Pharmacogenetics and pharmacogenomics in rheumatology

Abstract

Pharmacogenetics and pharmacogenomics deal with possible associations of a single genetic polymorphism or those of multiple gene profiles with responses to drugs. In rheumatology, genes and gene signatures may be associated with altered efficacy and/or safety of anti-inflammatory drugs, DMARDs and biologics. In brief, genes of cytochrome P450, other enzymes involved in drug metabolism, transporters and some cytokines have been associated with responses to and toxicity of NSAIDs, cortisosteroids and DMARDs. The efficacy of biologics may be related to alterations in cytokine, chemokine and FcγR genes. Numerous studies reported multiple genetic signatures in association with responses to biologics, however, data are inconclusive. More, focused studies carried out in larger patient cohorts, using pre-selected genes may be needed in order to determine the future of pharmacogenetics and pharmacogenomics as tools for personalized medicine in rheumatology.

Introduction

Rheumatoid arthritis (RA) is an autoimmune inflammatory rheumatic disease that affects approximately 0.5–1% of the population and causes chronic synovial inflammation eventually leading to joint destruction and disability (1). Early diagnosis and immediate, effective therapy are crucial in order to prevent joint deterioration, functional disability and unfavorable disease outcome (1–6). The optimal management of RA is needed within 3–6 months after the onset of disease, therefore a very narrow “window of opportunity” is present to achieve remission or at least low disease activity (LDA) (2, 3, 5, 7, 8).

RA patients do not form a homogenous population. Several clinical subsets of RA, such as erosive versus non-erosive, anti-citrullinated protein/peptide antibody (ACPA) seropositive versus seronegative, progressive versus mild-course, etc have been identified (9–13). Disease progression, outcome and the RA phenotype have been associated with genetic factors (10, 14, 15).

As not all RA patients are the same, there is an increasing need to establish personalized medicine in rheumatology, as well as in other fields in medicine. Various patients may respond differently to traditional disease-modifying drugs (DMARDs) or biologic agents. As recently very elegantly described by Cronstein (16) as an analogy to fashion design, “prêt-a-porter” (ready-to-wear) clothing is designed for different people with different body shapes, while “haute-couture” (made-to-order) clothing is individually designed for a specific person. Personalized medicine uses “haute-couture” drugs fitted to the needs of an individual RA patient, while “prêt-a-porter” compounds would work well for most patients, but would be ineffective for some of them (16).

Pharmacogenetics and pharmacogenomics associate expression of single nucleotide polymorphisms (SNPs) or multiple genetic signatures with responses to drugs (17–20) (Tables 1 and 2). There have been major advances in this field yet, we are still very far from finding the optimal genetic biomarkers. In this review, we will discuss the current understanding of pharmacogenomics based on recently published data. Antirheumatic drug responses have also been associated with non-genetic factors that include clinical, cellular, serologic, histologic and other biomarkers (21, 22). These latter, non-genetic predictors will not be discussed here. Due to space limitations, we will have to focus on the most relevant information, therefore, we apologize for any important contributions that have been omitted.

Table 1.

Pharmacogenetics of traditional disease-modifying drugs^*

Drug	Gene	Variant	Clinical effects
Methotrexate	SLC19A1 (RFC-1)	80G>A	increased or unaffected efficacy
	MTHFR	677C>T	increased toxicity in most studies
	MTHFR	1298A>C	controversy regarding toxicity and efficacy
	SHMT1	1420C>T	increased toxicity
	ABCB1 (MDR1)	3435C>T	decreased efficacy
	TYMS	5′-UTR repeat element	decreased efficacy, probably increased toxicity
	TYMS	3′-UTR deletion	increased efficacy
	ATIC	347C>G	increased efficacy and toxicity in most studies
	IL1RN	IL-1RN*3	decreased efficacy
Sulfasalazine	NAT2	NAT2*4	increased toxicity in slow acetylators in most studies increased efficacy in rapid acetylators (?)
Leflunomide	DHODH	C19A	increased efficacy and toxicity
	ESR1	SNP	increased efficacy
	CYP1A2	CYP1A2*1F	increased toxicity
Hydroxychloroquine	IL10	1082A>G 819C>T 592C>A	increased efficacy
	TNF	−308A>G	increased efficacy
Azathioprine	TPMT	TPMT*2, *3A, *3C	increased toxicity

^*See text for abbreviations and further explanations

Table 2.

