SUMOylation of HNRNPA2B1 modulates RPA dynamics during unperturbed replication and genotoxic stress responses

Summary

Replication Protein A (RPA) is a major regulator of eukaryotic DNA metabolism involved in multiple essential cellular processes. Maintaining appropriate RPA dynamics is crucial for cells to prevent RPA exhaustion, which can lead to replication fork breakage and replication catastrophe. However, how cells regulate RPA availability during unperturbed replication and in response to stress has not been well elucidated. Here, we show that HNRNPA2B1^SUMO functions as an endogenous inhibitor of RPA during normal replication. HNRNPA2B1^SUMO associates with RPA through recognizing the SUMO-interacting motif (SIM) of RPA to inhibit RPA accumulation at replication forks and impede local ATR activation. Declining HNRNPA2^SUMO induced by DNA damage will release nuclear soluble RPA to localize to chromatin and enable ATR activation. Furthermore, we characterize that HNRNPA2B1 hinders homologous recombination (HR) repair via limiting RPA availability, thus conferring sensitivity to PARP inhibitors. These findings establish HNRNPA2B1 as a critical player in RPA-dependent surveillance networks.

In brief

In this study, Zhu et al. show that HNRNPA2B1^SUMO associates with RPA to inhibit RPA accumulation at replication forks during normal replication. Declining HNRNPA2^SUMO induced by DNA damage will release nuclear soluble RPA to localize to chromatin and enable ATR activation.

Graphical Abstract:

Introduction

Heterogeneous nuclear ribonucleoprotein A2/B1 (HNRNPA2B1) is one of the most abundant RNA-binding proteins (RBPs) that belongs to the HNRNP protein family^1,2. HNRNPA2B1 is expressed as two alternatively spliced isoforms, HNRNPA2 and HNRNPB1. The shorter isoform HNRNPA2 lacks 12 amino acids at its N terminus and is the major isoform in most tissues^3,4. HNRNPA2B1 contributes to multiple aspects of RNA metabolism^5–8. Overexpression of HNRNPA2B1 has been described in many human cancers^9–11. Previous studies have linked HNRNPA2B1’s function in mRNA processing to cancer cell migration, epithelial-mesenchymal transition and aerobic glycolysis^10,12,13. In addition to prominent functions in RNA biology, HNRNPA2B1 also functions in telomere maintenance and infection by DNA viruses^14,15. However, the potential role of HNRNPA2B1 in the regulation of DNA replication stress and DNA damage response (DDR) has not yet been definitively established.

The Replication Protein A (RPA) heterotrimeric complex, comprised of the RPA70 (RPA1), RPA32 (RPA2) and RPA14 subunits, plays critical roles in almost all aspects of DNA metabolism¹⁶. In eukaryotic cells RPA is the major single-stranded DNA (ssDNA)-binding protein via its oligonucleotide-binding fold domains¹⁷. Obstacles for DNA replication forks often generate increasing amounts of ssDNA which are rapidly coated by RPA, thus forming a nucleoprotein structure that recruits the ATR-ATRIP kinase complex and its regulators¹⁸. RPA is also involved in homologous recombination (HR) repair. When a DNA double-strand break (DSB) occurs in a cell, the binding of RPA to the resected DNA end allows for the formation of RAD51 nucleoprotein filament to catalyze DNA strand exchange for high-fidelity repair of DSBs by HR¹⁹. Moreover, RPA exhaustion induced by ATR inhibition during replication stress will generate unprotected ssDNA and the ensuing DNA breakage²⁰, suggesting that keeping appropriate RPA dynamics is crucial to protect cells against replication catastrophe. Yet how cells regulate RPA availability during DNA replication under unperturbed and stressed conditions is largely unknown. HNRNPA2B1 was reported to displace RPA from telomeric ssDNA in a cell cycle-dependent manner, thus contributing to the prevention of persistent ATR activation at telomeres²¹. It remains elusive whether the dynamics of RPA in the contexts of DNA replication stress and DDR are also regulated by HNRNPA2B1.

Post-translational modification by the small ubiquitin-like modifier (SUMO), referred to as SUMOylation, is an important event for substrate protein regulation²². SUMOylation is regulated by the dedicated enzymes of SUMO-activating enzyme (E1, SAE1), SUMO-conjugating enzyme (E2, UBC9) and SUMO ligase (E3)²³. SUMO conjugates can be removed by Sentrin (SUMO)-specific proteases (SENPs)²⁴. The three ubiquitous SUMO isoforms (SUMO1, SUMO2 and SUMO3) can be covalently conjugated in the form of a single moiety (SUMO1) or in the form of polymeric chains (SUMO2 and SUMO3)²⁵. Mature SUMO2 and SUMO3 are ~95% identical in sequence and thus referred as SUMO2/3²⁶. Additionally, a large number of proteins have been described to non-covalently associate with SUMO modules through SUMO-interacting motifs (SIMs) implicated in a range of cellular processes^27,28.

In this study, we characterized HNRNPA2B1^SUMO as an endogenous inhibitor of RPA through its recognition of the SIM motif in RPA. Our findings elucidate how HNRNPA2B1^SUMO participates in replication and the genotoxic stress responses via regulating RPA dynamics.

Results

HNRNPA2 is SUMOylated at K108 by PIAS2

A previous study described that HNRNPA2B1 is SUMOylated in exosomes based on an observed shift in molecular weight and anti-SUMO1 signal⁸. SUMOylated HNRNPA2B1 plays a role in sorting miRNA into exosome, but whether SUMOylation of HNRNPA2B1 has a nuclear function or not is unclear. Given that SUMOylation plays a pivotal role in controlling the function of HNRNP protein family^29–31, it deserves deeper examination. To investigate whether HNRNPA2B1 is a bona fide SUMOylation substrate, FLAG-tagged SUMO1 or SUMO2/3 were transiently transfected into cells. HNRNPA2B1 is commonly detected as two alternatively spliced isoforms: HNRNPA2 and HNRNPB1. However, our Western blot analysis of HNRNPA2B1 from inputs of U2OS cells (Figure 1A) and HEK293T cells (Figure S1A) showed two obvious higher molecular weight bands (~10 kDa more) in comparison with the main HNRNPA2B1 bands (the expected normal size of HNRNPA2B1 is 37 kDa), indicating that HNRNPA2B1 might be post-translationally modified in cells. We observed that HNRNPA2B1 could be strongly detected by SUMO1 immunoprecipitation but not by the highly related proteins SUMO2/3 (Figures 1A and S1A), as shown by two specific shifted bands of SUMOylated HNRNPA2 (HNRNPA2^SUMO) and HNRNPB1 (HNRNPB1^SUMO).

Figure 1. HNRNPA2 is SUMOylated by PIAS2 at K108.

(A) U2OS cells transfected with indicated constructs were subjected to immunoprecipitation. Copurified HNRNPA2^SUMO and HNRNPB1^SUMO were examined by immunoblot using anti-HNRNPA2B1 antibody. (B) HEK293T cells were co-transfected with FLAG-HNRNPA2, HA-UBC9 and His-tagged SUMO1/2/3 as indicated. SUMO conjugated proteins were pulled down by Ni²⁺-NTA resin. (C) PLA assays were performed in U2OS cells using indicated antibodies. (D) GFP-HNRNPA2 and His-SUMO1, HA-UBC9 were transfected with or without FLAG-SENP1 into HEK293T cells. SUMO1 conjugated proteins were pulled down by Ni²⁺-NTA resin. (E) Knockdown of PIAS family SUMO E3 ligases by shRNAs and its effect on the expression of HNRNPA2B1^SUMO were examined in U2OS cells. (F) Recombinant HNRNPA2 was incubated with SAE1, UBC9, SUMO1, ATP and an increasing concentration of PIAS2 for in vitro SUMOylation assay. (G) Wild-type (WT) and mutants of HNRNPA2 with His-SUMO1, HA-UBC9 were transfected into HEK293T cells as indicated. SUMO1 conjugated HNRNPA2 proteins enriched by Ni²⁺-NTA resin were immunoblotted. (H) Schematic representation of the human HNRNPA2 showing the RNA recognition motif (RRM) and prion-like domain (PrLD). The position of SUMOylation site K108 and the SUMOylation consensus sequence are indicated. (I) Bacterially expressed GST-HNRNPA2 WT or K108R were subjected to in vitro SUMOylation reactions in the presence or absence of SUMO1, SAE1, UBC9 and ATP as indicated. (J) PLA of the association of SUMO1 with FLAG-tagged WT and K108R mutant of HNRNPA2 in U2OS cells. Scale bars, 10 μm. See also Figure S1.

To further validate that HNRNPA2B1 is modified by SUMOylation, we transfected FLAG-HNRNPA2 together with His-tagged SUMO1 or SUMO2/3 and the SUMO-conjugating enzyme E2 UBC9 into HEK293T cells. As shown in Figure 1B, His-SUMO1 conjugated HNRNPA2 was pulled down by Ni²⁺-NTA resin under denaturing conditions with a size of ~47 kDa, which suggests HNRNPA2 covalently conjugated with one molecule of SUMO1 (HNRNPA2^SUMO). However, we did not detect HNRNPA2 SUMOylation by His-tagged SUMO2/3. To examine that HNRNPA2B1 is also endogenously modified by SUMO1, co-immunoprecipitation (Co-IP) assays were performed with anti-SUMO1 antibody followed by immunoblotted with anti-HNRNPA2B1 antibody. The results showed that HNRNPA2B1 was modified by endogenous SUMO1 with a size of ~47 kDa (Figure S1B). N-Ethylmaleimide (NEM) can be used to stabilize SUMO conjugates, the absence of NEM results in almost complete loss of high molecular weight HNRNPA2^SUMO (Figure S1C). Cells treated with ML-792, a potent and selective SUMO E1 enzyme inhibitor, significantly decreased the HNRNPA2^SUMO level (Figure S1D).

To extend the above findings, we used a proximity-ligation assay (PLA) to analyze the association between HNRNPA2B1 and SUMO1 in cells. Fluorescent PLA dots indicative of close HNRNPA2B1/SUMO1 association were detected mainly in the nucleus (Figure 1C), where both HNRNPA2B1 and the SUMO machinery are enriched (Figure S1E). Altogether, these results demonstrated that HNRNPA2B1 is a SUMO1 substrate. Given that the shorter isoform HNRNPA2 is the main isoform accounting for ~90% of this protein in most tissues³, we focused on HNRNPA2 in the following studies.

