IFN-γ enhances cell-mediated cytotoxicity against keratinocytes via JAK2/STAT1 in lichen planus

Abstract

Lichen planus (LP) is a chronic debilitating inflammatory disease of unknown etiology affecting the skin, nails, and mucosa, and is histologically characterized by dense infiltration of T cells and epidermal keratinocyte apoptosis. Using global transcriptomic profiling, we demonstrate that LP is characterized by a type II interferon (IFN) inflammatory response. The type II IFN, IFN-γ is demonstrated to prime keratinocytes and increase their susceptibility to CD8+ T cell mediated cytotoxic responses through MHC class I induction, and we show that this process is dependent upon Janus kinase 2 (JAK2) and Signal transducers and activators of transcription 1 (STAT1), but not JAK1 or STAT2 signaling. Lastly, using drug prediction algorithms, we identify JAK inhibitors as promising therapeutic agents in LP, and demonstrate that the JAK1/2 inhibitor baricitinib, fully protects keratinocytes against cell-mediated cytotoxic responses. In summary, this work elucidates the role and mechanisms of IFN-γ in LP pathogenesis and provides evidence for the therapeutic use of JAK inhibitors to limit cell-mediated cytotoxicity in patients with LP.

One Sentence Summary:

IFN-γ enhances keratinocyte death in LP.

Introduction

Lichen planus (LP) is a chronic inflammatory skin disease of unknown etiology characterized by often widespread pruritic skin lesions, and less commonly painful oral erosions, or nail involvement. Several subtypes have been described based on the morphology of the lesions and the site of involvement (1). Among them, hypertrophic lichen planus (HLP) is characterized by pruritic hypertrophic or verrucous plaques typically on the lower limbs and tends to have a more chronic course (2). LP has a major impact on quality of life, particularly in those with oral LP (3). Histologically, LP is characterized by a dense band-like infiltration of lymphocytes in the upper dermis along the dermal-epidermal junction and keratinocyte apoptosis (4). In addition, both oral and cutaneous LP have been reported to have the potential to transform to malignancy (5).

The immune mechanisms involved in LP pathogenesis have been shown to involve the activation of infiltrated T cells and the immune responses against keratinocytes. Patients with LP have increased numbers of CD8+ T cells in the skin and blood (6, 7), and in skin, CD8+ T cells are located in close proximity to dyskeratotic keratinocytes (8), consistent with cell-meditated cytotoxicity as a central mechanism in LP pathogenesis. CD4+ T cells, including Th1, Th2, Th17, and T-regulatory cells have also been implicated as participants in LP (9), and epithelial-derived cytokines, chemokines, and costimulatory molecules may also play a crucial role, such as the Th1 chemokines chemokine (C-X-C motif) ligand (CXCL)9/10 (10).

Despite some recent progress, much is still unknown about the pathogenesis of LP, and there is urgent need for more effective treatments. Currently there are no FDA approved treatments for LP. In this study, we aimed to illustrate the predominant cytokine via global gene expression in skin lesions, and focused on the functions of interferon-γ (IFN-γ) and related potential targeted therapy for LP.

Results

Transcriptomic profiling of LP lesions reflects an IFN-γ-dominant inflammation

To investigate the specific gene regulations in LP, microarray profiling was performed on RNA extracted from paraffin embedded skin biopsy samples obtained from patients with LP (n=20), hypertrophic LP (HLP) (n=17), and healthy controls (n=24). Principal components analysis demonstrated a complete separation between LP/HLP and normal control samples but prominent overlap between LP and HLP (Fig. 1A). We identified 1447 and 1533 differentially expressed genes (DEGs) in skin lesions of LP and HLP compared to that in healthy control skin, respectively (FDR<0.1, FC>1.5 or <−1.5, Fig. 1B, Table S1). Of these, 1104 genes, including 890 upregulated genes and 214 downregulated genes, were shared between LP and HLP (Fig. 1B). Gene-ontology pathways showed very similar enrichment for LP and HLP except that keratinization was more prominent in HLP, which was consistent with the clinical and histologic presentation of this disease (Fig. 1C). Further enrichment in LP and HLP DEGs was for pathways involved in IFN-γ signaling, defense response, antigen processing and presentation, and apoptotic process (Fig. 1C), revealing a predominant IFN-mediated immune response in LP lesions. Therefore, we further assessed gene expression of type I, type II and type III IFN family members, IFN receptors, and IFN response genes. As expected, IFNG, the sole type II IFN, and IFNGR1 and IFNGR2 were dramatically increased in both LP and HLP samples, and IFN-induced genes (MX1 and OAS1) were all increased in LP/HLP skin lesions (Fig. 1D, E). In contrast, none of the type I and type III IFNs were increased as IFNG in LP/HLP skin lesions compared to healthy control skin (Fig. S1A, B). Immunofluorescent staining showed prominent IFN-γ in LP skin lesions and was predominantly co-localized with CD3 and CD8 (Fig. S1C). Type I IFNs including IFN-α/β, and in particular IFN-κ were also found to be increased in LP skin lesions compared to normal skin (Fig. S1D). Furthermore, the differentially expressed genes were enriched for IFN response genes, but not interleukin-4 (IL-4) or IL-17 response genes (Fig. 1F). These data were confirmed by quantitative real-time PCR (qRT-PCR) demonstrating that IFNG and MX1, but not IL17A or IL22, were increased in LP/HLP skin lesions compared to control skin (Fig. 1G). Therefore, we conclude that IFN-γ is the dominant interferon in LP inflammation.

Fig. 1. Transcriptomic profiling of LP lesions reflects an IFN-γ-dominant inflammation.

