Targeted next-generation sequencing for detection of PIK3CA mutations in archival tissues from patients with Klippel–Trenaunay syndrome in an Asian population

Background Klippel–Trenaunay syndrome (KTS) is a rare slow-flow combined vascular malformation with limb hypertrophy. KTS is thought to lie on the PIK3CA-related overgrowth spectrum, but reports are limited. PIK3CA encodes p110α, a catalytic subunit of phosphatidylinositol 3-kinase (PI3K) that plays an essential role in the PI3K/AKT/mammalian target of rapamycin (mTOR) signaling pathway. We aimed to demonstrate the clinical utility of targeted next-generation sequencing (NGS) in identifying PIK3CA mosaicism in archival formalin-fixed paraffin-embedded (FFPE) tissues from patients with KTS. Results Participants were 9 female and 5 male patients with KTS diagnosed as capillaro-venous malformation (CVM) or capillaro-lymphatico-venous malformation (CLVM). Median age at resection was 14 years (range, 5–57 years). Median archival period before DNA extraction from FFPE tissues was 5.4 years (range, 3–7 years). NGS-based sequencing of PIK3CA achieved an amplicon mean coverage of 119,000x. PIK3CA missense mutations were found in 12 of 14 patients (85.7%; 6/8 CVM and 6/6 CLVM), with 8 patients showing the hotspot variants E542K, E545K, H1047R, and H1047L. The non-hotspot PIK3CA variants C420R, Q546K, and Q546R were identified in 4 patients. Overall, the mean variant allele frequency for identified PIK3CA variants was 6.9% (range, 1.6–17.4%). All patients with geographic capillary malformation, histopathological lymphatic malformation or macrodactyly of the foot had PIK3CA variants. No genotype–phenotype association between hotspot and non-hotspot PIK3CA variants was found. Histologically, the vessels and adipose tissues of the lesions showed phosphorylation of the proteins in the PI3K/AKT/mTOR signaling pathway, including p-AKT, p-mTOR, and p-4EBP1. Conclusions The PI3K/AKT/mTOR pathway in mesenchymal tissues was activated in patients with KTS. Amplicon-based targeted NGS could identify low-level mosaicism from low-input DNA extracted from FFPE tissues, potentially providing a diagnostic option for personalized medicine with inhibitors of the PI3K/AKT/mTOR signaling pathway. Supplementary Information The online version contains supplementary material available at 10.1186/s13023-023-02893-1.

PIK3CA encodes p110α, a catalytic subunit of phosphatidylinositol 3-kinase (PI3K) that plays a role in cellular processes such as proliferation, motility, invasion, and death through its involvement in the PI3K/AKT/ mammalian target of rapamycin (mTOR) signaling pathway [10].Moreover, p110α is required for endothelial cell migration during angiogenesis [11,12], and its aberrant activation has been associated with the development of vascular malformations [13,14].Activation of PI3K leads to phosphorylation of AKT followed by mTOR and its downstream targets, including eukaryotic translation initiation factor 4E-binding protein 1 (4EBP1) [15].
The activation of mutations in PIK3CA is reported to play a role in many human cancers [16].Of the PIK3CA variants, more than 80% are found at three hotspots: the glutamates E542 and E545, located in the helical domain of exon 10, and the histidine H1047, located in the kinase domain of exon 21 [17].These three mutations exert the strongest effect on downstream signaling and enzymatic activation [17].In patients with KTS as well, E542K, E545K, H1047R, and H1047L are the most frequent (i.e., hotspot) variants [3,6,8,9].
These somatic gain-of-function variants, which arise in the postzygotic stage during embryonic development, result in a mosaic pattern in the affected lesion [5].Therefore, molecular testing of peripheral blood or saliva has been ineffective for detecting pathogenic variants in patients with PROS [3].Meanwhile, archival formalinfixed paraffin-embedded (FFPE) tissues can be a valuable resource for clinical genomic studies [18], but the DNA obtained from these tissues can have a wide range of quality depending on factors such as age, DNA-protein crosslinking, fixation conditions, and inhibitors, all of which can affect downstream genomic analyses [19].
Next-generation sequencing (NGS) involving an amplicon-based targeted sequencing method with high sensitivity can identify low-level mosaicism from low-input DNA extracted from FFPE tissues and provide a diagnostic option when affected tissue is available [20].Therefore, this study aimed to demonstrate the clinical utility of targeted NGS with a custom-designed panel for identifying PIK3CA mosaicism in archival FFPE tissues from patients with KTS, a relatively rare vascular malformation with limb hypertrophy.
Clinicopathological diagnoses were capillaro-venous malformation (CVM) (n = 8) and capillaro-lymphaticovenous malformation (CLVM) (n = 6).Geographic CMs were found in 8 patients.Lower limb discrepancy (LLD) in terms of length was found in 7 patients, including 2 patients with epiphysiodesis, while LLD in terms of girth was observed in all 14 patients (right-sided hypertrophy in 7 patients).Digital anomalies were found in 5 patients as macrodactyly of the foot.Patient 7 had bilateral lesions in the lower limbs (with CLVM in the left limb) and bilateral macrodactyly, and the longer right limb required epiphysiodesis.Thirteen patients had received treatment prior to resection for genetic analysis, including partial resection (n = 10), percutaneous sclerotherapy (n = 12), transcatheter arterial embolization for micro-arteriovenous shunts (n = 6), and high ligation of a lateral marginal vein in the thigh (n = 4).

