Sotos syndrome (MIM 117580) is a condition characterised by overgrowth, distinctive facial features, and learning disability.^1,2 Recently, we and others have shown that small intragenic NSD1 mutations and 5q35 microdeletions that delete one whole copy of NSD1 are responsible for the majority of Sotos syndrome cases.^3–6 There has been debate as to whether allelic or genetic heterogeneity is responsible for the remaining cases and it has been proposed that GPC3 mutations or 11p15 abnormalities, or both, may be responsible for some Sotos cases without NSD1 abnormalities.^7,8

Current diagnostic NSD1 screening strategies typically employ fluorescent in situ hybridisation with a single NSD1 probe to detect 5q35 microdeletions, together with mutational screening of the 23 NSD1 exons by heteroduplex analysis or direct sequencing.^3–6,9,10 These techniques detect whole gene deletions and small intra-exonic deletions/insertions or point mutations. However, they are rarely able to detect deletions or duplications that encompass one or more exons. Partial gene deletions/duplications have been identified in several conditions such as Fanconi anaemia group A, hereditary non-polyposis colorectal cancer (HNPCC), and hereditary breast-ovarian cancer syndrome (HBOS),^11–13 but have not been reported in Sotos syndrome.

Several methods for detecting partial gene deletions/duplications have been described, including Southern blotting, quantitative fluorescent polymerase chain reaction (QF-PCR), real time PCR, multiplex amplification and probe hybridisation (MAPH), and multiplex ligation dependent probe amplification (MLPA).¹⁴ MLPA is a high throughput, sensitive technique for detecting copy number variations in genomic sequences.¹⁵ Two adjacently hybridised probes are ligated and then PCR amplified using fluorescent labelled universal primers that correspond to sequence tags present in every probe. Each probe is designed to give a uniquely sized product resulting in a ladder of amplified products that can be quantified by fluorescent electrophoretic analysis.

We have analysed over 400 overgrowth cases, or cases with facial features similar to Sotos syndrome but no overgrowth, for intragenic NSD1 mutations and 5q35 microdeletions. The majority of negative cases after these analyses were clinically considered to have conditions other than Sotos syndrome because they lacked the characteristic facial features of this condition. However, in 18 cases we considered the facial features to be classic of Sotos syndrome. These cases were clinically very similar to the cases with NSD1 mutations and we suspected covert NSD1 abnormalities were responsible. We therefore developed an MLPA probe set that interrogates all NSD1 exons and analysed the 18 cases for exonic deletions/duplications.

METHODS

Cases

The research was approved by the London Multi-Research Ethics Committee and consent was obtained from all cases or their parents. DNA was extracted using standard methods. Intragenic NSD1 mutations were identified by methods described in Douglas et al,⁴ and 5q35 microdeletions were detected by microsatellite analyses, QF-PCR, or FISH, or combinations of these.¹⁶ Cases negative after these analyses were reviewed if clinical photographs were available. Cases with a clinical phenotype of Sotos syndrome were screened by MLPA.

MLPA analyses

MLPA reactions using 150 ng genomic DNA were carried out as previously described.¹⁵ We included known 5q35 microdeletion cases and control samples in all experiments, which were repeated in triplicate. MLPA PCR products were run on an ABI 3100 and analysed using Genescan and Genotyper software (Applied Biosystems, Foster City, California, USA). Chromatograms were initially checked individually by eye and samples were excluded from further analysis if the MLPA reaction control peaks were high. Sample peak areas were exported to an Excel template. For each sample, the peak area for each NSD1 and FGFR4 probe was normalised by dividing by the sum of the peak areas for all control probes (excluding the Y chromosome probe). The normalised ratio for each probe was then compared between samples. Deletion of an exon was indicated by a ratio that was approximately half that in controls and approximately equal to that obtained in 5q35 microdeletion cases.

Long range PCR analyses to define deletion breakpoints

We carried out long range PCR using BIO-X-ACT^TM long DNA polymerase with 2.5 mM MgCl₂ and a touch down 60–50° protocol. For exon 1–2 deletions we used “Hi-Spec Additive” according to suppliers instructions for GC-rich templates. We designed primers at approximately 2 kb intervals in the 20 kb flanking regions 5′ to exon 1 and 3′ to exon 2 (table 1). We used various combinations of these primers in cases with exon 1–2 deletions identified on MLPA and ran the PCR products on 1% agarose gels to detect shortened products. We purified the PCR products using QIAquick spin columns (Qiagen, Valencia, California, USA) and these were sequenced using the BigDye 3.1 cycle sequencing kit (Applied Biosystems). We designed additional internal primers to sequence across the breakpoints as necessary (primers and conditions available on request). For deletions of all other exons we used combinations of the forward and reverse primers for NSD1 mutation screening reported in Douglas et al.⁴

