Introduction

Growth retardation, a common condition leading to reduced height, is defined as the deviation of an individual's height of more than two SDs below the mean in the population or the estimated familial target height.1 The diagnosis is based on extent and type of growth retardation and other clinical signs and disorders.1, 2 Conditions with growth retardation include inborn errors of development, which also pose a risk of various additional health issues like cancer, stroke, and cardiac defects.3, 4, 5, 6 Along with these coexisting conditions, growth retardation might constitute a substantial emotional and clinical burden for affected individuals.

Short stature can be caused by nongenetic factors, such as nutrition, chronic systemic disorders, and emotional or psychosocial deprivation.7 Most forms of short stature, however, are based on genetic causes. Turner syndrome, SHOX defects, mutations in genes affecting the growth hormone signaling pathway, or rare skeletal dysplasias are well-known causes.1, 8, 9, 10 Nevertheless, the cause remains elusive in about 60–80% of patients, preventing early treatment of growth retardation and coexisting conditions as well as adequate genetic counseling.11, 12

Human height is a polygenic trait with a heritability of about 80%.13 Several genome-wide association studies (GWAS) have identified some 700 common variants explaining 20% of height variation in the normal population and recently an additional 1.7% have been shown to be caused by rare and low-frequency coding variants.14, 15 These studies probably explain some cases of short stature whereas recent studies suggest rare monogenic variants as the more common underlying cause.16, 17, 18, 19 In absence of a specific clinical phenotype, unbiased genome-wide approaches are necessary to uncover such rare variants. We therefore investigated 565 patients with unexplained growth retardation by exhaustive clinical characterization followed by targeted genetic testing and exome sequencing in a representative subset of 200 patients.

Materials and methdos

Patients

We enrolled 565 consecutive patients and their families referred by local medical specialists for evaluation of growth retardation/short stature (Table 1, Supplementary Figure S1 online). Of these, 551 patients were of European, 13 of Asian, and 1 of Arab descent. At time of enrollment, 83% of the patients were younger than 18 years. 81% presented with a height of 2 SDs below the age-related mean, whereas the remaining 19% were 2 SDs below the estimated target family height. Overall, 20% showed mild learning disabilities and 21% microcephaly. 30% underwent bone age evaluation and of those 84% had either delayed or accelerated bone ages. All 565 patients underwent extensive prior endocrinological and diagnostic workup to exclude defects of the growth hormone pathway and organic causes of their growth deficit. All procedures were in accordance with the ethical standards of the FAU Erlangen-Nürnberg and the Helsinki Declaration. Detailed recruitment information is provided in the Methods section in the Supplementary Data.

Table 1 Clinical characteristics of included patients with idiopathic short stature
Full size table

Systematic phenotyping and targeted testing

On enrollment, all 565 patients received extensive genetic evaluation including syndromic and radiographic assessment by a clinical geneticist according to a standardized questionnaire. All information was included in a database based on known phenotype terms, followed by assessment of published information and discussion with a review board of experts in clinical genetics and dysmorphology (systematic phenotyping, Methods section in the Supplementary Data).20 Targeted genetic testing was applied based on known disease frequencies and phenotypic characteristics.

Exome sequencing and variant assessment

After exclusion of individuals from the group of 565 patients in whom our genetic targeted clinical and diagnostic approach had led to the identification of an underlying genetic cause of growth retardation, we selected a representative group of 200 patients (Figure 1, Table 1 and Supplementary Figure S1). These 200 patients showed no statistically significant difference from the remaining group of patients regarding age, height distribution, stature type, development, bone age, and sex (Table 1 and Supplementary Figure S1; selection and clinical description of the 200 individuals for exome analysis is provided in the Methods section in the Supplementary Data).

