Type-1 cytokines regulate matrix metalloprotease-9 production and E-cadherin disruption to promote melanocyte loss in vitiligo

Loss of melanocytes is the pathological hallmark of vitiligo, a chronic inflammatory skin depigmenting disorder induced by exaggerated immune response, including autoreactive CD8 T cells producing high levels of type-1 cytokines. However, the interplay between this inflammatory response and melanocyte disappearance remains to be fully characterized. Here, we demonstrate that vitiligo skin contains a significant proportion of suprabasal melanocytes, associated with disruption of E-cadherin expression, a major protein involved in melanocyte adhesion. This phenomenon is also observed in lesional psoriatic skin. Importantly, apoptotic melanocytes were mainly observed once cells were detached from the basal layer of the epidermis, suggesting that additional mechanism(s) could be involved in melanocyte loss. The type-1 cytokines IFN and TNF induce melanocyte detachment through E-cadherin disruption, and the release of its soluble form, partly due to the matrix metalloproteinase MMP-9. MMP-9, whose levels are increased in vitiligo skin and patients’ sera, is produced by keratinocytes in response to IFN and TNF. Inhibition of MMP-9 or the JAK/STAT signaling pathway prevents melanocyte detachment in vitro and in vivo. Therefore, stabilization of melanocytes in the basal layer of the epidermis by preventing E-cadherin disruption appears promising to prevent the depigmentation occurring in vitiligo and during chronic skin inflammation.


Introduction
Vitiligo is the prototype of a depigmenting disease due to an exaggerated skin immune response.
However, depigmentation is not restricted to vitiligo and may occur during the course of other chronic skin inflammatory disorders, such as psoriasis (1). Vitiligo impacts the quality of life of patients dramatically and therapies so far remain limited (2). Its pathological hallmark is the loss of epidermal melanocytes, the cells located in the basal layer of the epidermis and responsible for pigmentation and skin phototype. Vitiligo is a complex multifactorial disease combining genetic predisposition and environmental triggers (e.g. friction, chemicals), together with metabolic, immunologic and inflammatory abnormalities (2,3). Depigmentation in vitiligo is consistently associated with infiltration of immune cells, in particular CD8 T cells, in close apposition to the remaining melanocytes, at least during disease progression (4)(5)(6). Importantly, vitiligo should be viewed as an immune memory skin disease, as recently highlighted by the role of resident memory T cells (TRM) during pathogenesis (7)(8)(9). A prominent type-1 cytokines skewed immune profile is characteristic of vitiligo patients' skin, as the majority of melanocyte-specific skin CD8 TRM cells express the chemokine receptor CXCR3 and produce elevated levels of the type 1-related cytokines IFN and TNF (7,10). The CXCR3/ IFN / TNF axis has been shown to be involved in the process leading to depigmentation in vitro and in vivo in a disease-prone mouse autoimmune model of vitiligo (11)(12)(13)(14), and IFN and TNF are involved in epidermal pigmentation homeostasis, inhibiting melanocyte function, phenotype and melanogenesis (4-6, 15, 16) However, although previous studies identified the role of these cytokines in the pathomechanism of the disease, the interplay between inflammation and the melanocyte loss characterizing vitiligo remains elusive.
The leading hypothesis is that melanocyte loss is mediated by autoreactive CD8 T cells (3,(17)(18)(19). Indeed, in vitro studies reported that CD8 T cells isolated from vitiligo patients' skin induced the apoptosis of autologous melanocytes (17,20,21). Nonetheless, clear evidence of the presence of apoptotic cells in vivo is still lacking and we recently showed that CD8 T cells from vitiligo skin display moderate cytotoxic activity comparable to that found in healthy skin and the lesional skin of psoriasis patients (7), suggesting that additional disease mechanism could be involved in melanocyte disappearance. Moreover, depigmentation in mouse models of vitiligo depends rather on IFN signaling, while perforin is dispensable (11,13), suggesting that the release of proinflammatory cytokines such as IFN in the skin is important in the process leading to depigmentation.
