Abstract
Purpose:
SouthSeq is a translational research study that performed genome sequencing (GS) for infants with symptoms suggestive of a genetic disorder. Recruitment targeted racial/ethnic minorities and rural, medically underserved areas in the Southeastern US that are historically under-represented in genomic medicine research.
Methods:
GS and analysis were performed for 367 infants to detect disease-causal variation concurrent with standard of care evaluation and testing.
Results:
Definitive diagnostic (DD) or likely diagnostic (LD) genetic findings were identified in 30% of infants and 14% harbored an uncertain result. Only 43% of DD/LD findings were identified via concurrent clinical genetic testing suggesting that GS testing is better for obtaining early genetic diagnosis. We also identified phenotypes that correlate with the likelihood of receiving a DD/LD finding, such as craniofacial, ophthalmologic, auditory, skin, and hair abnormalities. We did not observe any differences in diagnostic rates between racial/ethnic groups.
Conclusion:
We describe one of the largest-to-date GS cohorts of ill infants, enriched for African American and rural patients. Our results demonstrate the utility of GS as it provides early in life detection of clinically relevant genetic variation not identified via current clinical genetic testing, particularly for infants exhibiting certain phenotypic features.
Keywords: genome sequencing, infants, utility, diagnostic yield, genetic diagnosis
INTRODUCTION
Genome Sequencing (GS) holds potential value for critically ill infants with signs suggestive of a genetic disorder. Early diagnosis may be beneficial in the short-term for disease management and treatment but may also shorten the length of the “diagnostic odyssey” that often accompanies rare disease symptoms. Not only is GS capable of detecting a variety of genetic variant types (SNVs, indels, CNVs, aneuploidy), it also allows for assessment of genes in a phenotype-independent manner, a feature that is particularly important when testing young patients where clinical presentation may not be well-defined until later in life (e.g., intellectual disability). With the discovery of more than 4,000 genes that contribute to Mendelian diseases1,2, and with many more yet to be discovered, such comprehensiveness is valuable.
Although GS is being used to genetically diagnose pediatric patients with rare disease3–5, extensive use of GS to diagnose acutely ill infants is relatively lacking. A few groups have employed GS in a neonatal intensive care unit (NICU) setting6–8 and diagnostic rates range from ~20-50% depending on a variety of factors, including patient selection and testing/analysis methods. While GS holds promise for diagnosing and improving outcomes for ill infants9, many practical questions remain and additional studies on the use of GS as a first-line test are needed; this includes identifying phenotypic features that correlate with diagnostic success rate. Further, it is important to evaluate GS testing in underserved communities such as African Americans, who are sharply under-represented in existing NICU-based translational genomics studies.
SouthSeq is a clinical research study funded as part of the Clinical Sequencing Evidence-Generating Research (CSER) consortium and aims to use GS to detect causal genetic variation in a cohort of NICU infants with phenotypes suggestive of genetic disease (concurrent with standard of care) and evaluate GS as a first-line test to provide an early genetic diagnosis to improve outcomes of affected patients. SouthSeq targets enrollment of a diverse population of babies representing racial/ethnic minorities as well as those from rural, medically underserved areas. While ongoing, this report describes SouthSeq results from completed analysis of the first 367 probands enrolled across five different clinical sites in the Southeastern US between February 2018 to July 2020. Our study population (31 days mean age at enrollment) is sex-balanced (48% female) and enriched for individuals from diverse and medically underserved populations (74%). All 367 affected babies received GS, 30% of which received a genetic diagnosis and an additional 14% received results of uncertain significance. Our results highlight substantial diagnostic utility for GS, as 57% of GS-detected diagnostic variants were not detected by concurrent clinical genetic testing. Moreover, 15% (8 of 53) of infants who did not receive clinical genetic testing as part of standard care harbored a GS-detected diagnostic result. Further, we found that infants exhibiting abnormal craniofacial, ophthalmologic, auditory and/or skin/hair features were significantly more likely to receive a genetic diagnosis via GS than infants without such features. Finally, we show that although significant technical differences in the interpretation process do exist, diagnostic rates among African American infants are similar to those observed in European American babies.
MATERIALS AND METHODSs
Recruitment information
There was no public recruitment for this study. Participant infants were enrolled from the NICU, a high-risk prenatal clinic, or a pediatric unit at one of five clinical sites; University of Alabama at Birmingham/Children’s of Alabama (Birmingham, AL, USA), University of Mississippi Medical Center (Jackson, MS, USA), Woman’s Hospital (Baton Rouge, LA, USA), University of Louisville (Louisville, KY, USA), and Children’s Hospital New Orleans (New Orleans, LA, USA). At least one parent/legal guardian was required to consent for study participation. Antenatal consent was also an option for parents when phenotypic features that met enrollment criteria were detected prenatally. Translated consent documents and interpretation services were available for Spanish-speaking participants. A custom-built online platform was used to collect de-identified clinical information for dissemination to the research study team (e.g., consent documentation, demographics, birth history, phenotype, concurrent clinical genetic testing, and opt-in to secondary findings).