(A) Principal components analysis of microarray data from LP (n=20), HLP (n=17) and healthy controls (n=24). (B) The Venn diagram displays the intersection (upper panel) and correlation (lower panel) of differentially expressed genes (FDR<0.1, FC>1.5 or <−1.5) across the LP and HLP transcriptomes. The number of up-regulated and down-regulated genes for each data set is reported in parentheses. (C) Graphs demonstrate the most highly enriched Gene Ontology categories in the transcriptomes of LP and HLP. (D, E) Expression level of type II IFN, IFNG, and IFNGR (D), and the IFN-response genes MX1, MX2, OAS1 and OASL (E) in our microarray data. (F) Further analysis and classification of differentially expressed genes in LP and HLP skin lesions to cytokine-induced genes. (G) QRT-PCR was employed to detect the IFNG, MX1, IL17A, and IL22 expression level in skin lesions of LP (n=22) and HLP (n=16) and normal controls (NC, n=23). One-way ANOVA. Data are presented as the mean ± SEM of measurements obtained in each sample. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

IFN-γ increases keratinocyte sensitivity to cell-mediated cytotoxicity

Keratinocyte apoptosis is a major characteristic of LP skin lesions (11). We therefore, evaluated whether IFN-γ stimulation was associated with increased keratinocyte cell death. To address this, we utilized a culture system where keratinocytes from one individual were admixed with leukocytes from a different donor. Keratinocytes were primed with IFN-γ for 24h, medium replaced, and then co-cultured with CD3/CD28 activated peripheral blood mononuclear cells (PBMCs) from a healthy control for 72h (Fig. S2A). Annexin V-PI flow cytometry demonstrated that keratinocytes pre-treated with IFN-γ had increased level of cell death as measured by Annexin-V staining compared to those not pre-treated with IFN-γ (Fig. 2A, Fig. S2B), and this was observed in a dose-dependent manner (Fig. S2C). Moreover, non-activated PBMCs showed less cytotoxic activity against IFN-γ-induced keratinocytes than CD3/CD28 activated ones (PBMCs: keratinocytes=10:1) (Fig. S2C). Further, TdT-mediated dUTP nick end labeling (TUNEL) staining was confirmatory of the findings from flow cytometry (Fig. 2B). Notably, priming with Type I IFNs including IFN-α/β increased keratinocyte susceptibility to T cell-mediated cytotoxicity, but to a lesser extent than that observed with IFN-γ (Fig. S2D).

Fig. 2. IFN-γ increases keratinocyte susceptibility to cell-mediated cytotoxicity.

(A) Keratinocytes from one donor were co-cultured with CD3/CD28 microbeads-activated PBMCs from a second donor (n=2) and cell death was evaluated by Annexin-V PI staining. The representative flow cytometry data (left panel) and statistical analysis (right panel) are shown.(B) TUNEL staining and statistical analysis of TUNEL-positive cells was used to detect keratinocyte cell death. DAPI was used for nuclear staining (scale bar, 100μm). (C) PBMCs were primed with IFN-γ and then treated with CD3/CD28 microbeads, and then added to unstimulated keratinocyte cultures. (D) CD3/CD28 microbeads-activated PBMCs were pre-incubated with anti-CD4, anti-CD8, or anti-Nkp44 blocking antibody for 1 h, then added to in vitro co-culture model, and keratinocyte death was evaluated by Annexin-V PI staining. Data were analyzed using one-way ANOVA followed by Dunnett’s posttest. Data are presented as the mean ± SD of measurements obtained in triplicate or quadruplicate experiments. *P < 0.05, **P < 0.01, ****p < 0.0001.

In contrast, with the same experimental setup, but with PBMCs primed with IFN-γ for 24 hours prior to CD3/CD28 stimulation and co-culture with keratinocytes, no increase in keratinocyte cell death was seen (Fig. 2C, Fig. S2E).

To determine the cell type responsible for the cytotoxicity in vitro, we used blocking antibodies against CD4, CD8, and NK cells. Blockade of CD8 led to a inhibition of keratinocyte cell death whereas less inhibition was seen with CD4 blockade, and no change with NK-cell blockade (Fig. 2D). To characterize the nature of cell death in vitro, we assessed expression of cleaved-caspase 3, a protein marker of apoptosis, and two markers of necroptosis, phosphorylated receptor-interacting protein-kinase 3 (p-RIP3) and phosphorylated Mixed Lineage Kinase Domain Like Pseudokinase (p-MLKL) (Fig. S3). These data showed mixed apoptotic (cleaved caspase-3 positive) and necroptotic responses (p-RIP3 and p-MLKL positive), similar to what was seen in lesional LP skin in vivo (Fig. S4).

These above indicate that IFN-γ is a critical cytokine promoting epidermal cell death in LP lesions through direct effect on keratinocytes, and not through activation of infiltrating immune cells.

