Exome Sequencing and Array-Based Comparative Genomic Hybridisation Analysis of Preferential 6-Methylmercaptopurine Producers

Home , FAM46A, MOCS3, RCC2, Thiopurine

ORIGINAL ARTICLE Exome sequencing and array-based comparative genomic hybridisation analysis of preferential 6-methylmercaptopurine producers

EW Chua1,2, S Cree1, ML Barclay3,4, K Doudney5, K Lehnert6, A Aitchison1 and MA Kennedy1

Preferential conversion of azathioprine or 6-mercaptopurine into methylated metabolites is a major cause of thiopurine resistance. To seek potentially Mendelian causes of thiopurine hypermethylation, we recruited 12 individuals who exhibited extreme therapeutic resistance while taking azathioprine or 6-mercaptopurine and performed whole-exome sequencing (WES) and copy- number variant analysis by array-based comparative genomic hybridisation (aCGH). Exome-wide variant ﬁltering highlighted four genes potentially associated with thiopurine metabolism (ENOSF1 and NFS1), transport (SLC17A4) or therapeutic action (RCC2). However, variants of each gene were found only in two or three patients, and it is unclear whether these genes could inﬂuence thiopurine hypermethylation. Analysis by aCGH did not identify any unusual or pathogenic copy-number variants. This suggests that if causative mutations for the hypermethylation phenotype exist they may be heterogeneous, occurring in several different genes, or they may lie within regulatory regions not captured by WES. Alternatively, hypermethylation may arise from the involvement of multiple genes with small effects. To test this hypothesis would require recruitment of large patient samples and application of genome-wide association studies.

The Pharmacogenomics Journal (2015) 15, 414–421; doi:10.1038/tpj.2015.9; published online 10 March 2015

INTRODUCTION retrospective study (n = 1879), 2.5% of New Zealand individuals Azathioprine and 6-mercaptopurine structurally resemble endo- receiving thiopurine treatment were found to have a very high 8 genous purines, differing from the latter at the sixth carbon atom, 6-MMP/6-TGN ratio of 4100. The role of TPMT in thiopurine 8,9 where the substituent is replaced by a sulfhydryl-based functional hypermethylation is uncertain, and involvement of other 10,11 group. Hence, these compounds are categorically designated genes has not been firmly established. Overall, the mechan- ‘thiopurines’. Intracellularly, azathioprine and 6-mercaptopurine isms underlying such metabolite shunting remain unclear. are converted by successive metabolic steps into 6-thioguanine Whole-exome sequencing (WES), which enables simultaneous nucleotides (6-TGN), which are mis-incorporated into DNA during sequencing of all protein-coding regions within the genome, has fi DNA synthesis, disrupting the process and causing cytotoxicity.1 identi ed several novel pharmacogenetic markers such as those involved in the response to clopidogrel,12 escitalopram,13 Other pathways competing with the formation of 6-TGN include 14 15 methylation or oxidation of 6-mercaptopurine, mediated by clozapine and anticancer agents (pazopanib and everolimus). thiopurine methyltransferase (TPMT) and xanthine dehydrogenase We set out to investigate the possible genetic basis of metabolite (XDH) respectively, as illustrated in Figure 1. Historically, use of shunting observed in a group of patients who exhibited resistance fl towards thiopurine therapy, by using WES technology. These thiopurine compounds in in ammatory bowel disease has been ‘ ’ noted particularly for its steroid-sparing effects,2,3 despite patients are referred to as extreme shunters in this report. Our 4 main aim was to test whether this phenotype was a Mendelian potentially fatal immunosuppressive complications. trait due to mutation of a gene involved in thiopurine metabolism, Interindividual variability in response to thiopurine compounds transport or function. has been studied extensively. In addition to TPMT, new pharmacogenetic markers are emerging.5 However, much of the variability still eludes current understanding in pharmacogenetics. Thera- MATERIALS AND METHODS peutic monitoring in individuals receiving thiopurine therapy is Patient recruitment achieved by determining the relative concentrations of 6-TGN and 6-methylmercaptopurine (6-MMP),6 and dose adjustment can help This study was approved by the Southern Health and Disability Ethics Committee, New Zealand. A total of 12 patients who had 6-MMP/6-TGN improve drug response and minimise potential treatment 4 fi 7 ratios 100 while on thiopurine therapy were identi ed from a previous toxicity. Individuals who preferentially methylate thiopurine local study.8 These patients reported European descent, with the exception compounds into 6-MMP, as evidenced by a high 6-MMP/6-TGN of one male who reported mixed ancestry. Potential participants were ratio (420), are predicted to have a non-favourable response. In a contacted first by mail and were required to indicate interest to participate

1Carney Centre for Pharmacogenomics and Department of Pathology, University of Otago Christchurch, Christchurch, New Zealand; 2Faculty of Pharmacy, Universiti Kebangsaan Malaysia, Kuala Lumpur, Malaysia; 3Department of Clinical Pharmacology, Christchurch Hospital, Canterbury District Health Board, Christchurch, New Zealand; 4Department of Medicine, University of Otago Christchurch, Christchurch, New Zealand; 5Molecular Pathology, Canterbury Health Laboratories, Christchurch, New Zealand and 6School of Biological Sciences, University of Auckland, Auckland, New Zealand. Correspondence: Professor MA Kennedy, Carney Centre for Pharmacogenomics, University of Otago Christchurch, Christchurch, New Zealand. E-mail: [email protected] Received 16 June 2014; revised 15 January 2014; accepted 28 January 2015; published online 10 March 2015 Exome sequencing analysis of 6-methylmercaptopurine overproducers EW Chua et al 415

Figure 1. A simpliﬁed representation of major thiopurine metabolic pathways (adapted from Zaza et al.32).

