Neuropsychiatric phenotype in relation to gene variants in the hemizygous allele in 3q29 deletion carriers: A case series

Abstract Background Genetic risk variants in the hemizygous allele may influence neuropsychiatric manifestations and clinical course in 3q29 deletion carriers. Methods In‐depth phenotypic assessment in two deletion carriers included medical records, medical, genetic, psychiatric and neuropsychological evaluations, brain MRI scan and EEG. Blood samples were analyzed for copy number variations, and deep sequencing of the affected 3q29 region was performed in patients and seven first‐degree relatives. Risk variants were identified through bioinformatic analysis. Results One deletion carrier was diagnosed with learning difficulties and childhood autism, the other with mild intellectual disability and schizophrenia. EEG abnormalities in childhood normalized in adulthood in both. Cognitive abilities improved during adolescence in one deletion carrier. Both had microcytic, hypochromic erythrocytes and suffered from chronic pain and fatigue. Molecular and bioinformatic analyses identified risk variants in the hemizygous allele that were not present in the homozygous state in relatives in genes involved in cilia function and insulin action in the autistic individual and in synaptic function and neurosteroid transport in the subject with schizophrenia. Conclusion 3q29 deletion carriers may undergo developmental phenotypic transition and need regular medical follow‐up. Identified risk variants in the remaining hemizygous allele should be explored further in autism and schizophrenia research.

including bipolar disorder, schizophrenia, autism spectrum disorder (ASD), intellectual disability (ID) or developmental delay (DD), attention deficit hyperactivity disorder (ADHD), and epilepsy (Grayton, Fernandes, Rujescu, & Collier, 2012;Kirov, 2015;Torres, Barbosa, & Maciel, 2016). In deletion carriers, haploinsufficiency of critical dose-dependent genes may account for some of the syndromic features associated with the specific CNV, but this alone cannot explain the neuropsychiatric pleiotropy (Chenier et al., 2014). It is now of great importance to understand the contributions of the individual genes and gene variants within a CNV region to allow us to understand their role in deletions and reciprocal duplications.
Only longitudinal follow-up of the patients can help to elucidate the developmental trajectory of a certain CNV. We need to consider that the neuropsychiatric phenotype might undergo phenotypic transition during development, as has been documented for patients with the 22q11.2 deletion syndrome (Swillen & McDonald-McGinn, 2015). Regarding 3q29 deletions, a systematic neuropsychiatric evaluation with special emphasis on cognitive and behavioral assessment has only been published in one case report of four children by Citta et al. (2013). In adults, only self-report data from five patients are included in the 3q29 Deletion Registry (Glassford et al., 2016), and little is known about developmental phenotypic transition and the adult phenotype. Therefore, systematic clinical in-depth characterization of adult carriers of the 3q29 deletion is needed. The estimated prevalence of the deletion is 1:30,000 in the general population, which makes it necessary to look critically for feasible approaches to study phenotypic pleiotropy in this rare disorder.
Family studies are fundamental tools in the discipline of behavioral genetics, as they permit assessments of degrees of familial resemblance, or aggregation of physical, psychological, and behavioral characteristics. Adoption studies and comparisons between monozygotic and dizygotic twins are the gold standard of these studies. In studies of rare behavioral and pleiotropic phenotypes, a supplementary approach could be to study genetic variants and phenotypes in a context that minimizes as much as possible the impact of general genetic background and environmental influences. This minimization can be achieved by studying CNV carriers together with their first-degree relatives. Such an approach could allow the generation of well-founded hypotheses about genotype-phenotype associations, which could later be tested in larger patient samples.
In this study, we want to explore the hypothesis that genetic variation in the remaining hemizygous allele contributes to phenotype variability in 3q29 deletion carriers. We have therefore performed in-depth phenotyping of two patients with 3q29 deletion syndrome and targeted deep sequencing of the affected 3q29 region in patients and first-degree relatives.

| Ethical compliance
The study adhered to the tenets of the Declaration of Helsinki. The Regional Ethical Committee of South-Eastern Norway approved the study, and the participants or their legal representatives provided written and oral informed consent. All participants were informed that they could withdraw from the | 3 of 12 MALT eT AL. study at any time without any negative consequences. The participants and their legal guardian gave their consent for publication.

