EXPERIMENTAL PIK3CA Activating Mutations in Facial Infiltrating Lipomatosis Reid A. Maclellan, M.D., M.M.Sc. Valerie L. Luks, B.S. Matthew P. Vivero, B.A. John B. Mulliken, M.D. David Zurakowski, Ph.D. Bonnie L. Padwa, D.M.D., M.D. Matthew L. Warman, M.D. Arin K. Greene, M.D., M.M.Sc. Kyle C. Kurek, M.D., M.M.Sc. Boston, Mass.
Background: Facial infiltrating lipomatosis is a nonheritable disorder characterized by hemifacial soft-tissue and skeletal overgrowth, precocious dental development, macrodontia, hemimacroglossia, and mucosal neuromas. The authors tested the hypothesis that this condition is caused by a somatic mutation in the phosphatidylinositide-3 kinase (PI3K) signaling pathway, which has been indicted in other anomalies with overgrowth. Methods: The authors extracted DNA from abnormal tissue in six individuals, generated sequencing libraries, enriched the libraries for 26 genes involved in the PI3K pathway, and designed and applied a sequential filtering strategy to analyze the sequence data for mosaic mutations. Results: Unfiltered sequence data contained variant reads affecting ~12 percent of basepairs in the targeted genes. Filtering reduced the fraction of targeted basepairs containing variant reads to ~0.008 percent, allowing the authors to identify causal missense mutations in PIK3CA (p.E453K, p.E542K, p.H1047R, or p.H1047L) in each affected tissue sample. Conclusions: Affected tissue from individuals with facial infiltrating lipomatosis contains PIK3CA mutations that have previously been reported in cancers and in affected tissue from other nonheritable, overgrowth disorders, including congenital lipomatous overgrowth, vascular, epidermal, and skeletal anomalies syndrome, Klippel-Trenaunay syndrome, hemimegalencephaly, fibroadipose overgrowth, and macrodactyly. Because PIK3CA encodes a catalytic subunit of PI3K, and in vitro studies have shown that the overgrowth-associated mutations increase this enzyme’s activity, PI3K inhibitors currently in clinical trials for patients with cancer may have a therapeutic role in patients with facial infiltrating lipomatosis. The strategy used to identify somatic mutations in patients with facial infiltrating lipomatosis is applicable to other somatic mosaic disorders that have allelic heterogeneity. (Plast. Reconstr. Surg. 133: 12e, 2014.)
F
acial infiltrating lipomatosis is a rare nonhereditary disorder characterized by hemifacial overgrowth and fatty infiltration of soft tissues. Clinical features include bony and soft-tissue hypertrophy, mucosal neuromas, hemimacroglossia, macrodontia, and premature dental eruption.1,2 Histopathologically, nonencapsulated, mature, fat cells infiltrate adjacent muscle and soft tissue.1 Treatment consists of subtotal resection of hypertrophied soft tissue and bone; however, local recurrence can occur.2,3 From the Departments of Plastic and Oral Surgery, Orthopaedic Surgery, and Pathology and Howard Hughes Medical Institute, Boston Children’s Hospital; and the Department of Genetics, Harvard Medical School. Received for publication June 17, 2013; accepted July 12, 2013. Copyright © 2013 by the American Society of Plastic Surgeons DOI: 10.1097/01.prs.0000436822.26709.7c
12e
Several pathogenic mechanisms have been proposed, such as cytomegalovirus infection,4 abnormal angiogenesis/vasculogenesis,5 and a somatic mutation in adipose stem cells.2 The phenotypic features of facial infiltrating lipomatosis are sometimes seen in patients with congenital lipomatous overgrowth, vascular, epidermal, and skeletal anomalies (or CLOVES) syndrome.6,7 Somatic mutations affecting components of the phosphatidylinositide-3 kinase (PI3K) signaling pathway have been identified in patients with CLOVES syndrome8 and other disorders with overgrowth.9–14 Therefore, we targeted genes in the PI3K pathway and performed deep sequencing of affected tissue DNA from patients with facial infiltrating lipomatosis. Disclosure: The authors have no financial interest to declare in relation to the content of this article.
www.PRSJournal.com
Volume 133, Number 1 • Facial Infiltrating Lipomatosis PATIENTS AND METHODS Patient Ascertainment and Sample Collection After approval by the Committee on Clinical Investigation at Boston Children’s Hospital, abnormal subcutaneous tissue was collected from the cheeks of six patients with facial infiltrating lipomatosis at the time of a clinically indicated procedure. The disorder was diagnosed based on history, physical examination, imaging, and histopathologic analysis. DNA Preparation for Massively Parallel Sequencing Genomic DNA was extracted from each affected tissue specimen and used to prepare sequencing libraries as previously described.8 DNA (3 µg) was physically sheared to generate fragments between 100 and 200 basepairs long. Sequencing primers, which were given barcodes so as to distinguish between the different samples, were enzymatically ligated to the sheared DNA fragments. We performed 15 cycles of the polymerase chain reaction to increase the amount of DNA in each barcoded library, pooled the libraries, and hybridized the pooled DNA to a custom-designed capture array (Agilent Technology 1m SureSelect DNA Capture Array, Agilent Technologies, Santa Clara, Calif.) that contains the coding regions for 26 genes thought to function within the PI3K signaling pathway (AKT1, AKT2, AKT3, FKBP1A, IGF1, KCNH1, KCNH2, KCNK5, MDK, MTOR, NFATC1, NF1, NF2, NOS3, PIK3CA, PIK3CB, PIK3CG, PIK3R1, PIK3R2, PIK3R3, PPP3CA, PPP3CB, PPP3CC, PTEN, PTN, and SLC12A7). The DNA fragments captured by the array were again increased in abundance by performing 17 cycles of polymerase chain reaction and then used to obtain 100-basepair paired end reads on an Illumina HiSeq2 (HiSeq 2000; Illumina, Inc., San Diego, Calif.) sequencing apparatus. DNA Sequence Analysis Paired-end read data were de-barcoded with Novobarcode (Novocraft Technologies, Selangor, Malaysia), aligned to the University of California Santa Cruz human reference genome (GRCh37/ hg19), and filtered to remove polymerase chain reaction duplicates.15,16 Pileup files were generated for each of the six patients using SAMtools.16 VarScan variant detection software was used to list all chromosomal locations containing two or more alternative base calls (relative to the reference base) within the total depth of coverage at the particular base.17 We further filtered the data sequentially to reduce false-positive findings that could have
arisen during library preparation and HiSeq2 sequencing. We considered only bases for which there were 10 or more reads, of which at least three reads contained the variant base. We excluded variants that exhibited significant strand bias; we kept all variants that had fewer than 10 reads, but discarded variants with more than 10 reads if 80 percent or more of the reads occurred in a single direction. For example, we excluded a variant that was found 10 times in one direction and two times in the other direction (10 of 12, or 83.3 percent) and kept a variant that was found nine times in one direction and three times in the other direction (nine of 12, or 75 percent). We filtered variants that were in the 1000 Genomes database,18 the National Heart, Lung, and Blood Institute Exome Variant Server database (http:// evs.gs.washington.edu/EVS/; accessed March, 2013), and build 132 of the database of common single nucleotide polymorphisms,19 since these variants are highly unlikely to represent a somatic mutation, and filtered variants that were not in the coding sequence or at a splice site. Lastly, we ranked variants by their frequency in descending order. A variant with a frequency of 5 percent (e.g., 20 of 400 reads) was ranked higher than a variant with a frequency of 1 percent (e.g., four of 400 reads). We then visually inspected each highly ranked variant using the Integrative Genomics Viewer20 and determined whether any of the variants were present in the Catalog Of Somatic Mutations In Cancer (or COSMIC) database.21 Finally, we determined if any of the 26 genes in the array contained variants in three or more patient samples. Variants that satisfied all criteria were confirmed as true-positive mutations. We re-extracted DNA from affected tissue, polymerase chain reaction–amplified and subcloned the relevant amplimers, and screened individual subclones for the presence of the mutation.
