Spliceosomal Factor Mutations and Mis-splicing in MDS

Abstract

Somatic, heterozygous missense and nonsense mutations in at least eight proteins that function in the spliceosome are found at high frequency in MDS patients. These proteins act at various steps in the process of splicing by the spliceosome and lead to characteristic alterations in the alternative splicing of a subset of genes. Several studies have investigated the effects of these mutations and have attempted to identify a commonly affected gene or pathway. Here, we summarize what is known about the normal function of these proteins and how the mutations alter the splicing landscape of the genome. We also summarize the commonly mis-spliced gene targets and discuss the state of mechanistic unification that has been achieved. Finally, we discuss alternative mechanisms by which these mutations may lead to disease.

Introduction

Somatic mutations in over 30 splicing factors have been identified in hematological malignancies [1] with SF3B1, U2AF1, SRSF2, LUC7L2, PRPF8, DDX41 and ZRSR2 being the most common. Over 50% of MDS patients harbor mutations in splicing factors [2, 3]. The splicing factor mutations occur early in MDS and have been observed in CHIP, suggesting that they can act as founder mutations [4–6]. Splicing factor mutations are also observed in AML and MDS/MPN syndromes. They are more common in sAML than in pAML, and it has been proposed that the presence of splicing factor mutations may be a defining factor for differentiating between sAML and pAML [7]. Indeed, patients diagnosed with pAML, who harbor splicing factor mutations, are more similar in prognosis and disease progression to patients with sAML even if they had no MDS history [7].

The mutations in the splicing factors are generally mutually exclusive [3–5] although they co-occur with epigenetic modifiers [4, 8, 9]. This suggests that the splicing factor mutations operate on a common downstream target or pathway. However, SF mutations play different roles in the spliceosome and interact differently with pre-mRNA, therefore, it is likely that in there also are mutation-specific alterations explaining the morphological and clinical diversity of cases affected by mutations of SFs.

Pre-mRNA Splicing

Pre-mRNA splicing is the post-translational process of removing introns and ligating exons to form mature transcripts that are suitable for translation. Introns must be removed from a transcript before the transcript is exported to the cytoplasm for translation and this process is carried out by the spliceosome, a dynamic multimegadalton molecular machine. The spliceosome is composed of five small nuclear RNAs (snRNAs) and over 100 proteins [10]. The snRNAs and proteins are organized into five small nuclear ribonucleoproteins (snRNPs) named U1, U2, U4, U5 and U6.

Figure 1 shows a highly simplified diagram of the process of spliceosomal splicing as it pertains to the factors commonly mutated in myeloid neoplasms. A single intron separating two exons is shown traversing the splicing process from upper left to upper right to yield the spliced exons and the excised intron. The five snRNAs and their associated proteins are shown as colored balls while the various splicing factors mutated in disease are denoted by boxes located at the approximate step where they become stably bound to the spliceosome.

Figure 1:

Diagram of the spliceosome cycle.

Shown are some of the steps of pre-mRNA splicing of a single intron. The pre-mRNA is diagramed at the upper left prior to the binding of splicing factors. In this simplified diagram, the various snRNAs and their associated proteins are shown as colored balls. The gain and loss of these snRNPs provide the framework for this model. The various protein factors that are frequently mutated in MDS are shown in boxes at points in the cycle where they are thought to join the splicing complex. The cycle ends at the upper right with the release of the spliced exons and the dissociation of the spliceosome from the excised lariat intron.

Splicing begins with the recognition of the 5’ and 3’ splice sites plus the branch site (BS), in the E complex (Fig. 1). This process is highly regulated in higher eukaryotes through cis acting RNA sequence elements and trans acting RNA binding proteins (such as SRSF2) to select the appropriate splice sites for cell type and condition specificity [11]. Another mutated factor, LUC7L2, is thought to play a role in the binding of the U1 snRNP to the 5’ splice site [12, 13]. The formation of Complex A includes the binding of the U2 snRNP which is promoted by the binding of the U2AF factor that includes the mutated protein U2AF1. One of the proteins delivered along with U2 is the mutated protein SF3B1. Another mutation-prone protein, ZRSR2, appears to play a similar role as U2AF1 but in a subset of introns [14]. Complex B is formed with the addition of a large complex of three snRNAs, U4, U5 and U6 and their associated proteins including the large and highly conserved factor PRPF8 which is also frequently mutated. A significant rearrangement of Complex B to B* with the loss of many factors and the addition of others ultimately leads to the catalytic Complex C. One of the added proteins in this rearrangement is DDX41, another frequently mutated factor. The function of DDX41 in the spliceosome is unclear but it is observed to copurify with Complexes B*, C* and P [15].

