Stevens-Johnson Syndrome and Toxic Epidermal Necrolysis: Associations, Outcomes, and Pathobiology-Thirty Years of Progress but Still Much to Be Done

Nearly a century ago, Stevens and Johnson first described Stevens-Johnson syndrome (Heng 2015, Roujeau 1995). More than 60 years ago, Lyell first described epidermal necrolysis (Heng 2015, Su 2016). Most of the progress in defining the spectrum of disease now known as Stevens-Johnson syndrome, overlap syndrome, and toxic epidermal necrolysis (SJS/TEN) and differentiating it from other acute blistering diseases of skin and mucous membranes has occurred over the last 30 years. During these three decades, the causes, pathophysiology and treatment of SJS/TEN have been comprehensively explored, but the answer to many questions and effective treatment remain elusive.

Much of the work in this area has been done by the group at Hôpital Henri Mondor in Paris, led by Prof. Jean-Claude Roujeau and his disciples for nearly 30 years. Roujeau has also established numerous international collaborations, including with the University Medical Center in Freiburg, Germany and with investigators in Taiwan. This group’s recent joint effort assesses interleukin (IL)-15 levels in SJS/TEN (Su 2016). Su et al.’s paper in the JID demonstrates an association between IL-15 levels and the severity and mortality of toxic epidermal necrolysis. This work takes another step in illuminating the complex pathobiology of the disease. It may also provide a basis for new and much needed treatment. Su’s paper is best understood in the context of the progress made over the last three decades in:

1. Defining the clinical spectrum of SJS/TEN

2. Etiology

3. Pathobiology

4. Predicting outcomes

5. Prevention

6. Treatment

1. Defining the clinical spectrum of SJS/TEN

For any prognostic test to be useful, the disease to which it is applied must be precisely defined, i.e. there should be as close as possible to a gold standard definition of that condition. More than 40 years ago, clinical, bacteriologic and histologic criteria defined staph scalded skin syndrome and differentiated it from what we now know as toxic epidermal necrolysis. Thirty years ago, there was little consistency in the diagnostic criteria for erythema multiforme and Stevens-Johnson syndrome. Over these three decades, the understanding that Stevens-Johnson syndrome and toxic epidermal necrolysis are a spectrum of disease which vary in extent of epidermal detachment and severity but have common pathology and etiology has become widely accepted.

Under Roujeau’s leadership, the International SJS study developed and validated diagnostic criteria which provide the most precise diagnostic criteria for Stevens-Johnson syndrome and toxic epidermal necrolysis. This work also defined erythema multiforme, which differs from SJS/TEN clinically, demographically, and etiologically (Bastuji-Garin 1993).

Su’s work confirms that the population cytokine profile of SJS/TEN is different from that of “maculopapular” drug eruption DRESS syndrome and viral exanthem. They also found higher serum levels of IL-6 granulysin and TNF alpha levels among those diagnosed with toxic epidermal necrolysis, the more severe diagnosis within the SJS/TEN spectrum. However, concentrations of these cytokines varied greatly among the patients tested and are unlikely to be helpful additions to the clinical exam in differentiating SJS from TEN.

2. Etiology

With clinical and histologic criteria that differentiate SJS/TEN from other conditions, more robust epidemiologic studies became possible. Pioneering case series from Hôpital Henri Mondor demonstrated that a substantial majority of cases of SJS/TEN were due to drugs (Guillaume 1987). At the same time, interest in severe adverse cutaneous reactions to drugs increased among regulatory authorities and pharmaceutical industries.

Working with colleagues in Europe and the United States, Roujeau led in organizing the International SJS/TEN study, the first large case-control study of drugs as a cause of this spectrum of disease (Roujeau 1995). This and subsequent international collaborations have established the drugs most frequently associated with the development of SJS/TEN, as well as the usual timing between beginning a likely causative agent and the development of first symptoms of SJS/TEN. Analyses of these data indicate that drugs are responsible for most cases, but a substantial minority of cases of SJS/TEN are due to exposures other than medications (Heng 2015, Sassolas 2010).

