Motifs of the C-terminal domain of MCM9 direct localization to sites of mitomycin-C damage for RAD51 recruitment

The MCM8/9 complex is implicated in aiding fork progression and facilitating homologous recombination (HR) in response to several DNA damage agents. MCM9 itself is an outlier within the MCM family containing a long C-terminal extension (CTE) comprising 42% of the total length, but with no known functional components and high predicted disorder. In this report, we identify and characterize two unique motifs within the primarily unstructured CTE that are required for localization of MCM8/9 to sites of mitomycin C (MMC)-induced DNA damage. First, an unconventional “bipartite-like” nuclear localization (NLS) motif consisting of two positively charged amino acid stretches separated by a long intervening sequence is required for the nuclear import of both MCM8 and MCM9. Second, a variant of the BRC motif (BRCv) similar to that found in other HR helicases is necessary for localization to sites of MMC damage. The MCM9-BRCv directly interacts with and recruits RAD51 downstream to MMC-induced damage to aid in DNA repair. Patient lymphocytes devoid of functional MCM9 and discrete MCM9 knockout cells have a significantly impaired ability to form RAD51 foci after MMC treatment. Therefore, the disordered CTE in MCM9 is functionally important in promoting MCM8/9 activity and in recruiting downstream interactors; thus, requiring full-length MCM9 for proper DNA repair.

The MCM8/9 complex is implicated in aiding fork progression and facilitating homologous recombination (HR) in response to several DNA damage agents. MCM9 itself is an outlier within the MCM family containing a long C-terminal extension (CTE) comprising 42% of the total length, but with no known functional components and high predicted disorder. In this report, we identify and characterize two unique motifs within the primarily unstructured CTE that are required for localization of MCM8/9 to sites of mitomycin C (MMC)induced DNA damage. First, an unconventional "bipartite-like" nuclear localization (NLS) motif consisting of two positively charged amino acid stretches separated by a long intervening sequence is required for the nuclear import of both MCM8 and MCM9. Second, a variant of the BRC motif (BRCv) similar to that found in other HR helicases is necessary for localization to sites of MMC damage. The MCM9-BRCv directly interacts with and recruits RAD51 downstream to MMC-induced damage to aid in DNA repair. Patient lymphocytes devoid of functional MCM9 and discrete MCM9 knockout cells have a significantly impaired ability to form RAD51 foci after MMC treatment. Therefore, the disordered CTE in MCM9 is functionally important in promoting MCM8/9 activity and in recruiting downstream interactors; thus, requiring full-length MCM9 for proper DNA repair.
Homologous recombination (HR) of DNA involves multifaceted processes and pathways that respond to various types of DNA damage agents encountered during S/G2 phases of mitotic cells (1,2). Recombination occurring during meiosis can generate crossovers for genetic diversity and proper segregation in germline cells, utilizing many of the same enzymes (3,4). Therefore, HR is vital for genomic integrity and diversity required for organismal survival. Defects in either mitotic or meiotic HR can directly contribute to increased cancer susceptibility and infertility through improper chromosomal rearrangements that represent incomplete intermediates and are hallmarks of disease. Various DNA helicases contribute to several steps in the recombination pathways either facilitating or dissolving hybrid DNA recombinants (5). Their individualized roles and substrate specificities in HR are commonly overlapping, making absolute distinctions of function difficult.
MCM8 and MCM9 are recent additions to the roster of DNA helicases involved in HR (6). They are members of the ATPases associated with a variety of cellular activities (AAA + ) superfamily and the minichromosome maintenance (MCM) family of proteins that includes MCM2-7 as the heterohexameric helicase complex central to the replication fork. The MCM8/9 complex does not appear to interact directly with MCM2-7, nor is it essential for replication (7). However, MCM8/9 is commonly associated with the replication fork (8) and may be able to take over helicase activities upon depletion of MCM2-7 (9), suggesting a more active and dynamic role in elongation. Mounting evidence suggests that MCM8/9 is itself a heterohexameric complex involved in mediating unknown aspects of fork progression and/or downstream HR (10)(11)(12)(13).
Knockouts or knockdowns of MCM8 and/or MCM9 in mice and humans cause sex-specific tumorigenesis, defects in HR processing, and sensitivities to DNA damaging agents (11)(12)(13)(14)(15). This results in diminished DNA damage signaling as exhibited by decreased phosphorylated CHK1 (pCHK1) and increased double-strand breaks (DSBs) as indicated by H2AX foci in the presence of various fork stalling or cross-linking agents. The absence of functional MCM8 or 9 impairs HR mediated fork rescue after damage through decreased recruitment of Mre11/Rad50/Nbs1 (MRN), RPA, and RAD51. In fact, MCM8/9 has been shown to be required for MRN nuclease activity to generate single-strand DNA (ssDNA) for HR after treatment with cisplatin (Cis-Pt) (10). Even so, there are differing reports on the temporal association of MCM8/9 in relation to RAD51 after treatment with various DNA damage agents (9)(10)(11)(12). These differences may be related to differential activities of RAD51 in HR-mediated fork stability/ restart compared with that of direct DSB repair (16) from specific DNA damage agents utilized or different eukaryotic cell types.
Mutations in MCM8 and MCM9 in humans are linked to premature ovarian failure (POF) (17,18), amenorrhea, sterility (19), and cancer (20). In fact, deficiencies in MCM8 or 9 are phenotypically similar to Fanconi anemia (FA) patient mutations (21) (where 50% of patients are infertile (22)) and suggest an overlapping role in interstrand cross-link (ICL)coupled HR, but without the associated anemia. MCM9 mutations are also linked to hereditary mixed polyposis and colorectal cancer, commonly caused by loss of function in DNA mismatch repair (MMR) (23,24). The MMR connection may be more important in regulating microhomology mediated HR that requires mismatch repair for Holliday junction progression (24)(25)(26) or from overlapping recognition of duplex distorting lesions (27,28). Instead, several human cancer genomes show homo and heterozygous deletions in MCM9 coding regions, many missense mutations in MCM8 and MCM9, and altered expression levels that correlate with aggressive clinical features and poorer long-term survival in several human cancers (10,(29)(30)(31). Like that found for BRCA1/2-deficient cells, MCM8 or 9 deficient cells are hypersensitive to poly(ADP-ribose) polymerase (PARP) inhibitors indicating a link between fork progression and BRAC1/2-mediated HR repair that can be exploited in the clinic for synthetically lethal therapies with platinum-based DNA cross-linking agents (32).
MCM9 contains a unique and large C-terminal extension (CTE) not found in the other MCM family members and can be alternatively spliced to give a shorter MCM9 M isoform that retains the conserved helicase domains but removes the CTE (33). The CTE is a common feature in other HR helicases and is generally considered to be unstructured with scattered putative amino acid motifs that can impact protein interactions and affect proper function (6). However, no such motifs or role for the CTE has been identified for MCM9. Here, we can show that the CTE in MCM9 plays an essential role in nuclear import and formation of DNA repair foci after treatment with the cross-linking agent, mitomycin C (MMC). We have identified and validated a unique "bipartite-like" nuclear localization signal (NLS) within the CTE that directs the nuclear import of MCM8. Finally, we have also identified a BRCv motif that is required for the recruitment of RAD51 to sites of MMC-induced damage that is analogous to that found in other HR helicases (34). The overall results confirm an influential role for the CTE of MCM9 in importing the MCM8/9 complex into the nucleus, directing it to sites of cross-link damage, and recruiting RAD51 for downstream repair.

