The protein phosphatase EYA4 promotes homologous recombination (HR) through dephosphorylation of tyrosine 315 on RAD51

Abstract

Efficient DNA repair and limitation of genome rearrangements rely on crosstalk between different DNA double-strand break (DSB) repair pathways, and their synchronization with the cell cycle. The selection, timing and efficacy of DSB repair pathways are influenced by post-translational modifications of histones and DNA damage repair (DDR) proteins, such as phosphorylation. While the importance of kinases and serine/threonine phosphatases in DDR have been extensively studied, the role of tyrosine phosphatases in DNA repair remains poorly understood. In this study, we have identified EYA4 as the protein phosphatase that dephosphorylates RAD51 on residue Tyr315. Through its Tyr phosphatase activity, EYA4 regulates RAD51 localization, presynaptic filament formation, foci formation, and activity. Thus, it is essential for homologous recombination (HR) at DSBs. DNA binding stimulates EYA4 phosphatase activity. Depletion of EYA4 decreases single-stranded DNA accumulation following DNA damage and impairs HR, while overexpression of EYA4 in cells promotes dephosphorylation and stabilization of RAD51, and thereby nucleoprotein filament formation. Our data have implications for a pathological version of RAD51 in EYA4-overexpressing cancers.

Graphical Abstract

Graphical Abstract.

Introduction

DNA is susceptible to damage from both endogenous and exogenous genotoxic agents. Double-strand breaks (DSBs) are some of the most harmful lesions, as they can cause significant genomic rearrangements, chromosome fragmentation, or even result in loss of chromosomes if not appropriately repaired. In human, non-homologous end joining (NHEJ) and homologous recombination (HR) are the two primary pathways for repairing DSBs. NHEJ quickly connects the two broken ends of DNA with minimal processing of the DNA ends. In contrast, HR is a more accurate mechanism that relies on extensive homology between the broken DNA and its repair template to restore an intact DNA copy at the break site. This requires extensive resection of the DSB, which generates a single-stranded DNA (ssDNA) molecule that will be coated by RAD51 and used to locate the homology.

The major players of the two pathways that compete to repair DSBs have been well characterized (1,2). These pathways are tightly linked with cell cycle progression, ensuring that HR is initiated only in S or G2 when a sister chromatid is present. This prevents the formation of potentially harmful structures that could impair recombination and ligation (3–5). When a DNA break occurs, cell cycle-regulated kinases, such as ATM and ATR, sense the damage and phosphorylate DNA damage repair (DDR) effectors. Upon phosphorylation, many DDR proteins relocate to chromatin, including the recombinase RAD51 (6). The chromatin environment at the break site, including heavy modification of histones, plays a crucial role in DNA repair processes (7). One of the earliest phosphorylation events in the cascade that activates the DDR machinery is the phosphorylation of H2AX at Ser139 (8). γH2AX (pSer139) acts as a beacon for DSB repair, recruiting DDR factors to the break site. Histones can be extensively modified on their flexible tails, and subunits of the nucleosome can be substituted to incorporate histone variants specific to DNA replication or repair. Post-translational modifications of histones can signal overwhelming DNA damage and direct cells toward apoptosis rather than DNA repair. This is the case of the concomitant phosphorylation of Ser139 and Tyr142 on H2AX, which is associated with apoptosis (9,10).

Three phosphatases from the Eyes Absent family (EYA1/2/3) have been shown to dephosphorylate H2AX pTyr142 in response to DSBs and the fourth member, EYA4, was proposed by similarity to perform the same function. EYA4 contains a compact Ser/Thr phosphatase domain (residues 268–292) and a larger Tyr phosphatase domain that contains 80 amino acids, important to the catalytic activity of EYA4, and fragmented into four individual motifs (spanning residues 369–614) (Figure 1A). Both catalytic domains are well conserved across vertebrates (Figure 1A) and are atypical phosphatase domains (11) in that they contain acidic residues.

Figure 1.

Atypical protein phosphatase EYA4 depletion in cell lines causes genomic instability. (A) Two domains of EYA4 that possess phosphatase activity are well conserved among vertebrates. (B) Incomplete knockdown can be achieved by shRNA in HeLa cells and leads to a decrease in both protein and mRNA transcript levels. (C) Cells depleted for EYA4 present large numbers of nuclear aberrations (bridges, micronuclei, lost chromosomes; quantification n > 300). (D) CIN70: chromosomal instability score based on expression levels of 70 proteins, shows CIN linked to EYA4 expression. 90 cell lines from CCLE split into two equal-size groups: EYA4 low and EYA4 high with median EYA4 protein expression level used as cut-off. Means ± SEM are shown. P value: P= 0.0028, t-test. (E) EYA4 depletion leads to elevated γH2AX (pSer139) foci. Representative images (scale bar 10 μm) and quantification (mean ± SD; n ≥ 200) are shown. For all panels * P ≤ 0.05, **** P ≤ 0.0001, except when indicated.

EYA proteins are transcriptional coactivators involved in the development and maintenance of the eye and the cochlear organ of Corti. Defects in EYA proteins are linked with neural defects, deafness (12–14), cardiomyopathies (15,16), lung carcinogenesis (17) and oral dysplasia (18). EYA4 is not well characterized and only few possible substrates have been identified. By similarity, ERβ (pTyr36), H2AX (pTyr142) and WDR1 (pTyr238) are all valid potential targets of EYA4 Tyr phosphatase activity (19–21).

EYA4 is hypermethylated and possibly overexpressed in triple-negative breast cancer samples (22,23) prompting the further study of its cellular role. Using EYA4 recombinant proteins (full-length WT, fragments and mutants) we uncovered a novel ssDNA-binding activity for EYA4, which greatly stimulates its phosphatase activity. EYA4 is also chromatin-bound through direct interaction with histones. Upon DNA damage, EYA4 dephosphorylates H2AX on residue Tyr142 to promote repair and RAD51 on residue Tyr315, enhancing RAD51 polymerization and presynaptic filament formation. EYA4 is itself phosphorylated in response to irradiation, which causes it to detach from the chromatin, likely to facilitate access to DSBs by the DNA repair machinery. We found that cells depleted for EYA4 are sensitive to genotoxic stress and deficient for HR, while overexpression of EYA4 drives the accumulation of hyperactive and stable dephosphorylated RAD51 protein, which forms longer presynaptic filaments in vitro. Conversely, phosphorylation of RAD51 reduces its DNA binding activity, explaining the HR deficiency observed in EYA4-depleted cells. Taken together, our data indicate that RAD51 phosphorylation status controls the nature of the filament and the recombination function of RAD51. This makes EYA4 a key player in DDR. Overexpression of EYA4 in tumors could yield accumulation of RAD51, which is linked with hyperrecombination and drug resistance phenotypes, even in the absence of copy number variation.

