Phenotypic Expansion Illuminates Multilocus Pathogenic Variation

Abstract

Purpose:

Multilocus variation, pathogenic variants in two or more disease genes, can potentially explain the underlying genetic basis for apparent phenotypic expansion in cases for which the observed clinical features extend beyond those reported in association with a “known” disease gene.

Methods:

Analyses focused on 106 patients, 19 for which apparent phenotypic expansion was previously attributed to variation at known disease genes. We performed a retrospective computational re-analysis of whole exome sequencing data using stringent Variant Call File filtering criteria to determine whether molecular diagnoses involving additional disease loci might explain the observed expanded phenotypes.

Results:

Multilocus variation was identified in 31.6% (6/19) of families with phenotypic expansion and 2.3% (2/87) without phenotypic expansion. Intrafamilial clinical variability within 2 families was explained by multilocus variation identified in the more severely affected sibling.

Conclusions:

Our findings underscore the role of multiple rare variants at different loci in the etiology of genetically and clinically heterogeneous cohorts. Intrafamilial phenotypic and genotypic variability allowed a dissection of genotype-phenotype relationships in 2 families. Our data emphasize the critical role of the clinician in diagnostic genomic analyses and demonstrate that apparent phenotypic expansion may represent blended phenotypes resulting from pathogenic variation at more than one locus.

INTRODUCTION

The diagnostic path leading from clinical evaluation to genetic diagnosis is not always direct, and physicians have historically relied on hallmark patterns of history and physical examination findings, or phenotypic features, to formulate a differential diagnosis and guide clinical investigation. However, even ‘typical’ clinical presentations can be misleading, as illustrated by early reports of Fitzsimmons syndrome describing 6 individuals with a seemingly unique combination of spastic paraplegia, skeletal anomalies, dysarthria, and mild developmental delay.^1–3 Whole exome sequencing (WES) ultimately demonstrated that despite the consistent set of shared phenotypic features, a single genetic locus did not explain the full phenotypic spectrum of the condition.⁴ Multiple segregating pathogenic alleles at single^5,6 or multiple^7,8 loci in a family or clan can challenge both clinical and molecular diagnosis.

The occurrence of dual, or multiple, molecular diagnoses in an individual is not new, and reports dating back to the 1960s describe individuals with both hemolytic anemia due to glucose-6-phosphate dehydrogenase deficiency (OMIM #300908) and thalassemia minor, B-thalassemia, or sickle cell trait.^9,10 The true frequency of multiple molecular diagnoses has only begun to emerge with the availability of genome-wide investigations such as chromosome microarray analysis (CMA)¹¹ and WES; larger-scale analyses of diagnostic WES in a clinical setting have revealed that 3.2 to 7.2% of unselected diagnostic WES cases have two or more molecular diagnoses contributing to their observed clinical phenotype.^8,12–17 This rate can be even higher, 12% (5/41), in selected cohorts.¹⁸ From a clinical standpoint, recognition of multiple molecular diagnoses can have important implications for both treatment and estimates of familial recurrence risk.⁸

Genome-wide analyses have been instrumental for identifying molecular diagnoses in the setting of both atypical presentations of well-characterized syndromes,⁷ as well as cases for which the set of observed phenotypic features extended beyond those previously reported in association with a particular locus, termed ‘phenotypic expansion’.¹⁹ Examples of phenotypic expansion attributed to previously-known disease loci abound, with a striking 18.2% (174/956) of discoveries reported by the NIH-funded Centers for Mendelian Genomics (CMG) representing apparent phenotypic expansion.¹⁹ Given the relative frequency of both phenotypic expansion and multiple molecular diagnoses, we explored the hypothesis that cases of apparent phenotypic expansion may sometimes represent a blended phenotype resulting from variation at more than one locus. We tested this hypothesis using extensive phenotyping and personal genome data (WES) from 106 unrelated subjects from a well-characterized neurodevelopmental cohort, including 19 families with apparent phenotypic expansion that had previously been attributed to variation at a known disease gene.²⁰ A high prevalence of identity-by-descent enabled a focus on phenotypic and genotypic variability among siblings with shared genetic backgrounds.²⁰

MATERIALS AND METHODS

Patients

Patients provided informed consent, including consent to publish photographs, under the Baylor Hopkins Center for Mendelian Genomics (BHCMG) protocol H-29697 approved by the Baylor College of Medicine (BCM) Institutional Review Board.

