Cold medicine-related Stevens-Johnson syndrome/toxic epidermal necrolysis with severe ocular complications-phenotypes and genetic predispositions.

Stevens-Johnson syndrome (SJS) is an acute inflammatory vesiculobullous reaction of the skin and mucosa, such as the ocular surface, oral cavity, and genitals. In patients with extensive skin detachment and a poor prognosis, the condition is called toxic epidermal necrolysis (TEN). Severe ocular complications (SOCs) appear in some-but not all-SJS/TEN patients who are diagnosed by dermatologists, and cold medicines including multi-ingredient cold medications and nonsteroidal anti-inflammatory drugs are the main causative drugs particularly for SJS/TEN with SOCs and all SJS and TEN. In this review, we focus on the genetic predisposition of cold medicine-related SJS/TEN (CM-SJS/TEN) with SOCs. CM-SJS/TEN with SOCs was strongly associated with HLA-A*02:06 and significantly associated with HLA-B*44:03 in Japanese individuals, significantly associated with HLA-B*44:03 in Indian and Brazilian individuals, and associated with HLA-A*02:06 in Korean individuals. In the first genome-wide association study (GWAS), we found an association between the prostaglandin E receptor 3 (PTGER3) gene and SJS/TEN with SOCs. In this study, we focused on CM-SJS/TEN with SOCs and found that the association of CM-SJS/TEN with SOCs became stronger than all SJS/TEN with SOCs. In the second GWAS, we found an association between the IKZF1 gene and CM-SJS/TEN with SOCs not only in Japanese, but also in Korean and Indian populations. Moreover, we found that TSHZ2 gene single nucleotide polymorphisms (SNPs) also showed especially low p values in the Japanese population; however, this association was not found in the Korean population. Furthermore, we investigated the interaction between susceptibility genes, and found multiplicative interactions of HLA-A*02:06 and TLR3 SNPs and additive interactions of HLA-A*02:06 and PTGER3 SNPs.

1. Stevens–Johnson syndrome/toxic epidermal necrolysis in ophthalmology

Stevens–Johnson syndrome (SJS) is an acute inflammatory vesiculobullous reaction of the skin and mucosa, such as the ocular surface, oral cavity, and genitals. In patients with extensive skin detachment and a poor prognosis, the condition is called toxic epidermal necrolysis (TEN). Table 1 shows the diagnostic criteria for SJS and TEN in Japan.1 Adefinite diagnosis of SJS requires mucosal lesions, whereas a definite diagnosis of TEN does not. Thus, SJS/TEN with mucosal lesions consists of SJS and a part of TEN. Moreover, not all cases of SJS/TEN with mucosal lesions have severe ocular lesions such as severe conjunctivitis with pseudomembrane and ocular surface epithelial defects in the acute stage (Figure 1). Figure 2 shows a case of severe conjunctivitis with pseudomembrane and ocular surface epithelial defects in the acute stage. It is reported that about 40% of SJS/TEN cases had severe ocular complications (SOCs) with pseudomembrane and ocular surface epithelial defects.1

Table 1

Diagnostic criteria for Stevens–Johnson syndrome and toxic epidermal necrolysis by Japanese Ministry of Health, Labour, and Welfare (2005).

Diagnostic criteria

Stevens–Johnson syndrome

Clinical entity

A severe mucocutaneous disorder characterized by erythema, epidermal detachment, and enanthema accompanied by high fever

Essential criteria (required)

1. Severe hyperemic and/or hemorrhagic mucocutaneous lesions

2. Epidermal detachment involving <10% of the total body surface area

3. High-grade fever (>38.0°C) in the absence of antipyretic therapy

Supportive findings

1. Flat, atypical target lesions

2. Bilateral acute keratoconjunctivitis accompanied by ocular surface epithelial defect and/or pseudomembrane formation

3. Histologic evidence of epidermal necrosis

Diagnostic criteria

Toxic epidermal necrolysis

Clinical entity

A severe mucocutaneous disorder characterized by extensive erythema, epidermal detachment (including blisters and erosions), and enanthema accompanied by high fever. The extent of epidermal detachment is >10% of the total body surface area

Essential criteria (required)

1. Epidermal detachment involving >10% of the total body surface area

2. Exclusion of staphylococcal scalded skin syndrome

3. High-grade fever (>38.0°C) in the absence of antipyretic therapy

Supportive findings

1. Generalized macular or diffuse erythema

2. Enanthema including bilateral acute keratoconjunctivitis accompanied by ocular surface epithelial defect and/or pseudomembrane formation

3. Histologic evidence of marked epidermal necrosis

Note. From “Predictive factors associated with acute ocular involvement in Stevens–Johnson syndrome and toxic epidermal necrolysis,” C. Sotozono, M. Ueta, E. Nakatani, A. Kitami, H. Watanabe, H. Sueki, et al., 2015, Am J Ophthalmol, 160, p. 228–37 e2. Copyright 2015, Elsevier Science. Reproduced with permission.

Figure 1

Ophthalmologists often report both SJS and TEN with severe ocular complications as SJS. SJS = Stevens–Johnson syndrome; TEN = toxic epidermalnecrolysis. Note. From Ueta M. Genetic predisposition to Stevens–Johnson syndrome with severe ocular surface complications. Cornea. 2015;34(Suppl. 11):S158–S165. Reproduced with permission.