Genotype-phenotype analysis
All patients with geographic CM, histopathological LM, or macrodactyly of the foot had PIK3CA variants.However, we did not find any association with phenotype or its severity between hotspot and non-hotspot PIK3CA variants.

Discussion
This study investigated the largest cohort of molecularly diagnosed patients with KTS in an Asian population.Using archival FFPE tissues, we identified PIK3CA variants in 85.7% of our cohort with KTS, 66.6% of which were the hotspot variants E542K, E545K, H1047R, and H1047L.The non-hotspot variants Q546K and Q546R were also identified, despite being rare in patients with vascular malformations.To our knowledge, Q546K was previously unreported in patients with KTS but was found in a patient with a fibro-adipose vascular anomaly [6] and in 3 patients with LM [8,52].Q546R has been reported in a patient with KTS [7] as well as a patient with LM [8] (Fig. 2).Our mutational findings were in line with those of patients with KTS in Western populations [3,6,8].
The PI3K catalytic subunit p110α encoded by PIK3CA has five domains: a C2 domain, a helical domain, a kinase domain, an N-terminal adapter-binding domain, and a Ras-binding domain (Fig. 2) [56].In many cancers, mutations are found throughout the entire p110α protein, except for the Ras-binding domain, including the following hotspots: E542 and E545 in the helical domain and H1047 in the kinase domain [17].The PIK3CA variants in PROS, including KTS, have a similar profile to that in cancers (Table 2), and an association of hotspot variants  c, d, f, g, i, l, m), and macrodactyly (a, c, d, f, l).Magnetic resonance images are axial views at the arrowhead position in each clinical photograph with the lesions of high signal intensity with more severe hypertrophy has been suggested, with milder hypertrophy linked to rarer non-hotspot variants [7,27,30].However, in our cohort, patients who had non-hotspot variants did not exhibit milder hypertrophy compared with those who had hotspot variants.Thus, a meta-analysis should be conducted to clarify any potential genotype-phenotype correlations in this rare disease.To that end, molecular diagnoses may prove helpful in providing prognostic information on clinical manifestations [20].
Molecular genetic testing for the diagnosis of PROS requires clinically affected tissues, preferably fresh frozen tissues [4].Testing can be performed using FFPE tissues; however, unlike fresh frozen tissues, DNA obtained from FFPE tissues can vary widely in terms of quality [19].NGS with the use of a highly sensitive amplicon-based targeted sequencing method can identify low-level mosaicism from DNA extracted from widely available archival FFPE tissues [20].Although digital droplet PCR is a simple and highly sensitive and specific method for the detection and quantification of targeted DNA variants, entire exons must be sequenced by NGS in order to capture all the coding single-nucleotide variants as well as small insertion and deletion variants in rare diseases [4].Detection of rare variants remains a challenge because of the error-prone nucleotide changes resulting from sequencing errors.NGS combined with molecular barcodes can eliminate false-positive variants and enable detection thresholds of 0.1% VAF [57].
Using archival FFPE tissues, we were able to compare the genotype and histology of the lesions.Immunohistochemistry revealed that the vessels in all 14 patients and the adipose tissues in 13 patients expressed p-AKT, p-mTOR, and/or p-4EBP1 (Fig. 3; Table 3), indicating enhanced activation of the PI3K/AKT/mTOR pathway in mesenchymal tissues compared with that in normal tissues in these patients with KTS.These findings are in line with previous reports of PIK3CA variants detected in adipocytes in PROS [3] and abnormal vessels detected in fibro-adipose vascular anomaly [58].However, 2 patients (patients 13 and 14) with undetected PIK3CA variants showed some expression of p-AKT, p-mTOR, and/ or p-4EBP1 in vessels and adipose tissues.Neither had geographic CM, histopathological LM, or macrodactyly of the foot.Patient 13 can be diagnosed with CVM and congenital nonprogressive limb overgrowth [59] caused by somatic GNA11 mutation [60].Patient 14 might have common VM caused by a somatic TEK mutation [61] encoding TIE2 upstream of the PI3K/AKT/mTOR pathway.
test positive for the presence of one or more PIK3CA variants are eligible for treatment with the PI3K inhibitor alpelisib [62].With a limit of detection from 2.4 to 14.04% VAF [63], the therascreen kit would not have detected five cases in our cohort because of the low VAF (C420R with 1.6% VAF; E542K with 3.3% VAF; Q546R with 3.9% VAF and 5.9% VAF, respectively; and H1047R with 1.6% VAF) as well as one case with the Q546K variant, which is not targeted by the kit.Our findings have potential implications for the treatment of patients with KTS using inhibitors of the PI3K/ AKT/mTOR signaling pathway, which have shown promising results with mTOR inhibitor sirolimus [64,65], pan-AKT inhibitor miransertib [66], and selective class I PI3K inhibitor alpelisib [67] or taselisib [68].It is thus critical to obtain more detailed information regarding specific variants in order to identify of the options for targeted treatment [69].
This study has some limitations.First, we evaluated only patients with KTS who were diagnosed based on the triad of CM, VM, and hypertrophy of the affected limb [2], so there was no controlling for vascular Fig. 2 Distribution of frequent PIK3CA variants in PIK3CA-related overgrowth spectrum (PROS) [3, 6-8, 20, 24-48] and vascular malformations [6,8,42,[49][50][51][52][53][54][55] from the literature as well as our cohort (variants in PROS ≥ 5 patients in each variant).Right, variants found in patients with PROS (blue, n = 597) in the literature and Klippel-Trenaunay syndrome (KTS) in our cohort (orange, n = 12).Left, variants found in patients with vascular malformations except PROS (green, n = 300) in the literature.ABD, p85α-binding domain; RBD, Ras-binding domain; C2, C2 domain; Helical, helical domain; Kinase, kinase domain malformations except KTS in the immunohistochemical analysis.Second, we performed targeted sequencing of PIK3CA gene coding sequences using a panel consisting of amplicons with an overall coverage of 87.9%, so genes not on the panel would have been missed.

