Pathogenic variants in MAEA disrupt DNA replication fork stability and are associated with developmental abnormalities in humans

Abstract

Replication stress (RS) poses a threat to genome stability and drives genomic rearrangements. The homologous recombination (HR) pathway repairs stalled replication forks (RFs) and prevents such instability. Through an E3 ubiquitin ligase screen aimed at identifying regulators of RAD51, we identified macrophage erythroblast attacher (MAEA), a core component of C-terminal to Lish (CTLH) E3 ubiquitin ligase complex, as a regulator of the HR pathway. Loss of MAEA impairs RAD51 recruitment at stalled RFs, leading to increased sensitivity to RS-inducing agents and excessive degradation of nascent DNA strands. Mechanistically, MAEA associates with and mediates the ubiquitylation of Ku80, enabling its removal from RF ends and facilitating the loading of RAD51. Notably, MAEA deficiency is associated with a developmental disorder involving microcephaly, craniofacial abnormalities, ocular defects, and heart malformations. Functional assays show that disease-linked MAEA variants (R34C, E349G, Y394D, and M396R) are defective in RS response. These findings establish MAEA as an essential factor in RF protection and genome integrity.

MAEA regulates the homologous recombination pathway and plays a key role in maintaining genomic stability.

INTRODUCTION

Topoisomerase I (TOP1) alleviates topological stress encountered during DNA replication and transcription by forming a transient TOP1-DNA cleavage complex (TOP1cc) (1). Clinical TOP1 inhibitors (TOP1is), such as camptothecin (CPT) and its clinical derivatives, irinotecan and topotecan (TPT), stabilize TOP1ccs into cytotoxic protein-DNA cross-links, blocking replication fork (RF) progression and inducing replication stress (RS) (2). Stalled RFs undergo collapse, generating single-ended double-strand breaks (seDSBs) through either replication runoff or cleavage by the MUS81/EME nuclease (3–7). seDSBs that typically occur at collapsed RFs are challenging for the cell because of the lack of a second DNA end, which is normally used as a template for accurate repair (8). The primary mechanism for repairing seDSBs is the homologous recombination (HR) pathway (8).

HR initiates with DNA end resection: The MRN complex and C-terminal binding protein interacting protein (CtIP) create a nick adjacent to the break, followed by exonuclease 1 (EXO1)/DNA replication helicase/nuclease 2 (DNA2)–mediated nucleolytic processing to generate long single-stranded DNA (ssDNA) stretches (9–15). Central to HR is RAD51, which catalyzes the homology search and strand invasion steps (16). RAD51 activity is tightly regulated by mediators, including BRCA2, which promotes RAD51 filament assembly on ssDNA and stabilizes stalled RFs (17–19). RAD52, although primarily associated with single-strand annealing (SSA), supports RAD51 filament formation in BRCA-deficient contexts, underscoring its context-dependent role in genome maintenance (20, 21). Following strand invasion, HR proceeds through replication-coupled DNA synthesis, restoring genome integrity (22–24). Although HR is essential for maintaining genomic stability under genotoxic stress, its precise role in resolving TOP1i-induced RS remains unclear.

In response to RS-inducing agents, such as TOP1is, cells initiate RF reversal, a process where the RF structure is remodeled into a four-stranded DNA structure known as a “regressed RF structure” through the annealing of complementary nascent strands (25, 26). While fork reversal prevents collapse, unprotected regressed forks are susceptible to nucleolytic degradation, resulting in genome instability (27, 28). Protective mechanisms counteract this risk: BRCA2 recruits RAD51 to shield reversed forks from meiotic recombination 11 (MRE11)–dependent degradation (29–32), while RAD52 stabilizes forks in BRCA-deficient settings, highlighting its compensatory role (6, 33). Remodeling factors (e.g., SMARCAL1, ZRANB3, and HLTF) (29, 34–44) and tumor suppressors (BRCA1/2) orchestrate fork reversal (29–32), while resolvases (GEN1 and MUS81) enable replication to restart (45–47). These protective mechanisms are crucial for suppressing carcinogenesis (48), but paradoxically, they limit the efficacy of chemotherapeutics (49, 50). Furthermore, mutations in genes that preserve RF stability cause genetic disorders affecting human development and predisposing individuals to premature aging and cancer (51).

While HR is the primary pathway for repairing seDSBs, nonhomologous end joining (NHEJ) can also repair these breaks (22). However, the end processing of seDSBs by NHEJ is cytotoxic, resulting in chromosomal aberrations and genetic instability as a result of the direct ligation of distant DNA ends (52, 53). Stalled or collapsed RFs are initially sequestered by the NHEJ heterodimer KU (KU70/KU80), which initiates NHEJ and inhibits RAD51-mediated HR (52). KU is crucial for stabilizing the DSB end structure and preventing exonucleolytic degradation by EXO1 (54–56). Although DNA resection can begin in the presence of KU, prior research has shown that removing KU from seDSB ends is necessary for replication protein A (RPA)-RAD51 exchange on ssDNA to occur (52). In mammalian cells, KU is removed from CPT-induced seDSBs through the coordinated actions of MRN and CtIP in an ATM (ataxia-telangiectasia mutated)–dependent manner (52). This process prompts long-range resection by EXO1, akin to DSB processing during HR repair (56–62). In addition, yeast studies revealed that KU can bind to stalled RFs even without RF collapse or DSB formation (63, 64). This finding suggests that KU may interact with the RF ends that arise during RF reversal. However, the mechanism by which KU is removed from stalled RFs in mammalian cells and the subsequent impact on RF remain unclear.

In this study, we identified macrophage erythroblast attacher (MAEA) as a key regulator of the HR pathway through a genetic screen. Depleting MAEA disrupts the recruitment of RAD51 to sites of stalled RFs, impairs HR, and increases cellular sensitivity to agents that induce RS. In addition, cells lacking MAEA are unable to restart stalled RFs under RS conditions, highlighting MAEA’s essential role in RF recovery. The absence of MAEA also results in excessive degradation of nascent DNA strands at stalled RFs, emphasizing its protective function. Our findings further demonstrate that MAEA can associate with and mediate KU80 ubiquitylation, which may promote HR by modulating the removal of KU from stalled RF ends. Last, we establish a link between MAEA deficiency and a multisystem developmental disorder characterized by developmental abnormalities (including microcephaly), craniofacial dysmorphism, ocular defects, and cardiovascular malformations. Functional analyses demonstrate that pathogenic MAEA variants (R34C, E349G, Y394D, and M396R) disrupt RS responses, linking MAEA deficiency to human developmental pathology. Our work establishes MAEA as a central guardian of genome stability during RS.

RESULTS

An E3 ubiquitin ligase screen identified MAEA as a factor that promotes HR

RF reversal depends on the central HR factor RAD51, which is consistently present at RFs independently of their breakage. To gain deeper insights into the mechanisms by which ubiquitin signaling influences the RAD51 recruitment at challenged RFs, we conducted a loss-of-function screen targeting E3 ubiquitin ligases to pinpoint genes whose inhibition diminishes RAD51 foci in U2OS cells, as illustrated in Fig. 1A. We have constructed a comprehensive human small interfering RNA (siRNA) “ubiquitome” library. This library encompasses RNA interference (RNAi) smart pools designed to target all known and predicted human E3 ubiquitin ligases (table S1). We subjected U2OS cells to a short (1 hour) treatment with a low dose of the TOP1i CPT (100 nM) and subsequently immunostained them using RAD51 and γH2AX (phosphorylated H2AX at S139) antibodies. Previously, it was demonstrated that at nanomolar concentrations, CPT predominantly leads to a slowing of RFs and a frequent reversal of the RF direction (65). Under these circumstances, where the generation of DSBs is minimal, RAD51 foci likely indicate the occurrence of RF remodeling events mediated by HR at stalled or broken RFs. As expected, the knockdown of established HR proteins, including BRCA1, BRCA2, and PALB2, reduced RAD51 focus formation, affirming the effectiveness of the screen (Fig. 1B). Our screening results revealed the C-terminal to Lish (CTLH) E3 ubiquitin ligase complex subunit MAEA (66), whose inhibition resulted in a decreased number of CPT-induced RAD51 foci, suggesting an unexplored role in HR repair (Fig. 1B). Figure 1C presents candidates from our screen that have not been previously linked to HR repair and exhibit reduced RAD51 focus formation. We further used CRISPR-Cas9 gene editing to create an MAEA knockout (KO) in U2OS cells. Western blot analysis confirmed the successful elimination of MAEA (Fig. 1, D and E). We observed that MAEA KO cells, similar to cells with siRNA-mediated MAEA deficiency, exhibited impaired RAD51 focus formation in response to CPT treatment (Fig. 1F). This defect in RAD51 focus formation was rescued by the introduction of wild-type (WT) MAEA but not by a catalytically inactive MAEA variant (C314A) (Fig. 1, E and G) (67). The impairment in RAD51 focus formation is not specific to a particular cell line, as we have observed this defect across various cell lines (fig. S1, A to C). Time-course analysis showed that the impairment in RAD51 focus formation persisted in MAEA KO cells across all time points examined (fig. S1D). To confirm MAEA’s role in RAD51 recruitment to stalled RFs, we used the in situ detection of proteins at RFs (SIRF) assay to assess RAD51 recruitment to nascent DNA at stalled RFs. SIRF is a modified proximity ligation assay (PLA) coupled with 5′-ethylene-2′-deoxyuridine (EdU) click chemistry for labeling nascent DNA at RFs (68). In line with our previous immunofluorescence (IF) results, we observed that MAEA depletion led to a decrease in RAD51 SIRF foci in response to CPT (Fig. 1H).

Fig. 1. Identification of MAEA as an RS factor.

