Congenital anomalies, ranging from relatively minor to severe multi-system anomalies, affect approximately 3 of 100 live births and are responsible for 21% of perinatal deaths (Ely Driscoll, 2020). Identification of the cause of congenital anomalies is important for genetic counseling and clinical management. Ultrasonography, which provides information on the presence or absence of bones, fetal size, shape, and position, is routinely used in prenatal care (Dhamankar et al., 2020; Macedo et al., 2022). However, ultrasound has difficulties in clearly distinguishing between different types of anomalies, resulting in limited prenatal phenotype information.

Typically, fetuses with ultrasonographic anomalies can be associated with all types of genetic variation, such as chromosome aneuploidies, copy number variations, and single-base mutation. Generally, G-band karyotyping that detects chromosome aneuploidies and unbalanced rearrangements (>5–10 Mb) is recommended as the first-line test for prenatal diagnosis with a diagnostic yield of 32% in fetuses with a structural abnormality. In addition, chromosomal microarray analysis (CMA) that identifies smaller microdeletions and duplications is also offered to improve diagnosis by up to 6% over G-banded karyotyping (Wapner et al., 2012). However, only a proportion of fetal anomalies have clear molecular pathogenesis, and the cause of a large number of fetuses with abnormal ultrasound remains elusive, resulting in challenges in prenatal counseling.

Next-generation sequencing (NGS) technology has been widely used for identification of causative genes in various genetic disorders, including intellectual disability (Vrijenhoek et al., 2018), inherited peripheral neuropathies (Hartley et al., 2018), and epilepsy (Symondsand McTague, 2020). Most recently, a clinical guideline of the American College of Medical Genetics and Genomics (ACMG) highly recommended exome and genome sequencing as a first- or second-tier test for patients with developmental delay or intellectual disability (Manickam et al., 2021). Whole-exome sequencing (WES) plays an important role in postnatal diagnosis, while only recently has WES obtained importance in prenatal diagnosis. Carss et al. (2014), who first introduced the WES strategy into prenatal diagnosis in a cohort of 30 non-aneuploid fetuses, identified 35 de novo single-nucleotide variants, small indels, deletions, or duplications. Causative variants were identified in three out of 30 cases with a diagnostic yield of 10%. Several prior studies have examined the use of prenatal WES, where either karyotype analysis or CMA yielded negative results. These studies reported diagnostic yields ranging from 9.1% to 45.9% (Leung et al., 2018; Sparks et al., 2020; Vora et al., 2020; Huang et al., 2023). These findings highlight the significant contribution of WES toward improving the prenatal diagnostic rates.

Despite the numerous advances in WES technology, conclusions drawn from the majority of pilot studies are biased by small cohorts (Cao et al., 2022; Yang et al., 2022) or are limited to highly selected cases or other certain conditions (Liu et al., 2019; Lei et al., 2022). Further pilot studies involving different disorders, conditions, sample types, and analytic methods are necessary to expand our understanding in the application of WES on prenatal diagnosis and to establish the diagnostic yield of WES for all kinds of prenatal disorders.

Here, we present 60 prenatal cases with various ultrasonographic anomalies. All pregnant women underwent invasive testing, either via karyotyping or CMA or both. Cases with negative results of karyotyping or CMA tests were further subjected to WES for causative variant identification, with the aim of discovering more pathogenic mutations and clarifying the utility of WES in prenatal diagnosis.

In this study, WES was used to examine fetuses with abnormal ultrasound findings but could not be diagnosed by routine testing. Our findings demonstrated that WES increased the diagnostic yield in fetuses with abnormal sonographic findings by 23.3%, following negative results obtained through karyotype and CMA testing. These results are consistent with previously published series on published prenatal WES, which have reported diagnostic yields ranging from 15% to 25% (Vora et al., 2017; Vora et al., 2020; Lei et al., 2021). Our results further confirmed the potential of WES as a promising tool for extended prenatal diagnosis. The decreasing turnaround time and increasing accessibility, as well as the falling cost of next-generation sequencing will make it more feasible to use the WES strategy clinically. This will remarkably promote parental counseling and pregnancy management. Nevertheless, conventional testing cannot be replaced by WES due to limitations of WES in detecting large insertions/deletions, chromosomal rearrangements, and mutations in regulatory regions (Alotibi et al., 2023). Karyotype analysis or CMA in combination with WES would be a valuable strategy in prenatal testing.