Conclusions
We identified PIK3CA variants in 12 of 14 patients (85.7%) with KTS by using archival FFPE tissues, and 8 of these patients had the following hotspot variants: E542K, E545K, H1047R, and H1047L.The rarer non-hotspot PIK3CA variants Q546R and Q546K were also identified in 3 patients.Amplicon-based targeted NGS was able to identify low-level mosaicism from low-input DNA extracted from FFPE tissues, suggesting its potential as a diagnostic option for personalized medicine.

Patients
This retrospective study involved Japanese patients with vascular malformations with lower limb hypertrophy who underwent resection of the vascular malformations at Tonan Hospital between 2011 and 2020.Of the 17 patients identified, 14 provided written informed consent and were included in the analysis.KTS was diagnosed based upon the triad of CM, VM, and hypertrophy of the affected limb [2].In younger patients, VM was often less  , d, f, h, j, k, l, n, p) and negative (c, g, o).Scale bars: a-d, m-p = 250 μm, e-h = 100 μm, i-l = 50 μm conspicuous but was diagnosed based on veins that were subtly dilated relative to the unaffected limb [70].
The patients' medical charts were reviewed, and the following demographic information and medical history data were collected: sex, date and age at resection, lesion resected, type of tissues in resected specimens, clinical photographs, and radiologic studies.Geographic CMs were defined as those with sharply demarcated borders and saturated dark red/purple color throughout the entire stain [70].
Prior to resection, all patients underwent MRI and color duplex ultrasound to evaluate the characteristics, distribution, and extent of the lesions.Vascular malformations were diagnosed based on the clinical history as well as the physical examination, ultrasonography, and MRI findings.LLD was evaluated in terms of length and girth according to teleoroentgenography and crosssection on MRI, respectively.LLD in terms of length was defined as 5 mm longer compared with the unaffected limb in children and 1 cm longer in adults, while LLD in terms of girth was defined as 10% greater in cross-sectional area compared with the unaffected limb [71].

