Copy Number Variation in GCK in Patients With Maturity-Onset Diabetes of the Young

Abstract

Purpose

Next generation sequencing (NGS) methods to diagnose maturity-onset diabetes of the young (MODY), a monogenic autosomal dominant cause of diabetes, do not typically detect large-scale copy number variations (CNVs). New techniques may allow assessment for CNVs using output data from targeted NGS, without requiring additional sequencing. Using this technique, two kindreds of patients presenting with features of MODY were found to bear the same heterozygous large-scale deletion in GCK.

Methods

Patients suspected of having MODY but with negative targeted NGS pathogenic variant calling were reanalyzed using the CNV caller tool (VarSeq v1.4.3). Two patients were identified as having a possible heterozygous whole exon deletion affecting exon 1 of GCK. For confirmation and determination of the exact breakpoints, whole exome sequencing followed by Sanger sequencing were used. Familial samples from both affected and nonaffected first-degree relatives were then analyzed for each proband.

Results

A heterozygous whole-exon deletion spanning 4763 bp affecting the entire exon 1 of GCK was detected in two apparently unrelated patients with clinical features of MODY. This deletion showed segregation concordance across generations in affected and nonaffected family members.

Conclusions

Our findings confirm the utility of applying the CNV caller tool to screen for CNVs in GCK from NGS data. In so doing, we identified a deletion of exon 1 of GCK as likely causal for MODY. Our data indicate that incorporating CNV analysis routinely when assessing for MODY via targeted NGS may increase diagnostic yield and reduce false negative genetic testing rates.

A whole exon deletion in GCK was found in two unrelated families exhibiting a MODY phenotype using a bioinformatics approach to identify copy number variations.

Maturity-onset diabetes of the young (MODY) is a heterogeneous group of monogenic diabetes inherited in an autosomal dominant fashion; it is thought to be the underlying cause in 3% to 5% of all individuals diagnosed with diabetes (1–15), and up to 6.5% in select pediatric populations (16, 17). Suggestive features for MODY include: (i) age of diagnosis < 25 years; (ii) normal body mass index; (iii) noninsulin dependence or no episodes of diabetic ketoacidosis; (iv) a strong family history following an autosomal dominant inheritance pattern, although de novo mutations are possible (18); and (v) robust response to sulfonylurea treatment in some forms. A diagnosis of MODY can significantly alter the management and expected clinical course of diabetes (4). Consequently, suspecting and confirming a diagnosis can be important for optimal patient care. Confirmation of MODY is via genetic testing to rule out mutations in one of the 14 known causative genes (6–8, 14, 15, 19–21). Clinical prediction calculators, and other diagnostic tools, are also available to aid clinicians in selecting patients who may benefit from genetic testing (22, 23). Strategies that incorporate the use of fasting blood sugars and response to an oral glucose tolerance test to predict a potential MODY diagnosis may also be effective as a screening tool (24). GCK-MODY (alias MODY2) is caused by mutations in the GCK gene, encoding glucokinase, a hexokinase enzyme that catalyzes the phosphorylation of glucose to glucose-6-phosphate, the first step in glycolysis (4, 5, 19, 20, 25–27). Activity of this enzyme is induced by the presence of glucose and it functions in the pancreas to mediate glucose-stimulated insulin secretion and in the liver to catalyze glucose uptake and conversion to glycogen (4, 5, 19, 20, 25–27). Heterozygous mutations in this gene have been linked to a mild, nonprogressive form of diabetes that is usually asymptomatic with mild elevations in fasting blood sugar (5.49 to 7.99 mmol/L) and mildly increased postprandial glucose excursions (< 3 mmol/L). GCK-MODY is characterized by a minor increase in the threshold for release of insulin in response to serum glucose; off-treatment hemoglobin A1c rarely exceeds 7.5% (6). Mutations in this gene are a relatively common form of MODY, with an overall prevalence of ∼1 in 1000 individuals (4–6, 19, 20, 25–27).