Alternative Splicing

Alternative splicing (AS) is a process in which the primary mRNA transcript is spliced in more than one pattern. AS facilitates genetic complexity by allowing a single gene to encode multiple mature mRNAs that often result in multiple protein isoforms. Over 90% of transcripts are alternatively spliced in humans [1, 16]. Many mRNA isoforms are translated into proteins with distinct functions, therefore isoform expression is tightly regulated. The balance of expressed isoforms can be developmental stage specific, tissue specific or in response to cellular signals [17]. Over half of AS events differ between tissues and the dominant isoform is tissue specific for 35% of genes [18, 19].

Dysregulation of splicing in cancer

Dysregulated splicing of transcripts is a hallmark of cancer [20] and is caused by somatic mutations in splice sites, mutations in splicing factors, or dysregulated expression of splicing factors. Assessment of alternative splicing patterns in the TCGA cohorts revealed large scale mis-splicing across all cancers with the highest number of mis-splicing events being in detected in AML patients [21]. Similar patterns are seen in MDS as well (Hershberger 2020). In the context of cancer, dysregulated alternative pre-mRNA splicing may confer functional changes in the translated protein, potentially creating oncogenes or inactivating tumor suppressor genes. There are several examples of oncogenic isoforms that promote disease including CASP8, BCL, Caspac2 and TP53 [22–24].

Splicing factors implicated in myeloid neoplasms

In this review, we will summarize the current knowledge about the relationship between myeloid neoplasms and splicing factor mutations with an emphasis on how recurrent mutations alter the function of each factor and how this impacts AS patterns. The guiding hypothesis for this analysis is that some or all of these mutations lead to impacts on either a single gene target or targets within a single pathway. This hypothesis arises from the highly mutually exclusive occurrence of the major splicing factor mutations. It appears that a single mutation is disease driving while more than one is non-additive or possibly deleterious [25]. A discussion of potential common targets or mechanisms is discussed at the end.

SF3B1

The gene SF3B1 is found at 2q33.1 and encodes a ~146kDa protein that is a subunit of the splicing factor 3b complex that functions in the major spliceosome and is also a component of U11/U12 di-snRNP in the minor spliceosome [26]. SF3B1 is a core component of the spliceosome [26–28]. It contains 22 non-identical HEAT repeats [29]. In the major spliceosome, SF3B1 binds to the intron, 5’ of the branchpoint as the spliceosome transitions from E complex to A complex [30]. Upon the rearrangements of SF3a and SF3b, SF3B1 is phosphorylated, resulting in a transition to the catalytically active conformation of the spliceosome [31, 32].

SF3B1 is the most commonly mutated splicing factor in MDS [3, 4, 33] and the second most commonly mutated overall, surpassed in frequency only by TET2. Of the MDS/MPN overlap syndromes, it is predominant in MDS/MPN-RS-T and MDS/MPN-U and less common in CMML. SF3B1 is the least frequent splicing factor found to be mutated in AML. SF3B1 mutations occur early, suggesting that it is an initiating event in the formation of myeloid malignancies. Additionally it has been observed in CHIP and normal blood controls [34].

In MDS, patients who harbor SF3B1 mutations frequently present with RARS or RCMD-RS [35]. SF3B1 MDS patients have a favorable outcome in multivariate analyses [35–38] and SF3B1 mutations have such a distinct clinical phenotype that they constitute a separate nosologic entity [5, 33]. Although SF3B1 mutations are closely associated with ring sideroblasts, the SF3B1 mutation status alone is able to predict prognosis and the progression of disease [39].

SF3B1 is also mutated in other leukemias such as chronic lymphocytic leukemia [2, 40] and solid tumors such as breast cancer [41], pancreatic cancer [42] and uveal melanoma [43]. However, the distribution of mutational hotspots differs from that seen in MDS.