Among seven categories in this cohort, skeletal anomalies and head or neck anomalies were the top two positive predictors of monogenic disorder. Here, we provided a diagnostic rate by WES of 39.1% of fetuses with abnormalities in the skeletal system, which was comparable to previously published prenatal WES studies with 30.4% (Fu et al., 2022) and 32.7% (Kucińska-Chahwan et al., 2022). However, a significantly higher rate of 75% has been reported elsewhere (Liu et al., 2019; Yang et al., 2019). Skeletal anomalies are a heterogeneous group of disorders affecting bones, cartilage, tendons, and joints (Köhler et al., 2017); a highly heterogeneous phenotype may contribute to the differences in detection rates between these studies. In addition, differences in inclusion criteria, sample size, or the definition of a positive result might also be involved. In any case, a higher diagnostic yield would be obtained from these malformations by WES with optimal cost-effectiveness. Therefore, WES is strongly recommended for fetuses with abnormal ultrasound findings of skeletal anomalies and head or neck anomalies.

With regard to pathogenic genes identified by WES, FGFR3 was one of the most frequently diagnosed genes in the current study that is consistent with previous findings (Han et al., 2020; Zhang et al., 2021). However, different variants of FGFR3 were responsible for different disorders with various phenotypes and prognoses that were unable to distinguish from fetal sonographic indicators (Yang et al., 2019). For instance, FGFR3: c.1138G>A (p.G380R) mutation accounted for 90% of the condition in achondroplasia patients (Bellus et al., 1995), while the FGFR3: c.1620C>A (p.N540K) mutation occurred in 70% of individuals with hypochondroplasia (Bellus et al., 2000). Therefore, WES is of importance for diagnosis and genetic counseling to fetuses with variations in identical genes.

A homozygous nonsense variation c.2041C>T (p.R681*) in P3H1 was identified in case 14 with bilateral femoral bowing. The parent denied consanguinity, and the fetus was eventually terminated because this mutation is implicated in osteogenesis imperfecta. The same mutation has been reported in a British family, in which the fetus exhibited abnormal ultrasound findings of short, thick bowed femurs, evidence of fractures, short humerus, and abnormal shape (Chandler et al., 2018). In a consanguineous family of Arabic descent, this mutation was identified in a fetus with severe limbs and chest deformities (Baldridge et al., 2008). All these lines of evidence suggested that the homozygous non-sense variation c.2041C>T (p.R681*) in P3H1 may be a common variant among different races and usually result in severe fetal malformation.

The PKD2 gene, encoding polycystin-2, is associated with polycystic kidney disease, and the clinical symptoms usually do not appear until adulthood, but the disease starts in utero (Janssens et al., 2021). In one study, the author reported on a family carrying a mutation in the PKD2 gene perinatal death due to polycystic kidney disease occurred in the mother’s second and third pregnancies (Bergmann et al., 2008). In case 17, a mutation in the PKD2 gene was identified in a fetus with enhanced parenchymal echo in both the kidneys. This case was lost to follow-up, while the genetic result suggested a high risk of polycystic kidney dysplasia in adulthood; therefore, advanced interfere treatment is a matter of great urgency.

