ATR16 Syndrome: Mechanisms Linking Monosomy to Phenotype

Background Sporadic deletions removing 100s-1000s kb of DNA, and variable numbers of poorly characterised genes, are often found in patients with a wide range of developmental abnormalities. In such cases, understanding the contribution of the deletion to an individual’s clinical phenotype is challenging. Methods Here, as an example of this common phenomenon, we analysed 34 patients with simple deletions of ∼177 to ∼2000 kb affecting one allele of the well characterised, gene dense, distal region of chromosome 16 (16p13.3), referred to as ATR-16 syndrome. We characterised precise deletion extent and screened for genetic background effects, telomere position effect and compensatory up regulation of hemizygous genes. Results We find the risk of developmental and neurological abnormalities arises from much smaller terminal chromosome 16 deletions (∼400 kb) than previously reported. Beyond this, the severity of ATR-16 syndrome increases with deletion size, but there is no evidence that critical regions determine the developmental abnormalities associated with this disorder. Surprisingly, we find no evidence of telomere position effect or compensatory upregulation of hemizygous genes, however, genetic background effects substantially modify phenotypic abnormalities. Conclusions Using ATR-16 as a general model of disorders caused by sporadic copy number variations, we show the degree to which individuals with contiguous gene syndromes are affected is not simply related to the number of genes deleted but also depends on their genetic background. We also show there is no critical region defining the degree of phenotypic abnormalities in ATR-16 syndrome and this has important implications for genetic counselling.


Introduction
Cytogenetic, molecular genetic, and more recently, next generation sequencing 3 (NGS) approaches have revealed copy number variations (CNVs) in the human 4 genome ranging from 1 to 1000s kb (Iafrate et al., 2004;MacDonald et al., 2014). 5 CNVs are common in normal individuals and have been identified in ~35% of the 6 human genome (Iafrate et al., 2004). When present as hemizygous events, in used (Roche NimbleGen, Madison, WI, USA). Array design was based on NCBI 1 Build 36.1 (hg18) and used as previously described (Phylipsen et al., 2012). 2 Genomic DNA from patients SH(Ju) and SH(Pa) were tested using CytoScan HD 3 arrays (Affymetrix) and analysed using Karyoview software. aCGH was performed 4 with genomic DNA from patients CJ, IM and YA using the Sentrix Human CNV370 5 BeadChip (Illumina) and analysed using GenomeStudio software. 6 7 Whole Genome Sequencing (WGS) 8 WGS was carried out at Edinburgh Genomics, The University of Edinburgh. The 9 pathogenicity of each variant was given a custom deleterious score based on a six-1 0 point scale, (Fu et al., 2013) calculated using output from ANNOVAR (Wang et al., 1 1 2010). This was used to prioritise variants present in the hemizygous region of 1 2 chr16p13.3 in each patient and also genome wide. 1 3

2
Clinical Features of ATR-16 Syndrome 3 All individuals with ATR-16 syndrome have α -thalassemia because at two of the four 4 paralogous α -globin genes are deleted (--/αα) and this manifests as mild 5 hypochromic microcytic anaemia. In combination with a common small deletion 6 involving one α-gene on the non-paralogous allele (--/-α) patients may have a more 7 severe form of α-thlassaemia referred to as HbH disease (Harteveld & Higgs 2010). 8 In addition to α -thalassemia, which is always present, common features of ATR-16 9 syndrome include speech delay, developmental delay and a variable degree of facial  Deletions larger than 2000 kb including the PKD1 and TSC2 genes lead to severe 1 3 MR with polycystic kidney disease and tuberous sclerosis respectively (European 1 4 Polycystic Kidney Disease Consortium, 1994). 1 5 Eleven patients from 9 families are reported here for the first time (OY, LA, 1 6 TY(MI), TY(Mi), YA, SH(P), SH(Ju), NL, CJ, MY and BAR) and we refine the 1 7 breakpoints in 4 previously reported families (BA, TN, IM, LIN). We define 1 8 breakpoints at the DNA sequence level in 7 of the 13 families studied (Figure 1), 6 of 1 9 which have been repaired by the addition of a telomere or subtelomere. In the 2 0 remaining family (SH) the deletion is interstitial and mediated by repeats termed 2 1 short interspersed nuclear elements (SINEs). Identification of Co-Inherited Deleterious Loci 2 4 Three families (LA, YA and TN) have 16p13.3 deletions smaller than 1 Mb Mb may be unmasking deleterious mutations on the intact chromosome 16 allele in 2 7 severely affected patients, we performed whole genome sequencing (WGS) where 2 8 DNA was available (YA and the three affected members of the TN family) and 2 9 considered only coding variants in the hemizygous region of chromosome 16.