In myeloid malignancies, mutations typically occur at the K700E position in heat repeat 7, however mutations are observed at a lower frequency across heat repeats 5–9. These mutations have been shown to have a similar effect on alternative splicing [44–47]. Initial studies showed that these mutations in SF3B1 altered 3’SS selection due to the increased recognition of cryptic splice sites that occur between the branchpoint and canonical 3’SS [45]. Other studies suggested instead that mutant SF3B1 activated a different branchpoint upstream of the canonical 3’SS [44]. This usage of alternative 3’SS results in frequent frameshifts in the affected mRNAs and induced nonsense mediated decay [44]. Recently SF3B1 mutations have been recognized to also contribute to deregulated intron splicing [46, 48–51] with the main phenotype being enhanced intron exclusion [47].

Cell models have shown that inhibition or reduced expression of SF3B1 results in the formation of ring sideroblasts and heme deficiency [52, 53]. Additionally, depletion of SF3B1 impairs growth and differentiation [53].

Early mouse models of SF3B1 haploinsufficient mice displayed a phenotype of dysregulated hematopoiesis, but they never developed MDS [52, 54, 55]. The cellular phenotypes were subtle while the splicing defects were not recapitulated suggesting that the hotspot mutations conferred a change of function rather than a loss of function [52–55]. Mouse models harboring K700E/WT did not exhibit RS but showed symptoms of anemia, erythroid maturation defects, decreased hematopoietic stem cell numbers, and impaired repopulating abilities [48, 49]. The mouse models recapitulated the dysregulation of alternative 3’ splice site usage and intron exclusion patterns observed in patients and cell lines, but the affected genes showed little (~5%) overlap with mis-splicing events observed in humans [48, 49].

SRSF2

The SRSF2 (Serine/arginine rich splicing factor 2) gene is located at 17q25.1. SRSF2 encodes a 221 amino acid protein that promotes spliceosome assembly and coordination of the first step of splicing. As part of the SR family of splicing factors, SRSF2 contains an RNA-binding RRM domain as well as an argenine/serine rich RS domain [56]. SRSF2 has been shown to guide splice site selection for both constitutive and alternatively spliced exons and introns. It primarily binds within exonic splicing enhancers to degenerate purine-rich motifs to activate adjacent splice sites.

SRSF2 missense mutations and small in-frame deletions occur in 9–15% of MDS cases [3, 8]. Mutations in SRSF2 are also observed in AML (~10%), though more frequently in sAML than pAML. SRSF2 mutations occur at the highest proportion in MDS/MPN syndromes, with a frequency of 33% in CMML [3, 57]. SRSF2 mutations occur early in disease progression [4, 7]. In MDS patients, SRSF2 mutations co-occur with IDH1/2, TET2, RUNX1, ASXL1 and STAG2 [4, 8, 35, 58]. In CMML patients, SRSF2 is mutually exclusive with mutations in SF3B1, U2AF1 and EZH2, but co-occurs with TET2 mutations [59].

MDS patients with SRSF2 mutations present with neutropenia and thrombocytopenia [35]. SRSF2 mutations are predictive of shorter survival and increased risk of transformation to sAML [4, 8, 37, 57, 58]. SRSF2 mutations are also observed in primary myelofibrosis (8–17%) and the presence of a mutation is indicative of poorer prognosis [60, 61].

In myeloid malignancies, the somatic missense mutations and small in-frame deletions are virtually invariably found at P95 in the region between the RRM and RS domain. This hotspot pattern coupled with the absence of frameshift and nonsense mutations, suggests that the observed mutations lead to gain or altered function [58, 60]. SRSF2^P95H has been shown to cause altered splicing of pre-mRNA in transformed cells, murine models and MDS patients. These splicing changes can be attributed to a shift in RNA binding preference; the P95H SRSF2 has enhanced binding to the sequence UCCA/UG and decreased binding of the sequence UGG/UG. The preferred binding of P95H SRSF2 to UCCA/UG results in changes in constitutive and alternative splice site usage [62–64]. While a large number of genes were affected, the absolute magnitude of most isoform shifts was rather low.

