RUNX1 germline variants in RUNX1-mutant AML: how frequent?

TO THE EDITOR:

RUNX1 mutations are recurrent aberrations in acute myeloid leukemia (AML) that are either somatically acquired or originate in the germline. Monoallelic pathogenic germline RUNX1 variants cause familial platelet disorder with propensity to AML (FPD/AML), featuring predisposition to develop myeloid malignancies. Recently, Simon et al¹  found that 30% of RUNX1-mutated AML patients carried a (nonpolymorphic) RUNX1 germline variant. This high frequency contrasted previously reported frequencies in AML cohort studies.^2-5 

To obtain further insight into the frequency of germline variants within RUNX1-mutated AML patients, we retrieved data from a large independent cohort of AML patients that was previously interrogated by paired genetic sequencing at diagnosis and in complete remission (CR).⁶  Targeted next-generation sequencing (NGS) at diagnosis was performed in 763 AML patients, enrolled in the Dutch-Belgian Hemato-Oncology Cooperative Group (HOVON)102 clinical trial (2010-2013),⁷  using the TruSight Myeloid Sequencing Panel (Illumina). The HOVON102 trial was designed to investigate the added value of clofarabine in combination with standard remission-induction chemotherapy in adult (18-65 years) AML or myelodysplastic syndrome patients. NGS data were analyzed as previously described,⁶  and nonpolymorphic RUNX1 variants were annotated to RefSeq ID NM_001754.5.

At diagnosis, at least 1 RUNX1 mutation was detected in 115 of 763 AML patients (15.1%) for a total of 142 RUNX1 mutations (data not shown), corresponding to frequencies of 10% to 15% reported in literature.^3,4,8,9  We detected 2 or more mutations in 23 of 115 patients.

To distinguish germline variants from somatic mutations, we retrieved data from paired diagnostic and remission HOVON102 AML samples to identify persisting RUNX1 mutations. Remission samples of 287 CR patients were available for NGS analysis.⁶  Among these 287 AML patients, 48 RUNX1 mutations were detected in 37 patients (12.9%), of whom 9 carried 2 or more mutations (supplemental Tables 1 and 2, available on the Blood Web site). The distribution of the variant allele frequencies (VAF) at diagnosis of the RUNX1 mutations of the selected samples (n = 48 in 37/287 patients) was similar to the initial HOVON102 cohort (n = 142 in 115/763 patients) (Figure 1A). In the majority of cases (34/37), RUNX1 mutations were acquired because the VAF of these mutations (n = 45) was ≤10% in CR (Figure 1A). However, in 3 of the 37 AML cases (8.1%), a RUNX1 mutation was present at a VAF of 50% in both diagnostic and remission samples (Figure 1A), highly indicative for germline origin. Other cooccurring mutations present at diagnosis in these 3 AML patients were either cleared or persisted at much lower VAF in CR (Figure 1B-D), supporting the germline status of these RUNX1 variants. Thus, in the current cohort of 287 AML patients that attained CR, 3 (1.0%) harbored a germline RUNX1 variant.

We identified 2 different RUNX1 germline variants in these 3 AML patients. A single patient (no. 748) carried RUNX1 p.(Arg232Trp) (exon 7), a missense variant located between the runt homology domain and the transactivation domain (Figure 1E). The 2 other patients (no. 124 and no. 740) shared variant RUNX1 c.97+1G>A, affecting the splice donor site of exon 3 (Figure 1E). A familial bond between these 2 patients could not be confirmed.

In accordance to recently published guidelines by the ClinGen Myeloid Malignancy Variant Curation Expert Panel,¹⁰  we classified both identified RUNX1 germline variants as variants of unknown significance (supplemental Table 3). Neither of these have been reported as RUNX1 germline variants before in the context of FPD/AML traits,¹¹  although RUNX1 p.(Arg232Trp) has been recurrently reported as somatically acquired mutation in AML.^1,3,4  Both variants occur outside of the runt homology domain, in which most missense RUNX1 germline variants cluster.¹¹ RUNX1 c.97+1G>A is presumed to specifically affect isoform RUNX1C, 1 of 3 major RUNX1 isoforms, complicating its interpretation.¹⁰  However, deletions of exon 2 and 3, also putatively affecting RUNX1C specifically, have been reported in the context of FPD/AML before.¹² 

Additional somatic mutations were found in all 3 patients, most of which have been reported in the context of germline RUNX1-mutant AML before, including additional RUNX1 mutations and concomitant mutations in DNMT3A, FLT3, and GATA2.^1,11,13-15  Secondary mutations in RUNX1 are a frequent recurrent event in FPD/AML-related AML,¹¹  and cooccurred in patient no. 748 carrying the germline variant RUNX1 p.(Arg232Trp). Besides somatic mutations in GATA2, which were also reported to be overrepresented in germline RUNX1 AML,¹¹  none of the mutated genes were specific to the germline RUNX1 patients (supplemental Table 4). We revealed a somatic mutation in ETV6, a gene that has not been previously reported in AML with RUNX1 germline variants, in one of our cases (no. 740).

