A novel somatic mutation in GNB2 provides new insights to the pathogenesis of Sturge–Weber syndrome

Abstract Sturge–Weber syndrome (SWS) is a neurocutaneous disorder characterized by vascular malformations affecting skin, eyes and leptomeninges of the brain, which can lead to glaucoma, seizures and intellectual disability. The discovery of a disease-causing somatic missense mutation in the GNAQ gene, encoding an alpha chain of heterotrimeric G-proteins, has initiated efforts to understand how G-proteins contribute to SWS pathogenesis. The mutation is predominantly detected in endothelial cells and is currently believed to affect downstream MAPK signalling. In this study of six Norwegian patients with classical SWS, we aimed to identify somatic mutations through deep sequencing of DNA from skin biopsies. Surprisingly, one patient was negative for the GNAQ mutation, but instead harbored a somatic mutation in GNB2 (NM_005273.3:c.232A>G, p.Lys78Glu), which encodes a beta chain of the same G-protein complex. The positions of the mutant amino acids in the G-protein are essential for complex reassembly. Therefore, failure of reassembly and continuous signalling is a likely consequence of both mutations. Ectopic expression of mutant proteins in endothelial cells revealed that expression of either mutant reduced cellular proliferation, yet regulated MAPK signalling differently, suggesting that dysregulated MAPK signalling cannot fully explain the SWS phenotype. Instead, both mutants reduced synthesis of Yes-associated protein (YAP), a transcriptional co-activator of the Hippo signalling pathway, suggesting a key role for this pathway in the vascular pathogenesis of SWS. The discovery of the GNB2 mutation sheds novel light on the pathogenesis of SWS and suggests that future research on targets of treatment should be directed towards the YAP, rather than the MAPK, signalling pathway.


Introduction
Sturge-Weber syndrome [Mendelian Inheritance in Man (MIM): 185300] is a rare neurocutaneous disorder of dilated postcapillary venules in the eye, skin and the leptomeninges of the brain. Due to its well-circumscribed borders, absence of systemic symptoms and a generally unilateral location, it has been believed to be caused by a somatic mutation (1). This hypothesis was confirmed in 2013 by Shirley et al. (2), who found a single somatic missense mutation in the GNAQ gene (MIM: 600998) in 23 out of 26 patients with Sturge-Weber syndrome and in 12 out of 13 patients with port-wine birthmarks. Additional studies have confirmed that the somatic mutation in GNAQ is found in affected tissue in about 70-90% of the patients with Sturge-Weber syndrome and port-wine birthmarks (3)(4)(5)(6)(7)(8).
GNAQ encodes a G-protein alpha subunit (Gαq) of heterotrimeric G-proteins, and the mutation reported in Sturge-Weber syndrome is considered to be an activating mutation (NM_002072.5: c.548G>A, p.Arg183Gln, the protein variant will from here onward be referred to as GNAQ R183Q), where intrinsic GTPase activity increases downstream signalling. This finding is now shaping a novel understanding of the molecular cues that lead to Sturge-Weber syndrome and other vascular malformations. A longstanding hypothesis has been that Sturge-Weber syndrome results from aberrant early vascular development, due to a somatic mutation originating in a subset of angioblasts (9). Support for this assumption was found in a study by Couto et al. who sequenced DNA from purified dermal endothelial cells from patients with sporadic capillary malformations and revealed that 10 out of 13 samples harboured the c.548G>A mutation in GNAQ (3). Other studies have later corroborated this finding in brain endothelial cells (6).
The heterotrimeric G-protein complex consists of three subunits, the alpha, the beta and the gamma subunit. It is intracellularly located, associated with a G-protein-coupled receptor (GPCR) and responsible for transmitting extracellular signals. In an active state, the protein complex disassociates to an alpha unit and a beta-gamma dimer, which both affect several downstream signalling pathways. The main hypothesis concerning the functional impact of the GNAQ mutation in Sturge-Weber syndrome has been that the decreased GTPase activity results in a moderate increase in mitogen-activated protein kinase (MAPK) signalling (2). Emerging data from cancer research have also raised the possibility that the transcriptional co-activator YAP, a central mediator of the Hippo pathway, may be involved in a Hippo-independent manner (10,11), and the current understanding of the pathogenesis is that the GNAQ mutation causes excessive activation of several downstream pathways including MAPK and YAP (12).
In this study, we aimed to identify, characterize and localize the cellular origin of the genetic cause of Sturge-Weber syndrome in a cohort of Norwegian patients and to explore the underlying mechanisms of disease development. In our cohort, we found that five out of six patients harboured the GNAQ mutation and that it was enriched in endothelial cells. Intriguingly, in the GNAQ mutation-negative patient, we discovered a novel somatic mutation in GNB2 (MIM: 139390), which encodes a β subunit of the heterotrimeric G-protein complex. Importantly, this mutation was also enriched in endothelial cells and provides an alternative molecular basis for aberrant G-protein signalling in Sturge-Weber syndrome lesions. Indeed, ectopic expression of mutant and wild-type GNAQ and GNB2 induced differential MAPK phosphorylation, but similar changes in the YAP pathway, indicating that the latter pathway may be more relevant to the development of Sturge-Weber syndrome than the MAPK pathway.

