ARTICLE

Genome sequencing increases diagnostic yield in clinically

diagnosed Alagille syndrome patients with previously
negative test results
Ramakrishnan Rajagopalan, MS1,2, Melissa A. Gilbert, PhD1, Deborah A. McEldrew, BS1,
James A. Nassur, BA1, Kathleen M. Loomes, MD3, David A. Piccoli, MD3, Ian D. Krantz, MD4,5,
Laura K. Conlin, PhD, FACMG1 and Nancy B. Spinner, PhD, FACMG 1,6

Purpose: Detection of all major classes of genomic variants in a diagnostic yield of 0.9%. Furthermore, GS resolved two complex
single test would decrease cost and increase the efficiency of rearrangements, resulting in identification of a pathogenic variant
genomic diagnostics. Genome sequencing (GS) has the potential to in 97.5% (n = 396/406) of patients after GS.
provide this level of comprehensive detection. We sought to Conclusion: GS provided an increased diagnostic yield for
demonstrate the utility of GS in the molecular diagnosis of 18 individuals with clinically defined ALGS who had prior negative
patients with clinically defined Alagille syndrome (ALGS), who had or incomplete genetic testing by other methods. Our results show
a negative or inconclusive result by standard-of-care testing. that GS can detect all major classes of variants and has potential to
Methods: We performed GS on 16 pathogenic variant-negative become a single first-tier diagnostic test for Mendelian disorders.
probands and two probands with inconclusive results (of 406 ALGS
probands) and analyzed the data for sequence, copy-number, and Genetics in Medicine (2021) 23:323–330; https://doi.org/10.1038/s41436-
structural variants in JAG1 and NOTCH2. 020-00989-8
Results: GS identified four novel pathogenic alterations including
a copy-neutral inversion, a partial deletion, and a promoter variant Keywords: genome sequencing; Alagille syndrome; JAG1;
in JAG1, and a partial NOTCH2 deletion, for an additional NOTCH2, diagnostic testing

INTRODUCTION microarray analysis (CMA), multiplex ligation-dependent

Targeted applications of next-generation sequencing (NGS) probe amplification (MLPA), cytogenetic analysis, and
tests, such as NGS panels, and screening of all protein-coding fluorescence in situ hybridization (FISH). By design, ES does
genes by exome sequencing (ES), are the current standard-of- not usually identify deep intronic, promoter, and noncoding
care diagnostic tests for many suspected Mendelian dis- variants. Genome sequencing (GS) is poised to overcome
orders.1 While the use of targeted NGS tests has increased the many of the barriers faced by ES for several reasons. GS
overall diagnostic yield over the years, a molecular cause is provides read coverage across both intronic and intergenic
not identified in roughly 70% of individuals1 who present for regions of the genome, enabling the comprehensive detection
genomic testing. However, the diagnostic utility of NGS of all coding and noncoding genomic variants at nucleotide-
panels and ES varies based on the clinical indication, with level resolution, which greatly enhances clinical interpreta-
some diseases exhibiting much higher yields than others.2,3 tion. In addition, polymerase chain reaction (PCR)–free
Regardless, there is a need to push beyond the current protocols for GS eliminate amplification bias (a known
standard-of-care testing methods to increase diagnostic yield. confounder) to provide more uniform coverage compared
Although ES is able to detect exonic sequence variants (single- with ES.
nucleotide variants [SNVs] and insertion/deletions [indels]) We sought to demonstrate the clinical utility of GS
with high confidence, ES is less equipped to comprehensively in identifying pathogenic variants in individuals with
identify some of the major classes of genomic variation, clinically defined Alagille syndrome (ALGS; OMIM 118450)
namely copy-number and structural variants,1 which are using a cohort of 18 patients with previously negative
typically detected by other approaches such as chromosomal or inconclusive testing. ALGS is an autosomal dominant
1
Division of Genomic Diagnostics, Department of Pathology and Laboratory Medicine, The Children’s Hospital of Philadelphia, Philadelphia, PA, USA; 2School of Biomedical
Engineering, Health and Sciences, Drexel University, Philadelphia, PA, USA; 3Division of Pediatric Gastroenterology, Hepatology and Nutrition, Department of Pediatrics, The
Children’s Hospital of Philadelphia and the Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA; 4Roberts Individualized Medical Genetics Center,
Division of Human Genetics at the Children’s Hospital of Philadelphia, Philadelphia, PA, USA; 5Department of Pediatrics, Perelman School of Medicine at the University of
Pennsylvania, Philadelphia, PA, USA; 6Department of Pathology and Laboratory Medicine, Perelman School of Medicine at the University of Pennsylvania, Philadelphia, PA, USA.
Correspondence: Nancy B. Spinner (spinner@email.chop.edu)
Co-First Authorship: Ramakrishnan Rajagopalan, Melissa A. Gilbert.
Submitted 30 April 2020; revised 17 September 2020; accepted: 21 September 2020
Published online: 20 October 2020