Figure 1
figure 1

Flowchart gene discovery approach. We built a study group of 565 individuals with short stature or growth retardation (Table 1). Systematic phenotyping and targeted diagnostic testing of common and recognizable causes in all 565 patients led to a diagnostic yield of 13.6% (Supplementary Table 1). Detailed information about the systematic phenotyping is provided in the Methods section in the Supplementary Data. For 200 representative individuals where no underlying cause could be determined (Table 1), we performed whole-exome sequencing, which increased the diagnostic yield by 17% up to 33% by the identification of mutations in known short stature–associated genes (Table 2, Supplementary Tables 3–4). Genotype–phenotype re-evaluation confirmed that these patients present with only part of the characteristic symptomatology. The additional diagnostic yield already had direct effects in preventive measures, symptomatic and targeted treatment in 15.5% of the 200 exome-sequenced patients (Table 4 and Supplementary Table S6).

Full size image

We performed whole-exome sequencing in this group of 200 patients—100 patients and both parents (trio analysis) and 100 patients (affected-only analysis)—after enrichment by SureSelect targeted capturing (Figure 1, Supplementary Figures S1–S2). Exomes were analyzed with our custom NGS Variant Analyzer tool, which involves the semiautomatic selection and data quality inspection of variants followed by the interpretation in relation to the reported phenotypic spectrum. The veracity and segregation in the families of selected variants were confirmed using Sanger sequencing.

All 200 patients were analyzed with respect to 1,000 known growth related genes derived from OMIM and MedGen databases (Supplementary Table S2). We considered causal all variants in known short stature–associated disease genes when the variant was predicted pathogenic or likely pathogenic referring to American College of Medical Genetics and Genomics (ACMG) criteria21 and segregating with the phenotype in the family (Supplementary Data). All results were followed by a genotype–phenotype reevaluation. Detailed information about exome sequencing and interpretation of variants in known short-stature genes is provided in the Methods section in the Supplementary Data.

Data availability

Data for various analyses have been mentioned throughout the text and derived data supporting the findings of this study are available from the corresponding author upon request. Any other data associated with this study are available in the Supplementary Data. Additional methods are described in the Supplementary Data.

Results

Diagnostic yield of common and recognizable phenotypes in 565 individuals

We could establish a diagnosis after systematic phenotyping in 13.6% of patients including disorders due to variants in 10 genes, imprinting defects, large chromosomal aberrations, and 24 distinct copy-number variants (Supplementary Table 1). Copy-number variants ranging in size from 0.1 to 14.2 Mb were the most common cause of nonspecific syndromic short stature, identified in 6.9% of the patients. A syndromic form of short stature was suspected in 181 patients and confirmed in 26 of them (4.6% of the 565 patients), mostly SHOX-related short stature and Silver–Russell syndrome. Turner syndrome or Turner syndrome variants were found in 1.6% of the patients and mutations in genes associated with skeletal dysplasia in 0.5%. No diagnosis could be established in 491 (86.4%) of the patients.

Diagnostic yield of exome sequencing in 200 individuals

To confirm the power of family-based exome sequencing in providing a diagnosis in individuals with growth retardation we selected a representative group of 200 families of the remaining patients with unexplained growth retardation (Figure 1, Table 1, and Supplementary Figure 1). We considered genes associated with growth phenotypes according to OMIM and MedGen databases and in these confirmed variants with regard to segregation and the specific patient’s phenotype. Thereby, we identified 38 variants affecting 26 genes in 33 of 200 exome patients (16.5% of the exome group; Table 2, Supplementary Figures S3–S27 and Supplementary Tables S2–S4). Of the 38 variants, 27 (71%) were missense, 5 (13%) frameshift, 4 (11%) nonsense, and 2 (5%) canonical splice-site variants. Twenty-nine of the 38 variants were not reported in the Exome Aggregation Consortium database, and nine were rare, with a maximum frequency of 2 × 10−4. Following the ACMG guidelines,21 16 of the identified 38 variants (42%) were classified as likely pathogenic and the remaining 22 (58%) were defined as pathogenic variants (Materials and Methods section, Table 2, and Supplementary Table S4). The median height of these 33 patients was 2.7 SDs below the average height in the population, 21% showed microcephaly, and 41% were classified as syndromic (Supplementary Figure S1 and Supplementary Table S3). Modes of inheritance were autosomal dominant (65%), autosomal recessive (19%), and X-linked recessive (15%) (Table 2, Supplementary Tables S3–S4). Pathway analyses of the affected proteins revealed that 58% of them are involved in the main functional categories of cartilage formation, chromatin modification, and Ras-MAPK signaling (Table 3).

