lymphocyte transcellular migration occurs through

ART ICLES

NATURE CELL BIOLOGY VOLUME 8 | NUMBER 2 | FEBRUARY 2006 113

Lymphocyte transcellular migration occurs through recruitment of endothelial ICAM-1 to caveola- and F-actin-rich domainsJaime Millán1,2, Lindsay Hewlett3, Matthew Glyn4, Derek Toomre5, Peter Clark4 and Anne J. Ridley1,2,6

During inflammation, leukocytes bind to the adhesion receptors ICAM-1 and VCAM-1 on the endothelial surface before undergoing transendothelial migration, also called diapedesis. ICAM-1 is also involved in transendothelial migration, independently of its role in adhesion, but the molecular basis of this function is poorly understood. Here we demonstrate that, following clustering, apical ICAM-1 translocated to caveolin-rich membrane domains close to the ends of actin stress fibres. In these F-actin-rich areas, ICAM-1 was internalized and transcytosed to the basal plasma membrane through caveolae. Human T-lymphocytes extended pseudopodia into endothelial cells in caveolin- and F-actin-enriched areas, induced local translocation of ICAM-1 and caveolin-1 to the endothelial basal membrane and transmigrated through transcellular passages formed by a ring of F-actin and caveolae. Reduction of caveolin-1 levels using RNA interference (RNAi) specifically decreased lymphocyte transcellular transmigration. We propose that the translocation of ICAM-1 to caveola- and F-actin-rich domains links the sequential steps of lymphocyte adhesion and transendothelial migration and facilitates lymphocyte migration through endothelial cells from capillaries into surrounding tissue.

Tissue inflammation induces the recruitment of leukocytes from neighbouring blood vessels. A key event in this process is leukocyte extravasation across the endothelium — a multistep process involving a cascade of successive interactions between leukocyte and endothelial adhesion receptors1. The current paradigm of leukocyte transendothe-lial migration proposes that leukocytes first interact with the adhesion molecules E- and P-selectin and then with VCAM-1 and ICAM-1 on the apical endothelial plasma membrane. They subsequently crawl to cell–cell junctions where other receptors (PECAM-1, JAM1 or CD99) facilitate transmigration between cells2. However, there is some evidence that neutrophil adhesion occurs near favoured domains of diapedesis at endothelial cell corners3,4. In addition, the use of blocking antibod-ies suggests that ICAM-1 functions in transendothelial migration as well as adhesion5 and may guide leukocytes to sites of transmigration6. Although no mechanistic explanation for these observations has yet been found, taken together, they suggest that the apical endothelial plasma membrane is organized to facilitate or even orientate arrested leukocytes to optimal sites for transendothelial migration.

Leukocytes are normally thought to transmigrate through endothe-lial cell–cell junctions7. However, in vivo electron microscopy studies have shown that leukocytes can also follow a transcellular route by

opening endothelial transcellular channels that are formed by association of numerous non-coated vesicles — resembling the so-called vesiculo-vacuolar organelles — situated close to the cell borders8. Recent evidence indicates that some leukocytes also take a transcellular route in vitro9. As vesiculo-vacuolar organelles mediate the transendothelial passage of macromolecules and have been proposed to form by fusion of multiple caveolae, it has been suggested that leukocytes induce remodelling of endothelial caveolae or vesiculo-vacuolar organelles to form a channel across the cell10. Caveolae are non-coated vesicles that mediate the trans-port of small molecules (such as albumin) and some viruses (such as SV40) and they are formed from plasma membrane domains called ‘lipid rafts’ that are enriched in cholesterol and glycolipids. The localization of these vesicles is regulated by F-actin11 and caveolin-1, the scaffolding protein of caveolae, regulates many signalling pathways12.

Here we report that ICAM-1 translocated to F-actin- and caveolin-1-rich regions close to endothelial cell–cell borders when it was engaged, and was subsequently transcytosed via caveolae. When T lymphoblasts bound to endothelial cells, they extended pseudopodia down to the basal endothelial membrane and established a transcellular pathway through caveolin- and F-actin-enriched channels. Reducing caveolin-1 levels by RNAi specifically decreased transcellular but not paracellular

1Ludwig Institute for Cancer Research, Royal Free and University College School of Medicine, 91 Riding House Street, London W1W 7BS, UK. 2Department of Biochemistry and Molecular Biology and 3MRC Laboratory for Molecular Cell Biology, University College London, Gower Street, London WC1E 6BT, UK. 4Leukocyte Biology, National Heart and Lung Institute, Sir Alexander Fleming Building, Imperial College London SW7 2AZ, UK. 5Ludwig Institute for Cancer Research, Department of Cell Biology, Yale University School of Medicine, 333 Cedar St., PO Box 208002, New Haven, 06520–8002 CT, USA.6Correspondence should be addressed to A.J.R. (e-mail: [email protected])

Received 28 September 2005; accepted 28 November 2005; published online 22 January 2006; DOI: 10.1038/ncb1356

print ncb1356.indd 113print ncb1356.indd 113 17/1/06 1:20:18 pm17/1/06 1:20:18 pm

Nature Publishing Group ©2006


114 NATURE CELL BIOLOGY VOLUME 8 | NUMBER 2 | FEBRUARY 2006

A RT I C L E S

transendothelial migration of T lymphoblasts, suggesting that ICAM-1 translocation to caveolae regulated the transcellular route of transen-dothelial migration.

RESULTSICAM-1 translocates to F-actin- and caveolin-rich regions at endothelial cell poles upon engagementICAM-1 and VCAM-1 on endothelial cells can localize to F-actin-rich ‘docking structures’ around adherent leukocytes13,14. As ICAM-1 induces actin stress fibre assembly in endothelial cells15,16, we analysed how ICAM-1 engagement affected its localization in relation to the actin network. Human umbilical vein endothelial cells (HUVECs) stimulated with tumour necrosis factor (TNF)-α are elongated and have stress fibres that are organized into apical and basolateral filaments that run almost parallel to the elongated axis of the cell, converging at cell poles17 (Fig. 1a, data not shown). ICAM-1 is normally localized to microvilli and cell–cell junctions15 (Fig. 1a). Antibody ligation induced a marked enrichment of ICAM-1 in areas where stress fibres converged, often at poles of elon-gated cells and close to (but not at) endothelial cell–cell junctions (Fig. 1a, b, arrows; see Supplementary Information, Fig. S1b). A fraction (<20%) of ICAM-1 localized on vesicle-like structures in the perinuclear region (Fig. 1a, arrowhead). Similarly, antibody-crosslinked VCAM-1 localized to F-actin-rich regions, although polar segregation was not as marked as for ICAM-1 (Fig. 1b; see Supplementary Information, Fig. S1a). In con-trast, E-selectin did not colocalize with F-actin and was not segregated to the cell periphery, but had a perinuclear localization indicative of internalization (Fig. 1b; see Supplementary Information, Fig. S1a).