Figure 2

Severe conjunctivitis with pseudomembrane and ocular surface epithelial defects in the acute stage of Stevens–Johnson syndrome (SJS). Reproduced from Ueta with permission from Medical-Aoi Publications, Inc.

Furthermore, dermatologists see SJS/TEN patients only in their acute stage, whereas ophthalmologists encounter such patients not only in the acute stage but also in the chronic stage. Therefore, for ophthalmologists, it is not easy to render a differential diagnosis of SJS or TEN when patients present in the chronic stage because the vesiculobullous skin lesions expressed in the acute stage have healed by the chronic stage.2 Diagnosis of SJS/TEN in ophthalmology was based on a confirmed history of acute-onset high fever, serious mucocutaneous illness with skin eruptions, and involvement of at least two mucosal sites including the ocular surface.34567891011121314 SJS/TEN patients with SOC in the acute stage often experience severe ocular sequelae, such as vision loss and very severe dry eye for which these patients would not have been able to lead a normal life.15 Figure 3 shows severe ocular sequelae such as very severe dry eye, trichiasis, symbrepharon, scarring of lid conjunctiva, and conjunctival invasion of the cornea in the chronic stage. We defined patients with SOCs in the acute stage as those who manifested pseudomembranes and epithelial defects on the ocular surface (cornea and/or conjunctiva),16 and in the chronic stage as patients with ocular sequelae such as severe dry eye, trichiasis, symblepharon, and conjunctival invasion into the cornea.15 Thus, ophthalmologists tend to report both SJS and TEN with SOCs as “SJS” in a broad sense.2

Figure 3

Severe ocular sequelae such as very severe dry eye, trichiasis, symbrepharon, scarring of lid conjunctiva and conjunctival invasion of cornea in the chronic stage of Stevens–Johnson syndrome (SJS). Reproduced from Ueta with permission from Medical-Aoi Publications, Inc.

To summarize, SOCs appear in some–but not all–SJS/TEN patients who are diagnosed by dermatologists.

2. Causative drug and human leukocyte antigen analysis

SJS/TENs are rare with an annual incidence rate of 1–6 cases/1 million persons,11718 but SJS/TEN carry high mortality rates of 3% for SJS and 27% for TEN.19 Moreover, they often associated with inciting drugs.3451018202122 It was reported that allopurinol and anticonvulsants such as carbamazepine are the main inciting drugs for SJS/TEN23; however, we234510 and others2022 found that cold medicines including multi-ingredient cold medications and nonsteroidal anti-inflammatory drugs (NSAIDs) are also major causative drugs for SJS/TEN.

Furthermore, the association between human leukocyte antigen (HLA) genotypes and drug-induced severe cutaneous adverse reactions (SCARs) including SJS/TEN has been reported. Carbamazepine-induced SJS/TEN exhibited a very strong association with the HIA-B*15:02 allele [case n = 44, control (tolerant) n = 101, odds ratio (OR) = 2504, pcorrected = 3.1 × 10−27 in Taiwanese Han Chinese patients.24 The HLA-A*31:01 allele was also reported to be strongly associated with carbamazepine-induced SCAR including SJS/TEN and drug-induced hypersensitivity syndrome (DIHS) in Japanese [case n = 77, control (tolerant) n = 420, OR = 9.5, p = 1.1 × 10−16]25 and European individuals [case n = 145, control (normal) n = 257, OR = 15.0, p = 3.5 × 10−8],26 suggesting that different ethnic groups may have the different risk factors for carbamazepine-induced SCARs. Allopurinol, a uric acid-lowering drug, could also induce SCAR including SJS, TEN, and DIHS, and allopurinol-induced SCARs were strongly associated with HLA-B* 58:01 in Han Chinese [case n = 51, control (tolerant) n = 135, OR = 580, pcorrected = 4.7 × 10−24],27 Caucasian [case n = 27, control (normal) n = 1822, OR = 80, pcorrected <10−6],28 and Japanese patients (case n = 36, control (normal) n = 986, OR = 62.8, p = 5.4 × 10−12),29 suggesting that different ethnic groups may share the same risk factors for allopurinol-induced SCARs.

We have also reported that cold medicine-related SJS/TEN (CM-SJS/TEN) with SOCs was strongly associated with HLA-A*02:06 [case n = 151, control (normal) n = 639, OR = 5.6, p = 2.7 × 10−20] and significantly associated with HLA-B*44:03 in Japanese individuals [case n = 151, control (normal) n = 639, OR = 2.0, p = 1.3 ×10−3, and these HLA genotypes were irrelevant to CM-SJS/TEN without SOCs.3 Thus, genetic predisposition such as HLA genotype might be different between SJS/TEN with and without SOCs.3 Moreover, HLA-A*02:06 and HLA-B*44:03 are not associated with cold medicine unrelated (other medicine related) SJS/TEN with SOCs,3 suggesting that genetic predisposition, such as HLA genotype, might be different depending on their causative drugs.353031