Inclusion/exclusion criteria
For study inclusion, a baby must be inpatient (e.g., neonatal, surgical, cardiac, or pediatric intensive care unit), be in the first 12 months of life (one baby enrolled was 379 days old), and exhibit a pattern of congenital anomalies consistent with a genetic disorder and/or present with an unexplained major medical condition (e.g., seizures, metabolic abnormality). Babies were excluded from the study if they had findings consistent with a known chromosomal aneuploidy (e.g., Trisomy 13, 18, 21 and monosomy X), exhibited anomalies known to result in low diagnostic yield for genetic causes (e.g. isolated gastroschisis, hydronephrosis), or had findings consistent with confirmed teratogenic exposure (e.g. hydantoin, valproate) or congenital infection (e.g. TORCH). Stillborn infants, or babies who died soon after birth, were enrolled if they met inclusion criteria.
Genome sequencing
Peripheral or cord blood samples collected in EDTA tubes were sent to the HudsonAlpha Clinical Services Laboratory (CSL, a CAP/CLIA-certified genetic testing lab) for DNA extraction (QIAsymphony) and storage. Sequencing libraries were constructed from genomic DNA using the CSL’s custom genome library preparation protocol. DNA library fragments were sequenced from both ends (paired) with a read length of 150 base pairs using the Illumina HiSeq X or NovaSeq 6000, with targeted mean coverage depth of 30X and >80% of bases covered at 20X. Sequence reads were aligned to GRCh38 using DRAGEN10 or the Sentieon implementation11 of BWA-mem. SNVs/indels were called using DRAGEN and GATK12 or Strelka13.
GS copy number variant calling
Copy number variants (CNVs) were called from GS bam files using DELLY14, ERDS15, Manta16, and CNVnator17. Overlapping calls with at least 90% reciprocity or large calls (>100,000 bp) containing less than 75% segmental duplications were retained if they were observed in eight or fewer unaffected individuals in an in-house database and at less than 1% in population frequency databases (Thousand Genomes18, gnomAD19). CNVs that survived filtration were subsequently analyzed for potential disease relevance. All rare CNVs found within 5 kb of an established developmental delay/intellectual disability gene, within 5 kb of a MIM2 disease-associated gene, or intersecting one or more exons of any gene, were subject to manual curation. CNVs were classified according to ACMG/ClinGen guidelines20.
Annotation, filtering, and variant classification
Identified SNVs and indels were annotated, filtered, and visualized using an in-house software platform. Variants that survived filtration were manually curated and classified as pathogenic, likely pathogenic, or uncertain using ACMG/AMP guidelines21. Additionally, variants were assigned a case-level designation in the context of clinical presentation that linked variation to confidence in causation, using the terms definitive diagnostic (DD), likely diagnostic (LD), or uncertain. The CSER consortium, spanning a collaborative group of sites performing translational genomic research in a variety of settings, established this case-level classification system and a manuscript describing development and implementation is in preparation. Most infants harboring pathogenic variation received case-level designation of DD, while those harboring a likely pathogenic variant were classified as LD. However, for a few cases a pathogenic or likely pathogenic variant received case-level designation of uncertain due to insufficient zygosity (only one heterozygous variant was identified in a gene associated with an autosomal recessive or X-linked recessive condition with significant phenotypic overlap), phenotype mismatch, and/or unknown phase. In contrast, some probands that harbored a variant of uncertain significance received case-level designation of LD based on the specificity of the match between observed and expected phenotypes.
Variant validation
GS testing was conducted in a CAP/CLIA-certified laboratory, while variant analysis/interpretation was conducted as part of a research protocol. Variants deemed to be returnable were clinically tested (via Sanger or array) to confirm variant presence, determine variant inheritance (when parent samples were available), and generate a report with clinical interpretation. CNVs were returned in accordance with the research protocol if they were too small to be confirmed via clinical array testing. HGVS nomenclature of all returned variants was verified using VariantValidator22.
Return of Results
Parents or legal guardians of babies enrolled received GS results via a randomized clinical trial (NCT03842995) comparing standard of care return (genetic counselors) to return by trained non-geneticist healthcare providers (neonatologist, nurse practitioner). Results of this trial will be published elsewhere. Parents or legal guardians of babies also had the option to receive secondary findings; pathogenic or likely pathogenic variation in an ACMG SFv2.0 gene23. GS result reports were placed in the infant’s medical record and follow-up medical care was managed by the appropriate clinical care teams.
RESULTS
Demographics
We enrolled 367 babies (365 total families; two families each with two affected babies) with signs suggestive of an underlying genetic disorder (see Methods for study enrollment criteria; Table 1). Patient recruitment occurred at five clinical sites, including the University of Alabama at Birmingham/Children’s of Alabama (UAB, n=139, 38%), University of Mississippi Medical Center (UMMC; n=118, 32%), Woman’s Hospital in Baton Rouge (BR; n=61, 17%), Children’s Hospital in New Orleans (LSU; n=31, 8%) and University of Louisville (UL; n=18, 5%). UAB and UMMC represent clinical sites where recruitment first began and thus had higher enrollment totals.
Table 1.