Clinical and imaging characteristics of the patients
The clinical characteristics of the six enrolled patients are summarized in Table 1. Briefly, all had congenital facial port-wine birthmarks, three had eye involvement and all had findings compatible with vascular malformations on brain MRI. In particular, patient 3 with the GNB2 mutation had a classic portwine birthmark affecting her left eye and brow, and her MRI and CT images demonstrated classic Sturge-Weber syndromeassociated features (Fig. 1).

Targeted sequencing of GNAQ:c.548G>A
Samples from affected and unaffected dermis (patients 2-6), available cell cultures of endothelial cells, keratinocytes and fibroblasts (patients 1-5), as well as laser capture microdissected material (patients 2 and 6) and a mutation positive control sample were subject to targeted amplicon sequencing (with MiSeq) covering the GNAQ:c.548G>A variant (selected samples Table 2, all samples are shown in Supplementary Material, Table  S1). Five out of six affected dermal biopsies were positive for the GNAQ:c.548G>A mutation. Endothelial cell cultures from three out of the four patient samples that could be analysed were positive for GNAQ:c.548G>A, with mutation frequencies ranging from 0.2 to 14%. All keratinocyte and fibroblast cultures   were negative for the mutation. Samples with laser capture microdissected endothelium from patients 2 and 6 were subsequently sequenced, resulting in higher ratios of GNAQ:c.548G>A mutations, 30 and 20%, respectively.

Whole-exome sequencing of the GNAQ mutation-negative patient
Deep whole-exome sequencing of DNA from lesional dermis of patient 3 revealed two candidate variants. One candidate was the GNB2 (NM_005273.  Table 3). Sequencing thus revealed a considerably larger ratio of mutation-positive DNA in the endothelial culture than in the dermis. Another candidate variant from the exome data (ZNF518B (NM_053042):c.A641C:p.Glu214Ala) failed MiSeq validation. MiSeq deep sequencing of all exons and exon-intron boundaries of GNB2 and the homolog GNB1 in DNA from brain from five additional GNAQ mutation-negative patients was performed at the Kennedy Krieger Institute, but none were found to harbour mutations. Likewise, targeted amplicon sequencing with MiSeq of the coding sequences of GNB2 in DNA from blood and brain tissue from a Japanese GNAQ mutation-negative patient did not reveal any somatic mutations.