In case 23, the LP variant of CTNNA3 was identified in the fetus with patent ductus arteriosus, but the transmitting mother was not an affected individual. This is largely due to the incomplete penetrance of this mutation in different affected individuals (van Hengel et al., 2013). A maternally transmitted canonical splice variant c.809 + 1G>A in HNF1B was identified in case 16 with bilateral renal cortical cysts after birth, while the same mutation inherited from the father was also reported in a previous study, in which the fetus was born with a unilateral ureteropelvic junction and renal failure, apart from bilateral renal cortical cysts (Heidet et al., 2010). These findings indicated that the severity of the renal disease resulting from the HNF1B mutation was highly variable. In case 18, the fetus inherited a heterozygous missense variation c.160C>T (p.R54C) in the CRYAA gene from the father, and the parents refused to disclose any information about the newborn. Meanwhile, the father reported no notable ocular disease. Khan et al. (2007) reported that three carriers of c.160C>T (p.R54C) in the CRYAA gene were asymptomatic but had similar bilateral discrete punctuate lenticular opacities evident by a careful slit lamp examination (Khan et al., 2007). Thus, in our study, fetal case 18 was highly suspected of congenital cataract, and we recommended further ophthalmic examination to the transmitted father, but it was not adopted. A sustained follow-up is necessary. To the best of our knowledge, five pathogenic or likely pathogenic variants implicated in SEC24D, FIG4, CTNNA3, EPG5, and PKD2 (Table 2) had never been reported prenatally. These newly identified variants demonstrate how prenatal WES expanded the phenotypic spectrum of monogenic disorders in prenatal diagnosis. In a word, regarding the 14 cases with pathogenic or likely pathogenic genes (Table 2), the pregnancy outcomes further supported the main ultrasound findings only in two cases (cases 15 and 16). However, the concordance of the remaining cases was indeterminate, either because of pregnancy termination or loss to follow-up.

The annotation of VUS in the prenatal setting has been debated for years, and it is essential that VUS results are properly explained to avert potential misuse (Cornthwaite et al., 2022). Here, we identified 22 VUSes (Supplementary Table S3), of which three variants (JAG1:c.1136C>T(p.S379F), NBAS:c.6124A>G (p.M2042V), and DYNC2H1:c.7409C>T (p.A2470V)) had been reported previously at least once. We firmly believe that with the promotion of WES in prenatal diagnosis, more VUSes will be identified and reclassified, allowing a more accurate diagnosis will be possible with integrated genotype–phenotype information.

Although WES has many advantages over the traditional test in prenatal diagnosis, some major challenges have been actively debated as prenatal WES clinically is becoming part of routine testing. The lack of an accurate fetal phenotype achieved by ultrasonography, variable disease descriptions, and appearances of ultrasound findings further complicate the interpretation of the WES results (Best et al., 2018). Other challenges including the cost effectiveness of WES (Aaltio et al., 2022) and even social and ethical issues (Horn and Parker, 2018) are also taken into account. There are several limitations in the current study which need to be further improved and supplemented. First, we performed WES only for the proband instead of parental–fetal trios. Since trio-based WES testing would be helpful for variant filtering in initial data analysis and enable rapid identification of de novo variants. In particular, for WES negative cases, trio-based WES is useful for determining the causative genes. Thus, parental–fetal trio analysis is preferable, although it would increase the cost. Second, this is a retrospective study; some of the pregnancy outcomes were obtained through follow-up phone calls rather than confirmation of diagnosis by a clinician; thus, we are not sure whether the parents have given an objective statement of fact. Finally, the relatively small sample size is still a limitation to our study so that we are unable to identify statistical differences of diagnostic rates between different disorders. Therefore, additional efforts are urgent on large cohort studies to better clarify correlations between the fetal genotype and phenotype and provide refined prenatal counseling.

In conclusion, we confirmed the potential of WES to improve prenatal diagnosis in a cohort of 60 cases. Our results demonstrated that WES remarkably increased the prenatal diagnosis rate by 23.3% in fetuses with abnormal ultrasound findings but normal G-band karyotyping and CMA results. In addition, we suggested that WES may be recommended when conventional prenatal methods fail to provide a diagnosis in fetuses with ultrasonographic anomalies, particularly for disorders involving the skeletal system and head or neck systems. More importantly, this study identified five P/LP variants that had not been included in the HGMD. Additionally, three VUSes that had been reported at least once previously were recurred in this study. All of these findings further add to our current knowledge of phenotype–genotype relationships, which will improve the ability of genetic counselors to advice families.