Pharmacogenetics of biologics^*

Drug	Gene	Variant	Clinical effects
Anti-TNF agents	TNF	−308G>A	increased efficacy in most studies
	TNF	−238A>G	increased efficacy
	TNFRSF1B	196T>G	decreased or no effect on efficacy
	FCGR3A	158V>F	no effect on efficacy
	PTPRC	SNP	increased efficacy
	MAPK14	SNP	increased efficacy of anti-TNF antibodies (infliximab, adalimumab)
Rituximab	FCGR3A	158V>F	increased efficacy or no effect

^*See text for abbreviations and further explanations. Large gene signature studies are not included

Therapeutic responses to anti-inflammatory drugs

Non-steroidal anti-inflammatory drugs (NSAIDs)

The cytochrome P450 enzyme family is highly involved in the metabolism of NSAIDs. The CYP2D6 gene has at least 70 variable alleles that often differ from each other in only one nucleotide. Some of these variants have been associated with slower metabolism of NSAIDs leading to abnormally increased serum concentration. On the other hand, formation of the active compound by enzymatic reaction will also be disturbed causing less efficacy (23, 24).

The CYP2C9 gene also exerts a number of functionally relevant polymorphisms. This enzyme has a key role in the metabolism of diclofenac, piroxicam and celecoxib. Therefore some polymorphisms has been associated with altered drug efficacy (25).

Corticosteroids

Drug transport through the cell membrane is particularly important in the case of intracellular receptor targets. One of the key enzymes involved in transmembrane transport is P-glycoprotein (P-gp), which protein is expressed on most somatic cells and regulates the intracellular migration, as well as excretion of various compounds. Pgp-1 has been implicated in the membrane transport of corticosteroids, such as cortisol, as well as in that of tacrolimus, cyclosporine A, colchicin and other agents (26–28). Polymorphisms in the P-gp (multidrug resistance, MDR1) genes, also known as the ABCB1 (ATP binding cassette transporter B1) gene, may lead to impaired activity of the transport enzyme (28). For example, 3435C>T, 2677G>T and 1236C>T polymorphisms in the P-gp/MDR-1 gene have been related to impaired transport activity and slower responses to corticosteroids in rheumatoid arthritis (RA) and other autoimmune-rheumatic diseases (29–32).

SNPs associated with responses to traditional DMARDs

Methotrexate

Methotrexate (MTX) is the most commonly prescribed disease-modifying agent (DMARD) used in the therapy of RA, lupus, and other rheumatic diseases (2). Several SNPs affecting genes of folate pathway, drug transporter, nucleotide synthesis and cytokine proteins may be related to MTX efficacy or safety (33) (Table 1).

