Clinical parameters and biological markers associated with acute severe ocular complications in Stevens-Johnson syndrome and toxic epidermal necrolysis

Stevens-Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN) are severe cutaneous adverse drug reactions with high mortality rates. Its sequelae, such as blindness, persist even after recovery. Patients with SJS/TEN should be accurately diagnosed and receive appropriate treatment as soon as possible. Therefore, identifying the factors for severity prediction is necessary. We aimed to clarify the clinical parameters and biological markers that can predict acute severe ocular complications (SOCs) in SJS/TEN. This retrospective cross-sectional study enrolled 47 patients with SJS/TEN who were divided into two groups according to ocular severity at acute onset: non-severe ocular complications group (n = 27) and severe ocular complications group (n = 20). Multivariate logistic regression analysis revealed that disease severity (body surface area detachment ≥ 10%) was a predictive factor for acute SOCs, and older age (≥ 60 years) was marginally significantly predictive of SOCs. Serum biomarker levels of S100A8/A9 and granulysin were marginally significant and tended to increase in the SOC group. Therefore, during the early acute stage, focusing on disease severity, patient age, and serum inflammatory biomarkers (S100A8/A9 and granulysin) might help predict SOC progression in patients with SJS/TEN who need prompt and aggressive ocular management to prevent severe ocular sequelae.


Results
Patient characteristics and causative drugs of SJS/TEN. The patient characteristics are summarized in Table 1. From a total of 47 SJS/TEN cases involved in the study, 35 (74.47%) were classified as SJS, six (12.77%) as SJS/TEN overlap, and six (12.77%) as TEN. Forty-three of the 47 patients (91.49%) presented with ocular involvement in the acute phase. The acute ocular severity score ( Table 2) was grade 0 in four cases (8.51%), grade 1 in 11 cases (23.40%), grade 2 in 12 cases (25.53%), and grade 3 in 20 cases (42.55%). We classified patients with acute ocular complications into two groups: non-severe and severe. Twenty-seven patients with ocular severity grades 0-2 were included in the non-severe group, and the remaining 20 patients were included in the severe group (ocular severity grade 3). The mean SCORTEN score (mean ± SD) was 1.80 ± 1.27. Considering the SCORTEN score, the SOC group showed higher scores than the non-severe group (2.25 ± 1.37 vs. 1.44 ± 1.08). The mean initial visual acuity at the onset of SJS/TEN (logMAR_ ± SD) was comparable in both groups (0.37 ± 0.53 and 0.46 ± 0.57 in the non-severe group and severe group, respectively).

Analysis of serum biomarkers between patients with SJS/TEN and healthy controls.
The geometric means and the geometric mean ratios (GMR) of serum biomarker levels between healthy controls (HC) Table 1. Analysis of the correlations between patient characteristics (demographic and clinical parameters) and acute severe ocular complications in SJS/TEN patients. BSA body surface area, SJS Stevens-Johnson syndrome; TEN toxic epidermal necrolysis; SCORTEN Severity-of-Illness Score for toxic epidermal necrolysis; VA visual acuity; SD standard deviation; LogMAR logarithmic minimum angle of resolution; HIV human immunodeficiency virus; NSAIDs nonsteroidal anti-inflammatory drugs; DMARDs disease-modifying antirheumatic drugs. † Oral mucosa, ocular mucosa, genital mucosa and other mucosa. † † Causative drug. ‡ n = 45. * P < 0.1; **P < 0.05. www.nature.com/scientificreports/ and patients with SJS/TEN with different ocular severities are shown in Table 4 and Fig. 1. After excluding seven patients with sepsis to rule out possible confounders, we compared biomarker levels in the serum of patients with SJS/TEN (N = 40) with HC (N = 18). Compared to the serum from the HC group, the SJS/TEN group serum showed significant upregulation of 11 factors. Three biomarkers had P values < 0.001, including IP-10, IL-6, and IL-17A. Two biomarkers (SCF and S100A8/A9) had P values < 0.01, and five biomarkers (MCP-1, ICAM-1, PDGF-AA, CRP, OPN, and PDGF-BB) had P values < 0.05.