RESULTS Figures 1 and 2 show a photograph and a cranial magnetic resonance image depicting the clinical findings in a patient with facial infiltrating lipomatosis and photomicrographs of hematoxylin and eosin–stained sections representative of the excised affected tissues that were sequenced. All patients had histopathologic findings of mature adipose infiltrating deep into skeletal muscle through dermal and subcutaneous paucicellular fibroconnective tissue. Small foci of entrapped minor salivary glands and abnormal-appearing
13e
Plastic and Reconstructive Surgery • January 2014
Fig. 1. (Left) Four-year-old boy with facial infiltrating lipomatosis. (Right) T1-weighted axial magnetic resonance image of this child demonstrates increased subcutaneous fat and lipomatous infiltration of the right maxilla and mandible, right salivary glands, and muscles of mastication.
Fig. 2. Photomicrographs of excised affected tissues used for mutation detection from four study participants demonstrating mature adipose tissue infiltrating or surrounding (above, left) dermal and subcutaneous paucicellular fibroconnective tissue, (above, right) skeletal muscle fibers (asterisk), (below, left) salivary tissue (asterisk), and (below, right) skeletal muscle (single asterisk) and peripheral nerves (double asterisks). All photomicrographs were obtained from hematoxylin and eosin–stained histologic sections at 200 times magnification. Bar = 100 microns.
14e
76.1% 80.0% 37.8% 71.2% 79.6%
79.8% PCR, polymerase chain reaction.
85.3% 86% 75.9% 79.8% 85.5%
85.7%
5.2% 5.7% 3.1% 9.3% 5.5%
5.2%
1,028,375 (72.4%) 411 1,356,749 (89.6%) 543 531,969 (60.6%) 213 899,204 (93.6%) 360 1,207,544 (92.4%) 483
1,175,250 (93.4%) 470
1,419,655 1,514,952 878,209 1,258,736 960,564
2 1
1,307,510
No. of reads after removal of PCR duplicates Reads mapping to array (% of total reads mapping to array) Average depth of coverage 0× depth sequence coverage of coding sequence 50× depth sequence coverage of coding sequence 100× depth sequence coverage of coding sequence
Table 1. Sequence Coverage for Six Affected Tissue Samples
3
Patient
4
5
6
Volume 133, Number 1 • Facial Infiltrating Lipomatosis nerves were frequently observed in the lesional tissue. Table 1 summarizes the DNA sequence coverage obtained from each specimen. After removing polymerase chain reaction duplicates, more than 878,000 100-basepair paired-end reads were obtained for each sample. These reads were enriched for the 26 genes on the capture array, with 60 percent to 93 percent aligning to the targeted sequence. The average read depth ranged from 213× to 543× per base captured. With the exception of one patient’s tissue that had lower coverage, the samples had coverage of 100× or more across more than 70 percent of the coding sequence on the capture array. Table 2 summarizes the number of bases for which deep sequencing yielded two or more reads with the same single nucleotide variant (SNV). Table 2 also depicts how sequential filtering criteria reduced the number of bases that had candidate disease-causing single nucleotide variants. Note that without filtering, about one in eight bases analyzed, at 10× or more coverage, had at least two reads that contained a single nucleotide variant (e.g., 21,086 bases among the 165,744 bases sequenced had at least two variant reads in patient 1). By requiring that three or more reads contain a variant base call, the frequency dropped to about one in 20 (8398 variant bases among the 165,744 bases). Filtering for strand bias and for probable germline single nucleotide variants (i.e., variants present in the 1000 Genomes, Exome Variant Server, and common single nucleotide polymorphisms databases) had little effect on the frequency of variant bases (about one in 22, or 7512 among the 165,744 bases). Another modest reduction in frequency to about one in 29 (3787 among the approximately 120,000 bases of sequence that represent coding or splice sites) resulted when only exons and splice sites were considered. The greatest reduction occurred when we considered the percentage of reads that contained a variant. As shown in Table 2, the requirement that 1 percent or more of reads at a base contain a variant call reduced the frequency to about one in 118 (935 among the approximately 120,000 bases). Increasing the stringency to 5 percent or more reduced the frequency to about one in 12,200 (nine out of approximately 120,000 bases). Under the latter filtering criteria, 10 of the 26 genes in our targeting array had nonsynonymous, coding sequence variants in three or more patient samples. Only one gene, PIK3CA, had nonsynonymous variants in all patient samples. Importantly, each of the PIK3CA variants
15e
Plastic and Reconstructive Surgery • January 2014 Table 2. Variant Filtering for Six Affected Tissue Samples Patient Bases sequenced* Bases analyzed† Minimum of 10× coverage Total bp with 2 variant reads Total bp with ≥3 variant reads Variants passing strand-bias filter criteria Variants not in databases‡ Variants in exons/splice sites Variants in ≥1% of reads Variants in ≥2% of reads Variants in ≥5% of reads Genes with a ≥5% variant in ≥3 patient samples Genes with a ≥5% variant in all patient samples The ≥5% PIK3CA variants found in COSMIC
1
2
3
4
5
6
523,068 178,549 165,744 21,086 8398
491,196 169,669 152,770 14,163 4765
526,138 179,888 166,907 18,585 6818
541,753 183,788 163,079 8,053 1980
523,387 178,844 166,427 23,047 8925
527,233 179,528 166,081 16,280 5179
7645 7512 3787 935 136 9
4294 4248 2061 777 120 10
6038 5887 2829 882 130 5
1446 1307 599 466 423 38
7977 7814 3916 873 113 13
4500 4333 2045 1251 280 6
10 (AKT1, KCNH1, KCNK5, mTOR, NF1, NFATC1, NOS3, PIK3CA, PIK3R2, SLC12A7) 1 (PIK3CA) All (p.E453K, p.H1047R, p.H1047L, p.E542K)
bp, base pair; COSMIC, Catalog Of Somatic Mutations In Cancer database. *Capture array encompassed 580,000 bases: all exonic regions, 5’ and 3’ untranslated regions, and 90 bases of DNA upstream and downstream of each captured exon. †Bases analyzed included approximately 120,000 bases of coding sequence and splice sites, as well as flanking bases for which paired sequencing reads that contain exonic sequence overlap. ‡The 1000 Genomes, Exome Variant Server, and common variants in single nucleotide polymorphism databases.
we identified had previously been reported as a somatically occurring mutation in the Catalog Of Somatic Mutations In Cancer database.21 In the nine other genes that had variants at a frequency to 5 percent or more, only two of the 35 variants were in the Catalog Of Somatic Mutations In Cancer database; these two variants were in the same sample (patient 4). Table 3 summarizes the depth of coverage and frequency of variant calls for the PIK3CA mutations identified in the six patient samples. After removal of polymerase chain reaction duplicates, the depth of coverage ranged from 121× to 1091× over the sites of the mutations. When polymerase chain reaction duplicates were retained, the depth of coverage increased approximately 10-fold with little to no effect on the frequency of the mutant allele.
