Mutations that compromise the regulation of cellular proliferation and death by reducing apoptosis, or by allowing constitutive cell division, cause pathological tissue growth. The most extreme example is cancer, where numerous activating mutations of genes in growth-promoting pathways have been documented. A commonly affected pathway is the receptor tyrosine kinase/phosphatidylinositol-3-kinase/AKT (RTK/PI3K/AKT) pathway, which harbors activating mutations in most solid tumors1
Given the importance of RTK/PI3K/AKT for cell proliferation and metabolism, germline mutations upregulating the pathway would likely be lethal in embryonic life, and have not been reported. However, mosaic mutations can cause segmental, or patchy, overgrowth disorders. Germline loss-of-function mutations in PTEN, a negative regulator of PI3K signaling , cause overgrowth, and a somatic second hit in PTEN causes the Type II Segmental Cowden Syndrome2
. Proteus syndrome, a progressively deforming regional overgrowth syndrome that affects bones, adipose, and other mesenchymal tissues, is caused by a somatic p.Glu17Lys AKT1 mutation3
which constitutively activates PI3K/AKT signaling3,4
. These findings proved the hypothesis that segmental overgrowth disorders can be caused by somatic mosaicism for mutations5
, and suggested that similar disorders may also be caused by somatic activation of the RTK/PI3K/AKT pathway. We describe here the clinical and molecular characterization of 10 such patients.
Ten patients with a previously uncharacterized progressive segmental overgrowth syndrome were evaluated. Of those for whom records were available, 7 of 8 had congenital overgrowth. Their major manifestation was segmental progressive overgrowth of subcutaneous, muscular, and visceral fibroadipose tissue with skeletal overgrowth ( & ). The overgrowth was predominantly adipose tissue, or a mixture of adipose and fibrous tissues. We have thus designated this as ‘fibroadipose’ tissue. In some affected areas, this tissue was solitary, but in others there was admixture of the fibroadipose tissue with muscle, or investiture within other tissues (e.g., spinal canal). The range of severity and the natural history among the patients was remarkable. Patient C1 had massive overgrowth of her body from the waist down, weighing 117 kg, with leg circumferences of 100-110 cm. Her total adiposity assessed by DXA was 50%, accounted for mostly by her legs. Her overgrowth continued into adulthood. In contrast, C2 had overgrowth limited to an arm and thumb and N110 had overgrowth limited to two rays of one foot (). Patient N7 had an extensive lipoma of the left lower trunk, buttock and leg that necessitated amputation (). Patient N45 had fibroadipose overgrowth of the lower body including a pelvic mass that encased the rectum () with attendant constipation. The overgrowth in this patient continued into adulthood. However, the overgrowth in patient N68 mostly ceased after puberty.
Spectrum of Overgrowth in Patients with Activating PIK3CA Mutations
Segmental Skeletal and Fibroadipose Overgrowth in Patients with Activating PIK3CA Mutations
Eight of the 10 patients have undergone multiple surgical debulking or orthopedic procedures for overgrowth. Skeletal overgrowth was variable in character. Four had skeletal overgrowth with preserved architecture while others had distorting overgrowth (). Patient N45 underwent orthopedic surgery for a 5 cm leg length discrepancy. Bone ages, when assessed, were normal, and no dental anomalies were reported. Several patients had enlarged peripheral nerves, cutaneous vascular malformations, and testicular or epididymal cysts and hydroceles. Four patients had strikingly lipoatrophy in areas not affected by overgrowth, but no evidence of either insulin resistance or hypoglycemia was reported. There have been no malignancies, although patient N108 did have benign nephrogenic rests.
All patients had been suggested to have Proteus syndrome, however on our evaluation none met the clinical criteria for this condition6
(), and 7 of the 10 had congenital segmental overgrowth, which is rare in Proteus syndrome. None had a cerebriform connective tissue nevus. Another diagnostic consideration is Klippel-Trenaunay syndrome (KTS). However, the vascular anomalies in the present patients comprised only capillary vascular malformations and none had the lateral venous anomaly or port wine stains typical of KTS7
. Furthermore, the overgrowth in the present patients was not coincident with the vascular anomalies, which is characteristic of KTS. All patients had apparently normal intellectual and pubertal development and one patient has two unaffected children.
