Novel POLG variants in mitochondrial disease patients
In a search for previously undescribed variants, POLG coding sequence analysis was conducted for 60 patients with clinical diagnosis of myopathy (OMIM # 251900) (n = 18), encephalomyopathy (n = 5), PEO (OMIM #157640 and #258450) (n = 4), PEO and myopathy (n = 3), encephalopathy (n = 3), polyneuropathy (n = 2), myoclonic epilepsy associated with ragged-red fibres (MERRF) (OMIM #545000) (n = 2), Kearns-Sayre syndrome (OMIM #530000) (n = 1), myopathy, encephalopathy, lactic acidosis and stroke-like episodes (MELAS) (OMIM #540000) (n = 1), mitochondrial neurogastrointestinal encephalopathy (MNGIE) (OMIM #613662) (n = 1), sensory ataxic neuropathy, dysarthria and ophthalmoparesis (SANDO) (OMIM #607459) (n = 2), general suspicion of mitochondrial disease (n = 15) or Alpers syndrome (OMIM #203700) (n = 3). Sequencing of the POLG gene in these patients, followed by in silico translation, revealed 1 poorly uncharacterised and 5 previously unreported sequence changes located in the regions conserved between yeast and human (c.924 C>T/p.Arg309Cys and c.867 C>T/p.Arg290Cys—proband from family A, c.2604 C>T/p.Arg869Ter—proband from family B, c.2901 C>G/p.Gln968Glu—proband from family C, c.3155_3166delCA/p.Thr1053Argfs*6—proband from family D and c.3316 T>C/p.Val1106Ala—proband from family E). None of the novel variants were present in the Human Polymerase Gamma Mutation Database, NCBI dbSNP (Sherry 2001), the 1000 genomes project (Genomes Project Consortium 2012), or in the Exome Variant Server database (http://evs.gs.washington.edu/EVS/). These variants, along with the poorly characterised p.Arg309Cys, were therefore selected for further investigation, and their sequences deposited in GenBank (accession numbers are listed in “Materials and methods”).
Patients’ clinical presentation
The results of sequence analysis, clinical and molecular evaluation of the four patients carrying previously unreported POLG variants and one with the known, but poorly characterised variant, are summarized in Table 1. Proband from family A (AI2) (Fig. 1a) is a 54-year-old woman with SANDO syndrome and positive family history (her brother presented similar symptoms). From the age of 31, she suffered from progressive external ophthalmoplegia, ptosis, dysarthria, weakness of upper and lower limbs and sensory ataxic neuropathy. Additionally, mental retardation was diagnosed. Nerve conduction studies indicated axonal sensory and motor neuropathy. MRI showed brain atrophy. Skeletal muscle biopsy disclosed ragged-red fibres on light microscopy examination. Analysis of mitochondrial DNA revealed multiple deletions in muscle tissue. Along with the p.Arg309Cys variant, patient AI2 also had another previously unknown sequence change c.867 C>T (p.Arg290Cys) for which, due to a low level of similarity in this position between POLG and Mip1 (Fig. 2c), creating a yeast model was impractical. According to PolyPhen-2 (Adzhubei et al. 2010), a software tool which predicts possible impact of a single amino acid substitution on the function of human proteins, both these variants found in patient AI2 are probably damaging, with a score of 1.00. SIFT Human Protein DB tool also predicts that both variants p.Arg290Cys and p.Arg309Cys are not tolerated (SeqRep = 0.92) (Ng and Henikoff 2001). The proband’s brother (AI1) carries both p.Arg309Cys and p.Arg290Cys variants present in the proband AI2.
The proband from family B (BII1) was without any symptoms at infancy. At the age of 12 months, he became progressively ataxic and recurrent convulsions appeared, developing quickly to severe epileptic encephalopathy. Soon after introduction of anticonvulsant valproic acid administration, a fulminant liver failure was observed and the patient died at the age of 2.5 years. Severe mitochondrial depletion found in the liver tissue on autopsy (0.08 %, while the threshold for diagnosing depletion is set at <30 % of the normal mtDNA level) was in agreement with the clinical suspicion of Alpers syndrome. POLG sequencing revealed two in trans variants, one known pathogenic p.Trp748Ser substitution (Palin et al. 2010; Van Goethem et al. 2004), and a novel variant p.Arg869Ter.
