Poor Correlations in the Levels of Pathogenic Mitochondrial DNA Mutations in Polar Bodies versus Oocytes and Blastomeres in Humans

Abstract

Because the mtDNA amount remains stable in the early embryo until uterine implantation, early human development is completely dependent on the mtDNA pool of the mature oocyte. Both quantitative and qualitative mtDNA defects therefore may negatively impact oocyte competence or early embryonic development. However, nothing is known about segregation of mutant and wild-type mtDNA molecules during human meiosis. To investigate this point, we compared the mutant levels in 51 first polar bodies (PBs) and their counterpart (oocytes, blastomeres, or whole embryos), at risk of having (1) the “MELAS” m.3243A>G mutation in MT-TL1 (n = 30), (2) the “MERRF” m.8344A>G mutation in MT-TK (n = 15), and (3) the m.9185T>G mutation located in MT-ATP6 (n = 6). Seven out of 51 of the PBs were mutation free and had homoplasmic wild-type counterparts. In the heteroplasmic PBs, measurement of the mutant load was a rough estimate of the counterpart mutation level (R² = 0.52), and high mutant-load differentials between the two populations were occasionally observed (ranging from –34% to +34%). The mutant-load differentials between the PB and its counterpart were higher in highly mutated PBs, suggestive of a selection process acting against highly mutated cells during gametogenesis or early embryonic development. Finally, individual discrepancies in mutant loads between PBs and their counterparts make PB-based preconception diagnosis unreliable for the prevention of mtDNA disorder transmission. Such differences were not observed in animal models, and they emphasize the need to conduct thorough studies on mtDNA segregation in humans.

Main Text

Mitochondrial DNA (mtDNA) is a maternally inherited, small, circular molecule located in the mitochondria. The mtDNA content remains stable during the first days of embryonic development, until mtDNA replication resumes, after implantation in the uterus.¹ The abundance of mtDNA in the mature oocyte (≈300 000 mtDNA copy number) is therefore crucial for early embryonic development. Qualitative and quantitative mtDNA defects have been reported in fertilization failures,^2,3 oocyte maturation defects,^4,5 and impaired murine postimplantation embryonic development.⁶ Yet how mtDNA segregates during oogenesis remains largely unknown. A mitochondrial genetic “bottleneck,” with only a small number of mtDNA genomes giving rise to the genotype of the offspring, is thought to occur during female gametogenesis.^7–11 It causes the fixation of new mtDNA variants, such that deleterious mutations are either lost from the maternal line or reach very high levels and cause a severe disease in a few generations. In humans, data collected from primary oocytes and early embryos carrying mtDNA mutations have suggested that the bottleneck occurs at the end of gametogenesis.^12–16 The bottleneck size may be mutation specific,^12–16 and it would depend on individual parameters.¹⁶ No data on mtDNA segregation during the late stages of human meiosis are currently available. In mice, similar heteroplasmy levels have been found in the first and second polar body (PB) and the corresponding oocyte for both a polymorphism¹⁷ and a deletion,¹⁸ suggesting that mtDNA molecules segregate randomly at this stage.

Taking advantage of our preimplantation genetic diagnostic (PGD) service, we investigated mtDNA segregation during human meiosis by measuring heteroplasmy levels of three different pathogenic mtDNA mutations in 51 first PBs and their counterparts (oocytes or preimplantation embryos). Significant mutant-load differences were found between PBs and counterparts, with a negative correlation at high mutant loads. Our results clearly differ from those in murine models, outlining the necessity of conducting mtDNA segregation studies in humans.

