Pol II is a major component of a RecQL5 complex.
To purify the RecQL5 (referred to as the RecQL5β isoform)-associated complex, we generated a HeLa cell line stably expressing Flag-tagged RecQL5 (FRecQL5). Flag immunoprecipitation (IP) of FRecQL5 followed by SDS-PAGE revealed the presence of three major polypeptides (Fig. , compare lanes 2 and 3). The polypeptides were later determined by mass spectrometry (MS) to be RecQL5 and the two largest subunits of Pol II, RPB1 and RPB2. The fact that RPB1 and RPB2 are the major polypeptides in the FRecQL5 IP suggests that Pol II is a main component of a RecQL5-associated complex. This notion is further supported by the following evidence. First, immunoblotting confirmed the presence of Pol II in the FRecQL5 IP (Fig. , lane 6). Second, MS analyses of the entire FRecQL5 IP mixture detected the presence of six additional subunits of Pol II (RBP3, -4, -5, -8, -9, and -11) (see Fig. S1a in the supplemental material). Third, we performed reciprocal IP with a Pol II antibody using HeLa cells expressing FRecQL5 and identified RecQL5 as a major polypeptide of Pol II by silver staining and MS analyses (Fig. , lane 2, and see Fig. S1b in the supplemental material). Fourth, in cells lacking FRecQL5, we detected the presence of endogenous RecQL5 in the Pol II IP (Fig. , lane 6), although the level of RecQL5 was lower than that from cells overexpressing FRecQL5 (Fig. , compare lane 2 to lane 1). Our data are consistent with previous reports that RecQL5 associates with Pol II (3
FIG. 1. A major component of a RecQL5 complex is Pol II. (a) Silver-stained SDS-PAGE gel of the RecQL5-Pol II complex isolated by Flag IP from HeLa cells expressing Flag-RecQL5 (FRecQL5). The three major polypeptides identified by mass spectroscopy analysis are (more ...)
We analyzed the presence of other known RecQL5-interacting partners in our FRecQL5 IP by immunoblotting and observed the presence of Rad51, PCNA, and the MRN complex but not Topo3α and Topo3β (Fig. , lane 6, and see Fig. S2, lanes 8 and 10, in the supplemental material). Because these partners were not detected by mass spectrometry (unlike Pol II), they may associate with small fractions of RecQL5 compared to Pol II.
We also analyzed our Pol II IP for the presence of other RecQ helicases, such as WRN and BLM, but failed to detect any of them (Fig. , lane 6). This is consistent with early findings that the RecQL5-Pol II association is specific (3
). Moreover, this association was not significantly altered by the treatment of cells with drugs that induce replication stress, such as camptothecin (CPT) and hydroxyurea (HU) (the ratio of Pol II over RecQL5 in the RecQL5 immunoprecipitate changes no more than 2-fold) (Fig. ).
RecQL5 interacts with Pol II through conserved KIX and SRI domains.
We mapped the Pol II-interacting region(s) in RecQL5 by deletion mutagenesis. Various Flag-tagged deletion mutants (Fig. ) were transfected into HEK293 cells and analyzed by Flag IP and Western analyses. Two nonoverlapping regions outside the RecQL5 helicase domain were each found to be capable of associating with Pol II (Fig. ; summarized in Fig. ). The first region (residues 501 to 650) associated with both hypophosphorylated (IIa) and hyperphosphorylated (IIo) Pol II, whereas the second region (residues 745 to 991) interacted only with Pol IIo (Fig. ). These data are consistent with a recent finding that RecQL5 has two independent Pol II-interacting regions (21
FIG. 2. RecQL5 interacts with Pol II through two conserved domains, the KIX and SRI domains. (a) Schematic diagram of the full-length (FL) and truncated forms of Flag-RecQL5. The N-terminal helicase and RecQ C-terminal (RQC) domains are shown along with the newly (more ...)
Bioinformatic analyses revealed that each Pol II-interacting region contains a conserved protein interaction domain (11
). The one in the first region (aa 540 to 620) is homologous to the KIX domain found in several Pol II transcriptional regulators (32
), whereas the one in the second region (aa 909 to 991) is homologous to the SRI (Set2 Rpb1-interacting) domain present in the histone methyltransferase SetD2, which also regulates transcription (23
) (Fig. ). Interestingly, the SetD2-SRI domain is known to interact with the phosphorylated C-terminal domain (CTD) (pCTD) of Pol IIo, which is in complete agreement with the function of the predicted SRI domain in RecQL5.
