Looking beyond the exome: a phenotype-first approach to molecular diagnostic resolution in rare and undiagnosed diseases.

PurposeTo describe examples of missed pathogenic variants on whole-exome sequencing (WES) and the importance of deep phenotyping for further diagnostic testing.MethodsGuided by phenotypic information, three children with negative WES underwent targeted single-gene testing.ResultsIndividual 1 had a clinical diagnosis consistent with infantile systemic hyalinosis, although WES and a next-generation sequencing (NGS)-based ANTXR2 test were negative. Sanger sequencing of ANTXR2 revealed a homozygous single base pair insertion, previously missed by the WES variant caller software. Individual 2 had neurodevelopmental regression and cerebellar atrophy, with no diagnosis on WES. New clinical findings prompted Sanger sequencing and copy number testing of PLA2G6. A novel homozygous deletion of the noncoding exon 1 (not included in the WES capture kit) was detected, with extension into the promoter, confirming the clinical suspicion of infantile neuroaxonal dystrophy. Individual 3 had progressive ataxia, spasticity, and magnetic resonance image changes of vanishing white matter leukoencephalopathy. An NGS leukodystrophy gene panel and WES showed a heterozygous pathogenic variant in EIF2B5; no deletions/duplications were detected. Sanger sequencing of EIF2B5 showed a frameshift indel, probably missed owing to failure of alignment.ConclusionThese cases illustrate potential pitfalls of WES/NGS testing and the importance of phenotype-guided molecular testing in yielding diagnoses.

Introduction

Whole exome sequencing (WES) has revolutionized clinical genetics by providing a comprehensive and agnostic method for patient evaluation[1]. Diagnostic rates vary from 25–50% and WES has allowed new disease-gene identification and insights into the phenotypic and genetic heterogeneity of Mendelian disorders[2-4]. WES has quickly become part of the standard repertoire of genetic testing, with a prevailing sense that a negative result indicates that disorders in the differential diagnoses have been effectively excluded. We describe three individuals in whom WES and targeted next generation sequencing (NGS)-based testing were non-diagnostic. Phenotype reassessment and use of additional data, such as the single nucleotide polymorphism (SNP) microarray data, helped determine the next steps in the diagnostic process. Targeted single-gene Sanger sequencing and deletion/duplication analyses identified pathogenic variants for the clinically suspected genetic disorder in all three individuals. We provide insights into the reasons for negative WES results and, in an era when genomic technology tends to drive the diagnostic process, we highlight the importance of revisiting clinical information for additional targeted testing.

Patients, Methods, and Results

Individuals 1 and 2 were evaluated at the Duke clinical site of the NIH Undiagnosed Diseases Network (UDN) (https://undiagnosed.hms.harvard.edu) and individual 3 at the Duke Genome Sequencing Clinic.

Individual 1

An 18-month-old Mexican female with progressive joint contractures and related morbidity was evaluated due to extensive skin plaques, subcutaneous and gingival nodules, a biopsy demonstrated dermal accumulation of amorphous hyaline material. Review of a duodenal biopsy showed dilated lymphatics and mucosal edema, consistent with clinical symptoms of protein-losing enteropathy. She had intact cognitive skills, ruling out alternative diagnoses such as Farber and Winchester syndromes. The parents reported a common ancestor in Mexico.

This extended phenotype was consistent with infantile systemic hyalinosis (ISH, OMIM #228600), an autosomal recessive disorder due to loss of function variants in ANTXR2, leading to widespread progressive accumulation of hyaline material and childhood death[5].

Pertinent Previous Genetic Testing

A SNP microarray identified 64.2 Mb regions of homozygosity (ROH), including the ANTXR2 locus. An initial NGS-based sequencing of the exons and flanking splice junctions of ANTXR2, followed by a proband-only WES through a commercial laboratory were negative.

