The replication of the genome by exquisitely accurate DNA polymerases is critical for the maintenance of genomic stability. In eukaryotes, DNA polymerases Pol δ [1], [2], [3], [4] and Pol ε [5], [6], [7] replicate genomic DNA. There is a “division of labor” in that Pol δ is primarily involved in lagging strand synthesis, while Pol ε is involved in leading strand synthesis [5]. Pol δ also has major roles in gap filling in DNA repair processes (NER, BER, MMR) and in D-loop extension during HDR (homology directed repair) of DSBs (double-strand breaks) [1], [2], [3], [4], [8].

Extension of the invading strand in the D-loop is a key step in HDR of DSBs. Once a D-loop is created, the 3’ end of the invading strand forms what is a primer on a template strand. The extension of this primer in the D-loop by DNA polymerases has been variously referred to as displacement synthesis, homology associated DNA synthesis, or simply D-loop extension. D-loop extension is involved in several HDR subpathways: synthesis dependent strand annealing (SDSA), double-Holliday Junction (dHJ) repair [9], [10], [11], [12], [13], [14], [15], and Break-induced replication (BIR), which is utilized for repair of single-ended DSBs [16], [17]. BIR is also involved in responses to replication stress from a number of sources that include stalled replication forks, under-replicated DNA at common fragile sites (CFS) [18], G4 quadruplex structures and damaged telomeres [12], [19], [20]. This convergence through multiple avenues of replication stress that can cause potential harm to genomic stability is intimately connected to HDR [21].

It has been generally thought that Pol δ (rather than Pol ε) is the replication polymerase involved in HDR. Studies in yeast have shown that Pol δ (yPol δ), rather than Pol ε, is the dominant polymerase involved in D-loop extension [9], [11], [22]. This is consistent with the observations that yeast Pol δ, but not Pol ε, exhibits the ability for strand displacement [23] which is involved in the process of D-loop extension [9].

In human cells, there are two physiologically important forms of Pol δ. Human DNA polymerase δ is a heterotetrameric enzyme (Pol δ4) composed of the catalytic subunit (p125), and three non-catalytic subunits, p50 and p68, and p12. These four subunits are encoded by the POLD1, POLD2, POLD3, and POLD4 genes, respectively. The p12 subunit plays an important role in the regulation of human Pol δ activity [1], [2], [3], [4]. Its transient degradation leaves behind an intact trimeric form of Pol δ (Pol δ3) consisting of the p125, p50 and p68 subunits. This occurs in response to UV, alkylating agents and replication stress under the control of ATR [24], as well during the cell cycle, where p12 is degraded on entry to S-phase and is restored during G2/M [25], [26]. Pol δ3 and Pol δ4 exhibit different functional capabilities: Pol δ3 is an intrinsically more accurate polymerase than Pol δ4. It is adapted for Okazaki fragment processing in concert with Fen1 [27], and for gap-filling in excision repair, while Pol δ4 is adapted for synthesis through more complex templates and is required for HDR [1], [2], [3], [4], [28].

The CryoEM structure of the Pol δ4/PCNA complex shows that the C-terminal region of p12 (residues 42-107) forms a cuboidal structure that is wedged between p125 and p50, and allows structural insights into the basis for the differential functional characteristics of Pol δ4 and Pol δ3 [29]. The N-terminal 41 residues are not resolved in the p12 structure, indicating conformational flexibility. Its orbit of movement makes it feasible for the N-terminally located PIP box of p12 (residues 4-11) [30] to interact with PCNA [29]. The roles of Pol δ4 are less well defined, but it has been established through studies of CRISPR/Cas9 knockout of the POLD4 gene in cultured cells that it is required for HDR [1], [28]. The p12-knockout cells were HDR-deficient as they exhibited a ca. 60% reduction in repair of I-SceI induced DSBs in the DR-GFP plasmid reporter assay, and were highly sensitized to DNA damaging agents and PARP inhibitors [1], [28].

Studies of human Pol δ4 in a plasmid based D-loop assay system have demonstrated that it is capable of D-loop extension [31], [32], but the Pol δ3 form has not been studied in relation to its ability to perform D-loop extension. The inference from the p12-KO studies is that Pol δ3 is unable or less able to perform D-loop extension. Pol δ3 has little or no strand displacement activity [27], and this is a plausible basis for predicting a lack of function in D-loop synthesis [27]. Also, strand displacement by Pol δ4 is much slower than synthesis on an unimpeded template [27]. We have suggested that the D-loop synthesis by Pol δ4 might be slow and be facilitated by helicases [1]. Studies of yeast Pol δ (yPol δ) using a plasmid based D-loop reconstitution assay [10], [33] demonstrated a requirement for Pif1, a 5’ to 3’ helicase, for efficient D-loop synthesis [22], [34]. Pif1 binds to PCNA and this interaction is critical for its recruitment and participation in BIR [10], [22], [33], [34]. Studies of human Pol δ4 in a plasmid based D-loop assay have demonstrated that it is capable of D-loop extension [31], [32], but thus far no biochemical evidence for helicase participation in this process in a human reconstitution system has been reported.

In the hypothesis we proposed for a role of helicases in facilitating D-loop synthesis we emphasized the importance of their abilities to interact with Pol δ4 and/or PCNA [1]. Such protein-protein interactions would drive the affinity of the helicase for recruitment to the leading edge of the D-loop, bearing in mind that helicases have a broad and complex repertoire of activities that could also act to disassemble the D-loop to terminate D-loop extension [35]. In this model (Fig. 1), recruitment of a helicase (shown here as DHX9) to the site of D-loop extension (rather than to a site that leads to D-loop dissolution) is driven by protein-protein interactions with Pol δ4 and PCNA. This ability would be essential for helicase recruitment to the appropriate site of action.

DHX9 was chosen as a candidate for a potential role in D-loop extension because it has a long history of involvement with mammalian Pol δ functions. DHX9, together with Pol δ and other replication proteins, were bound on immobilized PCNA columns [36], and was also associated with Pol δ purified by immunoaffinity chromatography and subsequent gel filtration (Supplementary Material, Table SI). DHX9 is a multifunctional RNA/DNA 3’ to 5’ helicase, originally identified as an RNA helicase (RNA Helicase A) [37], and then as Nuclear DNA Helicase II [38], [39]. It has been implicated in transcription, translation, microRNA biogenesis, RNA processing and transport as well as in maintaining genomic stability in DNA replication (reviewed in [40], [41]). Its substrate specificity overlaps with those of the Bloom and Werner helicases, and include complex DNA structures, D-loops, R-loops as well as triplex DNA and RNA or DNA G-quadruplexes. A role for DHX9 in the DDR (DNA Damage Response) of DSBs formed during transcription through the recruitment of BRCA1 was also reported [41].

We provide evidence that DHX9 has a direct involvement in the process of DSB repair through the stimulation of Pol δ4-mediated DNA synthesis in a reconstitution assay. We also show evidence for protein-protein interactions between DHX9 with Pol δ4 via the p125 and p12 subunits, and between DHX9 and PCNA.