Table 1. Candidate genes with involvement in thiopurine transport, by filling in and returning an enclosed form. Face-to-face interviews were subsequently conducted to obtain written consent and collect relevant metabolism or therapeutic action medical history. The study information sheet and consent form included procedures for handling of incidental findings, which would be followed- Candidate genes Ref. up in consultation with a clinical geneticist. DNA was extracted from Thiopurine transport, metabolism or target: ABCC4, ABCC5, 32 peripheral blood leukocytes using a KingFisher Flex Magnetic Particle ADA, ADK, AHCY, AOX1, GART, GMPS, GSTA1, GSTA2, GSTM1, Processor, as per the manufacturer’s instructions (Thermo Fisher Scientific, HPRT1, IMPDH1, ITPA, NT5E, PPAT, PRPS1, RAC1, SLC28A2, Waltham, MA, USA). SLC28A3, SLC29A1, SLC29A2, TPMT, XDH Genes identified through gene expression profiling: CD1D, 11 TPMT activity CTSS, DEF8, FAM46A, FAM156A, FAR1, GNB4, HVCN1, IMPDH2, Available TPMT activity data were also documented for each participant. A LAP3, MAP3K1, PLCB2, SLX1A, SMAP2,TGOLN2,TOX4,TUSC2,UBE2A Synthesis of S-adenosylmethionine: TYMS, MAT1A, MAT2A, previous study had established the normal range of TPMT activity to be 33–35 9.3-17.6 IU ml − 1 in the local New Zealand population.16 The distribution of MTHFR Molybdenum cofactor activity: MOCOS 36 TPMT activity among New Zealand individuals is similar to that in other 37 populations.17 Modulation of TPMT activity: PACSIN2 Efflux of methylated metabolite: ABCB5 38 Others: GDA, MTAP, NT5C2 39 Exome sequencing Whole-exome enrichment and sequencing was done by New Zealand Abbreviation: TPMT, thiopurine methyltransferase. Genomics Ltd. (Dunedin, New Zealand). Twelve TruSeq (Illumina, San Diego, CA, USA) DNA libraries were prepared and sequenced on one lane of a HiSeq 2000 (Illumina) to produce paired-end reads that were 100 The reduced variant file was annotated using SeattleSeq Annotation 138 nucleotides long. (accessible at http://snp.gs.washington.edu/SeattleSeqAnnotation138/),27 build 9.02 (updated on 20 April 2014). The annotated file was then Quality control, sequence reads alignment and variant calling successively filtered based on the Combined Annotation-Dependent 18 Depletion (CADD) C scores, which include functional and conservation Sequence reads were aligned by Burrows-Wheeler Aligner v0.7.4 to the 28 human GRCh37.p13 reference assembly and processed with SAMtools metrics, and population MAFs documented in the HapMap and Exome v0.1.1919 and Picard v0.96 (http://picard.sourceforge.net). Reads originating Variant Server (EVS) databases. CADD C scores cannot be generated for from PCR duplicates were removed with Picard before and after local indels; as a result, this group of variants were treated separately and were filtered by their Genomic Evolutionary Rate Profiling (GERP) scores,29 in realignment around potential insertions/deletions (indels) with the 30 20 accordance with the recommended cut-off value. Other functional Genome Analysis Toolkit (GATK, v2.7.1). Illumina base quality scores 31 were recalibrated with GATK in the final alignments. Single-nucleotide annotations, including the PolyPhen-2 predictions, were also considered. variants (SNVs) and indels were called simultaneously via local de novo assembly of haplotypes using GATK’s HaplotypeCaller tool.21 Joint variant Candidate-gene analysis calling and variant quality score recalibration were supported by adding A comprehensive list of 52 candidate genes was compiled from a literature exome alignments from 24 European individuals obtained by the same search encompassing thiopurine transport, metabolism and targets, as well sequencing and alignment procedures. These 24 exomes originated from as the genes observed in an expression profiling study, and other genes another local study that was unrelated to thiopurine resistance and were with relevant functions11,32–39 (Table 1). UCSC Genome Browser’s Table used as a filter for false-positives that may be population- or platform- Browser was used to extract genomic coordinates for these candidate specific. For instance, variants that were present at a similar frequency in genes from the track ‘RefSeq Genes’. The resultant intervals file was used to both the extreme shunters and this control cohort were deemed likely to extract variant calls originating only from the candidate genes. Extracted be harmless. variant calls were subsequently annotated using SeattleSeq Variant Annotation 138. The output file was then filtered as described for whole- fi Whole-exome variant filtering and annotation exome variants, except that singleton variants were not removed. The nal gene lists generated by exome-wide filtering and candidate-gene analysis Multiple pre-annotation variant-reduction strategies were adopted, using ’ 22–24 were also cross-compared with known disease-related genes for Crohn s GATK built into the web-based tool Galaxy. First, variant sites that fell 40,41 – – disease, ulcerative colitis and autoimmune hepatitis, and these were within truth-sensitivity tranches of 99 99.9 and 99.9 100 were removed; highlighted, if present. we designated the retained variants as ‘pass-filter’ variants. Low-coverage sites (o100-fold) were also filtered out, by assuming a minimum acceptable read depth of 10-fold per sample. Then common variants Array-based comparative genomic hybridisation (aCGH) having a global minor allele frequency (MAF) 410%, as defined in the Whole-genome copy-number variation (CNV) was examined using the Phase I 1000 Genomes Project,25,26 were discarded in order to acquire a NimbleGen CGX12 Comparative CGH Microarray (Roche NimbleGen, manageable, condensed set of variants that are more likely to be Madison, WI, USA), as per the manufacturer’s instructions. Briefly, subject deleterious. The final step of variant-reduction comprised removal of DNA samples and reference DNA samples were first labelled with Cy3 and variants that were present in only one subject. Cy5 dyes, respectively. The reference DNAs used were pooled genomic

