TNF‐α induced extracellular release of keratinocyte high‐mobility group box 1 in Stevens‐Johnson syndrome/toxic epidermal necrolysis: Biomarker and putative mechanism of pathogenesis

Abstract Decreased epidermal high‐mobility group box 1 (HMGB1) expression is an early marker of epidermal injury in Stevens–Johnson syndrome/toxic epidermal necrolysis (SJS/TEN). Etanercept, an anti‐tumor necrosis factor therapeutic, is effective in the treatment of SJS/TEN. The objective was to characterize antitumor necrosis factor‐alpha (TNF‐α)‐mediated HMGB1 keratinocyte/epidermal release and etanercept modulation. HMGB1 release from TNF‐α treated (± etanercept), or doxycycline‐inducible RIPK3 or Bak‐expressing human keratinocyte cells (HaCaTs) was determined by western blot/ELISA. Healthy skin explants were treated with TNF‐α or serum (1:10 dilution) from immune checkpoint inhibitor‐tolerant, lichenoid dermatitis or SJS/TEN patients ± etanercept. Histological and immunohistochemical analysis of HMGB1 was undertaken. TNF‐α induced HMGB1 release in vitro via both necroptosis and apoptosis. Exposure of skin explants to TNF‐α or SJS/TEN serum resulted in significant epidermal toxicity/detachment with substantial HMGB1 release which was attenuated by etanercept. Whole‐slide image analysis of biopsies demonstrated significantly lower epidermal HMGB1 in pre‐blistered SJS/TEN versus control (P < 0.05). Keratinocyte HMGB1 release, predominantly caused by necroptosis, can be attenuated by etanercept. Although TNF‐α is a key mediator of epidermal HMGB1 release, other cytokines/cytotoxic proteins also contribute. Skin explant models represent a potential model of SJS/TEN that could be utilized for further mechanistic studies and targeted therapy screening.


| INTRODUC TI ON
Stevens-Johnson syndrome/toxic epidermal necrolysis (SJS/TEN) is a rare, immune-mediated, cutaneous blistering condition, most often caused by drugs and characterized by widespread keratinocyte death and epidermal detachment. Although much is known about the immunopathogenesis of SJS/TEN, gaps in our knowledge of how cell death and epidermal detachment occur still exist.
Previous work 1 demonstrated that epidermal expression of highmobility group box 1 (HMGB1) protein is able to discriminate between severe cutaneous adverse drug reactions (ADRs) (SJS/TEN) and maculopapular exanthema. HMGB1 is a damage associated molecular pattern (DAMP) protein which, in its unacetylated form, acts as a marker of both sterile toxicity and, in its acetylated form, as a marker of innate immune response. 2 Etanercept is an antitumor necrosis factor-alpha (TNFα) monoclonal antibody used for treating auto-inflammatory conditions including psoriasis and rheumatoid arthritis. A number of case studies have also suggested that it is an effective treatment for SJS/TEN 3 and reduces mortality versus current standard of care (high-dose prednisolone). 4 Despite reports of rapid re-epithelialisation in response to etanercept, 5 there is currently limited understanding of the specific mechanism of action of etanercept in SJS/TEN treatment. Studies suggest that T-cell derived TNFα contributes to keratinocyte cell death via Fas ligand. 6 However, necroptosis is the key cell death mechanism responsible for the significant keratinocyte death seen in SJS/TEN 7,8 but its modulation by TNFα is yet to be established. Furthermore, the role of TNFα in mediating keratinocyte HMGB1 release and the effect of etanercept is not understood.
The aim of this study was to determine if TNFα-mediated keratinocyte cell death is associated with HMGB1 release, both in vitro and ex vivo, and whether this could be ameliorated by etanercept treatment. Furthermore, digital pathology methodologies have been utilized to quantitatively assess previously reported decreased SJS/ TEN epidermal HMGB1 expression. Three patients were identified (1 × SJS/TEN, 9 1 × lichenoid dermatitis, and 1 × tolerant control; Supporting Information Table S2) whose sera were utilized for exposure to healthy skin explants. Serum TNFα concentrations were determined by enzyme-linked immunosorbent assay (ELISA; R&D Systems) according to the manufacturer's protocol.