Our discoveries pave the way for future investigations of EYA4 as a druggable target that could be used in cancer treatments to limit metastasis and combat drug resistance caused by elevated RAD51 levels or secondary mutations in HR genes that restore HR (24,25).

Materials and methods

Cell culture and maintenance, transfections, and stable cell lines establishment

Cell lines were obtained from ATCC, U2OS-DIvA gifted by Gaelle Legube and HEK293/DR-GFP gifted by Jeremy Stark were maintained in cell-adhesion treated vessels at 37°C in 5% CO₂ incubators. Cells were cultured in Dulbecco's Modified Eagle Medium (DMEM) (Gibco) supplemented with 10% fetal bovine serum (FBS) and passaged at 80% confluence or less. DIvA cells were supplemented with Glutamax (Gibco). Virus were produced as described before (26). 1.2 × 10⁶ HEK 293FT cells were reverse-transfected using Lipofectamine 2000 Reagent (Invitrogen) with MISSION TRC2 pLKO.5-Puro (Sigma-Aldrich) empty vector or MISSION TRC2 pLKO.5-Puro (Sigma-Aldrich) EYA4 shRNA constructs (shRNA1, TRCN0000244430; shRNA2, TRCN0000218273; shRNA3, TRCN0000244429) and Lenti-vpak plasmids from OriGene to create lentivirus particles. Viruses were harvested at 48 and 72 hours post transfection and filtered through a 0.45 μm filter, then used to infect cell lines with polybrene (4 μg/ml) in a 6 cm dish. Stable cell lines were selected after 48 hours using 1–2 μg/ml of puromycin. For complementation, stable cells expressing shRNA1 were transfected with the pcDNA3.1/nV5-DEST construct coding for EYA4-resistant (full-length, FL) and selected with 500 μg/ml geneticin. For DSB induction, DIvA cells were treated for 4 h with 300 nM hydroxytamoxifen (4OHT) added directly to the culture medium. The origin of all cells was confirmed by short tandem repeat (STR) analysis. All cells were regularly tested for Mycoplasma.

RT-qPCR and western blot analyses

Total RNA was isolated by phenol chloroform extraction (TRIzol, Invitrogen) followed by nucleic acid precipitation as described before (27). The GoScript Reverse Transcription System (Promega) was used to generate first-strand cDNA. Real-Time quantitative PCR (RT-qPCR) was performed using TaqMan probes for human EYA4 (Invitrogen, Hs01012406_mH) and human 18S (Invitrogen, Hs99999901_s1) to amplify 70 bp and 187 bp fragments, respectively. The relative expression of EYA4 was determined using 2^-ΔΔCT method with 18S as an endogenous control for normalization. Western blot analysis was conducted according to our standard procedures (28,29). Briefly, cells were lysed on ice in RIPA buffer (10 mM Tris-HCl (pH 8.0), 1 mM EDTA, 0.5 mM EGTA, 1% Triton X-100, 0.1% sodium deoxycholate, 0.1% SDS, 140 mM NaCl) supplemented with cOmplete Mini EDTA-free protease inhibitor cocktail (Roche), 1 mM PMSF, 1 mM Na₃VO₄, 1 mM NaF, 1 mM benzamidine and 0.025 U/μL benzonase, sonicated 2 minutes (40%) in an ultrasonic water bath. The primary antibodies were: EYA4 (Abcam), γH2AX (pSer139) (Millipore), H2AX (pTyr142) (Abcam), β-Tubulin (9F3, CST), β-Actin (C-4, Santa Cruz) and GAPDH (CST).

Indirect immunofluorescence

Indirect immunofluorescence was performed as described elsewhere (29,30). Stable cell lines expressing control or shRNA plasmid were grown on coverslips for 24 h and treated with 4 Gy X-rays. Cell nuclei were pre-extracted with nuclear extraction buffer (NEB; 10 mM PIPES (pH 6.8), 100 mM NaCl, 300 mM sucrose, 3 mM MgCl₂, 1 mM EGTA (pH 8), 0.5% Triton X-100) for 2 min at RT and fixed with 4% paraformaldehyde (PFA) for 10 min at 4ºC. Nuclei were blocked in PBS with 5% BSA and 0.3% Triton X-100 for 2 h at RT, immunoblotted with a primary antibody (1:500 dilution in PBS with 1% BSA and 0.3% Triton X-100) for 2 h at RT, followed by secondary antibody (2 μg/ml in PBS with 1% BSA and 0.3% Triton X-100) for 2 h at RT. DNA was counterstained with DAPI. Slides were viewed on an Olympus FV3000 confocal microscope. Primary antibodies were: γH2AX (Millipore), H2AX (pTyr142) (Abcam), EYA4 (Abcam), pRPA (S4/S8) (Bethyl), pRPA (S33) (Bethyl), RAD51 (H-92, Santa Cruz) and pRAD51 (Y315) (Abcam). Secondary antibodies were: α-Mouse (Abcam, Alexa Fluor 647), α-Rabbit (Abcam, Alexa Fluor 488), α-Mouse (Abcam ab, Alexa Fluor 488) and α-Rabbit (Santa Cruz, CFL-647). The number of nuclear foci and their co-localization was quantified using CellProfiler. When indicated, cells were irradiated (4 and 10 Gy) in regular growth medium and a Gammacell 40 Exactor (radiation source: caesium¹³⁷) unit. Specificity of the pTyr142 and pTyr315 was validated by staining cells expressing no H2AX or knocked down for RAD51 respectively (Supplementary Figure S1A, B).

