See related article on pg 2503

Complex Role of TNF Variants in Psoriatic Arthritis and Treatment Response to Anti-TNF Therapy: Evidence and Concepts Ulrike Hu¨ffmeier1 and Rotraut Mo¨ssner2 Psoriatic arthritis (PsA) is a chronic inflammatory disease affecting joints, and it may manifest as peripheral arthritis, dactylitis, enthesitis, spondylitis, or sacroiliitis. In the great majority of patients, PsA is accompanied by the most frequent psoriatic manifestation—psoriasis vulgaris. The major genetic risk factor for PsA is an HLA-C allele, and in recent genome-wide association studies few other susceptibility loci have as yet been identified. In this issue, Murdaca et al. (2014) describe an association of an intronic single-nucleotide polymorphism at the TNF locus ( þ 489) with PsA, disease severity, and treatment responses to tumor necrosis factor-a blockers. Journal of Investigative Dermatology (2014) 134, 2483–2485. doi:10.1038/jid.2014.294

With an increase in the therapeutic armamentarium for chronic inflammatory diseases, there is much interest in predicting the therapy best suited to an individual patient in order to achieve the best therapeutic result, as well as to avoid costs and side effects. In clinical trials for psoriatic arthritis (PsA), tumor necrosis factor-a (TNF-a) blockers have been shown to have excellent clinical efficacy and to prevent further structural damage to joints. However, the reasons why 30–60% of patients do not respond sufficiently to treatment with TNF-a blockers (as measured by a failure to achieve an ACR20 response after 12–16 weeks of treatment) are largely unknown. Different combinations of genetic factors are assumed to have a role. In this issue, Murdaca et al., 2014 report preliminary evidence for an association between the singlenucleotide polymorphism (SNP) þ 489 at the TNF locus and both susceptibility

to the development of PsA and to treatment responses to TNF-a blockers. Experience with other drugs has exemplified the potential of genetic predictors in the use of drug therapies. Good predictors of therapeutic response are found in tumor therapy, in which the presence of somatic mutations in tumors—responsible for dysregulation of cell growth and/or cell survival— may also predict responses to drugs tailored to impact the corresponding signal transduction pathways. For example, in metastatic melanomas, tumors with the activating BRAF mutation V600E respond to the BRAF inhibitors vemurafenib or dabrafenib in about 50% of patients, whereas these drugs are ineffective in patients without mutations at V600 (Jang and Atkins, 2014). These findings are reflected in the drugs’ license, which excludes patients whose tumors do not carry BRAF mutations at V600.

1 Institute of Human Genetics, Friedrich-Alexander-Universita¨t of Erlangen-Nu¨rnberg, Erlangen, Germany and 2Department of Dermatology, Venereology and Allergology, University Medical Center Go¨ttingen, Go¨ttingen, Germany

Correspondence: Rotraut Mo¨ssner, Department of Dermatology, Venereology and Allergology, University Medical Center Go¨ttingen, Robert-Koch-Strasse 40, D-37075 Go¨ttingen, Germany. E-mail: [email protected]

& 2014 The Society for Investigative Dermatology

COMMENTARY

In contrast, for germline mutations, predictive value has been established primarily for side effects of drugs. For example, for carbamazepine, severe drug reactions, such as Stevens-Johnson syndrome/toxic epidermal necrolysis, have been associated with HLAB*1502. The odds ratio for the occurrence of carbamazepin-induced Stevens-Johnson syndrome/toxic epidermal necrolysis in carriers of HLA-B*1502 is high: 236 (95% confidence interval 72– 778) in a meta-analysis of several Chinese populations (Yip et al., 2012).This finding has been translated into a recommendation by the US Federal Drug Administration to test for this MHC allele in patients of certain ancestry who are more likely to carry HLAB*1502 (for example, Han Chinese, Hong Kong Chinese, and patients of Thai origin). For these carriers, carbamazepine should be used only if benefits clearly outweigh the risk.