Results
Specific domains of MCM9 reciprocally affect nuclear localization and foci formation after MMC damage We and others have shown previously that MCM8 and MCM9 are generally localized to the nucleus and form nuclear foci after damage with MMC (11,17). However, we sought to examine which of the domains of MCM9 (Fig. 1A) are required for nuclear foci formation after MMC treatment. MCM9 contains a unique CTE that comprises 42% of the 1143 amino acids and 52 kDa of the total 127 kDa molecular weight of the full-length protein. The CTE is more hydrophilic compared with the rest of the protein, has a higher disorder probability, and is likely mostly unstructured (Fig. S1).
As the CTE of MCM9 is predicted to have high disorder and low secondary structure, we sought to directly measure the solution structure composition of various MCM9 CTE truncations ( Fig. 1B) using circular dichroism (CD). Spectra for both MCM9 643 to 900 and 680 to 900 show a pronounced minimum at 201 nm (Fig. 1C), which is consistent with significant disorder (35). There are shallow valleys from 215 to 230 nm indicative of minor α-helical and antiparallel β-sheet characteristics, but the overall spectrum is representative of a highly disordered (50%) protein, indicated as "other" in the quantification (Fig. 1D). For the limited secondary structure, the two truncations are highly similar with only 7% helical and 30% β-sheet.
Full-length MCM9-GFP (MCM9 L ) is nuclear and forms a significant number of nuclear foci upon treatment with the cross-linking agent MMC (Fig. 2, A-C). Interestingly, the transfection of the alternatively spliced MCM9 product (MCM9 M ) (33) showed primarily cytoplasmic staining, while MCM9 Cterm showed concise nuclear staining but an absence of repair foci with MMC. Nuclear and cytoplasmic extractions of various MCM9-GFP transfected constructs were used to validate these observations in a population of cells and quantified by calculating nuclear/cytoplasmic (N/C) ratios (Fig. 2D). MCM9 L is more nuclear than cytoplasmic; MCM9 M is primarily cytoplasmic; and MCM9 Cterm is more nuclear, while alpha-Tubulin and Lamin B1 were used as cytoplasmic and nuclear controls, respectively. Quantification of the nuclear/ cytoplasmic (N/C) ratios show values above 1 for both MCM9 L and MCM9 Cterm but below 1 for MCM9 M that lacks the CTE, which are statistically different. Although MCM9 Cterm does not form MMC-induced foci, it does interact with chromatin probably though the winged-helix domain contained in the CTE (Fig. S2). Therefore, the CTE directs MCM9 into the nucleus, but the helicase core is required for localization to DNA repair sites.
A "bipartite-like" NLS is present in the CTE of MCM9 to import MCM8/9 As MCM9 Cterm had predominantly nuclear staining but MCM9 M did not, we searched for nuclear localization sequences (NLS) in the CTE and found four conserved, highconfidence putative NLS sequences (pNLS) (Fig. S3). These four pNLS sequences were individually mutated in MCM9 L -GFP, transfected into U2OS cells, and the localization was examined by confocal microscopy (Fig. 3, A and B). U2OS cells were used because of their larger nuclei that can be easily distinguished from cytoplasmic areas. Mutation of pNLS1 or pNLS2 resulted in primarily cytoplasmic staining indicating a lack of nuclear import. Alternatively, mutation of pNLS3 and pNLS4 had no noticeable effect on nuclear import. Interestingly, mutation of either pNLS1 or pNLS2 alone had a clear effect on nuclear import, indicating that a bipartite NLS1/ NLS2 is required; however, the linker between NLS1 and NLS2 spans 67 a.a., much longer than a canonical bipartite NLS that generally has a 10 to 12 a.a. linker. Therefore, the import signal motif for MCM9 is "bipartite-like" consisting of dual NLS1/ NLS2 motifs connected by an extended linker region.
To unequivocally show the effect of mutating NLS1 or NLS2 in a population of cells, we performed nuclear and cytosolic extractions on MCM9 L -GFP construct transfected into 293T cells (Fig. 3C). Again, wild-type (WT) MCM9 L is confined more in the nuclear fraction than the cytoplasm as indicated by N/C = 1.5. However, mutation of either NLS1 or NLS2 shows a significant decrease in the N/C ratio. This result confirms a requirement for both NLS1 and NLS2 in the CTE of MCM9 for efficient import into the nucleus.
We also searched in silico for a pNLS in MCM8, but no high-confidence motifs were detected. Therefore, we hypothesized that the NLS1/2 in the CTE of MCM9 may be responsible for nuclear import of the MCM8/9 complex. Therefore, we created custom CRISPR-Cas9 MCM8 or MCM9 MCM9 Cterm motifs knockout cell lines from 293T cells with significant multiallelic indels and validated phenotypically with severe MMC sensitivity (Fig. S4). To confirm if knockout of MCM9 disrupts import of MCM8 into the nucleus, MCM8-GFP was transfected into the MCM9 KO cell line (Fig. 4B). In WT 293T cells, MCM8-GFP is primarily nuclear and forms foci with MMC treatment (Fig.4A); however, in MCM9 KO cells, the MCM8-GFP staining is diffuse throughout the cytoplasm, not localized to the nucleus, and fails to form MMC-induced nuclear foci (Fig. 4B). Transfection of MCM9 L -GFP or MCM9 M -GPF into MCM8 KO cells (Fig. 4C) shows essentially the same results (as in Fig. 2, B and C) with nuclear and cytoplasmic staining, respectively. Therefore, MCM8 is imported into the nucleus complexed with MCM9 as directed by the "bipartite-like" NLS1/NLS2.