Interestingly, caveolae often localized close to endothelial cell–cell junctions18 (Fig. 1c and data not shown). Caveolin-1 was distributed primarily at the periphery of cells in areas where stress fibre bundles converged (Fig. 1c, arrows). Following ICAM-1 ligation, ICAM-1 local-ized with caveolin-1 at specific areas in HUVECs (see boxed areas in Fig. 1c), and also in human dermal microvascular endothelial cells (HDMVECs; see Supplementary Information, Fig. S1d), which express similar levels of caveolin-1 to HUVECs. In addition, caveolin-1 in these areas appeared more clustered, suggesting that ICAM-1 might be affecting the distribution of caveolae. In contrast, ICAM-1 clusters did not colocalize with transferrin or with glycosylphosphatidylinositol (GPI)-anchored CD59, which associates with lipid rafts but not with caveolae (see Supplementary Information, Fig. S1c). Antibody-clustered transferrin receptor (TfR) or vascular endothelial growth factor recep-tor (VEGFR) did not colocalize with caveolin-1 (see Supplementary Information, Fig. S1d), indicating that the localization of ICAM-1 with caveolin-1 is not a non-specific consequence of receptor clustering on the apical surface of endothelial cells.

To distinguish whether these ICAM-1 clusters were internalized or remained at the cell surface, a biotin-coupled anti-ICAM-1 antibody and tetramethylrhodamine isothiocyanate (TRITC)-coupled streptavi-din were used to crosslink ICAM-1. Similarly to a secondary antibody, streptavidin promoted ICAM-1 accumulation at the cell periphery (data not shown). Cells were then incubated at 4 °C to prevent membrane traf-ficking, and a secondary antibody was added to detect surface-clustered ICAM-1. Specific areas of ICAM-1 internalization were identified by comparing streptavidin staining (total clustered ICAM-1) with second-ary antibody staining (surface clustered receptor). ICAM-1 internaliza-tion occurred predominantly in caveolin-1- and F-actin-enriched areas

(see Supplementary Information, Fig. S2c). In agreement with these observations, ICAM-1 colocalized with caveolin-1 in membrane invagi-nations (Fig. 1d), vesicles and multi-vesicular complexes (Fig. 1e) as shown by immunoelectron microscopy. Similarly to ICAM-1, clustered VCAM-1 (but not E-selectin) partially colocalized with caveolin-1 (data not shown). Consistently, ICAM-1 and VCAM-1 (but not E-selectin) were rapidly recruited to detergent-resistant lipid rafts and cofraction-ated with endogenous caveolin-1 within 10 min of antibody crosslinking (see Supplementary Information, Fig. S2d–h).

ICAM-1 translocates to the ends of stress fibres and to the basal plasma membraneThe observation that ICAM-1 accumulated in F-actin- and caveolin-1-rich regions near cell poles following crosslinking suggested that it may

ICAM-1

Caveolin-1ICAM-1X-ICAM-1

X-I

CA

M-1

ICAM-1F-actin F-actin

F-actin

Merge Merge

Merge

Per

iphe

ry: c

entr

e ra

tio

201612840

ICAM

-1VC

AM-1

E-se

lect

in

Ap

ical

Bas

olat

eral

a

b

d e

c

Figure 1 Antibody-crosslinked ICAM-1 translocates to F-actin- and caveolin-rich areas at the endothelial cell periphery. ICAM-1 was crosslinked on confluent TNF-α-stimulated HUVECs by incubation at 37 °C with anti-ICAM-1 antibodies followed by FITC-labelled secondary antibodies. Cells were stained with TRITC-phalloidin to visualize F-actin. (a) Apical and basolateral confocal images of ICAM-1 before (–) and after (X-ICAM-1) 90 min of antibody clustering. (b) Quantification of receptor segregation at cell periphery (see Methods). Values represent mean ± s.e.m. (c) Basolateral image of HUVECs stained with anti-caveolin-1 antibody (green) and TRITC-phalloidin (blue) before and after crosslinking of ICAM-1 (red) for 90 min. Scale bars represent 20 µm. (d, e) ICAM-1 was crosslinked for 15 min, as in a–c, but using an anti-mouse secondary antibody conjugated to 10-nm gold particles. Cells were fixed, permeabilized and stained with anti-caveolin-1 antibody followed by anti-rabbit secondary antibody conjugated to 5-nm gold particles. Cells were then processed for electron microscopy. An invaginating caveola is shown in d. Scale bar represents 70 nm. A group of internal caveolae, some of them fused into a caveosome, is shown in e. Scale bar represents 100 nm. White arrowheads indicate 10 nm gold particles corresponding to ICAM-1 localization.





A RT I C L E S

be actively translocated on the cell surface. The movement of clustered ICAM-1 with respect to F-actin was therefore investigated using time-lapse microscopy in HUVECs transfected with β-actin–yellow fluores-cent protein (YFP; Fig. 2a; see Supplementary Information, Movie 1). To detect early events following ICAM-1 engagement, cells were initially incubated with secondary antibody at 4 °C to prevent any receptor move-ment before filming at 37 °C. Under these conditions, antibody-engaged ICAM-1 formed a wave that moved towards the cell pole, and clusters of ICAM-1 accumulated at the end of stress fibres (Fig. 2a).