Subgroup analysis of serum biomarkers between non-severe and SOC in SJS/TEN. We have analyzed different groups of SJS/TEN to compare both non-severe (n = 24) and severe (n = 16) ocular complications (Table 4 and Fig. 1). Although there was no statistically significant difference between the two groups, many cytokines were upregulated in the severe ocular involvement group. Among the biomarkers that were upregulated in the severe group, S100A8/A9 (P = 0.067, 95% CI 0.95-4.61) and GYLN (P = 0.065, 95% CI 0.97-2.84) were marginally significant and tended to increase in the severe groups.

Discussion
Acute SOCs in SJS/TEN often lead to chronic SOCs [10][11][12][13] . A recent study from Taiwan found a positive correlation between acute SOCs and chronic SOCs grading score among SJS/TEN patients 14  and GYLN might play a role in the pathogenesis of ocular complications in SJS/TEN by finding that the corneal epithelium induced by TNF-α can produce IL-13 and GYLN in a dose-response relationship 25 .
From the univariate analysis, our results revealed that BSA detachment ≥ 10%, older age (≥ 60 years), antibiotics as causative drugs, and heightened SCORTEN were associated with acute SOCs. Among these factors, only BSA (detachment ≥ 10%) was identified as a predictive factor for acute SOCs in the multivariate analysis. Yip 29 and Sotozono et al. 27 reported no association between BSA detachment and SOC. Interestingly, Yip et al. used the old ocular severity grading, combining both acute and chronic ocular findings into the same analysis, which might not represent the acute phase in SJS/TEN 30 . In contrast, our study uses the severity score, which includes only ocular findings in the acute phase 27 . Furthermore, our study included cases with AOS grade 3 in the SOCs group compared to the study by Sotozono et al. 27 , where patients with both AOC grades 2 and 3 were enrolled in the SOC group, leading us to find a correlation between the disease severity and acute SOCs in this study. This association could be due to similar mechanisms of apoptosis occurring in the skin and eyes 29,31,32 . Table 4. The comparison of serum biomarkers levels between healthy controls and SJS/TEN patients with different ocular severity. (A) Comparison between healthy controls group and SJS/TEN group. (B) Subgroup comparison between non-severe and severe ocular complications group. GM geometric mean; GMR geometric mean ratios; %CV percent coefficient of variation; IP-10 interferon-γ-inducible protein 10; IL-6 interleukin 6; S100A8/A9 heterodimeric of S100 calcium binding protein A8 and S100 calcium binding protein A9; SCF stem cell factor; GYLN granulysin; CRP C-reactive protein; OPN Osteopontin; ICAM 1 intercellular adhesion molecule 1; PDGF AA platelet-derived growth factor AA; PDGF BB platelet-derived growth factor BB; MCP monocyte chemotactic protein 1. † GMR in column A = the quotient of SJS/TEN and Healthy controls geometric means. † † GMR in column B = the quotient of SJS/TEN with severe ocular complications and SJS/ TEN with non-severe ocular complications geometric means. *P < 0.05, **P < 0.01, ***P < 0.001. Previous studies have shown that younger age (< 50 years) is associated with acute SOCs 27,33 . They hypothesized that not only younger age but also medication for common cold and viral infections causing cold-like symptoms play an essential role in the development of SOCs 27,33 . In contrast, we found that older age and antibiotic use were associated with acute SOCs in our study; cold medications (CM) were not a risk factor for SOCs. The explanation for the disparity between the present study and the aforementioned Japanese studies might be the difference in rational drug use, which cold remedies are used less often in our country. Genetic diversity may also play an important role. Moreover, the criteria for identifying causative drugs differed between studies. Although Thailand and Japan are both located in East Asia, associations between HLA genotype and cold medicine-related SJS/TEN with SOCs in Thailand have been shown to differ from Japanese patients 34 . Besides, the incidence of CM-related SJS/TEN in our study was relatively low (n = 1, 2%); therefore, our sample size might have been inadequate to obtain statistical significance.

