Patients and samples

This study, including the corresponding consent forms, was approved by the Cyprus National Bioethics Committee (ΕΕΒΚ/ΕΠ/2020/38) and written informed consent was obtained from all participating families, allowing the use of their biological samples and publication of results.

Study design

This study was designed as a two-part investigation. A retrospective meta-analysis was carried out, examining data from 1016 spontaneously aborted fetuses to evaluate the gestational age at the time of loss and identify genetic abnormalities related to pregnancy loss at various developmental stages. Additionally, a cohort study was performed on 33 families with pregnancy loss and/or abnormal prenatal sonographic findings that had unrevealing routine genetic testing, implementing WES to identify potential genetic contributors.

Meta-analysis

Results obtained from genetic testing of 1016 spontaneously aborted fetuses (SAFs) referred to the department of Cytogenetics and Genomics of the Cyprus Institute of Neurology and Genetics due to pregnancy loss were extracted from the departmental database. The samples had all previously undergone chromosomal analysis and/or array-CGH and Quantitative Fluorescent (QF)-PCR as well as other forms of targeted testing. This data was accessed on the 23^rd of October 2020. Authors and researchers had no access to information that could identify individual participants. The samples were then categorised based on 1) trimester of pregnancy at the time of pregnancy loss, 2) normal vs. abnormal genetic test result and 3) the type of abnormality identified in each sample with abnormal genetic test result. The data was then analysed to identify the group of fetuses/families that would benefit the most from undergoing futher investigation with WES.

WES inclusion criteria

Based on the database meta-analysis, prenatal samples including amniotic fluid, chorionic villi or skin biopsies were collected from second-third trimester fetuses with abnormal ultrasound findings and/or prenatal death, referred to our laboratory for genetic investigation. The fetuses had all previously undergone some form of genetic testing, including chromosomal analysis and array-CGH with no causative findings. Recruitment of participating families started on the 2^nd of November 2020 and ended on the 26^th of September 2023. In total, 33 families provided written informed consents to be enrolled in this cohort to undergo trio-based WES. All clinical features identified in the fetuses were described using Human Phenotype Ontology (HPO) terms [26].

Genetic investigation

Fetal DNA was isolated from amniotic fluid using QIAGEN QIAamp DNA Mini Kit. Parental DNA was also isolated from peripheral blood using QIAGEN QIAamp DNA Blood Midi Kit according to the manufacturer’s instructions.

Library preparation was performed with Illumina DNA Prep with Enrichment kit and exome sequencing was carried out using Illumina NextSeq 500/550 High Output Kit v2.5 on Illumina NextSeq 500 System as per the manufacturer’s protocol. After completion of the WES run, trio check was carried out using Peddy software package v0.4.2 [27]to ensure genetic relatedness between the fetus and parents. Fastq files were loaded on commercial VarSome Clinical Saphetor SA platform [28] for alignment to hg19 reference genome, variant calling and annotation. SNVs, small variants and CNVs were predicted from the platform. Variants were filtered based on minor allele frequency (MAF), inheritance mode, coverage (>8 at the specific locus) and genomic location (variants targeted in coding and splicing regions ± 10 base pairs). The pathogenicity of each variant was assessed based on the recommendations of ACGS 2020 guidelines (Ellard et al., 2020), ClinGen sequence variant interpretation working group and was assisted by the automatic scoring ACMG classification of VarSome Clinical and VarSome. Variants classified as pathogenic, likely pathogenic and variants of unknown significance (VUS) with strong evidence of pathogenicity, were prioritized as they are more likely to have an effect on the protein. Conservation scores were also taken into account since evolutionary conserved regions of the genome tend to be of high importance for cellular function. Τhe phenotypic features of the probands were compared against other literature reports of either the same variant or different variants within the same gene. Variants were reported in cases where the proband’s phenotype matched the phenotype observed in other patients carrying variants in the same gene. Pathogenic and likely pathogenic variants that fit the fetal phenotype were reported back to the patients and referring physicians. VUS were only reported if there was some evidence of pathogenicity and/or they were identified in trans with a pathogenic or likely pathogenic variant in a gene following autosomal recessive inheritance. CNV analysis was also carried out using CNVkit and custom scripts in R to identify CNVs with a resolution of 100kb [29].

