POLD1 variants leading to reduced polymerase activity can cause hearing loss without syndromic features

Abstract

DNA polymerase δ, whose catalytic subunit is encoded by POLD1, is responsible for synthesizing the lagging strand of DNA. Single heterozygous POLD1 mutations in domains with polymerase and exonuclease activities have been reported to cause syndromic deafness as a part of multisystem metabolic disorder or predisposition to cancer. However, the phenotypes of diverse combinations of POLD1 genotypes have not been elucidated in humans. We found that five members of a multiplex family segregating autosomal recessive nonsyndromic sensorineural hearing loss (NS-SNHL) have revealed novel compound heterozygous POLD1 variants (p.Gly1100Arg and a presumptive null function variant, p.Ser197Hisfs*54). The recombinant p.Gly1100Arg polymerase δ showed a reduced polymerase activity by 30–40%, but exhibited normal exonuclease activity. The polymerase activity in cell extracts from the affected subject carrying the two POLD1 mutant alleles was about 33% of normal controls. We suggest that significantly decreased polymerase δ activity, but not a complete absence, with normal exonuclease activity could lead to NS-SNHL.

The correct functioning of the eukaryotic genome requires very efficient and accurate DNA replication and repair. In particular, DNA polymerase delta (Pol δ) is indispensable for high-fidelity chromosomal DNA replication (Chen et al., 2000; Fukui et al., 2004; Garg & Burgers, 2005). The Pol δ holoenzyme has four subunits: the catalytic subunit p125 (Pol3), p55 (Pol31), p40 (Pol32), and p18 (Cdm1) (Liu, Mo, Rodriguez-Belmonte, & Lee, 2000; MacNeill, Baldacci, Burgers, & Hubscher, 2001). The largest catalytic subunit (p125), which is encoded by POLD1 in humans and is highly conserved among eukaryotes, has several domains: polymerase and exonuclease active sites; protein–protein interaction sites, such as C-terminal zinc fingers; and a binding motif for processivity clamp proliferating cell nuclear antigen (PCNA) (Garg & Burgers, 2005) which is crucial for the high processivity of Pol δ (Burgers, 1989). Initially, the functions of each domain in the catalytic subunit of Pol δ were investigated in mouse mutants. Abolishing the 3'–5' exonuclease proofreading activity of either Pol δ or Pol ε led to a strong mutator phenotype with increased tumorigenesis in mouse mutants, despite otherwise normal development (Goldsby et al., 2002; Morrison, Johnson, Johnston, & Sugino, 1993). Mice carrying a critical mutation (p.Leu604Lys) in the polymerase active site of Pold1 showed embryonic lethality in homozygotes and genomic instability in heterozygotes (Venkatesan et al., 2007). Uchimura, Hidaka, Hirabayashi, Hirabayashi, and Yagi (2009) also proved that Pold1 expression is essential for early mammalian embryogenesis and the 3'–5' exonuclease activity from Pold1 is necessary for tumor suppression in mice (Uchimura et al., 2009). However, it is likely that the results of studies of Pold1 mutations using mice and in vitro biochemical assays cannot be fully extrapolated to human phenotypes associated with alterations of POLD1. Recently, important and interesting disease phenotypes associated with human POLD1 mutations have been reported. As predicted from mouse studies (Goldsby et al., 2001), single heterozygous missense mutations in the exonuclease domains led to a predisposition to colorectal (MIM# 612591) and endometrial cancer (Palles et al., 2013; Valle et al., 2014). In contrast, a single heterozygous p.Ser605del mutation that exclusively abolishes DNA polymerase activity, but largely spares the 3'–5' exonuclease activity, resulted unexpectedly in a multisystem disorder characterized by mandibular hypoplasia, deafness, and progeroid feature (MDPL syndrome, MIM# 615381), as well as progressive loss of subcutaneous adipose tissue with severe insulin resistance (Weedon et al., 2013). Data from mice and humans suggested that dysregulated coordination of exonuclease and polymerase activity would lead to a diverse phenotypic spectrum. However, the phenotypic outcomes from combinations of POLD1 genotypes, such as a combination of two mutant alleles of POLD1 with varying degrees of exonuclease and polymerase activity, have never been reported in humans. In this study, for the first time, we report a phenotype resulting from two recessive mutant alleles of POLD1 and extended the phenotypic spectrum of POLD1 mutations to include nonsyndromic sensorineural hearing loss (NS-SNHL).

