An Old Dog Has a New Trick: Somatic Exonic Deletions in RUNX1 Are Frequent in AML

In this issue of Clinical Cancer Research, Eriksson and colleagues (1) report a high frequency of somatic exonic RUNX1 deletions in a cohort of sporadic acute myeloid leukemia (AML) samples. In AML, loss-of-function mutations in the gene encoding the transcription factor, RUNX1, portend a poor prognosis. As such, patients with RUNX1 mutation are generally classified as high risk and recommended for allogeneic stem cell transplantation (2). Thus, the accurate detection of RUNX1 alterations is critical for risk stratification and therapeutic decision-making.

Most frequently, RUNX1 mutations are either missense mutations that cluster within the Runt homology domain and disrupt RUNX1-DNA binding or nonsense/frameshift mutations that result in nonsense-mediated decay and the loss of RUNX1 protein (3). In addition to somatic mutations observed in AML, there are germline RUNX1 mutations that cause a hereditary thrombocytopenia called RUNX1 familial platelet disorder with predisposition to myeloid malignancy (RUNX1-FPDMM; refs. 4, 5). While RUNX1-FPDMM is frequently characterized by missense and frameshift mutations similar to those reported in sporadic AML, about 5%–10% of patients with RUNX1-FPDMM harbor large exonic deletions in RUNX1 (6). Here, the authors sought to determine whether similar exonic RUNX1 deletions are also present in de novo AML. Importantly, typical diagnostic sequencing approaches are not adept at identifying large intragenic deletions, thus the authors employed multiplex ligation-dependent probe amplification (MLPA; ref. 7) on well-characterized AML patient samples to better capture deletions.

MLPA involves the annealing of adjacent oligonucleotides and quantitative PCR to detect copy-number changes or genetic deletions and rearrangements. The tiling of MLPA probes can allow interrogation of a focused genomic region for small deletions. For instance, MLPA has been commonly used clinically to define deletions in the α-globin locus that are associated with α-thalassemia (8). In addition, MLPA has identified RUNX1 exonic deletions in several studies of RUNX1-FPDMM families (9, 10). Here, the authors deployed a commercially available MLPA kit with probe sets designed to detect deletions and duplications in genes commonly associated with familial myelodysplastic syndrome and AML. Of relevance to the current study, this included probes detecting RUNX1 exons 1–7, intron 7, exon 8, and exon 9.

By applying the MLPA technology in combination with a standard next-generation sequencing (NGS) myeloid panel to 60 well-characterized AML patient samples, the authors found that 35 patients were wild type for RUNX1, 9 contained classical RUNX1 mutations, and 16 contained large exonic deletions in RUNX1. The presence of large deletions was confirmed by whole-genome sequencing (WGS) in 8 individuals and by microarray in 11 individuals. In addition, while germline material was available for only 9 of 16 patients to exclude the possibility that the reported deletions were germline, the remaining 7 had no family history of thrombocytopenia or AML, suggesting a high likelihood of somatic origin. Remarkably, this suggests that somatic RUNX1 alterations occurred in roughly 40% of the examined cohort. The authors admit that their cohort is a relatively aggressive one and given the established correlation between RUNX1 mutation and poor outcomes, it will be important to see whether such a high frequency of RUNX1 aberrations can be confirmed in studies with larger cohorts.

The RUNX1 mutations defined as “classical” mutations were identified by the standard NGS myeloid panel and had been previously reported in the literature (11). Most of these mutations caused nonsense/frameshift insertions or deletions in the RUNX1-DNA binding domain, which likely lead to nonsense-mediated decay and loss of RUNX1 protein or loss of RUNX1 activity due to compromised DNA binding activity (5). Similarly, RUNX1 exonic deletions centered around exons 3–7, which encode the RUNX1 DNA binding domain, and thus are expected to abrogate RUNX1 function. Interestingly, while deletions in this region are also observed in RUNX1-FPDMM (12), the most frequent germline deletions encompassing exons 1–2 (9) were not observed in this cohort. However, patient numbers remain low, and this could change as larger cohorts are evaluated.

Molecular comparisons between the RUNX1-mutated (n = 9) and the exon-deleted (n = 16) groups revealed a common co-occurrence of RUNX1 deletion and NPM1 mutation that was not observed with classical RUNX1 mutations. In addition, NPM1 mutations are not observed in RUNX1-deleted FPD [(12), bioRxiv 2023.01.17.524290] suggesting that this may be a feature unique to somatic RUNX1 deletions. The European LeukemiaNet (ELN) guidelines for risk stratification dictate that patients with RUNX1 mutation are stratified into an adverse risk category unless they have a co-occurring favorable risk marker (2). As such, all 9 patients with classical RUNX1 mutation had been classified as being adverse ELN risk. In contrast, only 6 RUNX1-deleted patients had been classified as such. Of the remaining 10 RUNX1-deleted patients, 4 had been classified as intermediate risk and 6 had been classified as favorable risk. It is likely that the co-occurrence with NPM1 mutation contributed to the frequent stratification of RUNX1-deleted patients as favorable or intermediate risk based on ELN criteria.

Using this initial stratification, the authors demonstrate that overall survival (OS) did not vary significantly between the intermediate and adverse ELN risk groups. Thus, they reclassified the RUNX1-deleted patients from the intermediate to adverse risk group, reflecting the presence of a RUNX1 aberration. This restratification of patients with an exonic RUNX1 deletion resulted in a significant improvement in the prognostic value of ELN risk stratification, with a significant increase in OS of the intermediate risk group over the newly stratified adverse risk group (Fig. 1). Thus, evaluation of patients with AML for exonic deletions in RUNX1 has significant implications for effective risk stratification. Moreover, these data suggest that like classical RUNX1 mutations that are associated with poor outcomes, the presence of exonic RUNX1 deletions also predicts a poor prognosis.

The authors comment on some key hurdles to the incorporation of diagnostics capable of adequately identifying recurrent exonic deletions in RUNX1. First, while MLPA can identify these deletions, it is not capable of providing information regarding clone size and may not be sensitive enough to identify deletions in small clones. Thus, the authors suggest that long-read, WGS approaches with analysis pipelines specifically adapted for the identification of such deletions may be the best approach moving forward. While such WGS approaches are currently outside of the scope of typical diagnostic workups, whether such a high frequency of RUNX1 aberrations is confirmed in larger cohorts or whether similar exonic deletions are identified in other common tumor suppressor genes, the argument to incorporate such approaches will become more compelling.

No disclosures were reported.

This work was supported by the RUNX1 Research Program and Alex's Lemonade Stand Foundation (K.R. Stengel), the Edward P. Evans Foundation (K.R. Stengel), and an NIH grant (R35 GM147213; K.R. Stengel).