Proband from family C (CII1) (Fig. 1c) is a 53-year-old woman with progressive external ophthalmoplegia and negative family history. At the age of 38, she observed impaired eye movements and ptosis. Nerve conduction studies were normal, and electromyography indicated myogenic changes. Cardiological assessment revealed no abnormalities. Skeletal muscle biopsy disclosed ragged-red fibres on light microscopy, and multiple abnormal mitochondria with paracrystalline inclusions on electron microscopy examination. Multiple mtDNA deletions were present in muscle tissue. POLG sequencing revealed a known p.Arg309Leu (Lamantea et al. 2002) mutation and a novel variant p.Gln968Glu, predicted to be not tolerated by SIFT (SeqRep = 0.97), and probably damaging with a score of 0.992 according to PolyPhen2.
In the proband from family D (DII1) (Fig. 1d), a psychomotor retardation and muscle hypotonia were observed since 2 months of age, and autistic behaviour was observed during early chilhood. At the age of 3 years, the patient was given valproic acid due to episodes of unconsciousness and abnormal EEG pattern and responded suddenly with an acute liver failure. He died 6 months later before the qualification for liver transplantation was completed. Severe depletion of mtDNA was found in the liver section taken at autopsy (3.4 %, ref. >30 %), as in the case of the patient BII1. In the POLG sequence, in addition to the known p.Trp748Ser mutation, a novel variant p.Thr1053Argfs*6 was found.
In spite of normal developmental milestones, the proband from family E (EII2) (Fig. 1e) started to have walking difficulties at the age of 13 years. Her condition deteriorated, and she developed ataxic gait and dysarthria. Two years later, she developed action-exacerbated myoclonus. When last seen at 15 years, she had no ophthalmoparesis. Nerve conduction studies showed sensory-motor polyneuropathy of lower limb nerves. We did not detect mtDNA deletions in a DNA sample isolated from peripheral blood. Since muscle tissue was not available, the decision to sequence POLG was undertaken based on the patient’s clinical presentation, and indicated that she was a compound heterozygote p.Trp748Ser/p.Val1106Ala. The novel variant (p.Val1106Ala) is designated as not tolerated (SIFT SeqRep = 0.98) and possibly damaging with a score of 0.85 (PolyPhen2). Interestingly, another amino acid change (valine to isoleucine) in the same position was described previously as probably pathogenic (Horvath et al. 2006).
Each of the novel variants coexisted in the patient with a known pathogenic recessive mutation, and all the patients’ parents were healthy; therefore, we suspected that all the variants, if pathogenic, were also recessive. For all the patients, DNA samples isolated from other members of their families were also available. To confirm in trans localization of novel versus known variants in patients AI2, BII1, CII1, DII1 and EII2, we performed family studies, and in one case cloning.
Familial and population study
To verify reciprocal localization of novel and known variants in patients, we acquired DNA samples from their family members, and amplified and subsequently sequenced the relevant POLG fragments (Fig. 1). In three families (B, D and E), the proband’s parents were each carrier of one of the two pathogenic/possibly pathogenic variants. In the case of proband AI2, the parents’ DNA was unavailable, but POLG sequence analysis of her son (AII1) proved him to be the carrier of the known, but poorly characterised p.Arg309Cys variant. Her symptomatic brother (AI1) had both variants p.Arg290Cys and p.Arg309Cys found in the proband. Additionally, we performed cloning of POLG alleles in patient AI2, which further confirmed that she was a compound heterozygote. The family of patient CII1 did not give consent to obtain their DNA from peripheral blood, but agreed to perform buccal swabs. Both her children are heterozygous carriers of the pathogenic (CIII1-p.Arg309Leu) or possibly pathogenic (CIII2-p.Gln968Glu) variant. The presence of the novel variant p.Gln968Glu in her father (CI1) was ruled out, but due to a low yield of DNA extraction, we were not able to confirm the presence of the p.Arg309Leu mutation. Thus, we confirmed that in all the affected individuals, the two variants were in trans.
The population frequency of all novel variants was estimated using PCR–RFLP or HRM tests. Because the population frequency of p.Arg309Cys was not previously assessed, we also conducted PCR–RFLP test for this variant. None of the analysed POLG sequence variants were found in 320 chromosomes of healthy adults, which means that the population frequency of each variant is significantly lower than 1 %.
The analysis of familial distribution and frequency in the general population suggests that the analysed POLG sequence changes are probably pathogenic and inherited in a recessive manner.
