Description of a novel subtype of acute myeloid leukemia defined by recurrent CBFB insertions

TO THE EDITOR:

Molecular analysis of pediatric and adult acute myeloid leukemia (AML) is used routinely to identify subtype-defining driver structural variants and mutations which may provide important information for risk stratification. For example, the core-binding factor (CBF) AML subgroup defined by t(8;21) RUNX1::RUNX1T1 or inv(16)/t(16;16) CBFB::MYH11 are associated with favorable outcomes.^1,2 Despite the extensive genomic characterization of pediatric and adult AML, there remains an important proportion of previously unclassified cases where new driver lesions are still being identified, including those that can influence patient management owing to their association with outcomes. This includes the recent identification of tandem duplications in UBTF in pediatric AML^3,4 and structural variants that dysregulate BCL11B in lineage-ambiguous acute leukemia.⁵ Herein we describe an additional new subtype of AML characterized by a recurrent insertion mutation in CBFB.

We initially reanalyzed a cohort of 553 pediatric AML transcriptomes from our previous study³ and identified 2 patients (PARANT:SJAML040573, PARUTH:SJAML040605) with similar gene expression profiles to CBFB::MYH11 AML, but without a CBFB::MYH11 fusion or other known leukemic driver by conventional testing, and without finding a CBFB::MYH11 fusion by manual inspection of RNA sequence data (supplemental Figure 1, available on the Blood website; Table 1). However, in both of these patients, we identified a somatic 9-base pair insertion in exon 3 of CBFB (NM_022845.3). The CBFβ protein encoded by CBFB forms the non–DNA-binding regulatory subunit of a heterodimeric transcription factor complex with a DNA-binding CBFα subunit (RUNX1, RUNX2, or RUNX3). Interestingly, both CBFB mutations were predicted to lead to the same amino acid change, substituting aspartic acid at position 87 (D87) for glycine, aspartic acid, serine, and tyrosine [p.(Asp87delinsGlyAspSerTyr); GDSY] within the N-terminal RUNX-binding domain⁶ (Figure 1A).

Given these findings, we hypothesized that CBFB mutations may be recurrent in AML and a defining feature of a novel subtype. Through a combination of published data, clinical sequencing, and screening driver-negative AML cohorts from independent sources (supplemental Methods), we identified an additional 16 cases with CBFB insertions involving D87, including 15 AML and 1 B/myeloid mixed phenotype acute leukemia, for a total of 18 cases (Table 1). These additional cases also lacked a CBFB::MYH11 fusion or other known leukemic driver alterations (supplemental Table 1). CBFB mutations were confirmed in both DNA and RNA sequencing when available, or Sanger sequencing, and were confirmed to be somatic in 11 of 11 cases where matched germline data was available (supplemental Table 1; supplemental Methods; supplemental Figure 2). Remarkably, we identified 10 different nucleotide insertions at codon 87 in these 18 cases; 9 out of 10 were predicted to encode for the same in-frame GDSY amino acid change (p.(Asp87delinsGlyAspSerTyr)), with the other nucleotide insertion leading to a GDTY amino acid change (p.(Asp87delinsGlyAspThrTyr)) (Figure 1A). This highly stereotyped change at the protein level (ie, GDXY) strongly implies a functional relevance.

We next integrated 8 of these additional cases into our transcriptome cohort and observed a tight cluster of cases with CBFB insertions adjacent to the CBFB::MYH11 cluster (Figure 1B). Gene set enrichment analysis confirmed broad similarities between AML with CBFB insertions and CBF AMLs (Figure 1C; supplemental Figure 3; supplemental Table 2). However, CBFB insertion cases showed uniquely high expression of BCL2L14, MEIS1, and HOXA cluster genes, demonstrating a more stem-like signature compared with CBFB::MYH11 AML.

We also examined the cooperating mutations in 10 CBFB insertion cases from RNA sequencing data and integrated these findings with mutational information from other studies (Figure 1D; supplemental Tables 1 and 3). Recurrent mutations were detected in BCORL1 (7/18 [39%]), FLT3 (7/18 [39%]), NRAS (6/18 [33%]), ETV6 (5/18 [28%]), KDM6A (4/18 [22%]), and NF1 (2/18 [11%]). FLT3 tyrosine kinase domain (TKD) mutations were most common, although internal tandem duplications and mutations outside the TKD were also observed. Overall, these mutations are different from the mutational spectrum previously reported for CBF leukemias,⁷ most notably the absence of KIT mutations and a higher frequency of FLT3-TKD and BCORL1 mutations. Recurrent chromosomal alterations in the CBFB insertion group included trisomy 6 and trisomy 22, whereas trisomy 8 was not observed (Figure 1D; supplemental Table 1). Additionally, the CBFB mutation was conserved at both diagnosis and relapse in 2 cases profiled at both time points (AML075 and 115225). Collectively, these data suggest that CBFB insertions are a subtype-defining lesion, and we have provisionally termed this group CBFB-GDXY.

Like CBF AML, AMLs with CBFB-GDXY mutations were observed in both children and adults, but were enriched in adolescent and young adult age groups (median age, 16.5 years; range, 9-27 years). Overall, this mutation in AML cohorts was rare, including 3 of 188 in the TARGET pediatric AML cohort, 5 of 1048 in the pediatric AAML1031 cohort and 1 of 350 in the Clinseq-AML Swedish adult cohort, whereas more than 2000 cases from multiple large cohorts composed primarily of adult AMLs did not harbor a CBFB insertion (supplemental Data for a description of cohorts screened). Additionally, unlike the typical myelomonocytic morphology with abnormal eosinophils (FAB M4 Eo) observed for CBFB::MYH11 AML, CBFB-GDXY AML had fewer mature morphologies (FAB M0, n = 1; FAB M1, n = 4; or FAB M2, n = 4), consistent with the stem-related expression profiles, where morphologic reports were available. However, an increase in eosinophils was still observed (Figure 1E). Like RUNX1::RUNX1T1 AML,⁸ we noted that CBFB-GDXY AML may express CD19 (supplemental Figure 4; supplemental Table 1). Further supporting this observation is the identification of the GDXY insertion in 1 case of B/myeloid mixed phenotype acute leukemia.

This cohort is small and collected from different sources with varied treatment protocols, precluding a definitive assessment of the impact of this mutation on outcomes. However, 8 of 17 patients (where data were available) had either relapsed or refractory disease after initial treatment, whereas patients with RUNX1::RUNX1T1 or CBFB::MYH11 AMLs commonly have a good outcome and these AMLs are considered favorable risks. To investigate the treatment response in one patient, we designed a CBFB mutation-specific ultradeep next-generation sequencing assay (supplemental Methods) for longitudinal tracking of measurable residual disease. This 25-year-old patient (SBJ00860) was treated with cytarabine and idarubicin induction (7 + 3), followed by high-dose cytarabine consolidation. Measurable residual disease assessment after each cycle of chemotherapy showed detectable but decreasing CBFB insertion variant allele frequency, becoming undetectable after the last cycle of therapy and remaining undetectable at 9 months of follow-up (Figure 1F).

In summary, we have reported a novel subtype of AML characterized by recurrent in-frame insertion mutations in CBFB, leading to a GDXY amino acid sequence change at position D87. Molecular characterization demonstrated transcriptional similarity to CBF AML, while also highlighting an enrichment of FLT3-TKD mutations, lack of KIT mutations, and stemness-related gene expression signature. Recognition of this subtype and further study in clinical trials, as well as investigation of the underlying leukemogenic mechanism of the CBFB insertion, will be important to understand the full clinical relevance of this novel entity.