A BRCv motif is present in the CTE of MCM9
The BRC repeat sequence is a structural motif in the breast cancer type 2 susceptibility protein (BRCA2) that is utilized to promote RAD51 nucleoprotein filament formation during fork reversal or recombination (36). The eight repeats in BRCA2 show a consensus BRC motif sequence (Fig. S5), and variants of these BRC motifs (BRCv) have been found in other proteins including RECQL5 and yeast helicases Srs2, Sgs1, Mph1, and Pif1 (34). The BRCv motif in RECQL5 is shown to be important for interacting with RAD51, directing D-loop and filament formation, and responding to cross-linking stress.
In our examination of the expansive and unique CTE of MCM9, we also identified a putative BRCv motif that is similar to that found in RECQL5 and adjacent to a predicted WH Figure 3. A "bipartite-like" NLS1/2 is required for nuclear localization. A, four putative nuclear localization sequences (NLS) were identified in silico. B, mutation of these NLS sequences in GFP-MCM9 L was tested for any effect on location to the nucleus in U2OS cells. C, nuclear and cytoplasmic extractions of GFP-transfected MCM9 constructs. GAPDH is used as a control for cytoplasmic proteins. Ratios of nuclear to cytoplasmic (N/C) are indicated below the image. M is for molecular weight markers indicated on the left of the blot.

