Here we report the identification of germline hemizygous deletions of MIR17HG, encoding the miR-17~92 polycistronic miRNA cluster, in patients with microcephaly, short stature and digital abnormalities. We demonstrate that haploinsufficiency of miR-17~92 is responsible for these developmental abnormalities by showing that mice harboring targeted deletion of the miR-17~92 cluster phenocopy several key features of the patients.

Feingold syndrome (FS, MIM164280) is an autosomal dominant syndrome whose core features are microcephaly, relative short stature and digital anomalies, particularly brachymesophalangy of the second and fifth fingers and brachysyndactyly of the toes^12,13. Less penetrant defects include oesophageal/duodenal atresia (observed 30–55% of cases), heart and kidney defects and variable learning disabilities. In approximately 70% of affected families, FS is caused by germline loss-of-function mutations of the MYCN gene (MIM 164840) at 2p24.1^14,15, but the genetic lesion(s) responsible for the remaining cases have yet to be identified.

We employed high-resolution comparative genomic hybridization (CGH) arrays to perform a genome-wide analysis of 10 index patients displaying skeletal abnormalities consistent with a diagnosis of FS, but lacking any mutation at the MYCN locus (Supplementary Table 1). This led to the identification of germline hemizygous microdeletions at 13q31.3 in two cases (AO39 II3 and AO70 II1, Fig. 1 and Fig. 2A). The deletion in AO39 II3 spans 2.98 Mb and encompasses three genes: LOC144776, MIR17HG, and Glypican-5 (GPC5). The deletion identified in AO70 II1 is more informative as it spans 165 kb and encompasses only miR-17~92 and the first exon of GPC5 (Fig. 2A). qPCR and direct sequencing of MIR17HG (including the promoter region) and of the GPC5 coding sequence failed to identify non-annotated sequence variations in the remaining eight index cases (data not shown, the primers used are listed in Supp. Table 2).

By genomic qPCR, we determined that the deletions detected in the two index cases segregate with the disease in the two families (Fig. 2B), thus lending support to the hypothesis that these two microdeletions are the causative mutations in these patients.

Next, we queried the DECIPHER database, which contains array CGH data from more than 6000 individuals with a variety of disorders¹⁶ and identified an additional individual (patient id: 248412) harboring a 180kb hemizygous 13q31.3 microdeletion encompassing the entire MIR17HG locus and the first exon of GPC5 (Fig. 1 and ​and2A).2A). This deletion was further confirmed by genomic qPCR (Figure 2B). Unfortunately, due to the lack of any genetic and clinical data from the parents of this individual, it is unclear whether the deletion was inherited or de novo. Although not classified as having FS, the patient presents a combination of features compatible with a diagnosis of FS (Fig. 1 and Supplementary Table 1). The only exception was the presence of unusually hypoplastic thumbs, a trait we also observed in patient A070P (Fig. 1 and Supplementary Table 1) and one that is rarely as severe in MYCN-mutant FS patients¹⁵. Analogous digital abnormalities were previously reported in some individuals presenting large 13q deletions^17,18; however, due to the large size of the deletions, the gene(s) responsible for the developmental defects could not be identified.

To determine whether hemizygous loss of MIR17HG in humans results in a detectable reduction of miR-17~92 expression, we performed RT-qPCR on total RNA extracted from white blood cells obtained from three individuals carrying the 13q31.3 microdeletions. In these patients, expression of all six microRNAs encoded by the miR-17~92 cluster was approximately 50% relative to control individuals (Fig. 2C). This indicates that hemizygous loss of miR-17~92 results in a significant reduction in the expression of its constituent miRNAs that is not compensated by up-regulation of the remaining allele.

Together, these findings suggest that hemizygous deletion of MIR17HG and/or GPC5 is responsible for the skeletal abnormalities observed in these patients, but do not define the relative contribution of either gene. To address this issue, we first queried the Database of Genomic Variants (DGV)¹⁹, which compiles structural variations detected in the genomes of healthy individuals (http://projects.tcag.ca/variation/?source=hg19). We identified two Yoruba control individuals²⁰ heterozygous for a deletion encompassing exon 4 of GPC5 (Supplementary Fig. 2), which is predicted to produce a loss-of-function allele, as exon 3 and exon 5 are not in frame. Moreover, the 1000 genomes browser (http://browser.1000genomes.org/index.html) reports a single nucleotide insertion in the coding sequence of GPC5 (rs34433071, c.1356_1357insC, p. Val454CysfsX7). Although the allelic frequency of this variant is currently unknown, the insertion is predicted to result in the loss of the 113 C-terminal amino acids of GPC5. In contrast, although a ~1.8 kbp deletion ending 1.4 kbp upstream of the first miRNA of the miR-17~92 cluster is reported in five normal individuals (Supplementary Fig. 2), no structural variants or polymorphisms directly affecting the miRNAs encoded by the cluster were identified in these databases. Collectively, these data indicate that hemizygous loss of GPC5 alone cannot account for the phenotypes observed in our patients. To determine whether this can be explained by miR-17~92 haploinsufficiency, we next examined the consequences of hemizygous deletion of this cluster in mice¹¹.