Melanocyte disappearance could also result from their detachment (22,23). The stabilization of melanocytes in the basal layer of the epidermis is dependent on the adhesion protein E-cadherin (24,25). Interestingly, impaired cell-surface E-cadherin expression was recently shown in the nonlesional skin of vitiligo patients (26,27). Several mechanisms could be involved in such disruption, such as inhibition of its expression, internalization in the endosomal structure, or cell surface cleavage into a soluble form (soluble E-cadherin). E-cadherin cleavage may be induced by several proteases, including matrix metalloproteases such as MMP-3, MMP-7, MMP-9, A disintegrin and MMP domain-containing protein 10 (ADAM10), which are all known to be involved in extracellular matrix remodeling and cell migration in various physiologic and pathologic processes (28,29).
MMPs are a family of Zn-dependent proteases with common functional and structural properties.
MMP-9, also known as 92kDa gelatinase/type IV collagenase, is constitutively expressed primarily in leukocytes, while most other cell types including keratinocytes express MMP-9 in response to various pro-inflammatory cytokines such as IFN, TNF, IL-1, IL-6 (30,31). Initially recognized for its collagen-remodeling function, MMP-9 activity is now known for its role in the control of immune responses, such as the migration of immunocompetent cells into and out of peripheral tissues, and the regulation of the activation of both CD4 and CD8 T cells (32)(33)(34).
Here we report that vitiligo perilesional skin is characterized by the basal detachment of melanocytes associated with the disruption of E-cadherin surface distribution and the release of soluble E-cadherin; melanocyte apoptosis was mainly noticed following their detachment from the basal epidermal layer. Strikingly, this phenomenon is not restricted to vitiligo disease but may also occur in psoriasis. We further show that the type-1 cytokines IFN and TNF induce melanocyte detachment in both in vitro and in vivo models, and that this effect is dependent on the inhibition of E-cadherin gene expression, internalization of E-cadherin, and its cleavage through the release of MMP-9 by keratinocytes. Strikingly, MMP-9 inhibitors and JAK inhibitors prevented the release of soluble E-cadherin, leading to the stabilization of melanocytes in the basal layer of the epidermis. Our results highlight an additional disease mechanism leading to the loss of melanocytes in vitiligo and more generally in skin inflammation, and suggest the need for further exploration of therapeutic strategies aiming to both dampen the inflammation and maintain melanocyte stability in the basal layer of the epidermis of vitiligo patients.

Detachment of basal melanocytes is a hallmark of vitiligo and psoriasis
We first used immunofluorescence to examine the distribution of melanocytes in the epidermis in the context of skin inflammation. We examined the perilesional skin of patients with both vitiligo and psoriasis, and the lesional skin of psoriasis patients (Figure 1, A and B, patient characteristics are displayed in Supplemental Tables 1 and 2). Psoriasis was used as the archetype of a skin inflammatory disease. Collagen VII staining was used to identify the basal layer of the epidermis.
While melanocytes were located in the basal layer of the epidermis in the control skin of unaffected individuals, a significant number of melanocytes were found in the suprabasal layers in vitiligo perilesional skin, not only in patients with stable or active disease but also in the acanthotic epidermis of the perilesional skin of patients with concomitant psoriasis and vitiligo. Strikingly, similar observations were made in the lesional skin of patients only with psoriasis. The remaining melanocytes in vitiligo perilesional skin were unable to proliferate irrespective of their localization within the epidermis, unlike the proliferating melanocytes observed in lesional psoriatic skin, as revealed by Ki67 expression (Supplemental Figure 1). This suggests that while the detachment of melanocytes is a common process seen in vitiligo and psoriasis, vitiligo is characterized by the absence of melanocyte regeneration, unlike psoriasis. Importantly, cleaved caspase 3 staining or TUNEL assay showed that only few basal melanocytes were apoptotic in these conditions. This was in contrast with the epidermal cell death observed in cutaneous lupus erythematosus and toxic epidermal necrolysis, two skin diseases associated with strong inflammation and epidermal cell death ( Figure 1, C and D and Supplemental Figure 2). Apoptotic suprabasal melanocytes were evidenced only in some patients with active vitiligo (Figure 1, C and D). This suprabasal localization of melanocytes in vitiligo and psoriasis skin seems to be the consequence of a defect in their adhesion to keratinocytes, as shown by the disrupted distribution of the major adhesion molecule mediating melanocyte adhesion to keratinocytes, E-cadherin (25), in both melanocytes and keratinocytes ( Figure 1E).