Summary of Clinical Features of Patients
Exome sequencing of dermal fibroblast DNA from the unaffected arm and affected leg of patient C1 identified 45 non-synonymous genetic variants unique to leg-derived fibroblasts and absent from dbSNP and 894 controls. Sanger sequencing of DNA from a freshly cultured dermal fibroblasts from a second skin biopsy confirmed only 2 of these leg-specific genetic variants: the heterozygous PIK3CA mutations c.3140A>T (rs121913279, which predicts p.His1047Leu) and USHBP1
c.280C>G (which predicts p.Glu94Gln) (, Supplementary Figure 1
encodes a putative Usher syndrome-related USH1C binding protein, highly expressed in the heart, and no link of USHBP1
to the phenotype was apparent from existing literature. In contrast, PIK3CA
encodes the p110α catalytic subunit of the growth factor signal-transducing PI3K, and p.His1047Leu has been reported in >130 cancers by COSMIC, while the p.His1047Arg mutation, the commonest cancer-associated PIK3CA mutation8
, stimulates both PI3K signaling and growth. This mutation was selected for further study.
Identification of PIK3CA mutations in affected cells and tissues
Sequencing of DNA from 5 tissues obtained from the amputated left leg of patient C1 confirmed the PIK3CA
mutation at all sites, with mutation burdens of 8% to 39% ( and Supplementary Table 1
) using a custom restriction enzyme-based assay (Supplementary Figure 2
and Supplementary Table 2
). To determine whether mutations at codon 1047 might underlie similar forms of overgrowth we undertook targeted sequencing in 10 other patients with overgrowth syndromes with overlapping clinical features to those of patient C1. In 9 of 10 patients mosaicism for mutations at PIK3CA codon 1047 were identified. The cancer-associated PIK3CA variant, p.His1047Arg, was identified in 7 of the 9 patients, with mutation burdens of <1% to 35% in affected tissues and fibroblast cultures (Supplementary Table 1
)). In 2 out of the 9 patients, the p.His1047Leu mutation was identified, with mutation burdens from 4% to 49%. The mutations were absent from blood and unaffected tissues from 9 of the reported patients, from both parents of 6 patients, and from 51 cell or tissue control samples.
mutations were absent in >5,000 samples in the NHLBI Exome Sequencing Project. In the ClinSeq™ dataset (712 exomes) there were 21 variant sequence reads at this position among >92,000 reads, although no sample had >1 variant read. In the 1000 genomes data9
(933 low coverage genomes), there were >5,000 wild type reads and one variant read.
Both the p.His1047Arg and p.His1047Leu PIK3CA variants have increased kinase activity due to enhanced lipid binding10-12
. We thus assessed PI3K activity in dermal fibroblasts from 3 patients by applying a mass spectroscopic assay for PIP3
) before and after stimulation of cells with EGF. PIP3
levels were increased 2-4X in affected cells at baseline and in response to EGF stimulation (), and basal PIP3
levels in affected cells were indistinguishable from those in controls after stimulation. Basal hyperphosphorylation of downstream AKT and p70 S6 kinase signaling was detected in mutant PIK3CA cells. No amplification of stimulated phosphorylation was observed in affected cells, reflecting the maximum stimulation capacity of the signaling cascade. There was no increased signaling through the MEK/ERK pathway (). We conclude that these patients harbor somatically mutated cells with enhanced basal activity of the PI3K/AKT pathway.
Hyperactivity of phosphatidylinositol-3-kinase in cells harboring PIK3CA mutations
Our description of 10 patients with a previously uncharacterized severe segmental overgrowth syndrome due to somatic occurrence of activating mutations in the p110α catalytic subunit of PI3 kinase consolidates the paradigm of non-Mendelian, non-malignant growth disorders caused by isolated cancer-associated mutations in the PI3K/AKT signaling pathway. Moreover, the finding of distinct and homogeneous underlying genetic defects in stringently-defined Proteus syndrome, which is caused by somatic activation of AKT1, and in the currently delineated syndrome offers a mechanism-based validation of existing diagnostic criteria for Proteus syndrome. The syndrome we delineate does not fulfill the specific diagnostic criteria for PS13
, instead falling within the ill-defined category of segmental overgrowth14
. Attempts have been made to subclassify this group, and the patients reported here may be within category III of a recently proposed system15
. However, we believe that future refinement of understanding of the phenotypic spectrum associated with PIK3CA
mutations affords the prospect of further rationalization of diagnostic schemes.