Proteins expression and purification

EYA4 full-length, cloned in pEGFP-C1 was expressed in mammalian cells using the expi293 expression system (ThermoFisher). 250 ml of culture was infected then grown for four days, and cell pellets were resuspended in lysis buffer (50 mM Tris–HCl (pH 8.9), 150 mM NaCl, 10% sucrose, 10% glycerol, 0.5 mM EDTA, 1 mM TCEP, 1 mM PMSF, 0.5% IGEPAL and protease inhibitors), sonicated 20 times for 15 seconds (50%) and the lysate was clarified by centrifugation (20 000 × g, 30 min). The supernatant was diluted in 50 mM Tris–HCl (pH 8.9), 10% glycerol, and loaded onto a 7.4 ml Source 30 Q column equilibrated in buffer A (lysis + 75 mM NaCl). The protein was fractionated in 4 ml fractions using a linear gradient to 100% of buffer B (buffer A + 1 M NaCl). Fractions containing the peak EYA4 were pooled and incubated with 1 ml of agarose resin anti-GFP for 2 h at 4ºC. The resin was collected, washed with 20 CV of buffer B, followed by 10 CV of buffer C (50 mM Tris–HCl (pH 8), 150 mM NaCl, 10% glycerol). GFP-EYA4 was either left on beads for interaction studies or eluted by cleavage of the GFP-tag with TEV protease (4ºC 2 h). EYA4 was collected the next day by flow in buffer C. TEV was removed by incubating elution on Ni-NTA resin for 1 h. Flow through and washes (buffer C) were then pooled, buffer exchanged in storage buffer (50 mM Tris–HCl (pH 8), 150 mM NaCl, 30% glycerol) and concentrated before storage. RAD51 (WT and mutants) was expressed in E.coli and purified as previously described (28).

Phosphatase assays

To measure phosphatase activity, increasing concentrations of EYA4 (0–200 nM, alone or pre-incubated with ssDNA) were incubated in the presence of 400 nM potential substrates (peptides sourced from Genscript, see list and sequence in figures) in 50 μl of reaction buffer A (50 μM MES pH 6, 2 mM MgCl₂, 50 μM DTT) for 1 h at 37ºC. At the end of the first incubation, the reaction mix containing malachite green was added to phosphatase or ATPase reactions, and the mixture was further incubated for 30 min at 37ºC. Plates were read at 620 nm in a plate reader. The quantity of phosphatase released was inferred from the standard curve prepared as per the manufacturer's recommendation (SIGMA).

Electrophoretic mobility shift assay (EMSA)

Electrophoretic mobility shift assay was performed with primers (IDT, see list in SI) or PhiX174 DNA (sourced from NEB) as previously described (53,55). Increasing concentrations of purified protein were incubated with fixed amounts of DNA in 10 μl of reaction buffer (50 mM Tris–HCl (pH 7.5), 150 mM NaCl, 1 mM MgCl₂, 1 mM dithiothreitol, 100 μg/ml BSA and 2.5 mM ATP) at 37ºC for 20 min. The reaction mixtures were resolved in 8% polyacrylamide gels in TBE buffer (100 mM Tris-borate (pH 8.3), 2 mM EDTA), for the oligonucleotide-based system, followed by imaging at 647 nm wavelength and quantification using QuantumStudio (BioRad). EMSA involving circular DNA were resolved on a 1% agarose gel, stained with ethidium bromide, destained in water, imaged under UV excitation on a BioRad imager, then quantified.

Microscale thermophoresis (MST)

A total of 12 concentrations of ligand proteins (e.g. EYA (1–365)) were serially diluted by 1:1 using MST buffer from 2.4 nM to 5 μM, 10 μl of each dose of ligand-protein was mixed with 10 μl of Cy5-labeled dsDNA/ssDNA supernatant. The mixture of EYA and DNA was loaded into premium capillaries for MST assay with parameters set up at auto-detected excitation power and medium MST power. Binding affinity was analyzed by MO. Affinity Analysis software (version 2.1.3, NanoTemper Technologies) using the signal from an MST-on time of 20 s. Three individual experiments were merged to generate standard deviation.

Electron microscopy

The reaction was carried out at 37ºC in buffer containing 2 mM ATP, 2.5 mM MgCl₂, 50 mM KCl and no BSA. To assemble nucleoprotein filaments, RAD51 (2.4 μM) was incubated with an 83mer oligonucleotide polydT (7.2 μM nucleotides) for 5 min and then the reaction mixtures were diluted 8-fold with the same buffer, and a 4 μl aliquot was applied to 400-mesh grids coated with carbon film and which had previously been glow-discharged in air. After staining for 30 seconds with 4% uranyl acetate, the samples were examined in a JEOL JSM-6610LV electron microscope equipped with a tungsten filament at 100 keV. Digital images were captured with a charge-coupled device camera at a nominal magnification of ×63 000.

Immunoprecipitation (IP) and mass spectrometry

1.2 × 10⁶ HeLa cells were reverse transfected with 1 μg of DNA (empty vector or pEGFPC1-EYA4). On day 2 media was changed, and on day 3 cells were irradiated (or not) with 10 Gy, incubated for 1 h, and proteins were extracted by autolyze on ice in 1 ml of extraction buffer. Resuspended cell extracts were sonicated and spun down at 14 000 rpm at 4ºC for 10 min. Lysates were rocked with 200 μl of resin (IgG anti-mouse or nanobody anti-GFP, immobilized on beads) for 2 h at 4ºC. Beads were washed 4 times in 50 mM Tris–HCl (pH 8), 750 mM NaCl, 10% glycerol and proteins bound were eluted in Laemmli buffer, loaded on a gel, digested, and subjected to analysis by mass spectrometry (UTHSCSA Core facility). Data were visualized and analyzed using Scaffold.

Affinity pull-down

Affinity pull-downs were performed as described elsewhere (31,32). 2 μg (2.5 μl) of GFP-tagged EYA4 fusion protein immobilized on beads was incubated with 2 μg of purified proteins (H2A, H2B) in 10 μl of reaction buffer (25 mM Tris–HCl (pH 7.5), 120 mM KCl, 1 mM β-mercaptoethanol) for 30 min at 4ºC. Unbound proteins were collected as supernatant (S). Beads were washed (wash W) three times with 100 μl of the same buffer and proteins complexed with EYA4 were eluted (E) in 20 μl of Laemmli buffer. 10 μl of the supernatant (S), first wash (W), and elution (E) were analyzed by SDS-PAGE.

Cancer cell line encyclopedia (CCLE) analyses

Protein-level expression data from cell lines in the CCLE dataset with available EYA4 protein levels (n = 90) were downloaded (33). Chromosome instability (CIN70) and homologous recombination deficiency (HRD) scores in corresponding cell lines were determined as previously described (34,35).

Chromosomal break in vitro reporter system and flow cytometry

The HEK293-puro-DR-GFP cell line has been described elsewhere (36,37). Cells were reversed co-transfected with 3 μg I-SceI expression plasmid (pCβASce) and 500 ng MISSION TRC2 pLKO.5-Puro empty vector or shRNA constructs, as indicated. Transfected cells were kept in regular growth medium and analyzed by flow cytometry after 72 h to measure the percentage of cells expressing GFP (28,32). For complementation, cells were co-transfected with shRNA1 and pcDNA3.1 Myc/His EYA4 or containing a mutant version of EYA4 (29).