(A) Diagram illustrating the E3 ubiquitin ligase gene siRNA screen to monitor CPT-induced RAD51 foci in γH2AX-positive cells. (B) Scatter plot displaying z scores derived from the screen outlined in (A). (C) Top-scoring genes not previously linked to HR, whose knockdown reduced CPT-induced RAD51 foci. (D) Immunoblot confirming MAEA loss in CRISPR-Cas9–generated MAEA KO U2OS cells. (E) Schematic of MAEA WT and catalytically inactive C314A mutant. Domains shown include LisH, CTLH, and RING, with amino acid boundaries indicated. (F and G) IF analysis of RAD51 focus formation in parental U2OS cells, MAEA KO cells (F), and MAEA KO cells reconstituted with either MAEA WT or C314A mutant (G). Cells were treated with or without CPT (100 nM, 1 hour) and immunostained for RAD51 and γH2AX. The number of large RAD51 foci colocalized with γH2AX foci was assessed in at least 100 positive cells per experiment (n = 3). RAD51 foci were quantified in only γH2AX-positive cells for all CPT-treated samples. Arrows in (F) highlight representative large RAD51 foci included in the quantification. (H) SIRF assay assessed RAD51 recruitment to nascent DNA in U2OS cells prelabeled with EdU (125 μM, 10 min) and treated with CPT (100 nM, 1 hour). The number of EdU-RAD51 SIRF foci was quantified in at least 100 EdU-positive (S phase) cells (n = 3). (I) DR-GFP assay in TRI-DR-U2OS cells transfected with siRNAs targeting luciferase (siCTRL) or MAEA (siMAEA) and complemented with FLAG-MAEA WT or C314A. I-SceI was induced with Dox. GFP-positive cells were quantified by flow cytometry (n = 4). Statistical significance was assessed by ordinary one-way ANOVA in Prism. Asterisks denote significance: ns, not significant; **P < 0.01; ****P < 0.0001. Error bars show the means ± SD. Scale bars, 10 μm.

DSBs can originate from the collapse of DNA RFs encountering obstacles such as single-strand breaks or protein-DNA complexes, including transient TOP1cc or inhibited/trapped poly(ADP-ribose) polymerase 1 (PARP1) enzymes (52, 69). In these scenarios, DSBs are formed and resolved via the HR pathway to restore stalled or collapsed RFs. This process necessitates HR using the homologous sister chromatid as a template, involving RAD51-dependent strand invasion. The observed impairment in RAD51 focus formation suggests a role for MAEA in HR repair (22). We used the direct repeat green fluorescent protein (DR-GFP) reporter assay to confirm the role of MAEA in promoting HR. The DR-GFP assay relies on HR to produce a functional GFP allele after introducing a chromosomal DNA break in one of the two impaired copies of GFP within the reporter cell line (70). TRI-DR-U2OS cells stably expressing doxycycline (Dox)–inducible I-SceI endonuclease were used. Subsequently, we assessed HR efficiency by flow cytometry, quantifying the frequency of GFP-positive cells. Our data revealed an impairment in HR-mediated repair in MAEA knockdown (siMAEA) cells compared to cells transfected with control siRNA (siCTRL) (Fig. 1I). This defect in HR was rescued by the introduction of siRNA-resistant WT MAEA but not by a catalytically inactive MAEA variant (C314A) (fig. S1E). Flow cytometry analysis confirmed that these HR deficiencies were not attributable to alterations in cell cycle distribution (fig. S1F). Next, we used the DSB Spectrum V3 reporter to further examine the effect of MAEA loss on the DSB repair pathway choice (71). The DSB Spectrum system is a fluorescent Cas9-based reporter that simultaneously measures the activity of multiple DSB repair pathways (71). This tool allowed us to parallelly quantify HR, mut-EJ (mutagenic alternative end joining), and SSA. Consistent with our earlier DR-GFP results, depleting MAEA reduced HR efficiency (fig. S2A). We also detected a moderate decrease (~13%, P < 0.05) in mut-EJ activity, while SSA remained unaffected (fig. S2A). Our results indicate that the loss of MAEA impairs HR repair, suggesting that HR deficiency could result in synthetic lethality when combined with PARP inhibition. To test this, we assessed whether silencing MAEA increases sensitivity to the PARP inhibitor olaparib (Ola). Consistent with this hypothesis, we observed that MAEA-deficient cells are more sensitive to olaparib in the clonogenic survival assay compared to control cells (Fig. 2A). Reintroduction of WT MAEA reduces the sensitivity of MAEA-deficient cells to olaparib (Fig. 2A). Collectively, these findings establish MAEA as a regulator of the HR pathway.

Fig. 2. MAEA loss increases cellular sensitivity to agents that induce DNA damage in the S phase.

Clonogenic survival assay of parental U2OS cells, MAEA KO cells, and MAEA KO cells stably reconstituted with GFP-MAEA WT. Cells were treated with the indicated drugs: (A) olaparib (Ola), (B) CPT, (C) topotecan (TPT), (D) HU, and (E) aphidicolin (APH) at the specified concentrations. Colony formation was assessed after 14 days. Survival data represent the means ± SD from five independent experiments.

MAEA cellular depletion sensitizes cells to agents that induce DNA damage during the S phase

Given that HR is a crucial pathway for repairing DSBs during the S phase, we explored the sensitivity of MAEA-deficient cells to agents that induce DNA damage during this phase, such as CPT and its derivative topotecan (72). In addition, we examined their response to agents that perturb DNA replication, such as aphidicolin (APH), a B family DNA polymerase inhibitor (73), and hydroxyurea (HU), which depletes the cellular deoxynucleotide triphosphate pool by inhibiting ribonucleotide reductase (74). A colony formation assay was performed to assess cell survival after DNA damage treatments. Consistent with the role of MAEA in HR, MAEA KO cells exhibited increased sensitivity to these DNA-damaging agents, a vulnerability that was reversed by reintroducing WT MAEA (Fig. 2, B to E, and fig. S2B). These data corroborate the role of MAEA as a regulator of the HR pathway and underscore its importance in sensitizing cells to S phase DNA damage–inducing agents.

MAEA stabilizes RFs during normal DNA replication

RFs pause at DNA lesions and can be restarted via the HR pathway or, if they collapse and form DSBs, repaired by HR (75). HR involves the synthesis of one nascent DNA strand by extending DNA ends within D-loops, while the second DNA strand is synthesized using the first nascent strand as a template (75). This process, while reminiscent of normal DNA replication in terms of processivity, rate, and capacity to duplicate extensive DNA regions, distinctively initiates at DSBs rather than at replication origins. Given these similarities between the two processes, we sought to determine the potential role of human MAEA in the process of DNA replication. We initially asked whether MAEA expression is regulated in a cell cycle–dependent manner. Given that the expression of the HR factor BRCA1 peaks during the S and G₂ phases (76, 77), we hypothesized that MAEA protein levels might exhibit a similar pattern. To test this, we examined MAEA expression in HeLa cells, which can be efficiently synchronized at the G₁-S boundary using a double thymidine block (78). The cells were released at various time points following synchronization to target specific phases of the cell cycle. Flow cytometric analysis confirmed that these time points were enriched for cells in the S, G₂, or G₁ phases (fig. S2C). We found that human MAEA is expressed throughout the cell cycle, with peak expression during the S-G₂ phases, paralleling cyclin A (Fig. 3A). Intriguingly, MAEA depletion led to a marked increase in bromodeoxyuridine (BrdU)–positive cells, as determined by flow cytometry, suggesting its involvement in facilitating efficient S phase progression (Fig. 3B). Next, we investigated the role of MAEA in RF progression using the DNA fiber assay (79). This technique allows the analysis of DNA replication dynamics at the single-molecule level. In the DNA fiber assay, cells are sequentially pulse labeled with two halogenated thymidine analogs, 5-iodo-2′-deoxyuridine (IdU) and 5-chloro-2′-deoxyuridine (CldU), to mark initiation/early elongation events and subsequent RF progression, respectively. After spreading the DNA fibers on a microscopic slide, newly replicated regions are visualized using fluorescently labeled antibodies specific to each thymidine analog. The length of the fluorescence signals in captured micrographs is then measured to determine the tract length (Fig. 3C). DNA fiber analysis revealed a notable increase in spontaneous RF stalling and a corresponding decrease in the number of active RFs in MAEA KO cells compared to WT MAEA cells (Fig. 3, D to F). In addition, increased RF asymmetry was observed in cells lacking MAEA, indicating RF instability (Fig. 3G). These findings suggest that the observed increase in BrdU-positive cells in MAEA-depleted samples may be indicative of an extended S phase duration as a result of stalled RFs, rather than a generalized reduction in DNA synthesis. These results highlight the crucial role of MAEA in maintaining RF stability during normal DNA replication.

Fig. 3. Role of MAEA in stabilizing the RFs under unperturbed DNA replication.

(A) Immunoblot analysis of MAEA expression level in different cell cycle phases. HeLa cells were left asynchronous (async) or synchronized by a double thymidine block, released for different time points to reach different cell cycle phases, and then extracted and analyzed with Western blot. Cyclin A served as a marker for S-G₂ phases. The numbers below the blots represent the MAEA levels (relative to async) after being normalized to tubulin. (B) Increased percentage of BrdU-positive cells in MAEA WT and MAEA KO cells. U2OS cells were prelabeled with BrdU (10 μg/ml, 4 hours) to identify those in the S phase, followed by fixation and denaturation with 2 M HCl for 30 min. After neutralization, immunostaining for BrdU was performed. The percentage of BrdU-positive cells was quantified from two independent experiments, with at least 4000 cells analyzed for each condition. (C) Schematic representations of different readouts used in (D) to (G) are shown. (D to G) DNA fiber analysis of parental U2OS, MAEA KO, and MAEA KO cells stably expressing GFP-MAEA WT or GFP-tagged MAEA C314A (CA) mutant under unperturbed conditions (n = 4). (D) Quantification of spontaneous stalled RFs, represented by the percentage of green-only (IdU) tracks. (E) Quantification of ongoing RFs, represented by the percentage of double-labeled tracks (IdU followed by CldU). (F) Quantification of new origin firing, represented by the percentage of red-only (CldU) tracks. (G) Quantification of RF asymmetry, represented by the ratio of long fork to short fork lengths. For all quantifications, an ordinary one-way ANOVA was used to assess statistical significance in Prism. Asterisks indicate statistically significant differences: *P < 0.05; **P < 0.01; ****P < 0.0001. Error bars represent the means ± SD.