Comparison of apical and basolateral images suggested that apical ICAM-1 not only accumulated at the cell periphery but also moved to a more basolateral plane (Fig. 1a; data not shown). To address the pos-sibility that ICAM-1 was translocated to the basal plasma membrane, we used combined time-lapse epifluorescence (EPI) and total internal reflection fluorescence microscopy (TIRFM). In TIRFM, a beam of light that undergoes total internal reflection near the coverslip–cell inter-face generates an ‘evanescent wave’ (<100-nm penetration depth) and selectively excites fluorophores on the lower cell surface19. TIRFM has been used to detect exocytosis of granules and vesicles19,20. As ICAM-1 translocated towards the cell pole TIRF was detected, indicating that it was less than 100 nm from the basal membrane (Fig. 2b, c, arrowheads; also see Supplementary Information, Movie 2). Other ICAM-1 clusters (open arrowhead) reached this basal plane at earlier times and moved towards the pole, suggesting that they were moving on, or close to, the basal membrane. Most TIRF-positive ICAM-1 remained at the basal membrane and was eventually delivered to internal vesicles between

15 min 30 min 45 min

45 min

66 min

10 min 20 min 30 min 40 min 50 min 60 min 70 min

68 min 70 min 71 min 73 min 75 min

60 min 75 min

75 min

90 min

EP

I−TI

RF

TIR

F

a

b

c

d

Figure 2 ICAM-1 crosslinking leads to transcytosis of ICAM-1 and caveolin-1. (a) HUVECs were transfected with pEYFP–actin (green) and after 36 h were stimulated with TNF-α for 15 h. ICAM-1 (red) was crosslinked with antibodies and its distribution followed for 90 min by time-lapse epifluorescence (EPI) microscopy. (b, c) TNF-α−stimulated HUVECs were incubated with anti-ICAM-1 antibody at 37 °C for 45 min followed by Cy2-coupled secondary antibody at 4 °C for 15 min. ICAM-1 movement was followed by simultaneous TIRF–EPI time-lapse microscopy at 37 °C. TIRF (green) only illuminates fluorophores near the basal cell surface whereas EPI (red) illuminates the entire cell. A general view of the monolayer is shown in b and details of ICAM-1 transcytosis are shown in c. Arrowheads show ICAM-1 clusters that are first detected by EPI and subsequently by TIRF(<100 nm from the basal membrane). Open arrowhead shows an ICAM-1 cluster that is already close to the basal membrane (detected by TIRF) as a control. The upper panels in c show the merged EPI and TIRF images and the bottom panel shows only the TIRF images represented as inverse black and white images. (d) TNF-α-stimulated HUVECs expressing caveolin-1–YFP were analysed by EPI time-lapse microscopy to follow the movement of caveolin-1–YFP (green) and antibody-clustered ICAM-1 (red). ICAM-1 is quickly recruited to a caveolin-1-rich area at the top of the images (arrowhead) whereas in the lower part of the images it forms a wave that is translocated (arrow). Time (min) after addition of the secondary crosslinking antibody is indicated (a–d). Scale bars represent 10 µm in a, b and d, and 5 µm in c.

a

b

c

Figure 3 ICAM-1 localizes to caveolae and is translocated to the basal plasma membrane. (a) TNF-α-stimulated HUVECs were incubated with anti-ICAM-1 antibody for 45 min at 37 °C, fixed and incubated with a gold (10 nm)-conjugated secondary antibody. Cells were then processed for electron microscopy. Note that non-crosslinked anti-ICAM-1 antibody cannot access the basolateral (lower) membrane. Arrowheads indicate caveola-like vesicles on apical and basal membranes. (b, c) ICAM-1 was crosslinked by incubation with anti-ICAM-1 antibody for 45 min followed by a gold-conjugated secondary antibody for 60 min at 37 °C. Arrowheads in b indicate caveola-like vesicles containing ICAM-1; the arrow in b shows ICAM-1 at the basolateral membrane. Detail of a group of caveolae containing ICAM-1 that appear to reach the basolateral membrane then fuse and release ICAM-1 is shown in c. Scale bars represent 200 nm in a and b, and 100 nm in c.





A RT I C L E S

460 s 380 s

380 s

600 s 840 s

480 s

480 s

560 s

560 s

640 s

640 s

840 s

520 s 540 s 740 s

480 s 500 s

500 s480 s

38 0 s

600 s 640 s 840 s

480 s 560 s

460 s

520 s 540 s 740 s

S R P S R P

TIR

F

TIR

F

Cav

eolin

-1IC

AM

-1

T cells−

EP

I−TI

RF

EP

I−TI

RF

EP

I−TI

RF

E

E

E

E

T

T

a

c

d

b

Figure 4 ICAM-1 and caveolin-1 are translocated to the basal membrane at endothelium–lymphocyte interaction sites. (a) Dual TIRF–EPI timelapse microscopy of HUVECs expressing ICAM-1–GFP and incubated with T-lymphoblasts. ICAM-1–GFP is recruited around the lymphocyte perimeter, as detected by EPI (red). Discrete small areas of ICAM-1–GFP are pushed down by the lymphocyte and detected by TIRFM close to the basal membrane (green in upper panels; indicated by arrowheads). At 740 s, the T-lymphoblast starts to move away from this region and the TIRF-positive areas begin to disappear. Arrows indicate background TIRFM detection due to ICAM-1–GFP localization close to, or on, the basolateral membrane. The discontinuous red line indicates the T-lymphoblast perimeter. Time after adding the T-lymphoblasts is indicated. (b) TIRF–EPI timelapse microscopy of HUVECs expressing caveolin–YFP and incubated with T-lymphoblasts. Bright-field images are shown to identify areas of contact between the endothelial cell and lymphocytes. Lower panels show an enlarged region of

the box marked in upper panel (560 s). Note the caveolin-1 ring detected by EPI approaches the basal membrane, as detected by TIRFM (bottom panel, inverse black on white). (c) TNF-α-stimulated HUVECs (1 × 106) were surface labelled with sulpho-NHS–biotin and incubated with or without 2 × 106 T-cells for 20 min. Cells were lysed and lipid rafts were isolated by selective solubilization with β-octyl-glucoside. ICAM-1 was immunoprecipitated, blotted and detected with peroxidase-conjugated streptavidin. Caveolin-1 was detected by western blotting. S, Triton X-100-soluble fraction; R, Triton X-100-resistant–β-octyl-glucoside-sensitive caveolin-1-rich fraction; P, pellet. (d) T-lymphoblasts were incubated with TNF-α-stimulated HUVECs for 20 min at 37 °C, fixed and processed for electron microscopy. T, T-lymphoblast; E, endothelial cell. Insert shows a two-fold magnification of the boxed area. Arrowheads indicate endothelial intercellular junctions. Arrows indicate endothelial cells. Scale bars represent 5 µm in a and b and 400 nm in c.





