Polymerase θ—what does it see, and why does it matter for cancer therapy?

Abstract

Polymerase theta (Pol θ) is a DNA repair factor that has drawn much recent interest as a target for cancer therapy, since its inhibition is well-tolerated in most cells but is lethal in cancers deficient in breast cancer-associated (BRCA) genes (“synthetic lethality”). Its normal biological functions, as well as how these functions change in BRCA-deficient cancers, are only recently becoming clear, however. We review here recent progress in our understanding of the cellular regulatory mechanisms at work in determining if Pol θ sees DNA damage. At the molecular scale Pol θ then must notably see and repair diverse classes of damage, including (at least) conventional double strand breaks (made by e.g. ionizing radiation), as well as several types of damage associated with replication stress. We speculate on the mechanisms that could explain this flexibility.

Graphical Abstract

Graphical Abstract.

Introduction

Polymerase theta (Pol θ, gene name POLQ) is present in most eukaryotes [1], including plants and animals, though interestingly not fungi. In mammals it is a 290 kD protein that can be roughly equally divided into an N-terminal superfamily 2 helicase (SF2)-like domain, a mostly disordered central domain (CenD), and a C-terminal DNA polymerase domain (PolD) (Fig. 1A). The CenD is not well-conserved, both in sequence and in size. For example, the CenD in Caenorhabditis elegans is one-third the size of the mammalian CenD.

Figure 1.

Mechanism for repair of conventional double strand breaks by Pol θ. (A) Pol θ is a three domain protein consisting of a helicase like domain ([1]-900) (teal, HelD), a poorly conserved and mostly disordered central domain (gray, CenD), followed by a A family polymerase domain (blue, PolD). (B) Steps in repair include (1) end pairing and microhomology (red nucleotides) alignment, (2) flap cleavage by Pol δ (pastel green) 3′>5′ exonuclease (exo), (3) Pol θ PolD (blue) mediated synthesis of 10–25 nucleotides, (4) processive DNA synthesis to fill remaining gap by Pol δ polymerase activity, together with PCNA (neon green), ending with (5) resolution of a strand-displaced 5′ flap by PCNA interacting Flap endonuclease 1 (Fen1) and either ligase 1 or ligase 3 (yellow green). Inset; a dimer of helicase like domains (light and dark teal subunits), with each domain monomer binding to different ssDNA molecules such that complementary sequence (red nucleotides, microhomology) at 3′ termini are juxtaposed (from 9BH9), [22].

The deployment of Pol θ under physiologic conditions must by tightly regulated to prevent genome instability. Conversely, these regulatory mechanisms likely go awry in genetic disease and cancer. This has been most effectively demonstrated in HR-deficient cancers, where the products of error-prone Pol θ-mediated repair are abundant [2]. Accordingly, Pol θ has also drawn considerable interest as a target for cancer therapy, since its inhibition is well-tolerated in normal cells but often lethal in BRCA-deficient cancers [3, 4]. However, the biological role of Pol θ in normal cells is still not fully understood, nor is it clear how this role changes during the progression of BRCA-deficient cancers.

We review here recent progress in identifying the mechanisms that regulate engagement of Pol θ, and how this differs (or is the same) for different types of DNA damage. We focus first on cell-context dependent mechanisms that impact Pol θ engagement, including those downstream of the cellular DNA damage response (DDR), as well as those dependent on cell cycle phase. We then address how Pol θ sees its diverse substrates at the molecular scale, first for two ended chromosomal double strand breaks generated by conventional means (e.g. by nucleases or ionizing radiation), then DNA damage associated with replication stress.

Conventional double strand breaks are a readily accessible experimental model. They can be introduced as a homogeneous type of damage at high efficiency at a specific site by meganucleases like Cas9, have well characterized Pol θ-dependent repair products, and has readily manipulated alternatives to Pol θ-dependent repair. Accordingly, there has been considerable progress addressing Pol θ function in recognition and repair of conventional double strand breaks (DSBs). By comparison, the damage associated with replication stress is more diverse (including gaps and both one ended and two ended DSBs), and the pathways that protect and resolve this damage less easily manipulated. Much of our discussion of Pol θ engagement of damage caused by replication stress is thus speculative, and derived in part from assumptions based on Pol θ activity at conventional DSBs.

