Prognostic relevance of molecular measurable residual disease detection in AML with mutated CEBPA

Mutations in the CCAAT/enhancer binding protein alpha (CEBPA) are found in 2%–15% (mean 5%) of de novo acute myeloid leukemia (AML) patients.¹ CEBPA encodes a transcription factor that is important for hematopoietic stem cell (HSC) self-renewal as well as myeloid differentiation of hematopoietic progenitors.² The characteristic mutations in the CEBPA protein involve frame-shift mutations in the N-terminal transactivation domains and in-frame mutations in the C-terminal basic leucine zipper (bZIP).² Recently, the in-frame CEBPA bZIP mutations were incorporated in the 2022 European LeukemiaNet (ELN) risk classification as a favorable risk factor,³ replacing the CEBPA double mutations (CEBPA^dm) as favorable marker in the preceding ELN2017 guidelines.⁴

Recent advances in molecular minimal residual disease (MRD) detection in complete remission (CR) have shown profound prognostic value of a selection of AML-specific gene mutations.^5-7 However, the prognostic impact of persisting CEBPA mutations in CR has not been thoroughly investigated in AML patients. Here, we explored the prognostic impact of mutant CEBPA MRD in a relatively large cohort of 84 AML patients with mutated CEBPA by deep next-generation sequencing (NGS).

AML patients enrolled in the Dutch-Belgian Cooperative Trial Group for Hematology-Oncology (HOVON) and Swiss Group for Clinical Cancer Research (SAKK) clinical trials HO42A, HO92, HO102, HO103, and HO132 were included. All trial participants provided written informed consent in accordance with the Declaration of Helsinki, and were treated according to their respective treatment protocol (www.hovon.nl). Patients were assessed for gene mutations on diagnostic bone marrow samples using the TruSight Myeloid Sequencing panel (Illumina) targeting 54 frequently mutated genes in AML.⁸ Since NGS quality and depth of sequencing of the CEBPA gene varies when using this gene panel, CEBPA targeted sequencing was additionally performed on DNA of these diagnostic samples using a custom four-amplicon polymerase chain reaction (PCR) approach (amplicons A, B, C1, C2; Supporting Information: Methods).⁹ A total of 144 CEBPA mutant patients out of 1913 AML cases was identified, of which 84 with available CR samples were included for mutant CEBPA MRD assessment. Targeted deep sequencing was performed on 100 ng of DNA obtained at CR, after two cycles of standard induction chemotherapy and pretransplant, using the four-amplicon PCR-based NGS approach.^{8, 9}

At diagnosis, 43 out of 84 cases harbored a mutation in the bZIP region (bzip), whereas 41 carried other mutations (non-bzip) (Supporting Information S1: Table 1). All CEBPA^bzip mutations were in-frame insertions. CEBPA^bzip patients were significantly younger, but no significant differences were present between CEBPA^bzip and CEBPA^non-bzip patients in terms of sex, blast, and white blood cell counts at diagnosis, consolidation therapy, or treatment protocol (Supporting Information S1: Table 2).

Mutations were subsequently classified according to the ELN2017 (single mutant: CEBPA^sm [n = 28] vs. double mutant: CEBPA^dm [n = 56]) and ELN2022 (CEBPA^bzip [n = 43] and CEBPA^non-bzip [n = 41]) risk stratifications (Supporting Information S1: Figure 1). All CEBPA^bzip patients remained part of the favorable risk group in ELN2022, whereas CEBPA^non-bzip AML patients were stratified into favorable (27%), intermediate (39%), or adverse (34%) risk groups according to the ELN2022 criteria. Compared to ELN2017, 17 out of 84 CEBPA mutant AML patients were re-stratified into a different risk ELN2022 category, that is, 15 CEBPA^dm AML patients (27%) did not carry a favorable in-frame bZIP mutation, whereas two CEBPA^sm AML patients did (Supporting Information S1: Figure 1).

In the complete CEBPA mutant AML cohort, TET2 was most frequently comutated (24%), followed by GATA2 (23%), NPM1 (17%), NRAS (17%), and DNMT3A (16%) (Supporting Information S1: Figure 2). Six CEBPA mutant AML patients did not have any known co-mutation. NPM1 (32%, p < 0.001), DNMT3A (27%, p = 0.006), SRSF2 (20%, p = 0.014), RUNX1 (17%, p = 0.028), IDH2 (17%, p = 0.005), ASXL1 (15%, p = 0.011), FLT3-TKD (15%, p = 0.011), and IDH1 (12%, p = 0.024) were significantly more often mutated in CEBPA^non-bzip patients, whereas mutations in GATA2 (40%, p < 0.001) and WT1 (23%, p = 0.026) were more frequent among CEBPA^bzip patients (Supporting Information S1: Figure 2).

Next, we examined the differences in clinical outcome between the different CEBPA mutant AML subgroups. Overall survival (OS) and cumulative incidence of relapse (CIR) were compared between the subgroups using Kaplan–Meier estimates for all AML patients with mutant CEBPA at diagnosis (n = 84). OS and CIR were calculated from the date of sampling in CR to the date of an event. As expected, the presence of a CEBPA^bzip mutation at diagnosis was associated with improved OS compared to CEBPA^non-bzip mutations (p = 0.05).^{10, 11} No significant difference was observed in CIR (Supporting Information S1: Figure 3).