Study demographics
| Individuals (n (%)) | Individuals with DD/LD Result (n (%)) | |
|---|---|---|
| Clinical Sites | Total (%) | Total (%) |
| University of Alabama at Birmingham and Children’s of Alabama (UAB) | 139 (38) | 42 (30) |
| University of Mississippi Medical Center (UMMC) | 118 (32) | 33 (28) |
| Woman’s Hospital (BR) | 61 (17) | 20 (33) |
| Children’s Hospital in New Orleans (LSU) | 31 (8) | 7 (23) |
| University of Louisville (UL) | 18 (5) | 7 (39) |
| Sex | Total (%) | Total (%) |
| Male | 191 (52) | 54 (28) |
| Female | 176 (48) | 55 (31) |
| Race/ethnicity | Total (%) | Total (%) |
| Black or African American | 126 (34) | 39 (31) |
| White or European American | 187 (51) | 50 (27) |
| More than one category | 34 (9) | 10 (29) |
| Hispanic/Latino(a) only | 13 (4) | 7 (54) |
| Other a | 7 (2) | 3 (43) |
| Racial/Ethnic minority OR medically underserved b | 271 (74) | 81 (30) |
| Family Structure | Total (%) | |
| Proband + biological parents | 234 (64) | -- |
| Proband + one biological parent | 104 (28) | -- |
| Proband only | 29 (8) | -- |
| Age | Average (range) | |
| Age of proband at enrollment (days) | 31 (0-379); median 14 ± 53 days | -- |
| Mother age at delivery (years) | 27 (16-49) | -- |
| Birth History | Average (range) | |
| Gestational age of proband (weeks) | 36 (22-42) | -- |
| Birth weight (g) | 2531.1 (310-5930) | -- |
| Birth length (cm) | 44.97 (20-57) | -- |
| OFC (cm) | 32.3 (18-52.6) | -- |
| Apgar scores (1 min; 5 min) | 5.9 (0-10); 7.5 (0-10) | -- |
| Top HPO terms | Total (%) | Total (%) |
| Abnormality of the face (HP:0000271) | 83 (23%) | 35 (42) |
| Intrauterine growth restriction (HP:0001511) | 80 (22%) | 26 (33) |
| Birth weight less than 10th percentile (HP:0001518) | 64 (18%) | 18 (28) |
| Atrial Septal Defect (HP:0001631) | 59 (16%) | 19 (32) |
| Polyhydramnios (HP:0001561) | 59 (16%) | 20 (35) |
Other includes Asian, Middle Eastern, North African or Mediterranean or Unknown;
Based on self-reported income, zip code, race/ethnicity
Participant babies had a mean age of 31 days (range 0-379 days) at time of enrollment and 52% were male. Forty-nine percent of participants (n=180) were self-reported non-white and 74% (n=271) represent racial/ethnic minorities and/or reside in rural, medically underserved areas24 (Table 1). With input from NICU providers and medical geneticists, we collected patient clinical information within 13 high-level NICU-relevant phenotypic categories. These categories included 112 pre-selected Human Phenotype Ontology25 (HPO) terms (average of seven HPO terms per category), of which 89 were observed in at least one infant in the study population (Table S1). The five most common HPO terms observed across participants included abnormality of the face (n=83, 23%), intrauterine growth restriction (n=80, 22%), birth weight less than the 10th percentile (n=64, 18%), atrial septal defect (n=59, 16%), and polyhydramnios (n=59, 16%; Tables 1 and S1).
GS was only performed for enrolled infants, but when available, parental blood samples were obtained for Sanger confirmation to assess inheritance of GS-identified potentially clinically relevant variants. 234 infants were enrolled along with both biological parents, 104 were enrolled with one biological parent, and 29 were enrolled as proband only (Table 1).
Diagnostic yield
Of the 367 sequenced probands, 160 (44% of our cohort) received a result associated with the primary indication for testing. Results designated as definitive diagnostic (DD) or likely diagnostic (LD) were identified in 109 babies (30%), while an additional 51 participants (14%) received an uncertain case-level result (Figure 1A; Table S2). Returnable variation was identified across 107 genes (64 harboring DD/LD findings, 43 harboring an uncertain finding). Eight genes were found to harbor unique DD/LD variants in two or more unrelated probands (Tables S2 and S3), with six babies genetically diagnosed with CHARGE syndrome (CHD7, MIM:214800) and six with Noonan syndrome (PTPN11, MIM: 163950). Turnaround time from enrollment to report generation averaged 73 days, which included ~30 days for Sanger confirmation in an independent CAP/CLIA laboratory.
Figure 1.
Utility of genome sequencing as a first-line genetic test for infants (n=367 babies). (A) Diagnostic yield of SouthSeq study population. Thirty percent (n=109) of study participants received a definitive diagnostic (DD) or likely diagnostic (LD) finding; 14% (n=51) received an uncertain finding, and no genetic findings of interest were identified in the remaining 56% (n=207). (B) Percentage of findings that fall within each mode of inheritance category, including de novo, compound heterozygous or homozygous, X-linked, inherited, or unknown. Unknown represents heterozygous variants (SNV or CNV) where one or both parents were unavailable for testing, or inheritance could not be determined (e.g. non-paternity). (C) Types of variation represented by DD/LD findings, including missense, nonsense, frameshift, splice, and CNV; one case of uniparental disomy (UPD) not shown.
Additionally, 89% percent of families consented to receive secondary results in an ACMG SFv2.0 gene23 with seven pathogenic/likely pathogenic (P/LP) variants detected across six infants (1.6%). One baby harbored two P/LP findings in two different genes, MSH6 and ACTC1 (Table S4).
