Synonymous Variants: Necessary Nuance in Our Understanding of Cancer Drivers and Treatment Outcomes

Abstract

Once called “silent mutations” and assumed to have no effect on protein structure and function, synonymous variants are now recognized to be drivers for some cancers. There have been significant advances in our understanding of the numerous mechanisms by which synonymous single nucleotide variants (sSNVs) can affect protein structure and function by affecting pre-mRNA splicing, mRNA expression, stability, folding, micro-RNA binding, translation kinetics, and co-translational folding. This review highlights the need for considering sSNVs in cancer biology to gain a better understanding of the genetic determinants of human cancers and to improve their diagnosis and treatment. We surveyed the literature for reports of sSNVs in cancer and found numerous studies on the consequences of sSNVs on gene function with supporting in vitro evidence. We also found reports of sSNVs that have statistically significant associations with specific cancer types but for which in vitro studies are lacking to support the reported associations. Additionally, we found reports of germline and somatic sSNVs that were observed in numerous clinical studies and for which in silico analysis predicts possible effects on gene function. We provide a review of these investigations and discuss necessary future studies to elucidate the mechanisms by which sSNVs disrupt protein function and play a role in tumorigeneses, cancer progression, and treatment efficacy. As splicing dysregulation is one of the most well-recognized mechanisms by which sSNVs impact protein function, we also include our own in silico analysis for predicting which sSNVs may disrupt pre-mRNA splicing.

Early understanding of protein structure and function assumed that the primary amino acid sequence contained all information relevant to correct protein folding. Consequently, synonymous single nucleotide variants (sSNVs) (1,2), which affect the coding sequence but do not alter the protein primary structure, were termed silent and deemed to be benign. However, studies from many disciplines (genetics, molecular evolution, biochemistry, and biophysics) have provided compelling evidence in the last few years that synonymous variants affect protein structure and function. Importantly, sSNVs have also been documented to affect more than 85 diseases.

Studies in the last decade demonstrate that the effects of sSNVs can occur at several steps in the transfer of information from the DNA to generate a correctly folded functional protein (Figure 1, A). sSNVs can affect micro-RNA (miRNA) binding and regulation of gene expression (3), pre-mRNA splicing (4-6), mRNA structure and stability (7,8), translation rate (9-12), co-translational folding (9-12), protein conformation (9-12), phosphorylation (13), protein function and localization (14), and cell fitness (15). [See (16-21) for more in-depth reviews.] Here, we focus on the impact sSNVs have in cancer biology and on patient-specific treatment outcomes.

Figure 1.

sSNVs disrupt several cellular processes. A) Examples of sSNVs observed in cancer are listed next to the process they are reported to affect. Green star indicates sSNV. A: sSNVs affect pre-mRNA splicing by changing recognition sites for splicing factors, ultimately decreasing expression of constitutively spliced mRNA tran scripts or creating alternatively spliced transcripts. B: sSNVs change mRNA structure and expression, increase mRNA decay rate, and can slow down or inhibit translation. C: sSNVs change miRNA: mRNA affinity and affect miRNA-mediated gene modulation. D: sSNVs change codon usage frequency and affect the rate of co-translational protein folding, ultimately resulting in misfolded protein. E: sSNVs change protein conformation, which can alter posttranslational modifications (PTM; eg, phosphorylation), affect binding to other proteins and ligands, and/or promote protein misfolding. These changes ultimately disrupt signaling pathways and cellular activities. B) Quantification of mechanisms by which sSNVs included in this review exert a biological impact indicated pre-mRNA splicing was the most common mechanism reported for disrupting protein function. However, this does not necessarily indicate that other cellular processes are not affected. Select variants had reports of more than 1 effect as indicated by the overlapping groups. Most of the variants that were discussed here with accompanying in vitro data indicate a change to at least 1 of the mechanisms of disrupting protein function that we have discussed and are illustrated in Figure 1, A. This figure was created with BioRender.com. miRNA = micro-RNA; sSNV = synonymous single nucleotide variant.

According to GLOBOCAN 2020, in the year 2020 there were an estimated 19.3 million new cancer cases and 10.0 million cancer deaths worldwide (22). Considering that 1 in 5 people will develop cancer (22), there is an increasing need to determine which genetic variants are pathogenic and for understanding the contribution of synonymous variants to cancer development. An analysis of synonymous variants in more than 3000 cancer samples (23) showed that synonymous variants tend to be enriched in oncogenes, which results in gain of exonic splicing enhancer (ESE) motifs, like binding sites for the splicing factor SRSF1 (also called SF2/ASF), and loss of exonic splicing silencer motifs, like binding sites for heterogenous nuclear ribonucleoprotein H2 splicing factor (23). Within oncogenes, approximately 6%-8% of all driver variants due to single nucleotide variants are synonymous, and these sSNVs often affect mRNA splicing (23). Identifying which variants are pathogenic will be useful for developing appropriate cancer-specific treatments for patients (24).

We compiled a group of 439 sSNVs from a literature search conducted for reports of sSNVs relevant to cancer. In this review, we discuss select examples from this list and how bioinformatics tools can predict the consequences of synonymous variants. Reported mechanisms by which sSNVs disrupt cellular processes associated with cancer are also discussed.

Using In Silico Methods to Predict Consequences of sSNVs

Experimentally determining the effects of all synonymous variants is impractical for several reasons. There may be more than one underlying mechanism for the observed effect. Experiments are time consuming and costly. There is no consensus for which methods should be used to detect effects of sSNVs. There are more than 21 million possible sSNVs, which could impact the human genome (coding sequences), and it is not feasible to screen them all. Therefore, bioinformatics tools are invaluable for predicting the effects of sSNVs. Zeng and Bromberg (25) reviewed the sSNV-specific prediction tools, SilVA (26), regSNPs-splicing (27), DDIG-SN (28), and IDSV (29), as well as a tool that assesses both sSNVs and intronic variants, TraP (30). The lack of experimentally validated deleterious sSNVs may diminish performance and accuracy of these machine learning–based methods (25). Recently, Tang et al. (31) published a new model for predicting deleterious synonymous variants, usDSM, which uses a larger training dataset and a more optimal sampling scheme than other machine-learning models and thus improved prediction accuracy.

Sharma et al. (32) created a database for synonymous mutations in cancer, SynMICdb, where they assigned scores based on several properties that are affected by synonymous nucleotide substitutions. To our knowledge, this is the only database that predicts functional impacts of synonymous cancer variants. Approximately 14% of sSNVs compiled for this review (Supplementary Data 1, available online) are found in SynMICdb.

sSNVs Affect mRNA Splicing

Common ways to detect splicing changes are by visualizing reverse transciption–polymerase chain reaction (RT-PCR) products from patient RNA or minigene samples on a gel (33) or by sequencing the cDNA from the patient’s RT-PCR product. Aberrant splicing is the most common mechanism we found for disruption of protein function by sSNVs (Figure 1, B). This does not necessarily indicate that splicing is the most common mechanism of disrupting protein function because of synonymous substitution but rather could indicate a reporting bias because of the relative ease of detecting aberrant splicing as opposed to other sSNV-induced changes that affect protein structure and function. Moreover, all mechanisms may co-occur with others that affect protein attributes.

The cancer-associated sSNVs reported to impact splicing are summarized in Table 1. Whereas in vitro assays are necessary to confirm whether sSNVs affect splicing, in silico prediction tools provide a rapid screen to identify sSNVs that warrant further experimental investigation. Based on this premise, we screened sSNVs found in human cancers for their effect on splicing using the in silico prediction tools (Supplementary Methods, available online) NNsplice (34), MaxEntScan (MES) (35), and ESEfinder (36). Of the 439 sSNVs analyzed, 79 (18%) were predicted to impact splicing based on changed NNsplice and MES scores (Figure 2, A and B; Supplementary Data 2, available online).

Table 1.