We performed exome sequencing (ES) for five members of family SB127: two affected siblings (SB127–277 & 219) who manifested nonprogressive moderate degree SNHL, two unaffected siblings (SB127–235 and 247), and their unaffected mother (SB127–497; Figure 1). The mean depth of coverage of the target regions in the ES data from the five members was 98.86×, with 95.9% over 10× (Table S1). Among 2,052,288 single nucleotide polymorphisms (SNPs), candidate variants were narrowed down based on the stepwise filtering previously validated to be effective for a sporadic or autosomal recessive moderate degree of SNHL in the Korean population (Kim et al., 2015;Table S2).Afterthe Phase I filtering process based on the dbSNP database, 791 variants were detected as candidates. We ruled out all the variants except two variants, c.584_585del:p.Ser197Hisfs*54 and c.3298G>A:p.Gly1100Arg in POLD1 (NM_001256849.1), based on autosomal recessive inheritance patterns. Thesevariantswerepredictedtobelikely pathogenic, according to the guideline for the interpretation of classifying pathogenic variants (p.Ser197Hisfs*54; PVS1_Strong, PM2, PP1 and p.Gly1100Arg; PS4_Supporting, PM2, PP1_Moderate, high CADD score of 24.1, https://cadd.gs.washington.edu/info) (Oza et al., 2018). Note that we excluded the possibility of any de novo variants having occurred in this family, because there are two affected siblings from unaffected parents (Figure 1). The compound heterozygote variants were confirmed by Sanger sequencing (Figure 1g,h). We also confirmed that these two variants were not detected in 2,120 control chromosomes from Korean control subjects with normal hearing. The two variant residues were well conserved among orthologs from diverse species, except yeast (Figure 1i). Specifically, the p.Gly1100 residue was physically close to the CysA, CysB, and C4 type zinc-finger motifs (Figure 1j). The possibility of the presence of significant copy number variations (CNVs) in this family was also explored from the ES data, using the EXCAVATOR tool. Several CNVs involving target regions were noted from each subject (Table S3); however, no CNV regions satisfying the inheritance pattern and segregation in family 127 were detected for any genes. Notably, the audiological phenotype of SB127 was similar to that of autosomal recessive deafness-16 (DFNB16), which is caused by STRC mutations. STRC has a pseudogene and the coverage of STRC by ES was not sufficient in our previous study, we used the ES data to determine the average coverage of the exonic regions of STRC. As expected, we confirmed that approximately 50% the regions of STRC were not captured well (<20×); therefore, we decided to perform Sanger sequencing for the uncovered exonic regions of STRC (Table S4). Additionally, we performed multiplex ligation-dependent probe amplification to ensure the absence of CNV-STRC. This confirmed that SB127 did not segregate STRC mutations.

FIGURE 1.

Pedigree and audiological phenotypes and Sanger sequencing traces of the p.Ser197Hisfs*54 and p.Gly1100Arg variants of family SB127 and the conservation of the two mutant residues among orthologs of POLD1: (a) Family SB127 shows the autosomal recessive inheritance of sensorineural hearing loss (SNHL). (b–d) SB127–497 (F/70 years old [YO]), SB127–235 (F/37 YO) and SB127–247 (F/35 YO) show normal hearing thresholds. (e,f) SB127–277 (F/45 YO) and SB127–219 (M/29 YO), carrying two mutant alleles of POLD1, show a moderate degree of SNHL. (g) Sanger sequencing traces of the c.584_585GA and c.3298G in the unaffected subject, SB127–247. (h) Sanger sequencing traces of c.584_585del and p.Gly1100Arg from the proband, SB127–219. (i) Two amino acid residues of POLD1 (p.Ser197 and p.Gly1100) are highly conserved from human to zebrafish among POLD1 orthologs, but not in yeast (Saccharomyces cerevisiae). (j) Schematic of the functional domains of POLD1 protein showing the relative locations of: nuclear localization signal in green; the exonuclease domain in red; the polymerase active site in blue and the ZnF modules (CysA, yellow; CysB, orange) in purple. MDP, mandibular hypoplasia, deafness, and progeroid feature

The affected proband (SB127–219; male, 27-year-old) presented clearly normal skeletal morphology, facial morphology, fat distributions, and gonadal feature/function, without any anomalous features. The results of serum lipid profile, glucose and insulin levels, fasting insulin and triglyceride levels, and the thyroid and liver function were also normal (Table S5).