| Cutaneous ADR skin biopsies
Formalin-fixed paraffin-embedded skin samples were identified from the histology archive database at Cleveland Clinic, OH, USA, from 2013 to 2020 using criteria previously described. 1 Briefly, an internal diagnosis or description search included the terms "Stevens-Johnson Syndrome" or "toxic epidermal necrolysis" and drug eruption/drug reaction (including "dermal hypersensitivity reaction"). Cases were selected where a diagnosis of drug-induced SJS/TEN or maculopapular exanthema was very strongly favored and was supported by clinical notes. Normal skin from excision specimens was utilized as healthy control skin. Suspected causal drugs were identified from clinical notes (Supporting Information Table S2).  Table S1). Sfil restricted fragments were ligated into pSB-Tet Blast, which contains a tetracycline inducible expression cassette flanked by Sleeping Beauty transposon inverted repeat/direct repeats (IR/DRs) containing ampicillin resistance and blasticidin selection markers.
Optical density at 595 nm was determined by multimode plate spectrophotometry (Beckman Coulter). Cell viability was expressed as a percentage normalized to untreated controls. Details of the flow cytometric analysis of cell death can be found in the Supporting Information.   Table S3. Membranes were developed and visualized using enhanced chemiluminescent substrate (ECL) and a Chemidoc Touch Imaging System (both Bio-Rad).

| Western blotting
Image densitometric analysis was undertaken using Image J software.

| Skin explants
As previously described and published, 10 4-mm biopsies were taken from healthy skin tissue (Tissue Solutions Ltd). Anakinra was utilized as a positive control as previous observations suggest that at 10 μg/ mL it induces a toxicity and epidermal separation phenotype in explants akin to SJS/TEN. This is likely comparable to immediate injection site reactions where high local anakinra concentrations induce mast cell degranulation 11 and can lead to TNF-induced neutrophil infiltration 12 and tissue injury. Explants were cultured for 72 h in 200 μL of media (X-VivoTM 10, Lonza) ±1 μg/mL etanercept (Benepali, Biogen Inc.) containing (i) single patient sera (SJS/TEN, lichenoid dermatitis, or control) diluted 1:10, (ii) 10 ng/mL TNFα (R&D systems), or (iii) 10 μg/mL anakinra (n = 3 for all groups; Sigma-Aldrich). A single replicate was snap-frozen and stored at −80°C in optimal cutting temperature media with the other two formalinfixed and paraffin-embedded for H&E staining and immunohistochemistry. Histological damage was graded according to the Lerner criteria. 13 Briefly, these are:
Supernatant total HMGB1 protein concentrations were determined by ELISA according to the manufacturer's protocol (IBL International GmBH). Samples failing to reach the manufacturer's lower limit of quantification (0.2 ng/mL) or where replicates were discordant by >15% were excluded. For explants treated with human sera, input HMGB1 (how much was in the original serum sample) was determined and subtracted from the final concentration. Detailed protocols for western blot analysis are available in the Supporting Information. Visualization was with diaminobenzidine (Agilent Technologies Ltd). Consecutive sections incubated with nonimmune rabbit serum served as negative controls. Positive reaction was represented by a distinct brown nuclear (or rarely cytoplasmic) reaction. Positive control was represented by epidermis and follicle in normal skin.

| Analysis of whole-slide images
HMGB1-stained slides were scanned using a Leica Aperio CS2 scanner at 40× magnification. Whole-slide image analysis was undertaken using QuPath software. 14 Epidermis was manually separated from dermis, annotating the area of interest, and appropriate classification attributed to each annotation. Positive cells were identified as those with an evident brown precipitate due to a 3,3'-Diaminobenzidine

| Statistical analysis
For serum and supernatant HMGB1 concentrations, a one-way analysis of variance with Bonferroni correction was undertaken. For western blot densitometry, a Mann-Whitney U test was utilized. All analyses were performed using Prism 5 software (GraphPad Inc.).

| TNFα induced HMGB1 release in vitro
Our initial work aimed to characterize TNFα induced cell-death and Inhibition of RIPK1 (necroptosis) by necrostatin resulted in a significant decrease in TNFα induced extracellular HMGB1 (P < 0.05) which was greater than that for zVAD, with use of both resulting in attenuated HMGB1 release (P < 0.01; Figure 1d).