sSNVs with in vitro evidence of altering mRNA splicing

Genea	Variant	rs#	In silico predicted impactb,c	Observed splice impact	Observed diseases
APC	c.1458T>C	rs2229992	Exonic splice enhancer (ESEfinder) SynMICdb: 1.5419	Exon skipping [minigene (50)]	Familial adenomatous polyposis (50–52) Lynch syndrome (53) Breast cancer (54) Colorectal cancer (55,56)
APC	c.1548G>A	rs879254090	Constitutive donor (MES and NNsplice) Exonic splice enhancer (ESEfinder) SynMICdb: 6.8033	Exon skipping [patient sample (57)]	Familial adenomatous polyposis (57)
APC	c.1869G>T	n.a.	Exonic splice enhancer (ESEfinder)	Exon skipping [patient sample (58)]	Familial adenomatous polyposis (58)
APC	c.1956C>T	rs1064793716	Constitutive donor (MES)	Exon skipping [patient sample (59)]	Familial adenomatous polyposis (59)
APC	c.1957A>C	rs1114167580	Constitutive donor (MES and NNsplice) Exonic splice enhancer (ESEfinder)	Exon skipping [patient sample (59)]	Familial adenomatous polyposis (59)
BAP1	c.936T>G	n.a.	Exonic splice enhancer (ESEfinder) SynMICdb: 3.1316	Exon skipping [minigene (48)]	Clear cell renal cell carcinoma (48)
BRCA1	c.132C>Td	rs876658362	Constitutive donor (MES and NNsplice) Cryptic donor at c.130 (NNsplice)	Aberrant splicing [minigene (60)]	—
BRCA1	c.165G>A	rs1597900495	Exonic splice enhancer (ESEfinder)	Exon skipping [minigene (61)] Functionale (62)	Bronchioloalveolar adenocarcinoma (63)
BRCA1	c.231G>Td	rs80356847	Constitutive acceptor (NNsplice) Exonic splice enhancer (ESEfinder) SynMICdb: 5.1901	Exon skipping [minigene (64)]	Esophageal and gastric adenocarcinomas (65)
BRCA1	c.591C>T	rs1799965	Constitutive donor (MES) Cryptic acceptor at c.593 + 1 (NNsplice)	Exon skipping [patient sample (43–46)]	Breast and/or ovarian cancer (43–46,66–68)
BRCA1	c.693G>A	rs62625298	Exonic splice enhancer (ESEfinder)	Exon skipping [patient sample (69) and minigene (70)]	Breast cancer (66,69)
BRCA1	c.4185G>A	rs80356857	Constitutive donor (MES and NNsplice)	Exon skipping [patient sample (43,71)]	Breast and/or ovarian cancer (43,71)
BRCA1	c.4992C>Td	rs142459158	None	Exon skipping [cBROCA sequencing (72)]	Breast, ovary, endometrium, or colon cancer (72) Breast cancer (73,74)
BRCA1	c.5022C>T	rs786203868	Exonic splice enhancer (ESEfinder)	Exon skipping [cBROCA sequencing (72)]	Breast, ovary, endometrium, or colon cancer (72)
BRCA1	c.5073A>Td	rs80356853	Constitutive donor (MES and NNsplice)	Exon skipping [patient sample (75)]	Breast cancer (75)
BRCA2	c.516G>A	rs80359790	Constitutive donor (MES and NNsplice)	Exon skipping [patient sample and minigene (76)]	Breast cancer (76)
BRCA2	c.7992T>A	rs80359800	None	Exon skipping [cBROCA sequencing (72) and patient sample (77)]	Breast, ovary, endometrium, or colon cancer (72) Breast cancer (77)
BRCA2	c.8331G>A	rs80359802	Constitutive donor (MES and NNsplice)	Exon skipping [minigene (61)]	Breast or ovarian cancer (61)
BRCA2	c.9117G>A	rs28897756	Constitutive donor (MES and NNsplice)	Exon skipping [cBROCA sequencing (72)]	Breast, ovary, endometrium, or colon cancer (72)
BRCA2	c.9234C>T	rs587782428	Exonic splice enhancer (ESEfinder)	Exon skipping [patient sample (61)]	Breast or ovarian cancer (61)
MLH1	c.543C>G	rs1481129490	Constitutive donor (MES and NNsplice)	Partial exon skipping [patient sample and minigene (78)]	Lynch syndrome (78)
MLH1	c.543C>T	rs1481129490	Constitutive donor (MES and NNsplice) Cryptic donor at c.541 (NNsplice) SynMICdb: 4.9738	Exon skipping [patient sample (79)]	Lynch syndrome (79)
MLH1	c.882C>G	rs63751707	Exonic splice enhancer (ESEfinder)	Exon skipping [minigene (80)]	Lynch syndrome (80)
MLH1	c.882C>T	rs63751707	Constitutive donor (MES and NNsplice) Cryptic acceptor at c.884 + 1 (MES) Exonic splice enhancer (ESEfinder)	Exon skipping [minigene (78,80–82) and patient sample (82)]	Lynch syndrome (78,80–82)
MLH1	c.1038G>A	rs63751715	Constitutive donor (MES and NNsplice)	Aberrant splicing [patient sample (83)]	Lynch syndrome (83,84)
MLH1	c.1731G>A	rs63751657	Constitutive donor (MES) Exonic splice enhancer (ESEfinder)	Exon skipping [patient sample (82,85,86) and minigene [81, 82)]	Lynch syndrome (82,84–89)
MLH1	c.1896G>A	rs63751632	Constitutive donor (MES and NNsplice)	Aberrant splicing [patient sample (90)]	Lynch syndrome (90–94)
MLH1	c.1989G>A	rs63751662	Constitutive donor (MES and NNsplice)	Aberrant splicing (patient sample (95,96)]	Lynch syndrome (95,96)
MLH1	c.2103G>A	rs63750603	Constitutive donor (MES and NNsplice)	Aberrant splicing (patient sample and minigene (78)]	Lynch syndrome (78)
MSH2	c.942G>A	rs587779197	Constitutive donor (MES and NNsplice)	Aberrant splicing [patient sample (97)]	Lynch syndrome (97)
MSH2	c.1275A>G	rs63751650	Constitutive donor (MES)	Aberrant splicing [patient sample (86,96,98) and minigene (96)]	Lynch syndrome (86,96,98) Colorectal cancer (99,100)
MSH2	c.2634G>A	rs63751624	Constitutive donor (MES and NNsplice)	Exon skipping [patient sample (101)]	Lynch syndrome (101)
MSH6	c.3417C>T	rs876660283	Cryptic donor at c.3415 (MES and NNsplice) Exonic splice enhancer (ESEfinder)	Aberrant splicing [patient sample (102)]	Lynch syndrome (102)
PMS2	c.825A>G	rs876659736	Cryptic acceptor at c.826 (MES and NNsplice) Exonic splice enhancer (ESEfinder) SynMICdb: 2.8002	Aberrant splicing [patient sample (78,103) and minigene (78)]	Lynch syndrome (78,89,104) Breast cancer (103)
RB1	c.264G>A	n.a.	Constitutive donor (MES and NNsplice)	Exon skipping [patient sample (105)]	Retinoblastoma (105)
TP53	c.375G>A	rs55863639	Constitutive donor (MES and NNsplice) Exonic splice enhancer (ESEfinder) SynMICdb: 5.4652	Aberrant splicing [minigene (23,106,107)]	Li-Fraumeni syndrome (106,107)
TP53	c.375G>C	rs55863639	Constitutive donor (MES and NNsplice) Exonic splice enhancer (ESEfinder) SynMICdb: 6.4608	Aberrant splicing [minigene (23)]	Li-Fraumeni syndrome (108)
TP53	c.375G>T	rs55863639	Constitutive donor (MES and NNsplice) Exonic splice enhancer (ESEfinder) SynMICdb: 8.7687	Aberrant splicing [minigene (23,106)]	Li-Fraumeni syndrome (106)
TP53	c.561T>A	n.a.	Constitutive acceptor (NNsplice)	Aberrant splicing, [minigenef (38)]	—
TP53	c.561T>C	n.a.	Constitutive acceptor (MES)	Aberrant splicing, [minigenef (38)]	Skin cancer (109)
TP53	c.561T>G	n.a.	None	Aberrant splicing-nonneutral [minigenef (38)]	—
TP53	c.591G>A	rs1567551885	Exonic splice enhancer (ESEfinder)	Aberrant splicing-nonneutral [minigenef (38)]	Mucosa-associated lymphoid tissue lymphoma (110)
TP53	c.594A>G	n.a.	None	Aberrant splicing-nonneutral [minigenef (38)]	—
TP53	c.597A>T	n.a.	Cryptic donor c.595 (MES and NNsplice)	Aberrant splicing-nonneutral [minigenef (38)]	Breast cancer (111)
TP53	c.606T>G	n.a.	Exonic splice enhancer (ESEfinder)	Aberrant splicing-nonneutral [minigenef (38)]	—
TP53	c.612G>A	rs749629973	None	Aberrant splicing-nonneutral [minigenef (38)]	Nasal NK/T-cell lymphoma (112)
TP53	c.618G>A	rs142813240	None	Aberrant splicing [minigenef (38)]	Nasal NK/T-cell lymphoma (112) Myelodysplastic syndromes (113)
TP53	c.654G>T	n.a.	Exonic splice enhancer (ESEfinder)	Aberrant splicing-nonneutral [minigenef (38)]	Bladder cancer (114)
TP53	c.666G>A	rs72661118	None	Aberrant splicing [minigenef (38)]	Li-Fraumeni syndrome (108)
TP53	c.666G>C	n.a.	Exonic splice enhancer (ESEfinder)	Aberrant splicing-nonneutral [minigenef (38)]	Esophageal cancer (115)
TP53	c.672G>A	rs267605076	Constitutive donor (MES and NNsplice) SynMICdb: 5.3475	Aberrant splicing [minigenef (23,38,116) and patient sample (116)]	Li-Fraumeni syndrome (116)
TP53	c.993G>A	rs11575996	Constitutive donor (MES and NNsplice) SynMICdb: 4.8388	Aberrant splicing [minigene (23)]	Li-Fraumeni syndrome (117,118)
VHL	c.414A>G	rs869025648	Exonic splice enhancer (ESEfinder)	Exon skipping [patient sample (40,119,120) and minigene (40)]	von Hippel-Lindau disease (40,119,120)
VHL	c.429C>T	rs773556807	Exonic splice enhancer (ESEfinder)	Exon skipping [patient sample and minigene (40)]	von Hippel-Lindau disease (40)

^aAccession numbers: APC NM_000038.6 BAP1 NM_004656.4 BRCA1 NM_007294.4 BRCA2 NM_000059.4 MLH1 NM_000249.4 MSH2 NM_000251.3 MSH6 NM_000179.3 PMS2 NM_000535.7 RB1 NM_000321.3 TP53 NM_000546.6 VHL NM_000551.4. MES = MaxEntScan; n.a. = not available; NK = natural killer; sSNVs = single nucleotide variants. “—” indicates the variant was not reported to be observed in patients.

^bAll constitutive and cryptic splice sites near position of variant were scored using MES (35) and NNsplice (34) and either wild type or sSNV-included transcript sequence (see accession numbers above). Variants that decreased constitutive splice site score by 10% or more were predicted to impact a constitutive splice site. Variants that increased a cryptic splice site score to at least 90% of the nearest constitutive splice site’s score were predicted to impact a cryptic splice site. Predicted gains or losses of exonic splice enhancers (ESEs) were determined by differences in ESEfinder (36) results from wild type and sSNV-included transcript sequences. See Supplementary Methods and Supplementary Data 2 (available online) for more details.

^cWhen available for a variant, SynMICdb scores (32) are listed, which indicate predicted likelihood the variant has a functional impact among all synonymous variants in SynMICdb: >0.89 = top 50%; >2.70 = top 10%; >4.38 = top 1%; >5.83 = top 0.1%; >8.08 = top 0.01%.

^dRefer to Table 2 for changes to mRNA properties.