The DNA damage response and cell cycle phase

A screen for factors required for ionizing radiation-dependent Pol θ foci identified several factors implicated in the general DDR—MDC1, TOPBP1, and 53BP1 [5]. Interestingly, damage dependent Pol θ foci can rely on factors required for repair of DSBs by homologous recombination (discussed also in section 2), including Rad51 [6] and genes important for loading Rad51 (BRCA1/2, BARD1) [5]. FANCD2 has also been implicated in Polθ recruitment, particularly in BRCA1/2 deficient cells [7], but the mechanistic basis for this recruitment remains to be to be resolved.

Essential roles for Pol θ in repair of replication associated damage that persists into M phase has been identified in several models (Xenopus extracts, Drosophila melanogaster, and mammalian cells) [5, 8–13]. M-phase Pol θ activity was notably identified as downstream of FANCD2 in Drosophila development [9]. The 9-1-1 complex (Rad9, Rad1, and Hus1) was also identified as important for Pol θ activity in mammals and C. elegans [8, 14], together with a factor that interacts with 9-1-1, RHINO, that also directly interacts with Pol θ [8]. M-phase specific activation of Pol θ activity in mammals was linked to phosphorylation of several sites within the Pol θ CenD by the M-phase active kinase Plk1 [5], which promotes interaction between Pol θ and TOPB1 (TOPB1 in turn also interacts with the 9-1-1 complex). Plk1, 9-1-1, and RHINO’s role in activating M phase specific activity of Pol θ were identified as essential for cellular survival when cells were deficient in BRCA genes [5, 8].

In contrast, Pol θ-dependent repair activity is low to nonexistent in G1-arrested cells [15], at least for conventional DSBs. Down-regulation of Pol θ-dependent repair can be attributed to multiple factors: G1-phase-specific reduction in the extent conventional DSBs is resected (see section 2, below), reduced levels of Pol θ mRNA [15] (see also [16]), and anaphase promoting complex mediated destruction of a co-factor for Pol θ repair, RHINO, as cells exit M phase [8].

Pol θ sees conventional double strand breaks

Pol θ is essential for an end joining pathway (TMEJ) that repairs conventional two ended DSBs (made by e.g. ionizing radiation or nucleases). In mammals, this pathway is nearly equivalent to previously defined alternative end joining (a-EJ) and microhomology-mediated end joining (MMEJ) pathways [2]. A critical step in DSB repair by TMEJ is the nucleolytic resection of both DSB ends to generate 3′ terminated single-stranded overhangs, which can range from 10 to 1000 s of nucleotides long (Fig. 1B).

End resection is sufficient to block DSB repair by one of the other DSB repair pathways, NHEJ [17]. Early models implied TMEJ must still compete for repair of resected ends with DSB repair by homologous recombination. As introduced above, though, HR effectors BRCA1 and BRCA2 promote recruitment of Pol θ to cellular DNA damage [5], and Pol θ foci that co-localize with the HR recombinase, Rad51, relies on direct interactions between these two factors [6]. These results argue HR is an enabler of TMEJ instead of a competitor, likely reflecting an important role for Pol θ in resolving repair of HR intermediates that can’t be resolved. It remains to be determined which “stuck” HR intermediates recruit Pol θ, as well as the mechanism by which these intermediates are redirected to repair by TMEJ.

TMEJ must recognize and pair two 3′ single stranded DNA (ssDNA) termini, then dynamically align the two ends to identify and anneal 2–6 nucleotide segments of complementary sequence (microhomology search). The strongest candidate for performing these early steps is the N-terminal helicase-like domain of Pol θ. To begin with, a role for the helicase-like domain in removing Replication protein A (RPA) from resected ends has been demonstrated, both in vitro and in cells [18]. A direct role for the helicase-like domain in then pairing the two RPA-stripped ends in vitro was also recently demonstrated using single molecule methods [19]. Importantly, the presence of microhomologies wasn’t essential for helicase-like domain mediated end pairing, but when present, end pairing was both more stable and less dynamic.