Traditional sequencing methods, such as Sanger sequencing and targeted next generation sequencing (NGS) panels for MODY generally identify potentially causative small nucleotide polymorphisms or small insertions/deletions, frameshift mutations, or null mutations (6). In contrast, these methods are not optimized to detect large-scale copy number variations (CNVs), which are genetic changes that involve either deletions or duplications of regions of DNA of at least 50 nucleotides in size, and which can often encompass part of or a whole exon, gene, or several genes (28).

Because the affected sequence of duplicated or deleted genomic DNA in a CNV appears qualitatively normal with alterations instead due to quantitative changes affecting the dosage of genetic material, CNVs can be difficult to detect and confirm with traditional sequencing methods that are optimized to detect small qualitative changes in the genetic code. Without a robust quantitative analytical tool, even when there appears to be an increase or decrease in the amount of genetic material replicated in certain DNA sections, it is impossible to distinguish a true deletion or duplication from the natural variability in chemical amplification of DNA that is used in most sequencing platforms (Fig. 1). Secondary nonsequencing-based dedicated targeted DNA analytical methods using specific probes to assess for CNVs are costly and of uncertain value in MODY, although a few gene deletions have been detected this way (20, 29).

Figure 1.

Using DOC ratios to identify suspected deletions or duplications. Following PCR amplification of a region of interest, amplified sequences from starting material containing either a deletion or duplication appear qualitatively normal. There is a natural degree of variation in the number of copies of each section of DNA that is amplified. It can therefore be difficult to ascertain if the number of copies amplified by PCR represents a normal complement state or a duplication or deletion state, making determination of CNV challenging. A statistically significant difference in the ratio of DOC (i.e., the number of copies of amplified DNA containing the region of interest) in the affected individual when compared with a reference population sequenced using the same panel and conditions is suggestive of a possible CNV. A ratio of ∼0.5 would be suggestive of a deletion. A ratio of 1.5 or higher would be suggestive of duplication.

Recently, new bioinformatic techniques have been developed to provide the robustness needed to assess for CNVs using NGS output data, without requiring additional sequencing. These methods take advantage of the fact that current NGS protocols generate large numbers of short partially overlapping DNA fragments that are assembled computationally to seamlessly reflect the genomic sequence of the source material (28). In addition, the total number of these synthetically generated DNA fragments reflects the amount of starting material in the genome. This enabled the development of new algorithms that, through tallying the numbers of chemically generated DNA fragments, impute deviations of the amount of starting material from the normal diploid two copies (i.e., maternal and paternal) for any particular chromosomal region. This approach has successfully been applied to detect CNVs using NGS data for several genes causing dyslipidemias (30–33). Here we describe application of this technique to identify a heterozygous large-scale deletion in two unrelated individuals with GCK-MODY that was not identified using traditional GCK sequencing.

Materials and Methods

Subjects

Clinical information and DNA from whole blood samples was collected from individuals who were clinically suspected to have MODY and their family members. No specific inclusion or exclusion criteria were applied prior to genetic testing. We focused on 57 individuals in whom our initial targeted NGS screening failed to detect likely or definitely causative DNA variants in MODY genes (3). Patients and family members provided informed consent for genetic testing and analysis and under a protocol approved by the University of Western Ontario Ethics Review Board (#07920E).

Targeted NGS

All individuals were assessed for mutations in known MODY-associated genes using the targeted NGS panel and bioinformatics pipeline known as LipidSeq (34), designed to test for clinically relevant mutations in 73 specific genes associated with metabolic disorders, including those associated with MODY. Targeted NGS was performed using standard operating procedures of the London Regional Genomics Centre (www.lrgc.ca). Sequencing reactions were designed to include all coding regions, as well as the flanking ∼150 bp of intronic DNA for each exon and ∼500 bp at the promoter and three untranslated regions. The average depth of coverage (DOC) generated using this method is ∼300-fold for each base, meaning that there are ∼300 partially overlapping, nonidentical small generated DNA fragments covering all coding regions of MODY-related genes; these can be quantified using bioinformatic analysis for CNVs.