© 2015 Macmillan Publishers Limited The Pharmacogenomics Journal (2015), 414 – 421 Exome sequencing analysis of 6-methylmercaptopurine overproducers EW Chua et al 416 DNAs from Promega (Madison, WI, USA). Equal amounts of subject and reference DNAs were mixed and hybridised onto the microarray for 48 h. ) The microarray was washed and then scanned using the SureScan 1 − Microarray Scanner (Agilent Technologies, Santa Clara, CA, USA). Raw scan data were processed in the NimbleScan v2.6 software and loaded onto the (IU ml

web-based Genoglyphix Analysis Software (PerkinElmer, Waltham, MA, TPMT activity USA) for viewing. Only CNV calls with a probe count 45 were evaluated. The predicted pathogenicity of identiﬁed CNVs was assessed by comparison with the Database of Genomic Variants (DGV), which contains a collection of benign structural variants found in healthy individuals.42 131.28 14.9 153.37 12.1 Statistical analysis 212.36 17.0 Case variant frequencies (allele counts) were compared with those (highest recorded)

recorded in the EVS database using the chi-square test of independence 6-MMP/6-TGN ratio with Yates’ correction.43 A P-value of o0.05 was considered statistically signiﬁcant.

RESULTS

Clinical characteristics ed ﬁ Clinical details of the participants are shown in Table 2. Nine Addition of allopurinol? discontinued modi females and three males were recruited for exome sequencing. discontinued The primary indications for initiation of thiopurine treatment were Crohn’s disease (n = 8), autoimmune hepatitis (n = 3) and ulcerative colitis (n = 1). All participants had previously been tested for TPMT activity and noted to be normal metabolisers. However, following administration of thiopurine compounds they exhibited extremely high 6-MMP/6-TGN ratios 4100, resulting in therapeutic modiﬁcations, which largely comprised addition of allopurinol.

Exome-wide filtering Exome analysis of the 12 participants revealed a total of 118,784 pass-filter variant sites located within the genomic regions directly Experienced adverse effect(s) when on thiopurine therapy? targeted by the exome capture kit. These variants were successively filtered by site coverage (4100), global MAF (⩽10%) and intra-group allele frequency (required to be present in at least two participants) and then annotated by the web server SeattleSeq Annotation 138, as illustrated in Figure 2. SNVs were prioritised based on their CADD C scores, and indels were filtered fi s disease Yes No; drug s disease No No; treatment s disease Yes Yes 192.87 13.5 s disease Unknown Yes 126.88 17.8 s disease Yes No 101.53 16.3 by their GERP conservation scores (Figure 2). This produced a nal s disease Unknown Yes 161.35 16.6 ’ ’ ’ ’ ’ ’ candidate-gene list, which consisted of 1176 variants most likely to be of functional significance. Putative functions of implicated thiopurine compounds genes were then manually curated from the Universal Protein Crohn Resource Knowledgebase (UniProtKB; accessed at http://www. uniprot.org/)44 and screened for possible association with thiopurine hypermethylation. To avoid promiscuous inclusion of a large number of genes that may seem functionally relevant, several key terms were applied, namely ‘purine/nucleoside transport or metabolism’, ‘urate transport or metabolism’, ‘folate transport or metabolism’, ‘erythrocyte membrane’, ‘Rho-type or ’ Ethnicity Primary indication for Rac GTPases and genes interacting with those already known to Scandinavian be involved in thiopurine disposition (Table 1). The majority of the genes identified via this whole-exome filtering approach appeared unrelated to the phenotype in question (Supplementary Table S1). However, five genes (ENOSF1, GDA, NFS1, RCC2 and SLC17A4) that may indirectly influence thiopurine metabolism, transport or therapeutic action were observed. Each of these genes contained heterozygous variants with a low population frequency (MAFo2%) that occurred in two or three patients (Table 3). recruitment, years Additional variants with a higher MAF (43%), listed in Supplementary Table S2, were likely to have arisen by chance (P40.05). It should be noted that GDA is one of the a priori candidate genes. Variants in RCC2, SLC17A4, ENOSF1 and NFS1 occurred at a frequency that was significantly higher than expected (Po0.05). Moreover, these variants were identified in Relevant clinical history of participants recruited for exome sequencing none of the individuals from the in-house control cohort, suggesting that they may have an equally low frequency in the Participant Sex Age at the time of ES001ES002ES003ES004 Female Female Female Female 49 56 40 63 European European European European Crohn's disease Autoimmune hepatitis Crohn's disease Crohn No No Yes Yes Yes Yes 145.94 180.86 198.80 15.9 11.9 11.4 ES005ES006 Male Male 57 49 European European Ulcerative colitis Crohn No Yes 163.34 13.3 ES007 Male 49 French, Arabic and ES008 Female 44 European Crohn ES009 Female 43 European Crohn ES011 Female 40 European Crohn ES010 Female 49 European Autoimmune hepatitis Yes No; drug ES012 Female 51 European Autoimmune hepatitis Yes Yes 170.78 9.0 New Zealand population. Table 2. Abbreviations: TPMT, thiopurine methyltransferase; 6-MMP, 6-methylmercaptopurine; 6-TGN, 6-thioguanine nucleotides.

