Coexisting Conditions Modifying Phenotypes of Patients with 22q11.2 Deletion Syndrome

22q11.2 deletion syndrome (22q11.2DS) is the most common genomic disorder with an extremely broad phenotypic spectrum. The aim of our study was to investigate how often the additional variants in the genome can affect clinical variation among patients with the recurrent deletion. To examine the presence of additional variants affecting the phenotype, we performed microarray in 82 prenatal and 77 postnatal cases and performed exome sequencing in 86 postnatal patients with 22q11.2DS. Within those 159 patients where array was performed, 5 pathogenic and 5 likely pathogenic CNVs were identified outside of the 22q11.2 region. This indicates that in 6.3% cases, additional CNVs most likely contribute to the clinical presentation. Additionally, exome sequencing in 86 patients revealed 3 pathogenic (3.49%) and 5 likely pathogenic (5.81%) SNVs and small CNV. These results show that the extension of diagnostics with genome-wide methods can reveal other clinically relevant changes in patients with 22q11 deletion syndrome.


Introduction
22q11.2DS is the most common chromosomal microdeletion syndrome in humans, with a prevalence of 1:2148 live births [1,2]. The morbidity in utero is higher with the frequency of 1:1000 fetuses [3,4] and 1:1500 miscarriage samples [5]. 22q11 deletion is a result of non-allelic homologous recombination between high identity low copy repeats (LCRs). Most of the affected individuals (~90%) carry a typical hemizygous 3 Mb deletion on chromosome 22q11.2, mediated by LCRs 22-A and 22-D, but smaller deletions mediated by other combinations of LCRs are also described [6]. This leads to a diminished dosage of nearly 60 genes, including coding and noncoding RNAs.
The characteristics of this contiguous gene deletion syndrome is highly variable phenotypic severity with many common significant physical and behavioral clinical findings described, as well as numerous rare significant features impacting clinical care, but none appear to be fully penetrant and each exhibits variable expressivity [2,7,8]. Variability is observed across unrelated patients and between affected family members who have inherited identical deletions, including amongst identical twins [9][10][11][12]. Approximately 10% of chromosome 22q11.2 deletions are parentally transmitted, and in those cases, high penetrance and wide expressivity are observed [11,13,14]. Frequently associated features of 22q11.2DS are congenital heart defects, palatal differences, immunodeficiency, endocrine abnormalities, developmental delay, cognitive deficits, and behavioral phenotypes [7,8,14,15]. Less commonly observed are structural ophthalmologic abnormalities, choanal atresia, hearing loss, laryngo-tracheo-esophageal abnormalities, gastrointestinal differences, congenital diaphragmatic hernia, vertebral anomalies, polydactyly, IUGR, idiopathic seizures, microcephaly, neural tube defects, ADHD, and autism [14]. Patients with 22q11.2DS have been reported to have a diminished adult life expectancy, with an increased risk of sudden death and developing psychiatric disorders [16,17].
It remains unknown why individuals with deletion of the same size present such a wide range of phenotypes. Some of the genes located within the standard deletion have major clinical impact, particularly T-box 1 (TBX1). The TBX1 gene is part of the larger family of T-box genes, which help to regulate tissue and organ formation during development. However, a different minority of patients harbor nested distal deletions but retain two copies of the TBX1 gene [2]. Therefore, the single dose of deleted genes alone cannot explain the tremendous variation in the severity and penetrance of the associated clinical features among affected individuals. Differences in deletion extents also cannot account for the huge variability in phenotypic severity among patients with 22q11.2DS [18]. Currently, there is no association between deletion size and the presence of the most common clinical symptoms, congenital heart disease or palate anomalies [19], although the common clinical features are less penetrant in smaller atypical nested deletions [14]. This points out that genetic variants, beyond the deleted genes themselves, may play a role in determining the 22q11.2DS clinical complications. Therefore, sequencing and microarray methods can help uncover, at least partially, genetic differences among patients with 22q11.2DS that can influence disease severity.