UBC9 is the only well-characterized E2 conjugation enzyme for SUMO modification³², leading us to hypothesize that HNRNPA2 may physically interact with UBC9. As expected, Co-IP results revealed that HNRNPA2 indeed interacted with UBC9 (Figure S1F). We also checked HNRNPA2 SUMOylation in the presence of the deSUMOylating enzyme SENP1. As shown in Figure 1D, His-SUMO1 conjugated HNRNPA2 was pulled down and then immunoblotted to detect HNRNPA2^SUMO. HNRNPA2^SUMO was significantly enhanced by UBC9 overexpression, but this increase was efficiently reversed by co-expression with SENP1. To identify the putative HNRNPA2B1-specific SUMO E3 ligase, each of the E3 ligases PIAS1-4 was depleted using specific shRNAs. Among them, we observed PIAS2 to be a major E3 ligase involved in HNRNPA2B1 SUMOylation (Figure 1E). Knocking down PIAS2 led to a considerable reduction of HNRNPA2^SUMO and HNRNPB1^SUMO. To further confirm this finding, we performed in vitro SUMOylation assay. Incubation of recombinant GST-tagged HNRNPA2 in the presence of SAE1 (SUMO E1), UBC9 (SUMO E2), SUMO1 and an increasing concentration of PIAS2 efficiently drove a concentration-dependent SUMOylation of HNRNPA2 (Figures 1F and S1G), establishing PIAS2 as a SUMO E3 ligase for HNRNPA2.

To determine the potential SUMOylation site(s) of HNRNPA2, we utilized two independent computational programs: SUMOplot^™ Analysis Program (https://www.abcepta.com/sumoplot) and GPS-SUMO³³, to identify putative SUMOylation sites of HNRNPA2. Three lysine residues K108, K125 and K305 were observed with high prediction scores (Figures S1H and S1I). Furthermore, we found that mutation of lysine 108 to arginine (K108R), but not those of other two sites (K125R/K305R), inhibited HNRNPA2 SUMOylation (Figure 1G). SUMOylation generally occurs on the lysine residue within the consensus sequence [Ψ]-[K]-[X]-[E/D]²², where Ψ is a large hydrophobic residue and X represents any amino acid. K108 (corresponding to K120 of HNRNPB1) is embedded within a consensus SUMO sequence (Figure 1H). To further confirm that K108 is the SUMOylation site of HNRNPA2, we performed in vitro SUMOylation assays using recombinant wild-type (HNRNPA2^WT) and K108R mutant (HNRNPA2^K108R) proteins. As shown in Figure 1I, in contrast to HNRNPA2^WT, the HNRNPA2^SUMO band was abolished in our HNRNPA2^K108R sample. Additionally, we obtained discrete PLA foci of HNRNPA2^WT and SUMO1 (Figure 1J), but no such signals were visualized in HNRNPA2^K108R expressing cells indicative of failed association. In summary, our results suggest that HNRNPA2 is SUMOylated at K108 by PIAS2. Moreover, our results are corroborated by previous mass spectrometry studies showing the presence of SUMOylation at HNRNPA2 (K108)^34,35.

RPA1 associates with HNRNPA2B1^SUMO through its SIM motif

HNRNPA2B1 is a well-known RNA-binding protein involved in multiple aspects of RNA metabolism. Interestingly, HNRNPA2B1 was listed as an RPA-ssDNA-binding candidate in a previous proteomic screening³⁶, indicating HNRNPA2B1 might contribute to the cellular functions of RPA. This is in contrast to the reported role of HNRNPA2B1 in competing with RPA in telomere DNA binding²¹. We also found that RPA1 and RPA2, two subunits of RPA complex, were able to bind HNRNPA2B1 in cell extracts (Figure 2A). We demonstrated that these interactions were resistant to DNase I or RNase A treatment (Figure S2A), suggesting the DNA or RNA-independent associations between HNRNPA2B1 and RPA. Furthermore, we observed robust PLA signals between endogenous HNRNPA2B1 and RPA1/2, implying close proximity between HNRNPA2B1 and RPA in live cells (Figure 2B). It was previously reported that HNRNPA2B1 expression changes throughout the cell cycle³⁷. Indeed, HNRNPA2B1 as well as HNRNPA2B1^SUMO displayed a fluctuating expression pattern: with upregulation at S phase followed by a decline at G2/M, and even less expressing at G1 (Figures S2B and S2C). Accordingly, the interactions between HNRNPA2B1 and RPA1/2 were more efficiently detected at the S phase (Figure S2D).

Figure 2. RPA1 associates with HNRNPA2B1SUMO through its SIM motif.

(A) U2OS cells were subjected to immunoprecipitation followed by immunoblots. (B) U2OS cells were analyzed by PLA with indicated antibodies. Scale bars, 10 um. (C) HEK293T cells transfected with HNRNPA2^WT or HNRNPA2^K108R were harvested for immunoprecipitation followed by immunoblots. (D-E) Schematic representation of the full-length (FL) RPA1 and the mutants used in this study (D). HEK293T cells transfected with the indicated constructs were prepared for the immunoprecipitation with anti-Myc antibody (E). (F) Schematic representation of the human RPA1 domains and sequence conservation of the SIM motif. (G) HNRNPA2^K108 recognizes SIM (red) motif of RPA1 (green). The structural model of HNRNPA2^SUMO-RPA1 complex was generated using AlphaFold2 online server by modeling a ternary complex of HNRNPA2-SUMO1-RPA1. (H) HEK293T cells transfected with the indicated constructs were collected for the immunoprecipitation. HNRNPA2B1^SUMO was examined with anti-HNRNPA2B1 and anti-SUMO1 antibodies. See also Figure S2.

Given that there have been many studies reporting the involvement of SUMOylation in cellular activities, we set out to determine whether the interaction with RPA is SUMO-dependent. While HNRNPA2^WT interacted with RPA, HNRNPA2^K108R failed to associate with both RPA1 and RPA2 (Figure 2C), indicating the interactions between HNRNPA2 and RPA were SUMOylation-dependent. Next, a series of truncated or internally deleted mutants of RPA1 were utilized to further characterize the HNRNPA2B1^SUMO-RPA1 interaction. RPA1 contains four oligonucleotide-binding (OB) folds designated F, A, B, and C domains, with A, B, and C domains being directly involved in ssDNA-binding¹⁷. The N-terminal F domain functions as an element of protein–protein interactions and participates in checkpoint activation¹⁷. As shown in Figures 2D and 2E, while full-length (FL) and other mutants of RPA1 could associate with HNRNPA2B1^SUMO, the D1 mutant lacking the F domain failed to bind HNRNPA2B1^SUMO, suggesting that RPA1 associates with HNRNPA2B1^SUMO via its F domain. Since the F domain of RPA1 was previously demonstrated to be a platform for assembling the ATR-ATRIP kinase and its activators^18,38, we hypothesized that the interaction of RPA1 with HNRNPA2B1^SUMO may impair the recruitment of ATR-ATRIP complex, thus inhibiting local ATR activation during unperturbed replication.

As SUMO can create an interface for the recruitment of protein cofactors harboring short peptide sequences known as SUMO-interacting motifs (SIMs)^27,28, we asked if RPA contains any amino acid motifs indicative of an interaction with the SUMO conjugation pathway. As expected, analysis of the RPA1 protein sequence using JASSA³⁹ identified a putative SIM motif (amino acids 93–96) within its N-terminal F domain that could regulate interactions with SUMO-modified proteins (Figure S2E). Most SIMs feature a loose consensus sequence consisting of 3~4 aliphatic residues commonly flanked by acidic or phosphorylatable residues²⁷, sequence alignment of RPA1 revealed a highly conserved VVIL consensus SIM motif across different species (Figure 2F). We next generated a structural model of HNRNPA2^SUMO-RPA1 complex through modeling a ternary complex of HNRNPA2-SUMO1-RPA1. By mapping the position of the SIM on the structure of the N-terminal F domain of RPA1, we found that the SIM of RPA1 is near the K108 residue of HNRNPA2^SUMO (Figure 2G), making it a possible module to connect HNRNPA2^SUMO. To prove that RPA1 requires this SIM motif to interact HNRNPA2^SUMO, we generated a SIM-deleted mutant (ΔSIM, RPA1^ΔSIM) as well as a SIM-inactivated mutant (VVIL-AAAA, RPA1^4A) of RPA1. Both the SIM-deleted mutant and the SIM-inactivated mutant showed an almost complete loss of association with HNRNPA2B1^SUMO as determined by Co-IP assays (Figure 2H). Consistently, compared with the wild-type RPA1, the SIM mutants significantly abolished the HNRNPA2B1/RPA1 PLA signals in cells (Figure S2F). These results indicate that RPA1 physically interacts with HNRNPA2B1^SUMO through its SIM motif. Nevertheless, the assembly of the RPA complex and the associations between RPA with other HR repair factors including MRE11 and RAD51, were not affected by SIM mutants (Figures S2G and S2H), indicating the high substrate specificity for RPA1^SIM.

To further assess the ability of RPA1^SIM to recognize SUMO, we performed in vitro pull-down assays. Bacterially expressed recombinant RPA1^WT or RPA1^4A were incubated with SUMO1 protein in the presence of increasing NaCl concentration and the bound SUMO1 was eluted for subsequent analysis. We observed that RPA1^WT and SUMO1 showed consistently strong binding even under high NaCl concentration conditions (Figure S2I). However, the RPA1^4A mutant failed to bind SUMO1 in vitro (Figure S2I). Interestingly, the RPA1 SIM mutant was defective for its interaction with ATRIP (Figure S2J), a SUMOylated partner induced by ultraviolet radiation⁴⁰, indicating the requirement of SIM for the binding of RPA1 with its SUMOylated substrate. Moreover, HNRNPA2 could be recognized by recombinant RPA1 as revealed by His pull-down assays (Figure S2K). Although the ability of HNRNPA2^K108R to recognize the recombinant RPA1 was severely compromised compared to the HNRNPA2 counterpart, a signal of HNRNPA2 was detected after long exposure (Figure S2K), implying that other sequence of HNRNPA2 besides SUMOylation site also contributes to the association between HNRNPA2 and RPA1. In sum, these results prompt us to conclude that RPA1 recognizes SUMOylated HNRNPA2B1 through its SIM motif and SUMOylation of HNRNPA2B1 plays a significant part in its association with RPA1. However, an uncharacterized module of HNRNPA2B1 might also be required to impart specificity for SUMOylated HNRNPA2B1.

