subsets contribute to the failed induction of cutaneous tolerance in the CD11c-cre; Ezh2 mice, we took advantage of the radio-resistant feature of LCs and performed a series of reciprocal bone marrow (BM) transplantation experiments to selectively delete Ezh2 in either LCs or dDCs. The complete depletion of recipient origin leukocytes (CD45.2) and the efficient reconstitution of recipient mice with donor cells (CD45.1) were controlled by flow cytometry staining (Figure S1D). Using the same tolerization and sensitization regimen as employed in the previous experiments, we observed an increased ear swelling in BM chimeric mice that lacked Ezh2 only in epidermal LCs (Figure 1G). In contrast, ear inflammation was significantly reduced in BM chimeras whose skin was populated by wild-type LCs and Ezh2-deficient dDCs (Figure 1H). These results suggest that expression of Ezh2 in LCs is required for the induction of cutaneous tolerance.

Ezh2 Is Required for the Migration of Epidermal LCs

As Ezh2 inhibitor-treated leukocytes have been reported to display inflammatory gene signatures (Kruidenier et al., 2012), failure to induce tolerance in CD11c-cre; Ezh2 mice could be the consequence of an enhanced inflammatory response by the Ezh2-deficient LCs. To determine if Ezh2 expression in LCs contributes to tolerance induction through classical epigenetic-dependent mechanism, we analyzed global gene expression using RNA sequencing. Surprisingly, control and Ezh2-deficient LCs show similar gene expression profiles (Figure S2A), and in agreement, the global level of tri-methylated H3K27 remained unchanged despite Ezh2 deletion (Figure S2B). This is consistent with our previous report that gene expression profiles are not significantly altered in Ezh2-deficient BM-derived DCs (Gunawan et al., 2015). Hence, it is likely that Ezh2 in LCs promotes tolerance via an epigenetic-independent mechanism.

To further investigate how Ezh2 in LCs mediates cutaneous tolerance, we analyzed the epidermal sheets by immunofluorescence staining. Our data revealed that both control and CD11c-cre; Ezh2 mice retained a regular network of major histocompatibility complex (MHC) II+ cells with typical LC morphology in the epidermis (Figure 2A). However, when we quantified resident LC numbers, we detected an increased density of Ezh2-deficient LCs in the untreated epidermis (Figure 2A, left, steady state). As adult LCs are derived from embryonic precursors that take residence in fetal skin in utero and undergo a proliferative burst shortly after birth (Hoeffel et al., 2012), we next assessed whether steady-state accumulation of epidermal LCs in CD11c-cre; Ezh2 mice could be due to differential proliferation of Ezh2-deficient CD11c+ precursors. However, when we evaluated LC density in 2-week-old mice, we observed no significant difference between control animals and CD11c-cre; Ezh2 mice. Instead, Ezh2-deficient LCs accumulated gradually in the epidermis with increasing age (Figures S3A and S3B), whereas the numbers of proliferating and apoptotic LCs in these mice were comparable to that in the control mice (Figures S3C–S3H), suggesting that Ezh2 expression in LCs could be critical for their steady-state emigration from the epidermis. It is likely that Ezh2-deficient LCs fail to emigrate from the epidermis upon DNTB application and thereby contribute to defective tolerance induction. Indeed, whereas control LCs migrated away from the epidermis in response to DNTB painting, Ezh2-deficient LCs were unable to exit the skin when exposed to the same treatment (Figure 2A, right, 0.1% DNTB). These data suggest that Ezh2 is likely to be required for the induction of cutaneous tolerance by regulating LC migration away from the epidermis.

Figure 2

Ezh2 Regulates LC Migration from Epidermis to Skin-Draining Lymph Nodes

(A) Immunofluorescence analysis of epidermal sheets isolated from mouse ears that were treated or not with 0.1% DNTB 18 hr before excision. LCs were identified by immunostaining with anti-MHC II antibody. Images shown are representative of 2 independent experiments (n = 3, total ≥11 fields per group). Scale bars, 50 μm. Scatterplot indicates LC numbers/mm2. Data shown are from 2 independent experiments (n = 3, ≥11 fields per group) and were analyzed by two-tailed Student's t test (from left to right ***p = 1.2 × 10−8, 2.5 × 10−6, 5.2 × 10−9).

(B) Immunofluorescence analysis of epidermal sheets isolated from untreated mice (left), cultured for 18 hr in medium (center), or obtained from oxazolone-treated mice 18 hr after application (right). LCs were identified by immunostaining with anti-MHC II antibody. Images shown are representative of 3 independent experiments. Scale bars, 50 μm. Scatterplot indicates LC numbers/mm2. Data shown are from 3 independent experiments (n = 3, total 5 fields per group) and were analyzed by two-tailed Student's t test (from left to right *p = 0.01; **p = 0.0007, 0.0003, 0.0009; ***p = 5 × 10−5).