The wild-type and all the mutant Pol δ proteins used for this study showed similar molar ratios of p125, p66, p55, and p12 except a de novo POLD1 variant (p.Ile1070Asn) which was reported to be associated with the newly clarified MDPL syndrome (Figure S1a, b; Elouej et al., 2017), indicating that stable complex formation or interaction among p125, p66, p55, and p12 requiring an intact cysteine-rich metal-binding motif B (CysB) (Netz et al., 2012; Sanchez Garcia, Ciufo, Yang, Kearsey, & MacNeill, 2004) was achieved by most mutants, including p.Gly1100Arg. The p.Ile1070Asn mutant, which lies within a region termed ZnF2 domain also called CysB motif located near the C-terminus, was expressed in E. coli (Figure S1d), but did not interact with p125, p66, p55, and p12 (Figure S1c), suggesting that disruption of interaction with other subunits in the assembly of the active polymerase causes impairment of polymerase function (Elouej et al., 2017). To compare the DNA polymerization activities of the Pol δ mutants, we first conducted a primer extension assay in the presence of replication Factor C (RFC)and PCNA. An initial experiment comparing the wild-type and four mutant enzymes showed that recombinant mutant enzymes carrying p.Pro327Leu and p.Ser478Asn (related to colorectal cancers) with a diminished exonuclease activity of Pol δ (Palles et al., 2013) had polymerase activities comparable to that of the wild-type (Figure 2a), whereas the p.Ser605del mutant recombinant enzyme, which is associated with MDPL syndrome, was completely defective for the polymerase activity, as reported previously (Weedon et al., 2013). Pol δ-p.Gly1100Arg appeared slightly less active compared with the wild-type and three Pol δ mutants in terms of DNA polymerization activity (Figure 2a). The wild-type and p.Gly1100Arg-mutant enzymes were expressed and purified in two independent preparations, and their polymerization activities were compared (Figures 2b, c). The results indicated that p.Gly1100Arg showed a 30–40% reduction in the polymerization activity (Figures 2b, c). The affinity of Pol δ for the primer/template DNA was also tested, and Pol δ-p.Gly1100Arg bound to the DNA substrate with efficiency similar to that of the wild-type (Figure S2). This suggested that the p.Gly1100Arg mutant Pol δ enzyme, which retains normal binding activity to the DNA substrate, could compete with the wild-type Pol δ enzyme, which might further reduce the Pol δ polymerization activity in a dominant-negative fashion in the normal carrier state. We tested our recombinant enzymes for exonuclease activity and found that the p.Gly1100Arg mutant Pol δ had a comparable activity to that of the wild-type enzyme, whereas the Pol δ exonuclease activity of the mutants p.Pro327Leu, p.Ser478Asn, and p.Ser605del were approximately 30%, 25%, and 40% of the activity of the wild-type, respectively (Figure 2d). Endogenous POLD1 levels from patient-derived lymphoblastoid cell lines of SB127 were also examined. The cell line from the normal hearing carrier (SB127–235) of p.Gly1100Arg showed levels of POLD1 (×1.12) comparable with that shown by the wild-type (controls 1, 2, 4, and 5), indicating that p.Gly1100Arg does not affect the stability of this protein significantly, at least under these experimental conditions. However, the cell line from the affected proband, SB127–219 carrying a truncation mutation (p.Ser197Hisfs*54), showed 33.1% of the POLD1 protein level compared with the control cell lines (Figure 2e). Next, we determined whether the results obtained with the recombinant Pol δ enzymes could be replicated in the patient-derived lymphoblastoid cell lines. Furthermore, we checked the overall enzymatic activity of Pol δ from two mutant alleles of POLD1 in lymphoblastoid cell lines derived from the normal hearing sibling (SB127–235) carrying p.Gly1100Arg, and patient (SB127–219) carrying the two mutant alleles of POLD1. By testing the SB127–235-derived cell line, we aimed to evaluate the potential competition of Pol δ-p.Gly1100Arg with coexpressed Pol δ-wild-type of POLD1. Pol δ-p.Gly1100Arg showed 60–70% of the activity of the Pol δ-wild-type enzyme (Figure 2c); therefore, the extracts from SB127–219 (affected) carrying both p.Gly1100Arg and a presumptive null function variant were predicted to have about 30–35% of the polymerase activity of normal controls. As predicted, the extracts from SB127–219 (affected) had severely reduced activity that was 33% ± 2 of that from the unrelated normal control subject. Interestingly, the Pol δ activity in the extracts from the normal carrier of p.Gly1100Arg, SB127–235 showed 52 ± 3% of the activity of the normal control, which was significantly higher than that of SB127–219 (affected), but was lower than the estimates of the remaining activity (80–85% = 50% [wild-type allele] + 0.5 × [60–70%] p.Gly1100Arg]) expected from a 30–40% reduction from p.Gly1100Arg recombinant mutant Pol δ (Figure 2f). This could be attributed to the potential role of the p.Gly1100Arg mutant Pol δ as a strong competitive inhibitor of wild-type Pol δ. Alternatively, this could have resulted from certain negative interactions between the two alleles that do not affect DNA synthesis functions, such as competition at the level of gene expression and/or holoenzyme formation, causing further significant decreases in the activity in the cell lines. In the same assay system, DNA synthesis by all three extracts was inhibited by aphidicolin, but not by dTTP, indicating that these activities are catalyzed predominantly by replicative enzymes (Figure 2f). Further, cell extracts from the three subjects have a difference only in the POLD1 genotype (wild-type, a single heterozygote, and compound heterozygote of POLD1 variants). Therefore, we believe that the observed differences in the enzyme activities of their extracts in the primer extension assay reflected mainly the Pol δ activities in each cell line. Moreover, the two mutant residues of Pol δ detected in our study were not conserved in other replicative enzymes, Pol α and Pol ε, which suggested a minimal effect of these variants on the function of Pol α and Pol ε. However, some contributions from Pols α and ε cannot be ruled out completely.