MTX is a structural analog of folic acid that inhibits dihydrofolate-reductase and is also involved in cellular folate depletion and the inhibition of the methylene-tetrahydrofolate-reductase (MTHFR) enzyme. This enzyme synthesizes 5-methyl-tetrahydrofolate (5-MTHF), which serves as a methyl donor during the conversion of homocysteine to methionin. Mutations in the MTHFR gene may interfere with these physiological events thus leading to unexpected side-effects associated with MTX toxicity (34–37). The most widely studied SNPs in the MTHFR gene are 677C>T (33, 36, 37) and 1298A>C (33, 37–39) mutations. The 677C>T polymorphism leads to the expression of a thermolabile enzyme, impaired enzyme activity and hyperhomocysteinemia (37, 40, 41). This polymorphism has been associated with increased toxicity of MTX, but data have been inconclusive. Increased MTX toxicity in association with the 677T variant has been described in RA (36, 42). Symptoms including elevated transaminases, stomatitis, nausea, vomiting, alopecia and rashes were associated with hyperhomocysteinemia (34, 36, 42). This polymorphism also led to more frequent discontinuation of MTX treatment due to side effects (relative risk 2.4) (36). In a recent metaanalysis carried out in 1500 RA patients, however, no association between the 677C>T SNP and MTX toxicity could be observed (37). The 1298A>C SNP has also been associated with lower enzyme activity and hyperhomocysteinemia (37, 43–45). In contrast to the other polymorphisms, the 1298A>C SNP, despite hyperhomocysteinemia, does not seem to increase the frequency and severity of MTX toxicity (37, 44). Interestingly, the 1298A>C mutation has been related to increased cardiovascular morbidity in RA (45). Because of the paucity of pharmacogenetic data, further studies are needed to determine the role of MTHFR polymorphisms in the toxicity and efficacy of methotrexate.

The importance of intracellular molecular transport has been described above. One of the key enzymes involved in MTX transport is RFC-1 (reduced folate carrier 1). Genetic polymorphisms involving the RFC-1 (SLC19A1) gene lead to impairment or loss of RFC-1 expression and function (46). One of the most widely studied SNPs in this gene is 80G>A in codon 27 that results in an Arg to His change. RA patients that are 80A/A homozygous respond 3-times better to MTX than do 80G/G wild type homozygous individuals (47, 48).

Serine hydroxymethyltransferase (SHMT) is also involved in the transport of MTX into the cell. A 1420C>T polymorphism in the SHMT1 gene has also been associated with increased influx and toxicity of MTX (33).

The Pgp-1 (MDR) efflux pump, the ABCB1 gene encoding MDR1 and the 3435C>T SNP in this gene have already been described in relation to corticosteroids. Pgp-1 is also involved in MTX transport and increases MTX efflux. The 3435C>T SNP in exon 26 has been related to the clinical course of RA, as well as responses to MTX. Indeed, the 3435T allele may have a “protective” role as RA patients carrying this genotype exert milder disease course and may respond better to MTX (33, 49).

Regarding nucleotide synthesis genes, a key enzyme of de novo thymidylate biosynthesis is TYMS (thymidylate synthetase) that converts dUMP to dTMP. MTX interferes with TYMS activity and thymidylate synthesis. An active derivate of MTX, polyglutamate-MTX directly inhibits TYMS. Inhibition of TYMS leads to dTMP depletion, aggravated uracyl DNA incorporation and thus DNA injury and cell death. A tandem repeat sequence has been described at the 5′-UTR (untranslated) end of the TYMS gene (50). RA patients carrying a triple repeat of this sequence in homozygous form (TSER*3/*3), exert elevated TYMS mRNA expression and require higher MTX doses in order to achieve therapeutic response. In contrast, a six-base pair deletion in the 3′-UTR region of this gene yields impaired TYMS mRNA stability and lower gene expression. Patients carrying this latter genotype will respond extremely well to conventional MTX doses (33, 51, 52).

Another key enzyme involved in MTX metabolism is ATIC (amino-imidazole carboxamide ribonucleotide transformylase) that converts AICAR (amino-imidazole carboxamide ribonucleotide) to 10-formyl-AICAR. MTX inhibits AICAR and inhibition leads to increased adenosine and AICAR levels. A 347C>G SNP in the ATIC gene resulting in a Thr-Ser change has been associated with increased MTX efficacy but also increased toxicity in most studies (48, 53, 54)

With respect to cytokine genes, a SNP in the IL1RN gene coding interleukin-1 receptor antagonist (IL-1Ra or IL-1RN) was also associated with decreased efficacy of MTX (33, 55).

