Genetic Variations in Calicheamicin Pathway Genes Are Predictors of Gemtuzumab Ozogamicin Response in AML Patients: Results from COG-AAML0531 Study

Background: The antigen currently most exploited for targeted therapy in AML is CD33. In particular gemtuzumab ozogamicin (GO), an immunoconjugate with CD33 antibody linked with a cytotoxin caleachimicin that causes DNA strand breaks that elicit a DNA repair response and, if damage is overwhelming, leads to apoptosis and cell death. Although GO was withdrawn from the market in 2010, results of follow up clinical trials including phase 3 studies indicate that the addition of GO to conventional chemotherapy significantly improves EFS, DFS and risk of relapse in a subset of newly diagnosed AML patients.

Objective: To identify genetic variation in genes involved in calicheamicin transport and metabolism (ABCB1, glutathione-S-transferases, CYP3A), DNA-repair/damage response pathway genes (DDB2, XRCC4/5, ERCCs, XPC, PARP1, LIG4, PIK3CA, AKT1/2, ATM, and ATR etc.), and apoptosis-related genes on response to GO-based therapy in patients enrolled on COG AAML0531.

Methods: Patients with newly diagnosed AML enrolled on COG trial AAML0531 who were randomized to receive either standard 5 course chemotherapy alone (Arm A) or chemotherapy with two doses of GO (3mg/m2/dose) during induction 1 and intensification (Arm B) were included in the study. Genomic DNA samples from 942 patients who consented for the study were genotyped for 158 SNPs. SNPs with MAF <0.1 and/or not following HWE were excluded and 123 SNPs were analyzed for association with outcome in Arm A and Arm B.

Results: Since AAML0531 was designed to test the benefit of GO in a randomized fashion; we identified variants that influence GO efficacy from those that have an impact on efficacy of standard chemotherapeutics. Most significantly our results identified SNPs in ABCB1, GSTM1 and SNPs in DNA damage repair genes (DDB2, PARP1, XRCC3, XRCC1, XPC) were significantly associated with outcome (DFS, OS, EFS, Risk of relapse) in patients in Arm B receiving GO based therapy as compared to patients in Arm A receiving standard therapy. Figure 1 shows an example of an intronic SNP in DDB2 (Damage –specific DNA binding Protein 2), that is predictive of DFS (p=0.0001) and risk of relapse (p=0.0006) in patients receiving GO based therapy but not in patients receiving standard therapy (p>0.05). DDB2 is a critical component of the nucleotide excision repair pathway. Additionally, we identified some SNPs that demonstrated risk group specific association with outcome; one such example being a SNP in PTEN that was predictive of DFS (p=0.008), OS (p=0.03) and Risk of Relapse (p=0.0028) after Induction 1 within standard risk group patients in Arm B (GO) but not in standard risk patients (p>0.5) treated in Arm A (No GO).

Conclusion: A detailed understanding of the interplay between SNPs and therapeutic response to GO and conventional chemotherapy will have significant consequences for disease prognostication and therapy. Given the fact that some SNPs demonstrate risk group dependent association; integration of SNP data into cytogenetic/molecular based risk classification models would facilitate our ability to both prognosticate and provide optimal risk-stratified treatments in AML.

Acknowledgments: Research reported was supported by the COG, the NCI of the NIH under award number U10CA180899, U10CA180886, U10CA98413 and U10CA098543 as well as NCI-R21CA155524 (Lamba and Walter). We are also thankful to Dr. Stanley Pounds and Dr. Xueyuan Cao for sharing the statistical analysis codes.

A disclaimer –"The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health"