MAEA is enriched at the RFs

To gain a deeper understanding of how MAEA influences DNA replication, we analyzed two published quantitative mass spectrometry studies of the ubiquitomes from MAEA WT and MAEA KO cells to identify MAEA substrates (80, 81). This analysis revealed several replication-related proteins as potential interactors, including elements of the minichromosome maintenance (MCM) helicase and the RPA trimeric protein complex substrates (80, 81). To validate these interactions, we performed coimmunoprecipitation experiments followed by immunoblotting, which confirmed the binding of GFP-MAEA to several replisome constituents, including MCM2, MCM7, and proliferating cell nuclear antigen (PCNA) (Fig. 4A). These results indicate a strong association between MAEA and replisome components. We used three distinct methodologies to explore the relationship between MAEA and DNA replication. First, we carried out PLAs (82) of MAEA with the replication protein PCNA. We found that MAEA is associated with PCNA, and this association did not increase after CPT treatment (Fig. 4B). Second, we used the SIRF technique (68) to visualize MAEA with nascent DNA at active and stalled RFs. Consistent with PLA results, we found MAEA to be enriched at the nascent DNA of active RFs, and this enrichment increases when cells are exposed to RS (Fig. 4C). Third, we used the isolation of proteins on nascent DNA (iPOND) combined with Western blotting (83) to ascertain whether MAEA is present on newly replicated DNA. iPOND is a robust method for detecting proteins at sites of newly replicated DNA (83). The iPOND assay uses “click” chemistry to attach biotin to a nucleoside analog (EdU) incorporated in newly synthesized DNA, enabling the analysis of proteins associated with RFs. MAEA was found to be enriched at RFs, akin to PCNA, when compared to mature chromatin (Fig. 4D). Furthermore, thymidine chase resulted in the loss of MAEA in iPOND samples (Fig. 4D), providing additional evidence that MAEA is a bona fide RF-associated protein. RS increased the association of MAEA with challenged RFs. PCNA was also enriched on nascent DNA at the RF but diminished in the thymidine-chased sample, confirming the specificity of the iPOND assay (Fig. 4D). Histone H2A protein levels were equivalent in all samples tested, supporting its expected role as a general chromatin-bound protein (Fig. 4D). Together, these findings strongly suggest that MAEA is a critical component of the replisome.

Fig. 4. MAEA is enriched at RFs and associated with replisome components.

(A) HEK293 cells were transfected with empty GFP or GFP-tagged MAEA and treated with either CPT (100 nM, 1 hour) or HU (2 mM, 4 hour) or left untreated. The cells were subsequently subjected to GFP pull-down, followed by immunoblotting using the specified antibodies. A portion of the cell lysates was saved before GFP pull-down as input controls. IP, immunoprecipitation. (B) PLA assay showing the interaction between MAEA and PCNA in U2OS cells, either treated with CPT (100 nM, 1 hour) or left untreated (unstim). An anti-MAEA antibody alone served as a negative control. (C) SIRF assay assessed MAEA recruitment to nascent DNA. U2OS cells were prelabeled with EdU (125 μM, 10 min) to mark newly synthesized DNA, followed by treatment with either CPT (100 nM, 1 hour) or HU (2 mM, 4 hour) or no treatment. No EdU condition was used as a negative control. The number of EdU-MAEA SIRF foci, indicating the interaction of MAEA with nascent DNA, was quantified in at least 100 EdU-positive cells per experiment (n = 3). For quantifications in [(B) and (C)], a one-way ANOVA was performed to determine statistical significance. Asterisks depict statistically significant differences: ****P < 0.0001. Error bars indicate the means ± SD of the data. Scale bars, 10 μm. (D) The iPOND experiment confirmed MAEA enrichment on newly synthesized DNA. HEK293 cells were labeled with EdU (10 μM, 10 min) and treated or not with CPT (100 nM, 1 hour) or HU (2 mM, 1 hour). For the thymidine chase condition, cells were incubated with thymidine (10 μM, 1 hour) before processing. The iPOND eluates were immunoblotted using the specified antibodies. The “no-click” control refers to samples that were processed without biotin azide.

Depletion of MAEA reduces the efficient activation of the cell cycle checkpoint

Our work so far indicated that MAEA plays a critical role during DNA replication processes. To expand upon this, we explored MAEA’s function in preventing RF stalling under conditions of induced RS. Cells depleted of MAEA demonstrated a marked increase in RF stalling when exposed to RS-inducing agents such as HU, in comparison to control cells (Fig. 5A). In addition, MAEA-depleted cells showed an inability to suppress new origin firing under these stress conditions, a process typically reflective of checkpoint activity (Fig. 5B). These defects in RF stalling and new origin firing can be rescued by reintroducing WT MAEA but not the MAEA C314A mutant (Fig. 5, A and B). These observations suggest that MAEA is crucial for the efficient activation of the intra–S phase checkpoint. To delve deeper into this relationship, we investigated checkpoint activation after ataxia telangiectasia–mutated and Rad3-related (ATR) inhibition [using VE821, an ATR inhibitor (ATRi)] (84), particularly in the context of HU-induced stress. We focused on the functionality of the ATR-dependent RS response in the absence of MAEA. We assessed ATR pathway activation in MAEA KO cells treated with ATRi alone, HU alone, or a combination of ATRi and HU. This was achieved through immunoblotting using phospho-specific antibodies that target known ATR substrates, such as checkpoint kinase 1 (CHK1). Our results showed that cells lacking MAEA were unable to efficiently phosphorylate CHK1 in response to HU (Fig. 5C). To determine whether the reduced ATR signaling was a result of decreased levels of RPA-coated ssDNA, the trigger for ATR activation, we examined their levels following exogenous RS. Contrary to expectations, MAEA-depleted cells exhibited elevated levels of RPA-coated ssDNA (Fig. 5D). We then investigated whether the defective recruitment of ATR or its interactors (e.g., TOPBP1 or ATRIP) underlay the p-CHK1 impairment in MAEA KO cells. Quantification of these factors at stalled RFs revealed no defect in TOPBP1 recruitment (fig. S2D). However, consistent with the p-CHK1 phenotype, ATR and ATRIP localization was reduced in MAEA KO cells specifically after HU treatment (fig. S2, E and F). Collectively, our data substantiate that MAEA is vital for both stabilizing stalled RFs and effectively activating intra-S cell cycle checkpoints under exogenous RS conditions.

Fig. 5. MAEA protects against excessive resection.

MAEA deficiency leads to increased stalled RFs (A) and new origin firing (B) under RS. A DNA fiber assay was performed in response to HU [2 mM, 4 hours (h)]. The labeling protocol used is shown at the top. The percentage of only-green (IdU) tracks, representing RS-induced stalled RFs (A), and that of red-only (CldU) tracks, representing RS-induced new origin firing (B), were measured from four independent experiments. (C) Immunoblot analysis of RPA phosphorylation at S4/S8 [p-RPA (S4/S8)] in U2OS cells treated with the ATRi VE-821 (20 μM, 1 hour), HU (2 mM, 4 hour), or a combination of both treatments. Whole-cell extracts were prepared and subjected to immunoblotting. (D and E) IF analysis of RPA (D) and BrdU foci (E) formation in parental U2OS and MAEA KO cells treated or not with CPT (100 nM, 1 hour). γH2AX staining was used as a marker for DNA damage. The right panels show quantifications of the mean intensity of RPA and BrdU foci per nucleus, while representative images from one experiment are displayed on the left panels. For all CPT-treated cells, only γH2AX-positive cells were quantified. Data represent at least 100 cells per replicate across three independent experiments. (F) Immunoblot of whole-cell extracts of parental U2OS and MAEA KO cells treated with either CPT (100 nM, 1 hour) or HU (2 mM, 1 hour) or left untreated. For all quantifications, an ordinary one-way ANOVA with multiple comparisons test was performed to determine whether differences between conditions were statistically significant: **P < 0.01; ***P < 0.001; ****P < 0.0001. Error bars indicate the means ± SD of the data. Scale bars, 10 μm.

MAEA safeguards stalled RFs from harmful DNA resection

Prior research has established that the absence of BRCA2 leads to the nucleolytic degradation of stalled RFs (29). This degradation impedes the forks’ amenability to HR repair, as detailed in (29, 41). Such excessive RF resection is a key factor in the increased chromosomal breakage observed in BRCA2-deficient cells. The nucleases MRE11 and DNA2 are implicated in this uncontrolled resection; notably, MRE11 inhibition can mitigate RF degradation in BRCA2-deficient cells (85). In our work, we investigated whether MAEA deficiency follows similar pathways, with a focus on its impact on DNA resection. We used a nondenatured BrdU-based assay to visualize the ssDNA products of resection. We found that MAEA ablation intensified the formation of CPT-induced native BrdU foci, consistent with an increase in ssDNA generation in these cells (Fig. 5E). These results are consistent with our earlier data showing a substantial increase in RPA loading onto stalled RFs in MAEA-depleted cells compared to controls (Fig. 5D). We also observed elevated levels of CPT-induced RPA2 phosphorylation at S4/S8, a recognized marker of DNA resection, in MAEA-depleted cells (Fig. 5F). Hyperphosphorylation of RPA2 was also evident in MAEA-deficient cells exposed to HU, indicating that this defect is not limited to CPT-induced stress conditions (Fig. 5F). To validate that the observed uncontrolled resection in MAEA-depleted cells specifically occurs at RFs, we adopted the methodology used by Schlacher et al. (29), focusing on the degradation of nascent DNA. Similar to the loss of BRCA2 (29), MAEA-deficient cells displayed an increased degradation of newly synthesized DNA at stalled RFs (Fig. 6A). These data indicate that MAEA is critical in preventing uncontrolled resection at stalled RFs.

Fig. 6. MAEA stabilizes stalled RFs by antagonizing EXO1.

(A) HeLa cells were transfected with either control siRNA (siCTRL) or siRNA against MAEA (siMAEA) for 48 hours. The labeling protocol, depicted at the top, involved sequential pulse labeling with IdU, followed by CldU, and then treatment with HU (2 mM, 4 hours) to induce RS. The graph shows the CldU/IdU ratio with mean values and SDs. At least 150 fibers were analyzed per condition, and the data shown represent three independent experiments. (B) A DNA fiber analysis was performed in HeLa cells transfected with siRNA against MAEA (siMAEA) or BRCA2 (siBRCA2) alone or in combination with siRNAs against indicated nucleases for 48 hours. The graph shows the CldU/IdU ratio with mean values and SDs. At least 150 fibers were analyzed per condition, and the data shown represent two independent experiments. (C) Representative RPA IF micrographs of parental U2OS and MAEA KO cells transfected with siRNA against indicated nucleases for 48 hours. All cells were treated with HU (2 mM, 4 hours). The mean intensity of RPA foci in γH2AX-positive cells was quantified in at least 100 positive cells per replicate. Data represent the means of at least three independent experiments. An ordinary one-way ANOVA with multiple comparisons test was performed to determine whether differences between conditions were statistically significant: *P < 0.05; ***P < 0.001; ****P < 0.0001. Error bars indicate the means ± SD of the data. Scale bars, 10 μm.

