Somatic Mutations in Vascular Malformations of Hereditary ... · mutations in KRAS or MAP2K1.15,16...

Somatic Mutations in Vascular Malformations of Hereditary Hemorrhagic TelangiectasiaResult in Biallelic Loss of ENG or ACVRL1

Daniel Snellingsa, Carol Gallionea, Dewi Clarkb, Nicholas Vozorisb,c, Marie Faughnanb,c, Douglas Marchuka,∗

aDepartment of Molecular Genetics and Microbiology, Duke University, NC, 27710, USAbToronto HHT Centre, Division of Respirology, St. Michael’s Hospital and Li Ka Shing Knowledge Institute, Toronto, Canada

cDepartment of Medicine, University of Toronto, Toronto, Canada

Abstract

Hereditary Hemorrhagic Telangiectasia (HHT) is a Mendelian disease characterized by vascular malformations including visceralarteriovenous malformations and mucosal telangiectasia. HHT is caused by loss-of-function mutations in one of 3 genes; ENG,ACVRL1 or SMAD4 and is inherited as an autosomal dominant condition. Intriguingly, the constitutional mutation causing HHTis present throughout the body, yet the multiple vascular malformations in HHT patients occur focally, rather than manifestingas a systemic vascular defect. This disconnect between genotype and phenotype suggests that a local event is necessary for thedevelopment of vascular malformations. We investigated the hypothesis that local somatic mutations seed the formation HHT-related telangiectasia in a genetic two-hit mechanism. We identified low-frequency somatic mutations in 9/19 telangiectasia usinghigh depth next-generation sequencing. We established phase for 7 of 9 samples using long-read sequencing, which confirm that thegermline and somatic mutations in all 7 samples exist in trans configuration; consistent with a genetic two-hit mechanism. Thesecombined data suggest that biallelic loss of ENG or ACVRL1 may be a required event in the development of telangiectasia, and thatrather than haploinsufficiency, vascular malformations in HHT are caused by a Knudsonian two-hit mechanism.

Keywords: HHT, hereditary hemorrhagic telangiectasia, somatic mutation, two-hit, ENG, ACVRL1

Introduction

Hereditary Hemorrhagic Telangiectasia (HHT) is aMendelian disease characterized by the development of mul-tiple focal vascular malformations consisting of arteriovenousmalformations in visceral organs and telangiectasia in mucosaland cutaneous tissue. The genetic etiology of HHT has beenestablished and is caused by mutations in ENG1, ACVRL12,and rarely SMAD43; all of which follow an autosomal dom-inant inheritance pattern. Despite our understanding of thegenetics of and downstream pathways involved in HHT, themolecular mechanisms that initiate HHT-related vascularmalformation are poorly understood. Early studies of thefunctional consequences of HHT causal mutations establishedthat these result in the loss of function of the gene product.These findings, in combination with autosomal dominantinheritance, led to the presumption that vascular malformationsresult from haploinsufficiency of the mutated gene product.4–6

However, haploinsufficiency does not account for why HHT-related vascular malformations occur as strictly focal lesions,despite the systemic presence of the causal germline mutation.This disconnect between genotype and phenotype led to analternative long-standing hypothesis that HHT-related vascularmalformations result from a Knudsonian two-hit mechanism,where a local somatic mutation in the wild type allele of thecausal gene seeds the formation of focal lesions.

∗Corresponding AuthorEmail address: [email protected] (Douglas Marchuk)

The only published study which directly addresses the two-hit hypothesis attempted to determine whether Endoglin waspresent on the endothelial lining of arteriovenous malforma-tions from an HHT patient with a causal mutation in the corre-sponding gene ENG.7 Endoglin immunostaining was visible inthe vessel lining, albeit at low levels. The presence of Endoglinin HHT-associated vascular malformations would contradict atwo-hit mechanism, however complete loss of staining mightnot be predicted to occur, especially with the heterogeneous –and potentially mosaic – tissue of an arteriovenous malforma-tion that may only contain a minority of cells that harbor thesomatic mutation. Previous attempts to address this hypothe-sis at the DNA level have been hampered by the limitationsof past sequencing technology. The advent of next-generationsequencing has drastically increased our sensitivity for detect-ing low-frequency somatic mutations. Somatic mutations havebeen identified in a diverse array of vascular malformations8–14

including recent evidence that sporadic arteriovenous malfor-mations, not associated with HHT, harbor somatic activatingmutations in KRAS or MAP2K1.15,16 Notably, a genetic two-hit mechanism is known to contribute to Cerebral CavernousMalformations (CCM)17,18 and Capillary Malformation – Ar-teriovenous Malformation Syndrome (CM-AVM)19; like HHT,both diseases are caused by autosomal dominant loss of func-tion mutations. Here we demonstrate that HHT-related telang-iectasia contain biallelic mutations in ENG or ACVRL1, result-ing in homozygous loss of function; evidence in support of thelong-standing hypothesis that telangiectasia pathogenesis fol-

Preprint submitted to bioRxiv August 9, 2019

All rights reserved. No reuse allowed without permission. (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint. http://dx.doi.org/10.1101/731588doi: bioRxiv preprint first posted online Aug. 10, 2019;

http://dx.doi.org/10.1101/731588

lows a genetic two-hit mechanism.

Materials and Methods

Sample Collection

Individuals were enrolled in the study after giving informedconsent (approved by either the St. Michael’s Hospital IRBcommittee or the Duke University Health System IRB Com-mittee). Diagnosis of HHT was based on identification of apathogenic germline mutation, or by exhibiting at least threeof the four symptoms as per the Curacao criteria (Supp Table1).20

Telangiectasia were resected using a 3mm punch biopsy, af-ter local anesthesia (1% xylocaine with epinephrine), with stan-dard aseptic technique. Samples 6005-1 was immediately for-malin fixed (10% formalin) and paraffin embedded (FFPE), andthen shipped at room temperature. All other samples were im-mediately frozen at -80 Celsius, and then shipped on dry ice.Saliva samples were obtained using Oragene DNA saliva kits atthe time of tissue collection. Blood from patient 6003 was ob-tained during a subsequent visit, shipped at room temperature,and immediately used for RNA extraction.

DNA/RNA Extraction

DNA from telangiectasia samples was extracted using theDNeasy Blood and Tissue Kit (Qiagen). DNA from FFPE sam-ple 6005-1 was extracted using QIAamp DNA FFPE Tissue Kit(Qiagen). Genomic DNA and RNA were extracted from periph-eral blood leukocytes from patient 6003 and from a non-HHTcontrol individual using the Gentra PureGene Blood Kit (Qia-gen) and TRIzol Reagent (Invitrogen) extraction protocols, re-spectively, as per the manufacturers’ directions. Nucleotides arenumbered with the conventional nomenclature of the A in thefirst ATG as c.1.

Targeted Sequencing

To enable the detection of somatic mutations in telangiecta-sia we used a high-depth next generation sequencing strategy.Somatic mutations involved in the pathogenesis of other vas-cular malformation diseases such as cerebral cavernous malfor-mations often have a low allele frequency due to somatic mo-saicism in the malformation. Somatic mosaicism is also presentin retinal AVMs from mouse models of HHT suggesting thatlow allele frequency may be a confounder when identifying so-matic variants in telangiectasia. In addition, the telangiectasiasamples collected for this study consist of bulk biopsied tis-sue which have not been enriched for any particular cell type.Considering these potential sources of normal (non-mutant) cellcontamination, we sequenced telangiectasia to >1000x cover-age and incorporated a unique molecular identifier to enablethe detection of variants as low as 0.1% allele frequency.

Eighteen fresh-frozen samples and one FFPE sample wassequenced using a custom Agilent SureSelect panel covering 16genes implicated in various vascular malformation disorders:ENG, ACVRL1, SMAD4, BRAF, CCM2, FLT1, FLT4, GNAQ,KDR, KRAS, KRIT1, MAP2K1, NRAS, PDCD10, PIK3CA, and

PTEN. To ensure the generation of high-quality sequencing li-braries, the Agilent NGS FFPE QC Kit was used to determinethe extent of DNA degradation in the FFPE sample. Sampleswith ∆∆Cq values >2 were excluded from the study as permanufacturer recommendation. Sequencing libraries were gen-erated using the Agilent SureSelect XT HS Kit. Samples werethen pooled and sequenced on an iSeq 100 (Illumina) withpaired-end 150bp reads. All samples were sequenced to a meandepth of >1000x.

Mutation DetectionSequencing data was processed and analyzed using a cus-

tom pipeline based on the GATK best practices for somaticshort variant discovery. Briefly, after analyzing the raw datawith fastQC to ensure high quality data, the adapter sequenceswere trimmed from reads using bbduk, reads were aligned tothe hg19 human reference genome using bowtie2, duplicateswere removed based on UMI sequence using fgbio, variantswere called using MuTect2 in tumor-only mode, and variantswere annotated using snpSift. The resulting variant call file(VCF) was filtered through several steps to identify somaticmutations. To identify variants that may change the protein se-quence or impact splicing, we selected for variants that occurwithin exons or within 10bp of an exon. From this set, we re-moved variants that are present in the population at >0.01% fre-quency by comparing to 3 databases (dbSNP, 1000 Genomesproject, and the Exome Aggregation Consortium [ExAC]); asthese variants are more common than the frequency of HHT (1-5 in 10000).21 We removed any variants present in <0.02% ofsequence reads, as this is the reported technical limit of detec-tion for the SureSelect XT technology. We also removed vari-ants in regions with <100x coverage, variants with <5 support-ing reads, variants that were strand specific, and variants where<50% of alternative bases had a quality score >30 (>Q30).

