Sequential evaluation of thiopurine methyltransferase, inosine triphosphate pyrophosphatase, and HPRT1 genes polymorphisms to explain thiopurines' toxicity and efficacy



This article is corrected by:

  1. Errata: Erratum Volume 26, Issue 6, 968, Article first published online: 29 August 2007

Dr V. Annese, U.O.C. Endoscopia Digestiva, Ospedale ‘Casa Sollievo della Sofferenza’–I.R.C.C.S, Viale Cappuccini, 1 71013 San Giovanni Rotondo, Italy.




To evaluate the polymorphisms of several genes involved in the azathioprine and mercaptopurine metabolism, in an attempt to explain their toxicity and efficacy in Crohn’s disease and ulcerative colitis.


In 422 consecutive patients (250 with Crohn’s disease and 172 with ulcerative colitis) and 245 healthy controls, single nucleotide polymorphisms of thiopurine methyltransferase, inosine triphosphate pyrophosphatase and hypoxanthine phosphoribosyl transferase (HPRT1) genes were related to the occurrence of adverse drug reactions (ADRs) and efficacy of therapy.


Seventy-three patients reported 81 episodes of ADRs; 45 patients did not respond to therapy. Frequency of thiopurine methyltransferase risk haplotypes was significantly increased in patients with leucopenia (26% vs. 5.7% in patients without ADRs, and 4% of controls) (P < 0.001); no correlation with other ADRs and efficacy of therapy was found. Conversely, the frequency of inosine triphosphate pyrophosphatase and HPRT1 risk genotypes was not significantly different in patients with ADRs (included leucopenia). Non-responders had an increased frequency of inosine triphosphate pyrophosphatase risk genotypes (P = 0.03).


The majority of azathioprine/mercaptopurine-induced ADRs and efficacy of therapy are not explained by the investigated gene polymorphisms. The combined evaluation of all three genes enhanced the correlation with leucopenia (43.5% vs. 23% in controls) (P = 0.008), at the expense of a reduced accuracy (60%).


Azathioprine (AZA) and its metabolite Mercaptopurine (MP) are immunosuppressive drugs widely used in the treatment of several conditions, such as autoimmune hepatitis, rheumatic or dermatologic diseases, organ transplant recipients and inflammatory bowel disease (IBD). More specifically, both drugs are effective in inducing and maintaining remission in patients with ulcerative colitis (UC) and Crohn’s disease (CD).1, 2 Unfortunately, the occurrence of adverse drug reactions (ADRs)3, 4 or a therapeutic inefficacy leads to withdrawal of therapy in up to one third of patients.2

The metabolism of these drugs is complex and requires intracellular activation, catalysed by several enzymes (Supplementary Fig. 1). AZA is converted via a non-enzymatic reaction to MP, which is subsequently metabolized either through thiopurine methyltransferase (TPMT) to methylmercaptopurine or, alternatively, by xanthine oxidase to thiouric acid. Furthermore, MP may be converted by hypoxanthine phosphoribosyl transferase (HPRT) to thioinosine monophosphate. Subsequently, the enzyme inosine triphosphate pyrophosphatase (ITPA) catalyses the futile cycle to thioinosine triphosphate and viceversa, to avoid the accumulation of thio-inosine monophosphate, and from guanosine monophosphate synthetase to thio-guanine nucleotides (TGNs). The TGNs are the active metabolites conferring clinical efficacy, but they also carry the toxicity.5

Over the last two decades, genetic studies have established that a polymorphism at the TPMT gene locus plays a significant role in the enzymatic activity and bone marrow toxicity (BMT) (reviewed6). In Caucasians, approximately one of 300 individuals is TPMT enzyme deficient (homozygous or compound heterozygous for mutant alleles), 11% have an intermediate activity (heterozygous for mutant allele), and the remaining 89% have high activity (homozygous for wild-type alleles). Consistent with the frequency of these risk alleles, only a minority of BMT due to thiopurines are explained by TPMT deficiency.6 Moreover, no clear correlation of TPMT deficiency with others ADRs has been demonstrated. Conversely, some wild-type patients with very high TPMT activity, so-called fast-methylators, develop suboptimal 6-TGNs concentrations, which have been associated with treatment failure.7

