The utility of pathologic examination and comprehensive phenotyping for accurate diagnosis with perinatal exome sequencing

Abstract

Objective:

Exome sequencing (ES) offers the ability to assess for variants in thousands of genes and is particularly useful in the setting of fetal anomalies. However, the ES pipeline relies on a thorough understanding of an individual patient's phenotype, which may be limited in the prenatal setting. Additional pathology evaluations in the pre- and postnatal settings can add phenotypic details important for clearly establishing and characterizing a diagnosis.

Methods:

This is a case series of prenatal ES performed at our institution in which pathology evaluations, including autopsy, dysmorphology examination, histology, and peripheral blood smear, augmented the understanding of the fetal phenotype. ES was performed at our institution and a multidisciplinary panel reviewed and classified the variants for each case.

Results:

We present four cases wherein pathology evaluations were beneficial for supporting a perinatal diagnosis identified with ES. In each of these cases, pathology findings provided additional data to support a more complete understanding of the relationship between the perinatal phenotype and variants identified with ES.

Conclusion:

These cases highlight challenges of perinatal ES related to incomplete prenatal phenotyping, demonstrate the utility of pathology evaluations to support diagnoses identified with ES, and further characterize the disease manifestations of specific genetic variants.

1 ∣. INTRODUCTION

When fetal abnormalities are found on prenatal ultrasound, identification of an underlying genetic diagnosis informs prognosis, perinatal management, and counseling. Next generation sequencing, including exome sequencing (ES), enables the detection of rare genetic conditions and has transformed prenatal diagnosis. ES provides an opportunity to evaluate thousands of genes and is particularly helpful when fetal anomalies are present but the phenotype does not clearly suggest a specific single gene etiology. While conventional genetic testing with karyotype or chromosomal microarray can detect an underlying genetic diagnosis in up to one quarter of fetuses with structural anomalies, ES has demonstrated incremental diagnostic yields above conventional genetic testing ranging from 9% to 29%.^1-5

With prenatal ES, DNA from the affected fetus (proband) is fragmented to create a library and the exome is amplified and sequenced. Bioinformatics pipelines compare the proband's DNA to reference sequences from unaffected individuals; genetic variants are then prioritized using data from affected individuals reported in genomic databases and the medical literature as well as from population databases of unaffected individuals. In silico models can add predictions about the impact of variants on downstream protein function, and variants are prioritized and classified in the context of the proband's phenotype. Human Phenotype Ontology (HPO) terms provide a standardized framework with which to describe phenotypic features for variant filtering algorithms, and are commonly utilized in pediatric and prenatal ES to aid in variant filtering.^6,7 Variants are then classified and reported following American College of Medical Genetics and Genomics guidelines.^8-11 Figure 1 provides an overview of the variant filtration process.

FIGURE 1.

An example of trio exome sequencing variant filtration as performed at our laboratory with estimated number of variants provided. AF, allele frequency; HPO, human phenotype ontology

A limiting factor in prenatal ES is that the fetal phenotype is often incompletely detectable by prenatal ultrasound or magnetic resonance imaging (MRI) due to early gestational age and the focus on structural rather than functional compromise. Further, some fetal features of genetic diseases are poorly understood and may be unique to the prenatal period, such as hydrops fetalis or polyhydramnios. These factors can hinder the complete understanding of genetic variants in relation to presenting fetal features, as well as accurate identification and classification of variants. As a result, utilizing all available data points to develop the most thorough the perinatal phenotype for each case is critical for accurate disease identification. In particular, pathology and postnatal examinations are useful adjuncts that can contribute phenotypic information not detected in utero in order to establish a diagnosis in up to 20% of otherwise unsolved cases.¹² Examples of evaluations that may yield additional useful data are fetal blood smears from percutaneous umbilical cord sampling (PUBS), special stains from tissue biopsies, gross evaluation of external features and dysmorphology, and internal examination of tissues via autopsy.

