Variants in the first methionine of RAD51C are homologous recombination proficient due to an alternative start site

Abstract

In the 20+ years since the discovery of RAD51C, scientists have been perplexed as to how missense variants in this tumor suppressor gene impacts its function and pathogenicity. With a strong connection to breast and ovarian cancer, classifying these variants as pathogenic or benign aids in the diagnosis and treatment of patients with RAD51C variants. In particular, variants at translational starts sites are disruptive as they prevent protein expression. These variants are often classified as pathogenic, unless an alternative translational start is shown to produce a functional isoform to rescue protein expression. In this study, we utilized the ribosome profiling database GWIPS-VIZ to identify two active translational start sites in human RAD51C at methionine one and methionine ten. This second translational start at methionine ten is both conserved in 97% of mammals and is the sole translational start in 80% of mammals. Missense variants at either methionine have been identified in 47 individuals, preventing expression from one of these two start sites. Therefore, we stably expressed both wildtype isoforms, as well as the RAD51C M1 and M10 variants in a RAD51C CRISPR/Cas9 knockout U2OS cell and compared their homologous recombination function. Surprisingly, we find that expression of human RAD51C from either start site can equivalently rescue homologous recombination of RAD51C CRISPR/Cas9 knockout U2OS cells through a sister chromatid recombination assay. Similarly, each of our RAD51C CRISPR/Cas9 KO cells stably complemented with RAD51C missense variants at either M1 or M10 are homologous recombination proficient. Together, our data demonstrate that RAD51C has two translational start sites and that variants in either methionine result in homologous recombination proficiency. With this critical discovery, individuals with variants at M1 will be more accurately informed of their cancer risk upon reclassification of these variants.

Graphical Abstract

Human RAD51C has two methionines, at position M1 and M10, that can be used as alternative start sites. In other mammalian species, the M10 residue is conserved whereas the M1 residue is not. Both the long and short RAD51C isoforms (RAD51C-Long and RAD51C-Short) are HR proficient. In addition, ClinVar patient identified variants in RAD51C in either M1 or M10 are similarly HR proficient. These findings have important implications for establishing pathogenicity of RAD51C variants in M1, as translation from M10 can rescue RAD51C expression and function.

Introduction

RAD51C is a tumor suppressor important during the homologous recombination (HR) DNA repair pathway [1]. HR is a high-fidelity double-strand break repair mechanism that uses the homologous template of the sister chromatid or homologous chromosome to repair damaged DNA [2]. Defects in HR are highly associated with an increased risk of breast and ovarian cancer and HR deficient cancers can be targeted through the use of precision therapeutics like PARP inhibitors [3]. RAD51C comes from a family of proteins that derive from a gene duplication event of the ATPase RAD51 and is therefore a RAD51 paralog [2]. RAD51C is unique in that it interacts with the other RAD51 paralogs to form two distinct complexes the BCDX2 complex with RAD51B, RAD51D, and XRCC2 and the CX3 complex with XRCC3 [2]. While RAD51C is the most well studied of the RAD51 paralogs, RAD51D has also been established as a tumor suppressor and its disruption has connections to both breast and ovarian cancer [1]. Like RAD51D, variants in RAD51C have been identified in several hereditary breast and ovarian cancer (HBOC) families [4–7]. Yet, the majority of these variants are classified as variants of unknown significance (VUS) and therefore cannot aid in risk assessment or expanding treatment options (ClinVar).

RAD51C variants currently classified as pathogenic are generally those which greatly alter protein structure and expression. This includes early stop codon and large deletions. However, variants in RAD51C’s ATG start codon are still not well understood. Confounding this issue, germline variants in the first methionine of RAD51C have been identified in 21 different individuals in ClinVar. Suggesting that there may be an alternative start site, a second methionine in RAD51C is located in frame at amino acid 10. In 2018, the ClinGen Sequence Variant Interpretation Working Group (SVI) released their updated criteria for interpretation of loss-of-function variant criteria [8]. This new guidance expands on the loss of protein expression criteria (pathogenic very strong [PVS1]) and notes that functional isoforms produced from an alternative translational start must be considered before use of this evidence criteria [8]. Therefore, without knowledge of RAD51C’s translational start, we cannot determine how variants in RAD51C’s translational start affect protein expression, RAD51C function or cancer risk.