Candidate somatic variants identified in the targetedsequencing data were then validated by sequencing am-plicons generated during a second, independent round ofPCR amplification. We designed primers for each sampleto specifically amplify the position of the somatic mutationand 100-200bp of flanking sequence. When possible, theprimers were designed such that they would capture theposition of both the germline and somatic mutations withina single amplicon. Each primer was synthesized with thefollowing Illumina flow cell adapter sequences such that theamplicons could be easily indexed and sequenced: Forward5’-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAG-[sequence-specific primer]-3’, Reverse 5’-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAG-[sequence-specific primer]-3’. Amplicons were prepared fromtelangiectasia DNA and constitutional DNA if available (SuppTable 1) using two rounds of PCR. The first round of PCR was25 cycles and served to amplify the target region from genomicDNA. Amplicons from the first round were purified usingAMPure XP beads (Beckman Coulter) and used for a secondround of PCR using 8 cycles to attach a sample index using theNextera XT Index Kit. Amplicons from the second round werepurified, pooled, and sequenced on an iSeq 100 with 150bp

2



http://dx.doi.org/10.1101/731588

paired-end reads to a depth of >10000x. The frequency of thesomatic mutation in telangiectasia and constitutional DNA wasdetermined using custom scripts, excluding bases <Q15.

As these amplicons were sequenced in the same run and athigh depth, it is possible for a low level of index misassignmentcausing switching of reads between samples. This could causesome reads from telangiectasia to be assigned as constitutionalreads and vice versa. To estimate the rate of index misassign-ment, we examined pairs of samples that target different ge-nomic locations and quantified the proportion of misassignedreads between these samples. Based on this we estimate thatthe rate of misassignment is 0.2 – 0.8%, relative to the sampleof origin. For example, if one sample has a somatic mutationwith a frequency of 1% and was sequenced to 100000x cover-age, then 2 – 8 reads containing the somatic mutation would bemisassigned to each of the other samples in the pool.

Establishing PhaseTo establish the phase of the somatic and germline muta-

tions we used either short-read sequencing with Illumina chem-istry, or long-read sequencing with PacBio chemistry depend-ing on the distance between the two mutations. For muta-tions <500bp apart, we generated amplicons during the valida-tion process that cover the positions of both the somatic andgermline mutations. These amplicons were sequenced on aniSeq 100 as described above.

For mutations >500bp we designed primers such thatthe resulting amplicon would cover the position of both thegermline and somatic mutations for each sample. Each primerwas synthesized with the PacBio ‘universal tag’ as follows toenable indexing and sequencing: Forward 5’- /5AmMC6/

GCAGTCGAACATGTAGCTGACTCAGGTCAC-[sequence-specific primer]-3’, Reverse 5’- /5AmMC6/

TGGATCACTTGTGCAAGCATCACATCGTAG-[sequence-specific primer]-3’. We generated amplicons spanning thesomatic and germline mutations using the LongAmp TaqDNA Polymerase Kit as per manufacturer instructions. Theseamplicons were purified with AMPure XP beads and used for asecond round of PCR to attach the sample index. This processused no more than 30 cycles of PCR total. These ampliconswere pooled and sequenced across one SMRT cell on a PacBioSequel System. The sequence reads were aligned to the hg19human genome using Minimap2 in ava-pb mode.

The single molecule resolution of these technologies al-lowed us to determine how the mutant alleles are arranged; ifthe mutations are in trans then reads will have either the somaticmutant allele or the germline mutant allele, if the mutations arein cis then reads will have either no mutant alleles or both mu-tant alleles. In total we generated mutation-spanning reads for7 telangiectasia, each with >100 reads which contained the so-matic mutation. The p-values reported for phase status werecalculated using a binomial distribution with the null hypoth-esis that a random mutation has an equal probability of cis ortrans configuration with a nearby variant.

The genomic distance between mutations varied greatlywith the closest mutations in sample 6005-1 with 26 bases be-tween mutations, and the most distant in 6001-10 with 18.7

kilobases between mutations. Of the 9 telangiectasia with iden-tified somatic mutations, the distance between mutations in6002-2, 6003-1, and 6005-1 was small enough to allow formutation-spanning reads using illumina chemistry (Table. 2).Mutation-spanning reads were generated for 6001-3, 6001-7,6001-8, and 6002-1 using PacBio chemistry. We were unableto generate amplicons spanning the mutations for 6001-1 and6001-10. The sequence downstream of the somatic mutation inthese telangiectasia contains several repetitive regions which,combined with the genomic distance and limited quantity of in-put DNA may have contributed to PCR failure.

One notable confounder in this analysis is the generation ofchimeric reads resulting from template switching during PCR.The generation of chimeric reads is known to interfere withamplicon-based haplotype phasing by switching a variant fromone strand to another, potentially generating new haplotypesnot present in the original sample.22 In practice, chimeric readsrandomize the arrangement of the somatic and germline mu-tations. The frequency of chimeric arrangements is highly de-pendent on the distance between mutations and the number ofPCR cycles used to make the amplicons. To reduce the num-ber of chimeric reads in our libraries we used no more than 30cycles for amplification. A previous study reports a chimericarrangement frequency of 6.5% for 29 cycles of amplificationfor mutations 9kbp apart.22 Chimeric reads may account for thevery few discordant reads in some of our samples, as shown inTable 2. Nonetheless, these were so minor in comparison to thegreat majority of the reads that phase could be unequivocallydetermined.

in vitro Splicing

A 3.8kb fragment of ACVRL1 genomic DNA spanning from431 bases upstream of exon 3 to 215 bases downstream fromexon 8 was amplified and this entire insert was ligated into theMCS of pSPL3, a splicing vector.23 Clones were sequenced toensure that no PCR-induced errors were present in the exonsand adjacent intronic regions of the insert. The specific mu-tation, c.625+4A>T, was introduced using site directed mu-tagenesis and again, clones were sequenced to verify that theonly sequence difference was at the intended site. Plasmid DNAfrom empty vector, wild type (control) vector and mutation-containing vector were transfected into HEK293T cells usingLipofectamine 3000 (ThermoFisher Scientific), incubated for24 hours, and then the RNA was extracted using TRIzol andDirect-zol RNA miniprep kit (Zymo Research).

Reverse-Transcription PCR

RNA extracted from patients’ blood and from transfectedcells was used as template for cDNA synthesis using the Max-ima H Minus First Strand cDNA kit (ThermoFisher Scientific).An RT primer in ACVRL1 exon 8 was used in the RNA frompatients’ blood while a vector-specific RT primer was used forthe RNA from transfected cells to ensure that only RNA fromthe transfected vectors was being used as template for cDNAsynthesis. cDNA from the patients was PCR amplified usingprimers in exons 3 and 6 while cDNA from the transfected cells

3



http://dx.doi.org/10.1101/731588

was PCR amplified using primers in exons 3 and 8. PCR re-actions were run on 1% agarose gels, the bands excised andSanger sequenced.

Results

To determine whether a genetic two-hit mechanism under-lies HHT pathogenesis, we tested three underlying expectationsof the two-hit mechanism: 1) telangiectasia contain a somaticmutation in the same gene as a germline mutation which causesHHT, 2) the somatic and germline mutations are biallelic, and3) both mutations result in loss-of-function.

Telangiectasia Harbor a Somatic Mutation in ENG or ACVRL1We used capture-based library preparations to sequence 19

telangiectasia for the three HHT causal genes (ENG, ACVRL1,and SMAD4) and 13 other vascular malformation-related genes(see Methods for identity of genes). The 13 non-HHT vascularmalformation genes were chosen with the possibility that theyalso might harbor somatic mutations, but primarily to serve ascontrol genes, since the two-hit mechanism requires a mutationin the corresponding HHT gene harboring the causal germlinemutation. Somatic mutations in these other genes may or maynot contribute to HHT pathogenesis, but absence of a somaticmutation in the casual HHT gene, would violate the first expec-tation of the genetic two-hit mechanism.

In each telangiectasia we identified a pathogenic germlinemutation in either ENG or ACVRL1. Although in most casesthe patient’s germline mutation was already known from clin-ical diagnostic sequencing, we intentionally remained blindedto this information until after our own sequence analysis of thetissue samples. In 6003-1, the patient harbors a silent germlinemutation which was found by the clinical lab and noted as avariant of unknown significance. Below we show that this vari-ant is indeed the pathogenic germline variant in this patient.