Structural analysis indicates that the GNB2 mutation disrupts a salt bridge bond between the Gα subunit and the Gβϒ subunit
The identified GNB2:c.232A>G mutation causes a substitution of lysine to glutamic acid at position 78 in the GNB2 protein (protein variant from here onward referred to as GNB2 K78E).
To make structural assumptions and assess possible functional consequences of the K78E mutation, we took advantage of the known crystal structure describing subunit interactions of the modified heterotrimeric complex of transducin (13). In a ribbon diagram of the transducin structure ( Fig. 2), lysine at position 78 of GNB1 (corresponding to the same position in GNB2) is located in the β-propeller domain of Gβ and faces the N-terminal helix of GNAT1 where it forms an intermolecular salt bridge with the aspartic acid at position 22 (corresponding to Asp32 in GNAQ) in the N-terminal helix of Gα (Fig. 2). We may then speculate that if lysine at position 78 of GNB2 is mutated to glutamic acid, cationic ammonium from Lys78 is lost and disrupts the salt bridge to the anionic carboxylate of Asp32, instead inducing a charge repulsion and impairing binding to the alpha chain. The novel somatic mutation identified in GNB2 K78E may therefore provide an alternative molecular basis for increased Gαq signalling.

GNAQ R183Q and GNB2 K78E reduce proliferation, but differentially affect endothelial cell migration
The vascular malformation of the Sturge-Weber syndrome lesion is one of dilated venules, revealing a dysregulated pattern of vascular morphogenesis and likely involving altered proliferation and/or migration of endothelial cells (14). To test whether the GNAQ R183Q and GNB2 K78E mutation may affect endothelial cell behaviour in a similar way, we next assessed how ectopic expression of mutants affected proliferation of human umbilical vein endothelial cells (HUVECs), using non-transduced cells and wild-type protein as controls. Adenoviral transduction of vectors encoding wtGNAQ, GNAQ R183Q, GNAQ:c626A>T (p.Gln209Leu, protein variant from here on referred to as GNAQ Q209L), wtGNB2 and GNB2 K78E revealed that all mutations showed reduced or a trend towards reduced proliferation rates compared to their corresponding wild-type control ( Fig. 3A and B, all P-values in Supplementary Material, Table  S2). We next assessed how mutants affected the endothelial cell migration rate by performing scratch assays. Surprisingly, these analyses revealed that while the GNAQ R183Q mutation  seen in the majority of Sturge-Weber syndrome patients reduced migration compared to wtGNAQ, GNB2 K78E enhanced migration relative to controls (Fig. 3C). The GNAQ Q209L also showed a trend towards enhanced migration (Fig. 3C).

GNAQ R183Q and GNB2 K78E reduce YAP levels while differentially affecting MAPK phosphorylation
Ectopic expression of the GNAQ R183Q or Q209L in HEK293 cells is reported to cause excessive activation of the MAPK pathway (2). In line with these results, we also observed strongly enhanced ERK phosphorylation in response to GNAQ Q209L transduction of both confluent and subconfluent HUVECs, and a weaker effect of GNAQ R183Q transduction (Fig. 4). By contrast, transduction of the GNB2 K78E resulted in a statistically significant reduction in pERK levels, which was more pronounced in subconfluent cells. We also analysed the MAPKs p38 and pJNK, as well as the MAPK-activated protein kinase pS6, and observed a tendency to enhanced phosphorylation by GNAQ R183Q in HUVECs, most pronounced in subconfluent cells (Fig. 4A). Notably, these results differ from reported findings in HEK293 cells (2). By contrast, and with the exception of a weak reduction of p38 in confluent cells, we observed no modulation of these kinases in cells transduced with the GNB2 K78E mutant (Fig. 4A). Thus, the GNB2 and GNAQ mutations appear to differ in their effects on MAPK signalling. We next investigated the transcriptional co-activator YAP, recently found to be phosphorylated in GNAQ mutant cells (10,15). Western blot analyses of YAP showed reduced levels of both total YAP and YAP phosphorylated at position Ser127 (from here on referred to as pYAP S127) in GNAQ R183Q and GNB2 K78E mutants when compared to wild-type controls (Fig. 5A). Phosphorylation at Ser127 prevents YAP translocation to the nucleus, where YAP can act as a transcriptional coactivator (10). Efforts to detect nuclear translocation of YAP by means of immunocytochemistry and western blot of nuclear fractions were inconclusive (data not shown). Interestingly, the ratio of phosphorylated/total YAP remained constant, suggesting that the mutations of GNAQ or GNB2 reduce expression of YAP in endothelial cells without changing the fraction of phosphorylated protein ( Fig. 5B and C).