For variant validation, bidirectional Sanger sequencing was employed, using specifically designed primers (Metabion- Munich, Germany) through the Primer3 web interphase tool flanking the mutation site [30–32]. Applied Biosystems™ BigDye™ Terminator v1.1 Cycle Sequencing Kit (ThermoFisher Scientific- Waltham, Massachusetts, USA) and Affymetrix ExoSAP-IT^®† (Fisher Scientific- Pittsburgh, Pennsylvania, USA) were used for this procedure.

Meta-analysis

Initially, 1016 SAFs previously referred for routine genetic testing (chromosomal analysis, array-CGH, QF-PCR and targeted testing) were selected from the departmental database. Out of 1016 samples, 822 (80.91%) fetuses were spontaneously aborted during the first trimester of pregnancy and only 194 (19.09%) fetuses were spontaneously aborted during the second-third trimesters of pregnancy. The majority of first trimester SAFs (n = 569, 69.22%) had an abnormal genetic test result which determined the cause of the SPL, while the rest (n = 253, 30.78%) had a normal genetic test result leaving the cause of SPL undetermined (Fig 1). In contrast, a substantially higher percentage of second-third trimester SAFs (84.02%, n = 163) remained genetically undiagnosed. Meanwhile only 15.98% (n = 31) of fetuses from this group received a genetic diagnosis consistent with the reason for pregnancy loss using conventional methods of genetic testing. Thus, the majority of second-third trimester pregnancy losses do not receive a diagnosis relevant to the reason of pregnancy loss using chromosomal analysis and array-CGH.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. (A) Bar chart showing the distribution of genetic test result in first trimester SAF.

In total 253 out of 822 (30.78%) first trimester aborted fetuses had a normal genetic test result and 569 of 822 (69.22%) first trimester aborted fetuses had an abnormal genetic test result (B) Bar chart showing the distribution of genetic test result in second-third trimester spontaneously aborted fetuses. In total 163 out of 194 (84.02%) second-third trimester aborted fetuses had a normal genetic test result and 31 of 194 (15.98%) second-third trimester aborted fetuses had an abnormal genetic test result SPL=spontaneous pregnancy loss.

https://doi.org/10.1371/journal.pone.0319052.g001

Out of all first trimester aborted fetuses with an abnormal genetic test result, the most commonly identified genetic abnormality was aneuploidy or mosaic aneuploidy. More specifically, aneuploidies were detected in a total of 455 out of 569 first trimester SAFs with an abnormal genetic test (79.96%) (Table 1). The most common aneuploidy detected in first trimester SAFs was monosomy X with a total of 90 fetuses, followed by trisomies 16 (n = 74) and 22 (n = 74). The rarest aneuploidies observed in this group of SAFs were trisomy Y (n = 1), trisomy 11 (n = 2) and trisomy 17 (n = 2). Aneuploidies involving the chromosomes 1, 5 and 19 were not identified. Aneuploidies were also detected in 20 out of 31 (64.52%) of second-third trimester aborted fetuses with an abnormal genetic result, making it the most frequently identified abnormality in this group of fetuses as well. Trisomy 21(n = 7) was most commonly identified in this group of fetuses, followed by trisomy 13 (n = 4). Monosomy X which was the most common aneuploidy in first trimester aborted fetuses was identified 3 times in second-third trimester SAFs along with trisomy 18. A single case of trisomy 10, trisomy 14, and trisomy Y was also detected. Polyploidies were also detected in both groups of SAFs. In particular, 95 out of 569 first trimester fetuses with an abnormal genetic test (16.7%) were spontaneously aborted as a result of a polyploidy. In total, six of 31 (19.35%) second-third trimester fetuses with an abnormal genetic result were aborted due either of the following polyploidies: 69,XXX or 69,XXY. Unbalanced structural abnormalities in full or mosaic form had also been identified in a total of 12 fetuses. Out of these, 11 were spontaneously aborted during the first trimester of pregnancy accounting for 1.93% of first trimester aborted fetuses with an abnormal genetic result and one was spontaneously aborted in the second-third trimester of pregnancy accounting for 3.23% of second-third trimester aborted fetuses with an abnormal genetic result. CNVs have been determined to be the reason of pregnancy loss in seven of 569 fetuses miscarried during the first trimester (1.23%), and four of 31 fetuses in the second-third trimester (12.9%). Finally, only one case of pregnancy loss caused by uniparental disomy (UPD) has been identified in this study in a fetus spontaneously aborted during the first trimester of pregnancy (0.18%) (Table 1).

