Clinical and molecular correlates of somatic and germline DDX41 variants in patients and families with myeloid neoplasms (2024)

Associated Data

Supplementary Materials

Abstract

The diagnosis of germline predisposition to myeloid neoplasms (MN) secondary to DDX41 variants is currently hindered by the long latency period, variable family histories and the frequent occurrence of DDX41 variants of uncertain significance (VUS). We reviewed 4,524 consecutive patients who underwent targeted sequencing for suspected or known MN and analyzed the clinical impact and relevance of DDX41_VUS in comparison to DDX41_path variants. Among 107 patients (44 [0.9%] DDX41_path and 63 DDX41_VUS [1.4%; 11 patients with both DDX41_path and DDX41_VUS]), we identified 17 unique DDX41_path and 45 DDX41_VUS variants: 24 (23%) and 77 (72%) patients had proven and presumed germline DDX41 variants, respectively. The median age was similar between DDX41_path and DDX41_VUS (66 vs. 62 years; P=0.41). The median variant allele frequency (VAF) (47% vs. 48%; P=0.62), frequency of somatic myeloid co-mutations (34% vs 25%; P= 0.28), cytogenetic abnormalities (16% vs. 12%; P=>0.99) and family history of hematological malignancies (20% vs. 33%; P=0.59) were comparable between the two groups. Time to treatment in months (1.53 vs. 0.3; P=0.16) and proportion of patients progressing to acute myeloid leukemia (14% vs. 11%; P=0.68), were similar. The median overall survival in patients with high-risk myelodysplastic syndrome/acute myloid leukemia was 63.4 and 55.7 months in the context of DDX41_path and DDX41_VUS, respectively (P=0.93). Comparable molecular profiles and clinical outcomes among DDX41_path and DDX41_VUS patients highlights the need for a comprehensive DDX41 variant interrogation/classification system, to improve surveillance and management strategies in patients and families with germline DDX41 predisposition syndromes.

Introduction

Germline pathogenic variants that predispose to familial cancers have been reported in several genes.¹ In 2016 the World Health Organization recognized hematological malignancies associated with germline predisposition syndrome as a distinct sub-group with prognostic implications.² DEAD/H-box helicase 41 gene (DDX41) is located on chromosome 5q35, is thought to be a tumor suppressor gene involved in the splicing of pre-mRNA and processing of ribosomal RNA. In experimental knockouts models, defects in DDX41 have been shown to contribute to the development of myeloid neoplasms (MN).^3-6 Unlike traditional germline predisposition syndromes, DDX41-associated germline predisposition syndrome has a late age of onset, commonly presents with myelodysplastic syndromes (MDS) or acute myeloid leukemia (AML), and is associated with variable family histories of MN.⁷ The pro-leukemogenic properties of DDX41 pathogenic variants is also supported by lack of concurrent cytogenetic or molecular abnormalities that are usually seen in MDS/AML.⁸ Several studies including ours have demonstrated favorable prognostic outcomes associated with DDX41 mutations in MDS/AML.^8-10 In spite of readily available genomic sequencing to assist with the diagnosis of MN, the diagnosis of DDX41 mutated MN is hindered by a long latency, variable family history and the frequent occurrences of variants of uncertain significance (VUS), given inherent difficulties to characterize the DDX41 gene/protein function. The accurate recognition of germline and somatic DDX41 variants is vital for prognostication and the management of DDX41 variant-associated MDS and AML, including allogeneic stem cell transplant donor selection. In this large study we report the mutational profile of 107 patients with DDX41 variants (pathogenic and VUS) and their clinical outcomes. We also discuss challenges in interrogating and performing functional analysis for DDX41 VUS.

