Clonality and altered behavior See related Commentary, pages 665–666.
of endothelial cells from hemangiomas
Eileen Boye,1 Ying Yu,2 Gretchen Paranya,2 John B. Mulliken,3
Bjorn R. Olsen,1 and Joyce Bischoff2
1Department of Cell Biology, Harvard Medical School, and Harvard-Forsyth Department of Oral Biology, 
Harvard School of Dental Medicine, Boston, Massachusetts, USA
2Department of Surgery, and
3Division of Plastic Surgery, Children’s Hospital, Boston, Massachusetts, USA
Address correspondence to: Joyce Bischoff, Department of Surgery, Children’s Hospital, 300 Longwood Avenue, 
Boston, Massachusetts 02115, USA. Phone: (617) 355-7865; Fax: (617) 355-7043; E-mail: joyce.bischoff@tch.harvard.edu.
Received for publication September 27, 2000, and accepted in revised form January 18, 2001.
Hemangioma, the most common tumor of infancy, is a benign vascular neoplasm of unknown eti-
ology. We show, for the first time to our knowledge, that endothelial cells from proliferating heman-
gioma are clonal, and we demonstrate that these hemangioma-derived cells differ from normal
endothelial cells in their rates of proliferation and migration in vitro. Furthermore, migration of
hemangioma endothelial cells is stimulated by the angiogenesis inhibitor endostatin, unlike the inhi-
bition seen with normal endothelial cells. We conclude that hemangiomas constitute clonal expan-
sions of endothelial cells. This is consistent with the possibility that these tumors are caused by
somatic mutations in one or more genes regulating endothelial cell proliferation.
J. Clin. Invest. 107:745–752 (2001).
Introduction be clonal, and carry a somatic mutation that causes
Cutaneous hemangiomas occur in 5–10% of Caucasian abnormal proliferation. Additionally, it predicts that
infants. The lesions are benign tumors of vascular these mutated cells would behave differently from nor-
endothelial cells (ECs) that exhibit a predictable evolu- mal ECs and maintain these differences independently
tion and duration. Typically, they are not apparent at of their normal physiological environment. The alter-
birth, but appear around the second week of life, grow native extrinsic hypothesis predicts that ECs in heman-
rapidly over the next 6–10 months (proliferating phase), gioma are polyclonal and behave similarly to normal
and then slowly regress over the next 7–10 years (invo- ECs when removed from their in vivo environment.
luting phase) (1). Most hemangiomas are small lesions, In an attempt to elucidate the molecular pathogene-
but about 10% grow rapidly and because of their size sis of hemangiomas, we tested the intrinsic hypothesis,
and/or location can be problematic and even life threat- i.e., that the tumors are caused by clonal expansion of
ening. The proliferating and involuting phases of the vascular ECs. We isolated ECs from proliferating
hemangioma “life cycle” are not distinct, but represent hemangioma in nine infants and from multifocal
a gradual shift in the balance of mitotic and apoptotic hemangioendothelioma lesions in one infant. We
activity in the local EC population. The mitotic rate is assayed cell samples from eight of the ten patients for
high during the proliferating phase, and extensive pock- monoclonality and found all to be clonal. We further
ets of rapidly proliferating ECs are observed. During the demonstrated that hemangioma-derived ECs differ
early involuting phase (1–2 years of age), mitosis gradu- from normal ECs in rate of proliferation and migration
ally diminishes and apoptosis predominates, with ECs in vitro, as well as in their response to an angiogenesis
making up about a third of the dying cells (2). inhibitor endostatin.
The nature and cellular location of the primary defect Our results indicate that hemangiomas do indeed con-
responsible for triggering, maintaining, and arresting stitute clonal expansions of abnormal ECs. The findings
this abnormal endothelial proliferation are unknown. are consistent with the possibility that hemangiomas are
The defect could either be an inherent characteristic of caused by somatic mutation(s) in a gene(s) that regulates
the ECs themselves (3) or a secondary response to an EC proliferation.
