First applications of a targeted exome sequencing approach in fetuses with ultrasound abnormalities reveals an important fraction of cases with associated gene defects

Introduction

Recognizable fetal ultrasound (US) abnormalities are observed in approximately 3–5% of all pregnancies, while congenital abnormalities account for 20–25% of perinatal deaths (Centers for Disease Control and Prevention, 2008; Romosan et al., 2009). In everyday clinical practice, prenatal detection and prevention of severe congenital disorders is typically achieved through various levels of fetal ultrasonographic examination in the 1st or 2nd trimester of pregnancy; however, precise diagnosis of the underlying genetic defect is often challenging and elusive due to considerable clinical and genetic heterogeneity (Rice et al., 2011; Ermito et al., 2009; Schramm et al., 2009), while an exasperating factor is the lack in most cases of prior family history or other identifiable predisposing risks (Long & Sprigg, 1998).

In these cases, advances in molecular genetics nowadays offer a more detailed prenatal genetic investigation, mainly for chromosomal abnormalities revealed through array comparative genomic hybridization (aCGH), although the diagnostic yield from these studies typically does not exceed ∼10–15% (Konialis & Pangalos, 2015; Hillman et al., 2013), leaving a considerable residual risk for a potentially debilitating genetic disorder. Furthermore, genetic testing for gene mutations associated with the observed ultrasound abnormalities is limited to targeted testing of very few distinct genes suspected to be involved in the particular case and therefore comprehensive testing for gene disorders in the course of pregnancy has been highly selective and thus largely ineffective (Milunsky & Milunsky, 2015).

As an example, skeletal malformations and dysplasias, often detected through routine prenatal ultrasound examination, constitute a phenotypic finding in more than 300 syndromic and non-syndromic disorders and their genetic etiology may be due to a wide variety of genetic aberrations, ranging from copy number variations to single gene mutations. Similarly, fetal limb deformities, such as structural abnormalities of the fingers and toes (clinodactyly, polydactyly, etc.), are relatively frequent ultrasonographic findings and are also associated with a variety of syndromic and non-syndromic genetic disorders (Warman et al., 2011; Porter & Herman, 2011). The above may be expressed as dominant, recessive, or X-linked disorders or they may be the result of spontaneous de novo mutations, while many forms present with variable phenotypic expression and are more difficult to diagnose compared to lethal forms. Therefore, it is more than evident that fetal structural malformations and other fetal abnormalities revealed through prenatal ultrasound are characterized by considerable phenotypic and genetic heterogeneity. For example, single gene such as DHCR7, may be associated with a variety of ultrasound findings, which may include skeletal malformations and/or genitourinary abnormalities (e.g., hypospadias), typically associated with Smith-Lemli-Opitz syndrome. Similarly, other genes involved together with DHCR7 in the cholesterol biosynthesis pathway, are also associated with a number of genetic disorders/syndromes (e.g., Antley-Bixler syndrome—POR gene), presenting with several common prenatal clinical manifestations (Porter & Herman, 2011).

As a rule, abnormal ultrasound findings require detailed genetic counseling, where the merits and limitations of available genetic testing options are discussed as well as the associated risks. However, an accurate diagnosis is a highly desirable prerequisite in this process, not only for parental decisions regarding the outcome of the current pregnancy but also for providing accurate counseling for the risk for future pregnancies and this is even more important in families with recurrent fetal ultrasonographic findings, pointing to an underlying inherited genetic disease.

Very recently, a rather limited number of retrospective studies, utilizing whole genome sequencing (Talkowski et al., 2012) or whole exome sequencing (WES) in small cohorts of neonates and aborted fetuses with various ultrasound and/or post-mortem structural abnormalities (Carss et al., 2014; Drury et al., 2015; Alamillo et al., 2015; Westerfield et al., 2015), provided initial proof-of-principle of large-scale prenatal next generation sequencing (NGS). The results from these studies have been the subject of a recent review and debate (Filges & Friedman, 2015; Chitty, Friedman & Langlois, 2016), where both the merits and limitations of WES applications in malformed fetuses are presented. In addition, it appears that the majority of couples have a positive attitude towards prenatal WES testing, particularly when confronted with troubling fetal ultrasound findings (Kalynchuk et al., 2015).