We next aimed to determine whether the uncontrolled resection observed in MAEA-deficient cells was facilitated by specific nucleases known for their roles in RF resection, namely MRE11, CtIP, DNA2, and EXO1 (29, 30, 32, 86). We performed parallel experiments comparing BRCA2-deficient cells with MAEA-deficient cells under identical conditions. Consistent with the established literature, we found that in BRCA2-deficient cells, RF degradation is predominantly rescued by MRE11 knockdown (with a milder effect from EXO1 depletion), validating the canonical role of MRE11 nuclease in HR-deficient contexts (Fig. 6B). We found that the siRNA-mediated knockdown of MRE11, DNA2, or CtIP did not alleviate the nucleolytic degradation of RFs in the absence of MAEA, as shown in Fig. 6B and fig. S3 (A and B). Depletion of EXO1 using siRNA suppresses HU-induced RF degradation in MAEA-deficient cells (Fig. 6B). In line with these results, depletion of EXO1, unlike CtIP, effectively suppressed the HU-induced overresection of stalled RFs (assessed by RPA foci) caused by MAEA depletion, as illustrated in Fig. 6C and fig. S3C. MAEA-deficient cells use lesion-specific exonuclease pathways during RS. CtIP and MRE11 mediate fork degradation after CPT treatment, while EXO1 is required following HU treatment (fig. S3D). Our findings demonstrate that MAEA maintains genomic stability during HU-induced RS by regulating EXO1 activity at stalled RFs.

Mechanism by which MAEA stabilizes RAD51 on stressed RFs

RAD51, renowned as a key player in HR, is also integral to the stabilization and restart of challenged RFs, as noted in previous studies (87, 88). Notably, the overresection of RFs observed in BRCA2-deficient cells can be counteracted by overexpressing an adenosine triphosphatase–dead RAD51 mutant, which stabilizes RAD51 nucleofilaments on ssDNA by inhibiting adenosine 5′-triphosphate–dependent dissociation (29, 41). Hence, RAD51’s recruitment to stalled RFs is critical in preventing uncontrolled nucleolytic degradation. Our data revealed that RAD51 focus formation following CPT exposure was impaired in MAEA KO cells, aligning with the hypothesis that RAD51 functionality is pivotal in mitigating excessive fork resection in the absence of MAEA (Fig. 1, F and G). This deficiency was not attributed to alterations in RAD51 protein levels (fig. S3E). We investigated potential mechanisms underlying the defects in RAD51 recruitment at RFs in MAEA KO cells. Upon examining the recruitment of BRCA1 and PALB2, which are crucial for RAD51 loading onto damaged chromatin (22), no defects were observed in MAEA KO cells (fig. S3, F and G). During the S phase, stalled RF ends are initially sequestered by the NHEJ heterodimer KU (KU70/KU80) to initiate NHEJ and hinder RAD51-mediated HR (52). Given KU’s antagonistic role in regulating RAD51 filament formation at stalled RF ends (52), we hypothesized that MAEA deficiency might lead to KU’s persistence in these regions during CPT-induced stress. To test this hypothesis, we used an SIRF assay to assess KU’s accumulation at RF ends. In line with MAEA’s role in the removal of KU from stalled RF ends, we detected an increase in KU SIRF foci in MAEA KO cells compared to their WT counterparts after CPT treatment (Fig. 7A and fig. S4, A and B). As a control, the knockdown of CtIP results in the persistence of KU foci, consistent with its known function in promoting KU removal (Fig. 7A). siRNA-mediated KU knockdown decreases the KU-SIRF signal, confirming its specificity in cells (Fig. 7A and fig. S4 A and B). In addition, IF experiments revealed persistent KU80 foci in MAEA KO cells in response to CPT treatment (Fig. 7B). To investigate the role of MAEA in removing KU from RF ends, we examined whether MAEA directly interacts with KU. We used the PLA to quantitatively assess the MAEA-KU interaction under both normal and RS conditions. We detected MAEA-KU80 PLA foci in the absence of DNA damage, and this interaction is enhanced following CPT treatment (Fig. 7C). In addition, coimmunoprecipitation experiments confirmed interactions between MAEA and KU (Fig. 7D). The hypothesis that MAEA removes KU from RF ends predicts that reducing KU levels in cells would restore the impaired RAD51 focus formation seen in MAEA-deficient cells. To test this, we assessed RAD51 foci in MAEA KO cells transfected with either control siRNA or siRNA targeting KU80 (siKU80). Supporting this notion, siRNA-mediated depletion of KU80 restores RAD51 focus formation in MAEA KO cells in response to CPT treatment (Fig. 7E), indicating that MAEA’s modulation of RAD51 foci is dependent on the presence of KU. In addition, KU80 knockdown specifically affected RAD51 foci in MAEA KO cells and did not influence RAD51 focus formation in WT cells under CPT treatment (Fig. 7E). Together, these findings highlight a mechanism by which MAEA facilitates KU displacement from stalled RF ends.

Fig. 7. MAEA mediates Ku removal from the stalled RF ends.

(A) SIRF assay assessed Ku80 recruitment on nascent DNA. U2OS cells were transfected with siRNAs targeting CtIP (siCtIP; positive control) or Ku80 (siKu80; negative control) for 48 hours or left untransfected. U2OS cells were prelabeled with EdU (125 μM, 10 min) and treated with or without CPT (100 nM, 1 hour). The number of EdU-Ku80 SIRF foci was assessed in at least 100 EdU-positive (S phase) cells (n = 4). (B) IF analysis of Ku80 foci in parental U2OS, MAEA KO, and MAEA KO cells reconstituted with WT MAEA and treated with or without CPT (100 nM, 1 hour). The graphs show the quantification of the number of Ku80 foci per nucleus. For all CPT-treated cells, only γH2AX-positive cells were quantified (n = 3). (C) PLA assay showing the interaction between MAEA and Ku80 in U2OS cells treated with CPT (100 nM, 1 hour) or left untreated (unstim). An anti-MAEA antibody alone served as a negative control. (D) HEK293 cells were transfected with GFP or GFP-MAEA for 24 hours, treated with CPT (100 nM, 1 hour) or HU (2 mM, 4 hours), or left untreated. GFP pull-down was performed, and eluates were immunoblotted with the indicated antibodies. (E) IF analysis of RAD51 foci in U2OS cells transfected or not with siKu80 for 48 hours and treated with or without CPT (100 nM, 1 hour) (n = 3). Quantification of RAD51 foci was performed as detailed in Fig. 1 (F and G). An ordinary one-way ANOVA with multiple comparisons was used to assess statistical significance: ***P < 0.001; ****P < 0.0001. Error bars represent the means ± SD. Scale bars, 10 μm. (F) Guanidine-HCl-His pull-down (His-PD) was performed on extracts from HeLa His-Ub cells transfected with siCTRL or siMAEA for 48 hours and treated with or without CPT (100 nM, 1 hour), followed by immunoblotting with indicated antibodies.

We next investigated MAEA’s ability to mediate KU ubiquitylation in CPT-treated cells. Using guanidine-HCl-His pull-down (His-PD) with HeLa cells stably expressing 6xHis-tagged ubiquitin (His-Ub), we demonstrated that KU80 undergoes ubiquitylation in MAEA-proficient cells following CPT exposure (Fig. 7F). This ubiquitylation was diminished in MAEA-deficient cells, indicating MAEA’s role in KU80 ubiquitylation during CPT-induced RS (Fig. 7F). As a control, siRNA-mediated depletion of KU80 abolished KU80 ubiquitylation, thereby confirming the specificity of the KU80 ubiquitin signal (fig. S4C). In summary, our findings indicate that MAEA is crucial for counteracting KU’s antagonistic effects on RAD51, thereby stabilizing RAD51 on chromatin and promoting HR-dependent restart and repair of damaged RFs.

AlphaFold3 modeling of MAEA and RMND5A

MAEA is part of the E3 ubiquitin ligase complex CTLH, which contains TWA1, Muskelin, ARMC8α and ARMC8β, WDR26, RMND5A, GID4, and RanBPM (66, 89). Although many subunits of the CTLH complex show conservation across eukaryotic lineages, and the RING domains in MAEA and RMND5A exhibit notable conservation with their yeast counterparts (90), further characterization is needed to understand the function of the complex. The RING zinc finger domains, found in two members of the CTLH complex, RMND5A and MAEA, confer E3 ubiquitin ligase activity, similar to other heterodimeric RING E3s like BRCA1/BARD1 (90, 91). To investigate the conservation of the MAEA protein, we conducted a sequence alignment across various species using Clustal Omega (92). The results indicate that MAEA is highly conserved overall. Notably, the C-terminal region of the MAEA protein, implicated in the E3 catalytic activity, exhibits notable conservation among the aligned sequences (Fig. 8A). To gain insights into the structure and functional mechanism of MAEA and RMND5A, we used AlphaFold3 to model MAEA with its binding partner RMND5A (Fig. 8B) (93). AlphaFold3 evaluates protein model accuracy using several metrics. The predicted template modeling (pTM) score assesses the overall fold accuracy (>0.5 suggests reliability), while the inter-pTM (ipTM) score measures subunit positioning (>0.8 indicates high confidence). The predicted local distance difference test score quantifies local confidence, with values >70 indicating reliable Cα positioning and >90 for side chains. The pTM scores were 0.83 (MAEA) and 0.78 (RMND5A), indicating reliable overall folds. The ipTM scores of 0.84 (MAEA) and 0.76 (RMND5A) suggest confident subunit positioning. High predicted local distance difference test averages (89.08 for MAEA and 85.85 for RMND5A) further support the structural reliability of these models. For unbound MAEA and RMND5A, the median predicted aligned error (PAE) values were 6.76 and 9.92 Å, respectively, indicating reliable predictions. The interprotein PAE between MAEA and RMND5A was 10.41 Å, further supporting model accuracy. Specifically, the PAE values for the N- and C-terminal regions were 3.66 and 6.04 Å, respectively. Overall, the models were predicted with a high degree of confidence (table S2 and figs. S5 and S6, A and B). Modeling reveals that the two proteins form an intricate network of interactions forming a heterodimeric complex, extending from the N-terminal coiled-coil domains to the C-terminal RING domains. It has been previously shown the catalytic function of the heterodimer is mediated by the RING-like domain of MAEA and the RING domain of RMND5A (81). The folding of the noncanonical RING domain of MAEA depends on its interactions with RMND5A, stabilized by an intermolecular zinc ion, coordinated by three highly conserved cysteine residues from MAEA (C302, C314, and C317) and a histidine (H331) residue from RMND5A (Fig. 8, B and C) (94).