RESULTS

NSD1 MLPA probe set

We designed target probes for each NSD1 exon that avoided polymorphisms identified through our NSD1 mutation analyses (table 2). To optimise the probe set we analysed normal controls and samples from 5q35 microdeletion cases, in which every NSD1 exon is deleted. We replaced probes that gave poor or inconsistent results such that the probes in the final mix robustly detected exonic deletions and produced normalised ratios of ∼0.5 for each exon in microdeletion cases compared with controls. Only the probe for exon 1 and occasionally exon 2 gave variable results, particularly if poorer quality DNA was used. To offset this variability we added another probe that is 1 kb 5′ to exon 1 and we recommend that exon 1 and 2 deletions are confirmed, where possible, by long range PCR using the primers in table 1. The final probe set includes probes for each of the 23 NSD1 exons, the probe that is 1 kb 5′ to NSD1, two probes for FGFR4, which is 35 kb centromeric to NSD1, probes for FLT4 and TRIM52, which are telomeric to NSD1, and 14 control probes (table 2).

Analyses of Sotos syndrome cases

We used the NSD1 MLPA probe mix to analyse 18 classic Sotos cases negative for NSD1 mutations and 5q35 microdeletions, of which 17 were from the UK and one was from Belgium. We identified eight partial NSD1 deletions involving one or more exons (figs 1 and 2). All eight cases had the classic facial and clinical features of Sotos syndrome (table 3). In six cases we were able to amplify a long range PCR product across the deletion, thus confirming the MLPA results. In the remaining two cases—COG567 and COG151—the substantial size of flanking sequence in which the breakpoints were potentially located precluded long range PCR analyses (fig 2). We did not identify any abnormalities on MLPA in the remaining 10 cases. We suspect these are caused by covert NSD1 abnormalities because the phenotype of these cases was very similar to that of NSD1 positive cases.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

 Multiplex ligation dependent probe amplification (MLPA) analysis in Sotos syndrome cases with partial NSD1 deletions, 5q35 microdeletion, and a control. The asterisks indicate deleted exons. Numbers at the bottom of the figure refer to the NSD1 exon; C indicates a control probe. The apparent deletion of exon 8 in COG011 is actually caused by an insertion of illegitimate sequence within the ligation sequence of the exon 8 probe.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

 Schematic representation of NSD1, drawn to scale, showing size and position of partial NSD1 deletions. “Alu” indicates that an Alu repeat element was present at the breakpoint with the subfamily indicated; diamonds indicate that the breakpoint was not defined.

Four cases showed deletions of exons 1–2 on MLPA and each had a unique deletion with different breakpoints (figs 1 and 2). In COG003 a de novo 23.8 kb deletion encompassing exons 1 and 2 had occurred between highly homologous, directly orientated Alu repeats. Perfect sequence alignment of 11 bp was present at the breakpoint junction (fig 3). This suggests that non-allelic homologous recombination (NAHR) between Alu elements was the likely mechanism of generation.^17,18 A similar mechanism is implicated in COG068 in whom an 11.4 kb deletion between different directly orientated Alu elements had occurred. Sequence overlap of 17 bp was present at the breakpoint junction. Parental samples were not available, but neither parent is affected and the deletion is therefore likely to have arisen de novo. In COG001 we were not able to generate fully readable sequence at the breakpoint junction, but the de novo deletion was ∼15 kb and the breakpoints were in directly orientated Alu repeats. In the final case with an exon 1–2 deletion, COG567, the two FGFR4 probes in the MLPA mix were deleted, as was the maternal allele at the microsatellite marker SOT3 in intron 2.⁴ This indicates that the deletion was maternally derived, at least 67 kb in size, and differs from the other cases with deletions of exons 1–2.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

 Characterisation of partial gene deletions in COG003 and COG079. (A) Chromatograms at the deletion breakpoints, with regions of sequence overlap indicated by black bars. (B) Sequence flanking the deletion breakpoints, with region of sequence overlap indicated by black boxes around the nucleotides. Identical residues are indicated by vertical bars.

