Rare dosage abnormalities flanking the SHOX gene

Transcriptional regulation of the SHOX gene is highly complex. Much of our understanding has come from the study of copy number changes of conserved non-coding sequences both upstream and downstream of the gene. Downstream deletions have been frequently reported in patients with Leri–Weill dyschondrosteosis or idiopathic short stature. In contrast, there are only four cases in the literature of upstream deletions that remove regulatory elements. Although duplications flanking the SHOX gene have also been reported, their pathogenicity is more difficult to establish. To further evaluate the role of flanking copy number variants in SHOX-related disorders, we describe nine additional patients from a large SHOX diagnostic cohort. The nine cases presented here include five with duplications (two upstream of SHOX and three downstream), one with a downstream triplication and three with upstream deletions. Two of the deletions remove a single conserved non-coding element (CNE-3) while the third does not remove any known regulatory element but is just 4 kb upstream of SHOX, and the deleted region may be important in limb bud development. We also describe six families with novel sequence gains flanking SHOX. Three families had increased dosage of a proposed regulatory element approximately 380 kb downstream of SHOX (X:970,000), including one family with the first ever reported triplication of this region. One family had two in cis downstream duplications co-segregating with LWD, and the two others had a duplication of just the upstream SHOX regulatory element CNE-5. This study further extends our knowledge of the range of variants that may potentially cause SHOX-related phenotypes and may aid in determining the clinical significance of similar variants.

There are three known CNEs upstream of SHOX which have transcriptional activity: CNE-5 (X:398,357-398,906), CNE-3 (X:460,279-460,664) and CNE-2 (516,610-517,229) [23]. Deletions upstream of SHOX are infrequently reported compared to downstream deletions with only four previously reported probands with upstream CNEs deletions, all in individuals with ISS [24][25][26]. Upstream SHOX duplications are also comparatively rare with only four reported to date, all in patients with ISS. One of these duplications involved all three upstream CNEs [27] while two other reported cases involved two of the upstream elements (CNE-5/CNE-3 and CNE-3/CNE-2) [28]. The only patient reported to date with a duplication of a single upstream CNE had a duplication of CNE-2 [29].
It is accepted that SHOX downstream deletions that remove regulatory elements can result in SHOX haploinsufficiency, but other, rarer flanking dosage abnormalities are more difficult to interpret, in part due to their rarity in the literature. In this paper, we describe five individuals with a SHOX upstream dosage abnormality from a large SHOX cohort, including two deletions and two duplications of a single upstream CNE, plus an additional upstream deletion that contains no known CNE but removes part of a region that may be involved in limb bud development. We also present four probands with SHOX downstream dosage abnormalities: three short stature probands with unique CNV gains containing the proposed X:970,000 regulatory element, including the first reported case of a triplication of this region, and an LWD family with two in cis downstream duplications.

Study cohort
Patients 1-8 were identified from a SHOX cohort of 1963 referrals to the Wessex Regional Genetics Laboratory (WRGL). All patients referred specifically for SHOX analysis were included in the cohort, regardless of the referral reason. The deletions, duplications or triplication identified below were not detected in 22,017 array comparative genome hybridisation (aCGH) referrals to the WRGL (predominantly referred with developmental delay) or two cohorts of anonymised healthy controls totalling over 800 individuals [22,27]. Patient 9 was identified after referral to the Switzerland Health Service. The probands presented here were consented for SHOX analysis as part of their routine clinical care within the UK National Health Service (Patients 1-8) and the Switzerland Health Service (Patient 9).
Purified genomic DNA obtained from blood samples was extracted according to standard protocols. Analysis of SHOX and its flanking regions was carried out using the following techniques:

SHOX multiplex ligation-dependent probe amplification (MLPA) analysis
MLPA [30] was performed using the current SHOX kit at the time of testing according to the manufacturer's protocol (P018-G1; MRC-Holland, Amsterdam, The Netherlands). Abnormal MLPA data were repeated on a separate run in order to confirm the results.

Sequencing of the SHOX coding region
Direct sequencing of all the coding exons (isoform A, NM_000451.3, exons 2 to 6a) was carried out to exclude the presence of single-nucleotide variants and small deletions/insertions in the SHOX coding sequences and intron/exon boundaries (primer sequences available upon request).

aCGH analysis
The sizes of the SHOX flanking region duplications in the probands 1-4 and 6 were further defined in the Salisbury laboratory using aCGH analysis. The aCGH was performed using Oxford Gene Technologies (OGT, Oxford, UK) 60-mer oligo-array printed in 8 × 60 K International Standard Cytogenomic Array (ISCA) Consortium configuration, according to manufacturer's instructions, using Kreatech's pooled control DNA as a reference (Kreatech Diagnostics, Amsterdam, Holland). Slides were scanned using a G2539A Agilent microarray scanner (Agilent Technologies, Wokingham, UK) and analysed using OGT's CytoSure Interpret (v3.6) microarray software. The duplications in Patient 5 and 9 were confirmed and sized using single-nucleotide polymorphism array (SNP array) testing in external laboratories.