Homology modeling predicts a strong structural similarity between RecQL5-KIX and other KIX domains.
Several KIX domain structures were previously determined, with all of them containing a KIX fold, a three-helix bundle with a conserved hydrophobic core (37
). Primary-sequence alignment revealed a high degree of similarity between RecQL5-KIX of different species and the KIX domains from several Pol II transcriptional regulators, CBP, p300, and Med15 (Fig. ). Secondary-structure predictions identified three potential α-helices in RecQL5-KIX, consistent with the presence of the KIX fold (Fig. and data not shown). Given the high degree of primary-sequence and secondary-structure similarity, a homology model for the RecQL5-KIX domain was generated by a standard protein comparison method using the mouse CBP (mCBP)-KIX domain (PDB accession number 1SB0) as a structural template (Fig. ) (2
FIG. 3. The RecQL5-KIX domain interacts with both Pol IIa and Pol IIo through a common interface utilized by other KIX domains. (a) Sequence alignment showing similarity between the RecQL5-KIX domain and KIX domain family members CBP, p300, and Med15. KIX family (more ...)
Our model predicts that RecQL5-KIX contains a three-helix bundle similar to that of mCBP-KIX. Invariant hydrophobic positions within the core of the KIX fold are well conserved in primary sequence and in tertiary space between the two proteins (Fig. ). In contrast, surface-exposed hydrophobic residues are not conserved between RecQL5 and mCBP (Fig. ).
The KIX fold from all proteins contains two invariant charged residues and one invariant aromatic residue, all three of which are strictly conserved in RecQL5-KIX (R550, E584, and Y597) (Fig. ). The corresponding residues in mCBP (R600, E636, and Y649) form a stabilizing hydrogen bond network with several nearby hydrophobic core residues (V595 and Y640) (37
) (Fig. ). Interestingly, all residues of this extended hydrogen bond network are well conserved in RecQL5, suggesting that a similar network is present in RecQL5-KIX (Fig. ).
Mutations in the RecQL5-KIX domain disrupt its Pol II association.
A previous study showed that a mutation of the invariant glutamate residue within the hydrogen-bound network of Med15-KIX abolishes its interaction with its partner (SREBP-1a) (45
). We found that mutations at comparable residues in RecQL5-KIX, E584D and E584A, also disrupted its interaction with the corresponding partners, Pol IIo and Pol IIa (Fig. and data not shown). The data support our model that RecQL5-KIX contains a hydrogen bond network similar to that of other KIX domains and suggest that this network may be essential for all KIX domains to associate with their partners.
We next examined whether the E584 mutations can affect the association between full-length RecQL5 and Pol II. Because full-length RecQL5 consists of two independent Pol II interaction domains, the E584 mutations are expected to abrogate only the Pol IIa association (which is mediated solely by the KIX domain) but not the Pol IIo association (which can occur through either KIX or SRI). Consistent with this prediction, both E584D and E584A mutants lost their associations with Pol IIa but maintained their associations with Pol IIo (Fig. , lane 3, and data not shown). Moreover, mutations of another invariant residue within KIX, Y597A, reduced the Pol IIa association by about 70% while retaining the normal association with Pol IIo (Fig. , lane 7). Together, the data indicate that the KIX domain and the integrity of its hydrogen bond network are indispensable for the interaction between RecQL5 and Pol IIa but dispensable for the interaction with Pol IIo.
We found that a mutation of the third invariant residue of KIX, R550, strongly destabilized RecQL5 (Fig. , right, lane 6), suggesting that R550 is critical for the proper folding of the full-length protein. Interestingly, a previous study showed that a 9-residue deletion mutant, which includes R550, drastically decreased the association between RecQL5 and Pol II in vitro
). Our RecQL5-KIX homology model suggests that this mutant lacked critical hydrophobic core residues and the first turn of helix 1, which could result in the destabilization of the native KIX domain.
RecQL5-KIX interacts with Pol II through a common protein interaction surface.