Sulfasalazine

A key enzyme in sulfasalazine (SASP) metabolism is N-acetyltransferase (NAT) that exerts two major types, NAT-1 and NAT-2. The process of acetylation is characterized by the activity of NAT. Individuals with low or high enzyme activity are slow and fast acetylators, respectively. Multiple SNP-s have been described in the NAT2 gene leading to decreased enzyme activity. In Caucasian and African populations the prevalence of slow acetylators is around 60%, while in Asian populations it is 20%. In slow acetylators, SASP therapy is associated with more frequent side effects including nausea, vomiting, headache, hemolytic anemia or reticulocytosis. Yet, conventional doses of SASP seem to be more effective in slow acetylator RA patients (33, 56, 57) (Table 1).

Leflunomide

Leflunomide is a synthetic isoxazole. Its active metabolite, A77 1726, is responsible for its immunosuppressive and anti-inflammatory effects. This compound inhibits dihydro-orotate dehydrogenase (DHODH) and thus the synthesis of pyrimidine nucleotides (58). Increased efficacy of leflunomide has been linked to variations in the DHODH (C19A), estrogen receptor (ESR1) or the cytochrome P450 (CYP1A2) genes (33) (Table 1).

Azathioprine

Azathioprine (AZA) and its metabolite, 6-merkaptopurine (6-MP) have been widely used to treat various autoimmune-inflammatory diseases, such as lupus, Crohn’s disease, juvenile arthritis, systemic vasculitis, other connective tissue diseases and, less frequently, RA. Despite its efficacy, AZA exerts several side-effects including gastrointestinal intolerance, myelosuppression and cytopenias. AZA is an inactive “prodrug” that is converted in the body to 6-MP. This transition is mediated by the glutathione-S-transferase enzyme, as well as by non-enzymatic processes. The active compound inhibits aminotransferases and thus purine biosynthesis (59–61).

Two enzymes, thiopurine-methyl-transferase (TPMT) and xanthine-oxydase (XO) are involved in the metabolism of 6-MP yielding less active or inactive metabolites, such as methyl-MP or thiourate. Genetic polymorphisms in the genes of these enzymes will lead to the accumulation of toxic products and drug side effects. The most severe toxic side effect of AZA, myelosuppression, is explained by disturbed purine metabolism due to TMPT SNP and deficient TMPT activity. In addition, patients with TMPT deficiency exert a 3-times increased risk for gastrointestinal side effects (33, 59, 61). Decreased TMPT activity due to this genetic alteration is observed in about 10% of Caucasians and Africans (59, 61) (Table 1).

Other DMARDs

Mutations in the IL10 (1082A>G, 819C>T, 592C>A) or TNF gene (−308A>G) yield to increased efficacy of hydroxychloroquine (HCQ) (33).

The role of Pgp-1 (MDR1) protein has been described above. The 2677G>T/A and 3435C>T SNPs in exon 21 and exon 26 of the MDR1 gene, respectively, have been associated with efficacy of cyclosporine A (27, 62).

Pharmacogenomics in biological therapy

SNP associations with responses to biologics

Recently, therapeutic responses to biologic agents have drawn more attention. According to the recent treat-to-target (5) and EULAR/ACR treatment recommendations (2), treatment strategies should aim at remission or at least LDA. While the administration of anti-tumor necrosis factor α (TNF-α) agents yields to sustained efficacy in many patients, remission or LDA cannot be reached in as much as 60% of RA patients (63). Reasons for switching biologics include primary lack and secondary loss of response, partial response or side effects (63). There is an immense need to find biomarkers including pharmacogenomic approaches in order to predict inefficacy or side-effects early (Table 2).

First, studies of single SNPs were conducted followed by more detailed pharmacogenomic investigations on multiple signatures. Most studies have been conducted on the −308G>A SNP in the TNF gene of TNF-α. This SNP has been associated with clinical responses to infliximab, etanercept and adalimumab in RA, as well as spondyloarthropathies. In most studies, the GG genotype was associated with poorer, while the AA genotype with better efficacy (reviewed in (21, 33)) (Table 2).