Download:

• PNG
  larger image
• TIFF
  original image

Table 1.. Distribution of abnormalities identified on genetic testing per trimester of pregnancy.

https://doi.org/10.1371/journal.pone.0319052.t001

Genetic investigation with WES

In total, 35 probands from 33 families were referred for genetic analysis using chromosomal analysis, array-CGH and WES after IUD/termination of pregnancy. The probands of each family were on average 23.58 weeks of gestation at the time of sample collection and referral for genetic analysis, and all of them were referred either during the second or third trimester of pregnancy. Amniotic fluid, chorionic villi and skin biopsies were collected from fetuses with phenotypic abnormalities in various organ systems including musculoskeletal (n = 5), cardiovascular (n = 2), central nervous (CNS) (n = 2), digestive (n = 1), urinary/excretory (n = 1), hydrops abnormalities (n = 2) and non-specific abnormalities including low birth weight and IUD (n = 3). More than half of the referred fetuses presented multisystem anomalies (n = 18) which is the presence of at least one major abnormality in two or more different organ systems (Fig 2). Out of the 33 families, six (18.18%) had a causative finding that was previously linked to a disorder relevant to the prenatal phenotype, which also followed the expected inheritance pattern. Out of these six diagnosed families, four fetuses were referred with multisystem anomalies and two fetuses were referred with musculoskeletal abnormalities. In addition to the six families that received a diagnosis from WES, pathogenic findings possibly implicated in the fetal phenotype were identified in two more families in heterozygosity in genes following an autosomal recessive inheritance pattern (Table 2). All three fetuses in these two families presented with multisystem anomalies (Fig 2).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Bar chart showing the distribution of WES diagnosis according to the affected organ system.

Orange bars indicate the diagnosed cases and blue bars indicate the undiagnosed cases. CNS=Central nervous system. Non-specific abnormalities include low birth weight and IUD..

https://doi.org/10.1371/journal.pone.0319052.g002

Download:

• PNG
  larger image
• TIFF
  original image

Table 2.. List of prenatal cases for which findings relevant to the reason of referral were identified.

https://doi.org/10.1371/journal.pone.0319052.t002

In this cohort, nine causative variants relevant to the phenotype of interest were identified in families 1–6. In addition to these findings, two possibly causative heterozygous variants relevant to the phenotype of interest were identified in families 7 and 8. All findings were identified in genes known to cause Mendelian disorders. The variants were classified as pathogenic or likely pathogenic based on ACGS 2020 guidelines, with the exception of the maternally inherited variant of family 4 and the paternally inherited variant of family 6 that were classified as VUS. Overall, four missense, four nonsense, two splicing and one frameshift variant were reported in families 1–8. Out of all identified variants, eight had previously been reported in ClinVar or relevant literature and three were novel. All identified variants that were reported as causative had a strong correlation with the referred fetus’ phenotype based on OMIM and relevant literature. The detailed clinical picture identified in the fetus of each family along with information relevant to the candidate variants can be found in Table 2. The remaining 25 families that were enrolled in this study did not receive a genetic diagnosis even after WES. All families that had no causative findings identified using karyotype, array-CGH and WES are presented in S1 Table.

The findings of families 1 and 2 have been described in detail in a previous publication [33].

Meta-analysis

The process of patient selection and the inclusion criteria established for each study, have a significant effect on subsequent WES detection rate [17,18]. It is therefore important to target the groups of fetuses that would benefit the most from WES and would have a higher chance of receiving a diagnosis. A meta-analysis of the departmental database was carried out, aiming to provide data that would aid in the selection of this target group, by investigating previous pregnancy loss referrals and their diagnostic outcomes.