external abnormality, such as the upregulation of an
angiogenic factor or the downregulation of an angio- Methods
static factor, in the immediate environment of the Patient material. Proliferating-phase hemangioma spec-
tumor (4). The former hypothesis would implicate a imens were obtained from nine female infants. The
somatic mutation in a factor involved in control of EC patients’ ages at the time of resection, the location and
proliferation. This hypothesis requires that all the ECs size of the lesion(s), any prior treatment, and other rel-
in the lesion originate from the same progenitor cell, i.e., evant clinical information are given in Table 1. In addi-
The Journal of Clinical Investigation | March 2001 | Volume 107 | Number 6 745
tion, a patient (patient 8) with multiple high-grade nonimmune rabbit IgG as a control; mouse monoclonal
hemangioendothelioma (heoma) lesions of variable anti-human E-selectin IgG1 (clone 4B12; 5 µg/ml; kindly
size was included in the study. Normal skin samples provided by D. Hollenbaugh, Bristol-Myers Squibb) (7)
were obtained from age-matched female infants under- mouse monoclonal anti-CD31/PECAM-1 IgG1 (DAKO,
going other surgical procedures in which these tissues Carpinteria, CA), with an isotype-matched control mouse
would have been discarded. The tissue specimens were IgG1 (5 µg/ml; Becton Dickinson, Bedford, Massachu-
identified by age and gender only, in accordance with setts, USA) as a control; biotinylated goat anti-rabbit
the protocol approved by the Committee on Clinical IgG1, biotinylated horse anti-mouse IgG, and fluorescein-
Investigation at Children’s Hospital, Boston, and were conjugated avidin (Vector Laboratories Inc.). In addition,
collected and processed immediately after resection. we used anti-KDR (Flk1; Santa Cruz Biotechnology Inc.,
Normal blood samples from 40 age-matched female Santa Cruz, California, USA), and mouse anti-human
infants, from a pool of samples left over from routine smooth muscle cell α-actin (Sigma Chemical Co.) to stain
hematological tests, were obtained from the Hematol- many of the EC and UBC samples.
ogy Laboratory at Children’s Hospital, Boston, in For immunostaining, cells were grown in EBM-B on
accordance with the protocol approved by the Com- gelatin-coated glass coverslips, washed in PBS contain-
mittee on Clinical Investigation. These samples were ing 1.26 mM CaCl2 and 0.8 mM MgSO4, and fixed with
identified by the child’s age and gender only. methanol for 10 minutes on ice. Immunostaining was
Isolation, culture, and characterization of hemangioma-derived performed at room temperature, and cells were washed
ECs. ECs were isolated from hemangiomas (hemECs) and three times in PBS after each step. PBS with 1% BSA
normal skin (HFSECs) by the method described previ- was used for antibody dilution. Cells were mounted
ously by Kräling and Bischoff (5) for human dermal with Fluoromount G (Southern Biotechnology Associ-
microvascular ECs (HDMECs). The only modification ates Inc., Birmingham, Alabama, USA) and observed at
was trypsinization of the hemangioma tissue for 4 min- 400× with an Axiophot II fluorescence microscope
utes at 37°C instead of 10 minutes. The cells were resus- (Zeiss, Oberkochen, Germany). Photographs were
pended in EBM-A (EBM131 (Clonetics, San Diego, Cali- taken with Kodak TMAX p3200 film (Eastman Kodak
fornia, USA), 20% heat-inactivated FBS (Hyclone Co., Rochester, New York, USA).
Laboratories, Logan, Utah, USA), 2 mM glutamine, 100 DNA isolation and clonality assay. The Puregene DNA
U/ml penicillin, 100 µg/ml streptomycin (Irvine Scien- isolation Kit (Gentra Systems Inc., Minneapolis, Min-
tific, Santa Ana, California, USA), 0.25 µg/ml ampho- nesota, USA) was used to extract DNA, according to the
tericin B (Life Technologies Inc., Gaithersburg, Maryland, manufacturer’s instructions, from peripheral blood
USA), 0.5 mM N-6,2′-O-dibutyryladenosine 3′:5′-cyclic leukocytes (PBLs) of 40 control individuals and from
monophosphate (Sigma Chemical Co., St. Louis, Mis- cultured ECs at passages 4–7 from ten hemangioma
souri, USA), 1.0 µg/ml hydrocortisone (Sigma Chemical lesions (1, 4, 5, 10, 12, 13, 17A, 17B, 20, and 21), four
Co.) and grown overnight at 37°C with 5% CO2, washed hemangioendothelioma lesions (8B, 8C, 8F, and 8G),
vigorously with PBS to remove unattached cells, and fed and five normal skin samples. In addition, DNA was
with fresh EBM-A. These primary cultures were grown to isolated from the cultured unbound fibroblast-like
preconfluence (5–7 days). The ECs were purified from the cells (UBCs) from hemangioma lesions 1 and 10.
primary culture using Ulex europaeus I lectin–coated (Vec- Paraffin-embedded tissue sections of myometrium
tor Laboratories Inc., Burlingame, California, USA) mag- and leiomyomata from hysterectomy specimens were
netic beads (Dynal A.S., Lake Success, New York, USA), as provided by B. Quade (Department of Pathology,
described by Jackson et al. (6). ECs bound to the lectin- Brigham and Women’s Hospital, Boston, Massachu-
coated beads were collected with a magnetic particle con- setts, USA). These samples, previously shown to be
centrator, and any unbound cells were removed with five clonal (8), were deparaffinized, proteinase K–treated,
washes in HBSS wash buffer (5% FBS, 100 U/ml peni- and used as a positive controls for the HUMARA assay.