Melanocytes were classified into three types according to the distribution of cell surface Ecadherin staining, as previously described (26): homogeneous (type 1), heterogeneous (type 2) and absence of E-cadherin labelling (type 3). Compared to healthy control skin in which melanocytes stained homogeneously for E-cadherin, melanocytes from vitiligo perilesional skin, particularly in the active phase of the disease, and lesional psoriasis skin displayed a discontinuous cell-surface E-cadherin expression ( Figure 1E and Supplemental Figure 3A). In addition, soluble E-cadherin levels were significantly higher in stable and active vitiligo patients' sera compared to those of healthy controls ( Figure 1F). These findings suggest that pro-inflammatory factors released by immune and epidermal cells during skin inflammation are able to regulate the distribution of Ecadherin on melanocytes and are responsible for their detachment from the basal layer of the epidermis.

Type-1 cytokines IFN and TNFα induce detachment of melanocytes and disrupt E-cadherin distribution
The immune response of vitiligo is predominantly associated with Th1/Tc1 cells infiltrating the skin together with an elevated production of both IFN and TNFα (3). This immune bias is also found to a lesser extent in psoriasis compared to the strong Th17 skewed immune profile (35).
We next used an in vitro 3D model of reconstructed pigmented human epidermis (RHPE) containing both keratinocytes and melanocytes to investigate whether TNF and IFN could be involved in the E-cadherin disruption observed in patients. We found that the combination of TNF and IFN induced the detachment of more melanocytes from the basal layer than each cytokine alone. Such detachment was also observed in vitro on cocultures of melanocytes and keratinocytes (Supplemental Figure 3, B and C). The process was mediated partly by an altered Ecadherin distribution in melanocytes ( Figure 2, A and B and Supplemental Figure 3D). It was not associated with prominent melanocyte cell death, as observed by TUNEL assay staining ( Figure   2C) or by assessing melanocyte viability in response to increasing concentrations of these cytokines (Supplemental Figure 4). In addition, only the combination of TNF and IFN was able to downregulate significantly the expression of the E-cadherin encoding gene CDH1. This inhibition was specific to melanocytes as CDH1 transcript levels were not regulated in keratinocytes ( Figure 2D). The decrease was dose-and time-dependent (Figure 2, E and F). TNF and IFN also decreased transcript levels of the melanocyte adhesion-related genes CCN3 and DDR1, albeit more weakly, while no significant regulation of the -catenin-encoding gene CTNNB1 was noted (Supplemental Figure 5). DDR1 was recently reported to stabilize the membrane localization of E-cadherin (36). Importantly, the two cytokines induced E-cadherin relocalization in melanocytes, as evidenced by the presence of E-cadherin in vesicular structures positive for LAMP-1 staining, a known marker for lysosomes and late endosomes ( Figure 2G). Furthermore, levels of soluble E-cadherin were also increased in cell-free supernatants of RHPE treated by both TNF and IFN ( Figure 2H), suggesting their ability to induce cleavage of Ecadherin indirectly.

MMP-9 levels are increased in the circulation and skin of vitiligo patients
Consequently, we explored which major protease(s) could be involved in the cleavage of Ecadherin and lead to destabilization of melanocytes. Serum levels of both zymogen and active forms of MMP-9 were higher in active vitiligo and psoriasis patients than in healthy controls ( Figure 3, A and B), while MMP-3, MMP-7, and ADAM10 serum levels were similar in all groups (Supplemental Figure 6, A-C). MMP-9 levels were significantly higher in vitiligo patients with active disease (Figure 3, A and B), and a positive correlation between total and active MMP-9 was observed in vitiligo patients (Supplemental Figure 6D) Figure 6G). This was in contrast with MMP-3, MMP-7, and ADAM10 gene and protein expression, which remained lower and not significantly different between healthy control skin, vitiligo and psoriasis skin in comparison to strong MMP-9 upregulation (Supplemental Figure 6, H-J). We then compared the inflammatory transcriptome profile of the perilesional skin of patients with stable or active vitiligo using the nCounter® inflammation panel (248 genes) to identify the genes that were the most differentially regulated. We found that the MMP9 gene was among the top ten genes upregulated in active vitiligo skin ( Figure 3G). The interactomic network of the upregulated transcripts showed MMP-9 to be at the core of the molecular signature identified, together with genes previously shown to be involved in depigmentation such as TNF and CXCL9 ( Figure 3H). Lastly, TNF and IFN strongly induced the expression and production of MMP-9 by epidermal cells, mainly by keratinocytes (Figure 3, I-K), while they had little or no effect on MMP-3, MMP-7 and ADAM10 expression (Supplemental Figure 7, A and B).