Bioinformatic analysis

Following library preparation and enrichment, .FASTQ files of sequence data were generated using the MiSeq personal sequencer platform (Illumina, San Diego, CA) and sequence alignments, variant calling (.VCF files), and target region coverage statistics (.BAM files) were generated using a custom automated workflow in CLC Genomics Workbench (CLC Bio, Aarhus, Denmark). Using this method, a variant is considered causative if it had been previously reported as causative in the Human Genome Mutation Database or if present in < 1% of general population and predicted to be pathologic using in-silico prediction models. Neither of the individuals we report here were found to have causative mutations for MODY using this method.

CNV detection using NGS data

LipidSeq data (in the form of .VCF and .BAM output files) were analyzed using the CNV caller function in VarSeq v1.4.3 (Golden Helix, Bozeman, MT). A .BED file defining the target region and probes used in the NGS panel is also required. The algorithm uses a ratio of DOC in each region compared with a healthy reference population of 73 samples that were subjected to the same NGS sequencing panel as the sample being analyzed to identify potential CNVs (Fig. 1). Ratios are calculated using sample DOC divided by mean reference sample coverage. Z scores measure the number of SDs, and the samples’ coverage is from the mean reference sample coverage. A ratio is considered suspicious for deletion if it is less than 0.75, and suspicious for duplication if the ratio is greater than 1.25 with a Z score of greater than 5 (or −5).

Confirmation of CNV state using whole exome sequencing

We confirmed each CNV call by next performing whole exome sequencing (WES) on samples of interest, using standard operating methods and procedures at London Regional Genomics Centre. The average DOC for WES in our facility is ∼100. We determined the extent of the deletions using the same CNV caller analysis tool and procedure applied to WES-generated data. The ratio used for CNV suspicion is the same as for the LipidSeq targeted NGS data; however, since the reference population for this analysis was derived using only 15 WES samples, the Z score significance threshold was decreased to 3 (or −3). Significance threshold for consideration of a CNV was lowered to include Z-scores between 2 and 3 (−2 to −3), if the P value was strongly significant.

Breakpoint analysis

The CNV caller analysis results allow for strong suspicion of a CNV deletion using probe-level data. These methods allow determination of the approximate size of the deletion and the approximate breakpoints on either end of the deletion. To determine the exact breakpoint, Sanger PCR-based probe analysis is required. Probes were designed to target either side of the suspected deletion in an attempt to identify the exact start and stop position of the deletion (reagents and conditions available upon request). The span of the wild-type strand is too great to amplify under the same conditions as the mutant allele that contains a deletion between the primers; therefore, if amplification occurs in the candidate sequence but not in the control, it suggests the primers hybridize to DNA that is relatively close to each side of the breakpoint (Fig. 2). PCR amplification and sequencing of the abnormal amplified fragment using the identified primer pair is then completed (reagents and conditions available upon request). Sequence alignment between the amplification product and the known reference sequence then allows for exact identification of the start and stop points for the deletion.

Figure 2.

PCR amplification of DNA subjected to primers designed on either side of suspected breakpoint in both probands and family members. The top section shows the GCK gene and location of primers used to confirm and sequence across the breakpoint. The normal sequence distance between primer pair F4 and R4 is 10,655 bp; however, PCR amplification in probands using primer pair F4 and R4 generated a product size of 5893 bp, suggesting a 4763-bp deletion. In the pedigree charts for proband 1 and 2, affected status is indicated by solid color and probands are indicated by arrows as shown. Gels show PCR amplified fragments aligned beneath each individual. The top gel shows amplification products generated using primers P1 located on the proximal side of the suspected breakpoint, and P2, within the deleted fragment. The middle gel contains amplification products generated using primers P4, located on the distal side of the suspected breakpoint, and P3 within the deleted fragment. All subjects demonstrated amplification for both the proximal (636 bp) and distal (288 bp) primer pairs. The bottom gel shows amplification products generated using primers P1 and P4 (581 bp). In individuals without the deletion, the span between these two primers would be too great to amplify under standard conditions; therefore, if amplification occurs, it confirms the presence of a large deletion between the two primers. Amplification between P1 and P4 is seen in both probands and all affected family members, but not in unaffected family members, confirming the presence of a deletion in these individuals. The normal amplification products generated in individuals carrying the deletion with both the proximal (P1 and P2) and distal (P3 and P4) primer pairs confirms these individuals carry one normal copy of the gene and are heterozygous for the deletion. Thus, the mother and both children in family 1, and the mother and proband in family 2 are all heterozygous for the normal and deleted GCK alleles. Primer (P) design: F4 (GTTCAGCCTCAGGTGTAGAAGCAG); R4 (AGGAACAGGACAGGAGTATACGTGG); P1 (TGAGTCAGTGGCTCCTGGAAAGG); P2 (CTGTCATTCCTCAGCTGAGCCAG); P3 (CTAGGGCTGTAAACTCTCCAGAG); P4 (AGGCTGAAGCTTCCTGAGCAGG). All PCR reactions used an annealing temperature of 60°C and a 5% DMSO solution.