HNRNPA2B1^SUMO prevents RPA accumulation from replication forks during unperturbed DNA replication

Since RPA plays essential roles during normal DNA replication, we wondered whether HNRNPA2B1^SUMO could affect RPA loading onto ssDNA during unperturbed replication. To this end, we performed isolation of proteins on nascent DNA (iPOND) assays to analyze the impact of HNRNPA2B1 loss by CRISPR/Cas9-based targeted knockout (KO) on the RPA bound to the nascent replication forks (Figure S3A). The result revealed that compared to WT counterparts, depletion of HNRNPA2B1 in U2OS cells markedly increased RPA accumulation at replication forks (Figure 3A). Notably, previous iPOND-MS studies identified the enrichment of HNRNPA2B1 on nascent DNA^41,42. Indeed, we observed that HNRNPA2B1^SUMO, but not the unmodified HNRNPA2B1, could be recruited to nascent DNA (Figure 3A), possibly via its interaction with RPA. We next performed DNA fiber assays to monitor the progression of individual replication forks in HNRNPA2B1 KO U2OS cells. Newly synthesized DNA was labeled with iododeoxyuridine (ldU) followed by a second labelling with chlorodeoxyuridine (CIdU) and the length of individual tracts was quantified. WT and HNRNPA2B1 KO cells revealed no significant difference in fork progression speed (Figure 3B) or in the percentage of replication origins firing (Figure S3B), suggesting that loss of HNRNPA2B1 did not affect unperturbed DNA replication.

Figure 3. HNRNPA2B1^SUMO prevents RPA accumulation from replication forks during unperturbed DNA replication.

(A) Replication fork bound proteins were isolated by the iPOND assay with HNRNPA2B1 WT and KO U2OS cells. (B) Schematic for measuring replication fork progression by DNA fiber analysis. Representative images were shown and 150 events were counted for each condition. (C) Schematic of the PLA utilized to detect the association of proteins with nascent DNA as described in the main text. (D-E) Representative images of PLA foci obtained in the WT or HNRNPA2B1 KO U2OS cells according to the protocol depicted in (C). Quantification of the number of PLA foci per cell was shown in (E). (F) HNRNPA2B1 KO U2OS cells were reconstituted with FLAG-tagged HNRNPA2 or HNRNPA2^K108R. iPOND assay was performed as described in (A). (G-H) HNRNPA2B1 KO U2OS cells were reconstituted with FLAG-tagged HNRNPA2 or HNRNPA2^K108R. PLA assay was performed as described in (D). Quantification of the number of PLA foci per cell was shown in (H). Values represent mean and s.e.m. of three independent experiments. Scale bars, 10 μm. Statistical analysis was performed using two-tailed unpaired t-tests. See also Figure S3.

To further demonstrate the inhibitory effect of HNRNPA2B1 on accumulation of RPA at forks, we applied a PLA-based approach that measures the association of proteins at nascent ssDNA located at replication forks. In this assay, U2OS cells were first labeled with EdU for 15 min. Biotin was then conjugated to EdU-labeled DNA by using click chemistry, and PLA analysis was performed to detect proteins that are bound to biotin-labeled nascent DNA (Figure 3C). Using this approach, we observed robust HNRNPA2B1/Biotin PLA foci (Figure S3C) and these PLA foci were clearly compromised by PIAS2 knockdown (Figures S3C and S3D). This suggests that HNRNPA2B1 associates with nascent DNA in a SUMOylation-dependent manner and fits with our observations from the iPOND data (Figure 3A). As expected, knockout of HNRNPA2B1 led to a remarkable increase in the number of RPA/Biotin PLA foci (Figures 3D and 3E), indicating that HNRNPA2B1 prevents RPA association with nascent DNA during unperturbed replication. Next, we estimated the percentage of RPA bound by HNRNPA2B1, our data revealed that as much as ~43% of RPA1 in human cells were bound by HNRNPA2B1 (Figure S3E), supporting HNRNPA2B1 as an essential regulator of RPA1 during DNA replication.

To further assess the requirement of HNRNPA2B1 SUMOylation for the association between HNRNPA2B1 and RPA at replication forks, we stably re-expressed HNRNPA2^WT or HNRNPA2^K108R in HNRNPA2B1 KO cells. iPOND analysis showed that HNRNPA2^WT but not the HNRNPA2^K108R mutant protein could be recruited to the nascent replication forks (Figure 3F). Importantly, reconstitution of HNRNPA2B1 KO cells with HNRNPA2^WT but not HNRNPA2^K108R restored the inhibitory effect of HNRNPA2B1 on RPA accumulation at forks as revealed by iPOND assay (Figure 3F) and the number of RPA/Biotin PLA foci (Figures 3G and 3H), without causing aberrations in the cell cycle (Figure S3F). Collectively, these data indicate that HNRNPA2B1^SUMO functions as an endogenous inhibitor of RPA to prevent RPA accumulation at forks during normal replication.

To determine whether the HNRNPA2B1-mediated RPA is connected to the ATR-based control of RPA at replication forks, we knocked down ATR in U2OS cells or treated cells with ATR inhibitors (ATRi) and analyzed the SUMOylation of HNRNPA2B1. As shown in Figure S3G, the SUMOylation of HNRNPA2B1 was not affected by ATR depletion or inhibition of ATR activity, suggesting that ATR is not required for HNRNPA2B1 SUMOylation. The associations between HNRNPA2B1 and RPA were not significantly altered by ATRi treatment (Figure S3H). Furthermore, depletion of HNRNPA2B1 consistently elicited a strong increase in RPA accumulation at forks in the presence of ATRi as illustrated by iPOND assay (Figure S3I) and RPA/Biotin PLA foci (Figures S3J and S3K), indicating that the HNRNPA2B1-mediated regulation of RPA at forks is not dependent on ATR pathway.

DNA damage dissociates RPA from HNRNPA2B1^SUMO by reducing HNRNPA2B1^SUMO

The intimate association of HNRNPA2B1 with RPA prompted us to hypothesize that HNRNPA2B1 might also participate in the DNA damage response (DDR). To test this, we analyzed the widely used genomic instability signature weighted genomic integrity index (wGII) in patients of multiple cancer types from the cancer genome atlas (TCGA) database. The results revealed that patients with breast cancer (Figure S4A), bladder cancer (Figure S4B), lung cancer (Figure S4C) or ovarian cancer (Figure S4D) with high HNRNPA2B1 expression exhibit higher wGII indicative of increased genomic instability. To further explore the performance of HNRNPA2B1 in the DDR, we employed laser micro-irradiation assay to examine the HNRNPA2B1 cellular localization following DNA damage. As shown in Figures 4A and S4E, HNRNPA2B1 was excluded from sites of DNA breaks, highlighting a potential functional role for HNRNPA2B1 in the DDR. Since HNRNPA2B1^SUMO was found to be recruited to nascent strand chromatin (Figures 3A and S3C), this raised the possibility that HNRNPA2B1^SUMO is repressed by DNA damage. In line with this hypothesis, cells treated with camptothecin (CPT), hydroxyurea (HU) or ionizing radiation (IR) showed a markedly decreased HNRNPA2B1^SUMO (Figures 4B and S4F). Similarly, HU treatment severely decreased His-SUMO1 conjugated to HNRNPA2 as demonstrated by Ni²-NTA pull down assays (Figure S4G). Extending this observation, treatment with CPT, HU and IR considerably reduced proximity labelling of endogenous HNRNPA2B1 and SUMO1 in cells (Figures 4C and 4D), which was consistent with our biochemical analysis.

Figure 4. DNA damage disrupts HNRNPA2B1^SUMO/RPA association by repressing HNRNPA2B1^SUMO.

(A) U2OS cells were subjected to laser microirradiation followed by immunofluorescence. (B) HEK293T cells transfected with FLAG-SUMO1 were treated with CPT (1 uM, 1h) or HU (10 mM, 2h) followed by immunoprecipitation. (C-D) U2OS cells either untreated or exposed to CPT (1 uM, 1h), HU (10 mM, 2h) or irradiation (10 Gy, 1h) were analyzed by PLA (C). 100 cells quantified in each condition were shown in (D). (E) HEK293T cells transfected with FLAG-HNRNPA2 were treated with HU as indicated. Immunoprecipitation was performed followed by immunoblots. (F-G) U2OS cells either untreated or exposed to CPT, HU or irradiation and then analyzed by PLA (F). 100 cells quantified in each condition were shown in (G). (H) HEK293T cells transfected with FLAG-HNRNPA2 were treated with HU as indicated. Immunoprecipitation was performed followed by immunoblots. (I-J) U2OS cells either untreated or exposed to CPT, HU or irradiation were analyzed by PLA (I). 100 cells quantified in each condition were shown in (J). (K) Schematic representation of FLAG-HNRNPA2^WT and FLAG-SUMO1-HNRNPA2^K108R proteins. (L) U2OS cells transiently transfected with GFP-HNRNPA2 or GFP-SUMO1-HNRNPA2^K108R were subjected to laser microirradiation followed by immunofluorescence. Scale bar, 10 μm (C, F and I); 5 um (A and L). Values represent mean and s.e.m. of three independent experiments. Statistical analysis was performed using two-tailed unpaired t-tests. See also Figure S4.

Given PIAS2 was the SUMO E3 ligase responsible for the HNRNPA2B1 SUMOylation, we next tested whether HNRNPA2B1 interacts with PIAS2 after DNA damage. As shown in Figure 4E, lysates from HEK293T cells transfected with FLAG-HNRNPA2 were used for Co-IP with anti-FLAG antibody. Subsequent immunoblotting analysis showed that HNRNPA2 indeed interacted with PIAS2. Of note, HU treatment dramatically reduced the association between HNRNPA2 and PIAS2, however, the HNRNPA2-UBC9 interaction was not affected by HU treatment (Figure 4E). We also applied PLA using primary antibodies against HNRNPA2B1 and PIAS2. In untreated cells, we obtained discrete proximity labelling of HNRNPA2B1 and PIAS2 (Figure 4F), corroborating the notion that these two proteins are in close proximity. Similarly, the HNRNPA2B1/PIAS2 PLA foci were markedly impaired by treatment with CPT, HU or IR in cells (Figures 4F and 4G). This suggests that DNA-damaging agents disrupt the interaction between HNRNPA2B1 and PIAS2, thus decreasing PIAS2-mediated SUMOylation of HNRNPA2B1.

Our finding that SUMOylation regulates a HNRNPA2B1-RPA interaction prompted us to interrogate this interaction under DNA damage. Interestingly, while HNRNPA2^SUMO is able to bind both RPA1 and RPA2, the interaction between HNRNPA2^SUMO and RPA in cells was clearly abolished by HU (Figure 4H), consistent with the decreased HNRNPA2/B1 SUMOylation. In addition to HU, CPT and IR also diminished HNRNPA2B1/RPA association as revealed by PLA (Figures 4I and 4J). In accordance with the cell cycle-dependent expression of HNRNPA2B1, the PLA foci were more efficiently detected in Cyclin A positive cells (Figures 4C, 4F and 4I). Altogether, these results indicate that DNA damage induced a decline of HNRNPA2B1^SUMO that drives the exclusion of HNRNPA2B1^SUMO from chromatin as well as dissociation of RPA from HNRNPA2B1^SUMO.

