Cohort characteristics

Trio exome sequencing was performed in 1618 fetuses. At testing time, the median maternal age was 29 years (range 18–47), and the median gestational age was 25 weeks (range 11–35). Based on the systems involved, fetal malformations were classified into 8 categories (central nervous, facial, chest, cardiovascular, abdominal, urogenital, skeletal, and multisystem). Isolated hydrops fetalis, fetal growth retardation (FGR), and increased NT were classified into separate categories. For fetuses with increased NT, if a new phenotype appeared in late pregnancy, they would be reclassified into the corresponding malformation categories. The most frequently affected organ referred for prenatal ES was the cardiovascular system (17.9%), followed by the central nervous (16.2%), skeletal (12.8%), and urogenital (11.4%) systems, altogether comprising more than half of all cases. The detailed clinical characteristics of the fetal cases are summarized in Additional file 2: Table S1. The turn-around time of prenatal ES was 4–10 weeks in the retrospective cohort and 1–4 weeks in the prospective cohort.

Positive diagnostic results and data reanalysis

In genotype-centric analysis step 1, P/LP variants were identified in 187 cases, representing an 11.6% diagnostic rate (187/1618). VUSes were identified in 55 cases (3.4%, 55/1618). Step 2 shows phenotype-driven analysis resulted in additional positive diagnostic variants in 28 cases (1.7%, 28/1618). In addition, VUSes potentially related to phenotypes were identified in 68 cases (4.2%, 68/1618).

In step 3 for data reanalysis, 295 cases obtained additional new phenotypes during the prenatal or perinatal period. One case received upgrades from inconclusive to positive based on further phenotypic information. After reanalysis, P/LP variants previously interpreted as incidental findings in steps 1 and 2 in 7 cases (0.4%, 7/1618) were reclassified as disease-related based on additional phenotypes in late pregnancy. The detailed information of fetuses with diagnostic or VUS results with new phenotypes is summarized in Additional file 2: Table S2. In addition, 3 cases (0.2%, 3/1618) were upgraded from negative to P/LP due to a new disease gene identified in reanalysis upon special clinical requests, 2 cases with intragenic copy number variants, and 1 case with a second allele identified by Sanger sequencing were revealed by reanalysis. Altogether, data reanalysis yielded an increased diagnostic rate of 0.9% (14/1618), of which 50.0% (7/14) attributed to new phenotypes reclassified from IFs to positive diagnoses, and 21.4% (3/14) attributed to new disease genes identified. In addition, eight cases were upgraded from negative to inconclusive based on the newly provided phenotype, yielding a VUS rate of 0.5% (8/1618). The overall positive diagnostic and VUS rates of each analysis step are summarized in Table 1.

Table 1 The overall positive diagnostic and inconclusive rates in each analysis step

In total, 253 different variants across 135 unique genes were identified as positive diagnoses in 229 fetal cases (Additional file 2: Table S3), with an overall diagnostic rate of 14.2% (229/1618). Of these, 98 diagnostic variants (38.7%, 98/253) were not previously reported. Of the 135 genes identified, 26 (19.3%) were revealed to expand the previously reported fetal phenotype spectrum; 11 were outside the list in the PAGE study [8]. Twenty-seven genes (20.0%, 27/135) were reported in prenatal cases for the first time, of which potential fetal phenotype expansion was identified in 13 genes (Table 2). Among the positive diagnostic cases, 172 (75.1%), 40 (17.5%), and 17 (7.4%) were associated with autosomal dominant, recessive, and X-linked disorders, respectively (Table 3). In our cohort, 134 cases had a family history record with prior affected pregnancies or relatives, including 100 with similar phenotypes (significant family history) and 34 with different phenotypes. The diagnostic rate was 48.0% (48/100) in cases with a significant family history, significantly higher than that for sporadic cases (11.8%, p<0.01).

Table 2 The 27 genes firstly reported in prenatal casesTable 3 Inheritance patterns of positive diagnostic and VUS cases

The diagnostic rate in the retrospective cohort was significantly higher than that in the prospective cohort (17.3% vs. 12.4%, p<0.01). Significantly higher diagnostic rates were obtained in fetuses with abnormalities in skeletal systems (30.4%) and multiple organ systems (25.9%) than in other subgroups (p< 0.05) (Fig. 2, Additional file 2: Table S4). No molecular diagnosis was made by pES for the 46 fetuses with chest malformations.