3. Pathobiology

Because of its rarity, early and subsequent samples of SJS/TEN are rarely available for analysis. Further, an animal model that reliably recapitulates human disease does not exist. As a result, the pathobiology of SJS/TEN remains elusive. However, ample evidence points toward T cell- and to a lesser extent NK cell- mediated pathogenesis. Blister fluid from SJS/TEN patients contains T cells, primarily CD8⁺ type, with lower numbers of NK and NKT cells. These cells demonstrate target cell killing in vitro (Chung, Hung et al. 2008). Also, several well-characterized HLA associations have been identified that predispose patients to specific delayed-type drug hypersensitivity reactions including SJS/TEN (White, Chung et al. 2015). Most known HLA associations are class I, supporting the role of CD8⁺ T cells in disease (White, Chung et al. 2015). CD8⁺ T cells isolated from peripheral blood of patients with SJS/TEN due to carbamazepine were shown to release the cytotoxic mediator granulysin in vitro in response to drug. These T cells also expressed a dominant clone, indicating antigen specific expansion (Ko, Chung et al. 2011). No such cells were found in patients who tolerated carbamazepine without disease (Ko, Chung et al. 2011)

However, the mechanism of T cell-mediated pathogenesis is not established. Both CD8⁺ T cells and NK cells kill target cells via release of granules containing the cytotoxic molecules perforin, granzymes and granulysin, via granule-independent secretion of granulysin, or via expression of Fas ligand (L) which binds Fas on target cells, with resultant target cell (in this case, keratinocyte) apoptosis. Initial studies pointed toward Fas:FasL-mediated cytotoxicity in TEN (Viard, Wehrli et al. 1998), but more recently granulysin, which is highly expressed in skin biopsies and blister fluid from patients during active disease, has been implicated as the major driver in epidermal death. Also, injection of granulysin into mouse skin induces blister lesions and necrosis similar to SJS/TEN (Chung, Hung et al. 2008). The involvement of cytotoxic CD8⁺ T cells and NK cells suggests that IL-15, a cytokine critical in CD8⁺ T cell and NK cell development, survival, and function, could play a major role in disease.

IL-15 is a widely expressed, pleiotropic cytokine produced by many cell types including immune cells (monocytes, macrophages and dendritic cells (DC)) and non-immune cells, including keratinocytes (Fehniger and Caligiuri 2001). In addition to its effects on T and NK cells, it also modulates B cells, monocytes, macrophages, DC, mast cells, and neutrophils (Patidar, Yadav et al. 2016). IL-15 is a member of the common gamma chain (γc) cytokine family, as its heterotrimeric receptor shares one of its three subunits, γc, with a number of other cytokines, IL-2, IL-4, IL-7, IL-9 and IL-21. It shares a second subunit, the IL-2/IL-15 receptor β (IL-15Rβ) with IL-2 and it has a third unique subunit, IL-15Rα (Figure 1) (Jabri and Abadie 2015). IL-15 exists in three functional forms: soluble IL-15, soluble IL-15:IL-15Rα complex and membrane-bound IL-15:IL-15Rα complex (Figure 1). This last form, membrane bound IL-15:IL-15Rα, is the most common and effective form of IL-15 presentation (termed trans-presentation) (Jabri and Abadie 2015). In this form, the membrane bound complex engages the IL-15Rβ/γc heterodimer expressed on the surface of a receiving cell. Cis-presentation of soluble IL-15 or soluble IL-15:IL-15Rα complex to receiving cells does occur but is believed to be of lesser importance to immune responses (Jabri and Abadie 2015).

Figure 1. IL-15 presentation and receptor binding.

IL-15 is most effectively presented in trans, meaning complexed to membrane bound IL-15Rα. This membrane bound complex on the surface of an IL-15-producing cell binds to IL-15R homodimer consisting of IL-15Rβ and γc subunits expressed on the surface of a CD8⁺ T cell or NK cell. Alternatively, soluble IL-15 complexed to soluble IL-15Rα can bind to the same receptor homodimer, or soluble IL-15 can bind the heterotrimeric receptor, consisting of IL-15Rα, IL-15Rβ and γc subunit expressed on the surface of a CD8⁺ T cell or NK cell. Binding of IL-15 to its receptor induces signaling via the γc subunit to Janus kinase (JAK) 3 which in turn stimulates STAT 5, and the IL-15Rβ subunit to JAK 1 which in turn stimulates STAT 3.