In 30% of DD/LD cases the returnable variant occurred de novo in a dominant disease gene, while 19% of DD/LD infants inherited compound heterozygous or homozygous variation in a recessive disease gene. Fourteen percent of DD/LD infants inherited a heterozygous variant in a dominant disease gene from a parent, some of which were affected (Figure 1B). An additional 2% of DD/LD cases were maternally inherited X-linked recessive findings in males. Finally, 35% of DD/LD cases were of unknown inheritance due to biological parent samples being unavailable (Figure 1B). DD/LD findings represent a variety of variant types; 33% of identified DD/LD variants result in missense, 17% in nonsense, and 17% in frameshift. Further, 7% were predicted to disrupt splicing (at or near a canonical splice site), and 26% represent a copy number variant (CNV; Figure 1C). There was also one DD case resulting from uniparental disomy.
Factors that drive diagnostic yield
We compared available clinical, demographic, and phenotypic variables to identify potential correlations with diagnostic yield, including site of enrollment, sex, race/ethnicity, access to care, and phenotype (Tables 1 and 2). We observed no difference in diagnostic yield between the two primary clinical sites that account for most study enrollments (30% at UAB vs. 28% at UMMC), and the rates at the other three nurseries are similar albeit more variable due to smaller enrollment totals (33%, 23%, and 39% at Woman’s Hospital, Children’s Hospital in New Orleans, and University of Louisville, respectively), suggesting roughly similar overall diagnostic yields across NICU locations (Table 1).
Table 2.
Diagnostic yield and phenotype. GS testing may yield increased diagnoses for infants with certain phenotypic features.
| Phenotype categories | # Probands | # Probands with DD/LD (%) | P-value (odds ratio) | # Probands with Uncertain (%) | P-value (odds ratio) |
|---|---|---|---|---|---|
| Study-wide totals | 367 | 109 (30) | --- | 51 (14) | --- |
| Prenatal b | 148 (40) | 50 (34) | 0.16 (1.40) | 21 (14) | 0.88 (1.04) |
| Craniofacial/Ophthalmologic/Auditory | 144 (39) | 60 (42) | <0.0001 (2.54)a | 23 (16) | 0.36 (1.32) |
| Cardiac/Congenital heart malformations | 142 (39) | 44 (31) | 0.73 (1.11) | 18 (13) | 0.64 (0.84) |
| Growth | 139 (38) | 43 (31) | 0.72 (1.09) | 15 (11) | 0.21 (0.65) |
| Skeletal/Limb abnormalities | 99 (27) | 34 (34) | 0.25 (1.35) | 18 (18) | 0.17 (1.58) |
| Neurological/Muscular | 91 (25) | 25 (28) | 0.69 (0.87) | 22 (24) | 0.003 (2.72)a |
| Genitourinary abnormalities | 86 (23) | 31 (36) | 0.17 (1.47) | 14 (16) | 0.48 (1.28) |
| Brain malformations/abnormal imaging | 66 (18) | 21 (32) | 0.66 (1.13) | 10 (15) | 0.70 (1.13) |
| Gastrointestinal | 50 (14) | 9 (18) | 0.06 (0.48) | 8 (16) | 0.66 (1.21) |
| Hematologic/Immunologic | 36 (10) | 7 (19) | 0.25 (0.56) | 2 (6) | 0.20 (0.34) |
| Metabolic | 30 (8) | 9 (30) | 0.99 (1.05) | 4 (13) | 0.99 (0.95) |
| Skin/Hair | 27 (7) | 13 (48) | 0.03 (2.43)* | 4 (15) | 0.78 (1.08) |
| Endocrine | 7 (2) | 0 (0) | 0.11 (0.00) | 0 (0) | 0.60 (0.00) |
Odds ratios and p-values calculated using Fisher’s exact test
denotes statistical significance.
IUGR, abnormal amniotic fluid levels, cystic hygroma, hydrops fetalis
We observed no difference in diagnostic yield between self-reported African American (AA) and European American (EA) babies (31% vs 27%, respectively). However, analysis of AA genomes required more manual curation and Sanger testing than EA genomes. After primary variant filtration, ~1.4X more variants were retained and required manual curation for an AA genome (average of 300 variants/proband, n=111 genomes) compared with an EA genome (average of 221 variants per proband, n=160 genomes). Additionally, we Sanger tested parent samples to determine inheritance/phase for candidate variants prior to final interpretation given that some ACMG/AMP evidence codes require such information. For EA babies that were enrolled with both biological parents, we averaged 0.79 Sanger tests/proband (116 Sanger tests across 146 EA probands) compared with 0.96 Sanger tests per proband for AA babies (54 Sanger tests across 56 AA probands; p=0.033, two-proportion z-test).
We also examined diagnostic yield in the context of phenotype. As previously mentioned, NICU providers and medical geneticists provided patient phenotype information within 13 high-level categories containing 112 NICU-relevant HPO terms (Tables 2 and S1). Because any given baby may exhibit more than one phenotype (HPO term) in each category, we conducted calculations based on total unique individuals in each. Phenotypic categories that were attributed to the largest number of participants included prenatal (e.g., IUGR, amniotic fluid levels, cystic hygroma; n=148, 40%), craniofacial/ophthalmologic/auditory (n=144, 39%), cardiac/congenital heart malformations (n=142, 39%), growth (e.g., birth weight/length, head circumference, failure to thrive; n=139, n=38%), and skeletal/limb abnormalities (n=99, 27%; Table 2).