(C) Flow cytometric analyses of LCs in steady-state epidermis and dermis. Contour plots for epidermis are pre-gated on singlet, live cells. Contour plots for dermis were pre-gated on singlet, live, CD45+ cells. CD11c+ MHC II+ DCs were further analyzed for CD103- EpCAM+ LCs. Data shown are representative of 7 independent experiments. Bar graphs indicate fold-change in percentage of LCs out of total epidermal and dermal cells (percentage of LCs in Ezh2 mice set as 1). Data shown are from 7 independent experiments (n = 7) and were analyzed by two-tailed Student's t test (*p = 0.01; **p = 0.0003).

(D) 0.5% oxazolone was painted onto mouse ears, and sdLNs were harvested 48 hr later. Dot plots were pre-gated on singlet, live cells. CD11c+ MHC IIhi migratory DCs were further analyzed for LCs (CD103− EpCAM+) and dermal CD103− and CD103+ DCs (CD103−−/+ EpCAM−/lo). Data shown are representative of 3 independent experiments. Scatterplot indicates the absolute numbers of migratory LCs and dermal CD103− and CD103+ DCs reaching sdLNs. Data are pooled from 3 independent experiments (n ≥ 6). From left to right *p = 0.01, 0.04, 0.03; **p = 0.001, 0.001; ***p = 9 × 10−7.

See also Figures S3 and S4.

We further verify the importance of Ezh2 in regulating LC migration using skin explant culture and oxazolone skin sensitization models. Similarly, whereas control LCs efficiently migrated away from the skin explants or oxazolone-painted skin, Ezh2-deficient LCs persisted in the epidermis despite exposure to these potent activating stimuli (Figure 2B). Flow cytometry analysis also revealed a reduction in migratory LCs (MHC IIhi CD11c+ EpCAM+ CD103– cells) in the dermal layers (Figure 2C) and sdLNs of untreated CD11c-cre; Ezh2 mice (Figure 2D), indicating a defect in their steady-state migration. Moreover, even though oxazolone sensitization induced a robust migration of control LCs to sdLNs within 48 hr of administration, this treatment failed to increase LC numbers in sdLNs of CD11c-cre; Ezh2 mice (Figure 2D). In contrast to LCs, migration of dDCs (MHC IIhi CD11c+ EpCAM– CD103−/+ cells) to sdLNs was not altered by Ezh2 deficiency either under steady-state conditions or following oxazolone sensitization (Figure 2D). In addition, we found a marked up-regulation of Ezh2 mRNA and protein expression in migratory LCs isolated from sdLNs compared with epidermis-resident LCs (Figures S3I and S3J), further supporting a role of Ezh2 in controlling LC localization and function. Together, these data strongly indicate that Ezh2 is required for LC migration away from the epidermis under both steady-state and immunostimulatory conditions.

To determine whether our findings are relevant in the human context, we analyzed LC migration from human skin treated with GSK126, a small molecule inhibitor of Ezh2 methyltransferase activity (McCabe et al., 2012). Indeed, co-culture of human skin explants with GSK126 resulted in a decline in LC migration, in which we observed a decrease in emigrated LCs present in the tissue culture medium (Figures 3A–3C) and a concomitant increase in LCs remaining in the epidermal sheet (Figure 3D). Hence, in agreement with our mouse model, our studies using human skin explants suggest that Ezh2's methyltransferase activity is also required for human LC migration away from the epidermis.

Figure 3

Ezh2 Regulates Migration of Human LCs

(A) Human skin explants were cultured overnight in medium containing DMSO (control) or increasing concentrations of GSK126. Cells were retrieved from the culture medium, and LCs were identified by flow cytometry. Graph indicates numbers of LCs retrieved from the culture medium (n > 3).

(B) Skin explants from 2 independent human samples were cultured for 2 days in medium containing DMSO (control) or 100 μM of GSK126. Cells were retrieved from the medium following 1 or 2 days of culture, and LCs were identified by flow cytometry. Graph indicates numbers of LCs retrieved from the culture medium (n = 2).

(C) Cells retrieved from the medium in (B) were analyzed by flow cytometry. Dot plots were pre-gated on CD45+ cells. CD11c+ HLA-DR+ DCs were further analyzed for CD1a expression to define LCs.

(D) Epidermal sheets were isolated from human skin explants in (B) and stained with anti-HLA-DR (green) and anti-Langerin (red). Scale bars, 100 μm.

See also Figure S4.

