The spectrum of genetic mutations in myelodysplastic syndrome: Should we update prognostication?

Abstract The natural history of patients with myelodysplastic syndrome (MDS) is dependent upon the presence and magnitude of diverse genetic and molecular aberrations. The International Prognostic Scoring System (IPSS) and revised IPSS (IPSS‐R) are the most widely used classification and prognostic systems; however, somatic mutations are not currently incorporated into these systems, despite evidence of their independent impact on prognosis. Our manuscript reviews prognostic information for TP53, EZH2, DNMT3A, ASXL1, RUNX1, SRSF2, CBL, IDH 1/2, TET2, BCOR, ETV6, GATA2, U2AF1, ZRSR2, RAS, STAG2, and SF3B1. Mutations in TP53, EZH2, ASXL1, DNMT3A, RUNX1, SRSF2, and CBL have extensive evidence for their negative impact on survival, whereas SF3B1 is the lone mutation carrying a favorable prognosis. We use the existing literature to propose the incorporation of somatic mutations into the IPSS‐R. More data are needed to define the broad spectrum of other genetic lesions, as well as the impact of variant allele frequencies, class of mutation, and impact of multiple interactive genomic lesions. We postulate that the incorporation of these data into MDS prognostication systems will not only enhance our therapeutic decision making but lead to targeted treatment in an attempt to improve outcomes in this formidable disease.


INTRODUCTION
Myelodysplastic syndrome (MDS) is a myeloid neoplasm that occurs predominantly but not exclusively in older adults, with an overall incidence of approximately one to five cases per 100,000, increasing to 20-75 cases per 100,000 in patients aged 65 and above with a median age of 76 [1,2]. The World Health Organization defines the syndrome based on cytopenias, dysplasia, and certain chromosomal abnormalities [3]. The development of MDS is thought to involve cumulative genetic changes within the hematopoietic stem cell, which can occur over months to years. The natural history of patients with MDS is dependent on the presence and magnitude of diverse genetic and molecular aberrations. The International Prognostic Scoring System F I G U R E 1 Prognostic impact of mutations in myelodysplastic syndrome (MDS) have independent prognostic significance ( Figure 1, Table 1). The addition of such molecular information would help to guide discussions of prognosis with patients and aid in determining treatment plans. Moreover, the identification of formative genetic lesions may help to design therapeutic combinations aimed at improving outcomes.

TP53 mutation
TP53 tumor suppressor gene provides instruction for creation of tumor protein p53, which acts as a tumor suppressor by mediating cell cycle arrest in response to a variety of cellular stressors [6]. TP53 mutations at the time of diagnosis are found in 6%-21% of MDS patients and are more frequently associated with complex chromosomal abnormalities, and prior exposure to alkylating agents or radiation (therapyrelated MDS) [7][8][9].
Equally important, there appears to be significant heterogeneity within TP53 mutations with respect to variant allele frequency (VAF) and number and class of TP53 mutation affecting survival. Higher VAFs (20%-50% and >50%) were associated with a worse prognosis, whereas VAF < 10% did not have a statistically different survival when compared to TP53 wt without complex cytogenetics [32]. Moreover, TP53 aberrations can present as a mutation, deletion, or copy-neutral loss of heterozygosity (cn-LOH). Patients whose MDS cells harbor a combination of mutation and deletion/cn-LOH constitute so-called biallelic or multi-hit TP53 disease. Interestingly, patients with biallelic TP53 lesions or losses have more severe cytopenias, higher bone marrow blast percentages, and shortened median OS (8.7 months vs. 42 months) relative to monoallelic TP53 mutation without deletion/cn-LOH in the other allele [33].

EZH2 mutation
Enhancer of zeste homolog 2 (EZH2) is a histone methyltransferase, forming part of a protein called polycomb repressive complex-2, and is involved in cell fate determination of immature forms [34]. These TA B L E 1 Prognostic information for driver mutations in myelodysplastic syndrome

DNMT3A mutation
This gene encodes DNA methyltransferase-3-alpha, which is involved in DNA methylation, promoting hematopoietic stem cell differentiation into progenitor cells [38]. DNMT3A mutations are among the most common, found in 8%-18% of cases. Patients are older (

ASXL1 mutation
An epigenetic regulator, additional sex-comb like-1′ gene, is frequently overexpressed in myeloid malignancies. This overexpression in mouse models impairs myeloid differentiation and induces MDS [46].