Here we asked whether RAD51C has two translational start sites and how germline variants in methionine one or ten impact RAD51C HR function. We found that 97% of mammalian species are conserved with human RAD51C (hRAD51C) at the M10 position. Importantly, this second methionine is the sole start site in 80% of mammals with complete RAD51C sequences. By creating RAD51C isoforms that translate RAD51C exclusively at M1 or M10 positions, we find that both RAD51C isoforms are HR proficient. Ribosome profiling reveals that, in multiple human cell lines, translation can initiate from two distinct ATG sequences. Consistent with this, we find that patient derived variants in M1 and M10 are HR proficient (ClinVar and gnomAD). Our results demonstrate that, in vivo, hRAD15C has two fully functional translational start sites and that protein translation from either start maintains HR proficiency. These findings have important implications for determining cancer risk and treatment strategies for patients with RAD51C variants in M1 or M10.

Materials and Methods

Conservation Analysis

Mammalian RAD51C protein sequences were gathered from UniProt. To prevent overcounting, the longest form of each protein was selected from 77 mammals to represent the RAD51C sequence for that species. Sequences were aligned using Clustal Omega and conservation determined [9]. Of the 77 RAD51C sequences, 12 were considered incomplete because they did not start with a methionine. The remaining sequences were split into two groups depending on whether they started at the conserved M10 (M10 group) or at an earlier methionine (M1 group). Finally, sequence logo maps of the region surrounding the conserved M10 in each group was created using WebLogo [10, 11]. This includes the −5 and +5 region in the M1 group and only the −5 region in the M10 group.

Ribosome Profiling

Ribosome profiling data was collected through GWIPS-VIZ and is sourced from Ji et al. 2015, Fijalkowska et al. 2017, Crappe et al. 2015, Gao et al. 2014, Zhang et al. 2017, Chen et al. 2020, Gawron et al. 2016 and Raj et al. 2016 mapped onto human Dec.2013 (GRCh38/hg38) assembly at chr17:58,692,626–58,692,906 [12]. Results from harringtonine treated, initiating ribosomes are shown.

Cell lines and plasmids

RAD51C isoform constructs and M1/M10 variants were created by Gene Universal. A list of plasmids used in this study is found in Table 1. The SCR#18 U2OS RAD51C CRISPR/Cas9 knockout cell line with the integrated sister chromatid recombination reported was provided by Mauro Modesti [13]. These cells were cultured in DMEM supplemented with 10% FBS, 1% PenStrep and 0.2% plasmocin prophylactic (InvivoGen). Cells were incubated at 37°C with 5% CO₂. Polyclonal stable cell lines were created by transfecting the plasmid of interest with TransIT-LT1 in the U2OS RAD51C CRISPR/Cas9 knockout cell line. Two days post- transfection, the growth media was replaced with G418-containing medium (800μg/ml) and selected for two weeks.

Table 1. Plasmids used in this study.

RAD51C plasmids used in this study are listed with plasmid name, nucleotide change, gene cloning and selection information. The plasmids were made by Gene Universal.

Plasmid	Gene	Tag	Cloning Site 5’	Cloning Site 3’	Selection (Mamm.)	Selection (Bact.)
pKB982-pCMV-51C-Long	RAD51C-Long	-	Noti	Sali	G418	KAN
pKB979-pCMV2B-51C-Short	RAD51C-Short		EcoRI	Sali	G418	KAN
pKB980-pCMV-51C-Short	RAD51C-Short	Flag (N)	Noti	Sali	G418	KAN
pKB1102-pCMV-51C-Long-M1I	RAD51C-Long-M1I	-	Noti	Sali	G418	KAN
pKB1103-pCMV-51C-Long-M1L	RAD51C-Long-M1L	-	Noti	Sali	G418	KAN
pKB1105-pCMV-51C-Long-M1V	RAD51C-Long-M1V	-	Noti	Sali	G418	KAN
pKB1104-pCMV-51C-Long-M1T	RAD51C-Long-M1T	-	Noti	Sali	G418	KAN
pKB1106-pCMV-51C-Long-M10A	RAD51C-Long-M10A	-	Noti	Sali	G418	KAN
pKB1108-pCMV-51C-Long-M10L	RAD51C-Long-M10L	-	Noti	Sali	G418	KAN
pKB1109-pCMV-51C-Long-M10R	RAD51C-Long-M10R	-	Noti	Sali	G418	KAN
pKB1110-pCMV-51C-Long-M10T	RAD51C-Long-M10T	-	Noti	Sali	G418	KAN