Cell cytotoxicity assay (MTT)

HeLa cells (empty vector EV or knocked down for EYA4) were seeded in 96-well plates. After 24 h, increasing concentrations of PARP inhibitor (olaparib) were added. Cell cytotoxicity was measured following manufacturer's protocol (Abcam ab211091). Briefly, 50 μl serum-free media and 50 μl MTT reagent were added to each well and incubated at 37°C for 3 h. 150 μl of MTT solvent was added and plates were incubated with agitation for 15 min. Absorbance was measured at 590 nm.

Results

EYA4-deficient cells accumulate DNA damage

Complete knockout of EYA4 is lethal in most mice models shortly after birth (38), and poorly or not tolerated in several cell lines ((17) and our work). However, using short hairpin RNAs (shRNAs), we were able to significantly decrease EYA4 expression and protein levels in HeLa cells (Figure 1B).

Upon knocking down EYA4, HeLa cells acquired nuclear defects indicative of genomic instability (Figure 1C). HeLa cells depleted for EYA4 exhibited aneuploidy, DNA bridges, and accumulation of micronuclei, consistent with various mitotic and especially metaphase defects previously observed upon EYA4 knockdown in Mitocheck screens (Supplementary Table S1) (39). To further assess this phenotype, we identified 90 cell lines in the Cancer Cell Line Encyclopedia (CCLE) dataset with available EYA4 protein-level data (33). The CIN70 signature can be used to evaluate the degree of genomic instability (34). In line with our observations, cell lines with low EYA4 protein levels show significantly higher protein expression levels of CIN70 signature proteins, a surrogate for the degree of genomic instability (39), when compared to cell lines with high EYA4 protein levels (P= 0.0028, t-test) (Figure 1D).

We investigated whether this genomic instability could result from defective DDR, leading us to investigate the level of γH2AX (pSer139) in HeLa control cells and cells depleted for EYA4 by any of three shRNAs. When compared to cells expressing a non-targeting shRNA, more than 50% of EYA4-depleted cells exhibited an overall 8-fold increase of γH2AX (pSer139) foci in normal growth conditions (5% CO₂, 37 °C) without any exogenous stress and contained 10 or more discrete γH2AX (pSer139) foci (Figure 1E). While the accumulation of γH2AX (pSer139) is a reliable marker for DSBs, we could not ignore the possibility that H2AX Ser139 could be a target of EYA4 phosphatase activity and thus remain phosphorylated regardless of the DNA damage and DNA repair status, in the absence of EYA4. To test this, we first induced DSBs with irradiation, then followed the stimulation and resolution of γH2AX (pSer139) foci. A time course of γH2AX (pSer139) foci by immunofluorescence (Figure 2A) and measurement of protein levels by Western blotting (Figure 2B) both show that γH2AX (pSer139) is constitutively elevated in HeLa cells depleted for EYA4, but still increases following irradiation. Interestingly, some DSBs are repaired in EYA4-depleted cells as shown by a modest decrease of γH2AX (pSer139) foci between 1 h (peak) and 4 h post damage (Figure 2A, B). Thus, γH2AX (pSer139) foci in EYA4-depleted cells indicate DSBs and are not efficiently repaired.

Figure 2.

EYA4 contributes to DNA repair signaling by dephosphorylating H2AX residue Tyr142. (A) Kinetic of γH2AX (pSer139) foci formation over time, in HeLa cells proficient and deficient for EYA4, following irradiation. Representative images are shown (scale bar 10 μm) and foci are quantified and plotted (mean ± SD; n ≥ 800). (B) γH2AX (pSer139) protein levels followed by western blot after irradiation (10 Gy). (C) Residues Ser139 and Tyr142 are on the extreme C-terminal position of the histone tail. (D) Protein purification scheme (left) used to purify GFP-EYA4. SDS PAGE and Coomassie stain (right) show the purified protein. (E) Schematic of the malachite green colorimetric assay (left) used to investigate possible substrates of EYA4 (right) such as residues pSer139 and pTyr142 in H2AX (mean ± SEM; n = 3). (F) Foci formation in HeLa EYA4-depleted cells using an antibody specific to phospho-Y142. Representative images are shown (scale bar 10 μm), and foci are quantified and plotted (mean ± SD; n ≥ 150). For all panels **** P ≤ 0.0001.

Histone variant H2AX pTyr142 but not pSer139 is a substrate of the EYA4 phosphatase activity

While PP4 (protein phosphatase 4) and WIP1 (wild-type p53-induced phosphatase 1) are known to dephosphorylate γH2AX (40,41) EYA4 is unlikely to target Ser139 but has been suggested to dephosphorylate Tyr142, similar to EYA1-3 proteins (11,17). The two residues are in close proximity (Figure 2C) and while pTyr142-H2AX is a low abundance modification (42), it is essential for the promotion of DNA damage repair over apoptosis (10,43). To confirm EYA4’s phosphatase activity on H2AX, we expressed and purified EYA4 as a GFP-fusion protein (Figure 2D, supplementary information and (44)) and incubated it with phosphorylated H2AX peptides containing either pSer139 or pTyr142 (GenScript; Figure 2E, Supplementary Table S2). Released phosphate was quantified by a colorimetric method using malachite green-molybdate (Figure 2E). Both Thr and Tyr phosphatase domains of EYA4 have been described as functional in vitro (44). The synthetic pTyr142 peptide was dephosphorylated 5-fold more efficiently than pSer139 peptides (Figure 2E) under conditions optimal for EYA4 Tyr phosphatase activity. Using conditions optimized for Ser/Thr phosphatase activity (described in supplementary methods), pSer139 was still not significantly dephosphorylated and pTyr142 was not dephosphorylated, confirming that EYA4 is not the phosphatase for γH2AX (pSer139) but can dephosphorylate pTyr142 on H2AX and promote DNA damage repair signaling. Surprisingly, the phosphatase activity of EYA4 was greatly stimulated by the addition of single-stranded DNA in the reaction (+/- DNA, Figure 2E). DNA alone added to EYA4 in the absence of phosphopeptides does not generate signal in this assay. We verified our in vitro findings in HeLa cells, where depletion of EYA4 led to the accumulation of pTyr142-H2AX by immunofluorescence and western blot (Figure 2F, Supplementary Figure S1). As previously reported, Tyr142 phosphorylation decreases in response to DNA damage (10), but we observed an accumulation of pTyr142-H2AX in EYA4-depleted cells, at 2 h following irradiation (Figure 2F and Supplementary Figure S1) suggesting that EYA4 targets H2AX pTyr142 for dephosphorylation and possibly contributes to its accumulation by targeting upstream kinases.