Genes found to be differentially spliced in mutant SRSF2 samples were enriched in cancer progression, development, and apoptosis pathways. Targets found in multiple model systems include genes involved in cell cycle (TBRG4, CDK10) and genes involved in apoptosis (ATF2, RAPGEF1). There are also several genes associated with cancer that are altered in SRSF2 cell line models, murine models and patient samples. Examples include ZMYND8 (T-Cell lymphoma oncogene), PICALM (gene mutated in AML), and TPD52L2 (Tumor Protein D52-like) and EZH2.

Early studies examined phenotypes in mice homozygous and heterozygous for SRSF2 knockout and mice harboring the missense P95H mutation. These studies revealed SRSF2 is necessary for hematopoiesis and that the P95H mutation confers a change of function, not loss of function. All three models exhibited impaired colony formation ability in vitro, leukopenia and anemia, but only the SRSF2 heterozygotes and homozygotes knockouts displayed bone marrow aplasia [62]. Srsf2^P95H mice alone developed myeloid and erythroid dysplasia that partially mimics features of human MDS [62].

SRSF2^P95H mouse models show blastic morphology; their bone marrow contains dysplastic neutrophils and erythrocytes [62, 65, 66]. Srsf2^P95H mice had increased number of HSCs in their bone marrow, but remained anemic due to an increase in apoptotic cells. Additionally, introduction of the P95H mutation resulted in impaired hematopoietic stem cell differentiation and/or impaired self-renewal [62, 65]. Although none of the SRSF2^P95H murine models develop overt MDS, upon transplantation, the recipient mice develop myelodysplasia [65, 66]. A recent study has shown cooperativity between SRSF2 and IDH2 mutations where mice transplanted with double mutant bone marrow develop severe MDS with shorter survival than mice with SRSF2^P95H alone [66].

U2AF1

U2AF1 (U2 Small Nuclear RNA Auxiliary Factor 1) is located at 21q22.3. U2AF1 is a 35kDa protein that interacts with U2AF2, a 65kDa protein to form the U2AF heterodimer. U2AF1 is recruited to 3’ splice sites by regulatory factor such as SRSF2. U2AF1 then interacts with the 3’ splice site (AG) and stabilizes the binding of U2AF2 to the polypyrimidine tract [67]. The heterodimer facilitates recruitment of the U2 snRNP to the branchpoint [68–70].

U2AF1 is mutated in 7–9% of patients with MDS. It co-occurs with ASXL1, DNMT3A and del20q [4, 8, 35]. It is mutated in 5–11% people with AML, more often in patients with sAML than pAML. It has also been shown to co-occur with complex [71, 72].

Patients with U2AF1 mutations often present with thromobocytopenia. Studies on the impact of U2AF1 on survival have yielded conflicting results with some indicating no effect on overall survival [8, 72, 73] while others demonstrate poorer prognosis [37, 73–75]. Some of these studies have indicated an association with progression to AML or a decrease in relapse free survival post-transplant [72, 75].

Mutations of U2AF1 at the S34 position have also been observed in 3–4% of lung adenocarcinomas, U2AF1 mutations have also been observed, albeit rarely, in bladder cancer and more frequently in uveal melanoma [76, 77]. Although only U2AF1^S34 mutations are observed in lung adenocarcinomas (as opposed to U2AF1^Q157), these patients exhibit some overlap of mis-spliced genes with myeloid neoplasm patients harboring the same mutation [76].

U2AF1 is required for the recognition of a subset of so called AG-dependent 3’SS [67]. U2AF1 contains the two zinc finger domains that include the S34 and Q157 mutational hotspots. U2AF1 mutations at either positions S34 and Q157 alter 3’SS recognition. The 3’ splice site has a consensus sequence of YAG/G with the / representing the intron/exon boundary. U2AF1S34 mutations alter the recognition of the Y at position −3 while Q157 mutations alter the recognition of the G at the first exon position. [76, 78–80]. This results in dysregulation of 5–10% of splice sites, causing usage of alternative 3’ splice sites and increased exon skipping [3, 73]. The two hotspot mutations in U2AF1 cause distinct effects on many of the same splice sites [76, 79, 80].

