The Journal of Immunology
Differential Roles for Endothelial ICAM-1, ICAM-2, and
VCAM-1 in Shear-Resistant T Cell Arrest, Polarization, and
Directed Crawling on Blood–Brain Barrier Endothelium
Oliver Steiner,* Caroline Coisne,* Rome ´o Cecchelli,†Re ´my Boscacci,* Urban Deutsch,*
Britta Engelhardt,* and Ruth Lyck*
Endothelial ICAM-1 and ICAM-2 were shown to be essential for T cell diapedesis across the blood–brain barrier (BBB) in vitro
under static conditions. Crawling of T cells prior to diapedesis was only recently revealed to occur preferentially against the
direction of blood flow on the endothelial surface of inflamed brain microvessels in vivo. Using live cell-imaging techniques, we
prove that Th1 memory/effector T cells predominantly crawl against the direction of flow on the surface of BBB endothelium
in vitro. Analysis of T cell interaction with wild-type, ICAM-1–deficient, ICAM-2–deficient, or ICAM-1 and ICAM-2 double-
deficient primary mouse brain microvascular endothelial cells under physiological flow conditions allowed us to dissect the in-
dividual contributions of endothelial ICAM-1, ICAM-2, and VCAM-1 to shear-resistant T cell arrest, polarization, and crawling.
Although T cell arrest was mediated by endothelial ICAM-1 and VCAM-1, T cell polarization and crawling were mediated by
endothelial ICAM-1 and ICAM-2 but not by endothelial VCAM-1. Therefore, our data delineate a sequential involvement of
endothelial ICAM-1 and VCAM-1 in mediating shear-resistant T cell arrest, followed by endothelial ICAM-1 and ICAM-2 in
mediating T cell crawling to sites permissive for diapedesis across BBB endothelium.
T cells has been characterized as a multistep process involving
T cell rolling along the vascular surface, T cell arrest and crawling
on the endothelium and, finally, diapedesis of T cells (1). During
interaction with the endothelium, T cells have to resist detachment
by the shear forces they are exposed to within the bloodstream.
Interestingly, recent studies documented that crawling of T cells,
as well as of neutrophils, is perpendicular or even against the di-
rection of blood flow in vivo (2, 3). Thus, crawling of T cells on
the luminal surface of the vasculature requires a spatiotemporal
regulation of the adhesion strength between the T cell and the
endothelium. It was shown that the a4-integrins a4b1 and a4b7
The Journal of Immunology, 2010, 185:
nteraction of T cells with the vascular endothelium is a criti-
cal step during T cell extravasation from blood into tissue
in immunosurveillance and inflammation. Extravasation of
and the b2-integrin LFA-1 control shear-resistant interactions of
T cells with the endothelium (4). In fact, the polarized distribution
of intermediate-affinity and high-affinity forms of LFA-1 along
the T cell axis has been observed during postarrest T lymphocyte
crawling on the endothelial luminal surface or on immobilized
ICAM-1 (5, 6).
T cell interactions with the blood–brain barrier (BBB) endo-
thelium may be unique because of the high specialization of these
endothelial cells in building a tight barrier to maintain CNS ho-
molecules as a result of an extremely low pinocytotic activity and
restrict the paracellular diffusion of hydrophilic molecules by an
between endothelial cells (8). When brain endothelial cells are
isolated and cultured in vitro, they rapidly lose many of their
BBB characteristics, including formation of proper tight junctions
and a permeability barrier, indicating that integrity of the BBB
This is confirmed by observations made with brain-derived endo-
thelioma cells, which maintain some BBB characteristics but lose
expression of a number of BBB-specific junctional molecules and,
therefore, their barrier characteristics (9). Barrier characteristics of
brain endothelium extend to the regulation of lymphocyte traf-
ficking into the CNS. During physiological conditions, lymphocyte
migration across the BBB is very low and restricted to highly ac-
tivated lymphocytes (10–12). However, during inflammatory dis-
eases of the CNS, such as multiple sclerosis (MS), circulating
immunocompetent cells readily gain access to the CNS, where
they induce inflammation, edema formation, and BBB breakdown,
which together establish the clinical picture of the disease (2, 11).
Using brain endothelial cells under static conditions, we showed
that endothelial VCAM-1 contributes to T cell adhesion, whereas
endothelial ICAM-1 and ICAM-2 are important for T cell adhesion
and diapedesis (13–16). However, the selective roles of endothe-
lial VCAM-1, ICAM-1, and ICAM-2 in mediating shear-resistant
*Theodor Kocher Institute, University of Bern, Bern, Switzerland; and
Lille-Nord de France, Lens, France
Received for publication November 19, 2009. Accepted for publication August 5,
This work was supported by the European Stroke Network (Grants EU FP7 No. 201024
and No. 202213 to R.C. and B.E.) and by grants from the Swiss Multiple Sclerosis
Society, the Novartis Foundation for Biomedical Research, and the SwissLife Founda-
tion (to R.L.). O.S. received a one-year fellowship from the French Multiple Sclerosis
Research Society, Fondation pour l’Aide a ` la Recherche sur la Scle ´rose En Plaques.
Address correspondence and reprint requests to Dr. Britta Engelhardt and Dr. Ruth
Lyck, Theodor Kocher Institute, University of Bern, Freiestrasse 1, 3012 Bern, Swit-
zerland. E-mail addresses: email@example.com and firstname.lastname@example.org
The online version of this article contains supplemental material.
Abbreviations used in this paper: BBB, blood–brain barrier; Dacc, accumulated total
distance of T cell movement; Dx, straight x-axis distance covered by the T cell; EAE,
experimental autoimmune encephalomyelitis; FOV, field of view; I1-KO, ICAM-1
deficient; I1/I2-KO, ICAM-1 and ICAM-2 deficient; I2-KO, ICAM-2 deficient; IF,
immunofluorescence; mICAM-1, mouse ICAM-1/Fc; mICAM-2, mouse ICAM-2/Fc;
MS, multiple sclerosis; mVCAM-1, mouse VCAM-1/Fc; Pe 3kDa, permeability co-
efficient for 3-kDa dextran; Pe 10kDa, permeability coefficient for 10-kDa dextran;
PLP, proteolipid protein; pMBMEC, primary mouse brain microvascular endothelial
cell; wt, wild-type; xFMI, forward migration index toward x-axis.
T cell arrest, versus T cell polarization and T cell crawling on the
surface of brain endothelium, under physiological flow conditions
have not been investigated. In this study, we specifically focused
on dissecting the individual contributions of these endothelial cell
adhesion molecules in the different steps of T cell extravasation
across the BBB in vitro. Using wild-type (wt), ICAM-1–deficient
(I1-KO), ICAM-2–deficient (I2-KO), or ICAM-1 and ICAM-2
deficient (I1/I2-KO) primary mouse brain microvascular endothe-
lial cells (pMBMECs) freshly isolated prior to each experiment
from the respective mice, we analyzed arrest, polarization, crawl-
ing velocity, crawling direction, and diapedesis of encephalitogenic
Th1 memory/effector cells across the highly specialized BBB
endothelium under physiological flow conditions. The specific
function of VCAM-1 for T cell arrest was substantiated by Ab
masking of endothelial VCAM-1 on pMBMECs. Recombinant
mouse ICAM-1, ICAM-2, and VCAM-1 were used to analyze the
individual roles of each cell adhesion molecule in supporting ar-
rest, polarization, and crawling against the direction of shear. In
summary, we demonstrate that endothelial ICAM-1 and VCAM-1,
but not ICAM-2, mediate shear-resistant T cell arrest on BBB
endothelium. Subsequently, endothelial ICAM-1 or ICAM-2, but
not VCAM-1, is sufficient to mediate T cell polarization and
crawling on the surface of the BBB, because both events are
completely abrogated in the absence of ICAM-1 and ICAM-2.
Endothelial ICAM-1 but not ICAM-2 enables T cells to crawl
against the direction of flow on the BBB endothelium in vitro.
Materials and Methods
pMBMECs were isolated from gender-matched 4–6-wk-old C57BL/6 mice
(Harlan Laboratories, Horst, The Netherlands), cultured as described,
and used nonpassaged on day 5 or 6 after isolation (9, 17). Stimulated
pMBMECs were cultured for 16–18 h in the presence of TNF-a (25 ng/
ml). All experiments were performed in migration assay medium (DMEM,
5% calf serum, 25 mM HEPES) at 37˚C. I1-KO and I2-KO mice (18, 19)
were backcrossed onto C57BL/6 mice for at least eight generations. I1/I2-
KO C57BL/6 mice were obtained by breeding I1-KO mice with I2-KO
C57BL/6 mice. All animal procedures were performed in accordance with
the Swiss legislation on the protection of animals and were approved by
the veterinary office of the Kanton of Bern.