| Patients
Two consecutive families with one first-degree family member diagnosed with 3q29 deletions were included in the study. Family 1 was of European ethnicity, and patient 1 was a 23year-old male. Family 2 was of South Central Asian ethnicity, and patient 2 was a 29-year-old female. Informed consent was obtained from all participants.

| Clinical and phenotypic assessments
Clinical information was collected from patients and firstdegree relatives by experienced staff, including a psychiatrist, a neuropsychologist, and clinical geneticists. For patients, all available lifetime medical records were collected, and a thorough developmental and general anamnestic history was taken. At inclusion, routine physical and neurological examinations were performed. Dysmorphic features and congenital anomalies were evaluated. ICD-10 diagnoses were assessed through a semi-structured neuropsychiatric interview, The M.I.N.I. Plus International Neuropsychiatric Interview (Norwegian Translation Version 6.0.0) (Leiknes, Malt, Malt, & Leganger, 2006), and a clinical interview with a parent. ASD was assessed at age 5 years 8 months in a psychiatric unit specialized in assessment of neurodevelopmental disorders in children. Assessments included the Autism Diagnostic Interview-Revised (ADI-R) (Lord, Rutter, & Couteur, 1994), the Autism Diagnostic Observation Schedule (ADOS), module 2 and 3 (Lord et al., 1989) and the ADI-R and Vineland Adaptive Behavior Scales (Sparrow, Balla, & Cicchetti, 1984). The ADOS is a diagnostic, semi-structured clinical assessment that directly observes behaviors associated with ASD in the areas of social communication, play and interaction, and restrictive and repetitive behaviors The ADI-R is a semi-structured comprehensive parent/caregiver interview designed to evaluate early developmental history and current and lifetime presentation of autism symptomatology. At inclusion verbal and nonverbal cognitive abilities were assessed using the Wechsler Adult Intelligence Scale-4th edition (WAIS-IV), the Norwegian edition with Scandinavian norms (Wechsler, Coalson, & Raiford, 2008), and Raven Matrices Practice Tests (Raven, 2000). Adaptive behavior was assessed by the semi-structured interview Vineland Adaptive Behavior Scales, 2nd edition (VABS-II) (Sparrow, Cicchetti, & Balla, 2005). A brain structural magnetic resonance imaging (MRI) scan with epilepsy protocol was performed in patient 1. Data were collected on a 1.5T magnet scanner. MRI scan had been performed previously and was not repeated in patient 2. Standard EEG was performed in both patients.
Peripheral blood samples were drawn for biochemical and pharmacogenetic analyses, and a neurometabolic workup was performed after the collection of overnight fasting urine and peripheral blood samples.
First-degree relatives underwent a brief clinical interview about psychiatric and somatic symptoms and disorders.

| Genetic counselling
Genetic testing and counselling of patients and first-degree relatives were conducted according to Act 2003-12-05-100 relating to human medical use of biotechnology.

| Genomic DNA extraction and molecular analyses
In patients, peripheral venous blood was analyzed for CGG repeats using the FMR1 with extra-long and repeated prime polymerase chain reaction. To assess genome-wide DNA dosage imbalance, array comparative genome hybridization was performed using Agilent 180K Sureprint G3 Human CGH array in the two patients. Subsequently, the presence of the 3q29 deletion was tested in relatives by MRC Holland SALSA MLPA probemix P245 (B1-0512). Genomic DNA from patients and first-degree relatives was extracted from peripheral blood using the DNeasy Blood & Tissue Kit (Qiagen). DNA was sheared using a focused-ultrasonicator (Covaris) and the fragment size of the sheared DNA was assessed by TapeStation (Agilent Technologies). An Agilent SureSelect Target Enrichment System was used to capture the genomic 3q29 region, including intergenic and intragenic regions, excluding repetitive sequences, prior to sample sequencing. The 3q29 region was sequenced by MiSeq, Illumina.

