ARTICLES

Soluble endoglin contributes to the pathogenesis

© 2006 Nature Publishing Group http://www.nature.com/naturemedicine

of preeclampsia
Shivalingappa Venkatesha1,6, Mourad Toporsian2,6, Chun Lam1,6, Jun-ichi Hanai1, Tadanori Mammoto1,
Yeon M Kim3, Yuval Bdolah1, Kee-Hak Lim1, Hai-Tao Yuan1, Towia A Libermann1, Isaac E Stillman1,
Drucilla Roberts4, Patricia A D’Amore5, Franklin H Epstein1, Frank W Sellke1, Roberto Romero3,
Vikas P Sukhatme1, Michelle Letarte2 & S Ananth Karumanchi1

Preeclampsia is a pregnancy-specific hypertensive syndrome that causes substantial maternal and fetal morbidity and mortality.
Maternal endothelial dysfunction mediated by excess placenta-derived soluble VEGF receptor 1 (sVEGFR1 or sFlt1) is emerging
as a prominent component in disease pathogenesis. We report a novel placenta-derived soluble TGF-b coreceptor, endoglin (sEng),
which is elevated in the sera of preeclamptic individuals, correlates with disease severity and falls after delivery. sEng inhibits
formation of capillary tubes in vitro and induces vascular permeability and hypertension in vivo. Its effects in pregnant rats are
amplified by coadministration of sFlt1, leading to severe preeclampsia including the HELLP (hemolysis, elevated liver enzymes,
low platelets) syndrome and restriction of fetal growth. sEng impairs binding of TGF-b1 to its receptors and downstream signaling
including effects on activation of eNOS and vasodilation, suggesting that sEng leads to dysregulated TGF-b signaling in the
vasculature. Our results suggest that sEng may act in concert with sFlt1 to induce severe preeclampsia.

Preeclampsia complicates 5% of all pregnancies worldwide and is a
major cause of maternal, fetal and neonatal mortality, especially in
developing nations. This condition is characterized by the onset of
hypertension and proteinuria in the third trimester of pregnancy1.
Severe preeclampsia can lead to the appearance of the HELLP
syndrome, seizures (eclampsia) and/or restriction of fetal growth1,2.
The placenta has a central role in preeclampsia, as evidenced by rapid
disappearance of disease symptoms after delivery. The clinical man-
ifestations of preeclampsia reflect widespread endothelial dysfunction,
resulting in vasoconstriction, end-organ ischemia and increased
vascular permeability1. Therefore, it has been suggested that
placenta-derived circulating factor(s) may induce the endothelial
defects leading to preeclampsia3.
We and others recently showed that high levels of circulating
sFlt1 (soluble fms-like tyrosine kinase, also known as soluble VEGF
receptor 1) of placental origin may contribute to the pathogenesis of
preeclampsia4–7. sFlt1 binds to and neutralizes the proangiogenic
actions of VEGF and placental growth factor (PlGF). High concentra-
tions of circulating sFlt1 along with decreased free VEGF and free PlGF
are seen not only during preeclampsia but before the onset of clinical
symptoms4,8–11. Overexpression of sFlt1 in rats leads to hypertension,
proteinuria and glomerular endotheliosis, the classical manifestations
of preeclampsia, suggesting that excess circulating sFlt1 may have a
causal role in preeclampsia4. sFlt1-treated animals, however, do not
develop hemolysis and thrombocytopenia, signs observed in the
HELLP syndrome, a subtype of severe preeclampsia2,12. Therefore,
we hypothesized that other placenta-derived soluble factors may act
in concert with sFlt1 to cause endothelial dysfunction, resulting in
severe preeclampsia.
Endoglin (Eng) or CD105, a cell-surface coreceptor for trans-
forming growth factor (TGF)-b1 and TGF-b3 isoforms, is highly
expressed in endothelial cells and syncytiotrophoblasts and
modulates actions of TGF-b1 and TGF-b3 (refs. 13–15). Mutations
in the gene encoding Eng, ENG, are the underlying cause of
hereditary hemorrhagic telangiectasia type 1 (HHT1), an autosomal
dominant disorder characterized by arteriovenous malformations
and focal loss of capillaries16. Eng–/– mice die at midgestation
because of defective angiogenesis and cardiovascular development,
whereas Eng+/– mice develop signs of HHT17,18. Recently, it was
found that Eng localizes to caveolae, where it can associate
with endothelial nitric oxide synthase (eNOS) and regulate its activity
and local vascular tone19. These data suggest the involvement
of Eng not only in cardiovascular development but also in
vascular homeostasis.

1Center for Vascular Biology, Departments of Medicine, Obstetrics and Gynecology, Surgery, and Pathology, Beth Israel Deaconess Medical Center and Harvard Medical
School, 330 Brookline Avenue, Boston, Massachusetts 02215, USA. 2Cancer Research Program, Hospital for Sick Children, and The Heart and Stroke Foundation
Richard Lewar Centre of Excellence, University of Toronto, 555 University Avenue, Toronto, Ontario M5G 1X8, Canada. 3Perinatology Research Branch, National
Institute of Child Health and Human Development, National Institutes of Health, Department of Health and Human Services, Bethesda, Maryland, and Wayne State
University School of Medicine, 3980 John R, Detroit, Michigan 48201, USA. 4Department of Pathology, Massachusetts General Hospital and Harvard Medical School,
55 Fruit Street, Boston, Massachusetts 02114, USA. 5Schepens Eye Research Institute and Harvard Medical School, 20 Staniford Steet, Boston, Massachusetts
02114, USA. 6These authors contributed equally to this work. Correspondence should be addressed to S.A.K. (sananth@bidmc.harvard.edu).
Received 9 March; accepted 5 May; published online 4 June 2006; corrected after print 16 June 2006; doi:10.1038/nm1429

642 VOLUME 12 [ NUMBER 6 [ JUNE 2006 NATURE MEDICINE

 ARTICLES

Figure 1 Expression of ENG mRNA and Eng

a b

ia

ia

ia
si
in placentae of normal and preeclamptic

ps

ps

ps
p
am

am

am

am
Normal Preeclampsia

al

Pr l
a
pregnancies. (a) Northern blot analysis for ENG

cl

cl

cl

cl
m

m
ee

ee

ee

ee
or

or
Pr

Pr

Pr
N

N
mRNA from normal and preeeclamptic placentae
5 kb
with their associated gestational ages (GA).
ENG
2 kb 18S ribosomal RNA levels were used as loading
controls. (b) Double immunofluorescence staining
18S of Eng (red) and smooth muscle actin (green) is
shown for preeclamptic placentae and corres-
GA (weeks) 31.3 33.0 30.1 39.2 38.1 38.5
25.6 weeks GA 25.2 weeks GA ponding control placentae derived from individuals
c Normal Preeclampsia with preterm labor. Original magnification,
© 2006 Nature Publishing Group http://www.nature.com/naturemedicine

200. (c) Representative western blot of Eng

(kDa)
90
immunoprecipitates shows full-length Eng
Eng
65
(90 kDa) and a smaller fragment (65 kDa), which
sEng
is more noticeable in preeclamptic placentae.
The normal and preeclamptic samples represent
44 β-actin
individual women. Equal loading for the lysates
GA (weeks) 39.4 30.1 34.1 38.1 28.0 33.0 40.5 weeks GA 41.6 weeks GA used for immunoprecipitation was confirmed by
actin western blots on the same lysates.