Results of post-WES Genetic Testing in UDN

Review of the SNP microarray did not identify deletions in the regions of ROH. Review of the NGS-based ANTXR2 sequencing and WES data confirmed that 98.5% of the coding regions of ANTXR2 gene were covered at >10×, except exon 1 (85.6%). Sanger sequencing revealed a homozygous pathogenic variant in exon 13 (c.1073dupC). Parental studies confirmed trans configuration. Upon discussion with the commercial laboratory that had performed the WES and NGS-based testing, it appeared that their variant calling software had not reported the variant. Its location adjacent to a homopolymeric repeat region and a common SNP could have contributed to low mapping quality, leading to failure of variant calling (Figure 1). Subsequently, we obtained the BAM files and manual inspection of the data by the UDN bioinformatician confirmed the presence of the variant.

Figure 1

Integrated Genome Variant browser showing condensed read alignment for internally realigned data (A), and the BAM file provided by the clinical laboratory (B). Reads are shown stacked together, with colors indicating a mismatch to the reference sequence. Almost all reads show the pathogenic insertion (c.1073dupC, 22:g.80905986dupG) as a purple mark. The insertion call may be considered low quality because of the adjacent homopolymer repeat and nearby variant. The nearby SNP (c.1069G>C, 22:g.80905990C>G) can be seen by the majority of reads containing the orange G nucleotide change. This variant can also contribute to lower mapping quality and can affect variant quality at the insertion.

Individual 2

A 3.5-year-old girl of Pakistani origin exhibited developmental regression at 16 months of age, cerebellar atrophy and a negative trio WES (proband and parents). She had lost the ability to cruise, crawl, sit, speak and eat by mouth. The parents were first cousins, and the proband had two first cousins once removed who died at age two years after neurodevelopmental regression.

The patient had optic atrophy, profound generalized hypotonia, minimal spontaneous movements, tongue fasciculations and diminished Achilles reflexes, in contrast to previously observed hypertonia with generalized hyperreflexia at age 2.5 years. Review of brain MRIs obtained at 2 and 2.5 years of age revealed stable white matter volume loss of the vermis and cerebellar hemispheres, a normal pons and no iron accumulation (Supplementary Materials and Methods, Figure 1S). The new clinical finding of peripheral nerve involvement led us to consider infantile neuroaxonal dystrophy (IND) (OMIM# 256600), a disorder of neurodevelopmental regression in childhood and early death, caused by biallelic variants in PLA2G6[6,7].

Metabolic laboratory tests and an ataxia gene panel (42 genes) were negative. Trio WES through a commercial laboratory detected a homozygous missense variant of unknown significance in RPGRIP1L, but the clinical course and brain MRI findings were not consistent with Joubert syndrome.

Results of Post- WES Genetic Testing in UDN

A review of the SNP microarray identified several ROH in >4.6% of the genome; 12 genes within these regions, including PLA2G6, were associated with cerebellar atrophy and developmental regression. Manual inspection of the WES BAM files found no functionally significant single nucleotide or copy number variants in these 12 genes. Sanger sequencing and multiplex ligation-dependent probe amplification (MLPA) for deletions/duplications was performed for the five genes within the ROH with the greatest phenotypic overlap: PLA2G6, AC02, BCS1L, NDUFA11 and ADSL. Sanger sequencing was normal for all. MLPA showed a novel homozygous deletion of the non-coding exon 1 in PLA2G6 (Figures 2A and 2S). Follow-up MLPA analysis of the parents confirmed that they were carriers of the deletion.

Figure 2

A. MLPA analysis of PLA2G6. Normalized MLPA data showing the homozygous deletion of the non-coding exon 1 of the PLA2G6 gene leading to total absence of amplification and hence a ratio of zero for the probe covering this region

B. qRT-PCR analysis of PLA2G6 mRNA expression in blood. Results are expressed as means ± SD; mRNA levels were quantitated with real-time PCR and normalized to the level of GAPDH. The figure represents real-time PCR quantification of PLA2G6 gene expression in the patient compared to controls. All results were done in triplicates. P-value was calculated by student t-test.

