Refinement of the critical genomic region for congenital hyperinsulinism in the Chromosome 9p deletion syndrome

Background: Large contiguous gene deletions at the distal end of the short arm of chromosome 9 result in the complex multi-organ condition chromosome 9p deletion syndrome. A range of clinical features can result from these deletions with the most common being facial dysmorphisms and neurological impairment. Congenital hyperinsulinism is a rarely reported feature of the syndrome with the genetic mechanism for the dysregulated insulin secretion being unknown. Methods: We studied the clinical and genetic characteristics of 12 individuals with chromosome 9p deletions who had a history of neonatal hypoglycaemia. Using off-target reads generated from targeted next-generation sequencing of the genes known to cause hyperinsulinaemic hypoglycaemia (n=9), or microarray analysis (n=3), we mapped the minimal shared deleted region on chromosome 9 in this cohort. Targeted sequencing was performed in three patients to search for a recessive mutation unmasked by the deletion. Results: In 10/12 patients with hypoglycaemia, hyperinsulinism was confirmed biochemically. A range of extra-pancreatic features were also reported in these patients consistent with the diagnosis of the Chromosome 9p deletion syndrome. The minimal deleted region was mapped to 7.2 Mb, encompassing 38 protein-coding genes. In silico analysis of these genes highlighted SMARCA2 and RFX3 as potential candidates for the hypoglycaemia. Targeted sequencing performed on three of the patients did not identify a second disease-causing variant within the minimal deleted region. Conclusions: This study identifies 9p deletions as an important cause of hyperinsulinaemic hypoglycaemia and increases the number of cases reported with 9p deletions and hypoglycaemia to 15 making this a more common feature of the syndrome than previously appreciated. Whilst the precise genetic mechanism of the dysregulated insulin secretion could not be determined in these patients, mapping the deletion breakpoints highlighted potential candidate genes for hypoglycaemia within the deleted region.

neonatal hypoglycaemia. Using off-target reads generated from targeted next-generation sequencing of the genes known to cause hyperinsulinaemic hypoglycaemia (n=9), or microarray analysis (n=3), we mapped the minimal shared deleted region on chromosome 9 in this cohort. Targeted sequencing was performed in three patients to search for a recessive mutation unmasked by the deletion. Results: In 10/12 patients with hypoglycaemia, hyperinsulinism was confirmed biochemically. A range of extra-pancreatic features were also reported in these patients consistent with the diagnosis of the Chromosome 9p deletion syndrome. The minimal deleted region was mapped to 7.2 Mb, encompassing 38 protein-coding genes. In silico analysis of these genes highlighted SMARCA2 and RFX3 as potential candidates for the hypoglycaemia. Targeted sequencing performed on three of the patients did not identify a second disease-causing variant within the minimal deleted region. Conclusions: This study identifies 9p deletions as an important cause of hyperinsulinaemic hypoglycaemia and increases the number of cases reported with 9p deletions and hypoglycaemia to 15 making this a more common feature of the syndrome than previously appreciated. Whilst the precise genetic mechanism of the dysregulated insulin secretion could not be determined in these patients, mapping the deletion breakpoints highlighted potential candidate genes for hypoglycaemia within the deleted region.

Amendments from Version 1
Following the requests of reviewers, the title of the manuscript has been changed to highlight that 9p deletions cause Congenital Hyperinsulinism. The description of the clinical characteristics of the cohort have been updated to confirm that hypopituitarism was excluded as a cause of hypoglycemia in these patients. The results section has been updated to clarify which patients had been screened for mutations in the known hyperinsulinism genes.
The manuscript now includes discussion on the disruption of imprinted genes by large deletions. Figure 1 has been updated to include patient 3, the ID for the patient listed on Decipher and the size of the minimal deleted region. The SDS birthweights for the cohort have been added to Table 1 and the age at remission for patient 3 has been included. Table 2 has been expanded to show whether disruption of genes residing within the minimal deleted region are known to cause human disease.

