Rare copy number variants are a common cause of short stature

Abstract

Human growth has an estimated heritability of about 80%-90%. Nevertheless, the underlying cause of shortness of stature remains unknown in the majority of individuals. Genome-wide association studies (GWAS) showed that both common single nucleotide polymorphisms and copy number variants (CNVs) contribute to height variation under a polygenic model, although explaining only a small fraction of overall genetic variability in the general population. Under the hypothesis that severe forms of growth retardation might also be caused by major gene effects, we searched for rare CNVs in 200 families, 92 sporadic and 108 familial, with idiopathic short stature compared to 820 control individuals. Although similar in number, patients had overall significantly larger CNVs (p-value<1×10(-7)). In a gene-based analysis of all non-polymorphic CNVs>50 kb for gene function, tissue expression, and murine knock-out phenotypes, we identified 10 duplications and 10 deletions ranging in size from 109 kb to 14 Mb, of which 7 were de novo (p<0.03) and 13 inherited from the likewise affected parent but absent in controls. Patients with these likely disease causing 20 CNVs were smaller than the remaining group (p<0.01). Eleven (55%) of these CNVs either overlapped with known microaberration syndromes associated with short stature or contained GWAS loci for height. Haploinsufficiency (HI) score and further expression profiling suggested dosage sensitivity of major growth-related genes at these loci. Overall 10% of patients carried a disease-causing CNV indicating that, like in neurodevelopmental disorders, rare CNVs are a frequent cause of severe growth retardation.

Figure 1. CNV discovery and characterization.

Molecular karyotyping was performed for 200 patients with short stature after thorough clinical evaluation and for 820 healthy control samples. Exclusion of common variants using the control samples and scoring CNVs above 50 kb resulted in an approx. reduction of 80% of the CNVs identified in the patient samples. 60.5% of these CNVs affect reference sequence gene regions. Functional characterization includes segregation analysis using parental arrays and/or MLPA as well as gene and CNV based evaluation.

Figure 2. Molecular karyotyping and MLPA confirmation of identified loci.

(A) Example representation of the copy number analysis of patient 1 using the Affymetrix Genotyping Console 3.0.2 software. The red bar shows the 2.2 Mb deletion region (CNSegments). (B) Graphical presentation of the deletion region including 33 candidate genes (modified from UCSC genome browser). (C) MLPA confirmation with a probe in the NFASC gene region. A relative quantity value (RQ) below 0.75 was considered as confirmation of a deletion, above 1.25 as confirmation of a duplication. Detailed data for the remaining patients is presented in the supporting information.

Figure 3. Higher incidence of CNVs with a length of above 100 kb in affected individuals.

(A) Presentation of the Odds Ratio (light blue) and –log10(p-value) (dark blue) for determination of the size threshold of the number of CNVs in patients vs. controls. The Odds Ratio and the –log10(p-value) confirms a CNV size cut-off at 99.2 kb (OR 1.26 and p-value 4.98×10⁻⁸). (B) Fraction of copy numbers segments in cases (grey) vs. control (white) quintiles (* p<0.005). Quintile borders were Q1: 68.3 kb, Q2: 99.3 kb, Q3 149.6 kb, Q4 298.1 kb, and Q5: 72,571.3 kb. The y axis presents the fraction of CNVs inside the corresponding quintile bin. Significance levels are calculated using Fisher's exact test. The figure shows a shift towards segments above 100 kb in patients (** p-value 1.188×10⁻⁷).

Figure 4. Height distribution of the study group.

Height distribution (SDS) of all 200 patients (blue) compared to the 20 patients with identified CNVs (green). The mean SD score for height was −3.34. Patients with identified CNVs showed a significant difference in the SDS distribution (p-value 0.03).

Figure 5. Genome-wide significant association of GWAS loci for height distribution in the 3 CNVs.

The values of r² are based on the CEU 1000genomes Nov 2010 samples. The blue line and right-hand y axis represent recombination rates. The SNPs with the min p values are highlighted as purple diamond. The figures were created using LocusZoom (http://csg.sph.umich.edu/locuszoom/).