RESULTS

We investigated the hypothesis that an observed clinical phenotype, initially interpreted as phenotypic expansion associated with pathogenic variation at one known disease gene locus, may actually represent a blended phenotype due to multilocus variation. We compared the frequency of multilocus variation between cases with and without apparent phenotypic expansion, using a pre-existing, well-phenotyped cohort for which WES had previously demonstrated a high rate of molecular diagnoses (84.4%, 108/128 families).²⁰ Of 108 families for whom a molecular diagnosis was identified, 19 families were classified as having phenotypic expansion due to the presence of observed clinical features that extended beyond those previously reported in association with variation at the identified locus (Table 1).²⁰ In the present pilot study, we performed a reanalysis of exome variant data from 106 of these 108 families - two families were removed from analysis as the exome variant data were previously generated at another institution using a different platform and capture.

Table 1.

Phenotypic features explained by multiple molecular diagnoses involving 26 known disease genes or loci and 2 candidate disease genes in 13 families. Phenotypes that are italicized indicate expanded phenotypes described in association with a particular molecular diagnosis: bold/red font indicates a phenotype due to a second molecular diagnosis; underscored/blue font indicates a phenotype due to a third molecular diagnosis. * De novo **Phenotype attributed to more than one molecular diagnosis. ***Severe phenotype proposed to result from burden due to combination of second and third molecular diagnoses. CC, corpus callosum; DD/ID, developmental delay/intellectual disability; PE, phenotypic expansion.

Group	Pedigree #	BAB#	1st Dx	MIM#	Patient Phenotypes	2nd Dx	MIM #	3rd Dx	MIM#
PE-KNOWN	HOU1341	BAB3405	GRIP1	617667	DD/ID, autistic features, agenesis of CC, subependymal heterotopia, hypoplastic vermis, dilated 3rd and 4th ventricles, hydrocephalus, delayed bone age, joint laxity	COL3A1	130050
	HOU1838	BAB4474	AP4B1	614066	DD/ID, microcephaly, agenesis of CC, Hypoplasia of bulbus and pons, blindness, protruduing tongue , atrial septal defect, cholelithiasis, hirsutism	AMPD2	615809
	HOU1857	BAB4635 BAB4636	C12ORF65	613559	ID, agenesis of CC, optic atrophy, sensorimotor neuropathy, distal contractures, cardiac valve abnormality	GPR126	616503
	HOU1965	BAB4918 BAB4919	PTCHD1	611603	DD/ID**, frontal lissencephalic cortical dysplasia, seizures, autism spectrum disorder	TUBA1A*	611603
	HOU2293	BAB3640	MCPH1	251200	DD/ID**, microcephaly, schizencephaly***	ALG8	608104	CLN5	256731
		BAB3636			DD/ID, microcephaly
	HOU2337	BAB6228	BCOR	300166	ID**, cortical atrophy **, agenesis of CC, microphthalmia, nystagmus, atrial septal defect, no walking	MED17	613668	SPG7	607259
		BAB6229	BCOR		ID, cortical atrophy, partial agenesis of CC, microphthalmia, ASD
PE-NOVEL	HOU2091 BAB5371	BAB5368	EPB41L1	614257	Frontotemporal dementia	SRR
	HOU2128	BAB5492	ITGA7	613204	DD/ID, microcephaly, agenesis of CC, cerebellar hypoplasia, seizures, hypotonia, chorioretinal lacunae, asymetric extremities, hypopigm macules, horseshoe kidney, scoliosis, hemivertebrae	MAGI3
NON PE-KNOWN	HOU2092	BAB5373	INA	n/a	ID**, microcephaly, seizures*, cerebral palsy**	GNAO1*	615473
	HOU2442	BAB6530	ATRX	309580	ID**, microcephaly**, epilepsy	CTNNB1	615075
PREV REPORT	HOU2215	BAB5804 BAB5805	RARS2 SNX14	611523 616354	ID, hypoplasia of CC, atrophy of bilateral cerebellum, hypoplastic vermis, seizures
	HOU2231	BAB5866 BAB5867	DHCR7 DMD ex	270400 310200	Smith-Lemli-Opitz, elevated creaktine kinase
	HOU2124	BAB5481	15q11.2 del ASXL3	615616 615485	DD/ID, seizures, microcephaly, diffuse cortical atrophy, gastroesophageal reflux