We found substantial variation in DD/LD and uncertain yield across phenotypic categories. For example, DD/LD rates ranged from 18% among babies with gastrointestinal findings to 48% for those with skin/hair findings. We also compared DD/LD or uncertain rates between babies who do and do not exhibit phenotypic features across each category. We found that babies with craniofacial/ophthalmologic/auditory abnormalities had the largest and most significant enrichment for DD/LD findings (OR=2.54, p<0.0001), followed by skin/hair abnormalities (OR=2.43, p=0.03). Our data also suggest that babies with neurological and/or muscular findings (OR=2.72, p=0.003) are more likely to receive an uncertain result (Table 2). We found no phenotypic category that is significantly depleted of DD/LD findings, although we observed a substantially lower diagnostic yield for babies who presented with abnormal gastrointestinal features (OR=0.48, p=0.06).
Utility of genome sequencing
The SouthSeq study design explicitly stated that standard clinical care and testing should be conducted regardless of SouthSeq participation. Reflecting this, the vast majority (86% percent, n=314 of 367) of SouthSeq babies received at least one clinical genetic test in parallel with SouthSeq GS, with an average of 1.7 (range 0-5) tests per baby. Most babies received clinical CNV testing (75%, n=234), while many received postnatal karyotype/FISH (39%, n=124), prenatal screening or testing (42%, n=132; e.g., quad screen, amniocentesis, maternal serum screen, noninvasive prenatal screening (NIPS)) or postnatal single-gene/panel testing (26%, n=81). A small fraction received clinical exome testing (3%, n=11; Table S5).
Only 43% (43 of 101) of GS-detected DD/LD findings were also found by clinical genetic testing (Tables 3, Table S5). Among these, 58% (25 of 43) were identified by CNV testing, 37% (16 of 43) by gene panel, and 5% (2 of 43) by clinical exome. Sixteen percent (8 of 49) of GS-detected uncertain results were also identified by clinical genetic testing (Tables 3, Table S5), mostly via CNV testing (88%, n=7). For 20/51 babies where GS and clinical genetic testing identified the same genetic variant (including uncertain cases), SouthSeq was able to provide families with inheritance information (including nine de novo and 11 inherited variants) not generated by standard testing.
Table 3.
Overlap of findings between GS findings and clinical genetic testing (GS, genome sequencing; CT, clinical genetic testing. Only infants that received clinical genetic testing (n=314) are included in the calculations.
| Case-level interpretation | SouthSeq CT performed (314 infants) | GS+/CT+ total (%) | GS+/CT− total (%) |
|---|---|---|---|
| DD/LD | 101 | 43 (43) | 58 (57) |
| Uncertain | 49 | 8 (16) | 41 (84) |
| No returnables | 164 | --- | --- |
Fifty-seven percent of GS-detected DD/LD events were not identified via clinical genetic testing (Tables 3 and S5), and in all cases this was because the ordered clinical genetic test was not capable of detecting the GS-identified variation. More specifically, 1) GS detected a SNV or indel but no single gene/panel test was ordered (43 of 58 cases); 2) GS detected a SNV or indel in a gene not on a clinically ordered single gene/panel test (13 of 58 cases); or 3) GS detected a CNV when clinical CNV testing was not ordered (2 of 58 cases). Case studies that highlight DD/LD results not detected by clinical genetic testing, and that demonstrate the benefit of GS testing, are provided in Table 4. Additionally, there were 10 GS positive cases (eight DD/LD and two uncertain) found among 53 infants who had no clinical genetic testing as part of standard of care.
Table 4.
Cases studies highlighting the utility of genome sequencing as a first-line genetic test.
| Participant | SS-47 | SS-231 | SS-51 |
|---|---|---|---|
| Clinical phenotype | Open lip schizencephaly, IUGR, congenital microcephaly, external ear malformation, central hypotonia, possible dysgenesis of corpus callosum, malformed large right ventricle, gliosis, calcification in right brain hemisphere and large right ventricle, high arched hard palate, poor head control, absent Moro reflex, upgoing toe signs, ankle clonus; TORCH suspected | Polyhydramnios, fetal bradycardia, open and flat anterior fontanelle, upturned nasal tip, tented upper lip, high arched palate, posterior hair whorl, central hypotonia, downslanted palpebral fissures, muscular hypotonia, low-set ears, hand clenching, poor reflexes, required intubation/failed extubation | IUGR, low set ears, undescended testis, jejunal atresia, mild cardiomegaly, rocker bottom feet, bilateral 5th finger clinodactyly, moderate petechiae on back, buttocks, and flank, café au lait spots in lumbar/sacral area, abnormal EEG suggestive of seizure activity |
| Clinical testing completed | TORCH screening, CMV testing, CNV testing-all non-diagnostic | Karyotype, CNV testing, SMA testing, methylation testing, gene panel testing (x2), repeat expansion testing, mitochondrial genome panel, brain MRI, muscle biopsy, biochemical studies - all non-diagnostic | Karyotype, CNV testing, seizure panel – all non-diagnostic |
| Findings of SouthSeq GS | NM_001845.5(COL4A1):c.3556G>A (p.G1186S) | GRCh38/hg38 (chr17:chr17:42690150-42694200)x1; NM_003632.2(CNTNAP1):c.2901_2902del (p.C968Ffs*11) | NM_001020658.1(PUM1):c.3439C>T (p.R1147W) |
| Associated disease (MIM#) | Brain small vessel disease with or without ocular anomalies (175780) | Lethal congenital contracture syndrome 7 (618186); Hypomyelinating neuropathy, congenital 3 (616286) | Spinocerebellar ataxia 47 (617931) |
| Variant inheritance | de novo | unknown; paternal (variants in trans) | de novo |
| Variant-level classification | Likely pathogenic (PS2, PM2, PP3) | Pathogenic (PVS1, PM2, PM3); Pathogenic (PVS1, PM2, PM3, PP1S) | Pathogenic (PS2, PS3, PS4M, PM2, PP3) |
| Case-level classification | Likely diagnostic | Diagnostic | Diagnostic |
Forty-one of the GS-identified uncertain findings were not detected by clinical genetic testing, likely owing at least partially to the inherent uncertainty related to these findings (e.g., questions about phenotype match, uncertainty of variant impact, etc.). Nine of the GS uncertain cases included a likely deleterious variant in a candidate gene not currently associated with disease, but which may become a disease gene in the future (e.g., via GeneMatcher submission26). An additional nine uncertain GS results arose from identification of a single P/LP variant in a gene associated with an autosomal recessive condition for which there was significant phenotypic overlap. For example, in SS-114 (Table S2) we identified a single paternally inherited nonsense variant in TMCO1 associated with autosomal recessive craniofacial dysmorphism, skeletal anomalies, and mental retardation syndrome (MIM: 213980). Although a second variant was not detected via GS, this result was returned due to significant phenotypic overlap between the proband and features reported for TMCO1-associated disease, including cleft palate, vertebral anomalies, hypotonia, hip dislocation, and genitourinary anomalies. While we returned such variation in SouthSeq, classified as case-level uncertain due to insufficient zygosity, none of these variants were returned via clinical genetic testing.