GENETICS in MEDICINE | Volume 23 | Number 2 | February 2021 323

 ARTICLE RAJAGOPALAN et al

disorder characterized by hepatic, cardiac, ocular, vertebral, at least three of five characteristic features of ALGS (Table
renal, vascular, and facial involvement.4 It has long been S2). Sixteen of 18 probands had previously negative genomic
recognized as a disease of Notch signaling deficiency, with testing and two had an inconclusive MLPA result for JAG1.
94.3% of individuals found to have a pathogenic variant in the These two were included to determine whether GS could fully
Notch ligand, JAGGED1 (JAG1) and 2.5% of individuals resolve an apparently complex pathogenic variant. Family
found to have a pathogenic variant in the Notch receptor, members of five probands (one quad, two trios, and two duos)
NOTCH2.5 Standard-of-care testing to identify pathogenic were also chosen for GS based on sample availability.
variants typically employs a serial testing strategy that includes
(1) sequencing JAG1 for SNVs and indels using genomic Ethics statement
DNA; (2) performing deletion/duplication analysis of JAG1 All probands and family members were enrolled and
using various strategies, but commonly MLPA and/or CMA; consented to participate in a research study approved by the
and (3) sequencing NOTCH2 for SNVs and indels using Institutional Review Board at CHOP.
genomic DNA.6 Functional studies strongly suggest haploin-
sufficiency as a disease mechanism for JAG1 pathogenic Standard-of-care testing
variants, with a majority leading to early protein truncations,5 We employed a serial testing strategy for our standard-of-care
and a handful of studied missense variants leading to the testing that included a minimum of three genomic tests to
translation of a nonfunctional or incorrectly trafficked protein assay for (1) small sequence variants such as SNVs and indels
product.5,7,8 The mechanism by which pathogenic NOTCH2 within the JAG1 coding region, including splice acceptor/
variants cause ALGS is less clear. No genes other than JAG1 donor variants; (2) full and partial gene deletion/duplication
1234567890():,;

and NOTCH2, including other Notch signaling genes, have analysis of JAG1; and (3) SNVs and indels within the
been identified to cause ALGS. Thus, we hypothesized that NOTCH2 coding region, including splice acceptor/donor
individuals with clinically consistent ALGS and without a variants. Previous studies have shown that these three testing
molecular diagnosis are likely to have a pathogenic variant strategies are capable of detecting up to ~97% of pathogenic
within JAG1 or NOTCH2 that is undetectable by current variants in ALGS.5,6
screening methodologies. Using a previously identified cohort
of well-characterized probands with clinically consistent Genome sequencing
ALGS, but with no confirmed JAG1 or NOTCH2 pathogenic DNA was extracted from either whole blood or lymphoblas-
variant, we aimed to assess the utility of GS in increasing the toid cell lines using the DNeasy Blood and Tissue Kit (Qiagen,
molecular diagnostic yield for this disease. Hilden, Germany). Short-read (2 × 150 bp) Illumina (Illu-
mina, San Diego, CA) GS was performed at the Broad
MATERIALS AND METHODS Institute Genomic Services (Boston, MA) using a PCR-free
Patient cohort protocol at a targeted mean sequencing depth of 30×.9
We have actively enrolled patients with suspected ALGS into BWA10-aligned CRAM files (hg38) produced by the GATK
our single-center ALGS research study (“Molecular Analysis best practices workflow11 were obtained from the Broad
of Alagille Syndrome”) at the Children’s Hospital of Institute. Initial quality control steps included the estimation
Philadelphia (CHOP) since 1992, amassing a convenience of coverage using the software tool indexcov,12 and the
series of 446 probands, with participant referral occurring pairwise relatedness and sex-check using somalier.13
both within CHOP and worldwide through physician out-
reach to our study team. For this prospective study, we Variant calling and prioritization
reviewed our database of 446 individuals with suspected SNVs and indels were called using the Strelka2 software14
ALGS and identified a cohort of 18 individuals enrolled and filtered using the genome intervals for JAG1
between July 1997 and July 2014 with a clinical diagnosis of (hg38 chr20:10628605-10683078) and NOTCH2 (hg38 chr1:
ALGS who had prior negative or inconclusive testing of 119911553-120069662) with 1 kb padding on either side to
sequence (via Sanger and/or NGS-based analysis) and copy- include variants in the promoter and untranslated regions.
number variants (by MLPA and/or FISH) in JAG1, and Annotation of SNVs/indels was performed using the software
sequence variants in NOTCH2 (Fig. 1 and Table S1). annovar.15 Genome-wide copy-number detection was per-
To establish our study cohort of 18 probands, we excluded formed using CNVnator16 and ERDS17 (mean read-depth
those who did not meet the clinical diagnostic criteria for using bins of length 1 kb) while structural variants were
ALGS after medical record review, those with insufficient identified using manta18 with default settings. We filtered for
clinical information to support a clinical diagnosis of ALGS, copy-number and structural variants overlapping the above
those with no remaining sample for study, and those with low genomic intervals including JAG1/NOTCH2. All identified
sample quality (Fig. 1). We also excluded individuals with a variants were manually inspected using the Integrative
pathogenic variant in either JAG1 or NOTCH2 previously Genomics Viewer (IGV) visualization software (Broad
identified by standard-of-care testing. All 18 probands in the Institute, Boston, MA). Rare SNVs and indels were filtered
remaining group underwent chart review by a clinical team of using a maximum minor allele frequency threshold of 0.1% in
two gastroenterologists and two geneticists at CHOP, and had gnomAD v2.1 (https://gnomad.broadinstitute.org/)19 with the