Figure 2. Whole-exome filtering strategies employed to identify novel candidate genes that may be related to response towards thiopurine compounds. Variants with excess heterozygosity, reference bias or no annotated gene name were excluded. The first branch point in the pathway reflects the fact that the Genome Analysis Toolkit cannot perform allele-frequency filtering for multiallelic variants. CADD, Combined Annotation-Dependent Depletion; EVS, Exome Variant Server; GERP, Genomic Evolutionary Rate Profiling; indel, insertion/deletion; MAF, minor allele frequency; SNV, single-nucleotide variant; UniProtKB, the Universal Protein Resource Knowledgebase.

Candidate-gene analysis GDA, ENOSF1 and NFS1) were validated by Sanger sequencing For analysis of the 52 predefined candidate genes, a slightly more (Supplementary Figure S1). relaxed filtering approach was adopted, whereby singleton variants were included. As shown in Table 4, a total of 13 variants Copy-number analysis by aCGH spread across 12 different genes were identified, not counting the CNVs are an additional source of genetic variation with potential GDA variant reported in Table 3. Some of these variants were pharmacogenetic significance,45 and these would not be readily noted to have a very low population MAFo1%. Most variants detected by exome sequencing. Therefore, we applied the listed in Table 4 were singletons. When a similar filtering approach method of aCGH to the DNA of the 12 extreme shunters. This was applied to the in-house control cohort, a comparable number analysis did not identify any potentially pathogenic or unusual of candidate-gene variants were identified in these individuals (13 CNVs. As shown in Supplementary Table S4, a total of 61 variants in 12 cases versus 29 in 24 control individuals). For duplication/deletion events were identified in our subjects, but instance, the GDA variant at the genomic position 9:74,840,668 the majority (n = 48) of these events are in the DGV database, was also identified in one individual from the control cohort. indicating that they are probably innocuous. For instance, a Therefore, the candidate-gene variants discovered in the prevalent CNV, which was present in 10 of the 12 extreme extreme shunters seem unlikely to be major contributors to the shunters, involved two genes of little known biological signifi- hypermethylation phenotype. All genotype calls of variants cance, namely ADAM5P and tMDC I. ADAM5P is a pseudogene, identified from exome-wide filtering and candidate-gene analysis whereas tMDC1 encodes non-functional transcripts,46 so they are were noted to be of high quality, having a genotype quality score unlikely to be relevant to thiopurine hypermethylation. The of 430.0 (Supplementary Table S3). As an additional quality- remaining 13 events were classified as possibly benign, because control check, the five exome-wide variants (RCC2, SLC17A4, they were larger in size, or were of a different CNV type, when

© 2015 Macmillan Publishers Limited The Pharmacogenomics Journal (2015), 414 – 421 Exome sequencing analysis of 6-methylmercaptopurine overproducers EW Chua et al 418 compared with their counterparts in the DGV database (Supple- mentary Table S4). 0.0001 o DISCUSSION

d Despite efforts to elucidate the complexity of thiopurine response,47,48 we still do not understand the biological process that gives rise to thiopurine hypermethylation. In the present

MAF (EVS) study, we recruited 12 individuals who exhibited extreme thiopurine resistance and performed WES to search for highly penetrant mutations that may cause the phenotype. The total Allele count P-value

c number of gene variants called was comparable to a previous report.14 Following whole-exome ﬁltering, we identiﬁed low- ed in none of the individuals from the in- Case

fi frequency variants (MAFo2%) in five genes, namely ENOSF1, NFS1, SLC17A4, RCC2 and GDA, which may potentially be associated with thiopurine metabolism, transport or therapeutic action, discussed as follows. were identi score GERP 4 2/24 (8.3%) 161/8578 (1.9%) 0.1172 5.6 2/24 (8.3%) 53/8600 (0.6%) 0.0005 5.62 2/24 (8.3%) 16/8600 (0.2%) 0 The ENOSF1 gene may influence thiopurine metabolism by TPMT via its regulation of thymidylate synthase’s activity.33–35,49 NFS1 Nfs1 acts as a sulphur donor in the synthesis of molybdenum

and 50

a cofactor, which is essential for the hydroxylating activity of XDH and aldehyde oxidase.51 A complete deficiency of molybdenum cofactor abrogates the functions of sulphite oxidase, XDH and ENOSF1 ling; MAF, minor allele frequency; NA, not available; UTR, untranslated , 52 fi aldehyde oxidase. The third variant was identified in SLC17A4, damaging PolyPhen-2 prediction damaging damaging which encodes a urate transporter,53 and urate is the end product of human purine metabolism. The effect of the fourth variant was SLC17A4 , uncertain, as it was discovered in the 3′ untranslated region of Protein functions were curated from the Universal Protein Resource Knowledgebase score 24.5 NA 3.12 2/24 (8.3%) 4/758 (0.5%) 0.0018 15.83 Probably 35 NA 0.718 3/24 (12.5%) 101/8600 (1.2%) 19.64 Probably b

