Phosphorylation Alters the Properties of Pol η: Implications for Translesion Synthesis.

There are significant ambiguities regarding how DNA polymerase η is recruited to DNA lesion sites in stressed cells while avoiding normal replication forks in non-stressed cells. Even less is known about the mechanisms responsible for Pol η-induced mutations in cancer genomes. We show that there are two safeguards to prevent Pol η from adventitious participation in normal DNA replication. These include sequestration by a partner protein and low basal activity. Upon cellular UV irradiation, phosphorylation enables Pol η to be released from sequestration by PDIP38 and activates its polymerase function through increased affinity toward monoubiquitinated proliferating cell nuclear antigen (Ub-PCNA). Moreover, the high-affinity binding of phosphorylated Pol η to Ub-PCNA limits its subsequent displacement by Pol δ. Consequently, activated Pol η replicates DNA beyond the lesion site and potentially introduces clusters of mutations due to its low fidelity. This mechanism could account for the prevalence of Pol η signatures in cancer genome.

Introduction

In non-stressed cells, the exclusive participation of high-fidelity DNA polymerases (Pol δ and Pol ɛ) in routine DNA replication processes is critical for genomic stability. However, mechanisms that limit the access of low-fidelity DNA polymerases, such as polymerase η (Pol η), to the replication fork are not well understood or characterized. Both classes of polymerases share a common mechanism to access the DNA primer/template (P/T) terminus by the possession of a PCNA-interacting peptide (PIP) motif (Choe and Moldovan, 2017).

Upon exposure to stresses that cause DNA lesions, maintenance of the integrity of the replication fork becomes a major priority (Segurado and Tercero, 2009). Replicative polymerases are blocked by bulky lesions on the template strand, leading to stalling of the replication fork. Failure to resolve the stalled fork in a timely manner leads to fork collapse and likely cell death (Cortez, 2015, Roos and Kaina, 2006). Under these circumstances, the DNA damage tolerance pathway is activated, leading to the recruitment of translesion synthesis (TLS) Pols (Goodman and Woodgate, 2013, Waters et al., 2009). These Pols displace replicative polymerases at the lesion sites to replicate across the lesion and rescue the stalled fork. Pol η is the most intensively studied of TLS Pols, and it replicates across UV-induced lesions, especially cyclobutane pyrimidine dimers (CPDs) formed between neighboring pyrimidine bases in DNA, with high accuracy (Johnson et al., 1999, Masutani et al., 2000). It is not essential for routine DNA replication and Pol η-null mice display normal size, fertility, and life expectancy in the absence of external DNA damage (Lin et al., 2006). However, germline loss of Pol η function results in increased UV sensitivity and propensity for cancer in a subset of patients with xeroderma pigmentosum (XP) (Cordonnier and Fuchs, 1999), the variant form (XP-V).

The ability of Pol η to replicate across bulky lesions arises from a relatively more spacious active site (Silverstein et al., 2010, Zhao et al., 2012) when compared with those of Pol δ and Pol ɛ. However, this property also contributes to its reduced replication fidelity (Kunkel, 2003). Pol η is about 10,000 times (error rate of ∼1/100) more error-prone than Pol δ or Pol ɛ when replicating normal templates, typified by the signature mutations of A to G or C to T transitions under certain sequence contexts (Chen and Sugiyama, 2017, Matsuda et al., 2001). Both types of mutations have been observed in a number of recently identified cancer mutational signatures (Rogozin et al., 2018, Rubin and Green, 2009, Supek and Lehner, 2017). Thus, it is prohibitive for unregulated participation of Pol η in routine DNA replication, a process that demands an error rate of less than one in a billion.

The mechanisms that regulate the recruitment of Pol η to DNA damage sites initially seemed to be clearly defined. It was discovered that monoubiquitination on K164 of PCNA (Ub-PCNA) by RAD6/RAD18 is critical for UV resistance in Saccharomyces cerevisiae (Hoege et al., 2002). Two subsequent studies demonstrated that yeast Pol η activity is stimulated by Ub-PCNA through its possession of a ubiquitin-binding domain (UBZ) in its non-catalytic C-terminal sequence (Garg and Burgers, 2005, Watanabe et al., 2004). These studies painted a clear picture of recruitment and regulation: monoubiquitination of PCNA induced by DNA damage stress recruits Pol η through enhanced affinity to the damage site to perform TLS (Kannouche et al., 2004). However, in vitro reconstitution studies (Acharya et al., 2007, Acharya et al., 2008, Bertoletti et al., 2017, Jung et al., 2011, Ma et al., 2017, Sabbioneda et al., 2009) provided contradicting results and failed to recapture the elegance of the proposed model (Hedglin et al., 2016). Furthermore, subsequent cellular studies failed to resolve the ambiguities (Acharya et al., 2010, Kannouche et al., 2001).

Meantime, accumulating evidence suggests that post-translational modifications (Chen et al., 2008, Dai et al., 2016, Göhler et al., 2011, Jung et al., 2011) of Pol η and interactions with partner proteins play major roles in its regulation (Maga et al., 2013, Tissier et al., 2010). Among these, two studies demonstrated that phosphorylation of Pol η by ataxia telangiectasia and Rad3 related (ATR) (Göhler et al., 2011) and protein kinase C (PKC) (Chen et al., 2008) is critical for its function in intact cells.

We aimed to resolve some of the aforementioned inconsistencies through comparative studies on wild-type (WT) and phosphomimetic mutants (PM) of Pol η. Our data show that phosphorylation of Pol η greatly potentiates its activity when Ub-PCNA is present, which reveals a mechanism whereby phosphorylation of Pol η enhances its affinity for Ub-PCNA. This mechanism sheds new light not only on the control of TLS but also on the avoidance of Pol η participation in normal DNA replication.