Discussion
This study substantially expands our understanding of Sturge-Weber syndrome pathogenesis by identifying a novel somatic missense mutation in the GNB2 gene, encoding a G-protein beta subunit (Gβ2). As the identified mutation GNB2 K78E is associated with a classical Sturge-Weber syndrome phenotype similar to the one caused by GNAQ R183Q (2), and as there is a well-characterized interaction of these G-protein subunits in GPCR-dependent intracellular signalling, it appears likely that the two mutations cause Sturge-Weber syndrome through the same downstream pathways. Intriguingly, the results of our study indicate that the MAPK pathway may be of less importance than commonly assumed and that YAP signalling, being similarly affected by the two mutations, is more relevant to the functional impairment that leads to Sturge-Weber syndrome.
Five out of the six patients in our study were positive for the GNAQ R183Q mutation, in line with the ratio of previous studies (2,4,8). In the patient negative for the known GNAQ mutation (patient 3), we instead identified a novel somatic mutation in the GNB2 gene (GNB2 K78E). The variant has not previously been reported in public databases, such as the COSMIC and gnomAD databases (https://cancer.sanger.ac.uk/cosmic, https://gnomad. broadinstitute.org/), nor has the GNB2 gene been implicated in Sturge-Weber syndrome or port-wine birthmark pathogenesis. Six additional GNAQ mutation-negative patients were tested for GNB2 K78E at the Kennedy Krieger Institute and Yokohama City University, but all were negative. However, strong evidence underpins the presumed pathogenicity of this mutation. First, the mutation was not observed in the unaffected biopsy of the same patient, while it was detected by both ultra-deep wholeexome and targeted amplicon sequencing in affected biopsies. Second, the phenotype of the patient was typical of Sturge-Weber syndrome, without aberrant features. Third, structural analysis (Fig. 2) and previous immunoprecipitation studies (16) indicate that GNB2 K78E interferes with the interaction between Gαq and Gβ2. Fourth, the recent discovery of a somatic missense mutation in the GNA11 gene, encoding another G-protein alpha q subunit, corroborates the crucial role of aberrant GTPase activity in the development of vascular lesions in atypical Sturge-Weber syndrome (17,18). Finally, the GNB2 mutation showed the same enrichment in endothelial cells as observed in GNAQ mutationpositive samples by us and others (3,6,19). Despite convincing evidence of pathogenicity, the fact that we have only identified the mutation in DNA from affected skin from a single patient and that we were unable to identify it in our six GNAQ mutationnegative controls, suggests that the mutation is a rare cause of Sturge-Weber syndrome. Replication in other patients and tissues is warranted.
The structure of heterotrimeric G-proteins and the roles of GNAQ and GNB2 within the complex explain how the two mutations may cause similar dysfunctional downstream signalling.
The G-protein heterotrimer consists of the GDP-bound alpha q subunit (Gα) associated with the beta and gamma subunits (Gβϒ) (Fig. 2) and is in its resting state bound to the GPCR. Ligandinduced activation of the GPCR leads to an exchange of GDP for a GTP in the Gα subunit and subsequent dissociation from the receptor and the Gβϒ subunits. The unbound Gα subunit mediates further signalling until its GTP is hydrolysed and it reassociates with the Gβϒ subunit and the GPCR (reviewed in (20)). The primary molecular mechanism underlying Sturge-Weber syndrome was identified following the discovery of a somatic mutation in GNAQ, located in the GTPase pocket, inhibiting the hydrolysis of GTP and thereby causing sustained Gαq signalling (2). The novel somatic mutation GNB2 K78E provides an alternative molecular basis for increased Gαq signalling. In Figure 2, which is based on the crystal structure of the Gαβγ heterotrimer with GDP (13), the lysine at position 78 is located on the surface of the β-propeller domain of Gβ and faces the N-terminal