We used the MuTect2 variant caller to detect variantspresent in the sequence data. To identify candidate somatic mu-tations, we removed variants based on several stringent filteringcriteria including briefly; intronic or intergenic variants, popu-lation frequency >0.01%, <0.1% or <5 total supporting reads,low coverage, strand specificity, and low base quality scores.

To validate or refute the authenticity of each candidate so-matic mutation we performed an independent round of ampli-fication using primers flanking each putative variant positionfor each sample, and sequenced to >10000x coverage. In eachtissue sample, the identical somatic variant was re-identified.Thus, these variants were bona fide somatic mutations exist-ing in the telangiectatic tissue. In total, we identified somaticvariants in 9 of 19 telangiectasia; 5 in ENG(NM 001114753) 4in ACVRL1(NM 000020) (Table 1) (Fig. 1) (See Methods). Ineach case, the somatic mutation was found in the same gene asthe pathogenic germline mutation. Somatic mutations were notfound in any of the other 15 genes sequenced, not even in one ofthe other HHT casual genes. Importantly, no telangiectasia har-bored more than a single somatic mutation. The lack of muta-tional noise suggests these mutations are pathobiologically sig-nificant. Importantly, all are consistent with strong mutations;

Figure 1: Low Frequency Somatic Mutations Detected in Telangiectasia.Visualization with IGV of deep next-generation sequencing data for one rep-resentative sample with a somatic mutation in ACVRL1. The somatic mutationin 6003-1, an A>T point mutation, is present at low frequency in DNA fromtelangiectatic tissue used for (A) discovery (capture-based sequencing) and (B)validation (amplicon-based sequencing). The mutation is below the level of se-quencing noise in (C) constitutional DNA (amplicon-based sequencing) con-firming the mutation is somatic.

five of the variants are small indels resulting in a frameshift,three are in-frame indels, and one is a point mutation 4 basesafter an exon-intron boundary that is predicted to impact RNAsplicing.

Although these variants fall well below the 50% allele fre-quency expected for germline variants, it is formally possiblethat these variants exist constitutionally as very rare, somaticmosaic variation in the patients. We next investigated whetherthese somatic mutations were present in constitutional DNAfrom the patient. A source of constitutional DNA (buccal ep-ithelial cells from saliva) was available for seven of the ninemutation-positive samples, and in each we find that the somaticmutation was completely absent or present at a level no higherthan technical sequencing noise in that sample (see Methods).Saliva was not available for patient 6001, but we obtained andsequenced DNA from multiple telangiectasia collected fromthis same patient. This enabled us to determine whether anyof the five somatic mutations that we identified in individualsamples was present in tissue of near identical pathobiologyfrom the same individual; compared with saliva as a control,this is a more powerful test for somatic mosaicism. We found

4



http://dx.doi.org/10.1101/731588

Table 1. Summary of Somatic Mutation Discovery and Validation Sequencing

Sample Germline Mutation Somatic Mutation Discovery Readsa Validation Readsa Constitutional Readsa

6001-1 ENGc.1080 1083delGACAfsX8

ENGc.293 304del

33/1318 (2.5%) 1067/100268 (1.1%) 0/26462 (0%)b

6001-3 same as above ENGc.1195 1196delAGfsX2

5/1080 (0.46%) 723/115963 (0.62%) 0/24357 (0%)b

6001-7 same as above ENGc.1237 1238insCAfsX7

27/5127 (0.53%) 341/115570 (0.30%) 0/23066 (0%)b

6001-8 same as above ENGc.578delCinsTGCGp.T193MR

111/4845 (2.3%) 1142/142572 (0.80%) 0/21315 (0%)b

6001-10 same as above ENGc.205delGfsX6

33/3389 (1.0%) 3575/326894 (1.1%) 0/22098 (0%)b

6001-* same as above NF

6002-1 ACVRL1c.1451G>A p.R484Q

ACVRL1c.349delGinsTTfsX52

20/2217 (0.90%) 309/24018 (1.3%) 0/65818 (0%)

6002-2 same as above ACVRL1c.1378-3 1402del19ins9

26/1649 (1.6%) 3189/202550 (1.6%) 6/155855 (0.0038%)

6003-1 ACVRL1c.474A>T p.G158G

ACVRL1c.625+4A>T

101/3392 (3.0%) 372/16303 (2.3%) 2/38924 (0.0051%)

6004-1 ACVRL1c.1232G>A p.R411Q

NF

6004-2 same as above NF

6005-1 ACVRL1c.1232G>A p.R411Q

ACVRL1c.1206delCfsX12

133/1664 (8.0%) 2671/189690 (1.4%) N/A

*Eight additional telangiectasia from patient 6001 with no identified somatic mutation

For multiple telangiectasia collected from one individual the sample ID is listed as (Patient#)-(Telangiectasia#, NF = None Found)aAllele frequency in other telangiectasia from 6001

that the somatic mutations identified in five of the telangiecta-sia for which we identified a mutation were entirely absent in allother telangiectasia from this same individual. Finally, Sample6005-1 is a single archived FFPE telangiectasia and no sourceof constitutional DNA is available. In total, we found that 9 of19 telangiectasia harbor a somatic mutation specifically in thesame gene as a pathogenic germline mutation and that thesemutations are not present constitutionally. The presence of so-matic mutations in telangiectasia fulfills the 1st expectation ofthe genetic two-hit mechanism.

Somatic and Germline Mutations are Biallelic

The 2nd expectation of the genetic two-hit mechanism isthat the somatic and germline mutations are biallelic; suchthat the somatic mutation occurs on the wild-type allele ofthe causal gene, in trans with the germline mutation. To deter-mine if the mutations are biallelic we examined whether theywere arranged in a cis or trans configuration by sequencingamplicons that cover the nucleotide positions of both somaticand germline mutations in a single molecule. The ampliconswere sequenced with either short-read (Illumina) or long-read(PacBio) chemistry, depending on the amplicon size, in order togenerate reads that would span the two mutations. In contrast totraditional Sanger sequencing which measures the populationaverage at each position, both Illumina and PacBio chemistriesoutput sequences of single DNA molecules. In total we gen-erated mutation-spanning reads for 7 telangiectasia, each with

Table 2. Phase of Somatic and Germline Mutation Pairs

Sample Total Reads Trans Reads Cis Reads P-value

6001-1 N/A N/A N/A N/A

6001-3 112 112 (100%)a 0 (0%) 1.9e-34

6001-7 155 153 (98.7%)a 2 (1.3%) 2.6e-43

6001-8 593 590 (99.5%)a 3 (0.5%) 1.0e-171

6001-10 N/A N/A N/A N/A

6002-1 125 120 (96.0%)a 5 (4.0%) 5.5e-30

6002-2 3189 3160 (99.0%)bc 29 (1.0%) 4.2e-890

6003-1 372 364 (97.8%)bc 6 (1.4%) 1.4e-99

6005-1 2671 2653 (99.3%)bc 18 (0.7%) 6.3e-759

aReads generated with PacBio long-read chemistry)bReads generated with Illumina short-read chemistrycThese reads were also used for validation shown in Table 1

more than 100 reads that contained the somatic mutation. Fromthese mutation-spanning reads we established that >95% ofreads with the somatic mutation possessed the wild type al-lele at the position of the germline mutation, showing that all 7mutation pairs are in trans configuration (Table 2) (Fig. 2A-B).Any two variants in a chromosome must be arranged in cis ortrans with an equal probability of either arrangement. Consid-ering this, our observation that 7/7 mutation pairs are arrangedin trans corresponds to a p-value of 0.008 demonstrating signif-

5



http://dx.doi.org/10.1101/731588

Figure 2: Somatic and Germline Mutations are Biallelic. IGV visualization of two samples showing both methods of establishing phase. Each panel shows readsfrom the initial discovery sequencing and the reads used to establish phase. (A) Somatic and germline mutations in 6003-1 are both A>T point mutations andhighlighted by the blue and green regions respectively. Since the distance between these mutations is relatively small (357bp) phase was established using Illuminashort-reads (also used for validation). Black lines between reads denote read pairs, showing that both reads originate from a single molecule of DNA. Each moleculewith the somatic mutation contains the wild-type allele at the germline mutation position proving the mutations are biallelic. (B) Somatic and germline mutationsin 6001-3, both small deletions. The genomic distance between these mutations is 4377bp. In long reads that span the two mutations, each read with the somaticmutation contains the wild-type allele at the position of the germline mutation.

icant bias towards a trans configuration. These data show thatthe somatic and germline mutations are biallelic, fulfilling the2nd expectation of the genetic two-hit mechanism.