Table 2 Genetic diagnosis obtained by exome sequencing in 200 individuals with idiopathic short stature
Full size table
Table 3 Diagnostic yield of exome sequencing in 200 patients
Full size table

We further wanted to know whether the selection of specific phenotypes or a combination of phenotypic features might reveal a higher diagnostic yield. A support vector machine approach resulted in 85.5% probability that it is not currently possible to accurately predict the presence of a pathogenic mutation based on one or a combination of clinical subgroups (height, occipitofrontal circumference, intellectual disability, syndromic versus isolated phenotype, prenatal growth retardation, and accelerated/decelerated bone age) (Methods section in the Supplementary Data).

Overall, considering only genes previously associated with growth retardation, the diagnostic yield of 13.6% achieved with systematic phenotyping and targeted testing was raised to 33% by additional exome sequencing (Figure 1). The most commonly mutated known short stature–associated genes identified were ACAN (2.5%), NPR2 (1.5%), KDM6A (1%), and COL2A1 (1%) (Table 2 and Supplementary Tables S3–S4).

Expansion of the phenotypic spectrum by exome sequencing

Some of the mutations identified by exome sequencing are located in genes known to be associated with syndromic intellectual disability or skeletal dysplasia (Table 2 and Supplementary Tables S3–S5). One patient with a novel pathogenic KRAS missense mutation presented with proportionate short stature and learning disability illustrating the mild end of the Noonan syndrome spectrum disorders.22 Two patients with KDM6A mutations demonstrated the phenotypic variability of Kabuki syndrome.23, 24 Neither showed any signs of developmental delay but they presented with part of the characteristic facial gestalt. Short stature is also an essential symptom of many skeletal dysplasias.9 We found mutations in FGFR3, COL2A1, and SLC26A2 in four patients presenting with no obvious specific skeletal involvement at initial clinical evaluation but consistent with the mild end of the spectrum of these entities upon re-evaluation. Finally, we identified a previously unreported likely pathogenic hemizygous missense mutation in the calcium/calmodulin-dependent serine protein kinase gene (CASK) in a male patient with mild short stature and microcephaly but no intellectual disability. Loss-of-function CASK mutations underlie several forms of X-linked intellectual disability syndromes and patients are reported to present with short stature.25

Carriers of recessive skeletal dysplasia mutations present with idiopathic short stature

In 3.5% of the 200 exome patients we identified heterozygous mutations in two genes (ACAN, NPR2) previously reported to cause the autosomal-recessive skeletal dysplasias spondyloepimetaphyseal dysplasia and acromesomelic dysplasia, respectively (Table 2, Supplementary Tables S3–S5).9, 26 Heterozygous carriers consistently show idiopathic short stature without dysmorphic findings.26, 27, 28, 29, 30 We found heterozygous mutations in ACAN in five and NPR2 in three patients who were not previously suspected to be carriers. The height of these patients ranged between −2.0 and −4.7 SDs (Supplementary Table S3). One patient carried both a variant in NPR2 and a variant in ACAN. This could indicate a blended phenotype as previously reported.31

Additional relevance of exome sequencing results for clinical management

Besides the increased diagnostic yield by exome sequencing in the 200 patients with idiopathic short stature, results with possible impact on treatment or additional preventive measurements occurred in 31 families (15.5% of 200 exome individuals, Table 4 and Supplementary Table S6). This led to preventive measures for osteoarthritis (ACAN, CUL7, MATN3) and neoplasias (NF1, PTPN11, TRIM37), as well as orthopedic support and regular developmental evaluation in affected individuals from 23 families (11.5%). Symptomatic treatment or screening for associated malformations (KAT6B, KRAS, MAP2K1, PTPN11), hearing loss (COL2A1, FLNB), or the risk for chronic kidney disease (IFT140, CLCN5) was recommended in nine families (4.5%). A treatment with recombinant growth hormone or IGF1 has been shown to be beneficial especially in the individuals with defects of the growth hormone pathway.32, 33, 34 We identified mutations affecting genes of this pathway (GHSR, IGF1R) allowing specific treatment in three families (1.5%). Interestingly, these have been missed by endocrine testing, but were confirmed by growth hormone stimulation tests and therefore considered at least to contribute in part to the patient’s growth phenotype. Also immediately relevant was the identification of a PDE3A mutation in one girl. Activating mutations in PDE3A led to hypertension and brachydactyly syndrome (MIM 123805) with the development of life-threatening hypertension later in life.35 A targeted treatment with phosphodiesterase inhibitors is indicated due to its potentially significant impact on life expectancy.36 Finally, knowledge about the inheritance pattern will aid evaluation of the recurrence risk and improve genetic counseling of families.