Three couples at risk of transmitting a pathogenic mtDNA mutation were referred for PGD and signed an informed consent for themselves and their embryos. This study was approved by the National Ethics Committee from the French Biomedical Agency. The m.3243A>G mutation in MT-TL1 (MIM 590050), was found at a heteroplasmic state (around 25% in blood, urinary tract, and buccal cells) in the unaffected female partner of couple 1. Her second child had encephalopathy, with stroke-like episodes, seizures, and ataxia, consistent with the diagnosis of MELAS syndrome (MIM 540000). The m.3243A>G mutation load was 60% in cultured skin fibroblasts from the proband. The couple had previously experienced two terminations of pregnancy for fetuses with a mutation load above 60%. Couple 2 requested PGD when the mother's brother passed away from myoclonic epilepsy with ragged-red fibers (MERRF syndrome [MIM 545000]). The female partner suffered from occasional myoclonic jerks, carried the m.8344A>G mutation in MT-TK (MIM 590060) , and had mutant loads of 75%, 85%, and 95% in blood, urinary pellet, and buccal cells, respectively. The first daughter of couple 3 presented in the first year of life with cerebellar ataxia and proximal tubulopathy. MRI showed symmetric lesions in the basal ganglia and hyperintensities in the cerebellum. A partial decrease of the ATP synthase activity was found on a skeletal muscle sample (40%) and was subsequently ascribed to the m.9185T>C mutation in MT-ATP6 (MIM 516060), with a 95% mutation load in skeletal muscle, urine, blood, and buccal cells. The proband's mother was found to have the mutation in blood (25% mutant load), urinary tract (14%), and buccal cells (17%).

PGD cycles were carried out via a standard in vitro fertilization (IVF) procedure in couples 1 (3 cycles), 2 (1 cycle), and 3 (1 cycle). Oocytes were retrieved by transvaginal ultrasound-guided aspiration. The zona pellucida was pierced with the use of a noncontact 1.48 μm diode laser system (Fertilase, MTM Technology), and the first PB was carefully aspirated from 57 ovulated metaphase II oocytes (33, 16, and 8 oocytes for patients with m.3243A>G, m.8344A>G, and m.9185T>C mutations, respectively). Fertilization by intracytoplasmic sperm injection resulted in 38 embryos, and their evaluation at day 4 or 5 enabled classification of “arrested” embryos (14 embryos containing fewer than 6 cells at day 3 or not continuing embryonic cleavage after the biopsy) or “developing” embryos (n = 24). A total of 28 out of 38 embryos containing at least six blastomeres were biopsied on day 3 after injection, as previously described.¹⁹ Overall embryos were either transferred (n = 8) or collected (n = 30) on day 3 or days 4–5. Unfertilized oocytes (n = 13) were analyzed separately. Oocytes, blastomeres, and overall embryos were rinsed in a drop of PBS supplemented with 0.1% polyvinyl alcohol (Sigma Aldrich, France) under a binocular microscope with the use of a mouth-controlled, finely pulled glass pipette. Cells and embryos were deposited in 3 μl of lysis buffer²⁰ under visual control with an inverted microscope and were lysed by heating for 10 min at 65°C. For each sample, a small volume of rinsing medium was transferred similarly and used as a negative template.

The mutations were quantitated via semiquantitative fluorescent PCR and restriction endonuclease digestion, as previously reported.^15,16,21 Such tests have been shown to be suitable for mutant-load assessment over a large range of mtDNA copy-number templates (from 10² to 10⁶ molecules), with good repeatability (± 2%).^16,21 Single-cell and embryonic DNA were amplified in a 25 μl reaction volume containing 0.2 μM of each primer (Eurogentec, Angers, France) (Table S1 available online) and 2× Master Mix (12.5 μl, QIAGEN Multiplex PCR Kit). Denaturation was carried out at 95°C for 15 min, followed by 30 cycles of 30 s at 94°C, 90 s at corresponding annealing temperature (Table S1), and 1 min at 72°C, and terminated with a final extension of 30 min at 60°C. Cross hybridization of oligonucleotide primers to genomic DNA was ruled out by PCR amplification on mtDNA-less Rho0 cells.²² The m.3243A>G amplification products (10 μl) were submitted to HaeIII digestion.¹⁶ The m.8344A>G mutation was quantitated via an Allele Creating Restriction Site (ACRS) method based on forward primer modification to create a Bgl1 restriction site in the mutant allele.²³ Taking advantage of the suppression of the MnlI restriction site by the m.9185T>G mutation, we submitted corresponding PCR products (10 μl) to Mnl1 digestion for 2 hr at 37°C. Digested products (1 μl) were suspended in a mix containing formamide (15 μl, genetics-analysis grade, Applied Biosystems) and ROX 400HD (0.3 μl, Applied Biosystems). After denaturation for 2 min at 95°C, samples were electrophoresed in an automated ABI3130 genetic analyzer (Applied Biosystems). Results were analyzed with the Genescan and Genotyper software (Applied Biosystems). The mutant load was calculated by dividing the mutant peak area by the sum of the mutant and wild-type peak areas. The mutant-load-detection threshold was 2% for the three mutations.