We present our initial findings and follow-up results, derived from a custom designed targeted exome sequencing strategy, as applied prospectively (on-going pregnancy) and retrospectively (abortuses) in a cohort of 14 euploid fetuses, all presenting with various ultrasonographic anomalies. This approach may afford a timely diagnosis in the course of pregnancy, while overcoming many of the pitfalls associated with large–scale prenatal NGS. We also discuss the potential benefits, the challenges and future developments of this testing strategy, through our central view of maintaining a careful and desirable balance between an increase in diagnostic potential and the undesirable ‘grey zones’ in prenatal diagnosis.

Materials and Methods

Results

Prenatal cases were referred and considered for testing based on the type and severity of the reported fetal ultrasound findings or other fetal abnormalities observed post-mortem.

Specifically for on-going pregnancies, during pre-test counseling and prior to providing their informed consent, couples were informed that testing is targeted, that it will not reveal or report incidental findings and that it will only evaluate and report variants/mutations included in the aforementioned categories (see Materials and Methods).

In order to be able to provide a timely diagnosis in on-going pregnancies, we evaluated and subsequently applied the use of the remaining DNA sample, extracted directly from uncultured amniotic fluid or CVS samples, following previous prenatal aCGH testing. As the Ion Ampliseq Exome RDY reagent typically requires no more than 100–130 ng of input DNA, this type of library preparation coupled to MPS on the Ion Proton PI chip afforded a rapid protocol, not necessitating cell culture and able to deliver interpretable results in less than 10 days, including confirmatory analysis. All 14 fetal samples yielded whole exome libraries within acceptable quality criteria. The technical analysis parameters from each sample are provided in Table S2. On average, a total of approximately 2,200 ± 50 variants per sample/case were detected corresponding to the 758 genes (∼1,030 variants occurring in exons, 1–3 variants in splice sites); following gene prioritization (based on the reported ultrasonographic abnormalities for each case) and variant filtering through the Fetalis algorithm, approximately 1–3 potentially significant variants/mutations were left for manual inspection and final clinical evaluation, with no incidental findings. As an example, applying the Fetalis algorithm in Case 7 (see Table 1), a total of 243 exonic and splice-site variants were identified in 134 genes (out of the 758 Fetalis genes), prioritized due to their association with genetic disorders which may possibly include abnormalities of the cerebellar vermis and/or with Dandy-Walker malformation. Further filtering and prioritization through the EMA-Fetalis built-in parameters for: (a) variant frequency against our local Greek variant database, (b) for variant frequency against international databases (1000 GP, EVS, ExAc) and (c) for pathogenicity score and variant class, resulted in 2 potentially significant missense variants/mutations requiring final manual curation and clinical evaluation (c.1375C > T–p.R459C in the CHD7 gene and the c.1208C > A–p.S403Y in the ZIC1 gene). The CHD7 gene missense mutation was quickly dismissed based on several factors (e.g., disease-phenotype association and relatively common allele frequency in ExAc), while the maternally-inherited ZIC1 missense variant was at the time reported as a VOUS (see note added in proof).

Even allowing for Sanger sequencing confirmation and segregation analysis, final results were available in less than 10 days, with the potential of attaining an even faster turn-around time (<5 days).

The Fetalis targeted (758 genes) exome sequencing strategy provided overall a clinical diagnosis in 5 of 14 cases (36%, Table 1, cases 1–5), involving 3 products of abortion (Cohort 1, cases 1–3) and 2 on-going pregnancies (Cohort 2, cases 4 and 5). In one further on-going pregnancy case (Table 1, case 6), a highly-likely diagnosis of PROKR2-related autosomal dominant Kallmann syndrome was made, involving a known pathogenic mutation. Also, in another on-going pregnancy (Table 1, case 7), a novel probably pathogenic missense variant/mutation was detected in the ZIC1 gene, which was maternally inherited and was reported as a variant of unknown clinical significance (VOUS). In the remaining 7 on-going-pregnancies (Table 1, cases 8–13), Fetalis testing did not reveal a known or ‘obligatory’ pathogenic mutation (see Materials and Methods) in the fetus, possibly associated with the reported ultrasound abnormalities.