There is high linkage disequilibrium within the densely packed genomic region at chromosome 6p21.3. When investigating genetic variations as predictors of therapeutic efficacy in complex diseases, the candidate gene approach is selected most commonly. Obvious candidate genes include genetic variations associated with the disease, as well as variations in genes that code for molecules involved in the pathway targeted by the drug. For TNF, the candidate gene chosen by Murdaca et al. (2014) in this issue, both conditions hold true. The authors describe a trend for a higher frequency of the TNF SNP þ 489G allele in responders to etanercept therapy compared with nonresponders, but not in patients treated with adalimumab, in whom, by contrast, the genotype SNP þ 489AA was more frequent. It might be expected that the genetic bases of treatment responses are similar for all of the TNF-a blockers; however, molecular differences may www.jidonline.org 2483

COMMENTARY

CCHCR1

TNF

~3

~1

00

00

kb

kb

HLA-C

Block 1 (1 kb) 2 3

1 8

3 1

4

2 2

0

5 1

1 47

27

+489

–238

–308

–857

–1031

PSORS1

1 kb

6 1

2 98

3 1

Figure 1. Upper part: Genomic arrangement of genes on chromosome 6p21.3. Thicker sections of the horizontal line represent the genes for coiled-coil helical rod protein 1 (CCHCR1), HLA-C, and tumor necrosis factor (TNF). Large solid boxes represent exons, and small solid boxes represent the 30 -untranslated region (30 -UTR) or 50 -UTR. Arrowheads represent the orientation of the genes. Vertical bars indicate 7 of the 8 relevant single-nucleotide polymorphisms (SNPs), which are located at CCHCR1 (n ¼ 2), and TNF promoter and first intron (n ¼ 5). Arrows in the middle part indicate correlations of relative positions of SNPs between the upper and lower parts of the figure. Lower part: Linkage disequilibrium (LD) structure of the PSORS1 risk haplotype as defined by Huffmeier et al. (2009) and of selected SNPs at the TNF locus in 226 PsA patients analyzed as described previously (Reich et al., 2007). Genotypes of intronic variant at þ 489 were determined in a genome-wide association study published previously (Huffmeier et al., 2010). (a) Physical relationship of variants, (b) name of selected variants, and (c) pairwise LD plot of the PSORS1 risk haplotype and selected SNPs (illustrated by Haploview (Barrett et al., 2005). Each square plots the level of LD between a pair of sites in the region; comparisons between neighboring sites lie along the first line. Black color indicates strong LD, gray indicates intermediate or uninformative LD, and white indicates weak LD. The black triangle indicates neighboring SNPs within an LD block.

explain some of the differences in therapeutic responses, and possibly genetic predispositions. This holds true especially for differences between the fusion protein etanercept on the one hand and the mAbs adalimumab and infliximab on the other. Etanercept is a fusion protein that binds lymphotoxin-a (also termed TNF-b) in addition to TNF-a. Also, there are differences in pharmacodynamics between the mAbs and etanercept: in contrast to infliximab and adalimumab, etanercept is believed to bind only soluble and membrane-bound TNF trimers, but not monomers and dimers, and it does not appear to fix complement. This has led to the speculation that etanercept— as opposed to adalimumab and infliximab—is unlikely to form aggregates on the surface of TNF-producing cells, which ordinarily would activate complement-dependent lysis and

antibody-dependent cell-mediated cytotoxicity (Furst et al., 2006). The search for genetic variations influencing therapeutic efficacy, however, may also be influenced by factors not related primarily to the mode of a drug’s effect. An important mechanism for treatment failure of biologics is the formation of anti-drug antibodies. Even so-called ‘‘humanized’’ therapeutic antibodies—not containing any murine parts—may be recognized as ‘‘non-self’’ proteins. Thus, some anti-drug antibodies can neutralize the target drug and prevent therapeutic effects. Such antibodies can often be detected within the first 28 weeks of treatment in patients treated with adalimumab or infliximab, and they are associated with lack of efficacy. In case of adalimumab, they have been observed in up to 53% of patients (van Schouwenburg et al., 2013). In contrast, antibodies against