More recently, the ITPase deficiency has also been associated with thiopurine toxicity. The ITPA gene is located on chromosome 20p; a decreased ITPase activity leads to accumulation of the potential toxic metabolite inosine triphosphate. Five SNPs of this gene have been identified so far6 and, more specifically, the C94A polymorphism (Pro32Thr) has been associated with thiopurines toxicity;8–12 conflicting data, however, have been reported.13, 14

Another potentially important gene is the HPRT1, located in the region Xq26–27.2. A number of mutations at different sites in the HPRT1 gene have been identified,15 and some of them by gene inactivation result in cellular resistance to purine analogues. This event occurs in the Lesch-Nyhan syndrome,16 that is characterized by abnormal metabolic and neurologic manifestations. No evaluation of HPRT1 gene polymorphisms in patients under thiopurines therapy has been reported yet.

As evaluation of TPMT polymorphism may explain only a minority of ADRs in patients using thiopurines, we investigated the likely implications of a number of polymorphisms of ITPA and HPRT1 genes, beside the TPMT polymorphisms. We hypothesized that the sequential evaluation of different polymorphisms could enhance the prediction of ADRs and, perhaps, explain the different efficacy of thiopurines therapy in IBD patients.

Materials and methods

Study population

Inflammatory bowel disease patients under AZA or MP therapy were consecutively recruited in five Italian referral centres: the IRCCS ‘Casa Sollievo della Sofferenza’ Hospital of San Giovanni Rotondo, the University Hospital Policlinico of Rome, the University Hospital of Padua, the University Hospital of Florence and the IRCCS of San Donato Milanese. Diagnosis of CD and UC was established according to accepted clinical, endoscopic, radiological and histological criteria.17 Patients were evaluated and classified based on the careful review of medical records as responder, non-responder or intolerant to thiopurines therapy.18 More specifically, evaluation of response to therapy was considered after at least six-months at the adequate dosage (2–2.5 mg/kg and 1–1.25 mg/kg for AZA and MP, respectively).19 Patients were considered responders when thiopurines were effective in induction or in maintaining the remission, with steroid sparing effect. Patients were considered intolerant when the occurrence of ADRs led to therapy withdrawal or dose reduction. Therefore, patients without ADRs but with less of 6 months of therapy were not included.

Outcome measures were the TPMT, ITPA and HPRT1 genotypes and their correlation with effectiveness of therapy and occurrence of thiopurines-related side effects, based on temporal relationship and exclusion of alternative explanations. BMT was defined according to the criteria used by Connel et al.20 as either leucopenia (white blood count <3.0 × 109/L) and/or thrombocytopenia (platelet count <100.00 × 106/L) resolving after treatment discontinuation or dose reduction. Hepatotoxicity was defined by serum alanine transaminase increase greater than twice the upper normal limit and resolution after withdrawal or dose reduction. Pancreatitis was defined as upper abdominal pain with pancreatic amylase and lipase greater than twice the normal upper limit. Flu-like symptoms were defined as the occurrence of arthralgia and/or myalgia and/or fever, that resolved after drug withdrawal.

Four hundred twenty-two IBD patients were included in the study; 250 were CD patients (134 male, mean age 35 ± 14 years), and 172 were UC patients (93 male, mean age 43.5 ± 15 years). Their clinical characteristics are summarized in Table 1.

Table 1.   Clinical and demographic characteristics of patients with Crohn’s disease (CD) and Ulcerative Colitis (UC). Values are expressed as mean ± s.d., or absolute values and percentages
 Crohn’s disease (n = 250)Ulcerative Colitis (n = 172)
  1. EIM, Extraintestinal manifestations; AZA, Azathioprine; MP, Mercaptopurine; ASA, Mesalazine; G-I, Gastrointestinal.

  2. *One patient upper G-I only.