How could the helicase-like domain promote end pairing? An early study solved the structure of the helicase-like domain alone and showed it could form a tetramer [20]. They then compared it to a related helicase, Hel308, bound to DNA. Like other superfamily 2 members, Hel308 forms a ring to wrap around ssDNA. When DNA bound Hel308 was used to model DNA in the Pol θ Helicase-like domain oligomer, two of the helicase-like domain monomers were arranged such that when the 3′ ends exited each of the ring’s central channel, they were well positioned to promote end pairing and microhomology annealing. Additional support for this model was provided by cryo-electron microscopy (cryo-EM) (Fig. 1B, inset) [21, 22]. Interestingly, the cryo-EM studies show that subdomain motions associated with binding of single-stranded DNA are incompatible with tetramerization, so that when two ends are bound, only dimers are observed. Whether either oligomeric state is biologically important remains to be seen. However, its notable that the other end joining pathway, NHEJ, similarly uses two “rings” to pair DSB ends [23].

Efficient Pol θ polymerase activity requires at least a few terminal primer-template base pairs (>2 bp microhomology). However, such microhomologies will typically be embedded after alignment by chance, leaving nonhomologous tails that must be removed by a nuclease before the Pol θ PolD can initiate synthesis [24]. TMEJ must also typically join 3′ ssDNA tails that are 100 to 1000 s of nucleotides long, and TMEJ isn’t sufficiently processive to fill the whole gap [19, 25]. Both missing activities are provided by Pol δ: this polymerases’ 3′>5′ proofreading exonuclease activity removes nonhomologous tails, and its proliferating cell nuclear antigen (PCNA) clamp-supported processive synthesis activity fills in the rest of the gap (Fig. 1B) [25]. A role for PCNA in TMEJ is also consistent with early studies suggesting that the resolution steps rely on PCNA-interacting Fen1 and Ligase I/III [2], the same machinery required for Okazaki fragment maturation.

Repair is thus altogether accomplished by alternating engagements of Pol θ and Pol δ. Accordingly, the successive engagements of these two polymerases, and coupling of pathway steps, rely in part on direct protein–protein interactions between Pol θ and Pol δ [25]. However, these two polymerases are appropriately not constitutively associated with each other. Instead, there is an additional DDR-dependent component to Pol θ–Pol δ association [25], though the mechanistic basis for this DDR-dependent component is unclear.

Pol θ sees replication stress associated damage

Pol θ also has a role in resolving replication stress from a variety of sources, including attempted passage of forks through template single strand breaks [26], interstand crosslinks [6, 27, 28], DNA–protein crosslinks [10], and at risk-DNA sequences (G4 DNA and fragile sites) [11, 13, 14, 29]. Pol θ activity in these contexts is typically most evident in cells deficient in other mechanisms that help mitigate replication stress. This includes deficiency in ATR [26], BRCA genes [11, 30–32], FANCJ [14, 29, 33, 34], translesion polymerases [35], or after inhibition of poly(ADP)ribose polymerase (PARPi) [11].

The replication stress-associated damage that’s seen and repaired by Pol θ likely includes at least two ended DSBs and lagging strand gaps. Pol θ engages two-ended DSBs after converging forks break [6, 10] (Fig. 2, left side), and likely can additionally engage the two ended DSBs generated after a replication fork transits single strand breaks in a lagging strand template (Fig. 2, center) [36–38]. Alternatively, Pol θ helps resolve gaps [30–32] or breaks [14] in the lagging strand after synthesis by Pol δ or Pol α is blocked (Fig. 2, right side), or after inhibition of PARP, a sensor of incomplete lagging strand synthesis [11]. Resolution of lagging strand gaps and breaks has primarily been associated with S-phase-specific Pol θ activity [30–32, 35], and shown to be especially important in BRCA-deficient contexts [30–32].

Figure 2.

Model for repair of lagging strand damage associated with replication stress by Pol θ. Lagging strand synthesis by PCNA (neon green) and Pol δ (pastel green) is arrested by lagging strand template blocks (red hexagon, left side) (step 1A) or breaks (red hexagon, right side) (step 1B). The DDR and direct protein interactions promotes recruitment of Pol θ to Pol δ+PCNA, as well as loading of the Rad9–Rad1–Hus1 (9-1-1) complex at downstream Okazaki fragments (step 2). The primer, bound by one subunit (light teal ring) of a Pol θ helicase like domain (HelD) dimer, is mobilized to downstream template near an Okazaki fragment by tethering of Pol θ to 9-1-1 through RHINO and TOPBP1 (step 3). The primer is then annealed to a downstream microhomology in template DNA (red nucleotides), as mediated by the other helicase-like domain in the dimer (dark teal ring), followed by initiation of synthesis by the PolD. The Okazaki fragment is then resolved by Fen1 and LIG1/3 (yellow) (step 4), resulting in a microhomology mediated deletions of 10–100 s of nucleotides in the daughter strand.