A RT I C L E S

the 2 and 3 h timepoints (data not shown). ICAM-1 trafficking to the basal membrane was dependent on crosslinking, as no TIRF signal was detected following incubation with a monomeric IgG (Fab) secondary antibody (data not shown). In contrast to ICAM-1, no TIRF signal was detected for clustered E-selectin, whesreas internalized dipyrromethene boron difluoride–low-density lipoprotein (BODIPY–LDL) was transcy-tosed as expected21 and detected by TIRF (data not shown).

ICAM-1 was similarly segregated and delivered to the basal plasma membrane in subconfluent cells, indicating that intercellular junc-tions were not necessary for ICAM-1 trafficking (see Supplementary Information, Fig. S3a and Movie 3). Confocal time-lapse microscopy was used to confirm the localization of clustered ICAM-1 near the ends of stress fibres in HUVECs expressing paxillin–green fluorescent pro-tein (GFP) as a focal adhesion marker (see Supplementary Information, Fig. S3b). Some overlap of ICAM-1 clusters with the rear of paxillin–GFP-labelled focal adhesions was detected. This suggests that ICAM-1 transcytosis occurs close to focal adhesions at the ends of stress fibres.

ICAM-1 is transcytosed via caveolae at the cell peripheryThe effect of ICAM-1 engagement on caveolin-1 distribution was assessed in live cells by transfection of caveolin-1–YFP into HUVECs. Caveolin-1–YFP showed a punctate and polarized distribution identi-cal to endogenous caveolin-1 (J.M., unpublished observations and ref. 22). Initially, ICAM-1 near caveolin-1–YFP-enriched areas was quickly clustered into these regions following antibody crosslinking (Fig. 2d, arrowhead; see Supplementary Information, Movie 4, top cell pole), whereas ICAM-1 situated in other areas of the cell surface was more slowly translocated to caveolin-1–YFP-rich areas (Fig. 2d, arrow; see Supplementary Information, Movie 4). Using combined TIRF–EPI microscopy, transcytosis of caveolin-1–YFP to basal areas at cell poles was detected in response to ICAM-1 crosslinking (data not shown).

To determine whether ICAM-1 was transcytosed via caveolae, we analysed the distribution of clustered ICAM-1 by electron microscopy. Without crosslinking, ICAM-1 was restricted to the apical membrane, with no labelling of internal vesicular structures or the basolateral mem-brane detected (Fig. 3a). In contrast, following sequential incubation with primary and gold-conjugated secondary antibodies, a significant fraction of ICAM-1 was detected on the basolateral membrane, in vesicular structures or basal membrane invaginations that were mor-phologically identical to caveolae (Table 1; Fig. 3b, c), or in aggregates of several caveolae (caveosomes), similar to those detected in Fig. 1e

(data not shown). Internalized ICAM-1 was also detected in lysosomes or multivesicular bodies, but little association with clathrin-coated pits and buds was observed (Table 1; data not shown).

T-lymphoblasts extend protrusions into endothelial cells and induce translocation of ICAM-1 and caveolin-1 To determine whether leukocytes affect the localization of ICAM-1 and caveolin-1, HUVECs expressing ICAM-1–GFP or caveolin–YFP were incubated with T-lymphoblasts and analysed by TIRF–EPI microscopy close to the endothelial cell borders. ICAM-1–GFP was initially enriched around attached lymphocytes (Fig. 4a). Lymphocytes then extended pro-trusions down towards the endothelial basal membrane and in these regions ICAM-1–GFP was detected by TIRF — indicating that it had moved onto or close to the basal membrane. Basal ICAM-1 disappeared once the lymphocyte had transmigrated (Fig. 4a). Ring-like caveolin-1 structures were also transiently detected under T-lymphoblasts by EPI and then TIRF microscopy in HUVECs expressing caveolin-1–YFP, suggesting that caveolae are reorganised around leukocyte protru-sions and redistributed to the basal plane (Fig. 4b). Consistent with this observation, T-lymphoblast binding induced recruitment of endothelial ICAM-1 to biochemically fractionated caveola-rich membrane domains (Fig. 4c). Furthermore, Fig. 4d shows an electron microscopy image of a T-lymphoblast extending a process into an endothelial cell in an area rich in caveolae. The presence of neighbouring intercellular junctions (arrowheads) suggests that the T-lymphoblast is initiating transmigra-tion away from the junctions.

Caveolin-1 and F-actin surround T-lymphoblasts following a transcellular route of diapedesisAs lymphocytes induced translocation of ICAM-1 and caveolin-1 to the basal endothelial membrane, we investigated the relationship between leukocytes, F-actin and caveolin-1 during transendothelial migration. Leukocytes have been reported to transmigrate across endothelial cells via either a paracellular or transcellular route9,23,24. We identified T-lym-phoblasts using a transcellular route through their localization away from intercellular junctions stained with anti-vascular endothelial (VE)–cadherin antibodies (Fig. 5a) or by staining the endothelial-cell surface with anti-ICAM-1 antibodies (Fig. 5b, c). We observed that 9.4 ± 1.8% of transmigrating T-lymphoblasts followed a transcellular route across TNF-α-stimulated HUVECs (Fig. 5d), and 30.6 ± 3.3% of T-lymphoblasts transmigrated transcellularly across HDMVECs (Fig. 6b). HDMVECs have been described as a mixture of blood and lymphatic microvascular endothelium25, and thus they were stained with anti-VEGFR3 or anti-lymphatic vessel endothelial hyaluronan receptor (LYVE-1) antibod-ies to distinguish blood from lymphatic HDMVECs25. Interestingly, no difference was found in the percentage of transcellular transendothelial migration between both cell types (data not shown), and thus lymphatic endothelial cells are also able to support the transcellular route.