0
However, only common variants (allele frequency >5%) were present 3 1 Table 3) suggesting the cause(s) of the relatively severe phenotypes 3 2 in these patients reside elsewhere in the genome. To identity rare variants we 3 3 considered only those absent from the publicly available databases. This analysis 3 4 yielded 14 variants shared between the three affected individuals of family TN 3 5 (Supplementary Table 4). Of these, only one (chr15:64,782,684 G>A) affects a gene 3 6 likely to be involved in the broader ATR16 phenotypic abnormalities. This change 1 leads to a R12X nonsense mutation in SMAD6, a negative regulator of bone 2 mophogenetic protein (BMP) signalling pathway. Heterozygous mutations in SMAD6 3 have been reported to underlie craniosynostosis, speech delay, global  Further evidence the effect of ATR-16 deletions is modified by other loci 1 3 comes from patients SH(Ju) and SH(Pa) who harbour the same chromosome 16 1 4 deletion. Patient SH(Pa) has developmental delay and skeletal abnormalities, 1 5 however, his mother SH(Ju) does not have craniofacial nor skeletal abnormalities nor 1 6 developmental delay although she suffers from severe anxiety and depression (see 1 7 Figures 1,2 and Supplementary Information). Genome wide array comparative 1 8 genomic hybridisation (aCGH) analysis revealed that both SH(Ju) and SH(Pa) 1 9 harbour a ~133 kb deletion on the short arm of chromosome 2 including exons 5 to involved in the formation of synaptic contacts and has been implicated in autism 2 2 spectrum disorder, facial dysmorphism, anxiety and depression, developmental delay 2 3 and speech delay (e.g. The Autism Genome Project Consortium 2007; Kirov et al., 2 4 2008). This finding offers an explanation for the differences in the phenotypic severity 2 5 of the ATR-16 syndrome affecting these patients as autism differentially affects 2 6 males and females (see Discussion). Chromatin Structure 2 9 Recent reports demonstrate chromosomal rearrangements, including  contact frequency is shown for the terminal 2 Mb of chromosome 16 in Figure 2A to 3 3 illustrate the effect of the deletions reported here on the chromatin structure. The 3 4 deletion in BA removes ~50% of the self-interacting domain in which CHTF18, 3 5 RPUSD1, GNG13, and LOC388199 reside thereby potentially removing cis-acting 3 6 regulatory elements of these genes, although the genes themselves remain intact. In  One explanation for the relatively mild abnormalities in many cases of ATR- 16. This has been described as part of the mechanism of genetic compensation, also 1 3 termed genetic robustness (El-Brolosy and Stainer, 2017). To assay for 1 4 compensatory gene transcription we used qPCR to measure expression of 12 genes 1 5 within the terminal 500 kb of chromosome 16 in lymphoblastoid cells from 20 normal 1 6 individuals and from 10 patients with monosomy for the short-arm of chromosome 16 1 7 and found no evidence of compensatory upregulation: transcripts of all deleted genes 1 8 were present at ~50% of the normal levels in these cells ( Figure 3B). It is possible 1 9 that other genes in downstream pathways affected by haploinsufficiency may be 2 0