^eFindlay et al. reported functionality as determined by a cell survival assay (62).

^fBhagavatula et al. used cells transfected with minigenes from a library containing various TP53 sSNVs fused with EGFP and used FACS to determine changes in p53 splicing and considered at least a 30% change as nonneutral (38).

Figure 2.

In silico predictions of the likelihood of sSNVs to affect pre-mRNA splicing using NNSplice and MES. A) Of the 439 sSNVs analyzed, 79 were predicted and 360 were not predicted to change constitutive or cryptic splice sites. B) The number of sSNVs predicted to affect constitutive donor or acceptor sites and/or activate cryptic donor and acceptor sites are shown. C) Comparison of in silico predictions using NNsplice and MES and in vitro observations of splicing impact indicated in silico analysis agreed with in vitro data for approximately 80% of analyzed sSNVs with available in vitro data in the literature. True positive is defined as in silico analysis predicted splicing changes, and in vitro data indicated aberrant splicing; true negative is defined as in silico analysis predicted no splicing changes, and in vitro data indicated no splicing changes; false positive is defined as in silico analysis predicted splicing changes, but in vitro splicing data did not find splicing changes; false negative is defined as in silico analysis did not predict splicing changes, but in vitro splicing data indicated aberrant splicing. MES = MaxEntScan; sSNV = synonymous single nucleotide variant.

Approximately 80% of our in silico predictions based on NNsplice and MES were confirmed by previously published in vitro data (Figure 2, C; Supplementary Data 3, available online), which supports the continued use of these in silico methods for preliminary screening of deleterious sSNVs. Although ESEfinder predictions do not reliably predict which variants impact splicing (Supplementary Data 3, available online), it remains a useful tool in screening for potential impact to ESEs, especially when observed splice changes cannot be attributed to strengthening or weakening of a specific splice site.

TP53 codes for p53, a tumor suppressor that is commonly mutated in cancer (37). Synonymous variants are found at higher frequencies in TP53 than in other tumor suppressor genes (23) and are especially pronounced near splice sites (Table 1; Supplementary Figure 1, available online). Bhagavatula et al. (38) created a p53 protein reporter assay to determine the effects of synonymous variants in exon 6 on splicing. They found 9 variants throughout the exon that caused nonneutral aberrant splicing (Table 1).

Individuals with Von Hippel-Lindau (VHL) disease carrying variants in the VHL tumor-suppressor gene are at an increased risk of developing tumors in many organ systems, and the most frequent causes of death are complications arising from renal cell carcinoma (RCC), hemangioblastoma, and/or pheochromocytoma (39). Two VHL sSNVs, c.429C>T and c.414A>G, were observed to be linked to 2 phenotypes, erythrocytosis and VHL disease, respectively. Both variants cause exon skipping. However, the c.429C>T (erythrocytosis associated) variant had a less severe impact on splicing than the c.414A>G (VHL disease associated) variant, and Lenglet et al. (40) suggest this could be the reason for the different phenotype observations.

Breast cancer type 1 and 2 susceptibility proteins (BRCA1 and BRCA2) are involved with DNA repair and cell-cycle checkpoints, among many other functions (41). Variants in BRCA1 and BRCA2 are known to increase susceptibility to breast and ovarian cancers (42). Several sSNVs in the BRCA1 and BRCA2 genes are reported to cause exon skipping (Table 1). BRCA1 c.591C>T has been reported by many researchers to affect splicing (43–45). However, the truncated BRCA1 transcript still has tumor suppressor activity, and this variant is not significantly associated with the risk of developing cancer (46,47). Thus, not all sSNV-mediated splice events have clinical consequences.

BRCA1-associated protein 1 (BAP1) is an epigenetic modifier and major driver for clear cell RCC tumor development (48). BAP1 c.936T>G causes exon skipping. BAP1 transcripts from a minigene reporter assay containing this variant were also observed to have extensive mRNA degradation (48). The patient with this variant had very poor prognosis despite being a good candidate for immune checkpoint inhibitor therapy.

Families with Lynch syndrome (also known as hereditary nonpolyposis colorectal cancer) have variants in DNA mismatch repair genes, which increases the likelihood of developing a variety of cancers, especially of the colon (49). Several sSNV in mismatch repair genes—MutL homolog 1 (MLH1), MutS homolog 2 (MSH2), MutS homolog 6 (MSH6), and Postmeiotic segregation increased 2 (PMS2)—have been reported to create alternatively spliced products (Table 1).

sSNVs Affect miRNA Binding Sites

Although most studies of miRNA regulation have focused on miRNA binding to 3′ untranslated regions (UTRs) (121), there is evidence that miRNA binding to coding sequences can also alter expression, protein abundance, and mRNA degradation rates (122,123). Although studies show that the effect of miRNA binding within the coding region is not as high as in 3’UTR, it may still modulate protein expression (124,125). Thus, sSNVs may affect miRNA activity by disrupting or creating binding sites. Assessing the impact of sSNVs on miRNA binding in silico would require tools capable of accurately predicting miRNA targets within coding sequences. Although many tools have been developed to identify miRNA targets [eg, miRDB (126), TargetScan (127), miRanda and miRanda-mirSVR (128)], these were developed based on 3′UTRs and thus may not be sufficiently accurate to assess the impact of synonymous variants on miRNA binding. Further in vitro studies confirming predictions of miRNA binding affected by sSNVs are necessary.

One way the effect of a variant on miRNA binding can be assessed in vitro is by measuring changes in mRNA and protein expression in cells with miRNA overexpression or knockdown in comparison to cells without the variant (3). Another in vitro method of detection would be performing tandem affinity purification of the mRNA (with and without the variant) and the miRNA (129).

The BCL2L12 variant c.51C>T, found in melanoma tumors, causes the loss of the binding site for hsa-miR-671-5p to BCL2L12 mRNA (Supplementary Table 1, available online) and results in increased expression of BCL2L12 mRNA and protein. This variant also increases binding of BCL2L12 to p53 and diminishes p53-induced Mdm2 transcription and apoptosis in response to ultraviolet radiation exposure (3).

KIT is a receptor tyrosine kinase. The KIT sSNV c.2586G>C is located within the binding sites for hsa-miR-146a and hsa-miR-146b (130). This sSNV is predicted to modify how the miRNAs bind to KIT and prevent miRNA-mediated inhibition of KIT expression, which may contribute to the development of papillary thyroid carcinoma (130) (Supplementary Table 1, available online).

We initially planned to assess all sSNVs presented in this review for their potential impact on miRNA binding. However, we found that many tools were developed with a focus on 3′UTR binding, and we were not convinced they would yield reliable predictions for miRNA binding in the coding sequence. When using miRDB, we found zero predicted changes from any variant included in this review, including BCL2L12 c.51C>T, which was confirmed in vitro to inhibit binding of hsa-miR-671-5p.

We attempted to find miRNA binding prediction tools that were specifically developed for coding region of genes so we could assess the potential impact of sSNVs presented in this review. Because they were not originally developed to predict the impact of sSNVs, it is not surprising that the DIANA-microT-CDS (131) and MinoTar (132) web tools can provide results based on only a reference gene sequence. In developing their paccmit-cds tool, Marín et al. (123) demonstrated its higher precision than DIANA-microT-CDS and its higher sensitivity than MinoTar. Ultimately, we did not analyze the sSNVs using miRNA prediction tools because none could produce reliable, consistent results for changed miRNA binding in coding sequences. Improving computational tools specifically designed to assess the effect of sSNVs on miRNA-mRNA interactions remains a vital area of research.

Changes in mRNA Abundance or Properties From sSNVs Result in Disease

sSNVs can affect local mRNA structure, which can affect its stability, result in premature mRNA degradation, and decrease protein expression. Changes to mRNA structure can be detected using circular dichroism (133). In vitro methods to investigate changes in mRNA stability include transcription inhibition combined with quantitative RT-PCR and endoribonuclease footprinting (134). In silico methods to predict changes in mRNA structure include RNAfold (135), remuRNA (136), Kinefold (137), mFold (138), Vienna RNA (139), and MutaRNA (140). Variants that were previously confirmed to impact mRNA abundance or properties are summarized in Table 2. Because splice aberrations can influence mRNA abundance, results of our in silico analysis with MES and NNsplice are also included.

Table 2.