helix of Gα where it forms an intermolecular salt bridge with the aspartic acid at position 22 (corresponding to Asp32 in Gαq). Replacement of lysine with glutamic acid will instead induce a charge repulsion, and it is therefore reasonable to assume that this mutation may disrupt the interaction between Gα and Gβϒ. Our analysis of structural data of these subunits provides strong evidence that the GNB2 K78E impairs Gαq reassociation to the Gβϒ subunit, which is likely to result in increased signalling by G-protein subunits. This is also supported by affinity precipitation-based experiments done by Yoda et al. (16), demonstrating that GNB2 K78E lacks the ability to bind Gαq, Gαi2, Gαi3 and Gα13.
The role of G-proteins in Sturge-Weber syndrome pathogenesis was first revealed by the discovery of GNAQ R183Q by Shirley et al. in 2013, who initially suggested that the mutation would lead to excessive activation of G-protein signalling by disabling the Gαq protein's ability to autohydrolyse GTP (2). In their mechanistic considerations, Shirley et al. pointed to previous studies of the GNAQ Q209L in uveal melanoma, showing that the mutation caused upregulation of several MAPK pathway components, including pERK, JNK and p38 (21). The GNAQ R183Q mutation, on the other hand, resulted in only a modest increase in pERK and had no significant effect on other MAPK components (2). Later, Feng et al. (10) found that several activating mutations in both GNAQ and its paralogue GNA11 found in uveal melanoma could also increase YAP signalling. In a review by Comi in 2016, she summarized that the current understanding of the pathogenesis of Sturge-Weber syndrome included the upregulation of both the MAPK pathway and YAP signalling, as well as mTOR indirectly (12). To identify shared downstream effects of the two Sturge-Weber syndrome mutations, we assessed the effect of GNAQ and GNB2 mutants on several MAPK pathway kinases in cultured endothelial cells, observing that GNAQ R183Q and Q209L cause a moderate to strong increase in pERK, whereas the GNB2 K78E seems to decrease pERK. We also found GNAQ R183Q and GNB2 K78E mutants to differentially affect migration as measured in scratch assays. However, both mutations found in Sturge-Weber syndrome and the GNAQ Q209L reduced proliferation in HUVECs. As our main hypothesis was that GNB2 K78E causes Sturge-Weber syndrome through the same mechanistic pathway as GNAQ R183Q, it is possible that altered proliferation is the underlying defect of the Sturge-Weber syndrome lesion and that ERK phosphorylation is not a key driver of altered endothelial cell behaviour involved in the Sturge-Weber syndrome phenotype. This prompted us to look for a possible involvement of the YAP pathway (10,11). Western blots of YAP showed both GNAQ R183Q and GNB2 K78E reduced levels of total YAP and pYAP S127 when compared to wild-type controls, although the ratio of phosphorylated/total protein remained constant. Nonetheless, this indicates that less nuclear YAP is available in the presence of mutated protein, which could explain the reduced proliferation mediated by both mutations. However, additional mechanisms control YAP activity in the cell, including phosphorylation (22,23), actin dynamics and mechanotransduction (24).
These aspects underline the complex regulation of YAP activity. YAP, together with TAZ (encoded by a paralogous gene of YAP1), are transcriptional co-activators that promote tissue growth through binding to transcriptional enhanced associated domains. They have been shown to have multiple effects on vascular cells and structures (8), in addition to regulating angiogenesis (25,26). YAP may be inhibited by the Hippo pathway through phosphorylation, in a process where GPCRs are believed to be crucial (15). However, YAP activity may also be regulated by Hippo-independent mechanisms, including mechanical stress and blood flow. Thus, whereas our findings suggest that YAP/TAZ is an essential pathway in Sturge-Weber syndrome pathogenesis, further studies are required to elaborate the specific mechanisms involved.