PI3K signaling activates the serine/threonine kinases AKT1, AKT2 and AKT3. AKT1 is most widely expressed, and is associated with growth16
, consistent with the Proteus phenotype, while AKT2 is highly expressed in insulin-responsive tissues including skeletal muscle, liver, and fat, and is more closely implicated in the metabolic actions of insulin17
. AKT3 is most highly expressed in brain and heart, with lower expression in the tissues affected in the current patients. Recently somatic occurrence of both AKT2 and AKT3 p.Glu17Lys mutants, paralogous to the Proteus-associated AKT1 mutation, have been described. The AKT2 mutation causes severe insulin-independent hypoglycemia, mild asymmetric overgrowth, and progressive obesity18
, while the AKT3 mutation was associated with brain overgrowth19
The role of AKT2 in inhibiting adipose lipolysis and stimulating glucose uptake may explain the aggressive adipose expansion seen with PIK3CA mutations compared to AKT1 mutation alone. We speculate that the adipose tissue paucity in the non-overgrown areas of the patients is caused by chronic negative energy balance of those adipose depots consequent upon the demands of the pathologically growing and energy-sequestering adipose tissue in affected regions. The patients we describe do not exhibit a composite of the phenotypes associated with pathological AKT1, AKT2 and AKT3 activation, lacking key features of each condition such as hypoglycemia, cerebriform connective tissue nevi, and brain overgrowth. Some differences may be attributable to the timing and location of the founder mutation during embryogenesis. Thus, the lack of hypoglycemia and brain overgrowth likely reflect lack of PIK3CA activation in the liver and brain respectively in the current patients. Quantitative differences in the degree of activation of common downstream signaling pathways by the different mutants may also play a role, as may PI3K stimulation of non-AKT-dependent responses including rac
-mediated cytoskeletal reorganization20
. Some of the explanation for the phenotypic discrepancies of PIK3CA and AKT activation may also lie in the complexity of the PI3K enzyme: p110α is only one of three catalytic subunits coupled to receptor tyrosine kinase activation, each of which dimerizes with one of five regulatory subunits, lending significant combinatorial complexity to PI3K function. Specificity of coupling of different PI3K heterodimers to downstream pathways is not fully understood.
Given the prevalence of PIK3CA codon 1047 cancer mutations, a critical consideration is whether patients with these mutations are at increased risk of malignancy. Several observations indicate that such mutations can initiate cancer. Circumstantial evidence comes from the observation of PIK3CA
mutations in the early stages of some human cancers21,22
, while more direct evidence comes from murine studies. Transgenic expression of the Pik3ca
p.His1047Arg mutation in lung23
, or breast epithelium24-26
of mice has been shown to produce malignant tumors. However, in these studies the Pik3ca
mutant was overexpressed, potentially exaggerating its oncogenicity. Expression of Pik3ca
p.His1047Arg at endogenous levels in mouse ovaries did not produce tumors after one year27
. It is thus possible that expression at endogenous levels in the cellular context of human mesodermal lineages has more benign consequences than implied by the mouse models overexpressing mutant Pik3ca. It is of note that codon 1047 oncogenic PIK3CA
mutations have been reported at high prevalence in benign seborrheic keratoses and epidermal nevi in humans28
, demonstrating that there is no obligate association of these mutations to malignancy. However, longitudinal studies are needed to properly assess this potential risk and to formulate surveillance recommendations, should such a risk be identified.
Treatment of segmental overgrowth disorders has relied upon surgical debulking6
and orthopedic procedures to limit growth29
. The identification of activated PI3K/AKT signaling suggests potential utility of targeted therapy, either through inhibition of PI3K, of AKT, or of downstream pathways such as mTORC1, using clinically available drugs including rapamycin, which has been reported to produce major benefits in a child with somatic PTEN deficiency30
. Intensive efforts are underway to develop novel inhibitors for use in cancer. When agents emerge with side effect profiles that are tolerable for long-term use in patients with non-malignant disease, their use in patients with severe progressive segmental overgrowth syndromes offers the prospect of rational mechanism-based therapy.