The proteolipid protein (PLP)-specific CD4+Th1 effector/memory T cell
line SJL.PLP7 raised against the PLP peptide aa139–153was described in
detail (13). T cells were used 3 d after the third or fourth restimulation with
their specific PLP Ag.
Permeability assays were performed in triplicates for one value within each
assay, as published (17), with minor adaptations: BBB endothelial cells
10 kDa, 10 mg/ml; LuBioScience, Luzerne, Switzerland) was used as per-
meability tracer. Diffused dextran was quantified usingthe Odyssey Imaging
System (LI-COR, Bad Homburg, Germany).
T cell adhesion and diapedesis under static conditions
T cell adhesion and diapedesis assays were performed, as described pre-
viously (16, 20), in triplicates for one value in each experiment.
Coating of cell-culture dishes with recombinant cell-adhesion
For live cell imaging of T cell interaction with recombinant purified
cell-adhesion molecules, mouse ICAM-1/Fc (mICAM-1), ICAM-2/Fc
(mICAM-2), and VCAM-1/Fc (mVCAM-1) (R&D Systems, Abingdon,
U.K.) were bound to culture dishes whose surfaces were coated (1 h at
37˚C, 20 mg/ml in PBS [pH 9]) with protein A (BioVision, Axxora Europe,
Lausen, Switzerland). Protein Awas overlaid with mICAM-1 or mVCAM-1
(10 mg/ml in PBS [pH 7.4]) or mICAM-2 (6.6 mg/ml in PBS, [pH 7.4] to
keep equal molarity with ICAM-1) for 2 h at 37˚C. Surfaces were blocked
with 1.5% BSA in PBS (pH 7.4) for 30 min before use. Lack of unspecific
T cell interactions with the substratum was tested in control experiments
using protein A surfaces blocked with BSA or protein A overlaid with the rat
IgG2a Ab 9B5 directed against human CD44.
Live cell imaging under flow conditions
For live cell imaging, a parallel flow chamber (21) connected to an auto-
mated syringe pump (Harvard Apparatus, Holliston, MA) was mounted on
TNF-a–stimulated pMBMECs or on immobilized mICAM-1, mICAM-2,
or mVCAM-1 and placed on the heating stage of an inverted microscope
(imaging of T cell interaction with pMBMECs: Axiovert 200; imaging of
T cell interaction with immobilized mICAM-1, mICAM-2 or mVCAM-1:
AxioObserver Z1; both microscopes from Carl Zeiss, Feldbach, Switzer-
land). Shear stress (dyn/cm2) was calculated according to the equation:
t = 3mQ/2a2b, where t is wall shear stress, m is coefficient of viscosity,
Q is volumetric flow rate, a is half channel height, and b is channel width
(22). T cells (5 3 105/ml) were allowed to accumulate for 4 min at low
shear stress (0.25 dyn/cm2). Then, dynamic T cell interactions with
pMBMECs or immobilized mICAM-1, mICAM-2, or mVCAM-1 were
recorded under physiological shear stress (1.5 dyn/cm2) at 3100 magni-
fication (pMBMECs: objective A-Plan 310/0.25; mICAM-1, mICAM-2, or
mVCAM-1: objective EC Plan Neofluar 310/0.3) using a monochrome
CCD camera (pMBMECs: Cohu, Poway, CA, connected to a digital video
recording system; mICAM-1, mICAM-2, or mVCAM-1: AxioCam MRm
Rev, Carl Zeiss). Time-lapse videos were created from one frame every
30 s (pMBMECs: iMovie, Apple, Cupertino, CA; mICAM-1, mICAM-2,
or mVCAM-1: AxioVision, Carl Zeiss).
Unless noted otherwise, data are presented as mean 6 SEM, and differ-
ences between two groups were analyzed by the unpaired Student t test.
For three or more groups, one-way ANOVA was performed, followed by
the Tukey multiple comparison test. A p value ,0.05 was considered
significant. Statistical analysis was done using GraphPad Prism 5 software
(GraphPad, San Diego, CA).
Absence of endothelial ICAM-1 and ICAM-2 has no influence
on barrier characteristics of pMBMECs in vitro
ICAM-1 versus ICAM-2 and VCAM-1 for T cell arrest and crawling
along the luminal surface of and their diapedesis across the highly
specialized BBB endothelium. Considering the unique barrier
characteristics of brain endothelium, we specifically established an
invitro BBB model with pMBMECs, as described previously (17).
To this end, pMBMECs were isolated from wt, I1-KO, I2-KO, and
endothelial monolayers without any apparent morphological dif-
ferences. Endothelial purity was confirmed for all four types of
pMBMECs by uniformly positive immunofluorescence (IF) stain-
ing for the endothelial marker protein von Willebrand factor
(Supplemental Fig. 1A). Because the molecular architecture of the
BBB endothelial intercellular junctions is highly specialized, we
carefully examined the molecular composition of the endothelial
junctions of wt, I1-KO, I2-KO, and I1/I2-KO pMBMECs by IF
analysis. We detected expression and proper junctional localization
of the endothelial adherens junction moleculevascular endothelial-
cadherin and of the tight junction proteins claudin-5, claudin-3,
occludin, and junctional adhesion molecule-A. Additionally, plate-
let endothelial cell adhesion molecule-1 and the intracellular
junction-associated scaffolding proteins zonula occludens-1 and -2
were found to be localized in the endothelial cell contacts, in-
dependent of the presence or absence of ICAM-1 and ICAM-2
(Supplemental Fig. 1A, 1B). Stimulation of wt pMBMECs with
TNF-a induced upregulation of ICAM-1 and VCAM-1 on the
endothelial cell surface, whereas cell-surface levels of ICAM-2
remained unchanged. We found ICAM-1, ICAM-2, and VCAM-1
The Journal of Immunology 4847
localized to the endothelial surface, with minor or no staining of
the intercellular junctions (Supplemental Fig. 1C). ICAM-1 or
ICAM-2 cell-surface expression was missing on the respective
ICAM-deficient pMBMECs under control and stimulatory con-
ditions (data not shown). By careful visual comparison of equally
stained and recorded IF images, we found comparable expression
levels of ICAM-1 or ICAM-2 under control and inflammatory
conditions on those pMBMECs expressing the respective mole-
cules (Supplemental Fig. 1D).
To demonstrate overall tightness of monolayers formed by
pMBMECs, we performed permeability assays with 3- and 10-kDa
dextran as low molecular paracellular tracers. First, we compared
line that has lost BBB-specific complexity of tight junction orga-
nization (9). For pMBMECs we determined a mean permeability
coefficient for 3-kDa dextran (Pe 3kDa) of 0.12 6 0.01 3 1023cm/
min and for 10-kDa dextran (Pe 10kDa) of 0.06 6 0.01 3 1023cm/
min. For bEnd5, Pe 3kDawas 2.20 6 0.29 3 1023cm/min and
Pe 10kDawas 0.91 6 0.16 3 1023cm/min. Therefore, the perme-
ability coefficients for 3- and 10-kDa dextran for pMBMECs were
18-fold (Pe 3kDa) or 15-fold (Pe 10kDa) lower than for bEnd5,
demonstrating the strikingly reduced paracellular permeability of
pMBMECs compared with bEnd5 (Supplemental Fig. 2). Thus,
pMBMECs form tight monolayers in vitro.
Next, we tested whether the absence of ICAM-1 and/or ICAM-2
influences pMBMEC barrier integrity. To this end, we compared
the paracellular permeability of wt pMBMECs with the respec-
tive paracellular permeabilities of I1-KO, I2-KO, and I1/I2-KO
pMBMECs. The Pe 3kDavalues for I1-KO, I2-KO, and I1/I2-KO
pMBMECs showed no significant differences compared with
wt pMBMECs (I1-KO Pe 3kDa, 0.10 6 0.02 3 1023cm/min; I2-KO
Pe 3kDa, 0.15 6 0.03 3 1023cm/min; and I1/I2-KO Pe 3kDa, 0.15 6
0.03 3 1023cm/min) (Fig. 1A). Likewise, permeability coefficient
values obtained for 10-kDa dextran confirmed the comparable
tightness of all four types of pMBMECs (I1-KO Pe 10kDa, 0.05 6
0.01 3 1023cm/min; I2-KO Pe 10kDa, 0.08 6 0.01 3 1023cm/min;
and I1/I2-KO Pe 10kDa, 0.08 6 0.01 3 1023cm/min) (Fig. 1B). In
summary, pMBMECs formed a tight BBB in vitro, independent of
the presence or absence of ICAM-1 or ICAM-2.