Western Blotting

One million cell expressing either RAD51C long construct, RAD51C short isoform construct or RAD51C variant complemented cell line were seeded in 60mm plates. The next day, cells were collected and lysed in RIPA buffer supplemented with PMSF, 1X PhosStop and 1X protease inhibitor cocktails (ThermoFisher) and 0.2 μL Benzonase (Sigma) for 30 min on ice. Lysates were spun at 17G for 30 min and supernatants were collected for Western blot. Protein samples were loaded into a 10% SDS-PAGE gel and run at 120V for approximately 75 minutes. Using the wet transfer method, protein was transferred to a PVDF membrane and run at 100V for 2 hours. Afterwards, the membrane was blocked in Oddessy Licor TBS blocking buffer for 1 hour. Blots were imaged on a Licor CLX scanner. RAD51C was visualized using a RAD51C antibody (Abcam Cat# ab55728, RRID:AB_945135, 1:500) and a secondary antibody (Licor IRDye 680RD goat anti-rabbit 926–68071). Tubulin, the loading control, was visualized using a anti-Tubulin antibody (Cell Signaling Antibody#2144S, 1:1000) and a secondary antibody (Licor IRDye 800CW goat anti-rabbit 926–32211). ImageStudioLite was used to image the western blots and adjust for brightness and contrast. Western blots were performed in triplicate.

Sister chromatid recombination assays

On day one, 200,000 U2OS RAD51C CRISPR/Cas9 knockout cells alone or stably expressing RAD51C-long, RAD51C-short or the RAD51C M1 or M10 variant of interest were seeded in a 6-well plate. The next day, each cell line was transfected with 4 μg of plasmid expressing the endonuclease I-SceI (pCBSceI) or a GFP control plasmid (peGFP) using 5 μL Lipofectamine2000. The transfection media was replaced with fresh media on day two post- transfection. Twenty-four hours later, the cells were collected and analyzed by flow cytometry (CytoFlex 6L). HR efficacy was determined by measuring GFP expression in 20,000 events. GFP expression was is shown as fold increase above the RAD51C CRISPR/Cas9 knockout cell. Experiments were repeated four times per RAD51C construct or variant.

Results

M10 in highly conserved and is the start of RAD51C major conservation.

Upon protein sequence analysis, we noticed that hRAD51C contains two methionine residues near each other at amino acid position one and ten. To determine if other mammals also have two methionines, we examined the protein sequences of RAD51C in additional organisms. We found that RAD51C is highly conserved at M10, particularly in comparison to M1. For example, we performed multiple sequence alignment from ten mammalian RAD51C sequences and M10 is conserved in each of these organisms (Fig. 1A). In contrast, M1 is not conserved. To determine the extent to which M10 is conserved in mammals, we gathered every mammalian RAD51C protein sequence available on UniProt (77 total) and analyzed RAD51C by multiple sequence alignment (Supplemental Fig. 1). Of the 77 mammalian RAD51C protein sequences, 75 mammals had conservation with the second methionine observed in hRAD51C (97.4%; Fig. 1B, top pie chart). In contrast, only one other mammalian RAD51C sequence was conserved at the first methionine in hRAD51C (Supplementary Fig. 1; Olive baboon).

Figure 1. Conservation of RAD51C-M1 and M10 among mammalian species.

(A) RAD51C sequences from 11 mammalian species organized into a multiple sequence alignment. Extent of conservation at each position is denoted as (.) somewhat conserved (weakly similar amino acid groups), (:) moderately conserved (strongly similar amino acid groups) or (*) fully conserved. (B) Top pie chart. Percentage of mammalian RAD51C sequences in which M10 is conserved. M10 conserved sequences are in red, non-conserved sequences are in blue.