Mutations are Consistent with Homozygous Loss of Function

The 3rd expectation of the genetic two-hit mechanism isthat the biallelic somatic and germline variants both result inloss of function. Due to the functional studies and extensiveallelic series of mutations in each of the HHT genes, HHT isknown to be caused by loss of function mutations. The germlinemutation in 3 of the 5 patients in this study has been iden-tified previously in an individual with HHT and are reportedin ClinVar (6001:VCV000213214.2, 6002:VCV000212796.2,6005:VCV000008243.2). There are also several publicationssupporting the pathogenicity of these mutations.24–28 These

are all therefore bona fide loss of function mutations. Thegermline mutation in 6003-1, a silent mutation in ACVRL1exon 4, has been identified before in an individual with HHT,however it was classified as a variant of unknown significance(VUS). We used the in silico tool Human Splicing Finder3.129 to analyze this variant and found that it was predictedto both disrupt an exonic splice enhancer and create an in-ternal splice donor site, potentially activating a cryptic splicesite. Based on this prediction, we extracted RNA from pe-ripheral blood leukocytes of patient 6003 and a control in-dividual and performed RT-PCR to examine the splicing ofACVRL1 transcripts. RNA from patient 6003 shows a newsplice variant that is not present in control wild-type RNA(Fig. 3B). As predicted by the Splice Finder program, theaberrant transcript is spliced precisely at the internal splice

6



http://dx.doi.org/10.1101/731588

donor created by the mutation. The resulting transcript is miss-ing the portion of exon 4 downstream of the germline mutationand skips exon 5 resulting in the in-frame deletion of 52 aminoacids (Fig. 3D-E). This deleted region contains several codonswith known pathogenic missense mutations, suggesting that the52 amino acid deletion would likely also result in loss of func-tion. It is possible that the skipping of exon 5 is due to alter-native splicing observed only in peripheral blood leukocytes,rather than a result of the mutation. If exon 5 is retained, themutation would then generate a protein lacking 17 amino acidresidues from exon 4 but then be frameshifted for the remain-der of the transcript. With this evidence, all of the identifiedgermline mutations meet the American College of Medical Ge-netics (ACMG) criteria for pathogenic mutations30, fulfillingthe first half of the 3rd expectation of the genetic two-hit mech-anism.

In contrast to the germline mutation, many of which havebeen previously identified, the somatic mutations we iden-tified are all novel. The ACMG guidelines for establishingpathogenicity are not applicable to somatic variants, howeverseveral lines of evidence support that all of the somatic muta-tions result in loss of function. Five of the nine somatic mu-tations result in a frameshift; all resulting in premature ter-mination codons which would generate transcripts susceptibleto nonsense-mediated decay. Frameshift mutations in ENG orACVRL1 are the most common mechanism for loss of function

Table 3. Evidence that Identified Germline and Somatic Mutations Result in Loss of Function

Sample Germline Mutation Somatic Mutation

6001-1 Frameshift: ACMG PVS16 supporting publications

In-Frame Deletion (-4 residues)PROVEAN: Deleterious (-14.903)deletes region in β-sheetgnomAD AF: 0

6001-3 same as above Frameshiftcommon ENG LOF mechanism, expect NMD


6001-8 same as above In-Frame Delins (-1 +2 residues)2 pathogenic in-frame indels overlapping this codon 31,32

PROVEAN: Deleterious (-6.106)gnomAD AF: 0


6002-2 same as above In-Frame Delins (-7 +1 residues)PROVEAN: Deleterious (-25.903)deletes region in α-helixgnomAD AF: 0

6003-1 Cryptic Splice Site: ACMG PS3in silico predicted to activatecryptic sitein vitro evidence (Fig. 4)

Splice Sitein silico predicted to disrupt donor sitegnomAD AF: 0

6005-1 Missense: ACMG PS116 supporting publications

Frameshiftcommon ACVRL1 LOF mechanism, expect NMD

AF = Allele Frequency; NMD = Nonsense Mediated Decay

Germline variant classification according to ACMG guidelines, not applicable to somatic variants 30

PVS1 = Very strong evidence for pathogenicity; PS1-4 = Strong evidence

PROVEAN scores below -2.5 are predicted deleterious

leading to HHT. Based on this, the5 somatic frameshift mutations likelyresult in loss of function. Other thanframeshift mutations, the other 4 so-matic mutations we identified con-sisted of 3 in-frame deletions and1 intronic mutation predicted to im-pact splicing. These 4 mutations arenot present in the genome aggre-gation database (gnomAD) showingthat the population allele frequencyof these variants is extremely low orzero. For the somatic in-frame dele-tion mutation found in 6001-8, thereare two reports of different in-framedeletions with overlap at this posi-tion which are known to cause HHT,suggesting that the somatic deletionin 6001-8 is likely to result in lossof function. The somatic mutations in6001-1 and 6002-2 also result in in-frame deletions, which delete 4 and7 amino acids respectively. Compar-ing the crystal structures of ENG andACVRL1 we determined that the so-matic mutations in 6001-1 and 6002-2 delete portions of a beta strandand helix respectively, potentially im-pacting protein folding (Table 3). Thein silico tool PROVEAN was usedto predict how the protein would tol-

erate these deletions. The threshold of -2.5 or lower (morenegative) is considered a deleterious change. The scores for6001-8, 6001-1, and 6002-2 were -6.106, -14.903, and -25.903respectively, strongly suggesting that all three are deleteri-ous. The remaining somatic mutation is intronic, and occurs 4nucleotides from the exon-intron boundary. We used HumanSplicing Finder 3.1 to predict the effect of this variant on splic-ing, and found that it is likely to disrupt the acceptor site. Thisprediction was confirmed by RT-PCR using an in vitro splicingconstruct which revealed that the somatic mutation prevents theformation of full-length ACVRL1 transcripts (Fig. 3C). In sum-mary we present evidence supporting that the biallelic germlineand somatic mutations all likely result in loss of function, ful-filling the 3rd expectation of the genetic two-hit mechanism.

Telangiectasia from the Same Individual Harbor Unique So-matic Mutations

We next sought to determine whether mutant cells indifferent telangiectasia derive from a somatic mutation in acommon ancestor cell, or whether the mutant cell popula-tion in each telangiectasia derives from an independent so-matic mutation event. To test this, we examined the somaticmutations present in multiple telangiectasia from single in-dividuals. In patient 6001, for which we had obtained 13different telangiectasia, we identified a somatic mutation in

7



http://dx.doi.org/10.1101/731588

5 telangiectasia tissue samples. In each case the somatic mu-tation in each telangiectasia was unique. Likewise, for patient6002, for which we had two telangiectasia, we identified aunique somatic mutation in each (Fig. 4). These results are con-sistent with independent mutation events rather than the so-

Figure 4: Germline and Somatic Mutations in 6003-1 Both Cause Aberrant Splicing of ACVRL1 RNA.(A) The gene structure of ACVRL1 exons 3-6 marked with the location of the germline and somatic mutationsfound in 6003-1. (B) RT-PCR showing ACVRL1 transcripts from blood leukocytes taken from wild-type (WT)and patient blood containing the germline mutation. The labeled bands were excised and sequenced. Full lengthtranscript (Band A) is present in both the control and the patient although the level in the patient is greatlyreduced. Band B in the control contains complete exons 3, 4 and 6. This splice variant has been seen previouslyand differs from band C in the patient which splices from the newly created splice acceptor site within exon 4directly to exon 6. (C) As the somatic mutation is only present in 3.0% of reads it would be challenging to detectmis-spliced RNA from the biopsied tissue. Therefore, we inserted wild-type and mutant sequence of ACVRL1into an in vitro splicing vector, pSPL3-ACVRL1, and used RT-PCR to visualize the impact of the mutation onsplicing. Only WT band A shows the full-length transcript containing exon 5; there is no corresponding full-length transcript from the plasmid containing the somatic mutation. (D) Exon structure of ACVRL1 transcriptsdetermined by sequencing the excised bands. (E) Sequence of DNA showing the nature of the germline mutation.In ACVRL1 transcripts containing the germline mutation, exon 4 is shortened due to the activation of a crypticsplice site.

matic mutation occurring in a pro-genitor cell or clonality due to ametastasis from a single initial lesion.

Disscussion

In this study we present strongevidence that vascular malformationsassociated with HHT, specifically,cutaneous telangiectasia, follow a ge-netic two-hit mechanism of patho-genesis. HHT is also associated witharteriovenous malformations in lung,liver, brain and the gastrointestinaltract, but these tissues are not avail-able for prospective collection. Wepostulate that the visceral, deepervascular malformations that occur inHHT also follow this two-hit mecha-nism.

The two-hit hypothesis forHHT pathogenesis has persisted fordecades without evidence, but theselow frequency somatic mutationsare challenging to identify usingtraditional sequencing methods. Theonly published study to address thistopic employed immunohistochemi-cal staining in an attempt to identifyendothelial cells lacking staining inthe lining of HHT-related arteriove-

nous malformations.33 However, the absence of staining as aproxy for the gain of a mutation could be difficult to discover,especially if only a fraction of a cells would exhibit this lack ofsignal.

Using deep next-generation sequencing with unique molec-ular identifiers we successfully identified somatic mutations in

Figure 3: Multiple Telangiectasia from Single Individuals Contain Unique Somatic Mutations. Schematic representation of exons in ENG and ACVRL1 withgermline and somatic mutations identified in (A) 5 telangiectasia collected from patient 6001 and (B) 2 telangiectasia collected from patient 6002. In each panel thecommon germline mutation is listed above the gene and somatic mutations in each telangiectasia below the gene. Gene structure and mutation position are to scale.