Methods

We reviewed 4,524 consecutive patients who underwent next-generation sequencing (NGS; MC OncoHeme 42-gene panel) assessments between January 2018 and May 2022 for suspected or known myeloid disorders. The study was conducted at the Mayo Clinic and approved by the Mayo Clinic Institutional Review Board (IRB) (Figure 1). The clinical information was stored in a de-identified database. Germline testing in patients and families with suspected or known germline predisposition syndrome were conducted under IRB approved protocol (Mayo Clinic IRB# 16-004173; clinical-trials gov. Identifier: NCT02958462). We identified 44 (0.9%) patients with DDX41 likely pathogenic/pathogenic (DDX41_path) variants and 63 (1.3%) patients with DDX41 VUS (DDX41_VUS) according to variant classification criteria from the American

College of Medical Genetics/ the Association of Medical Genetics/ the Association for Molecular Pathology (ACMG/AMP) guidelines.¹¹ We evaluated baseline characteristics, mutational profiles, and clinical outcomes of patients with DDX41_path and DDX41_VUS variants (Figure 1).

For the purpose of this paper, we operationally defined DDX41 variants with a variant allele frequency (VAF) of ≥40% to be presumed germline, as previously reported.^10,12DDX41 variants with positive germline testing on DNA derived from skin fibroblasts or hair follicles were defined as confirmed germline. Patients with DDX41 variants who had negative germline testing defined as confirmed somatic. In the absence of direct germline testing, variants with VAF <40% were defined as presumed somatic (Figure 1). We defined clonal cytopenia(s) of undermined significance (CCUS) as unexplained cytopenia(s) associated with known somatic pathogenic variants (VAF ≥2%), with bone marrow dysplasia <10% and bone marrow blast% <5%. Patients having cytopenia(s) with an isolated DDX41_VUS, without any somatic pathogenic variants, were classified as “cytopenia associated with DDX41_VUS alone”, as the pathogenicity of DDX41_VUS remain to be elucidated.^13,14 Responses to therapy were assessed according to the International Working Group (IWG) MDS response criteria.¹⁵

Clinical NGS testing was performed on DNA extracted from fresh bone marrow aspirates (102/110 [93%]) or peripheral blood samples (8/110 [7%]). The Mayo Clinic NGS panel included 42 genes (Online Supplementary Appendix) and has an accuracy of >99% and reproducibility of 100% for single-base substitutions and insertion/deletion events. The panel has a variant sensitivity of >5% VAF with a minimum depth coverage of 250x. All coding regions (exons 1-17) of DDX41 were covered by the panel. We manually reviewed the sequencing BAM files among patients with commonly occurring DDX41 pathogenic variants (M1I and D140fs) for the presence of DDX41 R525H somatic variants with low VAF (2-5%, below the report limit of the clinical laboratory). Germline testing was performed prospectively in our germline predisposition clinic, on DNA derived from skin fibroblasts or hair follicles as previously described.¹⁶ We also performed in depth curation assessments of frequently occurring DDX41_VUS to analyze allelic diversity and pathogenicity. In order to make a pathogenic prediction, based on available literature, we considered in silico CADD (combined annotation dependent depletion) and REVEL (rare exome variant ensemble learner) scores >25 and >0.8 as being likely pathogenic, respectively.^17,18

Results

Among 107 patients with DDX41 variants, we identified 17 unique DDX41_path variants and 45 unique DDX41_VUS (Figure 1). Eleven patients had both DDX41_path variants and DDX41_VUS. Twelve patients had two DDX41_path variants. Among 24 (22%) DDX41-mutated patients that had germline testing, 13 (54%) had DDX41_path, and 11 (46%) patients had DDX41_VUS, respectively (Online Supplementary Table S1;Figure 2). The previously reported DDX41 pathogenic variants and VUS by our group are marked with (*) in the Online Supplementary Table S1. In addition to previously described germline and somatic variants in MN,^10,12,19,20 novel germline variants identified in the current study (i.e., not previously reported in the literature or existing genetic databases) are summarized in the Online Supplementary Table S2.

Discussion

In this large cohort of patients who underwent NGS for a known or suspected myeloid neoplasm, we identified 17 (16%) unique DDX41_path variants and 45 (42%) unique DDX41_VUS. Majority of the DDX41 variants occurred in isolation (n=76/107 [71%]) without any additional somatic variant or clonal cytogenetic abnormality. Our observations validate previous reports that germline DDX41 associated MN have a later age of onset (median age 65 years), are male predominant (61% males), usually present with indolent cytopenias (27% in DDX41_path and 52% in DDX41_VUS have not yet needed treatment thus far), with variable family histories of MN (26.5%).^{7,10,12,21,22} Importantly, we did not find significant differences in clinical and demographic factors between patients with DDX41_path variants and DDX41_VUS.