Ezh2 Regulates LC Transmigration across the Basement Membrane

In order for LCs to reach the dermis and sdLNs, they must disengage from their epidermal niche before crossing the underlying basement membrane, which is primarily composed of a dense network of type IV collagen and laminin (Yurchenco, 2011). It is possible that trapping of Ezh2-deficient LCs in the epidermis is due to dysregulated expression of adhesion/migration-related molecules that anchor LCs to the surrounding keratinocytes. We therefore examined LC expression of molecules that have been reported to govern emigration from the epidermis, namely, E-cadherin, integrin α6, and CXCR4 (Price et al., 1997, Tang et al., 1993, Villablanca and Mora, 2008). In particular, E-cadherin was previously identified as a target gene of Ezh2-mediated epigenetic silencing in various cancer cells (Cao et al., 2008). However, the expression of these proteins in LCs during steady state and upon activation were not affected by Ezh2 deficiency (Figures S4A–S4C), which is in accordance with the comparable gene expression profiles of control and Ezh2-deficient LCs (Figure S2A). Lack of major changes in overall gene expression profiles is perhaps due to the compensatory effects of alternative histone methyltransferases. Indeed, the expression of a closely related histone methyltransferase—Ezh1, which shares some target genes with Ezh2 (Margueron et al., 2008, Shen et al., 2008), was increased in Ezh2-deficient BM-derived DCs (Figure S4D). Together, these results suggest that impaired migration of Ezh2-deficient LCs is not due to dysregulated expression of molecules involved in cell adhesion/migration.

We next investigated whether basement membrane represents a physical barrier that blocks transmigration of Ezh2-deficient LCs into the dermis. We first subjected excised skin samples to dispase digestion and cultured the isolated epidermis to facilitate LC activation in vitro. Dispase can disrupt the integrity of the basement membrane via specific cleavage of type IV collagen in the lamina densa, while leaving the epidermis intact (Stenn et al., 1989). In the absence of an intact basement membrane, activated Ezh2-deficient LCs were able to migrate away from epidermal sheet explants (Figures 4A and S4E). The observed reduction in epidermal LCs in this setting was not caused by increased cell death, as we recovered similar percentages of total viable LCs from both control and Ezh2-deficient tissue culture medium in these experiments (Figure S4F).

Figure 4

Ezh2 Regulates LC Transmigration across Basement Membrane

(A) Epidermal sheets were isolated by dispase digestion, and then cultured for 4 hr before staining with anti-MHC II antibody. Images shown are representative of 2 independent experiments (n = 3, total ≥5 fields per group). Scale bars, 50 μm. Scatterplot indicates LC numbers/mm2. Data shown are from 2 independent experiments (n = 3, total ≥5 fields per group) and were analyzed by two-tailed Student's t test (from left to right **p = 0.0007, 0.0002, 0.0002; ***p = 4 × 10−6).

(B) Immunofluorescence analysis of epidermal sheets isolated from untreated footpads. Images shown are representative of 2 independent experiments (n = 3, total ≥11 fields per group). Scale bars, 50 μm. Scatterplot indicates LC numbers/mm2. Data shown are from 2 independent experiments (n = 3, total≥11 fields per group) and were analyzed by two-tailed Student's t test (***p = 6 × 10−6).

(C and D) 0.5% oxazolone was painted onto footpad skin before harvesting 18 hr later. Untreated and oxazolone-painted footpad skin sections were stained with anti-laminin (red) and DAPI (blue). Endogenous CD11c-YFP+ LCs are visualized in green. Images shown are representative of 3 independent experiments (n = 3). Scale bars, 100 μm. Boxes outlined in (C) are enlarged in (D); scale bars, 10 μm in (D).

(E) Bar graph quantification of images in (C), showing distance of epidermal LCs (green) from the basement membrane (red) and LC diameter during steady state and after oxazolone administration (mean ± SEM of more than 200 cells per group). Data were analyzed by two-tailed Student's t test (from left to right ***p = 1 × 10−41, 1 × 10−11, 8 × 10−20, 6.5 × 10−51, 1.4 × 10−10, 4.5 × 10−5, 2.8 × 10−21).

See also Figure S4.

To further assess the importance of Ezh2 in regulating LC transmigration in vivo, we examined mouse skin by sectioning to visualize the longitudinal distribution of LCs throughout the epidermis. For these analyses, we used footpad rather than ear skin to obtain better resolution between the epidermal layer and basement membrane. Similar to our previous observations in ear skin, we detected a significant steady-state accumulation of epidermal LCs in footpad from CD11c-cre; Ezh2 mice (Figure 4B). Transverse sections of the untreated footpad revealed that the majority of control LCs displayed rounded morphology and were evenly distributed among the keratinocytes ∼8 μm above the basement membrane (Figures 4C–4E). Upon contact sensitization with oxazolone, these LCs became activated and were observed to migrate toward the basement membrane (Figures 4C–4E) and occasionally extend dendrites and/or cell body through into the dermis (Figure 4D). In contrast, Ezh2-deficient LCs were frequently found to line up against the basement membrane with atypical flattened morphology, with increased cell spreading upon activation with oxazolone (Figures 4C–4E). This apparent defect in transmigration of Ezh2-deficient LCs across the basement membrane could not be attributed to decreased matrix metalloproteinase (MMP) activity (Figures S4G and S4H) or increased expression of tissue inhibitors of MMP 2 and 3 (Figure S4I), which are reported target genes of Ezh2 in cancer cells (Shin and Kim, 2012, Xu et al., 2013). Taken together, these data reveal that Ezh2-deficient LCs are unable to transmigrate across the basement membrane to exit the skin.