The breakpoint junction of the PLA2G6 deletion was amplified by long-range PCR (Supplementary Methods). The deletion included 2431-bp in the 5’UTR region of PLA2G6 and revealed a 7-bp insertion at the breakpoint junction (c.−545_−46+1931delinsCGATCTC) (Figure 3S). Fine mapping analysis[8] demonstrated that the deletion encompassed a portion of the promoter region of the gene. Quantitative RT-PCR showed that mRNA expression of PLA2G6 was significantly lower in the patient’s blood compared with unaffected controls (p<0.01) (Figure 2B). Review of the WES data revealed that the capture kit did not include the non-coding exon 1.

Individual 3

An 8-year-old Caucasian female was evaluated for symptoms of ataxia, seizures and white matter disease. At age 3 years, she developed frequent falls and progressive decline in fine motor skills, speech and short-term memory. Generalized seizures started at 7.5 years. Exam revealed dysarthria, lower extremity spasticity, hyperreflexia, clonus and a wide-based gait. Brain MRI showed diffuse symmetric non-enhancing signal abnormalities involving both cerebral hemispheres with volume loss (Figure 4S). The features were consistent with leukoencephalopathy with vanishing white matter (VWM) (OMIM #603896), a disorder that presents with neurological regression, ataxia, spasticity, epilepsy and progressively vanishing white matter in brain MRI[9]. Variants in five genes (EIF2B1-B5) can cause the disorder, most frequently biallelic variants in EIF2B5[10].

SNP microarray analysis detected a paternally-inherited 416 Kb deletion at 10p12.31, interpreted as a benign variant. An NGS panel of 62 genes for VWM leukodystrophy showed a heterozygous c.338G>A, p. Arg113His pathogenic variant in the EIF2B5 gene. Subsequent deletion/duplication testing for the EIF2B5 gene via exon-targeted array-CHG was normal. Trio WES re-identified the heterozygous p. Arg113His variant. Manual inspection of the reads for the EIF2B5 gene did not reveal any additional variants, with coverage of >10× for 100% of this gene.

Results of Post-WES Genetic Testing

Due to the continued clinical suspicion of VWM leukodystrophy, and the detection of one pathogenic variant in the gene, Sanger sequencing of EIF2B5 was pursued. A heterozygous insertion, c.1694delAins45; p.Lys565Ilefs*38 was detected. Although not previously reported in patients with VWM, the insertion had been detected once by the commercial laboratory, in an affected individual. Subsequent parental testing confirmed trans configuration for each variant. Difficulty in alignment of indels larger than 20–50 bp was likely the reason for missing this variant by WES, since retrospective manual inspection of the BAM files failed to detect it.

Discussion

WES is increasingly used as the premier and first-line test for rare and undiagnosed Mendelian disorders[1,2,4,11-14]. The vast majority (>97%) of variants detected by Sanger sequencing can also be detected by WES and with increased detection of mosaicism, WES is a practical tool for comprehensive molecular evaluation[15]. When WES is negative, reanalysis of the data can provide resolution in 10%–30% of cases[16] and this is likely to increase with improvements in technology and new disease associations. However, there is little clarity on the diagnostic options if WES and reanalysis remain negative. Although whole genome sequencing (WGS) may be an option, it is not currently widely available clinically[17]. Therefore, in instances when WES is negative, clinicians may conclude that no diagnostic options remain for these patients.

WES is a complex high-throughput method, and data loss is possible at each step. Pathogenic variants in known disease causing genes may be missed because of decreased coverage, locus-specific features such as GC-rich regions, homopolymeric repeats, sequencing biases and indels that are >20–50 nucleotides[18]. The most common reason for variants being missed is a lack of sufficient sequence coverage depth[19]. Laboratories performing WES and NGS panels may use alternative methods to capture these[20]. Clinicians do reconsider the fit of a phenotype when interpreting variants of uncertain significance on WES, but when WES is negative, adequate coverage of selected genes or exons may lead to a belief that the WES was truly comprehensive. However, in all three of our individuals, coverage was adequate (10× at 98–100% of the bases), yet the pathogenic variants were missed. Thus, a negative WES result should be interpreted in the clinical context of the individual patient, to determine further testing.