We also describe the occurrence of MPN in patients with DDX41 variants. Recently, in a multicenter retrospective analysis by Li et al.¹² from a cohort of 176 patients (DDX41_path[n=116], DDX41_VUS [n=60]), 15 patients with MPN harboring DDX41 variants were reported (11 with DDX41_VUS), with 34% of these variants being somatic in nature. JAK2 and CALR mutations were not detected in patients with DDX41_path and were reported in 72% of patients with DDX41_VUS. We describe four patients with MPN, including two with JAK2 V617F mutant PV, and one each with triple-negative essential thrombocythemia and JAK2 V617F mutant primary myelofibrosis. In this cohort, three (75%) patients had presumed germline DDX41_path (c.337del; p. E113Kfs*14, c.465G>A; p.M155l and c.298+5G>A; p.?) and the remainder had presumed germline DDX41_VUS (c.138+5G>T p?). We are not able to comment on the prognostic impact of DDX41 mutations in MPN patients given the small sample size.

As previously reported, we did not find a significant impact of BM blast percentage (10-19% vs. ≥20%) on MDS/AML outcomes, in patients with DDX41 variants, germline or somatic.^3,12,23 In contrast, DDX41-mutated MN with normal karyotypes had a trend towards a better OS compared to those with abnormal karyotypes, with the most common abnormal karyotype being del 20 (q11.2q13.1) (3/11 [27%]). In addition, survival outcomes for patients with DDX41-mutated MDS/AML were relatively favorable from historical cohort of high-risk MDS/AML and not significantly different between the DDX41_path versus DDX41_VUS groups. In a recent report by Makishima et al. on 346 patients with pathogenic/likely pathogenic germline DDX41 variants, MDS patients with truncating DDX41 variants had rapid progressions to AML in comparison to those with non-truncating variants, without any significant impact on survival.²⁴ In our cohort, 25% (n=27) of patients had truncating DDX41 variants, among them three of 27 were DDX41_VUS (c.138+5G>T; p?, c.1732+4 A>G; p?, c.28-3C>T; p?). One DDX41_VUS was secondary to an inframe deletion (c.992_994del; p.Lys331del), making the adjudication as to whether or not this variant resulted in truncation of the protein challenging, hence it was excluded. We did not find significant differences in the rate of progression to AML between truncating and non-truncating DDX41 variants. The median OS was also not significantly different between DDX41 truncating and non-truncating variants. Further studies are needed to determine the clinical impact of truncating DDX41 variants in MDS and AML patients.