FIGURE 2.

(a,b) DNA polymerization activity of DNA polymerase Delta (Pol δ). (a) Primer extension by Pol δ wild-type (WT) or mutant enzymes. The relative activity of p.Pro327Leu, p.Ser478Asn, and p.Gly1100Arg (Prep 1) Pol δ to the WT Pol δ (Prep 1) at 100 fmol was measured as 78%, 73%, and 61%, respectively (single measurement). The activities at 300 fmol appear saturated. (b) Primer extension by Pol δ WT or the p.Gly1100Arg mutant from two independent preparations. (c) Polymerization activity is presented as the percentage of fully extended products among the total primer substrates: the relative activities of p.Gly1100Arg Pol δ to the WT Pol δ at 50, 100, and 200 fmol were calculated as 48%, 60%, and 69%, respectively. These data were from triplicate experiments (Figure 3b is one representative) and are plotted with standard deviations. (d) 3'→5' exonuclease activity of Pol δ. A 5'-labeled single-stranded oligonucleotide was incubated with the indicated Pol δ enzymes. The exonuclease activity was calculated as described in Supporting Information. Average activities and standard deviations from three reactions are presented in the right panel. (e) Cellular levels of POLD1 from patient-derived lymphoblastoid cell lines. The expression level was quantified using “Image J” and the relative amount adjusted to the amount from control 1 is provided. The normal-hearing subject 127–235 carrying p.Gly1100Arg as a single heterozygote showed a similar amount of POLD1 protein to that from the controls under these experimental conditions. The affected subject 127–219, however, showed only 33% of the protein amount compared with the controls and the normal carrier of p.Gly1100Arg. (f) DNA synthesis activity of whole-cell extracts from Pol δ-mutants and normal control cell lines. Left panel, primer extension by the cell extracts. These assays were conducted with 10 μg of protein from the indicated extracts, as described in Supporting Information. When indicated, 30 μg/ml aphidicolin or 100 μM ddTTP (in the presence of 40 μM dTTP) was added to the reactions. Right panel, relative normalized polymerization activity, with the normal control set as 100%. The polymerase activities of the SB217–219 and SB217–235 extracts were 33% ± 2% and 52 ± 3%, respectively. These data were from triplicate experiments and are plotted with standard deviations