Modelling the putative pathogenic mutations in yeast
To gain some insights into the effects of the novel variants in the human polymerase γ using the well established and accessible S. cerevisiae model, we first aligned the amino acid sequences of human POLG and its yeast ortholog Mip1p, and identified the residues that were mutated in patient sequences (Fig. 2). Of the 5 residues, three (Arg309, Arg869, and Val1106) were conserved in the yeast Mip1p. The site of the frameshift mutation p.Thr1053Argfs*6 also corresponds to a block of a nearly complete amino acid sequence identity. The Gln968 residue, also in a conserved region, corresponds to Arg770 in the yeast protein. In this case, in addition to the mutated allele (with Glu in this position), we also created the humanized allele (Arg770Gln).
The mutant mip1 alleles were created by site-directed mutagenesis and introduced into the Δmip1 background on centromeric (low copy number) vectors, as described in “Materials and methods”. The plasmid shuffling strategy was used to maintain functional mtDNA owing to the presence of the wild-type MIP1 gene on an URA3 vector, which could subsequently be eliminated by 5-FOA counterselection (Sikorski and Boeke 1991).
Influence of mip1 mutations on respiratory growth phenotype in the yeast model
Growth on a non-fermentable carbon source (glycerol or ethanol) is the fastest and simplest method to examine respiratory function in S. cerevisiae strains. Among the five tested strains carrying yeast equivalents of putative pathogenic mutations, three exhibited a severe respiratory deficiency. Mutants Arg265Cys, Arg672Ter and Thr809Ter were unable to support growth on glycerol at 30° (Fig. 3). Growth of the remaining strains did not differ from the wild-type. Incubation at a restrictive temperature (37°) contributed to an even more severe phenotype of mutants which were unable to grow on glycerol (Arg265Cys, Arg672Ter, Thr809Ter), causing slower growth even on a medium with fermentable carbon source. The elevated temperature did not significantly affect the rest of the tested strains.
Determination of petite frequencies in strains with mip1 mutations
Petite mutants of S. cerevisiae are unable to grow on non-fermentable carbon sources and form only small anaerobic colonies in the presence of glucose. While this phenotype can result from a point mutation in mtDNA or a mutation in a nuclear gene, most often it is caused by a complete loss (rho
0) or extensive deletions (rho
−) of mtDNA. Petite colonies appear at a low (1–5 %) rate spontaneously in cultures grown on fermentable carbon sources. Mutations in nuclear genes encoding proteins involved in mitochondrial function often result in a marked increase in the frequency of petite generation (Contamine and Picard 2000; Lipinski et al. 2010).
All the strains carrying mip1 alleles corresponding to the modelled human mutations show increased petite frequencies in comparison to the one with the wild-type MIP1 (Fig. 4a; Table 2). Furthermore, mutations Arg265Cys, Arg672Ter, Thr809Ter, which cause a complete respiratory deficiency, lead to a complete loss of functional mtDNA in the tested strains.
For the Arg770Glu mutation, petite accumulation is significantly (p < 0.05) higher than in its humanized allele mip1Arg770Gln and in the wild-type version of the protein (5.6, 1.6, and 1.4 % at 30°, respectively). Furthermore, there is no significant difference between wild-type Mip1p (1.4 % petite) and humanized Arg770Gln (1.6 %). Petite frequency of the strain carrying the Val863Ala allele (5.3 %) is approximately three times higher than in the wild-type strain at 30°, and this difference is statistically significant (p < 0.05).
To confirm that the observed respiratory negative phenotype was related to the loss of functional mtDNA, petite colonies from the strains carrying the Arg265Cys, Arg672Ter and Thr809Ter variants (50 colonies each) were crossed to a wild-type rho
0 tester, and the resulting diploids assayed for respiratory growth. This test confirmed that in all the petite colonies there was no functional mtDNA left (rho
0). Due to the presence of nonfunctional residual mitochondrial genome fragments in rho
− mtDNA, it is not practical to perform quantitative analysis of mtDNA levels in yeast cells using PCR-based methods. For the strains carrying the Arg265Cys, Arg672Ter and Thr809Ter variants, DNA staining with Hoechst dye followed by fluorescent microscopy imaging revealed that, unlike in the wild-type control, extranuclear DNA signal is mostly absent, suggesting that they are essentially rho
0 (Online Resource 2).
The severe phenotype of the Arg265Cys, Arg672Ter and Thr809Ter variants raised the possibility that these mutations abolished the expression of the MIP1 gene, and thus were equivalent to transcriptional nullomorphs. We performed an RT-PCR assay on total RNA isolated from the strains carrying the Arg265Cys, Arg672Ter and Thr809Ter variants as the only MIP1 alleles, and found no apparent change compared to the wild-type control, thus confirming that these mutations do not interfere with the expression of the MIP1 gene, at least on mRNA level (Online Resource 3).