DNA extraction
Surgical specimens were fixed with 10% buffered formalin and embedded in paraffin.Affected tissue was retrieved from archived FFPE tissue blocks.QIAamp DNA FFPE Tissue Kit (Qiagen, Germantown, MD) was used to extract genomic DNA from FFPE tissues.The positive and negative control DNA from the FFPE PIK3CA Reference Standard (E545K, HD112; H1047R, HD599; wild-type [WT], HD320; Horizon Discovery, Cambridge, UK) were extracted using a Maxwell RSC FFPE DNA Kit (Promega, Madison, WI).The concentration and quality of the extracted DNA were assessed using a 2200 TapeStation system with the Genomic DNA ScreenTape (Agilent, Santa Clara, CA) and a Qubit 3.0 Fluorometer with a Qubit dsDNA BR Assay Kit (Thermo Fisher Scientific, Waltham, MA), respectively.DNA controls were prepared for each condition: 5%, 1%, 0.5% and 0.1% of E545K/ H1047R mixture diluted in WT for the positive controls and only WT for the negative control.

Next-generation sequencing
We used an Ion AmpliSeq™ HD Made-to-Order Panel (IAH215884_374) to perform targeted sequencing of all PIK3CA gene coding sequences (3,607 bp).The panel consisted of 64 amplicons with an overall coverage of 87.9%.Deaminated cytosine residues were removed from 20 ng of the sample DNA and the target region was amplified using multiplex PCR with the Ion AmpliSeq™ HD Library Kit (Thermo Fisher Scientific).The primer sequence in the amplicon was partially digested, and the library was amplified using primers, to which barcode The expression of D2-40 in lymphatic endothelial cells was graded as positive (+) or negative (−).The cytoplasmic intensity of the immunohistochemical stains was graded as positive (+) or negative (−) in p-AKT, p-mTOR, and p-4EBP1.None, no small vessels observed sequences were added using Ion AmpliSeq™ HD Dual Barcode Kit (Thermo Fisher Scientific).After purification of the library DNA, the concentration and size of DNA were checked, and the library was mixed.Emulsion PCR was performed to clonally amplify the library DNA on beads and then template beads were collected and sequencing reactions were performed on the Ion S5™ XL system (Thermo Fisher Scientific) using an Ion Chef 550 Chip Kit (Thermo Fisher Scientific).

Bioinformatics analysis for detection of PIK3CA variant
The quality of the read data was checked, and the adapter sequences and poor-quality reads were removed.Then, the reads were mapped to reference sequences (hg19) using the torrent mapping alignment program of Torrent Suite ver.5.16.1(ThermoFisher Scientific) and variants were detected using Ion Reporter ver.

Fig. 1
Fig. 1 Clinical photographs (upper) and short-tau inversion recovery or fat-suppressed T2-weighted magnetic resonance images (lower) of the patients with Klippel-Trenaunay syndrome with detected or undetected PIK3CA variants.Various clinical manifestations were observed, including geographic capillary malformations (a, c-g, i, l), lower limb discrepancy in terms of length (c, d, f, g, i, l, m), and macrodactyly (a, c, d, f, l).Magnetic resonance images are axial views at the arrowhead position in each clinical photograph with the lesions of high signal intensity

Table 3
Immunohistochemical analysis of the lymphatic endothelial marker D2-40 and the PI3K/AKT/mTOR signaling pathway

Table 2
Summary of the frequent PIK3CA variants found in PROS and vascular malformations 5.18 (Thermo Fisher Scientific).Annotation information was assigned to the detected variants.The thresholds for the main parameters of Ion Reporter's variant detection were set as follows: Downsample to Coverage: 20,000; Minimum Allele Frequency of SNP (single nucleotide polymorphism), MNP (multiple nucleotide polymorphism), and INDEL (insertion or deletion of nucleotides): 0.05%; Minimum Variant Score of SNP and MNP: 6; Minimum Variant Score of INDEL: 10.