Fig. 8. AlphaFold3 modeling of MAEA and RMND5A ring domains with UBE2D1 (purple) and ubiquitin.

(A) Clustal Omega amino acid sequence alignment of the RING-like domain of MAEA with ortholog sequences. Amino acids of interest are highlighted in red rectangles: C314, C317, E349, Y394, and M396. (B) AlphaFold3 model of the heterodimeric complexes MAEA (green) and RMND5A (brown). All modeling was done using AlphaFold3 (PMID: 38718835). Models were visualized and analyzed, and images were made using PyMOL Molecular Graphics System, version 3.0, Schrödinger LLC. The zinc ions (gray) were modeled using the ortholog structure of the yeast heterodimer Gid9 (MAEA) and Gid2 (RMND5A) (Protein Data Bank ID: 7NS4). The square highlights the intermolecular zinc ion. (C) Close-up view of the boxed region from (B), noting the intermolecular zinc ion coordinated by C314, C317, and C302 from MAEA and H331 from RMND5A. (D) Model of MAEA and RMND5A RING domains modeled with UBE2D1 (purple) and ubiquitin (gray). The electrostatic/polar surface on ubiquitin is colored in pink, and the hydrophobic surface is blue. Three crucial residues emanating from MAEA are labeled to denote their involvement in coordinating ubiquitin: E349, Y394, and M396. The zinc ion is in gray. (E) Clustal Omega amino acid sequence alignment of residues in the coiled-coil domain. The red rectangle highlights the R34 part of the positive patch on MAE. (F) Overlay of AlphaFold3-modeled MAE-RMND5A (gray) with cryo–electron microscopy structures of MAE (deep teal), RMND5A (cyan), UBE2H (brown), and ubiquitin (pink). The square highlights additional interactions between the phosphorylated tail of UBE2H with MAEA. (G) Close-up view of the phosphorylated tail of UBE2H (depicted as sticks) alongside the positive patch on MAEA. Residues from both the AlphaFold3 model and experimental cryo–electron microscopy structure are displayed in stick representation.

RING domains play a crucial role in mediating the catalytic activity of E3 ligases by facilitating the recruitment of E2 ubiquitin and the transfer of ubiquitin to the substrate (95). Using AlphaFold3, UBE2D1 and ubiquitin were modeled with MAEA and RMND5A (Fig. 8D). The accuracy of the model was evaluated using the previously described metrics, and the models were predicted with a high degree of confidence (table S2 and figs. S5 and S6, A and B). This revealed a canonical RING binding interface between RMND5A and E2. However, no visible interactions were observed with MAEA. This finding is consistent with previously characterized heterodimeric RING E3 ligases, where one active RING directly interacts with the E2, and the inactive RING plays a role in other protein-protein interactions (96). Unlike the E2, the activated ubiquitin forms extensive contacts with both RMND5A and MAEA (Fig. 8D). MAEA’s C-terminal tail inserts into a binding groove within ubiquitin, and a loop from the RING domain packs against the surface of ubiquitin, implicating MAEA in the catalytic activity of the E3 ligase complex (Fig. 8D). Previous experimental mutations in the yeast ortholog of MAEA (GID9) at residues C434 and C437 (corresponding to C314 and C317 in humans) to alanine resulted in catalytic inactivation (Fig. 8, A and C) (97).

A recent paper used cryogenic electron microscopy to determine the structure of human MAEA and RMND5A bound to ubiquitin and phosphorylated UBE2H (98). This experimental structure was published after AlphaFold3’s training cutoff, meaning that it was not included in its training data (98). The phosphorylated C-terminal tail of UBE2H was observed to extend and interact on a positive patch found on MAEA (Fig. 8, F and G). This patch is found on the helix involved in the coiled-coil domain and CTLH domain. Notably, an overlay of the experimental structure onto our AlphaFold3 model of MAEA and RMND5A reveals that the conserved residue R34 is positioned in both structures to interact with the phosphorylated tail (Fig. 8, E and G).

AlphaFold3 modeling of the MAEA disease-causing missense mutations

Replication-associated genetic instability can lead to the loss of cell fitness, developmental defects, genetic diseases, and cancer (51). Wolf-Hirschhorn syndrome is attributed to the loss of multiple genes on the short arm of chromosome 4, including MAEA (99). In addition, a comprehensive whole-exome sequencing study involving 1113 children with undiagnosed developmental disorders identified missense mutations in the MAEA gene (100). These mutations spanned various domains of MAEA (Fig. 9A). Further examination of the Decipher UK (101) and PubMed databases revealed additional patients with missense mutations in MAEA of undetermined significance (fig. S6C). Notably, three of these mutations have been linked to developmental delays, suggesting a potential connection between MAEA variants and replication-associated developmental abnormalities. The clinical phenotypes of patients with MAEA mutations exhibited notable similarities, including abnormalities of the head and neck, as well as the nervous system. Patients with the R34C mutation exhibited milder clinical symptoms compared to those with Y394D and M396R variants. Beyond the R34C phenotype, patients with Y394D and M396R mutations displayed pronounced developmental defects, such as seizures, unilateral polymicrogyria, additional nipples, hypermetropia, concave nails, congenital hip dislocation, cardiovascular abnormalities, eye anomalies, and integument issues (101). These phenotypic patterns suggest a likely partial or complete loss of MAEA function because of these mutations.

Fig. 9. MAEA patient variants impair its function in DNA replication and RS response.

(A) Schematic representations of MAEA patient mutants used in subsequent experiments. For all experiments, parental U2OS, MAEA KO, and MAEA KO cells stably expressed GFP-tagged WT MAEA or GFP-tagged MAEA variants mentioned in (A). (B and C) A schematic representation of the assay conditions is presented at the top of each graph. DNA fiber analysis of the indicated cells under unperturbed conditions (B) and HU-induced RS (C). The graphs display the percentage of stalled RFs for both normal (B) and RS conditions (C) by quantifying the percentage of green-only (IdU) tracks. Data are expressed as the mean values with SDs from four independent experiments. (D) A DNA fiber assay was performed as indicated at the top of the graph. The graph displays the CldU/IdU ratio with mean values and SDs. At least 150 fibers were analyzed per condition, and the data shown represent three independent experiments. (E) DNA fiber analysis of the indicated U2OS cells expressing different MAEA variants under unperturbed (E) and HU-induced RS (F). The graphs show the percentage of new origin firing quantified by the percentage of red-only (CldU) tracks. The labeling protocol used is shown at the top of each graph. The data shown represent three independent experiments. (G) IF analysis of RAD51 focus formation in the indicated cells treated with CPT (100 nM, 1 hour). The number of large RAD51 foci colocalized with γH2AX foci was quantified in at least 100 cells per replicate across three independent experiments. An ordinary one-way ANOVA with multiple comparisons test was performed to determine whether differences between conditions were statistically significant: **P < 0.01; ****P < 0.0001. Error bars indicate the means ± SD of the data.

Next, we used our AlphaFold3 models to predict the effects of MAEA disease-causing missense mutations on the E3 ubiquitin ligase activity of the CTLH complex. Of the 18 MAEA variants identified in patients, we focused our efforts on four of them (R34, E349, Y394, and M396). Three relatively conserved residues (E349, Y394, and M396) are observed to pack against ubiquitin (Fig. 8, A and D). E349 and Y394 form electrostatic interactions and hydrogen bonds, respectively, and M396 forms hydrophobic contacts with ubiquitin (Fig. 8D). Notably, Y394 specifically is highly conserved (Fig. 8A). Mutations in Y394, a key non-RING priming element, also ablate MAEA’s catalytic activity (81, 94). Studies of the Y394A mutant showed that MKLN1, a known substrate of MAEA and RMND5A, is degraded in cells expressing MAEA WT but not in cells expressing MAEA Y394A (81). The patient mutation MAEA Y394D likely disrupts key hydrogen bonding with ubiquitin, causing steric clashes and disruption of the E2~ubiquitin thioester, which hinders the efficient nucleophilic attack by the amine on the substrate lysine. Y394, with the help of E349, and M396 likely play a crucial role in orienting the E2~ubiquitin thioester, promoting the efficient nucleophilic attack by the amine on the substrate lysine. In addition, the patient mutation E349G within the RING domain likely disrupts MAEA’s interaction with ubiquitin, as E349 forms crucial electrostatic interactions with positive residues on ubiquitin (Fig. 8D). The C-terminal residue M396 also plays a crucial role by tightly packing against ubiquitin and forming hydrophobic interactions that further stabilize and position ubiquitin (Fig. 8D). A patient mutation M396R would likely disrupt this interaction, further compromising MAEA’s function. A recent experimental structure of human MAEA-RMND5A bound to ubiquitin and UBE2H implicates an N-terminal region of MAEA in E2 binding (96). Specifically, R34, a highly conserved residue, has been reported as a patient mutation (R34C) (Figs. 8E and 9A). It is intriguing to consider that this mutation could lead to a destabilization of the interaction with the E2 tail and a reduction in the overall activity of the E2-E3 complex. Collectively, this points to a compelling mechanism that demands further experimental validation.

