A 16-Gene Signature Associated with High Levels of Wilms Tumor-1 (WT1) Expression Is an Adverse Prognostic Factor in Acute Myeloid Leukemia

Wilms Tumor-1 (WT1) expression level has long been found to be implicated in acute myeloid leukemia (AML) prognosis, though this is not reflected in current AML risk stratification. We hypothesized that a gene expression profile (GEP) associated with WT1 expression could be of prognostic value. We analyzed two publically available AML GEP series in order to identify a gene signature associated with high- WT1 expression (hi- WT1 ). The first, herein called Netherlands series, comprised of 524 younger adult patients who have been treated according to sequential Dutch-Belgian Hemato-Oncology Cooperative Group and the Swiss Group for Clinical Cancer Research (HOVON/SAKK) AML-04, 04A, 29, 32, 42, and 43 protocols (GSE14468). The second series, herein called Germany series, consisted of 562 younger and older AML patients who were treated in the German AMLCG 1999 trial (GSE37642). We identified the hi- WT1 gene sets by comparing GEP among the highest and lowest quartiles of WT1 expression in both AML studies. About 62% of the probe sets in the Netherlands hi- WT1 set were found to be common with the Germany hi- WT1 set; 97% differed in the same direction. Moreover, a high degree of correlation of the fold differences was found among the two hi- WT1 sets (r2 = 0.81, p < 10-18), collectively suggesting a biological relevance for hi- WT1 gene sets. In order to assess the prognostic implication of the hi- WT1 set, we used K-Nearest Neighborhood algorithm to generate various lists of hi- WT1 probe sets predicting event-free survival (EFS) as the favorable, and all others (dead, no remission, progressive disease/relapse) as the unfavorable events in the Netherlands series. Stepwise screening of the lists of 10 to 100 probe sets by Cox Regression identified a 16-gene subset of hi- WT1 set with distinct GEP and as the optimal predictor of overall survival (OS) and EFS in the Netherlands series. It comprised of GPR56, FAM30A, NGFRAP1, WBP5, LTK, PTP4A3, CD109, ZC3H12C, PYGB, CHIC1, HAVCR2, TMEM110, HAL, HDAC4, BLVRA, and P2RY2 . In this series, the hi- WT1 cluster of patients showed lower 5y-probability of OS (10% vs 44%) and EFS (6% vs 38%) as compared to the remaining clusters. Accordingly, the hi- WT1 cluster showed shorter median OS (8.3 [CI 6.7-9.9] vs 31.3 [CI 17.1-45.5] months, P = 6 x 10-18) and EFS (4.9 [CI 3.2-6.5] vs 14.5 [CI 9.4-19.5] months, P = 3 x 10-16). Although the hi- WT1 cluster was associated with some of the cytogenetic and molecular aberrations including FLT3 -ITD, it remained significant for both OS ( P = 3 x 10-5) and EFS ( P = 3 x 10-6) after adjustment for known AML risk factors. In order to validate our findings, we performed a supervised clustering of the Germany AML series using the 16-gene signature. The hi- WT1 cluster predicted both adverse OS and relapse-free survival (RFS) (Fig. 1), which remained statistically significant after adjustment for known AML risk factors (Table 1). The median OS was 7.1 (CI 5.7-8.5) months for the hi- WT1 cluster as compared to 20.1 (CI 15.2- 25.0) months for other cases ( P = 2 x 10-13), and the median RFS was 5.8 (CI 4.8-6.8) vs 20.3 (CI 10.6-30.0) months, respectively ( P = 2 x 10-11). Moreover, the rate of complete remission was significantly lower in hi- WT1 cluster as compared to other clusters (42% vs 61%, P = 2 x 10-5). The positive (PPV) and negative predictive value (NPV) of the marker for prediction of adverse OS were 90% and 34%, respectively. These values were found to be 88% and 38%, respectively, for prediction of adverse RFS. MetaCore analysis identified the Antigen Presentation by MHC-II as the most implicated biological pathway in hi- WT1 sets, with many genes downregulated in the pathway. In brief, we identified a 16-gene signature associated with WT1 expression and demonstrated its adverse and independent prognostic impact in adult AML patients. These promising results should be validated in further trials and provide new clues to the molecular mechanisms underlying WT1 regulation. ![Figure 1][1] Figure 1 | Variable | OS | RFS | | -------------- | ------- | ------------- | ----------- | | P val | HR | 95% CI for HR | P val | HR | 95% CI for HR | | Lower | Upper | Lower | Upper | | WT1 signature | .001 | 1.485 | 1.186 | 1.860 | .000 | 1.834 | 1.326 | 2.536 | | Age | .000 | 1.291 | 1.197 | 1.391 | .001 | 1.198 | 1.081 | 1.327 | | ELN2 | .000 | 2.395 | 1.743 | 3.291 | .000 | 3.147 | 2.124 | 4.663 | | ELN3 | .000 | 2.395 | 1.732 | 3.312 | .000 | 2.153 | 1.405 | 3.299 | | ELN4 | .000 | 3.306 | 2.376 | 4.599 | .000 | 6.401 | 4.027 | 10.175 | Table 1. Multivariate analysis of OS and RFS in the Germany AML series. Disclosures No relevant conflicts of interest to declare. [1]: pending:yes