Bottom pie chart. Percentage of full-length mammalian RAD51C sequences that start at the conserved M10 residue. Sequences starting at M10 are shown in red, sequences starting before M10 or not containing the conserved M10 are in blue. (C) Sequence logo map of the region around the conserved M10 in mammalian RAD51C sequences that contain the conserved M10 but start at an earlier methionine residue (top) and those that start at the conserved M10 (bottom). Position relative to the conserved M10 (in red at position 0) is seen in the X-axis. Common residues are shown at each position. Size of each residue is relative to conservation.

In some mammals (i.e. mouse, pig, goat and cat), RAD51C initiates at the conserved M10 residue (Fig. 1A). Therefore, we examined where RAD51C begins in all RAD51C mammalian protein sequences. Excluding 12 incomplete sequences, we found that 52 out of 65 RAD51C sequences analyzed start at the conserved M10 residue (Fig. 1B; bottom pie chart, 80%). The remaining 13 sequences (including human, rat, hybrid cattle, cow, horse, platypus, guinea pig, baboon, macaque, opossum, seal and Tasmanian devil; Supplementary Fig. 1) initiate at a methionine upstream of M10 (Fig. 1B; bottom pie chart, blue). In these 13 mammals, major conservation of RAD51C begins at the conserved M10. For example, in the sequence logo map of the region surrounding M10, only F and E are somewhat conserved at the −2 and −1 amino acid positions upstream of M10, with little to no sequence conservation in amino acid positions −5 to −3 (Fig. 1C). Starting at M10, the sequence QRDLV is well conserved among these seven species and all other RAD51C species (Fig. 1C).

RAD51C translation is initiated at both M1 and M10.

Our sequence analysis suggests that the first nine amino acids of hRAD51C may be dispensable for RAD51C function. Thus, we next determined whether RAD51C translation occurs at M1, M10, or at both methionines. To do this, we analyzed whole genome ribosome profiling data from GWIPS-VIZ [12, 14–19]. Using the harringtonine initiating ribosome data, we identified two translation initiation peaks. The first, and most consistent peak, is located at the canonical start of RAD51C, M1. We also identified a second peak at the conserved M10 residue in five of the eight cell lines analyzed (Fig. 2). These results suggest that in human cell lines, RAD51C can be translated at either the M1 and/or M10.

Figure 2. GWIPS-VIZ initiating ribosome profiling data identified translational start sites at both M1 and M10.

Harringtonine ribosome profiling data from Ji et al. 2015 (1), Fijalkowska et al. 2017 (2), Crappe et al. 2015 (3), Gao et al. 2014 (4), Zhang et al. 2017 (5), Chen et al. 2020 (6), Gawron et al. 2016 (7), and Raj et al. 2016 (8) mapped onto human assembly GRCh38/hg38 (Dec. 2013). M1 (blue) and M10 (red) start sites are indicated. Ribosome footprint reads indicated by the blue peaks. Cell lines used in each experiment are indicated to the right of the profile.

RAD51C isoforms perform HR at comparable proficiencies

We next sought to determine if initiation of RAD51C at either M1 or M10 results in a functional RAD51C protein. Using a U2OS RAD51C CRISPR/Cas9 knockout cell line, we created polyclonal stable cell lines of four hRAD51C constructs beginning at methionine one (RAD51C-long) or methionine ten (RAD51C-short isoform) (Fig. 3A and B). We analyzed both untagged and N-terminally Flag tagged versions of both isoforms of RAD51C. Since the long RAD51C isoform could potentially be translated into both a short and long isoform, we also created a RAD51C variant where M10 is mutated to alanine. In this M10A construct, only the longer version of RAD51C would be translated (Fig. 3A). Each of these constructs was stably integrated into the U2OS RAD51C CRISPR/Cas9 knockout cell line (Fig. 3A; Line 3). These RAD51C isoforms were expressed at differing levels in our RAD51C knockout cells. While the Flag-tagged isoforms express at equal levels (Flag-51C-long and Flag-51C-short), untagged RAD51C-Long and the RAD51C-M10A mutant consistently express at lower levels compared to untagged RAD51C-short (Fig. 3B; 51C-Long, 51C-Long-M10A, 51C-Short). Importantly, the untagged RAD51C-long isoform (51C-long) runs as a doublet or smear, which may be due to two forms of the RAD51C protein being translated (Fig. 3B). In contrast, the other RAD51C constructs are expressed as a single discernible band. In addition, we are unable to perform mass spectrophotometry analysis to detect the two isoforms due to several tryptic cut sites within the first ten amino acids of RAD51C. Together these results demonstrate that both RAD51C isoforms are expressed to varying levels (RAD51C-long and RAD51C-short) and this difference is abated with the addition of an N-terminal flag tag.