Introduction
Monosomy of part of the short arm of chromosome 9 causes the complex congenital condition chromosome 9p deletion syndrome (MIM: 158170) 1 . These large contiguous gene deletions can occur in isolation or form part of an unbalanced translocation 2 . The cardinal clinical features of the 9p deletion syndrome are craniofacial dysmorphisms, including trigonocephaly, midface hypoplasia, flat nasal ridge, long philtrum, short neck and developmental delay. Other common features include musculo-skeletal abnormalities, congenital heart defects, abdominal wall defects and disorders of sexual differentiation 3 . A further rare feature is hypoglycaemia, which has been described in 3 of the >100 genetically confirmed cases 4-6 .
The phenotypic heterogeneity observed between individuals with the chromosome 9p deletion syndrome is likely to reflect differences in the extent of the deletion, with individual features resulting from haploinsufficiency of a specific gene(s). An example is seen in males with 46,XY gonadal dysgenesis (MIM: 154230) which has been linked to disruption of the putative sex-determining genes DMRT1 and DMRT2 on 9p 7 .
Recent efforts have focussed on defining the critical region for the 9p deletion syndrome but there have been some differences in results. Swinkels et al. refined the critical region to a 300 kb stretch of DNA on 9p22.3; however, this region did not overlap with the critical region mapped by Faas et al. 3,8 . Given the differences in the craniofacial features between the cohorts reported it seems likely that there is not a single 'critical region' for the 9p deletion syndrome but rather that the syndrome represents a phenotypically and genetically heterogeneous group of disorders with the extent of the deletion, and in some cases the reciprocal trisomy, determining the phenotype.
Congenital hyperinsulinism is a rare condition of hypoglycaemia due to dysregulated insulin production from pancreatic beta cells 9 . Despite major advances in genetics the underlying cause of congenital hyperinsulinism is not identified in approximately 55% of patients 10 . Studying patients with congenital hyperinsulinism and the 9p deletion syndrome provides an opportunity to further unravel the genetic underpinnings of dysregulated insulin secretion in congenital hyperinsulinism.
In this study we investigated the clinical and genetic characteristics of 12 patients with congenital hypoglycaemia and a large deletion on chromosome 9p. We mapped the genomic breakpoints in all 12 patients which allowed for refinement of the critical region for hypoglycaemia to 7.2 Mb encompassing 38 genes. We sought to identify candidate genes for congenital hyperinsulinism in this region, an approach which has been successfully employed for gene discovery in other conditions 11 .

Cohort
A total of 12 patients with large deletions of the short arm of chromosome 9 were identified (as described below) following referral for genetic testing for congenital hyperinsulinism or a history of neonatal hypoglycaemia. Informed consent for publication of the patients' details was obtained. This study was approved by the North Wales Research Ethics Committee (517/WA/0327).
The DECIPHER database was searched for individuals who had hypoglycaemia and deletions of chromosome 9p which overlapped with the deletions identified in our cohort 6 .

Calling deletions
In nine patients multiple syndromic features had prompted microarray analysis leading to the identification of a 9p deletion. In the remaining three patients a deletion on 9p was detected using SavvyCNV (release 1) using off-target reads from the next-generation sequencing analysis of the known congenital hyperinsulinism genes. This technique calls 97.5% of true CNVs >1Mb 12 .
Break points were mapped in patients 1-9 using off-target reads from the targeted next generation sequencing data. In these patients analysis of the known hyperinsulinism genes did not identify a mutation. In patients 10-12 the breakpoints were mapped by microarray analysis, DNA was not available for targeted sequencing in these individuals. In 5/11 patients the 9p deletion formed part of an unbalanced translocation ( Table 1).

Sequencing of the deleted region
To search for recessive mutations unmasked by the deletion, next generation sequencing was performed in three patients following targeted capture of the Chr9p24 region (patients 4, 5 and 6,   Libraries were sequenced on an Illumina HiSeq 2000 using 100 base paired-end reads. Sequence data was analysed using an approach based on the GATK best practice guidelines 13 . Reads were aligned to the GRCh37/hg19 human reference genome with BWA mem (version 0.7.15) 14 followed by local re-alignment using GATK IndelRealigner (version 3.7.0) 15 . Large sections of the region are low complexity and while mean target coverage was 36X, 34X and 41X, only 56%, 56% and 58% of the minimal deleted region was covered at 10X or above in the three samples, respectively. Variants were called using GATK haplotype caller and annotated using Alamut Batch (Interactive Biosoftware version 1.11, Rouen, France) (an open-access equivalent is ANNOVAR). 16 . We excluded variants present in gnomAD 17 at a frequency greater than 1 in 27,000 -the highest published prevalence of hyperinsulinism in an outbred population 18 . Variants that were homozygous in internal controls (n = 65) and intronic variants that were not predicted to affect splicing by the in silico tools MaxEntScan 19 , SpliceSiteFinder-like 20 , and NNSPLICE 21 were excluded.

Evaluation of protein expression of candidate genes in deleted region
The expression of genes within the deleted region was assessed by the median transcripts per million value from the Genotype-Tissue Expression (GTEx) portal.