Cell line models of U2AF1 mutations exhibit decreased cell proliferation and increased G2/M arrest as well as enhanced apoptosis [3]. Yip et al. [81] studied the effects of U2AF1^S34F mutations on both granulocyte and erythrocyte differentiation by overexpressing U2AF1^S34F in CD34⁺ human cells. Differentiation into either lineage was impaired. Those that were differentiated into erythrocytes showed impaired growth and hemoglobinization as well as an increase in apoptosis. Cells differentiated into granulocytes also showed impaired growth and increased cell cycle arrest [81].

The earliest U2AF1 mouse models displayed engraftment, proliferation and differentiation defects [3]. Inducible mouse models of U2AF1^S34F exhibited leukopenia as well as a reduction of B cells and monocytes upon induction. They also showed increased HSPCs in the bone marrow compartment [82, 83]. However, the mice did not develop MDS or AML nor were they able to outcompete wild type cells in a competitive repopulation assay [82].

ZRSR2

The ZRSR2 gene, also known as U2AF1-RS2 and Urp, is located at Xp22.2. This gene encodes a 483 amino acid splicing factor with homology to U2AF1 that appears to contributes to U12-dependent minor class splicing. Like other U2AF1-related proteins, ZRSR2 contains an argenine/serine rich region near the c-terminus, two zinc-finger domains, and a U2AF homology motif (UHM) [84]. ZRSR2 also contains two stretches of sequence homologous with U2AF1, allowing it to bind U2AF2, however, ZRSR2 is not functionally redundant with U2AF1 [85]. In vitro, ZRSR2 is recruited to the 3’ splice sites of both major and minor-type introns although at different steps. The binding of ZRSR2 at U12-type 3’ splice sites is required for assembly of early splicing complexes. But the binding of ZRSR2 at U2-dependent 3’ splice sites is required for completion of the second step of splicing [84, 85].

More than 20 different missense and frameshift mutations have been reported throughout the ZRSR2 gene in patients with MDS [3, 8]. The frameshift mutations observed result in a truncated ZRSR2 protein and presumptive loss of function. While missense mutations were found in both males and females, the frameshift mutations were found exclusively in male patients. ZRSR2 mutations are present in MDS patients at a frequency of 3–8%. But ZRSR2 occurs less frequently (<2%) of patients with AML [3, 61]. Additionally, ZRSR2 expression is decreased in the bone marrow of patients with myeloid neoplasms compared to healthy individuals [86]. The impact of ZRSR2 mutations and lower expression on patient outcome in MDS is unclear. Thol et al. did not identify any association with adverse outcome [8] but other studies have shown that ZRSR2 mutations are associated with greater disease burden, higher blast counts and with greater risk of progression to AML [3, 35].

ZRSR2 mutations occur in 2% of TCGA samples and 1% of GENIE cases, across multiple cancer types. However, unlike myeloid neoplasm samples, the vast majority of these mutations are missense, rather than frameshift [87, 88].

Since ZRSR2 mutations are typically frameshift and occur throughout the gene, they are likely loss of function. Knockdown of ZRSR2 in the cell lines HEK293T, TF-1, and K562 resulted in decreased splicing efficiency of U12-dependent introns, although U2 introns were largely unaffected. Similarly, bone marrow cells derived from ZRSR2-mutant MDS patients have a greater incidence of abnormal splicing, including high levels of intron retention. 95% of U12-dependent introns are mis-spliced in bone marrow cells harboring a ZRSR2 mutation. There is also evidence of cryptic splice site activation of U12 splice sites and exon retention near U12 splice sites [14].

U12-type introns are found in a large number of essential genes. ZRSR2-deficiecy results in reduced splicing and/or mis-splicing of genes involved in pathways such as RNA transport, cell cycle, cellular response to stress, response to DNA damage stimulus, protein transport, protein serine/threonine kinase activity and ribonucleotide binding. Specific genes containing U12-type introns that were consistently mis-spliced included several E2F transcription factors, MAPK signaling proteins; MAPK1, MAPK3 and RAS-guanyl releasing proteins, and RAF serine/threonine protein kinases. Interestingly, the PTEN oncogene, loss of which known to impair HSC activity in mice, is also mis-spliced in ZRSR2-deficient cells [14].