RCC2 RCC2. The gene encodes a coordinator protein which limits the CADD C activity of Rac1,54 a therapeutic target of thiopurine compounds.32 The candidate gene GDA encodes a deaminase protein that metabolises 6-thioguanine to 6-thioxanthine.55 We also noted an abundance of relatively common variants in genes related to Rac activity (Supplementary Table S2), but these variants occurred at comparable frequencies among the in-house control individuals. Likewise, the variants found in the candidate genes do not seem to account for the thiopurine hypermethylator phenotype (Table 4). Overall, because the variants identiﬁed from exome- co-transport This was limited to the European population; where SNP information was not available from EVS, comparison was wide and candidate-gene analyses were each seen in only a few d + subjects, and the mechanistic links with thiopurine hypermethyla- b

30.0. tion are relatively weak, the contribution of these genes to

⩾ hypermethylation remains rather speculative. ed via multiple whole-exome variant-reduction approaches

ﬁ If a single major pathway modulating thiopurine methylation Active transport of phosphatecells into via Na May be a guanine-nucleotide exchange factor for Rac1 Hydrolytic deamination of guanine, producing xanthine and ammonia Regulation of thymidylate synthase activity Biosynthesis of molybdenum cofactor 33 Probably was affected by mutation in our participants, a substantially higher rate of variant discovery in relevant genes should have been observed, given an estimated population phenotype frequency of candidate genes. It is noteworthy that the variants in approximately 2.5%.8 Our results failed to demonstrate this but SLC17A4 RCC2 Gene Function GDA ENOSF1 NFS1 ruled out the involvement of large CNVs in this phenotype. The a priori alternative model, that the drug-response phenotype is complex and mediated by multiple genes, each having a small contribution, cannot be tested with the experimental approach applied

UTR here and would require genome-wide association study of a much ′ Amino-acid change larger cohort. The main limitation of our study is that of sample size coupled with the use of population MAFs documented in the public , which was one of the

All reported genotypes had a quality score database for statistical comparisons. Also, we have relied on c heuristic ﬁltering, which, of course, could have excluded GDA potentially causative variants. However, we argue that the utility Nucleotide change of exome sequencing lies in its power to detect highly penetrant variants, which likely have a low prevalence (MAF) in the general population. Moreover, given the low frequency (2.5%) of the extreme hypermethylation phenotype, deﬁned by a 6-TGN/6-MMP ratio of 4100, the use of a population control is appropriate for the detection of large-effect variants. A second limitation is that measurement of thiopurine metabolites in the erythrocytes may Variants of potential thiopurine pharmacogenetic interest identi Exome-wide hits included

a confound phenotype interpretation, because the metabolite levels found in the erythrocytes may not be truly reﬂective of those achieved in the peripheral leucocytes. Another limitation is 6:25,778,182 C/T Gln/Ter Genomic position (hg19; chromosome: nucleotide) 1:17,735,340 T/G 3 9:74,840,668 G/A Val/Met 18:694,303 G/A Ala/Val 20:34,278,459 T/C Lys/Arg (UniProtKB, http://www.uniprot.org/). based on data from the 1000 Genomes Project. house control cohort, suggesting that these variants may have an equally low frequency in the New Zealand population. Table 3. Abbreviations: CADD, Combined Annotation-Dependent Depletion; EVS, Exome Variant Server; GERP, Genomic Evolutionary Rate Pro region. that WES is certainly not a perfect technology. Downstream

Table 4. Potentially deleterious variants identiﬁed in the candidate genesa

b Genomic position (hg19; Nucleotide Amino-acid Gene Function CADD C PolyPhen-2 GERP Allele count P-value chromosome: nucleotide) change change score prediction score Casec MAF (EVS)d