Cytotoxic responses to IFN-γ-primed keratinocytes are MHC class I dependent

Cytotoxic responses against cells are known to be dependent upon MHC antigen processing and presentation. We first analyzed the mRNA expression of MHC class I and II molecules from our microarray data of healthy, LP, and HLP skin and demonstrate that expression of all the major MHC class I and MHC class II molecules were increased in LP and HLP skin compared to healthy control skin (Fig. 3A). Immunofluorescence using pan-anti-MHC class I antibody showed that MHC class I molecules were highly expressed in LP epidermis compared to healthy skin. In contrast, MHC class II staining, using a pan-anti-MHC class II antibody, was localized to the upper dermis co-localizing on the inflammatory infiltrate along the epidermal-dermal junction. No MHC class II staining was seen in the epidermis of LP/HLP lesions. Furthermore, no staining for MHC class II was seen in normal skin (Fig. 3B). To assess the role of IFN-γ in this process, we evaluated the expression level of all MHC molecules in keratinocytes after IFN-γ stimulation and found that IFN-γ primarily upregulated the mRNA expression of HLA-A, HLA-C, and HLA-DR robustly in keratinocytes, but not HLA-B, HLA-DP, and HLA-DQ which had very low expression level in both unstimulated and stimulated keratinocytes (Fig. 3C). Similarly, keratinocytes stimulated with Type I IFNs such as IFN-α/β had increased expression of MHC I molecules, but to a lesser extent than with IFN-γ stimulation (Fig. S5A). To address the role of MHC class I and class II responses, we blocked their function using MHC class I or MHC class II neutralizing antibodies in our model system. Intriguingly, treatment of IFN-γ-activated keratinocytes with a pan anti-MHC class I blocking antibody completely inhibited cytotoxic activity towards keratinocytes. Anti-HLA-DR partially inhibited cytotoxic responses against keratinocytes, while other anti-MHC class II antibodies (HLA-DP/DQ) were inefficient (Fig. 3D, Fig. S5B). Therefore, these data suggest that IFN-γ promotion of cytotoxicity in keratinocytes in LP is mainly through induction of MHC class I expression.

Fig. 3. Cytotoxic responses to IFN-γ-primed keratinocytes are MHC class I dependent.

(A) The mRNA expression level of all MHC molecules from microarray data from LP (n=20), HLP (n=17), and healthy control skin (n=24). (B) Representative immunofluorescence staining of MHC class I and MHC class II in LP/HLP lesions (n=3) and controls (n=3). DAPI was used to show nuclear staining (Scale bar, 100μm). (C) QRT-PCR showed mRNA expression of MHC I and II molecules in keratinocytes stimulated with IFN-γ (10ng/ml). (D) IFN-γ-primed keratinocyte were pre-incubated with anti-MHC class I and II monoclonal antibody for 1 h, and then co-cultured with CD3/CD28 activated PBMCs. Statistical analysis was done using one-way ANOVA followed by Dunnett’s posttest. Data are presented as the mean ± SD of measurements obtained in triplicate or quadruplicate experiments. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

IFN-γ induces keratinocytes to express MHCs through JAK2/STAT1 signaling

To get a deeper understanding of the contribution of JAK/STAT pathway to LP development, we performed literature-based network analysis of the genes regulated in the same direction in both LP/HLP lesions (FDR<0.1, FC>1.5 or <−1.5). This showed that highly connected nodes could be attributed to IFN signaling including JAK2 and STAT1 (Fig. 4A). STAT1 mRNA expression was the most highly expressed gene among all the STAT genes detected in our gene profile analysis (Fig. 4B), and this was confirmed for total and activated (phosphorylated) STAT1, which was primarily observed in the epidermis of LP lesions, but not in healthy control skin (Fig. 4C), whereas STAT2 and phosphorylated STAT2 were mostly expressed in the inflammatory infiltrates (Fig. S6A).

Fig. 4. IFN-γ signals through JAK2/STAT1 signaling in keratinocytes to induce MHC class I expression.

(A) Induced modules network analysis of the differentially expressed genes shared by LP and HLP. Node size correlates with the number of connected nodes and edges. Nodes with 5 connections are marked larger. (B) Expression level of STAT1-6 in LP (n=20), HLP (n=17) and healthy control skin (n=24) in our microarray data. (C) Representative immunofluorescence staining of STAT1 and phosphorylated (p)-STAT1 in LP/HLP skin lesions and controls (four samples were stained and analyzed per group). DAPI was used for nuclear staining (scale bar, 100μm). (D) Keratinocytes were stimulated with IFN-γ (10 ng/ml) for indicated time. Western blot for JAK1, JAK2, p-JAK1, p-JAK2, STAT1, STAT2, p-STAT1, and p-STAT2 are shown. (E) mRNA expression of MHC I molecules in JAK1, JAK2, STAT1, and STAT2 KO cells after 24hr IFN-γ (10 ng/ml) stimulation. One-way ANOVA. Data are presented as the mean ± SD of measurements obtained in triplicate experiments. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

IFN-γ signals through the JAK1/JAK2 and STAT1/STAT2 signal transduction pathway (12). To assess this pathway in keratinocytes, we examined the activation of JAK/STAT in response to IFN-γ stimulation. Western blotting of IFN-γ stimulated keratinocytes showed that the expression of phosphorylated STAT1 (p-STAT1), p-STAT2, p-JAK1, and p-JAK2 were nearly undetectable at 0 min but increased over the time-course (Fig. 4D, Fig. S6B). Notably, weak phosphorylated JAK1 (p-JAK1) was seen at 15 minutes, where JAK2 phosphorylation was more prominent and peaked at the 90 min time-point. In contrast STAT1 phosphorylation (p-STAT1) showed early and sustained activation after 15 minutes, whereas STAT2 phosphorylation (p-STAT2) was weaker and only detectable after 60 min (Fig. 4D, Fig. S6B).

To examine the role of JAK/STAT signaling in regulating MHC class I expression in keratinocytes, we used the JAK1, JAK2, STAT1, and STAT2 CRISPR/Cas9 plasmids to knock-out (KO) each member of the JAK/STAT signaling machinery involved in IFN-γ responses (Fig. S7A). Absence of the targeted protein was verified by western blotting (Fig. S7B). Importantly, KO of JAK2 or STAT1 substantially inhibited the induction of MHC class I (HLA-A/B/C) in keratinocytes in response to IFN-γ stimulation compared to control keratinocytes (Fig. 4E). In contrast, minimal suppression of MHC class I expression was seen in JAK1 or STAT2 KO lines (Fig. 4E). In addition, the mRNA level of IFN-γ-responsive genes such as MX1, OASL, IRF7, and IRF9 were also decreased in JAK2 and STAT1 KO keratinocytes (Fig. S6C).