cillin, 100 µg/ml streptomycin [Irvine Scientific, Santa We analyzed these DNA samples for clonality using
Ana, California, USA], and 0.25 µg/ml amphotericin B the X-linked human androgen receptor gene
[Life Technologies Inc.]). The ECs and unbound fibrob- (HUMARA) assay described by Allen et al. (9). Exon 1 of
last-like cells (UBCs) were separately resuspended in the HUMARA gene contains a highly (90%) polymor-
EBM-A and grown to confluence on gelatin-coated 60- phic CAG short tandem repeat (STR) (11–31 copies)
mm plates. To expand further the purified cultures after with adjacent HhaI and HpaI sites, at which methylation
P2, the cells were passaged 1:3 into a simplified medium correlates with X-chromosome inactivation. In females
EBM-B (EBM131, 10% heat inactivated FBS, 2 mM glut- heterozygous at this allele, STR size polymorphism dis-
amine, 100 U/ml penicillin, 100 µg/ml streptomycin, 2 tinguishes the maternal and paternal X chromosomes,
ng/ml bFGF [kindly provided by Scios Inc., Mountain and digestion with methylation-sensitive HhaI distin-
View, California, USA]) (5). guishes between the unmethylated, active X, which is
Indirect immunofluorescence staining was performed cut, and the methylated, inactive X, which is not.
to characterize the isolated cells. The reagents used A total of 5 µg of DNA was digested in a 50-µl reaction
included: rabbit polyclonal anti-human vWF (diluted with 20 units of HhaI (New England Biolabs Inc., Bever-
1/2,000; DAKO Corp., Carpinteria, California, USA) with ly, Massachusetts, USA) at 37°C for up to 24 hours. For
746 The Journal of Clinical Investigation | March 2001 | Volume 107 | Number 6
each sample, PCR was performed on 1 µl of the digest On the basis of other studies (10, 11) and a compari-
and 50 ng of undigested DNA. PCR reactions (50 µl) con- son with samples of leiomyomata previously shown to
tained 20 nM each of primers: H1 5′-GCT GTG AAG be clonal (see earlier discussion here), we defined sam-
GTT GCT GTT CCT CAT-3′ and H2 5′-TCC AGA ATC ples as clonal when there was significant skewing
TGT TCC AGA GCG TGC-3′ (9) and 0.5 µl α-33P dATP toward one allele after HhaI digestion, so that the
(1,000–3,000 Ci/mmol, 10 mCi/ml; ICN Radiochemicals amplification ratio of the minor allele to the predomi-
Inc., Irvine, California, USA). The amplification was car- nant allele was 0.45, corresponding to a ratio between
ried out with 0.5 units of Taq DNA Polymerase (QIAGEN the two alleles of about 30:70.
Inc., Valencia, California, USA) in a 10X buffer, contain- Cell proliferation assay. ECs were plated onto gelatin-
ing 15 mM MgCl2, supplied by the manufacturer. The coated 24-well plates at a density of 8,000 cells per 2
PCR conditions were as follows: an initial denaturation cm2 well in EBM131 medium with 5% FBS and 2 mM
at 93°C for 3 minutes, followed by 30 cycles of denatura- glutamine, 100 U/ml penicillin, and 100 µg/ml strep-
tion at 93°C for 40 seconds, annealing at 65°C for 40 sec- tomycin. The following day, the plating efficiencies of
onds, and extension at 72°C for 40 seconds. These steps the cells were determined and the proliferation of
were followed by a final extension at 72°C for 5 minutes. HDMECs and hemECs were compared after a 48-hour
The PCR products were electrophoresed in an 8% incubation with 1 ng/ml bFGF or 10 ng/ml VEGF
nondenaturing polyacrylamide gel (Accugel 29:1; (R&D Systems Inc., Minneapolis, Minnesota, USA) at
National Diagnostics, Atlanta, Georgia, USA) at 600 V 37°C with 5% CO2. Assays were performed in quadru-
for approximately 10 hours and visualized by exposure plicate, and cells were counted in a Coulter counter.
to autoradiographic film (Eastman Kodak Co.) or The results were expressed as mean ± SD.
phosphorimaging screens (Fuji Medical Systems Inc., Cell migration and endostatin response assays. Migration
Stamford, Connecticut, USA). assays for ECs were performed in a standard 48-well
An initial PCR was used to look for informative alle- chemotaxis chamber (Neuro Probe Inc., Gaithersburg,
les and homozygous individuals were excluded from Maryland, USA) according to the method described by
the study. The NIH Image (version 1.62) software was Yamaguchi et al. (12). HemECs and HDMECs were
used to obtain densitometric scans for each lane on the assayed in EBM131 medium containing 2 mM gluta-
gels. Defined bands representing the different CAG mine, 100 U/ml penicillin, 100 µg/ml streptomycin, and
alleles were seen as major peaks that could be easily dis- 0.025% BSA. A total of 10,000 cells per well were placed in
tinguished from “stutter” bands, seen as minor peaks the upper chamber and their migration across a Nucleo-
slightly out of phase with the main peaks. The intensi- pore polyvinylpyrrolidone-free polycarbonate membrane
ty of each band and its stutter bands was determined as (Corning-Costar Corp., Cambridge, Massachusetts, USA)
the total area under the corresponding peaks in the with 8-µm pores was stimulated with 5 ng/ml VEGF165
densitometric tracing. The ratio of the least intense to (R&D Systems Inc.) in EBM131 with 0.025% BSA, placed
the most intense CAG allele in each heterozygous sam- in the lower chamber. Assays were performed in quadru-
ple represented the amplification ratio. A ratio of 1.00 plicate for 4–6 hours, and the migrated cells, adhered to
represents random X-inactivation; a ratio of 0.00 rep- the underside of the membrane, were stained with Diff-
resents complete skewing and monoclonality. Amplifi- Quick stain (VWR Scientific Products, Bridgeport, New
cation ratios determined before and after HhaI diges- Jersey, USA); nuclei were counted using a light micro-
tion were compared as a measure of clonality. scope. Results were plotted as mean ± SD.