To establish a direct link between HNRNPA2B1^SUMO and DNA damage, we generated an N-terminal SUMO1-fusion of HNRNPA2^K108R (termed SUMO1-HNRNPA2) (Figure 4K). Interestingly, laser microirradiation showed that compared to the HNRNPA2, the SUMO1-HNRNPA2 fusion could not be excluded from the chromatin (Figure 4L). We then examined the potential of the SUMO1-HNRNPA2 to rescue the genetically ablated HNRNPA2B1. HNRNPA2B1 KO cells were stably complemented with FLAG-tagged HNRNPA2 (termed HNRNPA2) or SUMO1-HNRNPA2 and analyzed for cell viability (Figures S4H–S4J). Clonogenic analysis showed that HNRNPA2B1-deficient U2OS and MDA-MB-231 cells were more resistant to HU treatment, whereas re-expression of HNRNPA2 would reverse this phenomenon (Figures S4K and S4L). Intriguingly, cells restored of SUMO1-HNRNPA2 displayed significantly less survival following exposure to HU (Figures S4K and S4L). These data indicate that the turnover of HNRNPA2^SUMO is essential for proper cell response to DNA damage.

Forced expression of SUMO1-HNRNPA2 limits RPA availability during DNA replication stress

During DNA replication stress, RPA is the first responder in cells coordinating cellular response and checkpoint activation⁴³. Given that RPA is responsible for the majority of ssDNA-binding activity¹⁶, we next explored whether HNRNPA2B1 regulates RPA1 function during replication stress. To examine this, HNRNPA2B1 KO cells complemented with HNRNPA2 or SUMO1-HNRNPA2 were used for Co-IP assays. As shown in Figure 5A, HU treatment abolished the HNRNPA2-RPA interaction in HNRNPA2 rescued cells. Conversely, expression of SUMO1-HNRNPA2 in HNRNPA2B1 KO cells elicited a persistent association of HNRNPA2/RPA with or without HU treatment (Figure 5A). Furthermore, the chromatin accumulation of RPA was significantly reduced in SUMO1-HNRNPA2 reconstituted cells compared to HNRNPA2 expressing cells (Figure 5B). RPA-ssDNA functions as an essential upstream module in orchestrating ATR activation¹⁷, accordingly we observed almost abrogated chromatin accumulation of ATR and ATRIP in HU-treated SUMO1-HNRNPA2 expressing cells (Figure 5B). Intriguingly, we noted an increase in unmodified HNRNPA2 within the nuclear soluble pool of HNRNPA2 expressing cells after HU treatment (Figure 5B), reflecting the dynamics of HNRNPA2 in response to genotoxic stress.

Figure 5. Forced expression of SUMO1-HNRNPA2 impairs ATR activation and the replication stress response by limiting RPA availability.

(A) HNRNPA2B1 KO U2OS cells were complemented with HNRNPA2 or SUMO1-HNRNPA2 and then treated with HU followed by immunoprecipitation. (B) HNRNPA2B1 KO U2OS cells were complemented with HNRNPA2 or SUMO1-HNRNPA2. Chromatin and nuclear soluble fractions were separated after HU treatment for immunoblot analysis. (C) Diagram showing the principle of RPA protection assay as described in the main text. (D) HNRNPA2B1 KO U2OS cells complemented with HNRNPA2, SUMO1-HNRNPA2 together with or without RPA1 were subjected to RPA protection assay. Representative images were shown. (E) HNRNPA2B1 KO U2OS cells were stably complemented with HNRNPA2 or SUMO1-HNRNPA2. Cells were treated with or without HU and harvested for immunoblots. (F) HNRNPA2B1 KO U2OS cells were stably restored with HNRNPA2 or SUMO1-HNRNPA2. Cells were treated with or without HU for immunoprecipitation. (G) HNRNPA2B1 KO U2OS cells were stably complemented with HNRNPA2 or SUMO1-HNRNPA2. GFP-ATRIP transiently transfected cells were treated with or without HU before immunofluorescence analysis. (H-I) Schematic for measuring nascent DNA degradation at stalled replication forks by DNA fiber analysis and representative images of DNA fibers (H). 150 events quantified in each condition were shown in (I). (J-K) Schematic for measuring replication fork restart by DNA fiber analysis and representative images of DNA fibers (J). 150 events quantified in each condition were shown in (K). (L-M) Schematic for measuring fork progression in cells treated with low dose HU by DNA fiber analysis and representative images of DNA fibers (L). 150 events quantified in each condition were shown in (M). Values represent mean and s.e.m. of three independent experiments. Scale bars, 10 μm. Statistical analysis was performed using two-tailed unpaired t-tests. See also Figure S5.

The above findings prompted us to hypothesize that SUMO1-HNRNPA2 would limit the ability of nuclear soluble RPA to localize to chromatin upon DNA replication stress. To this end, we performed RPA protection assays in HNRNPA2B1 KO cells complemented with HNRNPA2 or SUMO1-HNRNPA2 as described previously^44,45. In this assay, replication stress-generated ssDNA fragments, which are not protected by RPA due to the limiting free RPA, would serve as a template for polymerization of 5-bromo-2’-deoxyuridine (BrdU)-labelled DNA by exogenous Taq polymerase (Figure 5C). Upon treatment with HU, anti-BrdU antibody detected few punctate foci in HNRNPA2 expressing cells, but significantly more in SUMO1-HNRNPA2 expressing cells indicating more unprotected tracks of ssDNA (Figure 5D). Notably, co-expression of RPA1 with SUMO1-HNRNPA2 could markedly reduce the ssDNA-derived BrdU foci (Figures 5D, S5A and S5B). Collectively, these results strongly support the idea that forced expression of SUMO1-HNRNPA2 limits RPA availability during DNA replication stress through restricting the access of nuclear soluble RPA to chromatin, thus generating high levels of RPA-unprotected ssDNA, implying that HNRNPA2B1^SUMO functions as a “monitor” of RPA that regulates RPA dynamics during normal replication and stress response.

Forced expression of SUMO1-HNRNPA2 impairs ATR activation and the replication stress response

If HNRNPA2B1^SUMO functions as a “sponge” of RPA, it should be vital for RPA-ssDNA-directed ATR activation. Immunoblotting analysis revealed that HNRNPA2B1 KO markedly enhanced the phosphorylation of CHK1^S345 and RPA2^S33 (Figure S5C), which are well-known substrates of ATR^46,47. Similar results were observed using a phosphorylation-specific antibody against RPA2^S4/S8 for fluorescent immunostaining in HU-treated cells. Depletion of HNRNPA2B1 exacerbated phospho-RPA2^S4/S8 foci formation (Figures S5D and S5E)^48,49. Furthermore, HNRNPA2 or SUMO1-HNRNPA2 was stably re-expressed in HNRNPA2B1 KO cells. We observed that restoring with SUMO1-HNRNPA2 resulted in reduced phosphorylation of CHK1^S345, RPA2^S33 and RPA2^S4/S8 after HU treatment (Figure 5E). Consistently, immunofluorescence staining revealed that phospho-RPA2^S4/S8 foci were efficiently suppressed by SUMO1-HNRNPA2 re-expression when compared with HNRNPA2 (Figures S5F and S5G).

An early event of ATR activation is the recruitment of the ATR-ATRIP complex to damaged sites which depends on RPA1-ATRIP interactions¹⁸. The N-terminus of RPA1 is specifically important for the activation of ATR¹⁷. Our finding that HNRNPA2B1^SUMO associates with the N-terminus of RPA1 led us to ask if lower HNRNPA2B1^SUMO is required for ATR-ATRIP recruitment to sites of DNA damage. Indeed, depletion of HNRNPA2B1 boosted HU-induced focus formation of GFP-ATRIP (Figures S5H and S5I). Co-IP assays were performed with anti-RPA1 antibody followed by immunoblotting using HNRNPA2B1 KO cells restored with HNRNPA2 or SUMO1-HNRNPA2. As shown in Figure 5F, in SUMO1-HNRNPA2 expressing cells the RPA1-ATRIP interaction was significantly disrupted compared to HNRNPA2 expressing cells. In addition, GFP-ATRIP foci were clearly induced by HU in HNRNPA2 expressing cells, but largely suppressed by SUMO1-HNRNPA2 expression (Figures 5G and S5J). Together, these data suggest that during replication stress, HNRNPA2B1 dissociates from RPA1 thus allowing for the assembly of the ATR-ATRIP complex at damage sites. Forced SUMO1-HNRNPA2 expression would restrict the access of ATRIP to RPA-ssDNA and undermine ATR activation.

ATR signaling plays a pivotal role in regulating cellular replication stress response⁵⁰. Thus, we carried out DNA fiber assay to monitor replication fork dynamics in HNRNPA2B1 KO cells complemented with HNRNPA2 or SUMO1-HNRNPA2. The ratio of CIdU to IdU track lengths was measured to reflect DNA synthesis in the presence of stress. Compared to HNRNPA2 expressing cells, stalled forks in SUMO1-HNRNPA2 expressing cells were prone to be shorter with HU (4 mM) treatment (Figures 5H and 5I) and failed to recover after release from replication stress (Figures 5J and 5K). Additionally, in the presence of low HU doses (0.5 mM) forks progressed particularly slowly in SUMO1-HNRNPA2 expressing cells compared to HNRNPA2 expressing cells (Figures 5L and 5M). Importantly, we observed that SUMO1-HNRNPA2 expressing-induced fork defects could be largely compensated by co-expression of RPA1 (Figures 5H–5M). Collectively, we concluded that forced expression of SUMO1-HNRNPA2 affects not only the availability of nuclear soluble RPA to chromatin, but also ATR activation by restricting ATR-ATRIP from the damage sites, suggesting that reduced HNRNPA2B1^SUMO and HNRNPA2B1^SUMO/RPA dissociation are required for efficient replication stress response.