Next, we used the RecQL5-KIX domain homology model to predict the Pol II interaction surface. Unlike residues that maintain the integrity of the KIX fold, many of the surface-exposed residues that mediate direct intermolecular protein interactions are not conserved between RecQL5 and mCBP (Fig. ). The largest contiguous RecQL5 conserved surface mapped along the interface of helix 1 and helix 3. Strikingly, mCBP-KIX and Med15-KIX use the same surface to bind their respective partners (Fig. ) (10
To determine whether the predicted interface in RecQL5 is important for its interaction with Pol II, we independently mutated four residues on this surface, C553A, L556D, L602D, and K598A, in full length RecQL5 (Fig. ). IP-Western analysis showed that mutations of the first 3 residues reduced the Pol IIa association by more than 80% (Fig. , right, lanes 8, 9, and 11), supporting our model that RecQL5 employs the same protein-interacting surface as other KIX domains. As expected, all mutants retained the association with Pol IIo due to redundant interactions with the SRI domain.
The strongest reduction of the Pol II association (>95%) was observed with the RecQL5 L602 mutation, which corresponds to residues A654 in mCBP and A60 in Med15 (Fig. ). Mutations of this residue in the latter KIX domains disrupt their interactions with their respective partners (37
). In mCBP, A654 specifically recognizes a conserved leucine found in several CBP partners. Although RecQL5 has a leucine rather than an alanine at this position, the role of this residue in mediating protein interactions is likely conserved.
The interaction between the RecQL5-KIX domain and Pol II is specific.
Previous studies have shown that a KIX domain can interact with multiple partners and that several KIX domains can interact with the same partner (45
). We examined whether other KIX domains can interact with Pol II by transfecting various Flag-tagged KIX domains into HEK293 cells, followed by IP-Western analysis. Only the KIX domain of RecQL5, but not those of CBP, p300, and Med15, associated with Pol II (Fig. ), suggesting that the interaction between RecQL5-KIX and Pol II is specific. This specificity is consistent with our homology model showing that several residues at the protein-interacting surface of RecQL5-KIX are distinct from those in other KIX domains.
SRI domains of RecQL5 and SetD2 likely use a homologous interface to bind Pol IIo.
Primary-sequence alignment and secondary-structure prediction revealed strong similarity between the SRI domains of RecQL5 and SetD2 (Fig. ). To further examine the relationship between the two domains, we generated a homology model of the RecQL5-SRI domain based on the structure of the human SetD2-SRI domain that was previously solved (PDB accession number 2A7O) (Fig. ) (2
). Our model predicts that RecQL5-SRI consists of the same structural fold, a three-helix bundle and a conserved hydrophobic core, as SetD2-SRI, although the length of the former is approximately 28 residues shorter than that of the latter. The regions that showed the strongest difference are the first two helices and the intervening loops, in which many residues conserved in SetD2-SRI are either different or absent in RecQL5-SRI (Fig. ). Many of these residues are located along the periphery of the hydrophobic core or in structurally divergent sequences (Fig. ). Despite such differences, the helical segments comprising the main SRI fold are well conserved in RecQL5-SRI, including the essential hydrophobic core residues (Fig. ). Importantly, the surface-exposed residues, especially those critical for Pol II-pCTD binding (27
), are invariant between the two SRI domains (Fig. ). We therefore hypothesize that RecQL5 binds Pol IIo using the same conserved surface as does SetD2-SRI (Fig. ).
FIG. 4. The RecQL5-SRI domain interacts only with Pol IIo, and its structure is similar to that of the human SRI domain. (a) The sequence alignment shows strong similarity between the RecQL5 and SetD2 SRI domains. The color usage is described in the legend of (more ...)
To test this hypothesis, we mutated two of the conserved surface residues (K939 and R943) of RecQL5-SRI to alanine. Mutations of the comparable residues in SetD2-SRI have been shown to strongly diminish its interactions with the Pol II-pCTD (27
). IP-Western analysis showed that RecQL5-SRI domains carrying either K939 or R943 mutations failed to associate with Pol IIo (Fig. ), supporting our hypothesis that RecQL5 binds Pol IIo-pCTD though the same interface utilized by SetD2.
Both the KIX and SRI domains in RecQL5 can independently bind Pol II.