Endothelial ICAM-1 is essential for T cell adhesion to and
diapedesis across the BBB under static conditions
To ensure the validity of our previous findings on the roles of
endothelial ICAM-1 and ICAM-2 in Th1 cell adhesion to and
diapedesis across immortalized brain endothelioma cells, which do
not form a tight permeability barrier in vitro, we reassessed the
involvement of endothelial ICAM-1 and ICAM-2 in T cell in-
teraction with the in vitro BBB (15, 16, 18). To this end, we in-
vestigated T cell adhesion and diapedesis across pMBMECs under
static conditions. First, we compared T cell interaction with un-
stimulated and TNF-a–stimulated wt pMBMECs. As expected,
stimulation of pMBMECs with TNF-a induced a strong upregu-
lation of T cell adhesion (2.1-fold increase) and T cell diapedesis
(2.5-fold increase) compared with unstimulated pMBMECs (Sup-
plemental Fig. 3).
Next, we compared T cell adhesion to I1-KO, I2-KO, and I1/I2-
KO pMBMECs with T cell adhesion to wt pMBMECs (100%)
under unstimulated and TNF-a–stimulated conditions (Fig. 2A).
Under both conditions, T cell adhesion to I1-KO pMBMECs was
significantly reduced (unstimulated, 37.5 6 4.8%; TNF-a stimu-
lated, 79.4 6 6.5%) compared with wt pMBMECs. In contrast,
T cell adhesion to I2-KO pMBMECs remained unchanged (un-
stimulated, 92.7 6 12.5%; TNF-a stimulated, 98.7 6 1.0%).
However, a greater reduction in T cell adhesion to I1/I2-KO
pMBMECs (unstimulated, 15.7 6 3.1%; TNF-a stimulated,
42.1 6 2.6%) demonstrated an additional role of endothelial
ICAM-2 in T cell adhesion to brain endothelium, which was only
detectable in the absence of endothelial ICAM-1 (Fig. 2A).
Comparing T cell diapedesis across I1-KO, I2-KO,and I1/I2-KO
pMBMECs with T cell diapedesis across wt pMBMECs (100%)
under unstimulated and TNF-a stimulated conditions (Fig. 2B), we
found results similar to T cell adhesion. T cell diapedesis across
I1-KO or I1/I2-KO pMBMECs was strongly reduced (I1-KO:
unstimulated, 35.2 6 8.9%, TNF-a stimulated, 51.3 6 3.8%;
I1/I2-KO: unstimulated, 43.6 6 4.4%, TNF-a stimulated 47.3 6
5.2%), but diapedesis of T cells across I2-KO pMBMECs re-
mained unchanged (unstimulated I2-KO: 91.7 6 11.9%, TNF-a–
stimulated I2-KO: 114.2 6 13.6%). Under these static conditions,
an additional contribution of ICAM-2 to T cell diapedesis in the
absence of ICAM-1 was not detectable, because the values ob-
tained for T cell diapedesis across I1/I2-KO pMBMECs were
similar to the values obtained for I1-KO pMBMECs.
Taken together, endothelial ICAM-1 fulfils a major function in
T cell adhesion to and in T cell diapedesis across the BBB under
static conditions. In the presence of endothelial ICAM-1, endo-
thelial ICAM-2 is only of minor relevance for T cell interaction
with pMBMECs. However, in the absence of endothelial ICAM-1,
ICAM-2 was found to contribute to T cell adhesion to the BBB
Endothelial ICAM-1 and VCAM-1, but not ICAM-2, mediate
T cell arrest on the BBB under flow
Intermediate- and high-affinity forms of LFA-1 were described
to control T cell dynamic interaction with the endothelium or on
purified ICAM-1 under physiological flow (5, 6, 23). The ex-
periments described above do not allow dissection of the role of
KO BBB endothelium. Pe 3kDa(A) and Pe 10kDa(B) for wt, I1-KO, I2-KO,
and I1/I2-KO pMBMECs. Permeability coefficients for endothelial mono-
layers were calculated from diffused Alexa Fluor 680-dextran (four time
points of 10 min each), as published (36). Bars represent mean 6 SEM of
at least three independent experiments. One-way ANOVA, followed by the
Tukey multiple-comparison test, confirmed the absence of significant dif-
ferences (all p values .0.05).
Unchanged permeabilities of wt, I1-KO, I2-KO, and I1/I2-
4848ROLE OF ICAM-1 AND -2 AND VCAM-1 IN T CELL–BBB INTERACTIONS
endothelial ICAM-1 and ICAM-2 in shear-resistant T cell arrest
and crawling on the BBB prior to their diapedesis across the BBB.
Therefore, we investigated the roles of endothelial ICAM-1 and
ICAM-2 in T cell interaction with the BBB in vitro by performing
live cell imaging in a flow chamber experimental setup. We spe-
cifically focused on inflammatory conditions of the BBB with
regard to the relevance for T cell trafficking across the BBB in
chronic inflammatory diseases, such as MS. T cells were perfused
over a monolayer of pMBMECs at low shear stress (0.25 dyn/cm2)
to allow for their accumulation. Subsequently, shear stress was
increased to physiological strength (1.5 dyn/cm2), and T cell in-
teraction with pMBMECs was recorded under constant flow over
30 min. To determine the number of T cells that arrested in a
shear-resistant manner to the pMBMEC monolayer during the
accumulation phase, we analyzed one image acquired after the
first 30 s of enhanced shear stress and counted the numbers of
arrested T cells per field of view (FOV). These T cells, which
initially adhered under low shear stress and resisted immediate
flow enhancement, were named arrested T cells.
We counted 88 6 7 arrested T cells per FOVon wt pMBMECs.
A significant contribution of endothelial ICAM-1 to T cell arrest
on the BBB was shown by a reduction in arrested T cells by
48 and 50% on I1-KO (46 6 4 T cells/FOV) and I1/I2-KO
pMBMECs (44 6 3 T cells/FOV), respectively, compared with
wt pMBMECs (Fig. 3A). Lack of ICAM-2 on I2-KO pMBMECs
did not result in significantly reduced numbers of arrested T cells
compared with wt pMBMECs (Fig. 3A). In contrast to static T
cell-adhesion experiments, we counted equal numbers of T cells
firmly arrested on I1-KO and I1/I2-KO pMBMECs, demonstrating
that endothelial ICAM-2 does not significantly contribute to initial
shear-resistant T cell arrest on pMBMECs.
To investigate the role of endothelial VCAM-1 in initial shear-
resistant T cell arrest on the inflamed BBB, we masked VCAM-1
on TNF-a–stimulated wt, I1-KO, I2-KO, and I1/I2-KO pMBMECs
with Ab prior to T cell perfusion. Although initial arrest of T cells
on wt pMBMECs was reduced only marginally and nonsignifi-
cantly, a complete inhibition of T cell arrest was observed after
blocking VCAM-1 on I1-KO and I1/I2-KO pMBMECs (Fig. 3B).
Taken together, these observations demonstrate that endothelial
ICAM-1 and VCAM-1, but not ICAM-2, mediated shear-resistant
T cell arrest on the inflamed BBB in vitro.
To confirm theindividualcontributions of ICAM-1 andVCAM-1
versus ICAM-2 in mediating shear-resistant T cell arrest in a cell-
independent context, we investigated the individual abilities of
immobilized recombinant murine ICAM-1, ICAM-2, and VCAM-1
to support shear-resistant T cell arrest. On immobilized ICAM-1 or
VCAM-1, 122 6 5 and 104 6 4 T cells arrested per FOV, re-
spectively, supporting the notion that both cell-adhesion molecules
efficiently support T cell arrest under flow (Fig. 3C). In contrast,
immobilized ICAM-2 was significantly less efficient in supporting
T cell arrest, with 60 6 4 T cells counted per FOV (Fig. 3C).