Figure 3. RAD51C short and long isoforms perform HR with the same efficacy.

(A) RAD51C isoform constructs analyzed include the N-terminally Flag-tagged constructs, untagged constructs and the RAD51C-M10A mutant. The resulting isoform is indicated at the right and is shown in blue for the long RAD51C isoform (Long) or red for the short RAD51C isoform (Short). (B) Stable expression of RAD51C N-terminally flag-tagged and untagged constructs were analyzed by western blot using αRAD51C antibody or αTubulin, as a loading control. The samples were run on 10 cm and 16 cm gels. The resulting long or short isoform is indicated in blue (Long) or red (Short). The experiment was performed in triplicate. (C) Schematic of SCR assay used to determine the fold increase in HR as measured by GFP. Fold increase in GFP relative to the U2OS RAD51C CRISPR/Cas9 knockout cell line was measured for 20,000 events and graphed. (D) RAD51C Flag-tagged and untagged short and long isoforms (Flag-51C-Short, 51C-Short, Flag51C-Long, 51C-Long, respectively) and the RAD51C-M10A control (51C-Long-M10A) were compared to that of the parental RAD51C CRISPR/Cas9 knockout cell line (black dotted line). The resulting isoform is indicated in blue (long) or red (short). The experiment was performed five times. Significance was determined by unpaired student-T-test and standard deviation is shown.

Next, we asked whether RAD51C beginning at either M1 or M10 would be HR proficient. To examine the HR proficiency of untagged and Flag-tagged RAD51C-long and RAD51C-short, we performed sister chromatid recombination (SCR) assays using a stably integrated GFP-based recombination assay. In this assay, the reporter contains two non-functional copies of GFP. The first is interrupted by a unique I-SceI restriction enzyme cut site. The second truncated GFP sequence is used as a repair template (Fig. 3C). After transfection of a plasmid containing the I-SceI enzyme, a DSB is induced. We found that both the RAD51C-long and RAD51C-short expressing cells are equally HR proficient (Fig. 3C). This similarity is seen between both the untagged constructs (p=0.9161) and the Flag-tagged constructs (p=0.6205). Furthermore, N-terminal Flag tag does not significantly impact HR proficiency differences between long or short RAD51C isoforms (Fig. 3C, p=0.7102 and p=0.8207, respectively). In support of both forms of RAD51C being HR proficient, the RAD51C-long construct with a mutation at methionine ten (RAD51C-M10A) that can only express the full-length protein is similarly HR proficient (Fig. 3C). These findings suggest that RAD51C expressed from either M1 or M10 result in a functional RAD51C protein.

RAD51C variants at M1 and M10 have been identified in several individuals

RAD51C variants have been identified in M1 or M10 in 47 different individuals (ClinVar; gnomAD). Collectively, 12 variants have been reported in either the M1 or M10 position of RAD51C and each of these are classified as VUS. These variants have been identified in ClinVar, which reports germline variants, and gnomAD, which reports single nucleotide polymorphisms in the population (Table 2). Therefore, understanding how these variants affect functionality is important to understand individual cancer risk, health monitoring and treatment options.

Table 2. M1 and M10 RAD51C variants.

RAD51C variants identified on ClinVar and gnomAD. The variants in black text were analyzed in this study whereas the variants in grey text were excluded from analysis since they were recently added to ClinVar. The variant, nucleotide change, population frequency from gnomAD, the number of affected individuals, and variant source are shown. Table last updated 8/25/23.

