Majority of the available drugs used for the treatment of RA were clinically evaluated in European ancestries, this raises a concern about their efficacy and toxicity in other ancestry groups globally29. Asian populations especially those in South East Asia were considerably under-represented in pharmacogenomic and pharmacogenetic studies of RA29.30. Hence, the present study evaluated the outcomes of MTX treatment in three major ancestry groups in Malaysia and their association with 6 SNPs from the enzymes involved in MTX metabolism. Comparing with the studies conducted by geographical locations, our study attempted to delineate ancestry specific risk factors that would increase the precision of the proposed association.
The MAF of ATIC T675C (rs4673993) recorded in Genome Aggregation Database (gnomAD) and 1000 Genomes is 0.3251 and 0.2855, respectively31,32. In our study, the allele frequency of ATIC T675C (rs4673993) for the overall cohort is 0.4 and it is 0.44, 0.31 and 0.54, respectively, in Malay, Chinese and Indian populations in Malaysia (Table 2). By comparing the allele frequency between our study and the public database, we noticed that our population is carrying a higher allele frequency of ATIC T675C (rs4673993). When comparing allele frequencies by ethnicity within our study cohort, there was a significant difference between Malay, Chinses and Indian for this SNP. Apparently, the allele frequency of ATIC T675C (rs4673993) in Indian and Malay subjects were significantly higher than that observed in Chinese subject and in the public databases.
Interestingly, our study suggested that the Malay RA patients with ATIC T675C (rs4673993) have a better treatment outcome upon MTX monotherapy. In other words, this minor allele was associated with an increased remission rate in Malay RA patients following the treatment of MTX. A few studies have also demonstrated the impact of ATIC T675C (rs4673993) on MTX treatment outcome. Prospective studies conducted by Lee et al. (2009) and Iannaccone et al. (2010) have shown that ATIC T675C (rs4673993) was significantly associated with low disease activity in RA patients with MTX monotherapy23,33. These two studies were conducted in the USA with 120 and 262 RA patients, respectively, as the subsets of Brigham and Women’s Hospital Rheumatoid Arthritis Sequential Study (BRASS). Moreover, a meta-analysis performed by Chen (2017) indicated that ATIC T675C (rs4673993) predicts the responsiveness of MTX treatment34. The authors combined two studies to yield a total of 698 Caucasians and observed a significant favouritism of ATIC T675C (rs4673993) in RA patients having response to MTX treatment. On the other hand, a retrospective study by Lima et al. (2014) gave a totally different conclusion35, whereby more than 4-fold increase in risk of MTX inefficacy was associated with ATIC T675C (rs4673993) in a population of 233 adults (\(\ge\) 18 y.o.) of Portuguese Caucasian RA patients. The result discrepancy may be due to different ancestral lineages of RA patients enrolled in the respective studies. In our case, the association of ATIC T675C (rs4673993) with the responsiveness to MTX treatment could only be observed in Malay but not in Chinese and Indian RA patients.
Current RA literature consistently highlights the hypothesis that the anti-inflammatory action of MTX is achieved through the indirect inhibition of AICAR. The ATIC T675C (rs4673993) SNP is positioned in the intronic region of ATIC. To our knowledge, there are no functional studies on this particular SNP in ATIC activity. Nevertheless, the intronic SNP either interferes the transcriptional regulation of the coding-enzyme or is in linkage disequilibrium (LD) with another coding SNP23,36−38. In the present study, since similar effect size of ATIC T675C (rs4673993) and ATIC C347G (rs2372536) was observed, both SNPs can be in LD. Nevertheless, this observation need further validation, since the current sample size is too small to perform a LD test and the lack of a reference panel for ATIC T675C (rs4673993) and ATIC C347G (rs2372536) in Malay patients.
Similar to ATIC T675C (rs4673993), the allele frequency of ATIC C347G (rs2372536) in Malay, Chinese and Indian populations is 0.45, 0.31 and 0.54, respectively; and the allele frequency of ATIC C347G (rs2372536) for the entire study cohort is 0.40 (Table 2). The allele frequency of ATIC C347G (rs2372536) retrieved from gnomAD, GO-ESP and 1000 Genomes are 0.3172, 0.2468 and 0.277831,32,39, respectively. Except the Chinese population, the allele frequencies observed for Malay and Indian are higher than the ones retrieved from the public databases.