| Bioinformatic analyses and variant classification
After base calling and removal of adaptors, raw reads were mapped to the reference genome GRCh37. Variant calling was performed using the GATK haplotype caller (v.3.7) to identify single-nucleotide polymorphisms (SNPs) and indels (McKenna et al., 2010). Variants were filtered using the GATK recommendations. In brief, SNPs were filtered according to variant confidence (QualByDepth, QD < 2.0). Fisher's exact test (FisherStrand, FS > 60.0) and StrandOddsRatio (SOR > 3.0) were used to detect strand bias. Mapping quality across samples was assessed using RMSMappingQuality (MQ < 40.0). MappingQualityRankSumTest (MQRankSum < −12.5) was used to assess mapping qualities (reads with reference bases vs. those with the alternate allele). ReadPosRankSumTest (ReadPosRankSum <−8.0), a ubased z-approximation from the Mann-Whitney rank sum test, was used to assess the distance from the end of the read for reads with the alternate allele. If the alternate allele was only seen near the ends of reads, the variant was discarded. Similarly, indels were filtered using the following cutoffs: QualByDepth (QD < 2.0), FisherStrand (FS > 200.0), ReadPosRankSumTest (ReadPosRankSum < −20.0), and StrandOddsRatio (SOR > 10.0).
SnpEff v 4.3 was used to annotate variants and classify them according to a predicted impact-high, moderate, low, and modifier (Cingolani et al., 2012). A variant was assigned to have high impact in cases when it was predicted to cause protein truncation, loss of function, or trigger nonsense-mediated decay; moderate impact when it was predicted to be nondisruptive, but possibly affecting protein function; low impact when it was predicted to be harmless; and modifier for noncoding variants or variants affecting noncoding genes where there is no evidence of impact. The variant lists were manually inspected to identify variants that segregated with the phenotype.

| RESULTS
In patient 1, array-CGH analysis identified a 1.6-Mb deletion, Arr(hg19)3q29 x1. The deletion was paternal and de novo. In patient 2, array-CGH identified a 1.5-Mb deletion, Arr(hg19)3q29  x1. The mother and sisters of patient 2 did not carry the deletion, and there is no information about the father, who is deceased. In both patients, the number of CGG repeats in FMR1 was within the normal range, and no other potentially pathogenic CNVs were detected.

| Clinical phenotypic characteristics
Clinical characteristics of the two patients are presented in Table 1. Patient 1 was diagnosed with childhood autism (ICD-10 F84.0), while patient 2 had early-onset psychosis developing to paranoid schizophrenia (ICD-10 F22.0) and was only partially responsive to treatment with antipsychotic drugs. Neither of the patients showed regression in cognitive function, and cognitive testing revealed improvement during adolescence in patient 1 with autism and stable function in patient 2 with schizophrenia. Adaptive abilities improved in patient 1 and declined in patient 2. Neither patient had epilepsy, but both patients' EEGs were abnormal and indicated cerebral dysfunction during childhood. Both patients' EEGs were normalized during adulthood. Both patients had hematological abnormalities persisting over time, with microcytic, hypochromic erythrocytes, and low levels of transferrin receptor. Hemoglobin was within the normal range in both patients, but serum iron was low in patient 2. Both suffered from chronic pain, especially headache, and fatigue. Patient 2 suffered from hypothyreosis and received hormonal substitution. Thyroid-stimulating hormone was slightly elevated without elevated autoantibodies (thyroid peroxidase, thyroid receptor antibody). Other general laboratory blood tests were normal in both patients.
Evaluations of first-degree relatives of both patients indicated no psychiatric, hematological, endocrine, or other serious somatic disorders. The clinical interview, academic performance, and occupational performance indicated cognitive abilities and adaptive skills within the normal range in all first-degree relatives.