eNOS is a Ca2+/calmodulin-regulated nitric oxide (NO) synthase
that can be activated by fluid shear stress and neurohumoral stimuli.
Endothelium-derived NO is a potent vasorelaxant that contributes to
the regulation of systemic blood pressure, vascular permeability and
angiogenesis. Indeed, the effects of VEGF on angiogenesis and vascular
tone are partly mediated by activation of eNOS, through Akt-
dependent phosphorylation of eNOS Ser1177 (ref. 20) and increased
association of eNOS with Hsp90 (ref. 21). Our recent demonstration
that placenta-derived sFlt1 in sera of preeclamptic individuals can
inhibit angiogenesis and induce hypertension4 may in fact reflect
impairment of VEGF-dependent activation of eNOS. The regulation
of eNOS activity is complex, involving dynamic regulation of its
phosphorylation status. While phosphorylation at Ser1177 is indica-
tive of agonist-induced eNOS activation, it is preceded by depho-
sphorylation at Thr495. Coordination of these reactions determines
the activity of eNOS in endothelial cells22.
We now report a novel placenta-derived 65 kDa soluble form of Eng
(sEng) present in sera of pregnant women, elevated in preeclamptic
individuals and increased with disease severity. sEng cooperated with
sFlt1 to induce endothelial dysfunction in vitro and severe preeclamp-
sia-like illness in vivo. We show that sEng interferes with
TGF-b1 signaling and eNOS activation in endothelial cells, thereby
disrupting key homeostatic mechanisms necessary for maintenance
of vascular health.
RESULTS
Upregulation of Eng in preeclamptic placenta
To identify novel placenta-derived factors involved in preeclampsia, we
performed gene expression profiling of placental tissue from pregnant
women with or without preeclampsia using Affymetrix U95A micro-
array chips. In addition to marked upregulation of mRNA encoding
sFlt1 and Flt1 in preeclamptic compared to normal gestational age-
matched placentae4, we noted a fourfold increase in ENG mRNA
which was confirmed by northern blot analysis (Fig. 1a). Unlike
sFlt1, however, no shorter, alternately spliced ENG mRNA was
detected. Immunostaining of placental sections confirmed enhanced
expression of Eng, particularly on the syncytiotrophoblasts of pre-
eclamptic placentae at 25 and 40 weeks relative to age-matched control
pregnancies (Fig. 1b). Western blot analysis of Eng immunoprecipi-
tates of normal and preeclamptic placentae showed the expected
90 kDa Eng monomer, characteristic of the integral plasma membrane
protein in both groups, albeit at higher levels in preeclamptic samples
(Fig. 1c). We also observed a smaller 65 kDa band, suggesting that a
soluble form of Eng (sEng) may be produced by the placenta.
Expression of sEng in preeclamptic placentae was fourfold higher
than in normal pregnancy (n ¼ 10 per group, P o 0.01). Although
both Eng and sEng were upregulated in preeclampsia, quantification
of the sEng /Eng ratio in these placental specimens showed no
significant difference between normal and preeclamptic samples,
suggesting that both Eng and sEng are proportionally increased
in preeclampsia.
Elevated serum sEng correlates with disease severity
After immunoblotting Eng immunoprecipitates for Eng from
sera of preeclamptic individuals, we observed a single 65 kDa band
(Fig. 2a). This protein was present at much lower levels in the sera of
normal pregnant women and barely detectable in nonpregnant
women. Purification from sera of preeclamptic individuals and
analysis by mass spectrometry showed several Eng-specific peptides
ranging from Gly27 to Arg393 (Fig. 2b), indicating a soluble
form (sEng) corresponding to the N-terminal region of the full-
length protein.
Our finding of elevated sEng concentrations in sera of preeclamptic
individuals suggested that this may serve as diagnostic or prognostic
marker for the disease. Quantification by ELISA of serum sEng
concentrations showed a two- and threefold increase in preterm and
term pregnancy, respectively, compared to nonpregnant states
(Fig. 2c), suggesting a role for sEng during normal pregnancy.
Concentrations of sEng were three-, five- and tenfold higher in
individuals with mild preeclampsia, severe preeclampsia and HELLP
syndrome, respectively, compared to gestational age–matched preterm
controls (Fig. 2c). The clinical characteristics of these individuals are
shown in Supplementary Table 1 online. Concentrations of sEng in
pregnant individuals correlated with those of sFlt1 (R2 ¼ 0.56) except
in the HELLP syndrome group, in which the level of sEng was higher
than that of sFlt1. Blood samples from a subset of women obtained
48 h after placental delivery showed a 70% reduction in mean sEng
circulating levels in preeclamptic compared to normal pregnant
women (Fig. 2d). Although these data suggest that the placenta is a
major source of sEng during pregnancy, other sources such as
maternal vasculature cannot be ruled out.
Effects of sEng on endothelial function
TGF-b and VEGF are crucial factors in angiogenesis. Recombinant
sEng inhibited endothelial tube formation on Matrigel in vitro
to the same extent as sFlt1. Moreover, sEng potentiated the

NATURE MEDICINE VOLUME 12 [ NUMBER 6 [ JUNE 2006 643

 ARTICLES

a Recomb.
sEng H
Non-
Normal Preeclampsia
b Recomb.
sEng Fraction # c *#
pregnant 100 sEng
(kDa) (kDa) 3 4 5 6

Concentration in serum (ng/ml)

116 116 sFlt1

82 82 80
sEng 64 sEng *
64 *
49 49 60 *
37
20 †† *
*
sEng level (units of optical density)

** 40

16
1 GASCSLSPTS LAETVHCDLQ PVGPERGEVT YTTSQVSKGC VAQAPNAILE 20
51 VHVLFLEFPT GPSQLELTLQ ASKQNGTWPR EVLLVLSVNS SVFLHLQALG
101
1 IPLHLAYNSS LVTFQEPPGV NTTELPSFPK TQILEWAAER GPITSAAELN
© 2006 Nature Publishing Group http://www.nature.com/naturemedicine

12
151 DPQSILLRLG QAQGSLSFCM LEASQDMGRT LEWRPRTPAL VRGCHLEGVA
201 GHKEAHILRV LPGHSAGPRT VTVKVELSCA PGDLDAVLIL QGPPYVSWLI 0
251 DANHNMQIWT TGEYSFKIFP EKNIRGFKLP DTPQGLLGEA RMLNASIVAS Non- Normal Normal Mild Severe HELLP
8 ## 301 FVELPLASIV SLHASSCGGR LQTSPAPIQT TPPKDTCSPE LLMSLIQTKC pregnant term preterm PE PE
351 ADDAMTLVLK KELVAHLKCT ITGLTFWDPS CEAEDRGDKF VLRSAYSSCG
401 MQVSASMISN EAVVNILSSS SPQRKKVHCL NMDSLSFQLG LYLSPHFLQA
4 451
501
SNTIEPGQQS
GNCVSLLSPS
FVQVRVSPSV
PEGDPRFSFL
SEFLLQLDSC
LHFYTVPIPK
HLDLGPEGGT
TGTLSCTVAL
VELIQGRAAK
RPKTGSQDQE
d
551 VHRTVFMRLN IISPDLSGCT SKGLVLPAVL GITFGAFLIG ALLTAALWYI
T=0

sEng concentration in serum (ng/ml)