Ezh2 Controls Adhesion Dynamics and Migration of LCs

We have previously reported that cytosolic Ezh2 acts on the extranuclear substrate Talin1 by catalyzing the addition of three methyl groups at Lys 2454 to disrupt binding to F-actin (Gunawan et al., 2015, Venkatesan et al., 2018). As Talin1 is a critical component of the focal adhesion complex (Franco et al., 2004), we hypothesized that Ezh2-mediated methylation of Talin1 contributes to the transmigration of LCs across the basement membrane. To test this possibility, we next isolated LCs for ex vivo assessment of their ability to attach to a laminin-coated surface that mimics the basement membrane in vivo. Immunostaining revealed that the majority of control LCs did not form notable adhesion structures in these assays, whereas Ezh2-deficient LCs frequently formed numerous and enlarged adhesion structures upon attachment to laminin (Figures 5A–5C). It is likely that Ezh2 regulates the turnover of adhesion structures in LCs via methylation of Talin1, similar to our previous observations in BM-derived DCs (Gunawan et al., 2015).

Figure 5

Ezh2 Regulates Adhesion Characteristics of Skin LCs

(A) LCs attached on laminin-coated slides were immunostained with DAPI (blue) and anti-Talin antibody (green). Images are representative of 4 independent experiments (n = 4). Scale bars, 10 μm.

(B) Scatterplot indicates distribution of adhesion structures in LCs. Data shown are from 4 independent experiments (n = 4) and were analyzed by two-tailed Student's t test (from left to right ***p = 5 × 10−7, 1 × 10−5, 3 × 10−5).

(C) Cumulative frequencies and numbers of adhesion structures as in (B). Data shown are from 4 independent experiments and were analyzed by Kolmogorov-Smirnov test (p = 1.4 × 10−9).

(D) Bar graph indicates surface area of LCs from (A) (mean ± SD). Data shown are from 4 independent experiments (n = 4) and were analyzed by two-tailed Student's t test (***p = 2 × 10−12).

(E and F) (E) Migration of CD11c-YFP+ LCs (green/outlined) on laminin-coated slides was tracked by time-lapse imaging. (F) Scatterplot indicates total distance moved by LCs in (E). Data shown are from 4 independent experiments (n = 4, 42 control and 35 Ezh2-deficient cells pooled from 4 experiments were scored). (**p = 0.002).

(G and H) Mouse ears were tape-stripped and infected with retroviral vector carrying different GFP-tagged Talin1 mutants (K2454, K2454Q, K2454F). 0.5% oxazolone was painted onto untransduced (UT) or retrovirus-infected mouse ears, and epidermis and sdLNs were harvested 18 and 48 hr later respectively. (G) Scatterplot indicates epidermal LC numbers/mm2. Data shown are from 3 independent experiments (n = 3, ≥7 fields per group) and were analyzed by two-tailed Student's t test (from left to right **p = 0.001, 0.009; ***p = 2.5 × 10−5, 7.8 ×10−5). (H) Bar graph indicates absolute number of LCs reaching sdLNs (mean ± SEM). Data shown are from 2 independent experiments (n = 3) and were analyzed by two-tailed Student's t test (from left to right *p = 0.03, 0.05, 0.02; **p = 0.009).

See also Figure S5.