We also reported that CM-SJS/TEN with SOCs was significantly associated with HLA-B*44:03 in Indian [case n = 20, control (normal) n = 55, OR = 12.3, p = 1.1 × 10−5 and Brazilian [especially Brazilian Caucasians; case n = 15, control (normal) n = 62, OR = 6.2, p = 3.7 × 10−3 individuals, and that HLA-A*02:06 was associated with CM-SJS/TEN with SOC in Korean individuals [case n = 31, control (normal) n = 90, OR = 3.0, p = 0.018].4

We have likewise reported that cold medicines including multi-ingredient cold medications and NSAIDs were the main causative drugs for SJS/TEN with SOCs in all SJS and TEN.12351031 About 80% of our SJS/TEN with SOC patients have developed SJS/TEN within several days after receiving treatment for common cold.210

Interestingly, allopurinol might induce SJS/TEN without SOCs,30 and not all cases of carbamazepine-induced SJS/TEN have SOCs.2

In summary, genetic predisposition, such as HLA genotype, might be different depending on the causative drugs and their phenotype, for example, with and without SOCs (Figure 4).23531

Figure 4

Associations between HLA type and the main causative drugs of SJS/TEN. HLA = human leukocyte antigen; NSAIDs = nonsteroidal anti-inflammatory drugs; SJS = StevenseJohnson syndrome; TEN = toxic epidermal necrolysis. Note. From “IKZF1, a new susceptibility gene for cold medicine-related Stevens–Johnson syndrome/toxic epidermal necrolysis with severe mucosal involvement,” by M. Ueta, H. Sawai, C. Sotozono C, et al., 2015, J Allergy Clin Immunol, 135, p. 1538–1545. Copyright 2015, Mosby. Reproduced with permission.

3. Genome wide association study

3.1. PTGER3: first genome-wide association study

We previously reported that a genome-wide association study (GWAS) using the Affymetrix GeneChip Mapping 500K Array Set (Affymetrix, Santa Clara, Calif) showed an association between prostaglandin E receptor 3 (PTGER3) gene and SJS/TEN with SOCs.10 Moreover, we analyzed a total of 38 single nucleotide polymorphisms (SNPs) of the PTGER3 gene using DigiTag2 assay3233 and found 20 SNPs associated with SJS/TEN with SOCs (Figure 5).34 Theprotein of PTGER3 gene is EP3. EP3 is one of the four receptors (EP1, EP2, EP3 and EP4) of prostaglandin E2 (PGE2).35

Figure 5

20 SNPs of PTGER3 associated with SJS/TEN with severe ocular complication. PTGER3 = prostaglandin E receptor 3; SJS = Stevens–Johnson syndrome; SNP = single nucleotide polymorphism; TEN = toxic epidermal necrolysis. Note. From “Epistatic interaction between Toll-like receptor 3 (TLR3) and prostaglandin E receptor 3 (PTGER3) genes,” by M. Ueta, G. Tamiya, K. Tokunaga, et al., 2012, J Allergy Clin Immunol, 129, p. 1413–1416. Copyright 2012, Mosby. Reproduced with permission.

Cold medicine such as NSAIDs (e.g., ibuprofen and loxoprofen) and cold medicine ingredients (e.g., acetaminophen) have the suppressive effect in the production of prostanoid, including PGE (Figure 6).35

Figure 6

Cold medicine such as nonsteroidal anti-inflammatory drugs (NSAIDs) and cold medicine ingredients have the suppressive effect of the production of prostanoid, including PGE2.

It was reported that PGE2 acts at EP3 in the airway epithelium and negatively regulates inflammation in the mouse asthma model.36 We also reported that EP3 negatively regulated the inflammation of ocular surface in the mouse allergy conjunctivitis model37 and skin in the mouse contact dermatitis model.38 Because PGE2 acts at EP3 and negatively regulates mucocuta-neous inflammation,363738 we suggest that the suppression of PGE2 production by cold medicine might contribute to the onset and pathogenesis of CM-SJS/TEN with SOCs.210 Therefore, we turned to CM-SJS/TEN with SOCs. When we focused on CM-SJS/TEN with SOCs, the association with PTGER3 gene became stronger than that in total SJS/TEN with SOCs; seven of 18 SNPs already reported to be associated with SJS/TEN were significantly associated with CM-SJS/TEN with SOCs after Bonferroni correc-tion.14 PTGER3 SNP rs1327464 (G vs. A) was most significantly associated with CM-SJS/TEN with SOCs; the OR for the major allele was 0.232 (p = 7.92 × 10−10)14. Figure 7 shows the p values of the 18 SNPs in the dominant models; PTGER3 SNP rs1327464 (GG vs. GA + AA); p = 1.90 × 10−8, OR = 0.223.14

Figure 7

PTGER3 SNPs associated with CM-SJS/TEN with SOCs. PTGER3 SNP rs1327464 was most significantly associated with CM-SJS/TEN with SOCs

We also previously reported that EP3 protein levels are much lower in the conjunctival epithelium of patients with SJS/TEN with SOCs than control individuals such as conjunctival chalasis patients and chemical burn patients (Figure 8).39 Thus, EP3 expression might be downregulated in the ocular surface of the SJS/TEN with SOCs patients.