Variants affecting mRNA expression or structure

Genea	Variant	rs#	In silico predicted impactb	In vitro observations	Observed diseases
BRCA1	c.132C>Tc	rs876658362	Constitutive donor (MES and NNsplice) Cryptic donor at c.130 (NNsplice)	Decreased mRNA expression; loss of functiond (62)	—
BRCA1	c.211A>C	rs80357382	Constitutive donor (MES and NNsplice)	Decreased mRNA expression; loss of functiond (62)	—
BRCA1	c.231G>Tc	rs80356847	Constitutive acceptor (NNsplice) SynMICdb: 5.1901	Decreased mRNA expression; loss of functiond (62)	Esophageal and gastric adenocarcinomas (65)
BRCA1	c.4938C>A	rs2052371169	Cryptic donor at c.4935 (NNsplice)	Decreased mRNA expression; loss of functiond (62)	—
BRCA1	c.4992C>Tc	rs142459158	None	Decreased mRNA expression; possible loss of functiond (62)	Breast cancer (73,74) Breast, ovary, endometrium, or colon cancer (72)
BRCA1	c.4995G>A	rs1412152746	None	Decreased mRNA expression; possible loss of functiond (62)	—
BRCA1	c.4998C>T	rs730882165	None	Decreased mRNA expression; possible loss of functiond (62)	—
BRCA1	c.5007C>T	rs751856943	None	Decreased mRNA expression; possible loss of functiond (62)	—
BRCA1	c.5073A>C	rs80356853	Constitutive donor (MES and NNsplice) SynMICdb: 3.7985	Decreased mRNA expression; loss of functiond (62)	Gastric adenocarcinoma (65)
BRCA1	c.5073A>Tc	rs80356853	Constitutive donor (MES and NNsplice)	Decreased mRNA expression; loss of functiond (62)	Breast cancer (75)
BRCA1	c.5073A>G	rs80356853	Constitutive donor (MES and NNsplice)	Decreased mRNA expression; functionald (62)	—
BRCA1	c.5079T>A	rs2051936328	None	Decreased mRNA expression; possible loss of functiond (62)	—
BRCA1	c.5127A>G	rs2051916034	None	Decreased mRNA expression; loss of functiond (62)	—
BRCA1	c.5130A>G	rs2051914890	None	Decreased mRNA expression; loss of functiond (62)	—
BRCA1	c.5277G>A	rs80356854	Constitutive donor (MES and NNsplice)	Decreased mRNA expression; functionald (62)	—
BRCA1	c.5406A>C	rs879255493	None	Decreased mRNA expression; functionald (62)	—
BRCA1	c.5430G>C	rs786201582	Cryptic acceptor at c.5438 (NNsplice)	Decreased mRNA expression; loss of functionc (62)	—
BRCA1	c.5469A>C	rs1597797238	None	Decreased mRNA expression; functionald (62)	—
BRCA1	c.5469A>T	rs1597797238	None	Decreased mRNA expression; functionald (62)	—
BRCA1	c.5475G>T	rs1057520941	None	Decreased mRNA expression; possible loss of functiond (62)	—
BRCA1	c.5550G>A	rs786201502	None	Decreased mRNA expression; possible loss of functiond (62)	—
BRCA1	c.5553C>T	rs80357326	None	Increased mRNA expression; possible loss of functiond (62)	—
ERCC1	c.354T>C f, g	rs11615	None	T allele-increases mRNA expression (146,147) No change in mRNA expression (148)	Colorectal carcinoma (146)
IDH1	c.315C>T g	rs11554137	None	Increases mRNA expression (149,150) No change in mRNA expression (151)	Gliomas (150,152–154) Acute myeloid leukemia (149,151,155–157)
KRAS	c.30A>C f	n.a.	SynMICdb: 6.091	Affects mRNA secondary structure (32)	Hepatocellular carcinoma (65)
KRAS	c.36T>C f	n.a.	SynMICdb: 6.0684	Increases mRNA expression (32)	Lung cancer (158,159)
KRAS	c.39C>Af	rs397517040	None	Increases mRNA expression (32)	Colorectal adenocarcinoma (65)
PDGFRA	c.2472C>Tf, g	rs2228230	SynMICdb: 3.7781	Decreased mRNA stability (160) Predicted to affect splicing by creation of ESS (161) Predicted to affect mRNA secondary structure (161)	Colorectal cancer (161,162) Cholangiocarcinoma (163) Renal cell carcinoma (164) Merkel cell carcinoma (skin cancer) (165) Acral melanoma (160) Adenosquamous carcinoma (166) Malignant solitary fibrous tumors (167) Neuroendocrine tumor (168) Peripheral nerve sheath tumors (169) Glioma (170) Core binding factor leukemia (171) Neuroblastoma (172)
RET	c.1890C>T	rs781145070	None	Increased mRNA expression (141)	—
TP53	c.30C>Te	rs568171603	SynMICdb: 0.6565	Affects mRNA binding to HDM2 (144)	Hepatocellular carcinoma (173) Dukes’ C colorectal carcinoma (174)
TP53	c.66A>G	rs748527030	None	Affects mRNA binding to HDM2 (142)	Chronic lymphocytic leukemia (175)
UGT1A6	c.627G>T	rs17863783	SynMICdb: -1.7254	Increases mRNA expression (176)	Decreases risk of bladder cancer (176)

^aAccession numbers: BRCA1 NM_007294.4 ERCC1 NM_202001.3 IDH1 NM_005896.4 KRAS NM_033360.4 PDGFRA NM_006206.6 RET NM_020975.6 TP53 NM_000546.6 UGT1A6 NM_001072.4. ESS = exonic splicing silencer; MES= MaxEntScan; n.a. = not available; sSNV = synonymous single nucleotide variant. “—” indicates the variant was not reported to be observed in patients.

^bAll constitutive and cryptic splice sites near position of variant were scored using MES (35) and NNsplice (34) and either wild-type or sSNV-included transcript sequence (see accession numbers above). Variants that decreased constitutive splice site score by 10% or more were predicted to impact a constitutive splice site. Variants that increased a cryptic splice site score to at least 90% of the nearest constitutive splice site’s score were predicted to impact a cryptic splice site. Predicted gains or losses of exonic splice enhancers (ESEs) were determined by differences in ESEfinder (36) results from wild type and sSNV-included transcript sequences. See Supplementary Methods (available online) and Supplementary Data 2 (available online) for more details. When available for a variant, SynMICdb scores (32) are listed, which indicate predicted likelihood the variant has a functional impact among all synonymous variants in SynMICdb: >0.89 = top 50%; >2.70 = top 10%; >4.38 = top 1%; >5.83 = top 0.1%; >8.08 = top 0.01%.

^cRefer to Table 1 for changes to pre-mRNA splicing.

^dFindlay et al. reported functionality as determined by a cell survival assay (62).

^eRefer to Table 3 for changes to protein properties.

^fRefer to Table 6 for effects on cancer prognosis and pharmacogenomics.

Findlay et al. (62) assessed the effects of all possible sSNVs in 13 BRCA1 exons. Several variants were characterized as loss of function or possible loss of function, many of which resulted in decreased mRNA expression (Table 2) (62). In addition to decreased mRNA expression, changes in BRCA1 splicing could also contribute to the observed loss of function of BRCA1 in cells with these variants as predicted by NNsplice and MES (Table 2). Interestingly, 5 sSNVs exhibited decreased mRNA expression and were still found to be functional based on cell survival (62). This comprehensive study demonstrates that sSNVs can mediate changes in mRNA expression; however, changed mRNA levels do not necessarily result in functional consequences.

Variants in RET, a proto-oncogene, can lead to medullary thyroid carcinoma. sSNV c.1890C>T, found in a patient with medullary thyroid carcinoma, was predicted to create a new ESE site. This variant was initially considered clinically irrelevant because it is synonymous. However, a subsequent minigene expression assay demonstrated that this variant resulted in increased mRNA and protein expression of the constitutive exon encoding the functional domain of the RET receptor (141). RNA immunoprecipitation experiments supported the prediction that this variant created new ESE motifs, which increased recruitment of splicing factors and lead to an increased abundance of correctly spliced mature mRNA (141).

TP53 c.66A>G is found in the internal ribosome entry sequence (IRES) hairpin structure of p53 and affects binding of the p53 regulators HDM2 and HDMX to p53 mRNA (142). Under normal conditions, the E3 ubiquitin ligase HDM2 targets p53 protein for 26S proteasomal degradation. In response to DNA damage, HDM2 binds to TP53 mRNA and mediates ataxia telangiectasia mutated (ATM) phosphorylation of the nascent p53 protein at Ser15, which ultimately stabilizes p53 because HDM2 does not bind to phosphorylated p53 and does not target it for 26S proteasomal degradation (13). HDMX binds TP53 mRNA and supports p53-HDM2 interaction (142). The TP53 c.66A>G sSNV results in p53 degradation even in the event of genotoxic threat (13,142). HDMX and HDM2 will not bind TP53 mRNA that carries this variant, as it changes the IRES conformation (making it unfavorable for binding IRES-interacting factors) and causes reduction in IRES activity (142,143). Similarly, TP53 c.30C>T reduces binding of HDM2 to TP53 mRNA in a co-immunoprecipitation assay and reduces the tumor suppressor function of p53 (144). C.30C>T may also alter TP53 translation kinetics because of the introduction of rare codons (21,145) (Table 3). Such changes in translation kinetics have been shown to affect protein structure and function (see next section).

Table 3.

Variants affecting protein expression, structure, or function

Genea	Variant	rs#	Codon change	Observationsb	Observed diseases	Codon usage per 1000c	RSCU
ERCC1	c.354T>Cd,e	rs11615	AAT> AAC	T allele increases protein expression (146)	Colorectal carcinoma (146)	15.07 > 20.58	0.85 > 1.15
ESR1	c.261G>C	rs746432	GCG> GCC	Affects ERα transactivation efficiency (184,185)	Breast cancer (188)	7.22 > 29.64	0.39 > 1.61
KRAS	c.30A>Cd	n.a.	GGA> GGC	Increases protein expression (32)	Hepatocellular carcinoma (65)	15.52 > 24.01	0.95 > 1.47
KRAS	c.36T>Cd	n.a.	GGT> GGC	Increases protein expression (32)	Lung cancer (158,159)	13.33 > 26.09	0.74 > 1.45
KRAS	c.36T>G	n.a.	GGT> GGG	Decreases protein expression (32) SynMICdb: 5.6839	Pancreatic cancer (189)	15.50 > 16.60	0.80 > 0.86
KRAS	c.39C>Ad	rs397517040	GGC> GGA	Increases protein expression (32)	Colorectal adenocarcinoma (65)	26.80 > 15.80	1.49 > 0.88
KRAS	c.39C>G	n.a.	GGC> GGG	Decreases protein expression (not mRNA level) (32)	Endometrial carcinoma (190)	26.16 > 16.39	1.53 > 0.92
PDGFRA	c.2472C>Td,e	rs2228230	GTC> GTT	Decreased protein expression (160) Altered signaling via MAPK and PI3K/AKT pathways (160) SynMICdb: 3.7781	Colorectal cancer (161,162)	15.87 > 10.83	0.99 > 0.67
					Cholangiocarcinoma (163)	15.68 > 10.52	0.99 > 0.66
RET	c.1827C>T	rs377767396	TGC> TGT	Less frequent codon—predicted to decrease translational speed (191)	Medullary thyroid carcinoma (191)	12.04 > 9.06	1.14 > 0.86
RET	c.2364C>T	rs1554819400	ATC> ATT	Less frequent codon—predicted to decrease translational speed (191)	Medullary thyroid carcinoma (191)	23.17 > 15.69	1.57 > 1.06
RET	c.2418C>T	rs553418132	TAC> TAT	Less frequent codon—predicted to decrease translational speed (191) SynMICdb: 2.8342	Medullary thyroid carcinoma (191)	16.55 > 11.39	1.18 > 0.82
RET	c.2673G>A	rs201620214	TCG> TCA	More frequent codon—predicted to increase translational speed (191)	Medullary thyroid carcinoma (191)	4.43 > 9.59	0.36 > 0.78
TP53	c.30C>Td	rs568171603	GTC> GTT	Reduces binding of HDM2/affects cell signaling (144) SynMICdb: 0.6565	Hepatocellular carcinoma (173)	15.87 > 10.83	0.99 > 0.67
					Dukes’ C colorectal carcinoma (174)	15.73 > 11.5	0.96 > 0.70

^aAccession numbers: ERCC1 NM_202001.3 ESR1 NM_001122740.1 KRAS NM_033360.4 PDGFRA NM_006206.6 RET NM_020975.6 TP53 NM_000546.6. ER = estrogen receptor; n.a. = not available; sSNV = synonymous single nucleotide variant.