Overall, the result of our meta-analysis was in agreement with previous research which had shown that approximately 50%–70% of SAFs present some form of cytogenetic abnormality with the most common being autosomal trisomies and monosomy X (approximately 80% of the cases combined) and polyploidies (20%) [3]. In our study, 59.06% of all the fetuses had an abnormal genetic test result of which aneuploidies appeared to be the leading cause of pregnancy loss (79.96% in the first and 64.52% in second-third trimester).

Chromosomal abnormalities tend to vary depending on the gestational age at the time of fetal demise. For example, miscarriages observed during early developmental stages often involve uncommon aneuploidies, whereas pregnancy losses occurring at more advanced developmental stages tend to present chromosomal abnormalities more commonly seen in live births [38]. Consistent with the above studies, we observed that the most common aneuploidy in first trimester fetuses was monosomy X (19.78% of all identified aneuploidies), followed by trisomies 16 and 22. As expected, fetuses presenting trisomies 16 and 22 were spontaneously aborted during the first trimester as these aneuploidies are incompatible with life. In contrast, the aneuploidies identified in the second-third trimester aborted fetuses were primarily trisomies 13, 18 and 21. Only a singular case of trisomies 10 and 14 were identified in second-third trimester aborted fetuses confirming the statement that unusual aneuploidies are not largely detected in more advanced developmental stages.

Following aneuploidies, polyploidies were the second most common abnormality identified in the referred fetuses. Both digynic and diandric polyploidies were identified in first trimester pregnancy losses, including 69,XXX, 69,XXY, 69,XYY, 92,XXYY, and 92,XXXX. In contrast, only 69,XXX and 69,XXY had been identified in second-third trimester losses. Although the origin of the second X chromosome in cases with 69,XXY was not assessed, the most probable scenario is that it is of maternal origin, as fetuses with diandric polyploidies are usually aborted during early stages of pregnancy [39].

Despite the high detection rate of combined chromosomal analysis and array-CGH for first-trimester pregnancy losses, 84.02% of spontaneously aborted second-third trimester fetuses do not receive a genetic diagnosis. A portion of these undiagnosed cases could be attributed to environmental factors such as maternal endometrial abnormalities and chronic conditions. In terms of genetic etiology, studies have shown that pregnancy losses occurring at more advanced developmental stages are more likely to be caused by smaller genetic abnormalities such as CNVs, indels and SNVs that cannot be detected by conventional methods of genetic testing and require additional genetic investigation [40,41]. Unlike fetuses with large chromosomal abnormalities that are incompatible with life and spontaneously aborted during the first trimester of pregnancy, fetuses with smaller genetic aberrations may be aborted at later developmental stages or even be born with phenotypic abnormalities that vary in severity, depending on the nature of the genetic abnormality and the effect it has on protein level [41]. For instance, a higher percentage of second-third trimester pregnancy losses (12.9%) in this study were caused by an CNV compared to first trimester pregnancy losses (1.23%).

The above data leads to the conclusion that, in combination with the ability of ultrasound to provide a more detailed and comprehensive phenotypic description in second-third trimester fetuses compared to those in the first trimester, the former group would derive the greatest benefit from WES in the absence of a previous genetic diagnosis. Therefore, a small study was conducted in which undiagnosed second-third trimester fetuses underwent WES to evaluate its utility for such cases.

Our WES investigation, along with prior research, highlights how the inclusion criteria can greatly impact the diagnostic yield. Previous studies show a 26% diagnostic yield when selecting cases of prenatal or perinatal deaths with thorough phenotypic data from autopsies. In contrast, studies focused solely on fetal structural abnormalities identified via ultrasound yield significantly lower (8.5%–10%) [42]. Here, the diagnostic yield, which was calculated at 18.18% (excluding fetuses from families 7, and 8), surpassed the 8.5%–10% anticipated when focusing solely on structural abnormalities. However, due to the absence of detailed phenotypic reporting in some cases and the lack of autopsies in most cases, the diagnostic yield fell short of the expected 26%. Therefore, in addition to patient inclusion criteria, successful WES diagnosis is also heavily dependent on detailed and accurate prenatal phenotypic descriptions [18].

WES findings

In this study, several inheritance patterns and types of variants were identified using WES. Out of eight families with findings related to the phenotype of interest, six families had findings in genes following an autosomal recessive inheritance and two had findings in genes with an autosomal dominant inheritance that occurred de novo. This displays the necessity of family-based WES, especially for countries like Cyprus that have a small population and thus an increased chance of recessive disorders.