Having established the contribution of necroptosis apoptosis to
HaCaT HMGB1 release in a TNFα exposed model, we next wanted to observe it in a "cleaner" model of both cell death mechanisms. In RIPK3 overexpressing cells increased MLKL expression was additionally seen (Figure 2c). Extracellular HMGB1 expression was significantly induced in all three dox-inducible cell-lines (compared to wild-type, P < 0.01), with MLKL expression appearing to exhibit the highest levels (Figure 2d).

| TNFα induced HMGB1 release and modulation by etanercept
Having establish a predominant role for necroptosis, in both our TNFα exposed and cell-death inducible models, we next wanted to quantify HMGB1 release and assess the effect on it of etanercept  Figure S3) were significantly higher in TNFα treated explants (7.83 ng/mL ± 1.2) versus controls (1.03 ng/mL ± 0.14, P < 0.01). Etanercept co-exposure with TNFα did not significantly alter HMGB1 levels (10.87 ng/ mL ± 1.70, P > 0.05).
Anakinra (positive control)-treated explants exhibited grade II/III toxicity or grade II in the presence of etanercept (Supporting Information Figure S3). HMGB1 supernatant levels were significantly higher in response to anakinra (59.64 ng/mL ± 20.17, P < 0.05)

F I G U R E 1 A combination of TNFα and BV-6-induced HMGB1 release in HaCaTs which is modulated by inhibitors of both necroptosis (necrostatin) and apoptosis (Z-VAD-FMK). (a) HaCaT cell viability determined by MTT assay, (b) representative flow cytometric scatter plot of AV-and PI-stained cells with (c) percentage cells positive stained and (d) extracellular HMGB1
following 24 h TNFα exposure ±40 μM necrostatin, 50 μM ZVAD (exemplar western blot image selected from n = 3). Densitometric analysis was performed on three separate blots and statistical analysis was undertaken using the Mann-Whitney U test (*P ˂ 0.05 and **P ˂ 0.01). (e) Supernatant HMGB1 levels determined by ELISA and (f) corresponding HaCaT viability by MTT assay (with BV-6 pre-treatment). Data represent mean normalized to untreated control (±SE) of three separate experiments conducted in triplicate (*P ˂ 0.05, **P ˂ 0.01, ***P < 0.001 and # < 0.05 vs. TNF-a/BV-6 treated cells). compared to untreated controls. Co-administration of etanercept did not result in a significant change in HMGB1 concentrations in response to anakinra (12.13 ng/mL ± 5.54; ns).
RIPK3 expression in the basal layer was not significantly higher in TNFα-treated explants compared to controls (Supporting Information Figure S1). However, there was a noticeable reduction in RIPK3 expression in explants (control and TNFα-treated) co-administered etanercept. No significant expression of cleaved caspase 3 was observed in any of the treatment groups (Supporting Information Figure S1).

| HMGB1 expression and supernatant release in healthy skin explants exposed to SJS/TEN patient serum
Having established the effect of TNFα on epidermal toxicity and HMGB1 release, the ex vivo work was expanded to look at the effect of exposing healthy skin to sera from healthy control, lichenoid dermatitis and SJS/TEN patients (acute phase). TNFα concentrations of the three sera were determined by ELISA as healthy control and was visibly lower in TNFα and SJS/TEN serum-treated explants compared to control, healthy, or lichenoid dermatitis serum treatment (Supporting Information Figures S1 and S2).
In addition to in situ skin HMGB1 expression, we also examined extracellular HMGB1 in the explant culture media (Supporting Information Figure S3). This was found to be significantly higher in explants treated with SJS/TEN patient serum (16.91 ng/mL ± 3.51) compared to the media-only control (1.03 ng/mL ± 0.14, P < 0.05).
There was no significant difference in those co-administered etanercept (19.00 ng/mL ± 2.62; P < 0.05). Explants treated with sera from either the tolerant or lichenoid dermatitis patient (4.36 ng/mL ± 1.83 and 1.82 ng/mL ± 0.22, respectively) did not demonstrate higher mean supernatant HMGB1 levels compared to control explants (ns) and there was no significant difference from either when coadministered etanercept (ns).