HNRNPA2B1 impedes HR repair via limiting RPA availability

While RPA is a crucial regulator of DNA replication and the replication stress response, RPA is also implicated in error-free repair of DNA double-strand breaks (DSBs) by HR¹⁹. Our finding that HNRNPA2B1 limits RPA availability raises an interesting possibility that HNRNPA2B1 might function in HR repair. To test this possibility, we analyzed HR deficiency (HRD) score and three independent measures of genomic instability acting as the biomarker of tumor HRD: large-scale state transitions (LST), loss of heterozygosity (LOH) and telomeric allelic imbalance (TAI), in breast cancer samples. we observed that high expression of HNRNPA2B1 is associated with an increased HRD score (Figure 6A), LST (Figure 6B), LOH (Figure 6C) and TAI (Figure 6D), indicating that HNRNPA2B1 inhibits HR repair. To further confirm this finding, we employed a well-established DR-GFP HR reporter assay in which expression of I-SceI endonuclease generates a DSB that when repaired by HR restores GFP expression (Figure 6E)⁵¹. In both HEK293T and U2OS cells, loss of HNRNPA2B1 enhanced HR efficiency (Figures 6F and S6A–S6D), directly suggesting that HNRNPA2B1 impedes HR repair. Notably, we did not find significant changes in cell cycle distributions in asynchronous cell populations (Figure S6E), suggesting that the observed effect was not a result of changes in cell cycle regulation.

Figure 6. HNRNPA2B1 impedes HR repair via limiting RPA availability.

(A-D) Analysis of HRD score (A), LST (B), LOH (C) and TAI (D) in breast cancer samples (See Methods section). (E) Schematic for the DR-GFP HR reporter system. (F) HR reporter assay was performed to assess HR activity in HNRNPA2B1 knockdown U2OS cells. (G) The sensitivity of HNRNPA2B1 knockdown U2OS cells to IR was assessed by clonogenic assay. (H) HNRNPA2B1 KO cells stably complemented with HNRNPA2 or SUMO1-HNRNPA2 were subjected to HR reporter assay. HR activity was normalized to WT cells. (I) HNRNPA2B1 KO cells were stably re-expressed with HNRNPA2 or SUMO1-HNRNPA2 to assess the sensitivity to IR by clonogenic assay. (J-K) HNRNPA2B1 KO U2OS cells were stably restored with HNRNPA2 or SUMO1-HNRNPA2 together with or without RPA1. Cells were exposed to IR (5 Gy) followed by neutral comet assay (J). 100 cells quantified in each condition were shown in (K). Values represent mean and s.e.m. of three independent experiments. Statistical analysis was performed using two-tailed unpaired t-tests (Figures 6F, 6G and 6H, 6I). See also Figure S6.

To further dissect the role of HNRNPA2B1 in the DDR, we examined both U2OS and MDA-MB-231 cells depleted of HNRNPA2B1 for sensitivity to IR and cisplatin (Figures S6C and S6F), which can induce DSBs directly or indirectly. Loss of HNRNPA2B1 caused cells resistance to both IR and cisplatin (Figures 6G and S6G–S6I), suggesting that HNRNPA2B1 is functional for cells responding to DSB-inducing agents. We next assessed the kinetics of IR-induced γH2AX foci, a marker of DNA damage, in HNRNPA2B1 depleted or overexpressed cells (Figure S6J). As shown in Figures S6K and S6L, HNRNPA2B1 knockdown cells displayed decreased accumulation of γH2AX foci at late time points (8 h) when compared with control cells. Conversely, ectopically expressed HNRNPA2 remarkably delayed the resolution of γH2AX foci at 16 h post IR (Figures S6M and S6N). Taken together, these results provide strong evidence showing that HNRNPA2B1 hinders DNA repair.

Next, we wondered if the function of HNRNPA2B1 in DNA repair is dependent on its SUMOylation. To this end, we restored HNRNPA2 and SUMO1-HNRNPA2 expression in HNRNPA2B1 KO cells, respectively, and performed HR reporter assays. We observed that re-expression of HNRNPA2 restored the HR capability to that of parental U2OS cells (Figure 6H). However, expression of SUMO1-HNRNPA2 severely compromised HR repair (Figure 6H). Furthermore, clonogenic analysis showed that introduction of HNRNPA2 restored sensitivity to IR in HNRNPA2B1 KO cells (Figure 6I), while expression of SUMO1-HNRNPA2 markedly reduced survival of HNRNPA2B1 KO cells following exposure to various doses of IR (Figure 6I). To provide direct evidence that HNRNPA2B1^SUMO plays a role in DSB repair in individual cells, we performed neutral comet assays in HNRNPA2B1 KO U2OS cells complemented with HNRNPA2 or SUMO1-HNRNPA2 that were exposed to IR. Depletion of HNRNPA2B1 suppressed IR-induced DSBs as measured by tail moments, indicating that HNRNPA2B1 prevents DSB repair (Figures 6J and 6K). Notably, restoration of SUMO1-HNRNPA2 significantly delayed the resolution of DSBs, as evidenced by larger tail moments than HNRNPA2 expressing cells. However, co-expression of RPA1 reduced the tail moment, indicating improved DSB repair (Figures 6J and 6K). Altogether, these results indicate that the HR inhibitory effect of HNRNPA2B1 is dependent on its SUMOylation in limiting RPA availability.

High HNRNPA2B1 expression predicts better patient survival in breast cancer

We next analyzed the cancer genome atlas (TCGA) database to study whether HNRNPA2B1 expression has an impact in the clinical setting. We observed that HNRNPA2B1 expression was upregulated across multiple tumor types (13 of 15), including breast cancer (Figures 7A and S7A). Moreover, based on Kaplan-Meier plot data, high expression of HNRNPA2B1 was associated with better overall survival in breast cancer (Figure 7B), bladder cancer (Figure S7B), lung cancer (Figure S7C) and ovarian cancer (Figure S7D). To further investigate the pathological relevance of HNRNPA2B1, we performed immunohistochemistry (IHC) staining using a human breast cancer tissue microarray containing tumor samples at different clinical stages (Figure S7E). Analysis of 89 breast tumor tissues and 61 adjacent non-tumor tissue samples revealed a pronounced upregulation of HNRNPA2B1 in breast tumors (Figures 7C and 7D). Upregulation of HNRNPA2B1 was observed in 62.9% (56 of 89) of breast tumors, whereas only 8.20% (5 of 61) of adjacent non-tumor tissues exhibited high expression of HNRNPA2B1 (Figure 7C). In addition, we assessed whether the expression of HNRNPA2B1 is associated with clinical outcomes based on our microarray analysis. We found that breast cancer patients with high HNRNPA2B1 expression had better overall survival (Figure 7E) as well as improved disease-free survival (Figure 7F). Additionally, Kaplan-Meier survival analysis showed that patients with lower HNRNPA2B1 expression were at increased risk of breast cancer-related death compared to those with higher HNRNPA2B1 expression (Figure 7G). Taken together, these results strongly suggest that HNRNPA2B1 expression may function as an independent predictor of good prognosis for breast cancer patients.

Figure 7. High HNRNPA2B1 expression predicts better survival in breast cancer patients.

(A) HNRNPA2B1 expression in breast cancer and matched normal tissues was shown based on TCGA data. P values were from Wilcoxon tests. (B) Kaplan Meier plotter reveals the prognostic value of HNRNPA2B1 expression in patients with breast cancer. (C-D) IHC staining of HNRNPA2B1 was evaluated in adjacent tissues and breast tumor tissues. The calculated results based on the staining intensity were shown in (C) and representative images of HNRNPA2B1 IHC staining were shown in (D). Scale bar, 100 uM. (E-F) Kaplan–Meier estimates the overall survival (E) and disease-free survival (F) of breast cancer patients with low or high expression of HNRNPA2B1. Scores of <= 9 were defined as low expression, scores of >= 12 were defined as high expression. The log-rank test was used to compare the survival curves between groups. (G) The survival and death rates for breast cancer patients with low and high HNRNPA2B1 expression were shown. (H-I) Sensitivity of control and HNRNPA2 overexpressed MDA-MB-231 (H) and HCC1806 (I) cells to Olaparib was assessed by clonogenic assay. (J) Working model depicting the role of HNRNPA2B1^SUMO during unperturbed DNA replication and stress response (See discussion). Values represent mean and s.e.m. of three independent experiments and statistical analysis was performed using two-tailed unpaired t-tests (Figures 7H and 7I). See also Figure S7.

Extensive studies have reported that DDR pathway alterations affect cancer chemotherapeutic response⁵². Defective HR has been demonstrated as an important therapeutic target as exemplified by the advent of poly(ADP-ribose) polymerase (PARP) inhibitors displaying synthetic lethality when utilized in HR-deficient cells^53,54. To explore whether HNRNPA2B1 impacts breast cancer response to chemotherapy, we overexpressed HNRNPA2 in two breast cancer cell lines MDA-MB-231 and HCC1806, which showed a relatively low expression of HNRNPA2B1 (Figures S7F–S7H). Clonogenic analysis revealed that overexpression of HNRNPA2 markedly sensitized cells to the PARP inhibitor Olaparib (Figures 7H and 7I), suggesting that high expression of the HNRNPA2B1 renders breast cancer cells vulnerable to PARP inhibitor. Therefore, we infer that HNRNPA2B1 may serve as a prospective biomarker for response prediction to PARP inhibitor in breast cancer.

Discussion

In this study, we uncovered that HNRNPA2B1^SUMO functions as an endogenous inhibitor of RPA during unperturbed replication (Figure 7J). We found that RPA1 contains a SIM motif located in its N-terminal F domain (Figure 2F). Interestingly, this SIM motif recognizes HNRNPA2B1^SUMO, which is SUMOylated at K108 by PIAS2 (Figure 1). During unperturbed DNA replication, HNRNPA2B1^SUMO and RPA1 form a stable nuclear soluble complex in the SUMO-dependent manner, preventing RPA accumulation at forks (Figures 2 and 3). Previous studies documented that the N-terminus of RPA1 binds to multiple checkpoint proteins, including ATRIP¹⁷. Our results support the view that HNRNPA2B1^SUMO/RPA interaction at chromatin would restrict the access of ATR-ATRIP complex to the RPA-ssDNA platform, such that the HNRNPA2B1^SUMO/RPA association impedes local ATR activation at active replication forks (Figure 7J). Following DNA damage, RPA dissociates from HNRNPA2B1^SUMO through reducing the levels of HNRNPA2B1^SUMO, thus enabling the release of RPA to bind ssDNA (Figures 4, 5 and 7J). Loss of HNRNPA2B1 or uncoupling of HNRNPA2B1^SUMO could license the recruitment of ATR activating complex to the RPA-ssDNA and allow for the ensuing ATR activation (Figure 7J). Consistent with this idea, enforced expression of a SUMO1-HNRNPA2 fusion could hinder ATR activation as well as the replication stress response (Figure 5). These data place HNRNPA2B1^SUMO at the core of ATR activation.