In the aforementioned eight families, 11 causative variants were identified, out of which eight had previously been reported in publicly available databases and three were novel. The reporting of novel variants in this study was mainly based on previous phenotypic descriptions relevant to the gene in which the variant of interest was identified. The successful diagnosis of these families was therefore dependent on previous causative variant reporting and the associated clinical picture described in the literature, reflecting the importance of enriching the prenatal phenotype of disorders as well as creating new genotype-phenotype associations. The absence of suitable databases and guidelines for fetal variant classification creates difficulties and disparities in diagnostic yield between different studies [42]. Although a more detailed phenotypic description can be obtained at later stages of pregnancy, a major challenge of prenatal diagnosis is the fact that the fetal phenotype of many disorders has not yet been described or may considerably vary from the corresponding postnatal phenotype. Reporting phenotypic abnormalities of disorders with prenatal onset is therefore of utmost importance to create valuable additions to preexisting phenotypic spectrums, confirm previous findings and thus assist health care professionals in recognizing various disorders prenatally [43].

Variants that have been characterized in the literature before and have previously been established as a cause of a genetic disorder were much more easily identified and reported than novel variants. As a result, the diagnosis was made in a shorter time span, allowing families more time for genetic counselling and family planning. For example, the variant identified in the fetus of family 1 [33] in the SCN2A gene, had previously been reported as pathogenic and had been associated with developmental and epileptic encephalopathy 11 (DEE11) (OMIM #613721) which expedited the diagnosis. DEE11 is a neurologic condition characterized by seizures that appear neonatally [44]. In the case of family 1, postnatal phenotypic characteristics suggestive of DEE11 were identified by the referring physician. However, the lack of reported prenatal manifestations in the literature would create difficulties with reaching a diagnosis antenatally had the variant not previously been reported. Nevertheless, in this instance, new correlations between genotype and prenatal phenotype were established, potentially aiding healthcare providers in managing similar prenatal cases in the future.

In some cases, interpreting genetic variants was challenging due to the discrepancies between prenatal and postnatal phenotypic manifestations of genetic disorders. In addition, several cases in this cohort did not present extensive phenotypic abnormalities either due to gestational age or imaging limitations, creating even more complications in the interpretation of variants and therefore diagnosis. For instance, causative variants in RYR1 were detected in the fetus of family 4, leading to the diagnosis of congenital myopathy 1B (OMIM #255320). The fetus presented polyhydramnios, clenched hands, hypokinesia and talipes on ultrasound. Although these phenotypic features are suggestive of a congenital myopathy, they are not specific to one myopathy, which can create uncertainties in cases where the candidate variants are classified as VUS or in cases where the phenotypic abnormalities could be a result of several candidate variants. In this case the pregnancy had been terminated due to the abnormalities identified on ultrasound. However, it is plausible that more disorder-specific phenotypic traits could have emerged at more advanced developmental stages. In such cases, fetopsy could also provide valuable insights into disorder-specific manifestations that can often not be assessed due to imaging limitations. Monitoring fetal phenotypic changes throughout pregnancy, neonatally, and post-mortem is crucial, as disorder-specific abnormalities usually manifest as the pregnancy progresses. In this cohort fetopsy was only employed in the fetuses of family 2. Prenatal characteristics identified on ultrasound in this case, were sufficient for the successful diagnosis of congenital myopathy 22B (OMIM #620369) which was assisted by a cohort in which prenatal phenotypic manifestations were reported [45]. Nevertheless, the autopsies conducted on the fetuses in family 2 not only validated the diagnosis but also uncovered valuable additions to the existing phenotypic spectrum of this myopathy, ultimately aiding healthcare providers in its diagnosis [33]. In these cases, where the phenotypic spectrum of a disorder has been expanded, WES failed to reveal any additional candidate variants that would account for the atypical features of the suspected genetic condition. However, the possibility of dual diagnosis cannot be ruled out since additional candidate variants may have been overlooked due to filtering or human error.