| Quantitative digital image analysis of HMGB1 skin expression
Previous work 1 , utilizing semiquantitative analysis of epidermal HMGB1 expression, showed a significant decrease in the SJS/ F I G U R E 3 Effect of 72 h of 10 ng/mL TNFα or cutaneous ADR patient serum (control, lichenoid dermatitis or SJS/TEN) exposure ±1 μg/ mL etanercept on healthy skin explant morphology (H&E stained). Images show HMGB1, immunohistochemical expression, and localization. Images are representative of n = 3 skin sample. For H&E images: black horizontal scale bar = 100 μm (400× magnification). For HMGB1: black horizontal bar =60 μm (zoomed 40× whole slide scanned image). Dashed lines represent the dermal epidermal junction. TEN epidermis. We therefore used image analysis to quantify dermal and epidermal cellular HMGB1 expression in healthy, drug-induced maculopapular exanthema, and SJS/TEN skin.

| DISCUSS ION
The findings from this study show a clear link between TNFα-induced cell death (predominantly necroptosis) and HMGB1 release in keratinocytes (Figures 1 and 2). Our previous observations 1 showed that HMGB1 is released from keratinocytes in SJS/TEN even prior to epidermal detachment. The suggestion that RIPK3 necroptotic keratinocyte death may be a predominant source of HMGB1 (Figures 1 and   2) is consistent with previous literature suggesting that necroptosis is a key mediator of epidermal injury in severe skin-blistering ADRs. 7,8 Our finding that TNFα plus BV-6 elicits keratinocyte cell death (and Interestingly, loss of HMGB1 expression appears to be a more sensitive marker of epidermal injury than both RIPK3 and caspase expression in explants treated with either TNFα or SJS/TEN serum ( Figure 3 and Supporting Information Figures S1 and S2). The data suggest that both TNFα and SJS/TEN serum lower RIPK3 expression and may be negatively correlated to HMGB1 expression in the epidermis. This is hypothesized to be due to the RIPK3 antibody used not cross-reacting with the phosphorylated RIPK3 formed during necroptotic cell death. 18 The lack of significant epidermal staining for cleaved caspase 3 is intriguing and seems to suggest that negligible apoptotic cell death is occurring. This supports the in vitro evidence that the predominant contributor to epidermal HMGB1 release is necroptotic cell death. We also observe apparent increased levels of HMGb1 expression in immune cells (Figure 3), which could also account for increased extracellular levels. This is supported by the observation that TNFα and SJS/TEN serum increased the num-  Information Table S4), which leads to elevated supernatant levels (Supporting Information Figure S3). However, whilst the addition of etanercept negated epidermal release, it did not reduce extracellular levels (Supporting Information Figure S3). TNFα is known to induce HMGB1 release from monocytes, 20 and it is therefore possible that the extracellular levels of HMGB1 in SJS/ TEN are due to TNFα-induced release by both epidermal and immune cells, and are not only due to epidermal loss of expression.
An alternative hypothesis is that etanercept itself increases HMGB1 release from activated immune cells 21 which counters the inhibition of HMGB1 release from injured keratinocytes by etanercept. This also suggests that etanercept has an effect on both inflammatory cells and keratinocytes. This is also true of anakinra, 21 which would explain the very large increase in HMGB1 that is seen and the lack of reduction in the presence of etanercept (Supporting Information Figure S3).
Epidermal expression of HMGB1 was significantly decreased by exposure to SJS/TEN patient serum at 1:10 dilution ( Figure 3) and marginally restored by etanercept. This mirrored the effect on toxicity where a marginal decrease was seen in response to etanercept (Supporting Information Table S4 and Figure 3). This suggests that TNFα might play a role in mediating HMGB1 release and toxicity in SJS/TEN but HMGB1 release is also attributable to non-TNFα-mediated pathways. A limitation of our studies is the small number of patient samples tested so data should be assessed with caution. However, the results obtained were quite distinct between sera obtained from the SJS/TEN patient, and sera from the tolerant control and the patient with maculopapular exanthema.