Several potential genetic mechanisms can underlie these highly variable phenotypic manifestations, including that they may be attributed to the hemizygous variants on the remaining allele, due to additional rare pathogenic variants elsewhere in the genome, as well as nongenetic environmental factors. Unmasking autosomal recessive disorders by having a deletion in one 22q11.2 allele and a mutation on the other non-deleted allele has been described in GP1BB (Bernard-Soulier syndrome type B) [20], SNAP29 (CEDNIK syndrome-Cerebral dysgenesis, neuropathy, ichthyosis, and keratoderma) [21], LZTR1 (Noonan syndrome) [22,23], SCARF2 (van den Ende-Gupta syndrome (VDEGS) [24], and CDC45 (CGS syndrome-craniosynostosis, gastrointestinal differences, short stature, and skeletal differences/Meier-Gorlin syndrome/Baller-Gerold syndrome) genes [25,26]. Mapping the variation in the remaining alleles in patients with 22q11.2DS also identified potential variants in PI4KA and PRODH [27]. Recently, known pathogenic genetic factors beyond the 22q11.2 region, including mutations and chromosomal structural aberrations, were described in~0.9% of patients with 22q11.2DS with atypical phenotypic features [28].
Our goal was to determine the presence of additional variants affecting the phenotype of patients with 22q11.2DS. To obtain the most detailed clinical description of the patients in each case, medical consultation was carried out by experienced clinical geneticists. A standard form (Supplementary Table S1) for deep phenotyping was prepared at the Institute of Mother and Child on the basis of review of the literature and included clinical features that may occur in individuals with 22q11.2DS. The clinical symptoms were grouped by category: cardiovascular system, immunodeficiency, autoimmune disease, palatal abnormalities/velopharyngeal insufficiency/dysfunction, laryngo-tracheo-esophageal anomalies, endocrinologic problems including hypocalcemia, short stature, intrauterine growth deficiency/failure to thrive (FTT), macrocephaly, microcephaly, structural eye defects and ocular disorders, skeletal abnormalities, dental problems, structural ear defects and hearing problems, structural CNS anomalies, unprovoked seizures, genitourinary tract anomalies, inguinal hernia, gastrointestinal anomalies and functional problems, umbilical hernia, craniofacial dysmorphic features, psychomotor and intellectual development, learning disabilities/cognitive deficits, behavioural abnormalities, and psychiatric illness. A detailed form with over 230 clinical features was completed where possible (Supplementary Table S2). The research was approved by the Bioethics Committee of the Institute of Mother and Child in Warsaw and the Institutional Review Board of the Children's Hospital of Philadelphia.

aCGH
We performed array comparative genomic hybridization (arrayCGH), with commercially available arrays such as the 60K CytoSure Constitutional v3 microarray (Oxford Gene Technology, Oxford, UK) according to manufacturer's instructions. CNVs that showed partial or complete overlap with known segmental duplications were excluded from further analysis, due to the high variability of copy number variations in those regions. CNVs were classified as pathogenic, likely pathogenic, likely benign, benign and of uncertain significance (VUS) based on clinical data in known CNV databases: Database of Genomic Variants (http://dgv.tcag.ca/dgv/app/home accessed on 18 November 2022), ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/ accessed on 18 November 2022), DECIPHER (https://www.deciphergenomics.org/ accessed on 18 November 2022), ClinGen (https://www.clinicalgenome.org/ accessed on 18 November 2022). Recurrent CNVs or CNVs associated with known microdeletion or microduplication syndromes were classified as pathogenic or likely pathogenic depending on the penetrance and clinical features present in probands. In 159 patients (82 prenatal and 77 postnatal), array CGH was performed as a first diagnostic test. Additionally, we conducted aCGH where exome sequencing revealed CNV, which could be confirmed (due to size and array coverage).

Exome Sequencing
Exome sequencing (ES) was applied to identify rare variants that could affect the phenotype in 86 patients with 22q11.2DS. Detailed phenotyping of all patients was performed (Supplementary Table S2) to estimate the correlation between ES results and specific phenotypes.