Moreover, consistent with our findings in longitudinal skin sections, Ezh2-deficient LCs tended to exhibit excessive spreading upon interaction with laminin, which gave rise to an increased surface area ∼2-fold greater than that of control LCs (Figures 5A and 5D). Consequently, when we evaluated the migratory capacity of LCs on a laminin-coated surface via time-lapse imaging, we observed that Ezh2-deficient LCs appeared tightly adhered and were largely immobile, whereas control cells were highly migratory under the same conditions (Figures 5E and 5F). We next assessed whether this migration defect in Ezh2-deficient LCs was due to an inability to methylate Talin1 at Lys 2454. To do this in vivo, we analyzed the adhesion dynamics and migratory capacity of LCs in mouse skin infected with retrovirus to express various Talin1 mutants (Figure S5A). Substitution of K2454 with a glutamine (K2454Q) residue that cannot undergo methylation resulted in impaired migration of control LCs away from the epidermis to sdLNs in response to oxazolone treatment similar to the Ezh2-deficient LCs (Figures 5G, 5H, and S5B). In contrast, replacement of K2454 with the methyl-mimicking residue phenylalanine (K2454F) restored the migratory capacity of Ezh2-deficient LCs (Figures 5G, 5H, and S5B). Consequently, Ezh2-deficient LCs expressing Talin1-K2454F were now able to reach the sdLNs at a comparable rate as control cells, whereas cells expressing wild-type Talin1 or Talin1-K2454Q were trapped in the epidermis (Figures 5G, 5H, and S5B). These data were consistent with our earlier in vitro observation that Talin1 methylation promotes turnover of adhesion structures and enhances cell migration, whereas unmethylatable Talin1 renders cells immobile (Gunawan et al., 2015). Taken together, these findings suggest that Ezh2 regulates LC transmigration across basement membrane via a non-epigenetic mechanism involving the methylation of Talin1 and subsequently affects the turnover of adhesion structures.

Mobilization of Ezh2-Deficient Dermal DCs with Enhanced Tolerogenic Activity Restores Cutaneous Tolerance

Our findings reported above indicate that Ezh2 regulates LC transmigration across basement membrane during induction of cutaneous tolerance. As other dDC subsets have also been implicated in tolerance induction, and our study revealed that migration of dDCs to sdLNs was unaffected by Ezh2 deficiency upon contact sensitization with oxazolone (Figure 2D), we were intrigued by the failed cutaneous tolerance induction observed in CD11c-cre; Ezh2 mice (Figure 1C), in which only the migration of LCs was impaired. Thus we investigated the migratory DC populations in sdLNs and observed that LCs was the only population to be mobilized to the sdLNs upon DNTB application in our experimental setting (Figures 6A and 6B). As high dosage of haptens has been shown to disseminate into the dermis enabling mobilization of dDCs (Noordegraaf et al., 2010), we treated control and CD11c-cre; Ezh2 mice with a 10-fold higher dose of DNTB. Interestingly, at high-dose DNTB (1%), we were able to restore tolerance in CD11c-cre; Ezh2 mice as indicated by reduced ear swelling (Figure 6C) and attenuated formation of dDC clusters (Figure 6D). An expansion of activated ICOS+ Tregs was also observed in the sdLNs of CD11c-cre; Ezh2 mice, and this number was comparable to that observed in the control mice treated with low-dose DNTB (0.1%) (Figure 6E). This effect was associated with an increased number of migratory Langerin– dDCs (MHC IIhi CD11c+ CD103– CD11b+), whereas the number of Langerin+ dDCs remained unchanged in both control and CD11c-cre; Ezh2 mice (Figure 6F). Moreover, upon high-dose DNTB treatment, Ezh2-deficient LCs were still trapped in the epidermis and the number of migratory LCs in sdLNs remained low (Figure 6F). These results suggest that upon treatment with a high dosage of haptens, Ezh2-deficient Langerin– dDCs are able to compensate for the reduced LC numbers in sdLNs of CD11c-cre; Ezh2 mice to promote tolerance.

Figure 6

Dermal DCs Exposed to High-Dose DNTB Restore Tolerance in CD11c-cre; Ezh2 Mice

(A) Flow cytometric analyses of migratory DCs in sdLNs from mice that were tolerized with or without 0.1% DNTB 48 hr prior. Dot plots were pre-gated on singlet, live cells. CD11c+ MHC IIhi migratory DCs were further analyzed for Langerin+ CD103− LCs, Langerin+ CD103+ dermal DCs, and Langerin– dermal DCs. Data shown are representative of 4 independent experiments (n ≥ 5).

(B) Scatterplot indicates absolute numbers of migratory DCs recovered from sdLNs of untreated and DNTB-painted mice. Data shown are from 4 independent experiments (n ≥ 5) and were analyzed by two-tailed Student's t test (from left to right *p = 0.05, 0.04; **p = 0.007).

(C) Mice were tolerized with 1% DNTB on shaved abdomen skin before treated as described in Figure 1A. Ear swelling response was determined daily over a period of 5 days (mean ± SEM). Data shown are from 2 independent experiments (n ≥ 8).

(D) CD11c-YFP+ dermal DC clustering was visualized in whole-mount ear tissues 48 hr after DNFB challenge. Data shown are representative of 3 independent experiments (n = 3). Scale bars, 50 μm.

(E) Scatterplot indicates absolute numbers of ICOS+ Tregs in sdLNs from untreated mice or mice tolerized with 0.1% or 1% DNTB. Data shown are from 3 independent experiments (n ≥ 5) and were analyzed by two-tailed Student's t test (from left to right *p = 0.02; **p = 0.003, 0.01, 0.0002; ***p = 6.1 × 10−5).