We present four cases from our fetal sequencing studies at the University of California, San Francisco (UCSF), in which pathology evaluations provided information critical to the understanding of fetal genetic disease. While in some cases, the molecular diagnosis either was or would have been identified prenatally based on fetal features alone, the pathology and other evaluations we present provided additional support for the classification and confidence in these diagnoses. All ES was performed at the UCSF Genomic Medicine Laboratory, with protocols previously described.¹ In brief, sequencing was performed using the Illumina NovaSeq 6000, or HiSeq 2500 Sequencer on rapid run mode. Our bioinformatics pipeline utilized Ingenuity Variant Analysis (Qiagen) before March 2020 and Moon (Diploid, Invitae) beginning in March 2020 for variant filtration. Suspicious variants were reviewed by a multidisciplinary team, represented by maternal–fetal medicine, medical genetics, molecular genetics, pathology, bioinformatics, and other subspecialties. Variants thought to be involved in the proband's presentation were then disclosed in a clinical report. Some of the cases presented here have been previously published, but details of pathology evaluations, specific phenotypes, and variant classification were not extensively discussed in prior publications.¹ This study was approved by the UCSF Institutional Review Board and all participants provided informed written consent.

2 ∣. CASE 1

A 37 year old G1P0 whose pregnancy was conceived via in vitro fertilization with low risk preimplantation genetic testing for aneuploidy underwent chorionic villus sampling for the indication of advanced reproductive age. A normal karyotype and chromosomal microarray resulted. The family history was notable only for heart block in the biological father of the fetus, who required placement of a pacemaker at 3 years of age. At 19 weeks' gestation, the fetus was noted to have echogenic bowel and long bones measuring 1 week behind dates with normal morphology and mineralization. Follow-up ultrasounds showed persistence of the shortened long bones; at 23 weeks' gestation, they were 2–3 weeks behind dates and the patient underwent an amniocentesis. A gene panel for skeletal dysplasias was sent prior to referral of the patient to our institution, which reported a variant of uncertain significance (VUS) in TRPV4. PCR from the amniocentesis was negative for cytomegalovirus, toxoplasmosis, and parvovirus. The patient was then referred to our institution, where ultrasounds between 21 and 26 weeks demonstrated increased cardiothoracic ratio (0.6), hypertrophic and dilated cardiomyopathy, a small pericardial effusion, and persistently shortened long bones.

At 30 weeks, there was persistent pericardial effusion, as well as development of ascites, and worsening of the hypertrophic and dilated cardiomyopathy (Figure 2A,B). The patient was admitted for close monitoring and ultimately underwent cesarean delivery at 30 weeks because of worsening fetal status. The neonate was intubated, developed refractory lactic acidosis, and in light of clinical deterioration, the family chose to withdraw care. Autopsy showed cardiomegaly with cardiac weight nearly double the expected value, dilation and hypertrophy of the right atrium and right ventricle, hypertrophy of the left ventricle, moderate narrowing of the ductus arteriosus, pleural effusions, moderate lung hypoplasia, shortened long bones, and anasarca. Electron microscopy demonstrated biventricular diffuse cardiomyocyte mitochondrial accumulation with mitochondrial swelling and focal subendocardial myocytolysis.

FIGURE 2.

Ultrasound images for case 1. (A) Hypertrophic and dilated cardiomyopathy seen on third trimester prenatal ultrasound. (B) Fetal ascites seen on third trimester prenatal ultrasound

Trio ES performed post-mortem on the banked DNA sample obtained from amniocentesis revealed two variants in ACAD9: a maternally inherited missense variant (NM_014049.4:c.796C>T:p. Arg266Trp) and a de novo frameshift variant (NM_014049.4:c.1109delC:p.Pro370fs*13). ACAD9 encodes a member of the acyl-CoA dehydrogenase family, which catalyzes the rate-limiting step in beta-oxidation of fatty acyl-CoA. Autosomal recessive phenotypes associated with ACAD9 variants include acyl-Coenzyme dehydrogenase 9 deficiency, mitochondrial complex I deficiency, and cardiomyopathy. The pathologist who performed the postmortem examinations confirmed that the cardiomyopathy and mitochondrial accumulations strongly fit the phenotype expected from disease-causing ACAD9 variants, and this phenotypic data was considered in the classification of the ACAD9 variants. The maternally inherited missense variant was a rare amino acid change that had been reported 4 times in gnomAD (4 of >250,000 alleles), was predicted to be damaging by in silico prediction models, had a high Combined Annotation Dependent Depletion (CADD) score of 34 (likely deleterious), and had been previously reported as pathogenic in a 12 month old with compound heterozygous ACAD9 variants, cardiac hypertrophy, and severe Complex I deficiency. As a result, this variant was classified as pathogenic. The de novo frameshift variant was predicted to result in truncation in the middle of the protein with loss of function and was not present in large population databases including gnomAD, ExAC, and 1000 genomes. This variant was also classified as pathogenic. The previously identified TRPV4 VUS was ultimately determined to be noncontributory.