Altogether, these characteristics demonstrate that our cohort of 84 AML cases is representative for CEBPA-mutated AML.^{10, 11}

The presence of mutant CEBPA MRD was determined using NGS deep sequencing (average read depth: 579,164×, range: 97,806×–1,566,187×) enabling detection of mutant CEBPA at a sensitivity up to 0.0004% VAF (VAF ≥ 0.0004% [indel, n = 58] and ≥0.03% [SNV, n = 3]; Supporting Information S1: Figure 4).⁹ Persistence of CEBPA mutations was demonstrated in 42 out of 84 mutant CEBPA AML cases. Allogeneic hematopoietic stem cell transplantation (HSCT) was carried out in 23 patients (27%, Supporting Information S1: Table 2), and censoring at allogeneic HSCT was performed in all survival analyses. In AML patients with detectable CEBPA MRD regardless of mutation type, relapse rates were increased and OS was inferior although not statistically significant (p = 0.24 and p = 0.49 respectively; Figure 1A,B). In addition, multiparameter flow cytometry (MFC) MRD was assessed in 68 out of 84 mutant CEBPA AML patients. Twelve had detectable MFC-MRD (18%), and did not correlate with mutant CEBPA MRD status. Detection of MFC-MRD did not lead to a significantly increased CIR or inferior OS in mutant CEBPA AML patients (Supporting Information S1: Figure 5). Due to the low number of MFC-MRD-positive cases, no further analyses could be performed.

We next addressed the association of MRD and outcome according to CEBPA mutation type. CEBPA^bzip MRD was detectable in 22 out of 43 patients (51%), whereas CEBPA^non-bzip MRD was present in 20 out of 41 patients (49%). Patient characteristics did not significantly differ between different MRD status within the CEBPA^bzip and CEBPA^non-bzip subgroups (Supporting Information S1: Tables 3 and 4). Persisting CEBPA^bzip in CR did not associate with changes in OS or CIR (Figure 1C,D). In contrast, detectable MRD in CEBPA^non-bzip AML patients showed a nonsignificant trend towards increased CIR and inferior OS (p = 0.12 and p = 0.24; Figure 1E,F). Importantly, an increase in 2-year relapse risk was seen for CEBPA^non-bzip AML patients with detectable MRD (2-year CIR: 59%) compared to patients without (2-year CIR: 26%), indicating that CEBPA^non-bzip MRD is potentially a strong prognostic factor for relapse risk, independent of allogeneic HSCT. In sensitivity analysis, no significant age-related or trial-related interactions were observed. The number of cases did not allow multivariable analyses.

The persistence of co-existing mutations in NPM1 (n = 13) and FLT3 internal tandem duplication (ITD) (n = 10) in CR was also determined using NGS deep sequencing.⁷ Out of 12 mutant NPM1/CEBPA^non-bzip AML patients, five had detectable mutant NPM1 MRD, of which three also had CEBPA^non-bzip MRD. Moreover, three out of five FLT3-ITD/CEBPA^non-bzip AML patients had detectable FLT3-ITD in CR, all in combination with detectable CEBPA^non-bzip MRD. None of the four FLT3-ITD/CEBPA^bzip patients had detectable FLT3-ITD MRD, while CEBPA^bzip persisted in three of these patients. All AML patients with detectable FLT3-ITD MRD and three out of five with detectable mutant NPM1 MRD experienced relapse, possibly explaining the trend in increased CIR seen in patients with CEBPA^non-bzip MRD (Figure 1E). However, larger studies are needed to demonstrate whether persistence of these secondary persisting mutations is associated with increased CIR.

In conclusion, we have studied mutant CEBPA MRD in a representative CEBPA-mutated AML cohort, that is, our data support previous findings that CEBPA^bzip mutations are present in younger AML patients, carry specific co-mutations and confer improved OS in CEBPA-mutated AML.^{10, 11} Here we show in a relatively large cohort of 84 AML patients, that mutant CEBPA MRD is not significantly associated with increased CIR or inferior OS. Importantly, we demonstrate that mutant CEBPA MRD in the ELN2022 favorable subtype of AML carrying CEBPA^bzip mutations does not have impact on outcome. However, AML patients carrying persistent CEBPA^non-bzip mutations in CR have increased CIR and inferior OS. We were unable to demonstrate whether this association was independent of other risk factors. In fact, mutant NPM1 or FLT3-ITD persisting in CR might be better indicators for impending relapse in CEBPA^non-bzip AML, however, similar studies in larger CEBPA-mutated AML cohorts are warranted.

Christian M. Vonk, Melissa Rijken, and Francois G. Kavelaars performed experiments. Christian M. Vonk, Emma L. Boertjes, Roxanne E. Cromwell, Francois G. Kavelaars, Jolinda M. L. Konijnenburg, and Tim Grob analyzed data. Christian M. Vonk, Emma L. Boertjes, and Tim Grob prepared the figures. Christian M. Vonk and Emma L. Boertjes drafted the manuscript, and all authors approved the final version of the manuscript.

The authors declare no conflict of interest.