Age (years)35 ± 1443.5 ± 15
Age at diagnosis (years)28 ± 1429 ± 14
Duration of follow-up (years)6 ± 46 ± 5
Sex (male/female)134/11693/79
Localization CD: n, (%)
 ileum91 (33.8) 
 ileo-colon83 (30.9) 
 colon75 (27.9) 
 upper G-I tract20 (7.4)* 
Localization UC: n, (%)
 rectum-sigmoid 33 (19.2)
 left colon 67 (38.9)
 pancolitis 72 (41.9)
Disease type CD: n, (%)
 inflammatory75 (30) 
 stricturing62 (24.8) 
 fistulizing113 (45.2) 
Smoking: n, (%)
 Active90 (36.4)17 (10.1)
 Ex smokers28 (11.3)46 (27.2)
 Never129 (52.2)106 (62.7)
Resective surgery: yes/no, (%)107/143 (42.8)24/151 (13.7)
EIM: yes/no, (%)140/110 (56)67/103 (39.4)
Arthropathy: yes/no, (%)99/150 (40)50/124 (29)
Perianal disease: yes/no, (%)43/142 (23)6/136 (3,5)
Family history: yes/no, (%)14/142 (9)10/119 (8)
Use of 5-ASA: yes/no, (%)202/48 (81)170/2 (99)
Duration of AZA/MP therapy (month)22 ± 1518 ± 13

Two hundred forty-five gender-matched healthy subjects were also evaluated. They were unrelated, asymptomatic subjects (blood donors, students and staff members) recruited in the same centres, all of Caucasian and non-Jewish descent. The Ethics Committees of each participating centres approved the study, and an informed consent was obtained from each subjects.


Genomic DNA samples were extracted from peripheral blood leukocytes, according to standard protocols21 and genotyped at the Molecular Laboratory of the Unit of Gastroenterology of the San Giovanni Rotondo Hospital, Italy.

The sequences of the primers used for TPMT and ITPA genotyping, designed by the Primer3 Input program (, were as follows:


Polymerase chain reaction (PCR) reactions (25 μL) were performed in 1X GeneAmp Buffer II, 2.5 mmol/L MgCl2, 3U AmpliTaq Gold DNA polymerase (Applied Biosystems, Foster City, CA, USA), 0.2 mmol/L for each primer, 0.5 mmol/L dNTPs and 30 ng of genomic DNA. After initial denaturation at 95 °C for 10 min, reaction was subjected to amplification for 35 cycles: 1 min at 95 °C, 1 min at 58 °C for TPMT_460, 62 °C for TPMT_719, and 62 °C for ITPA_Pro32Thr, 1 min at 72 °C. Final extension was 72 °C for 10 min. The G460A and A719G polymorphisms of TPMT gene were detected by using DNA sequencing analysis on ABI 310 DNA sequencer (Applied Biosystems), according to the manufacturers’ recommendations.

The C94A variation of the ITPA gene was analysed using PCR-restriction fragment length polymorphism assay. PCR product was digested for 2 h with restriction enzyme (XmnI), and analysed on 3% (w/v) agarose gel. An ITPA wild-type, heterozygote and homozygote mutant, validated by sequencing, was included as a control specimen in each set of genotyped samples.

The C__11680155_10, C__11680164_10, C__27862676_10 and C___8940525_10 (rs1468266) variations in the HPRT1 gene were analysed using the Taqman 7700 (Applied Biosystems), according to manufacturers’ recommendations. Briefly, PCR reaction (15 μL) was done in 1X TaqMan Universal PCR Master Mix, 1X Genotyping Assay Mix, and 50 ng of genomic DNA. After 50 °C for 2 min and initial denaturation at 94 °C for 10 min, reaction was subjected to amplification for 40 cycles: at 94 °C for 15 s, and 60 °C for 1 min.

Statistical analysis

Comparison of allele-genotype frequencies and genotype/phenotype association was performed by the chi-square or Fisher exact tests, when appropriate; Student’s t-test was used to compare means of continuous variables by the SPSS software ver 11.5 (Chicago, IL, USA). Test of Hardy–Weinberg equilibrium and marker linkage disequilibrium analysis was performed by the Arlequin software ver 2.0 (

The sensitivity, specificity, positive and negative predictive values for each risk genotypes and their combination were calculated after sorting patients in those with or without occurrence of leucopenia.


Genotypic frequencies for TPMT, ITPA and HPRT1 variants are reported on the supplementary Table S1 (available online). All frequencies were in Hardy–Weinberg equilibrium.