Therefore, our data demonstrate a dominant role for endothelial
ICAM-1 and VCAM-1 in mediating shear-resistant T cell arrest.
The moderate intrinsic ability of ICAM-2 to support T cell arrest
under flow is fully masked when ICAM-2 is displayed in its phys-
iological context on the pMBMEC membrane in the presence of
ICAM-1 and VCAM-1.
Endothelial ICAM-1 and ICAM-2, but not VCAM-1, mediate
T cell polarization and crawling on the BBB under flow
After their shear-resistant arrest on the endothelial surface, T cells
rapidly acquired a polarized cell shape, with a characteristic broad
lamellipodium at the cell front and a typical elongated and pro-
jected uropod at the trailing edge (Fig. 4A), and started crawl-
ing. During the 30-min observation period, T cells continuously
crawled on the surface of the pMBMECs or transiently crawled on
the surface of the brain endothelial cells, followed by diapedesis
through the monolayer (Supplemental Video 1). T cells that failed
to diapedese across the brain endothelium within the 30 min of
recording time covered strikingly longer distances on the brain
endothelial surface than T cells that underwent diapedesis.
To analyze the individual contributions of endothelial ICAM-1
and endothelial ICAM-2 to T cell polarization, crawling, and di-
apedesis, we performed a visual frame-by-frame offline analysis of
the dynamic behavior of T cells arrested on wt, I1-KO, I2-KO, and
I1/I2-KO pMBMECs andassigned each arrested T cell to one of the
following four groups: T cells that crawled and diapedesed, T cells
that continuously crawled without diapedesis, T cells that stayed
stationary, and T cells that detached and failed to maintain firm
adhesion under enhanced shear. The numbers of initially arrested
T cells on the respective pMBMECs under low shear stress were set
T cells. On wt pMBMECs, the majority (52.2 6 8.9%) of T cells
underwent diapedesis after a crawling phase, and 42.1 6 9.2% of
T cells crawled continuously during the observation period. Only
a minor fraction of T cells (4.6 6 1.3%) remained stationary on
wt pMBMECs, and a very small number (1.2 6 0.1%) failed to
maintain adhesion and detached from pMBMECs (Fig. 4B).
hesion and diapedesis under static conditions. T cell adhesion to (A) and
transmigration across (B) confluent wt, I1-KO, I2-KO, and I1/I2-KO
pMBMECs. Because it was not possible to simultaneously isolate all four
types of pMBMECs, we included wt pMBMECs in each experiment and set
the values obtained for unstimulated (w/o TNF-a) or TNF-a–stimulated
(+TNF-a) wt pMBMECs to 100%. One representative experiment each
for T cell adhesion and T cell diapedesis and showing raw values are pre-
sented in Supplemental Fig. 2. A, Numbers of T cells adherent towt, I1-KO,
I2-KO,and I1/I2-KOpMBMECs were countedperFOV(600 3 600 mm) at
3200 magnification and expressed as the percentage of adherent T cells on
wt pMBMECs. Data are mean 6 SEM of seven independent experiments.
B, T cell diapedesis across wt, I1-KO, I2-KO, and I1/I2-KO pMBMECs
measured in a static two-chamber–based experimental setup (6 h migration
time). Within one experiment, each value for diapedesed T cells was cal-
culated as percentage of input. For comparison among different experi-
mean 6 SEM of 12 independent experiments. ppp , 0.01; pppp , 0.001,
versus wt; one-way ANOVA, followed by the Tukey multiple-comparison
Essential contribution of endothelial ICAM-1 to T cell ad-
The Journal of Immunology4849
In the absence of endothelial ICAM-1 on I1-KO pMBMECs,
a significantly reduced percentage of T cells (28.2 6 5.6%)
crawled and diapedesed compared with wt pMBMECs. This was
accompanied by an increase in the percentage of T cells that
remained stationary (19.1 6 1.0%) or that detached from I1-KO
pMBMECs (6.0 6 3.1%). Interestingly, the fraction of T cells that
crawled continuously on I1-KO pMBMECs remained unchanged
(46.7 6 2.7%) compared with wt pMBMECs (Fig. 4B). Lack of
endothelial ICAM-2 on I2-KO pMBMECs did not result in any
significant changes in the dynamic behavior of T cells compared
with wt pMBMECs.
However, dramatic changes in the dynamic T cell behavior on
the endothelium were obvious in the absence of endothelial ICAM-
1 and ICAM-2. In contrast to wt, I1-KO, or I2-KO pMBMECs, the
T cells that arrested on I1/I2-KO pMBMECs failed to acquire
polarized cell morphology and did not crawl on the endothelium
(Fig. 4A). Most of these T cells (51.0 6 1.7%) remained stationary
throughout the entire observation period (Fig. 4B, Supplemental
Video 2). In addition, a significantly increased fraction of initially
arrested T cells detached from I1/I2-KO pMBMECs (22.8 6 3.6%),
indicating a defect in shear-resistant T cell firm adhesion on
pMBMECs lacking ICAM-1 and ICAM-2. Despite the lack of po-
larization and crawling, few T cells dislocated from their original
site of arrest. This movement of T cells on I1/I2-KO pMBMECs
was always along the direction of flow and resembled consecutive
sequences of detachment and parallel reattachment to the endo-
thelium. The particular T cell behavior on I1/I2-KO pMBMECs
differed markedly from the effective crawling of highly polarized
T cells observed on wt, I1-KO, or I2-KO pMBMECs; instead, it
resembled a recurrent arrest behavior of T cells. Therefore, we
assigned the T cell dislocation on I1/I2-KO pMBMECs (13.2 6
6.3% of T cells) to a different category that we named “recurrent
arrest” (Fig. 4B). Interestingly, a small fraction of T cells (13.1 6
2.6%) that remained stationary were able to diapedese across I1/
I2-KO pMBMECs at the site of their initial arrest. Taken together,
our findings demonstrate that endothelial ICAM-1 and ICAM-2
are essential for polarization and crawling of T cells on the in-
T cell arrest on pMBMECs under flow. T cells that arrested on the endo-
thelium during the accumulation phase and resisted immediate detachment
by enhanced shear stress were counted by evaluation of the first frame after
flow enhancement to 1.5 dyn/cm2. A, Numbers of arrested T cells per FOV
(963 3 642 mm) on TNF-a–stimulated wt, I1-KO, I2-KO, and I1/I2-KO
pMBMECs. B, wt, I1-KO, or I1/I2-KO pMBMECs were incubated with
blocking monoclonal anti–VCAM-1 Ab (MK2.7; 20 mg/ml) for 15 min
prior to the flow experiment. C, Numbers of arrested T cells per FOV
(865 3 650 mm) on immobilized recombinant mICAM-1, mICAM-2, or
mVCAM-1. Data are mean 6 SEM from three independent experiments
for each group. ppp , 0.01; pppp , 0.001, compared with wt; one-way
ANOVA, followed by the Tukey multiple-comparison test.
Essential roles for endothelial ICAM-1 and VCAM-1 in
larization, crawling, diapedesis, and resistance to detachment of T cells
interacting with BBB endothelium under flow. A, Different morphologies of
T cells on wt (left panel) or I1/I2-KO (right panel) brain endothelium.
Differential interference contrast images were taken of T cells that were in
contact with TNF-a–stimulated pMBMECs for 12 min under enhanced
flow and subsequently fixed with 4% PFA. Original magnification 3400;
scale bars, 10 mm. B, Characterization of dynamic T cell interactions with
TNF-a–stimulated wt, I1-KO, I2-KO, or I1/I2-KO pMBMECs under flow
conditions. Crawling with diapedesis represents T cells that polarized and
crawled until they finally crossed the endothelial cell monolayer (identified
by a characteristic change in brightness under phase-contrast illumination).
Continuous crawling represents T cells that polarized and crawled at least
two T cell diameters but did not diapedese across the endothelium. Sta-
tionary represents T cells that remained stationary (less than two T cell
diameters of movement from their site of arrest). Detachment represents
T cells that detached during the 30 min of recording under physiological
shear stress. Two additional categories were defined based on observations
of T cell behavior on I1/I2-KO pMBMECs: stationary with diapedesis
represents T cells that remained stationary but finally crossed the endo-
thelial cell monolayer, and recurrent arrest represents T cells that dislocated
from the site of arrest in the absence of polarization and crawling. Data are
mean 6 SEM from three independent experiments for each group. ppp ,
0.01; pppp , 0.001, compared with wt; one-way ANOVA, followed by the
Tukey multiple-comparison test.