Sequencing was performed using Illumina HiSeq (San Diego, CA, USA) instruments after exome capture with the Sure Select All Human V6 design. Raw sequence data were post-processed using the bc-bio pipeline (https://github.com/chapmanb/bcbio-nextgen accessed on 18 November 2022). The bc-bio pipeline performs mapping of the short reads against a human genome reference sequence (GRCh38) using the Burrows-Wheeler Alignment (BWA), BAM post-processing using GATK, and variant calling using the GATKHaplotype Caller. Finally, ANNOVAR [29] was used to annotate relevant information about gene names, predicted variant pathogenicity, reference allele frequencies and metadata from external resources, and then to add these to the Variant Call Format (VCF) file.
Next, we used HMZDelFinder algorithm [30] to search for small, hemizygous deletions within the 22q11.2 region and small, heterozygous deletions beyond 22q11.2 region. As a control data set, we used WES data from 164 samples sequenced at the same platform and processed using the same pipeline, as our patients with 22q11.2DS. To prepare input data for HMZDelFinder algorithm, for each sample for every exonic target from the capture design we computed RPKM values, i.e., number of reads per kilobase of transcript per million mapped reads. To detect heterozyous CNVs outside of the 22q11.2 region we used CoNIFER tool with its default parameters [30]. As an input we reused the same rpkm data previously precomputed to run HMZDelFinder. Each putative CNV identified by CoNIFER has been visualized and undergone manual inspection to remove common polymorphisms and likely false positive calls.

aCGH Analysis
In 21 patients, in addition to the chromosome 22q11.2 loss, we identified 31 copy number variants. Among those 31 CNVs, 5 were pathogenic, 5 were likely pathogenic, and 21 CNVs were classified as variants of uncertain significance (Table 1). In 8 patients, more than one additional CNV was observed. Ten out of thirty-one additional CNVs also occurred on chromosome 22q, but in different regions, however in 7 of those cases the additional deleted regions most likely were the extension of the typical or nested 22q11. Importantly, these variants would not be possible to identify using targeted methods, like FISH or MLPA.

Variants in Genes from Remaining 22q11.2 Region
Hemizygotic variants in the 22q11.2 deletion region were analyzed with a frequency of 0.01 in gnomAD database. Next, only exonic, stopgain, nonsynonymous, and frameshift variants were taken into analysis. Across all 86 patients, a total of 29 variants in 21 patients were identified in the remaining 22q11.2 region (Supplementary Table S3). None of the variants were previously described as pathogenic or likely pathogenic, therefore all variants were classified as variants of uncertain clinical significance. However, investigation of the remaining allele in our tested cohort revealed a rare nonsynonymous-damaging variant in CDC45 in one patient out of 8 with craniosynostosis. In addition to bilateral coronal craniosynostosis, this patient (GC028958) also presented with additional clinical features overlapping with the previously described MGORS7/CGS syndrome, such as an anteriorly placed anus.
HMZDelFinder identified two putative hemizygous deletions in the remaining 22q11.2 region. The first deletion encompasses two genes; DGCR6, PRODH (Chr22:18906447-18931248). This deletion is classified in ClinVar as likely benign and had no effect on the patient's phenotype, because the region is commonly deleted in control populations. However, this deletion was never described in patients with 22q11.2DS. Therefore, we cannot exclude the possibility that nullisomy of PRODH has an impact on the patient's long-term phenotype. This deletion was classified as uncertain clinical significance. The second deletion (hemizygous), also identified in the 22q11.2 region, encompasses 8 exons of TANGO2 (chr22:20043265-20064650). Bi-allelic truncating mutations in this gene have been associated with TANGO2-related metabolic encephalopathy and arrhythmias presenting with recurrent muscle weakness with rhabdomyolysis, metabolic crises, and cardiac arrhythmia [31]. Therefore, this variation was confirmed by a qPCR method, classified as pathogenic, and was consistent with the patient's phenotype ( Figure 1).  this gene have been associated with TANGO2-related metabolic encephalopathy and ar rhythmias presenting with recurrent muscle weakness with rhabdomyolysis, metaboli crises, and cardiac arrhythmia [31]. Therefore, this variation was confirmed by a qPCR method, classified as pathogenic, and was consistent with the patient's phenotype ( Figure  1).