^*De novo

^**Phenotype attributed to more than one molecular diagnosis.

^***Severe phenotype proposed to result from burden due to combination of second and third molecular diagnoses.

Using an objective, very stringent set of Variant Call File filtering criteria to parse the personal genome sequencing data, followed by manual review, we identified a set of rare variants predicted to be damaging to protein function in 51 loci in 37 families for which previously reported disease-gene associations indicated a phenotypic overlap with the proband’s observed clinical features. From this set of variants, candidate etiologic variants were selected if they followed expected segregation patterns in affected and unaffected available relatives and were confirmed through independent orthogonal dideoxy Sanger sequencing. This final analysis step identified variants in 10 loci in 8 of 106 (7.5%) families (Table 1, Table S1). Notably, the originally reported variants/loci were parsed and not identified in 60.2% (65/108) of families using these strict filtering criteria which increase the specificity of personal genome analysis but decrease the sensitivity, highlighting the limitations of utilizing a purely computational approach to vcf data filtration prior to manual variant review of individual genomes within a large dataset. Variants at 8 loci were identified in 6 families (6/19, 31.6%) classified as having phenotypic expansion, and variants at 2 loci were identified in 2 families (2/87, 2.3%) without phenotypic expansion, demonstrating that apparent phenotypic expansion is more frequently associated with multilocus variation (two-tailed Fisher’s exact p=0.0004). Notably, this analysis demonstrated a second variant in HOU1960 (BAB4892, C2CD3) that was interpreted to explain the full phenotype of the proband; because the phenotypic contribution of the originally identified IQSEC2 variant was less clear, this family was not considered to have multilocus variation. Candidate novel disease genes were identified in 2 additional families previously classified as having phenotypic expansion, raising the potential rate of multilocus variation in cases of apparent phenotypic expansion to 42.1% (8/19); these candidate genes are discussed in further detail below. Taken together with the 3 cases of multilocus variation described in the original study (listed in Table 1)²⁰, the overall rate of multilocus variation for this neurodevelopmental cohort is 12.0% (13/108).

In the phenotypic expansion cohort, variants at a second or third locus involving known disease genes were observed in 6 families (31.6%, 6/19), and included COL3A1 (HOU1341), AMPD2 (HOU1838), GPR126 (HOU1857), TUBA1A (HOU1965), ALG8 and CLN5 (HOU2293), and MED17 and SPG7 (HOU2337, Table 1). In each case, variation at the second and sometimes third locus provided a molecular etiology for the expanded objectively determined phenotype (Table 1). In the non-phenotypic expansion cohort, variants at a second locus related to the observed clinical phenotype were identified in 2 families: GNAO1 (HOU2092) and ATRX (HOU2442). Unlike the phenotypic expansion cohort, the initially reported (Table 1) variants identified in the non-phenotypic expansion group were interpreted as causative of all phenotypic features in each case. Objective re-analysis of variant data demonstrated that each of these cases had an additional rare variant in a second gene associated with a subset of the observed clinical features. Notably, one case (HOU2268, BAB5990) with macrocephaly with extended subarachnoid spaces, a thin corpus callosum, and hypotonia was previously reported as having biallelic variation in ZNF423.²⁰ In this re-analysis, a de novo AHDC1 frameshifting variant was identified and because of the comprehensive phenotypic match with AHDC1 pathogenic variation,^21,22 the proband’s reported clinical diagnosis of Joubert syndrome (OMIM 614844) was considered less likely. This case was not classified as having multilocus variation in the present study.