324 Volume 23 | Number 2 | February 2021 | GENETICS in MEDICINE

RAJAGOPALAN et al ARTICLE

Probands in “Molecular
Analysis of Alagille syndrome”
n=446

Confirm ALGS Excluded

n=40
-Could not confirm clinical diagnosis or
insufficient sample

Clinically consistent ALGS

probands
n=406

Standard of care Excluded

molecular testing n=388
-Pathogenic variant in JAG1 (n=378)
-Pathogenic variant in NOTCH2 (n=10)

GS of ALGS probands with no

(n=16) or indeterminate (n=2)
pathogenic variant
n=18
Excluded
Pathogenic variants n=2
previously missed -Previously missed JAG1 varriant

Complex structural
variants resolved Excluded
n=2
-Resolution of an indeterminate result

GS Analysis Cohort
n=14

Negative result–no identified Positive result–identification of a

pathogenic variant pathogenic variant
n=10 n=4

Fig. 1 Flow diagram of the study population. Genome sequencing (GS) was performed on a cohort of 18 individuals that were identified in our study,
Molecular Analysis of Alagille Syndrome (ALGS). Exclusion criteria and results of the study are indicated.

software slivar (https://github.com/brentp/slivar). Synon- used to map the soft-clipped reads across the breakpoint
ymous and intronic variants were annotated using the tool junctions to map the novel breakpoint junctions. Orientation
spliceAI.20 SpliceAI produces probability (DELTA score) for of the read-pairs with abnormal insert sizes was used to infer
loss or gain of an acceptor or donor site. We used the inversions.
recommended threshold of 0.5 to filter for cryptic splice
acceptor/donor variants. We used genomic coordinates from RESULTS
ORegAnno database to filter for variants that fall within Of the 446 individuals in our database with both suspected
potential JAG1/NOTCH2 regulatory regions. and molecularly confirmed ALGS collected within our
Molecular Analysis of ALGS research study at CHOP, we
Variant confirmation identified a cohort of 16 probands with prior negative
The deletion identified in proband 10, the inversion identified standard-of-care testing and two probands with inconclusive
in proband 12, and the promoter variant identified in MLPA results showing noncontiguous deletions who con-
proband 11 were confirmed using standard droplet digital fidently met the clinical classification of ALGS (Fig. 1, Table
PCR (ddPCR) assays, which are described in the Supplemen- S1). This cohort included ten males and eight females with a
tary Materials and Methods. median age at enrollment of 5.7 years (range 3 months to 26
years) and with a highly diverse geographical distribution
Complex rearrangements (Table 1). All 18 individuals presented with hepatic
Paired-end reads with abnormal insert sizes and soft-clipped manifestations of ALGS, which included the presence of one
reads that span the breakpoints were analyzed, and blat21 was or more of the following features: bile duct paucity,