Actively transmigrating lymphocytes (with a ‘squeezed’ morphology) were often surrounded by a ring of caveolin-1 and situated at the cell periphery in caveolin-1 and F-actin-rich regions (Fig. 5a, d; Fig. 6A, C). Caveolin-1 localized more frequently around transcytosing lymphocytes than lymphocytes taking a paracellular route (Fig. 5d). In contrast, other membrane proteins such as the TfR or VEGFR3 were not localized in the transcellular channels (Fig. 6A, C). Furthermore, neither the GPI-anchored protein Thy1 (a component of non-caveolar rafts that is

Table 1 Subcellular distribution of ICAM-1 after antibody-induced clustering

Cellular Location n = percentage of ICAM-1

Apical cell membrane 69.8 ± 4.4

Basolateral cell membrane 9.2 ± 2.9

Caveolae Total 10.7 ± 2.5

Apical 5.9 ± 1.3

Basolateral 2.34 ± 1.1

Internal 2.48 ± 1.4

Coated pits 0.3 ± 0.1

Multivesicular body 5.9 ± 2.1

Lysosomes 3.59 ± 2.5

The number of gold particles in 26 electron microscopy images from HUVECs with normal morphology were counted. For each image, the percentage of gold particles in each subcellular location was determined. No gold particles were found in the Golgi, endoplasmic reticulum or nucleus. Values represent mean ± s.e.m.





A RT I C L E S

Caveolin-1

Caveolin-1

Caveolin-1

Caveolin-1

F-actin

F-actin

F-actin

F-actin

ICAM-1

ICAM-1ICAM-1

VE−cadherin Merge

Merge

Rou

tes

of T

-lym

pho

bla

std

iap

edes

is (p

erce

ntag

e)C

aveo

lin-1

sur

roun

din

gT-

lym

pho

bla

sts

(per

cent

age)

Str

ess

fibre

s in

con

tact

with

tran

scel

lula

r p

assa

ges

(per

cent

age)

Merge

Merge

Apical

Ap

ical

Basal

Bas

al

100

80

80

60

60

40

40

20

20

0

100

80

60

40

20

0

0

Paracellular

ParacellularSpread Transcellular

Transcellular

+ −

Transcellularchannel

Endothelialintercellular junction





T-cell body(apical)

T-cell uropod(apical)

T-celluropod(apical)

T-cell leading edge(basal) T-cell leading edge

(basal)

T-cell lamellipodia(basal)

a, lower panels c, upper panelsb, upper panels

a c

b

d

e

Figure 5 Caveolin-1 and F-actin accumulate at transcellular passages of diapedesis. (a) T-lymphoblasts were incubated with TNF-α-stimulated HUVECs for 15 min at 37 °C. Cells were then fixed and F-actin (red), caveolin-1 (green) and VE–cadherin (blue) distribution was analysed by immunofluorescence. T-lymphoblasts were identified by phalloidin staining. Transmigrating lymphocytes were identified by taking confocal sections at different planes and by their characteristic transmigratory morphology (arrowheads). Upper panels: single projection of a z-stack of 15 confocal sections. Lower panels: two-fold magnification of selected confocal planes of one of the cells undergoing transcellular diapedesis. Note the caveolin-1 rings (arrows) are away from the cell border (indicated by VE–cadherin staining and the discontinuous line). (b) Selected z-stack confocal sections showing a T-lymphoblast following transcellular diapedesis and stained for caveolin-1 (green), ICAM-1 (blue) and F-actin (red). The basal section shows the lymphocyte lamellipodia spreading under the endothelial cell (F-actin staining) and the basal areas of the ICAM-1 and caveolin-1 transcellular passage (the discontinuous line indicates the endothelial cell perimeter); medial sections show the ICAM-1, F-actin and caveolin-1-enriched passage; the apical section shows

the microvillus-like docking structures devoid of caveolin-1. A rotated (85°) projection of this z-stack shows the T-lymphoblast (discontinous line) embedded in the endothelial cell (lower panels). (c) Projection of a confocal z-stack where a T-lymphoblast is undergoing transcellular diapedesis. Upper four panels are a general view of the transcellular passage (arrowhead). The T-lymphoblast perimeter is indicated in the merged image (discontinuous line). Lower four panels show an enlargement of the boxed area where the microvillus-like docking structures (arrows) are clearly separate from the caveolin-1-rich transcellular passage. (d) Upper graph: quantification of T-lymphoblast paracellular versus transcellular diapedesis across TNF-α-stimulated HUVECs (mean ± s.e.m. of five different experiments). Middle graph: frequency of localization of caveolin-1 around spread (non-transmigrating) T-lymphoblasts, or T-lymphoblasts undergoing paracellular or transcellular diapedesis (mean ± s.e.m. of five different experiments). Lower graph: percentage of transcellular passages in contact with stress fibres (mean ± s.e.m. of three different experiments). Scale bars represent 10 µm in a and c and 5 µm in b. (e) Schematic representation of cells at different stages of transcellular transmigration, as seen in the lower panels of a and the upper panels of b and c.





A RT I C L E S

Caveolin-1

Caveolin-1

Caveolin-1

Caveolin-1C

aveo

lin-1

VE−cadherin

F-actin

F-actin

F-actin

F-actin

F-actin

ICAM-1

PECAM-1

VEGFR3

TfR

Thy-1

Merge

Merge

Merge

Merge

Merge

Endothelialintercellularjunction

Transcellular channel

T-cell uropod(apical)

T-cell leadingedge (basal)

a

b

c

d

e

Rou

tes

of T

-lym

pho

bla

std

iap

edes

is (p

erce

ntag

e)

Tran

scel

lula

r ch

anne

ls(p

erce

ntag

e)