Thirty-two VUSs were detected via clinical genetic testing and, although detected via GS, were not returned by the study (Table S5). Most of these participants were GS negative (n=18), while nine had GS-identified DD/LD findings, and five had GS-identified uncertain findings that did not match the variants returned via clinical genetic testing. Reasons for discrepancies in return between the clinical lab and SouthSeq include: 1) variant in a highly penetrant disease gene inherited from an unaffected parent; 2) phenotype mismatch; 3) observed too frequently in population frequency databases; 4) VUS in a gene associated with an autosomal recessive condition without a match to observed phenotypes.
DISCUSSION
In SouthSeq, we used GS concurrently with standard of care, in a diverse cohort of infants from underserved and rural populations who presented with phenotypic features suggestive of a genetic disorder. Sanger testing was completed in parent samples (when available) to determine variant inheritance and aid interpretation. 44% of SouthSeq participants were found to harbor a genetic variant associated with indication for testing; 30% received a DD/LD result, and an additional 14% received a result of uncertain significance. This observed diagnostic rate is in line with previous reports from other groups6–8, even in the absence of sequencing trios. While we did not sequence parent samples in SouthSeq to aid in identification of causal variation, doing so may have potentially increased the diagnostic rate3,27. Returnable findings include different variant types (SNVs, indels, CNVs) affecting 107 unique genes, highlighting the comprehensiveness of GS. Finally, we identified seven P/LP variants across six babies (one baby with two findings, 1.6%) in an ACMG SFv2.0 gene.
We observed no significant difference in diagnostic rate between the two largest racial/ethnic categories in SouthSeq (31% diagnostic rate for AA babies, 27% for EA babies), which would be expected in the context of disease-associated ultra-rare variation. AA babies did, however, require more manual curation of variants that appeared to be rare and damaging, and also required more Sanger testing in parent samples to determine variant inheritance (0.79 Sanger tests/EA trio babies, 0.96 tests/AA trio babies). These results are consistent with previous reports of ancestry-associated differences in clinical genetic testing procedures and are likely to result, at least in part, from under-representation of African alleles in population frequency databases28,29. Increased population frequency data for minority populations would likely reduce this discrepancy.
Because resources necessary to conduct GS are limited, identification of patients most likely to benefit from GS testing could maximize GS utility. As such, we sought to identify patient attributes that correlate with the likelihood of receiving a genetic diagnosis (Table 1). We grouped infant participants into 13 high-level phenotype categories based on 112 pre-defined HPO terms and compared rates of different finding types (DD/LD and uncertain) between babies that exhibit features falling within a specific category and those that do not (Table 2). We found that infants that exhibited abnormal craniofacial, ophthalmologic, auditory, skin, or hair findings were ~2.5X more likely (p<0.0001) to receive a DD/LD result compared with babies who did not exhibit these phenotypic attributes. Conversely, although not statistically significant, babies with gastrointestinal phenotypes received a DD/LD result at ~50% the rate of babies that did not exhibit gastrointestinal abnormalities. Finally, individuals with neurological/muscular-associated features were 2.7X (p=0.003) more likely to receive a variant of uncertain significance, perhaps reflecting the fact that some neurological features are not observable in infancy30. Although others have conducted similar analyses31,32, future larger studies may help to further delineate clinical features most predictive of GS diagnostic potential. It is also important to note that some phenotypes in infants with a DD/LD result may not be obviously related to the GS-identified variant (Table 4), and conversely, not all phenotypes associated with a specific DD/LD result were observed in the given affected patient. Such incomplete phenotypic overlap may occur due to the presence of multiple underlying conditions but may also result from incompleteness in either the phenotypic spectra reported for rare genetic diseases or the phenotypes recorded as part of study enrollment.