601 YSHTRSPSKR EPVVAVAAPA SSESSSTNHS IGSTQSTPCS TSSMA 100
0 T = 48
Non- Normal Preeclampsia
pregnant 80
Figure 2 Increased sEng levels in sera from individuals with preeclampsia. (a) Representative western
blot and graph of Eng immunoprecipitates showing higher level of a soluble 65kDa fragment (sEng) in 60
sera of preeclamptic women (at 28, 33 and 32.2 weeks) compared to normal pregnant women (at 34.1,
28.2 and 39.4 weeks; **P o 0.01) and nonpregnant women (wwP o 0.01; n ¼ 3/group). sEng levels in 40
*
normal pregnant women were higher than in nonpregnant individuals (##P o 0.01). HUVEC (H) extracts
and recombinant human sEng served as positive controls (90 kDa and 75 kDa, respectively). Individuals 20
matched for gestational ages were randomly chosen from normal (term and pre-term) and preeclampsia *
(severe and HELLP) groups described in Supplementary Table 1. (b) sEng was purified from the serum of 0
Normal Preeclampsia
preeclamptic individuals. Fractions 4 and 5 eluted from the 44G4-IgG (Eng-specific) Sepharose, were
run on SDS-PAGE under reducing conditions and tested by western blot using a polyclonal antibody to Eng. The eluted fractions were subjected to mass
spectrometry analysis (three runs), and the peptides identified are in boldface and underlined in the sequence of human Eng. The underlined amino acids
562–586 represent the transmembrane domain of human cell-surface Eng. Sera from ten random individuals with preeclampsia (severe and HELLP) in
Supplementary Table 1 were pooled. (c) ELISA results for sEng and sFlt1 in sera of individuals with varying degrees of preeclampsia, control pregnancies
and four nonpregnant healthy volunteers. *P o 0.05 compared to preterm controls, #P o 0.05 compared to severe preeclampsia. All individuals described
in Supplementary Table 1 were analyzed. (d) ELISA results for sEng in a subset of pregnant individuals (normal, n ¼ 6; preeclampsia, n ¼ 11) described in
c with blood drawn before (0–12 h) or after (48 h) delivery. *P o 0.05 as compared to T ¼ 0 samples. All individuals described in Supplementary Table 1
from whom we were able to obtain a serum specimen 48 h after delivery were analyzed.

antiangiogenic actions of sFlt1, as seen by a larger reduction in the than that observed with sFlt1 (Table 1). Proteinuria was modest in
number of capillary-like structures in the presence of both proteins sEng-treated rats, but severe in the sFlt1-treated group. The
(Fig. 3a). This suggests that sEng and sFlt1 block the proangio- sFlt1+sEng group showed nephrotic-range proteinuria, severe hyper-
genic effects of TGF-b1 and VEGF, respectively, present in Matrigel. tension and biochemical evidence of HELLP syndrome (elevated
Capillary permeability is increased in pre-
eclampsia23, and preeclamptic individuals are
more prone to develop peripheral and pul- a b
monary edema. Therefore, we tested the role
Control
of sEng and sFlt1 in microvascular perme- sEng
#
ability using the Evans blue assay in BALB/c * sFlt1
mice pretreated with adenovirus expressing 0.8 sEng+sFlt1

sFlt1, sEng, sFlt1+sEng or mouse Fc protein

**
(optical units of density)

*
Absorbance 620 nm

Control: 139,249 sEng: 75,515 0.6

as a negative control. Capillary permeability * *
was increased by sEng and sFlt1 in the lungs, *
0.4
liver and kidneys (Fig. 3b). Notably, the
combination of sEng and sFlt1 showed an
0.2
additive effect in the liver, indicating that
these soluble receptors may act in concert to
0.0
disrupt endothelial integrity and induce con- Lung Liver Kidney
sFlt1: 74,570 sEng+sFlt1: 38,989
siderable vascular damage and leak.
Figure 3 sEng inhibits capillary formation and increases vascular permeability. (a) Angiogenesis assays
In vivo effects of sEng and sFlt1 were done using human umbilical vein endothelial cells (HUVECs) in growth factor–reduced Matrigel
To induce the clinical features of preeclamp- and performed in the presence of 1 mg of recombinant sEng, sFlt1 or both, and endothelial tube
sia in pregnant rats, we used adenoviral lengths quantified. A representative experiment (n ¼ 4) is shown, with average tube lengths (in pixels)
expression of sEng and sFlt1, alone or in indicated below the panels. Tube length results were quantified, showing a 62.2% ± 3.9, 62% ± 4.1
and 39.7% ± 4.2 reduction in sEng, sFlt1 and sFlt1+sEng, respectively (P o 0.01 versus controls).
combination. At 17–18 d of pregnancy, (b) Microvascular permeability assessed by Evans blue leakage in mice injected with adenovirus
hemodynamic and biochemical data indi- containing Fc (Control), sEng, sFlt1 or sFlt1+sEng. Leakage of Evans blue was quantified in the various
cated that sEng induced a significant change organs. Data represent a mean of four independent experiments. *P o 0.05 compared to controls,
in mean arterial pressure (MAP), albeit lower #P o 0.05 compared to sFlt1 alone.

644 VOLUME 12 [ NUMBER 6 [ JUNE 2006 NATURE MEDICINE

 ARTICLES

Table 1 Hemodynamic and biochemical data in pregnant rats

Groups n MAPa (mmHg) Urine albumin/creatinineb (mg/mg) Platelet count (
1,000/ml) LDH (U/L) AST (U/L) Fetal weightb (g)

Control (CMV) 6 83 ± 5 186 ± 94 1,098 ± 75 156 ± 32 54 ± 4 2.1 ± 0.5

sEng 6 104 ± 6* 432 ± 249 1,195 ±78 188 ± 46 110 ± 13* 1.6 ± 0.4
sFlt1 6 117 ± 7* 2,295 ± 867* 1,131 ± 91 172 ± 53 94 ± 4* 1.75 ± 0.4
sFlt1+sEng 6 121 ± 9* 9,029 ± 4,043* 615 ± 67* 1,952 ± 784* 210 ± 92* 0.75 ± 0.3*

Data are presented as mean ± s.e.m. *P o 0.05 compared to control group. aMAP ¼ diastolic pressure + 1/3 pulse pressure.bFetal weight is the average weight of the litter for each group.
Expression of sFlt1 and sEng were first confirmed in rat plasma by western blots (Supplementary Fig. 3) and circulating concentrations quantified using commercially available ELISA kits. Mean
plasma concentrations of sFlt1 in the control, sEng, sFlt1 and sFlt1+sEng groups were 0.64 ng/ml, 0.66 ng/ml, 249 ng/ml and 204 ng/ml, respectively. Similarly, the concentrations of sEng in
© 2006 Nature Publishing Group http://www.nature.com/naturemedicine

these four groups were 0.39 ng/ml, 129 ng/ml, 0.37 ng/ml and 123 ng/ml, respectively.

lactate dehydrogenase (LDH) and aspartate aminotransferase (AST),
and decreased platelet counts). We observed restriction of fetal growth
in litters born to the sFlt1+sEng group, probably related to the
placental vascular ischemia and damage (Table 1).
Renal histology by light and electron microscopy showed
focal endotheliosis in the sEng group compared to the control
group (Fig. 4a and Supplementary Fig. 1 online). Severe glomerular
endotheliosis was observed in the sFlt1 and the sFlt1+sEng groups
(Fig. 4a and Supplementary Fig. 1). We observed extensive vascular
damage of the placenta, including infarction at the maternal-fetal
junction in the sFlt1+sEng group, but not in control rats or in those
treated with either agent (Fig. 4b). We noted diffuse inflammation in
the giant-cell layer (corresponding to human invasive trophoblasts)
in the sFlt1 and sEng groups, and increased inflammation in the
sFlt1+sEng group compared to either the sFlt1 or sEng groups. Liver
histology showed signs of ischemia and areas of necrosis in the
sFlt1+sEng group, similar to those seen in individuals with the
HELLP syndrome (Fig. 4c). Evidence of hemolysis in the sFlt1+sEng
group was confirmed in peripheral blood smears, in which we
observed schistocytes and reticulocytosis (Fig. 4d). We also saw
signs of severe maternal vascular damage after injection of sFlt1 and
sEng together into nonpregnant rats, suggesting that the phenotype in
pregnant rats resulted from a direct effect on the maternal vessels and
does not require the placenta (data not shown).
sEng inhibits TGF-b–mediated NOS-dependent vasodilation
Given the known effect of VEGF on reducing vascular reactivity
through activation of eNOS and the recent demonstration that Eng
modulates eNOS-dependent vasomotor activity19, we assessed the
hemodynamic effects of TGF-b isoforms and sEng on isolated rat
renal microvessels. Both TGF-b1 and TGF-b3 induced a dose-depen-
dent increase in arterial diameter (Fig. 5a), whereas TGF-b2, which is
not a ligand for Eng, did not produce significant vasodilation (o2%
at 0.1 and 1 mg/ml). sEng significantly attenuated the effects of TGF-b1
and TGF-b3 (Fig. 5a). This acute effect of TGF-b1 and TGF-b3
isoforms on vascular tone has not been previously recognized to our
knowledge and was also seen in mesenteric vessels (Supplementary
Fig. 2 online). VEGF and TGF-b1 had additive effects on vasodilation,
which were blocked by the combination of sEng and sFlt1 at concen-
trations noted in individuals with preeclampsia (Fig. 5b). L-NAME
blocked the vasodilation mediated by TGF-b1 and VEGF, indicating a
NOS-dependent response (Fig. 5b). These data suggest that circulating