EYA4 is a DNA binding protein that interacts with DNA, histones, and nucleosomes

The unexpected stimulation of phosphatase reaction by DNA suggested that EYA4 may interact directly with DNA. To investigate this possibility, we used purified EYA4, GFP full-length (Figure 2D), N-terminally mCherry-tagged EYA4 (aa 1–365, Figure 3A and Supplementary Figure S2A), and GST-C-terminal EYA4 (aa 358–639; Figure 3A) and conducted electromobility shift assay (EMSA) using synthetic substrates to mimic ssDNA or dsDNA. We found that EYA4 binds to both ssDNA and dsDNA (Figure 3B, Supplementary Figure S2B) but prefers ssDNA in a competition assay. The DNA binding activity of EYA4 is localized to its N-terminus, which displays the same substrate specificity as the full-length protein (Figure 3B, Supplementary Figure S2C). Using oligonucleotides (Figure 3B top) or a circular ϕX174 virion DNA with no DNA end available (Figure 3B, bottom) yielded similar results. The C-terminus EYA4, which contains the Tyr domain (358–639) does not possess DNA binding activity (Supplementary Figure S2D).

Figure 3.

EYA4 is a DNA and histones binding protein. (A) EYA4 FL WT, or FL mutated (** indicate Ser209, Thr211 residues), N-terminal 1–365 or C-terminal 358–639 constructs were purified for testing DNA binding activity. (B) Increasing concentrations (0–450 nM) of EYA4 1–365 were incubated with ssDNA, dsDNA or a mixture of both. Electromobility shift assay was visualized by tracking the Cy5-DNA on gel under UV (Bio-Rad imager). DNA binding was also assessed using phage DNA (ssDNA ϕX174 virion DNA (12 μM nt) and dsDNA ϕX174 RF I DNA (6 μM bp)) incubated with increasing concentrations (75–1500 nM) of EYA4 1–365. Complexes were run on agarose gel and visualized after ethidium bromide staining. The percentage of bound DNA was quantified and plotted (mean ± SEM; n > 3). (C) IP-MS workflow and a subset of peptides pulled down with GFP-EYA4 in HeLa cells non-irradiated (0 Gy) or 1 h after irradiation (10 Gy). Note that core histones are identified with certitude but peptides can be shared among variants. Full list of identified peptide available in supplemental files. (D) Summary of Microscale thermophoresis (MST) measurements obtained for EYA4 1–365 and SQTQ double mutant. (E) MST measurements of EYA4 1–365 on ssDNA or dsDNA and the affinity constants are presented, showing the binding of EYA4 1–365 to ssDNA (K_D 303 nM) and dsDNA (K_D 732 nM). (F) A mutant of EYA4 that mimics phosphorylation on residues Ser209 and Thr211 reduced DNA binding as shown by MST (K_D > 2 μM). (G) In vitro pull-down with purified histones H2A and H2B (left) or nucleosomes (NEB kit).

As dephosphorylation of H2AX and DNA binding activity makes EYA4 a good candidate for a role in DNA repair, we sought to learn more about the EYA4 protein network in response to DNA damage. We performed immunoprecipitation followed by mass spectrometry in HeLa cells expressing GFP-EYA4 and exposed to either no irradiation or 10 Gy irradiation. Among other interactions, EYA4 was found strongly associated with chromatin, and bound to histones and associated proteins in the absence of IR. Interestingly, binding to histones is weakened or lost following irradiation (Figure 3C), suggesting that EYA4 could be a sensor of DNA damage and gets displaced, possibly to allow chromatin accessibility by repair machineries.

Looking at the EYA4 sequence for possible post-translational modification (PTM) sites, we identified an ATM-ATR consensus site SQTQ in position and created a phosphomimetic EYA4 double mutant (Ser209Asp, Thr211Asp) (Figure 3A and Supplementary Figure S2A). We repeated DNA binding experiments, this time by Microscale thermophoresis (MST), to compare binding capacity between the EYA4 N-terminal domain (1–365) and the SQTQ double mutant, and observed a near complete loss of DNA binding activity by SQTQ (Figure 3D-F). This observation partially explains why EYA4 loses interaction with the chromatin following DNA damage, likely due to phosphorylation by ATM. Next, we examined if EYA4’s binding to histones as observed in IP is direct or through DNA. Using GFP-EYA4 protein immobilized on GFP-nanobody beads (45) and purified histones, we performed in vitro pull-downs and identified strong interactions between EYA4 and H2A or H2B (Figure 3G). In addition to binding individual histones, we found that EYA4 binds strongly to nucleosomes assembled with an octamer of histones and 208 bp of dsDNA (Figure 3G).

EYA4 is part of DNA damage repair foci

To gain a deeper understanding of the behavior of EYA4 in cells, we monitored a GFP-tagged version of EYA4 over time. EYA4 forms spontaneous foci in cells with untreated cells having an average of 20, small discrete EYA4 foci randomly distributed in the nucleus (Figure 3C, left) and which fuse into larger structures in G2. When using an antibody directed against EYA4 in HeLa cells, we observed that irradiation promptly stimulates EYA4 foci formation, and these foci partially co-localize with γH2AX (pSer139) (Figure 4A). To control the introduction of DSBs, we used the DIvA system (DSB inducible via AsiSI (46)), generously provided by the Legube lab. U2OS cells do not express EYA4 and we transfected the cells with GFP-EYA4, probing for the GFP tag during immunolocalization. In untreated U2OS-DIvA cells, we observed a low basal level of γH2AX (pSer139) foci, and low or absent RAD51 foci, as expected (Figure 4B top panel). Cells expressing the highest levels of EYA4 also exhibited RAD51 foci, even in the absence of γH2AX (pSer139) (Figure 4B top panel). This suggests that EYA4 co-localizes with non-repair RAD51 foci, and also that RAD51 accumulates in the presence of EYA4, which is tested later. Following the addition of tamoxifen and subsequent translocation of AsiSI, DSBs were introduced as shown by increased γH2AX (pSer139), and EYA4 formed robust foci, a large proportion of which co-localized with RAD51 foci (54%, Figure 4B, bottom panel) and with γH2AX (pSer139) (25%).

Figure 4.