MCM9 Cterm motifs
motif important for DNA binding (Fig. 5, A and B). The BRCv consensus sequence within motif 1, FxTASxxxΦxΦS, is highly similar to the BRC repeats themselves FxTASGKxΦxΦS (where "Φ" represents hydrophobic residues). BRCv motif 2 occurs after a linker region in these HR helicases that is not present in BRCA2 and is less well conserved. However, a small hydrophobic/electrostatic patch in Motif 2, ΦL(D/E)-(D/E), is present in MCM9 and similar to RECQL5. Both BRCv motifs 1 and 2 in MCM9 are well conserved across Mammalia species, but less conserved in other metazoans (Fig. S5B).
Structural modeling of the MCM9 BRCv motif threaded onto BRCA2-BRC4 component in complex with RAD51 (34) shows the importance of the universally conserved F687 residue binding deep into a hydrophobic pocket of RAD51 ( Fig. 5C and Fig. S5C). Other conserved residues (T689, S690, S691, I695, and Y697) within motif 1 make backbone and sidechain contacts across a path of hydrophobic surface area. Residues within motif 2 make both hydrophobic and hydrogen bonding contacts with RAD51 that would loop out the linker region in between (Fig. S5, D and E). Leu714 is predicted to make hydrophobic contacts with RAD51 residues 254-5 and 258-9. Our model also predicts that Glu722 and Asn724 make H-bond contacts with Tyr205 and Arg250 of RAD51, respectively. Together MCM9 BRCv motifs 1 and 2 are predicted to make sufficient hydrophobic surface area contacts with RAD51 that are anchored by F687.

MCM9 recruitment is upstream of RAD51 during MMC treatment
As both BRCA1 and RAD51 form MMC-induced nuclear foci (37) similar to MCM9, we sought to examine whether there is colocalization of these two known ICL/HR proteins with MCM9 after DNA damage. BRCA1 has dual and independent functions in ICL repair and HR (38) but importantly acts upstream of RAD51. BRCA1 interacts with FANCI/D2 complex in the Fanconi cross-link repair pathway and works to stabilize stalled/reversed forks to facilitate ICL repair (39)(40)(41). Previously, it was shown that the MCM8/9 complex is rapidly recruited to direct DSBs caused by I-SceI and promotes the association of RAD51 directly with MCM9 for HR (9,12). However, an earlier report suggested that MCM8/9 acted downstream of RAD51 in DT40 cells treated with MMC or Cis-Pt (11), suggesting possible differences in cell types and ICL versus HR repair.
Using the cross-linking agent MMC, both BRCA1 and MCM9 nuclear foci are increased in 293T cells as expected (Fig. 6A). Furthermore, BRCA1 foci appear colocalized with MCM9 foci. MMC-induced MCM9 nuclear foci also colocalize with RAD51; however, this is incomplete as individual RAD51 foci are apparent without MCM9 (Fig. 6B). Interestingly, siRNA knockdown of RAD51 significantly reduced the RAD51 foci formation but left MCM9 foci intact ( Fig. 6B and Fig. S6). This suggests that MCM9 is acting upstream and is required (either directly or indirectly) for the recruitment of RAD51 in the MMC-induced DNA repair pathway.
To test the dependance of MCM9 in the temporal recruitment of other known effectors in the cross-link-HR pathway, we tested the ability of BRCA1 to form MMC-induced foci in MCM9 KO cells. In 293T cells, there is a significant increase in the number of BRCA1 foci per cell when treated with MMC (2.9 ± 0.6 versus 6.2 ± 0.5) as expected (Fig. 6, C and D). A similar increase in BRCA1 foci is observed in MCM9 KO cells when treated with MMC (3.4 ± 0.6 versus 7.8 ± 0.4) (Fig. 6, C and D). Although there is a slight increase in BRCA1 foci in MCM9 KO versus WT cells, this is not significant in our analysis. Therefore, the BRCA1 response to MMC treatment is not dependent and appears generally unaffected by the absence of MCM9.
As we had previously characterized MCM8 and MCM9 mutations of POF patients within consanguineous families, we examined whether these conditioned mutations in MCM9 impaired RAD51 foci formation with MMC treatment (Fig. 7, A and B and Fig. S7). 8AIV-3 designates patient EBV transformed lymphocytes from an affected MCM8 family (17) but is fully WT for both MCM8 and MCM9, acts as a suitable control, and shows significant RAD51 foci after MMC damage as expected. 9BII-4 designates lymphocytes from a heterozygous family member with one WT allele and one splice site mutation that eliminates the CTE of MCM9 (18). Even in these WT/MT cells, RAD51 foci formation is impaired and diffuses throughout the nucleus suggesting that MCM9 copy number and the CTE are important in promoting downstream repair by RAD51. Finally, 9AII-6 designates lymphocytes from a fully homozygous (MT/MT) affected patient where a nonsense mutation in exon 2 effectively creates a complete knockout of MCM9 (18). In 9AII-6 cells, there is also a complete lack of RAD51 foci and diffuse nuclear staining with MMC-induced damage.
To confirm this dependency of MMC-induced RAD51 foci on MCM9, we changed the experiment to quantify RAD51 foci in 293T WT cells versus their MCM9 KO counterparts. Treatment of 293T WT cells with MMC showed a significant increase in the number of foci per cell (6.1 ± 0.5 versus 21.2 ± 0.8), as expected (Fig. 7, C and D). Conversely, RAD51 foci in MCM9 KO cells were not significantly increased with MMC treatment (4.4 ± 0.2 versus 4.8 ± 0.2). Interestingly, the number of RAD51 foci were also significantly decreased in MCM9 KO cells compared with WT cells. Collectively, these data support the conclusion that RAD51 recruitment and loading for downstream repair of MMC-induced DNA damage are likely dependent on an interaction with the BRCv motif in MCM9.