Figure 8

Downregulation of EP3 in the conjunctival epithelium of SJS/TEN with severe ocular complications. EP3 (brown stain) is strongly downregulated in the conjunctival epithelium of SJS compared with chemical burns. Controls consisted of isotype control antibodies instead of anti-EP3 antibodies. SJS = StevenseJohnson syndrome; TEN = toxic epidermal necrolysis. Note. From “Prostaglandin E receptor subtype EP3 expression in human conjunctival epithelium and its changes in various ocular surface disorders,” by M. Ueta, C. Sotozono, N. Yokoi, T. Inatomi, S. Kinoshita, 2011, PLoS One, 6, p. e25209. Copyright 2011, Public Library of Science. Reproduced with permission.

Furthermore, because patients with CM-SJS/TEN with SOCs have developed this condition by taking cold medicines after having a common cold as a result of some viral or mycoplasma infections, we assume that not only cold medicine, but also some microbial infectious such as virus or mycoplasma, might be important and necessary to trigger the onset of SJS/TEN with SOCs.231

3.2. IKZF1: second GWAS

In the second GWAS, we focused on CM-SJS/TEN with SOCs, and performed a GWAS of Japanese 117 cases and 691 controls using the Affymetrix AXIOM Genome-Wide ASI 1 Array (Affymetrix, Santa Clara, Calif).5 The Manhattan plot of the GWAS indicated that the HLA-A region showed the strongest association with susceptibility to CM-SJS/TEN with SOCs (Figure 9),5 which was consistent with findings from our previous studies.31112

Figure 9

Manhattan plot of a genome-wide association study using the Affymetrix AXIOM Genome-Wide ASI 1 Array. HLA = human leukocyte antigen. Note. From “IKZF1, a new susceptibility gene for cold medicine-related Stevens–Johnson syndrome/toxic epidermal necrolysis with severe mucosal involvement,” by M. Ueta, H. Sawai, C. Sotozono C, et al., 2015, J Allergy Clin Immunol, 135, p. 1538–1545. Copyright 2015, Mosby. Reproduced with permission.

Outside of the HLA region, there were 60 SNPs with p < 10−3 (in the allele frequency, dominant model, or recessive model) in the GWAS.5 Of the 60 SNPs, 47 were p < 10−4, and 11 of these 47 were p < 10−5.5 Among the 11 SNPs of eight genes that were p < 10−5, IKZF1 showed especially low p values [rs897693: (C vs. T) OR = 4.3, p = 1.2 × 10−7; (CC + CT vs. TT), OR = 5.0, p = 2.1 × 10−8].5

Moreover, we genotyped the SNPs of the IKZF1 gene with all Japanese samples (149 SJS, 877 controls) with 16 additional Japanese cases, Korean samples (31 SJS and 90 controls), Indian samples (20 CM-SJS/TEN with SOCs and 56 controls), and Brazilian samples (39 CM-SJS/TEN with SOCs and 135 controls), and then found that a meta-analysis from the Japanese, Korean, Indian, and Brazilian samples showed a significant genome-wide association between CM-SJS/TEN with SOCs and IKZF1 [rs4917014 (G vs. T), OR = 0.5, p = 8.5 × 10−11.5 These findings show that IKZF1 may be a universal marker for susceptibility to CM-SJS/TEN with SOCs.5 Furthermore, we have performed a functional analysis for SNPs of the IKZF1 gene and found that the Ik2/Ik1 (both are splicing isoforms) ratio may be influenced by IKZF1 SNPs that are significantly associated with susceptibility to CM-SJS/TEN with SOCs.5 To elucidate the role of IKZF1 in the pathogenesis of CM-SJS/TEN with SOCs, we are going to do further investigations.

3.3. TSHZ2: second GWAS

Among the 11 SNPs of eight genes that were p < 10−5 in the second GWAS, TSHZ2 also showed especially low p values [rs4809905: (Avs. G), OR = 0.4, p = 5.6 × 10−7; (AA + AG vs. GG), OR = 0.3, p = 1.5 × 10−7].5 Furthermore, we have examined a total of 17 SNPs of the near region of TSHZ2 gene using 101 CM-SJS/TEN cases and 200 control samples, and found that seven SNPs (rs4811338, rs1555278, rs2092954, rs4371408, rs6021911, rs4809905, and rs6096940) showed a significant genome-wide association with CM-SJS/TEN with SOCs (Table 2). However, these SNPs were not associated with CM-SJS/TEN with SOCs in the Korean population (data not shown). TSHZ2 might be a susceptibility gene for CM-SJS/TEN with SOCs in the Japanese population.