While cells reserve ~8 fold higher of RPA than the amount needed to support normal replication, unscheduled firing of dormant origins can exhaust the RPA pool and result in replication catastrophe²⁰, suggesting an RPA threshold exists. Motivated by the phenomena that enforced expression of SUMO1-HNRNPA2 fusion leads to exposed ssDNA and compromises the replication stress response (Figure 5), we propose that cells sustain RPA availability during replication stress via reducing HNRNPA2B1^SUMO. A recent paper describes that the basal activity level of ATR monitors and regulates the amounts of RPA at replication forks during unperturbed replication⁵⁵. Here, our findings highlight HNRNPA2B1^SUMO as an endogenous RPA inhibitor via recognizing the highly conserved SIM motif of RPA1 (Figures 2 and 3). Thus, the presence of the SIM within RPA1 may provide a different modality to deal with genotoxic stress by coupling to SUMO modified components.

Genomic instability is a well-documented hallmark of most cancers⁵⁶. In this study, we observed that HNRNPA2B1 was upregulated in multiple tumor types (Figure S7A), and high expression of HNRNPA2B1 was linked to increased genomic instability (Figures S4A–S4D), suggesting a potential functional role of HNRNPA2B1 in tumor development. Based on the findings described here, we reasoned that the HR repair inhibition characteristic of HNRNPA2B1 elicits inefficient DNA damage repair, which contributes to genomic instability and predisposes to the progression of tumors. Genomic instability has previously been linked to unfavorable clinical outcomes⁵⁷, however, our analysis revealed that higher expression of HNRNPA2B1 was associated with better cancer patient survival (Figures 7B, 7E, 7F and S7B–S7D). This prompts us to hypothesize that tumors with a higher level of HNRNPA2B1 might be more sensitive to DNA-damaging agents considering its inhibitory role in the HR and replication stress. Indeed, forced expression of HNRNPA2B1 renders breast cancer cells sensitive to PARP inhibition (Figures 7H and 7I). More importantly, tumors highly expressing HNRNPA2B1 show HR repair deficiency and display potential vulnerability to the synthetic lethality therapy. Therefore, our work identifies HNRNPA2B1 as a prospective response biomarker for PARP inhibitor treatment, which may lead to more effective targeting therapeutic approaches for specific patient groups.

Limitations of the study

Although our data demonstrated that HNRNPA2B1 exists as a SUMOylated protein, our study lacks a specific HNRNPA2B1^SUMO antibody to unambiguously monitor the behaviors of HNRNPA2B1^SUMO in both the chromatin and nuclear soluble pools. This issue also limits us from further investigating the physiological role of HNRNPA2B1^SUMO in vivo. Another limitation of our study is the lack of direct evidence regarding how the association between HNRNPA2B1^SUMO and RPA counteracts the local activation of the ATR pathway during unperturbed replication. The use of SUMO fusion protein to mimic protein SUMOylation might not completely represent physiological SUMOylation. Finally, while we generated a model of HNRNPA2^SUMO-RPA1 complex, a more detailed structural study of the complex is still needed to understand its working mode during unperturbed replication and genotoxic stress response.

STAR★Methods

Resource Availability

Lead contact

Further information and requests for resources and reagents should be directed to Zhenkun Lou (Lou.Zhenkun@mayo.edu).

Materials availability

All unique reagents generated in this study are available from the lead contact with a completed Materials Transfer Agreement.

Data and code availability

• Original western blot images and microscopy data reported in this paper have been deposited at Mendeley. The DOI is listed in the Key Resources Table.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon reasonable request.

• This paper does not report original code.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse Anti-HNRNPA2B1	Santa Cruz	Cat# sc-374053; RRID: AB_10947257
Rabbit Anti-HNRNPA2B1	GeneTex	Cat# GTX127928; RRID: AB_2616069
Mouse Anti-phospho-Histone H2A.X (Ser139)	Millipore	Cat# 05-636; RRID: AB_309864
Mouse Monoclonal Anti-FLAG® M2	Sigma-Aldrich	Cat# F1804; RRID: AB_262044
Rabbit Anti-RPA1	Bethyl	Cat# A300-241A; RRID: AB_2180681
Mouse Anti-RPA2	Santa Cruz	Cat# sc-56770; RRID: AB_785534
Rabbit Anti-ATR	Cell Signaling Technology	Cat# 2790; RRID: AB_2227860
Rabbit Anti-ATRIP	Cell Signaling Technology	Cat# 2737; RRID: AB_823659
Mouse Anti-Cyclin A	Santa Cruz	Cat# sc-239; RRID: AB_627334
Mouse Anti-HA	Sigma-Aldrich	Cat# H9658; RRID: AB_260092
Rabbit Anti-His-Tag (D3I1O)	Cell Signaling Technology	Cat# 12698; RRID: AB_2744546
Rabbit Anti-PIAS1	Cell Signaling Technology	Cat# 3550; RRID: AB_1904090
Rabbit Anti-PIAS2	GeneTex	Cat# GTX115180; RRID: AB_11162657
Rabbit Anti-PIAS3	Cell Signaling Technology	Cat# 9042; RRID: AB_2797692
Rabbit Anti-PIAS4	Cell Signaling Technology	Cat# 4392; RRID: AB_10547884
Rabbit Anti-UBC9	Cell Signaling Technology	Cat# 4786; RRID: AB_10559206
Mouse Anti-CHK1	Cell Signaling Technology	Cat# 2360; RRID: AB_2080320
Rabbit Anti-Phospho-CHK1 (Ser345)	Cell Signaling Technology	Cat# 2341; RRID: AB_330023
Rabbit Anti-Phospho-RPA32 (Ser4, Ser8)	Bethyl	Cat# A300-245A; RRID: AB_210547
Rabbit Anti-Phospho-RPA32 (S33)	Bethyl	Cat# A300-246A; RRID: AB_2180847
Mouse Anti-BrdU	BD Biosciences	Cat# 347580; RRID: AB_10015219
Rat Anti-BrdU	Abcam	Cat# ab6326; RRID: AB_305426
Rabbit Anti-RAD51	GeneTex	Cat# GTX100469; RRID: AB_1951602
Mouse Anti-SUMO1	Thermo Fisher Scientific	Cat# 33-2400; RRID: AB_2533109
Mouse Anti-Biotin	Jackson ImmunoResearch Labs	Cat# 200-002-211; >RRID: AB_2339006
Rabbit Anti-Phospho-Histone H3 (Ser10)	Cell Signaling Technology	Cat# 9701; RRID: AB_331535
Mouse Anti- PCNA (Alexa Fluor 488 Conjugate)	Cell Signaling Technology	Cat# 8580; RRID: AB_11178664
Alexa Fluor 488-AffiniPure Goat Anti-Mouse IgG (H+L)	Jackson ImmunoResearch Labs	Cat# 115-545-062; RRID: AB_2338845
Alexa Fluor 488-AffiniPure Goat Anti-Rabbit IgG (H+L)	Jackson ImmunoResearch Labs	Cat# 111-545-045; RRID: AB_2338049
Rhodamine Red-X-AffiniPure Goat Anti-Mouse IgG (H+L)	Jackson ImmunoResearch Labs	Cat# 115-295-146; RRID: AB_2338766
Rhodamine Red-X-AffiniPure Donkey Anti-Rat IgG (H+L)	Jackson ImmunoResearch Labs	Cat# 712-295-153; RRID: AB_2340676
Bacterial and virus strains
BL21 (DE3) Escherichia coli	Transgen Biotech	Cat# CD601-02
Chemicals, peptides, and recombinant proteins
Cisplatin	Sigma-Aldrich	Cat# P4394
Olaparib (AZD-2281)	LC laboratories	Cat# O-9201
ML-792	MedChemExpress	Cat# HY-108702
Isopropyl b-D-1-thiogalactopyranoside (IPTG)	Thermo Fisher Scientific	Cat# 15529019
5-Iodo-2′-deoxyuridine	Sigma-Aldrich	Cat# I7125
5-Chloro-2′-deoxyuridine	Sigma-Aldrich	Cat# C6891
N-Ethylmaleimide	Sigma-Aldrich	Cat# E3876
BrdU	Biolegend	Cat# 423401
5-ethynyl-2’-deoxyuridine	Cayman Chemical	Cat# 20518
TransIT-X2® Dynamic Delivery System	Mirus	Cat# MIR6004
DAPI	Thermo Fisher Scientific	Cat# D1306
Dynabeads™ Protein G	Thermo Fisher Scientific	Cat# 10003D
Anti-Flag M2 Affinity Gel	Sigma-Aldrich	Cat# A2220
PrimeSTAR HS DNA Polymerase	TaKaRa	Cat# R040A
Hydroxyurea	Sigma-Aldrich	Cat# H8627
Biotin-azide	Cayman Chemical	Cat# 13040
Berzosertib (VX970)	Selleck	Cat# S7102
(+)-Sodium L-ascorbate	Sigma-Aldrich	Cat# A7631
SUMO1	Boston Biochem	Cat# UL-712
PIAS2	NOVUS Biologicals	Cat# H00009063-P01
Pierce™ Glutathione Agarose	Thermo Fisher Scientific	Cat# 16102BID
Ni-NTA Agarose	QIAGEN	Cat# 30210
SAE1	Boston Biochem	Cat# E-315
UBC9	Boston Biochem	Cat# E2-645
Critical commercial assays
Duo-link in situ PLA Kit	Sigma-Aldrich	Cat# DUO92101
QuickChange Site-Directed Mutagenesis Kit	Stratagene	Cat# 200519
CometAssay kit	Trevigen	Cat# 4250-050-K
Deposited data
Raw data files	This paper	Mendeley Data: https://doi.org/10.17632/vzgpmfmt9t.1
Experimental models: Cell lines
Human: HEK293T	ATCC	Cat# CRL-11268
Human: U2OS	ATCC	Cat# HTB-96
Human: MDA-MB-231	ATCC	Cat# HTB-26
Human: HCC1806	ATCC	Cat# CRL-2335™
Oligonucleotides
shHNRNPA2B1-1	Sigma-Aldrich	TRCN0000001058
shHNRNPA2B1-2	Sigma-Aldrich	TRCN0000001059
shATR	Sigma-Aldrich	TRCN0000196538
shPIAS1-1	Sigma-Aldrich	TRCN0000231898
shPIAS1-2	Sigma-Aldrich	TRCN0000231900
shPIAS2	Sigma-Aldrich	TRCN0000230128
shPIAS3	Sigma-Aldrich	TRCN0000274337
HNRNPA2B1 KO	GenScript	TCTCTTGCTACAGCACGTTT
HNRNPA2-forward	This paper	ATGGAGAGAGAAAAGGAACAG
HNRNPA2-reverse	This paper	GTATCGGCTCCTCCCACC
RPA1-forward	This paper	ATGGTCGGCCAACTGAGC
RPA1-reverse	This paper	CATCAATGCACTTCTCC
SUMO1-forward	This paper	ATGTCTGACCAGGAGGC
SUMO1-reverse	This paper	AACTGTTGAATGACC
shPIAS4-1	Sigma-Aldrich	TRCN0000231892
shPIAS4-2	Sigma-Aldrich	TRCN0000004115
Recombinant DNA
PLVX3-FLAG-HNRNPA2	This paper	N/A
PLVX6-EGFP-HNRNPA2	This paper	N/A
pGEX-4T2-HNRNPA2	This paper	N/A
PLVX3-FLAG-RPA1	This paper	N/A
PLVX3-FLAG-RPA1 (ΔSIM)	This paper	N/A
Myc-RPA1	This paper	N/A
pTEV-His-RPA1 (4A)	This paper	N/A
PLVX3-FLAG-RPA1 (4A)	This paper	N/A
PLVX3-FLAG-SUMO1	This paper	N/A
PLVX3-FLAG-SUMO1-HNRNPA2	This paper	N/A
PLVX3-FLAG-HNRNPA2 (K108R)	This paper	N/A
pFLAG-CMV-SENP1	Addgene	Cat# 17357
pCMV-HA-hUBC9	Addgene	Cat# 14438
Software and algorithms
GraphPad Prism	GraphPad	https://www.graphpad.com
ImageJ	ImageJ	https://imagej.nih.gov/ij/
FlowJo (10.1)	FlowJo LLC	http://www.flowjo.com/