We investigated the effect of SRI domain mutations on the association of full-length RecQL5 and Pol II. Two SRI domain deletion mutants (residues 1 to 900 and 1 to 650) retained normal associations with both forms of Pol II (Fig. ), indicating that the SRI domain is dispensable for full-length RecQL5 to bind Pol II. This result is expected because both SRI deletion mutants retain the intact KIX domain that is capable of binding both Pol IIa and Pol IIo. Together with the finding that the KIX point mutant (E584D) retained its association with Pol IIo due to the presence of the SRI domain (Fig. , lane 3, and 4f, lane 3), our data suggest that both the KIX and SRI domains can independently bind Pol II.
One prediction from the above-described suggestion is that the inactivation of RecQL5-Pol II interaction requires simultaneous mutations of both the KIX and SRI domains. We generated two such mutants, E584D K939A and E584D R943A, and found that they substantially reduced associations with both forms of Pol II (Fig. , lanes 4 and 5). The results are in complete agreement with the notion that KIX and SRI domains bind Pol II independently and further indicate that there are no additional Pol II-interacting domains in RecQL5.
The RecQL5-Pol II association is conserved in chicken DT40 cells.
Next, we utilized chicken DT40 cells to investigate the significance of RecQL5-Pol II interactions in vivo
. DT40 cells have a high gene-targeting efficiency and have been used widely for genetic analyses of DNA damage response factors (44
), including RecQL5, BLM, RecQL1 (31
), and multiple Fanconi anemia proteins. The findings from these analyses were invaluable and they led directly to the discovery of a new Fanconi anemia gene (26
). One reason for this is that human and chicken proteins are highly homologous in sequence, so their functions are well conserved.
For RecQL5, both its KIX and SRI domains are highly conserved between human and chicken, with about 60% identity and over 70% similarity (Fig. and ). More importantly, the residues predicted to interact with Pol II are strongly conserved, with over 70% identity and 93% similarity for the KIX domain (Fig. ) and 100% identity for the SRI domain (Fig. ). For Pol II, it is known to be one of the most highly conserved proteins in all eukaryotes. For example, the RBP1 subunit of Pol II is over 90% identical and 95% similar between human and zebrafish. Because both Pol II and its interaction domains in RecQL5 are highly conserved between human and chicken, we predict that their association is also conserved in DT40 cells.
FIG. 5. The association between RecQL5 and Pol II is conserved in chicken DT40 cells. (a and b) RecQL5-KIX (a) and RecQL5-SRI (b) human and chicken conserved Pol II interaction surfaces. (a) Ribbon model of human RecQL5-KIX shown as described in the legend of (more ...)
Consistent with this prediction, human RecQL5 transfected into DT40 cells coimmunoprecipitated with both Pol IIa and IIo (Fig. ). Moreover, various KIX and SRI mutants of RecQL5 displayed the same Pol II interaction patterns in DT40 cells as those in human HEK293 cells (compare Fig. with and ). For example, the KIX domain point mutant (E584D) in DT40 cells similarly exhibited a reduced association with Pol IIa but retained normal associations with Pol IIo. The SRI domain deletion mutant (ΔSRI) had associations with both Pol IIa and Pol IIo. The double point mutants of the KIX and SRI domains (E584D K939A) lost associations with both Pol II forms. These data suggest that the mechanism of RecQL5-Pol II interactions is conserved in chicken, which allows the use of DT40 cells for functional analyses of these interactions.
We found that human RecQL5 transfected into DT40 cells coimmunoprecipitated with several other interacting partners of RecQL5, including RAD51 and PCNA (Fig. , lane 5). The data suggest that the interactions between RecQL5 and its other partners are also conserved in chicken.
The KIX but not the SRI domain is required for the suppression of SCE by RecQL5.
A previous study showed that RecQL5−/−/BLM−/−
DT40 cells had an SCE level that was higher than that of BLM−/−
). We repeated these experiments and obtained similar results (Fig. and see Fig. S3 in the supplemental material). The introduction of human RecQL5 into RecQL5−/−/BLM−/−
cells restored the SCE level to that of BLM−/−
cells (Fig. ), indicating that RecQL5 is responsible for the elevated levels of SCE in these cells. This allowed us to test various RecQL5 mutants using the same assay to determine the importance of the different activities of RecQL5 in cells. Control experiments showed that all mutants were expressed at levels similar to or above that of the wild-type protein (Fig. ).
FIG. 6. Both helicase activity and KIX domain-mediated Pol IIa interactions are required for RecQL5 to suppress SCE and resist CPT-induced cell killing. (a) Summary of different human RecQL5 mutants and their corresponding activities in association with Pol II, (more ...)