DISCUSSION Certain disorders must be the result of somatic mosaicism because complete heterozygosity would be lethal or cannot be transmitted through the egg or sperm.22 Massively parallel sequencing technologies provide a means for identifying these somatic mosaic mutations and were used by us to identify the likely cause of facial infiltrating lipomatosis. We focused our attention on the PI3K signaling pathway that has been implicated in other overgrowth disorders, such as Proteus syndrome,9 CLOVES
16e
syndrome,8 Klippel-Trenaunay syndrome,8 hemimegalencephaly,11,13,14 macrodactyly,10 and lipoadenomatous hyperplasia.12 Furthermore, features seen in patients with facial infiltrating lipomatosis (e.g., fatty overgrowth and neuromas) occur in some patients with CLOVES syndrome. Therefore, our finding that patients with facial infiltrating lipomatosis have the same somatic PIK3CA mutations that occur in patients with CLOVES syndrome was not entirely unexpected. There are several possible explanations for how the same somatic mutation can cause a variety of clinical conditions. First, the phenotype may depend on when the somatic mutation arises during embryogenesis. For example, patients with CLOVES syndrome who have widespread involvement (e.g., face, trunk, extremities, and viscera) could have a somatic mutation that arises earlier in development than patients with regional overgrowth in the face or a digit. Perhaps facial infiltrating lipomatosis is the result of a mutation arising in the first pharyngeal arch, since derivatives of this structure form the maxilla, mandible, muscles of mastication, and the anterior twothirds of the tongue.23,24 A second explanation is that the phenotype may depend on the specific cell type that contains the mutation. Digital overgrowth in patients with isolated macrodactyly follows a “nerve territory,” and the affected tissue
Volume 133, Number 1 • Facial Infiltrating Lipomatosis Table 3. PIK3CA Mutation Frequency in Six Affected Tissue Samples Patient Mutation Total reads over mutation site after PCR duplicate filtering Variant reads over mutation site after PCR duplicate filtering (% mutant reads) Total read over mutation site without PCR duplicates filtering Variant reads over mutation site without PCR duplicate filtering (% mutant reads)
1
2
3
4
5
6
p.H1047R
p.H1047L
p.E452K
p.E542K
p.E453K
p.H1047R
794
318
121
358
70 (9)
73 (23)
36 (30)
58 (16)
7889
2490
1201
2759
743 (9)
577 (23)
362 (30)
497 (18)
1091 339 (31) 10,477 3421 (33)
621 71 (11) 14,644 1036 (7)
PCR, polymerase chain reaction.
is composed of enlarged peripheral nerves.10 A mutation arising in the neural crest cell component of primitive neuroectoderm25 could explain the phenotype in facial infiltrating lipomatosis, since these cells migrate into each of the facial structures affected in this disorder and patients frequently have mucosal neuromas. Genetic background is another possibility for the divergent phenotypes among individuals with the same PIK3CA mutation. Patients with CLOVES syndrome may have modifier alleles that amplify the effect of the postzygotic somatic PIK3CA mutation relative to patients with more regional phenotypes. Modifier alleles have been identified for Mendelian genetic diseases26 and for somatic mutations that give rise to cancer.27 There are two mouse strains with an inducible Pik3ca mutation28,29 that is the same as one of the mutations found in patients with facial infiltrating lipomatosis. These mouse strains could be used to address the first two potential mechanisms. Testing the last mechanism will require the genomes of many patients with PIK3CA-associated overgrowth syndromes to be sequenced and analyzed. Identification of PIK3CA mutations in affected tissue from patients with facial infiltrating lipomatosis utilized a strategy that focused on only 26 of the estimated 20,000 genes in the human genome. Since a priori candidate pathways may not exist for other diseases that are somatic mosaic in origin, would we also have been successful if the entire exome was sequenced? Current commercially available, whole-exome sequencing provides an average coverage of approximately 70×, in contrast to the average coverage of more than 400× we achieved with the targeted array. Nevertheless, despite the lower coverage of whole-exome sequencing, it is likely we would have detected the somatic PIK3CA mutations in most patients
because the mutant alleles would have survived our filtering process. For example, patients 1, 3, 4, and 5, who have PIK3CA mutant allele frequencies of 31, 23, 30, and 16 percent, respectively, would each have had more than three variant reads with more than 5 percent of the reads being variant in whole-exome data at 70× coverage. We had this experience applying whole-exome sequencing in patients with CLOVES syndrome.8 In the current study, patients 2 and 6, whose mutant allele frequencies were only 9 and 11 percent, respectively, should also have been detected with whole-exome sequencing, but they might have been missed due to stochastic variation in allelic sampling. We used a strict filtering strategy to minimize false-positive variants during our sequence analysis. Ultimately, ranking mutations based on frequency and only looking at those above a 5 percent threshold improved the positive predictive value of the variants we identified. For example, one-in-nine, one-in-10, and one-in-five of the variants that survived this filtering strategy in patients 1, 2, and 3, respectively, were the PIK3CA mutations. Setting a lower threshold would have increased the sensitivity for detecting low-frequency mosaic mutations but at the cost of having significantly more false-positive variants to consider as well. A 1 percent threshold was used to identify a somatic missense mutation in GNAQ as the cause of Sturge-Weber syndrome and capillary malformations; however, this search was facilitated since all patients had the identical causative variant.30 Our filtering strategy, which detected different somatic mutations affecting the same gene, should be helpful to other investigators studying phenotypes that result from somatic mosaic mutations with allelic heterogeneity. PIK3CA encodes the 110-kD alpha subunit of the enzyme phosphoinositide-3-kinase (PI3K).31,32 Upon activation, PI3K converts
17e
Plastic and Reconstructive Surgery • January 2014 phosphatidylinositol (3,4)-bisphosphate (PIP2) to phosphatidylinositol (3,4,5)-triphosphate (PIP3).33 PIP3 promotes phosphorylation of AKT family members, which regulate cell proliferation, adhesion, survival, and motility. The PIK3CA mutations we identified in patients with facial infiltrating lipomatosis have previously been identified in several cancers21,34 and have been shown to constitutively activate PI3K.35 Therefore, pharmacologic agents that inhibit the PI3K pathway may benefit patients with facial infiltrating lipomatosis. Several PI3K pathway inhibitors are currently in phase II and III clinical trials for cancers containing PIK3CA mutations.36 Although pharmacotherapy probably would not cure the disorders, it might prevent progression or recurrence. Arin K. Greene, M.D., M.M.Sc. Department of Plastic and Oral Surgery Boston Children’s Hospital 300 Longwood Avenue Boston, Mass. 02115 [email protected]
ACKNOWLEDGMENTS
This work was supported by the Manton Center for Orphan Disease Research at Boston Children’s Hospital and the Howard Hughes Medical Institute. The authors are grateful to their patients for participating in this study and to Dr. Peter Jezewski (University of Alabama at Birmingham) for his insights into craniofacial developmental biology. The authors thank the National Heart, Lung, and Blood Institute Grand Opportunity Exome Sequencing Project and its ongoing studies that produced and provided exome variant calls for comparison: the Lung Grand Opportunity Sequencing Project (HL-102923), the WHI Sequencing Project (HL102924), the Broad Grand Opportunity Sequencing Project (HL-102925), the Seattle Grand Opportunity Sequencing Project (HL-102926), and the Heart Grand Opportunity Sequencing Project (HL-103010). REFERENCES 1. Slavin SA, Baker DC, McCarthy JG, Mufarrij A. Congenital infiltrating lipomatosis of the face: clinicopathologic evaluation and treatment. Plast Reconstr Surg. 1983;72:158–164. 2. Padwa BL, Mulliken JB. Facial infiltrating lipomatosis. Plast Reconstr Surg. 2001;108:1544–1554. 3. Kang N, Ross D, Harrison D. Unilateral hypertrophy of the face associated with infiltrating lipomatosis. J Oral Maxillofac Surg. 1998;56:885–887. 4. Donati L, Candiani P, Grappolini S, Klinger M, Signorini M. Congenital infiltrating lipomatosis of the face related to cytomegalovirus infection. Br J Plast Surg. 1990;43:124–126. 5. Couto RA, Mulliken JB, Padwa BL, et al. Facial infiltrating lipomatosis: expression of angiogenic and vasculogenic factors. J Craniofac Surg. 2011;22:2405–2408.
