Nature | Vol 578 | 13 February 2020 | 293Cell–cell and cell–matrix adhesions represent two sites that are
highly mechanically active within ECs. These sites include the junc-
tional mechanosensory complex comprising PECAM-1, VEGFR2 and
VE-cadherin and integrins at the cytoskeleton–extracellular matrix
interface^13 ,^17 (Fig. 3a). We analysed the relationship between PLXND1
and these mechanical ‘hotspots’ in ECs. En face confocal imaging
revealed robust and similar expression of PLXND1 in ECs in both
the arch and descending aorta and colocalization with PECAM-1 at
cell–cell junctions (Extended Data Fig. 7a). Staining was specific as
it was not observed in Plxnd1iECKO aortas (Extended Data Fig. 7b).
SEMA3E was also observed in en face sections and expression of
SEMA3E was found to be lower in the arch (Extended Data Fig. 7c).
Co-immunoprecipitation experiments showed a flow-induced asso-
ciation of PLXND1 with components of the junctional mechanosen-
sory complex (PECAM-1, VEGFR2, VE-cadherin and the p85 subunit
of PI3K) (Extended Data Fig. 7d). To explore whether PLXND1 is just
another component of the junctional complex or whether it operates
upstream of the junctional complex, we used immunoprecipitation
to analyse complex formation at the level of the junctional mecha-
nosensory complex and integrin–matrix adhesions. We found that
responses at the junctional complex, such as shear-stress-induced
phosphorylation of VEGFR2 and association of the p85 subunit of
PI3K and VE-cadherin with VEGFR2^13 , were all abrogated by knock-
down of PLXND1 (Fig. 3b). Consistent with this observation, both the
inhibition of the VEGFR2 receptor kinase (Fig. 3d) and the deletion of
PECAM-1 abrogated force-induced signalling, suggesting that junc-
tional mechanosensory components are necessary intermediates
for the PLXND1-mediated force response (Extended Data Fig. 8a).
Similarly, flow-induced complex formation at integrin–matrix adhe-
sions (as assayed by association of Shc with integrin αvβ 3 )^18 ,^19 was also
strongly reduced with loss of PLXND1 (Fig. 3c). A previous study has
highlighted a role for PIEZO1-mediated and Gαq/G 11 -mediated mecha-
nosignalling, although there are conflicting reports as to whether
these pathways are linked^20 ,^21 or independent of each other^22 ,^23. Force
application on PLXND1 showed that loss of Gαq/G 11 abolished the
PLXND1 force response, whereas knockdown of PIEZO1 had no effect
(Extended Data Figs. 1e, f, 8b, c).
To further investigate the molecular mechanisms, we examined
the role of the PLXND1 coreceptor neuropilin-1 (NRP1). NRP1 is a
cell-surface transmembrane protein that acts as a SEMA3 and VEGF
coreceptor for PLXND1 and VEGFR2, respectively^24 , and its presence
in neurons switches the SEMA3E signal from repulsion to attraction^25.
We found that NRP1 was required for the PLXND1-mediated force
response, as both knockdown (Extended Data Fig. 1d) and inhibition
of NRP1 abolished the force-induced phosphorylation of vinculin
(Fig. 3d). We also observed that shear stress induced the formation
of a complex between PLXND1, VEGFR2 and NRP1 (Fig. 3e, f) and this
complex was dependent on NRP1 (Fig. 3g). Taken together, these
data show that PLXND1 associates with NRP1 and VEGFR2 in response
to flow and operates upstream of both the junctional complex and
integrins.