In individual 1, the well-characterized phenotype was consistent with only one diagnosis (ISH), and the ROH on the SNP array included ANTXR2, but molecular confirmation remained elusive despite repeated sequencing and adequate coverage. The single nucleotide insertion detected by Sanger sequencing is located adjacent to a complex repetitive region and a SNP, both of which could have decreased the mapping quality of the region around the insertion, and resulted in failure of the variant caller program to detect the insertion. Updated variant calling software and/or manual inspection of the reads would have identified this variant. Manual inspection is of particular importance when a limited set of specific genes is under consideration, but it is not standard practice in commercial laboratories.

Individual 2 had clinical features that changed over a year, leading us to strongly consider a diagnosis of infantile neuroaxonal dystrophy. In this instance, the WES did not capture the deletion because the non-coding exon 1 was not included in the capture kit, as is commonly the case with WES capture kits. Furthermore, structural variants of this size (2.3 kb) would not be detectable by WES. The clinical phenotype and the ROH containing the PLA2G6 gene led us to pursue Sanger sequencing and MLPA, which detected the deletion. It is possible that WGS might have detected this structural variant, but its limited availability and lack of validation of WGS structural variant callers make this an impractical option.

Individual 3 had features consistent with VWM leukoencephalopathy. Although indels <50bp are below the resolution of exon-level deletion/duplication analyses, we would expect detection by WES, if alignment works well. WES missed the 46 bp indel in the EIF2B5 gene completely, since the detection rate of indels decreases with sizes >20–50 bp.

Clinical geneticists are aware of WES being unsuitable for trinucleotide repeat disorders, mitochondrial DNA variants, epigenetic disorders and large structural variants. However, for Mendelian disorders in which SNVs and small indels are possible, the prevalent thinking is that if coverage of the genes of interest by WES is adequate, the disorders have been effectively excluded. Indeed, an increased depth of coverage would not have detected the missed variants in all three of our cases. The cases presented here underscore the importance of further testing if the clinical phenotype is strongly indicative of a specific condition when WES is negative.

In conclusion, these three case examples illustrate the importance a multi-pronged approach when WES is negative. These include: 1) Obtaining detailed clinical phenotyping to create an accurate differential diagnosis; 2) Reconsidering the family history and mode of inheritance; 3) Reassessing SNP microarray data to identify potential causal genes; 4) Manual inspection of the WES reads for genes that are of interest and obtaining information on capture kits and coverage; 5) Pursuing alternative sequencing methodologies such as Sanger sequencing and deletion/duplication testing to detect SNVs and indels that might have been missed with WES. Although WES is comprehensive, its limitations must be considered when negative results are obtained, and alternative diagnostic approaches should be pursued if the phenotype is compelling.