Filtering for SNVs Elsewhere in the Genome
Filtering for known pathogenic variants, defined as a frequency less than or equal to 0.001 in gnomAD genome database, autosomal dominant inheritance, and classified a pathogenic in ClinVar database, revealed 1 rare stopgain, 1 rare frameshift deletion, 1 rare UTR3 variant, and 4 rare non-synonymous variants. In total, 6 genes across 86 patient A B

Filtering for SNVs Elsewhere in the Genome
Filtering for known pathogenic variants, defined as a frequency less than or equal to 0.001 in gnomAD genome database, autosomal dominant inheritance, and classified as pathogenic in ClinVar database, revealed 1 rare stopgain, 1 rare frameshift deletion, 1 rare UTR3 variant, and 4 rare non-synonymous variants. In total, 6 genes across 86 patients carry a rare protein-altering variant. The clinical outcome of these variants was discussed with clinicians for all individuals (Table 2). For two cases, the clinical features were consistent with the predicted consequences of genomic variant and were classified as pathogenic. In five cases, the patients did not present expected phenotype features, and therefore the variants have been classified as potentially pathogenic.
Rare SNVs were selected using filtering with an allele frequency less than 0.001 in gnomAD and were then filtered against an in-house control dataset. Variants in genes associated with OMIM diseases, of autosomal and X-linked dominant inheritance or X-linked recessive inheritance in male patients, were chosen. Effects of missense variants were predicted using four in silico analysis tools (SIFT; Polyphen2_HDIV; Polyphen2_HVAR; MutationTaster). SNVs with deleterious predictions in at least 2 out of 4 prediction tools or SNV automatically deleterious in Mutation Taster were selected. Next, phenotypic features of patients with those SNVs were compared with clinical features related to the implicated genes. Only those variants, where the clinical features overlapped with predicted phenotypes, were included in the table (Table 3). The pathogenicity of all selected SNVs was estimated using Alamut-2.11-0 software.
Additionally known cancer susceptibility SNVs in genes associated with cancer risk have been identified in 5 patients (Supplementary Table S3). The variants have been revealed in 3 genes: ATM, BRCA1, and BRCA2. All patients were informed regarding these risk genes and were referred to the oncology clinic.
Overall, 885 variants of uncertain clinical significance were selected based on the frequency (less than 0.0001 in gnomAD database), function (exonic), and type of change (frameshift deletion, frameshift insertion, stopgain, or nonsynonymous (Supplementary Table S4).
In summary, we have identified 8 pathogenic variants, and 10 likely pathogenic variants in 225 patients, which is much more than previously expected (Table 5). Table 2. Pathogenic and likely pathogenic SNVs, of autosomal dominant inheritance, were selected using filtering with an allele frequency less than 0.001 in gnomAD, then filtered against an in-house control dataset. EBD-Epidermolysis bullosa dystrophica.      encodes γ-aminobutyric acid type A receptor alpha5 subunit; restricted expression toward brain; GABA is the major inhibitory neurotransmitter in the mammalian brain where it acts at GABA-A receptors, which are ligand-gated chloride channels GC028958 F 2 weeks chr15_26937134_26937376 exon 8 of GABRA5

ID
encodes γ-aminobutyric acid type A receptor alpha5 subunit; restricted expression toward brain; GABA is the major inhibitory neurotransmitter in the mammalian brain where it acts at GABA-A receptors, which are ligand-gated chloride channels

Discussion
In our study, we have analyzed genomes of 225 patients with 22q11.2 deletion syndrome, using array CGH and/or exome sequencing methodology. The aim of our research was to identify variants beyond the 22q11.2 deletion, which may contribute to the clinical complexity in our patients.