This case illustrates the importance of understanding the full perinatal phenotype and not limiting the genetic evaluation unnecessarily to a small set of genes. ES was deferred until the postnatal setting for this case per patient preference, and these ACAD9 variants may have been identified if ES had been pursued prior to delivery. However, the pathology findings allowed us to confidently counsel the family that the ACAD9 gene variants explained the severe fetal presentation.

3 ∣. CASE 2

A 29 year old G4P2012 conceived spontaneously and her pregnancy was initially uncomplicated with normal integrated screening, nuchal translucency (NT) measurement, and anatomical survey. However, at 33 weeks' gestation, she developed contractions and the ultrasound demonstrated pleural effusions, skin edema, absent stomach, hepatomegaly, clenched hands, polyhydramnios with amniotic fluid index of >50 cm, and abdominal circumference measuring <2 percentile. She underwent cesarean delivery at 34 weeks due to evidence of fetal distress, and the neonate passed away shortly after birth. Microarray performed postnatally was negative and trio ES based on the prenatal phenotype was inconclusive. Autopsy was notable for numerous additional phenotypic features: bilaterally low set ears, short upturned nose with anteverted nares, mild retrognathia, high arched palate, pectus excavatum, hypoplastic nipples, contractures of both elbows, mild clinodactyly of the fifth fingers, mild bowing of the legs, hypoplastic heart, adrenal hypoplasia, undescended testes, right hydronephrosis, and fetal liver with ongoing hematopoiesis.

The HPO terms were expanded using information from the autopsy, and reanalysis of the ES led to a de novo heterozygous missense variant in ACTA1 (NM_001100.3:c.521C>G: p.Pro174Arg). ACTA1 encodes skeletal muscle alpha-actin, which forms the core of the thin filament of the sarcomere. Variants in this gene are associated with autosomal dominant congenital myopathies, and de novo missense variants in ACTA1 had been reported with a fetal akinesia phenotype. This specific variant had been reported once in ClinVar as a VUS, but a different amino acid substitution in the same position within this conserved region was reported as likely pathogenic and nearby missense variants had been reported with nemaline myopathy. Further, in silico models predicted this variant to be deleterious, and this variant was rare and not reported in 1000 Genomes, ExAC, or gnomAD. Given the strong gene-disease and moderate variant-disease association, this variant was classified as likely pathogenic.

Postnatally identified contractures and other features on autopsy were integral in this case for solidifying the relationship between the ACTA1 variant with the perinatal presentation. The expertise of the pathologist performing the autopsy was particularly useful, as the postnatal features added further support for the diagnosis of congenital myopathy.

4 ∣. CASE 3

A 40 year old G2P1001 conceived spontaneously. Her pregnancy was initially uncomplicated, including low risk integrated serum screening results, normal NT, low risk cfDNA screening, and normal anatomic survey. She underwent a growth ultrasound for the indication of advanced reproductive age at 32 weeks' gestation, at which time mega cisterna magna and prominent cavum vergae were noted. A fetal MRI was performed, demonstrating multiple intracranial abnormalities including global cerebral volume loss, markedly diminished white matter, bilateral ventriculomegaly, hypoplasia of the corpus callosum, germinolytic cysts in the ganglionic eminences, and abnormal posterior fossa structures. Maternal serologies for infectious etiologies were negative (including parvovirus, cytomegalovirus, syphilis, toxoplasmosis, and herpes simplex virus). The patient underwent induction termination. Neuropathology evaluation at the time of autopsy was notable for multifocal astrogliosis involving the subcortical white matter of frontal and temporal lobes, multifocal nodules with focal calcification involving the subcortical white matter of these lobes, and partial loss of neurons in the cerebral cortex, Purkinje cell layer, and internal granular layer of the cerebellum. Postnatal chromosomal microarray was negative, and trio ES was performed following termination of pregnancy.

ES identified a de novo heterozygous missense variant in TUBB2A (NM_001069.2:c.1072C>A: p.Pro358Thr). TUBB2A encodes beta tubulin, a subunit of the microtubule, and is important in mitosis, intracellular transport, neuron migration during brain development, and ciliary and flagellar motility. Disorders in microtubule subunits, or tubulinopathies, have been associated with numerous complex brain malformations. Variants in TUBB2A have been associated with an autosomal dominant syndrome notable for cortical dysplasia and other brain malformations, seizures, and intellectual disability. This particular variant had not been previously reported, either in affected populations or in large population genome databases. However, given the detailed phenotype available and strong gene-disease correlation with a de novo gene variant, this variant was classified as likely pathogenic.