Our result suggested Malay RA patients with ATIC C347G (rs2372536) having a better response to MTX treatment as compared to Chinese and Indian RA patients. In fact, this is in alignment with the data previously presented by Dervieux et al. (2004)36 on a cross-sectional study of 108 RA patients (age ≥ 18 y.o.) from a local rheumatology clinic in Knoxville, USA36. In their study, patients carrying a homozygous GG of ATIC C347G (rs2372536) may have a higher ratio of good response to MTX compared with patients carrying a CC or CG genotype. Moreover, Kurzawski et al. (2016)40 studied 422 Caucasian RA patients in Poland who were treated with MTX therapy and found that GG minor genotype significantly exhibited a good response to MTX. However, the lack of association between rs2372536 polymorphism and the clinical response to MTX was also reported in some studies41,42. Recently, two meta-analyses were performed to investigate the association between ATIC C347G (rs2372536) and MTX response43,44. The first meta‐analysis was based on five studies of 1056 RA patients in which 722 were MTX responders and 334 were non-responders. This analysis found the difference of ATIC C347G (rs2372536) between Caucasions (Spain, Slovenia and Netherlands) and Asians (India), being associated with non-responsiveness to MTX treatment in Caucasians but not associated in Asians43. The second meta‐analysis combined two European (Spain and Netherlands), one East Asian (Japan) and two South Asian (India) studies with 458 MTX responders and 398 non-responders in total44. When combining five studies, ATIC C347G (rs2372536) demonstrated a significant association with non-responsiveness of MTX under the dominant and codominant models. Yet, geographical stratification showed that the association of ATIC C347G with MTX response was still observed in Europeans in pre-allele, dominant and codominant models but not in South Asian populations44.
Despite all studies above demonstrated a significant association between ATIC variants and MTX efficacy, the results were raher inconsistent. Common factors for inconsistency such as small sample size and insufficient statistical power, study design, medication dosage, grouping criteria, and patient condition could cause limitations in the association study. Moreover, gene-gene interactions within folate and adenosine biosynthesis pathways may complicate the association study between SNPs and MTX treatment outcomes45. In fact, RA has complex inheritance patterns and no single genetic variant has a decisive role in MTX efficacy or MTX toxicity in the treatment of RA. By using the MDR (Multifactor Dimensionality Reduction) method, a cohort of 255 RA patients treated with MTX in the USA was evaluated with the efficacy of MTX treatment, and the results showed that 53% MTX responders was associated with high-order interactions among SNPs in ITPA (C94A), RFC1 (G80A), and ATIC (C347G) genes45. Upon excluding the predisposing genotype combinations, a 3.8-fold lower efficacy was observed45. Later, the same researchers extended their study of gene-gene interactions using ITPA (C94A), RFC1 (G80A), and ATIC (C347G) to another 3 RA cohorts (USA, Dutch and Swedish)46. Both USA and Dutch cohorts (n=435) confirmed a predisposing genetic attribute significantly associated with methotrexate response [odds ratio (OR)=2.9, 95% confidence interval (CI): 1.9-4.2; P<0.001]. Although the association of combined SNPs with MTX responsiveness in the Swedish cohort (n=530) could not be determined, the association was observed after the non-genetic factors, age, sex and anti-citrullinated protein antibody (ACPA) status were included in MDR analysis46. Thus, individual variants of ATIC may not play a direct role in MTX efficacy, future studies shall map the ATIC variants to drug response as based on the detection of nonlinear multigene interactions, this may improve the accuracy of predicting the MTX efficacy. In addition, other non-genetic covariates should be considered because the association study between genetic variants and MTX efficacy sometimes seems oversimplified understanding the MTX response in RA.
AICAR transformylase contains two domains which are MGS (methylglyoxal synthetase) like domain and AICAR binding domain47. ATIC C347G (rs2372536) causes the substitution of threonine (Thr) with serine (Ser) at position 116. Thr116 lies in the binding pocket of MGS-like domain and is the first residue of α8 helix which likely serves as a N-cap residue stabilizing the helix by interacting with the amide groups from the main chain. We proposed that the side-chain hydroxyl group of Thr116 forms hydrogen bonds with the amide groups of Val117 and Glu118 (green arrowhead in Fig. 4), while its main-chain carboxyl group forms hydrogen bond with the amide group from Glu119 (blue arrowhead in Fig. 4). The methyl group of Thr116 might stabilize the hydrogen bond between Thr116 and the main chain. As Thr116 is substituted with serine, the methyl group can be removed and this results in a more flexible C-N rotation. In other words, Ser116 causes the rearrangement of the protein structure at N-cap and thus, potentially affects the substrate-binding affinity and AICAR transformylase enzyme activity. This explains why the RA patients with the minor allele of ATIC C347G (rs2372536) might have a phenotypic change in response to MTX.