These two classes of substrate—two ended DSBs associated with replication fork collapse, and especially lagging strand gaps—differ significantly from what Pol θ must see when it repairs “conventional” two-ended DSBs. Here we speculate on mechanisms that enable Pol θ to also see sites of replication stress associated damage.

We suggest the seeming preference for Pol θ in rescue of blocked or broken lagging strand templates (versus blocked leading strand synthesis) may rely in part on direct and DDR stimulated interactions between Pol θ and the lagging strand polymerase, Pol δ [25]. In accord with this idea, Pol θ and Pol δ both associate near sites of nascent DNA synthesis when Pol α and Pol δ’s activity were blocked by aphidicolin treatment [31]. Its already been suggested Pol θ might bypass such a blocked lagging strand primer by reinitiating synthesis with the same primer, except after the primer is mobilized to a microhomology site downstream (“microhomology-mediated gap sealing”) [30]. There is abundant precedent that Pol θ can do this: an important signature of Pol θ activity during DSB repair are templated insertions, which reflect successive initiations of Pol θ-dependent synthesis from the same primer, but at different template sites [39, 40]. Such a mechanism is a notable contrast to PRIMPOL-mediated rescue of blocked leading strand synthesis, where blocks are bypassed by reinitiation of synthesis from a wholly new primer located downstream of the block [41].

Pol δ, PCNA, and a “stuck” lagging strand primer terminus could thus define one end of a Pol θ substrate: we suggest there’s a need for a second end, by analogy to Pol θ’s role in repair of conventional two-ended DSBs. A strong candidate for the second “end” is lagging strand template DNA near the 5′ terminus of a downstream Okazaki fragment (Fig. 2). Okazaki fragments, together with the Rad9–Rad1–Hus1 complex (9-1-1), have already been implicated in helping define the downstream deletion boundaries for Pol θ-dependent rescue of lagging strand blocks in C. elegans [14]. We suggest this reflects loading of the 9-1-1 complex at the 5′ recessed ends of Okazaki fragments, and tethering of Pol θ to the 9-1-1 complex through Pol θ interactions with RHINO and TOPBP1, both of which also associate with 9-1-1 [5, 8]. In accord with the idea that TOPBP1 links Pol θ to Rad9 (and a downstream Okazaki fragment), the TOPBP1-mediated interactions noted above are mediated through different domains: its N-terminal region interacts with Rad9 [42], while its C-terminal region interacts with Pol θ [5].

We propose the annealing of the mobilized primer and downstream microhomology may again be mediated by a dimer of Pol θ helicase-like domain “rings” (Fig. 2, see also inset to Fig. 1B). Half of a Pol θ dimer may interact with PCNA and Pol δ at the stuck primer, ultimately “grabbing” the primer with its helicase-like domain ring. The other half of the helicase-like domain dimer could load on lagging strand template near the downstream Okazaki fragment, as dictated by DDR and cell cycle phase specific mechanisms that induce tethering of Pol θ to 9-1-1 (discussed above). Here, loading of the Pol θ helicase-like domain near the downstream Okazaki fragment would require its ring to open, then close again, since there’s no free 3′ ssDNA terminus. The helicase-like domain dimer could then juxtapose the mobilized primer with a downstream template microhomology (Fig. 2, red nucleotides) in much the same way microhomologies are annealed during repair of conventional DSBs. For both lagging strand replication blocks and conventional DSB repair, Pol θ helicase like domain promoted annealing of the microhomology is then coupled to microhomology primed synthesis by the Pol θ PolD.