Experimental Model and Subject Details

Cell culture conditions

MDA-MB-231, HCC1806, U2OS and HEK293T cell lines were purchased from ATCC. MDA-MB-231 and HEK293T were cultured in Dulbecco’s Modified Eagle Medium (DMEM) with 10% fetal bovine serum (FBS). U2OS and HCC1806 cells were cultured in McCoy’s 5 A and RPMI1640 with 10% FBS, respectively. The ER-mCherry-LacI-FokI-DD U2OS cell line was a gift from R. A. Greenberg (University of Pennsylvania). All cell culture reagents were obtained from Life technologies. All cell lines have been identified by the medical genome facility center of Mayo Clinic (Rochester, MN) and were routinely tested for mycoplasma contamination. All cells were incubated in a humidified Thermo Fisher incubator with 5% CO2.

Method details

Plasmid constructions

Full length HNRNPA2 was cloned from HEK293T cDNA and inserted into the pLVX3-3XFLAG, pLVX6-EGFP and pGEX-4T-2 vectors and confirmed by sequencing. Myc-tagged RPA1 truncation mutants were gifts from Jingsong Yuan (Columbia University). GFP-ATRIP plasmid was a gift from Shan Zha (Columbia University). His-RPA1 plasmid was a gift from Donna D. Zhang (University of Arizona). SUMO1 and SUMO2/3 plasmids were stored in our lab. HA-UBC9 (14438) and FLAG-SENP1 (17357) were got from Addgene. The RPA1^4A, RPA1 ^ΔSIM mutants and HNRNPA2^K108R were generated by site-directed mutagenesis and subcloned into pLVX3-3XFLAG or pGEX-4T2 vector as indicated. SUMO1 was subcloned into pLVX3-3XFLAG-HNRNPA2^K108R to get the SUMO1-HNRNPA2 fusion construct.

RNA interference

HNRNPA2B1, PIAS1, PIAS2, PIAS3, PIAS4 and ATR-specific shRNA constructs were commercially got from Sigma. Lentivirus for shHNRNPA2B1 as well as the control (shScr) were packaged with pMD2.G and pSPAX2. Briefly, 3 μg pMD2.G, 9 μg pSPAX2 and 12 μg lentivirus plasmids were co-transfected into HEK293T cells in 10 cm cell culture dish using ProFection Mammalian Transfection System (Promega). The lentivirus particles were harvested at 48h after transfection and concentrated. For infection, the concentrated lentivirus particles were added into the medium with 8 ug/ml polybrene and cultured at 37°C for 24–48 hours before they were selected with puromycin (8 μg/mL). The knockdown cells were confirmed by Western blot and used for further experiments.

HNRNPA2B1 knockout (KO) cells were generated using Lenti-CRISPR V2 (GenScript) that contains a guide (g)RNA targeting human HNRNPA2B1 (TCTCTTGCTACAGCACGTTT) by CRISPR/Cas9 approach. The lentiviruses were harvested at 48h post transfection and were utilized to infect indicated cells. The infected cells were then selected with puromycin (8 μg/mL). The KO cells were confirmed by Western blot and used for further experiments. For the HNRNPA2, HNRNPA2^K108R and SUMO1-HNRNPA2 reconstituted cells, following 48h post lentivirus transfection, cells were then screened and confirmed by Western blot. The indicated HNRNPA2B1 KO cells were maintained under the same conditions as their wild-type control cells.

Immunofluorescence

Briefly, cells were first fixed with 4% paraformaldehyde (PFA) then permeabilized with 0.25% Triton X-100. Then cells were blocked by 2% BSA followed by incubation with primary antibody. Fluorescent secondary antibody and DAPI were then incubated with cells to stain the targeted proteins and nucleus, respectively. The cover slips were mounted onto glass slides with Prolong Diamond Antifade Mountant (P36961, Thermo). Cells were visualized using ImageXpress Micro Confocal (Molecular Devices) and at least 100 cells were counted in each condition. To stain RPA2, cells were fixed with cold methanol /acetone (1/1) at −20 °C for 15 min, then washed twice using PBS and fixed again with 4% PFA at room temperature for 15 min before incubation with primary antibody.

In situ proximity ligation assay (PLA)

U2OS cells were fixed with 4% PFA for 30 min at room temperature followed by permeabilization with 0.25% Triton X-100 for 20 min. Then PLA experiments were performed using Duo-link in situ PLA kit (DUO92008, Sigma) according to the manufacturer’s protocol. Briefly, samples were first blocked with blocking solution at 37 °C for 1h and then incubated with the mixture of primary antibodies at 4 °C overnight. Then, the probes were incubated at 37 °C for 1h, followed by hybridization, ligation and amplification. Cellular nuclei were stained with DAPI. The cover slips were mounted onto glass slides with Prolong Diamond Antifade Mountant and visualized using ImageXpress Micro Confocal (Molecular Devices) and at least 100 cells were counted in each condition.

For EdU incorporation, cells were first labeled with EdU (10 uM, 15 min) followed by fixation and permeabilization. Then, click reactions (combine the following reagents in the listed order: 0.1 X PBS, 2 mM copper (II) sulfate, 5 uM biotin azide, 10 mM sodium ascorbate) were performed at room temperature for 15 min. Cells were then washed with PBS before PLA assay. Anti-Biotin antibody was used to detect EdU signal.

Laser microirradiation

U2OS or MDA-MA-231 cells were cultured on glass-bottom dishes (Wuxi NEST Biotechnology). Laser microirradiation was performed using a Micropoint Laser Illumination and Ablation system (Photonic Instruments). A high-energy ultraviolet laser (170 mJ at 10 Hz) generated DNA breaks in nuclei. Cells were then fixed with 4% PFA for primary antibody staining. Images were taken using the IX71 microscope with cellSens software (Olympus) or the Nikon Eclipse 80i fluorescence microscope.

wGII signature analysis

Segmented copy number data and HNRNPA2B1 RNA expression data as log-transformed transcripts per million (TPM) in TCGA breast cancer cohort (n = 1092 samples) were downloaded from the UCSC xenabrowser⁵⁸. The chromatin instability signature wGII for all samples was recalculated as described in Andrea et al⁵⁹. The samples were subgrouped based on the median expression level of HNRNPA2B1. Statistical significance was tested using two-sided Mann–Whitney–Wilcoxon tests.

Western blot and Co-immunoprecipitation (Co-IP)

Cells were harvested and lysed with RIPA lysis buffer (CST). After centrifugation at 12000 rpm for 10 min (4°C), the supernatants were collected for immunoblot with indicated antibodies.

For Co-IP assays, HEK293T or U2OS cells transfected with indicated constructs were collected, washed once in PBS and lysed in Cell Lysis Buffer (CST) or RIPA lysis buffer (CST) with freshly added protease inhibitor cocktail (Roche) for 10 min at 4°C. The lysates were briefly sonicated and clarified by centrifugation at 12000 rpm for 10 min (4°C). The supernatants were incubated with Dynabeads protein G beads (Life technologies) and indicated antibody at 4°C for 4 h with rotation. Beads were washed 3 times in lysis buffer and bound proteins were eluted by heating at 95°C for 10 min in 1× LDS sample buffer (Life technologies). The eluates were analyzed by immunoblotting.

RPA-protection assay

RPA-protection assays were performed as described previously^44,45. Briefly, the indicated U2OS cells cultured on glass coverslips were treated with HU to induce replication stress. The cells were first permeabilized with 0.25% Triton X-100 in PBS for 20 min on ice then fixed in 4% PFA at room temperature for 15 min. Then, the cells were inverted onto Parafilm placed in a sealed chamber and incubated with Taq DNA polymerase (M0273S, NEB), dNTP (0.2 mM) and BrdU (0.4 mM) at 72°C for 5 min. After that, the slides were incubated with a specific antibody against BrdU (347580, BD) at 4°C overnight, followed by incubation with a secondary antibody for 1h at room temperature before immunofluorescence microscopy.