Table 2

Association between CM-SJS/TEN with SOCs and TSHZ2 gene SNPs.

rs no. of SNP	Frequencies of genotypes (%)				Allele 1 vs. allele 2	Genotype 11 vs. 12 + 22	Genotype 11 + 12 vs. 22
	Genotypes		Control	Case	Pa OR (95% CI)	Pa OR (95% CI)	Pa OR (95% CI)
rs7360046	11	G/G	1/200 (0.5)	0/101 (0.0)	0.28	0.48	0.31
	12	G/A	27/200 (13.5)	10/101 (9.9)	–	–	–
	22	A/A	172/200 (86.0)	91/101 (90.1)	–	–	–
rs4811338	11	G/G	29/200 (14.5)	5/101 (5.0)	9.7 × 10−9	0.014	7.0 × 10−10
	12	G/T	109/200 (54.5)	27/101 (26.7)	0.31	0.31	0.21
	22	T/T	62/200 (31.0)	69/101 (68.3)	(0.2–0.5)	(0.1–0.8)	(0.1–0.3)
rs1555278	11	T/T	19/196 (9.7)	5/100 (5.0)	4.0 × 10−8	0.16	7.7 × 10−10
	12	T/C	96/196 (49.0)	16/100 (16.0)	0.29	–	0.19
	22	C/C	81/196 (41.3)	79/100 (79.0)	(0.2–0.5)	–	(0.1–0.3)
rs6096922	11	G/G	1/200 (0.5)	0/101 (0.0)	0.28	0.48	0.31
	12	G/A	27/200 (13.5)	10/101 (9.9)	–	–	–
	22	A/A	172/200 (86.0)	91/101 (90.1)	–	–	–
rs4811342	11	G/G	1/200 (0.5)	0/101 (0.0)	0.28	0.48	0.31
	12	G/A	27/200 (13.5)	10/101 (9.9)	–	–	–
	22	A/A	172/200 (86.0)	91/101 (90.1)	–	–	–
rs2092954	11	C/C	29/200 (14.5)	5/101 (5.0)	9.7 × 10−9	0.013	7.0 × 10−10
	12	C/T	109/200 (54.5)	27/101 (26.7)	0.31	0.31	0.21
	22	T/T	62/200 (31.0)	69/101 (68.3)	(0.2–0.5)	(0.1–0.8)	(0.1–0.3)
rs4811343	11	C/C	1/200 (0.5)	0/101 (0.0)	0.22	0.48	0.26
	12	C/T	21/200(10.5)	7/101 (6.9)	–	–	–
	22	T/T	178/200 (89.0)	94/101 (93.1)	–	–	–
rs4371408	11	G/G	29/200 (14.5)	5/101 (5.0)	9.7 × 10−9	0.013	7.0 × 10−10
	12	G/A	109/200 (54.5)	27/101 (26.7)	0.31	0.31	0.21
	22	A/A	62/200 (31.0)	69/101 (68.3)	(0.2–0.5)	(0.1–0.8)	(0.1–0.3)
rs6021911	11	G/G	20/200 (10.0)	5/101 (5.0)	4.0 × 10−8	0.13	9.8 × 10−10
	12	G/A	98/200 (49.0)	17/101 (16.8)	0.29	–	0.19
	22	A/A	82/200 (41.0)	79/101 (78.2)	(0.2–0.5)	–	(0.1–0.3)
rs4809905	11	A/A	28/200 (14.0)	5/101 (5.0)	6.7 × 10−9	0.018	5.3 × 10−10
	12	A/G	105/200 (52.5)	24/101 (23.8)	0.30	0.32	0.20
	22	G/G	67/200 (33.5)	72/101 (71.3)	(0.2–0.5)	(0.1–0.9)	(0.1–0.3)
rs8123906	11	A/A	3/200 (1.5)	0/101 (0.0)	0.0010	0.21	0.0011
	12	A/G	49/200 (24.5)	10/101 (9.9)	0.33	–	0.31
	22	G/G	148/200 (74.0)	91/101 (90.1)	(0.2–0.7)	–	(0.2–0.6)
rs6096940	11	G/G	28/200 (14.0)	5/101 (5.0)	9.3 × 10−9	0.018	9.1 × 10−10
	12	G/A	104/200 (52.0)	24/101 (23.8)	0.30	0.32	0.21
	22	A/A	68/200 (34.0)	72/101 (71.3)	(0.2–0.5)	(0.1–0.9)	(0.1–0.3)
rs968162	11	A/A	20/200 (10.0)	5/101 (5.0)	5.5 × 10−8	0.13	1.6 × 10−9
	12	A/G	97/200 (48.5)	17/101 (16.8)	0.30	–	0.20
	22	G/G	83/200 (41.5)	79/101 (78.2)	(0.2–0.5)	–	(0.1–0.3)
rs6021974	11	A/A	3/200 (1.5)	0/101 (0.0)	4.8 × 10−4	0.22	5.0 × 10−4
	12	A/G	49/200 (24.5)	9/101 (8.9)	0.29	–	0.28
	22	G/G	148/200 (74.0)	92/101 (91.1)	(0.1–0.6)	–	(0.1–0.6)
rs1474794	11	T/T	31/200(15.5)	5/101 (5.0)	1.2 × 10−6	0.0077	1.1 × 10−6
	12	T/C	105/200 (52.5)	34/101 (33.7)	0.39	0.28	0.30
	22	C/C	64/200 (32.0)	62/101 (61.4)	(0.3–0.6)	(0.1–0.8)	(0.2–0.5)
rs6068248	11	C/C	30/200 (15.0)	5/100 (5.0)	2.3 × 10−6	0.011	1.6 × 10−6
	12	C/T	106/200 (53.0)	34/100 (34.0)	0.40	0.30	0.30
	22	T/T	64/200 (32.0)	61/100 (61.0)	(0.3–0.6)	(0.1–0.8)	(0.2–0.5)
rs6068268	11	C/C	27/199(13.6)	33/101 (32.7)	5.7 × 10−6	9.2 × 10−5	4.4 × 10−4
	12	C/T	97/199 (48.7)	50/101 (49.5)	2.2	3.1	2.8
	22	T/T	75/199 (37.7)	18/101 (17.8)	(1.6–3.1)	(1.7–5.5)	(1.6–5.0)