RAD51 physically interacts with the BRCv of MCM9
In order to test whether the BRCv within the CTE of MCM9 is utilized for the direct interaction with RAD51, we switched to more explicit biochemical assays including a GST pulldown (Fig. 8A). A GST-RAD51 construct was expressed, purified, and bound to GST resin before the addition of the CTE of MCM9 (643-900 a.a.). Some MCM9 flowed through indicating either overloading or more likely a weaker interaction, but after seven washes, even more MCM9 was eluted after addition of glutathione. Further constructs of different lengths and truncations containing the BRCv mutation were prone to aggregation and insolubility issues with lower purification yields.
Therefore, we switched to a yeast two-hybrid (Y2H) assay to test the effect of a BRCv mutation on the interaction with RAD51 (Fig. 8, B-D). Both positive (1) and negative (2, 7-10) controls grew and behaved as expected in a filterbased β-galactosidase assay: blue-positive, white-negative. Both full-length MCM9 (3) and the CTE only (4) showed blue color indicative of an interaction with RAD51. Mutation of the BRCv motif in both constructs (5, 6) eliminated the positive blue color consistent with a disruption of the RAD51 interaction.

Discussion
The CTE domain of MCM9 is long and primarily unstructured; however, we identified two amino acid motifs that are necessary for both the nuclear import of the MCM8/9 complex and their response to cross-link damage (Fig. 9). The NLS motif is "bipartite-like" consisting of two positively charged amino acid stretches separated by a longer unstructured and unconventional 67 amino acid linker. The BRCv motif is also bipartite, utilized for localization to sites of MMC-induced DNA damage, and promotes the association of RAD51 for downstream repair. Interestingly, MCM9 appears to be present at the replication fork (8) and can likely respond early to MMC-induced damage by recruiting RAD51 for recombinational repair.

Unstructured C-terminal tails are common in DNA repair proteins
The CTE of MCM9 is translated primarily from the last exon (exon 12) and appears to be a later evolutionary addition to the MCM9 gene (33). MCM9 is almost universally present together with MCM8 and is distributed in most opisthokonts, excavates, and chromalveolates (42). Interestingly, MCM9 is fully present in plantae; however, the terminal exon giving rise to the CTE is absent in this eukaryotic supergroup. MCM9 members are absent in some opisthokonts, which include animals and fungi, but of those present, the primary isoform includes the CTE, while other isoforms are alternatively spliced for shorter proteins.
Several DNA replication and repair proteins contain large unstructured regions or domains that are required for proper function including several translesion synthesis DNA polymerases (43,44), DNA replication initiation proteins (45), various DNA repair helicases (6), mismatch repair proteins (MutLα) (46), nucleotide excision repair proteins (XPA) (47), the heterotrimeric single-stranded binding protein (RPA) (48), and even the tumor suppressor p53 (49). Some residues in these regions are posttranslationally modified, but there are also specific motifs that are either structured on their own or adopt structure in order to control their localization, interactions, and/or functions. Previously, a role for the CTE of MCM9 was unknown. No posttranslational modifications are known to occur in the CTE, but we have validated both an unconventional bipartite-like NLS and a bipartite BRCv motif that are required to direct function of MCM8/9 to facilitate MMC-induced DNA damage repair.