2:31,600,072 G/C Ser/Cys XDH Oxidation of hypoxanthine to xanthine and 21.4 Possibly 4.82 1/24 (4.2%) 19/8600 (0.2%) 0.0590 xanthine to uric acid damaging 5:56,168,70 A/G Thr/Ala MAP3K1 A component of a protein kinase signal 15.5 Possibly 5.97 1/24 (4.2%) 0/8198 (0%) 0 transduction cascade damaging 6:82,461,726 C/CTCCGCCG Frameshift FAM46A Poly(A) RNA binding NA NA 4.08 1/24 (4.2%) NA NA 7:20,685,484 C/A Gln/Lys ABCB5 Cellular efflux of doxorubicin 20.6 Probably 4.79 1/24 (4.2%) 46/7164 (0.6%) 0.3843 damaging 10:82,040,480 C/T Val/Ile MAT1A Synthesis of S-adenosylmethionine 21.6 Probably 4.53 1/24 (4.2%) 0/8600 (0%) 0 damaging 10:97,635,721 T/G 3′UTR, intron ENTPD1 Hydrolyses ATP and other nucleotides to 18.23 NA 0.137 3/24 (12.5%) 51/758 (6.7%) 0.4907 regulate purinergic neurotransmission 13:95,830,299 C/T Arg/Gln ABCC4 An organic anion pump relevant to cellular 35 Probably 5.8 1/24 (4.2%) 7/8600 (0.1%) 0.0013 detoxification damaging overproducers Chua 6-methylmercaptopurine EW of analysis sequencing Exome 14:21,960,826 G/A Val/Met TOX4 Control of chromatin structure and cell cycle 18.43 Probably 4.89 1/24 (4.2%) 2/8600 (0%) o0.0001 progression during the transition from mitosis damaging into interphase tal et 16:90,020,700 C/T His/Tyr DEF8 Intracellular signal transduction 27.8 Probably 5.11 1/24 (4.2%) 0/8600 (0%) 0 damaging 18:33,793,438 A/C His/Pro MOCOS Sulphuration of the molybdenum cofactor, 15.07 Possibly 5.56 1/24 (4.2%) 0/8600 (0%) 0 which is essential for the activity of xanthine damaging dehydrogenase and aldehyde oxidase 21:34,907,023 C/A Cys/Phe GART Biosynthesis of inosine monophosphate 27.9 Probably 5.93 1/24 (4.2%) 0/8600 (0%) 0 damaging 21:34,907,024 A/T Cys/Ser 33 Probably 5.93 1/24 (4.2%) 0/8600 (0%) 0 damaging X:118,717,547 C/T 3′UTR UBE2A Regulation of transcription 16.12 NA 5.52 1/24 (4.2%) NA NA Abbreviations: CADD, Combined Annotation-Dependent Depletion; EVS, Exome Variant Server; GERP, Genomic Evolutionary Rate Profiling; MAF, minor allele frequency; NA, not available; UTR, untranslated region. aA similar approach to exome-wide variant filtering was employed, except that singleton variants were not discarded. bProtein functions were curated from the Universal Protein Resource Knowledgebase (UniProtKB, http://www.uniprot.org/). cAll reported genotypes had a quality score ⩾ 30.0. dThis was limited to the European population; where variant information was not available from EVS, comparison was based on data from the 1000 Genomes Project. – 421 419 Exome sequencing analysis of 6-methylmercaptopurine overproducers EW Chua et al 420 bioinformatics processing is susceptible to errors such as incorrect 14 Tiwari AK, Need AC, Lohoff FW, Zai CC, Chowdhury NI, Muller DJ et al. Exome alignment, owing to inter-gene sequence similarity.56 Causative sequence analysis of Finnish patients with clozapine-induced agranulocytosis. mutations may lie within intronic or other regulatory regions that Mol Psychiatry 2014; 19:403–405. are not captured by WES or are silent exonic variants, the 15 Wagle N, Grabiner BC, Van Allen EM, Hodis E, Jacobus S, Supko JG et al. fi Activating mTOR mutations in a patient with an extraordinary response pathogenicity of which is dif cult to assess. Other pharmaco- on a phase I trial of everolimus and pazopanib. Cancer Discov 2014; 4: 13,14 genomics studies seem to support this. For instance, a 546–553. case–control study in Finnish patients with clozapine-induced 16 Sies C, Florkowski C, George P, Gearry R, Barclay M, Harraway J et al. Measurement agranulocytosis identified only SNVs with nominal statistical of thiopurine methyltransferase activity guides dose-initiation and prevents significance, none of which were non-synonymous variants.14 toxicity from azathioprine. NZMedJ2005; 118: U1324. Although it did not detect mutations of major effect underlying 17 McLeod HL, Lin JS, Scott EP, Pui CH, Evans WE. Thiopurine methyltransferase activity in American white subjects and black subjects. Clin Pharm Ther 1994; 55: the hypermethylator phenotype, this study provides clues for 15–20. future work. A more exhaustive search for mutations could be 18 Li H, Durbin R. Fast and accurate short read alignment with Burrows-Wheeler accomplished using whole-genome sequencing, to identify transform. Bioinformatics 2009; 25: 1754–1760. intronic or regulatory variants not detectable by WES. Alterna- 19 Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N et al. The Sequence tively, larger cohorts of patients with this phenotype could be Alignment/Map format and SAMtools. Bioinformatics 2009; 25: 2078–2079. used in a genome-wide association study to identify relevant loci 20 McKenna A, Hanna M, Banks E, Sivachenko A, Cibulskis K, Kernytsky A et al. The Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation of small effect size. In conclusion, our report illustrates the DNA sequencing data. Genome Res 2010; 20: 1297–1303. application of exome analysis and aCGH in the pharmacogenetic 21 DePristo MA, Banks E, Poplin R, Garimella KV, Maguire JR, Hartl C et al. A frame- setting and adds to the current body of evidence demonstrating work for variation discovery and genotyping using next-generation DNA the complexity of thiopurine response. sequencing data. Nat Genet 2011; 43: 491–498. 22 Blankenberg D, Kuster GV, Coraor N, Ananda G, Lazarus R, Mangan M et al. Galaxy: a web-based genome analysis tool for experimentalists. In: Ausubel FM, CONFLICT OF INTEREST Brent R, Kingston RE, Moore DD, Seidman JG, Smith JA et al. (eds). Current Protocols in Molecular Biology. John Wiley & Sons, Inc.: New Jersey, USA, 2010 pp fl The authors declare no con ict of interst 19.10.1–19.10.21 . 