Individual roles of endothelial ICAM-1 and ICAM-2 in po-
4850ROLE OF ICAM-1 AND -2 AND VCAM-1 IN T CELL–BBB INTERACTIONS
flamed BBB under flow conditions. Lack of endothelial ICAM-1
and ICAM-2 results in a dramatically reduced T cell diapedesis
across the BBB. Interestingly, however, a few diapedesis events still
occurred, independent of endothelial ICAM-1 and ICAM-2.
To further delineate the contributions of ICAM-1, ICAM-2, and
VCAM-1 to the support of T cell polarization and crawling, we
analyzed T cell behavior on immobilized recombinant mouse
ICAM-1, ICAM-2, and VCAM-1 after their initial arrest under
shear for an observation period of 15 min (Fig. 5A, Supplemental
Videos 3–5). On ICAM-1 and ICAM-2, 95.0 6 1.3% and 89.0 6
4.1%, respectively, of the arrested T cells polarized rapidly and
started to crawl (Fig. 5B), whereas only a few T cells (2.8 6 1.5
and 0%, respectively) remained stationary. These observations
support the assertion that both molecules support T cell polarization
and crawling equally well. A lower avidity of T cell interaction with
ICAM-2, compared with ICAM-1, was demonstrated by the ob-
servation that only 2.2 6 0.3% of T cells detached from ICAM-1
versus 11.1 6 4.1% from ICAM-2 during the observation period.
Similar to ICAM-1 and ICAM-2, few T cells (4.3 6 1.1%)
remained stationary on immobilized VCAM-1. However, in strik-
2, T cells failed to completely polarize, flatten, and crawl on
VCAM-1. Measuring the mean length of T cells interacting with
ICAM-1 versus VCAM-1, from the leading edge to the trailing
edge, further supported the notion that VCAM-1 fails to trigger
complete T cell polarization; the mean length of T cells adhering to
VCAM-1 was only 18.8 6 0.35 mm compared with 23.1 6 0.37
mm for those polarized and crawling on immobilized ICAM-1
(Supplemental Fig. 4A). Instead, most T cells interacting with
immobilized VCAM-1 (65.0 6 8.1%) repeatedly extended and re-
tracted cell protrusions in all directions, resembling the recurrent
arrest behavior of T cells observed on I1/I2-KO pMBMECs (Vid-
eos 2, 5). As a consequence, T cell interactions with VCAM-1
failed to establish stable adhesions over time, and a significantly
greater percentage of T cells (30.8 6 9.1%) detached from the
VCAM-1–coated surface compared with ICAM-1– and ICAM-2–
Taken together, our data clearly demonstrate that T cell polar-
ization and crawling under physiological shear are supported by
endothelial ICAM-1 and, in its absence, by endothelial ICAM-2.
In contrast, although VCAM-1 supports repeated events of shear-
resistant T cell arrest under physiological flow, it does not medi-
ate T cell polarization and crawling; thus, it establishes less stable
adhesive interactions with pMBMECs over time.
ICAM-1, but not ICAM-2, supports T cell crawling against the
direction of flow on BBB endothelium
While observing T cell crawling on wt pMBMECs, it became
obvious that the direction of crawling on the endothelium was not
random. Rather, T cell crawling paths were predominantly oriented
against the direction of flow (Fig. 6A), as recently observed for
T cell interaction with meningeal blood vessels in vivo (2). To
verify directed T cell crawling against the shear flow, we calcu-
lated a forward migration index toward the x-axis (xFMI), which
was along the direction of flow in the FOV. On wt pMBMECs, the
xFMI of T cells was found to be 0.13 6 0.03, which differed
significantly from 0, thereby proving T cell crawling against the
direction of flow (Fig. 6B). Control experiments showed that, in
the absence of shear flow, T cell crawling on wt pMBMECs oc-
curred randomly in all directions (xFMI: 0.06 6 0.03). Thus, flow
conditions specifically induced T cell crawling on wt pMBMECs
against the direction of shear forces.
and ICAM-2 to directed T cell crawling on BBB endothelium,
we compared xFMI values for T cell paths tracked on wt, I1-KO,
I2-KO, and I1/I2-KO pMBMECs (Fig. 6A, 6B). For I1-KO
pMBMECs, the xFMI was 0.0 6 0.04, demonstrating that endo-
thelial ICAM-1 is essential for the directed crawling of T cells
against the flow on the inflamed BBB. This is further supported by
the observation that I2-KO pMBMECs fully supported directed
T cell crawling against the flow, comparable to wt pMBMECs,
with the xFMI being 0.13 6 0.04. T cells failed to polarize and
crawl on I1/I2-KO pMBMECs. Rather, their behavior resembled
recurrent arrest and detachment events. Nevertheless, to demon-
1, support T cell crawling under flow. Analysis of
T cell interaction with immobilized recombinant
mICAM-1, mICAM-2, or mVCAM-1 is shown. A,
Representative images of crawling T cells on ICAM-
1 or ICAM-2 or of recurrent arresting T cells on
VCAM-1. Arrows indicate direction of shear flow
(original magnification 3100). Insets show enlarged
images of individual T cells marked with an asterisk
(scale bar, 10 mm). B, Analysis of dynamic T cell
1. Continuous crawling represents T cells that polar-
ized and crawled. Stationary represents T cells that
remained stationary (less than two T cell diameters of
movement from their site of arrest). Detachment
represents T cells that detached during the 15 min of
recording under physiological shear. Recurrent arrest
represents T cells that dislocated from the site of ar-
rest in the absence of polarization and crawling in
a fashion resembling multiple detachment and reat-
tachment cycles. Data are mean 6 SEM from three
independent experiments for each group. pp , 0.05,
Tukey multiple-comparison test.
The Journal of Immunology4851
strate the failure of T cells to withstand shear forces in the absence
of endothelial ICAM-1 and ICAM-2, we calculated the xFMI
for these T cell interactions on I1/I2-KO pMBMECs. As expected,
the xFMI for the direction of T cell movement on I1/I2-KO
pMBMECs was negative (20.43 6 0.07).
Obviously, therefore, crawling of T cells on the surface of wt
pMBMECs was mediated by endothelial ICAM-1 and ICAM-2.
Although only endothelial ICAM-1 allowed for directed crawl-
ing against shear forces on BBB endothelium, perpendicular crawl-
ing of T cells on I1-KO pMBMECs was dependent on the presence
of ICAM-2. To investigatewhether T cell crawling velocities onI1-
KO and I2-KO pMBMECs would also differ, we calculated mean
velocities of T cells crawling on wt, I1-KO, and I2-KO pMBMECs.
Because T cells failed to crawl on I1/I2-KO pMBMECs, we had to
omit them from this analysis. Although the average T cell crawling
velocity on wt pMBMECs was 4.0 6 0.1 mm/min, we observed
a significantly reduced T cell crawling velocity of 2.8 6 0.1 mm/
min in the absence of endothelial ICAM-1 on I1-KO pMBMECs
(Fig. 6C). The velocity of T cell crawling on I2-KO pMBMECs
remained unchanged compared with wt pMBMECs (Fig. 6C).
Thus, T cell crawling velocity on BBB endothelium is defined by
Purified immobilized ICAM-1 and ICAM-2, but not VCAM-1,
support T cell crawling against the direction of flow
a function of the endothelium or the T cell itself, we manually
tracked T cells duringtheir interaction withimmobilized rICAM-1,
rICAM-2, and rVCAM-1 (Fig. 7A). As the result of a different
experimental set-up usedforthisparticular analysis,shear flowwas
oriented opposite to the set-up described above for studying T
cell interaction with pMBMECs; thus, a negative xFMI describes
a predominant movement toward shear flow. As expected, in the
absence of shear flow, postarrest T cell crawling tracks on immo-
bilized ICAM-1 lacked any directionality (xFMI: 0.02 6 0.04).
When applying physiological flow, T cells crawled against the
direction of flow on immobilized ICAM-1 and, surprisingly, in a
similar manner on immobilized ICAM-2 (xFMI: 20.22 6 0.03
and 20.26 6 0.04, respectively) (Fig. 7A, 7B). In contrast, on
immobilized VCAM-1, T cells exhibited recurrent arrest move-
ments exclusively along the direction of flow (xFMI: 0.67 6 0.02;
Fig. 7A, 7B).