In summary, these observations demonstrate that IFN-γ induction of MHC class I expression in keratinocytes is largely JAK2/STAT1 dependent.

Targeting JAK signaling protects keratinocytes from cytotoxic responses

Next, we employed JAK2 or STAT1 KO keratinocytes in our model system to verify the roles of JAK2/ STAT1 signaling in priming keratinocytes for cytotoxic responses in LP. This result demonstrated that JAK2 or STAT1 KOs successfully protected IFN-γ-induced keratinocytes from cell-mediated cytotoxic responses (Fig. 5A). Notably, JAK1 and STAT2 KO demonstrated decreased cell-mediated cytotoxicity but much less than JAK2 or STAT1 KO did (Fig. S8A).

Fig. 5. Targeting JAK signaling protects keratinocytes from cell-mediated cytotoxicity.

(A) JAK2 and STAT1 KO cells were primed with IFN-γ and co-cultured with CD3/CD28 activated PBMCs. Keratinocyte cell-death was evaluated for Annexin-V positivity by flow cytometry. (B) The representative flow cytometry data and analysis of keratinocyte cell-death in the co-culture model with or without baricitinib. (C) The mRNA expression of MHC I molecules in IFN-γ-induced keratinocytes with or without baricitinib for 24 h. One-way ANOVA. Data are presented as the mean ± SD of measurements obtained in triplicate or quadruplicate experiments. *P < 0.05, ***P < 0.001, ****P < 0.0001.

To assess for enriched drug-targets amongst the genes differentially expressed in LP, we performed a drug target enrichment analysis as previously described by our group (13). Top predicted drug targets included the JAK-inhibitor tofacitinib (FDR p=5.4E-15), methylprednisolone (FDR p=3.7E-10), and chloroquine (FDR p=1.6E-08) (Table S2).

To further investigate the therapeutic potential of JAK blockade in LP, we used baricitinib. Baricitinib was selected instead of tofacitinib as it is more selective for JAK1 and JAK2, in contrast to tofacitinib, which is selective for JAK1 and JAK3 (14). IFN-γ-primed keratinocytes were treated with baricitinib during the IFN-γ priming (first 24 h) and during the 72 h co-culture of keratinocytes and PBMCs. Notably, baricitinib-treated keratinocytes were completely resistant to the cell-mediated cytotoxicity (Fig. 5B), and upregulation of MHC class I was inhibited by baricitinib in IFN-γ-activated keratinocytes, as demonstrated by qRT-PCR (Fig. 5C). Importantly, consistent with the qRT-PCR results, the western blot showed that the high protein level of MHC class I was induced by IFN-γ treatment for 24 h, which was dramatically inhibited by baricitinib (Fig. S8B). Together, these data suggest that drugs targeting JAK2/STAT1 signaling should be assessed for the treatment of LP.

High IFN-γ responses characterize diseases with prominent epidermal cell death

Diseases characterized by interface dermatitis (ID), such as LP, lupus erythematosus, graft-versus-host disease (GVHD), and Stevens-Johnson syndrome (SJS), share similar histopathological characteristics including keratinocyte apoptosis and inflammatory cell infiltration close to the basal membrane (15). However, it is unknown whether a common immune mechanism characterizes this type of inflammatory response in the skin. Thus, we investigated the overlap of diseases with interface dermatitis, including LP and lupus erythematosus, and compared against disease such as psoriasis, a T cell-dominant inflammatory disorder that lacks epidermal apoptosis. As shown by TUNEL staining assay, keratinocyte cell death in the epidermis was observed in skin lesions of LP and cutaneous lupus erythematosus (CLE), but not in psoriasis vulgaris (PV) and normal controls (NC) (Fig. 6A). To assess the nature of cell death in skin lesions of inflammatory diseases including LP and CLE, we performed tissue immunofluorescence demonstrating that the epidermal keratinocytes in LP and CLE skin lesions were mostly cleaved-caspase3 positive, compared with no staining for cleaved-caspase3 in epidermis of psoriasis vulgaris (PV) or normal skin. Moreover, both LP and CLE had positive staining for both p-RIP3 and p-MLKL (Fig. S4). This suggests, that features of both apoptosis and necroptosis are seen in the epidermis of LP and CLE, inflammatory dermatoses characterized by interface dermatitis.

Fig. 6. High IFN-γ responses characterizes diseases with prominent epidermal cell death.

(A) Representative TUNEL staining showing cell death in skin lesions of LP, CLE, psoriasis, and normal controls. DAPI was used for nuclear staining (scale bar, 100μm). (B) The mRNA expression of IFNG and MHC I molecules in chronic plaque psoriasis (PV) (n=12 or 7), LP (n=38 or 10), CLE (n=21 or 12) skin lesions, and healthy control skin (NC, n=11 or 15). One-way ANOVA. Data are presented as the mean ± SD of measurements obtained in each sample.(c) Representative immunofluorescence staining of MHC I and MHC II, STAT1 and p-STAT1 in PV and CLE skin lesions and normal controls. These staining were performed on more than 3 patient samples for each group. DAPI was used to highlight nuclear staining (scale bar, 100μm).

Moreover, our qRT-PCR analysis demonstrated that mRNA expression of IFNG and MHC I transcripts was much higher in LP and CLE than that in control or PV skin (Fig. 6B), which was also validated by western blot (Fig. S8C). Immunofluorescence staining for MHC molecules showed higher expression of MHC I in CLE compared to psoriasis or healthy control skin (Fig. 6C). Furthermore, STAT1 and p-STAT1 were all markedly observed in the epidermis of CLE, which was similar with the expression pattern of LP (Fig. 6C). Taken together, these data show that inflammatory skin diseases characterized by interface responses share a similar central IFN-γ immunological mechanisms.