Variant	Nucleotide Change	Population Frequency	Affected Individuals	Cancer Type	Source
M1I	3G>C 3G>A 3G>T	- 3.19 × 10−5 -	1 7 7	Not listed Breast Breast, Ovarian	ClinVar ClinVar, gnomAD, [22] ClinVar, [23]
M1L	1A>T	-	3		ClinVar
M1V	1A>G	-	4		ClinVar
M1T	2T>C	-	1		ClinVar
M10I	30G>T 30G>A	3.98 × 10−6 3.98 × 10−6	3 1		ClinVar, gnomAD ClinVar, gnomAD
M10K	29T>A	-	1		ClinVar
M10L	28A>T	3.19 × 10−5	5	Breast	ClinVar, gnomAD, [24]
M10R	29T>G	7.95 × 10−5	11	Breast, Ovarian	ClinVar, gnomAD, [25], [26], [27], [28]
M10T	29T>C	-	1		ClinVar
M10V	28A>G	-	2		ClinVar

RAD51C M1 and M10 patient-derived variants are HR proficient

To determine the function of RAD51C M1 and M10 patient-derived variants, we first created RAD51C M1I, M1L, M1V, MT, M10L, M10R and M10T in the RAD51C-long isoform construct (Fig. 4A). Next, each of these seven variants were stably expressed as a population in the U2OS RAD51C CRISPR/Cas9 knockout cell line. We found that the M1 variants, which express RAD51C-short isoform due to a mutation in the M1 residue, express more RAD51C protein compared to the M10 variants, which solely express the RAD51C-long isoform (Fig. 4B).

Figure 4. RAD51C-M1 and M10 variants are HR proficient.

(A) RAD51C variant constructs with wild-type (Line 1,6) or M1/M10 variant sequences (Lines 2–5 and 7–10). M1 and M10 variants were created in the untagged RAD51C-long construct. All variants resulting in the expression of a short RAD51C wild-type or shortened version of RAD51C are shown in red (M1I, M1L, M1V, M1T), and those resulting the expression of RAD51C-long wild-type or M10 variants are shown in blue (M10A, M10L, M10R, M10T). Each construct is labeled on the left. (B) Expression of RAD51C-M1 and M10 stably expressed variants was determined by western blot using αRAD51C antibody or αTubulin, as a loading control. The isoforms that express a short of long version of RAD51C are indicated with a red or blue line (short and long, respectively). (C) SCR assay was performed in stable cell lines expressing M1 and M10 RAD51C variants. RAD51C short and long isoforms and the M10A mutant were used as controls. Results are presented as fold increase above the parental RAD51C CRISPR/Cas9 knockout cell line (dotted line). All isoforms resulting in the expression of RAD51C-short are shown in red, those resulting the expression of RAD51C-long wild-type or M10 variants are shown in blue. The experiment was performed 5 times with 20,000 events being analyzed in each trial. Significance was determined by unpaired student-T-test and standard deviation is shown.

To determine the HR proficiency of RAD51C M1 and M10 variants, we performed SCR assays in each of the RAD51C variant expressing cell lines. We find that the M1 and M10 variants largely have similar HR efficacy compared to the control RAD51C-long and RAD51C-short isoforms (51C-long and 51C-short; Fig. 4C). In contrast, the RAD51C-M10T variant had elevated HR compared to both the RAD51C-long and -short isoforms (Fig. 4C; p = 0.029 and p = 0.014 respectively). Together these results suggest that both the short and long isoform of RAD51C are HR proficient and that patient or population-derived variants in these translational start sites do not result in HR deficiency.

DISCUSSION

RAD51C is important for genome stability and tumor suppression. By analyzing the isoforms of RAD51C, we determined that the first nine amino acids of RAD51C are dispensable HR function. After analyzing the N-terminus of RAD51C, we determined that the M10 position is highly conserved and the most common translational start for RAD51C in most mammals (Fig. 1). Upon functional analysis of the RAD51C-long and RAD51C-short isoforms and variants, we found that these proteins perform HR with equivalent efficacy (Fig. 3C, 4C). This is consistent with previous studies in Chinese hamster ovary cells [20]. The only considerable difference identified between the two RAD51C isoforms is that the shorter RAD51C isoform, beginning at M10, expresses higher protein levels than its longer counterpart (Fig. 3B, 4B). This suggests a decreased stability of the long isoform of RAD51C. Interestingly, the recently published BCDX2 complex structure shows RAD51C with the absence of amino acids one through eight. This suggests a solvent exposed disordered structure that is immune to crystal formation [21, 22]. Our findings are consistent with recent structural studies which demonstrated that the N-terminus of RAD51C located at position 1–9 are dispensable for mediating its protein interactions with other BCDX2 complex members [21, 22]. To date, missense variants in either M1 or M10 have been identified in 47 individuals. The findings of this study demonstrate that these variants have minimal impact on HR function, providing important information for patients harboring germline variants in the first methionine of RAD51C.