For this case, additional central nervous system anomalies were identified on fetal MRI, providing a useful adjunct to prenatal ultrasound. Further, the findings on postnatal neuropathology examination added important details for understanding the consequences of the tubulinopathy and counseling the family regarding certainty of the diagnosis.

5 ∣. CASE 4

A 39 year old G3P0020 had a spontaneous conception following 2 prior spontaneous first trimester losses. She had normal expanded carrier screening results and products of conception from the second miscarriage confirmed trisomy 21. At 14 weeks' gestation, during the index pregnancy, a cystic hygroma was noted. She later underwent amniocentesis and had normal results of karyotype, chromosomal microarray, and a RASopathy panel. At 19 weeks' gestation, there was evidence of persistent cystic hygroma and the phenotype further evolved to include bilateral pleural effusions, pericardial effusion, ascites, and skin edema. There was no evidence of alloimmunization or hemoglobinopathy based on maternal antibody testing and hemoglobin electrophoresis. Elevated middle cerebral artery Dopplers were noted at 20 weeks, measuring 1.6 multiples of the median. The patient subsequently underwent PUBS that confirmed fetal anemia with an opening hemoglobin of 4.9 g/dl. Due to persistent evidence of fetal anemia across gestation, the patient subsequently underwent six intrauterine transfusion procedures. Amniotic fluid was sent from these procedures and PCR was negative for CMV, parvovirus, and toxoplasmosis. A sample was also sent for trio ES and the results returned prior to delivery.

ES identified a de novo missense variant in PIEZO1 (NM_001142864.2:c.1792G>A: p.Val598Met). PIEZO1 encodes a regulator of electrical currents along the plasma membrane. Gain of function variants in this gene have been associated with dehydrated hereditary stomatocytosis, also known as hereditary xerocytosis, an autosomal dominant condition characterized by hemolytic anemia due to erythrocyte dehydration. Loss of function variants in PIEZO1, on the other hand, have been associated with an autosomal recessive condition known as lymphatic malformation 6. This condition has a strong association with nonimmune hydrops fetalis and generalized lymphatic dysplasia. As the variant detected in this case was a missense variant, downstream effect on protein function could not be assumed. However, this particular variant had been reported once prior in association with hereditary stomatocytosis. A different amino acid change in the same position had also previously been reported with hereditary stomatocytosis, and this variant was rare and not reported in large population sequencing projects. The CADD score was somewhat elevated at 23.5. Importantly, communication with the pathologist who reviewed the fetal blood smears from the PUBS procedures further revealed that the smears were notable for autoagglutination, spherocyte formation, polychromatophilia, and were overall suspicious for a red blood cell membrane disorder that affected the electrical currents in the cells. Taking all evidence into account (namely the de novo inheritance, absence from population databases, and previous reports in a patient with a similar genetic disease), the variant itself was deemed likely pathogenic.

The PIEZO1 variant in this case was identified prenatally, and communication with the pathologist along with additional data from the fetal blood smear provided further support for the diagnosis of dehydrated hereditary stomatocytosis. This highlights the utility of adjunctive evaluations and collaboration with a multidisciplinary team.

6 ∣. DISCUSSION

These cases demonstrate the important roles that pathology as well as both pre- and postnatal phenotyping play in the interpretation of ES results for cases of fetal anomalies. In each of these cases, additional features identified with perinatal pathology examinations augmented the prenatal findings and strengthened gene–phenotype relationships, particularly when variants were novel or underreported. While the genetic variants reported in these cases either were or may have been identified based on the prenatal phenotype alone, these cases highlight the value of pathology and complete perinatal phenotyping for disease characterization, accurate classification of variants with ES, and comprehensive counseling. Additionally, they highlight the utility of a multidisciplinary approach to the evaluation of fetal disease as well as the importance of considering a broad differential rather than limiting evaluations unnecessarily to a small panel of genes.