Unequal partitioning of mitochondria between two daughter cells of vastly different sizes (inheriting around 1% and 99% of the ooplasm, respectively)²⁴ generates a genetic drift (the so-called “sampling error”). A reliable statistical estimate of this drift can be calculated, but it is necessarily based on the standard assumption that the mtDNA molecules drawn into the PB are randomly distributed and sampled from the oocyte and that the sample size is much smaller than the mtDNA content of the oocyte. The sampling error, which is a function of the heteroplasmy rate (p) and the sample size (N_sample, in this case the number of mtDNA molecules within the PB), was assessed as $2\sqrt{p\left(1-p\right)/N_{sample}}$ with a 95% confidence interval. Since we could not experimentally ascertain the amount of mtDNA in the PB because of technical limitations, the sampling error was estimated over a range of possible N_sample values. Mean values are reported as mean ± 2 standard errors of the mean (SEM). Comparisons of mean values were performed with the use of a two-tailed t test assuming unequal variances, calculated in Excel 2007.

Successful amplification was achieved in 51 of 57 PBs (amplification rate 89%). A total of seven PBs were found to be mutation-free (six and one from the patients with m.3243A>G and m.9185T>G mutations, respectively), as were their corresponding oocytes (n = 3) or embryos (n = 4). Among the heteroplasmic PBs (n = 44), the proportion of mutant mtDNA ranged from 2% to 95% (mean 50%, median 53%). The corresponding mutant loads, tested on oocytes (n = 10), whole embryos (n = 28), or two blastomeres (transferred embryos, n = 6), ranged from 5% to 95% (mean 48%, median 53%; Figure 1). The mean mutant load was 37% (± 27%) and 44% (± 26%) in “arrested embryos” and “developing embryos,” respectively.

Figure 1.

Comparison of Mutation Level between First Polar Bodies and Their Counterparts

Values are shown for 3243 (black circle), 8344 (white circle), and 9185 (black star).

The curves show the 95% confidence interval for a sampling error calculated by $2\sqrt{p\left(1-p\right)/N_{sample}}$, in which N_sample is the mtDNA copy number in the polar body. The dotted line and the full line have been calculated for two values of N_sample (100 and 10 mtDNA copy number, respectively). The maximum sampling error occurs at p = 0.5. Seven PBs and counterparts were wild-type homoplasmic.

The mutant load of each PB was then compared to its counterpart value (Figure 1). In only half of the cases (n = 27 of 51), the PBs and their counterparts had similar mutant loads (± 10%). In particular, homoplasmic wild-type PBs were constantly associated with homoplasmic wild-type counterparts (n = 7). By contrast, in the remaining half of the cases, the first PBs differed from their corresponding oocytes or embryos. Indeed, 13 PBs had mutant loads 11% to 34% higher than their counterparts, whereas 11 PBs had mutant loads 13% to 34% lower than their counterparts. These differences were observed for the three mutations (Figure 1). A poor correlation in the levels of mutant mtDNA in PBs and their counterparts was therefore observed (coefficient of correlation R² = 0.68, p < 0.0001 for a linear fit). The coefficient of correlation was even lower when only heteroplasmic PBs were considered (R² = 0.52, p < 0.0001 for a linear fit). Considering that the mtDNA amount sampled from the ooplasm to the PB could account for the differences in heteroplasmy levels between the two cells, a statistical estimation of the mtDNA copy number in PBs was made. The spread of the observed variations was consistent with a very low mtDNA copy number in PBs (largely ten molecules; Figure 1).

At higher values of PB mutations, the mutation load appeared generally higher in the PB than in its counterpart (Figure 1). The PBs were then split into two groups: below and above a 60% mutation rate. In the first group (n = 34), the mean difference between the counterpart and the PB mutant loads was 3.5% ± 5% (Figure 2). The average difference, however, raised to −11.8% ± 7.8% in the second group (n = 17; Figure 2). The difference between the two groups (high versus low mutated PBs) was shown to be significant with the use of a two-tailed t test (p value = 0.0023).

Figure 2.

Apparent Influence of the Heteroplasmy Rate on the Correlation between PB and Counterpart Mutant Loads

The values shown are means ± 2 SEM of the difference between the counterpart and the polar body mutant loads, calculated from 34 and 17 PBs carrying less than or over 60% mutant loads, respectively.

