Expression of the nuclear oncogene Ski in patients with acute myeloid leukemia treated with all-trans retinoic acid

For acute myeloid leukemia (AML), a two-hit model has been proposed for leukemogenesis:1,2 the first mutation involving tyrosine kinase signaling such as a RAS mutation or activation of this pathway via Flt-3-mutation, the second aberration via truncated or fusion transcripts caused by chromosomal translocations, deletions or inversions.3

We have recently found that the nuclear oncogene Ski is overexpressed in AML, especially in AML with deletion of chromosome 7.4 Ski represses all-trans retinoic acid (ATRA) signaling and myeloid differentiation in AML cells in vitro. Expression of a Ski mutant unable to interact with N-CoR (nuclear receptor corepressor) and thus unable to co-operate with histone deacetylases (HDAC)5 did not inhibit ATRA signaling. In vitro treatment of cells with the HDAC inhibitor valproic acid abrogated the inhibitory effect of wild type Ski.4

ATRA has proved to be a highly efficient agent in the therapy of patients with acute promyelocytic leukemia (APL; FAB classification M3) characterized by the fusion protein PML-RARα, which results in impaired retinoic acid signaling.6 Discrepant results have been published regarding the use of ATRA in AML other than APL. A study from the British Medical Research Council (MRC) did not find a significant difference in survival of AML patients except AML-M3 up to the age of 55 years treated with chemotherapy versus chemotherapy plus ATRA.7 By contrast, a study of the German-Austrian AMLSG study group including 242 AML patients >60 years, excluding APL patients, revealed a significant advantage for the group receiving ATRA together with ICE (idarubicin, cytarabine, etoposide) versus ICE alone. The combination improved complete remission (CR) rates, event free survival and overall survival (OS) in these AML patients. However, in this study the ATRA schedule and age were different from the MRC trial.8

Figure 1.Cumulative incidence of relapse for the 48 elderly acute myeloid leukemia patients in complete remission treated with ICE versus ICE plus ATRA with low (A) or high (B) Ski expression level. The cut-off level was arbitrarily defined at the median Ski expression (low Ski ≤0.36; high Ski >0.36). Gray’s k-sample test for (A) is p=0.11, for (B) p=0.57.

The clinical data, showing a slight advantage for patients treated additionally with ATRA, together with our in vitro data, showing that Ski is a repressor of retinoic acid signaling, suggested the hypothesis that ATRA treatment should be more efficient in patients with low Ski expression compared with high Ski expression.

We retrospectively analyzed Ski expression in a sub-sample of 134 patients randomized into both arms of the AMLSG HD98B study8 for whom leukemia samples were still available. In 2 out of 134 patients, cDNA synthesis was not sufficient. Out of the remaining 132 patients, 65 patients had received ICE (12 mg/m idarubicin, day 1 and 3; 100 mg/m cytarabine, days 1 to 5; 100 mg etoposide, day 1 and 3) as induction chemotherapy, and 67 patients were treated with chemotherapy ICE plus oral ATRA (45 mg/m; days 3 to 5 and 15 mg/m days 6 to 28). Criteria for response and definition of relapse have been previously described.8

Figure 2.Relative expression level of Ski related to relapse free survival in the samples of the 48 elderly acute myeloid leukemia patients in complete remission. Open circles indicate ATRA-ICE treatment, events of relapse; closed circles indicate ICE treatment, events of relapse; open triangles indicate ATRA-ICE treatment, censored observations; closed triangles indicate ICE treatment, censored observations; open squares indicate ATRA-ICE treatment, cases of death due to other (competing) causes; closed squares indicate ICE treatment, cases of death due to other (competing) causes.

RNA extraction and cDNA synthesis were performed using standard protocols.4 Quantitative Ski PCR was performed as previously described.4 The staff of the molecular laboratory was blinded for the clinical data. Statistical analysis was carried out using a SPSS software package 12.1 as well as the software package R9 (http://www.R-project.org). OS was calculated using the Kaplan-Meier method and cumulative incidence of relapse (CIR) was analyzed by a competing risk analysis. Differences between groups were evaluated using the log-rank test and Gray’s k-sample test.10 The median Ski expression was chosen as the arbitrary cut-off level. Accordingly, low Ski expression was defined as ≤0.36.

Statistical analysis showed that both groups of patients (ATRA-ICE versus ICE) were comparable for age, cytogenetic risk, lactate dehydrogenase level and leukocyte count. There was no difference in expression of Ski in patients randomized to ICE only or ICE plus ATRA. A comparison of patients whose AML cells expressed Ski at high levels (Ski >0.36) with AML patient cells with low Ski levels (Ski <0.36) found no difference in OS (p=0.57). Out of the 132 patients, 48 went into CR. Response to chemotherapy (CR, refractory disease and early death) was not associated with Ski expression level.

According to our hypothesis, ATRA should best protect patients with low Ski expression from relapse. Therefore, CIR and relapse free survival were evaluated in both treatment arms according to low and high Ski expression. Analysis of CIR for patients treated with versus without ATRA (26 patients with ATRA-ICE vs. 22 patients with ICE) showed a distinct difference for the subgroup with high or low Ski expression (Figures 1A and 1B). These 48 patients were also analyzed for relapse free survival (RFS). There was a trend towards longer RFS in patients with low Ski expression randomized to ATRA therapy. Of 8 patients with RFS over 1,000 days, 7 were in the ATRA-ICE, and one in the ICE group. However, the Gray test for the low Ski group was not significant (p=0.11). Interestingly, the cells of the 5 long-term RFS patients randomized to ATRA had very low Ski levels compared with other AML patients (relative Ski expression of those 5 patients: 0.03; 0.059; 0.098; 0.12; 0.158; compared with range 0.023–4.39 in all 132 AML samples (median 0.36)). Of the 5 patients with longest RFS, 4 were in the low and one in the high Ski group (Figure 2). These data suggest that patients with very low Ski expression levels might be a special subgroup responsive to ATRA treatment.

To summarize, there was a trend to prolonged RFS for patients in CR with low Ski expression treated with additional ATRA compared with those who received ICE treatment only. Interestingly, long term survivors (>1,000 days) treated with ATRA had very low Ski expression levels. In AML, low Ski expression seems to mark a subgroup responsive to ATRA when given in combination with ICE during induction chemotherapy. This should be addressed in future trials.

Acknowledgments

we thank Prof. Konstanze Döhner and Prof. Hartmut Döhner (Department of Internal Medicine III, University of Ulm, Germany) for providing clinical data and leukemia specimens from the AMLSG HD98B trial. We thank Eva Bug and Dagmar Daniels for their expert technical assistance.

Footnotes

• Funding: this study was supported by Deutsche Forschungs-gemeinschaft, SFB TR17, by the BMBF (Leukemia and CancerNet NGFN-2; Kompetenznetz Leukämie), and the German José Carreras Leukämie Stiftung.