18e
6. Sapp JC, Turner JT, van de Kamp JM, van Dijk FS, Lowry RB, Biesecker LG. Newly delineated syndrome of congenital lipomatous overgrowth, vascular malformations, and epidermal nevi (CLOVE syndrome) in seven patients. Am J Med Genet A. 2007;143A:2944–2958. 7. Alomari AI. Characterization of a distinct syndrome that associates complex truncal overgrowth, vascular, and acral anomalies: a descriptive study of 18 cases of CLOVES syndrome. Clin Dysmorphol. 2009;18:1–7. 8. Kurek KC, Luks VL, Ayturk UM, et al. Somatic mosaic activating mutations in PIK3CA cause CLOVES syndrome. Am J Hum Genet. 2012;90:1108–1115. 9. Lindhurst MJ, Sapp JC, Teer JK, et al. A mosaic activating mutation in AKT1 associated with the Proteus syndrome. N Engl J Med. 2011;365:611–619. 10. Rios JJ, Paria N, Burns DK, et al. Somatic gain-of-function mutations in PIK3CA in patients with macrodactyly. Hum Mol Genet. 2013;22:444–451. 11. Lee JH, Huynh M, Silhavy JL, et al. De novo somatic mutations in components of the PI3K-AKT3-mTOR pathway cause hemimegalencephaly. Nat Genet. 2012;44:941–945. 12. Lindhurst MJ, Parker VE, Payne F, et al. Mosaic overgrowth with fibroadipose hyperplasia is caused by somatic activating mutations in PIK3CA. Nat Genet. 2013;44:928–933. 13. Rivière JB, Mirzaa GM, O’Roak BJ, et al.; Finding of Rare Disease Genes (FORGE) Canada Consortium. De novo germline and postzygotic mutations in AKT3, PIK3R2 and PIK3CA cause a spectrum of related megalencephaly syndromes. Nat Genet. 2012;44:934–940. 14. Poduri A, Evrony GD, Cai X, et al. Somatic activation of AKT3 causes hemispheric developmental brain malformations. Neuron. 2012;74:41–48. 15. Li H, Durbin R. Fast and accurate short read align ment with Burrows-Wheeler transform. Bioinformatics. 2009;25:1754–1760. 16. Li, H., Handsaker, B., Wysoker, A., et al. The Sequence Alignment/Map format and SAMtools. Bioinformatics 25: 2078–2079, 2009. 17. Koboldt DC, Chen K, Wylie T, et al. VarScan: variant detection in massively parallel sequencing of individual and pooled samples. Bioinformatics. 2009;25:2283–2285. 18. Abecasis, G. R., Auton, A., Brooks, L. D., et al. An integrated map of genetic variation from 1,092 human genomes. Nature 491: 56–65, 2012. 19. Sherry ST, Ward MH, Kholodov M, et al. dbSNP: the NCBI database of genetic variation. Nucleic Acids Res. 2001;29:308–311. 20. Robinson JT, Thorvaldsdóttir H, Winckler W, et al. Integrative genomics viewer. Nat Biotechnol. 2011;29:24–26. 21. Forbes SA, Bindal N, Bamford S, et al. COSMIC: mining complete cancer genomes in the Catalogue of Somatic Mutations in Cancer. Nucleic Acids Res. 2011;39(Database issue):D945–D950. 22. Happle R. Lethal genes surviving by mosaicism: a possible explanation for sporadic birth defects involving the skin. J Am Acad Dermatol. 1987;16:899–906. 23. Cobourne MT, Sharpe PT. Tooth and jaw: molecular mechanisms of patterning in the first branchial arch. Arch Oral Biol. 2003;48:1–14. 24. Müller F, O’Rahilly R. The prechordal plate, the rostral end of the notochord and nearby median features in staged human embryos. Cells Tissues Organs. 2003;173:1–20. 25. LaBonne C, Bronner-Fraser M. Molecular mechanisms of neural crest formation. Annu Rev Cell Dev Biol. 1999;15:81–112. 26. Cutting GR. Modifier genes in Mendelian disorders: the example of cystic fibrosis. Ann N Y Acad Sci. 2010;1214:57–69.
Volume 133, Number 1 • Facial Infiltrating Lipomatosis 27. Ruark E, Snape K, Humburg P, et al.; Breast and Ovarian Cancer Susceptibility Collaboration; Wellcome Trust Case Control Consortium. Mosaic PPM1D mutations are associated with predisposition to breast and ovarian cancer. Nature. 2013;493:406–410. 28. Kinross KM, Montgomery KG, Kleinschmidt M, et al. An activating Pik3ca mutation coupled with Pten loss is sufficient to initiate ovarian tumorigenesis in mice. J Clin Invest. 2012;122:553–557. 29. Yuan W, Stawiski E, Janakiraman V, et al. Conditional activation of Pik3ca(H1047R) in a knock-in mouse model promotes mammary tumorigenesis and emergence of mutations. Oncogene. 2013;32:318–326. 30. Shirley MD, Tang H, Gallione CJ, et al. Sturge-Weber syndrome and port-wine stains caused by somatic mutation in GNAQ. N Engl J Med. 2013.
31. Hiles ID, Otsu M, Volinia S, et al. Phosphatidylinositol 3-kinase: structure and expression of the 110 kd catalytic subunit. Cell. 1992;70:419–429. 32. Volinia S, Hiles I, Ormondroyd E, et al. Molecular cloning, cDNA sequence, and chromosomal localization of the human phosphatidylinositol 3-kinase p110 alpha (PIK3CA) gene. Genomics. 1994;24:472–477. 33. Hemmings BA, Restuccia DF. PI3K-PKB/Akt pathway. Cold Spring Harb Perspect Biol. 2012;4:a011189. 34. Samuels Y, Wang Z, Bardelli A, et al. High frequency of mutations of the PIK3CA gene in human cancers. Science. 2004;304:554. 35. Kang S, Bader AG, Vogt PK. Phosphatidylinositol 3-kinase mutations identified in human cancer are oncogenic. Proc Natl Acad Sci U S A. 2005;102:802–807. 36. Kurtz JE, Ray-Coquard I. PI3 kinase inhibitors in the clinic: an update. Anticancer Res. 2012;32:2463–2470.