GENETICS in MEDICINE | Volume 23 | Number 2 | February 2021 325

ARTICLE RAJAGOPALAN et al

Table. 1 Demographic and clinical features of the ALGS The pathogenic variants resolved by GS included an
cohort who underwent GS. inversion, a promoter variant, an exon 1 deletion in JAG1,
Patient characteristics n (%) and a deletion in NOTCH2. In proband 12, a 672-kb copy-
neutral inversion involving the first three exons of JAG1
Median age at enrollment, years (range) 5.7 (0.25–26)
(chr20:10,663,195–11,342,633) was identified (Fig. 2, Fig. S1),
Women 8 (44.4%)
which was inherited from his clinically affected father (Table
Geographic location of patient referral
S2). The 5’ end of the inversion mapped to intron 3 of JAG1
CHOP 3 (16.7%)
and the 3’ end mapped outside of the JAG1 gene, within a
New York 2 (11.1%)
gene desert. Gene expression analysis of JAG1 using ddPCR
Vietnam 2 (11.1%)
confirmed that JAG1 expression was reduced in both proband
Brazil 1 (~5.6%)
12 and his father (proband 12-F), suggestive of JAG1
Denmark 1 (~5.6%)
haploinsufficiency (Fig. 3a). In proband 11, a novel SNV
England 1 (~5.6%)
(c.-100G>A; chr20:10,673,630), of unknown inheritance, that
India 1 (~5.6%) is absent from public genomic variant databases (ExAC and
Massachusetts 1 (~5.6%) gnomAD; https://gnomad.broadinstitute.org/)19 was identi-
Mississippi 1 (~5.6%) fied in the promoter region of JAG1 (Fig. S2). Gene expression
North Carolina 1 (~5.6%) analysis of JAG1 using ddPCR confirmed that JAG1
Tennessee 1 (~5.6%) expression was reduced in proband 11, suggestive of JAG1
Turkey 1 (~5.6%) haploinsufficiency (Fig. 3a). The two deletion variants
Virginia 1 (~5.6%) included a maternally inherited (from an affected mother)
Washington 1 (~5.6%) 606-bp deletion involving exon 1 of JAG1 (chr20:10,673,044–
Clinical features 10,673,649), identified in proband 15 (Fig. S3), and a de
Hepatic 18 (100%) novo 5.9-kb deletion involving exons 31–34 of NOTCH2
Cardiac 16 (88.9%) (chr1:119,913,673–119,919,578), identified in proband 10
Posterior embryotoxon 9 (64.3%)a (Fig. 3b, Fig. S4).
Family history 6 (54.6%)a Additionally, we resolved the complex structural variants in
Butterfly vertebrae 6 (46.5%)a the two individuals (probands 8 and 14, both of unknown
Facies 6 (46.5%)a inheritance) with prior MLPA results that were suggestive of a
Renal 6 (46.5%)a complex rearrangement involving multiple breakpoints and
a
Of those reporting. See Table S1 for patient-specific phenotypes. required more precise characterization (Fig. 4a). In proband
8, we confirmed the presence of the two previously identified
noncontiguous deletions, involving exon 3 (9 kb) and exons
cholestasis, elevated liver enzymes, and cirrhosis (Table S2). 9–26 (23 kb), as well as a third deletion distal to the exon 3
Probands were also assessed for the presence of skeletal, deletion (4 kb), which was not known prior to GS (Fig. 4b, c
cardiac, renal, ocular, and facial phenotypes as well as family and Fig. S5). Analysis of the read-pairs with abnormal insert
history of ALGS (Table S2). All probands presented with at sizes and orientation further revealed an inversion between
least three of these clinical features. the two intragenic deletions within JAG1 (Fig. 4b, c and
GS was performed in this ALGS cohort of 18 probands, Fig. S5). CNVnator identified all three deletions (22.9 kb, 9 kb,
specifically focusing on sequence, copy-number, and struc- and 4.3 kb) while ERDS identified two of the three (9 kb
tural variants in JAG1 and NOTCH2, and resulted in the and 4.3 kb). Manta identified the inversion with precise
identification of a pathogenic variant in 6 of the 16 individuals breakpoints with an 18-bp insertion at the distal breakpoint
with prior negative testing, including two deletions, an (chr20:10690781) (Table S3).
inversion, a promoter SNV, a JAG1 missense variant Similarly, analysis of GS data from proband 14 revealed a
(c.401T>C; p.L134S), and a JAG1 frameshift variant comparatively simpler JAG1 rearrangement but with smaller
(c.1978del; p.E660Rfs*83). GS further resolved the breakpoint segments of DNA including an intragenic inversion involving
architecture of both complex structural variants that were exon 10 (246 bp) between two 1-kb deleted segments (exon 9
previously identified via MLPA (Table S2). Both the missense and exons 11–12) (Fig. 4b, c and Fig. S6). CNVnator
and frameshift variants identified in JAG1 were detectable by identified a single contiguous 2.3-kb deletion and missed the
standard-of-care sequencing, and subsequent review of the normal region (246 bp) situated between the two 1-kb
original raw sequencing data showed that both of these deletions. Manta identified two pairs of breakends and did
variants were present, but were missed by the analyst. Of the not classify the breakends into a variant type (e.g., deletion)
remaining four novel variants that were identified by GS, (Table S3).
three were within JAG1 and one deletion was within
NOTCH2. There were no cryptic splice site or potential DISCUSSION
regulatory variants identified by spliceAI (DELTA score ≥0.5) Molecular diagnosis of ALGS can be accomplished by
and ORegAnno database, respectively. screening for pathogenic variants in JAG1 or NOTCH2 using

326 Volume 23 | Number 2 | February 2021 | GENETICS in MEDICINE

RAJAGOPALAN et al ARTICLE
chr20:10,663,195 chr20:11,342,633

Exon 4 Exon 3 Exon 2 Exon 1 JAG1 Promoter

Region

Reference
Genome

Rearranged
Structure

JAG1 Promoter
Exon 4 Region Exon 1 Exon 2 Exon 3
679 kb Inversion

Fig. 2 Schematic of JAG1 inversion identified in proband 12. The reference genome (upper structure) depicts the 679-kb inverted region, encom-
passing JAG1 exons 1–3, bounded by dashed lines. The breakpoints extend from intron 3 to a gene desert upstream of the JAG1 promoter. The rearranged
structure is shown below. Paired-end reads with abnormal insert size and orientation were used to infer the approximate boundaries of the inversion and
soft-clipped reads at the ends of the inversion were used to precisely map the breakpoints at nucleotide-level resolution.