Clinical characteristics
Clinical characteristics of the cohort are provided in Table 1.
Hypoglycaemia (blood glucose <3.0 mmol/l) was diagnosed in 9/12 patients at birth and in three patients at the age of 3 days, 8 weeks and 20 weeks respectively. In all patients cortisol deficiency was excluded clinically and biochemically at diagnosis and no patients had evidence of growth failure. In 10/12 patient's in our cohort and one patient reported in the literature 4 detectable insulin at the time of hypoglycaemia confirmed a diagnosis of congenital hyperinsulinism which was treated with diazoxide. In the four remaining patients, two from our cohort (patients 3 and 12) and two from the literature, a diagnosis of congenital hyperinsulinism was not confirmed. In three of these patients insulin was either not measured or the results were not reported (patient 12, reference 5 and https://decipher.sanger.ac.uk/patient/249708). In the final patient (patient 3) insulin was measured at the time of hypoglycaemia but was suppressed (less than 6.0 pmol/L). The duration of hypoglycaemia varied considerably within our cohort with one child having transitory hypoglycaemia not requiring treatment yet another patient requiring ongoing diazoxide treatment at 8 years.
Two patients within the cohort (patients 10 and 11) were affected siblings; the remaining 10 patients were unrelated and had no family history of hypoglycaemia. Extra-pancreatic features previously reported in patients with Chromosome 9p deletions were observed in all individuals although there was no uniform phenotype. Common features reported in our cohort include cardiac anatomical defects in seven patients, facial dysmorphism in ten patients, digit/limb abnormalities in six patients and undervirilisation in four patients ( Table 1).
The minimal deleted region for hypoglycaemia is 7.   Table 1. A list of the NCBI RefSeq genes within the 7.2Mb minimal deleted region is provided in Table 2. Table 2. Data on the genes within the minimal deleted region (Chr9:0-7200000[hg19], 9p24.3-9p24.1). pLI scores were obtained from gnomAD. Pancreatic expression was obtained from the Genotype-Tissue Expression (GTEx) portal (gtexportal.org). NA indicates the gene was not found in this database. When diseasecausing mutations have been reported details of the associated syndrome and the inheritance of mutations are provided. To test whether the deletion was unmasking a second recessively inherited mutation on the opposite allele we performed targeted capture followed by next generation sequencing of the minimal deleted region in three unrelated individuals. No rare variants shared by all three samples were identified. We also searched for genes harbouring different rare variants in each of the three samples but did not identify any genes which met this criterion.

Discussion
Our cohort of 12 patients with 9p deletions and hypoglycaemia is the largest reported series and significantly widens the phenotypic spectrum over and above the three reported cases. In 10 of the 12 patients congenital hyperinsulinism was confirmed, whilst in two patients insulin was either not measured at the time of hypoglycaemia or was shown to be appropriately suppressed. Variability in extra-pancreatic phenotypes was observed in our cohort with specific features likely to be determined by the extent of the deletion in each patient and in five cases the reciprocal trisomy. This study allowed the refinement of the critical region for the Chromosome 9p deletion syndrome which features hypoglycaemia to 7.2 Mb. The critical gene(s)/regulatory region(s) within this locus are not known.
As insulin was appropriately suppressed in one patient in our cohort and was not measured in three further individuals we cannot be certain that hypoglycaemia results from dysregulated insulin secretion in all cases with a 9p deletion. If these four patients do have a different mechanism for hypoglycaemia compared to the congenital hyperinsulinism group, we would, however, not expect the size of the minimal deleted region for congenital hyperinsulinism calculated in this report to change given that the deletions in these patients were not critical for determining the boundaries on the 7.2 Mb region ( Table 1 and Figure 1).
There are four possible mechanisms by which large deletions can cause disease: 1) disruption of a gene at the breakpoint 2) haploinsufficiency of a gene within the deletion 3) unmasking a recessive mutation in a gene within the deleted region and 4) disruption of an imprinted gene. None of the genes within the deleted region are known to be imprinted and the breakpoints for the deletions varied between patients in our cohort, making it unlikely that the disruption of a gene at a breakpoint is the cause of the hypoglycaemia in these patients. We performed targeted sequencing of the minimal deleted region to search for recessive mutations but did not identify any variants which could explain the phenotype. Although it is possible that our approach may have missed a mutation, given that only 56% of the minimal deleted region was captured at ≥10X coverage in three patients, from our data the most likely explanation it that haploinsufficiency of one or more genes within the minimal deleted region is responsible for the hypoglycaemia. If this is true we would expect this aetiology to be associated with variable penetrance given that patients without hypoglycaemia and deletions over this region have been reported 3 . This variable penetrance would be similar to what is observed with the gonadal dysgenesis phenotype where 46,XY patients with 9p24 deletions and normal male external genitalia have been reported 3,7 .
Interestingly, the deletions in four patients within our cohort do not overlap with the 3.5 Mb minimal deleted region defined by Faas et al. 8 and only two of our patients had a deletion which overlapped with the 300 kb critical region identified by Swinkels et al. 3 . The majority of deletions in our patients were called from sequence data by savvyCNV 8 which maps breakpoints with an estimated accuracy of ±200 kb. Even including this margin of error, not all of our patients have deletions which overlap with either of the previously identified critical regions. This is in keeping with the 9p deletion syndrome being a genetically and phenotypically heterogeneous collection of overlapping syndromes.
In conclusion, our study identifies 9p deletions as an important cause of hypoglycaemia and refines the critical region for this phenotype to 7.2 Mb. Whilst we highlight potential candidate genes the genetic mechanism for the hypoglycaemia in our patients remains unknown. Further studies are required to investigate the cause of hyperinsulinism in these patients and in those with other copy number variant (CNV) syndromes which feature congenital hyperinsulinism such as Turner's syndrome where the causative gene(s) have also not been definitively identified 23,24 . These large deletions can be screened for by targeted panels using an off-target CNV caller such as SavvyCNV 12 .