IL-15 signals through the JAK-STAT pathway, with the γc subunit activating JAK 3 then in turn STAT5, and the IL-15Rβ subunit activating JAK1 then in turn STAT3 (Jabri and Abadie 2015). Multiple downstream signaling pathways are stimulated, including the PI3K/AKT/mTOR pathway, which appears critical for IL-15-mediated effects on NK cells (Ali, Nandagopal et al. 2015) and T cells (Powell and Delgoffe 2010).

IL-15 promotes T and NK cell responses both directly and indirectly (Figure 2). It directs T cell development, expansion and maintenance/survival, particularly of CD8⁺ memory T cells, and it acts as a chemoattractant (Fehniger and Caligiuri 2001). It stimulates T cells to produce pro-inflammatory cytokines/chemokines and increase cytotoxicity (Jabri and Abadie 2015). IL-15 orchestrates NK cell development, expansion, survival and activation (Fehniger and Caligiuri 2001). It also contributes to the homeostasis and expansion of NKT cells (Fehniger and Caligiuri 2001), a subset of T cells, found in blister fluid of SJS/TEN patients, that recognize CD1d rather than traditional MHC. IL-15 activates DC and macrophages resulting in increased antigen-presentation to T cells and production of pro-inflammatory cytokines which drive the adaptive immune response (Jabri and Abadie 2015, Patidar, Yadav et al. 2016).

Figure 2. Pleiotropic effects of IL-15 on immune cells thought critical to SJS/TEN pathogenesis.

IL-15 induces activation of skin antigen presenting cells including macrophages (MΦ) and DC. Activation enhances production of pro-inflammatory cytokines and antigen presentation. IL-15 promotes development, expansion, and maintenance/survival of CD8⁺ T cells and NK cells. It also stimulates CD8⁺ T cell and NK cell activation resulting in increased cytotoxicity and cytokine/chemokine production. IL-15 plays a role in NKT cell expansion and homeostasis though its effects on NKT cell activation are still under investigation.

To our knowledge, Su et al., are the first to demonstrate a substantive link between IL-15 and SJS/TEN. They show that IL-15 levels correlate with disease severity and mortality, implicating IL-15 as a possible mediator of disease. They demonstrate that addition of exogenous IL-15 stimulated TEN blister fluid cells to secrete granulysin in three samples and that neutralization of IL-15 reduced drug-induced cell stimulation in a TEN patient. Further work is needed to determine whether IL-15 drives T and NK cell responses in SJS/TEN. The current study measured soluble IL-15 in serum, however trans-presentation of IL-15, ie membrane bound IL-15 in immunologic synapse with a receiving cell may be more important for driving an immunologic response. Expression of IL-15 in skin, as well as activation of downstream signaling events to confirm IL-15 activity should be investigated. The discrepancy between serum levels and extent of tissue damage is a relevant one in SJS/TEN as granulysin levels in serum have not consistently shown correlation with disease severity or mortality (Su et al.,(Chung, Hung et al. 2008)) but granulysin levels within blister fluid do correlate with disease severity (Chung, Hung et al. 2008). Lymphocyte stimulation tests in SJS/TEN are also inconsistent. For example, Porebski et al attempted to enhance responses to lymphocyte stimulation tests by adding exogenous IL-7 and IL-15 but found it effective in only a few patients (Porebski, Pecaric-Petkovic et al. 2013). Ultimately, blocking IL-15 in patients or in an animal model of SJS/TEN with prevention/resolution of disease is likely necessary to support a primary causative role for IL-15.

4. Predicting Outcomes

Early work at Hôpital Henri Mondor identified factors associated with death among TEN patients (Revuz 1987). Subsequently, this group developed a specific severity of illness score (SCORTEN) for cases of toxic epidermal necrolysis which was shown to be a good predictor of survival for these SJS/TEN patients (Bastuji-Garin 2000).

SCORTEN incorporates seven independent prognostic factors. Only one of these, body surface area involved (greater than equal 10%), is directly related to body surface area detachment. Age (>40) and physiologic factors accounted for the remaining six independent prognostic factors (age, malignancy, tachycardia, serum urea, serum glucose and serum bicarbonate). Some of these physiologic factors could impact serum cytokine levels. Agreement between SCORTEN values and mortality vary over the first five days of admission (Guegan 2006).