standard techniques for sequencing and copy-number Proband 10 was found to have a deletion across exons
analysis, with a diagnostic rate of ~97%.5,6 In this study, we 31–34 in the NOTCH2 gene. Copy-number variants in the
used GS on 18 patients, in whom pathogenic variants were NOTCH2 gene have not been previously reported in ALGS
not identified by standard methods, drawn from a larger patients, and therefore copy-number analysis for NOTCH2
patient cohort of 406 individuals. Of these 18 probands, 2 has not been recommended for standard-of-care testing.5 The
were found to have a pathogenic variant that was missed by pathomechanism of NOTCH2 variants is less clear than that
Sanger sequencing, 2 had breakpoint mapping of complex for JAG1 variants, particularly since the majority of NOTCH2
rearrangements involving JAG1 to a resolution that was not variants are missense rather than protein-truncating.5 Trun-
attainable by MLPA, and 4 were found to have novel variants cating variants in the terminal exon (exon 34) of NOTCH2
that could not be detected by previous testing methods. have been implicated in Hajdu–Cheney syndrome, character-
Therefore, 8/18 patients with no confirmed pathogenic ized by focal bone destruction and osteoporosis, along with
variant identified by standard-of-care testing were resolved other features (OMIM 102500). Hajdu–Cheney associated
via GS, and 4 of these individuals were found to have a variant NOTCH2 pathogenic variants have been shown to escape
that would only be detectable through GS. nonsense-mediated decay and lead to gain-of-function
Each of the four novel pathogenic variants identified protein products,22 a pathomechanism that is distinct from
emphasizes a different diagnostic advantage of GS compared those proposed for NOTCH2 variants in ALGS. Minimal
with ES, highlighting its diverse and striking clinical utility. functional evidence from ALGS NOTCH2 variants is unable
Proband 15 was found to have a JAG1 exon 1 deletion, despite to confirm haploinsufficiency as a singular disease mechan-
having a previously normal MLPA result. MLPA assays are ism, and indeed a study examining the functional effect of a
limited to detect copy-number variants inside of a region handful of NOTCH2 variants found that despite all of them
bound by two probes. After reviewing the original MLPA displaying defective Notch signaling, one of the nonsense
data, we found that the distal breakpoint of this deletion variants studied was shown to escape nonsense-mediated
fell 10 bp outside of the second exon 1 probe in the MLPA messenger RNA (mRNA) decay.23 Thus, the pathomechanism
design. This highlights a limitation in MLPA testing and of NOTCH2 variants in ALGS appears to be varied, possibly
underscores the possibility of false negatives when utilizing including haploinsufficiency as well as other mechanisms. The
this technology. phenotype of proband 10 in our cohort is remarkably

GENETICS in MEDICINE | Volume 23 | Number 2 | February 2021 327

ARTICLE RAJAGOPALAN et al

a 1.5 JAG1 or NOTCH2 by conventional testing methodologies

were likely to have variants in regulatory regions,5 and finding
JAG1 Gene Expression

Exon 1 Probe
this promoter variant opens up the field to more exploratory
Exons 25-26 Probe research on JAG1 gene regulation, promising the potential to
1.0
find disease-causing variants outside of protein-coding
regions.
Lastly, the identification of an inversion involving the first
0.5
three exons of JAG1 highlights the advantage of GS when it
comes to detecting copy-neutral genomic rearrangements,
which are missed by chromosomal single-nucleotide poly-
0.0
Nega c.21 morphism (SNP) arrays. Moreover, the intronic position of
tive 22_2 Proband Proband Proband
125d
el
11 12 12-F one breakpoint requires GS technology, which provides equal
coverage across coding and noncoding regions, rather than ES
b 2.0 testing, which would fail to identify the intronic breakpoint.
GS also resolved previously unknown complex rearrange-
NOTCH2 Copy Number