Table 1
Clinical data: ages of patients at the time of resection of each lesion, location and size of the lesions, and other relevant information
Patient number Lesions Age at time of resection Location and size Comments
1 hemEC-1 8 months 1.8 × 0.8 cm, right eyelid Normal karyotypeA
4 hemEC-4 3 months 3 × 3 cm, right flank
5 hemEC-5 4 months 5 mm, lower abdomen Multiple hemangiomas
8 heomaEC-8B 2 years B - ankle Normal karyotypeA, multiple 
heomaEC-8C C - knee hemangioendothelioma lesions
heomaEC-8F F - right foot of variable size, new lesions still
heomaEC-8G G - inner thigh appearing at 2 years
10 hemEC-10 2 years 4 cm, forehead 
12 hemEC-12 4 months 1.5 cm, left cheek
13 hemEC-13 9 months 3 cm, left upper eyelid Other hemangiomas on neck and labia
17 hemEC-17A 10 months A: 3 cm, scalp
hemEC-17B B: 3.5 × 4 cm, forehead
20 hemEC-20 8 months 2.5 cm, neck Multiple hemangiomas
21 hemEC-21 9 months 6.5 × 5.5 cm, scalp Other hemangiomas on 
right ear and forehead
AECs from these patients showed a normal (46, XX) karyotype.
The Journal of Clinical Investigation | March 2001 | Volume 107 | Number 6 747
EC samples were pretreated with human recombi- ure 1c), providing further confirmation of an endothe-
nant endostatin (hES) prepared as described previous- lial phenotype. Diffuse background staining was
ly (12), at concentrations of 1, 10, and 100 ng/ml, by observed with a isotype-matched control IgG1 (Figure
incubation at 37°C for 30 minutes, and the migration 1d). HemECs from all patients showed similar patterns
assay was performed as already described here. of expression of vWF, CD31, and inducible E-selectin,
and in addition, expressed other markers of the
Results endothelial lineage such as KDR, TIE-2, and VE-cad-
Clinical data. As summarized in Table 1, the patients herin (data not shown) but did not express the mes-
selected for the study ranged in age from 3 to 26 months enchymal cell marker, smooth muscle α-actin.
at the time of resection of their hemangioma; three of In hemEC-12 and hemEC-17A, 10–30% of the cells
nine had multiple lesions. Patient 8 had more than 100 appeared to be vWF and CD-31 negative, indicating
rapidly proliferating vascular lesions at age 2 years, and that these cultures were contaminated with UBCs. Cul-
new lesions continued to appear. The lesions failed to tures of unbound cells from hemangiomas 1 and 10
respond to drugs used to treat hemangioma, such as cor- were also negative for EC markers vWF or CD31.
ticosteroid, IFNα-2b, and cyclophosphamide. Further- HemEC-12 was reselected with Ulex europaeus I–coated
more, resected tissue from the lesions did not react with beads to give an essentially pure EC culture.
anti-GLUT1 antibodies (H. Kozakewich, personal com- Clonality of hemangioma ECs. The experimental design
munication), as has been described for the majority of for assessing clonality, using the X-linked human
hemangiomas (13). Therefore, these lesions have been androgen receptor (HUMARA) gene is described in
described as hemangioendotheliomas. Additional evi- Methods. A DNA sample from each individual was first
dence that this patient’s tumors were not hemangiomas tested for heterozygosity of HUMARA STRs, by direct
was provided by the response of ECs to endostatin (see amplification using primers H1 and H2. Patients 5 and
later here), which is the opposite of that for the heman- 13, as well as ten of 40 control blood leukocyte samples,
gioma cell samples. were excluded because they were found to be homozy-
Characterization of isolated ECs. To verify that the iso- gous. This is about twice the 10% homozygosity previ-
lated Ulex europaeus I bead–bound cells (see Methods) ously reported for this allele (9, 14). Blood leukocyte
were indeed endothelial and free from contaminating DNA from the remaining 30 controls was digested with
non-ECs, we analyzed expression of endothelial-specif- HhaI and then amplified with H1 and H2 primers.