Because we determined that a significantly higher percentage of
(Fig. 5B), we speculated that the avidity of T cell interaction with
immobilized ICAM-1 is higher than with immobilized VCAM-1.
To further investigate this, we analyzed T cell detachment from
immobilized ICAM-1 or VCAM-1 under enhanced shear forces.
Stepwise enhancement of shear forces from 1.5–9.0 dyn/cm2min-
imally influenced the number of T cells adhering to immobilized
ICAM-1, with 90 6 10% of initially arrested T cells remaining in
the FOV with shear forces of 9 dyn/cm2(Supplemental Fig. 4B).
In contrast, on immobilized VCAM-1, T cells readily started to
detach in parallel with the enhanced shear forces, with only 38 6
5% of initially arrested T cells remaining attached at shear forces
of 9.0 dyn/cm2. Taken together, immobilized ICAM-1 and ICAM-
2, but not VCAM-1, were able to support directional T cell
crawling against the flow. The failure of VCAM-1 to support T
cell polarization and crawling correlated with a reduced avidity
of T cell interactions on immobilized VCAM-1 compared with
To address whether crawling velocity is influenced by shear and
differentially supported by ICAM-1 or ICAM-2, we calculated the
mean velocities of T cells crawling on immobilized ICAM-1 and
ICAM-2 (Fig. 7C). Interestingly, we found equal crawling ve-
locities on rICAM-1, independent of the presence (8.5 6 0.15 mm/
min) or absence (7.9 6 0.17 mm/min) of physiological shear. Un-
expectedly, on immobilized ICAM-2, T cells crawled with an
equal velocity (8.3 6 0.22 mm/min) compared with on immobi-
lized ICAM-1 (Fig. 7C).
In conclusion, the reduced T cell crawling velocity and lack of
preferentially directed T cell crawling against the flow observed on
I1-KO pMBMECs is not due to the intrinsic inability of ICAM-2 to
support these features; rather, it seems to depend on the different
accessibility of endothelial ICAM-2 in its physiological context on
the pMBMEC surface membrane.
In the current study, we dissected the individual contributions of
brain endothelial ICAM-1, ICAM-2, and VCAM-1 in mediating
shear-resistant T cell arrest, polarization, and directed crawling
on BBB endothelium. We found that endothelial ICAM-1 and
VCAM-1, but not ICAM-2, mediated shear-resistant T cell arrest
on the BBB in vitro. Subsequent polarization and crawling of
directed T cell crawling and crawling velocity on BBB endothelium under
flow. Evaluations of T cell tracks on TNF-a–stimulated wt, I1-KO, I2-KO,
and I1/I2-KO pMBMECs are shown. A, x/y diagrams of T cell tracks for
one representative experiment in each group. T cell tracks were divided into
two groups: T cells that underwent diapedesis (upper panels) and T cells
that remained on the endothelium without diapedesis (lower panels). For
each track, the site of arrest was set to the center point of the respective
diagram. End points of tracks are indicated by dots. B, Directionality of T
Dxis straight x-axis distance covered by the T cell and Daccis accumulated
flow was along the x-axis from plus to minus. Therefore, a positive xFMI
value represents a directed crawling against the orientation of shear flow,
whereas a negative xFMI value indicates movement along the shear forces.
xFMIwascalculatedfromTcell tracksofthreeindependent videospertype
of pMBMEC. Data are mean 6 SEM. pp , 0.05, compared with random
crawling (xFMI = 0); one-sample t test. C, Mean T cell crawling velocities
(micrometers per minute) were calculated from three independent videos
per type of pMBMEC (wt, I1-KO, I2-KO). Each data point represents the
velocity of one T cell. pppp , 0.001, one-way ANOVA, followed by the
Tukey multiple-comparison test.
Differential roles of endothelial ICAM-1 and ICAM-2 in
4852 ROLE OF ICAM-1 AND -2 AND VCAM-1 IN T CELL–BBB INTERACTIONS
T cells to sites permissive for diapedesis across the BBB endo-
thelium are mediated by endothelial ICAM-1 and ICAM-2; the
resistant crawling on pMBMECs. Although immobilized purified
ICAM-1 and ICAM-2 supported T cell crawling against the di-
rection of flow, the presence of endothelial ICAM-1 on the cell
surface of pMBMECs was strictly required for directed T cell
crawling. We observed rare diapedesis events of noncrawling T
cells across I1/I2-KO pMBMECs, suggesting that this particular
step can occur independently of endothelial ICAM-1 and ICAM-2.
In previous studies using brain-derived endothelioma cell lines,
we showed that both types of molecular interactions (a4-integrins
with endothelial VCAM-1 and LFA-1 with endothelial ICAM-1
and ICAM-2) mediated the adhesion of encephalitogenic T cells
to brain endothelium under static conditions in vitro. Although
a4–VCAM-1 interactions did not contribute to T cell diapedesis
across brain endothelium, endothelial ICAM-1 and ICAM-2 were
essentially required for T cell diapedesis across brain endothelium
under static conditions (13–16). In this study, we modeled the
BBB using pMBMECs that maintain barrier characteristics and
the adhesion molecule expression profile of BBB endothelium in
an in vitro setting (9, 17, 24). Using wt pMBMECs in static ad-
hesion and diapedesis experimental set-ups similar to our previous
studies, we showed a .2-fold increase in T cell adhesion and
diapedesis when the pMBMECs were stimulated with TNF-a
compared with unstimulated pMBMECs. In line with our previous
studies, we confirmed the essential role of endothelial ICAM-1
and ICAM-2 for T cell adhesion to and diapedesis across unsti-
mulated and stimulated pMBMECs under static conditions.
It has now been recognized that investigation of T cell inter-
actions with the endothelium requires live cell imaging tools to
detect postarrest T cell dynamicson the endothelium, suchas T cell
polarization and crawling on the endothelium to permissive sites of
diapedesis (5, 25). Moreover, accumulating data show the essential
stimulation and, thereby, strengthening of adhesive contacts be-
tween T cell-expressed integrins and their respective endothelial
ligands by shear forces physiologically exerted by the blood flow
(23). To elucidate these recent findings in more detail, we per-
formed live cell-imaging experiments under shear to delineate the
individual contributions of endothelial VCAM-1, ICAM-1, and
ICAM-2 to the distinct steps of T cell extravasation across the
inflamed BBB in vitro.
For all experiments, we used a murine PLP-specific CD4+Th1
effector/memory T cell line that has been well characterized for
its high predisposition to interact with the BBB; in vivo, these
T cells migrate across the BBB and induce experimental auto-
immune encephalomyelitis (EAE) when injected into susceptible
mouse strains (12, 13, 26). As described in this study, these spe-
cific highly active T cells are able to interact with the extracellular
domain of purified mouse ICAM-1, ICAM-2, or VCAM-1 without
coadsorbing any chemokine. This behavior stands in apparent
contrast to in vitro studies on the interaction of human T cells with
immobilized ICAM-1. These studies were most often executed
using freshly prepared peripheral blood T cells representing a
heterogeneous population of CD4+and CD8+T cells (5, 6, 27).
Obviously, human peripheral blood T cells require the presence of
a chemokine coimmobilized with ICAM-1 for efficient adhesion
and crawling. Whether this apparent difference in T cell behavior
is a consequence of different T cell activation status or an intrinsic
characteristic of the T cell subtype used remains to be elucidated.
By a thorough comparison of the dynamic interactions of en-
or I1/I2-KO BBB endothelium, we demonstrated that endothelial
ICAM-1, but not ICAM-2, mediates shear-resistant T cell arrest
on the BBB in vitro. Endothelial VCAM-1 serves as an alternative
ligand for T cell arrest on the inflamed BBB endothelium; blocking
as blocking of VCAM-1 in the absence of ICAM-1 on I1-KO or
I1/I2-KO pMBMECs completely abrogated T cell arrest. The spe-
cific roles of ICAM-1 and VCAM-1 in mediating T cell arrest were
further substantiated by analyzing T cell arrest on immobilized
purified ICAM-1, ICAM-2, and VCAM-1.