Conclusions
GNB2 is the third gene in which a somatic mutation causing Sturge-Weber syndrome has been discovered. Functional data from cultured endothelial cells indicate that the G-protein-YAP/TAZ Hippo signalling pathway is a key mechanism causing Sturge-Weber syndrome and port-wine lesions, while the MAPK pathway appears less essential. Moreover, our study confirms that GNAQ R183Q is the most common cause of Sturge-Weber syndrome and that somatic mutations in Sturge-Weber syndrome must include a localized subset of angioblasts, as the mutations were detected in all successful endothelial cultures and were enriched in laser capture microdissected blood vessels.

Patients
Patients were recruited from the National Centre for Rare Epilepsy-Related Disorders, where a registry of Norwegian patients with Sturge-Weber syndrome is located. Six patients with Sturge-Weber syndrome were enrolled in the study (Table 1). An additional six GNAQ mutation-negative patients with Sturge-Weber syndrome (five recruited from the Kennedy Krieger Institute and one from the Yokohama City University) were included for validation of novel candidate mutations. The study was approved by the Regional Ethical Committee for Medical and Health Research Ethics, South-Eastern Norway (2012/353). Informed written consent was obtained from each study participant according to the declaration of Helsinki.

Sample collection
Two punch biopsies (4 mm) were obtained from each patient; one from unaffected skin and the other from affected skin. Biopsies were taken after administration of local anaesthesia and immediately put into Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum (FCS), 100 U/ml penicillin and 100 μg/ml streptomycin.
Biopsies from patients 1-5 were treated according to a protocol described by Normand and Karasek (27) with some modifications. Briefly, samples were swabbed sequentially with 10% iodine, 70% alcohol and Hank's Balanced Salt Solution (HBSS) and placed with the epidermal side down in a Dispase solution (30 mg/ml in HBSS, 4 • C overnight) to gently remove the epidermis and process it to establish keratinocyte cultures. A curved forceps was then used to apply pressure to the dermis approximately 10 times to push out endothelial cells and vascular stalks into a small volume of DMEM and establish endothelial cell cultures. The remaining dermis was cut into small pieces, snap-freezing one piece (approximately 1-2 mm) without further processing for DNA extraction (except for patient 1) and using the remaining pieces to establish fibroblast cultures.
Biopsies from patient 6 were bisected, processing one half in formalin/paraffin for histology and laser capture microdissection (LCM) and leaving the other half in Dispase overnight to separate dermis and epidermis before snap-freezing and subsequently extracting DNA.

Cell cultures
Endothelial cultures. Medium containing vascular stalks from digested skin samples was centrifuged at 200×g for 5 min and the pellet resuspended in Human Endothelial-SFM (Gibco) supplemented with 0.25 mg/ml cAMP (Sigma-Aldrich, Germany), 10 ng/ml recombinant human (rh) vascular endothelial growth factor (VEGF) (R&D systems), 10 ng/ml rhEGF (R&D Systems, Minneapolis, MN) and 1 ng/ml rh fibroblast growth factor (rhFGF) (Peprotech, Europe) and seeded in a single gelatin-coated well of a 24-well plate. The next day, debris was removed by gentle washing with warm phosphate-buffered saline before adding fresh medium. Endothelial cells were obtained from four patients, but a morphologically pure endothelial culture that was allowed to passage three times before harvesting was possible in only one patient. For the remaining three patients, colonies of endothelial cells were established, but due to lack of growth these were harvested from the primary culture. In one culture, endothelial colonies failed to grow and the sample was discarded.
Fibroblast culture. Dermal explants were placed in 6-well plates and centrifuged for 2 min at 110×g in order for them to attach to the well floor. Wells were then gently filled with 3 ml of DMEM +10% FCS and placed in a cell incubator. Medium was changed every 3-4 days and cells passaged 3-4 times to obtain a confluent T75 culture flask.