Enrolled babies received GS concurrent with standard of care, including clinical genetic testing. Because of this, we were able to compare the results from standard clinical genetic testing with GS. We found that most GS DD/LD results (57%) were not detected via clinical genetic testing, and 8 DD/LD results were found among 53 infants (15%) who received no genetic testing as part of standard care; conversely, no DD/LD findings from clinical genetic testing were undetected by GS, although it should be noted that positive clinical testing results were made available to the SouthSeq study team for a few cases at time of GS analysis. As would be suspected, most of the DD/LD variation missed by standard testing resulted from the fact that the clinical tests ordered were not capable of detecting the relevant variants. In contrast to single-gene or gene-panel tests, for example, GS testing is independent of hypotheses about which specific disease genes may be involved. Further, early use of GS can reduce the number of tests required to obtain a diagnosis; for example, on average almost two in-parallel non-GS genetic tests were ordered per SouthSeq baby despite leading to an overall yield less than half that of GS. As such, GS has considerable potential to prevent the multiplicity of testing associated with the “diagnostic odyssey”33,34 that many rare-disease patients experience.
In conclusion, we have conducted GS testing for a cohort of infants affected with suspected congenital anomalies and enriched for individuals from medically underserved and historically underrepresented groups in genomics research. We show that infants with certain phenotypic features may benefit more from GS testing and highlight the comprehensiveness of GS and its benefit beyond standard clinical genetic testing. Our data strongly support using GS as a first-line genetic test for seriously ill infants.
Supplementary Material
Acknowledgements
We are grateful to the patients and their families who contributed to this study. We thank the HudsonAlpha Software Development and Informatics team and the Clinical Services Laboratory who contributed to data acquisition and analysis. The SouthSeq project (U01HG007301) is supported by the Clinical Sequencing Evidence-Generating Research (CSER) consortium which is funded by the National Human Genome Research Institute (NHGRI) with co-funding from the National Institute on Minority Health and Health Disparities (NIMHD) and the National Cancer Institute (NCI). More information about CSER can be found at https://cser-consortium.org/.
Footnotes
Ethics Declaration
All authors declare no competing financial interests in relation to the work described.
The review board at the University of Alabama at Birmingham (IRB-300000328) approved and monitored the study. All study participants (parent or legal guardian) were required to give written consent to participate in the study. All individual-level data was de-identified to the research team. The authors received and archived written patient consent to publish individual data.
Data availability
The sequencing data generated by this work is available through AnVIL/dbGaP35 via https://www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/study.cgi?study_id=phs002307.v1.p1. Interpreted variants have been placed into ClinVar36 under the study name ‘CSER-SouthSeq’ and can be accessed through https://www.ncbi.nlm.nih.gov/clinvar/?term=CSER-SouthSeq.
REFERENCES
- 1.Chong JX et al. The Genetic Basis of Mendelian Phenotypes: Discoveries, Challenges, and Opportunities. Am J Hum Genet 97, 199–215, doi: 10.1016/j.ajhg.2015.06.009 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Amberger JS, Bocchini CA, Scott AF & Hamosh A OMIM.org: leveraging knowledge across phenotype-gene relationships. Nucleic Acids Res 47, D1038–D1043, doi: 10.1093/nar/gky1151 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Bowling KM et al. Genomic diagnosis for children with intellectual disability and/or developmental delay. Genome Med 9, 43, doi: 10.1186/s13073-017-0433-1 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Stavropoulos DJ et al. Whole Genome Sequencing Expands Diagnostic Utility and Improves Clinical Management in Pediatric Medicine. NPJ Genom Med 1, doi: 10.1038/npjgenmed.2015.12 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Wright CF, FitzPatrick DR & Firth HV Paediatric genomics: diagnosing rare disease in children. Nat Rev Genet 19, 253–268, doi: 10.1038/nrg.2017.116 (2018). [DOI] [PubMed] [Google Scholar]
- 6.French CE et al. Whole genome sequencing reveals that genetic conditions are frequent in intensively ill children. Intensive Care Med 45, 627–636, doi: 10.1007/s00134-019-05552-x (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Kingsmore SF et al. A Randomized, Controlled Trial of the Analytic and Diagnostic Performance of Singleton and Trio, Rapid Genome and Exome Sequencing in Ill Infants. Am J Hum Genet 105, 719–733, doi: 10.1016/j.ajhg.2019.08.009 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Wang H et al. Optimized trio genome sequencing (OTGS) as a first-tier genetic test in critically ill infants: practice in China. Hum Genet 139, 473–482, doi: 10.1007/s00439-019-02103-8 (2020). [DOI] [PubMed] [Google Scholar]
- 9.Friedman JM et al. Genome-wide sequencing in acutely ill infants: genomic medicine’s critical application? Genet Med 21, 498–504, doi: 10.1038/s41436-018-0055-z (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Miller NA et al. A 26-hour system of highly sensitive whole genome sequencing for emergency management of genetic diseases. Genome Med 7, 100, doi: 10.1186/s13073-015-0221-8 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Kendig KI et al. Sentieon DNASeq Variant Calling Workflow Demonstrates Strong Computational Performance and Accuracy. Front Genet 10, 736, doi: 10.3389/fgene.2019.00736 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Van der Auwera GA et al. From FastQ data to high confidence variant calls: the Genome Analysis Toolkit best practices pipeline. Curr Protoc Bioinformatics 43, 11 10 11–11 10 33, doi: 10.1002/0471250953.bi1110s43 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Kim S et al. Strelka2: fast and accurate calling of germline and somatic variants. Nat Methods 15, 591–594, doi: 10.1038/s41592-018-0051-x (2018). [DOI] [PubMed] [Google Scholar]