ments in probands 8 and 14. We used two read-depth based

1.5 copy-number variant callers (CNVnator and ERDS) and one
split-read caller (manta) to identify these structural variants.
1.0 CNVnator was the most sensitive among the read-depth
callers finding all deletions (n = 5/5) while ERDS failed to
detect deletions smaller than 2 kb. In proband 14, there were
0.5
two deletions (918 bp and 1.2 kb) with a small 246-bp region
of two-copy DNA between them. CNVnator failed to
0.0 recognize the 246-bp normal region between the two
Proband 10 Mother Father deletions, most likely due to the fact that the size of the
Fig. 3 RNA expression (JAG1) and copy number (NOTCH2) is reduced normal region is lower than the threshold used for binning
in patient cell lines harboring novel pathogenic variants. (a) Droplet the read-depth (1000 bp). Manta was able to identify the
digital polymerase chain reaction (ddPCR) performed on complementary inversions overlapping these breakpoints. Read-depth callers
DNA (cDNA) made from RNA extracted from lymphoblastoid cell lines of provided approximate breakpoints while manta provided
affected individuals showing reduced JAG1 gene expression in the indivi-
dual with the promoter variant (proband 11) as well as the individual with
exact breakpoints involved in both structural variants. Thus, it
the inversion (proband 12) and his affected father (proband 12-F), who also is important to use a combination of read-depth and split-
has the inversion. Parental samples were not available for proband 11. The read-based callers to characterize complex rearrangements
negative control is the average of four unaffected individuals with no using GS data along with manual work to reconstruct the
pathogenic JAG1 variant. An individual with a pathogenic frameshift variant entire complex rearrangement.
(c.2122_2125del) that is predicted to truncate the JAG1 protein was
included as a positive control. Two separate primer/probe sets were used for
In addition, we further identified two JAG1 variants that
confirmation, one designed in exon 1 and one designed to cross exons were missed by conventional Sanger sequencing. The frame-
25–26. Values for all samples were normalized to the internal control, shift variant (c.1978del; p.E660Rfs*83) has not previously
TBP. Error bars for the negative control are plotted as standard deviation. been reported, but is expected to be pathogenic since it results
(b) ddPCR showing NOTCH2 copy number in proband 10 and her unaf- in early protein truncation. The missense variant (c.401T>C;
fected parents.
p.L134S) was previously reported in another individual within
our ALGS cohort.5 Upon review of the original sequencing
different compared with Hajdu–Cheney syndrome and data, both of these variants were found to be present, and thus
includes cholestasis, peripheral pulmonic stenosis, posterior were missed by the original study, highlighting an analytical
embryotoxon, and classic ALGS facies. While we confirm the limitation of manual sequencing review.
reduced copy number of NOTCH2 in proband 10, more Our study was limited to include only those individuals that
functional evidence may be required to substantiate haploin- had an available sample and that met a very conservative
sufficiency as the pathomechanism. Although we predict that clinical diagnostic requirement, as determined by a team of
NOTCH2 copy-number variants are a rare cause of ALGS, our two gastroenterologists and two geneticists. The identification
results suggest that current testing guidelines should be of JAG1 and NOTCH2 pathogenic variants in individuals with
reconsidered to include copy-number analysis for NOTCH2. mild ALGS, or who have less than three clinical symptoms,
Our finding of a JAG1 promoter variant in proband 11, and has been documented.24 In our testing scheme, these
the subsequent confirmation of reduced JAG1 gene expression individuals were excluded from our GS analysis. Although
in this proband, is particularly interesting as it provides new this was a limitation of our study, we felt it was necessary to
evidence that variants in JAG1 regulatory regions are capable apply the most stringent clinical diagnostic guidelines to
of causing ALGS. We have previously suggested that ALGS evaluate GS as a genomic tool in a population that was most
patients who were not found to have an identified variant in likely to truly have ALGS. A second potential limitation for

328 Volume 23 | Number 2 | February 2021 | GENETICS in MEDICINE

RAJAGOPALAN et al ARTICLE
a
Proband 8 Proband 14
1.5 1.5
Gain Gain
Probe Ratio

Probe Ratio
1.0 1.0

Loss Loss
0.5 0.5

0.0 0.0
1 3 5 7 9 11 13 15 17 19 21 23 25 26 1 3 5 7 9 11 13 15 17 19 21 23 25 26
Exon Exon

b Exon 26 Exon 9 Exon 3 Exon 13 Exon 12 Exon 11 Exon 10 Exon 9

JAG1 JAG1

A B C D E F G A B C D E

23 kb 10 kb 9 kb 20 kb 4 1 kb 246 1 kb
kb bp
10,628,528 10,661,832 10,690,719 10,648,536 10,649,454 10,650,866
10,651,408 10,670,875 10,694,980 10,649,700

c
C C
A E G A E

Fig. 4 Genome sequencing (GS) resolves complex structural rearrangements in individuals with clinically defined Alagille syndrome
(ALGS). (a) JAG1 multiplex ligation-dependent probe amplification (MLPA) results for probands 8 and 14. Probe ratio is plotted for each exon. A threshold
below 0.75 was used to classify losses and above 1.25 was used to classify gains. Circles represent copy-neutral ratios while squares represent deletions. The
SALSA MLPA probemix (P184-C3 JAG1) was purchased from MRC Holland (Amsterdam, Netherlands) and details, including quantification, normalization,
and controls can be found through this link: https://www.mrcholland.com/products/18527/Product%20description%20P184-C3-0317%20JAG1-v12.pdf.
(b) Schematic of the genomic structure of JAG1 for proband 8, who had noncontiguous deletions of exon 3 and exons 9–26 demonstrated by MLPA, and
proband 14, who had noncontiguous deletions of exon 9 and exons 11–12. Dashed lines denote the genomic coordinates (hg38) of the breakpoints and red
bars indicate the deleted regions, which are interspersed with nondeleted portions of the gene (shown in gray). (c) Schematic of the genomic rearrangement
for probands 8 and 14.