| Molecular and bioinformatic analyses of genetic risk variants
Identified high-and moderate hemizygous genetic risk variants in patients found only in the heterozygous state in first-degree relatives are shown in Table 2. These variants were not associated with disease in any of the family members. Patient 1 carried a splice donor variant in TCTEX1D2 predicted by the SnpEff algorithm to be a high-risk splice variant. However, other supporting laboratory data are required in order to assess whether this variant may have functional consequences.
Patient 1 also carried a missense mutation (rs1134986) annotated by the ExAC browser to be a missense mutation in  PCYT1A, but upon careful inspection of the validated gene predication in this region we found that this variant is not located in the PCYT1A gene or any other gene. Consistently, the transcript has not been experimentally verified. It cannot be excluded, however, that this variant may contributes to the phenotype of the patient through a regulatory function. Patient 2 carried a hemizygous a missense, structural interaction moderate-risk variant rs1134986 in DLG1. This missense mutation introduces and amino acid substitution (Arg278Gln) in a conserved region of the protein. It is classified as at moderate-risk variant as it substitutes a large and basic amino acid to a medium sized and polar residue (Carroll et al., 2011). Moreover, the variant is located in the PDZ1-domain which mediates interaction with a range of proteins and lipids. Modeling this substitution based on a crystal structure of the human DLG1 protein (PMID:21,858,148) (Zhang et al., 2011) supports that the mutation may impair interaction with Glu281 and thereby introduce a higher degree of surface flexibility. In addition, the substitution will promote local negative surface charge that could affect interaction with protein partners ( Figure  1).
Patient 2 also carried a moderate-risk missense variant in SLC51A, a C to T transition corresponding to the high-frequency polymorphism rs939885 and leading to substitution of valine for isoleucine at position 202. This variant has relatively high frequency in the normal population and the amino acids have similar chemical properties. Neither variant is reported as having known clinical significance.
Both patients shared a missense variant (rs6776064) in CEP19. This variant, classified as having moderate effect, leads to a Met3Val N-terminal amino acid change in a region with no known sequence motifs or protein interaction sites.
Some moderate-risk variants in TFRC, TM4SF19, and PIGZ found in the patients and RNF168 were present in the homozygous state in first-degree relatives. As expected, many variants classified as either as mild risk or as modifiers were identified in the patients. Similarly, intergenic variants in predicted regulatory regions that segregated with the disease phenotype were identified in both patients; however, the impact of these is currently unknown.