Mercedes	E.	Alejandro	Baylor College of Medicine (Clinical)
Carlos	A.	Bacino	Baylor College of Medicine (Clinical)
Ashok		Balasubramanyam	Baylor College of Medicine (Clinical)
Lindsay	C.	Burrage	Baylor College of Medicine (Clinical)
Gary	D.	Clark	Baylor College of Medicine (Clinical)
William	J.	Craigen	Baylor College of Medicine (Clinical)
Shweta	U.	Dhar	Baylor College of Medicine (Clinical)
Lisa	T.	Emrick	Baylor College of Medicine (Clinical)
Brett	H.	Graham	Baylor College of Medicine (Clinical)
Neil	A.	Hanchard	Baylor College of Medicine (Clinical)
Mahim		Jain	Baylor College of Medicine (Clinical)
Seema	R.	Lalani	Baylor College of Medicine (Clinical)
Brendan	H.	Lee	Baylor College of Medicine (Clinical)
Richard	A.	Lewis	Baylor College of Medicine (Clinical)
Mashid	S.	Azamian	Baylor College of Medicine (Clinical)
Paolo	M.	Moretti	Baylor College of Medicine (Clinical)
Sarah	K.	Nicholas	Baylor College of Medicine (Clinical)
Jordan	S.	Orange	Baylor College of Medicine (Clinical)
Jennifer	E.	Posey	Baylor College of Medicine (Clinical)
Lorraine		Potocki	Baylor College of Medicine (Clinical)
Jill	A.	Rosenfeld	Baylor College of Medicine (Clinical)
Daryl	A.	Scott	Baylor College of Medicine (Clinical)
Alyssa	A.	Tran	Baylor College of Medicine (Clinical)
Jing		Zhang	Baylor College of Medicine (Clinical)
Tiphanie	P.	Vogel	Baylor College of Medicine (Clinical)
Bret	L.	Bostwick	Baylor College of Medicine (Clinical)
Shan		Chen	Baylor College of Medicine (Clinical)
Susan	L.	Samson	Baylor College of Medicine (Clinical)
Hugo	J.	Bellen	Baylor College of Medicine (MOSC)
Michael	F.	Wangler	Baylor College of Medicine (MOSC)
Shinya		Yamamoto	Baylor College of Medicine (MOSC)
Christine	M.	Eng	Baylor College of Medicine (Sequencing)
Donna	M.	Muzny	Baylor College of Medicine (Sequencing)
Patricia	A.	Ward	Baylor College of Medicine (Sequencing)
Yaping		Yang	Baylor College of Medicine (Sequencing)
David	B.	Goldstein	Columbia University
Nicholas		Stong	Columbia University
Yong-hui		Jiang	Duke University
Allyn		McConkie-Rosell	Duke University
Loren	DM.	Pena	Duke University
Kelly		Schoch	Duke University
Vandana		Shashi	Duke University
Rebecca	C.	Spillmann	Duke University
Jennifer	A.	Sullivan	Duke University
Nicole	M.	Walley	Duke University
Alan	H.	Beggs	Harvard University
Lauren	C.	Briere	Harvard University
Cynthia	M.	Cooper	Harvard University
Laurel	A.	Donnell-Fink	Harvard University
Elizabeth	L.	Krieg	Harvard University
Joel	B.	Krier	Harvard University
Sharyn	A.	Lincoln	Harvard University
Joseph		Loscalzo	Harvard University
Richard	L.	Maas	Harvard University
Calum	A.	MacRae	Harvard University
J.	Carl	Pallais	Harvard University
Lance	H.	Rodan	Harvard University
Edwin	K.	Silverman	Harvard University
Joan	M.	Stoler	Harvard University
David	A.	Sweetser	Harvard University
Chris	A.	Walsh	Harvard University
Cecilia		Esteves	Harvard University (CC)
Ingrid	A.	Holm	Harvard University (CC)
Isaac	S.	Kohane	Harvard University (CC)
Paul		Mazur	Harvard University (CC)
Alexa	T.	McCray	Harvard University (CC)
Matthew		Might	Harvard University (CC)
Rachel	B.	Ramoni	Harvard University (CC)
