Mitochondrial DNA quantification correlates with the developmental potential of human euploid blastocysts but not with that of mosaic blastocysts

This study was approved by the Ethics Committee of the First Affiliated Hospital of Anhui Medical University, China (No. 20211217). This study was a retrospective analysis of archived data using clinical samples obtained from patients referred to the Reproductive Medical Center of Anhui Medical University for the purpose of PGT-A. Written informed consent for the analysis of archived data was obtained from patients with the approval of the Ethics Committee.

Embryos in this study were subjected to TE biopsy at the request of the patients for chromosomal aneuploidy assessment. For this purpose, the standard procedure in our laboratory involved whole-genome amplification (WGA; SurePlex, Illumina, CA, USA) followed by next-generation sequencing (NGS) on the Ion Proton (Thermo Fisher Scientific, MA, USA) or NextSeq550 (Illumina, CA, USA) platform. The experimental procedure described herein involved the analysis of archived data following the completion of PGT-A. The embryos were not subjected to additional manipulations, and the clinical treatment of the patients was not changed as a result of their participation in this study.

The study included 2,814 biopsied blastocysts from all 664 PGT-A cycles (576 couples, among whom five completed three cycles and 78 completed two cycles) performed from June 2018 to June 2021. Of the 2,814 biopsied samples subjected to NGS, 1,460 were processed on a semi-conduct sequencing platform (Ion Proton), whereas 1,354 were processed on an Illumina sequencing platform (NextSeq550). All samples were de-identified prior to the initiation of this study. Table 1 summarizes the characteristics of the study population, and no significant differences were observed between samples processed on different NGS platforms.

Oocyte retrieval and intracytoplasmic sperm injection were performed as previously described [34]. For embryo micromanipulation and trophectoderm biopsy, embryos were transferred from blastomere medium (COOK, Sydney, Australia) to blastocyst medium (COOK, Sydney, Australia) at 10:00 on the third day after insemination and cultured until they reached the blastocyst stage on days 5 or 6. The biopsy was conducted between 11 a.m. and 12 p.m. on day 5 or 6. First, an enlarged blastocyst was attached with a holding pipette (Sunlight Medical Inc, Jacksonville, USA) at the 9 o'clock position, and then a laser was used to produce a tiny hole in the zona pellucida (Hamilton Thorne LYKOS, Beverly, MA, USA). Subsequently, the biopsy pipette was constantly pressed against the blastocyst's perimeter until the TE shrunk and detached from the inner surface of the zona pellucida. Multiple laser pulses were then utilized to bore a 5 μm hole at the far end of the zona pellucida from the inner cell mass (ICM). The biopsy pipette was then introduced into the blastocyst through the hole, and the constricted TE was aspirated under negative pressure. Some TE cells were simultaneously extracted from the hole, and a laser was used to chop TE cells along the flat mouth of the pipette. All procedures were carried out on a heated micromanipulator (ECLIPSE Ti2, Nikon, Japan). Following biopsy, the blastocysts were cryopreserved. The TE samples were placed in a 200μL PCR tube with 2 L phosphate-buffered saline (Thermo Fisher Scientific). According to the manufacturer's guidelines, the DNA isolated from TE samples was submitted to WGA using SurePlex (Illumina), followed by NGS, using the Ion Proton (Thermo Fisher Scientific) or NextSeq550 (Illumina) systems. The reads were aligned to the NCBI GRCh37 human reference genome using in-built analysis software and converted to BAM files.

To analyze chromosomal abnormalities, the genome was partitioned into non-overlapping windows of constant size ranging from 600 kb to 1 Mb. Subsequently, a set of 30 reference values was constructed by averaging the proportion of mapped reads attributed to each window in a series of euploid samples. The ratio of the percentage of reads derived from a chromosomal window > 10 Mb in an embryo (test) sample to the reference value corresponding to the same window was calculated. Copy number variations with a length of > 4 Mb were considered positive. Ratios > 1.25 were associated with chromosomal duplication, whereas ratios < 0.75 were associated with deletion [35].

To determine mosaicism, read counts in each window (RC) were normalized to avoid the local GC content and RC biases induced by genomic mappability. Briefly, RCs were corrected using the observed coverage deviation for a specific GC percentage and mappability score using Eq. 1.where, RCi denotes the read counts for the i-th window, mGC denotes the median read counts for all windows with the same GC percentage as the i-th window, mMAP denotes the median read counts for all windows with the same mappability score as the i-th window, and m denotes the overall median of all windows [36].