The most evident genotype-phenotype associations were observed in families 3 and 5. In family 3, the fetus displayed Intrauterine growth retardation (IUGR), polyhydramnios, hypokinesia, and clenched hands, all of which are features suggestive of myopathy. A candidate variant, classified as likely pathogenic, was found in homozygosity in the gene KLHL40 in the fetus, leading to the diagnosis of nemaline myopathy 8 (OMIM #615348), which is inherited in an autosomal recessive manner. In line with previous discoveries in fetuses with pathogenic variants in KLHL40, which commonly lead to either prenatal or neonatal demise, the fetus in family 3 died prenatally [46,47]. This variant had previously been reported in patients diagnosed with nemaline myopathy 8, providing additional evidence for the diagnosis. In family 5, the fetus exhibited characteristic features of thanatophoric dysplasia type I (TDI) (OMIM #186700), including shortened and bowed long bones, also known as a “telephone-receiver like deformity.” The candidate variant (Ser249Cys) which was identified in the gene FGFR3, is one of five known disease-causing variants associated with thanatophoric dysplasia type I, thereby leading to a diagnosis [36,48].

It is also important to note that initially, the paternally inherited variant in the CPLANE1 gene of family 6 was not reported due to insufficient evidence supporting its pathogenicity. However, upon re-analysis, the variant was reclassified as a VUS with more significant evidence of pathogenicity. This highlights the value of re-analyzing WES data in cases with partial or no findings, as variants of interest may be reclassified as new information emerges.

Potentially causative heterozygous variants have been identified in the genes USP18 and CC2D2A in the fetuses of families 7 and 8 respectively. Although the variants appear to be associated with the corresponding phenotype of each fetus, these genes typically follow an autosomal recessive mode of inheritance, thus a single causative hit would not be sufficient to cause an abnormal phenotype. The presence of a second pathogenic hit within these genes, which has not been identified by array-CGH and WES analysis cannot be excluded. This could be due to unsuccessful variant calling by the analysis platform, a cryptic indel or structural variant or a variant not classified as pathogenic. Unsuccessful variant calling due to inadequate coverage has been excluded, since all the regions within this gene have been sufficiently covered and manually inspected. These cases could be subjected to additional methods of investigation such as targeted gene long-read sequencing and whole genome sequencing techniques for the identification of previously unidentified causative variants.

Importantly, similarly with the results of previous studies, it was observed that the fetuses receiving a WES diagnosis in this cohort exhibited either multisystem or musculoskeletal anomalies [16,49,50]. This study therefore provides additional confirmation that WES analysis of prenatal cases without causative findings from chromosomal analysis and array-CGH is more likely to uncover the genetic cause of pregnancy loss when the fetal phenotype consists of multisystem or skeletal anomalies [50]. Fetuses presenting phenotypic abnormalities isolated to other organ systems could potentially benefit more from targeted forms of genetic testing such as NGS panels.

Challenges and pitfalls of prenatal WES

Overall, studies have shown that WES is a highly valuable tool for postnatal diagnosis and is slowly being incorporated into the prenatal diagnostic setting as well, despite the existing obstacles [24]. A main challenge of prenatal WES is turnaround time (TAT). Although this was not a factor to be taken into consideration in this study since all pregnancies ended in IUD/termination or neonatal death prior to their enrolment, timely WES is essential in prenatal diagnosis to allow time for decision making [49]. Incidental findings relevant to the fetus and parents also complicate the use of WES in prenatal diagnosis. Participants in this study could opt in or out of being informed of such findings. One mother was found to have a known frameshift pathogenic BRCA1 variant (NM_007300.4:c.3018_3021del) relating to breast and ovarian cancer familial 1 (AD), emphasizing the need for strict regulations and consent forms to manage incidental findings and avoid delays in diagnosis.

Prior research in Cyprus shows WES has a 43% diagnostic yield for postnatal patients with multiple malformation syndrome (MMS) [51]. However, in this study, only 22.22% (4/18) of fetuses with multisystem anomalies received a diagnosis, likely due to incomplete phenotypic reporting and imaging limitations. Fetopsies could improve diagnostic yield by revealing traits undetectable via ultrasound. Accurate diagnosis requires detailed prenatal phenotypic reporting, highlighting the need for databases cataloging fetal-specific variants and phenotypes, as prenatal manifestations often differ from postnatal ones.