Family member analysis

Following breakpoint determination in the two probands, the designed primer pair was used to conduct targeted assessment of available DNA for first-degree relatives of each proband to detect either the presence or absence of the deletion.

Results

Likely CNVs were seen in GCK in two male individuals with clinically suspected MODY using CNV caller analysis of LipidSeq output data. These were confirmed by WES-generated data in one of the two individuals (Table 1).

Table 1.

Suspected MODY Patients With Confirmed GCK CNVs

MODY Subtype	Gene	Affected Exons	Probes Affected Chr:Position Range	CNV State	VarSeq Average Ratio	VarSeq Average Z Score	WES Average Ratio	WES Average Z Score	Deletion Size	Breakpoint
MODY 2	GCK	5UTR -altExon 1	7:44228257–44229272	Del, het	0.57531	−7.69375	0.58120	−2.94078	4763 bp	chr7: 44224750–44229512
MODY 2	GCK	5UTR -altExon 1	7:44228257–44229272	Del, het	0.57517	−7.02283	N/A	N/A	4763 bp	chr7: 44224750–44229512

The two patients with CNVs affecting GCK (MODY2) were confirmed to have deletions at precisely the same breakpoint, suggesting this may have been a single ancestral event. As far as we are aware, there was no direct familial relationship between the two individuals. The GCK deletions spanned the 5′UTR – altExon 1 in both individuals, affecting probes 7:44228257-44229272. In the first individual, the CNV average ratio for this deletion was calculated at 0.57531 with Z score of −7.69375 for the LipidSeq-generated data and an average ratio of 0.58120 with a Z score of −2.94078 for the WES data. In the second individual, the CNV average ratio for this deletion was calculated at 0.57517 with Z score −7.02283 for the LipidSeq-generated data. WES was not conducted in this individual due to low volume of DNA. Breakpoint analysis confirmed the presence of a 4763-bp deletion in both individuals spanning chr7: 44224750 to 44229512 (Fig. 3). Because the deletion encompasses the promoter region, initiation codon and all of exon 1, neither transcription of RNA nor translation of mature enzyme are predicted from this allele.

Figure 3.

Determination of deletion breakpoint for GCK. (A) Screen capture of targeted NGS-generated data from proband 1 processed by the CNV caller tool identifies a potential deletion (shown by the star), as indicated by a significant drop in DOC ratio to less than 0.75 and a Z score less than −5, when compared with DOC in a reference population. The bottom section maps the extent of involvement of GCK coding sequence, which is oriented 5′ to 3′ right-to-left, indicating that exon 1 is involved; (B) WES-generated data from the same sample using the same tool to confirm and also determine the approximate extent of the deletion involving exon 1 and the 5′-flanking region of GCK but not neighboring genes; (C) Sanger sequencing electropherogram tracings showing normal DNA sequences in the vicinity of the 3′ (left side; letter codes shaded blue) and 5′ (right side; letter codes shaded yellow) breakpoints of the deletion. Internal sequence that is missing in the deleted allele is shaded gray; (D) Sanger sequencing electropherogram tracings mutated DNA sequence in which the 3′ (shaded blue) and 5′ (shaded yellow) regions flanking the deletion breakpoint, with absence of the intervening 4763 nucleotides.

