CNVs resulting from genome rearrangements typically involve sub-microscopic insertions, duplications, and deletions within the genome[32, 33]. Most carriers with minor CNVs exhibit no clinical phenotype, whereas a small subset of individuals manifest neuropsychiatric abnormalities or other conditions, including cancer, syndromic aortopathy, inherited retinal degeneration, and osteoporosis[10, 34-36]. Consequently, the identification of CNVs is crucial for early diagnosis and prevention of diseases.
Advances in genetic testing technologies have made CNVs testing convenient and efficient. Array comparative genomic hybridisation (aCGH), which was initially employed over two decades ago for CNV identification, has become the most frequently used detection method[37]. However, microarray platforms have several limitations[38]. In the contemporary era, NGS technologies have revolutionised biological research[39]. Emerging CNV detection methods based on NGS are superior to those based on aCGH in terms of sequencing speed, detection cost, and reproducibility of sequencing data[40]. As genetic testing technology continues to evolve, carriers of CNVs now have the option to prevent the onset of related diseases through PGT. Recently, various studies have explored the use of aCGH, NGS, and other CNVs detection techniques. However, most of these studies are in the form of case reports, with few analysing the differences of these detection techniques[41-44].
In the present study, we tested 122 blastocysts derived from 16 couples undergoing PGT cycles for CNVs and euploidy using four distinct methods. These methods include multiplex PCR sequencing of the target SNP loci, whole-genome SNP genotyping, and low-coverage whole-genome NGS at resolutions of 4 Mb and 1 Mb. Multiplex PCR sequencing of target SNP loci relied on linkage analysis through NGS-based SNP haplotyping and screening for aneuploidy in WGA products. This technique selectively amplifies sequence-specific genomic regions containing the target SNP loci for family-based analyses. Whole-genome SNP genotyping was performed using the Illumina HumanKaryomap-12 microarray which encompasses approximately 300,000 SNP markers. Karyomapping employs genomic biomarkers to assess the likelihood of an embryo carrying a genetic variant, confirming the presence or absence of specific alleles within CNVs, based on an SNP array from parental DNA and translocation-associated unbalanced embryos. Low-coverage whole-genome NGS identifies CNVs by analysing the depth of coverage of sequencing reads across different genomic regions, where the DNA copy number correlates with the number of filtered reads mapped to the genome. The NextSeq550 platform used in our study had an average read length of 2×75 base pairs, with a minimum sequencing depth of 0.01X at 4 Mb resolution and 0.1X at 1 Mb resolution.
Among the four cohorts, a significant difference was observed in the age of females at the time of oocyte retrieval (P=0.000). However, no significant differences were noted in other aspects, such as paternal age, number of retrieved oocytes, and fertilisation rates.
There were notable differences in the proportion of unaffected blastocysts among the four groups (P=0.027). In addition, the group employing multiplex PCR sequencing of the target SNP loci exhibited a remarkably higher blastocyst quality control failure rate of 9.2%, surpassing the 0% observed in the other three groups (P=0.034). Multiplex PCR sequencing of target SNP loci imposes stringent quality requirements on amplification products; however, the occurrence of undetermined results is relatively low, indicating a higher detection rate. In this assay, the WGA of biopsied samples was performed first, followed by linkage analysis through NGS-based SNP haplotyping, and aneuploidy screening. Insufficient DNA quantity, contamination, unsuccessful WGA, errors in DNA library preparation, and sequencing processes may be responsible for samples failing quality control[45].
Despite advancements in detection technologies, a subset of blastocysts has not yet been definitively characterised. Factors potentially contributing to an undetermined diagnosis include poor embryo quality, varying days of embryo biopsy, technical issues during sample preparation, amplification failure, and errors in sequencing or data analysis[46, 47]. Previous studies have concluded that the inconclusive diagnosis rate of PGT-A ranges from 0.86 to 3.8%[48]. In our study, the proportion of undetermined blastocysts was higher in the group using the NGS-based low-coverage WGS technique at 1 Mb resolution than that of the group at 4 Mb resolution. However, the differences among the four groups did not reach statistical significance (P=0.078), possibly because of the small sample size. Further investigations with larger sample sizes are warranted for more robust comparisons. Furthermore, both the lc-WGS (4 Mb) (9.1%) and lc-WGS (1 Mb) (12.5%) groups exhibited undetermined diagnostic rates higher than those at the general level. In addition, the sensitivity of whole-genome sequencing can be enhanced by increasing the sequencing depth, thereby reducing the proportion of results that cannot be determined[49, 50]. Addressing these factors and minimising the number of undetermined blastocysts are urgent challenges that must be addressed.