We’ve proposed a model where Pol θ overcomes obstacles to lagging strand gaps and breaks through a primer mobilization mechanism, and that its mechanism has interesting parallels to Pol θ’s end-pairing function in repair of conventional two-ended DSBs. As noted above, it takes into account available data on deletion patterns and Pol θ interactions. It also addresses two issues specific to Pol θ activity associated with sites of replication stress. It allows bypass of template strand blocks that can’t be bypassed by classical translesion synthesis, like G4 structures, template strand breaks, and DNA–protein and interstrand DNA crosslinks. Additionally, it allows Pol θ to resolve lagging strand gaps and breaks with a relative short synthesis tract (10 s of nucleotides), consistent with its relatively low ability to synthesize long stretches of DNA in a single binding event (processivity) [19, 25]. If (or when) any of this model is relevant remains to be seen. If it is relevant during resolution of gaps, the fate of the looped-out DNA also isn’t clear. If small enough it could be resolved by mismatch repair, or it could persist into the next S phase, with only one of the daughter cells manifesting the deletion.

Summary, future directions, and relevance to cancer therapy

Together, recent findings hint at complex mechanisms at work to ensure Pol θ appropriately navigates different DNA repair challenges. It must integrate diverse cell cycle cues and DNA damage contexts. At the structural level Pol θ’s biological role is likely defined by its modular nature and its ability to coordinate activities of its helicase like domain and its PolD. We propose that when coupled together, the two domains are uniquely suited to promote primer mobilization and synthesis across barriers to synthesis that are remarkably diverse and often impassable by other mechanisms. These activities are further controlled through DDR and cell cycle phase regulated interactions with partner proteins, beginning with (but likely not ending with) 9-1-1, RHINO, TOPBP1, and Pol δ.

We anticipate the many unanswered mechanistic questions raised above and elsewhere will require a far more comprehensive understanding of Pol θ structure, especially the full-length protein, as well as how Pol θ interacts with its both its cofactors and apparently diverse biological substrates. This in turn will help us identify and experimentally apply Pol θ variants that cleanly ablate different damage recognition mechanisms, especially HelD oligomerization, HelD ring opening/DNA binding, and DDR-dependent interactions. Experiments addressing how Pol θ sees the diverse damage types associated with replication stress are an especially difficult challenge, but will benefit from recently developed experimental models using defined damage (e.g. [6, 13, 36, 38]).

We need a much better understanding of what Pol θ sees for the safe, effective deployment of Pol θ inhibitors in cancer therapy. To begin with, much of the rationale for use of Pol θ inhibitors in HR deficient cancers requires the assumption that there isn’t much danger to normal, HR-proficient cells. However, there’s abundant evidence that this danger is not zero: Pol θ does have a role in normal cells, as first made clear by identification of Pol θ in a screen for suppressors of spontaneous micronucleus formation [43]. A more comprehensive exploration of the role of Pol θ in HR proficient cells, both what it sees, and what happens when it doesn’t see it, is essential to assessing therapy risk.

A clearer understanding of the damage Pol θ sees, and how it sees it, will also help in designing combinations of Pol θ inhibitor and other therapies that are most effective. For example, there is a complicated relationship between what Pol θ sees, and what another therapeutic target, PARP, sees. PARP acts in part upstream [3, 4, 44], and in part in parallel pathway(s) [11, 30, 31], relative to Pol θ activity, and this seems to depend on the triggering damage (conventional DSBs for the former relationship, unfinished replication for the latter). Loss of antagonists of resection, like 53BP1, can also promote resistance of cancers to PARP inhibition but sensitize them to inhibitors of Pol θ [45, 46]. The comprehensive untangling of the relationship between PARP and Pol θ, in both normal and cancer-specific contexts, will be essential to identifying when there’s a benefit to combined therapy. Finally, the emerging evidence that Pol θ is especially important when cells enter M-phase with unfinished replication suggests cancers with disrupted cell cycle checkpoint control (either pharmacologically induced or associated with oncogenesis) may be better targets for Pol θ inhibition.

There are inhibitors for both the polymerase and helicase-like domain catalytic activities [46–50], and preclinical models indicating both can be effective [46, 50]. As we learn more about how Pol θ functions during its many roles there may be reason to inhibit different domains in different contexts. Ultimately, convergence of mechanistic studies with translational research using clinical-grade Pol θ inhibitors will be required to fully realize the therapeutic potential of targeting this pathway in cancer.

Conflict of interest

None declared.

Funding

D.A.R. and G.P.G. were supported by P01 CA247773 from the National Cancer Institute.

Data availability

No new data were generated or analyzed in support of this research.