Discussion

The identification of the underlying genetic cause in fetuses with US abnormalities, especially in non-consanguineous families, is a challenging task and prenatal genetic diagnosis is typically limited to the investigation of possible chromosomal imbalances, either through conventional fetal karyotype analysis or, more recently, through aCGH. However, as aCGH will only uncover the underlying genetic cause in <15% of these cases (Konialis & Pangalos, 2015; Hillman et al., 2013), an accurate diagnosis is rarely achieved and the necessary advice regarding the precise risks associated with the current pregnancy as well as recurrence risk in subsequent pregnancies is elusive and relies on empirical knowledge. Therefore, the decision to terminate the pregnancy is based solely on the type of fetal ultrasound findings and how the associated risks for a possible severe disorder are perceived by the couple.

In this report we describe the first application of Fetalis, a targeted 758 genes exome sequencing approach, in three products of abortion (Cohort 1) and prospectively in 11 on-going pregnancies (Cohort 2), all with diverse abnormal ultrasound findings and of a troubling but non-extreme type, often encountered in routine prenatal diagnosis. Notwithstanding known and unavoidable technical limitations inherent to NGS testing (e.g., incomplete coverage), a highly-likely or definitive diagnosis was achieved in all three abortuses and in three out of 11 on-going pregnancies, resulting in an overall diagnosis rate of ∼36–43%. Although this figure is obviously a very preliminary assessment from a limited number of cases, it nonetheless provides an initial proof-of-principle regarding the merits of the approach described herein. Most importantly, follow-up data of apparently healthy newborns, born out of on-going pregnancy cases where Fetalis was performed as part of prenatal risk assessment, confirmed the high degree of reassurance provided from the negative results of the test.

There are several important issues we wish to address and discuss, regarding the possible clinical use of the Fetalis testing strategy, as described in this report. Although there is no doubt that WES in fetuses with ultrasound abnormalities is in principle expected to provide a considerable increase in our current prenatal diagnostic capabilities, albeit with many caveats (Chitty, Friedman & Langlois, 2016), the limited data from the few recent studies (Carss et al., 2014; Drury et al., 2015; Alamillo et al., 2015; Westerfield et al., 2015) and from our own initial efforts provide valuable insights to several important limitations regarding its clinical utility in a prenatal setting. Firstly, interpretation of the results following wide scale WES or WGS testing is hampered by the sheer complexity of data analysis and the concomitant difficulties and time required for variant prioritization and final clinical evaluation. Secondly, analysis of WES data will inevitably lead to the uncovering of a large number of incidental findings, unrelated to the reported US findings, leading to serious counseling and ethical issues. The third and most important point is related to the clinical evaluation of numerous WES variants of unknown significance, a very common occurrence in these types of studies.

The targeted exome testing strategy described in this report overcomes many of these limitations and concerns. The volume of data (variants) derived from the 758 genes is an order of magnitude less than the corresponding WES data and this fact, coupled to the rapid laboratory protocol and the custom-developed EMA-Fetalis prioritization pipeline, affords a highly cost-effective, more simplified and timely diagnosis (even <1 week) during the course of pregnancy. In addition, the decision to clinically evaluate and report in an on-going pregnancy only known pathogenic variants or ‘obligatory’ pathogenic variants (see Materials and Methods) reduces dramatically the number of incidental findings (see Results) and the reporting of dubious variants of unknown significance (VOUS), thus simplifying both pre- and post-test genetic counseling. All the above become even more important, if not paramount, when investigating fetuses with less severe US abnormalities, a common troubling occurrence and request in prenatal diagnosis. Although one may argue that the primary purpose and value of prenatal exome sequencing lies in the investigation of highly pathological US anomalies (Filges & Friedman, 2015), in everyday clinical practice this is not common as both the obstetrician and the parents wish to ascertain whether a less severe US finding (e.g., ambiguous genitalia) may only be ‘the tip of the iceberg,’ hiding underneath other much more serious phenotypic features, not yet apparent through US examination of the developing fetus. In such a case, through precise knowledge of the diseases excluded following our targeted testing approach, a negative test result provides a highly desirable reassurance.