Concerning the HPRT1 gene, only the most informative variant was evaluated (rs14682666) as the others had a major allele frequency higher than 99% (data available on request). The genotypic frequencies of the TPMT, ITPA, and HPRT1 polymorphisms were similar in the IBD, UC, and CD subgroups of patients, compared to controls. Accordingly, also the frequency of risk haplotypes of the TPMT gene (3*A, 3*B and 3*C) did not differ between patients and healthy controls.

Patients were stratified according to the occurrence of ADRs, and efficacy of thiopurines therapy (Table 2). Seventy-three patients reported 81 episodes of ADRs, classified as leucopenia (n = 23), flu-like symptoms (n = 18), pancreatitis (n = 16), hepatoxicity (n = 12), nausea/vomiting (n = 9) and skin reactions (n = 3).

Table 2.   Correlation between thiopurine methyltransferase (TPMT), inosine triphosphate pyrophosphatase (ITPA) and HPRT1 risk variants with occurrence of adverse drug reactions (ADRs) and response to thiopurines therapy. P values, Odd ratios and confidence intervals obtained by comparison with healthy controls (HC). Please note that 81 ADRs were reported in 73 patients
3*A + 3*B + 3*C
HPRT1 rs1468266
ADRs (n = 81)Odds ratio (CI)2.9 (1.04–8.2)1.20.9
Heterozygotes (n)7115
Homozygotes (n)208
Leucopenia (n = 23)Odds ratio (CI)8.3 (2.3–29)1.21.1
Heterozygotes (n)452
Homozygotes (n)202
Flu-like symptoms (n = 18)Odds ratio1.30.41.1
Heterozygotes (n)112
Homozygotes (n)001
Pancreatitis (n = 16)Odds ratio01.81.1
Heterozygotes (n)031
Homozygotes (n)002
Hepatotoxicity (= 12)Odds ratio2.11.50.9
Heterozygotes (n)120
Homozygotes (n)002
Nausea/vomiting (n = 9)Odds ratio002.5
Heterozygotes (n)000
Homozygotes (n)003
Skin reactions (n = 3)Odds ratio11.702.4
Heterozygotes (n)100
Homozygotes (n)001
No ADRS (n = 349)Odds ratio0.51.41.3
Heterozygotes (n)205340
Homozygotes (n)0129
Responders (n = 304)Odds ratio1.41.31.2
Heterozygotes (n)174334
Homozygotes (n)0123
Non-responders (n = 45)Odds ratio (CI)1.62.5 (1.1–5.8)0.8
Heterozygotes (n)3106
Homozygotes (n)016
Healthy controls (n = 245)Heterozygotes (n)102824
Homozygotes (n)0017

The 3*A, 3*B and 3*C haplotypes of TPMT polymorphisms were grouped together. The frequency of TPMT risk genotypes was significantly increased in ADRs group of patients (11%), compared with healthy controls (4.1%, OR = 2.9, CI = 1.04–8.2, P = 0.038). Moreover, TPMT risk genotypes were significantly more frequent in patients who experienced leucopenia (26.1%), compared with healthy controls (OR = 8.2, CI = 2.3–29, P = 0.0001), and patients without ADRs (5.7%, OR = 5.8, CI = 1.8–17.9, P = 0.001). Two patients, homozygotes for TPMT 3*A/3*A, experienced severe BMT (white blood count <1.0 × 109/L) after only 2 weeks of therapy; they completely recovered 1 month of after drug withdrawal. Conversely, no significant correlation with other ADRs was found. No significant difference of TPMT risk genotypes was also found, when comparing responders with non-responders to thiopurines therapy.

Regarding the C94A polymorphism of the ITPA gene, the frequency of risk genotypes in the ADRs group (13.6%) was not significantly different from that reported in the non-ADRs group (15.5%), and healthy controls (11.4%). Furthermore, no significant association of risk genotype with any specific ADRs was demonstrated, with the exception of a trend towards an association with leucopenia (21.7% vs. 11.4% in healthy controls, P = 0.2). Non-responders to thiopurines had an increased frequency of risk genotype (24.4%) compared to healthy controls (11.4%, P = 0.034, OR = 2.5, CI = 1.06–5.85) (Table 2). One patient was homozygote for the ITPA risk allele but did not experience ADRs.

No difference of the HPRT1 polymorphism was found in patients with or without ADRs (16.5% vs. 20.8%), and controls (16.7%). No significant difference for any specific ADRs and response to therapy was also found, with the exception of a trend towards an increased frequency of this polymorphism in patients with nausea/vomiting (33%, P = 0.3).