Family analysis revealed the presence of the identical deletion in an affected mother of proband 1, whereas no deletion was detected in the unaffected father. Similarly, analysis from the family of proband 2 revealed the presence of the deletion in the affected mother and sister, but not in the unaffected father (Fig. 2).

The clinical presentation in the two GCK-MODY probands was similar (Table 2). Both were of French Canadian descent and each presented in the ninth year of life. Both had fasting hyperglycemia, mild elevations in glycated hemoglobin and negative auto-antibodies; proband one was being treated with metformin and proband 2 was following a diabetic diet. Both had an affected mother and genotypically and phenotypically normal father. Both individuals presented with polyuria, which is atypical for GCK-MODY. The sister of proband 1 carried the same mutation but presented with a milder phenotype.

Table 2.

Clinical Features of GCK-MODY Probands

	Proband 1	Sister of Proband 1	Proband 2
Gender	Male	Female	Male
Age at diagnosis	8.5 y	11.8 y	8.9 y
Duration of symptoms prediagnosis	Longstanding polydipsia	Asymptomatic	9–12 mo of polyuria, polydipsia
Presentation	Fasting hyperglycemia and abnormal OGTT	Elevated fasting BG (6–7 mmol/L)	Polyuria, polydipsia
	Fasting BG 7 mmol/L
Family history	Mother: impaired glucose tolerance	Mother: impaired glucose tolerance	Mother: impaired glucose tolerance
	Paternal grandmother: T2DM	Paternal grandmother: T2DM	Maternal aunts: DM early adulthood
Ethnic background	French Canadian	French Canadian	French Canadian
Weight at diagnosis	Weight 95th percentile	Weight 95th percentile	23 kg (25th percentile)
	Height 75th percentile	Height 50th percentile
Anti-GAD antibodies	<1	N/A	<1
Evolution	Fasting BG 6–9 mmol/L	Fasting BG 5–7 mmol/L	No decompensation
OGTT (at 120 min)	BG 11.1 mmol/L	BG 10.0 mmol/L	BG 7.4 mmol/L
			Peak insulin 149 pmol/L
Treatment	Metformin	Metformin	Diet only
Autoimmunity	No thyroid antibodies	No thyroid antibodies	No thyroid antibodies
HbA1c	6.4%–6.9%	6.0%–6.2%	5.9–7.0%
Family	M: positive for mutation	M: positive for mutation	M: positive for mutation
	F: negative for mutation	F: negative for mutation	F: negative for mutation
	S: positive for mutation	B: positive for mutation

Abbreviations: B, Brother; BG, blood glucose; DM, diabetes; F, Father; GAD, glutamic acid decarboxylase; HbA1c, glycosylated hemoglobin; M, mother; N/A, not available; OGTT, oral glucose tolerance test; S, Sister; T2DM, type 2 diabetes.

Disclosure of the genetic results to the sending physician resulted in a significant change to her management approach, and recommendations were made to the affected probands and family members that they could safely discontinue all therapeutic interventions.

Discussion

Bioinformatic analysis of NGS data detected the identical heterozygous deletion of 4763 bp in GCK in two unrelated individuals suspected to have MODY, but in whom targeted NGS analysis did not detect a small-scale nucleotide variant or mutation. The deletion was confirmed in WES-generated sequencing data, with identical breakpoints found in both individuals, confirmed by Sanger sequencing. Pedigree analysis confirmed cosegregation of the deletion with a phenotype consistent with MODY. This deletion encompasses the promoter region, initiation codon and exon 1 of GCK, and is predicted to be deleterious and disease-causing. The findings indicate that expanding analysis of NGS data to include both quantitative and qualitative attributes of the targeted regions can increase the yield of molecular diagnosis, here by ∼4% of patients who were not previously diagnosed molecularly.