8



http://dx.doi.org/10.1101/731588

multiple telangiectasia from different HHT patients. The so-matic mutations we identified were present at frequencies rang-ing between 0.46% and 8.0% in the tissue, with an average of2.3%. The low allele frequency is likely a result of two maincontributing factors; the presence of normal tissue in the skinbiopsy, and somatic mosaicism within the telangiectasia. Thetelangiectasias in this study were sampled as skin punch biop-sies and although some of the surrounding skin tissue was re-moved before DNA extraction, an undetermined amount of nor-mal tissue invariably remained. However, after removing thesurrounding tissue, the enrichment for the vascular componentof the tissue was subjectively greater than the often low so-matic mutation allele frequency might suggest. We posit thata second explanation for the low mutant allele frequency is thattelangiectasia are mosaic for the somatic mutation. This agreeswith existing data from mouse models of HHT showing thatinduced retinal AVMs are mosaic: consisting of both heterozy-gous and homozygous null cells.34 It is also consistent with theheterogeneity seen in Cerebral Cavernous Malformations35 andthe low mutant allele frequency in somatic mutations of othervascular malformation disorders.8–18 Vascular malformations,similar to tumors in cancer, appear to be seeded by somaticmutations; however unlike tumors, vascular malformations donot appear to consist of pure populations of clonally expandedmutant cells, but contain a substantial percentage of unmutatedcells.

In addition to the data presented here, the two-hit hypothe-sis of HHT-related vascular malformations is supported by ev-idence from mouse models of the disease. Whereas constitu-tional loss of both copies of Eng or Acvrl1 in mice is embry-onic lethal, mice heterozygous for constitutional deletion of ei-ther gene show extremely mild phenotypes with relatively few,if any, detectable vascular malformations.36,37 Robust mousemodels of HHT that recapitulate vascular malformation phe-notypes require the use of Cre-Lox technology to delete bothcopies of Eng in a temporally controlled (postnatal), cell-typespecific (endothelial cells) manner.

In these mouse models of HHT it is also required that bial-lelic KO of the HHT gene occur in endothelial cells. Severalgroups have experimented with expressing Cre recombinase indifferent vascular-related cell types including pericytes (NG2-Cre), vascular smooth muscle cells (Myh11-Cre), and endothe-lial cells (Scl-Cre & Pdgfb-Cre); however, mice only developvascular malformations when Cre is expressed in endothelialcells.38–41 In addition, human ENG and ACVRL1 are expressedprimarily in endothelial cells, supporting that the somatic mu-tation must occur in an endothelial cell to seed a malformation.We have attempted to confirm that the somatic mutations weidentified in human lesions occur in the endothelium by us-ing laser capture microdissection, however these efforts havebeen hampered by the small quantity of tissue in telangiectasiabiopsies and the difficulty of isolating a single layer of cells bymicrodissection. This question may be more easily addressedin larger arteriovenous malformations; however, these sampleshave been thus far inaccessible due to the rarity of their removalin HHT patients.

Interestingly, in addition to the local requirement for loss of

both alleles, vascular malformations in this model only developafter injury such as an ear punch, or by VEGF injection.42 Thisrequirement of an angiogenic stimulus is consistent in mousemodels for all HHT genotypes: Eng, Acvrl1 and Smad4.43 Therequirement for knockout of both copies of the gene supportsthe genetic two-hit mechanism we describe here. In addition,the necessity for an angiogenic stimulus suggests that loss ofboth copies of the relevant HHT gene is necessary, but not suf-ficient, for the development of the vascular malformation.

We might have expected to find somatic mutations in everytelangiectasia, however we only found somatic mutations in 9of the 19 we sequenced. The next-generation sequencing strat-egy we employ for discovering somatic mutations is extremelysensitive for the detection of point mutations and small indels.However, there are several other types of genetic alterationsthat would result in biallelic loss of function due to loss of het-erozygosity (LOH). LOH is a common occurrence in many tu-mors in cancer, and is a predominant mechanism of somaticloss/mutation. LOH can occur due to a variety of genetic mech-anisms: large deletions, chromosome loss, and mitotic recom-bination. Given the apparent capacity for even a low fraction ofsomatically mutant cells to initiate the vascular malformation,it follows that the same would be true for LOH-associated mu-tational events; therefore, these mutations would appear insteadas allelic imbalance rather than outright LOH. But if the levelof allelic imbalance is as low as the frequency of somatic mu-tations we have observed in this study, we might expect linkedmarker haplotype ratios in the range of 48% to 52% at nearbymarkers; this slight and even trivial imbalance would be diffi-cult if not impossible to detect and validate. It is also possiblethat non-genetic mechanisms such as loss of expression due toepigenetic silencing account for biallelic loss of function. Thisprocess, like the LOH associated events, would not be detectedby our sequencing strategy; a problem that is only exacerbatedby low allele frequency. Thus, it may not be surprising that weidentified a somatic mutation in approximately only 19 of theVMs that were sequenced. We postulate it is highly likely thatall of the 10 telangiectasia with no identified somatic mutationhave biallelic loss by one of these other mechanisms.

One consequence of the genetic two-hit mechanism thatmight appear to be improbable is that, if true, a new somaticmutation must occur in every one of the numerous vascularmalformations in HHT patients. For example, some HHT pa-tients have dozens or more visible telangiectasia on the skinand mucocutaneous surfaces alone.44–46 An attractive hypothe-sis to reconcile this conundrum would be if a somatic mutationfirst occurs in a circulating progenitor cell which then prolifer-ates and seeds the formation of multiple telangiectasia. There isprecedence for this mechanism in another vascular malforma-tion syndrome, Blue Rubber Bleb Nevus syndrome. These pa-tients display multiple small vascular lesions which harbor anidentical somatic double-mutation in the TIE2 gene. These vas-cular lesions appear to be anatomically-dispersed clones arisingfrom an original, dominant, large lesion.9 In HHT-related vas-cular malformations, we report evidence that contradicts thishypothesis: different telangiectasia collected from the same in-dividual harbor different, unique somatic mutations. This ob-

9



http://dx.doi.org/10.1101/731588

servation does not exclude the possibility that circulating cellsmay in some cases spread telangiectasia, however the data thusfar suggests that the primary mechanism is independent somaticmutation events.

The dilemma of the requirement for numerous indepen-dent somatic mutation events in a single gene can be resolvedwith a probabilistic argument. Considering the size of the hu-man genome (3.23•109 bp), the probability that a random so-matic mutation occurs in the coding sequence of ENG (3201bp)in trans (50%with a pathogenic ENG germline mutation is0.00005%. Compounding this value with empirical evidencethat single cells have anywhere from 100-1500 somatic muta-tions per cell47–49, and an estimate that 5.66% of exonic somaticmutations result in LOF47; we calculate a conservative estimatethat 0.00028% of cell have biallelic LOF ENG mutations. Anadult human has at least 6•1011 endothelial cells50, thereforewe estimate that an individual with HHT and a germline muta-tion in ENG has biallelic LOF ENG mutations in 1.5 millionendothelial cells. It is clear that each cell with ENG biallelicLOF does not result in vascular malformation as individualswith HHT have at most hundreds, not millions, of telangiecta-sia. This disparity is consistent with the idea that telangiectasiaonly form under very specific conditions: likely that the somaticmutation must occur in a specific type of vascular bed, in an en-dothelial cell, and followed by local angiogenic stimulus.

Our samples consisted of telangiectasia from HHT patientsfrom an HHT Centre of Excellence in Toronto, Canada. Thisgroup of patients had germline mutations in either ENG orACVRL1, and somatic mutations were identified in telangiecta-sia from both genotypes. Mutations in SMAD4 cause the com-bined syndrome HHT and Juvenile Polyposis (JP-HHT) , how-ever individuals with SMAD4 mutations only account for 2% ofHHT cases.3 Unfortunately, no individuals with JP-HHT werepresent in our cohort, but we believe it is likely that JP-HHT-related telangiectasia follow an identical genetic two-hit mech-anism, resulting from somatic mutations in SMAD4.

As HHT is caused by germline mutations in any one of 3genes, an interesting question is whether the compound effectof a LOF mutation in two different causal genes could drivepathogenesis (e.g. a germline mutation in ENG and a somaticmutation in ACVRL1). However, thus far all of the somatic mu-tations we identified in HHT-related telangiectasia occur in thesame gene as that harboring the germline mutation. This hu-man genetic data is consistent with the observation that micewith combined deficiency for one allele each of Acvrl1 and Engare fertile, viable and are not teeming with vascular malforma-tions51 (Srinivasan and Marchuk, unpublished), as might be ex-pected if trans-heterozygosity of mutations in these two HHT-genes could initiate vascular malformation development.