19e
Background: Facial infiltrating lipomatosis is a nonheritable disorder characterized by hemifacial soft-tissue and skeletal overgrowth, precocious dental development, macrodontia, hemimacroglossia, and mucosal neuromas. The authors tested the hypothesis that this condition is caused by a somatic mutation in the phosphatidylinositide-3 kinase (PI3K) signaling pathway, which has been indicted in other anomalies with overgrowth. Methods: The authors extracted DNA from abnormal tissue in six individuals, generated sequencing libraries, enriched the libraries for 26 genes involved in the PI3K pathway, and designed and applied a sequential filtering strategy to analyze the sequence data for mosaic mutations. Results: Unfiltered sequence data contained variant reads affecting ~12 percent of basepairs in the targeted genes. Filtering reduced the fraction of targeted basepairs containing variant reads to ~0.008 percent, allowing the authors to identify causal missense mutations in PIK3CA (p.E453K, p.E542K, p.H1047R, or p.H1047L) in each affected tissue sample. Conclusions: Affected tissue from individuals with facial infiltrating lipomatosis contains PIK3CA mutations that have previously been reported in cancers and in affected tissue from other nonheritable, overgrowth disorders, including congenital lipomatous overgrowth, vascular, epidermal, and skeletal anomalies syndrome, Klippel-Trenaunay syndrome, hemimegalencephaly, fibroadipose overgrowth, and macrodactyly. Because PIK3CA encodes a catalytic subunit of PI3K, and in vitro studies have shown that the overgrowth-associated mutations increase this enzyme’s activity, PI3K inhibitors currently in clinical trials for patients with cancer may have a therapeutic role in patients with facial infiltrating lipomatosis. The strategy used to identify somatic mutations in patients with facial infiltrating lipomatosis is applicable to other somatic mosaic disorders that have allelic heterogeneity. (Plast. Reconstr. Surg. 133: 12e, 2014.)
F
acial infiltrating lipomatosis is a rare nonhereditary disorder characterized by hemifacial overgrowth and fatty infiltration of soft tissues. Clinical features include bony and soft-tissue hypertrophy, mucosal neuromas, hemimacroglossia, macrodontia, and premature dental eruption.1,2 Histopathologically, nonencapsulated, mature, fat cells infiltrate adjacent muscle and soft tissue.1 Treatment consists of subtotal resection of hypertrophied soft tissue and bone; however, local recurrence can occur.2,3 From the Departments of Plastic and Oral Surgery, Orthopaedic Surgery, and Pathology and Howard Hughes Medical Institute, Boston Children’s Hospital; and the Department of Genetics, Harvard Medical School. Received for publication June 17, 2013; accepted July 12, 2013. Copyright © 2013 by the American Society of Plastic Surgeons DOI: 10.1097/01.prs.0000436822.26709.7c
12e
Several pathogenic mechanisms have been proposed, such as cytomegalovirus infection,4 abnormal angiogenesis/vasculogenesis,5 and a somatic mutation in adipose stem cells.2 The phenotypic features of facial infiltrating lipomatosis are sometimes seen in patients with congenital lipomatous overgrowth, vascular, epidermal, and skeletal anomalies (or CLOVES) syndrome.6,7 Somatic mutations affecting components of the phosphatidylinositide-3 kinase (PI3K) signaling pathway have been identified in patients with CLOVES syndrome8 and other disorders with overgrowth.9–14 Therefore, we targeted genes in the PI3K pathway and performed deep sequencing of affected tissue DNA from patients with facial infiltrating lipomatosis. Disclosure: The authors have no financial interest to declare in relation to the content of this article.
www.PRSJournal.com
Volume 133, Number 1 • Facial Infiltrating Lipomatosis PATIENTS AND METHODS Patient Ascertainment and Sample Collection After approval by the Committee on Clinical Investigation at Boston Children’s Hospital, abnormal subcutaneous tissue was collected from the cheeks of six patients with facial infiltrating lipomatosis at the time of a clinically indicated procedure. The disorder was diagnosed based on history, physical examination, imaging, and histopathologic analysis. DNA Preparation for Massively Parallel Sequencing Genomic DNA was extracted from each affected tissue specimen and used to prepare sequencing libraries as previously described.8 DNA (3 µg) was physically sheared to generate fragments between 100 and 200 basepairs long. Sequencing primers, which were given barcodes so as to distinguish between the different samples, were enzymatically ligated to the sheared DNA fragments. We performed 15 cycles of the polymerase chain reaction to increase the amount of DNA in each barcoded library, pooled the libraries, and hybridized the pooled DNA to a custom-designed capture array (Agilent Technology 1m SureSelect DNA Capture Array, Agilent Technologies, Santa Clara, Calif.) that contains the coding regions for 26 genes thought to function within the PI3K signaling pathway (AKT1, AKT2, AKT3, FKBP1A, IGF1, KCNH1, KCNH2, KCNK5, MDK, MTOR, NFATC1, NF1, NF2, NOS3, PIK3CA, PIK3CB, PIK3CG, PIK3R1, PIK3R2, PIK3R3, PPP3CA, PPP3CB, PPP3CC, PTEN, PTN, and SLC12A7). The DNA fragments captured by the array were again increased in abundance by performing 17 cycles of polymerase chain reaction and then used to obtain 100-basepair paired end reads on an Illumina HiSeq2 (HiSeq 2000; Illumina, Inc., San Diego, Calif.) sequencing apparatus. DNA Sequence Analysis Paired-end read data were de-barcoded with Novobarcode (Novocraft Technologies, Selangor, Malaysia), aligned to the University of California Santa Cruz human reference genome (GRCh37/ hg19), and filtered to remove polymerase chain reaction duplicates.15,16 Pileup files were generated for each of the six patients using SAMtools.16 VarScan variant detection software was used to list all chromosomal locations containing two or more alternative base calls (relative to the reference base) within the total depth of coverage at the particular base.17 We further filtered the data sequentially to reduce false-positive findings that could have
arisen during library preparation and HiSeq2 sequencing. We considered only bases for which there were 10 or more reads, of which at least three reads contained the variant base. We excluded variants that exhibited significant strand bias; we kept all variants that had fewer than 10 reads, but discarded variants with more than 10 reads if 80 percent or more of the reads occurred in a single direction. For example, we excluded a variant that was found 10 times in one direction and two times in the other direction (10 of 12, or 83.3 percent) and kept a variant that was found nine times in one direction and three times in the other direction (nine of 12, or 75 percent). We filtered variants that were in the 1000 Genomes database,18 the National Heart, Lung, and Blood Institute Exome Variant Server database (http:// evs.gs.washington.edu/EVS/; accessed March, 2013), and build 132 of the database of common single nucleotide polymorphisms,19 since these variants are highly unlikely to represent a somatic mutation, and filtered variants that were not in the coding sequence or at a splice site. Lastly, we ranked variants by their frequency in descending order. A variant with a frequency of 5 percent (e.g., 20 of 400 reads) was ranked higher than a variant with a frequency of 1 percent (e.g., four of 400 reads). We then visually inspected each highly ranked variant using the Integrative Genomics Viewer20 and determined whether any of the variants were present in the Catalog Of Somatic Mutations In Cancer (or COSMIC) database.21 Finally, we determined if any of the 26 genes in the array contained variants in three or more patient samples. Variants that satisfied all criteria were confirmed as true-positive mutations. We re-extracted DNA from affected tissue, polymerase chain reaction–amplified and subcloned the relevant amplimers, and screened individual subclones for the presence of the mutation.