CI = confidence interval; CM-SJS = cold medicine-related Stevens–Johnson syndrome; OR = odds ratio; SNP single nucleotide polymorphism; SOC = severe ocular complication; TEN = toxic epidermal necrolysis.

Bold shows the genome wide significance (p < 5 × 10−8).

ap value for allele or genotype frequency comparison between case and control using the chi-square test.

4. Interaction between susceptibility genes

4.1. HLA-A*02:06 and TLR3

SNPs have been widely used as genetic markers for identifying human disease susceptibility genes for the past decade. Moreover, it has recently become apparent that gene–gene interactions are meaningful in addition to major single-locus effects.34 We previously reported that, in addition, HLA-A*02:06,31112 PTGER3,1034 and IKZF1,5 TLR3934 SNPs showed significant associations with SJS/TEN with SOCs, although the expression of TLR3 protein on the ocular surface was not different between SJS/TEN and controls such as conjunctival chalasis and chemical burn (Figure 10).34

Figure 10

TLR3 expression in the conjunctival epithelium of SJS/TEN with severe ocular complications. The expression of TLR3 protein (brown stain) on the ocular surface was not different between SJS/TEN and controls such as conjunctival chalasis and chemical burn. Controls consisted of isotype control antibodies instead of anti-EP3 antibodies. SJS = Stevens–Johnson syndrome; TEN = toxic epidermal necrolysis. Note. From “Epistatic interaction between Toll-like receptor 3 (TLR3) and prostaglandin E receptor 3 (PTGER3) genes,” by M. Ueta, G. Tamiya, K. Tokunaga, et al., 2012, J Allergy Clin Immunol, 129, p. 1413–1416. Copyright 2012, Mosby. Reproduced with permission.

When we focused on CM-SJS/TEN with SOCs, five of the seven SNPs, which were significantly associated with SJS/TEN with SOCs in our previous report,1334 were still associated with CM-SJS/TEN with SOCs (Table 3).

Table 3

Association with TLR3 SNPs in CM-SJS/TEN with SOC in Japanese individuals.

rs no. of SNP	Frequencies of genotypes (%)				Allele 1 vs. allele 2	Genotype 11 vs. 12 + 22	Genotype 11 + 12 vs. 22
	Genotypes		Control	Case	pa OR (95% CI)	pa OR (95% CI)	pa OR (95% CI)
rs4861699	11	A/A	28/221 (12.7)	8/86 (9.3)	0.0156	0.4102	0.0066
	12	A/G	105/221 (47.5)	29/86 (33.7)	0.62	0.71	0.50
	22	G/G	88/221 (39.8)	49/86 (57.0)	(0.4–0.9)	(0.3–1.6)	(0.3–0.8)
rs6822014	11	G/G	9/214 (4.2)	13/131 (9.9)	0.0673	0.0349	0.2377
	12	G/A	72/214 (33.6)	45/131 (34.4)	1.40	2.51	1.30
	22	A/A	133/214(62.1)	73/131 (55.7)	(1.0–2.0)	(1.0–6.0)	(0.8–2.0)
rs11732384	11	A/A	16/220 (7.3)	9/133 (6.8)	0.1737	0.8575	0.1081
	12	A/G	91/220 (41.4)	44/133 (33.1)	0.78	0.93	0.70
	22	G/G	113/220 (51.4)	80/133 (60.2)	(0.6–1.1)	(0.4–2.2)	(0.5–1.1)
rs3775296	11	T/T	12/220 (5.5)	18/133 (13.5)	0.0604	0.0083	0.3436
	12	T/G	94/220 (42.7)	53/133 (39.8)	1.37	2.71	1.23
	22	G/G	114/220 (51.8)	62/133 (46.6)	(1.0–1.9)	(1.3–5.8)	(0.8–1.9)
rs5743312	11	T/T	10/219 (4.6)	17/133 (12.8)	0.0492	0.0050	0.3195
	12	T/C	90/219(41.1)	51/133 (38.3)	1.40	3.06	1.24
	22	C/C	119/219(54.3)	65/133 (48.9)	(1.0–2.0)	(1.4–6.9)	(0.8–1.9)
rs7668666	11	A/A	27/212 (12.7)	26/133 (19.5)	0.0986	0.0876	0.2795
	12	A/C	101/212 (47.6)	62/133 (46.6)	1.30	1.66	1.28
	22	C/C	84/212 (39.6)	45/133 (33.8)	(1.0–1.8)	(0.9–3.0)	(0.8–2.0)
rs3775290	11	A/A	25/220(11.4)	26/133 (19.5)	0.1554	0.0340	0.6324
	12	A/G	110/220 (50.0)	59/133 (44.4)	1.25	1.90	1.11
	22	G/G	85/220 (38.6)	48/133 (36.1)	(0.9–1.7)	(1.0–3.4)	(0.7–1.7)

CI = confidence interval; CM-SJS = cold medicine-related Stevens–Johnson syndrome; OR = odds ratio; SNP = single nucleotide polymorphism; SOC = severe ocular complication; TEN = toxic epidermal necrolysis.