An important role for VCAM-1 in T cell arrest on the inflamed
BBB in vivo was suggested by a number of previous studies. We
recently demonstrated that T cells depend on b1-integrins to firmly
arrest within the inflamed spinal cord microvasculature during
EAE and, therefore, b1-integrin–deficient T cells fail to invade
the CNS (28). Similarly, the therapeutic success of the anti–a4-
integrin Ab natalizumab in the treatment of MS relies on the in-
hibition of a4-integrin–mediated arrest of human T cells on the
ICAM-1 and ICAM-2 but not on VCAM-1. Evaluation of T cell tracks on
immobilized rICAM-1, rICAM-2, or rVCAM-1 under physiological shear.
Note that because of a different experimental set-up, shear flow direction is
opposite to the direction of flow shown in Fig. 6. A, x/y diagrams of T cell
tracks are depicted for one representative experiment in each group. For
each track, the site of arrest was set to the center point of the respective
diagram. End points of tracks are indicated by dots. B, Directionality of
T cell movements under shear flow is expressed as xFMI. In this particular
experimental set-up, a negative xFMI demonstrates predominant movement
against the direction of flow, and a positive xFMI describes a direction of
movement alongthe shear flow. ThexFMIwas calculatedfrom T cell tracks
of three independent videos each. Data are mean 6 SEM. pp , 0.05,
compared with random crawling (xFMI = 0); one-sample t test. C, Mean
T cell crawling velocities (micrometers per minute) were calculated from
three independent movies each. Each data point represents the velocity of
one T cell.
T cells crawl against the direction of flow on immobilized
The Journal of Immunology4853
inflamed BBB in vivo (29, 30). Our present study demonstrates an
additional significant involvement of endothelial ICAM-1 in me-
diating T cell arrest on the inflamed BBB, which suggests that
efficient inhibition of T cell arrest on the inflamed BBB in vivo
requires blocking of a4b1-integrin–VCAM-1 and LFA-1–ICAM-
1 interactions. This might explain the finding that, unlike in other
animal models of EAE, blocking a4-integrins was not sufficient to
inhibit actively induced EAE in the C57BL/6 mouse (31). Indeed,
only the simultaneous blockade of LFA-1 and a4-integrins, but
not blocking of either integrin alone, led to instantaneous de-
tachment of encephalitogenic T cells from the vascular walls of
meningeal CNS vessels during EAE (2).
Subsequent to shear-resistant T cell arrest, we observed rapid
polarization of T cells and their crawling preferentially against the
direction of flow on the inflamed BBB. Therefore, our in vitro
observations exactly mimic T cell interactions observed in men-
ingeal blood vessels in vivo during EAE: encephalitogenic T cells
were observed to crawl against the direction of blood flow to sites
permissive for diapedesis (2). In our present study, we determined
that ICAM-1 and ICAM-2, but not VCAM-1, mediate T cell po-
larization and crawling on the endothelium. The role of endo-
thelial ICAM-2 in supporting T cell polarization and crawling
became obvious only in the absence of endothelial ICAM-1 on
pMBMECs or when studying T cell crawling on immobilized
purified ICAM-2. The absence of both ICAMs was required to
completely abrogate T cell polarization and crawling on the BBB
in vitro. The failure of VCAM-1 to mediate T cell polarization and
crawling was confirmed by studying T cell interactions with im-
mobilized purified VCAM-1. T cells seemed to continuously re-
peat the arrest phase, failed to establish high-avidity interactions,
and, therefore, were observed to be pushed along the direction of
flow. These findings are in line with our recent studies demon-
strating that a4b1-integrins mediate T cell arrest on inflamed
BBB microvessels in vivo during EAE (28, 29).
crawling on the BBB endothelium against the direction of flow.
Because T cells were able to crawl against the direction of flow on
immobilized purified ICAM-2, we speculate that this ability of
endothelial ICAM-2 is masked within the physiological context of
the endothelial cell surface membrane, probably because of its
shorter length compared with ICAM-1 or VCAM-1. Alternatively,
the unique ability of endothelial ICAM-1 to support directed T cell
crawling against shear on pMBMECs could be explained by
a lower molecular density of ICAM-2 compared with ICAM-1 on
the surface of the inflamed BBB. Although we showed that the
levels of ICAM-1 and ICAM-2 are similar in equally processed IF
images, a precise analysis of molecular density remains to be
T cell was described in in vitro studies performed under flow
conditions (5). Flow conditions exert shear forces on the attached
T cell, as well as on endothelial ICAM-1 and ICAM-2 bound to
their ligand LFA-1. LFA-1 exists in low-, intermediate-, and high-
affinity forms on the T cell surface; it is tightly regulated in its
spatial distribution, with intermediate LFA-1 at the leading edge
and high-affinity LFA-1 at the mid-zone or on adhesive filopodia
capable of invading the endothelial cell (5, 6). Therefore, a dif-
ferential spatial distribution of endothelial ICAMs on the BBB
during their interaction with the T cell could provide an expla-
nation for the different capacities of ICAM-1 and ICAM-2 to
support directed crawling. Several recent studies on T cell in-
teraction with HUVECs described enrichment of endothelial
ICAM-1, but not of ICAM-2, around T cells in the process of
diapedesis, irrespective of whether their route was transcellular or
paracellular (25, 32–34). Because ICAM-1 and ICAM-2 represent
the main endothelial ligands for T cell crawling, they must be able
to withstand traction forces exerted by the crawling motions of the
T cell; therefore, a tight anchorage of ICAM-1 and ICAM-2 to the
endothelial cytoskeleton would be reasonable. This assumption is
supported by our observation that T cells crawl efficiently against
the direction of flow on immobilized purified ICAM-1 and ICAM-2
that are laterally fixed and, in consequence, cannot be clustered.
Therefore, mechano-sensing of the direction of flow is an intrinsic
characteristic of the T cells, and it might be transduced by dif-
ferent signaling molecules intracellularly associated with the in-
dividual affinity conformations of LFA-1 (reviewed in Ref. 35).
In summary, our data demonstrate a continuous role for en-
dothelial ICAM-1 in mediating shear-resistant T cell arrest, polar-
ization, and directed crawling on the BBB, whereas VCAM-1
supports T cell arrest but is not involved in the other steps of
T cell extravasation. In the absence of endothelial ICAM-1, T cell
polarization and crawling on the BBB are mediated by ICAM-2.
Thus, blocking a single endothelial adhesion molecule is not
sufficient to completely inhibit T cell recruitment into the CNS.
We thank Mark Liebi for technical assistance.
The authors have no financial conflicts of interest.
1. Ley, K., C. Laudanna, M. I. Cybulsky, and S. Nourshargh. 2007. Getting to the
site of inflammation: the leukocyte adhesion cascade updated. Nat. Rev. Immu-
nol. 7: 678–689.
2. Bartholomaus, I., N. Kawakami, F. Odoardi, C. Schlager, D. Miljkovic, J. W.
Ellwart, W. E. Klinkert, C. Flugel-Koch, T. B. Issekutz, H. Wekerle, and
A. Flugel. 2009. Effector T cell interactions with meningeal vascular structures
in nascent autoimmune CNS lesions. Nature 462: 94–98.
3. Phillipson, M., B. Heit, S. A. Parsons, B. Petri, S. C. Mullaly, P. Colarusso,
R. M. Gower, G. Neely, S. I. Simon, and P. Kubes. 2009. Vav1 is essential for
mechanotactic crawling and migration of neutrophils out of the inflamed mi-
crovasculature. J. Immunol. 182: 6870–6878.
4. Luster, A. D., R. Alon, and U. H. von Andrian. 2005. Immune cell migration in
inflammation: present and future therapeutic targets. Nat. Immunol. 6: 1182–
5. Shulman, Z., V. Shinder, E. Klein, V. Grabovsky, O. Yeger, E. Geron,
A. Montresor, M. Bolomini-Vittori, S. W. Feigelson, T. Kirchhausen, et al. 2009.
Lymphocyte crawling and transendothelial migration require chemokine trig-
gering of high-affinity LFA-1 integrin. Immunity 30: 384–396.
6. Stanley, P., A. Smith, A. McDowall, A. Nicol, D. Zicha, and N. Hogg. 2008.
Intermediate-affinity LFA-1 binds alpha-actinin-1 to control migration at the
leading edge of the T cell. EMBO J. 27: 62–75.
7. Abbott, N. J., A. A. Patabendige, D. E. Dolman, S. R. Yusof, and D. J. Begley.
2010. Structure and function of the blood-brain barrier. Neurobiol. Dis. 37:
8. Wolburg, H., and A. Lippoldt. 2002. Tight junctions of the blood-brain barrier:
development, composition and regulation. Vascul. Pharmacol. 38: 323–337.
