Reply (to Dr. Riksen and colleagues)

To the Editor:

We thank Dr. Riksen and colleagues for their comments on our pharmacogenetics study. Although the precise mechanism of action of MTX in rheumatoid arthritis (RA) is unclear, we agree with Dr. Riksen that the effects are probably contributed by endogenous adenosine release and the interactions of MTX with enzymes involved in the folate pathway.

Dr. Riksen is right that the mutant 34T allele of AMPD1 leads to reduced activity of AMPD. In skeletal and heart biopsy specimens, heterozygous (CT) individuals have ≥50% reduced enzyme activity, whereas homozygous (TT) individuals have ≤1% residual activity.

Riksen et al also speculate that the mutant variant of AMPD reduces the interaction with AICAR. As a consequence of this assumption, it would be less likely that carriage of the 34T allele increases adenosine release. However, there are no data to support their assumption that the affinity of AMPD for AICAR is affected by the mutant 34T allele variant. In addition, studies involving other folate pathway enzymes, such as dihydrofolate reductase and folylpolyglutamate synthase, showed that the activity of the mutant enzyme was reduced, whereas the affinity of the mutant enzymes for their substrates, including antifolates, remained similar or was affected only for other substrates such as L-glutamate. Therefore, we think that the reduced activity of AMPD due to the mutant AMPD1 34T allele in the enzyme and the inhibition of AMPD via AICAR may act complementarily in enhancing adenosine release.

Biochemical studies are urgently needed to investigate the effect of the AMPD1 34C >T variant in health and disease. The effect of the AMPD1 34C>T variant itself on plasma adenosine is hardly documented. The scarce evidence indicates that the T allele probably increases adenosine: during exercise (not during rest), venous adenosine levels increase in carriers of 34TT. Furthermore, the AMPD1 34C >T variant improves survival of patients with established congestive heart failure or patients with established coronary artery disease (9, 10), probably by promoting release of adenosine, a well-known cardioprotective substance (11). How AMPD1 34C >T interacts with MTX in RA (factors that influence purine enzyme kinetics) is purely speculative.

Second, Dr. Riksen and colleagues raise concerns about confounding in our results due to the efficacy cutoff points used in our data analysis. Our approach increases the power of the study, because it categorizes all patients with a Disease Activity Score (DAS) into 2 groups instead of 3 groups. If we use the precise EULAR response criteria, patients are categorized into 66 responders, 41 nonresponders, and 79 moderate responders. Comparing the precise EULAR definition of nonresponse and good response with our definition of response and nonresponse with a DAS of ≤2.4 shows that only 4 patients were misclassified. The genotype analysis with the precise EULAR definition resulting in 66 responders and 41 nonresponders shows similar results. The association for the ATIC CC genotype remains significant, while the AMPD1 and ITPA genotypes are nonsignificant due to the reduced numbers of responders and nonresponders.