Single nucleotide polymorphisms in ABCB1 and CBR1 can predict toxicity to R-CHOP type regimens in patients with diffuse non-Hodgkin lymphoma

The prediction of high-grade toxicities in cancer patients would clearly help their overall management by intensifying the surveillance and monitoring specific parameters in patients at risk. We studied the relation between selected genetic polymorphisms and treatment-associated grade 3 or higher toxicities in 760 patients with diffuse large B-cell lymphoma receiving R-CHOP or R-ACVBP, and found significant correlation between rs2229109 in ABCB1 and vomiting (P=0.003) and diarrhea (P=0.007), and between rs20572 and rs9024 in CBR1 and anemia, thrombocytopenia and diarrhea (approx. P=0.02). These results suggest that genotyping of peripheral blood cells could help predict severe toxicity in patients receiving R-CHOP type regimens.

Both functional and non-functional polymorphisms have been shown to be correlated to the toxicity of R-CHOP-based treatment of non-Hodgkin lymphoma (NHL).21 In order to confirm such correlations, we studied 14 polymorphisms in 11 genes previously shown to be involved in the metabolism and cellular response to drugs used in this treatment (Online Supplementary Table S1). Genotypes were determined as described before3 or by TaqMan SNP Genotyping Assays (Life Technologies, Carlsbad, CA, USA) on blood samples from 760 patients out of a total of 1703 patients included in prospective randomized trials (clinicaltrials.gov identifier: 00140660, 00140595, 00144807, 00169143, 00144755, 0187424 and 00135499) organized by the LYSA, and correlated with grade 3 toxicities prospectively evaluated for each cycle using the NCI CTC V3 scale.

Patients’ characteristics are listed in Table 1. Initial analyses on the occurrence and distribution of toxicities showed that (R)-ACVBP was associated with a significantly (P<0.001) higher incidence of toxicity as compared to R-CHOP, with a higher frequency of anemia (48.0% vs. 24.0%), thrombocytopenia (50.5% vs. 19.3%), febrile neutropenia (50.8% vs. 22.2%), and mucositis (26.6% vs. 7.7%), but not for diarrhea and vomiting or for delay in treatment administration (Table 2). The call rate for each polymorphism was 99.5%–100% except for rs2740574 in CYP3A4 (96.4%), and their distribution was consistent with Hardy-Weinberg equilibrium except for GSTP1 (P=0.003).

Table 1.Clinical and biochemical characteristics of patients included in the current study.

Table 2.Comparison of grade ≥3 toxicity and delay in treatment in patients treated with R-CHOP or (R)-ACVBP.

As the toxicity profile differed between R-CHOP and R-ACVBP, we first analyzed for correlation with toxicity in the two treatment groups ((R)-ACVBP and R-CHOP) separately. Here, we observed correlations for febrile neutropenia (P=0.039), and vomiting (P=0.043) with rs2229109 in ABCB1, and for febrile neutropenia with rs1695 (P=0.030) in GSTP1 in (R)-ACVBP-treated patients. In R-CHOP-treated patients, we observed correlations for diarrhea (P=0.041), vomiting (P=0.031), and mucositis (P=0.004) with rs2229109 in ABCB1, for febrile neutropenia (P=0.031), and treatment delay (P=0.040) with rs20572 in CBR1, for febrile neutropenia with rs9024 (P=0.044) in CBR1, and for treatment delay with rs714368 (P=0.015) in SLC22A16.

In a second series of analyses, we searched for correlations between grade 3 or higher toxicities and all genotypes in the complete cohort (Table 3). SNP in ABCB1 and both SNP in CBR1 correlated with the occurrence of grade 3–4 toxicities. rs2229109 (Ser400Asn) in ABCB1 was associated with increased risk of high-grade diarrhea (P=0.007) and vomiting (P=0.003) in patients with CT genotype, whereas there were no patients with TT genotype for this SNP. The rare homozygous genotypes of the two silent SNP in CBR1 (Ala209Ala and 3′-UTR) were found to be associated with higher incidence of grade 3–4 anemia (P=0.018 and 0.026, respectively), thrombocytopenia (P=0.015 and 0.017), and diarrhea (P=0.019).