TranscellularParacellular

70

60

50

40

30

20

10

0

TfR

Thy1

VEG

FR3

A

B C80

60

40

20

0

Figure 6 Transcellular transendothelial migration in human microvascular endothelial cells. (a) T-lymphoblasts were incubated with TNF-α-stimulated HDMVEC for 12 min. Cells were then fixed, permeabilized and the distributions of different proteins analysed by immunofluorescence. T-lymphoblasts were identified by TRITC–phalloidin staining (red). Transmigrating lymphocytes were identified by taking confocal sections at different planes and by their characteristic transmigratory morphology (arrowheads). Each panel shows a single projection of a z-stack of between 6 and 8 confocal sections. In merged images, caveolin-1 and VE–cadherin are shown in green, F-actin in red and other proteins in blue.Schematic representations of cells undergoing transcellular transmigration are shown on the right. (a) Caveolin-1 and ICAM-1 were found surrounding the transcellular channel whereas Thy1 (b) is not concentrated around

the passages and remains evenly distributed in endothelial cells that are supporting several simultaneous transcellular diapedesis events. Although many lymphocytes underwent transcellular diapedesis close to intercellular junctions, the junctional proteins VE–cadherin and PECAM-1 did not localize around the passages (c). VEGFR (d) or TfR (e) were not accumulated in the passages either, although an increase in TfR staining can be detected in the area close to the passage due to the T-cell TfR, that appears concentrated at the uropod pole. Scale bars represent 20 µm (B) Quantification of T-lymphoblast paracellular versus transcellular diapedesis across TNF-α-stimulated HDMVECs (mean ± s.e.m. of three different experiments). (C) Frequency of localization of TfR, Thy1, VEGFR3 and caveolin-1 around T- lymphoblasts undergoing transcellular diapedesis. (mean ± s.e.m. of three different experiments).





A RT I C L E S

recruited to caveolae on antibody clustering) nor PECAM-1 (involved in paracellular diapedesis) were enriched in the transcellular passages (Fig. 6A, C). This indicates the distinct composition of transcellular pas-sages compared with the paracellular route.

Higher-magnification images revealed that transcellular passages were surrounded by caveolin-1, F-actin and ICAM-1. Caveolin-1 appeared around the border of, and/or as punctae close to, the transcellular chan-nels (Fig. 5b). ICAM-1 partially colocalized with caveolin-1 at the pas-sages (Fig 5b, basal and medial panels and Fig. 5c, arrowhead) but was also found in apical docking structures (Fig. 5b apical panels and Fig. 5c, arrows), as previously described13,14. The F-actin–caveolin-rich tran-scellular passages were clearly morphologically separate from the api-cal docking structures. In addition, stress fibres were localized around 87 ± 7.1% of the passages, suggesting that the F-actin ring may originate from these structures (Fig. 5b–d). Together, these results indicate that human T-lymphoblasts can follow a transcellular route of transmigration through HUVECs and that translocation of ICAM-1 to caveolin- and F-actin-enriched domains may be a mechanism that promotes the opening of transcellular channels.

Knockdown of caveolin-1 specifically reduces transcellular transendothelial migrationThe role of caveolin-1 in the formation of these transcellular passages was investigated by using three different short-interfering (si)RNAs that knockdown caveolin-1 levels with different efficiencies in HUVECs. Cavlow siRNA transfection induced a small decrease in caveolin-1 expres-sion. Cavmed decreased caveolin-1 levels by around 70%, whereas Cavhigh

siRNA transfection reduced caveolin-1 by more than 80% (Fig. 7a–d). Knockdown of caveolin-1 concomitantly reduced the expression of caveolin-2 (Fig. 7a, b), consistent with studies on caveolin-1-null mice26. Knockdown of caveolin-1 expression reduced transcellular transend-othelial migration, and the reduction was proportional to the level of knockdown (Fig. 7d; see Supplementary Information, Fig. S4). However, total transendothelial migration was not significantly altered, indicating that caveolin-1 is not required for paracellular transendothelial migra-tion (the route taken by the majority of lymphocytes; Fig.7c). Cavhigh

siRNA induced a small decrease in the level of ICAM-1 (around 30%; Fig. 7a, b), possibly because caveolin-1 contributes to signalling pathways regulating ICAM-1 levels, but this did not affect total transendothelial migration (Fig. 7c, d), indicating that this amount of ICAM-1 is sufficient to support lymphocyte transendothelial migration.

DISCUSSIONICAM-1 and VCAM-1 are the major receptors responsible for leukocyte adhesion to the vascular endothelium through engagement of leukocyte integrins. In addition, it has been suggested that ICAM-1 functions in transendothelial migration. Here we describe recruitment of endothelial ICAM-1 to caveola- and F-actin-rich regions following engagement and its translocation, with caveolin-1, to the basal plasma membrane. As caveolin-1, ICAM-1 and F-actin surround the transcellular channels around transmigrating lymphocytes, our data indicate that lymphocyte-induced ICAM-1 recruitment to caveolae initiates the formation of a transcellular passage for lymphocytes to cross endothelial cells.

The majority of research on transendothelial migration has concen-trated on the paracellular route, but analysis of electron microscopy sec-tions indicates that leukocytes can follow a transcellular route in vivo8, and transcellular diapedesis of leukocytes has recently been described in vitro9,27. Although transcellular transendothelial migration occurs in HUVECs (but relatively infrequently observed in vitro), we have found that the fraction of leukocytes taking a transcellular route is much higher in microvascular endothelial cells — which mediate the majority of leu-kocyte transmigration in vivo. Thus, it is possible that there are condi-tions and sites in vivo where transcellular transendothelial migration is the predominant route. Computerized three-dimensional reconstruc-tions of electron microscopy sections suggest that leukocytes cross the endothelium in vivo through endothelial pores resembling vesiculo-vacuolar organelles8. Whether or not vesiculo-vacuolar organelles are formed by the fusion of caveolae is unclear10. However, the morphology of the aggregates of caveolae or caveosomes that we observe associated with ICAM-1 in endothelial cells resemble these vesiculo-vacuolar organelles and contain caveolin-1. It will be interesting to establish the inter-relationship between vesiculo-vacuolar organelles and caveolae.