| DISCUSSION
This is the first report that combines neurodevelopmental, indepth phenotypic, and genetic data to describe longitudinal phenotypic development in adult carriers of the 3q29 deletion. Both patients showed previously described traits associated with the syndrome (Cox & Butler, 2015). However, we have also identified some new features in both patients that have not been described before. These include EEG abnormalities in childhood and enduring hematological abnormalities. It is also important and worrying that both patients suffered from chronic pain and fatigue.
In line with previous observations of pleiotropy, the probands had different neuropsychiatric phenotypes (Cox &   Only variants for which no first-degree relatives were homozygous are presented. A variant was assigned to have high impact in cases when it was predicted to cause protein truncation, loss of function, or trigger nonsense-mediated decay; it was assigned moderate impact when it was predicted to be nondisruptive, but possibly affecting protein function. a hg19, GRCh37. Butler, 2015). One was diagnosed with childhood autism, while the other had paranoid schizophrenia. Patient 1 was diagnosed with ID at 5 years of age, but his cognitive abilities were found to be within the normal range at 23 years of age. The impaired cognitive function in patient 2 was stable over time, although adaptive functions had declined considerably. These findings indicate that the 3q29 deletion might undergo phenotypic transition during development, as has been described previously in patients with other pathogenic CNVs. This underscores the importance of longitudinal follow-up of 3q29 deletion carriers. Our hypothesis was that genetic risk variants in the remaining hemizygous allele may contribute to diverse neuropsychiatric phenotypes. It is therefore interesting that the genetic analyses identified highly relevant risk variants not shared by the two patients.
Patient 1 carried a possible splice variant in TCTEX1D2, which codes for a ciliary protein. TCTEX1D2 is found in primary cilia and has a role in intraflagellar transport, a bidirectional microtubule-based transport system that operates between the ciliary tip and base (Youn & Han, 2018). Primary cilia play a role in key developmental signaling pathways, and mutations that affect the structure or function of primary cilia may result in ciliopathies. The pleiotropic features of distinct ciliopathies often overlap and may include brain malformation, neurologic impairment, retinal degeneration, skeletal anomalies, congenital heart diseases, hepatic fibrosis, and cystic-fibrotic kidney disease. Brain phenotypes frequently manifest, including encephalocele, holoprosencephaly, microcephaly, polymicrogyria, heterotopia, intracerebral cysts, hippocampal dysgenesis, corpus callosum agenesis, hydrocephalus, cerebellar hypoplasia, cognitive deficits such as ID/DD, and ASD (Youn & Han, 2018). Regarding neuropsychiatric disorders, it has actually been suggested that primary cilia represent a conserved cellular structure at which the effects of diverse neuropsychiatric risk genes converge (Marley & von Zastrow, 2012).
Biallelic loss of function mutations in TCTEX1D2 cause the ciliopathy Short rib thoracic dysplasia, which is characterized by a constricted thoracic cage, short ribs, and shortened tubular bones (Gholkar et al., 2015;Mukhopadhyay, 2015;Schmidts et al., 2015). Nonskeletal involvement can include cleft lip/palate, as well as anomalies of major organs. Our patient did not display most of these features. However, he had restricted thoracic cage leading to restrictive respiratory disease.
TCTEX1D2 has broad expression in the brain. In mice, Tctex1d2 has been found to have a physiological function related to insulin-stimulated glucose uptake through the regulation of Glut4 receptor translocation (Shimoda, Okada, Yamada, Pessin, & Yamada, 2015). Glut4 receptor translocation is implicated in hippocampus-dependent memory, as well as in the integration of hormonal and nutritional cues to mediate the actions of insulin on energy balance in the CNS (Pearson-Leary & McNay, 2016;Ren, Lu, McGraw, & Accili, 2015). It is therefore possible that TCTEX1D2 may influence cognitive function, as well as the nutritional problems with "failure to thrive" seen in our patient and others with 3q29 deletion (Glassford et al., 2016). Altogether, supporting evidence suggest that impaired TCTEX1D2 expression and function are consistent with some features seen in our patient, but further experimental work too confirm that the variant affects splicing is warranted before we can suggest that TCTEX1D2 variants may have a potential role in ASD.
Patient 2 carried a hemizygous predicted high-risk variant in DLG1 (rs1134986). The gene has not been connected to a Mendelian disease, as of today. DLG1 belongs to a family of scaffolding proteins, the membrane-associated guanylate kinases (MAGUKs), that are highly enriched in the postsynaptic density of synapses and play an important role in organizing protein complexes necessary for synaptic development and plasticity (Gupta, Uner, Nayak, Grant, & Kalb, 2018). DLG1 regulates synaptic delivery of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors in hippocampal neurons (Nakagawa et al., 2004). It has been suggested that DLG1 may be important for the pruning process of neuron branches that occurs during adolescence and that alterations in this early brain remodeling mechanism may be the basis for the altered neural connectivity and functioning that can result in schizophrenia later in life (Cocchi, Drago, & Serretti, 2016).