2484 Journal of Investigative Dermatology (2014), Volume 134

etanercept do not neutralize the drug’s effect and thus do not lead to treatment failure. Thus, anti-drug antibodies are relevant for infliximab and adalimumab but not for etanercept. Formation of anti-drug antibodies in itself may vary between carriers and non-carriers of certain genetic factors, and these factors may also contribute to genetically based lack of efficacy. Interestingly, a recent large multinational genome-wide association study analyzing anti-TNF-a drug responses in 2,703 patients with rheumatoid arthritis, using a multistep strategy, did not reveal an association between variants of TNF and treatment responses (Umicevic et al., 2013). Subgroup analyses for treatment with different TNF-a blockers in that study were not pursued (Umicevic et al., 2013). From the genetic point of view, PsA belongs to the category of ‘‘complex’’

COMMENTARY

diseases: an individual carrying several genetic risk factors develops the disease under the additional influence of (an) environmental risk factor(s). PsA’s main genetic risk factor is an HLA-C allele (HLA-Cw*06), explaining a considerable portion of heritability in PsA and even more in psoriasis vulgaris. In many independent candidate association studies followed by genome-wide approaches performed in PsA/psoriasis, this genetic risk factor has almost always been recognized as a major factor: (a) by genome-wide linkage studies as PSORS1 (psoriasis susceptibility locus 1) or (b) by genome-wide association studies as a very defined association signal at the HLA-C/HLA-B locus. According to the current state of knowledge, 144 variants are diseasecausing candidates, and, because of the location of a subset in an enhancer element, an influence on the expression of the HLA-C risk allele is suspected (Clop et al., 2013). However, the exact mechanism by which HLA-Cw*06 or other highly significantly associated HLA-C/HLA-B risk alleles and/or haplotypes (Chandran et al., 2013) contribute to PsA and psoriasis remains to be elucidated. In the past era of candidate gene studies, psoriasis vulgaris, but also PsA, has been independently associated with variants in the promotor region of the TNF gene coding for TNF-a (Reich et al., 2007; Giardina et al., 2011). As TNF is known to locate quite closely to the MHC at chromosome 6p21.3 (Figure 1a), it could be shown that association of the promotor variant at  238 is dependent on carriage of the HLA-C risk allele; the latter was more significantly associated with psoriasis and PsA and therefore the more probable disease-causing risk factor (Reich et al., 2007). This dependence is referred to as linkage disequilibrium. In contrast, the promotor variant at  857 was associated with PsA independently from carrier status of HLA-Cw*06. Interestingly, current data provide evidence that the promotor variant at  857 is in strong linkage disequilibrium to variant þ 489 (r2 ¼ 0.98, Figure 1b). Therefore, both variants at

the TNF locus, or perhaps their combination, are candidate variants. There is a high linkage disequilibrium within the densely packed genomic region at chromosome 6p21.3. Therefore, an isolated candidate approach at the TNF locus, without considering linked risk alleles, might lead to false-positive associations (Daly and Day, 2001). In this respect, it has been shown that the rarer allele of intronic variant þ 489 at TNF is also in linkage disequilibrium with certain HLA-B and HLA-D alleles (Low et al., 2002). Therefore, the association of intronic variant þ 489 with PsA susceptibility and better responses to TNF blockage in PsA is interesting (Murdaca et al., 2014), although there is as yet no evidence for a causal relationship. More comprehensive genetic studies are necessary to dissect the responsible candidate variants. A functional validation of the latter could then provide a biological link. Possibly in the future, large consortiums will be formed to investigate genetic predictors of therapeutic response to therapies in PsA, analogous to those investigating the genetic basis of the disease. The potential of saving costs of therapies likely to be ineffective or not tolerated by patients—in addition to a desire to provide the optimal therapy for an individual patient—may be sufficient incentive for governmental agencies to support such research. CONFLICT OF INTEREST

RM has received payments for conduction of clinical trials, as invited speaker or advisor, or travel grants from the following companies: Abbvie, Abbott GmbH & Co. KG, Biogen IDEC GmbH, Essex Pharma GmbH, Janssen-Cilag, Leo Pharma GmbH, Lilly GmbH, Merck Serono GmbH, Novartis Pharma GmbH, Pfizer GmbH, and Wyeth Pharma GmbH.