The hidden Markov model (HMM) was employed to determine mosaicism. The HMM matrix was established based on known negative samples, and eight chromosomal status codes were set as follows: 0, 5, 10, 15, 20, 25, 30, and 40, representing homozygous deletion, deletion mosaicism, haploid, deletion mosaicism, diploid, duplication mosaicism, triploid, and tetraploid, respectively. The transition probability between the same states of the initial matrix was 0.999, and the transition probability between different states was 0.025. The observed state transition probability matrix was calculated using the RC value for each window of the known negative sample. The state probability of each window was calculated using the above matrix and the corrected RC value of the sample to be tested. For windows that were reported as anomalies by HMM, the mosaic ratio was calculated using Eq. 2.where, RC_j is the read count of the j-th window and RC_N is the value of each window of the known negative sample.

Mosaicism of > 80% or < 20% cannot be distinguished from technical noise; therefore, samples with > 80% mosaicism were categorized as aneuploid, those with < 20% mosaicism were categorized as euploid, and those with mosaicism between 20 and 80% were categorized as mosaic. Mosaicism-positive chromosomes were analyzed in detail using our NGS validation data.

To calculate the mtDNA content, an optimized algorithm was used to process the output dataset generated by PGT-A analysis, which contained a mixture of mtDNA and nuclear DNA (nDNA) readings. To determine the relative copy number of mtDNA in embryos, the number of filtered reads mapped to the mitochondrial genome was divided by the number of reads mapped to the nuclear genome, and then multiplied by 100. This enabled batch normalization, thereby reducing the variability of NGS studies.

We extended the algorithm developed by Victor et al. [33]. Specifically, the total number of reads mapped to the mitochondrial genome (r_m) was divided by the total number of reads mapped to the nuclear genome (r_n). The corrected mtDNA score m_NGS was defined as the product of the r_m r_n dividant multiplied by the correction factor F_NGS (Eq. 3).where, F_NGS considers two parameters to adjust for the number of cells examined: embryo sex (sex correction factor, SCF) and ploidy (ploidy correction factor, PCF) using Eq. 4.where, SCF for female embryos equals 1 and SCF for male embryos equals 0.9842, because the length of male genomes is 98.42% that of female genomes. Aneuploid or mosaic embryos possess altered genetic material compared with euploid embryos, resulting in increased mtDNA copy numbers in monosomies and nullisomies, or decreased mtDNA levels in polysomies if not corrected. To correct for this, the mtDNA value for each embryo is multiplied by PCF based on the chromosomal composition and degree of mosaicism (Eq. 5). RGL represents the length of the reference genome, which is defined as the length of the diploid female at 6,072,607,692 bp. Parameter k represents the degree of mosaicism, which varies between 30 and 80%.

When embryos contain many lesions, the correction factors for each lesion should be multiplied. Thus, the corrected mtDNA score for embryos with aneuploidy mosaicism in more than one (n ≥ 1) chromosome segment was calculated using Eq. 6.

Statistical analysis was performed using SPSS 24 (IBM Corp., USA). Categorical data were compared using the Chi-squared test and Fisher’s exact test, where appropriate. Continuous data were compared using the Mann–Whitney U test and Wilcoxon signed-rank test, where appropriate. Linear regression analysis was applied to test the correlation between maternal age and mtDNA contents. Linear mixed models (LMMs) with a random cohort effect were used to test the effects of patient-specific characteristics, ovarian stimulation variables, and embryo parameters on the mtDNA content. Models were simplified by removing nonsignificant interaction terms based on a likelihood ratio test. Results were considered statistically significant at P < 0.05 and indicated high significance at P < 0.001.

The mitochondrial DNA (mtDNA) scores obtained from both the NextSeq550 and Ion Proton platforms exhibited a non-normal distribution, as illustrated in Fig. 1a and b. Specifically, the data sourced from the NextSeq550 platform displayed a positive asymmetry of 6.242 ± 0.066 and kurtosis of 55.154 ± 0.133, while the data sourced from the Ion Proton platform exhibited a positive asymmetry of 5.450 ± 0.064 and kurtosis of 48.744 ± 0.128. The median and interquartile range (IQR) values for mtDNA scores obtained from NextSeq550 were 0.074 and 0.066, respectively. On the other hand, the median and IQR values for mtDNA scores obtained from Ion Proton were 0.082 and 0.072, respectively. Statistical analysis using the Mann–Whitney test revealed a significant difference between the mtDNA scores obtained from Ion Proton and NextSeq550 platforms (P < 0.001). To eliminate the possibility of differences arising from variations in samples, 100 randomly selected samples that were tested using the Ion Proton platform were re-examined using the NextSeq550 platform. The Wilcoxon Signed Ranks test still showed a significant difference between the mtDNA scores obtained from the two different sequencing platforms (P < 0.001).