DLG1 has previously been associated with schizophrenia (Sato, Shimazu, Yamamoto, & Nishikawa, 2008;Toyooka et al., 2002;Uezato et al., 2012Uezato et al., , 2015Uezato et al., , 2017. The variant found in our patient was identified in a study of 234 unrelated cases with schizophrenia and 272 unrelated controls from the United Kingdom (Carroll et al., 2011). The allele frequency was >1% and did not differ between patients and controls (Carroll et al., 2011). This finding, as well as the presence of the allele in our patient´s healthy mother, indicates that the variant is not associated with schizophrenia when present in the heterozygote state. However, the effect of the variant in hemi-or homozygous states should be explored in greater depth. None of the patients carried the intronic variant rs3915512, which has previously been associated with reduced cortical expression of a splicing variant in DLG1 in patients with nonearly-onset schizophrenia (Uezato et al., 2015), and with neurocognitive function in schizophrenia in general (Xu et al., 2018). Neither carried rs9843659, which has been associated with schizophrenia in male patients in a recessive model (Uezato et al., 2012).
Patient 2 also carried a predicted moderate-risk variant (rs939885) in SLC51A. SLC51A is an essential component of the Ost-alpha/Ost-beta complex, a heterodimer that acts as an organic solute transporter of the sulfated neurosteroids pregnenolone sulfate (PREGS) and dehydroepiandrosterone | 9 of 12 MALT eT AL. sulfate (DHEAS) (Grube et al., 2018). Ost-alpha deficient mice exhibited changes in serum DHEA and DHEAS levels, and in tissue distribution of administered DHEAS (Fang et al., 2010). In the human brain, the Ost-alpha/Ost-beta transporter is expressed in Purkinje cells and hippocampal neurons (Fang et al., 2010). After delivery to their target sites, they act as endogenous neurotransmitters or modulators on γ-aminobutyric acid (GABA), particularly type A (GABA A ), N-methyl-aspartate (NMDA) glutamate, serotonin (5-HT3), and alpha-1 receptors (Akwa, Ladurelle, Covey, & Baulieu, 2001;Smith, Gibbs, & Farb, 2014;Tuem & Atey, 2017;Whittaker, Gibbs, & Farb, 2008). PREGS may also act as a negative allosteric modulator of the AMPA, kainate, and glycine receptors, potassium channels, and voltage-gated sodium channels; it may also inhibit voltage-gated calcium channels and interact with nicotinic cholinergic receptors and sigma receptors (Rajagopal, Soni, & Meltzer, 2018). In this way, they exert potent effects on neuronal excitability and neurotransmitter receptor function. Modeling of the variant was not possible as no crystal structure is available, but the valine to isoleucine change is generally believed to be a conservative amino acid change. We suspect therefore that the present amino acid substitution might be too minor to lead to the functional impairment of enzyme activity.
Roles for PREGS and DHEAS in the etiopathogenesis of schizophrenia have been suggested (Vuksan-Cusa, Sagud, & Rados, 2016;Wong, Sze, Chang, Lee, & Zhang, 2015). Blood DHEA and DHEAS levels of schizophrenia patients and healthy subjects have been found to differ across studies, ranging from normal to low and to high levels. Although several randomized, double-blind, placebo-controlled clinical trials have been conducted with DHEA(S) for the augmentation treatment of schizophrenia, their results have been conflicting (Vuksan-Cusa et al., 2016). In a dopamine transporter knockout mouse model of schizophrenia, acute treatment with PREGS was able to rescue behavioral anomalies through the NMDA receptor-mediated AKT/GSK3β signaling pathway (Wong et al., 2015). The current body of evidence supports further research with PREGS as a cognitive enhancer, as well as for treating negative symptoms in schizophrenia and related disorders (Rajagopal et al., 2018). The role of the identified risk variant in SLC51A in schizophrenia should be explored further.
Some of the risk variants we identified were also present as homozygous in first-degree relatives. We have no indication that these variants are relevant to the patients' phenotype.
In addition to genetic risk variants, haploinsufficiency of dose-response genes may contribute to a certain phenotype (Costain et al., 2019). Our patients are hemizygous for TFRC. The receptor is primarily involved in iron homeostasis by regulating cellular iron uptake, and haploinsufficiency of TFRC is known to influence iron homeostasis with hematological abnormalities similar to those of our patients (Levy, Jin, Fujiwara, Kuo, & Andrews, 1999;Matak et al., 2016). Whether disturbed iron homeostasis and hematological abnormalities may be related to the chronic fatigue experienced by our patients is currently unknown.
The small sample size consisting of only two families is an important limitation of our study. The patients have different population background, which potentially may impact psychiatrically relevant phenotypes. However, the 3q29 deletion is known to be associated with ASD and schizophrenia both in Caucasian and Asian populations (Glassford et al., 2016;Kushima et al., 2018).
In conclusion, our in-depth characterization of two adult carriers of the 3q29 deletion have broadened the phenotypic spectrum of this syndrome and shown that the deletion carriers may undergo a developmental phenotypic transition. The chronic pain and fatigue experienced by both patients underscores the need for regular clinical follow-up and support. We have identified some risk variants in the remaining hemizygous allele that should be explored in greater depth regarding a possible role in neuropsychiatric disorders in a broader patient population. Our study points at genes involved in cilia function and neurosteroid transport, in addition to neural connectivity and excitability, neurotransmitter receptor function, and synaptic plasticity. Improved knowledge of the functional roles of genes and gene variants in the 3q29 chromosomal region may aid us in devising individually targeted, personalized interventions in 3q29 deletion carriers and contribute to our understanding of serious neuropsychiatric disorders such as autism and schizophrenia in general.

ACKNOWLEDGMENTS
We thank the participating families for their willingness to contribute to this research. We also thank Bjørn Dalhus at the South-Eastern Norway Regional Health Authority Regional Core Facility for Structural Biology for structure modelling and Jon K Laerdal at the Bioinformatics Core Facility, Department of Informatics, University of Oslo, Oslo, Norway for assistance to validate predicted transcripts.