(F) Bar graphs indicate absolute numbers of migratory DCs reaching sdLNs of untreated mice or mice treated with 0.1% or 1% DNTB (mean ± SEM). Data shown are from 4 independent experiments (n ≥ 5) and were analyzed by two-tailed Student's t test (from left to right *p = 0.02, 0.05, 0.05, 0.04, 0.04; **p = 0.01, 0.007).

Even though Langerin– dDCs have been reported to harbor ALDH activity required for Treg differentiation (Guilliams et al., 2010), they are unable to promote DNTB-mediated cutaneous tolerance in the absence of LCs (Gomez de Aguero et al., 2012). As increased numbers of migratory Ezh2-deficient Langerin– dDCs in sdLNs were correlated with restored tolerance in the absence of inflammation-induced migratory LCs (Figures 6C, 6E, and 6F), Ezh2-deficient Langerin– dDCs are likely to exhibit higher tolerogenic capacity than control counterparts. In fact, even though the overall H3K27me3 levels were comparable between control and Ezh2-deficient DCs and our previous microarray analysis did not show a major difference in gene expression patterns, Aldh2 is among a handful of up-regulated genes in Ezh2-deficient DCs (Gunawan et al., 2015). Therefore, we further investigated ALDH activity and found that a higher percentage of Ezh2-deficient CD11b+ dDCs (Langerin– dDC-equivalent cell population) exhibited ALDH activity than control cells (Figures 7A and 7B). A similar increase was observed in Ezh2-deficient BM-derived DCs after lipopolysaccharide stimulation (Figures 7C and 7D, mature). Treatment of BMDCs with diethylaminobenzaldehyde, an inhibitor of ALDH activity, was included to determine the specificity of AldeFluor staining (Figure 7C). Moreover, the expressions of ALDH protein and mRNAs were also higher in Ezh2-deficient BMDCs than in control cells (Figures 7E and 7F). Next, to determine whether Ezh2 regulates ALDH expression and activity via the classical epigenetic-dependent mechanism, we performed chromatin immunoprecipitation in BMDCs with an antibody recognizing tri-methylated H3K27 (H3K27me3) and observed that the association of H3K27me3 with the Aldh1 or Aldh2 promoters in control cells (Figure 7G). The amount of H3K27me3 associated with these promoters was significantly reduced in Ezh2-deficient DCs (Figure 7G), which is consistent with the reduced transcriptional repression of Aldh genes and increased number of cells with ALDH activity. Taken together, these data reveal that upon exposure to a high-dose hapten, Ezh2-deficient Langerin– dDCs, which express high level of ALDH, are efficiently mobilized and are likely to substitute LCs in promoting cutaneous tolerance.

Figure 7

Ezh2 Represses Aldehyde Dehydrogenase Activity of Dendritic Cells

(A) ALDH activity of individual DC subsets in sdLNs. Dot plots were pre-gated on singlet, live cells. CD11c+ MHC IIhi migratory DCs were further analyzed for CD103+ DCs (CD103+ CD11b–/+), CD11b+ DCs (CD103− CD11b+ EpCAM–), LCs (CD103– CD11b+ EpCAM+), and CD103– CD11b– DCs (CD103– CD11b– EpCAM–) (top panel). Data shown are representative of 3 independent experiments.

(B) Bar graphs indicate percentages of AldeFluor+ cells. Data shown are from 3 independent experiments (n = 3) and were analyzed by two-tailed Student's t test (**p = 0.001).

(C) ALDH activity of immature and mature BMDCs. Cells treated with ALDH inhibitor (diethylaminobenzaldehyde) were included to determine specificity of staining. Dot plots were pre-gated on singlet, live cells. Data shown are representative of 3 independent experiments (n = 3).

(D) Bar graph indicates percentages of AldeFluor+ cells. Data shown are from 3 independent experiments (n = 3) and were analyzed by two-tailed Student's t test (***p = 0.0001).

(E) Expression of Aldh2 in mature BMDCs by intracellular staining. Data shown are representative of 3 independent experiments (n = 3).

(F) Aldh1a2 and Aldh2 expression in control and Ezh2-deficient mature BMDCs (mean ± SEM). Gene expression was normalized against Hprt. Data shown are representative of 3 independent experiments (n = 3) and were analyzed by two-tailed Student's t test (from left to right ***p = 0.0001, 0.0001).

(G) H3K27me3 chromatin immunoprecipitation assay of Aldh1a2 and Aldh2 in mature control and Ezh2-deficient BMDCs (mean ± SEM). Hoxa11 is included as a non-target control. Primer sets 1 and 2 for both Aldh genes are located distal and proximal to the transcription start sites, respectively. Data shown are representative of 3 independent experiments (n = 3) and were analyzed by two-tailed Student's t test (from left to right *p = 0.02; **p = 0.0006, 0.0005; ***p = 0.0001).