EYA4 takes part in DNA repair foci assembly. (A) EYA4 (green) and γH2AX (pSer139, red) foci were observed by indirect immunofluorescence on samples fixed after no irradiation (0 Gy) or 4 Gy irradiation, at the indicated time points (scale bar 10 μm). Foci were quantified and plotted (mean ± SEM; n ≥ 250). (B) GFP-EYA4 transfected U2-OS-DIvA cells, with no DSB (-4OHT) or after induction of DSB by hydroxytamoxifen (+4OHT) were imaged and co-localization of EYA4, RAD51, and γH2AX (pSer139) at the break was quantified and plotted. (C) γH2AX (pSer139) and pRPA foci (S4/S8) were observed over time, after no (0 Gy) or 4 Gy irradiation, in HeLa controls (EV) or cells depleted for EYA4 (shRNA3). Foci were imaged (scale bar 10 μm) and quantified (mean ± SD; n ≥ 300; **** P ≤ 0.0001). Induction is plotted (mean ± SEM). (D) BrdU incorporation in control cells or silenced for EYA4 after irradiation allows indirect measurement of ssDNA accumulation.

In the absence of EYA4, homologous recombination efficiency is compromised

As all data indicate that EYA4 might be a hitherto unidentified player of HR, we investigated the repair kinetics of DSBs in the presence and absence of EYA4, after no or 4 Gy ionizing irradiation (IR). By following ionizing radiation-induced foci formation by γH2AX (pSer139), pRPA and RAD51, we found that HeLa EYA4-depleted cells accumulate γH2AX (pSer139) foci and replicative pRPA (phosphorylated at residue Ser33; Supplementary Figure S3A) but fail to form pRPA (Ser4/Ser8) foci that are surrogate markers of end resection at the break (Figure 4C, Supplementary Figure S3B). This suggests that the signaling of the break is perturbed in the absence of EYA4 and resection of the break might be decreased.

We tested this by measuring BrdU incorporation in native conditions (47). Without a denaturing step, exclusively ssDNA is measured, as an imperfect readout for resection efficiency. Use of this technique evidenced that cells depleted for EYA4 exhibit diminished ssDNA accumulation compared to control cells, in response to IR (Figure 4D). Consistent with this, RAD51 was found recruited less efficiently to the break in HeLa EYA4-depleted cells, and later than in control cells (Figure 5A, Supplementary Figure S4A). Using HeLa cells stably expressing the FUCCI system (26), in which clover-geminin and mKO-cdt1 label cells in green when in S-G2-M and in red when in G1 respectively, we verified that EYA4 depleted cells present no significant defect in the cell cycle that would explain decreased RAD51 foci formation (Supplementary Figure S4B).

Figure 5.

EYA4 is essential for efficient homologous recombination repair of double-strand breaks. (A) γH2AX (pSer139) and RAD51 foci were observed over time, after no (0 Gy) or 4 Gy irradiation, in HeLa controls (EV) or cells depleted for EYA4 (shRNA3). Foci were imaged (scale bar 10 μm) and quantified (mean ± SD; n ≥ 300; *** P ≤ 0.001, **** P ≤ 0.0001). Induction is plotted (mean ± SEM). (B) The DR-GFP system in HEK293 cells was used (left) and the percentage of GFP-positive cells was quantified by flow cytometry on >100 000 cells to estimate HR efficiency. shRNA-derived lentiviruses used are indicated and complemented with an shRNA-resistant EYA4. (C) 87 cell lines from CCLE split into two equal-size groups: EYA4 low and EYA4 high with median EYA4 protein expression level used as a cut-off. Means ± SEM are shown. P value: P= 0.0167, t-test. HRD score: homologous recombination deficiency score, as defined by Peng et al. (35). (D) An MTT assay was used to measure survival in response to the PARP inhibitor olaparib.

The lack of RAD51 foci confirmed that EYA4 might be involved in recombination-directed DSB repair, which can be tested using the in vitro DR-GFP system developed by the Jasin group (37,48). HEK293 EYA4-depleted cells were found less proficient in repairing DSBs by HR (Figure 5B).

To assess whether reduced EYA4 protein levels might cause HR deficiency more broadly, we determined previously defined HR deficiency scores (35) in 87 CCLE cell lines (39). This revealed that cell lines with low EYA4 protein levels show significantly higher HR deficiency compared to cell lines with high EYA4 protein levels (P= 0.0167, t-test) (Figure 5C). In accordance with an HR repair defect, HeLa cells depleted for EYA4 were also found sensitive to the PARP inhibitor olaparib in an MTT assay (Figure 5D).