The large majority of patients investigated were under concomitant 5-ASA therapy (81% of CD and 98% of UC patients), with dosage ranging from 1.6 to 3.2 g/day. Frequency of concomitant 5-ASA therapy was equally distributed in subgroups of patients with or without ADRs, responders and non-responders to thiopurines (data not shown).

When relating carriers of risk genotypes to the occurrence of leucopenia, the sensitivity, specificity, positive and negative predictive values are reported in Table 3. The greatest accuracy was accounted by the TPMT polymorphism (90%).

Table 3.   The sensitivity, specificity, positive (PPV), negative predictive values (NPV), and accuracy for each risk genotypes and their combination calculated after sorting patients in those with or without occurrence of leucopenia.
  1. TPMT, thiopurine methyltransferase; ITPA, inosine triphosphate pyrophosphatase; HPRT1, hypoxanthine phosphoribasyl transferase.


To verify the hypothesis that combining different risk genotypes could enhance the prediction of ADRs, and more specifically that of leucopenia, data were stratified according to the presence of at least one risk genotype of the three investigated genes. When the TPMT and ITPA risk genotypes were combined, one risk genotype was found in 18 of 81 episodes of ADRs (22%), 10 of 23 episodes of leucopenia (43.5%), 8 of 58 of other ADRs episodes (14%), 73 of 349 patients without ADRs (21%) and 38 of 245 controls (15.5%). Data were significant only for the occurrence of leucopenia in comparison with patients without ADRs (OR = 2.9, CI = 1.1–7.1, P = 0.02), and controls (OR = 4.2, CI = 1.5–11.1, P = 0.002). After combining all three risk genotypes, al least one risk genotype was found in 29 of 81 episodes of ADRs (36%), 13 of 23 episodes of leucopenia (56.5%), 16 of 58 of other ADRs episodes (27.5%), but also in 139 of 349 patients without ADRs (40%), and 79 of 245 healthy controls (32%). These differences were significant only in the subgroup of patients with leucopenia when compared to controls (OR = 2.7, CI = 1.1–7.1, P = 0.03).

As a whole, two risk genotypes were demonstrated in four out of 81 episodes of ADRs (5%), and four out of 349 patients without ADRs (1%)(P = 0.06).

Therefore, when considering the most life threatening adverse event, i.e leucopenia, the sequential genotyping of TPMT, ITPA and HPRT1 polymorphism ‘paradoxically’ decreased the accuracy of prediction from 90% (only TPMT), to 77% (TPMT+ITPA), and 60% (all three) (Table 3).


Since its introduction 40 years ago, AZA and its metabolite MP, have become the most widely used immunosuppressive drugs for the treatment of IBD. Unfortunately, in up to one third of patients, therapy must be withdrawn owing to the occurrence of potential life threatening adverse events or treatment failure.

Until now, at least 21 variant alleles of the TPMT gene have been reported to lower the enzymatic activity.6, 10 Two of them, both non-synonymous coding single nucleotide polymorphisms, determining the haplotype TPMT*3A, and resulting in G460A/Ala154Thr (Exon 7) and A719 G/Tyr204Cys (Exon 9), are the most common variant in Caucasians.8, 9, 22, 23 Two other haplotypes, the TPMT*3B with the G460A variation, and the TPMT*3C haplotype with the A719G variant also determine a decreased enzymatic activity but are less frequently detected among Caucasians. The haplotypes TPMT*3A and TPMT*3B result in a virtual lack of enzymatic activities (1.6% and 1.7%, respectively), compared with the wild-type genotype,10 due to disruption and misfolding of the enzyme, protein aggregation and aggregosome formation.24, 25

Several studies have demonstrated that TPMT-deficient patients are at high risk for severe and often fatal BMT (see for review)26 (supplementary Table 2).27–31 In IBD patients, this may occur as early as during the initial weeks of instituting the therapy.3 Among patients on continuous thiopurines therapy, however, TPMT genotype does not fully predict bone marrow suppression, that has been related more often to other interfering factors (i.e. viral infections, use of other drugs interfering with thiopurines metabolism).6 In the largest study so far available on AZA-treated CD patients developing mielotoxicity, only 11 out of 41 patients (27%) had one or two mutant TPMT alleles.32 Therefore, the clinical relevance of TPMT genotyping has been questioned,3 although patients with higher risk could be identified.33