To test whether PLXND1 (and its molecular partners) is sufficient
to confer mechanosensitivity in a heterologous cell line, we trans-
fected COS-7 cells with plasmids expressing PLXND1, NRP1 and/
or VEGFR2 and applied shear stress (Fig. 3h). These cells do not
express any of the components of the junctional complex (that is,
PECAM-1 or VE-cadherin) and are therefore an ideal system to moni-
tor mechanical responses that are specifically due to PLXND1. COS-7
cells expressing all three proteins (VEGFR2, NRP1 and PLXND1)
showed activation of early signalling responses, including phos-
phorylation of VEGFR2, association of VEGFR2 with Src tyrosine
kinase, and PLXND1–VEGFR2–NRP1 complex formation in response
to shear stress. Notably, none of these responses occurred in the
absence of PLXND1, thus providing further evidence that PLXND1
is a specific and direct force sensor (Fig. 3i). Overall, these data(^0025025)
1
2
3
4
5
- – +
Force 0 2 5 0 2 5 0 2 5 0 2 5IP: VEGFR2abd(^0051030051030)
1
2
3
(^4) **
(^0051030051030)
2
4
6
8
**
(^0051030051030)
5
10
15
p-Akt
Akt
p-ERK1/2
ERK1/2
p-eNOS
eNOS
Shear 0 5 10 30 0 5 10 30
WT Mutant WT Mutant WT Mutant
Shear
time (min)
h
WT Mutant
VEGFR2
g
0
1
2
3
4
(^5)
Vehicle SEMA3E No force Force
WT
Mutant
WT
Mutant
e f
0
0.5
1.0
1.5 ***
WT
Mutant
Sema3E – + – + Force – + – +
WT
Mutant
Shear
stress
pVEGFR2
Src
(^0025025)
1
2
3
(^4)
(^0025025)
2
4
6
(^8) *
WT Mutant WT Mutant WT Mutant
Force time (min)
p-Akt
Akt
p-ERK1/2
ERK1/2
p-VEGFR2
VEGFR2
NRP1
PLXND1
c
i
time
(min)
0 2 5 0 2 5
time 0 5 10 30 0 5 10 30
(min)
0 2 5 0 2 5 0 2 5 0 2 5 0 2 5 0 2 5
0 5 10 30 0 5 10 30
1 SS SEMA PSI1IPT1PSI2IPT2PSI3IPT3IPT4IPT5IPT6 1925
TM
Y517C
A1135C
C
PLXND1(Y517C/A1135C)
Cys
Cys
SS
0 5 10 30 0 5 10 30
0 5 10 30 0 5 10 30 0 5 10 30 0 5 10 30
Normalized FA Normalized FA
Fold change Fold change Fold change
Fold change Fold change Fold change
Fig. 4 | PLXND1 f lexion is required for mechanotransduction. a, Schematic
domain organization of PLXND1 spanning amino acids 1–1925. SS, signal
sequence; TM, transmembrane region; c, cytoplasmic region.
b, Representative negative-stain class averages of the PLXND1 ectodomain
and corresponding structural models showing the ring-like and open
conformations. Scale bar, 10 nm. Two-dimensional class averages were
obtained by classifying 1,357 particles into 10 classes. c, Model of opening of
the ring-like ectodomain, which confers the mechanosensory functions of the
PLXND1. d, Design of the PLXND1 mutant with an intramolecular disulfide bond
to lock the ring-like structure. The magnified view shows the disulfide bond
between the SEMA domain (domain 1) and IPT5 domain (domain 9). e, f, Bovine
ECs in which endogenous PLXND1 was knocked down were infected with
adenoviruses expressing wild-type or mutant PLXND1, treated with SEMA3E
for 30 min or incubated with anti-PLXND1 paramagnetic beads followed by
force application (10 pN; 30 min). Cells were immunostained with anti-vinculin
antibodies. Focal adhesion numbers were quantified using ImageJ; n = 30 cells
across either 4 (e) or 3 (f) biological replicates. ***P < 0.0001. Scale bar, 10 μm.
g, COS-7 cells were transfected with wild-type or mutant PLXND1, NRP1 and
VEGFR2 before shear-stress application for 2 min and VEGFR2 was
immunoprecipitated. Shear-stress sensitivity was assessed by analysing the
levels of phosphorylated VEGFR2, the complex formation between VEGFR2
and Src and the complex of PLXND1, VEGFR2 and NRP1. n = 3. h, Mouse ECs in
which endogenous PLXND1 was knocked down were infected with
adenoviruses expressing wild-type or mutant PLXND1 and incubated with anti-
PLXND1 paramagnetic beads followed by force application (10 pN).
Phosphorylation of Akt, ERK1/2 and VEGFR2 was determined. n = 3. P < 0.05
relative to the no force condition; #P < 0.05 relative to the force time point of the
respective wild-type protein. i, Mouse ECs in which endogenous PLXND1 was
knocked down were infected with adenoviruses expressing wild-type or
mutant PLXND1 and subjected to f luid shear stress. Phosphorylation of Akt,
ERK1/2 and eNOS was determined. n = 3 biological repeats. *P < 0.05 relative to
the static condition; #P < 0.05 relative to the respective shear time point of the
wild-type protein. Data are mean ± s.e.m. P values were obtained using two-
tailed Student’s t-test using GraphPad Prism.