our study is that GS for all of the identified genetic variants in Prior to GS, our diagnostic yield using standard-of-care
addition to their orthogonal variant confirmation utilize testing for our clinically consistent ALGS probands within our
DNA/RNA from lymphoblastoid cell lines rather than Molecular Analysis of ALGS study was 96.6% (JAG1 n = 382/
primary tissue. However, the use of lymphoblastoid cell lines 406, 94.1%; NOTCH2 n = 10/406, 2.5%; pathogenic variant
in the medical genetics literature is well established, and has negative n = 14/406, 3.4%). We report an additional diagnostic
had a very high success rate. While the possibility that somatic yield of 0.9% after applying GS to our testing strategy (JAG1
variants may arise during cell culture exists, the reported n = 385/406, 94.8%; NOTCH2 n = 11/406, 2.7%; pathogenic
somatic variant rate is quite low (0.3%).25 variant negative n = 10/406, 2.5%). We excluded both samples
Although we hypothesize that the remaining ten individuals with complex rearrangements previously identified by MLPA
who were not found to have a pathogenic variant identified (probands 8 and 14) and both samples with SNVs that were
may have variants in as-yet unidentified regulatory regions of, missed by Sanger sequencing (probands 17 and 18) from our
or cryptic splice variants in, JAG1 or NOTCH2, or were increase in diagnostic yield since standard-of-care would be
missed by the current bioinformatics methods, there is the diagnostic for these four individuals.
possibility that some of these individuals may have a different Major deterrents to the utilization of GS as a first-tier
disorder. Clinical phenotypes can emerge over time, and it is genomic test include the higher sequencing costs and the
possible that new information may point to other molecular burden of data analysis. By using a “genome slice” and
diagnoses in these patients.26,27 However, we attempted to analyzing only the two known disease genes known to cause
minimize the likelihood of this outcome by choosing ALGS, we significantly reduced the burden of data analysis yet
individuals with phenotypic features that were highly we are able to detect all previously reported JAG1 and
characteristic for ALGS. In the future, total mRNA sequen- NOTCH2 pathogenic variants in ALGS as well as novel
cing of these individuals might help identify pathogenic structural variants, intronic variants, promoter variants, and
alternatively spliced variants in JAG1/NOTCH2 missed by regulatory variants, with the capability to reflex back to the
DNA sequencing. whole genome if necessary. Current diagnostics for ALGS