Figure 4 Renal, placental and hepatic

histological changes and peripheral blood
a
Control sEng sFlt1 sFlt1+sEng

smears in pregnant rats after sEng and sFlt1

treatment. (a) Renal histology (plastic
sections) of control, sEng, sFlt1 and sFlt1+sEng
groups. sEng-treated rats showed mild focal
endotheliosis (arrowhead, b) not seen in the
control glomeruli. sFlt1-injected rats showed
moderate to severe endotheliosis with complete b
occlusion of capillary lumens. sFlt1+sEng-treated
rats showed extremely swollen glomeruli and
marked endotheliosis with protein resorption
droplets in the podocytes. Scale bar, 50 mm.
(b) Placental histology (H&E stain) of control,
sEng, sFlt1 and sFlt1+sEng groups. Both the c
sEng- and sFlt1-treated rats showed diffuse
inflammation (arrowheads) at the maternal-
fetal junction that was not seen in controls.
There was hemorrhagic infarction and
fibrinoid necrosis with lumen obstruction
of a maternal vessel (arrow) in the decidua
of the sFlt1+sEng-treated placenta. Scale bar,
d
200 mm. (c) Liver histology in the control,
sEng, sFlt1 and sFlt1+sEng groups. Ischemic
changes with multifocal necrosis (arrowhead)
were noted in the sFlt1+sEng group. Control
group and rats given sEng or sFlt1 showed no
changes. Scale bar, 200 mm. (d) Peripheral
blood smear (Wright stain) of control, sEng, sFlt1 and sFlt1+sEng groups. A representative smear in the sFlt1+sEng group showed active schistocytes
(arrowheads) and reticulocytosis (arrows). No hemolysis was seen in the other groups.

NATURE MEDICINE VOLUME 12 [ NUMBER 6 [ JUNE 2006 645

 ARTICLES

Vehicle C sEng
a TGF-β1
b c

Percent change in vascular diameter

125
Percent change in vascular diameter

TGF-β3 16 [I ] TGF-β1 6 12 25 50 100 100 6 12 25 50 100

25
TGF-β1 + sEng (pM)
TGF-β3 + sEng 12 80 kDa TβRII
20
8
15
4 Vehicle *

Percent maximal TβRII binding

* 100 sEng
10
** * 0
80 *
** –4
5 60
* –8 40
0
© 2006 Nature Publishing Group http://www.nature.com/naturemedicine

0.2 2 20 200 VEGF + – + + + –

Concentration (ng/ml) 20
TGF-β1 – + + + – +
sFlt1 – – – + – – 0
sEng – – – + – –
0 25 50 75 100
L-NAME – – – – + + 125
[I ] TGF-β1 (pM)
d 25

* e
Relative light intensity units

20 Vehicle
100 sEng

Percent phospho Thr495 /

Vehicle sEng
15 TGF-β1 (pM) 0 125 250 125 250 90

total eNOS level

(kDa) eNOS 80
140 pThr495
10 70
eNOS * *
140 pSer1177 60
5
50
eNOS
140 Total 40
0
0 0 15 15 TGF-β1 (ng/ml) 0
0 125 250
0 100 0 100 sEng (ng/ml) TGF-β1 (pM)

Figure 5 Recombinant sEng attenuates TGF-b1 binding and has effects on vasodilation through activation of eNOS. (a) Microvascular reactivity of rat renal
microvessels was measured in the presence of TGF-b1 or TGF-b3 from 200 pg/ml to 200 ng/ml, and with or without recombinant sEng at 1 mg/ml (n ¼ 4
per experiment; *P o 0.05 as compared to TGF-b1 or TGF-b3 alone). (b) Microvascular responses of renal microvessels to VEGF, TGF-b1 (1 ng/ml each) or
the combination of the two. The effects of 100 ng/ml each of sFlt1 and sEng on the combined response are shown (n ¼ 4 experiments). Also shown is the
blocking effect of L-NAME on TGF-b1– and VEGF-stimulated responses. (c) Representative autoradiogram and graph showing a dose-dependent increase in
[I125]TGF-b1 (6–100 pM) binding to TbRII on mouse endothelial cells. Treatment with 5 nM recombinant sEng significantly reduced binding at 50 pM and
100 pM (*P o 0.05 compared to untreated group). Competition with 40
excess cold TGF-b1 in cells treated with 100 pM [I125]TGF-b1 abolished receptor
binding and served as background control (C). (d) Graph showing significantly increased TGF-b-induced activation of the Smad 2/3-dependent CAGA-Luc
reporter construct transfected in human umbilical vein endothelial cells (HUVECs) and inhibition by treatment with sEng (n ¼ 3 experiments, **P o 0.01
compared to sEng untreated group). (e) Representative western blots and graph (n ¼ 4) showing significant dephosphorylation at eNOS Thr495 after
treatment with TGF-b1 and attenuation by sEng (*P o 0.05 compared to untreated). Phosphorylation was unchanged at Ser1177 and total levels of eNOS
remained constant throughout the experiments.

sFlt1 and sEng may oppose the physiological NO-dependent vasodi- mesenteric resistance vessels, we explored its immediate effects on
latation elicited by these angiogenic growth factors, contributing to the eNOS activation. Although TGF-b1 had no effect on phosphorylation
development of hypertension seen in preeclampsia. of eNOS Ser1177, it induced a significant (P o 0.01 versus baseline
control) dephosphorylation at Thr495 (Fig. 5e), suggesting that
sEng inhibits TGF-b1 binding and signaling in endothelial cells TGF-b regulates the phosphorylation status of a key residue involved
Given that endoglin is a coreceptor for TGF-b1 and TGF-b3 isoforms, in eNOS activation. This effect was significantly attenuated by
we hypothesized that sEng acts by interfering with binding of sEng (Fig. 5e).
cell-surface receptors. Preincubating radiolabeled TGF-b1 with recom-
binant sEng significantly reduced its binding to TGF-b receptor DISCUSSION
type II (TbRII) at both 50 and 100 pM (Fig. 5c). Thus, sEng Maternal endothelial dysfunction caused by placenta-derived soluble
competes for TGF-b1 binding to its receptors on endothelial cells. factors such as sFlt1 is emerging as a major component in the
To test whether this leads to impaired signaling, we assessed the pathogenesis of preeclampsia26,27. We report a novel soluble form of
activity of a CAGA-Luc reporter construct in human endothelial Eng of placental origin that is present in the sera of pregnant women,
cells. TGF-b1 induced the activation of the Smad 2/3-dependent elevated in preeclamptic individuals and correlated with disease
CAGA-Luc reporter, and this response was abolished by treatment severity. In addition to its potential use as a biomarker of preeclamp-
with sEng (Fig. 5d). sia, we have shown that sEng disrupts formation of endothelial tubes
in vitro and induces vascular permeability and hypertension in vivo.
sEng blocks TGF-b1–mediated activation of eNOS Notably, sEng can act in concert with sFlt1 to amplify endothelial
Although previous studies have shown that chronic exposure of dysfunction and induce clinical signs of severe preeclampsia, including
endothelial cells to TGF-b induces expression of the eNOS gene and development of the HELLP syndrome and restriction of fetal growth.
protein24,25, the immediate effects of TGF-b1 on the activation We show that TGF-b1 induces vasorelaxation through activation of
of eNOS have not yet been characterized. Given our findings that eNOS by triggering dephosphorylation of Thr495, providing a novel
TGF-b1 induces NOS-dependent vasorelaxation in both renal and mechanism for endothelium-dependent vasoregulation. sEng