^bMaxEntScan (MES) and NNsplice did not predict any changes in splicing for these sSNVs. When available for a variant, SynMICdb scores (32) are listed, which indicate predicted likelihood the variant has a functional impact among all synonymous variants in SynMICdb: >0.89 = top 50%; >2.70 = top 10%; >4.38 = top 1%; >5.83 = top 0.1%; >8.08 = top 0.01%.

^cCodon usage and relative synonymous codon usage (RSCU) were computed as previously described—CancerCoCoPUTs (145).

^dRefer to Table 2 for changes to mRNA properties.

^eRefer to Table 6 for effects on cancer prognosis and pharmacogenomics.

sSNVs Disrupt Co-translational Folding and Result in Altered Protein Expression

In addition to affecting mRNA structure, sSNVs can also affect protein conformation and structure because of alterations in translation rate arising from differences in codon usage between the variant and wild-type codons. Synonymous variants in cancer often result in a more preferred or higher frequency codon (177). This change is advantageous for cell growth, as preferred codons allow for a faster rate of translation (177). Protein translation dynamics are influenced by mRNA structure and by the relationship between codon usage frequencies and the available transfer RNA (tRNA) pool. One way to detect changes in translation speed because of variants is by ribosome profiling (178). The interactions between tRNA availability, codon usage, and synonymous substitutions in cancer continue to be under investigation (145). Table 3 summarizes sSNVs detected in human cancers that alter codon usage, impact protein expression, or alter binding to other signaling molecules.

Increased expression of the proto-oncogene KRAS is associated with some cancers, and low KRAS expression is critical for decreasing the likelihood of developing cancer (179,180). Rare codons attenuate KRAS transcription and translation efficiency. Wild type KRAS is rich in rare codons and has poor translation efficiency. When rare codons were replaced with common codons, transcriptional rates were increased because of histone modifications and recruitment of transcriptional coactivators (181,182) increasing the risk of cancers. Previous studies demonstrated that the KRAS protein structure was also affected by sSNVs that change codon usage (182), and several KRAS sSNVs have been reported to modulate KRAS protein expression [Table 3; Supplementary Table 2, available online; (145)].

Estrogen receptor alpha (ERα) is a transcription factor (encoded by ESR1) that translocates into the nucleus upon activation and triggers signaling pathways that activate other transcription factors. Variants in ERα play a substantial role in the development of breast cancer (183). ESR1 c.261G>C prevents nuclear translocation of the ERα protein, although its mRNA expression level does not change. In an in vitro assay, c.261G>C prevented ERα from binding and activating an estrogen response element reporter gene (184,185). The tRNAs that pair to the GCC codon have a gene copy number that is 5 times higher than tRNAs pairing with the GCG codon (186). By eliminating a rare codon, this variant could affect ERα conformation by altering translational velocity and impacting co-translational folding. This change in conformation could in turn affect ligand-inducible activation of ERα as protein structure is strongly linked to protein function (187).

sSNVs Found in Cancer With No Reported Underlying Mechanism

In our literature search, we found numerous sSNVs that are observed in cancer that are predicted to affect protein function (Table 4) or have a statistically significant association with cancer (Table 5) for which in vitro mechanistic investigations are lacking. Selected variants are discussed here.

Table 4.

sSNVs predicted to affect pre-mRNA splicing or have a functional impact

Genea	Variant	rs#	Observed diseases and predicted mechanisms	In silico predicted impactb
APC	c.450A>G	rs116020626	Familial adenomatous polyposis (51)	Cryptic acceptor at c.451 (MES and NNsplice)
APC	c.1005A>G	rs3797704	Breast cancer—no observed association (54) Colorectal cancer (215)	SynMICdb: 2.6099
APC	c.1635G>A	rs351771	Colorectal cancer (55,56) Familial adenomatous polyposis (51) Breast cancer—no observed association (54)	Constitutive acceptor (NNsplice)
APC	c.2205G>A	rs141001261	Familial adenomatous polyposis (51)	SynMICdb: 4.1843
APC	c.4479G>A	rs41115	Craniopharyngioma (216) Breast cancer—no observed association (54) Colorectal cancer (55,56,217) Ependymoma (218) Lynch syndrome (53) Familial adenomatous polyposisc (50,52) Predicted to alter splicing—ESRc (50,52)	Cryptic acceptor at c.4481 (MES and NNsplice)
APC	c.5034G>A	rs42427	Colorectal cancer (55,56) Familial adenomatous polyposis (50–52) Lynch syndrome (53) Predicted frame shifting—ESR (50)	None
APC	c.5268T>G	rs866006	Colorectal cancer (55,56) Lynch syndrome (53) Familial adenomatous polyposis (50–52) Predicted frame shifting—ESR (50)	None
APC	c.5880G>A	rs465899	Colorectal cancer (55,56) Lynch syndrome (53) Familial adenomatous polyposis (50,51) Brain metastasis from lung carcinoma (219) Predicted frame shifting—ESR (50)	None
BRCA1	c.63C>T	rs1555600921	Loss of functiond (62) Lung adenocarcinoma (63)	SynMICdb: 4.5287
BRCA1	c.213G>A	rs1441240938	Functionald (62)	Constitutive acceptor (MES and NNsplice)
BRCA1	c.261G>A	rs757971617	Loss of functiond (62)	None
BRCA1	c.1878A>G	rs8176154	Breast and ovarian cancer (220)	Cryptic donor at c.1878 (NNsplice)
BRCA1	c.4920A>C	rs2052378025	Possible loss of functiond (62) Lung adenocarcinoma (221)	SynMICdb: 1.464
BRCA1	c.4938C>T	rs2052371169	Possible loss of functiond (62) Lung adenocarcinoma (65)	SynMICdb: 2.3857
BRCA1	c.5076T>C	rs2051937547	Functionald (62)	Constitutive acceptor (MES)
BRCA1	c.5193G>A	rs876660702	Functionald (62)	Constitutive donor (MES)
BRCA1	c.5280C>T	rs750040616	Functionald (62)	Cryptic acceptor at c.5286 (NNsplice)
BRCA1	c.5331A>T	rs1411246255	Functionald (62)	Constitutive donor (MES)
BRCA1	c.5409T>A	rs2050995166	Functionald (62)	Constitutive acceptor (NNsplice)
BRCA1	c.5409T>G	rs2050995166	Functionald (62)	Constitutive acceptor (NNsplice)
BRCA2	c.9117G>C	rs28897756	Breast or ovarian cancer (222)	Constitutive Donor (MES and NNsplice)
EGFR	c.1509C>T	rs17336800	Oral squamous cell carcinoma (223)	Cryptic donor at c.1507 (MES and NNsplice)
EGFR	c.2361G>Ae	rs1050171	Thyroid cancer (carcinoma showing thymus-like elements) (224) Cervical adenosquamous carcinoma (166) Non-small cell lung carcinoma (225) Adamantinomatous craniopharyngioma (216) Cutaneous squamous cell carcinoma (226) Ependymoma (218) Colorectal cancer (227) Lynch syndrome (53) Oropharyngeal cancer (228) Lung squamous cell carcinoma (229) Lung adenocarcinoma—no observed association (230) Colorectal cancer—no observed association (231)	SynMICdb: 2.0095
EGFR	c.2982C>T	rs2293347	Lung cancer (232)	Cryptic acceptor at c.2981 (NNsplice) SynMICdb: 1.7773
ERCC2	c.468A>C	rs238406	A allele—esophageal cancer (192) A allele—bladder cancer (193) A allele—basal cell carcinoma (195) A allele—lung cancer (194) C allele—colorectal cancer (196)	Cryptic acceptor at c.477 + 6 (NNsplice)
FGFR3	c.1644C>T	rs756321317	Lynch syndrome (53)	Constitutive donor (MES and NNsplice)
FGFR3	c.1647G>T	rs3135897	Lynch syndrome (53)	Cryptic donor at c.1652 (NNsplice) SynMICdb: 3.3413
FLT3	c.1776T>C	n.a.	Ependymoma (218)	Cryptic acceptor at c.1785 (MES)
KIT	c.2394C>T	rs55789615	Neuroblastoma (172) Gonadoblastoma and dysgerminoma (233) Merkel cell carcinoma (skin cancer) (165)	SynMICdb: 4.9301
KRAS	c.36T>A	n.a.	Metastatic colorectal cancer (234) Predicted to be Pathogenic (234)	None
KRAS	c.39C>T	rs397517040	Metastatic colorectal cancer (234) Predicted to be Pathogenic (234)	SynMICdb: 4.357
MET	c.534C>T	rs35775721	Craniopharyngioma (216)	Cryptic donor at c.528 (NNsplice)
MSH2	c.1746C>T	rs786201486	Breast or ovarian cancer (235)	Cryptic acceptor at c.1759 + 1 (NNsplice)
MSH2	c.1863A>T	rs786203119	Ovarian cancer (236)	Cryptic donor at c.1857 (NNsplice)
MSH3	c.2511G>A	rs149628160	Colorectal cancer (55)	Cryptic acceptor at c.2513 (MES and NNsplice)
MSH6	c.1401C>T	rs1558661556	Lynch syndrome (88)	Cryptic acceptor at c.1412 (NNsplice)
PDGFRA	c.939T>G	rs4358459	Peripheral nerve sheath tumors (169) Renal cell carcinoma (164)	SynMICdb: 4.5806
PDGFRA	c.2481A>T	n.a.	Colorectal cancer (161) Predicted changes to splicing and mRNA secondary structure (161)	None
PDGFRA	c.2496G>A	n.a.	Colorectal cancer (161) Predicted changes to splicing and mRNA secondary structure (161)	None
PDGFRA	c.2514C>T	n.a.	Colorectal cancer (161) Predicted changes to splicing and mRNA secondary structure (161)	None
PDGFRA	c.2517G>T	rs1213039385	Colorectal cancer (161) Predicted splice changes (161)	SynMICdb: 4.0463
PDGFRA	c.2520C>A	n.a.	Colorectal cancer (161) Predicted changes to splicing and mRNA secondary structure (161)	None
PMS2	c.30A>G	rs876660608	Lynch syndrome (88)	Cryptic acceptor at c.29 (NNsplice)
POLE	c.1323G>A	rs116573514	Endometrial cancer; predicted pathogenic (237,238)	None
RET	c.2307G>T	rs1800861	Craniopharyngioma tumor (216) Lynch syndrome (53) Ependymoma (218)	SynMICdb: 4.7067
RET	c.2712C>G	rs1800863	Medullary thyroid carcinoma (239)	Cryptic donor at c.2711 (MES and NNsplice)
TP53	c.108G>A	rs1800370	Breast cancer (198,240) Li-Fraumeni syndrome-osteosarcoma (108) Chronic lymphocytic leukemia (199)	Constitutive acceptor (NNsplice)
TP53	c.108G>T	n.a.	Skin cancer (109)	Constitutive acceptor (NNsplice)
TP53	c.333G>A	rs1465965835	Skin cancer (109)	Cryptic acceptor at c.335 (NNsplice)
TP53	c.465C>T	n.a.	Breast cancer (197)	SynMICdb: 3.428
TP53	c.816G>A	rs756421198	Breast cancer (197)	Cryptic donor at c.813 (MES and NNsplice)
TP53	c.894G>A	rs756123992	Breast cancer (197)	Cryptic acceptor at c.905 (MES)
TP53	c.900C>T	rs767356182	Breast cancer (197)	Cryptic acceptor at c.905 (NNsplice)