Fcγ receptors (FcγR) expressed on mononuclear cells are involved in several mechanisms underlying cellular immunity including cell-mediated and complement-dependent cytotoxicity, apoptosis and immune complex clearance. Structural changes in the FcγR molecule alter cellular interactions and functions. A polymorphism in the FCGR3A gene encoding FcγRIIIA at position 158, leading to an amino acid change from Val to Phe (158V>F) results in weaker binding and altered response to biologics. The homozygous F/F phenotype confers more favorable responses to infliximab, adalimumab and etanercept (64–66) (Table 2).

Other polymorphisms associated with clinical outcome during anti-TNF treatment include other changes in the TNF locus (e.g. −238A>G), 196T>G change in the TNFRSF1B gene and a SNP in the receptor-type tyrosine-protein phosphatase C gene (PTPRC) (reviewed in (33)). SNPs in the MAPK14 gene yield to increased efficacy of anti-TNF antibodies, but not etanercept (33, 67) (Table 2).

Responses to the B-cell inhibitor rituximab have been linked to SNPs in the FCGR gene encoding FcγRIIIA (158V>F) already described above (68), a SNP in the IL6 gene (174G>C) (69) and in the gene of the B lymphocyte stimulator (BLyS) (70) (Table 4) (Table 2).

Genetic signatures that may determine responses to biologics

Expression patterns of multiple genes and genetic signatures have been studied in association with responses to biologics using microarray technique. Most studies have been conducted on infliximab and etanercept. Unfortunately, these studies yielded very inconclusive results as different groups reported different signatures. These differences could come from biological and technical reasons; different source of samples were examined e.g. synovial tissue/cells or blood samples and not the same microarray type was used in different experiments. There have been many studies, but we present only a few examples.

Genes that have been associated with responses to etanercept include primarily cytokine genes (TNF, lymphotoxin α [LTA], IL10, TGFB1, IL1RN), cytokine receptor genes (TNFRSF1A, TNFRSF1B) and Fc receptor alleles (FCGR2A, FCGR3A, FCGR3B) (71). In the a study conducted in 19 patients treated with etanercept, good clinical response was associated with the CCL4, IL8, IL1B, TNFAIP3 and some other genes (72). Four SNPs (TNF, IL10, TGFB1, IL1RN) have been studied in relation to etanercept responses in 123 RA patients. Some TNF and IL10 SNPs could be associated with better responses, while the combination of IL1RN and TGFB1 SNPs indicated unfavorable responses (73). In 457 early RA patients, HLA-DRB1, TNF and LTA SNPs were related to clinical response to etanercept (71).

With respect to infliximab, in a study conducted on 44 RA patients, clinical responses were associated with eight genes including HLADRB3 (74). In another cohort, 41 genes including HLADPB1 and PTPN12 could be linked to infliximab response. The function of most genes related to clinical responses has not yet been identified in RA (75). Polymorphisms in absolutely different genes including chemokine and chemokine receptor genes (CCL4, CX3CR1) and interferon-induced genes were associated with responses to infliximab in yet another study (76). Pharmacogenomic studies in infliximab-treated patients have been conducted in several other cohorts (77).

To date the largest published pharmacogenomic GWAS study on biologics included 566 RA patients. Changes in altogether 460000 SNPs were assessed in patients treated with etanercept, infliximab or adalimumab. Prediction analysis revealed 7 loci that may be related to responses to these biologics. Two loci, PDZD2 encoding PDZ domain-containing protein 2 and EYA4 encoding the protein eyes absent 4 showed the strongest correlation with biologic responses. Yet, the role of these two genes in RA is unknown (19, 78, 79).