Intrafamilial variability allows a dissection of genotype-phenotype correlation

Two families initially classified as having phenotypic expansion in particular demonstrated phenotypic variability between affected siblings. In pedigree HOU2293, the proband (BAB3640) was reported to have intellectual disability, neurodevelopmental delay, microcephaly, and schizencephaly; this individual’s sister (BAB3636), however, had intellectual disability, neurodevelopmental delay, and microcephaly without schizencephaly (Figure 1). Initial WES analysis demonstrated a molecular diagnosis of primary microcephaly (OMIM 251200) due to biallelic pathogenic, frameshifting variants in MCPH1 in both siblings.²⁰ Reanalysis of WES data in the proband and sibling demonstrated additional molecular diagnoses of congenital disorder of glycosylation type Ih (OMIM 608104) and neuronal ceroid lipofuscinosis 5 (OMIM 256731) due to rare biallelic variants in ALG8 and CLN5 (Figure 1) present only in the more affected proband; notably, none of the three unaffected siblings were homozygous for these variants. Pathogenic variation at both of these loci has been associated with structural brain anomalies including brain atrophy,^23,24 and we propose that the additional mutational burden conferred by rare variation at ALG8 and CLN5 led to the schizencephaly phenotype identified in the more severely affected proband.

FIGURE 1. Intrafamilial genotypic and phenotypic variability in HOU2293.

Pedigree structure of family HOU2293. Both affected siblings share a homozygous variant in MCPH1; however, the proband (BAB3640) has additional homozygous variants in ALG8 and CLN5, resulting in a more severe blended phenotype.

In pedigree HOU2337, the proband (BAB6228) was reported to have cortical atrophy, agenesis of the corpus callosum, microphthalmia, and severe neurodevelopmental delay characterized by the inability to walk, even with support. This individual’s brother (BAB6229) had a milder phenotype with cortical atrophy, partial agenesis of the corpus callosum, and microphthalmia; despite neurodevelopmental delay, he was eventually able to walk independently. A maternally inherited hemizygous variant in BCOR consistent with a molecular diagnosis of syndromic microphthalmia 2 (OMIM 300166) was identified in both siblings (Figure 2).²⁰ Re-analysis of WES data further demonstrated biallelic variants in MED17 and SPG7, consistent with molecular diagnoses of postnatal progressive microcephaly with seizures and brain atrophy (OMIM 613668) and autosomal recessive spastic paraplegia (OMIM 607259), in the more severely affected proband only (Figure 2). Although the proband does not demonstrate the full set of features for each of these diagnoses, both loci were hypothesized to provide genotypic correlation in the form of additional mutational burden to the more severe neurodevelopmental phenotype observed in the proband. Siblings pairs demonstrating phenotypic variability allow a true functional analysis of shared and private variation at genomic loci associated with Mendelian disease and support the contention that all affected individuals in a family or clan should be thoroughly investigated from a clinical phenotyping perspective.

FIGURE 2. Intrafamilial genotypic and phenotypic variability in HOU2337.

Pedigree structure of family HOU2337. Both affected siblings share a hemizygous variant in BCOR; the proband (BAB6228) has additional homozygous variants in MED17 and SPG7, resulting in a more severe phenotype due to the additional mutational burden. Cort, cortical; agCC, agenesis of the corpus callosum.

