Gene Expression Analyses At Time Of Diagnosis Indicate Biomarkers Predictive Of Therapeutic Response In Acute Myeloid Leukemia

In this study we aimed to identify biomarkers predictive of clinical response in acute myeloid leukemia (AML). For this purpose mRNA was isolated from diagnostic samples from 42 AML patients (“training cohort”) and subjected to Affymetrix^® gene expression analysis. All patients entered complete remission (CR) after high-dose induction chemotherapy, reaching a median CR duration of 161 (range 12-3701) days. Samples from patients with “short CR duration” (<6 months, n=24) and “long CR duration” (>6 months n=18), respectively, were pooled and compared. Gene expression analyses revealed 383 genes to be up-regulated and 610 genes down-regulated more than two fold in samples from patients with short, as compared to those with long CR duration. Ten genes were found to be up-regulated >30 times, with the runt-related transcription factor 1; translocated to 1 (cyclin D-related) (RUNX1T1) gene showing the highest differential expression (116-fold), while annexin 1 (ANXA1) was the most down-regulated gene (58-fold). Significantly higher transcript levels of RUNX1T1 were confirmed in the poor outcome group when performing quantitative real time polymerase chain reaction (qRT-PCR) on individual samples (n=20, p<0.002). Subjecting our data to pathway analysis (Ingenuity^®) comparing the same groups, we focused on RUNX1T1 and created a network of RUNX1T1 interacting molecules. Utilizing the IPA database to create a network over interacting molecules of RUNX1T1, we identified a majority of these to be transcriptional regulators and among them the transcription factor 3 (TCF3) to be up-regulated 5-fold in patients with short CR duration.

Our training cohort data were validated in silico extracting information from an independent study by Metzeler et al, publicly available from Oncomine^® (www.oncomine.org) and encompassing diagnostic samples from 162 AML patients. Among genes differentially and similarly regulated in poor responders in both the training and validation cohorts we observed TCF3, chemokine (C-X-C motif) ligand 3 (CXCL3), Zinc finger, MIZ-type containing 1 (ZMIZ1) (up-regulated) and Peroxiredoxin 2 (PRDX2) (down-regulated). Analyses of clinical outcome revealed that AML patients with a high ZMIZ1 expression had a significantly decreased overall survival (OS) as compared to that of patients with a low ZMIZ1 expression (p <0.03). ZMIZ1 has been reported to be involved in tumor growth in general and suggested to interact with activated NOTCH in inducing leukemia, but its more precise role in AML is still unclear.

In conclusion, we report clear differences in gene expression in diagnostic samples from AML patients with subsequent poor vs. better long-term clinical outcome to therapy, thus to constitute possible novel predictive biomarkers for response. In our training set RUNX1T1 was the most differentially expressed gene, while ZMIZ1 was upregulated in both the training and validation sets and significantly associated with survival. Further, functional studies of differentially expressed genes in clinical subsets of AML patients appear warranted.