Bold shows p < 0.05.

ap value for allele or genotype frequency comparison between case and control using the chi-square test.

We previously also reported that HLA-A*02:06 with TLR3 SNPs exerted more than additive effects in SJS/TEN with SOCs.13 When we focused on CM-SJS/TEN with SOCs, more than additive effects of HLA-A*02:06 with TLR3 SNPs in CM-SJS/TEN with SOCs were still observed. HLA-A*02:06 and TLR3 SNP rs3775296T/T or TLR3 SNP rs5743312T/T exerted more than additive effects (p < 0.00005, OR = 37.7, Woolf's correction or p < 0.00005, OR = 37.5, Woolf's correction): 10 of 133 patients (7.5%) manifested both HLA-A*02:06 and TLR3 rs3775296T/T SNP or TLR3 rs5743312T/T SNP, whereas none of the controls did (Table 4). There was a strong linkage disequilibrium (LD) between rs.3775296 and rs.5743312. Moreover,HLA-A*02:06 and TLR3 SNP rs3775290A/A exerted additive effects (p < 0.00005, OR = 9.9): 16 of 133 patients (12.0%) manifested both HLA-A*02:06 and TLR3 rs3775290A/A SNP, whereas only three of 220 controls (1.4%) did (Table 4).

Table 4

Interaction between HLA-A*02:06 and TLR3 SNPs in CM-SJS/TEN with SOCs.

HLA-A*02:06 and TLR3 rs3775296
HLA-A*02:06	rs3775296	感冒薬SJS	Controls	pa	OR (95% CI)
+	+	10/133 (7.52%)	0/220 (0.00%)	3.69 × 10−5	37.6* (2.19–647.7)
+	–	52/133 (39.1%)	30/220 (13.6%)	4.04 × 10−8	4.07 (2.42 6.83)
–	+	8/133 (6.02%)	12/220 (5.45%)	0.8253	1.11 (0.441–2.79)
–	–	63/133 (47.4%)	178/220 (80.9%)	5.34 × 10−11	0.212 (0.132 0.343)
HLA-A*02:06 and TLR3 rs5743312
HLA-A*02:06	rs5743312	感冒薬SJS	Controls	pa	OR (95% CI)
+	+	10/133 (7.52%)	0/219 (0.00%)	3.84 × 10−5	37.5b (2.18e644.7)
+	–	52/133 (39.1%)	30/219 (13.7%)	4.61 × 10−8	4.04 (2.41–6.80)
–	+	7/133 (5.26%)	10/219 (4.57%)	0.7674	1.16 (0.431e3.13)
–	–	64/133 (48.1%)	179/219 (81.7%)	3.75 × 10−11	0.207 (0.128–0.336)
HLA-A*02:06 and TLR3 rs3775290
HLA-A*02:06	rs3775290	感冒薬SJS	Controls	pa	OR (95% CI)
+	+	16/133 (12.0%)	3/220 (1.36%)	1.68 × 10−5	9.89 (2.82–34.6)
+	–	46/133 (34.6%)	27/220 (12.3%)	5.28 × 10−7	3.78 (2.21–6.48)
–	+	10/133 (7.52%)	22/220 (10.0%)	0.4314	0.732 (0.335–1.60)
–	–	61/133 (45.9%)	168/220 (76.4%)	6.00 × 10−9	0.262 (0.165e0.416)

CI = confidence interval; CM-SJS = cold medicine-related Stevens–Johnson syndrome; OR = odds ratio; SNP = single nucleotide polymorphism; SOC = severe ocular complication; TEN = toxic epidermal necrolysis.

a p value for allele or genotype frequency comparison between case and control using the chi-square test.

b Woolf's correction.

By interaction analysis, HLA-A*02:06 and TLR3 SNP rs3775296T/T, which were in strong LD with TLR3 SNP rs5743312T/T, manifested more than additive effects; therefore, the multiplicative interactions of HLA-A and TLR3 gene might be required for the onsetof CM-SJS/TEN with SOCs.

HLA-A is a component of HLA Class I, which resides on the surface of all nucleated cells and alerts the immune system that the cell may be infected by a virus, thereby targeting the cell for destruction, and TLR3 recognizes viral double-stranded RNA.40

As the onset of SJS/TEN was associated not only with the administration of drugs but also with putative microbial infection, multiplicative interactions of HLA-A and TLR3 gene might contribute to the characteristic and pathogenic immune response to the microbial infection.