Tissue preparation and LCM
Formalin-fixed paraffin-embedded skin samples were cut in 6 μm thick sections on MembraneSlide 1.0 PEN glasses (Carl Zeiss MicroImaging GmbH, Göttingen, Germany) pretreated with UVlight at 254 nm for 30 min. The sections were deparaffinized, rehydrated, haematoxylin-and eosin-stained and dehydrated according to a protocol described by Espina et al. (29) with the modification that tap water was used instead of Scott's Tap Water Substitute. LCM was performed shortly after staining with a PALM Combisystem duoflex microscope system (Zeiss, Germany) equipped with Zeiss objective lenses and using Palm-Robo V4.6 software. A 30 μl of Maxwell Incubation Buffer (from Maxwell 16 FFPE Plus LEV DNA Purification Kit, Promega cat. no. AS1135) was inserted in the lid of the collecting tubes (Micro-Tube 500, Carl Zeiss MicroImaging GmbH, Göttingen, Germany). Endothelial cells from affected tissue of patient 2 and patient 6 were marked and dissected using the laser at 200× magnification (x20 LD plan-NEOFLUAR objective lens) ( Supplementary  Material, Fig. S1). Controls were taken from affected epidermis of the same biopsy (patient 2) and an unaffected FFPE skin biopsy (patient 6). No more than 30 min of dissection time was used for each collecting tube, to prevent the fluid in the lid from evaporating. Cells were microdissected from two sections from patient 2 and from five sections from patient 6. Collected cells from each slide were put into separate collecting tubes that were centrifuged at 12 000×g at 4 • C for 5 min. The affected samples from each patient were pooled, with estimated 3000-5000 cells/pooled sample. The lids of the collecting tubes were washed with additional Maxwell Incubation Buffer to achieve a total volume of 180 μl and centrifuged for 1 min. The samples were transferred to DNase/RNase free tubes and incubated at 70 • C overnight in 20 μl of Proteinase K (20 μg/μl, from Maxwell kit) before mixing in 400 μl of Lysis Buffer (from Maxwell kit).

Library preparation and sequencing
Libraries for amplicon sequencing with MiSeq were prepared using a two-step PCR approach as published by Shirley et al.  Table S4. DNA was amplified in a 20 μl reaction by 0.4 units Phusion High Fidelity Polymerase (New England Biolabs, cat. no. M0530S) supplemented with HF-buffer, 0.5 μm P1 primer, 0.5 μm P2 primer and 200 μm dNTPs (Applied Biosystems/Ther-moFisher Scientific, cat.no. 4303442). If available, 20 ng of DNA was used as input (applying to most of the samples). For a few samples, DNA input was in the range 0.4-10 ng. Cycling conditions were as follows: 98 • C for 30 s, followed by 25 cycles of 98 • C for 10 s/annealing for 20 s/72 • C for 10 s and finally 72 • C for 5 min. Annealing temperatures were 70, 69 and 65 • C for the GNAQ, GNB2 and ZNF518B amplicons, respectively. Reactions were purified by Agencourt Ampure XP beads (Beckman Coulter, Brea, CA) according to in-house procedures, with 25 μl beads per 20 μl reaction. One-third of each purified product (or more, adjusting for DNA input) was subject to the second-step PCR using primers P3 and P4 with conditions as for first-step PCR (annealing temperature 70 • C), except that number of cycles was set to 8. A few samples (mainly the LCM samples) were subject to an additional 2-4 cycles of the second PCR in order to get enough material for sequencing. Reactions were purified by AmPure XP beads. Amplicon size was evaluated on TapeStation (Agilent), using the D1000 kit (Agilent cat. no. 5067-5582/3), and concentrations were measured on Qubit. Final libraries were sequenced on Illumina MiSeq, either 50 bp single read or 150 bp paired end. The sequencing service was provided by the Norwegian Sequencing Centre (www.sequencing.uio.no).

Bioinformatics processing of amplicon sequences
We replicated the analysis method described by Shirley et al. (2). In detail, the sequences from amplicon sequencing were demultiplexed according to barcode sequence and aligned against hg19 human reference genome by BWA-MEM (v0.7.12) (30). A pileup format file was generated by SAMtools (v0.