Table 3.Statistically significant correlations between grade ≥3 toxicities and genotypes in all patients comparing common homozygotes (HH) with heterozygotes (Hh) and rare homozygotes (hh). Correlations were performed with the Cochran Mantel Haenszel test with stratification on type of treatment [(R-CHOP or (R)-ACVBP)]. Other studied polymorphisms were rs172378 in C1QA, rs1001179 and rs10836235 in CAT, rs2740574 in CYP3A4, rs1695 in GSTP1, rs12210538 and rs714368 in SLC22A10, rs4880 in SOD2, rs8175347 in UGT1A1 and copy number variations for GSTM1 and GSTT1.

Results presented here suggest that polymorphisms in ABCB1 (coding for the ATP binding cassette efflux protein Pgp) and CBR1 (coding for carbonyl reductase 1) could be used to identify NHL patients with a high risk of myeloid or digestive toxicity after treatment with R-CHOP-type regimens. Both of these genes have been reported to be involved in anthracycline transport or metabolism, indicating a possible mechanistic role in the correlation.4

ABCB1 is well described as a membrane transporter of anthracyclines, as well as other hydrophobic drugs used in the treatment of NHL patients, such as etoposide or vinca alkaloids.5 Several studies have correlated ABCB1 polymorphisms with sensitivity to therapeutic compounds. Interestingly, Yao et al. reported on a set of tag SNP in ABCB1 and did not observe any correlation to grade 3 or higher hematologic or intestinal toxicity in cyclophosphamide- and doxorubicin-treated breast cancer patients.6 However, these tag SNP did not cover the position of rs2229109 that was shown to be correlated in our study. This discrepancy could be due to the fact that rs2229109 is a non-synonymous polymorphism, and that the two protein variants potentially do not have the same activity.

Carbonyl reductase 1 (CBR1) is an anthracycline-metabolizing enzyme. It was reported that increased expression of CBR1 was associated with reduced sensitivity of fresh AML blasts to daunorubicin in vitro and was positively correlated with increased intracellular level of daunorubicinol, a major catabolite of daunorubicin.7 CBR1 has also been suspected to be involved in the occurrence of anthracycline-related toxicities as non-synonymous SNP were associated with reduced metabolism of doxorubicin and daunorubicin, and as these metabolisms were correlated to the expression of carbonyl reductases.98 In addition, polymorphisms in CBR1 were correlated with altered pharmacokinetics with increased exposure to doxorubicin.10 Finally, the cardioprotectant flavonoid 7-monohydroxyethyl rutoside was shown to behave as a CBR1 inhibitor.11 Our data support the role of CBR1 variants as predictors of anthracycline-related toxicity. The two SNP in CBR1 are situated very close to each other on chromosome 21 and are in high linkage disequilibrium, explaining the similar results obtained for the two variants.7

An important question when considering correlations of SNP with drug-induced toxicity is whether the polymorphisms actually impact on the expression level or functionality of the corresponding proteins. The effect of SNP on ABCB1 has been reviewed by Leschziner et al.12 Most studies concluded that there is no effect of rs2229109 (C1199A) on mRNA or protein expression. Concerning variants in CBR1, rare variants for both rs20572 and rs9024 were shown to be associated with lower mRNA expression than the most frequent alleles.13 The rare variant of rs9024 was also shown to be associated with resistance to regulation by hsa-miR-291 but not by has-miR-574-5p.14

The strengths of our studies include the number of patients, the prospective collection of toxicity data in a randomized setting, and the fact that the SNPs analyzed have already been reported to be associated with toxicity, thus making some of these results a validation of previous hypothesis-generating studies. Limitations include the low number of patients in some groups and the number of statistical analyses performed without any correction for multiple tests. Differences in the toxicity parameters, which appear to be significantly correlated in the different analyzed groups (R-CHOP, R-ACVBP or all patients), are most probably due to the differences in the numbers of patients and their associated statistical power.

This study indicates the role for ABCB1 and CBR1 polymorphisms in the occurrence of severe myeloid and digestive toxicity in patients receiving CHOP-like regimens for the treatment of NHL. While confirmation of our results in other patient cohorts is required, the study design, with, in particular, the literature-based SNP selection, gives a partial validating value to our results already. Our study also confirms that the (R)-ACVBP regimen induces more toxicity in lymphoma patients than R-CHOP,15 suggesting that markers predictive for toxicity would be of particular interest in patients receiving high-dose R-CHOP type regimens or in patients with pre-existing comorbidities.