MMP-9 (also known as gelatinase B) belongs to the gelatinase subgroup of MMPs, which also includes MMP-2 (also known as gelatinase A). We therefore investigated whether MMP-2 was deregulated in our experimental setting. In contrast to MMP-9, levels of MMP-2 were decreased in vitiligo patients' sera and a similar, although not significant, tendency was observed in psoriasis serum (Supplemental Figure 8A). Nonetheless, MMP-2 serum levels did not correlate negatively with MMP-9 serum levels or the body surface area involved in vitiligo patients (Supplemental

MMP-9 is involved in melanocyte detachment induced by type-1 cytokines in vitro and in vivo
We next explored the contribution of MMP-9 to melanocyte destabilization in RHPE and observed that active MMP-9 induced a dose-dependent increase in the proportion of suprabasal melanocytes, associated with an increase in soluble E-cadherin levels ( Figure 4, A-C). Subsequent experiments using the gelatinase inhibitor SB-3CT or a selective inhibitor of MMP-9 (ab142180) investigated the impact of MMP-9 inhibition in RHPE treated with TNF and IFN. We observed a modest and robust dose-dependent inhibition of melanocyte detachment induced by type-1 cytokines with both of the inhibitors tested ( Figure 4, D and E). This effect was associated with a decrease in soluble E-cadherin levels ( Figure 4F). To fully confirm this mechanism in vivo, both TNF and IFN were injected intradermally into C57BL/6 wild-type mice every day for 6 days ( Figure 4G).
Owing to the low number of melanocytes in mouse skin, the mouse tail was used for injection, because melanocytes are located in the basal layer of epidermis, thus reproducing human pigmentation. Consistent with human data, concomitant intradermal injection of IFN and TNFα induced a significant detachment of melanocytes associated with disruption of E-cadherin expression ( Figure 4, H and I and Supplemental Figure 9). In support of our results obtained in vitro, MMP-9 inhibition with SB-3CT was associated with a significant stabilization of epidermal melanocytes in the basal layer of the epidermis (Figure 4, H and I), thus providing further evidence of the involvement of MMP-9 in the process leading to melanocyte loss.

MMP-9 reduction
To further decipher the mechanism involved in melanocyte destabilization, we studied the impact of JAK signaling using tofacitinib (a JAK1/3 inhibitor) or ruxolitinib (a JAK1/2 inhibitor) in our in vitro and in vivo models of melanocyte detachment induced by TNF and IFN. Both JAK inhibitors led to significant melanocyte stabilization in RHPE treated with TNF and IFN ( Figure   5, A and B), associated with a decrease in levels of soluble E-cadherin and active MMP-9 ( Figure   5, C and D). A decrease in MMP9 levels was also observed in vitro on vitiligo perilesional epidermis explants following tofacitinib treatment ( Figure 5E). Inhibition of JAK signaling also prevented melanocyte detachment induced by type-1 cytokines in vivo ( Figure 5 F and G), and no significant difference was noted between the two inhibitors tested in our models.

