Klippel–Trenaunay syndrome (KTS) is a rare congenital vascular disorder affecting approximately one in 100,000 live births [Lorda-Sanchez et al., 1998]. It is characterized by: (i) cutaneous vascular malformations of the capillary (i.e., port-wine stains), venous or lymphatic vessels; (ii) disturbed growth (most commonly hyperplasia) of the affected extremity; and (iii) limb anomalies including polydactyly, syndactyly, and hip dislocation [Oduber et al., 2008]. A more detailed description of the clinical features of KTS in included in Supplementary Material 2. Gastrointestinal or genitourinary hemorrhage in patients with visceral involvement can cause serious morbidity and mortality [Cha et al., 2005]. Rarely, pulmonary embolism and hemorrhage due to intracranial arteriovenous malformations can be fatal [Sadiq et al., 2014]. Odd phenotypes of the syndrome may also occur. Rarely, hypo-trophy instead of hypertrophy (inverse KTS) [Danarti et al., 2007], or mixed patterns of increased and decreased growth occur [Ruggieri et al., 2014].

Most KTS cases are assumed to be sporadic; however, several studies have reported familial occurrence [Aelvoet et al., 1992]. Isolated vascular malformations or isolated hemi-hypertrophy have been identified in family members of KTS patients, suggesting that it might be an inherited syndrome with phenotypic variability or incomplete penetrance [Aelvoet et al., 1992; Craven and Wright, 1995; Lorda-Sanchez et al., 1998]. KTS has also been reported in association with chromosomal anomalies, including balanced trans-locations and a supernumerary chromosome 18, as well as a susceptibility variant in the angiogenic factor VG5Q/AGGF1 [Whelan et al., 1995; Wang et al., 2001; Tian et al., 2004; Timur et al., 2004].

The goal of this study was to identify novel, potentially causal, copy number variants (CNVs) that would enhance our understanding of the genetic mechanisms producing KTS.

In the NYS population, the birth prevalence of KTS was 1 per 192,000 live births. KTS cases were compared on maternal demographic factors with all live births from NYS without major birth defects (Table I). KTS was observed to be significantly more common in children of underweight mothers (BMI < 18.5) when compared to all other mothers (BMI > 18.5) [RR: 6.62, 95%CI: 1.37–31.85] but not when compared to only normal weight mothers (BMI 18.5–24.9) [RR: 4.19, 95%CI: 0.85–20.78]. We did not identify any differences in maternal age at delivery or race/ethnicity between cases and controls. Additionally, the sex ratio did not differ significantly between cases and controls, and no differences in birth weight or gestational age were identified.

The microarray analysis of all 17 KTS samples resulted in 2,256 PennCNV and 906 cnvPartition calls. After applying the exclusion criteria described in the methods section, we identified 15 candidate CNVs in seven cases. Of those, 12 CNVs in six cases were validated in the laboratory using qPCR. Validated CNVs included four duplications and two heterozygous deletions intersecting genes, two duplications and one heterozygous deletion encompassing at least one GENCODE transcript, and three intergenic heterozygous deletions (Table II). The CNVs ranged in size from 21 Kbps to 600 Kbps.

Three cases had rare CNVs overlapping genes with functions related to chromatin modification. One case (Case ID 1, Table II) had a 72 Kbps heterozygous deletion at 7p21.1 overlapping two different transcripts of HDAC9 (Fig. 1a and b). A second case (Case ID 2, Table II) was found to have a partially overlapping deletion during qPCR validations. The first case also had a 600 Kbps duplication at 18q23 upstream of SALL3. We also identified a 98 Kbps duplication in a third case (Case ID 3, Table II) overlapping ING5. However, a duplication overlapping both TaqMan copy-number assays used for validation (KT_1343_CXVI3RU [custom designed] and Hs01490888_cn) and ING5 was also found in one control.

In this first genome wide investigation of KTS, we identified CNVs in three cases encompassing genes that are involved in chromatin modification. This process is critical in regulating angiogenesis during embryonic development and is known to be associated with several genetic defects [Hendrich and Bickmore, 2001; Ohtani and Dimmeler, 2011]. Interestingly, chromatin modifying genes have been implicated in the pathogenesis of defects with focal and asymmetric growth such as Silver Russell and Beckwith Weide-mann syndromes by altering tissue specific expression of other genes [Bliek et al., 2006; Delaval et al., 2006]. KTS is postulated to be the result of defective vascular development during embryogenesis [Oduber et al., 2008]; therefore, alterations in chromatin modifying genes may contribute to the pathogenesis of the syndrome.

It is noteworthy that in two cases we identified overlapping heterozygous deletions at 7p21.1 that encompass HDAC9 transcripts. This gene is important in blood vessel growth and proper vascular organization. HDAC9 is a member of the Class II histone deacetylase (HDAC) family that regulates angiogenesis by modulating the expression of the anti-angiogenic microRNA 17–92 cluster. In endothelial cells, silencing this gene inhibits VEGF-A expression, and reduces vascular growth. Hdac9 knockout zebrafish and mice embryos display reduced angiogenic sprouting, and defective vascular formation highlighting this gene’s importance in embryonic angiogenesis [Kaluza et al., 2013]. HDAC9 controls the angiogenic activity of endothelial cells at transcription by encoding multiple splice variants which display distinct functional and localization patterns [Petrie et al., 2003]. The deletion found in Case 1 overlaps exon 1 of isoform 6, a variant that seems to have the essential deacetylase domains but is excluded from the nucleus. It also overlaps intron 2 of isoform 8, a variant that is missing the catalytic subunit and probably lacks deacetylase activity (Fig. 1a and b). It has been suggested that nuclear localization and deacetylation activity transcriptionally control angiogenic activity of endothelial cells in vitro. It is still unclear, however, whether HDAC9 regulates expression of the miRNA 17–92 cluster through histone modification or whether it targets other transcription factors, activators, or repressors of the cluster [Sato, 2013]. During qPCR validations, Case 2 was found to have a heterozygous deletion using validation assay Hs03646420_cn and was normal for assay Hs03638416_cn; since TaqMan copy-number assays target a single genomic location, we can determine that Case 2 carries a deletion that overlaps at chr7:18,267,017 but not its size.