Su’s work demonstrates that a number of factors, including IL-6, IL-8, TNF alpha, and most notably IL-15, were upregulated in SJS/TEN patients. In supplemental figures 3, 4 and 5, the authors also present data that strongly suggest an association between the evolution of individual patients’ clinical course, extent of cutaneous involvement and IL-15 levels. However, these findings raise question about the utility of IL-15 levels in determining prognosis.

Prognosis implies a forecast of the likely course or outcome of the disease. As noted above, SCORTEN is such a validated measure, with good performance characteristics based on the area under its receiver operator characteristics curve. Comparable data are not presented for IL-15. However the data presented suggest that only a few of the patients who died had notably high initial IL-15 levels. The majority of those with high IL-15 levels also had SCORTEN values equal to three or greater; values associated with high mortality (Bastuji-Garin 2000, Guegan 2006). Many individuals who died had only modest initial IL-15 levels suggesting that early on this cytokine may not be a sensitive predictor of mortality.

Although only presented for a limited number of patients (see Supplemental Figures), an individual’s IL-15 serum level over time seems to be correlated with their clinical course and outcome. Additional data may show that clinical worsening or mortality is better predicted on the basis of changes in IL-15 levels rather than as a single value. The apparent correlation of IL-15 levels over time in an individual with their clinical status provide further evidence that IL-15 may play an important role in SJS/TEN.

5. Prevention

While case-control and registry studies have identified the medications with the highest associated risk of SJS/TEN and patient characteristics associated with a higher risk of these reactions, the absolute risk of most medications in most populations is low. There are notable exceptions. Perhaps the most striking example is the identification of a specific HLA associated with a greatly increased risk of SJS/TEN in Han Chinese starting carbamazapines (Chung 2004). This landmark research changed prescribing practices and provided hope that genetic screening could greatly reduce the risk of these reactions with other drugs and in other populations. Unfortunately, subsequent genetic studies revealed such strong associations appear limited to only certain drugs in certain populations (Lonjou 2006).

6. Treatment

Prompt withdrawal of possible causes and supportive care remain the mainstay of treatment (Garcia-Doval 2000, Roujeau 2011). The association of IL-15 levels with patients’ clinical course and mortality suggests IL-15 may be another therapeutic target for treatment of this often fatal disease. However, past experience with targeted therapies has largely been disappointing. TNF alpha is increased in SJS/TEN. One of the few randomized placebo controlled studies of a therapy for SJS/TEN examined the effect of thalidomide, a potent inhibitor of TNF alpha, on mortality. The group treated with thalidomide experienced higher death rates (Wolkenstein 1998). Initial enthusiasm for many treatments including IVIG, a targeted therapy for SJS/TEN, has also largely waned as more evidence suggests most treatments had little or no effect on outcome (Roujeau 2011). If its correlation with clinical course is more robustly demonstrated, IL-15 may be a therapeutic target, and monitoring IL-15 levels may provide a useful surrogate endpoint for evaluating treatments.

If IL-15 is indeed the major cytokine orchestrating SJS/TEN, a number of novel therapeutics hold promise for efficacy in a disease that desperately requires a viable treatment option. Tofacitinib and ruxolitinib are novel JAK-STAT inhibitors, inhibiting JAK 1 and 3, and JAK 1 and 2, respectively. Tofacitinib has FDA approval for rheumatoid arthritis and ruxolitinib is approved for myelofibrosis and polycythemia vera. Both are under investigation for cutaneous inflammatory disorders such as psoriasis, alopecia areata and vitiligo. Several new JAK inhibitors are under development or already in clinical trials. There is a humanized monoclonal antibody that blocks IL-15Rβ, termed Hu-Mik-β-1, currently in clinical trials for T cell large granular lymphocytic leukemia (Waldmann, Conlon et al. 2013), HTLV-1-associated complications (ClinicalTrials.gov NCT00076843) and refractory celiac disease (Waldmann 2013). Alternatively, since IL-15 activates the mTOR pathway which in turn promotes NK and T cell responses, mTOR inhibitors such as sirolimus or everolimus could theoretically mitigate SJS/TEN.

The work from the Taiwanese, French and German group demonstrates the power of international cooperation in the study of this rare but devastating condition. Given the limited resources that have been available for research in this area, the progress made in defining SJS/TEN, determining its most frequent causes, understanding its pathobiology and evaluating its treatment has been exceptional. Prof. Jean-Claude Roujeau’s ideas and contributions are evident in much of the most important work done over the last 30 years.