Multilocus variation involving candidate novel disease genes

Variants at a second locus suggesting a potentially novel disease gene were observed in 2 families (10.5%, 2/19, Table 1). The proband in HOU2091 has frontotemporal dementia. Initial WES analysis revealed a heterozygous variant in EPB41L1 associated with Mental Retardation AD 11 (OMIM 614257). On reanalysis, a rare heterozygous variant at a second locus, SRR encoding serine racemase, was identified in the affected proband (BAB5368), as well as the affected mother (BAB5369) and affected maternal uncle (BAB5371). SRR is expressed in mammalian brain, where it converts L-serine to D-serine, a relevant compound for glutamatergic neurotransmission. Studies of Srr knockout mice have demonstrated both a reduction of D-serine in the brain and a reduction of NMDA-induced neurotoxicity, the latter of which has been proposed to be a potential mechanism for the development of the dementia observed in Alzheimer disease.²⁵ Based on the known function of serine racemase and its expression pattern, we propose that the identified SRR variant may function in a gain of function manner, resulting in a predisposition to frontotemporal dementia.

The proband of family HOU2128 has microcephaly, agenesis of the corpus callosum, cerebellar hypoplasia, seizures, horseshoe kidney, scoliosis, hemivertebrae, asymmetric extremities, and hypopigmented skin macules. This individual was initially found to have a rare and potentially deleterious variant in ITGA7, conferring a molecular diagnosis of congenital muscular dystrophy (OMIM 613204, Table 1). Reanalysis of the WES data demonstrated variation in MAGI3, encoding membrane-associated guanylate kinase with inverted structure 3, and not previously associated with disease. The membrane associated guanylate kinase proteins, MAGI1, MAGI2, and MAGI3, are expressed in the mammalian brain.²⁶ MAGI2 in particular has been described as causative of mental retardation, autosomal dominant 10 (OMIM 614256) in humans.²⁷ Disruption of MAGI2 in humans can lead to a seizure phenotype, as atypical 7q11.23 deletions of the Williams-Beuren syndrome (OMIM 194050) region that extend more distally to include MAGI2, are associated with seizures and infantile spasms.²⁸ MAGI3 enhances the ability of PTEN to regulate AKT kinase activity, particularly under conditions of low PTEN expression, ²⁹ and also plays a role in efficient trafficking of TGF-alpha to the cell surface in polarized epithelial cells.²⁶ Based on the expression pattern and function of MAGI3, and the known disease association of MAGI2, we propose that MAGI3 may be a novel candidate gene associated with structural and functional neurologic abnormalities and also potentially with asymmetric extremities.

Absence of Heterozygosity (AOH)-Mediated Mutation Burden

One of the advantages of the cohort chosen for this analysis is the high rate of reported consanguinity, enabling pathogenic variants to be reduced to homozygosity within a single generation,³⁰ and facilitating study of recessive disease, particularly in sibling pairs sharing a similar genetic background. Considering the 8 cases with a second and sometimes third molecular diagnosis identified in a known disease gene, there were a total of 10 variants in four families found within regions of absence of heterozygosity, determined based on the calculated B-allele frequency from exome variant data using BafCalculator (https://github.com/BCM-Lupskilab/BafCalculator) (Table S2).³¹ The calculated total AOH for probands of each of the 19 cases with phenotypic expansion ranged from 62.0 Mb to 542.2 Mb, and this was not significantly different from that of the studied brain malformation cohort (128 original cases), which ranged from 28.3 Mb to 866.7 Mb. Although the relatively small number of cases with phenotypic expansion and/or multilocus variation was not enriched for AOH, the studied brain malformation cohort did demonstrate an enrichment for AOH compared to all other Turkish individuals in the CMG, and the Turkish cases within the CMG (~1,000 exomes) were similarly enriched for AOH compared to non-Turkish individuals in our local CMG exome variant database (~6,000 exomes) (Figures S1, S2).