Isogenic modeling of MAEA variants in individuals with undefined developmental delay

As we explained earlier, AlphaFold3 modeling of human MAEA revealed that MAEA missense mutations could have a deleterious impact on the E3 ubiquitin ligase activity of MAEA and its interaction with ubiquitin. Thus, we suspect that pathogenic MAEA variants can disrupt the function of MAEA in DNA replication and RS response, leading to developmental disorders in affected individuals. To test this, we used an isogenic cell system, using MAEA KO cells, to model the mutations identified in these patients. We created stable cell lines expressing each patient-specific MAEA mutation to assess the impact on MAEA’s role in DNA replication and HR repair. Immunoblot analysis indicated a reduction in MAEA protein levels in cells expressing the M396R mutation, unlike the R34C, E349G, and Y394D mutations, suggesting the posttranscriptional modulation of protein levels (fig. S6D). To explore the relationship between MAEA’s role in RF stability and the phenotypes of patients with MAEA mutations, we examined replication dynamics and genomic stability in MAEA KO cells expressing each variant. The isogenic cell lines, which expressed GFP-tagged MAEA variants, demonstrated increased spontaneous RF stalling, heightened RF stalling following HU treatment, greater fork degradation, and inability to suppress new origin firing under unperturbed and RS-induced conditions compared to cells expressing WT MAEA (Fig. 9, B to F). Crucially, these defects were rescued by expressing WT MAEA (Fig. 9, B to F). This supports the pathogenic nature of the R34C, K349G, Y394D, and M396R variants. To directly assess the impact of MAEA missense mutations on DNA repair and cell sensitivity to S phase–specific DNA-damaging agents, we measured RAD51 focus formation, HR repair, and cell survival in MAEA KO cells expressing each of the MAEA missense mutations. Cells harboring the R34C, K349G, Y394D, and M396R variants exhibited impaired RAD51 focus formation, defects in DSB repair via HR, and increased sensitivity to CPT (Figs. 9G and 10, A and B). These findings highlight the harmful effects of MAEA mutations on its role in DNA replication and repair, further supporting our conclusion that MAEA acts as a guardian of genome stability during RS. Defective MAEA’s function leads to genome instability, which underlies the developmental defects associated with MAEA deficiency.

Fig. 10. Model of how MAEA functions in the RS response.

(A) DR-GFP reporter assay in TRI-DR-U2OS cells transfected with siRNAs targeting luciferase (siCTRL) or MAEA (siMAEA) and complemented with indicated siRNA-resistant MAEA patient variants. Cells were treated with Dox to induce I-SceI expression. GFP-positive cells were quantified by flow cytometry. Results are based on three independent biological replicates. (B) Clonogenic survival assay of parental U2OS, MAEA KO, and MAEA KO cells stably expressing GFP-tagged WT MAEA or GFP-tagged MAEA variants mentioned in Fig. 9A, treated with different concentrations of CPT. Colony formation was assessed after 14 days. Survival data represent the means ± SD from five independent experiments. (C) Proposed model for the function of MAEA in the RS response. The model schematic was created with BioRender. Ismail, I. (2025) (https://biorender.com/m31g536). Content is licensed under a BioRender agreement.

DISCUSSION

In response to RS, replisomes are rapidly modified with ubiquitin chains through the coordinated actions of E3 ligases that either travel with the RF or are recruited when RF stalling occurs (102). However, much remains unknown about how ubiquitylation precisely regulates replisome processes during RS. Here, we identify MAEA as a regulator of the HR pathway. Several lines of evidence support MAEA’s role in HR and highlight its importance as a subject for further study. First, by promoting HR, MAEA plays a key role in maintaining the balance between NHEJ and HR. This balance is critical for genomic stability, as errors in DSB repair pathway choice can lead to gene rearrangements that drive cancer, while defective repair can result in lethal chromosomal aberrations (103). Second, MAEA ranks among the top 40 synthetic lethal genes identified in PARP1 KO and PARP1 inhibitor screens (95), mirroring the effects observed in the depletion of key HR proteins such as BRCA1 and BRCA2 (104, 105). Third, MAEA has also been associated with developmental disorders, a phenotype commonly linked to defective HR repair pathways, including ATR and CtIP mutations (106, 107). Last, genome-wide CRISPR KO screenings have identified MAEA as a potential gene conferring sensitivity to HU, thereby reinforcing its role in the RS response (108). These findings establish MAEA as an essential regulator of HR, making it a compelling candidate for further investigation and potential therapeutic targeting.

The precise mechanism through which MAEA ensures RF stability and facilitates checkpoint activation remains unclear. Our data suggest that MAEA plays a previously unrecognized role in enhancing the ATR-CHK1 response to RS. This is supported by the finding that MAEA-deficient cells exhibit impaired CHK1 phosphorylation upon exogenous RS despite higher levels of RPA-coated ssDNA, the primary trigger for ATR activation. Instead, we observed a reduction in the recruitment of ATR-ATRIP to stalled RFs in MAEA KO cells following HU treatment. These results align with the established model wherein ATR must localize to chromatin to phosphorylate CHK1 efficiently. However, whether MAEA facilitates ATR-ATRIP recruitment directly (e.g., through physical interactions) or indirectly (e.g., via RF remodeling) requires further investigation.

RAD51-dependent HR plays an essential role in stabilizing, protecting, and promoting the restart of stalled or damaged RFs. RAD51 loading on RFs stabilizes intermediates and prevents harmful nucleolytic processing (29, 41, 87). The loss of this protective activity cripples the repair and restart of damaged RFs, compromising genomic integrity. We observed that MAEA-depleted cells exhibit increased RF degradation, similar to BRCA1/2-deficient cells, suggesting that defects in RAD51 recruitment and/or stabilization led to RF degradation in the absence of MAEA. We found that RF degradation in MAEA-depleted cells was alleviated by co-depleting EXO1. In contrast, knocking down MRE11 and CtIP had no effect on RF degradation in the absence of MAEA. We found that in MAEA-deficient cells under RS, the exonuclease pathways regulating RF stability are dependent on the DNA lesion. Specifically, CPT triggers fork degradation through CtIP/MRE11, while HU activates EXO1-mediated degradation (Fig. 6C and fig. S3D). These findings highlight a fundamental divergence: BRCA2 and MAEA operate in distinct pathways to suppress exonuclease activity at stalled RFs (30, 33, 42). While BRCA2 safeguards forks by limiting MRE11 access, MAEA antagonizes EXO1 to prevent pathological resection. This mechanistic separation underscores that RF protection is not a monolithic process but rather governed by context-specific factors.

During the S phase, the NHEJ heterodimer KU initially sequesters the DSB ends, initiating NHEJ and inhibiting RAD51-mediated HR (53). NHEJ processing of DSBs during the S phase is cytotoxic, causing chromosomal aberrations and genetic instability by ligating distant DNA ends (52). Understanding how KU is regulated at RF ends is crucial. Recent studies have shown that KU transiently binds to stalled RFs induced by CPT and is later removed through the coordinated actions of MRN and CtIP in an ATM-dependent manner (52, 53). In addition, KU may undergo posttranslational modifications to reduce its affinity for DNA ends, a possibility supported by observations in yeast separation-of-function mutants (60). Alternatively, KU might be degraded at RF ends through mechanisms akin to those facilitating KU release from DNA upon completion of DSB repair (109, 110). Our previous work indicates that the E3 ubiquitin ligase RNF138 functions in conjunction with MRE11 to promote KU release (111). We found that MAEA associates with Ku and mediates its ubiquitylation, potentially promoting HR by modulating NHEJ components. Supporting this, depleting KU via siRNA restored RAD51 focus formation in MAEA KO cells following CPT treatment, suggesting that MAEA’s role in RAD51 focus modulation depends on the presence of KU. Although the depletion of MAEA and RNF138 similarly affects KU ubiquitylation, they exhibit distinct phenotypes. RNF138 operates during DNA end resection, while MAEA functions during the RPA-RAD51 exchange step, indicating that MAEA and RNF138 have distinct roles in HR.

Cells depleted of MAEA exhibit increased new origin firing following the induction of RS. We hypothesize that this elevated origin firing is a cellular response to an inability to complete DNA replication caused by the uncontrolled resection of stalled RFs as a result of a failure to stabilize RAD51. Whether RAD51 defects alone promote new origin firing is currently unclear. Increased new origin firing does not occur in human cells depleted of RAD51 or in BRCA2 null CHO cells following HU (87, 112), but increased new origin firing has been shown in BLM (Bloom syndrome helicase)–deficient cells and those lacking PALB2 (113, 114). It therefore remains to be determined whether the inability of MAEA-depleted cells to retain RAD51 at stalled RFs contributes to the increase in origin firing. Despite this, we predict that the increased origin firing in MAEA-deficient cells could contribute to genome instability, perhaps due to collisions between newly fired origins and damaged RFs lying in close proximity.

Our analysis of the DECIPHER and PubMed databases identified missense mutations in the MAEA gene in patients with developmental disorders, establishing these variants as a recurrent genetic cause of developmental pathology. Among these mutations, the M396R variant was found to decrease protein levels posttranscriptionally. AlphaFold3 modeling of human MAEA revealed that MAEA mutations could have a deleterious impact on the E3 ubiquitin ligase activity of MAEA and its interaction with ubiquitin, key functions for its role in DNA replication and RS response. In functional studies, MAEA KO cells expressing patient-derived variants exhibited three hallmark defects: impaired RAD51 focus formation after CPT treatment, increased RF stalling during both normal DNA replication and RS conditions, and enhanced RF degradation compared to WT MAEA-expressing cells. These findings provide a direct mechanistic link between MAEA mutations and replication-associated genome instability. The developmental phenotypes observed in patients may stem from two interrelated mechanisms. First, similar to hypomorphic ATR mutations that cause microcephaly and growth retardation (107), MAEA deficiency likely causes proliferation defects because of impaired RF stability, leading to reduced organismal size. Second, the exceptionally high replication demands of neural progenitor cells during brain development may make them particularly vulnerable to MAEA dysfunction, explaining the predominance of neurological symptoms in affected individuals (115). While our engineered cell models provide robust experimental evidence for MAEA’s role in RF stability, we acknowledge that the analysis of patient-derived samples would strengthen the clinical correlation. Our findings establish a compelling model in which pathogenic MAEA variants disrupt RF stabilization, leading to RS-induced genome instability that disproportionately affects developing tissues, particularly the rapidly dividing neural progenitor population. This work provides the foundation for future studies using patient-derived samples to further validate these findings.

In summary, our data led us to propose the following model (Fig. 10C): MAEA is a crucial component of the RF protection system. When an RF stall (for example, because of CPT exposure), KU binds to the DNA ends generated during RF reversal. Similar to DSB end processing in HR repair, limited nucleolytic resection facilitates the removal of KU from the reversed forks, enabling RAD51-dependent HR and the repair or restart of the stalled RF. MAEA stabilizes RAD51 at these structures and restricts extensive DNA end resection by EXO1. In the absence of MAEA, RAD51 is destabilized on ssDNA, making the reversed fork vulnerable to excessive resection by EXO1, which can lead to genome instability. Our data identify MAEA as a disease-associated gene crucial for regulating cellular replication and cell cycle checkpoints. Further exploration into the mechanisms by which MAEA operates will yield fundamental insights into how cells safeguard RF integrity and mitigate the risk of human disease.