RESULTS Figures 1 and 2 show a photograph and a cranial magnetic resonance image depicting the clinical findings in a patient with facial infiltrating lipomatosis and photomicrographs of hematoxylin and eosin–stained sections representative of the excised affected tissues that were sequenced. All patients had histopathologic findings of mature adipose infiltrating deep into skeletal muscle through dermal and subcutaneous paucicellular fibroconnective tissue. Small foci of entrapped minor salivary glands and abnormal-appearing
13e
Plastic and Reconstructive Surgery • January 2014
Fig. 1. (Left) Four-year-old boy with facial infiltrating lipomatosis. (Right) T1-weighted axial magnetic resonance image of this child demonstrates increased subcutaneous fat and lipomatous infiltration of the right maxilla and mandible, right salivary glands, and muscles of mastication.
Fig. 2. Photomicrographs of excised affected tissues used for mutation detection from four study participants demonstrating mature adipose tissue infiltrating or surrounding (above, left) dermal and subcutaneous paucicellular fibroconnective tissue, (above, right) skeletal muscle fibers (asterisk), (below, left) salivary tissue (asterisk), and (below, right) skeletal muscle (single asterisk) and peripheral nerves (double asterisks). All photomicrographs were obtained from hematoxylin and eosin–stained histologic sections at 200 times magnification. Bar = 100 microns.
14e
76.1% 80.0% 37.8% 71.2% 79.6%
79.8% PCR, polymerase chain reaction.
85.3% 86% 75.9% 79.8% 85.5%
85.7%
5.2% 5.7% 3.1% 9.3% 5.5%
5.2%
1,028,375 (72.4%) 411 1,356,749 (89.6%) 543 531,969 (60.6%) 213 899,204 (93.6%) 360 1,207,544 (92.4%) 483
1,175,250 (93.4%) 470
1,419,655 1,514,952 878,209 1,258,736 960,564
2 1
1,307,510
No. of reads after removal of PCR duplicates Reads mapping to array (% of total reads mapping to array) Average depth of coverage 0× depth sequence coverage of coding sequence 50× depth sequence coverage of coding sequence 100× depth sequence coverage of coding sequence
Table 1. Sequence Coverage for Six Affected Tissue Samples
3
Patient
4
5
6
Volume 133, Number 1 • Facial Infiltrating Lipomatosis nerves were frequently observed in the lesional tissue. Table 1 summarizes the DNA sequence coverage obtained from each specimen. After removing polymerase chain reaction duplicates, more than 878,000 100-basepair paired-end reads were obtained for each sample. These reads were enriched for the 26 genes on the capture array, with 60 percent to 93 percent aligning to the targeted sequence. The average read depth ranged from 213× to 543× per base captured. With the exception of one patient’s tissue that had lower coverage, the samples had coverage of 100× or more across more than 70 percent of the coding sequence on the capture array. Table 2 summarizes the number of bases for which deep sequencing yielded two or more reads with the same single nucleotide variant (SNV). Table 2 also depicts how sequential filtering criteria reduced the number of bases that had candidate disease-causing single nucleotide variants. Note that without filtering, about one in eight bases analyzed, at 10× or more coverage, had at least two reads that contained a single nucleotide variant (e.g., 21,086 bases among the 165,744 bases sequenced had at least two variant reads in patient 1). By requiring that three or more reads contain a variant base call, the frequency dropped to about one in 20 (8398 variant bases among the 165,744 bases). Filtering for strand bias and for probable germline single nucleotide variants (i.e., variants present in the 1000 Genomes, Exome Variant Server, and common single nucleotide polymorphisms databases) had little effect on the frequency of variant bases (about one in 22, or 7512 among the 165,744 bases). Another modest reduction in frequency to about one in 29 (3787 among the approximately 120,000 bases of sequence that represent coding or splice sites) resulted when only exons and splice sites were considered. The greatest reduction occurred when we considered the percentage of reads that contained a variant. As shown in Table 2, the requirement that 1 percent or more of reads at a base contain a variant call reduced the frequency to about one in 118 (935 among the approximately 120,000 bases). Increasing the stringency to 5 percent or more reduced the frequency to about one in 12,200 (nine out of approximately 120,000 bases). Under the latter filtering criteria, 10 of the 26 genes in our targeting array had nonsynonymous, coding sequence variants in three or more patient samples. Only one gene, PIK3CA, had nonsynonymous variants in all patient samples. Importantly, each of the PIK3CA variants
15e
Plastic and Reconstructive Surgery • January 2014 Table 2. Variant Filtering for Six Affected Tissue Samples Patient Bases sequenced* Bases analyzed† Minimum of 10× coverage Total bp with 2 variant reads Total bp with ≥3 variant reads Variants passing strand-bias filter criteria Variants not in databases‡ Variants in exons/splice sites Variants in ≥1% of reads Variants in ≥2% of reads Variants in ≥5% of reads Genes with a ≥5% variant in ≥3 patient samples Genes with a ≥5% variant in all patient samples The ≥5% PIK3CA variants found in COSMIC
1
2
3
4
5
6
523,068 178,549 165,744 21,086 8398
491,196 169,669 152,770 14,163 4765
526,138 179,888 166,907 18,585 6818
541,753 183,788 163,079 8,053 1980
523,387 178,844 166,427 23,047 8925
527,233 179,528 166,081 16,280 5179
7645 7512 3787 935 136 9
4294 4248 2061 777 120 10
6038 5887 2829 882 130 5
1446 1307 599 466 423 38
7977 7814 3916 873 113 13
4500 4333 2045 1251 280 6
10 (AKT1, KCNH1, KCNK5, mTOR, NF1, NFATC1, NOS3, PIK3CA, PIK3R2, SLC12A7) 1 (PIK3CA) All (p.E453K, p.H1047R, p.H1047L, p.E542K)
bp, base pair; COSMIC, Catalog Of Somatic Mutations In Cancer database. *Capture array encompassed 580,000 bases: all exonic regions, 5’ and 3’ untranslated regions, and 90 bases of DNA upstream and downstream of each captured exon. †Bases analyzed included approximately 120,000 bases of coding sequence and splice sites, as well as flanking bases for which paired sequencing reads that contain exonic sequence overlap. ‡The 1000 Genomes, Exome Variant Server, and common variants in single nucleotide polymorphism databases.
we identified had previously been reported as a somatically occurring mutation in the Catalog Of Somatic Mutations In Cancer database.21 In the nine other genes that had variants at a frequency to 5 percent or more, only two of the 35 variants were in the Catalog Of Somatic Mutations In Cancer database; these two variants were in the same sample (patient 4). Table 3 summarizes the depth of coverage and frequency of variant calls for the PIK3CA mutations identified in the six patient samples. After removal of polymerase chain reaction duplicates, the depth of coverage ranged from 121× to 1091× over the sites of the mutations. When polymerase chain reaction duplicates were retained, the depth of coverage increased approximately 10-fold with little to no effect on the frequency of the mutant allele.