(Supplementary
transcriptionally upregulated, however, the mechanisms underlying this are complex 2 1 and beyond the scope of this study. Telomere Position Effect 2 4 Previous work in human cells has shown that telomeres may affect chromatin 2 5 interactions at distances of up to 10 Mb away from the chromosome ends (Robin et 2 6 al, 2014) reducing expression of the intervening genes. This phenomenon, termed 2 7 telomere position effect (TPE) is thought to be mediated by the spreading of 2 8 telomeric heterochromatin to silence nearby genes. In budding yeast this effect can 2 9 extend a few kb towards the sub-telomeres, although in some cases yeast telomeres  To determine the effect of telomere proximity on genes adjacent to telomere- 3 3 healed breakpoints we measured their expression relative to the allele present in a 3 4 normal chromosomal context. To achieve this, we screened them for informative 3 5 single nucleotide polymorphisms (SNPs) in EBV transformed lymphoblastoid cells 3 6 generated from ATR-16 patients. The phase of polymorphisms was established 3 7 using mouse erythroleukemia (MEL) cells fused to patient cells and selected to 1 contain a single copy of human chromosome 16, generated as previously described 2 (Zeitlin & Weatherall 1983). Expressed coding polymorphisms were present in genes 3 whose promoters are <60 kb away from breakpoints in 3 patients: TY, MY and BA. 4 For TY the nearest gene expressed in lymphoblastoid cells containing a 5 coding polymorphism is WDR90, the promoter of which is ~43.1 kb from the 6 abnormally appended telomere ( Figure 3A). For BA, CHTF18 is the closest 7 expressed polymorphic gene with the promoter ~16.3 kb away from the breakpoint. sequenced amplified fragments containing informative polymorphisms. We compared 1 3 peak heights of polymorphic bases in chromatograms derived from cDNA and 1 4 genomic DNA. None of the alleles assayed in the three patients tested showed any 1 5 evidence of a repressive effect ( Figure 3A). size, however, there is no clear correlation. 8 The deletions in patients reported here range from ~0.177 Mb to ~2 Mb. 9 Previous studies suggest the critical region leading to abnormalities in addition to α-1 0 thalassemia is an 800 kb region between ~0.9 and ~1.7 Mb from the telomere of SOX8 may not lead to MR with complete penetrance and any "critical region" for MR 1 5 starts after this point (Bezerra et al., 2008; Family "F" in Figure 2). Supporting this we 1 6 report patients NL and BAR, who have deletions of ~1.14 Mb and ~1.44 Mb 1 7 respectively and show no abnormalities beyond α-thalassemia. 1 8 By contrast, we find LA (deletion ~408 kb) has speech delay and YA (deletion 1 9 ~748 kb), has speech and developmental delay and facial dysmorphism ( Figure 2).

0
Family members of YA also have omphalocele, umbilical hernia and pyloric stenosis 2 1 suggesting there are other loci rendering YA susceptible to developmental 2 2 abnormalities. BA (deletion ~762 kb), who has a similarly sized deletion to YA, has 2 3 developmental delay but no other abnormalities. Four other patients with deletions <1 2 4 Mb (YA, TN(Pa), TN(Pe) and TN(Al)) have speech delay and facial dysmorphism. 2 5 This suggests the risk of developmental and neurological abnormalities arises from 2 6 much smaller terminal chromosome 16 deletions (~400 kb) than previously reported. 2 7 In SH(Pa) we have identified a strong candidate for the discordant 2 8 abnormalities: SH(Pa) also has a deletion of NRXN1, disruptions of which cause 2 9 autism and a range of neurological disorders. There is a higher incidence of autism in 3 0 males than in females, with a ratio of 3.5 or 4.0 to 1 (reviewed in Volkmar et al., 3 1 2004). This phenomenon is also specifically found in individuals with autism resulting 3 2 from rearrangements of NRXN1: Kirov and colleagues (2008) reported two affected 3 3 siblings who inherited a deletion of NRXN1 from their unaffected mother. It is 3 4 therefore possible Sh(Ju) is protected by her gender from the effects of NRXN1 3 5 disruption while the neurological and skeletal abnormalities in Sh(Pa) arise from the 3 6 complex interaction of NRXN1 perturbation with his gender and coinheritance of the present elsewhere in the genome. These may be rare variants (such as those 1 1 identified in the TN and SH families) or common variation; a recent study that shows 1 2 that common genetic variants (allele frequency >5% in the general population) 1 3 contribute 7.7% of the variance of risk to neurodevelopmental disorders (Niemi et al., 1 4 2018), highlighting the complexity of this area. 1 5 Together these observations suggest that monosomy for 16p13.3 unmasks 1 6 the effects of other variants genome-wide. This is supported by findings in SCH who 1 7 has a very similar deletion to BAR and may be more severely affected owing to the 1 8 presence of other CNVs (Scheps et al., 2016). At the other end of the spectrum, 1 9 large ATR-16 deletions may be associated with relatively mild abnormalities. In LIN abnormalities are very mild and there is no evidence of language delay. Here we 2 3 propose chromosome 16p13.3 deletions larger than 400 kb predispose to MR and 2 4 associated developmental abnormalities, however, we find no evidence for critical 2 5 regions that incrementally worsen ATR-16 syndrome abnormalities.