ZRSR2-deficient cell models exhibit proliferation defects and reduction in colony formation. In human CD34+ hematopoietic stem cells enriched from cord blood, ZRSR2 knockdown results in changes in myeloid and erythroid lineage differentiation capacity [14]. In mouse models, K562 ZRSR2 knockdown cells have the capacity to form spontaneous subcutaneous tumors in NOD-scid-gamma mice, but produce smaller overall tumors compared to tumors formed from K562 ZRSR2-WT cells [14].

PRPF8

PRPF8 is a large core splicing factor located on 17p13.3 that facilitates the completion of the second step of splicing [89, 90]PRPF8 contains an RRM domain and nuclear localization signal [91]. It is the most evolutionarily conserved among nuclear proteins, is expressed in all tissue types and is used for both U2 and U12 intron removal [92]. As a component of the core, associates with U5 as well as the tri-snRNP (U5*U4/U6) and interacts with both the 5’ and 3’ splice sites as well as the branchpoint [91, 93].

PRPF8 mutations are found in 1–4% of myeloid neoplasms, they are mostly mis-sense mutations but rarely occur in the same position [37, 94]. Some patients lose a copy of PRPF8 through deletion of 17q; 24/447 cases from Kurtovic-Kozaric study [94] and 12 cases in the TCGA AML dataset. PRPF8 mutations co-occur with TET2, CBL and TP53 and PRPF8 is sometimes co-deleted with TP53 [94]. The majority of cases of PRPF8 mutations and deletions in MDS patients were indicative of RS and PHA (Pelger-Huet anomaly) [94]. PRPF8 mutations and del17p deletions are observed at even greater frequencies in AML and are associated with a poor prognosis [5, 94].

Germline and somatic mutations in PRPF8 may also have a pathological effect outside of bone marrow. Somatic alterations are observed in 3% of all patients across the TCGA cohorts [87, 88]. Additionally germline mutations in the C terminus lead to autosomal dominant retinitis pigmentosa [95].

Although mis-sense mutations are not correlated with changes in expression, some myeloid neoplasm patients have decreased PRPF8 expression due to chromosomal abnormalities. Decreased expression of PRPF8 alters splicing of introns with weak 5’ splice sites [96]. Knockdown of PRPF8 in both K562 cells and CD34+ primary bone marrow cells resulted in increased proliferation and clonogenicity [94].

Other patients have somatic mis-sense mutations in PRPF8. Cells harboring PRPF8 mutations are more susceptible to treatment with the splicing inhibitor meayamycin, confirming that PRPF8 mutations contribute to dysfunctional splicing [94]. Since the mis-sense mutations in PRPF8 do not occur in a hotspot location, the impact on splicing may differ depending on the location of the mutation. Nine PRPF8 mutations have been tested for deregulated splicing effects in yeast. The experiments in yeast, in combination with experiments in human cells, indicated that, although both haploinsufficiency and mutations in PRPF8 have a downstream effect on splicing, mutated PRPF8 has improved recognition of weak 3’ splice sites, indicating a change of function rather than a loss of function [94]. These observations were corroborated using RNA-Seq data comparing AML patients with PRPF8 mutations and those without.

DDX41

DDX41 encodes a 622 amino acid protein and is located at 5q35.3. It contains an N-terminal DEAD domain, c-terminal ATP-dependent helicase domain and a zinc finger domain suggesting that it is a member of the DEAD-box RNA helicase family. It has been identified by IP-Mass Spec as a component of the mammalian spliceosome catalytic core although its precise role in splicing is unclear [97]. It also appears to plays a role in innate immunity [98–101].

DDX41 was first found to be mutated in MDS and AML by Polprasert et al in 2015 [102]. Since then, both germline and somatic mutations have been identified in patients with myeloid neoplasms [102–111]. DDX41 is the only splicing factor mutated in myeloid neoplasms that harbors germline mutations. The germline mutations are typically frameshift or initiating codon mutations [102]. Genomic mutation carriers are asymptomatic until late in life and the acquisition of additional somatic mutations. Approximately half of patients acquire a highly recurrent R525H hotspot mutation in their normal DDX41 allele. Notably, somatic R525H mutations can occur either in the presence or absence of a frameshift mutation [102].