IDH mutations
There are two isocitrate dehydrogenase genes, IDH1 and IDH2, both of which produce enzymes that convert isocitrate to 2-ketoglutarate to produce cellular energy. Mutations in these genes are common in several cancers, resulting in overproduction of (R)-enantiomer of 2hydroxyglutarate, which is thought to promote leukemogenesis [51].
In general, these are uncommon mutations in patients with MDS; IDH2 mutations occur in 2.1%-4.0% compared with IDH1 mutations in 0.6%-3.6% of patients.
It is unclear how this mutation affects patient outcomes. IDH1 mutations have been linked with inferior LFS and OS when controlling for factors such as karyotype, transfusion dependence, and IPSS [52][53][54].
While IDH2 mut compared with IDH2 wt confers inferior survival and increased probability of relapse following HSCT, poor survival and increased rate of relapse associations for IDH2 mut remained statistically significant in only one of five datasets [15,18,20,22,55].

TET2 mutation
The Ten-Eleven translocation gene is a probable tumor suppressor gene altered in hematologic malignancies and may have key functions in normal hematopoiesis and hematopoietic stem cell biology [58].
While TET2 mutations occur relatively frequently in MDS, detected in 11%-33% of cases, the impact of these mutations on prognosis is unclear. TET2 mutations have been independently associated with a shortened survival following allogeneic HSCT [19]. In contrast, TET2

RUNX1
This gene produces runt-related transcription factor 1 that interacts with core-binding factor beta and binds DNA to prevent it from being degraded. RUNX1 protein activates genes to control hematopoietic stem cells [66]. RUNX1 mutations occur in 8%-23% of MDS, most commonly in the setting of therapy-related MDS, and are frequently detected in patients who transform to AML [67]. Given these associations, it is not surprising that outcomes in this patient population are quite grim. Controlling for factors such as MD Anderson Prognos-

Other transcription factors
The ETV6 gene encodes a hematopoietic transcription factor that functions in early hematopoiesis in the bone marrow [69]. It is an uncom-

MUTATIONS IN RNA SPLICING
Next-generation sequencing approaches have identified mutations in multiple RNA splicing genes. Collectively, these mutations have been observed in 85% of neoplasms with MDS features, and approximately half of MDS patients carry at least one somatic mutation in a spliceosome gene [72][73][74]. Those mutations likely affect the core components of initial steps in the RNA splicing machinery, but the link to leukemogenesis remains elusive [40].

SF3B1 mutation
This gene encodes for the splicing factor 3b subunit 1, and it occurs in approximately 25% of patients. Based on a recent update from the international working group of experts in MDS, SF3B1-mutant MDS now refers to a new subtype of the disease. It is defined by the mutation, cytopenias, morphologic dysplasia (with or without ringed sideroblasts), bone marrow blasts < 5%, and peripheral blasts < 1%. This distinction was made as SF3B1 mut is classically characterized by a ringed sideroblastic phenotype with ineffective erythropoiesis but an indolent clinical course [74]. Undoubtedly, patients whose cells harbor this genetic alteration live longer (7.5 vs. 4.1 years) with less progression to AML, independent of age, IPSS-R, or other somatic mutations [18,27,30,72,75]. Further confirming the association, SF3B1 mut compared to SF3B1 wt found this mutation to possess a superior OS [5,15,30].

SRSF2 mutation
This gene encodes serine/arginine-rich splicing factor 2 that belongs to the serine/arginine rich protein family, which is important for splicesite selection, splicesome assembly, and both constitutive and alternative splicing [76]. This subtype occurs in 4%-23% of patients and is associated with excessive blasts, mild thrombocytopenia (88,000 vs. 165,000), and decreased neutrophils (1100 vs. 2300) [40].

U2AF1/ZRSR2 mutation
These genes (U2AF1-U2 small nuclear RNA auxiliary factor 1; ZRSR2-U2 small nuclear ribonucleoprotein auxiliary factor 35 kDa subunitrelated protein 2) are involved in the spliceosome pathway and are reported to be mutated in approximately 5%-16% and 3%-11% of MDS cases, respectively. Both mutations have been linked to inferior outcomes, however in comparison to other genetic lesions the amount of evidence is small. U2AF1 mut patients have been linked to inferior survival when controlling for IPSS-R/other genetic mutations [18], and increased rate of leukemic transformation when compared to wildtype patients [79]. Conversely, inferior survival noted in U2AF1 mut compared to U2AF1 wt disease did not meet statistical significance when considering other prognostic variables [16]. Similarly, ZRSR2 mutations have been linked to inferior OS and increased rate of AML transformation [40], but this observation is unconfirmed to date.

CBL mutation
The Casitas B-lineage (CBL) lymphoma gene encodes a tumor suppressor protein with E3 ubiquitin ligase activity that acts as a negative regulator of receptor tyrosine kinase. As such, a lack of this signal drives an oncogenic pathway [80]. Mutant CBL is thought to play an important role in development of myeloid malignancies [81], and in MDS the mutation occurs in 1.5%-5.1% of cases. Regardless of its incidence, CBL mutant patients have poor outcomes compared with wild-type CBL. Importantly, independent inferior survival was noted despite controlling for other prognostic markers such as age, transfusion dependence, IPSS, and other mutations [18,22,64].