RAD51 Tyr315 is a target of EYA4

RAD51 phosphorylation on residues Tyr54 and Tyr315 can influence its function (28,29). The Tyr kinase c-Abl, or the oncogenic fusion protein BCR/ABL, phosphorylates RAD51 sequentially, on residue Tyr315 followed by Tyr54 (49), and phosphorylation of these residues control RAD51 recruitment and strand exchange activity at DSBs (50). While phosphorylation of Tyr54 is found to stimulate RAD51 strand exchange activity (44) in vitro at high concentration, this modification is never detected in nuclear extracts (43). Moreover, persistent phosphorylation of Tyr315, and consequently double-phosphorylated status of RAD51, is inhibitory of HR, due to a defect in nucleofilament formation (51). Using a specific antibody recognizing pTyr315 (described in (43) and re-validated in Supplementary Figure S1), we observed that HeLa cells depleted for EYA4 exhibit elevated foci of pTyr315-RAD51, compared to control cells (Figure 6A). Interestingly, pRAD51 is present at a low level in unstressed cells and it decreases in response to IR in control cells, while it remains elevated in cells depleted for EYA4 (Figure 6B, Supplementary Figure S5A). Using the in vitro malachite green phosphatase assay, we identified Tyr315, but not Tyr54, as a substrate for EYA4 (Figure 6C). The dephosphorylation of RAD51 by EYA4 was again found stimulated by the addition of ssDNA suggesting a common activation mechanism for its phosphatase function. Biochemical characterization of Tyr54 and Tyr315 have been performed independently (49–51) and elegant studies from the Spies and Fleury laboratories among others evidence the role of these residues in RAD51 recombinase activity. There are many documented instances of dual phosphorylation having different effects in cells than single PTM, as exemplified by Ser139, Tyr142, in H2AX (9). Therefore, we investigated the combined action of these two residues by purifying the double-mutant RAD51(Tyr54Phe, Tyr315Phe) that cannot be phosphorylated, as well as the double phospho-mimetic RAD51(Tyr54Asp, Tyr315Asp) that cannot be dephosphorylated (Figure 6D). First, we wanted to investigate the ability of RAD51(Tyr54Phe, Tyr315Phe) to polymerize onto ssDNA and form nucleoprotein filaments. Surprisingly, we observed by electron microscopy and negative staining (Figure 6E) that this mutant is much more proficient at polymerizing than the WT protein (Figure 6E). It can form filaments across ∼10 lengths of ssDNA (83dT) in the mutant (Figure 6E), while WT RAD51 coats ssDNA as individual filaments, without any tethering activity (Figure 6E). In light of these surprising data, we purified and used a Thr309Ala mutant that is less efficient in HR (52) but does not affect filament formation. This mutant did not exhibit significant differences when compared to WT RAD51. The RAD51(Tyr54Asp, Tyr315Asp) phospho-mimetic formed globular structures that are reminiscent of the heptameric ring. Minimum variations in the ATPase activity were observed between RAD51 WT and mutants (Supplementary Figure S5B), which cannot fully explain the phenotype observed. DNA binding activity is central to presynaptic filament formation, and we conducted DNA binding experiments by microscale thermophoresis (MST) (Figure 6F). Surprisingly, we found that Tyr54Phe, Tyr315Phe binds to ssDNA with an affinity comparable to the WT RAD51, however, it is much more avid for dsDNA than the WT. In this experiment, the hill coefficient was very high at 7.07 (Supplementary Figure S5C), which is consistent with cooperative binding, phase separation and/or condensation. These data indicate a different DNA binding modality for the Tyr54Phe, Tyr315Phe mutant, possibly by aggregation onto the DNA. Even more surprisingly, the phospho-mimetic Tyr54Asp, Tyr315Asp was found to bind neither ssDNA nor dsDNA (Figure 6F and Supplementary Figure S5C) with a good affinity. While this was consistent with our EM observations, we could not rule out that this phospho-mimetic can bind to nucleic acid. Indeed, in two independent protein preparations (no added DNA and no ATP nor MgCl₂ in the storage buffer), we observed rare, long filament structures (Supplementary Figure S5D). Whether RAD51 (Tyr54Asp, Tyr315Asp) co-purifies with specific sequences of DNA, RNA, or forms spontaneous filamentous polymers on its own remains to be elucidated. While Mass Photometry analysis (Refeyn) confirmed that the WT RAD51 forms hexameric rings in solution at the expected ∼200 kDa mark, the mutants and mimics readings indicated possibly different configurations (Figure 6G). When introducing GFP-fused WT; Tyr54Phe, Tyr315Phe or Tyr54Asp, Tyr315Asp constructs in cells either expressing EYA4 (EV) or not (shRNA3), we observed that the WT forms structures as previously described for GFP-RAD51 (53) in the EV cells. However, the Tyr54Asp, Tyr315Asp is found non-nuclear, consistent with its inability to bind to DNA (Figure 6H). In cells transfected with shRNA3, both the WT and the phospho-mutant constructs form condensates that are consistent with the MST data, namely that dephosphorylated RAD51 is prone to phase separation on DNA. We had previously observed that the overexpression of EYA4-WT in cells led to an increase of RAD51 staining in otherwise untreated cells (see Figure 4B). Since EYA4 overexpression leads to accumulation of RAD51 in vivo (Figure 4) and dephosphorylates RAD51, we wondered whether RAD51 (Tyr54Phe, Tyr315Phe) might be more stable than RAD51 WT or RAD51 (Tyr54Asp, Tyr315Asp). We subjected cells expressing GFP-RAD51 WT to photobleaching experiments. Analysis of fluorescence intensity over time after photodamage showed that RAD51 WT mobility is diminished in cells depleted for EYA4 when compared to control cells expressing EYA4 (Supplementary Figure S5E).

Figure 6.

The phosphorylation status of RAD51 controls its activity. (A) RAD51 phosphorylation at residue Tyr315, in non-damaged conditions, can be observed in HeLa cells using a specific antibody. (B) EYA4 depletion leads to the accumulation of RAD51 pTyr315 under normal growth conditions. For (A) and (B), representative images are shown (scale bar 10 μm) and foci are quantified (mean ± SD; n ≥ 100; *** P ≤ 0.001, **** P ≤ 0.0001). (C) EYA4 dephosphorylates residue pTyr315 of RAD51, but not pTyr54, in a malachite green assay and the activity is stimulated by DNA (mean ± SEM; n = 3). (D) RAD51 WT, phospho-mutant Tyr54Phe, Tyr315Phe and phospho-mimic Tyr54Asp, Tyr315Asp were purified near homogeneity. (E) RAD51 nucleoprotein filaments were formed using 83(dT). Pre-synaptic filament lengths were assessed by electron microscopy. Representative electron microscopy micrographs of the RAD51 proteins on DNA and quantification (median; n ≥ 50) are shown. (F) Microscale thermophoresis (MST) measurements of RAD51 proteins on ssDNA or dsDNA and the affinity constants are presented. (G) Mass photometer measurements of apparent sizes for RAD51 polymers in suspension are shown. Proteins were analyzed alone or in the presence of DNA. (H) The introduction of RAD51 proteins as GFP-constructs in HeLa control or EYA4-depleted cells allows live imaging and the tracking of RAD51 structure formation and their subcellular localization. Representative images are shown (scale bar 10 μm).

The combination of this increase in stability coupled with high avidity for dsDNA and a slightly decreased ATPase activity of RAD51 (Tyr54Phe, Tyr315Phe) could explain strong staining of RAD51 in cells overexpressing EYA4.

Taken together, our data demonstrate that EYA4 promotes DNA repair by homologous recombination. EYA4 is phosphorylated upon DNA damage, dephosphorylates histone H2AX to promote repair over apoptosis, and contributes to HR repair through the dephosphorylation and subsequent stabilization of RAD51.

Discussion

Phosphorylation is a frequent post-translational modification on proteins, and it plays key roles in a wide range of essential cellular functions. This includes regulation of DNA damage repair pathways, as kinases trigger cascades of protein activation and orchestrate their timely recruitment to DNA breaks and adducts. Most phosphorylation events in DNA repair, which are well characterized and now utilized as readout of repair progression and efficiency, involve serine or threonine residues. Here, we described that tyrosine residues on H2AX and RAD51, which have been described before as phosphorylated by the c-Abl kinase, are favored substrates of the protein phosphatase EYA4. Interestingly, the phosphatase activity of EYA4 is greatly stimulated by the addition of DNA to in vitro reactions. We identified a consensus site for ATM-ATR phosphorylation (SQTQ) in EYA4, which when mutated to resemble phosphorylation, severely impairs EYA4 ability to bind DNA. From these observations, we can draw a model (Supplementary Figure S6) where EYA4 is nuclear and located at the chromatin in dividing cells. EYA4 interacts with the chromatin through direct interaction with DNA and with histones H2A and H2B. Following DNA damage, EYA4 dephosphorylates pTyr142 on H2AX, detaches from the chromatin, likely after being phosphorylated by ATM, and dephosphorylates RAD51 to enhance presynaptic filament formation, thus increasing the stability of RAD51 and promoting DSB repair by homologous recombination.