A genetic two-hit mechanism for HHT pathogenesis hastherapeutic implications. Certain efforts to develop therapies forHHT assume a model of haploinsufficiency of the relevant HHTgene. These strategies attempt to increase the amount of the af-fected gene product by increasing the level of transcript/proteinarising from the wild-type allele.52 We show here that somefraction of cells within the malformation do not possess a wild-type allele. Even if expression in surrounding heterozygous

cells could be increased, the null cells would remain devoid ofprotein, suggesting that this avenue of therapy may be ineffec-tive. By contrast, a more effective strategy may be gene replace-ment, reintroducing a fully wild-type allele into the mutatedcells53, as this would simultaneously provide an extra copy ofthe gene to both the heterozygous and null cells. This strategyis particularly attractive as it might inhibit new VM formationby adding back a second wild-type copy of the mutated gene.

Acknowledgments

We thank all patients and their families who participatedin this study. We thank Dr. Nicolas Devos and the Duke Se-quencing and Genomic Technologies Core for assistance de-signing experiments and sequencing. This study was supportedby U.S. Department of Defense grant (W81XWH-17-1-0429)and a Fondation Leducq Transatlantic Network of ExcellenceGrant in Neurovascular Disease (17 CVD 03). MEF also re-ceived financial support from the Nelson Arthur Hyland Foun-dation.

Declaration of Interests

The authors declare no competing interests.

References

[1] K. A. McAllister, K. M. Grogg, D. W. Johnson, C. J. Gallione, M. A.Baldwin, C. E. Jackson, E. A. Helmbold, D. S. Markel, W. C. McKinnon,J. Murrell, et al., Endoglin, a tgf-beta binding protein of endothelial cells,is the gene for hereditary haemorrhagic telangiectasia type 1, Nat Genet8 (4) (1994) 345–51. doi:10.1038/ng1294-345.URL https://www.ncbi.nlm.nih.gov/pubmed/7894484

[2] D. W. Johnson, J. N. Berg, M. A. Baldwin, C. J. Gallione, I. Marondel,S. J. Yoon, T. T. Stenzel, M. Speer, M. A. Pericak-Vance, A. Diamond,A. E. Guttmacher, C. E. Jackson, L. Attisano, R. Kucherlapati, M. E. Por-teous, D. A. Marchuk, Mutations in the activin receptor-like kinase 1 genein hereditary haemorrhagic telangiectasia type 2, Nat Genet 13 (2) (1996)189–95. doi:10.1038/ng0696-189.URL https://www.ncbi.nlm.nih.gov/pubmed/8640225

[3] C. J. Gallione, G. M. Repetto, E. Legius, A. K. Rustgi, S. L. Schelley,S. Tejpar, G. Mitchell, E. Drouin, C. J. Westermann, D. A. Marchuk,A combined syndrome of juvenile polyposis and hereditary haemor-rhagic telangiectasia associated with mutations in madh4 (smad4), Lancet363 (9412) (2004) 852–9. doi:10.1016/S0140-6736(04)15732-2.URL https://www.ncbi.nlm.nih.gov/pubmed/15031030

[4] N. Pece, S. Vera, U. Cymerman, J. White, R. I., J. L. Wrana, M. Letarte,Mutant endoglin in hereditary hemorrhagic telangiectasia type 1 is tran-siently expressed intracellularly and is not a dominant negative, J ClinInvest 100 (10) (1997) 2568–79. doi:10.1172/JCI119800.URL https://www.ncbi.nlm.nih.gov/pubmed/9366572

[5] S. A. Abdalla, N. Pece-Barbara, S. Vera, E. Tapia, E. Paez, C. Bernabeu,M. Letarte, Analysis of alk-1 and endoglin in newborns from familieswith hereditary hemorrhagic telangiectasia type 2, Hum Mol Genet 9 (8)(2000) 1227–37. doi:10.1093/hmg/9.8.1227.URL https://www.ncbi.nlm.nih.gov/pubmed/10767348

[6] R. Ola, S. H. Kunzel, F. Zhang, G. Genet, R. Chakraborty,L. Pibouin-Fragner, K. Martin, W. Sessa, A. Dubrac, A. Eichmann,Smad4 prevents flow induced arteriovenous malformations by in-hibiting casein kinase 2, Circulation 138 (21) (2018) 2379–2394.doi:10.1161/CIRCULATIONAHA.118.033842.URL https://www.ncbi.nlm.nih.gov/pubmed/29976569

10



http://dx.doi.org/10.1101/731588

[7] N. Pece-Barbara, U. Cymerman, S. Vera, D. A. Marchuk, M. Letarte,Expression analysis of four endoglin missense mutations suggests thathaploinsufficiency is the predominant mechanism for hereditary hemor-rhagic telangiectasia type 1, Hum Mol Genet 8 (12) (1999) 2171–81.doi:10.1093/hmg/8.12.2171.URL https://www.ncbi.nlm.nih.gov/pubmed/10545596

[8] L. Al-Olabi, S. Polubothu, K. Dowsett, K. A. Andrews, P. Stadnik, A. P.Joseph, R. Knox, A. Pittman, G. Clark, W. Baird, N. Bulstrode, M. Glover,K. Gordon, D. Hargrave, S. M. Huson, T. S. Jacques, G. James, H. Kon-dolf, L. Kangesu, K. M. Keppler-Noreuil, A. Khan, M. J. Lindhurst,M. Lipson, S. Mansour, J. O’Hara, C. Mahon, A. Mosica, C. Moss,A. Murthy, J. Ong, V. E. Parker, J. B. Riviere, J. C. Sapp, N. J. Sebire,R. Shah, B. Sivakumar, A. Thomas, A. Virasami, R. Waelchli, Z. Zeng,L. G. Biesecker, A. Barnacle, M. Topf, R. K. Semple, E. E. Patton, V. A.Kinsler, Mosaic ras/mapk variants cause sporadic vascular malformationswhich respond to targeted therapy, J Clin Invest 128 (4) (2018) 1496–1508. doi:10.1172/JCI98589.URL https://www.ncbi.nlm.nih.gov/pubmed/29461977

[9] J. Soblet, J. Kangas, M. Natynki, A. Mendola, R. Helaers, M. Uebelhoer,M. Kaakinen, M. Cordisco, A. Dompmartin, O. Enjolras, S. Holden, A. D.Irvine, L. Kangesu, C. Leaute-Labreze, A. Lanoel, Z. Lokmic, S. Maas,M. A. McAleer, A. Penington, P. Rieu, S. Syed, C. van der Vleuten,R. Watson, S. J. Fishman, J. B. Mulliken, L. Eklund, N. Limaye, L. M.Boon, M. Vikkula, Blue rubber bleb nevus (brbn) syndrome is caused bysomatic tek (tie2) mutations, J Invest Dermatol 137 (1) (2017) 207–216.doi:10.1016/j.jid.2016.07.034.URL https://www.ncbi.nlm.nih.gov/pubmed/27519652

[10] N. Limaye, J. Kangas, A. Mendola, C. Godfraind, M. J. Schlogel,R. Helaers, L. Eklund, L. M. Boon, M. Vikkula, Somatic activating pik3camutations cause venous malformation, Am J Hum Genet 97 (6) (2015)914–21. doi:10.1016/j.ajhg.2015.11.011.URL https://www.ncbi.nlm.nih.gov/pubmed/26637981

[11] N. Limaye, V. Wouters, M. Uebelhoer, M. Tuominen, R. Wirkkala, J. B.Mulliken, L. Eklund, L. M. Boon, M. Vikkula, Somatic mutations in an-giopoietin receptor gene tek cause solitary and multiple sporadic venousmalformations, Nat Genet 41 (1) (2009) 118–24. doi:10.1038/ng.272.URL https://www.ncbi.nlm.nih.gov/pubmed/19079259

[12] M. D. Shirley, H. Tang, C. J. Gallione, J. D. Baugher, L. P. Frelin, B. Co-hen, P. E. North, D. A. Marchuk, A. M. Comi, J. Pevsner, Sturge-webersyndrome and port-wine stains caused by somatic mutation in gnaq, NEngl J Med 368 (21) (2013) 1971–9. doi:10.1056/NEJMoa1213507.URL https://www.ncbi.nlm.nih.gov/pubmed/23656586

[13] J. A. Couto, M. P. Vivero, H. P. Kozakewich, A. H. Taghinia, J. B. Mul-liken, M. L. Warman, A. K. Greene, A somatic map3k3 mutation is as-sociated with verrucous venous malformation, Am J Hum Genet 96 (3)(2015) 480–6. doi:10.1016/j.ajhg.2015.01.007.URL https://www.ncbi.nlm.nih.gov/pubmed/25728774

[14] V. L. Luks, N. Kamitaki, M. P. Vivero, W. Uller, R. Rab, J. V. Bovee, K. L.Rialon, C. J. Guevara, A. I. Alomari, A. K. Greene, S. J. Fishman, H. P.Kozakewich, R. A. Maclellan, J. B. Mulliken, R. Rahbar, S. A. Spencer,r. Trenor, C. C., J. Upton, D. Zurakowski, J. A. Perkins, A. Kirsh, J. T.Bennett, W. B. Dobyns, K. C. Kurek, M. L. Warman, S. A. McCarroll,R. Murillo, Lymphatic and other vascular malformative/overgrowth dis-orders are caused by somatic mutations in pik3ca, J Pediatr 166 (4) (2015)1048–54 e1–5. doi:10.1016/j.jpeds.2014.12.069.URL https://www.ncbi.nlm.nih.gov/pubmed/25681199