ic markers vWF, CD31/PECAM-1, and E-selectin by The results demonstrated that 30% (9/30) of the 30
indirect immunofluorescence. hemEC-1 exhibited normal control samples had skewed X chromosome
punctate cytoplasmic staining with anti-vWF (Figure inactivation in their blood cells, confirming a previous
1a), consistent with localization of vWF in Weibel- report (15), and suggesting that X-inactivation patterns
Palade bodies, a definitive feature of ECs. are tissue specific. Thus, PBLs would be poor controls
CD31/PECAM-1 was seen concentrated at cell-cell bor- for study of clonality in cutaneous endothelium. There-
ders (Figure 1b), consistent with its role as a cell adhe- fore, we decided to use human dermal derived ECs
sion molecule. To upregulate E-selectin, hemEC-1 cells from normal female skin (HFSECs) as controls. Ethical
were treated with LPS before immunostaining. E- considerations prevented us from obtaining such cells
selectin was readily detected upon LPS induction (Fig- from the hemangioma patients in the study, so we
Figure 1
Characterization of hemangioma-
derived endothelial cells by indirect
immunofluorescence. Hemangioma-
derived endothelial cells (hemEC-1)
were fixed with methanol and incubat-
ed with anti-vWF (a), anti-
CD31/PECAM-1 (b), anti–E-selectin
(c), or isotype-matched control mouse
IgG1 (d). Endothelial cells in c were acti-
vated with 0.2 ng/ml LPS for 5 hours to
upregulate E-selectin (22). Bar, 10 µm.
748 The Journal of Clinical Investigation | March 2001 | Volume 107 | Number 6
obtained them from skin of age-matched unrelated
female patients, as described in Methods. Isolated ECs
from hemangioma patients and five controls were sub-
jected to HUMARA assay. Cells were analyzed at
sequential trypsin passages (P4, P5, P6, and P7) to
ensure that the cells were not becoming clonal as a
result of the prolonged culturing process. For each
DNA sample, two independent HhaI digests and prod-
ucts of at least two independent amplification reac-
tions from each digest were analyzed. Table 2 shows the
mean amplification ratios of the HUMARA alleles and
standard deviations; a ratio of 0.45 or less was consid-
ered to represent clonality. In three of five (60%) of the
control EC samples (HFSEC 3, 4, 11), both X chromo-
somes were equally amplified, at all passages tested (P4
to P7, Figure 2a), whereas two of the control samples
(HFSEC 1 and 10) exhibited skewing. ECs from all
eight (100%) hemangioma lesions showed similarly sig-
nificant skewing to one allele, indicating clonality of
those cell populations (Figure 2b). In contrast, UBCs
from hemangiomas 1 and 10 showed almost equal
amplification of both X chromosomes (Figure 2c).
Hemangioma lesions 1, 4, 10, 17B, 20, and 21 showed
complete absence of one allele (Figure 2b). Lesions 12
and 17A initially showed significant, but not complete,
skewing toward one of the two alleles. Staining of
hemEC-12 and -17A suggested some contamination
with UBCs (see earlier discussion here). To determine
whether this contamination could contribute to the
incomplete skewing, hemEC-12 was reselected (see
Methods) and reassayed for clonality. After reselection, Figure 2
Results of HUMARA assay on DNA from ECs of control infants’ skin,
hemEC-12r showed complete skewing (Table 2). hemangioma lesions, and unbound fibroblast-like cells. Amplification
DNA from five of eight lesions exhibited skewing to was performed before (–) or after (+) HhaI digestion. For each cell sam-
the larger of the two alleles; the remaining three ple shown, the individual is heterozygous for CAG repeat size and has
showed skewing to the smaller allele (Table 2). Two two different alleles before HhaI digestion, represented by two major PCR
lesions from patient 17 had discordant skewing. In products on gel. Fainter (“stutter”) bands are products of slippage by
summary, these data demonstrate that all heman- DNA polymerase in STR region. (a) Amplification of DNA from ECs cul-
giomas studied here are clonal. tured at P4 to P7 from skin of two control female infants HFSEC-3 (left)
Differences in rates of proliferation and migration between and HFSEC-11 (right). In each sample, two alleles are equally amplifi-
able both before and after digestion. (b) Amplification of DNA from cul-
HDMECs and hemECs. Proliferation and migration of tured hemECs from patient 1 (left) at P4 and P6 and patient 21 (right)
some hemECs were measured in comparison to that of at P5, P6, and P7. In both cases, one allele completely disappears after
HDMEC samples from infant foreskin that had been HhaI digestion. This selective amplification of one allele indicates that it
isolated and cultured under identical conditions. Given is always methylated and not subject to digestion by HhaI and, therefore,
that it is well known that clones of normal ECs can is the only inactive allele in cell population. C is a positive control for the
exhibit variations in growth rate and respond differ- HUMARA assay. (c) Amplification of DNA from cultured unbound,
ently to growth factors (16), hemEC samples were com- fibroblast-like cells of hemangioma 10 at P4 (UB-P4). For comparison,
pared with six different control samples (HDMECs) for DNA from cultured hemEC-10 at P5 (EC-P5) is shown at right.
proliferation and/or migration properties.