Determination of petite frequencies in heteroallelic strains
To determine the dominance/recessivity of the mutations, we assayed the frequency of petite formation in heteroallelic yeast strains bearing each of the mutant alleles as well as the wild-type MIP1 allele (Fig. 4b; Table 2). A hemiallelic strain containing the empty YCplac111 vector as well as the vector carrying the wild type MIP1 (vector/MIP1) was also used as a control to normalise the allele dosage and differentiate between antimorphy and haploinsufficiency. The petite frequency in the hemiallelic strain is slightly, but significantly (p < 0.05) higher than in the MIP1/MIP1 strain, and this effect is more pronounced at the elevated temperature (37°). MIP1 haploinsufficiency is evident when a diploid strain heterozygous for the Δmip1 allele is used (Baruffini et al. 2006). The haploid deletant strain with plasmid-borne wild-type and mutant MIP1 alleles used in this study is clearly less sensitive to haploinsufficiency.
In the Val863Ala/MIP1 strain, the petite frequency was slightly, yet significantly (p < 0.05) higher than in the vector/MIP1 strain at 30°. No increase in petite frequency could be observed at the elevated temperature (37°). Even though the difference at 30° was statistically significant, the petite frequency in this heteroallelic strain remained low (well below 5 %); the dominant negative effect of this allele is thus negligible.
The petite accumulation frequency for the heteroallelic strain Arg770Glu/MIP1 strain, as well as for the humanized Arg770Gln/MIP1, was not significantly elevated above the level observed in the hemiallelic vector/MIP1 control at either temperature; this mutation is, therefore, recessive.
Strains heteroallelic for the three mutations that showed the most severe phenotype (complete respiratory deficiency and loss of functional mtDNA) in the monoallelic setting, Arg265Cys/MIP1, Arg672Ter/MIP1, and Thr809Ter/MIP1, accumulated petites at a much higher frequency than the hemiallelic vector/MIP1 control strain (at both temperatures), although they still retained >80 % of functional mtDNA. These alleles thus have a partially dominant negative effect on the mtDNA replication process (Fig. 4b; Table 2). To verify whether this effect is dependent on allele dosage, we also constructed strains carrying the mutant alleles on multicopy (episomal) plasmids, and the wild-type allele on a centromeric plasmid (Fig. 4c). For Arg265Cys and Arg672Ter, no significant difference was observed between the low copy number and multicopy vectors, for the Thr809Ter allele; however, increasing the copy number of the mutated variant resulted in a significantly (p < 0.01) more pronounced phenotype of the heteroallelic strain, suggesting a true antimorphic character of this mutation.
Frequency of mtDNA point mutations in the yeast model
Two of the five mutated mip1 alleles, Arg770Glu and Val863Ala, showed only a partial decrease in overall mtDNA stability (measured by the frequency of petite colony formation), whereas the humanized Arg770Gln variant was indistinguishable from the wild type in that aspect. No effect on the respiratory growth was observed in these three strains. To test whether the replicative function of the mitochondrial DNA polymerase in these strains was not affected in a more subtle way, we performed experiments assessing the point mutation rates in mtDNA.
Quantification of point mutation accumulation in mtDNA is based on estimating the frequencies of spontaneous oligomycin- (OliR) and erythromycin-resistant (EryR) mutants that result from substitutions in mitochondrial genes encoding the ATPase subunits or the large subunit rRNA, respectively. This assay can indicate changes in the fidelity of mtDNA replication, such as defects in the proofreading activity of the mitochondrial DNA polymerase.
Two mutations (Arg770Glu, Val863Ala) and the humanized allele (Arg770Gln) were examined in comparison to the wild-type strain (Table 2; Fig. 5). The strain carrying the Val863Ala allele showed a slight, but still statistically significant (p < 0.05, Mann–Whitney U test) increase in the frequency of both OliR and EryR mutations. The effect of the Arg770Glu allele, on the other hand, was more pronounced, and significantly (p < 0.05) higher both in comparison with the wild-type MIP1, and with the Arg770Gln humanized variant (differences between the humanized allele and the wild type version were not significant). Interestingly, this substitution (as well as Val863Ala) is located in the polymerase domain (Table 2). It has to be noted, however, that the effect observed for this variant, while statistically significant (p < 0.05), is far less evident than that reported for an MIP1 allele devoid of the corrective exonuclease domain activity (exo
−), where the EryR mutation frequency increased by 550-fold (Stumpf et al. 2010).