^aAccession numbers: APC NM_000038.6 BRCA1 NM_007294.4 BRCA2 NM_000059.4 EGFR NM_005228.5 ERCC2 NM_000400.4 FGFR3 NM_000142.4 FLT3 NM_004119.3 KIT NM_000222.3 KRAS NM_033360.4 MET NM_00112750.3 MSH2 NM_000251.3 MSH3 NM_002439.5 MSH6 NM_000179.3 PDGFRA NM_006206.6 PMS2 NM_000535.7 POLE NM_006231.4 RET NM_020975.6 TP53 NM_000546.6. MES = MaxEntScan; n.a. = not available; sSNV = synonymous single nucleotide variant.

^bAll constitutive and cryptic splice sites near position of variant were scored using MES (35) and NNsplice [34] and either wild type or sSNV-included transcript sequence (see accession numbers above). Variants that decreased constitutive splice site score by 10% or more were predicted to impact a constitutive splice site. Variants that increased a cryptic splice site score to at least 90% of the nearest constitutive splice site’s score were predicted to impact a cryptic splice site. Predicted gains or losses of Exonic Splice Enhancers (ESEs) were determined by differences in ESEfinder [36] results from wild type and sSNV-included transcript sequences. See Supplementary Methods and Supplementary Data 2 (available online) for more details. When available for a variant, SynMICdb scores (32) are listed, which indicate predicted likelihood the variant has a functional impact among all synonymous variants in SynMICdb: >0.89 = top 50%; >2.70 = top 10%; >4.38 = top 1%; >5.83 = top 0.1%; >8.08 = top 0.01%.

^cLiu et al. identify APC c.4479G>A as c.4425G>A NM_001127511.3 (50).

^dFindlay et al. reported functionality as determined by a cell survival assay (62).

^eRefer to Table 6 for effects on cancer prognosis and pharmacogenomics.

Table 5.

sSNVs with reported associations to cancer

Genea	Variant	rs#	Observed diseases and in silico predicted impactb
ESR1	c.30T>Cc	rs2077647	C allele reduces distant recurrence risk of breast cancer (204) C allele protective against breast cancer (203) C allele protective against endometrial cancer (201) C allele protective against bladder cancer (202) C allele increases risk-renal cell carcinoma (205) C allele increases risk of colon cancer (202) Prostate cancer—no statistically significant association (241) Breast cancer (188) SynMICdb: 3.6756
ESR1	c.975G>C	rs1801132	G allele protective against breast, colorectal, and bladder cancers; hepatocellular carcinoma; and acute myeloid leukemia (202) G allele—decreased breast cancer risk in meta-analysis (206) G allele—protective against lymph node metastasis in breast cancer (203) Prostate cancer—no statistically significant association (241) G allele—associated with family history of breast cancer (188) SynMICdb: 0.8289
ESR1	c.1782G>Ac	rs2228480	G allele—breast cancer (202, 207, 208) Breast cancer (188) SynMICdb: -0.0514
HABP2	c.183C>T	rs3740530	C allele—increased risk of papillary thyroid carcinoma (210)
HRAS	c.81T>C	rs12628	Ependymoma (218) Cutaneous squamous cell carcinoma (226) Chronic myeloid leukemia (242) Bladder cancer (243) Melanoma (244) Gastric cancer (245) Cancer—statistically significant association (meta-analysis) (246) Colorectal cancer (162) SynMICdb: 2.8013
NLRP6	c.2229G>A	rs11246050	A allele—increased risk of papillary thyroid carcinoma (210)
TERT	c.915G>A	rs2736098	Increased risk of papillary thyroid carcinoma (210) Predicts poor prognosis in renal cell carcinoma (211) Lynch syndrome (53) Lung cancer (247) Increases cancer risk (213, 214) Increased risk of lung and bladder cancers (212) SynMICdb: -2.0727
TERT	c.3039C>T	rs33954691	Increased risk of radioiodine-refractory papillary thyroid carcinoma (210)
TLR2	c.597T>C	rs3804099	T allele—colon cancer (248) T allele—gastric cancer (249) T allele—hepatocellular carcinoma (250) C allele—decreases risk of cancer (251) Predicted to affect splicing (248)
TP53	c.639A>G	rs1800372	Skin cancer (109) Li-Fraumeni syndrome-osteosarcoma (108) Non-small cell lung cancer—statistically significant association (200) Myelodysplastic syndromes (113) Chronic lymphocytic leukemia; marker of unfavorable prognosis of disease (199) Breast cancer (111, 197, 198, 240, 252) No splicing changes [minigened (38)]

^aAccession numbers: ESR1 NM_001122740.1 HABP2 NM_004132.5 HRAS NM_005343.4 NLRP6 NM_138329.2 TERT NM_198253.3 TLR2 NM_001318787.2 TP53 NM_000546.6. MES = MaxEntScan; sSNVs = synonymous single nucleotide variants.

^bRefer to Table 1 legend for details on SynMICdb scores (32). MES and NNsplice did not predict any changes in splicing for these sSNVs.

^cRefer to Table 6 for effects on cancer prognosis and pharmacogenomics.

^dBhagavatula et al. used cells transfected with minigenes from a library containing various TP53 sSNVs fused with EGFP and used FACS to determine changes in p53 splicing and considered at least a 30% change as nonneutral (38).

Excision repair cross-complementation group 2 (ERCC2, also known as XPD) is involved in nucleotide excision repair. The 2 alleles of the sSNV c.468A>C are associated with different cancers. The A allele has a statistically significant association with esophageal cancer (192), bladder cancer (193), lung cancer (194), and basal cell carcinoma (195). However, the C allele was found to have a statistically significant association with colorectal cancer (196). Our in silico analysis predicts the C allele may activate a cryptic acceptor site, leading to aberrant ERCC2 splicing.

A variety of studies observe several TP53 sSNVs (109,115,197,198), some of which have been predicted to be disease causing by Mutation Taster (115). Our in silico analysis predicted that several TP53 sSNVs will affect splicing and may promote expression of TP53 alternative splice products, resulting in diminished abundance of functional p53 protein (Table 4). In chronic lymphocytic leukemia, carriers of TP53 c.639A>G had shorter median overall survival compared with patients without TP53 variants (199). C.639A>G was reported to have a statistically significant association with non-small cell lung cancer (NSCLC) (200). This variant does not induce changes in splicing (38); therefore, other mechanisms should be studied.

Findlay et al. (62) reported several BRCA1 sSNVs to be functional according to their in vitro reporter assay based on cell survival. However, some of these functional sSNVs may still affect splicing (for example, c.165G>A) (Table 1). Other functional sSNVs are predicted to affect splicing according to MES and NNsplice (Table 4). It is possible that although the cells in their assay are functional, there are modest splicing changes that could affect BRCA1 structure, function, or abundance.

In a study of APC in breast cancer patients, the sSNVs c.1005A>G, c.1635G>A, and c.4479G>A were found to occur in breast cancer patients but not at higher frequencies than they occurred in healthy patient controls (54). However, these variants are predicted to affect splicing (our analysis) or cellular function (32) (Table 4).