Additionally, some patients lose a single copy of DDX41 through deletion of a portion of 5q. Five to ten percent of MDS patients have deletion of 5q. Deletion of 5q alone is referred to as 5q syndrome, but many patients also harbor mutations in TP53 [112]. Patients with 5q syndrome are predominantly female, have a low rate of progression to AML, and respond well to treatment with lenalidomide. Although the sizes of the 5q deletions vary, the most commonly deleted region includes the critical genes RSP14, CSNK1A1, APC, DDX41 and Mir-145–146a [113].

Unlike patients with 5q syndrome, Patients with DDX41 mutations are typically younger and present with a more aggressive phenotype [102]. Large cohorts such as TCGA and BeatAML report very low variant allele frequencies of the R525H mutation (~15%) suggesting that once it is acquired, the disease progresses very quickly [87, 88].

DDX41 aberrations are present in 2% of the TCGA cohorts suggesting that it may play a role in other cancers. R525H has only been observed in myeloid malignancies, but R525C has been seen in colon adenocarcinoma and R525S has been reported in cutaneous melanoma [87, 88].

Little is known about the effects of DDX41 mutations on splicing. However, the germline frameshift mutations that are observed often occur early in the gene, suggesting a loss of function, rather than the production of a truncated protein. Loss of a single copy of DDX41 alone is not sufficient to induce bone marrow failure, but no DDX41 null patients have been observed, nor have DDX41 null mice been bred, suggesting that some expression is essential for life. Gene knockout screens for genes required for cell survival in culture have shown that DDX41 is essential. In addition to its probable role in splicing, DDX41 also appears to have a cytoplasmic role as a component of the innate immune system through its interaction with the STING protein [100].

LUC7L2

The LUC7L2 gene is located at 7q34. It encodes a 392 amino acid protein that is predicted to play a role in pre-mRNA splicing. Although the function of LUC7L2 has not yet been characterized, it colocalizes with components of the U1 snRNP in the nucleus [114]. LUC7L2 shares 30% sequence homology with the yeast protein Luc7p, an essential component of the yeast U1 snRNP that is involved in recognition of non-consensus 5’ splice sites [13]. In yeast, Luc7p binds the Sm proteins of U1 snRNP through its N-terminal alpha helix and contacts the 5’SS-U1 alpha helices with both zinc fingers to stabilize the binding of weak 5’ splice sites [115]. The N-terminus as well as two zinc finger domains are highly conserved between LUC7p and LUC7L2. In Luc7p, these domains have been shown to have necessary and distinct functions. Mutations or deletions of either the N-terminal domain or zinc finger domains result in impaired in vivo splicing of SUS1 pre-mRNA. [12] Human LUC7L2, also contains a C-terminal arginine and serine rich domain, common to the SR family of splicing factors.

Deletion of the q arm of chromosome 7 (Del7q) is a common feature in MDS, affecting 5–10% of patients with de novo MDS and 50% of therapy related cases. It is associated with poor prognosis. LUC7L2 is located in the most frequently deleted region of this chromosome which also includes EZH2, CUX1 and MLL3 [116].

Expression of LUC7L2 is decreased in 14% of patients with MDS through mutation or deletion of a portion of 7q. Low expression of this protein is correlated with a decrease in overall patient survival [117]. Additionally, expression of LUC7L2 is lower in the bone marrow of patients with myeloid neoplasms (MDS, AML, MDS/MPN-U, MDS/MPN-RS-T, CMML) than in the bone marrow of healthy individuals, regardless of mutational status [86].

LUC7L2 is affected by mutation, amplification or deletion in 2% of patients in the TCGA cohort indicating that it may play a role in other cancers as well. Additionally, LUC7L2 is significantly downregulated in 5 cancer types of the TCGA cohort (ACC, BLCA, LAML, UCEC, UCS) and upregulated in 5 others (DLBC, GBM, LGG, PRAD, THYM) [118].