Discussion
Our findings shed light on a so far unknown mechanism regarding the interplay between the inflammatory response and melanocyte loss in vitiligo. We demonstrate that destabilization of melanocytes from the basal layer of the epidermis is an important event leading to their loss. This destabilization results from alteration of membrane E-cadherin in response, at least in part, to the upregulated production of active MMP-9 by epidermal cells, especially keratinocytes, induced by the type-1 cytokines IFN and TNFα, which are two pro-inflammatory cytokines involved in vitiligo. This phenomenon could be reproduced by administering active MMP-9 or prevented by using MMP-9 inhibitors or by inhibiting the JAK/STAT signaling pathway with JAK inhibitors ( Figure 6). Importantly, the suprabasal localization of melanocytes within the epidermis is not restricted to vitiligo, and the mechanism hereby identified could occur in the hypopigmentation associated with other causes of chronic skin inflammation, as in psoriasis. Indeed, we recently reported that hypopigmentation could occur in 10% of patients with psoriasis and was observed in areas previously affected by the disease (37). However, in contrast to vitiligo, complete and durable depigmentation was not observed in psoriasis, perhaps due to 1/ the proliferative capacity of melanocytes, which is greater in lesional psoriatic skin than in vitiligo perilesional skin and 2/ to apoptosis of suprabasal melanocytes in vitiligo, as also previously reported (26) and which could involve the recently identified innate lymphocyte-induced CXCR3B-mediated apoptosis (38).
Another explanation could be that melanocytes in vitiligo patients have intrinsic abnormalities that cause a regenerative deficiency (2,3). Indeed, the activation, maturation, proliferation and recruitment of melanocyte precursors in vitiligo skin are defective (39), explaining the persistence of white patches. This contrasts with lesional psoriatic skin which harbors a high proportion of proliferating melanocytes, explaining not only the absence of complete depigmentation in patients despite melanocyte detachment from the basal layer of the epidermis, but also the fact that depigmented lesions are able to recover rapidly under therapies. This is consistent with previous reports showing an increased number of melanocytes both in psoriasis lesions and in resolved psoriasis skin (40,41).
Our data suggest that the death of melanocytes, especially in vitiligo patients, could occur following their detachment and relocalization to the outer layers of the epidermis (melanocytorrhagy), as previously observed (26). Vitiligo is characterized by clinically undetectable and pathologically mild inflammation with infiltration of melanocyte-specific resident memory CD8 T cells producing elevated levels of pro-inflammatory cytokines, while production of cytotoxic markers was similar to healthy skin and lesional psoriasis skin (7).
Interestingly, it was recently demonstrated that a melanocyte antigen can trigger auto-immunity in psoriasis with skin infiltration of epidermal melanocyte-specific CD8 T cells producing proinflammatory cytokines involved in the development of psoriasis, such as IL-17 and IFN (42).
However, while granules containing granzyme B were identified in a psoriasis skin CD8 T cell subset, the authors failed to detect any signs of cell death in melanocytes. In view of our findings, we hypothesize that IFN and TNF produced by resident memory T cells in vitiligo skin contributes to melanocyte destabilization. Our observations contrast with what may be observed in other inflammatory skin disorders, such as the group of lichenoid dermatitis associated with cutaneous lupus disease and lichen planus. These diseases are characterized by strong immune infiltration of the basal layer leading to the destruction of epidermal cells such as melanocytes, the release of melanin in the dermis (incontinence) and to definitive cicatricial pigmentary changes (43).
Our data reveal a previously unknown role of MMP-9 during depigmentation and highlight its prominent role in inducing melanocyte destabilization by the shedding of E-cadherin and the release of its soluble form. Besides type 1 cytokines, additional cytokines upregulated in vitiligo patients' blood and/or skin and known to increase MMP-9 expression could also be involved in its upregulation, including IL-1, IL-17, or IL-6 (30,31,44). In line with a study suggesting the low production of MMP-9 by melanocytes in vitiligo patients (45), we found that keratinocytes were the main epidermal source of MMP-9. In contrast, another gelatinase, MMP-2, was found to be decreased in vitiligo patients' sera and its production downregulated by IFN and TNFα. MMP-2 is known to be involved in the migration of melanocyte precursors for their optimal epidermal replenishment (46). MMP-2 reduction, together with the increase in MMP-9 production, could therefore impact simultaneously the stabilization of epidermal melanocytes and their replenishment from the melanocyte precursor reservoir, thus leading to durable depigmentation in vitiligo. However, while we could not identify any increased expression of some other proteases (e.g. MMP-3, MMP-7, and ADAM10) known to play a role in E-cadherin cleavage, we cannot rule out the possibility that other proteases not evaluated in this study are involved in this process. This is supported by the fact that the treatment of RHPE with active MMP-9 induced a significant but lower melanocyte detachment than that observed in response to IFN and TNFα. Additionally, it would be relevant in psoriasis to test the role of Th17-related cytokines alone or in combination with Th1-related cytokines in this phenomenon.