Discussion

Here, we have described the global transcriptomic changes in LP and its variant hypertrophic LP and identified IFN-γ as the dominant inflammatory pathway in its pathogenesis. We demonstrate that IFN-γ may play a major role in priming keratinocytes towards CD8+ T cell mediated cytotoxic cellular responses, which is characteristic of LP lesions, through induction of MHC I expression. We further show the importance of JAK2/STAT1 signaling in this process and demonstrate that JAK inhibition inhibits these responses and therefore can be considered as a potential therapeutic option for the treatment of LP. Furthermore, we demonstrate that heightened IFN-γ expression and activity is a common feature for diseases characterized by an interface inflammatory reaction, broadening the implications of our findings.

IFN-γ is the only member of the type II interferon class with a critical role in both innate and adaptive immunity and has a wide-range of pro-inflammatory effects. It enhances expression of the chemokines CXCL9/10/11 to promote T cell infiltration (16). Furthermore, IFN-γ is an important activator of macrophages (17), and an inducer of MHC class II expression (18). IFN-γ can also mediate MHC class I antigen presentation in epithelia (19). While earlier reports claimed an important role for type I interferons in the pathogenesis of LP (20–22), the focus has recently shifted towards IFN-γ and its involvement in processes such as keratinocyte necroptosis (20). Furthermore, polymorphisms in the IFNG gene have been shown to be associated with oral lichen planus (23). However, the mechanism involved in IFN-γ driven pathogenicity have remained unclear.

LP is characterized by a cytotoxic immune response against keratinocytes of the basal layer (known as interface dermatitis), an immune response pattern not observed in psoriasis and atopic dermatitis (24). The etiology of LP is unclear, but many investigators consider LP to be an autoimmune disease (25, 26), particularly on the basis of its association with HLA-DR (27, 28). In addition, several reports have suggested that intracellular bacteria (29) or infected human papillomavirus (30) might be the source of such antigens. While our data is consistent with LP being driven by T-cell responses against a specific or limited set of antigens, our findings do not address the nature of this antigen, and additional studies are required to address this. Several clinical variants of LP are known but the most common ones are lichen planus and oral lichen planus, which can be a debilitating disease with extensive erosions in the mouth and esophagus (1). HLP is a subtype of lichen planus and characterized by marked hyperkeratosis, which is consistent with what we see in our transcriptomic analysis where HLP had greater enrichment for processes such as keratinization (Fig. 1C), but otherwise exhibited near identical features. Despite the well-known clinical manifestations of LP, the immune pathogenesis of this disease remains mostly unknown. One of the more characteristic features of LP is the dense T cell infiltration along the dermal-epidermal junction of the skin often obscuring the boundary between the dermis and the epidermis. This inflammatory pattern is not restricted to LP but is also found in other skin diseases where it is typically accompanied by mild to sometimes marked epidermal cytotoxic cell-death. This includes diseases such as cutaneous lupus (31), GVHD (32), and drug reactions such as Steven Johnson’s syndrome (33). Despite extensive investigation, the mechanisms involved in the interface inflammatory reaction have not been elucidated. It has been reported that T cells from LP lesions are more cytotoxic against autologous lesional keratinocytes than T cells from adjacent clinically healthy skin, and this could be partially blocked with anti-MHC class I antibody (34). Furthermore, obvious accumulation of Langerhans cells and T cells, particularly of CD4+ and CD8+ subsets, are found in both epidermis and dermis of LP lesions compared with non-lesional and healthy skin (35, 36). CXCR3+ cytotoxic T cells that can be attracted by CXCL9 and CXCL10 are reported as the major effector cells that located in the junctional zone of LP (37, 38). There CD4+ T cells may be activated by MHC class II positive cells, further activating cytotoxic CD8+ T cells, against epidermal keratinocytes (9), where they recognize antigens associated with MHC class I molecules, and subsequently triggering keratinocyte apoptosis via the perforin/granzyme pathway or the Fas/FasL pathway (36).

The manner of cell death of epidermal keratinocytes in these skin diseases is under debate. It is universally recognized that destruction of basal keratinocytes is due to apoptosis in LP (22, 39, 40). However, one study revealed that both cleaved caspase 3 positive cells, indicative of apoptosis, and p-RIP3 positive cells, consistent with necroptosis, can be observed in LP and CLE epidermis (20). Our findings are consistent with these observations with positive staining for markers of both apoptosis, as defined by presence of cleaved caspase 3 in the epidermis, and markers of necroptosis, including p-RIP3, and p-MLKL positivity in epidermal keratinocytes, suggesting that both of these cell death mechanisms contribute to inflammation in LP and a wide range of skin diseases (41).

Based on the above observations, MHC class I and II molecules, which are both overexpressed in LP lesions, may contribute to its pathogenesis. It has been reported that keratinocyte expression of HLA-DR, but not HLA-DP or -DQ by keratinocytes in LP may be induced by the lymphocytic infiltrates. We were unable to confirm these observations but found instead MHC class II molecules to be predominantly localized to the inflammatory infiltrate (Fig. 3). Notably, in our study we observe that blocking HLA-DR, but not HLA-DP or DQ, partially prevents keratinocyte cell death in our model system (Fig. 3), which is of interest as HLA-DR gene frequency is closely associated with LP in several populations (27, 28). In contrast to MHC class II, MHC class I blockade showed a much greater inhibitory effect on T-cell-mediated keratinocyte cell death, suggesting involvement of antigen presentation through MHC class I, and involvement of CD8+T cells, as suggested by our results (Fig. 2D, Fig. S1C). Such a scenario is consistent with our findings, given the strong induction of MHC class I by IFN-γ in keratinocytes in vivo, and the markedly increased expression of MHC class I molecules in LP epidermis. Besides increasing MHC class I molecules, IFN-γ is likely to also contribute by enhancing the activity of the antigen-processing machinery (42) in keratinocytes.