MCM9 Cterm motifs
MCM9 has an unconventional "bipartite-like" NLS Classic NLS (cNLS) motifs are generally recognized as clusters of basic amino acids that are either monopartite or bipartite (50,51). Bipartite NLS motifs are conventionally described as being separated by 10 to 12 residues based on the initial characterization of the cNLS of Xenopus laveis nucleoplasmin (52). However, other bipartite cNLS with longer or extended linkers (20-25 residues) has been characterized in Smad4 (53), Aspergillus topoisomerase II (54), and XRCC1 (55,56), and systematic extension of the linkers in bipartite cNLS has shown that longer linkers can be utilized in nuclear import (57,58). Yet, the linker region between NLS1 and NLS2 of MCM9 is 67 amino acids and is to our knowledge one of the longest, prompting us to term this "bipartite-like" NLS1/2.
Nuclear import of a bipartite NLS involves the binding of the upstream and downstream basic residues to minor and major binding pockets, respectively, for importin-α (59,60). Both backbone and side-chain interactions can be made from the linker region to a region between binding pockets. Once bound, Importin-α transports the cargo protein along with Importin-β through the nuclear pore complex depositing into the nucleus (61). It is hard to imagine that the longer unstructured linker between NLS1 and NLS2 in MCM9 would make specific contacts in this region, although this remains to be determined. However, we would predict that both NLS1 and NLS2 are required to bind to the binding pockets of Importin-α to facilitate import.
Based on our results, the "bipartite-like" NLS1/2 in MCM9 is also responsible for the nuclear import of MCM8. MCM8 is presumed to be constitutively bound to MCM9 after translation to form a higher-order oligomeric complex. MCM8 is devoid of any identifiable NLS and, based on its large molecular weight, would require an association with a partner for nuclear import. This is reminiscent of DNA ligase 3 (Lig3α) being bound to XRCC1 for cotransport into the nucleus for DNA repair (62). Another example includes the bipartite NLS in Smad4 being utilized for import of other complexed and phosphorylated Smad proteins for transcriptional regulation (53). Interestingly, several studies have shown that knockdown or knockout of MCM8 also eliminates MCM9, but that  . Model for MCM8/9 recruitment of RAD51 to sites of damage. The C-terminal bipartite-like NLS1/2 motif in MCM9 is responsible for nuclear import of MCM8/9, and the BRCv motif is required for recruiting and loading of RAD51 to sites of MMC-induced DNA damaged forks.

MCM9 Cterm motifs
knockdown of MCM9 has more of a singular effect (12,14). Therefore, we would hypothesize that MCM9 is stabilized by the presence of MCM8 in an MCM8/9 complex and MCM8, although stable on its own, is unable to be imported into the nucleus. This would represent an important regulatory mechanism for ensuring that MCM8/9 is in a complex and not represented as individual components.

MCM9 has a BRCv motif in the CTE
The BRC repeats are hallmarks within the BRCA2 protein required for interactions with RAD51 during HR and are not typically found in other proteins. A variant BRC motif (BRCv) has been characterized in the human DNA repair helicase, RECQL5, and other BRCv motifs are predicted to be present in other DNA repair helicases including yeast Srs2, Mph1, Sgs1, and Pif1 (34). The BRCv in MCM9 shows high homology to the motifs 1 [FXTA(S/T)] and 2 [Φ(L/F)XX(D/E)] in RECQL5 with an equivalent 14 amino acid linker region between both motifs. Like that for RECQL5, BRCv motif 1 appears to be more conserved and makes more intimate contacts with the oligomerization interface of RAD51. Besides the conserved F687 and residues adjacent, I695 in MCM9 is homologous with L672 in RECQL5 as a small hydrophobic residue that is important for RAD51 association (34). When modeled for MCM9, L672 makes significant contacts in the hydrophobic patch adjacent to F687 making up a lynchpin for binding. Motif 2 is just adjacent to the proposed binding site for motif 1 and acts to strengthen bipartite binding to RAD51 through conserved central hydrophobic residues that include L720. Interestingly, the linker region is not conserved between these helicases and would represent an extruded unstructured area only required as a spacer for proper positioning of motif 2. Therefore, like RECQL5, our data support the conclusion that the BRCv motif in MCM9 is required to regulate RAD51 recruitment and filament formation to mediate fork restart or recombination activities at the replication fork during times of stress.