Discussion

Here we have reported that Ezh2 controls skin tolerance by regulating different subsets of skin DCs via distinct mechanisms. Ezh2 is required for the regulation of LC transmigration across epidermal basement membrane via direct methylation of Talin1 and thereby promotes adhesion turnover, consequently affecting tolerance induction against cutaneous allergens. Moreover, we also identify a subset of Ezh2-deficient dDCs with enhanced ALDH activity, which could potentially rescue tolerance when LCs are unable to exit the skin compartment. These data reveal previously unappreciated abilities of Ezh2 to regulate LC migration non-epigenetically and ALDH activity of DCs epigenetically, which collectively contribute to its role in controlling cutaneous tolerance.

Ezh2 is a well-established histone methyltransferase that mediates gene silencing via regulation of chromatin compaction. Surprisingly, loss of Ezh2 did not exert a significant influence on the gene expression profiles in any DC subset analyzed in our study and the global levels of H3K27 methylation remained comparably low in both Ezh2-deficient and wild-type LCs as well as in BMDCs (Gunawan et al., 2015). These data suggest that loss of Ezh2 epigenetic regulation in DCs is likely compensated for by the up-regulation of Ezh1. However, Ezh2 can also methylate a selected number of non-histone substrates (Dasgupta et al., 2015, He et al., 2012, Jung et al., 2013, Lee et al., 2012). We have identified previously that Ezh2 is localized in the cytoplasmic compartment of several immune cells and has potential functions in cytosol (Gunawan et al., 2015, Su et al., 2005). In our current study, we report that cytoplasmic Ezh2-mediated tri-methylation of the integrin adaptor protein Talin1 controls the adhesion dynamics and migratory potential of both human and mouse skin LCs. Our data indicate that Ezh2 is essential for LC trafficking to the sdLNs under both steady-state and inflammatory conditions and therefore plays a pivotal role in cutaneous tolerance induction. In the absence of Ezh2, LCs were frequently found to be trapped on the basement membrane and displayed atypical turnover of adhesion structures, leading to increased cell spreading and defective integrin-dependent migration on laminin-coated surfaces. Accordingly, CD11c-cre; Ezh2 mice displayed progressive accumulation of LCs in the epidermis with increasing age, which could not be explained by Ezh2 effects on basal proliferation or apoptotic rates. It is therefore likely that LC accumulation over time is due to the impaired steady-state emigration of Ezh2-deficient cells from the skin.

Maintaining constant DC numbers in peripheral tissues and lymphoid organs is critical to ensuring a balance between protective and pathological immune activation, but it remains poorly understood how this process is regulated in vivo. A previous study suggested that loss of DC from the tissues triggers a feedback mechanism that stimulates recruitment and differentiation of pre-cDCs (Hochweller et al., 2009). However, we observed that Ezh2-deficient LCs continued to replenish themselves in steady state even when cell numbers exceeded normal population numbers. Since LCs' down-regulation of E-cadherin and migration toward basement membrane were not affected by Ezh2 deficiency, it is possible that the stalled Ezh2-deficient LCs on basement membrane is sufficient to stimulate LCs or their precursors to proliferate/differentiate and occupy the new niche. Proximity to keratinocytes may therefore be an important influence on sensing and control of LC numbers in the epidermis. A role for Ezh2 in the homeostatic regulation of LC numbers remains possible, whereas steady-state DC numbers in other organs such as the spleen and thymus, where a physical barrier like basement membrane is not present, are unaffected by Ezh2 deficiency (Gunawan et al., 2015) (data not shown). We therefore propose that the observed age-dependent accumulation of Ezh2-deficient LCs in the epidermis is due to reduced steady-state egress from the skin.

LCs are thought to depend on integrin-based adhesions to migrate and must cross the basement membrane to reach sdLNs, whereas dDCs migrate via actomyosin contractions and can directly access the lymphatic vessels, which contain little or no basement membrane (Lammermann et al., 2008). As such, only the migration of Ezh2-deficient epidermal LCs, but not dDCs, to sdLNs is impaired. In line with these data, control and CD11c-cre; Ezh2 mice were observed to respond equally to DNFB-mediated CHS upon tolerization with high-dose DNTB. This also suggests that pro-inflammatory gene expression in DCs is not released from repression by loss of the transcriptional silencer Ezh2, unlike what has been reported in the case of other Ezh2 inhibitor-treated leukocytes (McCabe et al., 2012). Although this redundancy of LCs with dDCs during cutaneous tolerization is consistent with earlier reports (Honda et al., 2010, Noordegraaf et al., 2010), studies by others have yielded seemly conflicting observations (Bobr et al., 2010, Kaplan et al., 2005). Their studies suggest a non-redundant role of LCs in tolerance induction by depleting LCs using human diphtheria toxin receptor/diphtheria toxin system. In our experimental model, the Ezh2-deficient LCs could not migrate and deliver allergen/tolerogen to sdLNs and the dDCs were also Ezh2-deficient with potentially altered functions. When we treated the animals with high-dose DNTB, we promoted, as reported (Noordegraaf et al., 2010), the dissemination of hapten into the dermis and the migration of a specific subset of dDCs (Langerin– or CD11b+) in both control and CD11c-cre; Ezh2 mice. However, Ezh2-deficient dDCs were able to rescue tolerance induction in the absence of LCs. Hence, apart from regulating the migration of LCs, Ezh2 plays a role in controlling the tolerogenic properties of dDCs as well.