Underlying data
The genotype data could be used to identify individuals and so cannot be made openly available. Access to data is open only through collaboration. Requests for collaboration will be considered following an application to the Genetic Beta Cell Research Bank (https://www.diabetesgenes.org/currentresearch/genetic-beta-cell-research-bank/). Contact by email should be directed to the Lead Nurse, Dr Bridget Knight (b.a.knight@exeter.ac.uk).

Acknowledgments
This study makes use of data generated by the DECIPHER community. A full list of centres who contributed to the generation of the data is available from http://decipher.sanger.ac.uk and via email from decipher@sanger.ac.uk. Those individuals who carried out the original analysis and collection of the DECIPHER data bear no responsibility for the further analysis or interpretation of it in this study. This study makes use of data generated by the Genotype-Tissue Expression (GTEx) Project, which was supported by the Common Fund of the Office of the Director of the National Institutes of Health, and by NCI, NHGRI, NHLBI, NIDA, NIMH, and NINDS. The data used for the analyses described in this manuscript were obtained from the GTEx Portal on 04/25/19.
3. Figure 1 illustration of deletion seems not consistent with the deletion data shown in table 1. Either one should have error.

1.
Did author test those genes associated with hypoglycemia to rule out pathogenic variants on those genes for these cases? 2.
The NGS data for the sequencing didn't achieve enough depth and coverage to rule out the third possible mechanism mentioned by author: unmasking a recessive mutation. I understand authors already discussed a little bit about his, but a more comprehensive analysis of NGS data with enough depth and coverage will strength the conclusion of this paper. However, I wouldn't say this is necessary to index. 3.

If applicable, is the statistical analysis and its interpretation appropriate? Yes
Are all the source data underlying the results available to ensure full reproducibility?
No source data required

Are the conclusions drawn adequately supported by the results? Yes
Competing Interests: No competing interests were disclosed. In the manuscript "Refinement of the critical genomic region for hypoglycemia in the Chromosome 9p deletion syndrome", authors studied 12 patients with chromosome 9p deletion syndrome who had a history of neonatal hypoglycemia. Study of such cases might be helpful to understand the disease mechanism.
The study was well designed and the manuscript was well structured and written. I will recommend to index the MS after comments below are addressed.
1. Figure 1 illustration of deletion seems not consistent with the deletion data shown in table 1. Either one should have error.
We thank the reviewer for their appraisal of our manuscript. We are grateful for highlighting the discrepancies between table 1 and figure 1. A new version of figure 1 has now been uploaded which includes patient 3 (omitted by mistake in the previous version).
The tables and figure have now bee double checked.
2. Did author test those genes associated with hypoglycemia to rule out pathogenic variants on those genes for these cases?
We apologise that this was unclear. We had excluded all known causes of congenital hyperinsulinism in 9 of the 12 patients. In the three remaining patients samples were unavailable to targeted next-generation sequencing. We have updated the methods to reflect this point.
The paper is well written and provides novel information. However, there are several issues that require attention: There are discrepancies between Fig. 1 and Table 1:a) Patient 3 is reported to have a 17.6 Mb deletion on table 1 (larger than the deletion of patient 2) but according to figure 1, the deletion is smaller than patient 2.b) Deletion size is similar between patients 6-7-8-9 and yet patient 9 is shown in fig-1 to have a larger deletion. Patient 4 and 5 deletions should be similar in size according to the table. 1.
In Fig. 1: patient 12 is not shown. Why there is a single base for 9/10? 2. 3.
Indicate on the table that the glucose levels were at the time of onset of the hypoglycemia. 4.
Please add whether each patient was AGA, SGA or LGA according to their birth weight and gestational age.