4.2. HLA-A*02:06 and PTGER3

Next, we looked for interactive effects between these seven SNPs of the PTGER3 gene and HLA-A*02:06 in CM-SJS/TEN with SOCs, because the PTGER3 SNPs were significantly associated with the patients.14 We found an interaction with additive effects between HLA-A*02:06 and the high-risk genotypes PTGER3 rs1327464 GA or AA (OR = 10.8, p = 2.56 × 10−7; Table 5).14 In the Japanese population, although PTGER3 rs1327464 GA/AA alone showed OR = 4.48 and p = 1.90 × 10−8, and HLA-A*02:06 alone showed OR = 5.46 and p = 1.39 × 10−11, the combination of PTGER3 rs1327464 GA/AA and HLA-A*02:06 could show a higher OR (OR = 10.8, p = 2.56 × 10−7) than each allele alone.

Table 5

Interaction between HLA-A*02:06 and PTGER3 rs1327464 GA/AA among Japanese individuals.

HLA-A*02:06	rs1327464 GA or AA	CM-SJS/TEN with SOC	Controls	pa	OR (95% CI)
+	+	22/131 (16.8%)	4/218 (1.83%)	2.56 × 10−7	10.8 (3.63–32.1)
+	–	39/131 (29.8%)	26/218 (11.9%)	3.38 × 10−5	3.13 (1.80–5.45)
–	+	26/131 (19.9%)	21/218 (9.63%)	6.80 × 10−3	2.32 (1.25–4.33)
–	–	44/131 (33.6%)	167/218 (76.6%)	1.74 × 10−15	0.154 (0.0956–0.249)

CM-SJS = cold medicine-related Stevens–Johnson syndrome; HLA = human leukocyte antigen; OR = odds ratio; PTGER3 = prostaglandin E receptor 3; SOC = severe ocular complication; TEN = toxic epidermal necrolysis.

Note. From “HLA-A*02:06 and PTGER3 polymorphism exerts additive effects in cold medicine-related Stevens–Johnson syndrome with severe ocular complications,” by M. Ueta, K. Tokunaga, C. Sotozono, H. Sawai, K.C. Yoon, M.K. Kim, et al., 2015, Hum Genome Var, 2, p. 15023. Copyright 2015, Nature Publishing Group. Reproduced with permission.

Moreover, we found that even in Korean populations, there was an additive effect between the SNPs of the PTGER3 gene and HLA-A*02:06 (Table 6).14 In the Korean population, the combination of PTGER3 rs1327464 GA/AA and HLA-A*02:06 could show a higher OR (OR = 14.2, p = 5.58 × 10−6) than each allele alone, although PTGER3 rs1327464 GA/AA alone showed OR = 4.07 and p = 0.00101, and HLA-A*02:06 alone showed OR = 2.50 and p = 0.0412.

Table 6

Interaction between HLA-A*02:06 and PTGER3 rs1327464 GA/AA among Korean individuals.

HLA-A*02:06	rs1327464 GA or AA	CM-SJS/TEN with SOC	Controls	pa	OR (95% CI)
+	+	8/30 (26.7%)	3/120 (2.50%)	5.58 × 10−6	14.2 (3.49–57.7)
+	–	2/30 (6.67%)	17/120(14.2%)	0.269	0.433 (0.0943–1.99)
–	+	5/30 (16.7%)	16/120(13.3%)	0.638	1.30 (0.435–3.89)
–	–	15/15 (50.0%)	84/120 (70.0%)	0.0386	0.429 (0.190–0.968)

CM-SJS = cold medicine-related Stevens–Johnson syndrome; HLA = human leukocyte antigen; OR = odds ratio; PTGER3 = prostaglandin E receptor 3; SOC = severe ocular complication; TEN = toxic epidermal necrolysis.

Note. From “HLA-A*02:06 and PTGER3 polymorphism exerts additive effects in cold medicine-related Stevens–Johnson syndrome with severe ocular complications,” by M. Ueta, K. Tokunaga, C. Sotozono, H. Sawai, K.C. Yoon, M.K. Kim, et al., 2015, Hum Genome Var, 2, p. 15023. Copyright 2015, Nature Publishing Group. Reproduced with permission.

ap value for allele or genotype frequency. Comparison was between patients and controls using the chi-square test (Pearson).

Thus, the interaction between HLA-A*02:06 and PTGER3 SNP might strongly contribute to the onset of CM-SJS/TEN with SOCs.

5. Application strategy in future

Because CM-SJS/TEN is a rare condition that probably has a complex genetic background, it is reasonable to posit multiplicative interactions of genes such as HLA-A and TLR3,13 and HLA-A and PTGER3.14

Notably, after removing samples with both HLA-A*02:06 and TLR3 SNP rs3775296 T/T, the additive effect between HLA-A*02:06 and PTGER3 rs1327464 GA/AA persisted in CM-SJS/TEN with SOCs, suggesting that the interactions are independent of each other.14

These findings might show that the combinations of the two CM-SJS/TEN with SOCs associated polymorphisms, such as HLA-A*02:06 and TLR3 SNP and HLA-A*02:06 and PTGER3 SNP, could give improvements for a genetic testing compared with using only one susceptibility gene.14

Based on all of our previous observations, we suggest that in addition to microbial infections and cold medicines, the combination of multiple gene polymorphisms and their interactions contribute strongly to the onset of CM-SJS/TEN with SOCs.14