The utility of AOH for the study of multilocus variation is perhaps best illustrated by family HOU1838 for which there are biallelic variants at 2 distinct loci, each with evidence that they contribute to the phenotype, and each within the same 44.8 Mb region of AOH on chromosome 1 (Figure 3). BAB4474, the proband of HOU1838, was initially described as having a novel pathogenic variant in the gene AP4B1, providing a molecular diagnosis of spastic paraplegia 47, autosomal recessive (OMIM 614066). This diagnosis was consistent with the observed phenotypes of DD/ID, microcephaly, and agenesis of the corpus callosum, as well as the segregation pattern observed in the pedigree. Extensive re-analysis of the WES data demonstrated additional pathogenic variation at AMPD2, providing etiologic molecular diagnoses potentially explaining the pontocerebellar hypoplasia, thin retina, and protruding tongue (Pontocerebellar hypoplasia 9, OMIM 615809 associated with pathogenic variants in AMPD2, Figure 3). Notably, a homozygous splice site variant in NOTCH2 (c.4512–3T>G) could be consistent with the features of cholelithiasis, thin retina, atrial septal defect and hirsutism (Alagille syndrome, OMIM 610205 due to variants in NOTCH2), however the zygosity of the identified variant is not consistent with the autosomal dominant inheritance observed with Alagille syndrome, and we propose that the potential for this locus to be involved in both autosomal recessive and dominant modes of disease trait inheritance requires further exploration.^32,33

FIGURE 3. Absence of heterozygosity mediates mutational burden at a single locus.

(A) Pedigree structure of family HOU1838 and segregation of the variants identified in AMPD2, AP4B1 and NOTCH2. (B) Images of proband (BAB4474) demonstrate narrow forehead, esotropia of the left eye, short nose and protruding tongue. (C) B-allele frequency calculated from exome variant data demonstrates several regions of AOH, marked by grey zones. Variants in AMPD2, AP4B1, and NOTCH2 are located within the same region of AOH. (D) Blended phenotypic features and associated syndromes are presented. ASD, atrial septal defect.

In family HOU1857, siblings BAB4635 and BAB4636 were found to have potentially etiologic variants at two unlinked loci within regions of AOH (Figure 4). Both siblings presented with agenesis of the corpus callosum, optic atrophy, sensorimotor neuropathy, distal contractures, and a cardiac valve anomaly. After the initial identification of variants in C12ORF65 (Combined oxidative phosphorylation deficiency 7, OMIM 613559), homozygous variants in GPR126 (Lethal congenital contracture syndrome, OMIM 616503) were identified as causative of the expanded phenotype (Figure 4A, B). Both the C12ORF65 and GPR126 variants are located in distinct but overlapping regions of AOH on chromosome 12 and 6: C12ORF65 within a region of 11.3 Mb (BAB4635) and 11.0 Mb (BAB4636); GPR126 within a region of 9.5 Mb (BAB4635) and 12.1 Mb (BAB4636, Figure 4C).

FIGURE 4. Absence of heterozygosity mediates mutational burden at more than one locus.

(A) Pedigree structure of family HOU1857 and segregation of variants identified in C12orf65 and GPR126. (B) Both siblings had esotropia, progressive sensorimotor polyneuropathy with impaired walking and resultant contractures and deformities in their extremities. (C) B-allele frequency calculated from exome variant data demonstrates several regions of AOH within chromosome 6, marked by grey zones. Variants in GPR126 (red line) are located in a region of AOH. (D) B-allele frequency calculated from exome variant data demonstrates several regions of AOH within chromosome 12, marked by grey zones. Variants in C12ORF65 (red line) are located in a region of AOH.