For Nextseq550 data, the median value of euploid embryos was 0.064, compared with 0.080 for aneuploid embryos and 0.073 for mosaic embryos. For Ion Proton data, the median value of euploid embryos was 0.077, compared with 0.086 for aneuploid embryos and 0.081 for mosaic embryos. Differences between the euploid and aneuploid groups were significant for both NextSeq550 (P = 0.0001) and Ion Proton (P = 0.0006) data (Fig. 2a and b). NextSeq550 data revealed significant differences between mosaic and euploid samples (P = 0.0029), whereas no significant difference from mosaic samples was observed in either euploid or aneuploidy samples sourced from Ion Proton data (Fig. 2a and b). Day 5 embryos tended to carry higher levels of mtDNA than day 6 embryos (P = 0.00001 for both NextSeq550 and Ion Proton data) (Fig. 3a and b).

Samples were divided into two groups based on maternal age above or below 35 years old at oocyte retrieval. Significant differences in the proportion of aneuploid embryos between both groups were observed using both NextSeq550 (P = 0.001) and Ion Proton (P = 0.005) data (Supplementary fig. S1). Linear regression analysis of mtDNA scores among euploid, mosaic, and aneuploid embryos revealed no difference between embryos derived from oocytes of different maternal ages for either NextSeq550 or Ion Proton data (Fig. 4a–f).

To analyze other factors that could be associated with embryonic mtDNA content, we employed an LMM to test the possible effects of patient-specific characteristics, ovarian stimulation variables, and embryo parameters on the mtDNA level of blastocyst TE biopsies. The model was simplified by removing nonsignificant interaction terms based on a likelihood ratio test. Blastulation rate (P = 0.021), day of biopsy (P = 0.000), ploidy of TE cells (P = 0.000), and NGS platform (P = 0.049) appeared to be associated with mtDNA score (Table 2). There was no evidence of an association between other patient-specific or embryo morphological variables and mtDNA scores.

Among the 2,814 recruited samples, 868 embryos were transferred, 772 were euploid, and 96 were mosaic. To investigate the effect of mtDNA content on clinical outcomes, given the significant effect of biopsy day, ploidy, and NGS platform on the mtDNA copy number, we first evaluated the median and third quartile for different cohorts of embryos (Supplementary Table S1 and S2). Subsequently, we stratified embryos into groups by mtDNA content (above and below the median, or above and below the third quartile; Tables 3 and 4). Between the euploid and mosaic cohorts, there were no significant differences in positive beta-human chorionic gonadotropin (hCG), implantation, live birth, biochemical miscarriage, or miscarriage rates in embryos stratified by the median mtDNA content (Tables 3 and 4). The cutoff was set at the third quartile. Euploid embryos with higher mtDNA contents presented significantly higher miscarriage rates (P = 0.012) and lower live birth rates (P = 0.024), whereas no significant difference was observed in the mosaic cohort (Tables 3 and 4).

Furthermore, we investigated the factors associated with the viability of euploid and mosaic embryos. In the euploid cohort, a better ICM grade was associated with higher live birth rate (P = 0.023) (Table 3); a better trophectoderm grade was associated with a higher positive beta-hCG rate (P = 0.004), implantation rate (P = 0.026), and live birth rate (P = 0.020) (Table 3). Compared to blastocysts biopsied on day 5, blastocysts biopsied on day 6 were associated with a lower positive beta-hCG rate (P < 0.001), implantation rate (P < 0.001), and live birth rate (P = 0.003) (Table 3). Endometrial thickness was associated with higher live birth rate (P = 0.018) (Table 3). Oocyte age were not associated with the outcomes of euploid embryos (Table 3). In the mosaic cohort, we observed no significant associations between the aforementioned factors and outcomes (Table 4).

Mosaic embryos were associated with poorer outcomes than euploid embryos: positive beta-hCG rates, 55.2% versus 65.7% (P = 0.043); implantation rate, 52.1% versus 60.1% (P = 0.132); miscarriage rate, 28.0% versus 16.2% (P = 0.036); live birth rate, 37.5% versus 50.4% (P = 0.017; Table 5). We further investigated the potential effects of mosaic parameters on clinical outcomes. In our dataset, blastocysts with only whole chromosome mosaicism were associated with a considerably higher miscarriage rate (P = 0.005) and lower rate of live birth (P = 0.064) than euploid blastocysts, whereas exclusively sub-chromosomal mosaicism showed no significant difference compared with euploid blastocysts (Table 5). Notably, blastocysts carrying a single mosaic lesion site were associated with a higher miscarriage rate (P = 0.021) and lower livebirth rate (P = 0.046) than euploid blastocysts, whereas no significant results were observed in blastocysts carrying multiple mosaic lesion sites compared with euploid blastocysts (Table 5). The degree of mosaicism, which is an estimate of the proportion of aneuploid cells in the biopsy sample, appeared to be associated with different rates of positive beta-hCG (P = 0.005), implantation (P = 0.007), miscarriage (P = 0.027) and live birth (P = 0.002) (Table 5).