Further work will be needed to assess interindividual variability in response across multiple different explants and serum samples. It should be noted that the SJS/TEN case from which sera was used was an ICI-treated patient. However, previous work has shown that the reaction was due to exposure to iodinated contrast media precipitated by ICI therapy rather than the actual ICI 9 so the pathogenesis of this case is akin to a "normal" small-molecule-induced SJS/ TEN phenotype.
The clinical benefit of TNFα inhibition by etanercept in treating SJS/TEN has been demonstrated in multiple case reports 3 and in a randomized control trial. 4 However, the actual mechanism of action of etanercept in SJS/TEN has not been established. We were unable to assess the role of etanercept on pathogenesis in the model in isolation or co-administered with anakinra, which is a limitation although we would not expect any significant modulation in either scenario. Our findings suggest that etanercept, whilst attenuating TNFα-induced keratinocyte death, also significantly reduces HMGB1 release (Figure 1). The suggestion that TNFα has a role in the SJS/TEN pathogenesis is not new 22,23 but the link to HMGB1 is novel and could represent a new insight into the downstream effects of TNFα and the mechanism of action of etanercept in SJS/ TEN. It should be noted that the action of TNFα on keratinocytes in this model is not specific and indeed it is likely that, as part of the observed pathogenesis, it is mediating its effect via skin-resident CD8+ T-effector cells. 24 We used digital pathology to build on previous observations of decreased epidermal HMGB1 expression in SJS/TEN skin. 1 A statistically significant decrease in HMGB1 positive cells was observed in pre-blistered SJS/TEN versus healthy skin (Figure 4), although the difference between maculopapular exanthema and SJS/TEN serum did not reach significance. The data confirm that HMGB1 nuclear > cytosol > extracellular translocation is indicative of early epidermal stress in SJS/TEN.
The data further underline the potential utility of epidermal HMGB1 release as an early marker of epidermal cell death, and potentially detachment as we have previously proposed. 1 HMGB1 may also, through its isoform-dependent immunomodulatory functions, 2 exacerbate tissue damage. This will require further study. The data presented further suggest that HMGB1 keratinocyte release may be a viable proxy biomarker for the onset of SJS/TEN and that, at least in keratinocytes, it is related to both necroptotic and apoptotic cell death. Explant data suggest that epidermal HMGB1 may be a more sensitive marker than both RIPK3 and cleaved caspase, although there is a suggestive correlation between HMGB1 and RIPK3. Future work will look to further establish this is both treated explants and clinical biopsy samples.
We have shown that the exposure of healthy skin explants to both TNFα and serum from an SJS/TEN patient was able to evoke a skin phenotype akin to SJS/TEN in both morphology and HMGB1 expression. We have also shown that etanercept was able to attenuate toxicity and, at least for TNFα, reverse HMGB1 cellular release in the epidermis. The lack of effect of etanercept in SJS/TEN serumtreated explants is likely to be due to the multitude of other molecules present which exert cytotoxic effects independently of TNFα, in keeping with the fact that the pathogenesis of SJS/TEN is multifaceted and complex. However, this physiologically-relevant skin explant model has the potential to be used for assessing the pathogenic role and mechanism of other immune-derived cytotoxic mediators implicated in SJS/TEN, such as granulysin, perforin, granzyme B, and LL37. [25][26][27][28] Furthermore, by inducing an SJS/TEN-like event in healthy skin explants with TNFα, and inhibiting it with etanercept, we have a valid model system with the potential to be used for screening targeted therapies for potential therapy in SJS/TEN. The results presented here have be interpreted with some caution. Given the small sample size, it is not possible to determine interindividual variability in response, both between skin donors and between serum samples. This will be the focus of future work.
To conclude, we have demonstrated that keratinocyte-derived HMGB1 release is a useful biomarker for keratinocyte injury in early SJS/TEN. The skin explant model described may be a useful model to identify not only other proteins involved not only in causing epidermal damage, but also in screening novel therapies to treat SJS/TEN. This will, however, need further validation.