Comparison of biomarkers between Healthy controls group and SJS/TEN group
Similar to earlier reports, we did not find an association between SCORTEN and SOCs 7,29 . SCORTEN tends to use clinical parameters that reflect overall patients' systemic conditions to predict the severity and morbidity. Since the ocular surface has a unique and privileged immune response, the discordance between ocular and systemic manifestations might explain why this score was not directly related to the severity of ocular manifestations in acute SJS/TEN. Different inflammatory mediators and cell types have been proposed to regulate the immunopathology of SJS/ TEN. In our study, S100A8/A9 levels were upregulated in patients with SJS/TEN compared to those in the control group and tended to increase in patients with severe ocular involvement. The S100A8/A9 complex is a critical alarmin that is upregulated in numerous inflammatory diseases such as rheumatoid arthritis, chronic inflammatory bowel disease, psoriasis, systemic lupus erythematosus, and atopic dermatitis [35][36][37] . Currently, no data are available on the potential function of S100A8/A9 in SJS/TEN. S100A8/A9 is expressed in monocytes, which   Figure 1. Scatter plot graphs of serum biomarker levels (geometric means) in healthy controls, and those from patients with SJS/TEN with non-severe and severe ocular complications. Representative scatter plot graphs of IP-10, IL-6, S100A8/A9, PDG-AA, PDG-BB, and granulysin levels. The geometric means with the 95% confidence interval of each biomarker in healthy controls (HC), SJS/TEN with non-severe ocular complications, and SJS/TEN with severe ocular complications group are shown. The Mann-Whitney test was used to compare biomarker levels between the HC and SJS/TEN groups and between the non-severe and severe ocular complications groups. IP-10, IL-6, S100A8/A9, PDG-AA, and PDG-BB were significantly higher in SJS/TEN group than in HC. When comparing the non-severe and severe ocular complication groups, any biomarkers showed significant differences. S100A8/A9 and granulysin were marginally significant and tended to increase in the severe ocular complications group. IP-10 interferon-γ-inducible protein 10; IL-6 interleukin 6; S100A8/A9 heterodimeric of S100 calcium binding protein A8 and S100 calcium binding protein A9; PDG-AA PLATELETderived growth factor AA; PDGF-BB platelet-derived growth factor BB. *P < 0.05, **P < 0.01, ***P < 0.001. www.nature.com/scientificreports/ are present in the epidermis of SJS/TEN skin lesions 38 , and can also be produced by epidermal keratinocytes 39 . While previous studies have implicated a role for S100A8 and S100A9 in causing accelerated inflammation 40,41 , opposite effects were reported in models of type IV-c hypersensitivity reaction 42 . S100A8/A9 inhibits dendritic cell maturation and antigen-presenting ability, leading to decreased T-cell activation and reduced intensity of immune responses in contact dermatitis 42 . Due to contradictory evidence on whether S100A8/A9 amplifies inflammation or exhibits an anti-inflammatory effect, the role of S100A8/A9 expression in the pathogenesis of SJS/TEN needs to be further elucidated. S100A8 and S100A9 are found to be upregulated in the epidermis of the skin during the active epidermal regeneration process 43 , and S100A8/A9 could alter the skin barrier proteins in atopic dermatitis 36 . In the present study, the elevation of serum levels of S100A8/A9 possibly arose from epithelial damage to the skin in SJS/TEN. Further studies are required to investigate S100A8/A9 expression and function in SJS/TEN skin lesions.
In the eye, S100A8/A9 enhances inflammation in the ocular surface by promoting PMN infiltration and the expression of IL-1b, IL-6, and TNF-α, as an inflammatory amplifier 44,45 . Moreover, there is evidence that elevated serum S100A8/A9 levels correlate with clinically active joint and eye inflammation in autoimmune uveitis 46 . As primary ocular manifestations in acute SJS/TEN are intense ocular inflammation and epithelial sloughing, S100A8/A9 probably plays a role in ocular inflammation during the acute phase of SJS/TEN.
Our study demonstrated the elevation of several mediators in the serum of patients with SJS/TEN compared to those in HC. These results confirm the previous finding that IP-10 and IL-6 levels were elevated in the serum of patients with SJS/TEN 21,38,47,48 . IL-6 levels were also found to be higher in tears from SJS/TEN patients, although no correlation with the severity of ocular involvement was observed 20,22,49 . Studies have demonstrated that GLYN is probably a key mediator in keratinocyte apoptosis in SJS/TEN, and corneal epithelium could produce GYLN in the presence of TNF-α in an ex vivo report 25,26 . However, the fact that GYLN levels were marginally significantly higher in those with SOCs compared to those with non-severe reactions in our study might be due to the limited sample size, or the roles of GYLN in SOCs may be, in fact, not as important as that of S100A8/A9 in SJS/TEN. The functions of PDGFs in the pathogenesis of SJS/TEN remain to be explored.