Strikingly, our data demonstrates the discordant role of JAK1 and JAK2 in regulation of MHC class I expression in keratinocytes. Whereas JAK1 KO had minimal to no effect on MHC class I expression, JAK2 KO had more marked effect on suppressing, although not completely inhibiting, the effect of IFN-γ on MHC class I expression. In contrast there was an absolute requirement for STAT1 for MHC class I expression, whereas there was minimal effect of STAT2 KO (Fig. 4). This matches the data from western blotting showing robust and sustained activation of JAK2/STAT1 than JAK1/STAT2 in keratinocytes. These data demonstrate the relative importance of the two JAKs and two STATs in type II responses, which is likely to have implications with the use of future and more selective JAK-inhibitors for inflammatory skin disorders.

Currently there are no FDA approved medications for the treatment of LP. Our observations support the use of JAK inhibitors for LP, and possibly other diseases characterized by interface reaction. The JAK/STAT pathway is down-stream of the IFN-γ receptor, and small molecules that inhibit JAKs can reduce IFN-γ-induced immune reactions (43). Our drug prediction analysis identified tofacitinib, a selective JAK1/JAK3 inhibitor, as the top therapeutic agent based on differential gene expression in LP skin. Several JAK inhibitors are currently available, and drug targeting information on these is still fairly limited. Given the dependency of IFN-γ on JAK1/JAK2 signaling, we selected baricitinib, an oral and reversible inhibitor specific to JAK1 and JAK2, as the drug of choice to test in our model system. Baricitinib was recently approved by the FDA for the treatment of moderately to severely active rheumatoid arthritis (RA) in adults (44), and has a favorable safety profile (45, 46). Baricitinib is a promising treatment for a wide-range of dermatologic diseases, based on evidence from ongoing clinicial trials, including SLE (47), psoriasis (48), atopic dermatitis (49), alopecia areata (50), and GVHD (51). Our data further indicate that baricitinib may become a promising biological treatment for LP and other diseases characterized by interface reaction, and a clinical trial utilizing JAK inhibitors in this setting is warranted.

One limitation of this work is the lack of an autologous in vitro system to test responses of T cells against keratinocytes. However, this is difficult in execution for several reasons, including difficulties in recruiting untreated patients for extensive skin sampling, growth inhibitory effect of IFN-γ on cell growth, use of T cell clones, and use of immortalized E6 and E7 virally transfected keratinocytes for immunological studies. However, our observations are in agreement with a previous study looking at in vitro cytotoxic activity of T cell clones from LP patients against immortalized autologous epidermal keratinocytes, where cytotoxicity was dependent upon CD8+ T cells (34). Furthermore, this study did not address the clinical efficacy of baricitinib or other JAK1/2 inhibitors in patients with LP. Future clinical studies are needed to address this.

In conclusion, our findings support targeting IFN-γ or its downstream signaling as a therapeutic strategy for LP. In addition, we provide evidence that diseases characterized by the interface inflammatory reaction may utilize IFN-γ as a common pathway, and therefore the implications of our findings are likely to apply to a wide range of inflammatory skin diseases.

Materials and Methods

Study design

The aim of this study was to demonstrate that IFN-γ plays a pathogenic role in LP and to validate the JAK as a therapeutic target. The first objective was illustrated by microarray profiling performed on skin biopsy samples demonstrating that IFN-γ and IFN-γ-regulated genes were enriched in the LP skin transcriptome, and by cell co-culture model showing that IFN-γ increased keratinocytes susceptibility to CD8+ T cell-mediated cytotoxicity. The second objective was investigated by in vitro cell co-culture model studies using CRISPR/Cas9 KO cells and JAK inhibitor baricitinib.

The microarray profiling experiment was carried out on RNA extracted from paraffin embedded skin biopsy samples obtained from patients with LP (n=20), HLP (n=17), and healthy controls (n=24). TUNEL staining images were quantified in a blinded fashion. Details of sample number and experimental replicates are provided in each figure legend.

Human Subjects

The study protocol was approved by the Institutional Review Board of the University of Michigan Medical School, and the study was carried out in accordance with the Declaration of Helsinki principles. All patients and controls gave written, informed consent. LP patients had both clinical and pathologic confirmation of diagnosis. Demographic information of patient and normal controls is provided in Table S3. Normal controls were recruited by advertisement.