In recent years, it has become clear that RAD51 plays an important role in DNA metabolism, beyond vegetative double-strand break repair, and its function is not always mediated by BRCA2 alone (54–56) or even BRCA-dependent. There has been an accumulation of conflicting reports on the role of phosphorylation on residues 54 and 315 (20,50,51,57–59). Previous reports showed robust evidence that RAD51 tyrosine 54 and 315 are co-regulated (49), and that phosphorylation of tyrosine 54 decreases DNA binding and nuclear localization. However, whether the phosphorylation of RAD51 by c-Abl increases or decreases its activity in all of RAD51 functions at double strand breaks and at damaged forks, remains to be established, and is likely to be dose-, substrate-, and cell cycle timing- dependent. Here, we studied both residues together, by combining phospho- mutant or -mimic mutations in the same construct. In addition, we conducted in vitro and in vivo analysis. This approach allowed us to identify that dephosphorylated RAD51 is more stable than its phosphorylated version, in cells and in vitro, and more proficient to nucleate onto ssDNA. Dephosphorylated RAD51 exhibits DNA binding modalities that diverge strongly from the WT protein and was found to accumulate rapidly onto dsDNA. The lack of preference for ssDNA by RAD51 phospho-mutant, while RAD51 is known to bind preferentially to ssDNA over dsDNA, could have implications for later stages of HR post nucleofilament formation; not only for strand exchange, but also in duplex capture and homology search. Whether this hyperactive and stable filament promotes duplex capture for homologous recombination or targets different DNA structures warrants further studies, and will require biophysical characterization of the nucleoprotein filaments it can assemble. RAD51 is often upregulated in p53 deficient cancers and overexpressed in breast, cervical, ovarian, pancreatic, and other cancers (60,61), where it causes drug resistance (24,25,62). One can wonder if the stabilization of RAD51 at the protein level, through its dephosphorylation, could be the reason for its apparent overexpression in tumors, rather than a true genetic or epigenetic up-regulation. If lack of phosphorylation by c-Abl or excessive dephosphorylation of RAD51 by EYA4 leads to its accumulation and subsequent pathogenicity, targeting EYA4 to prevent aberrant or untimely dephosphorylation of RAD51 could offer valid and novel therapeutic options, and help reverse both drug resistance and hyper recombinant phenotypes.

Conversely, others have found before us that the phosphorylation of RAD51 on tyrosine 54 or the generation of a mutant that mimic it and cannot be dephosphorylated (pTyr54, Tyr54Asp or Tyr54Glu) severely decreases DNA binding affinity of RAD51(49–51). In this study, we take these observations further, as we demonstrate that RAD51 Tyr54Asp, Tyr315Asp binds both ssDNA and dsDNA feebly, but that when this phospho-mimic is purified, a small fraction of it can be found forming long filaments. This intriguing discovery could explain why the Spies lab observed that RAD51 Tyr54Asp is proficient in strand exchange despite DNA binding deficiency, and probably BRCA2 independent. Further studies will understand what type, if any, of nucleic acid favors the nucleation of this RAD51 variant.

The exact mechanism by which EYA4 itself undergoes post-translational modifications in response to DNA damage and how it regulates its interaction with other DNA repair proteins remains to be investigated. However, overall, our data start explaining the link between EYA4 and carcinogenesis. Decreased EYA4 levels in breast cancer samples might help identify HR deficiency, while overexpressed or hyperactive tyrosine phosphatase such as these of the EYA family could predict RAD51 stabilization, accumulation, and resulting drug resistance. Targeting EYA4 could be of interest for the future development of novel cancer therapies, especially those aimed at decreasing uncontrolled DNA damage repair in drug-resistant tumors.

Notes

Present address: Romain Tropée, Southern RNA, Springfield Central, QLD 4300, Australia.

Present address: Nicolas Paquet, Scorpius Biomanufacturing, San Antonio, TX, USA.

Data availability

All data are available as supplementary files and are published alongside this article. The MS data have been deposited to MassIVE under identifier MSV000091844, and to ProteomeXchange under identifier PXD041953.

Supplementary data

Supplementary Data are available at NAR Online.

Funding

NBCF [ECR13-04]; Cancer Council Queensland [APP1099791]; San Antonio Area Foundation (to E.D.); CIHR FDN-388879 (to J.Y.M.); J.Y.M. is Canada Research Chair in DNA Repair and Cancer Therapeutics; P.H.G.D. is supported by Cancer Council New South Wales [RG 21-13]; Mays Cancer Center is supported by National Cancer Institute cancer center support core grant [P30 CA054174]; Flow Cytometry Shared Resource at UT Health San Antonio is supported by the National Cancer Institute grant from Mays Cancer Center [P30CA054174]; Cancer Prevention & Research Institute Texas [RP210126]; National Institutes of Health [1S10OD030432-01A1]; UTHSA GCCRI TIF and MCC DDSBSR Facilities are partially supported by P30CA054174 from National Institutes of Health and RP160844 from Cancer Prevention & Research Institute Texas; University of Texas Health Science Center at San Antonio Institutional Mass Spectrometry Laboratory, directed by Susan T. Weintraub, is supported in part by NIH [P30 CA54174-26] (S.T. Weintraub, Mays Cancer Center Mass Spectrometry Shared Resource) and NIH [1S10RR025111-01] for purchase of the Orbitrap mass spectrometer (to S.T. Weintraub); E.D. and P.D. were recipients of ECR fellowships from the National Breast Cancer Foundation, B.dlP.A. and R.T. were recipients of a Research Fellowship by the Princess Alexandra Research Foundation; J.W.L. is supported by grants from NIH [NIGMS: R35GM137798, NCI: R01CA244261]; American Cancer Society [RSG-20-131-01-DMC and TLC-21-164-01-TLC]. Funding for open access charge: National Institutes of Health.

Conflict of interest statement. None declared.