23 Giardine B, Riemer C, Hardison RC, Burhans R, Elnitski L, Shah P et al. Galaxy: a platform for interactive large-scale genome analysis. Genome Res 2005; 15: ACKNOWLEDGMENTS 1451–1455. This work was supported by Jim and Mary Carney Charitable Trust and University of 24 Goecks J, Nekrutenko A, Taylor J. Galaxy: a comprehensive approach for sup- Otago Research Grant. porting accessible, reproducible, and transparent computational research in the life sciences. Genome Biol 2010; 11: 2010–2011. 25 Abecasis GR, Altshuler D, Auton A, Brooks LD, Durbin RM, Gibbs RA et al. Amapof REFERENCES human genome variation from population-scale sequencing. Nature 2010; 467: 1061–1073. 1 Lennard L. The clinical pharmacology of 6-mercaptopurine. Eur J Clin Pharmacol 26 Abecasis GR, Auton A, Brooks LD, DePristo MA, Durbin RM, Handsaker RE et al. An 43 – 1992; : 329 339. integrated map of genetic variation from 1,092 human genomes. Nature 2012; 2 Rosenberg JL, Wall AJ, Levin B, Binder HJ, Kirsner JB. A controlled trial of aza- 491:56–65. thioprine in the management of chronic ulcerative colitis. Gastroenterology 1975; 27 Ng SB, Turner EH, Robertson PD, Flygare SD, Bigham AW, Lee C et al. Targeted 69 – :96 99. capture and massively parallel sequencing of 12 human exomes. Nature 2009; 3 Lennard-Jones JE. Azathioprine and 6-mercaptopurine have a role in the treat- 461:272–276. 26 – ment of Crohn's disease. Dig Dis Sci 1981; : 364 368. 28 Kircher M, Witten DM, Jain P, O'Roak BJ, Cooper GM, Shendure J. A general 4 O'Donoghue DP, Dawson AM, Powell-Tuck J, Bown RL, Lennard-Jones JE. Double- framework for estimating the relative pathogenicity of human genetic variants. blind withdrawal trial of azathioprine as maintenance treatment for Crohn's dis- Nat Genet 2014; 46: 310–315. 2 – ease. Lancet 1978; : 955 957. 29 Cooper GM, Stone EA, Asimenos G, Green ED, Batzoglou S, Sidow A. Distribution fi 5 Chouchana L, Narjoz C, Beaune P, Loriot MA, Roblin X. Review article: the bene ts and intensity of constraint in mammalian genomic sequence. Genome Res 2005; fl of pharmacogenetics for improving thiopurine therapy in in ammatory bowel 15:901–913. disease. Aliment Pharmacol Ther 2012; 35:15–36. 30 Cooper GM, Goode DL, Ng SB, Sidow A, Bamshad MJ, Shendure J et al. Single- 6 Gearry RB, Barclay ML. Azathioprine and 6-mercaptopurine pharmacogenetics nucleotide evolutionary constraint scores highlight disease-causing mutations. and metabolite monitoring in inflammatory bowel disease. J Gastroenterol Nat Methods 2010; 7:250–251. Hepatol 2005; 20: 1149–1157. 31 Adzhubei IA, Schmidt S, Peshkin L, Ramensky VE, Gerasimova A, Bork P et al. A 7 Smith M, Blaker P, Patel C, Marinaki A, Arenas M, Escuredo E et al. The impact of method and server for predicting damaging missense mutations. Nat Methods introducing thioguanine nucleotide monitoring into an inflammatory bowel 2010; 7: 248–249. disease clinic. Int J Clin Pract 2013; 67: 161–169. 32 Zaza G, Cheok M, Krynetskaia N, Thorn C, Stocco G, Hebert JM et al. Thiopurine 8 van Egmond R, Chin P, Zhang M, Sies CW, Barclay ML. High TPMT enzyme activity pathway. Pharmacogenet Genomics 2010; 20: 573–574. does not explain drug resistance due to preferential 6-methylmercaptopurine 33 Rosmarin D, Palles C, Pagnamenta A, Kaur K, Pita G, Martin M et al. A candidate production in patients on thiopurine treatment. Aliment Pharmacol Ther 2012; 35: gene study of capecitabine-related toxicity in colorectal cancer identifies new 1181–1189. toxicity variants at DPYD and a putative role for ENOSF1 rather than TYMS. Gut 9 Chouchana L, Roche D, Jian R, Beaune P, Loriot MA. Poor response to thiopurine 2014; 64: 111–120. in inflammatory bowel disease: how to overcome therapeutic resistance? Clin 34 Karas-Kuzelicki N, Milek M, Mlinaric-Rascan I. MTHFR and TYMS genotypes influ- Chem 2013; 59: 1023–1026. ence TPMT activity and its differential modulation in males and females. Clin 10 Roberts RL, Gearry RB, Barclay ML, Kennedy MA. IMPDH1 promoter mutations in a Biochem 2010; 43:37–42. patient exhibiting azathioprine resistance. Pharmacogenomics J 2006; 7:312–317. 35 Milek M, Smid A, Tamm R, Kuzelicki NK, Metspalu A, Mlinaric-Rascan I. Post- 11 Haglund S, Almer S, Peterson C, Soderman J. Gene expression and thiopurine translational stabilization of thiopurine S-methyltransferase by S-adenosyl-L- metabolite profiling in inflammatory bowel disease - novel clues to drug targets methionine reveals regulation of TPMT*1 and *3C allozymes. Biochem Pharmacol and disease mechanisms? PloS One 2013; 8: e56989. 2012; 83: 969–976. 12 Price MJ, Carson AR, Murray SS, Phillips T, Janel L, Tisch R et al. First pharmaco- 36 Smith MA, Marinaki AM, Arenas M, Shobowale-Bakre M, Lewis CM, Ansari A genomic analysis using whole exome sequencing to identify novel genetic et al. Novel pharmacogenetic markers for treatment outcome in azathioprine- determinants of clopidogrel response variability: results of the Genotype Infor- treated inflammatory bowel disease. Aliment Pharmacol Ther 2009; 30: mation and Functional Testing (GIFT) exome study. J Am Coll Cardiol 2012; 59: E9. 375–384. 13 Tammiste A, Jiang T, Fischer K, Mägi R, Krjutškov K, Pettai K et al. Whole-exome 37 Stocco G, Yang W, Crews KR, Thierfelder WE, Decorti G, Londero M et al. PACSIN2 sequencing identifies a polymorphism in the BMP5 gene associated with SSRI polymorphism influences TPMT activity and mercaptopurine-related gastro- treatment response in major depression. J Psychopharmacol 2013; 27:915–920. intestinal toxicity. Hum Mol Genet 2012; 21: 4793–4804.