Induced pluripotential stem cells (iPSCs) derived from the hematopoietic cells of patients harboring del7q recapitulate the hematopoietic differentiation defects observed in MDS. A deletion spanning the 20MB between 7q32.3–7q36.1 is sufficient to phenocopy the differentiation defects of del7q. LUC7L2 is located at 7q34 and is one of four genes able to partially rescue hematopoietic differentiation [119]. Therefore, it is predicted that downstream targets of LUC7L2 are involved in the process of differentiation [119]. The contribution of LUC7L2 to the pathology of MDS may be through impairment of hematopoietic stem cell differentiation [119].

The search for commonly affected genes or pathways

The highly recurrent mutations seen in only a few of the many splicing factor genes as well as their mutually exclusive distribution has led to the hypothesis that they all affect a single or small set of genes and/or a single disease pathway. Several analyses have identified gene candidates as listed in Table 1. These represent genes that are reproducibly mis-spliced in multiple studies of a single mutated splicing factor. In a few cases, they are mis-spliced in multiple mutants. However, there is no apparent overlap among all or even a majority of splicing factor mutants. Pathway analyses have shown the dysregulation of oxidative phosphorylation and DNA repair as well as the Sirtuin pathway [46]. Our work implicates the mis-splicing of detained introns that may impair differentiation of progenitor cells into the myeloid lineages [86]. Other suggestions have included a role for enhanced R-loop formation in the presence of splicing factor mutations leading to increased genome instability [120, 121]. As of now, no clear commonality has emerged from the research that clearly connects the splicing factor mutations to the disease process.

Table 1.

Reproducible/Overlapping Target Genes

ASXL1	SF3B1	[52], [53]
ATR	U2AF1	[82], [80], [76], [81], [48]
BCOR	SRSF2	[63], [62], [86]
CBL	SF3B1	[52], [53]
EZH1	SF3B1	[52], [53], [62], [86]
EZH2	SRSF2	[63], [62], [46],[47], [86]
	U2AF1	[46],[47], [86]
FANCA	SF3B1	[50]
	U2AF1	[76], [86]
H2AFY	U2AF1	[76], [81], [86]
MAP3K7	SF3B1	[127], [48], [129]
NF1	SF3B1	[127], [48]
PDS5	SF3B1	[127], [48], [86]
SERBP1	SF3B1	[129], [48], [86]
SLC25A37	SF3B1	[128], [53], [86]
TMEM14C	SF3B1	[47], [129]
ANKHD1-EIF4EBP3	SF3B1	[129], [48], [86]

Exploiting splicing factor mutations for therapy

The observations that the more common spliceosomal factor mutations rarely, if ever, co-occur in a single disease clone as well as evidence from cell culture that multiple mutations lead to cell death [25] has led to the hypothesis that splicing factor mutant MDS clones might be hypersensitive to drugs that further inhibit the spliceosome [122]. Two compounds, E7107 and H3–8800 have shown enhanced cell inhibition when tested on cells expressing spliceosomal mutations [123, 124]. Clinical trials of these compounds were not successful, however, with E7107 causing ocular side effects [125] and H3–8800 failing to show clinical efficacy [126]. Both compounds target the same factor (SF3B1). Additionally, mis-splicing itself provides potential targets for personalized therapy through the creation of neojunctions that can be translated to neoantigens [86]. Future development of drugs targeting other factors or splicing-derived neoantigens may validate these therapeutic modes of action.

Practice points.

1. Somatic mutations in spliceosomal factor genes are present in 60–70% of patients with myeloid neoplasms. These can be founder mutations or seconday hits.

2. Spliceosomal factor mutations are often mutually exclusive but can co-occur with most other types of commonly observed mutations.

3. There are some phenotypes related to specific mutations such as the association of RARS with mutations in SF3B1.

Research agenda.

1. While it is believed that the various spliceosomal factor mutations must converge on a single target or pathway, this has not been clearly identified.

2. Further study of the molecular phenotypes of different mutations are needed, perhaps looking beyond the obvious effects on pre-mRNA splicing.

3. Recent data shows that MDS cells without known splicing factor mutations show aberrant splicing suggesting that the mutations may be amplifying a general miss-splicing phenotype. This could be a targetable mechanism for therapy.

4. Splicing defects can generate novel combinations of protein coding sequence that could be expressed as neoantigens. These might be of use in immune-therapy applications.