[15] S. I. Nikolaev, S. Vetiska, X. Bonilla, E. Boudreau, S. Jauhiainen,B. Rezai Jahromi, N. Khyzha, P. V. DiStefano, S. Suutarinen, T. R.Kiehl, V. Mendes Pereira, A. M. Herman, T. Krings, H. Andrade-Barazarte, T. Tung, T. Valiante, G. Zadeh, M. Tymianski, T. Raura-maa, S. Yla-Herttuala, J. D. Wythe, S. E. Antonarakis, J. Frosen, J. E.Fish, I. Radovanovic, Somatic activating kras mutations in arteriovenousmalformations of the brain, N Engl J Med 378 (3) (2018) 250–261.doi:10.1056/NEJMoa1709449.URL https://www.ncbi.nlm.nih.gov/pubmed/29298116

[16] J. A. Couto, A. Y. Huang, D. J. Konczyk, J. A. Goss, S. J. Fishman,J. B. Mulliken, M. L. Warman, A. K. Greene, Somatic map2k1 muta-tions are associated with extracranial arteriovenous malformation, Am JHum Genet 100 (3) (2017) 546–554. doi:10.1016/j.ajhg.2017.01.018.URL https://www.ncbi.nlm.nih.gov/pubmed/28190454

[17] A. L. Akers, E. Johnson, G. K. Steinberg, J. M. Zabramski,

D. A. Marchuk, Biallelic somatic and germline mutations in cere-bral cavernous malformations (ccms): evidence for a two-hit mecha-nism of ccm pathogenesis, Hum Mol Genet 18 (5) (2009) 919–30.doi:10.1093/hmg/ddn430.URL https://www.ncbi.nlm.nih.gov/pubmed/19088123

[18] D. A. McDonald, C. Shi, R. Shenkar, C. J. Gallione, A. L. Akers, S. Li,N. De Castro, M. J. Berg, D. L. Corcoran, I. A. Awad, D. A. Marchuk,Lesions from patients with sporadic cerebral cavernous malformationsharbor somatic mutations in the ccm genes: evidence for a common bio-chemical pathway for ccm pathogenesis, Hum Mol Genet 23 (16) (2014)4357–70. doi:10.1093/hmg/ddu153.URL https://www.ncbi.nlm.nih.gov/pubmed/24698976

[19] C. F. Macmurdo, W. Wooderchak-Donahue, P. Bayrak-Toydemir, J. Le,M. B. Wallenstein, C. Milla, J. M. Teng, J. A. Bernstein, D. A. Stevenson,Rasa1 somatic mutation and variable expressivity in capillary malforma-tion/arteriovenous malformation (cm/avm) syndrome, Am J Med GenetA 170 (6) (2016) 1450–4. doi:10.1002/ajmg.a.37613.URL https://www.ncbi.nlm.nih.gov/pubmed/26969842

[20] C. L. Shovlin, A. E. Guttmacher, E. Buscarini, M. E. Faughnan, R. H. Hy-land, C. J. Westermann, A. D. Kjeldsen, H. Plauchu, Diagnostic criteriafor hereditary hemorrhagic telangiectasia (rendu-osler-weber syndrome),Am J Med Genet 91 (1) (2000) 66–7.URL https://www.ncbi.nlm.nih.gov/pubmed/10751092

[21] S. D. Grosse, S. L. Boulet, A. M. Grant, M. M. Hulihan, M. E. Faughnan,The use of us health insurance data for surveillance of rare disorders:hereditary hemorrhagic telangiectasia, Genet Med 16 (1) (2014) 33–9.doi:10.1038/gim.2013.66.URL https://www.ncbi.nlm.nih.gov/pubmed/23703685

[22] T. W. Laver, R. C. Caswell, K. A. Moore, J. Poschmann, M. B. Johnson,M. M. Owens, S. Ellard, K. H. Paszkiewicz, M. N. Weedon, Pitfalls ofhaplotype phasing from amplicon-based long-read sequencing, Sci Rep 6(2016) 21746. doi:10.1038/srep21746.URL https://www.ncbi.nlm.nih.gov/pubmed/26883533

[23] D. M. Church, C. J. Stotler, J. L. Rutter, J. R. Murrell, J. A. Trofatter,A. J. Buckler, Isolation of genes from complex sources of mammaliangenomic dna using exon amplification, Nat Genet 6 (1) (1994) 98–105.doi:10.1038/ng0194-98.URL https://www.ncbi.nlm.nih.gov/pubmed/8136842

[24] D. W. Johnson, J. N. Berg, C. J. Gallione, K. A. McAllister, J. P.Warner, E. A. Helmbold, D. S. Markel, C. E. Jackson, M. E. Porte-ous, D. A. Marchuk, A second locus for hereditary hemorrhagic telang-iectasia maps to chromosome 12, Genome Res 5 (1) (1995) 21–8.doi:10.1101/gr.5.1.21.URL https://www.ncbi.nlm.nih.gov/pubmed/8717052

[25] A. D. Bossler, J. Richards, C. George, L. Godmilow, A. Ganguly, Novelmutations in eng and acvrl1 identified in a series of 200 individuals un-dergoing clinical genetic testing for hereditary hemorrhagic telangiectasia(hht): correlation of genotype with phenotype, Hum Mutat 27 (7) (2006)667–75. doi:10.1002/humu.20342.URL https://www.ncbi.nlm.nih.gov/pubmed/16752392

[26] C. J. Gallione, D. J. Klaus, E. Y. Yeh, T. T. Stenzel, Y. Xue, K. B.Anthony, K. A. McAllister, M. A. Baldwin, J. N. Berg, A. Lux,J. D. Smith, C. P. Vary, W. J. Craigen, C. J. Westermann, M. L.Warner, Y. E. Miller, C. E. Jackson, A. E. Guttmacher, D. A. Marchuk,Mutation and expression analysis of the endoglin gene in heredi-tary hemorrhagic telangiectasia reveals null alleles, Hum Mutat 11 (4)(1998) 286–94. doi:10.1002/(SICI)1098-1004(1998)11:4¡286::AID-HUMU6¿3.0.CO;2-B.URL https://www.ncbi.nlm.nih.gov/pubmed/9554745

[27] N. Ricard, M. Bidart, C. Mallet, G. Lesca, S. Giraud, R. Prudent, J. J.Feige, S. Bailly, Functional analysis of the bmp9 response of alk1 mutantsfrom hht2 patients: a diagnostic tool for novel acvrl1 mutations, Blood116 (9) (2010) 1604–12. doi:10.1182/blood-2010-03-276881.URL https://www.ncbi.nlm.nih.gov/pubmed/20501893

[28] C. Olivieri, F. Pagella, L. Semino, L. Lanzarini, C. Valacca, A. Pilotto,S. Corno, S. Scappaticci, G. Manfredi, E. Buscarini, C. Danesino, Analy-sis of eng and acvrl1 genes in 137 hht italian families identifies 76 differ-ent mutations (24 novel). comparison with other european studies, J HumGenet 52 (10) (2007) 820–9. doi:10.1007/s10038-007-0187-5.URL https://www.ncbi.nlm.nih.gov/pubmed/17786384

[29] F. O. Desmet, D. Hamroun, M. Lalande, G. Collod-Beroud, M. Claus-

11



http://dx.doi.org/10.1101/731588

tres, C. Beroud, Human splicing finder: an online bioinformatics toolto predict splicing signals, Nucleic Acids Res 37 (9) (2009) e67.doi:10.1093/nar/gkp215.URL https://www.ncbi.nlm.nih.gov/pubmed/19339519

[30] S. Richards, N. Aziz, S. Bale, D. Bick, S. Das, J. Gastier-Foster, W. W.Grody, M. Hegde, E. Lyon, E. Spector, K. Voelkerding, H. L. Rehm, A. L.Q. A. Committee, Standards and guidelines for the interpretation of se-quence variants: a joint consensus recommendation of the american col-lege of medical genetics and genomics and the association for molecularpathology, Genet Med 17 (5) (2015) 405–24. doi:10.1038/gim.2015.30.URL https://www.ncbi.nlm.nih.gov/pubmed/25741868

[31] C. L. Shovlin, J. M. Hughes, J. Scott, C. E. Seidman, J. G. Seidman, Char-acterization of endoglin and identification of novel mutations in heredi-tary hemorrhagic telangiectasia, Am J Hum Genet 61 (1) (1997) 68–79.doi:10.1086/513906.URL https://www.ncbi.nlm.nih.gov/pubmed/9245986

[32] L. Argyriou, S. Twelkemeyer, I. Panchulidze, L. E. Wehner, U. Teske,W. Engel, K. Nayernia, Novel mutations in the eng and acvrl1 genes caus-ing hereditary hemorrhagic teleangiectasia, Int J Mol Med 17 (4) (2006)655–9.URL https://www.ncbi.nlm.nih.gov/pubmed/16525724