Table 4 Intervention in affected 200 exome individuals with mutations in known short-stature genes
Full size table

Discussion

Even though the heritability of short stature is considered high, the large number of genes implicated and the nonspecific clinical phenotype have led to poor yield of diagnostic genetic testing.13, 16, 18, 19, 37, 38, 39 To demonstrate the diagnostic outcome we enrolled 565 carefully characterized patients with short stature and their families and established an interdisciplinary systematic phenotyping with additional targeted gene testing (Figure 1). This led to a diagnostic yield of 13.6% including copy-number variants, chromosomal aberrations, and monogenic causes (Supplementary Table S1). We next performed unbiased exome sequencing in 200 representative patients with short stature of unknown origin (Table 1 and Supplementary Figure S1). The variant evaluation and classification in accordance with the ACMG guidlines21 led to the identification of mutations in known short stature–associated genes in 33 affected individuals (16.5%) (Table 2, Supplementary Tables S3–S5). In a previous smaller study of 14 highly selected patients a diagnostic yield of 36% was reported,16 which might be explained by the study’s strict inclusion criteria and smaller sample size. Genotype–phenotype re-evaluation confirmed that the individuals from our exome study were lacking most of the characteristic features of the entities, indicating that current descriptions are biased and that the phenotypic spectrum needs to be expanded. This is especially true for genes associated with syndromic forms of short stature and intellectual disability as well as skeletal dysplasias. There are parallels to the field of intellectual disability, also plagued by high genetic heterogeneity and unspecific clinical presentation, where both a similar mutation yield and inheritance modes were reported.40 However, in this study clinical subgrouping did not reveal any correlation between a specific phenotypic feature or combination of them and the probability to identify a mutation, thus supporting an unbiased approach such as exome analysis for all patients with idiopathic short stature.

One remarkable aspect was the detected frequency of heterozygous mutations in genes previously implicated in autosomal-recessive skeletal dysplasias (ACAN, NPR2), confirming a dosage effect of cartilage matrix proteins in growth development.26, 27, 29, 30 In our exome study, ACAN was the most commonly mutated known short stature–associated gene with a frequency of 2.5% (Supplementary Table S2), whereas the previously reported most common single-gene defect affects SHOX with a frequency of 2.4%.8

An important aspect of the clinical application of whole-exome sequencing in short stature concerns prognosis, prevention, and treatment (Table 4 and Supplementary Table S6). In 11.5% of the families, the identification of the molecular cause of the disease by exome sequencing prompted further preventive action. The girl diagnosed with a mutation in PDE3A might benefit from current targeted treatment with phosphodiesterase inhibitors to reduce her high risk of life-threatening coronary artery disease and essential hypertension after puberty.35 Also 1.5% of the affected individuals became eligible for targeted treatment for growth retardation itself. Finally, 5% of the patients could benefit from treatment or screening for associated malformations. None of these clinical applications were considered prior to genetic diagnosis by exome sequencing, but they have now been applied to the patients’ care.

In conclusion, we demonstrated that systematic phenotyping combined with targeted genetic testing and exome sequencing increases the diagnostic yield in short stature up to 33% with concomitant improvement in treatment and prevention. As height has a proposed heritability of about 80%, and as the ongoing rate of discovery in other entities suggests,40 we expect that future identification of potential candidate genes, as well as their analysis in additional patients, will increase the diagnostic yield.