DISCUSSION

The availability of WES has enabled a genomic interrogation of phenotypes in a hypothesis-free manner, leading to increasing discoveries of both phenotypic expansion associated with a single genomic locus, and blended phenotypes resulting from multilocus variation. To investigate whether cases of apparent phenotypic expansion may have an increased frequency of multilocus variation, we focused on a cohort of well-characterized neurodevelopmental phenotypes , a subset of which were previously categorized as having phenotypic expansion, occurring in the setting of a high rate of reported consanguinity.²⁰ This strategy enabled a focus on predominantly recessive disease and greater likelihood of shared phenotypes amongst siblings with similar genetic backgrounds resulting from identity by descent.³⁰ We found that 31.6% (6/19) of cases characterized as demonstrating ‘phenotypic expansion’ actually represent blended phenotypes⁸ with more than one molecular diagnosis, compared to 2.3% (2/87) of cases not characterized as demonstrating phenotypic expansion (two-tailed Fisher’s exact p=0.0004).

Such clinical cases are difficult to recognize as potential dual molecular diagnoses, because blended phenotypes may present as phenotypic expansion, more severe disease along a spectrum, or an apparently novel disease.⁸ Our findings suggest that cases of apparent phenotypic expansion are more likely to present with a blending of distinct disease phenotypes due to multiple molecular diagnoses. We further demonstrate that cases that do not have apparent phenotypic expression can also result from multilocus variation, and may present as a manifestation of more severe disease. Considering this neurodevelopmental cohort in its entirety, which includes 3 cases with multilocus variation reported in the initial study²⁰, a total of 108 cases have one or more molecular diagnoses, of which 13 cases (12.0%, 13/108) have two or more molecular diagnoses resulting from known or potential novel disease genes, underscoring the frequency with which multiple molecular diagnoses can occur. Notably, the overall rate of multilocus variation in this selected cohort (12.0%) is greater than the 4.9% reported in a diagnostic cohort of mixed phenotypes from an outbred population⁸; this observation may reflect the genomics of the Turkish population, the high rate of consanguinity, and/or the studied phenotype. As the field moves toward study of oligogenic disease and multi-locus pathogenic variation, we will rely heavily on clinical expertise to untangle the relationships between phenotypic expression observed in the clinic and molecular diagnoses identified by genome-wide assays.

Intrafamilial variability enables a dissection of genotype-phenotype correlation

The interplay among genomic variants at more than one locus and the resulting impact on phenotypic expression can be difficult to predict, and a genotype-forward approach to analysis of phenotypic expression is limited by the rarity of each condition, and combination of conditions, studied. Here, we report 2 families with intrafamilial clinical variability that was ultimately explained by genotypic variability. In family HOU2337, we identified extremely rare and potentially deleterious homozygous variants in both MED17 and SPG7 genes in patient BAB6228, in addition to a hemizygous mutation in the BCOR gene that he shares with his less severely affected brother (Figure 2). Similarly, in family HOU2293, we identified variants in ALG8 and CLN5 in BAB3640, who additionally shared a homozygous MCPH1 variant with her less severely affected sibling (Figure 1). In both families, the expanded phenotype observed in the probands was in fact explained by variation at additional loci not present in their less severely affected siblings. These cases demonstrate the importance of a thorough clinical history and examination of all affected family members. Genomic analysis in families with more than one affected individual allows a dissection of phenotypes and phenotypic severity attributed to variation at each identified locus within the context of a shared genetic background.

AOH-mediated recessive mutation burden as a tool

In both families described above, AOH contributed to a recessive mutation burden, with total AOH, as determined from the calculated B-allele frequency from exome variant data,³¹ measuring 509.0 Mb in BAB6228 and 493.1 Mb in BAB3640 (Table S3). In 14 known or potential autosomal recessive loci identified in this cohort, 11 loci had variants that were homozygous and located within regions of AOH. This finding is particularly striking for patient BAB4474, for whom analysis revealed rare, homozygous variants in three genes within a single region of 44.8 Mb AOH: AP4B1 (OMIM 614066), AMPD2, and a potential contribution from NOTCH2 (OMIM 610205, Figure 3, Table 1). Although Alagille syndrome has previously been described as an autosomal dominant condition, BAB4474 is homozygous for the identified variant, which was inherited from apparently unaffected parents. Several examples of genes for which both mono- and biallelic variation lead to disease have been previously described,^32–35 and we propose that this NOTCH2 variant may represent a mild or hypomorphic allele for which biallelic variation is required for phenotypic expression.