Because POLD1 mutations cause moderate NS-SNHL in the family SB127, we examined the localization of POLD1 in the mouse inner ear by immunohistochemical staining and in situ hybridization. The inner ear includes a compartment named the scala media, which is filled with endolymph is characterized by its unique composition: high potassium and high electrical potential. Positive POLD1 staining was observed in cells from the structures lining the scala media, including Reissner's membrane, the spiral limbus, and supporting cells in the organ of Corti, as well as cells from the spiral ligament (Figure S3a–e). Additionally, the POLD1 was also detected in ß III-tubulin (Tub3)-positive spiral ganglion neurons, the primary sensory neurons in the cochlea (Figure S3f–h). The POLD1 in cells of the spiral ligament was exclusively concentrated in the nucleus, where DNA polymerase is localized. In contrast, strong nonnuclear signals were observed at stria vascularis (SV) and pillar cells, which were considered false positives (Figure S3a). In addition, our in situ hybridization data also demonstrated that POLD1 was expressed in cells from the structures lining Reissner's membrane, the interdental cells, and cells of the organ of Corti (Figure S3i–k).

In the present study, we report, for the first time, that POLD1 mutations can cause a moderate degree of NS-SNHL in an autosomal recessive fashion. Previously, a de novo POLD1 variant (p.Ser605del and p.Ile1070Asn) at the polymerase active site was reported to be associated with the newly clarified MDPL syndrome, which included SNHL as one of the symptoms and signs (Elouej et al., 2017; Shastry et al., 2010; Weedon et al., 2013). However, the affected subjects from family SB127 in this study did not have any clinical features indicating MDPL, suggesting that SNHL caused by POLD1 mutations can be NS-SNHL. Despite the fact that SB127 is not a large family and that we have not been able to recruit the second, unrelated NS-SNHL family or individuals carrying recessive POLD1 variants yet, we suggest that compound heterozygous recessive mutations of POLD1 cause NS-SNHL in SB127. First, our group reported that approximately 45% of sporadic cases of moderate NS-SNHL in a Korean pediatric population could be diagnosed genetically by ES and a customized filtering step (Kim et al., 2015). In the present study, we performed ES not only from a proband but also from all five surviving subjects in family SB127. Our filtering step indicated clearly that POLD1 variants were the only candidate variants accounting for SNHL. Additionally, we considered the possibilities of CNVs or capture failures. Second, SNHL has been reported to be related to POLD1 alterations, albeit in a syndromic form (Weedon et al., 2013). Consistently, previous reports on MDPL have shown an association with SNHL (Freidenberg et al., 1992; Shastry et al., 2010). However, they were unable to determine the molecular etiology of this syndrome. Recent reports provided evidence that the p.Ser605del mutant of Pol δ at polymerase active sites were documented to be completely deficient in polymerase activity (Weedon et al., 2013), suggesting that decreased replicative activity of Pol δ is somehow related to the development of SNHL. The present study showed that the POLD1 missense variant (p.Gly1100Arg) had a significant reduction in replicative polymerase activity. Although Gly1100Arg variant has been found in colorectal cancer, this variant is neither present in population databases (ExAC no frequency) nor has been reported in the literature in individuals with a POLD1-related disease. The mechanism by which the p.Gly1100Arg variant exerts a pathogenic potential on the polymerase replicative activity is not clear. This variant resides in the C-terminal domain of the catalytic subunit of Pol δ and is physically close to the CysA and CysB motifs. The CysB motif, crucial for the formation of a stable complex between the subunits of Pol δ, was not affected by p.Gly1100Arg, as shown by the similar molar ratios of p125, p66, p55, and p12 between the wild-type and p.Gly1100Arg enzymes in the present study. Therefore, we asked whether the p.Gly1100Arg variant could affect the CysA motif, which was reported to be the key binding site for PCNA. However, we did not observe any correlation between the reduced activity and the quantity of PCNA (Figure S4), which was significantly different from the data shown in the paper by Netz et al. (2012), suggesting that the 30–40% reduction in the activity of