The Pharmacogenomics Journal (2015), 414 – 421 © 2015 Macmillan Publishers Limited Exome sequencing analysis of 6-methylmercaptopurine overproducers EW Chua et al 421 38 Blaker PA, Peters van Ton AM, Arenas Hernandez M, Smith MA, Smith CH, Irving P 48 Blaker PA, Arenas-Hernandez M, Smith MA, Shobowale-Bakre EA, Fairbanks L, et al. PWE-234: optimising the response to thiopurine therapy: a search for novel Irving PM et al. Mechanism of allopurinol induced TPMT inhibition. Biochem explanations for thiopurine hypermethylation. Gut 2012; 61: A393. Pharmacol 2013; 86: 539–547. 39 Duley JA, Somogyi AA, Martin JH. The future of thiopurine pharmacogenomics. 49 Karas-Kuzelicki N, Jazbec J, Milek M, Mlinaric-Rascan I. Heterozygosity at the TPMT Pharmacogenomics 2012; 13: 1549–1552. gene locus, augmented by mutated MTHFR gene, predisposes to 6-MP related 40 Hindorff LA, MacArthur J, Morales J, Junkins HA, Hall PN, Klemm AK et al. toxicities in childhood ALL patients. Leukemia 2009; 23: 971–974. A Catalog of Published Genome-Wide Association Studies. 2014 (accessed 18 50 Marelja Z, Stöcklein W, Nimtz M, Leimkühler S. A novel role for human Nfs1 in August 2014). Available from www.genome.gov/gwastudies. the cytoplasm: Nfs1 acts as a sulfur donor for MOCS3, a protein involved in 41 Czaja AJ. Autoimmune hepatitis – approach to diagnosis. MedGenMed 2006; 82:55. molybdenum cofactor biosynthesis. J Biol Chem 2008; 283: 25178–25185. 42 MacDonald JR, Ziman R, Yuen RKC, Feuk L, Scherer SW. The database of genomic 51 Hille R. The mononuclear molybdenum enzymes. Chem Rev 1996; 96: 2757–2816. variants: a curated collection of structural variation in the human genome. Nucleic 52 Johnson JL, Waud WR, Rajagopalan KV, Duran M, Beemer FA, Wadman SK. Inborn Acids Res 2014; 42: D986–D992. errors of molybdenum metabolism: combined deficiencies of sulfite oxidase and 43 Preacher KJ. Calculation for the chi-square test: an interactive calculation tool for xanthine dehydrogenase in a patient lacking the molybdenum cofactor. Proc Natl chi-square tests of goodness of fit and independence (computer software), 2001. Acad Sci USA 1980; 77: 3715–3719. Available from http://quantpsy.org. 53 Togawa N, Miyaji T, Izawa S, Omote H, Moriyama Y. A Na+-phosphate cotran- 44 The UniProt Consortium. Activities at the Universal Protein Resource (UniProt). sporter homologue (SLC17A4 protein) is an intestinal organic anion exporter. Am Nucleic Acids Res 2014; 42: D191–D198. J Physiol Cell Physiol 2012; 302: C1652–C1660. 45 He Y, Hoskins JM, McLeod HL. Copy number variants in pharmacogenetic genes. 54 Humphries JD, Byron A, Bass MD, Craig SE, Pinney JW, Knight D et al. Proteomic Trends Mol Med 2011; 17:244–251. analysis of integrin-associated complexes identifies RCC2 as a dual regulator of 46 Frayne J, Hall L. The gene for the human tMDC I sperm surface protein is non- Rac1 and Arf6. Sci Signal 2009; 2: ra51. functional: implications for its proposed role in mammalian sperm-egg recogni- 55 Bronk JR, Lister N, Shaw MI. Transport and metabolism of 6-thioguanine and tion. Biochem J 1998; 334:171–176. 6-mercaptopurine in mouse small intestine. Clin Sci (Lond) 1988; 74:629–638. 47 Matimba A, Li F, Livshits A, Cartwright CS, Scully S, Fridley BL et al. Thiopurine 56 Drögemöller BI, Wright GE, Niehaus DJ, Emsley R, Warnich L. Next-generation pharmacogenomics: association of SNPs with clinical response and functional sequencing of pharmacogenes: a critical analysis focusing on schizophrenia validation of candidate genes. Pharmacogenomics 2014; 15:433–447. treatment. Pharmacogenet Genomics 2013; 23:666–674.

Supplementary Information accompanies the paper on the The Pharmacogenomics Journal website (http://www.nature.com/tpj)