Microarray analysis of skin lesions

Biopsies of LP and HLP cases were identified through the University of Michigan Pathology Database. Control blocks were obtained from healthy volunteers. Patients who met both clinical and histologic criteria for LP and HLP were included in the study. Validation of clinical and pathologic LP diagnosis was made via review of dermatology notes for each case. RNA was isolated from 10pm sections of formalin-fixed paraffin embedded blocks of identified skin biopsies. RNA was extracted using the E.N.Z.A. FFPE RNA Kit (Omega Bio-tek). Complementary DNA was prepared and biotinylated using the NuGEN Encore Biotin Module (Encore Biotin Module Manual, P/N M01111 v6). Labeled cDNA was hybridized at 48°C to Affymetrix Human Gene ST 2.1 array plates, which were then washed, stained and scanned using the Affymetrix GeneTitan system (software version 3.2.4.1515) with the assistance of the University of Michigan DNA Sequencing Core. Quality control and RMA (Robust Multi-array Average) 23 normalization of CEL files were performed in R software version 3.1.3 using custom CDF version 19 and modified Affymetrix_1.44.1 package from BrainArray (http://brainarray.mbni.med.umich.edu/brainarray/default.asp). Log2 expression values were batch corrected using Combat implemented into GenePattern (http://www.broadinstitute.org/cancer/software/genepattern/). The baseline expression was defined as minimum plus one standard deviation of the median of all genes. A variance filter of 80% was then applied. Literature pathway network analysis was performed using Genomatix Pathway System (GePS) (www.genomatix.de). A transcriptional network from the genes differentially regulated (FDR<0.1, FC>1.5 or <−1.5) as defined by our filter criteria was generated using the literature based Genomatix Pathway System software (GePS, www.genomatix.de). GePS allows visualization of dependencies among genes in pathways, networks and processes derived from literature-based knowledge and genome-wide sequence analysis. We applied a function-word filter, meaning that to be displayed in the transcriptional network, two genes have to be co-cited in the same sentence with a function word (e.g. gene A activates gene B).

Drug target enrichment analysis

We conducted the drug target enrichment analysis among the differentially expressed genes using an algorithm we described previously (13). We compiled the drug-gene relationships by combining interaction data from the Comparative Toxicogenomics Database (CTD) (52), DrugBank (53), and PharmGKB (54). We only used the drug targets expressed in our data, and removed drugs targeting more than 100 genes. We then used hypergeometric test to examine the statistical enrichment for differentially genes among each drug’s targets. False discovery rate <= 5% and observed/expected ratio >= 2 were used as criteria to nominate candidate drugs which have at least 5 drug targets overlapping with the differentially expressed gene list.

RNA extraction, qRT-PCR

RNA was isolated from cells using Qiagen RNeasy plus kit (74136). QRT-PCR was performed on a 7900HT Fast Real-time PCR system (Applied Biosystems) with TaqMan Universal PCR Master Mix (4304437, ThermoFisher Scientific). Primers (ThermoFisher Scientific) used in this study were: HLA-A; Hs01058806_g1; HLA-B; Hs00818803_g1; HLA-C; Hs00740298_g1; HLA-DPA1; Hs00410276_m1; HLA-DPB1; Hs03045105_m1; HLA-DQA1; Hs03007426_mH; HLA-DQB1; Hs03054971_m1; HLA-DRA; Hs00219575_m1; MX1, Hs00895608_m1; OASL, Hs00984387_m1; IRF7, Hs01014809_g1; IRF9, Hs00196051_m1; RPLP0, Hs00420895_gH.

Cell culture and stimulation

N/TERTs, an immortalized keratinocytes line (55), was grown in Keratinocyte-SFM medium (17005-042, ThermoFisher) supplemented with 30 μg/ml bovine pituitary extract, 0.2 ng/ml epidermal growth factor, and 0.3 mM calcium chloride. Cells at proper confluency were subsequently treated with recombinant human IFN-γ (10 ng/mL; R&D Systems), IFN-α (20 ng/mL, 14276, Sigma), and IFN-β (10 ng/mL, 8499-IF-010, R&D Systems) for indicated time prior to RNA and protein extraction.

Isolation of PBMCs

PBMCs were obtained from normal volunteers after obtaining informed consent. In short, the blood samples were diluted with 1× Hank’s buffer at 1:1 and underlayed with Histopaque®-1077 (10771, Sigma). The tube was centrifuged at 800×g for 20 min, and the lymphocyte interphase was carefully aspirated, collected, and washed with Hank’s buffer twice and centrifugated at 800×g for 5 min, and the cells were resuspended in Keratinocyte-SFM medium and added to the cell co-culture model.

Co-culture model of keratinocytes and PBMCs

Upon reaching semi-confluence, N/TERT cells (about 5 × 10⁴ per well) in 12 well plates were washed and stimulated with 10 ng/mL IFN-γ, 20 ng/mL IFN-α, or 10 ng/mL IFN-β for the first 24 h, followed by replacing the medium with 1 ml isolated and suspended PBMCs (5 × 10⁵/ml) with/without CD3/CD28 microbeads (130-091-441, Miltenyi Biotec; 1 ml CD3/CD28 microbeads for 1 × 10⁸ PBMCs) stimulation for 72 h. After the co-culture, N/TERT cells were then washed with PBS twice and collected for flow cytometry of Annexin V-PI analysis or TUNEL staining for the cell death. To block the functions of immune cells, isolated PBMCs were incubated with blocking antibodies targeting CD4 (1:100, 300516, BioLegend company), CD8 (1:3000, GTX19718, GeneTex company), and NK cells (1:100, 325104, BioLegend company) for 1 h, and then added together in the co-culture model for 72h. In another blocking experiments, the anti-HLA A/B/C (LS-C190912, LifeSpan BioSciences), HLA-DP (ab20897, Abcam), HLA-DQ (ab23632), HLA-DR (ab136320) neutralizing antibody at 10μg/ml and control IgG were added together with PBMCs and remained in the co-culture model for 72h. To inhibit the JAK signaling pathway in keratinocytes, baricitinib (10μM, 1187594-09-7, Sigma) was also added in the co-culture model.

Annexin V-PI

Apoptotic cells were detected by staining with FITC labeled Annexin-V and propidium iodide (PI) (BD Pharmingen) according to the manufacturers’ instructions. The data was statistically analyzed by FlowJo v10.