GENETICS in MEDICINE | Volume 23 | Number 2 | February 2021 329

ARTICLE RAJAGOPALAN et al

involves sequential testing of at least three tests, and with our 7. Morrissette JD, Colliton RP, Spinner NB. Defective intracellular transport
identification of a novel copy-number variant within the and processing of JAG1 missense mutations in Alagille syndrome. Hum
Mol Genet. 2001;10:405–413.
NOTCH2 gene, a fourth test involving copy-number evalua- 8. Tada M, Itoh S, Ishii-Watabe A, Suzuki T, Kawasaki N. Functional analysis
tion of NOTCH2 should be added to this testing schema. The of the Notch ligand Jagged1 missense mutant proteins underlying
variety of variants identified in the cohort we describe here Alagille syndrome. FEBS J. 2012;279:2096–2107.
9. Myles RB. Familial short oesophagus. Br J Radiol. 1939;12:645–647.
highlights the ability of GS to detect all major classes of 10. Cornell University. Aligning sequence reads, clone sequences and
variants, allowing for single-test diagnostics rather than serial assembly contigs with BWA-MEM. https://arxiv.org/abs/1303.3997.
testing strategies. As GS becomes more available as a clinical Accessed 16 October 2020.
11. Van der Auwera GA, Carneiro MO, Hartl C. From FastQ data to high
testing option, it is reasonable to recommend that it replace confidence variant calls: the Genome Analysis Toolkit best practices
NGS panels and deletion/duplication analysis as a first-tier pipeline. Curr Protoc Bioinformatics. 2013;43:11.10.1–11.10.33.
testing strategy for ALGS. A reduction in the time and costs 12. Pedersen BS, Collins RL, Talkowski ME, Quinlan AR. Indexcov: fast
coverage quality control for whole-genome sequencing. Gigascience.
associated with multiple tests may prove to be advantageous 2017;6:1–6.
over the challenges in implementing GS as a first-tier 13. Pedersen BS, Bhetariya PJ, Brown J, et al. Somalier: rapid relatedness
diagnostic test for ALGS, and a similar advantage may be estimation for cancer and germline studies using efficient genome
sketches. Genome Med. 2020;12:62.
found in employing this testing strategy with other mono- 14. Kim S, Scheffler K, Halpern AL, et al. Strelka2: fast and accurate calling of
genic or oligogenic Mendelian disorders that require focused germline and somatic variants. Nat Methods. 2018;15:591–594.
GS data analysis of only one or a few genes. 15. Yang H, Wang K. Genomic variant annotation and prioritization with
ANNOVAR and wANNOVAR. Nat Protoc. 2015;10:1556–1566.
16. Abyzov A, Urban AE, Snyder M, Gerstein M. CNVnator: an approach to
SUPPLEMENTARY INFORMATION discover, genotype, and characterize typical and atypical CNVs from
family and population genome sequencing. Genome Res.
The online version of this article (https://doi.org/10.1038/s41436- 2011;21:974–984.
020-00989-8) contains supplementary material, which is available 17. Zhu M, Need AC, Han Y, et al. Using ERDS to infer copy-number variants
to authorized users. in high-coverage genomes. Am J Hum Genet. 2012;91:408–421.
18. Chen X, Schulz-Trieglaff O, Shaw R, et al. Manta: rapid detection of
structural variants and indels for germline and cancer sequencing
DATA AVAILABILITY applications. Bioinformatics. 2016;32:1220–1222.
19. Karczewski KJ, Francioli LC, Tiao G, et al. The mutational constraint
The GS datasets generated and analyzed during the current study
spectrum quantified from variation in 141,456 humans. Nature.
are not publicly available as they are patient samples with unique 2020;581:434–443.
mutations that are impossible to guarantee anonymity for. 20. Jaganathan K, Kyriazopoulou Panagiotopoulou S, McRae JF, et al.
Predicting splicing from primary sequence with deep learning. Cell.
2019;176:535–548. e524
ACKNOWLEDGEMENTS 21. Kent WJ. BLAT-the BLAST-like alignment tool. Genome Res.
This work was funded by the Fred and Suzanne Biesecker 2002;12:656–664.
22. Penton AL, Leonard LD, Spinner NB. Notch signaling in human
Pediatric Liver Center at the Children’s Hospital of Philadelphia (N. development and disease. Semin Cell Dev Biol. 2012;23:450–457.
B.S., K.M.L., D.A.P) and by R01-HG009708 (R.R. and L.K.C.) and 23. Kamath BM, Bauer RC, Loomes KM, et al. NOTCH2 mutations in Alagille
R01-DK081702-05 (N.B.S.). syndrome. J Med Genet. 2012;49:138–144.
24. Guegan K, Stals K, Day M, Turnpenny P, Ellard S. JAG1 mutations are
found in approximately one third of patients presenting with only one or
DISCLOSURE two clinical features of Alagille syndrome. Clin Genet. 2012;82:33–40.
25. Sie L, Looneg S, Tan EK. Utility of lymphoblastoid cell lines. J Neurosci Res.
The authors declare no conflicts of interest. 2009;87:1953–1959.
26. Ansari M, Poke G, Ferry Q, et al. Genetic heterogeneity in Cornelia de
Publisher’s note Springer Nature remains neutral with regard to Lange syndrome (CdLS) and CdLS-like phenotypes with observed and
jurisdictional claims in published maps and institutional predicted levels of mosaicism. J Med Genet. 2014;51:659–668.
27. Rexach J, Lee H, Martinez-Agosto JA, Nemeth AH, Fogel BL. Clinical
affiliations. application of next-generation sequencing to the practice of neurology.
Lancet Neurol. 2019;18:492–503.
REFERENCES
1. Abou Tayoun AN, Krock B, Spinner NB. Sequencing-based diagnostics for
pediatric genetic diseases: progress and potential. Expert Rev Mol Diagn. Open Access This article is licensed under a Creative Commons
2016;16:987–999. Attribution-NonCommercial-NoDerivatives 4.0 International
2. Retterer K, Juusola J, Cho MT, et al. Clinical application of whole-exome License, which permits any non-commercial use, sharing, distribution and
sequencing across clinical indications. Genet Med. 2016;18:696–704. reproduction in any medium or format, as long as you give appropriate credit to
3. Yang Y, Muzny DM, Reid JG, et al. Clinical whole-exome sequencing for the original author(s) and the source, and provide a link to the Creative
the diagnosis of mendelian disorders. N Engl J Med. Commons license. You do not have permission under this license to share
2013;369:1502–1511. adapted material derived from this article or parts of it. The images or other third
4. Mitchell E, Gilbert M, Loomes KM. Alagille syndrome. Clin Liver Dis. party material in this article are included in the article’s Creative Commons
2018;22:625–641. license, unless indicated otherwise in a credit line to the material. If material is not
5. Gilbert MA, Bauer RC, Rajagopalan R, et al. Alagille syndrome mutation included in the article’s Creative Commons license and your intended use is not
update: Comprehensive overview of JAG1 and NOTCH2 mutation permitted by statutory regulation or exceeds the permitted use, you will need to
frequencies and insight into missense variant classification. Hum Mutat. obtain permission directly from the copyright holder. To view a copy of this
2019;40:2197–2220. license, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
6. Gilbert MA, Spinner NB. Genetics of Alagille syndrome. In: Kamath BM,
Loomes KM, editors. Alagille syndrome: pathogenesis and clinical
management. Cham, Switzerland: Springer; 2018. p. 33–48. © The Author(s) 2020

330 Volume 23 | Number 2 | February 2021 | GENETICS in MEDICINE