[33] A. Bourdeau, U. Cymerman, M. E. Paquet, W. Meschino, W. C. McK-innon, A. E. Guttmacher, L. Becker, M. Letarte, Endoglin expression isreduced in normal vessels but still detectable in arteriovenous malforma-tions of patients with hereditary hemorrhagic telangiectasia type 1, Am JPathol 156 (3) (2000) 911–23. doi:10.1016/S0002-9440(10)64960-7.URL https://www.ncbi.nlm.nih.gov/pubmed/10702408

[34] Y. Jin, L. Muhl, M. Burmakin, Y. Wang, A. C. Duchez, C. Betsholtz,H. M. Arthur, L. Jakobsson, Endoglin prevents vascular malformation byregulating flow-induced cell migration and specification through vegfr2signalling, Nat Cell Biol 19 (6) (2017) 639–652. doi:10.1038/ncb3534.URL https://www.ncbi.nlm.nih.gov/pubmed/28530660

[35] M. R. Detter, D. A. Snellings, D. A. Marchuk, Cerebral cav-ernous malformations develop through clonal expansion of mu-tant endothelial cells, Circ Res 123 (10) (2018) 1143–1151.doi:10.1161/CIRCRESAHA.118.313970.URL https://www.ncbi.nlm.nih.gov/pubmed/30359189

[36] A. Bourdeau, D. J. Dumont, M. Letarte, A murine model of heredi-tary hemorrhagic telangiectasia, J Clin Invest 104 (10) (1999) 1343–51.doi:10.1172/JCI8088.URL https://www.ncbi.nlm.nih.gov/pubmed/10562296

[37] S. Srinivasan, M. A. Hanes, T. Dickens, M. E. Porteous, S. P. Oh,L. P. Hale, D. A. Marchuk, A mouse model for hereditary hemorrhagictelangiectasia (hht) type 2, Hum Mol Genet 12 (5) (2003) 473–82.doi:10.1093/hmg/ddg050.URL https://www.ncbi.nlm.nih.gov/pubmed/12588795

[38] S. Tual-Chalot, S. P. Oh, H. M. Arthur, Mouse models of hereditaryhemorrhagic telangiectasia: recent advances and future challenges, FrontGenet 6 (2015) 25. doi:10.3389/fgene.2015.00025.URL https://www.ncbi.nlm.nih.gov/pubmed/25741358

[39] E. J. Choi, W. Chen, K. Jun, H. M. Arthur, W. L. Young, H. Su,Novel brain arteriovenous malformation mouse models for type 1hereditary hemorrhagic telangiectasia, PLoS One 9 (2) (2014) e88511.doi:10.1371/journal.pone.0088511.URL https://www.ncbi.nlm.nih.gov/pubmed/24520391

[40] E. M. Garrido-Martin, H. L. Nguyen, T. A. Cunningham, S. W. Choe,Z. Jiang, H. M. Arthur, Y. J. Lee, S. P. Oh, Common and distinctivepathogenetic features of arteriovenous malformations in hereditary hem-orrhagic telangiectasia 1 and hereditary hemorrhagic telangiectasia 2 an-imal models–brief report, Arterioscler Thromb Vasc Biol 34 (10) (2014)2232–6. doi:10.1161/ATVBAHA.114.303984.URL https://www.ncbi.nlm.nih.gov/pubmed/25082229

[41] M. Mahmoud, K. R. Allinson, Z. Zhai, R. Oakenfull, P. Ghandi, R. H.Adams, M. Fruttiger, H. M. Arthur, Pathogenesis of arteriovenous mal-formations in the absence of endoglin, Circ Res 106 (8) (2010) 1425–33.doi:10.1161/CIRCRESAHA.109.211037.URL https://www.ncbi.nlm.nih.gov/pubmed/20224041

[42] E. J. Choi, E. J. Walker, F. Shen, S. P. Oh, H. M. Arthur, W. L. Young,H. Su, Minimal homozygous endothelial deletion of eng with vegf stimu-lation is sufficient to cause cerebrovascular dysplasia in the adult mouse,Cerebrovasc Dis 33 (6) (2012) 540–7. doi:10.1159/000337762.

URL https://www.ncbi.nlm.nih.gov/pubmed/22571958

[43] Y. H. Kim, S. W. Choe, M. Y. Chae, S. Hong, S. P. Oh, Smad4deficiency leads to development of arteriovenous malformations inneonatal and adult mice, J Am Heart Assoc 7 (21) (2018) e009514.doi:10.1161/JAHA.118.009514.URL https://www.ncbi.nlm.nih.gov/pubmed/30571376

[44] C. D. Gonzalez, S. D. Cipriano, C. A. Topham, D. A. Stevenson, K. J.Whitehead, S. Vanderhooft, A. P. Presson, J. McDonald, Localization andage distribution of telangiectases in children and adolescents with heredi-tary hemorrhagic telangiectasia: A retrospective cohort study, J Am AcadDermatol (2019). doi:10.1016/j.jaad.2018.11.014.URL https://www.ncbi.nlm.nih.gov/pubmed/30819528

[45] T. G. Letteboer, H. J. Mager, R. J. Snijder, D. Lindhout, H. K. Ploos vanAmstel, P. Zanen, K. J. Westermann, Genotype-phenotype relationshipfor localization and age distribution of telangiectases in hereditary hem-orrhagic telangiectasia, Am J Med Genet A 146A (21) (2008) 2733–9.doi:10.1002/ajmg.a.32243.URL https://www.ncbi.nlm.nih.gov/pubmed/18831062

[46] H. Plauchu, J. P. de Chadarevian, A. Bideau, J. M. Robert, Age-relatedclinical profile of hereditary hemorrhagic telangiectasia in an epidemi-ologically recruited population, Am J Med Genet 32 (3) (1989) 291–7.doi:10.1002/ajmg.1320320302.URL https://www.ncbi.nlm.nih.gov/pubmed/2729347

[47] B. Milholland, X. Dong, L. Zhang, X. Hao, Y. Suh, J. Vijg, Differencesbetween germline and somatic mutation rates in humans and mice, NatCommun 8 (2017) 15183. doi:10.1038/ncomms15183.URL https://www.ncbi.nlm.nih.gov/pubmed/28485371

[48] M. A. Lodato, M. B. Woodworth, S. Lee, G. D. Evrony, B. K. Mehta,A. Karger, S. Lee, T. W. Chittenden, A. M. D’Gama, X. Cai, L. J. Lu-quette, E. Lee, P. J. Park, C. A. Walsh, Somatic mutation in single hu-man neurons tracks developmental and transcriptional history, Science350 (6256) (2015) 94–98. doi:10.1126/science.aab1785.URL https://www.ncbi.nlm.nih.gov/pubmed/26430121

[49] V. Lo Sardo, W. Ferguson, G. A. Erikson, E. J. Topol, K. K. Baldwin,A. Torkamani, Influence of donor age on induced pluripotent stem cells,Nat Biotechnol 35 (1) (2017) 69–74. doi:10.1038/nbt.3749.URL https://www.ncbi.nlm.nih.gov/pubmed/27941802

[50] R. Sender, S. Fuchs, R. Milo, Revised estimates for the number of hu-man and bacteria cells in the body, PLoS Biol 14 (8) (2016) e1002533.doi:10.1371/journal.pbio.1002533.URL https://www.ncbi.nlm.nih.gov/pubmed/27541692

[51] N. M. Eleftheriou, J. Sjolund, M. Bocci, E. Cortez, S. J. Lee, S. I. Cunha,K. Pietras, Compound genetically engineered mouse models of cancerreveal dual targeting of alk1 and endoglin as a synergistic opportunityto impinge on angiogenic tgf-beta signaling, Oncotarget 7 (51) (2016)84314–84325. doi:10.18632/oncotarget.12604.URL https://www.ncbi.nlm.nih.gov/pubmed/27741515

[52] L. Ruiz-Llorente, E. Gallardo-Vara, E. Rossi, D. M. Smadja, L. M.Botella, C. Bernabeu, Endoglin and alk1 as therapeutic targets for heredi-tary hemorrhagic telangiectasia, Expert Opin Ther Targets 21 (10) (2017)933–947. doi:10.1080/14728222.2017.1365839.URL https://www.ncbi.nlm.nih.gov/pubmed/28796572

[53] T. Seki, J. Yun, S. P. Oh, Arterial endothelium-specific activinreceptor-like kinase 1 expression suggests its role in arterializa-tion and vascular remodeling, Circ Res 93 (7) (2003) 682–9.doi:10.1161/01.RES.0000095246.40391.3B.URL https://www.ncbi.nlm.nih.gov/pubmed/12970115

12



http://dx.doi.org/10.1101/731588



http://dx.doi.org/10.1101/731588

Somatic Mutations in Vascular Malformations of Hereditary ... · mutations in KRAS or MAP2K1.15,16...

Documents

Transcript of Somatic Mutations in Vascular Malformations of Hereditary ... · mutations in KRAS or MAP2K1.15,16...