646 VOLUME 12 [ NUMBER 6 [ JUNE 2006 NATURE MEDICINE


interferes with TGF-b receptor binding and downstream signaling in
endothelial cells, and attenuates eNOS activation. We propose that the
contributions of sEng and sFlt1 to the pathogenesis of maternal
preeclampsia are, at least in part, related to their inhibition of
VEGF and TGF-b stimulation of endothelial-dependent NO activation
and vasomotor effects.
Increased Eng immunoreactivity has been reported in plasma of
individuals with angiogenic tumors28, but its nature has remained
elusive. We have now characterized a circulating form of Eng observed
in normal pregnancy and upregulated in preeclampsia. The absence of
© 2006 Nature Publishing Group http://www.nature.com/naturemedicine
alternate splice variants in placenta and the partial peptide sequence of
purified sEng suggest that it is an N-terminal cleavage product of full-
length Eng. Given that betaglycan, another TGF-b coreceptor with
partial sequence identity to Eng, can be shed by membrane-type
metalloprotease-1 (MT1-MMP)29, we speculate that sEng may be
generated by a similar mechanism.
We show that sEng prevents binding of TGF-b1 to TbRII on
endothelial cells, leading to decreased signaling. As circulating TGF-
b1 is complexed with latency-associated peptide and latent TGF-b1
binding protein, it cannot bind its receptors unless activated. It is
therefore likely that sEng only inhibits the effects of TGF-b1 locally,
where active TGF-b1 is generated. Hence sEng would not impact
circulating concentrations of TGF-b1. When circulating active TGF-b1
was measured in our cohort of pregnant women, it was undetectable
during pregnancy, confirming that the majority of circulating TGF-b1
is inactive. Although several clinical studies reported modest changes
in total circulating TGF-b1 in preeclampsia30–32, our data do not show
significant differences between normal and preeclamptic pregnancies
(35.95 ng/ml versus 37.74 ng/ml in serum, respectively). Eng is also
capable of binding other ligands of the TGF-b family, such as activins
and BMPs33, although no functional role of Eng in mediating their
effects has been described.
In this study, we identified a direct effect of TGF-b1 and TGF-b3 on
vascular reactivity, consistent with the known specificity of cell-surface
Eng for these isoforms, thereby suggesting its essential role in TGF-b–
dependent vasodilatation. This effect is NOS dependent and supports
our findings that Eng interacts with eNOS and stabilizes its activa-
tion19. Together, these data suggest a crucial role for Eng in linking
TGF-b receptor activation to synthesis of NO. We have found that
TGF-b1 dephosphorylates eNOS at Thr495, a regulatory step necessary
to increase the Ca2+ sensitivity and enzyme activity and preventing
uncoupling and production of eNOS-derived superoxide. Although
activation of eNOS also involves a coordinated increase in phosphor-
ylation of Ser1177, this was not observed in response to TGF-b1.
TGF-b may specifically modulate dephosphorylation of Thr495 and
synergize with factors such as VEGF, which can activate eNOS by
phosphorylating Ser1177. We observed that TGF-b and VEGF
have additive effects on NOS-dependent vasodilatation that are com-
pletely reversed by sEng and sFlt1 at concentrations seen in individuals
with preeclampsia.
Our functional studies suggest that sEng and sFlt1 act in concert to
induce vascular damage and HELLP syndrome by interfering with
TGF-b1 and VEGF signaling, respectively. One molecule likely to be
involved in the pathogenesis of hypertension34 and central to both
VEGF and TGF-b1 signaling is NO. Several studies support the
hypothesis that decreased biologically available NO is central to the
pathogenesis of preeclampsia35. The inhibition of eNOS activation by
both sFlt1 and sEng suggests a molecular basis for the elevated MAP.
Furthermore, as both TGF-b1 and VEGF stimulate expression of
eNOS25,36, it is likely that sEng and sFlt1 inhibit expression of
eNOS chronically and that their excess circulating concentrations
NATURE MEDICINE VOLUME 12 [ NUMBER 6 [ JUNE 2006
ARTICLES
may be responsible for hypertension noted in preeclampsia. Their
effects in decreasing activation of eNOS may also account for the
increased vascular permeability observed in preeclampsia37. Another
molecule that may be central to the pathogenesis of the procoagulant
state and thrombocytopenia in sFlt1+sEng-treated rats is the antith-
rombotic factor prostacyclin (PGI2). Both VEGF and TGF-b1 stimu-
late production of PGI238–40, and clinical studies have shown
decreased endothelial production of PGI2 even before the onset of
clinical preeclampsia41. As it is known that TGF-b1 induces produc-
tion of VEGF by pericytes42, it is also possible that sEng interferes
with VEGF functions indirectly through impaired production of
VEGF by pericytes.
The placentae of women with severe preeclampsia are characterized
by shallow invasion of the cytotrophoblasts and impaired spiral
arteriolar remodeling. During normal placentation, cytotrophoblasts
undergo a program of pseudovasculogenesis by acquiring the
endothelial markers VE-cadherin and avb3 integrin, a process
impaired in preeclampsia43,44. This impaired placentation and accom-
panying ischemia are thought to be the primary events leading to the
elaboration of soluble factors into the circulation. Treatment of villous
explant cultures with ENG antisense oligonucleotide enhanced tro-
phoblast outgrowth and migration45, suggesting that cell-surface Eng
negatively regulates this process. We speculate that sEng is produced
by the placenta as a compensatory mechanism to limit the effects of
cell-surface Eng. In preeclampsia, excessive production of cell-surface
endoglin would lead to increased sEng in the maternal circulation,
which in turn may be responsible for the clinical manifestations
of preeclampsia.
Our findings on the role of sEng in preeclampsia have important
diagnostic and therapeutic implications. Analogous to sFlt1 (ref. 9),
circulating sEng starts rising 6–10 weeks before clinical symptoms of
preeclampsia (S.A.K. & R.R., unpublished observations). We posit that
a predictive test measuring sEng, sFlt1 and PlGF in the serum will have
enhanced sensitivity and specificity, and provide a powerful tool in the
prevention of preeclampsia-induced mortality. In addition, neutraliz-
ing antibodies to sEng and sFlt1 may be beneficial in the treatment of
preeclampsia. Therapeutic monoclonal antibodies targeting surface
Eng are currently being developed for the treatment of metastatic
cancers46. Our data suggests that sEng may be another way to
interfere with angiogenically active tumors that express vast amounts
of TGF-b. Similar strategies have been employed in the VEGF
signaling cascade using VEGF-trap, a modified sFlt1 that is currently
in cancer clinical trials47.
In summary, we have shown that excess circulating concentrations
of sEng and sFlt1 in individuals with preeclampsia contribute to the
pathogenesis of hypertension, proteinuria, glomerular endotheliosis,
HELLP syndrome and restriction of fetal growth—all hallmarks of
severe preeclampsia. Understanding the regulation of sEng and sFlt1
production in placenta and their effects on systemic and placental
vascular function should lead to better insights into their roles during
normal pregnancy and the pathogenesis, prediction, treatment and
prevention of preeclampsia.
METHODS
Reagents. We purchased recombinant human sEng (1–587 amino acids,
corresponding to the extracellular region of cell-surface Eng), human sFlt1,
mouse sEng and sFlt1, human TGF-b1 and TGF-b3, mouse VEGF-164 and
human VEGF-165 from R&D Systems. Mouse monoclonal antibody to human
Eng (clone P4A4) and polyclonal antibody (H-300) to the N-terminal region of
human Eng were from Santa Cruz. ELISA kits for human and mouse sFlt1, and
human recombinant sEng were obtained from R&D systems.
647
 ARTICLES
Subjects. All clinical studies were approved by the Beth Israel Deaconess
Medical Center Committee on Clinical Investigations where all subjects were
recruited and provided informed consent. Clinical information is described in
Supplementary Table 1. Preeclampsia was defined by the American College of
Obstetricians and Gynecologists criteria48. Individuals with HELLP syndrome
are presented as a separate severe preeclamptic group. HELLP syndrome was
diagnosed when individuals had thrombocytopenia (o100,000 cells/ml) and