Whole-exome sequencing of the GNAQ mutation-negative patient
For patient 3, who was negative for the GNAQ-mutation, the unprocessed affected and unaffected dermis was used for whole-exome sequencing to a depth of approximately 400×. Exome capture was done using the Agilent SureSelect Human All Exon V5 kit (Agilent Technologies, Santa Clara, CA), and subsequent sequencing of the samples was done on Illumina HiSeq (Illumina, San Diego, CA), 125 bp paired end reads. The sequences were aligned to the reference human genome (hg19) by using BWA-MEM (v0.7.12) (30). The alignments were then refined by the Genome Analysis Toolkit (GATK, Version 3.4) (32-35) and PCR duplicates were marked by Picard (v1.124) (https://github.com/broadinstitute/picard). Low-quality reads were then removed from the alignments by GATK. Lowquality reads include reads without proper mate or mapping quality, or with aberrant CIGAR strings, PCR duplicates, zero or low mapping quality. The four bases were then counted on the cleaned alignment BAM file by using bam-readcount (v0.7.4) (http://github.com/genome/bam-readcount). A variant was called wherever three or more reads showed the same non-reference allele. The variant file was annotated by Annovar (v2015Jun17) (35).

Variant filtration
The annotated variants were analysed using the software FIL-TUS (36) (http://folk.uio.no/magnusv/filtus.html). Variants from affected tissue were filtered on the following criteria: alternative read count at least 15, allele frequency below 1% in the 1000 Genomes Project database (http://www.1000genomes.org) and GERP score above 5. The remaining variants were then filtered against the variants from unaffected tissue with an alternative read count over 5. This yielded a candidate list of 55 variants, which were inspected manually in IGV (https://www.broadinsti tute.org/igv/). Most variants resulted from reads of poor quality and had low base calling quality. Only two variants were found to be non-artefacts, stemming from long reads of high quality with good single base calling quality.

Cell proliferation assay
Transduced HUVECs were seeded in 96-well plates with detachable strips of eight wells at 19 × 10 3 cells/cm 2 in three replicate wells, fixing cells daily in 100 μl cold 4% paraformaldehyde (Chemi-Teknik, Oslo, Norway) at room temperature, 10 min. Cells were subsequently dried for 20 min and stored at 4 • C until the end of the experiment. Fixed cells were next stained with 70 μl freshly made 0.1% crystal violet (Apotekforeningen, Oslo, Norway) in distilled water (4 min, RT), before extensively washing them under cold tap water. Nuclear dye was eluted with 100 μl 33% acetic acid (Sigma-Aldrich, Germany) and absorbance was read at 550 nm in an Epoch microplate reader with software Gen5.0.10.1 (BioTek, Winooski, VT).

Cell migration assay
Transduced HUVECs were seeded to 24-well plates at 50 × 10 3 cells/cm 2 in at least three wells per experimental condition. The cell migration assay was performed at confluence by scraping a straight line with a sterile 200 μl pipet tip in the cell monolayer. Detached cells and debris were replaced with fresh medium before incubation at 37 • C for 6 h. Images from one region of the scraped area in each well were taken under a fluorescence microscope (Evos ® FL Cell Imagine System, Thermo Fisher Scientific, Waltham, MA) at 10× magnification at 0, 2, 4 and 6 h (Supplementary Material, Fig. S2). To determine the rate of cell migration, images were analysed using ImageJ 1.50i software (National Institutes of Health, Bethesda, MD) (ImageJ, NIH), subtracting the free area after migration from the area after scraping, according to Buachan et al. (37).

Statistical analysis
In the functional experiments, statistical comparisons were performed with the two-tailed Student's t-test after passing tests for normal distribution using GraphPad Prism 8.0.2 software. Dispersion of results was expressed as mean ± SEM or mean ± SD as appropriate and indicated in figure legends.

Supplementary Material
Supplementary Material is available at HMG online.