On the protein interaction front, PDIP38 proved to be an intriguing partner. PDIP38 interacts with the C-terminal UBZ domain (Tissier et al., 2010) of Pol η and was proposed to facilitate the recruitment of Pol η to UV-induced DNA damage sites. In in vitro reconstitution studies, PDIP38 stimulated the activities of Pol η (Maga et al., 2013). However, there has been no evidence showing that PDIP38 localizes to DNA damage foci with TLS Pols. In fact, PDIP38 is translocated to the spliceosome upon UV exposure (Wong et al., 2013). Our studies show that PDIP38 acts as a key regulator in restricting Pol η participation in DNA replication in unstressed cells.

Results

Phosphorylation Activates Pol η

To address the effects of ubiquitination of PCNA and phosphorylation on Pol η activities, we performed in vitro experiments using recombinant Pol η. Human Pol η possesses a UBZ domain that readily unfolds when stripped of a required zinc ion by EDTA (Woodruff et al., 2010); thus variable loss of UBZ function could give rise to a range of variations in Pol η properties. We took specific steps to maintain the functional integrity of its UBZ domains. We generated recombinant Pol η tagged at the C-terminus with six histidine residues by overexpression in E. coli strain Rosetta 2 (DE3). The expression level and solubility were further optimized by co-expression with ubiquitin, which can interact with the UBZ domain and stabilize Pol η. Zinc-charged chelating resin was used for purification of Pol η, instead of the traditionally used Ni-charged resins, to avoid metal ion exchange and maintain proper UBZ folding. Detailed protocols are provided in Supplemental Information. All recombinant Pol η preparations were over 90% pure and monomeric upon characterization by gel filtration chromatography (Figure S1). A phosphomimetic mutant of Pol η was also expressed to investigate the effects of phosphorylation on its activity. This phosphomimetic mutant (Pol ηPM) bears glutamate residues at the aforementioned phosphorylation sites for PKC (S587, T617) (Chen et al., 2008) and ATR (S601) (Göhler et al., 2011).

To validate our enzyme production and purification procedure, we conducted single nucleotide incorporation assays using protocols adopted from reports in the literature (Guilliam et al., 2016, Maga et al., 2013, Washington et al., 2003). Steady-state kinetic analysis indicates that Pol η exhibits similar catalytic constants in terms of Km (0.15 μM for dATP opposite of dTTP) as reported in the literature, and the kcat value (0.22 s−1) is similar to that reported by Washington et al. (2003).

Upon validation of our purified enzymes, we sought to investigate the role of Pol η in two cellular processes: routine DNA replication and TLS. Pol η replicates native and CPD-lesioned templates with similar efficacy (Johnson et al., 2005, Masutani et al., 2000, Yagi et al., 2005), with the determining factor being the efficacy with which Pol η binds to the P/T terminus. Therefore, a single pair of primer and template without lesions was used in both assays for fair comparison. The substrate consisted of a 70mer template with a 34mer primer radiolabeled at the 5′ end by 32P (Figure 1A). The template was modified at both ends with biotin and blocked with streptavidin to allow quantitative pre-loading of PCNA (Figure 1B) or Ub-PCNA (Figure 1C) by human replication factor C (RFC) (Lin et al., 2013). Concentrations of RFC were chosen such that P/T > RFC > PCNA or Ub-PCNA to favor the complete loading of PCNA and Ub-PCNA to the P/T termini. Assay conditions were chosen to mimic the cellular ionic environment in terms of pH, ion type, and strength. Specifically, the assay buffer contained 150 mM KCl and 50 mM NaCl (the ionic environment of nuclei consists of 150–260 mM KCl and 40 to 50 mM NaCl [Dick, 1978, Paine et al., 1981]).

Figure 1

Phosphomimetic Pol η Exhibits Increased Activity

(A–C) (A) Schematic representation of 32P-labeled primer/template (P/T). P/T pre-loaded with PCNA (B) or Ub-PCNA (C). The template was modified by biotinylation on both ends and blocked with streptavidin.

(D) Representative sequencing gel showing resolved extension products at 1 and 2 min. The labeled primer and the full-length product are indicated as 34 and 70, respectively.

(E) Relative quantification of full-length 70mer products of Pol ηPM with Ub-PCNA normalized to Pol η with PCNA at 2 min. Data shown as mean ± SD of 3 individual experiments and differences between groups analyzed by unpaired t test, **p ≤ 0.01.

(F and G) Extension products of increasing concentrations of Pol η or Pol ηPM (10–100 nM) titrated against a fixed concentration of either PCNA-loaded P/T (+PCNA) or Ub-PCNA-loaded P/T (+Ub-PCNA) at 2 min. The 70mer extension products formed by Pol ηPM on Ub-PCNA-loaded P/T in (F) were analyzed and fitted to one-site binding kinetics (G).

The primer extension activities of Pol η and Pol ηPM in the absence or presence of either pre-loaded PCNA (Figure 1B) or pre-loaded Ub-PCNA (Figure 1C) were compared. The reactions were initiated by the addition of Pol η and terminated at 1 and 2 min. A representative autoradiogram of the primer extension products is shown in Figure 1D. The activities of both Pol η and Pol ηPM are stimulated by PCNA, and even more so by Ub-PCNA (Figure 1D). The results are consistent with the models in which the PIP-motifs in Pol η are functionally effective in stimulating Pol η activity in the presence of PCNA, and causing a greater increase in Pol η activity (Figure 1D, first 3 lanes) with the binding of the added ubiquitin on Ub-PCNA. However, Pol ηPM is much more active than Pol η, as can be seen by the production of the full-length products (70mer) (Figure 1D, last 3 lanes). These findings show for the first time that phosphorylation of Pol η is a mechanism that leads to a significant activation of Pol η. This is illustrated by comparison of the activity of Pol η on a PCNA-loaded P/T and Pol ηPM on an Ub-PCNA loaded P/T (Figure 1E). The Pol ηPM/Ub-PCNA combination forms ∼7× more 70mer products than the Pol η/PCNA.

To investigate the mechanisms leading to the activation of Pol η, a series of titration studies were performed in which the concentrations of Pol η or Pol ηPM were increased from 10 to 100 nM, whereas the substrate concentrations were kept constant (Figure 1F). A similar pattern of activity changes as found in the previous assay (Figure 1D) was observed. Pol η was more active with Ub-PCNA than PCNA, and in both cases, Pol ηPM was much more active than Pol η. A plot of 70mer product formation against Pol ηPM for substrate P/T/Ub-PCNA showed that Pol ηPM exhibits one-site binding kinetics with an apparent Kd of 12.4 ± 1.7 nM (Figure 1G). Analysis for Pol η/Ub-PCNA or Pol ηPM/PCNA combinations indicated that the plots did not reach the saturation part of the curve (indicating Kd values of greater than 100 nM, Figure S1). Taken together, the above-mentioned observations indicate that optimal DNA TLS by Pol η requires phosphorylation of the enzyme as well as monoubiquitination of PCNA.

Phosphomimetic Pol η Binds Tightly to Ub-PCNA

The titration studies presented above demonstrated that Pol ηPM has an enhanced affinity for Ub-PCNA pre-loaded onto a P/T terminus. This affinity could reflect the affinity of Pol ηPM for either Ub-PCNA per se or the P/T/Ub-PCNA complex. To distinguish between these possibilities, we directly assessed the association of Pol ηPM with Ub-PCNA in the absence of DNA. Various size exclusion chromatographic techniques are commonly used to characterize both dynamic and stable protein-protein interactions (Bai, 2015, Beeckmans, 1999). The PIP box-PCNA interactions are examples of the dynamic interactions that facilitate Okazaki fragment maturation process (Choe and Moldovan, 2017, Lin et al., 2013). Thus, a majority of PCNA interactions are transient in nature, with the affinities (Kd) between PIP boxes and PCNA ranging from 10−4 M to 10−7 M (Bruning and Shamoo, 2004, Pedley et al., 2014). On the other hand, subunits of a holoenzyme generally form stable complexes with dissociation half-lives of hours or longer, and dissociation rates are difficult to measure by quantitative methods. The classical size-exclusion technique is suited for the analysis of these stable complexes. Fast protein liquid chromatography gel filtration (Supplemental Information) was used to investigate the strength of interaction between Pol ηPM and Ub-PCNA. High-ionic-strength buffers (350 mM NaCl) were used to eliminate any potential non-specific interactions.

Upon establishing individual elution profiles of Pol ηPM and Ub-PCNA (Figures 2A–2D), we mixed Pol ηPM with Ub-PCNA in a molar ratio of 1:2 (6 μM and 12 μM, calculated as Ub-PCNA trimer, in a total volume of 100 μL) and applied the mixture to a Superdex 200 10/300 GL column (GE). The elution profiles as well as SDS-PAGE analysis (Figures 2E and 2F) revealed that Pol ηPM and Ub-PCNA formed a stable complex in a 1:1 molar ratio. Incidentally, the SDS gel for the column fractions (Figure 2F) demonstrated that practically all of the Pol ηPM is engaged in complex formation, confirming the integrity of the preparation with regard to the functionality of UBZ domains. The absolute requirement for an intact UBZ domain for this binding was demonstrated by the inclusion of EDTA (Figures S2A and S2B), which acts to remove the zinc required for the folding of UBZ domain (Woodruff et al., 2010). The disruption of ubiquitin-Pol ηPM interaction by EDTA was further demonstrated by pull-down assays with glutathione S-transferase (GST)-ubiquitin (Figures S2C and S2D).

Figure 2

Pol ηPM and Monoubiquitinated PCNA form a High-Affinity Complex

Elution profiles of Superdex 200 10/300 GL column are shown on the left, and the corresponding fractions analyzed by SDS-PAGE analysis and Coomassie blue staining are shown on the right.

(A–J) (A and B) Ub-PCNA, (C and D) Pol ηPM, (E and F) Pol ηPM and Ub-PCNA mixture, (G and H) Pol η and Ub-PCNA mixture, and (I and J) phosphomimetic mutant Pol η2E (S601E and T617E) and Ub-PCNA mixture.

These interactions depend on the phosphomimetic mutation of Pol η on the aforementioned ATR and PKC sites. Neither the WT enzyme nor a Pol η mutant (Pol η2E) with two of the residues mutated to glutamates (S587E and S601E) formed a tight complex with Ub-PCNA when assayed under identical conditions (Figures 2G–2J). However, the Pol η2E mutant shows a slight shift on gel filtration (Figure 2J) and may indicate enhanced yet dynamic interaction with Ub-PCNA. These studies demonstrate that Pol ηPM and Ub-PCNA form a complex in a 1:1 molar ratio with affinities reminiscent of those between subunits of holoenzymes, which have a dissociative half-life of hours or longer (Bai, 2015, Beeckmans, 1999).

Phosphorylated Pol η at the Primer/Template/Ub-PCNA Terminus Cannot be Readily Displaced by Pol δ

The preceding experiments show that Pol ηPM binds to Ub-PCNA with high affinity. Under physiological conditions, the binding of phosphorylated Pol η would occur in the presence of Pol δ as a competitor. To investigate the possible exchange of Pol η with Pol δ, we constructed a synthetic P/T with Ub-PCNA (or PCNA) covalently linked near the primer terminus by N-hydroxysuccinimide(NHS)-psoralen (Figures 3A and 3D; Supplemental Information and S3). The complex was immobilized to streptavidin beads through the biotin-modified template at the 5′ end. We titrated the binding capacity of both beads: P/T/Ub-PCNA beads bind Pol η (60–90 ng/μL) and Pol δ (170–200 ng/μL). The P/T/PCNA beads bound about 2.5× greater amounts of these proteins than the /P/T/Ub-PCNA beads.

Figure 3

Pol ηPM on P/T/Ub-PCNA Terminus Is Resistant to Displacement by Pol δ

(A) Schematic representation of Ub-PCNA cross-linked to a primer/template (P/T).

(B) Lanes 1–3 (triplicates) contain 0.5 μg each of Pol δ and Pol η incubated with ∼100 ng binding capacity of Ub-PCNA-loaded P/T beads for 2 hr at 4°C. The beads were then washed with a stringent protocol. Lanes 4–6 are control reactions with 0.5 μg of individual proteins incubated with Ub-PCNA-loaded P/T beads. Bound proteins were detected by western blotting with antibodies against Pol η and Pol δ (p68 subunit).

(C) Increasing amounts of Pol δ (0.5–2.5 μg) incubated with 0.5 μg of Pol ηPM (lanes 1–5). Lanes 6 and 7 are control lanes with each protein individually. Lanes 8 and 9 are input.

(D) Schematic representation of PCNA cross-linked to a P/T.

(E) Binding of a mixture of Pol δ and Pol η to P/T/PCNA beads was performed as in (B). The beads were washed 3× with PBS, and bound proteins detected by western blot.

The initial experiments utilized P/T/Ub-PCNA beads to mimic a stalled replication fork post-UV irradiation. A limiting amount of P/T/Ub-PCNA beads (1 μL, <100 ng capacity) was used to bind a mixture of Pol ηPM (500 ng) and Pol δ (500 ng) in addition to control samples where individual proteins were used (Figure 3B). The binding buffer contained 160 mM KCl and 50 mM NaCl to maintain ionic strength comparable to that of nucleus (25 mM HEPES, pH 7.5). In addition, zinc (5 μM) and citrate (10 mM) were added to maintain a constant zinc ion concentration. The zinc-citrate complex has an apparent dissociation constant of 1.17 × 10−12 M2 (Krezel and Maret, 2016), and therefore our additions produce conditions close to cellular soluble zinc ion concentrations (Krezel and Maret, 2006). After incubation at 4°C for 2 hr, unbound proteins were washed by binding buffer with a stringent washing protocol (5×, 10-min incubation each). Bound proteins were analyzed by western blot analysis to detect Pol η and Pol δ (p68 subunit). Regardless of the presence (Figure, 3B, lanes 1–3, triplicates) or absence (lane 6) of Pol δ, Pol ηPM bound to full capacity of the beads, whereas Pol δ could not be detected when Pol ηPM was present (lanes 1 to 3). Under identical conditions, Pol η alone showed minimal binding (Figure 3B, lane 5) and Pol δ alone bound to 20%–30% of beads capacity (Figure 3B, lane 4). To characterize the interactions further, we increased the concentration of Pol δ while keeping the concentration of Pol ηPM constant (Figure 3C). Regardless of how high the Pol δ concentration was, Pol ηPM still bound to the full capacity of the beads, whereas no detectible Pol δ was bound (Figure 3C, lanes 1–5). These observations indicate that ATR- and PKC-phosphorylated Pol η will bind P/T/Ub-PCNA terminus regardless of the presence of related replicative Pols.

The binding preference of P/T/PCNA beads (Figure 3D) was also examined with a milder wash method, as the stringent washing protocol reduced the signal beyond detection. The aim of these experiments was to evaluate the extent to which Pol η can interfere with routine DNA replication process by Pol δ. Under milder washing conditions, Pol δ bound to DNA to a similar extent in the presence (Figure 3E, lanes 1–3, triplicates) or absence of Pol η (lane 4). Pol η also displayed similar binding to P/T/PCNA in the presence (Figure 3E, lanes 1–3) or absence of Pol δ (lane 5). In essence, bindings of Pol δ or Pol η to P/T/PCNA terminus are dynamic processes, which is consistent with studies of polymerase transactions on PCNA (Hedglin et al., 2016). These data do pose an important question: how is Pol η kept away from participation in routine DNA replication process under cellular conditions? Our in vitro assays mentioned above have shown that replication termini loaded with PCNA apparently recruited Pol δ and Pol η with similar efficacies in the absence of DNA damage stress. There must be additional mechanisms that prevent Pol η from participating in routine DNA replication.

Phosphorylation Is Responsible for Pol η Persistence at the Damage Foci

Having demonstrated that Pol ηPM and Ub-PCNA form a tight complex, which does not readily dissociate, we sought to replicate these findings under cellular conditions. We attempted to generate stable cell lines with ectopic expression of GFP-Pol ηPM as well as WT and a phosphorylation-deficient mutant (all three residues to A, Pol η3A). Cells expressing the WT enzyme were stable as established in many laboratories (Bienko et al., 2010, Kannouche et al., 2001, Watanabe et al., 2004), as were Pol η3A-expressing cells. However, repeated attempts to establish a stable GFP-Pol ηPM expression cell line failed. We tried immortalized XPV cells (GM02359-hTERT and XP30RO), 293T cells, and MRC5 lung cells with either antibiotic selection or lentiviral infection. Initially, all three constructs were transfected with similar efficiency judging by the fluorescence signals. However, cells expressing Pol ηPM gradually died and no cells with fluorescent signals survived after 2 weeks. Thus, the overexpression of Pol ηPM is clearly toxic to the cells. Therefore, we compared the fluorescent signal and localization of Pol η and Pol ηPM on a transient transfection basis within 48 hr in both 293T and MRC5 cells. Examination under the microscope at 36 hr after transfection revealed that ectopic expression of GFP-Pol ηPM resulted in the formation of bright foci. Even though WT Pol η has been shown to form some relatively large foci, the Pol ηPM foci are more intense with sharper boundaries and are significantly more in number (Figures 4A and 4B, MRC5 cells). We analyzed approximately 60 GFP cells in a single continuous field. Pol ηPM -expressing cells formed 4 times as many foci per cell (52.5) as Pol η-expressing cells (13.7), and with more cells (56.5% versus 21.7%) with 20 or more foci.

Figure 4

Efficient Retention of Pol η on UV-Induced DNA Damage Foci Requires Phosphorylation

(A and B) MRC5-SV2 cells were transfected with GFP- Pol η (A) or GFP- Pol ηPM (B) and fixed with 4% paraformaldehyde 36 hr later and imaged at 66X.

(C–E) MRC5-SV2 cells were transfected with GFP- Pol η, GFP-Pol ηPM, or GFP- Pol η3A and 36 hr later exposed to 30 J/m2 of UV-C. Cells were allowed to recover for 5 hr before subsequent fixation by 4% paraformaldehyde and imaged.

(F–S) 293T cells transfected with the GFP-Pol η (F–J), GFP- Pol ηPM (K–O), and GFP- Pol η3A (P–S) were exposed to UV irradiation 36 hr later. After recovery for 5 hr, cells were extracted with 1%Triton X-100-containing buffer at room temperature (∼22°C) before fixation by methanol at −20°C. Subsequently cells were imaged at 66× magnification. GFP signal exposures were maintained constant across samples. The strength of individual foci fluorescence for panels F and K were mapped using the 3D surface plot plug-in of ImageJ (J and O). Merge indicates the stacking of the GFP (green) and the PCNA (red) fluorescence signals, GFP intensity adjusted to indicate localization.

Upon UV-C irradiation (30 J/m2), Pol η (Figure 4C) as well as Pol ηPM (Figure 4D) formed bright foci after fixation with formalin. Even Pol η3A showed some diffused foci (Figure 4E). Our in vitro studies with Pol ηPM suggest that the binding of phosphorylated Pol η to Ub-PCNA at the damage sites on chromatin would no longer be a dynamic process, i.e., phosphorylated Pol η-Ub-PCNA or Pol ηPM-Ub-PCNA foci will have limited dynamic exchange with non-chromatin bound Pol η. However, technical proof of this phenomenon is complicated by the fact that Pol η/PCNA foci could arise from many combinations. Phosphorylated Pol η (or Pol ηPM) also possess tendencies to be recruited by PCNA (as shown without UV stress, Figure 4B). In addition, Ub-PCNA can also recruit Pol η to a greater extent than PCNA regardless of the phosphorylation status (Figure 1D). Indeed, earlier studies using fluorescence quenching methods showed that there are mixed populations of Pol η at the foci as well (Sabbioneda et al., 2008).

We took a different approach to investigate the stability of the Pol ηPM and Ub-PCNA complex. We perforated the cytoplasmic and nuclear membranes with buffers containing 1% Triton X-100 (Kannouche et al., 2001) and high salt concentrations (150 mM KCl and 50 mM NaCl) before fixation with formalin. We hypothesized that proteins not bound to chromatin or bound to chromatin dynamically would be washed away at room temperature with excess buffer; hence only proteins stably bound to chromatin would be observed. Upon perforation, transfected cells were washed extensively at room temperature with shaking for 5 to 8 min. The effectiveness of the washing protocol is illustrated by the total disappearance of fluorescence signal from the nuclei of GFP-Pol η3A cells (Figure 4P). In either 293T or MRC5 cells, this wash protocol has minimal impact on the number of foci in cells expressing WT Pol η or Pol ηPM (Figures 4F and 4K). Some loss of fluorescence intensity, especially in WT-Pol η-expressing cells (Figure 4F, all figures are adjusted for equal exposure) was observed. The individual foci were mapped for Pol η and Pol ηPM green fluorescence with ImageJ (Figures 4J and 4O). The fluorescence signals from the foci of the whole-cell populations were further quantified by laser scanning cytometry (Figure S4). Taken together, these data suggest that transfected Pol ηPM is recruited to the replication fork by PCNA-loaded primer ends even in the absence of UV stress, which could account for observed cell death. Post UV exposure, phosphorylated Pol η binds to Ub-PCNA and forms a stable complex that does not readily dissociate.

PDIP38 Sequesters Pol η from Participation in Routine DNA Replication in Unstressed Cells

The preceding studies show the importance of excluding Pol η from DNA replication except when absolutely needed. The mechanism(s) that exclude Pol η from the DNA are not fully understood. One likely mechanism is the association of Pol η with PDID38, based on previous observations that PDIP38 binds Pol η (Maga et al., 2013, Tissier et al., 2010). However, PDIP38 is subsequently enriched in the spliceosomes instead of the DNA repair foci in response to UV stress (Wong et al., 2013). Therefore, we hypothesize that PDIP38 sequesters Pol η away from replication fork to prevent its participation in DNA replication in unstressed cells.

Our laboratory has observed that PDIP38 is phosphorylated post-UV irradiation (Figure S5) in an ATR-dependent manner. PDIP38 possesses two potential ATR phosphorylation sites at Ser147 and Ser150. Ectopic expression of Flag-tagged Wt PDIP38 and its phosphorylation-null mutant, PDIP38AA (S147A/S150A), followed by 2D gel electrophoresis (Figure 5A) showed that these two residues are targeted for UV-induced phosphorylation.

Figure 5

PDIP38 Sequesters Pol η from Normal Replication Fork

(A) 2D gel electrophoresis of nuclear extract of 293T cells transfected with Flag-tagged PDIP38 (full-length with C-terminal tag) and its phosphorylation-deficient mutant, PDIP38AA. 24 Hours after transfection, cells were exposed to 30 J/m2 of UV-C irradiation and allowed to recover for 5 hr. Nuclear fractions were prepared and subsequently resolved by 2D gel electrophoresis before being transferred to nitrocellulose. PDIP38 was identified by anti-Flag antibody. The pH range (5–8) of the isoelectric focusing strip range is indicated above.

(B) Effects of PDIP38 (200 nM) on Pol η activity in primer extension assays using P/T/PCNA (see Figure 1B). The extension products were analyzed on sequencing gel and quantified (70mer product, 2 min) (Figure S5A). Data shown as mean ± SD of 3 individual experiments, and differences between groups analyzed by unpaired t test. **p < 0.01.

(C) Effects of PDIP38PM (200 nM) on Pol ηPM in primer extension assays using P/T/Ub-PCNA (see Figure 1C). A representative extension gel is included in Figure S5B. The extension products were analyzed on sequencing gel and quantified (70mer product, 2 min). Data shown as mean ± SD of 3 individual experiments and differences between groups analyzed by unpaired t test. ns indicates not significantly different.

(D) Pull down of recombinant Pol η using GST-tagged PDIP38 and its phosphomimetic mutant, PDIP38PM. Pol η (500 ng) was pulled down with GST, GST-PDIP38, or GST-PDIP38PM. To demonstrate the requirement for an intact UBZ domain, the pull-down assays were performed in the presence of 2 mM EDTA. Ponceau staining of GST-tagged proteins as loading controls is shown. Pol η bound to beads was analyzed by western blot.

(E and F) Effects of PDIP38 on binding of Pol η to P/T/PCNA beads. Each condition had 0.5 μg Pol η incubated with 100 ng binding capacity of PCNA-loaded P/T beads in the presence of increasing concentration of PDIP38 (50 nM–500 nM) for 2 hr at 4°C. After subsequent washes with PBS, the beads were boiled and bound proteins detected by western blot against Pol η to detect the amount bound to PCNA cross-linked P/T. PCNA was also detected as a loading control. Data shown as mean ± SD of 3 individual experiments, and differences between groups analyzed by unpaired t test, **p < 0.01, *p < 0.05.

(G and H) Effects of PDIP38PM on binding of Pol ηPM to P/T/Ub-PCNA beads. Each condition had 0.5 μg Pol ηPM incubated with 100 ng binding capacity of Ub-PCNA-loaded P/T beads in the presence of increasing concentration of PDIP38PM (50–500 nM) for 2 hr at 4°C. After subsequent washes with PBS, the beads were boiled and bound proteins detected by western blot against Pol η to detect the amount bound to Ub-PCNA cross-linked P/T. Ub-PCNA was also detected as a loading control by using antibody against PCNA. Data shown as mean ± SD of 3 individual experiments, and differences between groups analyzed by unpaired t test, ns indicates not significantly different. Statistical comparison between presence of 50 nM PDIP38PM and 500 nM PDIP38PM is indicated.

In the next set of experiments, we investigated the impact of direct PDIP38-Pol η interaction on the enzymatic activities in conditions mimicking pre- and post-UV irradiation. To conduct reconstitution studies, we optimized the production of the recombinant PDIP38. Using standard protocols, PDIP38 expressed in E. coli does not fold well and is present mainly as insoluble inclusion bodies (Maga et al., 2013). PDIP38 purified from the soluble fraction did not behave consistently in our hands. Therefore, we optimized the expression system and discovered that osmotic stress (0.5 M NaCl and 1 mM betaine) (Oganesyan et al., 2007) provided the optimal conditions. Under these conditions, the recombinant PDIP38 was soluble and exhibited minimal degradation (Figure S1B). We purified PDIP38 and its phosphomimetic mutant, as well as the GST fusion variants of these proteins (Figure S1B). The phosphomimetic mutant (PDIP38PM) was constructed by mutation of serine 147 and 150 to glutamates within the ATR consensus sequence (Kim et al., 1999) SQRSQ.

The effects of PDIP38 and PDIP38PM on Pol η and Pol ηPM activities were examined using the P/T substrates loaded with PCNA or Ub-PCNA (see Figures 1B and 1C). Representative gels are shown in Figures S6A and S6B. When PCNA loaded P/T was used in the activity assays, which represented routine DNA replication, PDIP38 inhibited Pol η activity (Figure 5B), whereas PDIP38PM did not show any significant inhibition (Figure 5B). When Ub-PCNA loaded P/T was used, which mimics TLS post-UV stress, neither PDIP38 nor PDIP38PM showed any inhibition of Pol ηPM activities (Figure 5C). These results were consistent with GST-fusion protein pull-down assays (Figure 5D). GST-PDIP38 showed direct interaction with Pol η, whereas GST-PDIP38PM showed diminished interaction with Pol η (Figure 5D). Interestingly, the PDIP38-Pol η interaction depended on an intact UBZ domain, as the inclusion of 2 mM EDTA disrupted the binding (Figure 5D).

Inhibition of polymerase activity can arise from two scenarios: direct inhibition of catalysis or exclusion of Pol η from binding to the P/T/PCNA terminus. To distinguish between these possibilities, we determined the impact of increasing concentrations of PDIP38 on Pol η binding to P/T/PCNA terminus (Figure 5E). Synthetic chromatin (P/T/PCNA) was incubated with constant concentrations of Pol η and increasing concentrations of PDIP38 (50–500 nM). The amount of Pol η bound to chromatin was analyzed by western blot analysis (Figures 5E and 5F). Under these conditions, PDIP38 inhibited Pol η binding to the P/T/PCNA terminus in a concentration-dependent manner. By contrast, PDIP38PM had no impact on Pol ηPM binding to P/T/Ub-PCNA terminus (Figures 5G and 5H).

At the cellular level, the hypothesis predicts that, upon loss of sequestration by PDIP38, some Pol η will be recruited to the replication fork and form foci even in the absence of UV irradiation. To test this, we generated PDIP38-null A549 cells by CRISPR methods (Figure 6B). Parental and PDIP38-null cells were then transfected with GFP-Pol η for ectopic expression and visualization. PDIP38-null cells showed significantly more Pol η foci co-localized with PCNA, whereas the parental cells display more diffused green fluorescence, typical of Pol η distribution in non-stressed cells (Göhler et al., 2011, Kannouche et al., 2001, Kannouche et al., 2004) (Figure 6A).

Figure 6

Loss of PDIP38 Leads to Pol η Recruitment to Replication Foci without UV Stress

(A) Parental and PDIP38-null A549 cells were transfected with GFP-Pol η and imaged 36 hr later for localization of Pol η and PCNA. Merge channel shows the superimposition of the GFP (green) and PCNA (red) fluorescence signals.

(B) CRISPR knockout of PDIP38 in A549 cells. β-Actin as loading control is shown.

Collectively these studies show that PDIP38 interaction with Pol η prevents its participation in DNA replication process in the absence of DNA damage. Upon UV irradiation, both PDIP38 and Pol η are phosphorylated by ATR and the phosphorylated forms lose the ability to interact with each other.

Discussion

Mechanism for the Recruitment of Pol η to a Stalled Replication Fork

Our present studies permit a reappraisal of the mechanisms involved in the recruitment of Pol η to stalled replication forks. This is illustrated in Figure 7. In unstressed cells (Figure 7, I), DNA replication fidelity is the sole priority. Consequently, low-fidelity Pol η is sequestered by PDIP38 to prevent its participation in DNA replication. Post-UV exposure, phosphorylation of Pol η by ATR and PKC is necessary and sufficient for its recruitment to the lesion site with monoubiquitinated PCNA regardless of whether the terminus is released from Pol δ (Figure 7, II). The high affinity of phosphorylated Pol η for Ub-PCNA can overcome the binding of other molecules that occupy the terminus through the PCNA-PIP interactions. The prompt recruitment of Pol η relieves replication fork stalling and promotes cell survival.

Figure 7

Proposed Model of Pol η Regulation by Phosphorylation and Sequestration

(I) Before DNA damage stress, Pol η is sequestered by PDIP38 away from the replication fork to ensure chromosome duplication is conducted by high-fidelity polymerases. (II) Upon UV irradiation, phosphorylation by ATR and PKC endows Pol η with overpowering affinity toward P/T/Ub-PCNA terminus so that it has sole access to conduct TLS and to maximize cell survival. PDIP38 is shuttled to spliceosomes for alternative splicing. (III) The high affinity between phosphorylated Pol η and Ub-PCNA protects the P/T terminus post-UV stress at the price of mutation clusters.

There are multiple potential PIP motifs and Rev1-interacting motif in addition to UBZ domain at the C-terminus of Pol η (Haracska et al., 2001, Pozhidaeva et al., 2012). The PIP motif is required for its translesion function and is most likely the initial anchor for its recruitment to PCNA or Ub-PCNA (Goodman and Woodgate, 2013). Our study suggests that the C-terminal non-catalytic domain of Pol η could adopt different conformations upon phosphorylation. How the PIP motif, Rev1-interacting region, and the UBZ domain interact with each other should be explored further, although caution should be exercised while interpreting the experimental results. Alterations in one of the motifs could very likely alter the conformations of the other motifs.

Exclusion of Pol η from Routine DNA Replication Is as Vital to Cell Fate as Timely Recruitment upon Lesion Encounter

Research on translesion polymerases has focused on the mechanisms that recruit them to the lesion sites and provide subsequent efficacy and specificity in TLS. There is scant information on the exclusion mechanisms that prevent TLS polymerases from unregulated participation in routine DNA replication process, which could result in far direr organismal or cellular fates due to their low fidelity and processivity in replication.

Our studies reveal two avenues that lead to the restriction of Pol η from unwanted participation in routine DNA replication. The first comes from differential affinities of Pol η and its UV-activated form. Pol η has low activities in terms of being recruited to P/T/PCNA terminus until being phosphorylated by ATR and PKC under the regulation of the DNA damage tolerance pathway (Figure 7, II). It seems likely that activation mutations of Pol η, or deregulation of the control of its phosphorylation, could have significant impact on cell fate. Our studies revealed that ectopic expression of activation mutant of Pol η (Pol ηPM) leads to cell death.

The other safeguard involves sequestration by an interacting protein, PDIP38 (Figure 7, I). Our results indicate that a single mechanism regulates both safeguard pathways: ATR phosphorylation induced by UV irradiation (Figure 7). Upon UV irradiation, PDIP38 is phosphorylated by ATR and loses its affinity toward Pol η and hence the ability to sequester. On the other hand, ATR and PKC phosphorylation of Pol η increases its affinity toward Ub-PCNA to facilitate TLS (Figure 7, II). The sequestration mechanism demands that there should be adequate quantity of PDIP38 under cellular conditions to bind Pol η, although it is impossible to assess all cells. In U2OS cells, it is estimated that there are about 70,000 molecules of PDIP38 and less than 500 molecules of Pol η (Beck et al., 2011). However, caution should be taken as most of PDIP38 molecules are localized in the cytosol and mitochondria, although PDIP38 can be shuttled between the nucleus and cytosol (Hernandes et al., 2017).

Our proposed model suggests that loss of either safeguard could lead to a dire consequence: unregulated participation of low-fidelity Pol η in routine DNA replication processes. Indeed, ectopic expression of the phosphomimetic mutant, Pol ηPM, leads to cell death. In some respects, Pol ηPM-expressing cells have a similar phenotype as primary PDIP38-null fibroblasts in that they can only replicate a few generations before dying (Brown et al., 2014). The lethal phenotype of PDIP38-null mice could originate from multiple functions endowed by PDIP38. It is tempting to suggest that the phenotype of PDIP38-null mice results from Pol η participation in routine DNA replication process. Our study shows that PDIP38 knockout leads to Pol η loading to PCNA foci in the absence of UV irradiation. Consistent with our studies, Tissier et al. demonstrated that PDIP38 knockdown by small interfering RNA also results in increased Pol η foci formation (Tissier et al., 2010) in the absence of UV damage.

The proposed model suggests that overexpression of Pol ηPM or suppression of PDIP38 may lead to a mutator phenotype, although these studies are challenging due to the fragile status of the resulting cell lines. The availability of PCNAK164R cell lines (Arakawa et al., 2006, Hendel et al., 2011) used in antibody diversification studies might provide a means for additional investigation. The inability to form Ub-PCNA in these cells might provide resistance to the formation of stable complexes of Pol η or Pol ηPM at the replication form and increase the chances of survival.

Implications in Cancer Etiology

The stability of the complex formed between Ub-PCNA and phosphorylated Pol η at the lesion site has important implications in determining which enzyme finishes the post-translesion DNA synthesis on the lagging strand. It raises questions of when or whether the second switch, i.e., Pol δ displacing Pol η post-lesion bypass in the classical TLS model, really occurs. Pol ηPM in combination with Ub-PCNA can processively replicate a fragment the size of Okazaki fragment within minutes, as our results showed. Post-UV irradiation, phosphorylated Pol η resides at the foci for a significant duration and is resistant to triton extraction (∼10 min) before fixation, in our study and has also been previoulsy reported (Kannouche et al., 2001). That is more than enough time for phosphorylated Pol η to complete the replication the Okazaki fragment. It is interesting to note that in yeast, lack of TLS in S-phase results in DNA gaps in G2 and these are no larger than Okazaki fragments under low to moderate UV exposure (Daigaku et al., 2010). In addition, post-UV stress, both phosphorylated Pol η (Chen et al., 2008, Göhler et al., 2011) and monoubiquitinated PCNA exist for significant duration, with Ub-PCNA levels detectible up to 72 hr in MRC5 cells (Niimi et al., 2008). When and how, or if the TLS signals are terminated in S-phase is still to be determined. The fact that post-replication repair can be separated from S-phase (Daigaku et al., 2010) may suggest a more globally activated termination signal in late-S or G2. The high affinity between phosphorylated Pol η and Ub-PCNA can exert a protective role for each individual terminus until all TLS operations are near completion for a more synchronized termination. Our data suggest that phosphorylated Pol η can and likely does replicate rest of the Okazaki fragment post-lesion bypass and introduces mutation clusters in that fragment (Figure 7, III).

Pol η is evolved to counter the effect of the most ancient DNA damaging agent and carcinogen: sunlight. This enzyme is ubiquitously expressed and involved in resolving many other bulky lesions. In fact, it can be argued that any DNA damage or replication stress that activates the ATR pathway could lead to Pol η participation in gap filling, as ATR activation causes the phosphorylation of Pol η and monoubiquitination of PCNA. The low fidelity of Pol η, estimated at about 1/100 (Chen and Sugiyama, 2017, Johnson et al., 2000), can lead to clusters of mutations. Statistically, if Pol η fills a gap of 100 bases, there is a probability of more than 25% of two or more mutations within the hundred bases following probability mass function of binomial distribution.where n is the gap size, k is the number of mutations within n, and p is the fidelity of polymerase.

The odds of clustered mutations increase with gap size. In a gap of 150 bases, typical of an Okazaki fragment, the odds increase to about 45% of two or more mutations. Compromises in mismatch repair system could further exacerbate Pol η-induced mutation clusters. Thus, our data suggest a mechanism that could explain the prevalence of Pol η signatures in cancer mutations (Nik-Zainal et al., 2012, Supek and Lehner, 2017).

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.