5.
Dose of diazoxide is not included in all cases. 6.
Please change age at diagnosis to 'onset' to make clear it refers to the hypoglycemia. 7. Fig. 1 should include the size of the deletions and an Ensembl/USC snapshot of the genes included in the minimally deleted interval.

8.
Is any of the genes in the interval imprinted? Imprinting would be a fourth mechanism that the authors can include in the discussion. Moreover, there are defects of imprinted genes known to cause hyperinsulinism (e.g., 11p15 region) and neonatal diabetes.

9.
Discuss whether haploinsufficiency for the genes in the interval are responsible for any known syndromes and whether these syndromes have been associated to hypoglycemia.

10.
Can the authors estimate the frequency of the hypoglycemia among all reported cases of 9p deletions? This estimate would help understanding whether mutations on the undeleted allele might be implicated. Given the apparently higher frequency of hypoglycemia in 9p deletion cases, one would predict that the underlying mechanism for hypoglycemia it cannot be due to unmasking of a recessive allele.

11.
Add DECIPHER ID for the case included on fig. 1. 12.

Are sufficient details of methods and analysis provided to allow replication by others?
The table has been updated as requested.
5. Please add whether each patient was AGA, SGA or LGA according to their birth weight and gestational age.
The birth weight standard deviation score for each patient has been added to table 1. We would prefer to add SDS scores than AGA/SGA/LDA.

Dose of diazoxide is not included in all cases.
We had included all of the information that we have available to us regarding diazoxide dose. When the dose is unavailable we had noted this in the manuscript.

Please change age at diagnosis to 'onset' to make clear it refers to the hypoglycemia.
The table has been updated as requested.

should include the size of the deletions and an Ensembl/USC snapshot of the genes included in the minimally deleted interval.
We have added the size of the minimal deleted region to the figure as requested. The size of each of the individual deletions is recorded in table 1. We felt that annotating them to the figure reduced clarity. Due to the minimal deleted region containing 38 protein coding genes (listed in table 2) we also felt that these would be difficult to include in figure 1 in a way that could be clearly read. We have not updated the legend to help highlight that this information is available within the tables. 9. Is any of the genes in the interval imprinted? Imprinting would be a fourth mechanism that the authors can include in the discussion. Moreover, there are defects of imprinted genes known to cause hyperinsulinism (e.g., 11p15 region) and neonatal diabetes.
We agree that disruption of imprinted genes is a possible mechanism of disease in individuals with large CNVs. We had checked the genes in the minimal deleted region but none are known to be imprinted (Baran et al The landscape of genomic imprinting across diverse adult human tissues. Genome Res. 2015 Jul;25 (7):927-36). We have now added a sentence to the discussion to reflect this point.
10. Discuss whether haploinsufficiency for the genes in the interval are responsible for any known syndromes and whether these syndromes have been associated to hypoglycemia.
Further information to highlight which of the genes within the deleted region are reported to cause disease when disrupted have now been included in table 2. With the exception of SLC1A1 none of these genes have been associated with hypoglycaemia. The OMIM record for SLC1A1 list fasting hypoglycaemia as a feature however this information is taken from a case report published in 1974 (Teijema et al: Dicarboxylic aminoaciduria: An inborn error of glutamate and aspartate transport with metabolic implications, in combination with a hyperprolinemia, Metabolism 23: 115,1974) prior to discovery of the SLC1A1 gene (Bailey et al Loss-of-function mutations in the glutamate transporter SLC1A1 cause human dicarboxylic aminoaciduria. J. Clin. Invest. 121: 446-453, 2011).
11. Can the authors estimate the frequency of the hypoglycemia among all reported cases of 9p deletions? This estimate would help understanding whether mutations on the undeleted allele might be implicated. Given the apparently higher frequency of hypoglycemia in 9p deletion cases, one would predict that the underlying mechanism for hypoglycemia it cannot be due to unmasking of a recessive allele.
We do not believe that it would be accurate to offer a prediction on the frequency of hypoglycaemia in the 9p deletion syndrome based on published cases as this does not provide an unbiased cross-sectional cohort. We agree with the reviewer that unmasking of a recessive allele is unlikely, we hope we have addressed this point adequately in the discussion. fig. 1. The ID for the Decipher case has now been included on figure 1.