We also describe a family (HOU1857) for which AOH-mediated recessive mutation burden involves two causative loci located within unlinked regions of AOH (Figure 4). In this case, the C12ORF65: c.248delT: p.V83fs (Spastic Paraplegia 55, AR OMIM 615035) variant explained the majority of the clinical features including intellectual disability, optic atrophy, agenesis of the corpus callosum, sensorimotor neuropathy and contractures (Figure 4, Table 1). A second locus, GPR126, has been shown to be associated with Schwann cell and peripheral nerve development and myelination. We propose contribution of the homozygous GPR126 variant (chr12:123738468_GT>G; c.1880A>G; p.K627R) at this locus to the neuropathy and contractures observed in this family (Figure 4C).^36,37 In addition, the role of Gpr126 in cardiovascular development has been identified in several animal model studies and may be consistent with the cardiac finding of valvular anomaly in this family.³⁸ While total AOH was not significantly greater in cases with multilocus variation and/or phenotypic expansion compared to the remainder of the neurodevelopmental cohort (Figures S1, S2), the cohort demonstrated enrichment for AOH compared to other Turkish individuals and non-Turkish individuals within our local exome database. The burden of homozygous variants identified in regions of AOH in this cohort demonstrate that AOH-driven recessive disease is one genetic mechanism leading to multiple molecular diagnoses and mutational burden in a patient.

Candidate novel genes

We propose two novel candidate disease genes, each with the potential to impact our understanding of the pathophysiology of neurologic disease and perhaps delineate a precision medicine-guided pathway to potential therapeutic interventional strategies. In HOU2019, a rare variant in SRR was identified as contributing to the frontotemporal dementia phenotype in the proband (BAB5368), affected mother (BAB5369), and maternal uncle (BAB5371). SRR is a serine racemase that impacts levels of L- and D-serine in the brain, and is thought to be important for glutamatergic neurotransmission, which has been proposed to be involved in the development of complex neurodegenerative disease, in particular Alzheimer disease.²⁵ We propose that SRR missense variation results in a gain-of-function allele affecting glutamatergic neurotransmission and resulting in excitatory neurotoxicity and dementia. Additionally, the role of SRR in this phenotype suggests that therapy aimed at modification of glutamatergic neurotransmission may be of potential clinical benefit. In the proband of HOU2128, a rare variant in MAGI3 was identified and may provide both a molecular explanation for the Aicardi phenotype as well as the observed hemihypertrophy. Hemihypertrophy and overgrowth syndromes can be associated with variation in PTEN, and MAGI3 has been demonstrated to impact PTEN regulation of AKT activity.²⁹ These examples demonstrate that re-analysis of cases characterized as phenotypic expansion can lead to potential novel candidate disease gene discovery and potentially inform the molecular pathophysiology of disease.

Conclusion

In conclusion, each genome is truly “personal” and phenotypic expression, what the clinician observes, can be impacted by variation at more than one locus, often resulting in more than one molecular diagnosis.⁸ In some cases, phenotypic expansion may be a hallmark of a blended phenotype resulting from multilocus variation. This variation can be driven by the genetic substructure of the family, clan or population (i.e. isolated, consanguinity), leading to AOH-mediated recessive disease burden. Implementation of WES has enabled genomic medicine to move from individual study of single gene Mendelian disorders to a more comprehensive understanding of the interplay of genetic variation at more than one locus and resulting phenotypic expression. These findings further emphasize the important role of the clinician and robust physical examination and clinical characterization of each patient in order to recognize blended phenotypes due to multilocus variation. Such blended phenotypes may present clinically as distinct phenotypes manifesting as apparent phenotypic expansion associated with the first identified locus, or overlapping phenotypes that may manifest as an apparent worsening of disease severity.