DNA fiber assay

DNA fiber assays were performed as described by Quinet et al⁶⁰ with some changes. In brief, the indicated U2OS cells were pulse labeled with 25 μM IdU for 30 min, washed with PBS (37 °C), refed with media (37 °C) containing 250 μM CldU and cultured for 30 min. Cells were then washed in PBS and diluted with cold PBS. Cells (~100,000) were lysed in 15 μl fiber lysis buffer (200 mM Tris-HCl, pH 7.4, 50 mM EDTA, and 0.5% SDS) on a microscope slide, and the slide was tilted to a 15° angle to allow fibers to spread along the length of the slide. Fibers were fixed in 3:1 methanol/acetic acid for 10 min and then air-dried. The slides were washed twice with PBS and denatured in 2.5 M hydrochloric acid overnight at 4 °C. Slides were washed twice with PBS, blocked in 2% BSA in PBS for 30 min, and stained with primary antibodies overnight at 4 °C: Mouse anti-BrdU (1/25, BD Biosciences, 347580), Rat anti-BrdU (1/1000, Abcam, ab6326) diluted in 5% BSA in PBS). The next morning, slides were washed twice with PBS, stained with secondary antibodies Alexafluor 594 anti-mouse (1/500, Invitrogen, A11032) and Alexafluor 488 anti-rat (1/500, Invitrogen, A110066) for 1h at room temperature and washed three times with PBS. SlowFade Gold 30 (μL) (ThermoFisher, S36936) was added, and coverslips were applied. Images of fibers were acquired at 63× magnification on a Nikon eclipse 80i Fluorescence microscope. For the DNA replication elongation rates, the lengths of the DNA fibers (only CldU tracks connected with IdU tracks) were measured and DNA replication elongation rates were calculated as fiber length divided by pulse labeling time. Fiber lengths were measured using ImageJ, and micrometer values were expressed in kilobases using the following conversion factor: 1 μm = 2.59 kb.

iPOND (isolation of proteins on nascent DNA)

iPOND assays were performed according to protocol with some changes. Briefly, about 1× 10⁸ cells per sample were labeled with 10 μM EdU for 15 min and then washed with PBS for two times. Cells were then fixed with 1% formaldehyde in PBS for 20 min at room temperature followed by quench crosslinking reaction by adding 1.25 M glycine. Cells were then washed and permeabilized with 0.25% Triton X-100 for 1h at 4 °C before incubating with the click reaction buffer (10 uM biotin azide, 2 mM CuSO4 and 10 mM sodium L-ascorbate in PBS) for 2 h at room temperature. After click reaction, cell pellets were resuspended in lysis buffer (1% SDS in 50 mM Tris-HCl, pH 8.0) and subjected to sonication. Cell lysates were then incubated with streptavidin beads and rotated the biotin captures overnight at 4 °C before being washed twice with cold lysis buffer, once with 1 ml of 1 M NaCl and repeated the cold lysis buffer washes two more times. Beads were incubated with 1× LDS loading buffer and heated at 95 °C for 10 min. The eluates were collected for SDS-PAGE analysis and immunoblotted with the indicated antibodies.

Expression and purification of recombinant proteins

HNRNPA2^WT and HNRNPA2^K108R were subcloned into pGEX-4T2 vector and transformed into BL21(DE3) Escherichia coli cells. Protein expression was induced with isopropyl-β-D-thiogalactoside (IPTG, 0.4 mM) at 16 °C for 16h. The bacteria were sonicated in the lysis buffer (20 mM Tris-HCl pH 8.0, 500 mM NaCl, 1 mM EDTA, 0.5 mM PMSF and 1% Triton X-100). Lysates were cleared by centrifugation at 4 °C for 30 min and applied to batch purification columns containing 1 ml bed volume of glutathione agarose (GE) and incubated for 4 h at 4 °C. The flow through was discarded and the beads were washed 5 times with lysis buffer. Proteins were eluted using elution buffer containing 100 mM Tris-HCl pH 8.0, 150 mM NaCl and 30 mM reduced glutathione (GSH). Proteins were dialyzed against buffer containing 50 mM Tris-HCl pH 7.5, 100 mM NaCl and 10% glycerol for subsequent experiments.

For the His-tagged RPA1 protein purification, the bacteria were sonicated in lysis buffer (20 mM Tris-HCl pH 8.0, 600 mM NaCl, 1 mM EDTA, 10 mM imidazole, 0.5 mM PMSF and 1% Triton X-100). Lysates were cleared by centrifugation (14000 rpm) at 4 °C for 45 min and applied to DNase I (5 ug/ml) treatment on ice for 15 min. Then, lysates were incubated with 1 ml Ni-NTA His-affinity resin and rotated for 2h at 4 °C. The flow through was discarded and the beads were washed 5 times with lysis buffer. Proteins were eluted using elution buffer containing 100 mM Tris-HCl pH 8.0, 150 mM NaCl and 250 mM imidazole. Proteins were dialyzed against buffer containing 50 mM Tris-HCl pH 7.5, 100 mM NaCl and 10% glycerol for subsequent experiments. For the His-RPA1 pull-down assay, purified His-RPA1 was incubated with SUMO1 (Boston Biochem) in NETN buffer for 2h at 4 °C, washed and eluted before subsequent analysis.

SUMOylation assay under denature conditions

SUMOylation analysis by Ni²⁺-NTA pull-downs was performed as previously described with some changes⁶¹. Briefly, the indicated cells were harvested and pellets were washed once in PBS. Then cells were lysed in lysis buffer (8 M Urea, 0.1 M NaH₂PO4, 300 mM NaCl and 10 mM Tris pH 8.0). Lysates were briefly sonicated to shear DNA and incubated with Ni²⁺-NTA agarose beads (QIAGEN) for 1h at room temperature. Beads were washed 5 times with lysis buffer. Input and beads were boiled for 10 min in 1 X LDS loading buffer and subjected to immunoblotting.

The SUMOylation analysis by Co-immunoprecipitation (Co-IP) under denaturing conditions was conducted to detect both exogenous and endogenous SUMOylation of HNRNPA2B1. Briefly, cells were lysed with lysis buffer S (20 mM sodium phosphate pH 7.4, 150 mM NaCl, 1% SDS, 1% Triton, 0.5% sodium deoxycholate, 5 mM EDTA, 5 mM EGTA, 20 mM N-ethylmaleimide (NEM), 5 mM DTT and Protease inhibitor cocktail), boiled for 15 min and then sonicated until the lysate became fluid. The lysates were clarified by centrifugation at 13000 rpm for 10 min at 4°C, diluted 1:10 with lysis buffer S without SDS, and then incubated with the indicated antibodies and Protein G beads 4h at 4°C. Beads were washed 5 times with lysis buffer, boiled for 10 min in 1 X LDS loading buffer and subsequently analyzed by Western blotting.

In vitro SUMOylation assay

In vitro SUMOylation assays were performed as previously described with some changes⁶¹. In brief, SUMO modification reactions were typically performed in a total volume of 100 μl with 200 ng of SUMO-activating enzyme (SAE1) (Boston Biochem), 100 ng of UBC9 (Boston Biochem), 1 μg of SUMO1 (Boston Biochem), 100 ng of PIAS2 (NOVUS), and recombinant GST-HNRNPA2 bound to GST-sepharose (20 μg). The 10X reaction buffer (Boston Biochem) was added with 5 mM ATP-Mg. Reactions were incubated at 37°C for 1h and stopped by adding 10 mM EDTA. The beads were washed with NETN buffer (20 mM Tris-HCl, pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.5% Nonidet P-40 with protease inhibitors) for 3 times and boiled in 1X LDS loading buffer. Finally, the samples were analyzed by immunoblotting.

DR-GFP HR reporter assay

To examine the repair of I-SceI-generated DSBs by HR in cells, the indicated HEK293T or U2OS cells were transfected with DR-GFP, pCBA-I-SecI and pmCherry. The HR efficiency was determined at 48h post-transfection. The double positive (GFP and mCherry-positive) cells were quantified by Attune NxT Flow Cytometry (Thermo Fisher). Results were normalized to control group. The gating strategies were shown in Supplementary Figure. S5B.

Colony-formation assay

U2OS, MDA-MB-231 or HCC1806 cells (1000) stably expressing the indicated constructs were plated in triplicates in each well of six-well plates for 24h before treated with indicated DNA-damaging agents for 10–14 days at 37 °C to allow colony formation. Colonies were stained with Giemsa and counted. Results were normalized to plating efficiencies.

Cell cycle analysis

The indicated cells were prepared and fixed in 70% ice-cold ethanol. Then the fixed cells were washed with PBS and resuspended in PI/RNase solution (Thermo Fisher Scientific) at room temperature for 30 min. The samples were analyzed using Attune NxT Flow Cytometer (Thermo Fisher Scientific) and the results were analyzed by FlowJo v.10. For cell synchronization by double thymidine block and release experiments, U2OS cells or the indicated cells were treated with 2 mM thymidine for 18h and released for 9h, followed by the second 18h treatment before being released into thymidine-free medium for the indicated periods of time. Cells were collected at 0, 4, 9h for analysis of cell cycle by DNA staining using PI or analysis of protein by Western blot using indicated antibodies.

Neutral Comet Assay

Neutral comet assays were performed using a Single Cell Gel Electrophoresis Assay Kit (TREVIGEN, 4250-050-K) based on the manufacturer’s protocols. In brief, the indicated cells were prepared for irradiation (5 Gy) and recovered for 4h. Then, 1 × 10⁵/ ml cells were combined with LMAgarose (at 37°C) 1% at a ratio of 1:10 (v/v) and immediately pipetted onto slides. Slides were immersed in the lysis solution at 4°C overnight followed by electrophoresis at 28 V for 1h and stained with SYBR Gold (Invitrogen, 1/10000) for 30 min at room temperature. Images were acquired on a Nikon Eclipse 80i Fluorescence microscope. Comet tail moment were analyzed using OpenComent plugin.

HRD score, LOH, LST and TAI analysis

TCGA breast cancer cohort HRD scores and the three subscores (TAI, LST, LOH) have been published previously and were downloaded from Genomic Data Commons (GDC)^62,63. Expression data (n = 1153 samples) as log-transformed transcripts per million (TPM) were downloaded from the UCSC xenabrowser⁵⁸. Samples were subgrouped by median expression level of HNRNPA2B1. For the results shown in Figure 6, the box outlines show the 25th and 75th percentiles, the solid lines show the median value and the whiskers extending to the most extreme data point that is no more than 1.5 times the interquartile range. Statistical significance was tested using two-sided Mann–Whitney–Wilcoxon test.

Structural modelling

The structural model of HNRNPA2^SUMO-RPA1 complex was generated using AlphaFold2 online server (https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/AlphaFold2.ipynb) by modeling a ternary complex of HNRNPA2-SUMO1-RPA1. A set of 5 models was generated. Interface between HNRNPA2 and RPA1 was converged in all 5 models, while SUMO1 adopted different conformations. The one exhibiting closest distance between Lys108 of HNRNPA2 and Gly83 of SUMO1 was used for further analysis. Figure was prepared using PyMol v.2.3.2.

Quantification and statistical analysis

Values in bar and line graphs were presented as mean ± S.E.M of three independent experiments. Statistical analysis was performed using two-tailed Student’s t-tests unless mentioned otherwise. All statistical analysis was performed using Prism v.8.0 (GraphPad) or Microsoft Excel 2019.

Highlights.

• RPA1 associates with HNRNPA2B1^SUMO through its SIM motif

• HNRNPA2B1^SUMO prevents RPA accumulation at unperturbed replication forks

• HNRNPA2B1^SUMO limits RPA availability – impacting the replication stress response and HR

• Declined HNRNPA2B1^SUMO is required for ATR activation during DNA replication stress