To further illustrate the aforementioned points, one could perhaps compare our approach to the application of a targeted aCGH, as opposed to a high-resolution aCGH, in prenatal diagnosis (Konialis & Pangalos, 2015; Ahn et al., 2014; Oneda et al., 2014; Ganesamoorthy et al., 2013). The Fetalis targeted exome sequencing strategy could be viewed as a targeted aCGH, focusing on regions of known pathogenicity, similar to the implementation of ‘focused’ lower resolution prenatal aCGH (Ahn et al., 2014). Although it may miss certain abnormalities, possibly picked-up by a more in-depth WES analysis, it has the advantage of avoiding: (a) VOUS, (b) extensive family studies or costly WES-trio analysis, (c) complicated genetic counseling and perhaps these drawbacks cannot be offset by the possibly higher detection rate.

Finally, the Fetalis testing strategy is flexible and its diagnostic yield will definitely increase, without compromising clarity. Although the Fetalis pipeline is currently targeting and evaluating variants detected only in the 758 genes, the initial step in the analysis involves construction of a whole exome (WES) library and therefore whole-exome variant data are readily available. Hence, as we gain more detailed knowledge, the pipeline may continuously incorporate new genes and pathogenic variants, identified through postnatal WES investigations in severely affected newborns and children.

Conclusions

There is little doubt that prenatal exome sequencing has the potential of becoming soon a routine diagnostic tool. However, as it is applied in a very sensitive and vulnerable setting, we should take the necessary care and precautions so that we do not extend its limits to the point where it might become more of a problem and less relevant and focused to the questions it was set out to resolve.

To our knowledge, this is the first report exploiting the diagnostic potential of a novel targeted and well-defined exome sequencing strategy, which may be deployed rapidly in a clinical prenatal setting and particularly in the course of pregnancy, with minimal ambiguous results and incidental findings. Although the number of cases presented in this report is too small to afford an accurate figure relating to the diagnostic yield, to this end we have initiated an extensive collaborative study with the aim of resolving questions relating to the diagnostic potential of our approach across different categories of US abnormalities, such as heart malformations, skeletal dysplasias, etc. We believe that the overall Fetalis strategy and the initial data presented in this report provide a sound, affordable and encouraging basis for routine clinical implementation.

NOTE ADDED IN PROOF: Since the submission of the manuscript, new developments have emerged regarding Case 7 (see Results and Table 1). The couple and their attending obstetrician informed us recently that a new (third) pregnancy is currently at the 23rd week of gestation, with identical fetal ultrasound brain MRI anomalies as the previous two pregnancies and requested clinical genetic evaluation. Suspecting a possible involvement of the maternal ZIC1 gene variant, we suggested an immediate brain MRI examination of the mother, where it was revealed that she presents with the same structural brain malformations detected in all three fetuses. We subsequently performed a first clinical genetic evaluation of the mother, where a mild Crouzon syndrome-like phenotype was noted, with mild dysmorphic features and irregular head shape. At the time of the initial Fetalis results assessment of Case 7 in May 2015, there had been no published reports directly linking ZIC1 gene variants to a recognizable pathological phenotype and thus, in accordance with our testing and reporting guidelines, the maternally inherited ZIC1 gene variant was reported as a VOUS. However, a recent publication (Twigg et al., 2015) has now established an association between ZIC1 gene mutations and coronal craniosynostosis accompanied with structural brain malformations, very similar to those observed in all three fetuses and the mother of Case 7. Further to this recent report, a more detailed clinical evaluation of the mother and of the current pregnancy is under way. We believe that this recent development provides further supporting evidence regarding the specificity and clinical validity of our approach described herein.

Supplemental Information