The results of proliferation assays are shown in Fig-
ure 3. HemEC-1, hemEC-4, heomaEC-8, and hemEC- maEC-8, and hemEC-17 also proliferated three times
17 cells proliferated approximately 2.5 times faster faster than HDMECs in response to 10 ng/ml VEGF.
than did HDMECs, both in the presence or absence of The increased proliferation rate was also observed
exogenous bFGF at 1 ng/ml. This difference was main- when the cells were plated in gelatin-, fibronectin-, or
tained with continued passaging of cells, at least up to type IV collagen–coated culture wells (data not shown).
passage 11 (data not shown). HemECs and HDMECs In migration assays, hemEC-1 (Figure 3b), hemEC-
did not appear to differ in the degree to which they 10, and hemEC-21 cells showed an approximately 3.5-
were stimulated by bFGF, however, indicating that fold better response than did different HDMECs in the
hemEC cells are not producing saturating concentra- presence of 5 ng/ml VEGF. These data demonstrate
tions of bFGF. HemEC-1 (Figure 3a), hemEC-4, heo- that hemECs derived from infantile hemangiomas
The Journal of Clinical Investigation | March 2001 | Volume 107 | Number 6 749
Figure 3
Comparison of in vitro properties of HDMECs vs. hemEC-1. (a) Comparison of proliferation rates of HDMECs vs. hemEC-1 in response to
VEGF. HemEC-1 cells proliferate approximately 2.5-fold faster than HDMECs, both in the presence (10 ng/ml) or absence of exogenous
VEGF. (b) Comparison of VEGF-induced migration of HDMECs vs. hemEC-1. HemEC-1 cells migrate approximately 3.5-fold faster than do
HDMECs with (10 ng/ml) or without VEGF stimulation.
retain altered proliferation and migration responses were stimulated in their migration toward VEGF in a
when removed from their in vivo environment. This is dose-dependent manner, even at concentrations of 1
consistent with the clonality data and supports the ng/ml of endostatin.
hypothesis that the defect is intrinsic to the ECs. Patient 8. Patient 8 was included in the study on the
Altered migratory response of hemECs to endostatin. In basis of having multiple rapidly proliferating vascular
light of these observations, we tested whether the lesions described as high-grade hemangioendothe-
increased migration was influenced by the angiogene- lioma (heoma). Clonality analysis showed skewing
sis inhibitor, endostatin (hES). Figure 4 shows the sur- toward the smaller allele in heomaECs isolated from a
prising result obtained. In contrast to the inhibition number of lesions (8B, 8C, 8F, and 8G) resected from
of migration effected by hES on HDMECs (Figure 4a) different parts of her body. ECs from this patient pro-
and other EC types in the presence of VEGF (12), liferated and migrated faster than normal, like the
hemEC-1 (Figure 4b), hemEC-10, and hemEC-21 cells hemECs, but they also behaved similarly to control
ECs, in that they were inhibited by hES in their VEGF-
stimulated migration (Figure 4c).
Table 2
Results of clonality assay showing the mean amplification ratios of the Discussion
smaller to larger HUMARA alleles, obtained at various passages, for Identification of factors that trigger and arrest the
each sample. The predominant allele in each case is also indicated.
growth of hemangiomas will be invaluable in our
Cell sample Average Standard Predominant understanding of vasculogenic and angiogenic process-
amplification ratio deviation allele es and related diseases. We hypothesize that whatever
HFSEC-1 0.00 0.00 S these factors are, they control EC proliferation and are
HFSEC-3 0.84 0.10 - altered as a result of somatic mutations. The results pre-
HFSEC-4 0.84 0.07 - sented here are consistent with this hypothesis. We have
HFSEC-10 0.02 0.00 S isolated cells that express EC specific markers, such as
HFSEC-11 0.96 0.06 - vWF, E-selectin, and CD31/PECAM-1, from proliferat-
hemEC-1 0.02 0.00 L ing hemangiomas and we have found that seven of
hemEC-4 0.02 0.01 L seven patients have lesions that are clonal in nature. We
heomaEC-8B 0.02 0.03 S provide further support for this hypothesis by showing
heomaEC-8C 0.06 0.01 S
heomaEC-8F 0.01 0.01 S that hemangioma cells proliferate and migrate more
heomaEC-8G 0.02 0.05 S rapidly than normal ECs, as would be predicted from
hemEC-10 0.05 0.08 S the nature of the lesions. These characteristics are main-
hemEC-12 0.43 0.17 L tained even when the cells are removed from their phys-
hemEC-12r 0.01 0.01 L iological environment and are consistent when com-
hemEC-17A 0.43 0.10 S pared with several different normal EC samples, which
hemEC-17B 0.05 0.06 L themselves may exhibit a large degree of variation in
hemEC-20 0.04 0.08 L growth rate and response to growth factors (16).