The cost per PGT cycle is primarily related to the number of biopsied embryos and testing cost of each sample. In our study, the cost-effectiveness of NGS-based whole-genome sequencing was evident, primarily because of the lower sequencing depth. Among the four testing assays, lc-WGS, with a resolution of 4 Mb emerged as the most economical technique. As the sequencing depth increased, lc-WGS at a 1 Mb resolution became slightly more expensive than lc-WGS (4 Mb). Conversely, the cost of single-sample testing via whole-genome SNP genotyping is relatively high, with resolution playing a crucial role in determining the price of SNP array-higher resolution. While multiplex PCR sequencing of target SNP loci is affordable, it does not facilitate whole-genome sequencing or aneuploidy screening.
To investigate pregnancy outcomes following the transfer of normal blastocysts identified using four different techniques, we analysed various indicators, including biochemical pregnancy, clinical pregnancy, clinical miscarriage, ongoing pregnancy, delivery, and continued pregnancy. No significant differences were observed between the four groups, indicating a comparable level of accuracy for these techniques.
To further validate the precision of the PGT results, we recommend invasive prenatal diagnosis using amniocentesis in patients who achieve clinical pregnancy. Concordance of amniotic fluid genetic testing results with PGT results was noted in eight of nine cases analysed, with the remaining case from F12 revealing a de novo 1.38 Mb micro-deletion in the q33.3-q34 region of chromosome 13 with uncertain clinical significance. Since chromosome 13 was outside the target region for multiplex PCR sequencing and the size of the deletion was greater than the detection threshold of lc-WGS (4 Mb), the microdeletion in this instance was not detected by PGT. Further genetic analysis is crucial for a comprehensive understanding of the potential implications of the detected chromosomal abnormalities. This de novo variant may result in poor clinical outcomes if it is pathogenic. The patient in this case underwent caesarean delivery and ultimately gave birth to a phenotypically normal female baby. Additional genetic testing of the peripheral blood of infants is necessary to reach a conclusive diagnosis. However, these data are currently flawed, as the patient did not undergo genetic detection, and we are still conducting follow-up assessments. It may be possible to identify tiny CNV fragments by deepening whole-genome sequencing and enhancing bioinformatics analysis procedures. Nevertheless, when determining the appropriate sequencing depth, it is essential to re-evaluate the cost-benefit ratio as well as the time and technological requirements, given that the majority of CNVs at the kilobase level are benign and represent nonsense variants[49].
To the best of our knowledge, this is the first study comparing different PGT strategies for CNVs. The detection rates of the four CNV assays were comparable. However, multiplex PCR sequencing at the target SNP loci exhibited a higher frequency of sample quality control failures and a relatively low incidence of undetermined blastocysts. In contrast, lc-WGS (1 Mb) and lc-WGS (4 Mb) exhibited a lower rate of quality control failures, but a higher prevalence of undetermined blastocysts. Targeted detection using multiplex PCR sequencing for small fragments of CNV is limited to specific loci and does not provide comprehensive genome-wide detection. Additionally, the detection of aneuploidy requires the assistance of lc-WGS, which leads to increased costs. Moreover, the quality of amplification products must be high to obtain reliable results. Despite these challenges, the relatively low proportion of indeterminate results suggests a high detection rate. Genome-wide SNP genotyping incurs a higher cost but requires no lc-WGS assistance and has comparatively laxer quality standards for amplified products, higher detection rates, and a decreased chance of inconclusive results. Whole-genome sequencing with low coverage is more cost-effective and has relaxed the requirements for improved quality of the amplified product. Owing to the relatively established analysis methodology for 4 Mb, the detection rates were higher, and the complexity was reduced. Contrastingly, 1 Mb analysis is more difficult and has a higher rate of indeterminate results. For CNVs larger than 4 Mb, lc-WGS represented the optimal PGT solution for comprehensive cost benefits. Currently, whole-genome SNP genotyping is the best option for handling CNVs at the kilobase level without considering the cost of sequencing. As sequencing and analytical technologies improve, sequencing costs decrease, and the development of long-read sequencing technologies progresses, lc-WGS holds the potential to become the optimal choice for 1–4 Mb CNV-PGT in the future.
A limitation of our study is that the statistical significance of the results may have been affected by the limited sample size. In the future, a larger sample size will be appropriate for an in-depth analysis of CNV detection techniques.