ACKNOWLEDGMENTS UH is supported by grants from the Deutsche Forschungsgemeinschaft (DFG 2163/1-1) and from the ELAN-Fonds of the Friedrich-Alexander-Universita¨t Erlangen-Nu¨rnberg (ELAN-13-03-09-1UH).

REFERENCES Barrett JC, Fry B, Maller J et al. (2005) Haploview: analysis and visualization of LD and haplotype maps. Bioinformatics 21:263–5 Chandran V, Bull SB, Pellett FJ et al. (2013) Human leukocyte antigen alleles and susceptibility

to psoriatic arthritis. Hum Immunol 74: 1333–8 Clop A, Bertoni A, Spain SL et al. (2013) An in-depth characterization of the major psoriasis susceptibility locus identifies candidate susceptibility alleles within an HLA-C enhancer element. PLoS One 8:e71690 Daly AK, Day CP (2001) Candidate gene casecontrol association studies: advantages and potential pitfalls. Br J Clin Pharmacol 52: 489–99 Furst DE, Wallis R, Broder M et al. (2006) Tumor necrosis factor antagonists: different kinetics and/or mechanisms of action may explain differences in the risk for developing granulomatous infection. Semin Arthritis Rheum 36:159–67 Giardina E, Huffmeier U, Ravindran J et al. (2011) Tumor necrosis factor promoter polymorphism TNF*-857 is a risk allele for psoriatic arthritis independent of the PSORS1 locus. Arthritis Rheum 63:3801–6 Huffmeier U, Lascorz J, Bohm B et al. (2009) Genetic variants of the IL-23R pathway: association with psoriatic arthritis and psoriasis vulgaris, but no specific risk factor for arthritis. J Invest Dermatol 129:355–8 Huffmeier U, Uebe S, Ekici AB et al. (2010) Common variants at TRAF3IP2 are associated with susceptibility to psoriatic arthritis and psoriasis. Nat Genet 42:996–9 Jang S, Atkins MB (2014) Treatment of BRAFmutant melanoma: the role of vemurafenib and other therapies. Clin Pharmacol Ther 95:24–31 Low AS, Azmy I, Sharaf N et al. (2002) Association between two tumour necrosis factor intronic polymorphisms and HLA alleles. Eur J Immunogenet 29:31–4 Murdaca G, Gulli R, Spano F et al. (2014) TNF-a gene polymorphisms: association with disease susceptibility and response to anti-TNF-a treatment in psoriatic arthritis. J Invest Dermatol 134:2503–9 Reich K, Huffmeier U, Konig IR et al. (2007) TNF polymorphisms in psoriasis: association of psoriatic arthritis with the promoter polymorphism TNF*-857 independent of the PSORS1 risk allele. Arthritis Rheum 56: 2056–64 Umicevic MM, Cui J, Vermeulen SH et al. (2013) Genome-wide association analysis of anti-TNF drug response in patients with rheumatoid arthritis. Ann Rheum Dis 72: 1375–81 van Schouwenburg PA, Krieckaert CL, Rispens T et al. (2013) Long-term measurement of anti-adalimumab using pH-shift-anti-idiotype antigen binding test shows predictive value and transient antibody formation. Ann Rheum Dis 72:1680–6 Yip VL, Marson AG, Jorgensen AL et al. (2012) HLA genotype and carbamazepineinduced cutaneous adverse drug reactions: a systematic review. Clin Pharmacol Ther 92: 757–65

www.jidonline.org 2485