hemEC-21 0.02 0.04 S The finding of discordance of X-inactivation patterns
UBC-1 0.85 0.04 - in two lesions from the same individual (patient 17)
UBC-10 0.91 0.07 -
could be explained by a mutation that occurred rela-
L, larger allele, S, smaller allele. tively early in development before the establishment of
750 The Journal of Clinical Investigation | March 2001 | Volume 107 | Number 6
Figure 4
Effect of hES on VEGF-induced migration of HDMECs,
hemEC-1, and heomaEC-8B. (a) Migration of HDMECs is
inhibited by hES. Complete inhibition is achieved at approxi-
mately 10 ng/ml hES. (b) HemEC-1 cells exhibit the opposite
response to the other cells; their migration is strongly stimu-
lated by hES even at 1 ng/ml. (c) Migration of heomaEC-8B is
inhibited by hES.
permanent X-inactivation; alternatively, it could be the
result of two independent mutations. Because heman-
gioma occurs in up to 10% of Caucasian infants, two
hemangiomas caused by independent mutations in
one individual would be relatively common as well.
Previous studies (17, 18) demonstrate that normal
tissues can contain fairly large patches of cells that (20). Changes in the balance of these factors, due to
are uniform with respect to X-chromosome inactiva- alterations in the factors themselves or in regulatory
tion. These include the aorta and coronary arteries molecules, can effect these angiogenic responses. For
(19), where clonal patches of developmental origin example, in hemangioma, an anomalous stimulus pro-
exist for smooth muscle cells. Such monoclonal duced by an abnormality of regulation or expression of
regions can explain why a significant fraction (∼ 30%) a specific gene that controls EC proliferation could
of aortic tissue samples show skewing in HUMARA cause abnormal angiogenesis, with subsequent activa-
assays (19). Although such data are not available for tion of EC apoptosis (2).
vascular ECs, we assume that similar developmental Walter et al. (21) presented evidence for linkage in
patches may exist. This would explain why two of five three familial cases of hemangioma to chromosome 5q,
(40%) of the normal EC samples showed monoclon- in a region containing three genes involved in vessel
ality. Given this frequency of clonality in the control growth: PDGF-Rβ, FGFR-4, and FLT-4. Several other
samples, the likelihood that all the hemangioma genes and their protein products are known to be
samples would show clonality by chance is less than involved in EC proliferation and angiogenic regulation,
10–3. We conclude, therefore, that the uniform X- including inducers, such as FGFs, VEGF, and ANG1,
inactivation patterns seen in hemangiomas are con- and inhibitors, such as angiostatin, endostatin, and
sequences of the disease process. Non-ECs within the thrombospondin (20). These are all obvious candi-
lesions are likely to be normal, as the unbound dates, mutations in which could produce the abnormal
fibroblast-like cells isolated from several lesions vessel growth seen in hemangioma. Alternatively,
showed random X-inactivation patterns. mutations in upstream or downstream targets of these
Endothelium is generally quiescent under normal genes could produce the same effect.
physiological conditions after birth, but will undergo The response of hemECs to the angiogenesis
localized proliferation in response to specific physio- inhibitor hES was unexpected. VEGF-induced migra-
logical or pathological stimuli, such as injury or tumor tion in three of three hemEC samples tested was stim-
growth. Angiogenesis can be halted and the quiescent ulated instead of inhibited by hES. It has been report-
state restored by antiangiogenic stimuli, suggesting ed that VEGF-induced migration of ECs is inhibited by
that it is a tightly regulated process, controlled locally hES at various concentrations and under different con-
by the levels of angiogenic and antiangiogenic factors ditions (12), suggesting that hES and VEGF may regu-
The Journal of Clinical Investigation | March 2001 | Volume 107 | Number 6 751
late a common signaling pathway for migration. In this 3. Mulliken, J.B., Fishman, S.J., and Burrows, P.E. 2000. Vascular anomalies.
pathway, hES may control a rate-limiting step that In Current problems in surgery. S.A. Wells and L.L. Creswell, editors. Mosby.St. Louis, Missouri, USA. 517–584.
results in the inhibition of VEGF-stimulated migra- 4. Berard, M., et al. 1997. Vascular endothelial growth factor confers a growth
tion. Conceivably, alterations in signaling pathways advantage in vitro and in vivo to stromal cells cultured from neonatal
hemangiomas. Am. J. Pathol. 150:1315–1326.
downstream of the VEGF-receptor in hemangioma 5. Kräling, B.M., and Bischoff, J. 1998. A simplified method for growth of
ECs, as a result of a somatic mutation, could convert human microvascular endothelial cells results in decreased senescence and
the inhibitory effect of endostatin to one of stimula- continued responsiveness to cytokines and growth factors. In Vitro Cell. Dev.Biol. Anim. 34:308–315.
tion. The inhibitory effect of hES on ECs from patient 6. Jackson, C.J., Garbett, P.K., Nissen, B., and Schrieber, L. 1990. Binding of
8 may prove useful as an experimental tool, as it human endothelium to Ulex europaeus I-coated Dynabeads: application
to the isolation of microvascular endothelium. J. Cell Sci. 96:257–262.
enabled us to clearly distinguish this patient’s tumors 7. Kräling, B.M., et al. 1996. E-selectin is present in proliferating endothelial
from other typical hemangiomas. cells in human hemangiomas. Am. J. Pathol. 148:1181–1191.