MATERIALS AND METHODS

Plasmids and siRNA transfection

All siRNAs that were used in this study were obtained from Sigma-Aldrich, Thermo Fisher Scientific, and Qiagen (table S3). siRNA transfections were performed according to the manufacturer’s instructions; all siRNA transfections were performed with 20 to 60 nM siRNA using RNAiMAX (Invitrogen) as a transfection reagent. All point mutants were generated using the Q5 site-directed mutagenesis kit (New England Biolabs) and verified by Sanger sequencing (table S4). The GFP-MAEA WT and MAEA patient variants were cloned into the eGFP-C1 vector using standard protocols. The guide RNA sequences for the CRISPR KO of MAEA in U2OS cells are listed in table S5.

Cell culture

U2OS cells were purchased from American Type Culture Collection and cultured in a Dulbecco’s modified Eagle’s medium (DMEM) or DMEM/F12 medium containing 10% fetal bovine serum at 37°C and 5% CO₂. Human embryonic kidney (HEK) 293T DSB-Spectrum_V3 cells were cultured in DMEM containing 10% fetal bovine serum. Cells were routinely tested for mycoplasma using DAPI (4′,6-diamidino-2-phenylindole) and were all mycoplasma-free. CPT, aphidicolin, and hydroxyurea were purchased from Sigma-Aldrich. The drugs were dissolved in dimethyl sulfoxide (DMSO), DMSO, and water, respectively, and stored at −20°C.

Assessment of z scores for siRNAs

To score siRNAs during the primary screen according to their potency in affecting RAD51 focus formation, we ranked each duplicate transfection according to their z score (Z), which allowed correcting for plate-to-plate variations. For each RNAi, we derived a population mean X and SD (111). To center X around the nontargeting control (ctr), we calculated an adjustment factor alpha D median(X_ctr)/sd(X_ctr) and used the nontargeting control as follows: Z = (X − μ) × α/σ, where μ is the median nontargeting control between wells per plate, and σ is the SD of the sample population of X (111).

IF staining

IF staining was performed as described previously (111, 116). After CPT/HU treatment, cells were washed with cold phosphate-buffered saline (PBS) and fixed with 4% paraformaldehyde (PFA) at room temperature (RT). For RAD51, RPA, and BrdU immunodetection, cells were pre-extracted with cold extraction buffer (25 mM Hepes, pH 7.9, 300 mM sucrose, 50 mM NaCl, 1 mM EDTA, 3 mM MgCl₂, and 0.5% Triton X-100) for 3 min at RT before being fixed at the indicated incubation time points after CPT/HU treatment. Cells were washed with PBS once before fixation with 4% (w/v) PFA in PBS for 20 min at RT. Cells were washed with PBS and permeabilized in 0.5% Triton X-100 in PBS for 3 min. Cells were incubated with a primary antibody for 1 hour (table S6). The unbound primary antibody was removed by rinsing with 0.1% Triton X-100 in PBS at RT followed by three washes with PBS and incubation with the appropriate secondary antibodies (table S7) for 1 hour at RT. Slides were then washed three times in PBS before mounting with Vectashield mounting medium (Vector Laboratories). For Ku staining, cells were washed with cold PBS and extracted with cold Ku extraction buffer (10 mM Pipes, pH 7.0, 100 mM NaCl, 300 mM sucrose, and 3 mM MgCl₂) containing 0.7% Triton X-100 and ribonuclease A (0.3 mg/ml) two times for 3 min at RT. Cells were then washed two times with PBS and fixed with 2% (w/v) PFA in PBS for 20 min at RT. Cells were washed with PBS and permeabilized in 0.5% Triton X-100 in PBS for 3 min. Cells were incubated with primary and secondary antibodies for 1 hour each at RT (tables S6 and S7). Coverslips were observed using an upright fluorescence microscope (Zeiss AxioImager.Z1) with a Plan Neofluar 1.3–numerical aperture (N.A.) ×40 oil immersion objective. The brightness and contrast were scaled evenly among all samples within an experiment.

Western blotting

Cells were lysed in 1× SDS–polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer [125 mM tris, pH 6.8, 4% SDS, 20% glycerol, 0.1% bromophenol blue, and 5% 2-mercaptoethanol (BME)] supplemented with 1× protease inhibitor (cOmplete EDTA-free tablet, Roche) and 1× phosphatase inhibitor (PhosSTOP, Roche). After transferring on a nitrocellulose membrane (0.2 μm) (110 V; 90 min) in transfer buffer (25 mM tris-HCl, 0.2 M glycine, and 20% methanol), membranes were blocked with 4% fish skin gelatin (FSG) or 5% skimmed milk diluted in TBS-T buffer (tris-buffered saline buffer with Tween 20, 0.1%) for at least 1 hour at RT and next incubated overnight at 4°C or 1 hour at RT with primary antibodies diluted in 2% FSG in TBS-T. A panel of commercially available primary antibodies directed against various DNA damage proteins was used (table S6). Next, the membranes were washed three times with TBS-T for 30 min and incubated with secondary horseradish peroxidase–conjugated or infrared-labeled (IRDye 680RD) secondary antibodies (all LI-COR Biosciences) (table S7) diluted in 2% FSG in TBS-T. After three more washes with TBS-T, horseradish peroxidase activity was detected using the Odyssey Fc Imaging System (LI-COR Biosciences) after adding the substrate for the Amersham ECL Prime reagent (Cytiva).

GFP selector coimmunoprecipitation

Pellets of cells expressing the GFP construct from a 100-mm dish were resuspended in ice-cold NETN-150 (50 mM tris, pH 8.0 at 4°C, 150 mM NaCl, 0.5% IGEPAL CA-630, and 1 mM EDTA) supplemented with fresh 2× cOmplete, 1.25× PhosSTOP, and 50 mM N-ethylmaleimide and shaken on ice at 250 rpm for 20 min. The lysate was then centrifuged at 20,000g for 15 min at 4°C, after which the pellet was discarded. A total of 10% of the supernatant was reserved as an input control, mixed with 2× SDS-PAGE sample buffer, and denatured at 95°C for 5 min. The remaining 90% of the supernatant was mixed with 20 μl of GFP Selector agarose beads (NanoTag Biotechnologies) that had been prewashed twice in ice-cold NETN-150 buffer and then incubated end over end at 4°C for 1 hour. To eliminate nonspecific binding, the beads were washed twice with ice-cold NETN-150 buffer. Each wash involved vortexing for 20 s, centrifuging at 3000g for 2 min, and aspirating the supernatant. Bound proteins were eluted by adding 2× SDS sample buffer with 5% BME to the beads and heating on a ThermoMixer at 95°C for 30 min at 1200 rpm. Both the input and eluate fractions were subsequently analyzed by SDS-PAGE and immunoblotting. For the eluate fraction, the modification of interest (e.g., phosphorylation) was first detected, followed by stripping the membrane and reprobing to identify the immunoprecipitated GFP-tagged protein.

Isolation of proteins on nascent DNA (iPOND)

iPOND was performed as previously described (83, 117). HEK293T cells were pulse labeled with 10 μM EdU (Life Technologies) for 10 min. For the thymidine chase experiments, cells were washed three times with complete medium and then incubated for 60 min in a medium supplemented with 10 μM thymidine (Sigma-Aldrich). Protein-DNA cross-linking was carried out using 1% formaldehyde for 15 min at RT, followed by quenching with 0.125 M glycine for 5 min and washing three times with PBS. Cells were then permeabilized with 0.25% Triton X-100 in PBS for 30 min and washed once with 0.5% bovine serum albumin (BSA) in PBS and then once with PBS. The cells were incubated in click reaction buffer (10 mM sodium l-ascorbate, 20 μM biotin azide, and 2 mM Cu₂SO₄) for 2 hours at RT. In the “no-click” control sample, DMSO was used instead of biotin azide in the click reaction buffer. After washing once with 0.5% BSA in PBS and then with PBS, cells were resuspended in lysis buffer (50 mM tris-HCl, pH 8.0, and 1% SDS) supplemented with protease/phosphatase inhibitors (Roche). Chromatin was solubilized by sonication for 2 min (30 s on and 30 s off) at 4°C using an amplitude of 30 on a Model 705 Sonic Dismembrator fitted with a microtip probe (Thermo Fisher Scientific). Samples were centrifuged for 10 min at 5000g, and the supernatants were diluted 1:1 with PBS (vol/vol) containing protease/phosphatase inhibitors and incubated overnight at +4°C with streptavidin-agarose beads (EMD Millipore). Beads were washed twice with lysis buffer, once with 1 M NaCl, and twice more with lysis buffer and eluted by boiling in 2× SDS sample buffer (Life Technologies) containing 200 mM dithiothreitol for 10 min at 95°C. Proteins were then resolved by electrophoresis and detected by Western blotting.

Cell cycle synchronization

HeLa cells were synchronized at the G₁-S boundary using a double thymidine block (118). Specifically, the cells were seeded to 40 to 50% confluency, and 3 to 4 mM deoxythymidine was added to the culture media for 14 to 18 hours (first block). After this initial block, the cells were washed twice with PBS and allowed to progress through the cell cycle in warmed DMEM for 12 hours. A second addition of 4 mM deoxythymidine was made to the culture media for another 18 hours (second block). Last, the cells were released by washing twice with PBS and then replaced with warmed culture media for the duration required to reach the desired cell cycle phase (e.g., 3 hours for the S phase, 7 hours for the G₂ phase, and 12 hours for the G₁ phase).

Cell cycle analysis

Asynchronous and synchronized cells were trypsinized, washed once with PBS, and then resuspended in 100 μl of PBS. Ice-cold 70% ethanol in 1× PBS was added dropwise, and the samples were incubated overnight at −20°C. The following day, the samples were washed once with PBS and treated with ribonuclease A (100 μg/ml) in PBS containing 3.8 mM sodium citrate for 30 min at 37°C with agitation. Propidium iodide was then added to achieve a final concentration of 50 μg/ml, and the cells were incubated for an additional 30 min. Flow cytometry was performed on the processed samples using a BD FACSCanto II (BD Biosciences) to detect propidium iodide fluorescence, with gating set for forward and side scatter.