Kimberly		Splinter	Harvard University (CC)
David	P.	Bick	HudsonAlpha
Camille	L.	Birch	HudsonAlpha
Braden	E.	Boone	HudsonAlpha
Donna	M.	Brown	HudsonAlpha
Daniel	C.	Dorset	HudsonAlpha
Lori	H.	Handley	HudsonAlpha
Howard	J.	Jacob	HudsonAlpha
Angela	L.	Jones	HudsonAlpha
Jozef		Lazar	HudsonAlpha
Shawn	E.	Levy	HudsonAlpha
J.	Scott	Newberry	HudsonAlpha
Molly	C.	Schroeder	HudsonAlpha
Kimberly	A.	Strong	HudsonAlpha
Elizabeth	A.	Worthey	HudsonAlpha
Jyoti	G.	Dayal	NIH
David	J.	Eckstein	NIH
Sarah	E.	Gould	NIH
Ellen	M.	Howerton	NIH
Donna	M.	Krasnewich	NIH
Laura	A.	Mamounas	NIH
Teri	A.	Manolio	NIH
John	J.	Mulvihill	NIH
Anastasia	L.	Wise	NIH
Tiina	K.	Urv	NIH
Ariane	G.	Soldatos	NINDS
Matthew		Brush	OHSU Metabolomics
Jean-Philippe	F.	Gourdine	OHSU Metabolomics
Melissa		Haendel	OHSU Metabolomics
David	M.	Koeller	OHSU Metabolomics
Jennifer	E.	Kyle	PNNL Metabolomics
Thomas	O.	Metz	PNNL Metabolomics
Katrina	M.	Waters	PNNL Metabolomics
Bobbie-Jo	M.	Webb-Robertson	PNNL Metabolomics
Euan	A.	Ashley	Stanford University
Jonathan	A.	Bernstein	Stanford University
Annika	M.	Dries	Stanford University
Paul	G.	Fisher	Stanford University
Jennefer	N.	Kohler	Stanford University
Daryl	M.	Waggott	Stanford University
Matthew	T.	Wheeler	Stanford University
Patricia	A.	Zornio	Stanford University
Patrick		Allard	UCLA
Hayk		Barseghyan	UCLA
Esteban	C.	Dell'Angelica	UCLA
Katrina	M.	Dipple	UCLA
Naghmeh		Dorrani	UCLA
Matthew	R.	Herzog	UCLA
Hane		Lee	UCLA
Stan	F.	Nelson	UCLA
Christina	GS.	Palmer	UCLA
Jeanette	C.	Papp	UCLA
Janet	S.	Sinsheimer	UCLA
Eric		Vilain	UCLA
Julian	A.	Martínez-Agosto	UCLA
Neil	H.	Parker	UCLA
Brent	L.	Fogel	UCLA
Emilie	D.	Douine	UCLA
Allen		Lipson	UCLA
Allison		Zheng	UCLA
Ascia		Eskin	UCLA
Sandra	K.	Loo	UCLA
Ani		Dillon	UCLA
Christopher	J.	Adams	UDP
Elizabeth	A.	Burke	UDP
Katherine	R.	Chao	UDP
Mariska		Davids	UDP
David	D.	Draper	UDP
Tyra		Estwick	UDP
Trevor	S.	Frisby	UDP
Kate		Frost	UDP
Valerie		Gartner	UDP
Rena	A.	Godfrey	UDP
Mitchell		Goheen	UDP
Gretchen	A.	Golas	UDP
Mary	G.	Gordon	UDP
Catherine	A.	Groden	UDP
Mary	E.	Hackbarth	UDP
Isabel		Hardee	UDP
Jean	M.	Johnston	UDP
Alanna	E.	Koehler	UDP
Lea		Latham	UDP
Yvonne	L.	Latour	UDP
C.	Christopher	Lau	UDP
Denise	J.	Levy	UDP
Adam	P.	Liebendorfer	UDP
Ellen	F.	Macnamara	UDP
Valerie	V.	Maduro	UDP
Thomas	C.	Markello	UDP
Alexandra	J.	McCarty	UDP
Jennifer	L.	Murphy	UDP
Michele	E.	Nehrebecky	UDP
Donna		Novacic	UDP
Barbara	N.	Pusey	UDP
Sarah		Sadozai	UDP
Katherine	E.	Schaffer	UDP
Prashant		Sharma	UDP
Sara	P.	Thomas	UDP
Nathanial	J.	Tolman	UDP
Camilo		Toro	UDP
Zaheer	M.	Valivullah	UDP
Colleen	E.	Wahl	UDP
Mike		Warburton	UDP
Alec	A.	Weech	UDP
Guoyun		Yu	UDP
Andrea	L.	Gropman	UDP, CNMC
David	R.	Adams	UDP, NHGRI
William	A.	Gahl	UDP, NHGRI
May Christine	V.	Malicdan	UDP, NHGRI
Cynthia	J.	Tifft	UDP, NHGRI
Lynne	A.	Wolfe	UDP, NHGRI
Paul	R.	Lee	UDP, NINDS
John	H.	Postlethwait	UO MOSC
Monte		Westerfield	UO MOSC
Anna		Bican	Vanderbilt University
Rizwan		Hamid	Vanderbilt University
John	H.	Newman	Vanderbilt University
John	A.	Phillips III	Vanderbilt University
Amy	K.	Robertson	Vanderbilt University
Joy	D.	Cogan	Vanderbilt University