Long-range PCR
As Patient 3 had a deletion below the detection limits of aCGH, long-range PCR of the deletion region was carried out in order to define the breakpoints. Essentially, PCR forward primers were designed at approximately 1-kb intervals across the potentially deleted region, and these were used in conjunction with a reverse primer situated immediately downstream of the maximum extent of the deletion as defined by aCGH. One primer pair produced a PCR product that was not present in normal controls, and this was subsequently Sanger sequenced to determine the breakpoints. Long-range PCR was also required for patient 2, but that involved a single pre-determined primer pair. All chromosomal location data are based on the Ensembl hg19 assembly. As the SHOX gene lies within the pseudo-autosomal region, the dosage abnormalities may be on the X or the Y chromosome in males, but the locations given here refer to the X chromosome for all individuals to aid comparison.

Results
All patients underwent SHOX diagnostic testing, comprising sequencing of the SHOX coding exons and MLPA. All variants were initially detected by MLPA and then further defined using aCGH, SNP array testing and Sanger sequencing where applicable. The minimum and maximum breakpoints of the CNVs are shown in Fig. 1 and Table 1. We cannot rule out that the additional material is located elsewhere in the genome, and the exact location and orientation of the triplicated and duplicated segments could not be determined.
Sanger sequencing excluded the presence of a SHOX pathogenic coding sequence variant in all individuals. Although some individuals had additional molecular or biochemical testing, none underwent whole exome sequencing so other potential causes of short stature have not been excluded. The Z-score (standard deviation) given for all patient heights is based on the relevant World Health Organisation length-for-age charts.
The clinical information, inheritance patterns and molecular results are summarised in Table 1, and the relevant pedigrees of the four families where samples from additional family members were available are shown in Fig. 2.