p.Gly1100Arg must be intrinsic to the Pol δ protein, and not caused by the interaction with PCNA. Taking our results into account, we propose that some degree of reduction, but not complete abolishment, of replicative POLD1 enzymatic activity, can lead to NS-SNHL. The other truncation mutant that we detected, p.Ser197Hisfs*54, was predicted to be a null mutation. Even though the presence of some replicative function from the first 80 amino acids of the N-terminal portion of POLD1 was reported previously (Schumacher, Stucki, & Hubscher, 2000), the 50% reduction of polymerase activity measured from the extracts of the affected subject SB127–235 carrying p.Ser197Hisfs*54 strongly suggested that p.Ser197Hisfs*54-Pol δ was nearly devoid of any function. The property of p.Ser605del, which is likely to stall the replication forks (Weedon et al., 2013), might further decrease the replication efficiency (Branzei & Foiani, 2010). It is possible that p.Ser605del has more severe effects than expected from a simple estimated reduction in the total Pol δ activity, such as blocking DNA synthesis and leaving long stretches of single-stranded regions. The p.Ser605del mutant recombinant Pol δ showed a complete lack of replicative polymerase activity under our experimental conditions (Figure 2a). However, the DNA-binding activity of the p.Ser605del mutant recombinant Pol δ was reported to be preserved (Weedon et al., 2013). Our biochemical Pol δ assay might not represent all of the properties of the mutant proteins during in vivo replication. Therefore, the phenotypic difference between the two single heterozygotes (p.Ser197Hisfs*54/+ [SB127–497] vs. p.Ser605del/+) could be attributed mainly to the dominant-negative effects on the wild-type allele exerted from the p.Ser605del/+allele. Thus, the total enzyme activity from the cells carrying p.Ser605del in such a state would be significantly less than what was expected from the heterozygous null allele of POLD1 (50%) (Figure 2). Measurement of replicative polymerase activity from cell extracts of p.Ser605del carrier will clarify this. The p.Ile1070Asn variant in the Zinc Finger 2 (ZnF2) domain of the protein is predicted to cause a major impairment of POLD1 activity (Elouej et al., 2017). Our data showed that p.Ile1070Asn mutant within the ZnF2 domain lies did not interact with p125, p66, p55, and p12 (Figure S1c). This ZnF2 domain has been reported to form an iron–sulfur cluster of the 4Fe-4S type and to be crucial for the assembly of the active DNA polymerase d complex and, the p.Ile1070Asn mutant is predicted to cause disruption of the Fe–S cluster, since this has been shown to be crucial for interaction with other subunits in the assembly of the active polymerase impairment of polymerase function (Elouej et al., 2017). POLD1 germline variants (p.Pro327Leu and p.Ser478Asn) in the exonuclease domain have been associated with colorectal cancer or endometrial carcinoma (Palles et al., 2013; Valle et al., 2014) were documented to be located near the DNA interface of the domain or to impair the function of exonuclease, which was consistent with our data showing significant impairment of the exonuclease activity of Pol δ (Figure 2d). Despite the fact that the exonuclease activity of p.Ser605del was significantly decreased to an extent comparable to that of exonuclease domain variants such as p.Pro327Leu or p.Ser478Asn (Figure 2d), the clinical phenotype of these two groups was strikingly different. The polymerization and proofreading activity of POLD1 is obviously necessary for the maintenance of the normal functions of all organs in the human body. However, the reasons why SNHL is the most penetrant feature resulting from the defective polymerase activity of Pol δ remain unclear. We speculate that the cochlea needs high polymerase activity because of the unique existence of the extracellular fluid endolymph. Endolymph maintains a highly polarized resting potential and high potassium concentrations, which can be toxic to neighboring cells. It might be that the cells lining this endolymphatic space need to replicate rapidly, thereby requiring the high polymerase activity of POLD1. Indeed, the immunolocalization and RNA in situ hybridization data indicate that POLD1 is mainly expressed in cells surrounding the endolymph in the cochlea (Figure S3).

In summary, we report that NS-SNHL can result from a significant reduction, but not a null function, of the polymerase d activity, thereby extending the phenotypic spectrum of defective POLD1 activity.