Variant curation of DDX41 is challenging since many aspects of its protein function are incompletely understood and for which functional assays or strong heritability links are not yet described. Our molecular hematology laboratory uses standard ACMG criteria, with the help of genetic counselors for evaluating DDX41 germline variants. We tend to be conservative with our curation approach, to not assign potential risk allele and disease causality without sufficient evidence. Thus, several variants classified as VUS in this manuscript have been reported in ClinVar with different and sometimes conflicting ACMG classification.¹¹ We are cautious about using ClinVar entry assertions when case counts are very limited. There are many inaccuracies in ClinVar and many entries with assertions provide little to no data for a designation of likely pathogenic variants. An example of this is the p.R339 site, which is now increasingly recognized as a recurrent event and a likely disease predisposing germline variant. It is also evident that these alterations have associations in SNP databases (e.g., gnomAD) with variable frequency among ethnic populations (overall frequency 0.003%). We feel this is a prudent approach given that data continues to emerge on the true effects of different DDX41 variants in relation to disease outcomes and currently there remains a large gap in our understanding due to a lack of functional data. Li et al.¹² recently described a valuable DDX41-specific classification approach but included assumptions on certain ACMG criteria to enable a more dichotomous classification as causal/likely causal, versus uncertain. Specifically PM2 was applied in the situation that the polymorphic association of a new possible germline causal variant was less frequent in the “general population” than the most common two DDX41 germline variants (c.465G>A; p.M1I and c.415_418dup; p.D140fs). While the basis for this approach is reasonable, it may inadvertently segregate rare polymorphic variants of limited or no significance with potential disease-associated risk variants. We also apply this criterion (PM2) in our process, but not at its full value. Similarly, this group applied PM3 (typically used in recessive disorder assessment) in a modified fashion when considering occurrence of DDX41 somatic genetic variants. While these maneuvers are certainly logical, they are not necessarily definitive. Another recent paper by Duployez et al.²⁵ applied ACMG/AMP criteria without apparent modification, although they did consider accompanying DDX41 somatic variants as strong evidence for a germline finding to be causal. Both papers identify many potential germline variants in patients with hematologic malignancies. In our cohort we find overlapping alterations, specifically with c.773C>T; p.P258L, c.1016G>A; p.R339H and c.992_994del; p.K331del variants, that have been described in several publications with a prevalence now exceeding that in ‘control’ patients with hematological malignancies. While the accumulating data supports the association of these rare alleles with disease risk, the standard ACMG classification would still render these as VUS calls. It is notable that in a more recent large international study Makishima et al.,²⁴ the c.992_994del; p.K331del variant was not identified in our reading of the paper. While this may reflect an effect of different ethnic group distribution, this finding also supports a conservative approach to curation, along with appropriate interpretive commentary in our reporting.

The pathogenicity of germline VUS associated with adult MN is difficult to determine through analysis of patient observational data alone, since these diseases often present with a complex array of co-mutations and cytogenetic abnormalities that may be epistatic to the effects of the DDX41_VUS. Thus, experimental analysis of the effect of these variants on gene function is necessary, in conjunction with analysis of available patient data to confirm pathogenicity. Secondly, the effect of VUS on the tumor suppressive activities of DDX41 likely depends on the effect of each variant on the structure and function of the DDX41 protein. Most VUS are missense or cause deletion of a single amino acid and, thus their effect on tumor suppression activity is likely dependent on the role of the effected amino acid in the protein structure and function. The challenge in determining the effect of a variant on tumor suppression is that DDX41 has multiple functions and variants may not affect all functions equally. Experimental testing of the effect of each VUS on all known functions of DDX41 is required to resolve this question. However, this experimental analysis can be tedious because each variant must be analyzed separately. DDX41 is particularly difficult to examine experimentally since it is an essential gene, causing DDX41-knockout cell lines to grow inefficiently in culture and making it difficult to engineer cells where the VUS can be studied in isolation from wild-type DDX41. Furthermore, the precise function(s) of DDX41 that are responsible for its role as a tumor suppressor remain incompletely defined, adding additional complication to functional analysis of variants. Importantly, DDX41 is an essential gene for hematopoiesis and potentially other physiological processes relying on cell proliferation and it is unlikely that inhibition of DDX41 would provide clinical benefit as a therapeutic approach. However, since DDX41 mutant MDS/AML is known to have favorable treatment outcomes and slower progression rates than other adult MDS/AML subtypes, understanding the effect of each DDX41 variant on protein function would allow for further classification and risk stratification with individualized medicine approaches.⁵

In summary, we provide a comprehensive genomic landscape, including germline and somatic pathogenic variants and VUS in patients with DDX41 variant-associated hematological disorders. We report on the likely pathogenicity of several unique DDX41_VUS and provide a detailed analysis on their clinical course and outcomes. We show that patients with DDX41_path and DDX41_VUS had similar clinical characteristics and clinical outcomes, underscoring the need for better variant interrogation and classification methods.

Supplementary Material

Funding Statement

Funding: The study was funded by the The Henry J Predolin Leukemia foundation. We thank the Mayo Clinic Cancer Center Support Group (P30 CA015083) and the Mayo Clinic Center for Individualized Medicine for sponsoring the germline predisposition clinic.

References