After physicians confirmed the diagnosis of SJS/TEN, all patients were registered with the ThaiSCAR. This systematic registration not only provides better holistic care for patients but also helps physicians to collect comprehensive data for each individual. This study identified predictive factors for SOCs in SJS/TEN patients and found that BSA and older age were correlated with SOCs. In addition, we analyzed the correlation between inflammatory cytokines and SOCs. Previous studies have examined local specimens, including tears and conjunctival swabs, but because of the complexity involved in collecting local specimens and the lack of a standard protocol for handling samples, we decided to interpret biomarkers from serum. We found that S100A8/A9 and GYLN levels tended to increase in the SOC group. To our knowledge, this is the first report indicating that S100A8/A9 is upregulated and might be involved in the pathogenesis of SJS/TEN. This study had a few limitations, including its retrospective design. Second, the sample size was limited due to the rarity of the disease, with a relatively small number of patients being treated at one center. Finally, all identified causes of SJS/TEN in the ThaiSCAR registry were medications; infections or other unidentified factors that could be potential etiologies of SJS/TEN causing SOCs were not included. Future large-scale, multicenter cohort studies, and the analysis of tear biomarkers or impression cytology evaluation of conjunctiva with their temporary changes are needed to verify the risk factors associated with SOCs in SJS/TEN.
In summary, these identified factors consisting of clinical manifestations and upregulated biomarkers would be beneficial for developing screening protocols or universal tools that will help physicians recognize high-risk patients, leading to prompt management and prevention of chronic severe ocular sequelae.

Conclusion
From the clinical factor analysis, disease severity (BSA detachment ≥ 10%) and older age were predictive factors for acute SOCs in SJS/TEN. From the serum biomarker analysis, the levels of S100A8/A9, IP-10, IL-6, PDGF-AA, and PDGF-BB were increased in the SJS/TEN group; they may be potential markers that could differentiate between HC and those with SJS/TEN. The S100A8/A9 and GYLN levels tended to increase in the severe ocular involvement group, suggesting that S100A8/A9 and GYLN are involved in the pathogenesis of SOCs and serve as helpful biomarkers to predict acute SOCs in SJS/TEN. Taken together, two clinical factors, namely BSA and older age, and increased levels of serum biomarkers, namely S100A8/A9 and GYLN, may guide clinicians in identifying high-risk groups that could develop acute SOCs. Prompt management in these patients will minimize the chance of developing chronic ocular sequelae, leading to vision loss.

Methods
Subjects and study design. This study followed the tenets of the Declaration of Helsinki and was approved by the Institutional Review Ethics Committee of the Faculty of Medicine, Chulalongkorn University. This cross-sectional study was conducted at King Chulalongkorn Memorial Hospital (KCMH), Bangkok, Thailand, between July 2020 and January 2021. Forty-seven patients diagnosed with SJS/TEN at KCMH were enrolled in this study. These patients were part of the Thailand Severe Cutaneous Adverse Reactions (ThaiSCAR) cohort registered at ClinicalTrials.gov (NCT02574988). The study protocol was approved by the Institutional Review Board of the Faculty of Medicine, Chulalongkorn University (approval no. 1272/2020, IRB No. 454/63). Informed consent was obtained from all participants. The classification criteria of Bastuji-Garin et al. were used to describe SJS, TEN, and SJS/TEN overlap syndrome 1 and the suspected culprit drugs were assessed according to the algorithm of drug causality for epidermal necrolysis (ALDEN) 50  www.nature.com/scientificreports/ According to the ThaiSCAR protocol, SJS/TEN patients were admitted after the onset of symptoms. Patients' serum was then collected, and referrals were made to the ophthalmologist. The ocular complications were evaluated primarily within 24 h of admission.
Demographic and clinical data. We retrospectively reviewed the medical records and electronic database of patients with SJS/TEN. The inclusion criteria were as follows: (1) patients with fully accessible medical records of the acute phase for evaluation of prognostic factors (2) no history of other ocular surface disorders or ocular surgeries, and (3) age > 18 years. For simplicity, patients with SJS/TEN overlap were assigned to the TEN group.