MCM9 aids in the recruitment of RAD51 to MMC-induced sites of damage
There appears to be two conflicting viewpoints for when MCM8/9 temporally responds to various types of DNA damage in relation to RAD51. Several lines of evidence examining the cellular role of MCM8/9 come primarily from responses to the cross-linking agent (Cis-Pt) and correlate to a purported activity in later stages of HR downstream of RAD51 synapsis (9,11). In those studies, RAD51 foci formed after Cis-Pt treatment are still present in MCM8 or 9 knockout cells, indicating no dependence of RAD51 on MCM8/9. This directly contradicts work here, which shows severely decreased RAD51 foci in patient deficient MCM9 cells after treatment with MMC, indicating a codependence for MCM8/9 with RAD51. Instead, these results correlate with MCM8/9 acting upstream of RAD51 after treatment with MMC, consistent with other studies in human/Xenopus systems (10,12). MCM8/9 are present early in the ICL/HR repair pathway during fork stalling, reversal, stabilization, and prior to RPA binding or synapsis (10,12,14) but are not dependent on BRCA1. MCM8/9 also colocalize with BRCA1 and a subset RAD51 foci in humans; however, MCM8/9 are required for RAD51 foci formation but not for that of BRCA1. To reconcile these discrepancies, it is probable that MCM8/9 are responding to yet unknown specificities and overlapping aspects of DNA recombination associated with fork stalling, ICLs, and DSBs dependent on RAD51 that will require further experimental depth.
Cross-linking agents such as Cis-Pt and MMC are commonly thought of in the same vein, and although both typically only cause about 5 to 10% ICLs, they have different targets and effects on duplex structure (63,64), which may be responsible for different temporal recruitment of MCM8/9. Cis-Pt cross-linking occurs at N7 of guanine at either 5'-GpG or 5'-ApG resulting primarily in intrastrand cross-links and <8% ICLs (65). MMC reacts with the N 2 exocyclic amine of guanine preferentially at 5'-CpG sites forming monoaducts that insert into the minor groove and can form ICLs across in about 10% of products (64,66,67). However, the major difference between Cis-Pt and MMC is their effects on duplex structure. Cis-Pt induces major helical distortion with both monoadducts and ICLs underwinding the helix, bending of the strands inward, and flipping out of the unpaired cytosines (68). On the other hand, MMC causes minimal distortion, no bending, and maintains the C-G base pairs even with the G-mito-G ICL (69). These differences will be important in understanding MCM8/9 specificities, recruitment, and effects on RAD51 pathways moving forward.
Single monoadducts or intrastrand cross-links on either the leading or lagging strands are blocks to DNA synthesis decoupling the replisome, while ICLs are total replisome blocking adducts. In these scenarios, various mechanisms of recombination would be utilized including fork reversal, template switching, and single-strand breaks repaired by sister chromatin recombination. It is probable that Cis-Pt and MMC direct MCM8/9 to act differentially in recruiting RAD51 in multiple pathways for repair. Damage from Cis-Pt is primarily recognized and repaired by nucleotide excision repair (NER) (70,71), but because the duplex is distorted, the mismatch repair (MMR) pathway also contributes (72,73). On the other hand, MMC lesions may only be recognized during active DNA replication processes and repaired by various translesion synthesis (TLS), HR, and NER pathways (74). Therefore, our work here has provided evidence that the CTE within MCM9 is important in regulating MCM8/9 activities at sites of DNA damage. Entry into the nucleus after translation is directed through an unconventional "bipartitelike" NLS in the CTE of MCM9 that imports MCM8/9 as a complex. Once there, MCM8/9 associates with the replisome to actively respond to DNA damage encountered during synthesis. Recruitment of RAD51 occurs through a bipartite BRCv motif in the CTE of MCM9 that carries out downstream recombination. Even so, further efforts are needed to MCM9 Cterm motifs better understand the specific enzymatic functions and interactions of MCM8/9 in relation to different DNA damage agents to more specifically describe its reported helicase function in relation to the many other analogous HR helicases.

Materials
Oligonucleotides were purchased from Sigma or IDT. ON-TARGET plus siRNA to RAD51 from Dharmacon. Mission siRNA and universal negative control #1 was from Sigma-Aldric. Mitomycin C (MMC) was from Thermo Fisher. Restriction enzymes from New England Biolabs. All other chemicals were analytical grade or better.

Protein purifications
GST-tagged RAD51 was purified as described previously (81). C-terminally truncated MCM9 constructs (643-900 or 680-900) with GST at the N-terminus were transformed into C43 pLysS (Lucigen) and induced with 0.25 mM IPTG at OD 600 0.5 for 4 h at 37 C. Pellets were resuspended in lysis buffer (20 mM HEPES pH 8.0, 300 mM NaCl, 10 mM BME, 10 mg/ml lysozyme, 0.1% Triton-X 100, 1 mM PMSF, 1 mM EDTA, 4 μg/μl pepstatin A) on ice. The clarified lysate was loaded onto the GSTrap column in Buffer A (20 mM HEPES pH 8.0, 300 mM NaCl, 10 mM BME, 1 mM EDTA, 10% glycerol) and eluted with Buffer B (Buffer A + 10 mM glutathione). Fractions were concentrated using a spin concentrator and rocked overnight at 4 C with TEV protease. The cleaved product was applied again to a GSTrap column, and the flow through collected. Purity was confirmed to be greater than 95% by SDS-PAGE, and the concentration was determined by absorbance using ε = 12,615 M −1 cm −1 .