Because ES evaluates thousands of genes simultaneously and generates data on tens of thousands of variants that are benign, of uncertain significance, or pathogenic, phenotypic information is critical for prioritizing these variants. However, prenatal diagnosis has unique limitations. One is limited ability to assess the fetal phenotype with ultrasound and MRI technologies, particularly at earlier gestational ages. In studies evaluating the concordance of prenatal ultrasound and autopsy, autopsy provided additional information in 22%–58% of cases and changed the initial diagnosis in 1%–33% of cases.^13,14 In studies assessing the concordance of prenatal MRI and postnatal diagnoses made with postnatal imaging or autopsy, prenatal MRI missed relevant findings in 4%–37% of cases with variation by organ system.^15-18 Further, one study reanalyzed prenatal cases of anomalies that had normal results of ES based on the prenatal phenotype alone, and found a missed genetic diagnosis in 20% of cases when the postnatal or pediatric phenotype was incorporated into the sequencing analysis.¹² There are undoubtedly many opportunities to improve our phenotyping abilities with prenatal imaging, although some limitations may be unavoidable, such as early stage of a developing fetus. Over time, improved phenotyping through increasing utilization of first trimester anatomy ultrasound, 3D ultrasound, fetal MRI, and focused assessments for dysmorphic features and abnormal functions, such as hypotonia, are likely to increase the diagnostic yield of prenatal ES.

Beyond the accuracy of prenatal imaging, there are additional limitations inherent to prenatal phenotyping. Certain phenotypes may not yet have developed or may not be evident in the prenatal period, such as seizures, hypotonia, or neurodevelopmental delays. Subtle facial and other dysmorphologies may be difficult to appreciate in utero. While fetal blood can be assessed at the time of PUBS, tissue biopsies are generally not available in utero, leaving important clues undetected that would be seen with postnatal pathology such as abnormal-appearing mitochondria. In addition, many childhood and adult genetic diseases have not been well described in utero, and for many, the prenatal phenotype may not match the generally accepted postnatal phenotype. As prenatal sequencing and phenotyping become more commonly performed, we will begin to understand not only the phenotypic features of genetic diseases that are unique to the fetal period but also the variants capable of leading to severe in utero presentations of diseases. For these reasons, a multidisciplinary team approach, postnatal follow-up, pathology evaluations, and potential reanalysis of ES data are important to consider to ensure accurate results of ES and to further characterize the perinatal phenotypes associated with specific genetic variants.

Additionally important to consider are accuracy and consistency in descriptions of phenotypic features that are used for variant filtering algorithms. HPO terms offer a standardized vocabulary of phenotypic abnormalities, including that of clinical manifestations as well as imaging and pathology findings. These terms are refined to a specific level of detail, such as muscular versus perimembranous versus restrictive ventricular septal defect. Use of HPO terms has been shown to increase the diagnostic yield of ES, and HPO terms are utilized for all cases undergoing ES at our institution.^20,21 However, as fetal features of genetic diseases are increasingly understood, prenatal imaging abilities are rapidly improving, and genetic variants leading to fetal disease are continuously being discovered, there is an urgent need for adding and refining existing HPO terms to improve the descriptions of fetal phenotypes and thus the diagnostic yield of prenatal ES. Additionally, it is important to consider who assigns HPO terms and how much phenotypic information is provided to the team performing variant curation and analysis. At our institution, natural language processing is performed on clinician notes and imaging reports to generate initial HPO terms. The clinical team of maternal–fetal medicine and medical genetics subspecialists works directly with the molecular genetics laboratory team to provide additional HPO terms that may not be captured through this process and refine the ultimate list of terms that are utilized.

In summary, perinatal ES is a powerful tool for pregnancies complicated by fetal anomalies. However, there are limitations to prenatal phenotyping, knowledge of fetal features of genetic diseases, and characterization of perinatal disease manifestations associated with specific genetic variants. Multidisciplinary team approaches, pathology evaluations, and postnatal phenotyping are essential components that may either support a prenatally identified genetic diagnosis or lead to the discovery of a diagnosis that would otherwise have been missed. Future studies should focus on improving and standardizing prenatal phenotyping, elucidating the unique fetal phenotypes of genetic diseases, and uncovering the genetic variants capable of leading to in utero disease in order to optimize diagnostic accuracy. Such data will also be key as prenatal diagnostic testing expands to whole genome sequencing and other technologies in the near future.

Key points.

What is already known about this subject?

• Prenatal Exome sequencing (ES) offers the ability to identify single gene disorders that would be missed by traditional karyotype or microarray.

What does this study add?

• This case series highlights the importance of histologic and anatomic pathology examination and complete phenotyping in order to maximize the yield of ES in the prenatal setting.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.