6
We could not detect compensatory up-regulation of the homologues of authors speculate that haploinsufficiency of the tumour suppressor AXIN1 may have 3 0 contributed to the neuroblastoma. Our finding that the remaining AXIN1 allele shows 3 1 no compensatory expression supports this hypothesis. 3 2 Terminal chromosome deletions are the most common subtelomeric 3 3 abnormalities (Ballif et al., 2007). The 16p deletions reported here are among the 3 4 most common terminal deletions along with 1p36 deletion syndrome, 4p terminal 3 5 deletion (causing Wolf-Hirschhorn syndrome), 5p terminal deletions (causing Cri-du- 3 6 chat syndrome), 9q34 deletion syndrome and 22q terminal deletion syndrome.  The presence of high-and low-copy-number repeats at breakpoints may play leading to terminal deletion. Following breakage, chromosomes can acquire a 1 4 telomere by capture or de novo telomere addition, which is thought to be mediated 1 5 by telomerase and this is stimulated by the presence of a telomeric repeat sequence 1 6 to which the RNA subunit of telomerase can bind (reviewed in Hannes et al., 2010). 1 7 We found 5 out of 6 telomere healed events share microhomology with appended 1 8 telomeric sequence. This is the same ratio (5 out of 6 breakpoints with may also be that telomerase binding to internal binding sites may inappropriately add 2 4 telomeres and thereby contribute to the generation of the breakpoints.

5
The lack of evidence for TPE in silencing gene expression is surprising and at 2 6 variance with previous findings (Stadler et al., 2013), which show that TPE can 2 7 influence gene expression at least 80 kb from the start of telomeric repeats.

8
However, TPE is likely to be context and cell type dependent. Additionally, because 2 9 of the lack of informative expressed polymorphisms in the patients studied here it was not possible for us to assay expression of genes immediately adjacent to 3 1 telomeres and a more comprehensive screen may reveal TPE mediated gene 3 2 silencing closer to the telomere. Additionally, when the area of chromatin interaction 3 3 (visualised by HiC) is considered (Figure 2), contact domains for many genes 3 4 adjacent to chromosomal breaks are severely disrupted. This is likely to include the 3 5 loss of cis-acting regulatory elements and may bring the genes under the control of 3 6 illegitimate regulatory elements (Franke et al., 2016). Therefore, it is likely that genes 1 adjacent to breakpoints would be incorrectly spatiotemporally expressed. phenotypic abnormalities this has important implications for genetic counselling. 8 Analysis of patients with uncomplicated deletions also revealed unexpected 9 background genetic effects that alter phenotypic severity of CNVs. Acknowledgements: 1 3 The authors thank Markissia Karagiorga-Lagana, MD, ex Director, Thalassaemia 1 4 Unit, "Aghia Sophia" Childrens Hosital Athens Greece for referring case BAR. This  assessed patients and clinical data. All authors commented on the manuscript.