MODY remains an underrecognized cause of diabetes, especially in the United States (3, 4, 22, 35). Moreover, genetically confirmed MODY diabetes can have significant implications for the management of patients with this condition. Optimal management of MODY diabetes is highly dependent on the underlying genotype. Generally, GCK-MODY (MODY2) is benign and nonprogressive and can be managed without medication, or other therapeutic measures, including diet, with the exception of special monitoring during pregnancy (2, 3, 36). However, it is important to distinguish it from other forms of MODY and from type 1 or type 2 diabetes to optimize management decisions (4). A recent study from the United States found that 49% of patients with GCK-MODY were unnecessarily being treated with glucose-lowering agents (4). As patients with MODY often present at a young age and with a normal body mass index, they are frequently misdiagnosed as having type 1 diabetes and are started on multiple daily injections of insulin (12, 22, 35). For example, two recent large cohort studies identified MODY mutations in up to 6.5% of children with diabetes, many of whom had not previously been identified (16, 17). Reclassifying some of these patients as GCK-MODY could allow for safe discontinuation of insulin, and generally all other treatments, which can have a profound impact on quality of life as well as health care costs and burden.

Emerging evidence suggests CNVs may be a more significant contributor to MODY than previously thought. One report identified deletions causing MODY in GCK and HNF1A using a dedicated multiplex ligation-dependent probe amplification assay (20), and large-scale CNVs in HNF4A have previously been reported as significant proportion of causative mutations for MODY5 (29). Case reports of partial gene deletions in GCK have also been reported in the literature (37, 38). Our data provide further support the presence of CNVs in at least 3% to 4% of previously undiagnosed MODY patients, while demonstrating a simpler, more cost-effective approach to make this molecular diagnosis.

Clinically, there is some suggestion that this large-scale deletion may present with a slightly more severe phenotype than other forms of GCK-MODY. Both probands experienced polyuria, a symptom not generally associated with GCK-MODY, which is usually asymptomatic. Proband 1 also had higher than expected fasting blood glucose, up to 9 mmol/L. However, both affected mothers and the sister of proband 1 appeared to have only mild phenotypes, with the mother of proband 1 diagnosed with only impaired glucose tolerance and the mother of proband 2 with isolated gestational diabetes. The sister of proband 1 presented with features more classically consistent with GCK-MODY, being asymptomatic, with lower fasting blood glucose (5–7). Both families were also French Canadian, and may have shared a common ancestor. The milder phenotype in the females harboring the deletion raises the possibility that there may be gender differences associated with phenotype expression, or this difference could be related to other factors, such as variable penetrance, environmental or lifestyle influence, or simple chance.

Although CNVs have not traditionally been screened for or reported as causal mutations for MODY, they may be the underlying cause for a clinically meaningful proportion of patients with suspected MODY. In a prior study, deletions were detected in 1/29 (3.5%) of patients tested for large-scale deletions in GCK and 6.7% (4/60) of patients tested for HNF1A (20). The overall proportion of MODY cases that can be attributed to CNVs may become more apparent as CNV analysis is incorporated routinely when assessing for MODY via NGS.

Our results have some limitations. Although these large deletions would be expected to have deleterious effects, and they seem to segregate with phenotype in family members, no functional studies were conducted to directly confirm lack of in vivo activity. Also, although the CNV caller tool was validated using our specific NGS panel and reference population, this may not be universally applicable. Also, given our small sample size, generalizations about any phenotype differences with this large-scale deletion compared with causes related to pathogenic small nucleotide polymorphisms or small inserts/deletions is limited.

Conclusion

We report the validation of the CNV caller tool for the identification of CNVs in the MODY gene GCK and its application to detect a heterozygous whole-exon deletion of exon 1 and 5′-flanking sequence as likely causal for MODY in two unrelated patients. We propose that CNVs may be an important and currently underrecognized genetic cause of MODY diabetes. The addition of routine CNV testing to targeted NGS or WES output may improve the accuracy of MODY diagnostic testing.

Glossary

Abbreviations:

CNV

copy number variation

DOC

depth of coverage

MODY

maturity-onset diabetes of the young

NGS

next generation sequencing

WES

whole exome sequencing