Hemizygous variants are one of the potential genetic mechanisms that could underlie this clinical variability. Allelic variations of genes within a critical region of the non-deleted chromosome can have an impact on phenotypes by unmasking recessive diseases, which has already been proven for several genes within the 22q11.2 region, like SNAP29 and GP1BB [21,32]. Recent studies have mainly focused on mapping mutations on the remaining allele [27,33]. This strategy has demonstrated high efficiency in cases with atypical and less common clinical features. For example, biallelic mutations in CDC45 are thought to be the cause of craniosynostosis in Meier-Gorlin syndrome (MGORS7, OMIM 617063) [25], a rare autosomal recessive primordial dwarfism disorder, characterized by microtia, short stature, and absent or hypoplastic patellae. Recently, in patients with 22q11.2DS, the alteration of CDC45 was associated with the pathogenesis of craniosynostosis, as well as phenotypic features more frequently reported in association with Baller-Gerold (OMIM 218600) and RAPADILINO syndrome (OMIM 266280), as well as unique previously-unreported features [26]. In fact, rare hemizygous variants in the CDC45 gene were found in 5/15 (33%) of the patients reported by Unolt et al., with both a 22q11.2 deletion and atypical clinical features in 3/7 of the patients (43%) including craniosynostosis (2 bicoronal, 1 metopic) and none of the 133 22q11.2DS patients from the control group (Fisher's test p value < 0.01). These patients also had important structural anomalies not typically associated with 22q11.2DS, most significantly some combination of anal/intestinal anomalies, limb differences, short stature, and craniofacial anomalies. Investigation of the remaining allele in our tested cohort revealed a rare nonsynonymous damaging variant in CDC45 in one patient out of eight with craniosynostosis. In addition to bilateral coronal craniosynostosis, this patient also presented other clinical features overlapping with MGORS7/CGS, such as an anteriorly placed anus.
In two patients, we found additional deletions on the remaining allele. In one patient, an intragenic deletion of 8 exons in TANGO2 was identified following a postmortem study. Mutations in this gene are responsible for TANGO2-related metabolic encephalopathy and arrhythmias, an autosomal recessive recurrent metabolic encephalomyopathic crises, rhabdomyolysis, cardiac arrhythmia, and intellectual disability syndrome (OMIM 616878). A large proportion of patients described in the literature have succumbed in childhood due to cardiac arrest related to arrhythmias or following seizures related to a hypoglycemic crisis. Some patients, with variants in the TANGO2 gene, presented primarily with neurologic features [34]. In our second patient, we identified a deletion encompassing the whole PRODH gene. Mutations or small deletions in PRODH cause recessive hyperprolinemia (OMIM 239500) with neurologic manifestations, including seizures. It is well known that individuals with 22q11.2DS are at high risk for developing schizophrenia and schizoaffective disorder. PRODH was suggested to play a role in the later stages of neural development [35]. Although the latest research does not support the theory that high proline level is responsible for psychosis, it cannot be excluded that nullisomy of the PRODH gene has no clinical effect. Despite the association between variants in PRODH and an increased risk of schizophrenia shown in some studies, the potential role of this gene in psychiatric diseases remains unclear [36,37].
Hemizygous variants on a second allele may explain only part of the clinical variability associated with 22q11.2DS. The phenotypic variability may also be associated with additional copy number variations, as was shown in our study.