Roodi et al. (188) were the first to report the occurrence of ESR1 c.30T>C, c.975G>C, and c.1782G>A (we report these variants as they appear in dbSNP [rs2077647, rs1801132, and rs2228480]) in breast cancer. However, they found a statistically significant association only with the G allele of c.975G>C to familial history of breast cancer. Since then, others have also studied the possible contribution of these variants to breast cancer. The C allele of c.30T>C was found to be protective against cancer (201–203) and to reduce the distant recurrence risk of breast cancer (204). However, the C allele has also been reported to increase the risk of other cancers (202,205).

ESR1 c.975G>C is located within the hormone-binding domain of ESR1, which is important for receptor dimerization, chaperone binding, and recruitment of coregulators transcription (202). Although this variant site is a potential target for ESEs sc35 and sf2 [202], it has not been confirmed to impact splicing in vitro. Anghel et al. (202) found the G allele to be protective against breast, colorectal, and bladder cancers; hepatocellular carcinoma; and acute myeloid leukemia (AML). A separate meta-analysis also reported the G allele decreased the risk of breast cancer (206). The G allele is reported to be protective against lymph node metastasis in breast cancer (203).

ESR1 c.1782G>A is located in the ligand-dependent transactivation domain TAF-2 within the F structural domain of ERα. Several studies have reported that the G allele is associated with increased risk of breast cancer (202,207,208). However, the precise effect of this sSNV on ERα structure and function is not known yet, and Anghel et al. (202) suggest it may recruit coregulators.

TERT encodes for telomerase reverse transcriptase, the catalytic subunit of DNA polymerase (209). TERT c.915G>A has been associated with an increased risk of papillary thyroid carcinoma (210), worsened prognosis in renal cell carcinoma (211), and increased risk of lung and bladder cancers (212). A meta-analysis reported c.915G>A had a statistically significant association with increased cancer risk (213,214).

sSNVs Affect Cancer Prognosis and Pharmacogenomics

There are numerous reports on associations of sSNVs to cancer prognosis, although few report possible mechanisms that could explain clinical observations. Furthermore, genetic polymorphisms also affect cancer treatments. The effects of sSNVs on cancer pharmacogenomics have been previously reviewed (253), and here, we highlight some recent studies in the field and those that affect cancer prognosis (Table 6). We also discuss sSNVs associated with patient-specific responses to cancer treatments.

Table 6.

sSNVs affect cancer prognosis and pharmacogenomics

Genea variant rs#	Observations and in silico predicted impactb
CYP19A1 c.240A>G rs700518	Letrozole treatment for breast cancer lowers triglycerides (270) Aromatase inhibitor treatment for breast cancer decreases fat-free mass and increases truncal fat mass (271) Musculoskeletal adverse events in breast cancer with tamoxifen or letrozole (269) Breast cancer treated with tamoxifen associated with longer cancer-free interval compared with letrozole (269) SynMICdb: 2.0974
EGFR c.2361G>Ac rs1050171	G/G genotype with metastatic colorectal cancer treated with cetuximab and/or panitumumab—associated with longer progression-free survival (227) G/A head and neck squamous cell carcinoma (HNSC) cells—more sensitive to gefitinib treatment than G/G HNSC cells (272) A/A squamous cell carcinoma primary culture cells—decreased transcription of EGFR-AS1 lncRNA (273) SynMICdb: 2.0095
ERCC1 c.354T>Cd,e rs11615	T allele—increases risk of colorectal cancer (196,254,255) T allele—increased risk of lymph node metastasis in esophageal cancer (257) C/C genotype—decreased mortality in colorectal cancer compared with C/T and T/T genotypes (256) T allele—improves lung cancer survival (258) T allele—lower risk of cervical cancer (259) T/T genotype—increased neutropenia while receiving chemotherapy in NSCLC (260) T/T genotype—increased stomatitis while receiving chemotherapy in colorectal cancer (196) C allele—increased nausea while receiving chemotherapy in colorectal cancer (196) T/T genotype—attenuated response to chemotherapy in esophageal cancer (257) C allele—improves chemotherapy response in esophageal cancer (meta-analysis) (261) C allele—improves chemotherapy response in lung cancer (meta-analysis) (262) T/T genotype improves response to chemotherapy in colorectal cancer (263) T allele—platinum-resistance less frequently observed in ovarian cancer (264)
ESR1 c.30T>Cf rs2077647	C/C and C/T genotypes—breast cancer letrozole treatment associated with decreased risk for adverse bone events compared with tamoxifen treatment (204) SynMICdb: 3.6756
ESR1 c.1782G>Af rs2228480	Luminal B–like breast cancer—tamoxifen resistance (267)
IDH1 c.315C>Td rs11554137	Glioblastoma—no different in survival (154) Gliomas—poorer prognosis (150) Acute myeloid leukemia—poorer prognosis (149,151,157) Gliomas—statistically significant association (152) Grade III gliomas—better prognosis (153) Adamantinomatous craniopharyngioma (216)
KRAS c.519T>C rs1137282	Decreased risk of distant metastasis in papillary thyroid carcinoma (210) SynMICdb: 4.0511
PDGFRA c.2472C>Td,e rs2228230	Acral melanoma—decreased mortality (160) Renal cell carcinoma—worse disease-free survival (164) SynMICdb: 3.7781

^aAccession numbers: CYP19A1 NM_000103.4 EGFR NM_005228.5 ERCC1 NM_202001.3 ESR1 NM_001122740.1 IDH1 NM_005896.4 KRAS NM_004985.5 PDGFRA NM_006206.6. EGFR = epidermal growth factor receptor; MES = MaxEntScan; NSCLC = non-small cell lung cancer; sSNV = synonymous single nucleotide variant.

^bMES and NNsplice did not predict any changes in splicing for these sSNVs. When available for a variant, SynMICdb scores (32) are listed, which indicate predicted likelihood the variant has a functional impact among all synonymous variants in SynMICdb: >0.89 = top 50%; >2.70 = top 10%; >4.38 = top 1%; >5.83 = top 0.1%; >8.08 = top 0.01%.

^cRefer to Table 4 for in silico predictions of pre-mRNA splicing.

^dRefer to Table 2 for changes to mRNA properties.

^eRefer to Table 3 for changes to protein abundance.

^fRefer to Table 5 for associations with other cancers.

Excision repair cross-complementation group 1 (ERCC1) is a single-stranded DNA endonuclease involved in DNA repair. Reports vary about which allele of ERCC1 c.354T>C (we report c.354T>C as it appears in dbSNP [rs11615]) is associated with cancer susceptibility. The prevalence of the 2 alleles, T and C, differs among the studied cancer populations and could explain conflicting reports. The T allele is dominant in European populations, and the C allele is dominant in Asian populations (196). The T allele has been linked to increased risk of colorectal cancer (196,254,255), worse survival rate in patients with colorectal cancer (256), and increased risk of lymph node metastases formation in patients with esophageal cancer (257). In contrast, the T allele was associated with better survival of NSCLC (258) and lower risk of cervical cancer (259).

ERCC1 c.354T>C may also be correlated to the prevalence of chemotherapy side effects (260). Colorectal cancer patients with the T/T genotype receiving chemotherapy (with or without oxaliplatin) had a statistically significant association with stomatitis compared with individuals with T/C or C/C genotypes, and the C allele was associated with nausea (196). In another study, the T/T genotype increased the risk of neutropenia in NSCLC patients receiving chemotherapy (260).

Resistance to platinum-based chemotherapeutic agents is a major problem for treating many tumors. Therefore, it is advantageous to find biomarkers, like genetic variants, that are predictive of drug response. Esophageal cancer patients with the T/T genotype of ERCC1 c.354T>C were found to have attenuated response to cisplatin and 5-fluorouracil treatment compared with T/C and C/C groups (257). Meta-analyses also reported that the C allele improves the response to chemotherapy in patients with esophageal cancer (261) and lung cancer (262). However, others have reported the opposite trend wherein the T/T genotype improved the response to oxaliplatin treatment in colorectal cancer patients (263). Kang et al. (264) observed ovarian cancer patients with the C/T or T/T genotypes are responsive to platinum-based therapies. Because high ERCC1 mRNA and protein levels have been associated with resistance to platinum-based chemotherapies (265,266), previous findings about elevated ERCC1 mRNA levels in T/T genotype cells (146,147) point to a possible mechanism by which the c.354T>C variant modulates resistance to platinum-based chemotherapeutic agents.

Tamoxifen is a selective estrogen receptor modulator that is commonly used to treat hormone receptor–positive breast cancer. However, approximately 30% of patients do not respond to tamoxifen (267). Babyshkina et al. reported that the ESR1 c.1782G>A sSNV was prevalent in patients with luminal B–like subtype breast cancer who were tamoxifen resistant and suggested that this variant may be used to predict if a patient will be responsive to tamoxifen treatment (267). The association was not observed between other ESR1 sSNVs (c.30T>C or c.975G>C) and luminal B–like subtype breast cancer. Considering that ESR1 mRNA expression is lower in tamoxifen-resistant tumors compared with tamoxifen-sensitive tumors (267), additional studies that characterize the impact of c.1782G>A on ESR1 mRNA expression could help elucidate the mechanism(s) by which this variant confers tamoxifen resistance.

CYP19A1 encodes the aromatase enzyme involved in the synthesis of estrogens. Breast cancer is often treated with the selective estrogen receptor modulator tamoxifen or the aromatase inhibitor letrozole. Although no statistically significant association between CYP19A1 c.240A>G and breast cancer risk was observed (268), this sSNV is clinically relevant. Breast cancer patients with the c.240A>G variant who were treated with tamoxifen had a longer breast cancer–free interval than those who were treated with letrozole (269). There also seems to be an association with the G allele and adverse musculoskeletal events (arthralgia and myalgia) regardless of tamoxifen or letrozole treatment (269). In a study of postmenopausal women with breast cancer, letrozole treatment had a statistically significant association with lower triglyceride levels in patients with the c.240A>G variant, whereas there was no association between letrozole treatment and changes in triglycerides in all genotypes (270). Treatment with aromatase inhibitors in postmenopausal breast cancer patients with the G/G genotype resulted in a statistically significant decrease in fat-free mass and increase in truncal fat mass relative to women carrying the T allele (271). These changes may increase the risk of cardiovascular events in women with the variant with long-term exposure to aromatase inhibitors (271).