The clinical characteristics and course of the 3 patients harboring germline RUNX1 variants is summarized in Table 1. All 3 patients were transplanted. Patient no. 124 had a therapy-related AML with a t(9;11) cytogenetic abnormality after breast cancer therapy. She was transplanted with hematopoietic cells from a sibling donor, relapsed 14 months after achieving CR, and died 4 months later. The origin of the relapse (donor/host) is not known. The other 2 patients achieved long-term survival without relapse after HSCT (of an unrelated donor [no. 748] and cord blood [no. 740]). Interestingly, at the time of study, a familial platelet disorder and a presumptive diagnosis of FPD/AML had been recognized in only 1 of 3 patients (no. 748).

Taken together, data from this independent cohort reveal an 8.1% frequency of germline variants in RUNX1-mutated AML. This frequency is lower than that reported by Simon et al (30%),¹  and consistent with previously reported frequencies in RUNX1-mutated AML in smaller series as assessed by DNA sequencing in buccal DNA at diagnosis or peripheral blood DNA in CR (8.6%, 10%, and 9.5%).^3-5  Schnittger et al²  did not detect any germline RUNX1 variants in remission samples of 60 noncomplex karyotype AMLs.

The apparent discrepancies in germline RUNX1 mutation frequencies in RUNX1-mutant AML between the previously reported (0%-10%) as well as ours (8.1%) and reported by Simon et al (30%) may reflect population differences, selection biases, technical differences in detecting genetic variants, as well as the inherent small group sizes.

Differences in techniques used to detect RUNX1 mutations may have contributed to the differences between our results and those reported by Simon et al.¹  RNA sequencing, as applied by Simon et al, is dependent on gene expression and detection of RUNX1 mutations with lower VAFs may therefore be more challenging by NGS-based RNA sequencing than DNA sequencing. This may possibly explain the higher proportion of patients with RUNX1 mutations at VAF <30% in our cohorts (26/115, 22.6% in the entire HOVON102 cohort; 8/37, 21.6% in the CR subcohort) vs 4.3% in Simon et al.¹  Differences in sensitivity for detection of these low VAF cases could potentially lead to an overestimation of the frequencies of RUNX1 germline variants in RUNX1-mutant AMLs. RNA sequencing may further fail to detect mutations introducing premature stop codons leading to nonsense mediated decay of transcripts, which would result in underestimation of the total number of RUNX1 mutations. Because this could potentially affect the detection of both germline and somatic variants, the overall effect on germline frequency is uncertain.

We cannot formally exclude the possibility that the frequency of germline variants in our study may be biased by a potentially differential response to induction therapy of patients harboring germline RUNX1 mutations. In the complete NGS-analyzed HOVON102 cohort of 763 AML patients, 668 (88%) achieved CR after 2 cycles of high-intensity induction chemotherapy (data not shown). RUNX1-mutated patients achieved CR less frequently (91/115, 79.1%) compared with RUNX1 wild-type patients (577/648, 89%), with a relative risk of refractory disease of 1.90 (95% confidence interval, 1.25-2.89), which is in line with the proposed relative chemoresistance of RUNX1-mutated AML.^3,4,8,9,16  No difference, however, was found in CR rates between patients carrying a RUNX1 mutation at a VAF ≥30% (70/89, 78.7%) and patients carrying RUNX1 mutations at a VAF <30% (21/26, 80.8%) at diagnosis, with a relative risk of refractory disease of 1.11 (95% confidence interval, 0.46-2.68), perhaps arguing against a potential bias introduced by limiting analyses to patients in CR.

On a final note, it is worth considering that not all reported variants are necessarily disease-causing mutations. Therefore, the frequency of RUNX1 germline variants reported in literature may not directly translate to the frequency of FPD/AML. Moreover, even if the latest classifying algorithms are applied, as exemplified in Simon et al and in the current analysis, pathogenicity of certain variants remains uncertain. Efforts to refine classifying algorithms by initiatives such as ClinGen, as well as continued reporting and curating of variants in databases like ClinVar and RUNX1db, are vital to improve the collective knowledge of variants and thus the ability to more accurately diagnose FPD/AML.

We conclude that the frequency of germline RUNX1 variants in RUNX1-mutated AML may be lower than recently reported, depending on multiple variables including cohort variation and RUNX1 detection methods. Future studies in larger cohorts are required to definitively establish this frequency. This does not negate the absolute necessity of germline testing in RUNX1-mutated AML (with high VAFs), as stressed by the clinical cases described herein.