We found that the KIX domain E584D mutant partially suppressed the abnormal SCE phenotype of RecQL5−/−/BLM−/− cells (Fig. ). In contrast, the SRI domain deletion mutant (ΔSRI) fully suppressed the abnormal SCE phenotype (Fig. ). Moreover, the KIX-SRI double mutant (E584D K939A) suppressed SCE to a level similar to that by the KIX domain single mutant (E584D) (Fig. ). These data demonstrate that the two Pol II interaction domains in RecQL5 are functionally nonequivalent in the suppression of SCE: the KIX domain is essential, whereas the SRI domain is dispensable.
The KIX but not the SRI domain is required for cellular resistance to camptothecin.
cells have been shown to exhibit hypersensitivity to CPT, a topoisomerase I inhibitor that blocks replication (19
). However, neither RecQL5−/−
DT40 cells showed significant CPT sensitivity, whereas RecQL5−/−/BLM−/−
double-mutant cells displayed the sensitivity (Fig. ). The feature is reminiscent of the SCE data in that only the double mutant, but not the RecQL5 single mutant, displayed a defective phenotype. The data suggest that RecQL5 and BLM have redundant functions not only in the suppression of crossover recombination but also in the resistance of replication stress.
We tested RecQL5 and its mutants using the CPT sensitivity assay. The transfection of human RecQL5 into RecQL5−/−/BLM−/− cells largely restored cellular resistance to CPT (Fig. ). In addition, the KIX domain mutant (E584D) partially restored CPT resistance, whereas the SRI domain mutant (ΔSRI) almost fully corrected CPT resistance (Fig. ). Furthermore, the double mutant of both domains, E584D K939A, had an effect similar to that of the KIX single mutant. These data are in complete agreement with findings from the SCE assay and suggest that the KIX domain is required for not only SCE suppression but also CPT resistance, whereas the SRI domain is dispensable for both.
The RecQL5 helicase and KIX domains are both required for full suppression of SCE and resistance to CPT.
RecQL5 possesses a highly conserved helicase domain and an associated helicase activity, but their importance in vivo
remains unclear. To address this issue, we generated two helicase mutants and investigated if they can complement the abnormal SCE and CPT sensitivity phenotypes of RecQL5−/−/BLM−/−
cells. The first mutant (K58R) was shown previously to lack ATP-dependent helicase activity (15
), whereas the second mutant (ΔHel) had a majority of its helicase domain deleted (residues 1 to 241) so that it should lack both ATP-dependent helicase activity and ATP-independent DNA-binding activity. Both mutants retained normal associations with Pol IIa and IIo (Fig. , lanes 5 and 7, and d, lane 7), which is expected given the presence of the intact KIX and SRI domains. Notably, both mutants only partially restored the abnormal SCE level and CPT resistance of RecQL5−/−/BLM−/−
DT40 cells compared to the wild-type protein (Fig. ), suggesting that the helicase activity is important for RecQL5 to function in vivo
. The fact that these helicase mutants had partial activity in correcting the SCE and CPT phenotypes suggests that RecQL5 can act through a mechanism distinct from that of its helicase activity.
We noticed that the Pol IIa interaction mutant (E584D) also had partial activity in restoring SCE levels and CPT resistance of RecQL5−/−/BLM−/− DT40 cells, a feature reminiscent of that of the helicase mutants. This implies that the helicase activity and Pol IIa interaction are parallel in protecting genome stability. To test this hypothesis, we created a double mutant, K58R E584D, to inactivate both its helicase activity and Pol IIa interaction. RecQL5−/−/BLM−/− cells complemented by this mutant had SCE levels and CPT sensitivity indistinguishable from those of double-null cells (Fig. ). The data suggest that RecQL5 can promote genome stabilization through either its helicase activity or Pol IIa association. Only when both activities are disrupted does RecQL5 completely lose its function.
As a control, the E584D, K58R, and K58R E584D mutants associated with normal levels of RAD51, PCNA, and the MRN complex by IP-Western analyses of HEK293 cell extracts (see Fig. S2, lanes 9 to 12, in the supplemental material). Moreover, the same mutants also associated with normal levels of RAD51 and PCNA in DT40 cell extracts (Fig. , lanes 5 to 8). These data argue that a loss of function of these mutants is not due to the disruption of their association with other partners of RecQL5.