RAS pathway mutations
This group includes NRAS, KRAS and HRAS that encode GTPases that regulate cell proliferation, differentiation, and survival [82]. While mutations in the RAS gene family are found at a rate of 6.2%-8.8% of MDS, there does not appear to be definitive relation to prognosis.
Multiple separate analyses show lack of a prognostic impact when controlling for age, cell counts, IPSS-R, and other mutations [5,16,29].
Alternatively, two studies have associated these mutations with inferior OS when controlling for established clinical risk factors [30,42].
Other RAS pathway mutations, PTPN11 and JAK2, are uncommon in MDS, but also part of this signal transduction pathway. No independent survival impact was found for PTPN11 or JAK2 in similar analyses [5,29].

COHESIN COMPLEX MUTATIONS
The cohesin complex includes a multitude of genes that act together in the control of cell division through regulating dissociation of sister chromatids in mitosis or meiosis [83]. Mutations that occur within this complex in MDS are SMC1A, RAD21, SMC3, and STAG2. STAG2 mut MDS appears to be the most frequently mutated gene in the complex, occurring in 5.9%-7.5% of cases. Collectively, the impact of cohesin complex mutations on prognosis has yet to be defined. STAG2 mut disease has been associated with inferior survival independent of age, gender, and IPSS [30,84]. Despite this finding other data suggest the association with poor survival is lost when controlling for known prognostic markers, even when considering cumulative mutations in the entire cohesin complex (STAG2, SMC3, RAD21/ SMC1A) [5,18,24]. In a large analysis, RAD21 mutations were found to predict inferior survival when controlling for age and IPSS-R, however this has not been duplicated elsewhere in the literature [30].

CLINICAL APPLICATION OF GENETIC MUTATIONS
Importantly, both the specific mutational profile and the overall muta-  compared to those without molecular abnormalities [24]. In addition, the rate of leukemic transformation (median rate 4 months with ≥5 mutations) and time spent free of disease progression (HR 3.364 for PFS) were inferior for those with higher mutational burden [15,17,24,42].
Recent data suggest that driver mutations emerge heterogeneously in different patient subgroups. VAF, or the percentage of a specific genetic variant, is gaining traction for mutational prognostication [85].
For example, in the case of TP53, patients with VAF < 10% are not found to have the same poor outcomes as those with higher VAFs [32].
In contrast, patients who harbor TET2 mutations with a VAF > 18% have been shown to have a 57% reduction in lifespan, when compared to patients with VAF < 18% or TET2 wild-type disease. Additionally, SF3B1 mutations with a VAF > 15% were not only seen to have supe-rior survival, but more likely to be classified as MDS with ringed sideroblasts [62].
In 2012, the IPSS was updated to include cytogenetics [86]. Using comparable OS and leukemic free survival data, we postulate a scoring system ( Tables 2-4 detail a proposed change to the IPSS-R, by comparing median survival, hazard ratios for overall survival, and transformation to AML, to similar cytogenetic data that was used to create the initial IPSS-R model in 2012. Please refer to Table 1 for full list of citations regarding overall survival and leukemic free survival for each individual mutation. A similar algorithm was recently created from 1471 MDS patients, incorporating the standard prognostic variables (karyotype, platelet count, hemoglobin, bone marrow blast percentage, age) and seven discrete genetic mutations (TP53, STAG2, RUNX1, RAD21, SRSF2, ASXL1, and SF3B1) that had independent impact on overall and leukemic free survival [30]. Other changes to the standard prognostication models include streamlined whole-genome sequencing of patients. A study of patients with AML and MDS using whole-genome sequencing detected all standard genetic alterations by traditional cytogenetic analysis, as well as identify a new genomic event in 25% of this population, leading to a change in prognostic risk category for 16.2% of the cohort [88].
A separate analysis of older (≥60) AML patients was able to show dramatically different outcomes for patients treated with traditional 7 +

GENOMIC-BASED THERAPEUTICS
Understanding genetic drivers of MDS provides opportunity to target these lesions ( Group for the prognosis of MDS. These trials will shed light on the net impact of driver mutations on clinical outcomes which should be incorporated into future prognostic scoring systems. In addition, impact of VAFs and class of mutation (frameshift vs. missense mutation vs. deletion vs loss of heterozygosity) for each specific genetic lesion will play a role in the biology of disease and response to treatment, allowing us to distinguish driver versus passenger mutations. We postulate that these data will enhance our prognostic capabilities and our therapeutic decision making to improve patient outcomes.

CONFLICT OF INTEREST
The authors declare no conflict of interest.

AUTHOR CONTRIBUTIONS
Michael R. Cook and Catherine Lai designed review concept. Michael R. Cook collected literature and summarized analysis. All the authors wrote and edited the manuscript, and have agreed to the final submitted version.