Concomitant WT1 mutations predict poor prognosis in acute myeloid leukemia patients with double mutant CEBPA

Acute myeloid leukemia (AML) with double mutant CCAAT/enhancer binding protein α (CEBPA^dm) is a new entity in the 2016 World Health Organization (WHO) classification with unique biologic features and prognostic implications.^1,2 The incidence of CEBPA^dm ranges from 7.5% to 11% in AML.^1,3,4 CEBPA^dm AML patients, when treated with standard chemotherapy, achieve a high complete remission (CR) rate. However, relapse occurs in 40% of patients who attain CR.¹ This has raised the clinically relevant question whether concomitant genetic alterations influence the prognosis of CEBPA^dm patients. Apart from GATA2, the prognostic impact of other concomitant gene mutations is largely unsettled because limited patient numbers preclude informative analyses.⁵ Given that AML is a heterogeneous disease, risk-adapted treatment may not only improve the prognosis, but also reduce toxicity from the therapy. Allogeneic hematopoietic stem cell transplantation (allo-HSCT) in first CR is not beneficial for cytogenetically normal AML (CN-AML) patients with CEBPA^dm.⁶ If any concomitant mutations adversely affect the clinical outcome of CEBPA^dm patients, it will be interesting to know whether allo-HSCT should be performed for these patients. As yet, there is no data to answer this question.

In this study, the aim was to identify additional mutations in CEBPA^dm AML patients that conferred prognostic significance. Furthermore, we investigated the role of allo-HSCT in CEBPA^dm patients with concurrent adverse-risk mutations. Mutation analyses in CEBPA and 19 other relevant genes, including FLT3-ITD, FLT3-TKD, NRAS, KRAS, KIT, PTPN11, RUNX1, GATA2, MLL/PTD, ASXL1, IDH1, IDH2, TET2, DNMT3A, SF3B1, SRSF2, U2AF1, NPM1, WT1, and TP53 were performed by Sanger sequencing for patients (n=500) diagnosed from 1994 to 2007.^7,8 For patients (n=256) diagnosed after 2008, Ion Torrent next generation sequencing (NGS) (Thermo Fisher Scientific, MA, USA) was performed. The WT1 mutations detected by NGS were all confirmed by Sanger sequencing.

We identified 102 (13.5%) CEBPA-mutated patients from 756 patients with newly diagnosed de novo AML (Online Supplementary Table S1); 33 (4.4%) had CEBPA single mutation (CEBPA^sm) and 69 (9.1%), CEBPA^dm. Sixty-nine CEBPA^dm patients were found to have 109 distinct mutations (Figure 1A, Online Supplementary Table S2). All patients had a combination of one N-terminal and one C-terminal mutation. Most (53 of 56, 94.6%) of the N-terminal mutations were frame-shift mutations, while most (42 of 53, 79.2%) of the C-terminal mutations were in-frame mutations with internal tandem duplications clustered in the junction between the basic region and the leucine zipper.

Figure 1.

CEBPA double mutations (CEBPA^dm) in de novo AML patients. (A) The diagram of concurrent mutations in patients with CEBPA^dm. (B) The distribution of concomitant mutations in AML patients with either CEBPA^dm or wild-type CEBPA. (C) Kaplan-Meier plots for OS (left) and DFS (right) according to WT1 mutation status in CEBPA^dm patients.

CEBPA^dm patients were significantly younger and had higher hemoglobin levels at diagnosis than CEBPA^sm and CEBPA wild-type patients. All except one CEBPA^dm patient had intermediate-risk cytogenetics (P<0.0001) (Figure 1A). The most frequent intermediate-risk cytogenetic change was del(9) (n=4, 5.8%), and CN- AML occurred in 81.2% of CEBPA^dm patients (n=56).

Fifty (72.5%) of the CEBPA^dm patients had additional genetic alterations (Online Supplementary Table S3). Among them, 29 (58%) had one, 17 (34%) had two, 3 (6%) had three and 1 (2%) had four changes. The most common concurrent molecular event in CEBPA^dm patients was GATA2 mutation (33.8%), followed by FLT3-ITD (14.5%), NRAS (14.5%), TET2 (13.2%), and WT1 (11.8%) mutations. GATA2 was more frequently mutated in CEBPA^dm patients than in CEBPA wild-type patients (33.8% vs. 2.8%, P<0.0001). In contrast, CEBPA^dm patients less frequently harbored NPM1, ASXL1, IDH2, DNMT3A and RUNX1 mutations (Figure 1B).

Survival analyses were restricted to 530 patients, including 62 CEBPA^dm patients and 468 others (22 with CEBPA^sm and 446 CEBPA wild-type), who received standard intensive chemotherapy. The CR rate was 90.2% for CEBPA^dm patients and 72.2% for others (P=0.003). In multivariate analysis, CEBPA^dm was an independent favorable prognostic factor for OS and DFS (RR 0.420, 95% CI 0.246–0.718, P=0.002 and RR 0.544, 95% CI 0.351–0.842, P=0.006, respectively, Online Supplementary Table S4). Of the 56 CEBPA^dm patients who achieved first CR, 10 received allo-HSCT and 46 had postremission chemotherapy alone. The reasons for frontline allo-HSCT were persistent residual leukemia cells in 4 patients, concurrent FLT3-ITD in 3 patients, initial hyper-leukocytosis in 2 patients and complex cytogenetics in 1 patient. Intriguingly, the relapse rate was 45.7% in the postremission chemotherapy group and 0% in the allo-HSCT group (P=0.009). DFS was significantly better in the allo-HSCT group (median, not reached (NR) vs. 59.4 months, P=0.023) than in the chemotherapy group, while OS was not different (P=0.247) (Online Supplementary Figure S1).