We now need to focus on two major goals in vitiligo therapy: first, dampening the activation of the immune response responsible for the loss of melanocytes and the maintenance of depigmentation; and second, maintaining melanocytes in the basal layer and promoting their regeneration from melanocyte precursors. Like other MMPs, MMP-9 is known to be upregulated in autoimmune and inflammatory disorders and to play a role in modulating the innate and adaptive immune response, the production of chemokine ligands and cytokine activity (47). Therefore, in vitiligo and other chronic depigmenting disorders associated with inflammation, the inhibition of MMP-9 activity with topical or systemic agents could dampen both the immune response and help to stabilize melanocytes in the basal layer of the epidermis. Moreover, our study provides evidence that targeted therapies such as topical or systemic JAK inhibitors, which have shown promising results in vitiligo (48)(49)(50)(51), could inhibit the type II IFN response, thereby maintaining melanocytes in the basal layer of the epidermis. Hence, the association of JAK inhibitors together with MMP-9 inhibitors could be a way to lower the dose of each component for either systemic or topical applications, ideally in combination with a therapy stimulating melanocyte regeneration such as phototherapy. While our study provides new insights into the targeting of MMP-9 in vitiligo, it is now critical to confirm this strategy in preclinical models of vitiligo. Mouse models of skin depigmentation following infiltration of melanocyte-specific CD8 T cells are available (57) and could be used to fully validate a strategy to inhibit melanocyte detachment through targeting of MMP-9. Showing that the use of a MMP-9 inhibitor alone or in combination with immunomodulating agents could either prevent depigmentation and/or induce repigmentation would be of great interest before going to clinical trials. We now need to set the therapeutic objective of stabilizing melanocytes in the basal layer of the epidermis with specific therapies able to restore the expression of membrane E-cadherin or to inhibit its cleavage.  (52), notably for disease activity. Briefly, they were classified using Wood's lamp examinations, as previously reported (53,54). Patients with a total spreading score 3 according to the VETF scoring system and/or the presence of hypomelanotic lesions with poorly defined borders and/or confetti-like lesions were considered active, while those with a total spreading score 1 and/or the absence of new lesions over the past 12 months were considered stable. Skin biopsies (4mm diameter) were obtained from the perilesional area. Lesional psoriatic skin biopsies were obtained as a control of skin inflammatory disorder. Paraffin-embedded skin sections from patients with cutaneous lupus and toxic epidermal necrolysis were obtained from the Department of Pathology at Bordeaux University Hospital, France. Unaffected control skin was obtained as discarded human tissue from cutaneous plastic surgery (Bordeaux University Hospital, France).

Subjects
Blood from unaffected subjects was obtained from volunteers exempt of autoimmune or inflammatory disorders.
Cell cultures and cytokine treatment. Primary Human Melanocytes were isolated from healthy children's foreskin as previously described (55) and maintained in melanocyte growth medium (MGM) supplemented with 1 ng/ml basic recombinant human fibroblast growth factor, 0.5 μg/ml hydrocortisone, 4 μg/ml bovine pituitary extract, 5 μg/mL recombinant human insulin (all from Promo Cell), 100 U/ml penicillin and 100 μg/ml streptomycin (Eurobio). Primary human keratinocytes were obtained from surgical samples of healthy breast skin as previously described (56). Cells were cultivated in keratinocyte serum-free medium (           TNFa and IFNg produced by activated T RM cells induce an E-cadherin defect in melanocytes. TNFa and IFNg induce the production of MMP-9 by epidermal cells, especially keratinocytes, that cleave E-cadherin (E-cad) to release its soluble form. Ecadherin cleavage leads to melanocyte destabilization. This effect is inhibited in the presence of MMP-9 or JAK inhibitors. TRM cells