Generation of KO keratinocytes by CRISPR/cas9

Guide RNAs were developed using a web interface for CRISPR design (http://crispr.mit.edu). The pSpCas9 (BB)-2A-GFP (PX458) was a gift from Feng Zhang (48138, Addgene plasmid) and used as cloning backbone. The details about target sequences, scores of the sgRNA were as follows: JAK1: TGATCTTCTATCTGTCGGAC (Score 91); JAK2: CATTTCTGTCATCGTAAGGC (Score 90); STAT1: ATTGGGCGGCCCCCCAATAC (Score 93); STAT2: TGGCAGCAGTAGCTCGATTA (Score 92). We followed the CRISPR/Cas9 protocol as we previously published (56). The chromatograms for the KO cell lines were provided in Fig. S7A.

Immunostaining

Formalin fixed, paraffin-embedded tissue slides obtained from patients with LP, lupus, psoriasis, and normal controls were heated for 30 min at 60°C, rehydrated, and epitope retrieved with Tris-EDTA, pH 6. Slides were blocked, incubated with primary antibody STAT1 (9172), STAT2 (4594), p-STAT1 (9167), and p-STAT2 (4441) (all from Cell Signaling), and MHC I (M00194-1, Boster), MHC II (ab157210, Abcam), IFN-α (LS-C390384, LifeSpan BioSciences), IFN-β (PA5-20390, Invitrogen Antibodies), CD4 (300516, BioLegend company), CD3 (ab17143),CD8 (ab199016), IFN-γ (ab9657), cleaved Caspase-3 (ab13847), p-RIP3 (ab209384), p-MLKL (ab187091) (all from Abcam), and Rabbit IgG Isotype Control overnight at 4 °C. Slides were incubated with biotinylated secondary antibody (biotinylated goat anti-rabbit IgG Antibody, Vector Laboratories BA1000; biotinylated horse anti-mouse IgG Antibody, Vector Laboratories BA2000) and then incubated with fluorochrome-conjugated streptavidin. Slides were prepared in mounting medium with 4′,6-diamidino-2-phenylindole (DAPI) (VECTASHIELD Antifade Mounting Medium with DAPI, H-1200, VECTOR). Images were acquired using Zeiss Axioskop 2 microscope and analyzed by SPOT software 5.1. Images presented are representative of at least three experiments.

TUNEL assays

For TUNEL staining of cells, keratinocytes that co-cultured with activated PBMCs were washed gently with PBS for three times, and then fixed with 4% paraformaldehyde and permeabilized with 0.1% sodium citrate and 0.1% Triton X. DNA fragmentation was determined by TUNEL (12156792910, Roche Applied Science) as described by the manufacturer. Per cent TUNEL+cells were quantified using CellC (cells positive for Red (TUNEL) and DAPI staining/cells DAPI positive).

For skin biopsies, TUNEL staining was performed according to manufacturer protocol. In short, paraffin-embedded skin slides were dewaxed and rehydrated according to standard protocol and then treated with proteinase K solution. Slides were then treated with TUNEL reaction mixture in a humidified chamber followed by PBS washing. Then slides were mounted by DAPI. Images were acquired using Zeiss Axioskop 2 microscope and analyzed by SPOT software V.5.1. Images presented are representative of three experiments.

Western Blot

Total protein was isolated from cells using Pierce RIPA buffer (89900, ThermoFisher) with PMSF Protease Inhibitor (36978, Sigma) and PhosSTOP (04906845001, Roche), and run on precast gel (456-1094S, Bio-Rad). The membrane was blocked with 3% BSA and then probed by primary antibodies including STAT1 (9172), STAT2 (4594), p-STAT1 (9167), p-STAT2 (4441), JAK1 (3344T), JAK2 (3230T), p-JAK1(74129), p-JAK2 (4406T) (all from Cell Signaling), and β-Actin (A5441, Sigma), followed by secondary antibodies (anti-mouse or rabbit IgG, AP-linked Antibody, Cell Signaling), then washed for 3 times, and substrate added (45-000-947, Fisher Scientific). Membrane was scanned on Molecular Dynamics STORM 860 PhosphorImager (GE Health Care, STORM 860).

Statistical Analysis

For statistical analysis, data obtained from at least three independent experiments were performed using GraphPad Prism software versions 6 (GraphPad software). Statistical significance was determined using Student’s unpaired two-tailed t test or analysis of variance (ANOVA) as indicated in the legend (*P < 0.05, **P < 0.01, ***P < 0.001, ****P <0.0001). Flow cytometry data was analyzed using FlowJo v10. The number of sampled units, n, is indicated in the Figure legends. For microarray, false discovery rate (FDR) was used to control the multiple testing.

Supplementary Material

Figure S6

Fig. S6. IFN-γ induces keratinocytes to express MHCs through JAK2/STAT1 signaling.

Figure S7

Fig. S7. Chromatograms and validation for JAK1, JAK2, STAT1, and STAT2 KO cell lines.

Figure S8

Fig. S8. Targeting JAK signaling protects keratinocytes from cell-mediated cytotoxicity.

Figure S5

Fig. S5. Cytotoxic responses to IFN-γ-primed keratinocytes are MHC class I dependent.

Figure S2

Fig. S2. IFN-γ increases keratinocyte susceptibility to cell-mediated cytotoxicity.

Figure S4

Fig. S4. Epidermal keratinocyte death in LP skin lesions is both apoptosis and necroptosis.

Figure S1

Fig. S1. Transcriptomic profiling analysis and tissue immunofluorescence of IFNs expressions in LP/HLP lesions.

Figure S3

Fig. S3. Keratinocvte death in in vitro co-culture system is mediated by apoptosis and necroptosis.

Table S1

Table S1. The list of differentially expressed genes based on microarray data in skin lesions.

Table S2

Table S2. Drug target enrichment analysis amongst the genes differentially expressed in LP.

Table S3

Table S3. Skin donor demographics.