Circular dichroism
CD spectra of 5 μM purified MCM9 643 to 900 and 680 to 900 were collected using a Jasco J-815 spectropolarimeter at room temperature (23 C) in 10 mM potassium phosphate buffer (pH 7.0) containing 100 mM (NH 4 ) 2 SO 4 using a 0.1 cm quartz cell. Five individual spectra were collected using a 1 nm bandwidth at 0.5 nm intervals at a speed of 50 nm/min before averaging. The raw data in measured ellipticity (ε) was then analyzed using BeStSel (http://bestsel.elte.hu/index.php) for conversion to Δε units and determination of secondary structure (82,83). All data points analyzed were collected under a HT measurement of 800 V. Secondary structure predictions had NRMSD values <0.05.

Cell transfection and treatment
HEK293T or U2OS cells plated on poly-lysine-coated coverslips were transfected with the respective EGFP-MCM9 constructs using linear polyethyleneimine (LPEI) or lipofectamine 3000 (Thermo Fisher) according to manufacturer's directions. Briefly, transfection reagent and plasmids were incubated at RT for 10 min and added to cells in Opti-MEM serum-free media (ThermoFisher) or DMEM 10% FBS with no antibiotics. siRNA RAD51 knockdown or the universal negative control #1 was transfected alone or in combination with a plasmid using Dharmafect1 or DharmafectDuo (Dharmacon). The transfection mixture was removed after 6 h and replaced with DMEM/10% FCS for up to 24 h. The cells were then treated with indicated concentrations of MMC (Sigma-Aldrich) for 6 h. Posttreatment, the cells were immediately prepped for western blotting, direct confocal microscopy, or immunofluorescence. Western blots were probed with RAD51 (H-92 or sc-8349) and α-tubulin (TU-02) antibodies (Santa Cruz Biotechnology Inc). Fluorescent secondary antibodies were α-rabbit IgG-Alexa488 or α-mouse IgG-Alexa647 conjugates (ThermoFisher). Membranes were scanned using the LAS 4000 imager (Cytiva).

Immunofluorescence
Adherent cells were washed in PBS (2 times), fixed in 4% paraformaldehyde in PBS for 10 min, and permeabilized with 0.1% Triton X-100 in PBS (PBST) for 15 min. Cells were blocked overnight with 5% BSA in PBST at 4 C and then incubated with 1:50 RAD51 primary antibody (sc-8349, Santa Cruz or ab63801, Abcam) or 1:50 BRCA1 primary antibody (sc-6954, Santa Cruz) in 2.5% BSA in PBST for 1 h at 37 C. Cells were washed three times in PBST and incubated with 1:50 dilution of the appropriate fluorescent secondary antibody followed by wash with PBST (three times). Cells were mounted in DAPI mountant (Prolong Gold, Thermo Fisher) and sealed with clear polish, and imaged under a FV-1000 epifluorescence or FV-3000 confocal laser scanning microscope (Olympus Corp). Images were processed with vendor included Fluoview (v.4.2 b) or CellSens (v2.2) software. For suspension of EBV transfected lymphocytes, cells were immunostained similar to adherent cells but were harvested by centrifugation after MMC treatment, and immunostaining was performed in suspension. To mount the cells, they were resuspended in 1 μl of mountant and placed on a coverslip, which was then picked up using a microscopic slide, and sealed with clear polish. RAD51 foci from epifluorescence images were automatically counted from individually gated cells using identical thresholds that eliminated background noise using Image J (84). Foci per cell are presented in a dot plot, averaged, and the standard error reported.

Subcellular fractionation
HEK293T cells transfected with EGFP constructs containing MCM9 truncations/mutations were cultured in 6-well plates and harvested. Fractionation into nuclear and cytoplasmic portions was carried out according to manufacturer's directions using a Nuclear/Cytosol Fractionation Kit (K266, BioVision). Equal volumes of protein were loaded onto a 12% acrylamide gel and transferred to PVDF membranes.

Affinity pull-down
GST-tagged RAD51 (4 μM) was added to glutathione agarose (UBPBio) on ice for 30 min before spinning down and removing the supernatant. Then, an equal molar amount of purified MCM9 (643-900) was added and incubated for 30 min, mixing every 10 min. Samples were spun down, and supernatant was reserved as the flow-through. The GSTagarose was washed with an equal volume of GST-binding buffer (20 mM HEPES [pH 7.5], 75 mM NaCl, 0.1 mM EDTA, 1 mM BME, and 10% glycerol) and centrifuged for 60 s at 8000g at least seven times. Proteins were eluted with an equal amount of GST elution buffer (20 mM HEPES [pH 8.0], 300 mM NaCl, 1 mM EDTA, 10 mM BME, and 10% glycerol, and 10 mM reduced glutathione). Eluted samples were separated using 10% SDS-PAGE gels and stained with Coomassie.

Data availability
All data are contained within the article.
Acknowledgments-We acknowledge the Baylor Molecular Bioscience Center (MBC) and the Center for Microscopy and Imaging (CMI) for providing instrumentation and resources aiding this project.  Conflicts of interest-The authors declare that they have no conflicts of interest with the contents of this article.