increased LDH (4600 IU/L) and AST (470 IU/L). Healthy pregnant women
were included as controls; eight individuals with pre-term labor served as
gestational age controls. Blood specimens from four nonpregnant healthy
individuals in the age groups (30–45 years) were also included. Placental
© 2006 Nature Publishing Group http://www.nature.com/naturemedicine
samples were obtained immediately after delivery. The placental samples used
in the study were randomly chosen from the preeclampsia (mild, severe,
HELLP) and controls (term and pre-term) groups described in Supplementary
Table 1, and the only criterion used for selection was that the controls were
matched in gestational age to the individuals with preeclampsia. Serum
was collected at the time of delivery (0–12 h before) and at 48 h (±8 h)
after delivery.
Northern blots, ELISA, western blots, immunohistochemistry and immu-
noprecipitation. We performed northern blots, ELISA, western blots, immu-
nohistochemistry and immunoprecipitation as described in Supplementary
Methods online.
Purification of sEng and analysis by mass spectrometry. We sequentially
applied serum (10 ml) from preeclamptic individuals onto CM Affi-gel blue
and protein A Sepharose (Bio-Rad) columns to remove albumin and immu-
noglobulins, respectively. We slowly applied the flow-through to a 2.5 ml
column of monoclonal antibody 44G4 IgG to human Eng, conjugated to
Sepharose. We eluted bound fractions with 0.02 M diethylamine (pH 11.4) and
immediately neutralized them with 1 M Tris pH 7.8. We pooled fractions 4 and
5 with elevated absorbance at 280 nm, reduced them with 10 mM DTT for 1 h
at 57 1C and alkylated them with 0.055 M iodoacetomide. We then completely
digested the samples with trypsin (1:100). We resuspended the lyophilized
sample in 0.1% trifluoroacetic acid and injected it into a CapLC (Waters) high-
performance liquid chromatography instrument. We separated peptides using a
75 mm Nano Series column (LC Packings) and analyzed them using a Qstar XL
MS/MS system. We searched data using the Mascot search engine (Matrix
Science) against the human protein database NCBInr.
Endothelial tube assays and microvascular permeability experiments. We
performed endothelial tube assays and microvascular permeability experiments
as described in Supplementary Methods.
Generation of adenoviruses. Adenovirus vectors containing sFlt1 and Fc have
been previously described4,49. To generate sEng adenovirus, we used the Adeasy
Kit (Stratagene). Briefly, we amplified human sEng (Thr27–Leu586) corre-
sponding to the complete extracellular region with PCR using human cDNA
encoding full-length Eng clone (Invitrogen) as template and the following
primers: forward, 5¢-ACGAAGCTTGAAACAGTCCATTGTGACCTT-3¢ and
reverse, 5-’TTAGATATCTGGCCT TTGCTTGTGCAACC-3¢. We subcloned
amplified PCR fragments into pShuttle-CMV vector (Stratagene) and con-
firmed expression by western blotting. We amplified the confirmed sEng clone
in 293T cells and purified it on a CsCl2 density gradient (Qbiogene). Ad-CMV
(control) was obtained from Qbiogene.
Rat model of preeclampsia. All animal protocols were approved by the Beth
Israel Deaconess Medical Center Institutional Animal Care and Use Commit-
tee. We intravenously injected pregnant Sprague-Dawley rats with 2
109
plaque-forming units of Ad-CMV (control), Ad-sFlt1, Ad-sEng or Ad-sFlt1+Ad
sEng. We injected females at day 8 or 9 of pregnancy and measured intracarotid
MAP at 17–18 d under anesthesia as previously described4. We confirmed levels
of circulating sFlt-1 and sEng by western blotting and quantified them using
commercially available ELISA kits (R&D Systems). We measured urinary
albumin by standard dipstick and quantified it using the Nephrat ELISA kit
(Exocell). We measured urinary creatinine using the Metra creatinine assay kit
(Quidel Corp). We measured AST and LDH using commercial kits from
Thermo Electron. We estimated platelet counts by hemocytometry (Hemavet
648
850). We detected schistocytes by staining peripheral blood smears with Wright
stain. Rats were killed, litters counted and individual placentae and fetuses
weighed. We used harvested kidneys and placentae for histopathology and
electron microscopy as previously described4.
TGF-b1 binding to endothelial cells. We affinity-labeled confluent endothelial
cell monolayers for 4 h at 41C with increasing concentrations of [I125]TGF-b1
(DuPont NEN) after preincubation with or without 2.5 nM recombinant sEng
or cold TGF-b1 (40-fold excess), washed them and cross-linked them with
disuccinimidyl suberate (Pierce). We solubilized cells in buffer containing 1%
Triton X-100 and separated extracts under reducing conditions by SDS-PAGE
on 4–12% gels13. We visualized levels of [I125]TGF-b1 bound to TbRII with a
STORM Phosphorimager and quantified them by densitometry using Image-
Quant software.
TGF-b1–dependent activation of Smad2/3. We transfected human umbilical
vein endothelial cells (HUVECs, Clonetics; passage 3) with (CAGA)12-Luc
plasmid (from K. Miyazano, University of Tokyo) in the absence or presence of
TGF-b1 and/or sEng and measured luciferase activity as previously described50.
Microvascular reactivity experiments. We performed microvascular reactivity
experiments as previously described4 using rat renal or mesenteric microvessels
(internal diameter, 70–150 mm). In all experimental groups, we examined the
relaxation responses of kidney microvessels after precontraction with U46619
(thromboxane A2 agonist) to 40–60% of their baseline diameter at a distending
pressure of 40 mmHg. Once the steady-state tone was reached, we examined
the responses to TGF-b1, TGF-b3 or VEGF in a standardized order. We
administered all drugs extraluminally. When required, we pretreated vessels
with sEng, sFlt1 or 10–5 M L-NAME for 30 min.
TGF-b1–dependent phosphoregulation of eNOS. We incubated confluent
mouse endothelial cells in serum-free media for 2 h. We stimulated cells with
TGF-b1 in the presence and absence of sEng (100 ng/ml) for 15 min and
extracted proteins in 2% SDS buffer supplemented with 1 mM Na8VO4, 10 mM
Na4P2O7, 25 mM NaF and protease inhibitors (Roche Molecular Biochemicals).
We quantified proteins and immunoblotted them with polyclonal antibodies to
eNOS Thr495, Ser1177 or monoclonal antibody specific for total eNOS.
Statistical analysis. Results are presented as mean ± s.e.m. and comparisons
between multiple groups were made by analysis of variance using ANOVA.
Significant differences are reported when P o 0.05.
Note: Supplementary information is available on the Nature Medicine website.
ACKNOWLEDGMENTS
We thank all the staff in the Department of Obstetrics at the Beth Israel
Deaconess Medical Center for help with patient identification and recruitment.
We thank B. Furie’s laboratory for help with measurement of platelet counts in
rats, J. Min and the Physiology Core laboratory for help with blood pressure
measurements, J. Li for vascular reactivity experiments, R. Mulligan for help with
the production of sFlt1 adenoviruses, L. Zhang for technical assistance with mass
spectrometry, B. Sachs, R. Levine, R. Thadhani, J. Flier, W. Aird and S. Pennathur
for discussions. This work was funded by US National Institutes of Health grants
DK064255 and HL079594 to S.A.K., Department of Medicine, Obstetrics and
Gynecology seed funds to S.A.K. and V.P.S., and Heart and Stroke Foundation
of Ontario grant T5016 to M.L.
AUTHOR CONTRIBUTIONS
S.V.: mRNA and protein expression studies of Eng/sEng, generation and
expression of sEng adenoviruses, all animal studies; substantial contribution to
writing, editing and generation of figures. M.T.: biochemical characterization of
sEng, TGF-b effects on eNOS dephosphorylation, studies of sEng on TGF-b
effects; substantial contribution to writing and editing of manuscript and
generation of figures. C.L.: enrollment and collection of all human material,
ELISA studies on human samples, in vitro angiogenesis assays and biochemical
studies of urine and plasma obtained from animals. J.H.: TGF-b–Smad promoter
studies. T.M.: vascular permeability studies. Y.M.K.: immunohistochemistry of
human placental samples. Y.B.: animal studies. K.H.L.: enrollment and collection
of all human material, expertise in clinical preeclampsia. H.Y.: in vitro
angiogenesis assays. T.A.L.: Affymetrix microarray experiments. I.E.S.: all rat
histological studies, including electron microscopy. D.R.: rat placental histological
VOLUME 12 [ NUMBER 6 [ JUNE 2006 NATURE MEDICINE