Discussion
Deletions flanking the SHOX gene are an established cause of SHOX-related phenotypes. However, CNVs near the SHOX gene should be interpreted with caution as the PAR1 region is highly repetitive and prone to structural rearrangements. The difficulty in distinguishing between a benign CNV and a clinically relevant CNV with variable phenotypic expression is demonstrated by the common 47.5 kb SHOX downstream deletion where six of the 14 probands with the X:780,550-828,092 deletion (43%) in one cohort had inherited the variant from a phenotypically normal parent [22]. Nevertheless, the study of such cases is important to increase our understanding of SHOX regulation and to try to identify the aetiology of these rearrangements.
We present nine novel CNVs flanking the SHOX gene in probands referred specifically for diagnostic SHOX testing; the frequency was 8/1963 for patients tested in the WRGL SHOX cohort. None of these CNVs were identified in 22,017 patients referred to the WRGL for diagnostic aCGH testing (although only eight of the nine variants are detectable by aCGH), and no comparable deletion or duplication has been reported in the Database of Genomic Variants (http:// proje cts. tcag. ca/ varia tion/) or in over 800 anonymised healthy controls analysed by MLPA [22,27]. While it is not possible to unequivocally demonstrate causality, the identification of rare novel  CNVs provides evidence for the importance of this class of abnormality and may help to define critical regulatory regions.
In this study, we describe three patients with small deletions upstream of SHOX, all in individuals with ISS. All of the four previously reported upstream deletions were also seen in patients with ISS [24][25][26]; therefore, upstream deletions may be associated with a more subtle phenotype than downstream deletions which can often result in LWD. The deletions in Patients 1 and 2 remove part or all of CNE-3. All four previously reported upstream deletions involved the loss of CNE-3, plus CNE-2 and/or CNE-5 [24][25][26]. Therefore, CNE-3 appears to be the critical upstream element with the loss of CNE-3 sufficient to cause ISS.
In contrast, the deletion in Patient 3 does not contain any known regulatory elements. However, the deleted interval contains a segment strongly conserved down to zebrafish, and a predicted layered H3K27Ac mark, often found near active regulatory elements or chromatin loops [31], is located close to the deletion. Figure 3 shows the UCSC genome browser output (http:// genome-euro. ucsc. edu) from CNE-5 to X:970,000 including sequence conservation and predicted layered H3K27Ac marks. The layered H3K27Ac mark approximately 10 kb upstream of the deletion in Patient 3 does not correspond to any known regulatory element. Enriched H3K27 acetylated regions in developing human limbs [31] are shown in greater detail in the lower half of Fig. 3. The deleted region removes part of region that is acetylated at three of the four developmental stages for which data are available (E33, E41, E44 and E47 datasets) [31]. Therefore, the deleted interval may be essential for normal SHOX expression.
The two duplications in Patient 8 segregate with LWD in the proband and two other affected family members. In contrast, assignation of causality for the duplications in Patients 4-7 and 9 is more difficult; they were  ascertained through their short stature which is a common phenotype with multiple causes. Since flanking CNVs are generally associated with mild SHOX-related phenotypes, they could result in reduced expression rather than complete haploinsufficiency. Within SHOX cohorts, duplications of upstream CNEs have previously been reported exclusively in individuals with ISS [25,28], although a duplication of all three upstream CNEs has been reported in one control individual from a study of the SHOX region [29] and one individual from a study of genome-wide CNVs [32]. Similarly, duplication of the proposed X:970,000 regulatory element [20] has also been reported as a copy number variant in another large study [33]. However, as the common 47.5 kb downstream deletion [17,22] is associated with a highly variable phenotypic effect ranging from normal to Leri-Weill dyschondrosteosis (with 43% of carrier parents being phenotypically normal [22]), the presence of a SHOX flanking variant in a phenotypically normal individual does not preclude that variant from being causative but with a variable phenotype. SHOX phenotypic variability may also be influenced by modifier genes such as CYP26C1 [34].
In contrast with deletions, the causative mechanism(s) for duplications is less clear as there is no loss of sequence material. However, there is evidence to suggest that the duplication of only the downstream CNE9 regulatory element is sufficient to cause ISS [18,28,29]. There are a number of potential explanations that can be considered. Firstly, gene regulation depends not only on the presence of regulatory elements and associated complexes but also on the nuclear positioning, chromatin conformation and integrity of the flanking chromosomal segments [35,36]. Insertion of duplicated material between SHOX and a given CNE would alter the normal chromatin structure and potentially affect gene expression. Similarly, the CNVs in patients 3, 4, 7 and 9 contain, or are close to, a predicted active H3K27Ac mark and could disrupt normal chromatin loop formation that links enhancers and promotors [37,38].
Secondly, the presence of additional copies of a regulatory element may act to reduce the availability of transcription factors that bind to these elements. For example, as three copies of the X:970,000 region exist on one chromosome in Patient 4, transcription factors that bind to this region will only bind in the "optimal place" one third of the time. A similar mechanism could be postulated for Patients 5, 7 and 9.
Thirdly, although approximately 95% of large duplications genome-wide are reported to be tandem [39], there could be undetected complexity, particularly in Patient 8 who has two in cis downstream duplications. An inverted tandem duplication may allow single-stranded DNA to form a quasipalindromic loop which could block any enhancers contained within the region.
Fourthly, downstream duplications could disrupt the adjacent topologically associated chromatin domain (TAD) boundary. TADs are discrete compartments of approximately 1 Mb in size, which restrict regulatory chromosomal interactions [38,40]. TAD boundary regions contain insulators that block interactions across adjacent TADs [41] and variants that disrupt TAD structures can cause malformation syndromes through de novo enhancer-promoter interactions and mis-expression [40]. The SHOX 3' TAD boundary has been mapped immediately upstream of the CRLF2 gene (i.e. close to X:1,314,890) [38], although other publications describe the SHOX TAD as covering an approximate region of either X:284,600-1,355,600 [25] or X:350,001-1,035,000 [37]. The duplication in Patient 9, and possibly the triplication in Patient 4, spans the SHOX TAD and would create an extra copy of the TAD boundary. In addition, the duplications seen in Patients 7 and 8 are contained within the SHOX TAD, so they would move the TAD boundary further from SHOX which may alter essential 3-D genomic architecture.

Conclusions
Although the dosage abnormalities presented here were all formally classified and reported as variants of uncertain clinical significance, they are rare or absent in control populations. The results from Patient 3 suggest an additional SHOX critical region near X:581,000, and the deletions in Patients 1 and 2 provide further evidence that the loss of CNE-3 can cause ISS. Also, the triplication seen in Patient 4 and the two novel duplications in Patients 7 and 9 involving the X:970,000 region provide further evidence that copy number gains containing X:970,000 may be a cause of short stature. Therefore, this study further extends our knowledge of the range of variants that may potentially cause SHOX-related phenotypes.