Epidermal growth factor receptor (EGFR) c.2361G>A has been observed in many different cancers (53,166,216,218,224–229). However, others have reported that the frequency of this sSNV was no different in the cancer or healthy patient control groups (230,231). In metastatic colorectal cancer, patients treated with cetuximab and/or panitumumab, the G/G genotype was associated with longer progression-free survival. In the absence of treatment, no difference was observed in survival (227). It has been shown that in head and neck squamous cell carcinoma cell lines, sensitivity to the tyrosine kinase inhibitor gefitinib is attenuated by higher EGFR protein and mRNA expression (272). Some cell lines were found to be heterozygous (G/A) for the c.2361G>A variant and were more sensitive to gefitinib treatment than G/G genotype cells. Although no statistically significant difference was confirmed, there was a trend for the G/A cells to have lower EGFR protein and mRNA expression than G/G cells, which would be in line with the previous observation (272). A later study with patient-derived squamous cell carcinoma primary culture cell lines attributed the increase in gefitinib sensitivity (A/A genotype) to decreased transcription of the long noncoding RNA, EGFR-AS1 (273). EGFR-AS1 is a driver for EGFR addiction, where the cancer is physiologically dependent on EGFR expression and activity (274). The same effect is seen with other tyrosine kinase inhibitors but not with monoclonal antibody treatment (cetuximab). These results suggest the EGFR c.2361G>A variant can be used as a biomarker for treatment efficacy with EGFR inhibitors (273).

IDH1 is a member of the isocitrate dehydrogenase family of enzymes involved in nicotinamide adenine dinucleotide phosphate production (275). C.315C>T has a statistically significant association with poorer prognosis in AML patients (149,151,157). However, in gliomas, there are conflicting reports on whether the presence of this variant affects prognosis. There was no difference between survival in glioblastoma (glioma grade IV) patients with or without the variant (154). In a patient population including grades II-IV malignant gliomas, the variant had a statistically significant association with poorer prognosis (150). However, in a different cohort from the same study, a statistical significance was observed only in grade III gliomas (not grade II or IV) (150). In a different study, c.315C>T was found to be 3 times more frequent in patients with brain tumors than population controls, and the variant was most prevalent in grade III tumors compared with grades II and IV (152). In contrast to the previous reports, a recent study of grade III gliomas reported the variant was associated with better prognosis (153). The conflicting reports could be due to differences in stages of glioma and glioblastoma included in the studies. Higher IDH1 mRNA expression may lead to altered nicotinamide adenine dinucleotide phosphate production in AML (149), but reports on how c.315C>T affects mRNA expression are conflicting (149–151) (Table 2). Wang et al. (150) reported that the increased mRNA expression was not correlated to the outcome in glioblastoma.

Platelet-derived growth factor receptor α (PDGFRA) is involved in tumor proliferation and cancer progression via signaling pathways, including PI3K/AKT and RAS/MAPK. PDGFRA is a tyrosine kinase receptor and a common cancer drug target (276). PDGFRA c.2472C>T is located within one of the tyrosine kinase domains of PDGFRα (171). Although numerous studies only report the observation of this sSNV, 1 report indicates c.2472C>T reduces mRNA stability and decreases protein expression by increasing proteolytic degradation and decreasing protein synthesis (160). Furthermore, this sSNV affects the signaling activity of PDGFRα via the MAPK and PI3K/AKT pathways, by reducing expression of the phosphorylated forms of PDGFRα, AKT, and ERK (160). This results in reduced expression of PDGFRα and decreased mortality in patients with acral melanoma and no observed difference in cutaneous melanoma (160). However, in a study with RCC patients, c.2472C>T had a statistically significant association with worse disease-free survival in patients with RCC (164).

Cancer genome studies are a tool for improving methods for identifying cancer risk and managing cancer treatment. A challenge for applying this tool for preventing cancer and managing treatment is identifying which variants to monitor (24). Genome-wide association studies have been used to identify SNVs and loci that are genetic risk modifiers for an inheritable disease. A limitation of these studies is the inability to detect rare variants for which there is lack of biological relevance to a disease (226,277). Next generation sequencing (NGS) is a more recent method to detect inheritable genetic risk modifiers (277,278) and has been used to identify several sSNVs, which we have included in this review. For example, NGS was used to identify SNVs in high-risk cutaneous squamous cell carcinoma that would have not been detected in genome-wide association studies (226), including KIT c.2586G>A (Supplementary Table 1, available online), EGFR c.2361G>A (Table 4), and HRAS c.81T>C (Table 5). An NGS study of families with Li-Fraumeni syndrome found a statistically significant association of TP53 c.639A>G (Table 5) with metastasis at diagnosis of osteosarcoma (108). Another NGS study indicated an associated risk of HABP2 c.183C>T, NLRP6 c.2229G>A, TERT c.915G>A, and TERT c.3039C>T (Table 5) with papillary thyroid carcinoma (210).

Genomic instability is one of the underlying factors contributing to tumorigenesis (279). In addition to microsatellite and chromosomal instability, nucleotide instability also contributes to genomic instability (280,281). Considering how genomic instability measurements have potential to be used as a factor in assessing patient prognosis and managing therapeutic strategies (280), it is necessary to include sSNVs in improved genomic instability assessments.

Discussion

Historically, synonymous variants were described as silent and disregarded as a possible genetic cause for disease. Because the significance of synonymous variants has been recognized only in recent decades, several studies did not investigate sSNVs observed in human cancers. For example, the sSNVs of PDGFRA c.1701A>G, c.1809G>A, and c.2472C>T and KIT c.2394C>T were observed in neuroblastoma cell lines and tumors and labeled as “not activating,” with no explanation (172). The mechanistic understanding of how sSNVs affect protein structure and function (9–12) as well as methods to study the effects of sSNVs (25–31) has exploded in the last decade. Here, we have reviewed the current understanding of the role of sSNVs in human cancers.

Codon usage affects translational efficiency in part because of tRNA availability. The expression of specific tRNAs varies among tissue types (282) and the differentiation state of the cell (283). In addition, many cancers can also change tRNA regulation, affecting protein expression and ultimately impacting disease progression (284). Differences in tRNA availability could thus be used as a biomarker for tumorigenesis and tumor progression (285). Identifying minimal residual disease by analyzing circulating tumor DNA has been reported to be a predictive marker for cancer recurrence following adjuvant therapy (286–289). When determining which variants to detect when analyzing circulating tumor DNA, sSNVs should also be considered because of the evidence presented in this review.

A diverse set of mechanistic pathways drive the sSNV-mediated changes in protein expression levels, protein structure, and functions (290–292). Tools to study these underlying mechanisms vary in their cost and technical complexity. Consequently, we assume available data on the mechanistic underpinnings of sSNV-associated cancer risk largely reflect the accessibility of the experimental tools. For example, there are numerous reports of cancer risk or prognosis associated with mRNA splicing and mRNA structure, and these phenomena are relatively easy to assess using in silico and in vitro tools. However, it would be incorrect to assume that most sSNVs associated with cancers affect mRNA splicing and mRNA structure rather than other mechanisms like miRNA binding or translation kinetics. There remains an unmet need for technical methods of detecting changes in protein structure and miRNA binding and the standardization of assays so that these can be routinely deployed.

Because of the large numbers of sSNVs in human cancers, in silico tools are indispensable in screening and formulating hypotheses with respect to underlying mechanisms. Although in silico tools cannot replace in vitro and ex vivo methods, they can be used to screen large numbers of variants to identify a smaller subset that can undergo experimental evaluation.

As an illustration of the utility and limitations of in silico tools, we analyzed possible splicing alterations for the sSNVs. We identified 8 sSNVs for which in silico analysis predicted changes in pre-mRNA splicing, but in vitro reports indicated otherwise (Supplementary Table 3, available online). Overall, approximately 80% of our predictions with MES and NNsplice were accurate for variants with reported in vitro splice data (Figure 2, C). Although ESRs are an important component of splicing, our prediction accuracy did not improve by incorporating ESEfinder results. These discrepancies do not negate the use of in silico prediction tools, but rather they demonstrate the importance of in vitro studies to confirm the mechanism(s) by which a specific sSNV elicits a disease phenotype. Tools that predict disease-causing effects for synonymous variants are limited because many of the training datasets used to develop them have historically not included synonymous variants. In silico tools for predicting deleterious variants in cancer should use data sets with synonymous variants (293).

It is evident that synonymous variants have an impact on cancer development and prognosis. When investigating genetic causes for cancer, synonymous variants should not be overlooked. In this review, we have included reports of in vitro studies that provide an explanation for how sSNVs alter gene function; however, other mechanisms should not be ruled out. In addition, many studies have reported only an association or observance of an sSNV in cancer, and there is still a need for investigations involving in vitro mechanisms to elucidate the role of these sSNVs in cancer formation and prognosis. As more evidence is available for disease-causing variants, better therapeutics can be developed to treat patients.

Funding

This work was supported by funds from the US Food and Drug Administration CBER Coronavirus (COVID-19) Supplemental Funding and CBER operating funds.

Notes

Role of funders: The funders had no role in the design of the study, interpretation of the data, the writing of the manuscript or the decision to submit it for publication.

Disclosures: The authors have declared no conflict of interest.

Author contributions: NMK, DM, CKS: Methodology, Writing—review & editing; NMK, DM: Data curation, Investigation, Writing—original draft; DM: Formal Analysis; NMK: Visualization; CKS: Conceptualization, Funding acquisition.

Acknowledgements: We thank Dr Katarzyna Jankowska, CBER FDA, and Dr Zuben E. Sauna, CBER FDA, for helpful comments and feedback.