In our population, TPMT risk genotypes were found only in six out of 23 patients with BMT (26%) and 20 out of 349 patients (5.7%) without ADRs. While this difference was significant (P = 0.001), in keeping with the majority of studies,26 this figure determines a positive and negative predictive value for the occurrence of leucopenia of 23% and 95%, respectively. Moreover, the accuracy for predicting other ADRs was even lower, with a positive and negative predictive value of 13% and 85%, respectively.

Recently, another enzymatic deficiency due to the polymorphisms of the ITPA gene has been associated with thiopurines toxicity. At least three SNPs have been studied, and the C94A (Pro32Thr) polymorphism, encoding for a Pro32Trh exchange, has been significantly associated with thiopurine toxicity; however, data are conflicting.8–14 Homozygotes for the C94A missense mutation lack enzymatic activity, while heterozygotes possess 22.5% of normal enzymatic activity.8 In some studies.11, 12 but not in others,13, 14 this polymorphism has been associated with AZA toxicity. Moreover, the possible correlation between ITPAse deficiency and specific ADRs is still unclear, as the risky genotypes have been associated with leucopenia in some studies12 and with influenza-like symptoms, rash and pancreatitis in another.11

In our study, no difference in either genotype or allelic frequencies of the C94A polymorphism in IBD compared with controls was found, with the exception of a slight positive trend in patients with BMT (P = 0.2).

In the attempt to predict thiopurine toxicity better, we also investigated the polymorphism of the HPRT1 gene. Three investigated polymorphisms occurred at a very low frequency (less than 1% both in patients and controls), whereas the remaining one was very common, and apparently unrelated to the development of ADRs. Only a trend towards an increased frequency in patients with nausea and vomiting was found (P = 0.3).

In carriers of at least one risk genotype of the three genes under investigation, an increased frequency was found in the group of patients with ADRs (39%), and more importantly in those who experienced leucopenia (43.5%) compared to healthy controls (23%). However, because of the high rate of ‘false positive’ results, the accuracy of prediction was reduced compared to TPMT testing alone (Table 3).

The possible effect of enzymatic deficiencies on the efficacy of therapy has been suggested only for TPMT polymorphisms,7 while data on ITPA and HPRT1 are lacking. In this study, 13% of patients using thiopurines without ADRs were considered non-responders. No influence of TPMT and HPRT1 genotypes was found; however, non-responder patients had an increased frequency of ITPA risk genotype (24.5% vs. 11.4% in controls, P = 0.03). This finding needs further confirmation in larger series because besides polymorphisms in specific drug-metabolizing enzymes, heterogeneity in genetic background and clinical features (i.e. location, behaviour) can influence response to treatment. Moreover, a further limitation of this finding is that it faces the issue of multiple testing in a low number (45) of non-responders and the lack of thiopurine metabolite measurements to assess compliance to therapy.

In conclusion, this study has been undertaken in a large population of IBD patients treated with thiopurines with a frequency of ADRs of about 20%. We found that the majority of ADRs episodes were not causally linked to the investigated polymorphisms of TPMT, ITPA and HTRP1 genes. Although genetic testing of TPMT is suggested by the current Food and Drug Administration recommendations34 and considered cost effective by decision analysis35–38 before embarking in the therapy with AZA/MP, prospective studies are lacking. In our population, its positive predictive value is low (23%) and the negative predictive value (95%) is comparable to the a priori probability of leucopenia occurrence (about 5%). In the absence of the TPMT risk alleles, the evaluation of C94A ITPA polymorphism may explain more episodes of leucopenia, but does not improve the prediction accuracy. Both polymorphisms and the consequent enzymatic deficiencies are weakly correlated with other ADRs; however, ITPA polymorphism might influence the efficacy of the therapy.


We thank Mr Corritore Giuseppe for genotyping assistance, and Mrs De Santo Ermelinda for technical support. We are also grateful for the participation and support of the families and patients with IBD. Declaration of personal interests: None. Declaration of funding interests: The work was supported by the Italian Minister of Health (RC0602GA27).