studies. P.A.D.: provided expertise in vascular biology and TGF-b signaling.
F.H.E.: provided expertise in clinical preeclampsia. F.W.S.: all microvascular
reactivity experiments. R.R.: immunohistochemistry of human placental samples;
provided expertise in preeclampsia. V.P.S.: interpretation of microarray
experiments, microvascular permeability studies; provided overall expertise in
vascular biology and preeclampsia; final editing of the manuscript. M.L.: provided
expertise on Eng, analysis and interpretation of studies involving biochemical
characterization of sEng, TGF-b effects on eNOS dephosphorylation, studies
of sEng on TGF-b effects; substantially contributed to writing and editing of
manuscript. S.A.K: principal investigator of the study; overall design and
concept, analysis and interpretation of all data, drafting and final editing
of the manuscript.
© 2006 Nature Publishing Group http://www.nature.com/naturemedicine
COMPETING INTERESTS STATEMENT
The authors declare competing financial interests (see the Nature Medicine website
for details).
Published online at http://www.nature.com/naturemedicine/
Reprints and permissions information is available online at http://npg.nature.com/
reprintsandpermissions/
1. Sibai, B., Dekker, G. & Kupferminc, M. Pre-eclampsia. Lancet 365, 785–799
(2005).
2. Weinstein, L. Syndrome of hemolysis, elevated liver enzymes, and low platelet count: a
severe consequence of hypertension in pregnancy. Am. J. Obstet. Gynecol. 142,
159–167 (1982).
3. Roberts, J.M. et al. Preeclampsia: an endothelial cell disorder. Am. J. Obstet. Gynecol.
161, 1200–1204 (1989).
4. Maynard, S.E. et al. Excess placental soluble fms-like tyrosine kinase 1 (sFlt1) may
contribute to endothelial dysfunction, hypertension, and proteinuria in preeclampsia.
J. Clin. Invest. 111, 649–658 (2003).
5. Zhou, Y. et al. Vascular endothelial growth factor ligands and receptors that regulate
human cytotrophoblast survival are dysregulated in severe preeclampsia and hemolysis,
elevated liver enzymes, and low platelets syndrome. Am. J. Pathol. 160, 1405–1423
(2002).
6. Ahmad, S. & Ahmed, A. Elevated placental soluble vascular endothelial growth
factor receptor-1 inhibits angiogenesis in preeclampsia. Circ. Res. 95, 884–891
(2004).
7. Chaiworapongsa, T. et al. Evidence supporting a role for blockade of the vascular
endothelial growth factor system in the pathophysiology of preeclampsia. Young
Investigator Award. Am. J. Obstet. Gynecol. 190, 1541–1547; discussion
1547–1550 (2004).
8. Taylor, R.N. et al. Longitudinal serum concentrations of placental growth factor:
evidence for abnormal placental angiogenesis in pathologic pregnancies. Am. J. Obstet.
Gynecol. 188, 177–182 (2003).
9. Levine, R.J. et al. Circulating angiogenic factors and the risk of preeclampsia. N. Engl.
J. Med. 350, 672–683 (2004).
10. Chaiworapongsa, T. et al. Plasma soluble vascular endothelial growth factor receptor-1
concentration is elevated prior to the clinical diagnosis of pre-eclampsia. J. Matern.
Fetal Neonatal Med. 17, 3–18 (2005).
11. Hertig, A. et al. Maternal serum sFlt1 concentration is an early and reliable predictive
marker of preeclampsia. Clin. Chem. 50, 1702–1703 (2004).
12. Romero, R. et al. Clinical significance, prevalence, and natural history of thrombocy-
topenia in pregnancy-induced hypertension. Am. J. Perinatol. 6, 32–38 (1989).
13. Cheifetz, S. et al. Endoglin is a component of the transforming growth factor-beta
receptor system in human endothelial cells. J. Biol. Chem. 267, 19027–19030
(1992).
14. Gougos, A. et al. Identification of distinct epitopes of endoglin, an RGD-containing
glycoprotein of endothelial cells, leukemic cells, and syncytiotrophoblasts. Int.
Immunol. 4, 83–92 (1992).
15. St-Jacques, S., Forte, M., Lye, S.J. & Letarte, M. Localization of endoglin, a
transforming growth factor-beta binding protein, and of CD44 and integrins in placenta
during the first trimester of pregnancy. Biol. Reprod. 51, 405–413 (1994).
16. McAllister, K.A. et al. Endoglin, a TGF-b binding protein of endothelial cells, is the
gene for hereditary haemorrhagic telangiectasia type 1. Nat. Genet. 8, 345–351
(1994).
17. Bourdeau, A., Dumont, D.J. & Letarte, M. A murine model of hereditary hemorrhagic
telangiectasia. J. Clin. Invest. 104, 1343–1351 (1999).
18. Li, D.Y. et al. Defective angiogenesis in mice lacking endoglin. Science 284,
1534–1537 (1999).
19. Toporsian, M. et al. A role for endoglin in coupling eNOS activity and regulating
vascular tone revealed in hereditary hemorrhagic telangiectasia. Circ. Res. 96,
684–692 (2005).
20. Dimmeler, S., Dernbach, E. & Zeiher, A.M. Phosphorylation of the endothelial nitric
oxide synthase at ser-1177 is required for VEGF-induced endothelial cell migration.
FEBS Lett. 477, 258–262 (2000).
NATURE MEDICINE VOLUME 12 [ NUMBER 6 [ JUNE 2006
ARTICLES
21. Garcia-Cardena, G. et al. Dynamic activation of endothelial nitric oxide synthase by
Hsp90. Nature 392, 821–824 (1998).
22. Fleming, I., Fisslthaler, B., Dimmeler, S., Kemp, B.E. & Busse, R. Phosphorylation of
Thr(495) regulates Ca(2+)/calmodulin-dependent endothelial nitric oxide synthase
activity. Circ. Res. 88, E68–E75 (2001).
23. Brown, M.A., Zammit, V.C. & Lowe, S.A. Capillary permeability and extracellular fluid
volumes in pregnancy-induced hypertension. Clin. Sci. (Lond.) 77, 599–604
(1989).
24. Inoue, N. et al. Molecular regulation of the bovine endothelial cell nitric oxide synthase
by transforming growth factor-beta 1. Arterioscler. Thromb. Vasc. Biol. 15, 1255–1261
(1995).
25. Saura, M. et al. Smad2 mediates transforming growth factor-b induction of endothelial
nitric oxide synthase expression. Circ. Res. 91, 806–813 (2002).
26. Bdolah, Y., Sukhatme, V.P. & Karumanchi, S.A. Angiogenic imbalance in the
pathophysiology of preeclampsia: newer insights. Semin. Nephrol. 24, 548–556
(2004).
27. Redman, C.W. & Sargent, I.L. Latest advances in understanding preeclampsia. Science
308, 1592–1594 (2005).
28. Li, C. et al. Plasma levels of soluble CD105 correlate with metastasis in patients with
breast cancer. Int. J. Cancer 89, 122–126 (2000).
29. Velasco-Loyden, G., Arribas, J. & Lopez-Casillas, F. The shedding of betaglycan is
regulated by pervanadate and mediated by membrane type matrix metalloprotease-1.
J. Biol. Chem. 279, 7721–7733 (2004).
30. Benian, A., Madazli, R., Aksu, F., Uzun, H. & Aydin, S. Plasma and placental levels
of interleukin-10, transforming growth factor-b1, and epithelial-cadherin in pre-
eclampsia. Obstet. Gynecol. 100, 327–331 (2002).
31. Muy-Rivera, M. et al. Transforming growth factor-b1 (TGF-b1) in plasma is associated
with preeclampsia risk in Peruvian women with systemic inflammation. Am. J.
Hypertens. 17, 334–338 (2004).
32. Hennessy, A. et al. Transforming growth factor-b 1 does not relate to hypertension in
pre-eclampsia. Clin. Exp. Pharmacol. Physiol. 29, 968–971 (2002).
33. Barbara, N.P., Wrana, J.L. & Letarte, M. Endoglin is an accessory protein that interacts
with the signaling receptor complex of multiple members of the transforming growth
factor-beta superfamily. J. Biol. Chem. 274, 584–594 (1999).
34. Shesely, E.G. et al. Elevated blood pressures in mice lacking endothelial nitric oxide
synthase. Proc. Natl. Acad. Sci. USA 93, 13176–13181 (1996).
35. Lowe, D.T. Nitric oxide dysfunction in the pathophysiology of preeclampsia. Nitric
Oxide 4, 441–458 (2000).
36. Papapetropoulos, A., Garcia-Cardena, G., Madri, J.A. & Sessa, W.C. Nitric oxide
production contributes to the angiogenic properties of vascular endothelial growth
factor in human endothelial cells. J. Clin. Invest. 100, 3131–3139 (1997).
37. Predescu, D., Predescu, S., Shimizu, J., Miyawaki-Shimizu, K. & Malik, A.B. Con-
stitutive eNOS-derived nitric oxide is a determinant of endothelial junctional integrity.
Am. J. Physiol. Lung Cell. Mol. Physiol. 289, L371–L381 (2005).
38. Tatsumi, M., Kishi, Y., Miyata, T. & Numano, F. Transforming growth factor-beta(1)
restores antiplatelet function of endothelial cells exposed to anoxia-reoxygenation
injury. Thromb. Res. 98, 451–459 (2000).
39. Ristimaki, A., Ylikorkala, O. & Viinikka, L. Effect of growth factors on human vascular
endothelial cell prostacyclin production. Arteriosclerosis 10, 653–657 (1990).
40. He, H. et al. Vascular endothelial growth factor signals endothelial cell production of
nitric oxide and prostacyclin through flk-1/KDR activation of c-Src. J. Biol. Chem. 274,
25130–25135 (1999).
41. Mills, J.L. et al. Prostacyclin and thromboxane changes predating clinical onset of
preeclampsia: a multicenter prospective study. J. Am. Med. Assoc. 282, 356–362
(1999).
42. Darland, D.C. et al. Pericyte production of cell-associated VEGF is differentiation-
dependent and is associated with endothelial survival. Dev. Biol. 264, 275–288
(2003).
43. Zhou, Y. et al. Human cytotrophoblasts adopt a vascular phenotype as they differenti-
ate. A strategy for successful endovascular invasion? J. Clin. Invest. 99, 2139–2151
(1997).
44. Zhou, Y., Damsky, C.H. & Fisher, S.J. Preeclampsia is associated with failure of human
cytotrophoblasts to mimic a vascular adhesion phenotype. One cause of defective
endovascular invasion in this syndrome? J. Clin. Invest. 99, 2152–2164 (1997).
45. Caniggia, I., Taylor, C.V., Ritchie, J.W., Lye, S.J. & Letarte, M. Endoglin regulates
trophoblast differentiation along the invasive pathway in human placental villous
explants. Endocrinology 138, 4977–4988 (1997).
46. Fonsatti, E., Altomonte, M., Arslan, P. & Maio, M. Endoglin (CD105): a target for anti-
angiogenetic cancer therapy. Curr. Drug Targets 4, 291–296 (2003).
47. Holash, J. et al. VEGF-Trap: a VEGF blocker with potent antitumor effects. Proc. Natl.
Acad. Sci. USA 99, 11393–11398 (2002).
48. ACOG practice bulletin. Diagnosis and management of preeclampsia and eclampsia.
Number 33, January 2002. Obstet. Gynecol. 99, 159–167 (2002).
49. Kuo, C.J. et al. Comparative evaluation of the antitumor activity of antiangiogenic
proteins delivered by gene transfer. Proc. Natl. Acad. Sci. USA 98, 4605–4610
(2001).
50. Akiyoshi, S. et al. c-Ski acts as a transcriptional co-repressor in transforming growth
factor-beta signaling through interaction with smads. J. Biol. Chem. 274, 35269–
35277 (1999).
649
 E R R ATA A N D C O R R I G E N D A