Fig. 2

Diagnostic rates based on malformation classification. The highest diagnostic rates were obtained in fetuses with multiple organ and skeletal anomalies and the lowest in fetuses with chest anomalies

Among the diagnosed cases, genes associated with neurodevelopmental disorders were found in all patients with central nervous anomalies (n=33). Positive cases with multisystem or cardiovascular abnormalities showed significantly higher probability (85.7% and 86.5%) of harboring neurodevelopmental disorder-related variants than cases with skeletal (44.4%), facial (25.0%), and urogenital (6.7%) anomalies (p<0.05).

Genotype-phenotype correlation analysis

In this study, clinical features were converted into standard HPO terms for all cases to facilitate the genotype-phenotype correlation analysis. A disease gene was considered potentially relevant to the fetal anomalies if its associated clinical phenotypes meet one of the following criteria: (1) match HPO entry of the fetal phenotype; (2) match the superclass based on HPO or clinical synopsis in Online Mendelian Inheritance in Man (OMIM) database; (3) be reported in previous cases manifesting the same or similar phenotypes of the fetuses.

Among the 229 diagnostic cases, 195 disease genes (85.2%, 195/229) matched at least one HPO term, and 27 (11.8%, 27/229) fit the superclass based on HPO or clinical synopsis in OMIM, showing atypical phenotypes. Variants in 4 genes (1.7%, 4/229) were initially considered as incidental findings and reclassified as diagnostic results due to similar phenotypes reported in the literature (cases 29, 39, 208, and 229). Three genes (KMT2C, KCNT1, NFIB) identified in fetuses with increased NT have not been reported prenatally (cases 169, 170, and 172).

In the 229 diagnosed cases, 49 had additional new phenotypes during prenatal and/or postnatal periods. P/LP variants identified in 7 fetuses (3.1%, 7/229) were considered as incidental findings based on initial fetal anomalies and reclassified as diagnostic variants due to new phenotypes in data reanalysis (Additional file 2: Table S2). One case (0.4%, 1/229) was upgraded from VUS to LP due to new phenotypes (case 207).

Frequent molecular diagnosis of disease genes

The most frequent diagnostic genes in 4 or more cases were FRFR3 (n=15), COL1A1 (n=12), KMT2D (n=11), COL2A1 (n=6), PTPN11 (n=6), TSC2 (n=6), FGFR2 (n=5), FLNA (n=4), NIPBL (n=4), HNF1B (n=4), and COL1A2 (n=4) (Fig. 3).

Fig. 3

Frequent positive and potential diagnostic genes. The number of cases by disease genes in different malformation categories is shown

Intragenic copy number variants (CNVs)

While gross copy number abnormalities were ruled out by karyotype/CMA for this cohort and CNV analysis was not part of the routine pES testing, focused CNV analysis on highly suspected genes based on phenotypes was performed for selected cases upon clinicians’ requests in the reanalysis step. The potential significant CNVs were confirmed by orthologous methods for a definitive diagnosis. Such assessments resulted in 2 additional positive cases involving intragenic deletions in the ATRX and SHOX genes (0.3%; case 126 and case 162).

Diagnostic rates in relation to NT measurement

In our cohort, NT measurement results were available for 690 cases, including 121 patients with isolated increased NT (≥3.5 mm) without concomitant anomalies (the isolated group) and 569 cases with increased NT (47 with NT ≥3.5 mm, and 522 with NT between 3.0 and 3.4 mm) associated with other structural abnormalities (the associated group; Additional file 2: Table S5). The pES detection rates were 5.0% in the isolated group and 25.5% in the associated group with NT ≥3.5 mm, respectively (p<0.05). For cases with isolated increased NT between 3.5 and 4.9 mm, the total diagnostic rate was 3.8% (3/78). The diagnostic rates increased with increasing NT measurement, although the correlation was not statistically significant (p=0.08). No such correlation was observed in the associated group.

Number of candidate variants analyzed

The numbers of potential diagnostic/candidate variants in step 1 (genotype-driven) and step 2 (phenotype-driven) in 525 representative cases (including all or at least 50 cases from each malformation class) are analyzed and listed in Additional file 2: Table S6. The numbers of candidate variants based on malformation class and overall result category (positive, inconclusive, and negative), are summarized in Additional file 2 Table S7. The average total number of variants closely reviewed per case was 1.7, and the mean number of variants interpreted as irrelevant after a quick review was 34.3. Overall, the negative cases had a significantly lower number of variants analyzed for all 6 statistical indexes, with either candidate variants ruled out after close review or no candidates for review (p<0.01). Moreover, cases with multisystem malformations had the highest number of variants closely reviewed (mean 3.1) and the highest number of variants quickly ruled out (mean 55.3).