We further analysed the prognostic significance of concomitant gene mutations with a frequency above 10% in CEBPA^dm patients. WT1-mutated patients tended to have a lower CR rate (71.4% vs. 92.5%, P=0.14) and a higher relapse rate (80% vs. 34%, P=0.047) compared to those with wild-type WT1 (Online Supplementary Table S5). With a median follow up of 69.7 months (range, 1.2–230 months), WT1-mutated patients had a significantly shorter OS and DFS than WT1-wild patients (median, 14 months vs. NR, P=0.021; 7.8 months vs. NR, P=0.008, respectively; Figure 1C). According to the 2017 European LeukemiaNet (ELN) classification, the AML patients were stratified into three risk groups (Figure 2A). Integration of WT1 mutations could further divide the ELN favorable-risk cohort into three subgroups: CEBPA^dm WT1-mutated patients, CEBPA^dm WT1-wild patients and others. As shown in Figure 2B, CEBPA^dm patients with WT1 mutations had worse outcome than other ELN favorable-risk patients. Sequential analyses of WT1 mutations revealed that the mutations in three WT1-mutated patients in the study were lost at CR, but regained at relapse. The mutation burden could either increase or decrease at relapse. Of the 116 WT1-wild patients studied, three acquired a novel mutation at relapse (Online Supplementary Table S6).

Figure 2.

Risk stratification of the ELN favorable group according to the status of CEBPA and WT1 mutations. (A) Kaplan-Meier plots for OS and DFS stratified by the 2017 ELN risk categories. (B) ELN favorable group could be further separated into three subgroups according to the status of CEBPA and WT1 mutations. CEBPA^dm patients with concurrent WT1 mutations had OS and DFS poorer than other ELN favorable-risk patients, but similar to those with the ELN intermediate (CR 76.8%, relapse rate 56.0%, median OS 26.0 months, median DFS 10.2 months) or unfavorable-risk category (CR 53.7%, relapse rate 61.3%, median OS 11.6 months, median DFS 2.1 months). ELN:European LeukemiaNet.

Regarding other concomitant gene mutations, GATA2 mutation was correlated with a trend of longer DFS (median, NR vs. 16.1 months, P=0.078). FLT3-ITD, NRAS and TET2 mutations seemed not to have implications on the clinical outcome (Online Supplementary Table S7 and Online Supplementary Figure S2).

To our knowledge, this is the first study to demonstrate the prognostic impact of concurrent WT1 mutations on CEBPA^dm patients. WT1 mutation occurs in 6–10% of AML patients and is associated with poor prognosis in CN-AML and non-selective AML patients.^9,10 Intriguingly, WT1 mutations are frequently identified in CEBPA^dm patients.¹¹ We distinctly found that WT1 mutations were associated with poor clinical outcome in CEBPA^dm patients. Furthermore, we showed that integration of WT1 mutations could refine the ELN risk stratification in favorable-risk subgroups. The prognostic impact of concomitant mutations in CEBPA^dm patients have not been widely assessed with the exception of GATA2 and TET2 mutations.^3,5,12 Grossmann et al. showed that the presence of TET2 mutations correlated with worse survival. In contrast, we did not find the prognostic impact of TET2 mutations in CEBPA^dm patients.

A high frequency of TET2 co-mutation (around 34%) in CEBPA^dm patients was reported previously,^3,5 while it was only 13.2% in our study. The reason that our results were very different from those reported in other geographical areas might partly be explained by the difference in patient characteristics. The CEBPA^dm patients in the studies of Grossmann et al.⁵ and Fasan et al.³ were significantly older than ours (median age, 57.5 and 56.3 vs. 40 years). It is well documented that TET2 mutations occur more frequently in elderly AML patients than younger ones¹³ and this was reflected in the different prevalence of TET2 mutations between our cohort and the other two. Furthermore, for TET2 missense mutations, the missense mutations with unknown biologic significance were censored, which would possibly lead to lower frequency of TET2 mutations in this study.¹³ The ethnic difference might be another influencing factor. Recently, CSF3R mutation was found closely associated with CEBPA mutation in both adult and pediatric AML patients.¹⁴ Unfortunately, CSF3R mutation was not included in our panel.

According to the current ELN guidelines, allo-HSCT is not routinely recommended in CEBPA^dm patients in first CR. Indeed, though postremission chemotherapy alone in first CR correlated with a significantly higher relapse rate and shorter DFS as compared with allo-HSCT, the high relapse rate in the chemotherapy subgroup did not translate into a significant inferior OS because relapsed patients still showed a high second CR rate.^6,15 The relapse rate of CEBPA^dm patients after first CR in this study was 37.5% in total CEBPA^dm patients and 45.7% in the postremission chemotherapy subgroup, which was comparable with that reported previously (36.2%–41%).^1,6 Surprisingly, all WT1-mutated CEBPA^dm patients, if not transplanted in first CR, encountered disease relapse (Table 1). The second CR rate was only 25% after re-induction, which was much lower than that (around 80%) in the total CEBPA^dm cohort.⁶ Taken together, it is suggested that CEBPA^dm patients with WT1 co-mutation receive HSCT in first CR given the high relapse rate and gravid prognosis if relapse occurs. Further prospective randomized studies are warranted to validate the point.

Table 1.

Clinical characteristics and treatment outcome of CEBPA^dm patients with concomitant WT1 mutations.

This study clearly demonstrates the heterogeneous clinical outcome of CEBPA^dm patients and provides useful clinical information on refining the 2017 ELN risk categorization. Concomitant WT1 mutations suffice to be a marker for dismal prognosis in CEBPA^dm patients and help in our understanding of the process of leukemogenesis in this group. More importantly, allo-HSCT in first CR may be indicated for long-term disease control of this poor-risk entity.