The remaining partial gene deletions each involved different exons. In COG056 recombination between an Alu-Sp repeat in intron 8 and an Alu-Sq repeat in intron 13 resulted in deletion of 14.5 kb including exons 9–13. Perfect sequence overlap of 23 bp was present at the junction. In COG079 there was deletion of 7.9 kb including exons 19–21 between intron 18 and intron 22. The proximal breakpoint was not in a recognised repetitive element and the distal breakpoint was in an Alu-Y element. There was overlap of only a single thymidine residue at the breakpoint and there was little homology between sequences flanking the breakpoint (fig 3). This indicates that non-homologous end joining (NHEJ) is the likely mechanism of generation.¹⁹ NHEJ is also implicated in the de novo 2.5 kb deletion of exon 22 in COG280. Neither breakpoint was in a known repeat or a recombination associated motif²⁰ and the sequence overlap at the junction was restricted to three base pairs, “ACT”. In the final case, COG151, a de novo deletion of exons 3–5 was identified. Owing to the extensive size of intron 2 we did not attempt long range PCR analyses to define the precise breakpoints.

In one case, COG011, the MLPA analysis indicated a de novo deletion of exon 8 (fig 1). However, long range PCR showed that in fact an insertion of ∼190 bp of Alu-Y sequence had occurred within the ligation sequence of the exon 8 MLPA probe. On reviewing the heteroduplex analysis from the NSD1 mutation screen in this individual, an aberrant band was visible but had not been noted because the size of the insertion had resulted in the heteroduplex band being very substantially displaced from the wild type allele.

DISCUSSION

Our data show that partial deletions of one or more NSD1 exons are a novel cause of Sotos syndrome and are readily identifiable by MLPA. Furthermore, MLPA is a robust method for detecting 5q35 microdeletions which occur in around 10% of non-Japanese Sotos cases and are the commonest cause of Sotos syndrome in Japan.^4,21

Each of the eight partial gene deletions we identified was unique. Alu mediated recombination was the likely cause in at least four cases, similar to other conditions with partial deletions.¹⁷NSD1 is enriched for Alu elements with a density of 40.2% compared with 10.6% for the genome generally.²² In particular, intron 2—the largest intron in NSD1—contains 115 Alu repeats and the sequence 5′ to NSD1 is also highly enriched with Alu elements. This probably explains why the most frequent partial deletion in our series encompassed exons 1–2. Despite the high density of Alu repeats in NSD1, two of six cases in which we defined the exact deletion breakpoints were not mediated by NAHR between Alu elements. In these, NHEJ was the likely mechanism as the breakpoints were not in repetitive elements, there was little homology between the breakpoint flanking sequences, and the region of sequence overlap at the junction was very short.¹⁹

As none of the parents was affected it is very likely that all the partial deletions occurred as a result of de novo events, as Sotos syndrome is essentially a fully penetrant disorder.² We confirmed this in six cases. Together with the observation that all the partial deletions were of unique size, these data suggest that founder or recurrent partial deletions are unlikely to be prevalent in Sotos syndrome and that strategies to identify such deletions will need to analyse the full gene. This differs from other conditions such as HNPCC and HBOS in which founder deletions are well recognised and specifically targeted in screening analyses.^23–25 This also contrasts with Sotos syndrome caused by 5q35 microdeletions in which a recurrent ∼2 Mb deletion causes the majority of Japanese Sotos syndrome cases and a lesser proportion of non-Japanese cases.^16,26

We have not seen false positive results for whole gene deletions or deletions involving more than one exon, and reproducibility of the deletion in a second separate MLPA analysis should be sufficient confirmation for such abnormalities.²⁷ For reproducible single exon deletions, sequencing of the relevant exon should be undertaken. This is because intra-exonic mutations within the probe sequences can prevent hybridisation or ligation, resulting in the profile of a single exon deletion, as exemplified by COG011 in this study. If MLPA were used as the primary NSD1 screening method before mutational screening, small deletions, insertions, or point mutations may be detected in this way. If the exon sequence is wild type we recommend confirmation of a single exon deletion by long range PCR.

To date, we have undertaken all NSD1 testing in clinically diagnosed cases of Sotos syndrome in the UK; seven of 124 cases were caused by partial NSD1 deletions and 12 by 5q35 microdeletions. Our data therefore suggest that MLPA (or a similar assay) could usefully be undertaken as the primary screen in individuals with a clinical diagnosis of Sotos syndrome to detect partial and whole gene deletions. MLPA analyses are less expensive than fluorescent in situ hybridisation (FISH) and can be carried out on an aliquot of the DNA obtained for mutation screening, so they do not require a separate sample to be taken. We estimate that primary MLPA analyses would detect around 15% of NSD1 abnormalities in non-Japanese individuals and a high proportion of NSD1 abnormalities in individuals from Japan.