The existence of multiple N-terminal methionines may be an evolutionary mechanism of preventing catastrophic protein loss. By providing several translational start sites in the 5’ region, a secondary ATG sequence may be able to prevent loss of the entire protein through random mutation of the first start [23]. This pattern has been identified in several critical DNA repair genes. For example, RAD51, the essential HR protein, contains three in frame methionines within the first nine amino acids. Therefore, if the first start codon is lost or mutated, an equally functional, in-frame version of the protein can still be expressed. The possibility of producing two interchangeable isoforms would allow for tolerance of variants at start codons. While this is a logical hypothesis, it is puzzling why hRAD51C has multiple methionines whereas most mammals do not.

In contrast, translation from different start codons may produce unique protein isoforms. Leaky scanning of the 40s ribosomal subunit allows the ribosome to skip over the first start codon due to the suboptimal Kozak sequence and result in protein expression from the first (canonical expression) and the second (leaky scanning) start codon [24–26]. This process would then allow for the production of two discrete protein isoforms. Such isoforms have already been identified in several other DNA repair proteins like p53. Twelve p53 isoforms have been identified in humans, each of which is required to prevent premature aging and disease [27]. Similar distinctive isoforms have even been identified in, SWSAP1, a RAD51 paralog and member of the Shu complex [28]. Variants within the start codon of these functional isoforms may then alter cellular function and promote disease. In contrast, the other human RAD51 paralogs do not have multiple start codons (RAD51B, RAD51D, XRCC2, XRCC3). Understanding the function of each of RAD51C’s protein isoforms will be key to preventing RAD51C dysfunction related diseases.

Due to the extensive conservation of M10 in RAD51C, we were curious to determine the true translational start of the protein. Using GWIPS-VIZ data, we identified two translational start sites for RAD51C [12, 14–19]. The first site is located at M1, the canonical start site. The second translation initiation site is located at the conserved M10 (Fig. 2). When we compare the translation from these two sites, we find that the M1 is used as the translational start with high initiating ribosome reads in each cell line examined. However, in certain cell types translation from M10 matches translation at M1 (Fig. 2; HEK293 and MCF10A). In particular MCF10A, a BRCA WT breast epithelial cell line, expresses both of these RAD51C isoforms. These results suggest that dual expression from these two initiation sites may play a role in breast cancer prevention. Future studies are required to determine how translation of RAD51C-short is activated in these cell types and whether a similar pattern exists in ovarian cells.

The identification of active translational start site has been an issue plaguing the scientific community since completion of the human genome project in 2003. With this, comes the identification of the physiological isoform of each protein, starting with the ATG start codon in mammalian species. Most of the time, the longest sequence translating from the first start codon is assumed to be the primary isoform. However, genes can be misannotated which will drastically impact our understanding of the activity of the related protein [29]. Furthermore, the identification of cancer variants in the first methionine of RAD51C without considering the second methionine, could result in errors in classification and misdiagnosis for patients. In fact, we found limited functional differences between the HR activity of both M1 and M10 variants. In addition to the possible rescue effects of M1 variants, expression from M10 may also provide a means of rescue for other types of devastating variants, such as nonsense or frameshift variants, upstream of M10. Therefore, comparing the functional activity of these protein isoforms is pertinent to understand these proteins and their effects of their dysfunction.

Highlights.

• Ninety seven percent of mammals are conserved at methionine 10 of human RAD51C.

• RAD51C is translated from both methionine 1 and methionine 10 in human cells.

• Both isoforms of RAD51C are equivalently homologous recombination proficient.

• Missense variants in methionine 1 and methionine 10 are functional.