DNA fiber assays

Fiber assays were conducted as previously described (118). Briefly, asynchronous cells were labeled with thymidine analogs: 20 μM IdU followed by 200 μM CldU. The specific labeling scheme and treatments for each experiment are detailed at the top of each graph. Cells were collected by trypsinization, washed, and resuspended in 60 μl of PBS. Next, 2 μl of the cell suspension was placed onto a polarized slide (Denville Ultraclear), and cell lysis was performed in situ by adding 10 μl of lysis buffer (200 mM tris-HCl, pH 7.5, 50 mM EDTA, and 0.5% SDS). The resulting DNA spreads were air dried for 40 min, followed by fixation for 5 min in a 3:1 methanol/acetic acid solution, and then refrigerated overnight. For immunostaining, the stretched DNA fibers were denatured with 2.5 M HCl for 30 min, washed three times for 5 min each in PBS, and then blocked with 5% BSA in PBS for 60 min at RT. Antibodies used to visualize CldU- and IdU-labeled tracts included rat anti–CldU/BrdU, chicken anti-rat Alexa 488, mouse anti–IdU/BrdU, and goat anti-mouse IgG1 Alexa 547 (tables S6 and S7). Labeled tracts were visualized using an upright fluorescence microscope (Zeiss AxioImager.Z1) with a Plan Neofluar 1.3-N.A. ×40 oil immersion objective, and tract lengths were measured using MetaXpress 6 software (Molecular Devices LLC). A total of 100 to 400 fibers was analyzed for each condition in every experiment, and all analyses were conducted with sample blinding. The statistical analysis of tract lengths was performed using GraphPad Prism.

Colony formation assay

Parental or MAEA KO cell lines were transfected as specified in the figure legends for 24 hours. Cells were then seeded in duplicate or triplicate into six-well plates at a density of at least 800 cells per well and allowed to settle at 37°C for 8 to 10 hours. Subsequently, they were treated with drugs at the indicated concentrations for 24 hours at 37°C. After treatment, the medium in each dish was replaced, and colonies were allowed to form over 14 days at 37°C. The colonies were then fixed and visualized using 0.5% crystal violet in 25% methanol. Colonies containing at least 50 cells were scored and counted, and the surviving fraction was calculated accordingly (64).

Ni-NTA pull-down assay

HeLa HB-ubiquitin cells (119) from 150-mm dishes were harvested as described above, but before pelleting for flash freezing, the cells were resuspended in ice-cold PBS, and 10% of each sample was set aside as the input control. All samples were then pelleted by centrifugation at 525g for 5 min at 4°C and flash frozen in liquid nitrogen before being stored at −80°C. The 10% input control was processed separately for lysis to prepare the whole-cell extract (as described above), while the remaining samples were processed for nickel affinity purification, as detailed here. All buffers were prepared no more than 4 hours before use. Washes involved vortexing in the specified buffer for at least 20 s, followed by centrifugation at 750g for 2 min and the removal of the supernatant via vacuum aspiration. Cell pellets were resuspended in ice-cold guanidine lysis buffer (6 M guanidine-HCl, 100 mM sodium phosphate buffer, pH 8.0, 10 mM tris, 5 mM imidazole, and 5 mM BME) by vortexing and sonicated for 1 min (amplitude of 25, Thermo Fisher Scientific Model 705 Sonic Dismembrator with a microtip probe). The lysate was then mixed with 150 μl of Ni-NTA agarose beads (Qiagen) that had been prewashed three times in guanidine lysis buffer. The mixture was agitated on a rocker for 4 hours at RT. Nonspecific interactions were eliminated through sequential washes at RT: first in guanidine wash buffer (6 M guanidine-HCl, 100 mM sodium phosphate buffer, pH 8.0, 10 mM tris, 10 mM imidazole, 0.1% Triton X-100, and 5 mM BME), followed by a wash in pH 8 urea wash buffer (8 M urea, 100 mM sodium phosphate buffer, pH 8.0, 10 mM tris, 10 mM imidazole, 0.1% Triton X-100, and 5 mM BME), and lastly three washes in pH 6.3 urea wash buffer (8 M urea, 100 mM sodium phosphate buffer, pH 6.3, 10 mM tris, 0.1% Triton X-100, and 5 mM BME). Bound proteins were eluted from the beads using Ni-NTA elution buffer (150 mM tris, pH 6.7, 200 mM imidazole, 5% SDS, 30% glycerol, 0.05% bromophenol blue, and 5% BME) at 60°C for 30 min on a ThermoMixer F1.5 (Eppendorf) set to 1200 rpm, and the proteins were resolved by SDS-PAGE for subsequent immunoblotting.

In vivo HR reporter assay

For each condition, ~4 × 10⁶ TRI-DR-U2OS cells were electroporated with 60 nM siRNA and, if necessary, 2 μg of Flag-tagged MAEA constructs using a 4D-Nucleofector X Unit (program CM-104) along with the SE Cell Line 4D-Nucleofector X Kit L (both from Lonza Bioscience) (116). Eight hours posttransfection, Dox (1 μg/ml) was added to the culture medium for 24 hours to induce I-SceI expression. The culture medium was then replaced, and the cells were cultured without Dox for an additional 24 hours. Cells were collected following the cell harvesting procedure outlined above; however, instead of being flash frozen, the cells were resuspended in 2% PFA in PBS and incubated for 20 min for fixation. The cells were then washed three times in PBS. The frequency of GFP-positive cells was measured by flow cytometry (FACSCanto II, BD Biosciences), analyzing at least 100,000 cells.

DSB Spectrum assay

The DSB-Spectrum_V3 assay was conducted with minor modifications to the original protocol (71). HEK293T cells, stably expressing the DSB-Spectrum_V3 reporter (provided by M. B. Yaffe), were transfected twice at 24-hour intervals with the specified siRNAs using Lipofectamine RNAiMAX (Invitrogen). Twenty-four hours following the second transfection, cells were transfected with either pX459-Cas9-sgAAVS1-iRFP or pX459-Cas9-sgBFP-iRFP plasmids using TransIT-293 (Mirus). On the subsequent day, cells were replated and allowed to proliferate for an additional 72 hours. Postincubation, cells were trypsinized and subjected to flow cytometry analysis using a BD FACSymphony A1 (BD Biosciences). The frequency of each fluorescent subpopulation in the AAVS1sg-transfected cells was subtracted from the frequency of the corresponding population in the BFPsg-transfected cells. The resulting background-corrected frequencies were then expressed as a fold change, normalized to the control siRNA.

SIRF and PLA assays

U2OS cells were grown on coverslips to 50 to 60 % confluency for PLA and SIRF assay. On the day of the experiment, cells were labeled with EdU (125 μM) for 10 min to visualize replicating cells in the S phase. Next, they were washed two times with warm PBS and treated with or without CPT (100 nM, 1 hour) and HU (2 mM, 4 hours). After treatment, cells were washed with ice-cold PBS and extracted for 3 min at RT. Next, cells were fixed with 2% PFA for 15 min at RT and washed three times with PBS. Then, they were permeabilized with 0.5% Triton X-100 in PBS for 15 min, followed by three washes with PBS. For MAEA/Ku PLA, cells were first extracted with CSK (cytoskeleton) buffer for 3 min and then fixed with 2% PFA for 15 min at RT. Subsequently, cells were incubated with a fresh Click-iT reaction cocktail (2 mM copper sulfate, 10 μM biotin azide containing 1:10 Alexa 488 azide, and 100 mM sodium ascorbate in PBS) at RT for 1 hour. Following the Click-iT reaction, the next steps followed the manufacturer’s protocol (Duolink In Situ, Sigma-Aldrich) using the appropriate primary and secondary antibodies (tables S6 and S7).

AlphaFold3 modeling and analysis

AlphaFold3 modeling was performed using the latest implementation available at https://alphafoldserver.com (93). Protein sequences were retrieved from UniProt with the following accession codes: human MAEA (Q7L5Y9), human RMND5A (Q9H871), UBE2D1 (P51668), and ubiquitin (P0CG48). These sequences were input into the AlphaFold3 web server, and each model underwent 10 recycling iterations. The top 5 models were ranked on the basis of pTM and ipTM values, and the top ranked model was further characterized. Structural characterizations and analyses were conducted using PyMOL (The PyMOL Molecular Graphics System, version 3.0, Schrödinger LLC). Confidence metrics were plotted using Microsoft Excel (version 2404) and Morpheus (https://software.broadinstitute.org/morpheus).

Sequence alignments

Protein sequence alignments were performed using Clustal Omega (European Bioinformatics Institute, European Molecular Biology Laboratory) (92). Amino acid sequences of MAEA orthologs were obtained from UniProt with the following accession codes: human (Q7L5Y9), mouse (Q4VC33), chicken (Q5F398), Xenopus laevis (Q6GR10), Danio rerio (Q7SXR3), and Saccharomyces cerevisiae (P40492). Sequence alignment visualization and annotation were conducted using Jalview (version 2.11.0).

Image processing and statistical analysis

All IF images were analyzed using the cell analysis tool in Imaris software (Oxford Instruments). The region of interest was designated for colocalization studies to identify nuclei, and a separate channel was created to capture the colocalized signal. Focus counts were then analyzed and compared from this channel. Composite figures of the collected images were assembled in Adobe Photoshop 2022 (Adobe) and labeled in Illustrator 2017 (Adobe). Higher-magnification images were obtained using a 1.4-N.A. Plan-Apochromat 63× or 100× objective lens (Carl Zeiss Inc.). All microscopic and immunoblot images were adjusted for brightness and contrast in Adobe Photoshop and then organized and labeled in Adobe Illustrator. Bar graphs and scatter plots were created using Prism (GraphPad) to display the means and SDs (error bars). To assess statistical significance, two-tailed, unpaired, nonparametric Student’s t tests (Mann-Whitney) and a one-way analysis of variance (ANOVA) with multiple comparisons were conducted in Prism. Asterisks indicate statistically significant differences: ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Schematic diagrams in Fig. 10 were prepared using BioRender.

Supplementary Materials

The PDF file includes:

Figs. S1 to S6

Legend for table S1

Tables S2 to S7

Other Supplementary Material for this manuscript includes the following:

Table S1