The amount of time for reviewing these candidate variant(s) for each case was approximately 15 min on average, ranging from 5 to 30 min for the vast majority of cases.

Inconclusive results

Variants of uncertain significance were detected in 131 fetuses (Additional file 2: Table S8), including 2 cases combined with positive diagnostic variants and 15 with significant family history. ES reanalysis resulted in upgrades from negative to inconclusive for 8 cases due to new phenotypes. Therefore, the inconclusive rate in this study was 8.1% (131/1618). Twenty-four inconclusive cases (18.3%, 24/131) with fetal phenotype consistency and variants predicted to be deleterious or reported in previous cases were considered as high-risk inconclusive, including 5 cases (cases 244, 267, 347, 346, and 357) with prior similarly affected fetuses in each family.

Incidental and secondary findings

Incidental findings with childhood-onset disease gene were revealed in 8 fetuses (0.5%, 8/1618) and secondary findings were detected in 13 fetuses (0.8%, 13/1618) according to ACMG recommended list (Additional file 2: Table S9). All the incidental findings were reported based on consensus between the laboratory and clinicians. Of note, secondary findings with childhood-onset diseases in 3 fetuses (3/13, 23.1%) were also included in the reports.

Candidate genes

Candidate gene analyses focused on de novo etiology and prioritized 33 variants based on the combined considerations of gene function and variant type (1 nonsense, 4 frameshift, and 28 missense changes) in 31 cases (Additional file 2: Table S10). In addition, compound heterozygous missense variants in the ASXL3 gene were identified in a family having 3 children with congenital heart defects [22]. Of these 32 cases, 16 had cardiovascular anomalies, 1 had fetal hydrops, 5 had urogenital anomalies, 4 had skeletal abnormalities, and 2 had central nervous or chest or multisystem anomalies, respectively.

Pregnancy outcomes and assessment of the clinical impact

Pregnancy outcomes were available in 1462 (90.4%) of the 1618 cases. Ninety-five cases were lost to follow-up (5.9%, 95/1618), and 61 were still in pregnancy until February 2022. For the remaining cases, 579 were terminated, 5 were fetal demise, and 878 were live birth, including 19 with neonatal death (Additional file 2: Table S11).

In the retrospective cohort, 98 cases (17.3%, 98/565) obtained molecular diagnosis contributing to recurrence risk assessment and reproductive planning. For the 5 cases with incidental or secondary findings with childhood-onset disease genes in the retrospective cohort, 2 were terminated due to fetal anomalies, and 3 (0.5%, 3/565) were live birth with the ES results implicated for future clinical surveillance and medical management.

In the prospective cohort, clinical impacts were evaluated in all cases (Table 4). For terminated cases with diagnostic findings, 25.7% (27/105) were terminated due to fetal anomalies prior to ES results, and 74.3% (78/105) made the decision based on positive ES results. For diagnostic cases with live birth, the clinical decision of continuation of pregnancy was made in 35.0% of cases (7/20) due to non-neurodevelopmental consequences, 20.0% (4/20) due to inherited from either parent, and 40.0% (8/20) due to both effects, respectively. Overall, positive ES results contributed to clinical decision making on termination (59.5%) or continuation of pregnancy (14.5%) in 97 cases (74.0%, 97/131). Inconclusive results had a predominantly clinical impact in one case (1.2%, 1/81), manifesting increased NT and pleural effusion with decision making on termination (reported in the previous case with intellectual disability and behavior disorder [23]), and 10 cases (12.3%, 10/81) with decision making on the continuation of pregnancy due to inherited from either parent. Incidental findings had a clinical impact in 2 cases (40%, 2/5) with decision making on the continuation of pregnancy due to non-neurodevelopmental consequences and implications for clinical management. Secondary finding with a childhood-onset disease was reported in 1 case, contributing to clinical surveillance and management with live birth. For negative cases, the clinical decision to continue pregnancy was made in 60.2% of cases (502/834) based on negative ES results and severity and treatability of fetal anomalies. In total, ES results showed an overall clinical impact of 61.5% (648/1053) on decision making regarding termination or continuation of pregnancy in the prospective cohort.

Table 4 Clinical impacts of ES results in the prospective cohort