The clinical parameters and ocular findings were collected during the acute phase of SJS/TEN, within 8-12 days of symptom onset 51 .When the acute ocular severity differed between the eyes, the eye with greater severity was chosen for evaluation. If both eyes were symmetrically severe, the right eye was selected. The acute ocular severity score was defined as previously described by Sotozono et al. 27 (Table 2). Patients with ocular severity scores greater than grade 2 were considered to have SOC.
Evaluation of clinical parameters for predicting SOC in SJS/TEN. The study population was divided into two groups according to the severity of ocular involvement: STS/TEN patients with non-SOC (acute ocular severity grade 0,1, and 2; n = 27) and SJS/TEN patients with SOC (acute ocular severity grade 3; n = 20). The disease severity (SJS: BSA detachment < 10%, SJS/TEN overlap: BSA detachment 10%-30%, TEN: BSA detachment > 30%), age, sex, laboratory results, SCORTEN, mucosal involvement, causative drugs, flu-like symptoms, underlying diseases, initial visual acuitiy, sepsis complications, and HIV infection in the acute phase were analyzed. Serum collection. The serum samples of patients with SJS/TEN were cryopreserved from a ThaiSCAR prospective study.
Serum (30 ml) was collected from patients registered in the ThaiSCAR database while admitted to KCMH during the acute phase of SJS/TEN, within 8-12 days of symptom onset 51 . Undiluted samples were stored at − 80 °C until biomarker measurement. Patients in this study underwent serum collection and ocular examination within a maximum of 3 days following admission.
Evaluation of biological markers associated with SJS/TEN. Because sepsis appears to be a possible confounder, seven patients with sepsis were excluded. Serum from 40 patients with SJS/TEN and 18 HC was analyzed to explore the relationship between serum biomarkers in SJS/TEN and HC. A non-significant difference was noted in the mean age between the SJS/TEN and HC groups. After thawing, the samples were analyzed using a bead-based multiplex immunoassay. Panels of 42 biomarkers were used to identify biomarkers with prognostic potential for SJS/TEN. All samples were analyzed in duplicate. Biomarkers with a GMR of more than two were considered clinically significant.
Subgroup analysis to determine biological markers associated with SOC in SJS/TEN. To identify the biological markers related to acute SOCs, we analyzed and compared the data between SJS/TEN patients with non-SOC (n = 24) and SOC (n = 16). ). 25 μl of assay buffer was added into each well. Then, 25 μl of diluted standard or plasma were added in standard or sample wells. After that, 25 μl of mixed beads and 25 μl of detection antibodies were added into each well and incubated for 2 h at room temperature on an orbital plate shaker. After incubation, 25 μl of streptavidin-PE solution was added and then incubated for 30 min at room temperature on an orbital plate shaker. The plates were centrifuged at 1,000 rpm for 5 min. After decanting liquid and wash more one time with wash buffer, all wells added 150 μl of wash buffer and shaked for 2-3 min prior to analyze by flow cytometry (BD FACSCalibur™, Becton Dickinson, USA). Statistical analysis. Data were analyzed as means and standard deviations for continuous variables and counts and percentages for categorical variables. Univariate and multivariate logistic regression analyses were used to investigate prognostic factors related to SOCs. Multivariate logistic regression analysis was performed by stepwise backward elimination approach with variables whose P value were less than 0.2 in univariate analysis. The biomarker levels were log-transformed and described as geometric means (GM) with the percent coefficient of variation (%CV). Differences between groups was presented as geometric mean ratios (GMR). Normal distribution was determined using histograms and the Shapiro-Wilk test for normality after logarithmic transformation. Linear regression analysis was used to evaluate the differences in logarithmic mean values of the biomarker concentrations among various groups (healthy controls group vs. SJS/TEN group and non-severe ocular involvement group vs. severe ocular involvement group). Mann-Whitney test was used for comparison between HC and SJS/TEN groups and between SJS/TEN with non-severe and severe ocular complications groups. Statistical www.nature.com/scientificreports/ significance was defined as P < 0.05. Scatter plot graphs were created by GraphPad Software. All analyses were performed by STATA Statistical software version 15.1 (StataCorp, LLC, College Station, Texas, USA).

Data availability
The data analyzed from the patient's clinical parameters and serum biomarkers that support the results of this study is available upon reasonable request from the corresponding author V.P.