Novel pharmacogenetic markers for treatment outcome in azathioprine-treated inflammatory bowel disease

M.A. Smith; A.M. Marinaki; M. Arenas; M. Shobowale-Bakre; C. M. Lewis; A. Ansari; J. Duley; J.D.  Sanderson


Aliment Pharmacol Ther. 2009;30(4):375-384. 

In This Article


The impact of common polymorphism in TPMT on clinical outcome with azathioprine treatment has become a classic example of the application of pharmacogenetics and is one of the first developments in this field to be widely adopted in clinical practice.[43] The only other gene which has been subject to significant enquiry in the context of AZA treatment is ITPA. It seems likely that ITPA polymorphism is responsible for a proportion of ADRs experienced on AZA, including flu-like symptoms[10,41,44–46] and perhaps some cases of neutropaenia.[46,47] A few other candidate enzymes have been subject to preliminary studies, for example glutathione s-transferase,[48] 5'nucleotidase,[49–51] inosine monophosphate dehydrogenase[52] and methylene tetrahydrofolate reductase.[53–55] However, the associations discovered have not yet found a place in clinical practice.

Our study has demonstrated that the presence of the coding region SNP AOX1 c.3404A > G predicts non-response to AZA therapy. This information can be usefully combined with the result of TPMT activity testing to provide even more information about an individual's chance of success on AZA treatment. This finding could have important applications in clinical practice, not just in the field of IBD, but potentially also for the use of thiopurines in rheumatology, dermatology, transplant medicine and oncology. Knowledge of an individual's chance of response to AZA will prompt an early review of treatment efficacy, allowing a timely switch to an alternative immunosuppressive agent. It could be argued that those with a poor chance of responding to AZA (TPMT activity > 35 pmol/h/mgHb and AOX1 3404G variant) should be offered an alternative treatment as first-line therapy, which might include reduced dose azathioprine in combination with allopurinol, a combination which has been shown to circumvent the problem of hyper-methylation in some patients.[28]

In patients with Crohn's disease, the next immunomodulator considered for treatment would usually be methotrexate. It is possible that the same polymorphism AOX1 c.3404A > G could also affect an individual's chance of response to methotrexate, as AO is known to metabolize methotrexate producing a 7-hydroxy metabolite, thought to be inactive. This relationship should be examined in the light of our data.

Our findings have implications for the current understanding of AZA's mechanism of action. Genotype variants, which have a functional impact, most commonly decrease the activity of the affected enzyme. If this is true for the AOX1 3404G variant, then the association with lack of clinical response would suggest that AO metabolites of AZA have immunosuppressive activity. The only functional study of the metabolites produced by the action of AO on AZA showed that 8-hydroxy-6MP did not slow the growth of rat sarcoma.[56] However, AO produces several other AZA metabolites on which no functional work has been carried out (Figure 1) and rat sarcoma is not a model for immunomodulation in IBD. The other possibility is that AO activity is increased in the presence of the AOX1 3404G variant. In this case, it is possible that overactive AO removes and inactivates a higher proportion of the ingested drug, resulting in decreased efficacy. In this instance however, one would expect carriers of the AOX1 sequence variant to have lower TGN levels. This was not the case in our cohort, but this could have been confounded by the high degree of inter-individual variability in TGN measurements and the relatively small number of individuals carrying the AOX1 3404G variant.

The XDH and MOCOS SNPs examined here are the first examples of protective pharmacogenetic influences reported in AZA treatment. However, these results must be interpreted with caution bearing in mind the borderline significance of the findings. Confirmation of the association should be sought in other cohorts. The two XDH SNPs which have been the subject of previous functional work (c.2107A > G and c.1936A > G) were not associated with significant changes in metabolite concentrations.[23] Neither sequence variant was associated with a significant clinical impact in our study. The fact that the SNP XDH c.837C > T does not encode a change in an amino acid raises the possibility that it is not this SNP itself which exerts the effect but rather that it is in linkage disequilibrium (LD) with another polymorphism. From the analysis of HapMap databases, we found 2 other SNPs (rs17011353 and rs17011359) within a 200 kbp region spanning from 50 kbp upstream of the 5'UTR to 69 kbp downstream of the 3'UTR, which were in LD with SNP XDH c.837C > T with an r2 > 0.5. Both of these SNPs are intronic. We cannot exclude the possibility of LD to an SNP located in a regulatory element. It is possible that one of these SNPs or a neighbouring intronic SNP has an effect on mRNA splicing, as seen with ITPA.[57]

Our findings are important for the understanding of thiopurine pharmacology. We had not expected that polymorphism in XDH or its cofactor gene MOCOS would protect against ADRs. Rather, we had anticipated that by limiting inactivation of AZA, they would improve response or possibly increase the risk of dose-dependent ADRs such as myelotoxicity. The unexpected association with reduced side effects suggests that the products of XDH metabolism of thiopurines may be toxic and attenuated XDH activity is therefore an advantage. This would be consistent with the theory that XDH activity creates damaging free radicals, which would be predicted to cause adverse events.[29] This is consistent with in vitro studies which showed oxidative damage resulting from the action of XDH on thiopurines[58] and the use of allopurinol in combination with low dose AZA to circumvent hepatotoxicity in some patients.[59]

Our findings are based on a well-powered and well-documented prospective cohort and should therefore be robust. However, for some of the SNPs we assessed, with a lower allele frequency, it may be that the numbers involved were too small to demonstrate significant effects. The cohort was mostly made up of Caucasians and the relevance of our findings to other ethnic groups is not clear. In particular, the AOX1 3404G variant has a lower reported frequency in other ethnic groups and none of the non-Caucasians in our study had this SNP present.

In conclusion, we have identified a novel pharmacogenetic marker of nonresponse to AZA: AOX1 3404G variant. The ability of this marker, particularly in combination with TPMT activity, to stratify each individual's chance of responding to AZA therapy could be used in clinical practice, allowing individualized prescription of immunomodulation to improve patient outcomes. We also report a weak protective effect of polymorphism in XDH and MOCOS against ADRs on AZA. These findings require replication in other cohorts and suggest future directions for investigation, particularly the relevance of AO metabolites for the action of thiopurines.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: