Identification of Three Human POLH Germline Variants Defective in Complementing the UV- and Cisplatin-Sensitivity of POLH-Deficient Cells

DNA polymerase (pol) η is responsible for error-free translesion DNA synthesis (TLS) opposite ultraviolet light (UV)-induced cis-syn cyclobutane thymine dimers (CTDs) and cisplatin-induced intrastrand guanine crosslinks. POLH deficiency causes one form of the skin cancer-prone disease xeroderma pigmentosum variant (XPV) and cisplatin sensitivity, but the functional impacts of its germline variants remain unclear. We evaluated the functional properties of eight human POLH germline in silico-predicted deleterious missense variants, using biochemical and cell-based assays. In enzymatic assays, utilizing recombinant pol η (residues 1—432) proteins, the C34W, I147N, and R167Q variants showed 4- to 14-fold and 3- to 5-fold decreases in specificity constants (kcat/Km) for dATP insertion opposite the 3’-T and 5′-T of a CTD, respectively, compared to the wild-type, while the other variants displayed 2- to 4-fold increases. A CRISPR/Cas9-mediated POLH knockout increased the sensitivity of human embryonic kidney 293 cells to UV and cisplatin, which was fully reversed by ectopic expression of wild-type pol η, but not by that of an inactive (D115A/E116A) or either of two XPV-pathogenic (R93P and G263V) mutants. Ectopic expression of the C34W, I147N, and R167Q variants, unlike the other variants, did not rescue the UV- and cisplatin-sensitivity in POLH-knockout cells. Our results indicate that the C34W, I147N, and R167Q variants—substantially reduced in TLS activity—failed to rescue the UV- and cisplatin-sensitive phenotype of POLH-deficient cells, which also raises the possibility that such hypoactive germline POLH variants may increase the individual susceptibility to UV irradiation and cisplatin chemotherapy.

Inherited pol η deficiency in humans results in a genetic disease, xeroderma pigmentosum variant (XPV), characterized by an increased skin cancer risk and sunlight sensitivity [1]. Diverse pathogenic mutations in the POLH gene, including nonsense, frameshift, premature stop, and missense alterations, have been identified in XPV patients [1,[9][10][11]. XPV patients suffer from severe adverse effects following cisplatin chemotherapy [12]. XPV fibroblasts are sensitive to UV light and cisplatin but are corrected by transfection of Int. J. Mol. Sci. 2023, 24, 5198 2 of 13 POLH cDNA [1,13]. In this context, it is reasonable to postulate that human germline POLH variants can alter the TLS activity and thus modify the susceptibility to toxic effects of UV and cisplatin in genetically affected individuals.
To date, a total of~510 missense germline single nucleotide variants in the human POLH gene have been listed in the Ensembl variation database [14], but their functional effects remain uncertain. In silico tools, such as SIFT [15] and Polyphen-2 [16], have been developed to predict the functional effects of missense variants. However, these predictions are not accurate enough to substitute for experimental functional assays, as shown in our previous work on the three other human Y-family pols ι, κ, and REV1 [17][18][19][20]. Therefore, experimental approaches are required to assess the functional effects of unstudied variants to validate the dysfunctional ones.
In this study, we selected eight human germline missense POLH variants, positioned in polymerase core domains, and predicted in silico to be deleterious, and then investigated their functional effects using both biochemical and cell-based assays. First, we evaluated catalytic activities of the pol η variants by experiments with primer extension, steady-state kinetics of single nucleotide incorporation, and pol-DNA binding assays using recombinant pol η (1-432) proteins. Next, we confirmed rescue abilities of pol η variants for the UV-and cisplatin-sensitive phenotype of POLH-knockout (KO) cells, by cell-based complementation assays. Here we report that C34W, I147N, and R167Q pol η variants, with considerably diminished activity, could not rescue the POLH-KO cells, while the five other variants-with slightly elevated activity-rescued the cells. These findings are discussed in the context of understanding the potential functional consequences of catalytically altered pol η variants.

Selection of Human Germline POLH Gene Variants to Study
We chose eight human germline missense POLH variants (Table 1 and Figure 1) that are expected to alter enzyme function on the basis of their location in polymerase core (finger, palm, thumb, and PAD) domains, and deleterious or damaging predictions by SIFT [15] and/or Polyphen-2 [16] from the Ensembl variation database [14].

Effects of Eight POLH Variants on Catalytic Activity of Pol η
To assess the alterations in catalytic activity of eight pol η variants, we performed "standing-start" primer extensions with wild-type pol η (1-432) and variants, using 17mer/25-mer duplexes containing a TT or CTD at template position 18-19 from the 3′ end, with all four dNTPs. The C34W, I147N, and R167Q variants generated extension products across the TT or CTD to a substantially lesser extent than the wild-type, while the other five variants yielded slightly more products (Figure 2). These results coincide with the steady-state kinetic data ( Table 2). The C34W, I147N, and R167Q variants showed 4-to 14fold decreases opposite the 3′-T of the CTD and 3-to 5-fold decreases opposite the next 5′-T in kcat/Km (specificity constant, a measure of efficiency) for correct dATP insertion, compared to the wild-type, while the other five variants showed 2-to 4-fold increases in those values. A similar trend of results was observed with unmodified TT templates. The misinsertion frequencies (a measure of fidelity) of eight variants with incorrect dGTP were not very different from those of the wild-type. is drawn using PyMOL (http://www.pymol.org (accessed on 8 September 2020)). The finger, palm, thumb, and PAD domains are colored yellow, red, green, and blue, respectively. The eight variant residues are indicated in the upper schematic domain diagram and shown as purple spheres in the structure.

Effects of Eight POLH Variants on Catalytic Activity of Pol η
To assess the alterations in catalytic activity of eight pol η variants, we performed "standing-start" primer extensions with wild-type pol η (1-432) and variants, using 17-mer/25-mer duplexes containing a TT or CTD at template position 18-19 from the 3 end, with all four dNTPs. The C34W, I147N, and R167Q variants generated extension products across the TT or CTD to a substantially lesser extent than the wild-type, while the other five variants yielded slightly more products (Figure 2). These results coincide with the steady-state kinetic data ( Table 2). The C34W, I147N, and R167Q variants showed 4-to 14-fold decreases opposite the 3 -T of the CTD and 3-to 5-fold decreases opposite the next 5 -T in k cat /K m (specificity constant, a measure of efficiency) for correct dATP insertion, compared to the wild-type, while the other five variants showed 2-to 4-fold increases in those values. A similar trend of results was observed with unmodified TT templates. The misinsertion frequencies (a measure of fidelity) of eight variants with incorrect dGTP were not very different from those of the wild-type.    Table 2. Steady-state kinetic parameters for dATP incorporation opposite the 3 -and 5 -T of a TT or CTD by human wild-type pol η (1-432) and variants.

DNA Template
Template Base

Effects of Eight POLH Variants on DNA Substrate Binding of Pol η
To assess the changes in DNA substrate binding affinities of eight pol η variants, we performed fluorescence polarization experiments ( Table 3). The K d,DNA of each pol η for CTD-containing DNA was similar to that of unmodified DNA, indicating that a CTD placed at the primer-template junction does not affect the DNA-binding affinity of pol η. The K d,DNA values of eight variants were not very (≤2-fold) different from the wild-type, indicating that those variants did not considerably alter the DNA-binding affinity of pol η. Table 3. DNA binding affinities of human wild-type pol η (1-432) and variants.

Complementation of UV and Cisplatin Sensitivity of POLH-KO Cells by Wild-Type Pol η and
Mutants D115A/E116A, R93P, and G263V We developed POLH-KO cell-based complementation assays to evaluate the capability of each pol η variant to rescue the UV-and cisplatin-sensitive phenotype in POLH-KO cells. First, the POLH-KO HEK293 cell line was generated using a CRISPR/Cas9 system, and verified at the gene and protein level ( Figure 3A,B). Second, we confirmed the distinct phenotype of POLH-KO cells, i.e., the enhanced sensitivity to UV and cisplatin, compared to wild-type cells, by CCK8 cell viability assays ( Figure 3C), as similarly reported earlier with XPV fibroblasts [1,13,21]. This phenotype was readily discernible in our assay condition without caffeine. Third, to validate this assay, we confirmed that the ectopic expression of wild-type pol η can reverse the UV-and cisplatin-sensitivity of POLH-KO cells to the wild-type cell level but, in sharp contrast, that of a catalytically inactive D115A/E116A mutant [22], and two known XPV-pathogenic defective mutants R93P and G263V [9,10], could not reverse the sensitivity ( Figure 3D), indicating that the catalytically intact pol η is required for the resistance of cells to UV and cisplatin. These features were also clearly demonstrated by comparison of the relative IC 50 values of UV and cisplatin ( Figure 3E), which were useful as indicators of cell sensitivity to genotoxic agents in cell-based assays with POLI-KO HEK293 cells [18]. Ectopic expression of wild-type pol η, but not that of three defective mutants, reversed both IC 50 values of the POLH-KO cells treated with UV and cisplatin to the wild-type cell level. In this assay, the protein levels of ectopically expressed pol η were similar to the endogenous level of HEK293 cells ( Figure 3D, middle).

Capabilities of Eight POLH Variants to Rescue the UV-and Cisplatin-Sensitivity of POLH-KO Cells
We employed this POLH-KO cell complementation assay to evaluate the capability of each pol η variant to rescue the UV-and cisplatin-sensitivity of POLH-KO cells. The C34W, I147N, and R167Q variants did not rescue the UV-and cisplatin-sensitivity ( Figure 4A), with no significant improvements in either the IC 50 UV or cisplatin values ( Figure 4B). In contrast, the other five variants fully rescued the UV-and cisplatin-sensitivity of POLH-KO cells ( Figure 4A), with significant improvements in their relative IC 50 UV and cisplatin values, nearly to the wild-type cell level ( Figure 4B). Under this assay condition, the protein levels of ectopically expressed pol η were similar to the endogenous level of HEK293 cells ( Figure 4A  features were also clearly demonstrated by comparison of the relative IC50 values of UV and cisplatin ( Figure 3E), which were useful as indicators of cell sensitivity to genotoxic agents in cell-based assays with POLI-KO HEK293 cells [18]. Ectopic expression of wildtype pol η, but not that of three defective mutants, reversed both IC50 values of the POLH-KO cells treated with UV and cisplatin to the wild-type cell level. In this assay, the protein levels of ectopically expressed pol η were similar to the endogenous level of HEK293 cells ( Figure 3D, middle).   Figure 3D were normalized to wild-type cells. Data are shown as mean ± SEM from three independent experiments. *** p < 0.001 (ANOVA with Tukey's test). C34W, I147N, and R167Q variants did not rescue the UV-and cisplatin-sensitivity ( Figure  4A), with no significant improvements in either the IC50 UV or cisplatin values ( Figure  4B). In contrast, the other five variants fully rescued the UV-and cisplatin-sensitivity of POLH-KO cells ( Figure 4A), with significant improvements in their relative IC50 UV and cisplatin values, nearly to the wild-type cell level ( Figure 4B). Under this assay condition, the protein levels of ectopically expressed pol η were similar to the endogenous level of HEK293 cells (Figure 4A middle).   Figure 4A were normalized to wild-type cells. Data are shown as means ± SEM from three independent experiments. * p < 0.05, ** p < 0.01, *** p < 0.001 (ANOVA with Tukey's test).

Discussion
In this study, we evaluated the functional properties of eight germline in silicopredicted deleterious missense variants of human pol η, at the molecular and cellular level, and identified three hypoactive variants as defective in rescuing the UV-and cisplatinsensitivity of POLH-deficient cells. Together with enzymatic analyses, we employed cell-based assays to assess the complementation ability of each variant, based on the capability of the ectopically expressed POLH gene (but not the defective mutant genes) to rescue the UV-and cisplatin-sensitive phenotype of POLH-KO cells (Figure 3). Here we report that the C34W, I147N, and R167Q variants, which were respectively 74-, 35-, and 10-fold reduced in the overall TLS efficiency (i.e., k cat /K m for dATP insertion opposite 3 -T× that opposite 5 -T, Table 2) opposite a CTD, failed to rescue the POLH-KO cells from the enhanced sensitivity to UV and cisplatin (Figure 4), suggesting that at least such hypoactive (≥10-fold reduction in the overall TLS efficiency) pol η variants might not be sufficient to protect cells from UV and cisplatin. This is the first report, to our knowledge, to identify the functionally defective germline missense variants of human POLH gene in a "bottom-up" approach from the Ensembl variation database [14], without using the XPV patient samples, in which in silico-predicted deleterious variants are screened biochemically and then confirmed by cell-based assays. This strategy was also successfully applied to identify dysfunctional POLI variants in our previous work [18]. Among eight variants that are predicted to be deleterious by the SIFT and/or PolyPhen-2 algorithms, only three variants were found to be dysfunctional in our assays. Although not perfect, the PolyPhen-2 algorithm (based on both sequence and structural features) [16] appeared to show a higher percentage (60%) of correct prediction of deleterious variants than the sequence homology-based SIFT algorithm (38%) [15]. We also note that all eight variants had combined annotation dependent depletion (CADD) scores (ranging from 22 to 29) higher than a possible cutoff value of 20, indicating that they are predicted to be among the top 1% most deleterious variants in the human genome by the CADD algorithm, integrating many diverse functional annotations [23]. Such discordance between functional assay results and in silico prediction effects is also observed with the TP53 and BRCA1/2 missense variants [24,25]. These observations highlight the necessity of experimental assays to validate the possibly deleterious genetic variants predicted by in silico tools.
The eight studied POLH variants can be divided into two types, according to the rescue capability ( Figure 4). The first type is the functionally defective ones (C34W, I147N, and R167Q), which were incapable of rescuing POLH-KO cells, with substantial impairments in catalytic activity. Noticeably, the C34W variant caused a severe catalytic impairment. It is plausible that the substitution to a bulky hydrophobic Trp at the beginning of the β2-strand in the finger domain would perturb the conformation of the β2-strand, at the end of which Gln-38 stabilizes the nascent base pair by hydrogen bonding with the template base [7], and thus destabilize the pol η active site. Interestingly, the moderately hypoactive R167Q variant was also deprived of the rescue ability, implying a demand of a certain minimum activity of pol η for providing tolerance to UV and cisplatin in cells. However, there also exists the possibility that these hypoactive variants are still at least partially functional in certain tissue or cell types, because their actual outcomes likely depend on the stress conditions and the levels of pol η and other TLS polymerases, which may vary with tissue or cell type. The other five variants belong to the functionally competent type, which could fully rescue POLH-KO cells, albeit with slightly increased TLS activities. The I272T variant, observed in several melanomas [26], was revealed to be functionally competent. Interestingly, all these slightly hyperactive variants did not show "over-rescue" (i.e., greater than wild-type) effects, which was similarly observed with three pol ι variants that were slightly hyperactive against H 2 O 2 sensitivity [18]. This finding agrees with the earlier observation, that a large overexpression of pol η (~59-fold above the endogenous mRNA level of human fibroblasts) restores the UV cytotoxicity of XPV fibroblasts nearly to the range obtained with normal fibroblasts [27]. These observations suggest that the protective effect of pol η against UV and cisplatin, is likely saturated at endogenous levels of pol η in those cells. It might also be attributed to the finding that the actual functioning of pol η is tightly regulated by multiple post-translational mechanisms including phosphorylation, ubiquitination, and PCNA monoubiquitination in cells [28][29][30].
In conclusion, our results suggest that three human germline POLH variants may substantially impair the TLS activity of pol η and thus lead to deprivation of its protective function against UV and cisplatin stresses in cells, which might possibly serve as predisposing factors for individual susceptibility to UV radiation and cisplatin chemotherapy. Although not conclusive yet, a genetically hypoactive status of pol η might potentially increase a cancer risk in humans, in that heterozygous POLH-deficient mice show an increased incidence of UV-induced skin cancer [31]. Our POLH-KO cell-based functional assays seem to be fairly quick and easy, and thus would also be useful for initial screening of unstudied non-synonymous POLH variants, in advance of the biochemical assays that reveal the mechanistic details. The exact clinical implications of the human germline dysfunctional POLH variants remain unclear and further evaluation of in vivo outcomes of these and other undetermined POLH variants would allow a better understanding of the role of POLH variants in interindividual variability in cancer risks and platinum drug responses.

Enzyme Assays and Steady-State Kinetic Analysis
DNA polymerase reactions and steady-state kinetic analyses were performed as described previously [5]. The reactions contained 50 mM Tris-HCl (pH 7.5), 5 mM dithiothreitol, 100 µg mL −1 bovine serum albumin (w/v), 10% glycerol (v/v), 5 mM MgCl 2 , and 100 nM DNA substrates (i.e., 5 -32 P-labeled 17-mer (or 18-mer) primers annealed to 25-mer templates containing a TT or CTD). Reactions were started by the addition of dNTPs and MgCl 2 to preincubated polymerase/DNA mixtures and ended with six volumes of a solution of 20 mM EDTA in 95% formamide (v/v). For steady-state kinetic analysis, the primer-template was extended in the presence of 0.4−1 nM pol η, with increasing concentrations of individual dNTPs, for 10 min, where the maximal product formation was ≤20% of the substrate concentration. Products were separated by 8 M urea-16% PAGE and analyzed with a PMI system (Bio-Rad, Hercules, CA, USA), as described previously [5]. Graphs of the product formation rates versus dNTP concentration were fit to the Michaelis−Menten equation in GraphPad Prism 7.0 (GraphPad Software, San Diego, CA, USA), for the determination of k cat and K m values.

Fluorescence Polarization
The 13-FAM-mer/25-mer (2 nM) was incubated with varying concentrations of pol η. The binding reactions contained 50 mM HEPES-KOH (pH 7.5), 10 mM potassium acetate, 2 mM β-mercaptoethanol, 0.1 mg/mL −1 BSA, and 5 mM MgCl 2 . Fluorescence polarization was measured with a Synergy Neo plate reader (Biotek, Winooski, VT, USA), using 485 and 528 nm excitation and emission filters, respectively, and K d,DNA (equilibrium dissociation constant for DNA binding) values were estimated as described previously [19]. The fluorescence polarization data (as a function of enzyme concentration) were plotted to estimate K d,DNA by fitting to a quadratic equation: where P is the measured change in polarization (in units of millipolarization), P 0 is the initial polarization (DNA alone), P max is the maximum polarization, D t is the total DNA concentration, and E t is the total enzyme concentration, using the GraphPad Prism 7.0 software.

Cell Viability Assay
Cells were seeded at 1.0 × 10 4 cells/well on 96-well plates, cultured overnight, and exposed to UV radiation or cisplatin (for 48 h) at varying doses. For UV radiation, cells were resuspended in PBS buffer, exposed to UV (254 nm) using a CL-1000 crosslinker (UVP, Upland, CA, USA), and incubated with fresh medium for 24 h. After treatment, cell viability was measured using CCK-8 (CK04; Dojindo, Kumamoto, Japan) following the manufacturer's instructions.

Statistical Analysis
Statistical comparisons were performed using Student's t-test or one-way analysis of variance (ANOVA) with Tukey's multiple comparison test. p < 0.05 was considered statistically significant.

Conflicts of Interest:
The authors declare no conflict of interest.

Abbreviations
CTD cis-syn cyclobutane thymine dimer EV empty vector IC 50 concentration that induces 50% inhibition of cell viability KO knockout TLS translesion DNA synthesis UV ultraviolet light WT wild-type