A previous in vivo DC-targeting study suggested the superior ability of Langerin+ dDCs to induce Treg conversion (Idoyaga et al., 2013), whereas in our system we did not observe these cells to migrate to the sdLNs in response to high-dose DNTB. Given that the Langerin+ DC population of the dermis is relatively small compared with the Langerin– DC population, their influence on tolerance induction via hapten exposure may be limited. Similarly, a previous study utilizing the same tolerization and sensitization regimen observed migration of both LCs and Langerin– DCs, but not Langerin+ DCs, to the sdLNs in response to DNTB application. However, they reported that only LCs are capable of inducing CD8+ T cell anergy and Foxp3+ Tregs activation to promote cutaneous tolerance (Gomez de Aguero et al., 2012). Here, we observed that increased number of migratory Ezh2-deficient Langerin– dDCs was correlated with restored tolerance in CD11c-cre; Ezh2 mice upon treatment with high-dose DNTB, because they possess higher ALDH activity and expression than wild-type counterparts, which catalyzes retinoic acid synthesis and thereby mediates Treg differentiation (Guilliams et al., 2010). Even though the overall gene expression profile of DCs was unaffected by Ezh2 deficiency, Aldh is one of the few genes found to be up-regulated in Ezh2-deficient DCs. It is likely that the increased expression of ALDH enhances the tolerogenic capacity of Ezh2-deficient Langerin– dDCs and thereby enables cutaneous tolerance in the absence of LCs upon exposure to high-dose DNTB. Our findings therefore suggest that modulation of the Ezh2 level in Langerin– dDCs may serve as an approach for the therapeutic induction of cutaneous tolerance. However, further study will be required to fully elucidate the tolerizing properties of these cutaneous DC subsets before effective treatment strategies can be designed to enhance host protection against skin allergy and autoimmune diseases.

Taken together, our data establish a critical role for Ezh2 in the control of LC transmigration across the basement membrane and the regulation of ALDH activity in Langerin– dDCs, thereby promoting the induction of skin tolerance toward innocuous haptens. Our findings that Ezh2 regulates LC- and DC-mediated immune responses may also extend to other inflammatory skin disorders in which LC migration and dDC tolerogenic potential have been identified as critical determinants of disease outcome, such as psoriasis and Candida albicans infection (Cumberbatch et al., 2006, Igyarto et al., 2011). To further support the discovery made in our animal studies, we observed in publicly available microarray expression database a trend of reduced Ezh2 expression in skin sample derived from patients with dermatitis and psoriasis compared with healthy controls (De Benedetto et al., 2011), highlighting the potential correlation between Ezh2 expression and cutaneous health. Hence, novel strategies that modulate the nuclear and cytosolic functions of Ezh2 in DCs may represent effective therapeutic approaches for various types of cutaneous diseases.

Limitations of Study

In this study, we demonstrated the association between epigenetic de-repression of Aldh and enhanced tolerogenicity of Ezh2-deficient dDCs. However, we did not directly address the contribution of high ALDH activity in Ezh2-deficient Langerin− dDCs to the rescue of tolerance in CD11c-cre; Ezh2 mice upon treatment with a high dosage of haptens. This will require the generation of additional mouse model to conditionally inactivate all Aldh loci in Ezh2-deficient dDCs of chimeric mice with ALDH-expressing Ezh2-deficient LCs. Such genetic tool is currently not available, and chemical inhibition or short hairpin RNA-mediated knockdown is likely to affect both LC and dDC at the same time in any in vivo setting. Nevertheless, since ALDH-associated tolerogenic capacity of Langerin– (CD11b+) dDCs in promoting Treg differentiation has been demonstrated in previous study (Guilliams et al., 2010) and the Langerin– (CD11b+) dDCs were the only Ezh2-deficient skin DC population that migrated into sdLN upon high-dose DNTB treatment, Ezh2-deficient Langerin– dDCs are very likely to contribute to the rescued tolerance induction in the absence of LCs.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.