Erratum: Proton NMR analysis of plasma is a weak predictor of coronary artery

disease
H L Kirschenlohr, J L Griffin, S C Clarke, R Rhydwen, A A Grace, P M Schofield, K M Brindle & J C Metcalfe
Nat. Med. 12, 705–710 (2006); published online 28 May 2006; corrected after print 19 June 2006

In the version of this article initially published, the layout of the data in Tables 1 and 2 is incorrect.
The error has been corrected in the HTML and PDF versions of the article.
© 2006 Nature Publishing Group http://www.nature.com/naturemedicine

Erratum: Mig6 is a negative regulator of EGF receptor–mediated skin

morphogenesis and tumor formation
I Ferby, M Reschke, O Kudlacek, P Knyazev, G Pantè, K Amann, W Sommergruber, N Kraut, A Ullrich, R Fässler & R Klein
Nat. Med. 12, 568–573 (2006); published online 30 April 2006; corrected after print 16 June 2006

In the version of this article initially published, Errfi1 was incorrectly referred to as Erffi1 in several instances, and in Figure 1h the middle panels
of the immunoblot were labeled Egf instead of Hgf.
The errors have been corrected in the HTML and PDF versions of the article.

Corrigendum: Soluble endoglin contributes to the pathogenesis of

preeclampsia
S Venkatesha, M Toporsian, C Lam, J Hanai, T Mammoto, Y M Kim, Y Bdolah, K-H Lim, H-T Yuan, T A Libermann, I E Stillman, D Roberts,
P A D’Amore, F H Epstein, F W Sellke, R Romero, V P Sukhatme, M Letarte & S Ananth Karumanchi
Nat. Med. 12, 642–649 (2006)

Due to an editing error, some reference citations in the text are incorrect. In particular, the second citation of ref. 22 (on p. 644) should be ref. 23
and refs. 23–49 should be refs. 24–50.
This has been corrected in the HTML and PDF versions of the article.

862 VOLUME 12 | NUMBER 7 | JULY 2006 NATURE MEDICINE