DISCUSSION Certain disorders must be the result of somatic mosaicism because complete heterozygosity would be lethal or cannot be transmitted through the egg or sperm.22 Massively parallel sequencing technologies provide a means for identifying these somatic mosaic mutations and were used by us to identify the likely cause of facial infiltrating lipomatosis. We focused our attention on the PI3K signaling pathway that has been implicated in other overgrowth disorders, such as Proteus syndrome,9 CLOVES
16e
syndrome,8 Klippel-Trenaunay syndrome,8 hemimegalencephaly,11,13,14 macrodactyly,10 and lipoadenomatous hyperplasia.12 Furthermore, features seen in patients with facial infiltrating lipomatosis (e.g., fatty overgrowth and neuromas) occur in some patients with CLOVES syndrome. Therefore, our finding that patients with facial infiltrating lipomatosis have the same somatic PIK3CA mutations that occur in patients with CLOVES syndrome was not entirely unexpected. There are several possible explanations for how the same somatic mutation can cause a variety of clinical conditions. First, the phenotype may depend on when the somatic mutation arises during embryogenesis. For example, patients with CLOVES syndrome who have widespread involvement (e.g., face, trunk, extremities, and viscera) could have a somatic mutation that arises earlier in development than patients with regional overgrowth in the face or a digit. Perhaps facial infiltrating lipomatosis is the result of a mutation arising in the first pharyngeal arch, since derivatives of this structure form the maxilla, mandible, muscles of mastication, and the anterior twothirds of the tongue.23,24 A second explanation is that the phenotype may depend on the specific cell type that contains the mutation. Digital overgrowth in patients with isolated macrodactyly follows a “nerve territory,” and the affected tissue
Volume 133, Number 1 • Facial Infiltrating Lipomatosis Table 3. PIK3CA Mutation Frequency in Six Affected Tissue Samples Patient Mutation Total reads over mutation site after PCR duplicate filtering Variant reads over mutation site after PCR duplicate filtering (% mutant reads) Total read over mutation site without PCR duplicates filtering Variant reads over mutation site without PCR duplicate filtering (% mutant reads)
1
2
3
4
5
6
p.H1047R
p.H1047L
p.E452K
p.E542K
p.E453K
p.H1047R
794
318
121
358
70 (9)
73 (23)
36 (30)
58 (16)
7889
2490
1201
2759
743 (9)
577 (23)
362 (30)
497 (18)
1091 339 (31) 10,477 3421 (33)
621 71 (11) 14,644 1036 (7)
PCR, polymerase chain reaction.
is composed of enlarged peripheral nerves.10 A mutation arising in the neural crest cell component of primitive neuroectoderm25 could explain the phenotype in facial infiltrating lipomatosis, since these cells migrate into each of the facial structures affected in this disorder and patients frequently have mucosal neuromas. Genetic background is another possibility for the divergent phenotypes among individuals with the same PIK3CA mutation. Patients with CLOVES syndrome may have modifier alleles that amplify the effect of the postzygotic somatic PIK3CA mutation relative to patients with more regional phenotypes. Modifier alleles have been identified for Mendelian genetic diseases26 and for somatic mutations that give rise to cancer.27 There are two mouse strains with an inducible Pik3ca mutation28,29 that is the same as one of the mutations found in patients with facial infiltrating lipomatosis. These mouse strains could be used to address the first two potential mechanisms. Testing the last mechanism will require the genomes of many patients with PIK3CA-associated overgrowth syndromes to be sequenced and analyzed. Identification of PIK3CA mutations in affected tissue from patients with facial infiltrating lipomatosis utilized a strategy that focused on only 26 of the estimated 20,000 genes in the human genome. Since a priori candidate pathways may not exist for other diseases that are somatic mosaic in origin, would we also have been successful if the entire exome was sequenced? Current commercially available, whole-exome sequencing provides an average coverage of approximately 70×, in contrast to the average coverage of more than 400× we achieved with the targeted array. Nevertheless, despite the lower coverage of whole-exome sequencing, it is likely we would have detected the somatic PIK3CA mutations in most patients
because the mutant alleles would have survived our filtering process. For example, patients 1, 3, 4, and 5, who have PIK3CA mutant allele frequencies of 31, 23, 30, and 16 percent, respectively, would each have had more than three variant reads with more than 5 percent of the reads being variant in whole-exome data at 70× coverage. We had this experience applying whole-exome sequencing in patients with CLOVES syndrome.8 In the current study, patients 2 and 6, whose mutant allele frequencies were only 9 and 11 percent, respectively, should also have been detected with whole-exome sequencing, but they might have been missed due to stochastic variation in allelic sampling. We used a strict filtering strategy to minimize false-positive variants during our sequence analysis. Ultimately, ranking mutations based on frequency and only looking at those above a 5 percent threshold improved the positive predictive value of the variants we identified. For example, one-in-nine, one-in-10, and one-in-five of the variants that survived this filtering strategy in patients 1, 2, and 3, respectively, were the PIK3CA mutations. Setting a lower threshold would have increased the sensitivity for detecting low-frequency mosaic mutations but at the cost of having significantly more false-positive variants to consider as well. A 1 percent threshold was used to identify a somatic missense mutation in GNAQ as the cause of Sturge-Weber syndrome and capillary malformations; however, this search was facilitated since all patients had the identical causative variant.30 Our filtering strategy, which detected different somatic mutations affecting the same gene, should be helpful to other investigators studying phenotypes that result from somatic mosaic mutations with allelic heterogeneity. PIK3CA encodes the 110-kD alpha subunit of the enzyme phosphoinositide-3-kinase (PI3K).31,32 Upon activation, PI3K converts
17e
Plastic and Reconstructive Surgery • January 2014 phosphatidylinositol (3,4)-bisphosphate (PIP2) to phosphatidylinositol (3,4,5)-triphosphate (PIP3).33 PIP3 promotes phosphorylation of AKT family members, which regulate cell proliferation, adhesion, survival, and motility. The PIK3CA mutations we identified in patients with facial infiltrating lipomatosis have previously been identified in several cancers21,34 and have been shown to constitutively activate PI3K.35 Therefore, pharmacologic agents that inhibit the PI3K pathway may benefit patients with facial infiltrating lipomatosis. Several PI3K pathway inhibitors are currently in phase II and III clinical trials for cancers containing PIK3CA mutations.36 Although pharmacotherapy probably would not cure the disorders, it might prevent progression or recurrence. Arin K. Greene, M.D., M.M.Sc. Department of Plastic and Oral Surgery Boston Children’s Hospital 300 Longwood Avenue Boston, Mass. 02115 [email protected]
ACKNOWLEDGMENTS
This work was supported by the Manton Center for Orphan Disease Research at Boston Children’s Hospital and the Howard Hughes Medical Institute. The authors are grateful to their patients for participating in this study and to Dr. Peter Jezewski (University of Alabama at Birmingham) for his insights into craniofacial developmental biology. The authors thank the National Heart, Lung, and Blood Institute Grand Opportunity Exome Sequencing Project and its ongoing studies that produced and provided exome variant calls for comparison: the Lung Grand Opportunity Sequencing Project (HL-102923), the WHI Sequencing Project (HL102924), the Broad Grand Opportunity Sequencing Project (HL-102925), the Seattle Grand Opportunity Sequencing Project (HL-102926), and the Heart Grand Opportunity Sequencing Project (HL-103010). REFERENCES 1. Slavin SA, Baker DC, McCarthy JG, Mufarrij A. Congenital infiltrating lipomatosis of the face: clinicopathologic evaluation and treatment. Plast Reconstr Surg. 1983;72:158–164. 2. Padwa BL, Mulliken JB. Facial infiltrating lipomatosis. Plast Reconstr Surg. 2001;108:1544–1554. 3. Kang N, Ross D, Harrison D. Unilateral hypertrophy of the face associated with infiltrating lipomatosis. J Oral Maxillofac Surg. 1998;56:885–887. 4. Donati L, Candiani P, Grappolini S, Klinger M, Signorini M. Congenital infiltrating lipomatosis of the face related to cytomegalovirus infection. Br J Plast Surg. 1990;43:124–126. 5. Couto RA, Mulliken JB, Padwa BL, et al. Facial infiltrating lipomatosis: expression of angiogenic and vasculogenic factors. J Craniofac Surg. 2011;22:2405–2408.