Here, we explored the mtDNA segregation during the last stage of oogenesis by comparing the ratios of mutated versus wild-type mtDNA populations in the first PB and in its counterpart (oocyte, or blastomere, and/or whole embryo) for three pathogenic mtDNA mutations. This study shows that the human first PB is not predictive of the egg mutant load, precluding the use of PB testing for the diagnosis of mtDNA disorders. Preconception diagnosis (PCD), based on the first PB²⁵ and/or second PB testing,²⁶ is routinely applied to nuclear gene disorders, in countries where cleavage-stage embryo selection is banned, or in couples declining potential embryo discarding. It was therefore considered as an attractive diagnostic procedure for mtDNA disorders. By fixing the mutant-load threshold under which an embryo is transferred at 30%, a first-PB-based decision would have led to the transfer of four embryos carrying mutant loads over 40% and to the discarding of five embryos with mutant loads less than 30%, giving a predictive positive value of the PB test at 0.84 but a negative predictive value at 0.63. The first PB is thus not suitable as a diagnostic material, unless the purpose of PCD is the selection of “mutation free” oocytes.

The mechanism for unequal partitioning of mitochondria between PB1 and the oocyte is unclear. Under the random-sampling hypothesis, we have computed that only the very low number of ten mtDNA molecules directed to the PB could explain the different mutant loads between human PBs and their oocytes or embryos. This estimate is two orders of magnitude less than that from a previous report, mentioning an mtDNA copy number of approximately 1000 in ten human first PBs from healthy women.²⁷ Thus, mtDNA copy number could be lower in mutant than in wild-type PBs, due to some unknown biological process impacting the mtDNA amount during the course of gametogenesis. Very little is indeed known about mtDNA turnover during the human gametogenesis and early embryonic development,²⁸ but a dramatic effect of a small increase in mutant load over the mtDNA copy number has been reported in a human placenta.²⁹ Another, more likely, hypothesis is that the above estimate of the mtDNA copy number based on the sampling error mathematical modeling is not fully reliable, owing to a nonrandom distribution of mtDNA populations in the primordial oocyte and therefore a nonrandom segregation between the two daughter cells.

Moreover, we observed that the mutant-load distribution between PBs and ooplasm is apparently influenced by the mutant load in the PBs (Figure 2). The mutation levels were indeed significantly lower in the oocytes or embryos than in the PBs (by 12% on average) at high mutant loads (> 60%), whereas they were not significantly lower at lower mutant-load values. This observation suggests a purifying effect against highly mutated cells acting at some stage of gametogenesis or early embryonic development, possibly resulting from a replicative advantage to wild-type DNA.³⁰ This selection process would conceivably counterselect highly mutated oocytes, whose subsequent embryonic development would be compromised by the respiratory chain defect.⁶ Yet, we did not observe any correlation between mutant loads and early embryonic development (developing versus arrested embryo), at least under the mutation rates of 70%, 90%, and 54% for m.3243A>G, m.8344A>G, and m.9185T>G, respectively.

It is worth noting that the results of the few similar studies in mice are quite different. A close correlation between PBs and their counterparts was indeed found for an mtDNA polymorphism (correlation coefficient at 0.99)¹⁷ and an mtDNA deletion (correlation coefficient at 0.95).¹⁸ These discrepancies between these previous reports in mouse models and the present one in humans raise several issues. One can hypothesize an influence of mtDNA point mutations versus neutral polymorphisms or large rearrangements on meiotic mtDNA segregation, irrespective of species. One can also hypothesize that the processes governing mtDNA segregation during meiosis differ between humans and mice. Given that the perinuclear redistribution of mitochondria during oocyte maturation is similar in mice and humans,³¹ species specificity of the mitochondrial trafficking could be considered.

Whatever their physiological bases, such apparent discrepancies between the animal studies and ours have to be kept in mind, especially in the context of preimplantation or preconception diagnoses.³² These results emphasize the limits of animal models and the subsequent need to conduct mtDNA segregation studies in humans before selecting a reproductive option for patients carrying mtDNA mutations.

Web Resources

The URLs for data presented herein is as follows:

• Online Mendelian Inheritance in Man (OMIM), http://www.ncbi.nlm.nih.gov/Omim/