We believe our data provide strong support for the 8. Quade, B.J., et al. 1997. Disseminated peritoneal leiomyomatosis. Clonali-
ty analysis by X chromosome inactivation and cytogenetics of a clinically
hypothesis that hemangiomas are caused by an intrin- benign smooth muscle proliferation. Am. J. Pathol. 150:2153–2166.
sic abnormality of ECs. However, it is possible, that all 9. Allen, R.C., Zoghbi, H.Y., Moseley, A.B., Rosenblatt, H.M., and Belmont, J.W.
hemangiomas are not due to the same underlying 1992. Methylation of HpaII and HhaI sites near the polymorphic CAGrepeat in the human androgen-receptor gene correlates with X chromo-
defect. Thus in some cases, the primary defect could some inactivation. Am. J. Hum. Genet. 51:1229–1239.
exist external to the proliferating ECs as suggested by 10. Murry, C.E., Gipaya, C.T., Bartosek, T., Benditt, E.P., and Schwartz, S.M.
1997. Monoclonality of smooth muscle cells in human atherosclerosis. Am.
Berard et al. (4). It is also possible that even in heman- J. Pathol. 151:697–705.
giomas caused by somatic mutations in ECs, different 11. Lee, S.D., et al. 1998. Monoclonal endothelial cell proliferation is present
genes may be involved in different patients. Elucidation in primary but not secondary pulmonary hypertension. J. Clin. Invest.101:927–934.
of the mutated genes in each case will enhance our 12. Yamaguchi, N., et al. 1999. Endostatin inhibits VEGF-induced endothelial
understanding of the molecular control of EC prolif- cell migration and tumor growth independently of zinc binding. EMBO J.
18:4414–4423.
eration, and the potential for antiangiogenic treatment 13. North, P.E., Waner, M., Mizeracki, A., and Mihm, M.C., Jr. 2000. GLUT1: a
of hemangiomas. newly discovered immunohistochemical marker for juvenile hemangiomas.
Hum. Pathol. 31:11–22.
14. Edwards, A., Hammond, H.A., Jin, L., Caskey, C.T., and Chakraborty, R.
Acknowledgments 1992. Genetic variation at five trimeric and tetrameric tandem repeat loci
We thank Brad Quade for providing us with control in four human population groups. Genomics. 12:241–253.
15. Gale, R.E., Wheadon, H., Boulos, P., and Linch, D.C. 1994. Tissue specifici-
samples and helpful advice for the HUMARA assay. ty of X-chromosome inactivation patterns. Blood. 83:2899–2905.
This work was supported by a grant from the NIH (AR 16. Tsuboi, R., Sato, Y., and Rifkin, D.B. 1990. Correlation of cell migration, cell
36820 to B.R. Olsen). E. Boye was supported by the John invasion, receptor number, proteinase production, and basic fibroblast
growth factor levels in endothelial cells. J. Cell Biol. 110:511–517.
Butler Mulliken Foundation. Y. Yu, G. Paranya, and J. 17. Linder, D., and Gartler, S.M. 1965. Distribution of glucose-6-phosphate
Bischoff were supported by grants from the Gackstat- dehydrogenase electrophoretic variants in different tissues of heterozy-
ter Foundation and the Charlotte Geyer Foundation. gotes. Am. J. Hum. Genet. 17:212–220.18. Tsai, Y.C., et al. 1995. Mosaicism in human epithelium: macroscopic mon-
oclonal patches cover the urothelium. J. Urol. 153:1697–1700.
19. Chung, I.M., Schwartz, S.M., and Murry, C.E. 1998. Clonal architecture of
1. Mulliken, J.B. 1988. Diagnosis and natural history of hemangiomas. In normal and atherosclerotic aorta: implications for atherogenesis and vas-
Vascular birthmarks: hemangiomas and malformations. J.B. Mulliken and A.E. cular development. Am. J. Pathol. 152:913–923.
Young, editors. W.B. Saunders Co. Philadelphia, Pennsylvania, USA. 20. Talks, K.L., and Harris, A.L. 2000. Current status of antiangiogenic factors.
41–62. Br. J. Haematol. 109:477–489.
2. Razon, M.J., Kräling, B.M., Mulliken, J.B., and Bischoff, J. 1998. Increased 21. Walter, J.W., et al. 1999. Genetic mapping of a novel familial form of infan-
apoptosis coincides with onset of involution in infantile hemangioma. tile hemangioma. Am. J. Med. Genet. 82:77–83.
Microcirculation. 5:189–195. 22. Bevilacqua, M.P., and Nelson, R.M. 1993. Selectins. J. Clin. Invest. 91:379–387.
752 The Journal of Clinical Investigation | March 2001 | Volume 107 | Number 6