9. Lyck, R., N. Ruderisch, A. G. Moll, O. Steiner, C. D. Cohen, B. Engelhardt,
V. Makrides, and F. Verrey. 2009. Culture-induced changes in blood-brain barrier
transcriptome: implications for amino-acid transporters in vivo. J. Cereb. Blood
Flow Metab. 29: 1491–1502.
10. Wekerle, H., B. Engelhardt, W. Risau, and R. Meyermann. 1990. Passage
of lymphocytes across the blood-brain barrier in health and disease. In In
Pathophysiology of the Blood-Brain Barrier. B. B. Johansson, C. Owman, and
H. Widner, eds. Elsevier p. 439–445.
11. Engelhardt, B. 2008. Immune cell entry into the central nervous system: in-
volvement of adhesion molecules and chemokines. J. Neurol. Sci. 274: 23–26.
12. Vajkoczy, P., M. Laschinger, and B. Engelhardt. 2001. Alpha4-integrin-VCAM-1
binding mediates G protein-independent capture of encephalitogenic T cell
blasts to CNS white matter microvessels. J. Clin. Invest. 108: 557–565.
13. Laschinger, M., and B. Engelhardt. 2000. Interaction of alpha4-integrin with
VCAM-1 is involved in adhesion of encephalitogenic T cell blasts to brain en-
dothelium but not in their transendothelial migration in vitro. J. Neuroimmunol.
14. Reiss, Y., and B. Engelhardt. 1999. T cell interaction with ICAM-1-deficient
endothelium in vitro: transendothelial migration of different T cell populations is
mediated by endothelial ICAM-1 and ICAM-2. Int. Immunol. 11: 1527–1539.
15. Lyck, R., Y. Reiss, N. Gerwin, J. Greenwood, P. Adamson, and B. Engelhardt.
4854ROLE OF ICAM-1 AND -2 AND VCAM-1 IN T CELL–BBB INTERACTIONS
endothelium in vitro: the cytoplasmic tail of endothelial ICAM-1 is necessary for
transendothelial migration of T cells. Blood 102: 3675–3683.
16. Reiss, Y., G. Hoch, U. Deutsch, and B. Engelhardt. 1998. T cell interaction with
ICAM-1-deficient endothelium in vitro: essential role for ICAM-1 and ICAM-2
in transendothelial migration of T cells. Eur. J. Immunol. 28: 3086–3099.
17. Coisne, C., L. Dehouck, C. Faveeuw, Y. Delplace, F. Miller, C. Landry,
C. Morissette, L. Fenart, R. Cecchelli, P. Tremblay, and B. Dehouck. 2005.
Mouse syngenic in vitro blood-brain barrier model: a new tool to examine in-
flammatory events in cerebral endothelium. Lab. Invest. 85: 734–746.
18. Gerwin, N., J. A. Gonzalo, C. Lloyd, A. J. Coyle, Y. Reiss, N. Banu, B. Wang,
H. Xu, H. Avraham, B. Engelhardt, et al. 1999. Prolonged eosinophil accumu-
lation in allergic lung interstitium of ICAM-2 deficient mice results in extended
hyperresponsiveness. Immunity 10: 9–19.
19. Xu, H., J. A. Gonzalo, Y. St Pierre, I. R. Williams, T. S. Kupper, R. S. Cotran,
T. A. Springer, and J. C. Gutierrez-Ramos. 1994. Leukocytosis and resistance to
septic shock in intercellular adhesion molecule 1-deficient mice. J. Exp. Med.
20. Ro ¨hnelt, R. K., G. Hoch, Y. Reiss, and B. Engelhardt. 1997. Immunosurveillance
modelled in vitro: naive and memory T cells spontaneously migrate across
unstimulated microvascular endothelium. Int. Immunol. 9: 435–450.
21. Stein, J. V., S. F. Soriano, C. M’rini, C. Nombela-Arrieta, G. G. de Buitrago,
J. M. Rodrı ´guez-Frade, M. Mellado, J. P. Girard, and C. Martı ´nez-A. 2003.
CCR7-mediated physiological lymphocyte homing involves activation of a ty-
rosine kinase pathway. Blood 101: 38–44.
22. Lawrence, M. B., C. W. Smith, S. G. Eskin, and L. V. McIntire. 1990. Effect of
venous shear stress on CD18-mediated neutrophil adhesion to cultured endo-
thelium. Blood 75: 227–237.
23. Alon, R., and M. L. Dustin. 2007. Force as a facilitator of integrin conforma-
tional changes during leukocyte arrest on blood vessels and antigen-
presenting cells. Immunity 26: 17–27.
24. Coisne, C., C. Faveeuw, Y. Delplace, L. Dehouck, F. Miller, R. Cecchelli, and
B. Dehouck. 2006. Differential expression of selectins by mouse brain capillary
endothelial cells in vitro in response to distinct inflammatory stimuli. Neurosci.
Lett. 392: 216–220.
25. Carman, C. V., P. T. Sage, T. E. Sciuto, M. A. de la Fuente, R. S. Geha,
H. D. Ochs, H. F. Dvorak, A. M. Dvorak, and T. A. Springer. 2007. Transcellular
diapedesis is initiated by invasive podosomes. Immunity 26: 784–797.
26. Engelhardt, B., M. Laschinger, M. Schulz, U. Samulowitz, D. Vestweber, and
G. Hoch. 1998. The development of experimental autoimmune encephalomy-
elitis in the mouse requires alpha4-integrin but not alpha4beta7-integrin. J. Clin.
Invest. 102: 2096–2105.
27. Takesono, A., S. J. Heasman, B. Wojciak-Stothard, R. Garg, and A. J. Ridley.
2010. Microtubules regulate migratory polarity through Rho/ROCK signaling in
T cells. PLoS ONE 5: e8774.
28. Bauer, M., C. Brakebusch, C. Coisne, M. Sixt, H. Wekerle, B. Engelhardt, and
R. Fa ¨ssler. 2009. Beta1 integrins differentially control extravasation of inflam-
matory cell subsets into the CNS during autoimmunity. Proc. Natl. Acad. Sci.
USA 106: 1920–1925.
29. Coisne, C., W. Mao, and B. Engelhardt. 2009. Cutting edge: Natalizumab blocks
adhesion but not initial contact of human T cells to the blood-brain barrier
in vivo in an animal model of multiple sclerosis. J. Immunol. 182: 5909–5913.
30. Engelhardt, B., and L. Kappos. 2008. Natalizumab: targeting alpha4-integrins in
multiple sclerosis. Neurodegener. Dis. 5: 16–22.
31. Kerfoot, S. M., M. U. Norman, B. M. Lapointe, C. S. Bonder, L. Zbytnuik, and
P. Kubes. 2006. Reevaluation of P-selectin and alpha 4 integrin as targets for the
treatment of experimental autoimmune encephalomyelitis. J. Immunol. 176:
32. Milla ´n, J., L. Hewlett, M. Glyn, D. Toomre, P. Clark, and A. J. Ridley. 2006.
Lymphocyte transcellular migration occurs through recruitment of endothelial
ICAM-1 to caveola- and F-actin-rich domains. Nat. Cell Biol. 8: 113–123.
33. Carman, C. V., and T. A. Springer. 2004. A transmigratory cup in leukocyte
diapedesis both through individual vascular endothelial cells and between them.
J. Cell Biol. 167: 377–388.
34. Barreiro, O., M. Yanez-Mo, J. M. Serrador, M. C. Montoya, M. Vicente-
Manzanares, R. Tejedor, H. Furthmayr, and F. Sanchez-Madrid. 2002. Dynamic
interaction of VCAM-1 and ICAM-1 with moesin and ezrin in a novel endo-
thelial docking structure for adherent leukocytes. J. Cell Biol. 157: 1233–1245.
35. Smith, A., P. Stanley, K. Jones, L. Svensson, A. McDowall, and N. Hogg. 2007.
The role of the integrin LFA-1 in T-lymphocyte migration. Immunol. Rev. 218:
36. Cecchelli, R., B. Dehouck, L. Descamps, L. Fenart, V. V. Bue ´e-Scherrer, V,
C. Duhem, S. Lundquist, M. Rentfel, G. Torpier, and M. P. Dehouck. 1999.
In vitro model for evaluating drug transport across the blood-brain barrier. Adv.
Drug Deliv. Rev. 36: 165–178.
The Journal of Immunology 4855