Our data suggest a mechanism by which leukocytes initiate transcel-lular migration. Lymphocytes induce transient ICAM-1 clustering14,28. If this occurs in areas with a high density of caveolae and actin stress fibres the ICAM-1 may associate with, and induce fusion of, caveolae result-ing in the formation of a transcellular pore. Lymphocytes will search for sites where they can transcytose by pushing down protrusions that constantly probe the endothelial surface. Our TIRFM data support this model, as T-cell contact induces local transcytosis of ICAM-1 and caveolin-1. Actin stress fibres would have an important function in this model — localizing caveolae and guiding ICAM-1 receptors towards cellular areas reinforced by the actin cytoskeleton, and therefore able to

Con

trol

Con

trol

− Cav

low

Cav

low

Cav

med

Cav

med

Cav

high

Cav

high

Con

trol

Cav

low

Cav

med

Cav

high

Con

trol

Cav

low

Cav

med

Cav

high

Caveolin-1

Cav

eolin

-1 e

xpre

ssio

n(p

erce

ntag

e si

RN

A c

ontr

ol)

Caveolin-2

Vimentin

ICAM-1

120

100

80

60

4020

1401201008060

40200

0

120

100

80

60

4020

0

Tran

scel

lula

r TE

M

TEM

a b

c d

Figure 7 Knockdown of caveolin-1 with siRNA decreases T-lymphoblast transcellular transendothelial migration. (a) Effect of transfection of different siRNAs on caveolin-1 levels. HUVECs were transfected with the indicated siRNAs or oligofectamine only (–). Cells were lysed and levels of different proteins monitored by immunoblotting. (b) The graph represents the mean levels of caveolin-1 in four different experiments, taking caveolin-1 levels in cells transfected with siRNA control (luciferase G2) as 100%. Quantification was normalized using vimentin as a loading control (mean ± s.e.m.). (c, d) Effect of the different caveolin-1 siRNAs on total T-lymphoblast transendothelial migration (TEM; c; measured by transwell assays and confirmed in parallel by time-lapse microscopy) and transcellular transendothelial migration (d), measured as in Fig. 5 (mean ± s.e.m. of four different experiements, except for cavhigh, three experiments).





A RT I C L E S

withstand the pronounced remodelling of transcellular passage open-ing. Indeed, F-actin is localized around transcellular pores, presumably to provide structure under the membrane. ICAM-1 could be linked to F-actin through ezrin–radixin–moesin proteins, which bind to the ICAM-1 tail and to actin filaments29. Caveolin-1 has also been shown to associate directly with the actin-binding protein, filamin (which can crosslink and thereby strengthen actin filament networks30).

Caveolae contribute to vascular permeability and clathrin-independ-ent endocytosis and transcytosis in endothelial cells31,32. In addition to their role in the internalization of small molecules, some viruses (such as SV40) enter the cell through caveolae. Reduction of caveolin-1 expres-sion using siRNA clearly diminished the proportion of T-lymphoblasts found in transcellular channels. Importantly, the extent of inhibition of T-lymphoblast entry into endothelial cells was similar to the reported effect of caveolin-1 siRNA on SV40 infection22,33, suggesting a similar involvement of caveolin-1 in both processes and a role for lipid rafts in this pathway of lymphocyte diapedesis.

Caveolae are also important in signal transduction as they compart-mentalize signalling complexes in response to various physiological stimuli34, particularly in focal adhesions35. Some caveolin-associated signalling proteins (such as Src and PLC-γ) are activated downstream of ICAM-1 (refs 16, 36), suggesting that ICAM-1 may signal through its association with caveolae. In addition, ICAM-1 crosslinking stimulates the small GTPase, RhoA15, which has been found associated with lipid rafts37, stress fibre formation and focal adhesion kinase (FAK) activa-tion26,38. As we observe ICAM-1 translocation to the ends of stress fibres and close to focal adhesions, it may signal from this location to focal adhesion proteins such as FAK.

Leukocyte interaction with endothelial docking structures formed by the protrusion of F-actin-rich microvillus-like endothelial membrane protrusions that surround the adhered leukocyte has been described13,14. These protrusions are associated with both paracellular and transcellular diapedesis9. Our observations suggest that a distinct and additional proc-ess occurs to initiate transcellular, rather than paracellular, diapedesis that involves the association of ICAM-1 with F-actin and its translo-cation into caveolae. Caveolae regulate cholesterol homeostasis12, and thus the direct links between caveolae, adhesion-receptor internalization and leukocyte transmigration described here may provide a mechanism whereby the pathological alteration of cholesterol levels — as occurs in artherosclerosis — would directly affect the mechanisms of leukocyte recruitment.

Note added in proof: an accompanying manuscript by Nieminen, M. et al. (Nature Cell Biol. 8, 156–162 (2005)) is also published in this issue.

METHODSMaterials. Human fibronectin, TRITC–FITC–phalloidin, filipin, methyl-β-cyclodextrin, cytochalasin D, phytohemagglutinin, human holotransferrin, interleukin-2, protein G-coupled agarose, mouse anti-β-actin antibody and all chemicals, unless otherwise stated, were from Sigma-Aldrich (Gillingham, UK); sulpho-NHS-succinimido-biotin (sulpho-NHS-biotin) and peroxidase-coupled streptavidin were from Pierce (Rockford, IL); Fugene 6 was from Boehringer Mannheim (Lewes, UK); mouse anti-ICAM-1 (cloneBBIG-I1), anti-VCAM-1 (clone BBA-5) and anti-E-selectin (clone BBIG-E4) antibodies were from R&D Systems (Abingdon, UK); rabbit anti-human transferrin was from Dako (Ely, UK); rabbit anti-caveolin-1 (N-20), mouse anti-transferrin receptor (CD71; 3B8 2A1) and goat anti-PECAM-1 (M-20) were from Santa Cruz Biotechnology (Santa Cruz, CA); mouse anti-VE-cadherin was from Pharmingen (Lowley, UK); anti-mouse IgG conjugated to 10-nm gold particles and anti rabbit IgG

conjugated to 5-nm particles were from British Biocell (Cardiff, UK). Anti-CD59 monoclonal antibodies MEM-43 and MEM43/5 were kind gifts of V. Horejsi (Prague, Czech Republic). Anti-Thy1 monoclonal antibody (clone K117) was a gift from M.A. Alonso (Madrid, Spain). Anti-VEGFR3 mouse monoclonal anti-bodies and anti-LYVE1 rabbit polyclonal antibody were a kind gift from K. Alitalo (Helsinki, Finland). ICAM-1–GFP and Caveolin-1–YFP plasmids were kind gifts of F. Sanchez-Madrid (Madrid, Spain) and A. Helenius (Zürich, Switzerland), respectively. The pEYFP–actin vector was from Clontech (Cambridge, UK).