Thus, polymorphisms in the TNF-LTA region, the IL10 gene, the IL-1 receptor antagonist gene IL1RN, members of the MAP kinase signalling network, TLR and NFκB pathways, FcγRIIIA (FCGR) polymorphism or other genes of unknown function in RA have been associated with anti-TNF treatment responses (33, 66, 71, 76, 79–81). Interestingly, HLA-DRB1 and PTPN22 alleles were not associated with treatment responses (79, 82).

Regarding other, non-anti-TNF biologics, few studies have been performed in rituximab-treated patients. In one of these studies, baseline expression of type I interferon signature genes was associated with non-response to rituximab (83).

Very recently, we have conducted the first pharmacogenomic study on response to the IL-6 receptor antibody tocilizumab. In 13 tocilizumab-treated RA patients, the expression of 59 genes showed significant changes between baseline and after 4 weeks of treatment. Four genes (DHFR, CCDC32, EPHA, TRAV8) determined responders after correction for multiple testing. Again, very little is known about the function of these genes in RA. Dihydrofolate reductase (DHFR) was identified as putative predictor for methotrexate response, T cell receptor alpha variable 8-3 (TRAV8-3) is involved in CD8+ T-cell responses, the ephrin receptor A4 (EPHA) plays a role in the nervous system, while the functionality of CCDC32 in arthritis is unknown (84).

Practical aspects of the clinical use of pharmacogenetics

As described above, there have been certain controversies regarding the applicability of pharmacogenetics and pharmacogenomics in rheumatology. Apart from associations with some well-characterized SNPs, numerous studies of multiple genetic patterns yielded results difficult to interpret. The role of some genes identified in relation to drug responses has not yet been determined in rheumatology. Therefore at least the further issues should be addressed in future studies to be conducted in larger arthritis cohorts:

• which are the functionally most important genes and signatures;

• the expression of which genes and gene profiles would change upon therapy;

• what baseline gene expression patterns would predict the therapeutic response;

• what is the relevance of identified genes with yet unknown function in (75, 78);

• are there geographical differences in pharmacogenetic associations;

• how would epigenetic, environmental factors (e.g. smoking) alter these pharmacogenetic associations (85),

• how could population-based pharmacogenomic results be adapted for the individual patient: is there a real future for personalized medicine?

Most modern techniques needed to address these issues are at our disposal. It is simple to collect peripheral blood, isolate mononuclear cells (PBMC) and use these cells for pharmacogenetic studies. Circulating PBMCs are highly involved in the initiation and maintenance of inflammation. PBMC-based microarray systems have been validated and are readily available. As usually not a single gene but multiple genes are involved in the pathogenesis of inflammatory diseases, gene expression profiling and signature studies should lead us to the better understanding of therapeutic responses (71, 79, 86–88). Pre-selection of genes with known function in arthritis, rather than the use of platforms with thousands of unselected genes could yield more specific results (71, 73, 87). Numerous previous pathogenic and recent genomic studies suggest that mostly genes of pro-inflammatory cytokines, chemokines and FcγRs should be investigated in the first place (18, 78, 79, 88).

Due to the increasing number of studies focusing on next generation sequencing for the diagnosis and stratification of medical conditions, the approach of analyzing single genes, variants or sets of genes might shift towards analyzing the whole genome at once in the near future. Another option that has been used in numerous papers in the last few years is finding interactions between microRNAs and changes in the phenotype of the disease with a special focus on rheumatology (89). Incorporating this huge amount of data into pharmacogenomnics decision support is clearly a challenge in this process (90).

Conclusions

Pharmacogenetics and pharmacogenomics may bring us closer to personalized medicine, however, recent studies yielded to inconclusive results reporting substantially different gene patterns in various cohorts. In addition, although numerous genes have been picked up that may predict therapeutic responses, the role of many of these genes in the pathogenesis of arthritis is unknown. Therefore, the breakthrough is yet to come. Further studies, possibly using pre-selected, well-characterized gene sets are needed to determine the value of pharmacogenomics in rheumatology.