18e
6. Sapp JC, Turner JT, van de Kamp JM, van Dijk FS, Lowry RB, Biesecker LG. Newly delineated syndrome of congenital lipomatous overgrowth, vascular malformations, and epidermal nevi (CLOVE syndrome) in seven patients. Am J Med Genet A. 2007;143A:2944–2958. 7. Alomari AI. Characterization of a distinct syndrome that associates complex truncal overgrowth, vascular, and acral anomalies: a descriptive study of 18 cases of CLOVES syndrome. Clin Dysmorphol. 2009;18:1–7. 8. Kurek KC, Luks VL, Ayturk UM, et al. Somatic mosaic activating mutations in PIK3CA cause CLOVES syndrome. Am J Hum Genet. 2012;90:1108–1115. 9. Lindhurst MJ, Sapp JC, Teer JK, et al. A mosaic activating mutation in AKT1 associated with the Proteus syndrome. N Engl J Med. 2011;365:611–619. 10. Rios JJ, Paria N, Burns DK, et al. Somatic gain-of-function mutations in PIK3CA in patients with macrodactyly. Hum Mol Genet. 2013;22:444–451. 11. Lee JH, Huynh M, Silhavy JL, et al. De novo somatic mutations in components of the PI3K-AKT3-mTOR pathway cause hemimegalencephaly. Nat Genet. 2012;44:941–945. 12. Lindhurst MJ, Parker VE, Payne F, et al. Mosaic overgrowth with fibroadipose hyperplasia is caused by somatic activating mutations in PIK3CA. Nat Genet. 2013;44:928–933. 13. Rivière JB, Mirzaa GM, O’Roak BJ, et al.; Finding of Rare Disease Genes (FORGE) Canada Consortium. De novo germline and postzygotic mutations in AKT3, PIK3R2 and PIK3CA cause a spectrum of related megalencephaly syndromes. Nat Genet. 2012;44:934–940. 14. Poduri A, Evrony GD, Cai X, et al. Somatic activation of AKT3 causes hemispheric developmental brain malformations. Neuron. 2012;74:41–48. 15. Li H, Durbin R. Fast and accurate short read align ment with Burrows-Wheeler transform. Bioinformatics. 2009;25:1754–1760. 16. Li, H., Handsaker, B., Wysoker, A., et al. The Sequence Alignment/Map format and SAMtools. Bioinformatics 25: 2078–2079, 2009. 17. Koboldt DC, Chen K, Wylie T, et al. VarScan: variant detection in massively parallel sequencing of individual and pooled samples. Bioinformatics. 2009;25:2283–2285. 18. Abecasis, G. R., Auton, A., Brooks, L. D., et al. An integrated map of genetic variation from 1,092 human genomes. Nature 491: 56–65, 2012. 19. Sherry ST, Ward MH, Kholodov M, et al. dbSNP: the NCBI database of genetic variation. Nucleic Acids Res. 2001;29:308–311. 20. Robinson JT, Thorvaldsdóttir H, Winckler W, et al. Integrative genomics viewer. Nat Biotechnol. 2011;29:24–26. 21. Forbes SA, Bindal N, Bamford S, et al. COSMIC: mining complete cancer genomes in the Catalogue of Somatic Mutations in Cancer. Nucleic Acids Res. 2011;39(Database issue):D945–D950. 22. Happle R. Lethal genes surviving by mosaicism: a possible explanation for sporadic birth defects involving the skin. J Am Acad Dermatol. 1987;16:899–906. 23. Cobourne MT, Sharpe PT. Tooth and jaw: molecular mechanisms of patterning in the first branchial arch. Arch Oral Biol. 2003;48:1–14. 24. Müller F, O’Rahilly R. The prechordal plate, the rostral end of the notochord and nearby median features in staged human embryos. Cells Tissues Organs. 2003;173:1–20. 25. LaBonne C, Bronner-Fraser M. Molecular mechanisms of neural crest formation. Annu Rev Cell Dev Biol. 1999;15:81–112. 26. Cutting GR. Modifier genes in Mendelian disorders: the example of cystic fibrosis. Ann N Y Acad Sci. 2010;1214:57–69.
Volume 133, Number 1 • Facial Infiltrating Lipomatosis 27. Ruark E, Snape K, Humburg P, et al.; Breast and Ovarian Cancer Susceptibility Collaboration; Wellcome Trust Case Control Consortium. Mosaic PPM1D mutations are associated with predisposition to breast and ovarian cancer. Nature. 2013;493:406–410. 28. Kinross KM, Montgomery KG, Kleinschmidt M, et al. An activating Pik3ca mutation coupled with Pten loss is sufficient to initiate ovarian tumorigenesis in mice. J Clin Invest. 2012;122:553–557. 29. Yuan W, Stawiski E, Janakiraman V, et al. Conditional activation of Pik3ca(H1047R) in a knock-in mouse model promotes mammary tumorigenesis and emergence of mutations. Oncogene. 2013;32:318–326. 30. Shirley MD, Tang H, Gallione CJ, et al. Sturge-Weber syndrome and port-wine stains caused by somatic mutation in GNAQ. N Engl J Med. 2013.
31. Hiles ID, Otsu M, Volinia S, et al. Phosphatidylinositol 3-kinase: structure and expression of the 110 kd catalytic subunit. Cell. 1992;70:419–429. 32. Volinia S, Hiles I, Ormondroyd E, et al. Molecular cloning, cDNA sequence, and chromosomal localization of the human phosphatidylinositol 3-kinase p110 alpha (PIK3CA) gene. Genomics. 1994;24:472–477. 33. Hemmings BA, Restuccia DF. PI3K-PKB/Akt pathway. Cold Spring Harb Perspect Biol. 2012;4:a011189. 34. Samuels Y, Wang Z, Bardelli A, et al. High frequency of mutations of the PIK3CA gene in human cancers. Science. 2004;304:554. 35. Kang S, Bader AG, Vogt PK. Phosphatidylinositol 3-kinase mutations identified in human cancer are oncogenic. Proc Natl Acad Sci U S A. 2005;102:802–807. 36. Kurtz JE, Ray-Coquard I. PI3 kinase inhibitors in the clinic: an update. Anticancer Res. 2012;32:2463–2470.
19e