Case 1 also had a 600 Kbps duplication spanning predicted regulatory elements 450 Kbps upstream of SALL3, a spalt-like transcription factor expressed in the embryonic mesoderm. Mutations in several members of the spalt-like family are associated with congenital malformations, indicating that it plays an important role in embryogenesis [Parrish et al., 2004; Barrio and de Celis, 2008]. SALL3 interacts with DNMT3A, a DNA methyltransferase that catalyzes de novo methylation and participates in establishing DNA methylation patterns in early embryogenesis [Okano et al., 1999; Miranda and Jones, 2007]. Experiments in embryonic cells have shown that SALL3 binds to DNMT3A and inhibits de novo CpG island methylation [Shikauchi et al., 2009]. Interestingly, DNMT3A also interacts with HDAC1, another member of the histone deacetylase family, postulated to be involved in angiogenesis [Fuks et al., 2001; Kim et al., 2001].

We also identified a 98 Kbps duplication at 2q37.3 spanning inhibitor of growth 5 (ING5). We are the second group to report on a CNV in this region associated with KTS [Puiu et al., 2013]. ING5 encodes a transcriptional coactivator and a histone acetylation regulator essential for chromatin remodeling and genomic transcription during embryonic development. More particularly, ING5 is a component of the of MOZ/MORF complex and is necessary for the histone acetylation activity of the complex as it binds to histone H3K4 [Carlson and Glass, 2014; Yang, 2015]. We also found a similar duplication in one healthy control (based on predicted breakpoints, at least 83% overlap with the duplication identified in Case 3). This finding could also indicate that this variant is not pathogenic, or it that it has incomplete penetrance.

Mosaic somatic mutations have been found in KTS [Lorda-Sanchez et al., 1998; Timur et al., 2004; Happle, 2009]. Mutations in PIK3CA, a phosphatidylinositol 3-kinase have been implicated in the pathogenesis of KTS and other overgrowth syndromes such as CLOVES and Proteus syndromes [Kurek Kyle et al., 2012; Youssefian et al., 2015; Vahidnezhad et al., 2016]. A noteworthy study reported an association between KTS and somatic mutations in PIK3CA [Luks et al., 2015]. Mutations in this gene were identified in affected tissue from cases with KTS and other lymphatic and vascular disorders. Because there was no germline tissue available to these investigators, we investigated whether, this was a germline mutation by searching for copy number variants overlapping PIK3CA but did not identify any. Additionally, another somatic mutation in RASA1, a negative regulator of the RAS signaling pathway, have also been associated with KTS and other capillary malformations [Whelan et al., 1995; Eerola et al., 2003; Hershkovitz et al., 2008]; this relationship; however, was not confirmed by other studies [Revencu et al., 2013].

One novel finding of our study was the higher prevalence of KTS in offspring of underweight mothers (BMI < 18.5). KTS was found to be significantly more common in children of underweight mothers when compared to all other mothers. Because there is no evidence that being overweight or obese is protective against malformations, we believe that this finding did not reach statistical significance when only normal weight mothers were used as the comparison group due to the smaller sample size. Pre-pregnancy BMI is thought to be an important determinant of fetal programming and development. Maternal obesity is strongly associated with several birth defects. Maternal pre-pregnancy underweight has been associated with gastroschisis [Lam et al., 1999], as well as atrial septal defects and hypospadias although the data for the latter two defects are mixed [Happle, 2009; Rankin et al., 2010; Block et al., 2013; Cai et al., 2014]. The association between maternal underweight and KTS requires confirmation.

Our study has several strengths. To our knowledge, this is the first whole genome investigation of KTS. The large NYS population from which data were drawn allowed us to accumulate enough cases to examine the role of CNVs in the pathogenesis of this syndrome. Additionally, to our knowledge, this is the first population-based investigation of KTS. By comparing cases to three million births without major birth defects we were able to show that underweight mothers had an increased risk of having a child with KTS. One limitation of our study is the lack of parental DNA in order to evaluate whether, the identified CNVs are inherited or de novo. It should be emphasized; however, that these are rare CNVs that are not present in our control populations. Another limitation of our study should also be noted. Although physicians and hospitals are required to report birth defects to the NYS CMR, there is a possibility of under-ascertainment or incorrect coding. KTS may not be diagnosed in the first 2 years of the infant’s life and, therefore, not reported to the CMR. Our prevalence was lower than previous studies have reported. We reported birth prevalence. Other studies might have reported prevalence in older children resulting in more complete ascertainment. It should be noted; however, that ascertainment of selected birth defects by the CMR has been shown to be excellent [Honein and Paulozzi, 1999].

In conclusion, we identified several CNVs in genes involved in chromatin modification, a key regulatory process of embryonic vascular development, which suggests that such genes may be involved in the pathogenesis of KTS. We also identified low maternal BMI as a potential risk factor for this syndrome. Although pathogenic CNVs do not appear to be a frequent cause of KTS, our results suggest avenues for future investigation of the role of chromatin modifying genes in KTS.