Animals harboring targeted deletions of a single miR-17~92 allele (miR-17~92^Δ/+) are viable and fertile but significantly smaller than wild type controls¹¹ (Fig. 3A), a feature also observed in both MYCN-mutant FS patients and patients harboring 13q31.3 microdeletions. Skeletal analysis of the limbs from age- and sex-matched wild-type and miR-17~92^Δ/+ adult mice revealed a striking shortening of the mesophalanx of the fifth finger in heterozygous animals (Figs. 3B and C), another feature observed in patients with 13q31.3 microdeletions and in virtually all FS patients¹⁵. Other long bones in the hands of miR-17~92^Δ/+ mice were only marginally shorter (Fig. 3B and data not shown) and syndactyly was not observed in any of the hemizygous animals. Analysis of the skull revealed shortening of the anterior-posterior axis and an overall reduction in size, which are both consistent with microcephaly (Fig. 3D). Importantly, though Gpc5 and miR-17~92 are also closely linked in the mouse genome, targeted deletion of miR-17~92 does not negatively affect the expression of Gpc5 in the forelimbs of developing mouse embryos or in mouse embryo fibroblasts (Supplementary Fig. 3 and data not shown), further demonstrating that miR-17~92, but not Gpc5 is responsible for the key features observed in miR-17~92^Δ/+ mice and patients harboring del13q31.3.

Prompted by these findings, we examined the consequences of complete loss of miR-17~92 function on skeletal development. Because homozygous deletion of the cluster leads to perinatal lethality in mice¹¹, we analyzed animals at embryonic day 18.5 (E18.5). Skeletal preparations of miR-17~92^Δ/Δ embryos revealed a severe and general delay of endochondral and membranous ossification (Fig. 4). Strikingly, the main limb defects observed in FS patients were grossly exacerbated in these animals. More specifically, we observed the complete absence of the mesophalanx of the fifth digit, the presence of a small mesophalanx of the second digit, and hypoplasia of the first digital ray (Fig. 4A and Supplementary Fig. 4). In addition, all embryos examined presented with fusion of cervical vertebrae (Fig. 4B and Supplementary Fig. 4) and microcephaly (Figs. 4C and D). Additional skeletal defects consistently observed in these embryos included dysmorphic zeugopods and fusion of the proximal carpal bones (Fig. 4A and Supplementary Fig. 4). In sum, the human genetic data and the analysis of miR-17~92^Δ/Δ and miR-17~92^Δ/+ mice demonstrate that miR-17~92 haploinsufficiency is responsible for developmental abnormalities in humans and highlight a previously unappreciated role for miR-17~92 in normal growth and skeletal development.

Based on the similarities between the skeletal defects observed in FS patients harboring MYCN mutations and in patients with hemizygous deletion of MIR17HG, it is tempting to speculate that these two genes may be components of the same developmental pathways and that miR-17~92 may be an important target of MYCN in skeletal development. Indeed, several lines of evidence indicate a close genetic and functional interaction between the MYC family of transcription factors (MYC, MYCN and MYCL) and the miR-17~92 cluster. Both MYC and MYCN can activate the transcription of miR-17~92 (refs. ^2,6,21–24, Supplementary Figure 5 and Supplementary Note) and can directly bind to the miR-17~92 promoter region in human and murine cells (refs. ^21,24,25 and Supplementary Fig. 6). Furthermore, ectopic expression of miR-17~92 cooperates with Myc in murine models of B-cell lymphoma and colorectal cancer^4,5,26. A possible cooperation between MYCN and miR-17~92 in modulating developmental processes is further supported by the remarkable similarity of the phenotypes observed in miR-17~92^Δ/Δ mice and in mice carrying hypomorphic or null Mycn alleles (Supplementary Table 3)^27–29.

One hypothesis emerging from these observations is that MYCN regulates various aspects of mammalian development via transactivation of miR-17~92. This hypothesis can be further refined by considering the phenotypic differences between mice carrying mutant alleles of Mycn and miR-17~92 and by comparing the consequences of MYCN and miR-17~92 loss in patients. For example, while FS patients with MYCN haploinsufficiency frequently present with gastrointestinal atresia (55% of cases)¹⁵, this phenotype was not observed in any of the six miR-17~92 mutant patients described here, nor was it observed in miR-17~92^Δ/+ or miR-17~92^Δ/Δ mice (data not shown). Because abnormal gut development is reported in Mycn^Δ/Δ mice (refs. ^30,31 and Supplementary Table 3), it seems plausible that MYCN controls gastrointestinal development independently of miR-17~92, or that miR-17~92 is functionally redundant. In contrast, microcephaly, short stature and brachymesophalangy are seen in MYCN- and miR-17~92-mutant patients and mice, indicating functional cooperation between these two genes in skeletal development and growth.

Because none of the individuals with a MIR17HG deletion have gastrointestinal atresia, whether they should be classified as true FS cases or as affected by a novel form of brachydactyly with short stature and microcephaly remains an open question. Nevertheless, our results provide a strong rationale for testing all patients displaying skeletal features of FS for mutations in both MYCN and MIR17HG.

At present, the detailed molecular mechanisms and the targets through which miR-17~92 modulates skeletal development remain unknown and will be the subject of future studies. Yet, it is worth noting that miR-17~92 has been shown to modulate the TGF-beta and Sonic Hedgehog axes^6,10,32–34, two of the most important signaling pathways controlling skeletal development and limb patterning.

Although our analysis of miR-17~92^Δ/+ mice clearly indicates that reduced dosage of this cluster can generate many of the skeletal phenotypes observed in FS patients with MYCN mutation and in patients harboring 13q31.3 microdeletions, we cannot exclude the possibility that GPC5 haploinsufficiency also contributes to the pathogenesis of some of these phenotypes. In particular, it is possible that GPC5 haploinsufficiency plays a role in the severe thumb hypoplasia observed in some individuals carrying 13q31.3 microdeletions. Addressing this issue will require the identification and characterization of additional families carrying deletions of this region and the generation of miR-17~92 and Gpc5 compound-mutant animals.

1