Analysis of other microdeletions with variable expressivity suggests that a two-hit model may be more generally applicable to neuropsychiatric disease [38]. Additional CNVs were already suggested to be genomic modifiers in patients with a 22q11.2 deletion or duplication [39,40]. In our study we have identified 31 additional CNVs in 21 patients, among 159 tested by array CGH. It is worth noting that more than one CNV was observed in 8 patients in our cohort, worsening the overall clinical picture. In 10 cases the additional CNV was located on chromosome 22q, however in 7 of those cases the additional deleted regions most likely were the extension of the typical or nested 22q11 deletions. It is particularly interesting that additional CNVs occurred in 7 patients with nested 22q11 deletions and 14 patients with A-D typical deletion. Taking into account the ratio of nested to the typical deletions (28:131 in our study), the result can suggest that if the smaller deletions have a milder phenotype than the classical ones, they may more likely have additional CNVs affecting phenotype. However, to draw final conclusions, more cases with different sizes of the deletion need to be studied. Moreover, in 3 cases, recurrent deletions with variable expressivity on chromosome 15 and 16 were revealed. Recurrent proximal 16p11.2 microdeletion found in case GC034796 can also influence some phenotypic features like global developmental delay, hypotonia, or cardiac anomalies. Rare CNVs outside of the 22q11.2 region may modify risk for congenital cardiac defects in some patients, with 22q11.2DS probably being due to the content of genes influencing heart development [41,42].
In patient GC034823, who demonstrated cognitive regression and dementia, both of which are unusual for 22q11.2DS, the deletion region is expanded and encompasses the TUBA8 gene. Mutations in TUBA8 were associated with recessive polymicrogyria [43]. Recently, Dantas et al. reported downregulation of this gene in patients with 22q11.2 deletion syndrome [44]. Another CNV encompassing gene associated with an intellectual disability is partial duplication of the MID2 gene, which was revealed in male patient GC034808. Previously, a duplication overlapping MID2 was reported in a boy with FG syndrome, an X-linked multiple congenital anomalies syndrome (OMIM 300581), who had hypotonia and developmental delay [45]. A missense mutation in MID2 was described in a large Indian family with global developmental delay and minor facial features [46]. Recently, an Xq22 duplication including MID2 was described in a patient whose phenotypic features were not consistent with FG syndrome [47]. Another mechanism modifying patients' clinical features is the influence of single nucleotide variations (SNVs) in genes that reside outside of the deleted region, e.g., SNVs in genes that lie on the same pathway with genes from the deleted region can modify the severity of symptoms. A good example of such a mechanism demonstrates patient GC034900, with a variant in the TLL1 gene. Heterozygous mutations in the TLL1 gene are responsible for atrial septal defect (ASD6). In about 1% of patients with 22q11.2DS. another known pathogenic genetic factor, such as mutations, chromosomal structural aberrations. or mosaic aneuploidy, was observed, resulting in dual diagnoses [28]. An example of a dual diagnosis in our cohort is patient GC030951, with both 22q11.2DS and a pathogenic variant in the NF1 gene resulting in Neurofibromatosis as well as 22q11.2DS.
Historically, medical evaluation of genetic conditions relies upon pattern recognition, which recognizes a phenotype, signs, or symptoms to guide targeted genetic testing. The clinical features of many genetic conditions and 22q11.2 deletion syndrome may overlap. Therefore, most often, when a 22q11.2 deletion is identified, further diagnosis is frequently not considered in most patients, and additional or rare clinical features are attributed to the presence of the deletion. 22q11.2DS is a multi-system disorder caused by a single deletion that is genetically complex and can be seen as a model for gene-gene interactions and phenotype-genotype correlations. Recent advances in genetics and systems biology provide excellent opportunities to gain novel insights into pathways underlying the morbidities associated with 22q11.2DS across the lifespan.
Our study demonstrates that secondary diagnoses should be taken into account in all patients with 22q11.2DS. Only a whole-exome/whole-genome study of large cohorts of patients with 22q11.2DS will provide a robust opportunity to determine genotype-phenotype correlations for individual genes with the 22q11.2 deletion and beyond. Knowing how often additional variants can modify the patient's phenotype, it is recommended to use the aCGH/SNP microarray methodology in cases of suspected 22q11.2 deletion at minimum.   Informed Consent Statement: Informed consent was obtained from all subjects involved in the study. Data Availability Statement: Any data requests can be directed to the corresponding author.