- 14.Rausch T et al. DELLY: structural variant discovery by integrated paired-end and split-read analysis. Bioinformatics 28, i333–i339, doi: 10.1093/bioinformatics/bts378 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Zhu M et al. Using ERDS to infer copy-number variants in high-coverage genomes. Am J Hum Genet 91, 408–421, doi: 10.1016/j.ajhg.2012.07.004 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Chen X et al. Manta: rapid detection of structural variants and indels for germline and cancer sequencing applications. Bioinformatics 32, 1220–1222, doi: 10.1093/bioinformatics/btv710 (2016). [DOI] [PubMed] [Google Scholar]
- 17.Abyzov A, Urban AE, Snyder M & Gerstein M CNVnator: an approach to discover, genotype, and characterize typical and atypical CNVs from family and population genome sequencing. Genome Res 21, 974–984, doi: 10.1101/gr.114876.110 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Sudmant PH et al. An integrated map of structural variation in 2,504 human genomes. Nature 526, 75–81, doi: 10.1038/nature15394 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Collins RL et al. A structural variation reference for medical and population genetics. Nature 581, 444–451, doi: 10.1038/s41586-020-2287-8 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Riggs ER et al. Technical standards for the interpretation and reporting of constitutional copy-number variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics (ACMG) and the Clinical Genome Resource (ClinGen). Genet Med 22, 245–257, doi: 10.1038/s41436-019-0686-8 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Richards S et al. Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology. Genet Med 17, 405–424, doi: 10.1038/gim.2015.30 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Freeman PJ, Hart RK, Gretton LJ, Brookes AJ & Dalgleish R VariantValidator: Accurate validation, mapping, and formatting of sequence variation descriptions. Hum Mutat 39, 61–68, doi: 10.1002/humu.23348 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Kalia SS et al. Recommendations for reporting of secondary findings in clinical exome and genome sequencing, 2016 update (ACMG SF v2.0): a policy statement of the American College of Medical Genetics and Genomics. Genet Med 19, 249–255, doi: 10.1038/gim.2016.190 (2017). [DOI] [PubMed] [Google Scholar]
- 24.Health Resources and Services Administration, <https://data.hrsa.gov/> (2020).
- 25.Kohler S et al. The Human Phenotype Ontology in 2021. Nucleic Acids Res 49, D1207–D1217, doi: 10.1093/nar/gkaa1043 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Sobreira N, Schiettecatte F, Valle D & Hamosh A GeneMatcher: a matching tool for connecting investigators with an interest in the same gene. Hum Mutat 36, 928–930, doi: 10.1002/humu.22844 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Tan TY et al. A head-to-head evaluation of the diagnostic efficacy and costs of trio versus singleton exome sequencing analysis. Eur J Hum Genet 27, 1791–1799, doi: 10.1038/s41431-019-0471-9 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Bien SA et al. The Future of Genomic Studies Must Be Globally Representative: Perspectives from PAGE. Annu Rev Genomics Hum Genet 20, 181–200, doi: 10.1146/annurev-genom-091416-035517 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Manrai AK et al. Genetic Misdiagnoses and the Potential for Health Disparities. N Engl J Med 375, 655–665, doi: 10.1056/NEJMsa1507092 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Ceyhan-Birsoy O et al. Interpretation of Genomic Sequencing Results in Healthy and Ill Newborns: Results from the BabySeq Project. Am J Hum Genet 104, 76–93, doi: 10.1016/j.ajhg.2018.11.016 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Marinakis NM et al. Phenotype-driven variant filtration strategy in exome sequencing toward a high diagnostic yield and identification of 85 novel variants in 400 patients with rare Mendelian disorders. Am J Med Genet A 185, 2561–2571, doi: 10.1002/ajmg.a.62338 (2021). [DOI] [PubMed] [Google Scholar]
- 32.Gubbels CS et al. Prospective, phenotype-driven selection of critically ill neonates for rapid exome sequencing is associated with high diagnostic yield. Genet Med 22, 736–744, doi: 10.1038/s41436-019-0708-6 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Deuitch NT et al. “Doctors can read about it, they can know about it, but they’ve never lived with it”: How parents use social media throughout the diagnostic odyssey. J Genet Couns, doi: 10.1002/jgc4.1438 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Wu AC, McMahon P & Lu C Ending the Diagnostic Odyssey-Is Whole-Genome Sequencing the Answer? JAMA Pediatr 174, 821–822, doi: 10.1001/jamapediatrics.2020.1522 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Mailman MD et al. The NCBI dbGaP database of genotypes and phenotypes. Nat Genet 39, 1181–1186, doi: 10.1038/ng1007-1181 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Landrum MJ et al. ClinVar: public archive of interpretations of clinically relevant variants. Nucleic Acids Res 44, D862–868, doi: 10.1093/nar/gkv1222 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The sequencing data generated by this work is available through AnVIL/dbGaP35 via https://www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/study.cgi?study_id=phs002307.v1.p1. Interpreted variants have been placed into ClinVar36 under the study name ‘CSER-SouthSeq’ and can be accessed through https://www.ncbi.nlm.nih.gov/clinvar/?term=CSER-SouthSeq.