Cell culture and transfection. HUVECs were isolated from umbilical cords as previously described28, or obtained from BioWhittaker (Wokingham, UK). They were cultured in Nunclon flasks pre-coated with 10 µg ml–1 human fibronectin in EBM-2 (BioWhittaker) supplemented with 2% foetal bovine serum (FBS), endothelial cell growth supplement EGM-2 (BioWhittaker) in an atmosphere of 5% CO2 and 95% air. Unless otherwise indicated, confluent HUVECs were starved for 5 h in EBM-2 medium supplemented with 1% FCS and then stimulated with 10 ng ml–1 TNF-α for 15 h before experiments.

HDMVECs were obtained from PromoCell (Heidelberg, Germany) and cul-tured in Nunclon flasks with the manufacturer’s endothelial cell growth medium MV in an atmosphere of 5% CO2 and 95% air. Cells were plated at confluency on fibronectin-coated dishes for 24 h, then stimulated with 10 ng ml–1 TNF-α for 15 h before experiments.

HUVECs were transiently transfected 24 h after plating on 35-mm dishes with 1–2 µg plasmid DNA using Fugene 6 (Roche, Mannheim, Germany) or Nucleofector (Amaxa Biosystems, Cologne, Germany) according to the manufac-turer’s instructions, and used for experiments 24–72 h post-transfection. siRNA transfection is described below.

T-lymphoblasts were prepared from isolated human peripheral blood mono-nuclear cells (PBMC). Non-adherent PBMC were stimulated with 0.5% phytohae-magglutinin for 48 h and maintained in interleukin-2 as previously described39. T-lymphoblasts were used in experiments after culturing for 7–12 d.

Receptor clustering and confocal microscopy. Anti-ICAM-1, VCAM-1 or E-selectin antibodies (1 µg ml–1) were added to TNF-α-stimulated HUVECs for 45 min. Cells were rinsed and 1 µg ml–1 fluorophore-coupled secondary antibodies added for between 0 and 90 min. All incubations and washes were performed in starving medium containing 10 ng ml–1 TNF-α. Cells were fixed with 4% para-formaldehyde for 20 min, blocked with TBS (25 mM Tris pH 7.4, 150 mM NaCl) for 10 min, permeabilized for 5 min with PBS containing 0.2% Triton X-100 at 4 °C, and incubated at 37 °C with antibodies or 1 µg ml–1 TRITC–FITC-labelled phalloidin. Specimens were mounted in DAKO fluorescent mounting medium (DAKO, Ely, UK).

Confocal laser scanning microscopy was carried out with an LSM 510 (Zeiss, Welwyn Garden City, UK) mounted over an Axioplan microscope (Zeiss) using a × 40 1.3 NA oil immersion objective. To obtain z stacks, 5–20 optical sections were taken over 4 µm.

To quantify ICAM-1, VCAM-1 or E-selectin localization, 10 confocal images per coverslip, stained with antibodies to each receptor and TRITC–phalloidin, were taken from three different experiments. Each image contained an average of 7–8 cells. Images were analysed using ImageJ software (National Institutes of Health, Bethesda, MA) and the mean fluorescent intensity of receptor staining was analysed from regions in the cell centre (top of nucleus or perinuclear region) and at cell poles, where stress fibres converged.

Total internal reflection fluorescence microscopy (TIRFM). TIRFM and epif-luorescence timelapse microscopy and analysis was achieved using an objective-type set-up. An Olympus IX-70 inverted microscope was fitted with an Olympus × 60 1.45 NA Plan-apochromat TIRFM oil-immersion lens, mounted on a piezo-electric focus drive (Physics Instruments, Karlsruhe, Germany). Cells were plated on MatTek coverslips previously coated with 10 µg ml–1 fibronectin and were imaged at 37 °C. For TIR illumination at 488 nm, a 100-mW argon laser (Mells Griot, Carlsbad, CA) was electronically shuttered and coupled via a single-mode fibre (KineFLEX; PointSource, Southampton, UK) to a dual fibre TIR–mono-chromator condenser (Till Photonics, Gräfelfing, Germany). TIR illumination with this configuration had a stimulated penetration depth (1/e) of about 93 nm, assuming refractive index ncytosol = 1.36 and TIR critical angle θmax = 69.7° (see ref. 40 for calculations). Epi-illumination was performed using an electrically shuttered





A RT I C L E S

Polychrome IV monochromator (Till Photonics). GFP and CFP–YFP–Cy5 dich-roic filter blocks (Chroma, Rockingham, VT) were used for TIR–epi-illumination of GFP and epi-illumination of YFP–Cy5 dyes, respectively. Light was detected with an Imago QE-cooled charge-coupled device (CCD camera; PCO; full chip, 1,376 × 1,040 pixel, 6.45 µm2 per pixel, 16-MHz scan rate). Rapid sequential (less than 10 ms delay) TIR and EPI illumination was accomplished using Till Photonics hardware controlled by TillVision software. Typically, dual images were acquired at 1–2 Hz with 2 × 2 binning and exposure times of 400 ms and 100 ms for TIR and EPI illumination, respectively. Images were collected between 5 and 10 s intervals. Images were exported as TIFF format and processed and con-verted to different movie formats using Metamorph software (Universal Imaging, Sunnyvale, CA) and Photoshop.

Electron microscopy. TNF-α-stimulated HUVECs were incubated with anti-ICAM-1 antibody for 45 min. They were then either incubated with gold-con-jugated anti-mouse antibody for 60 min, or fixed with 4% paraformaldehyde in PBS followed by incubation with gold-conjugated anti-mouse antibody. They were then fixed for electron microscopy in 2% paraformaldehyde and 1.5% glu-taraldehyde in 0.1 M sodium cacodylate buffer, followed by 1% osmium tetroxide and 1.5% potassium ferricyanide and treated with tannic acid as described previ-ously41. The samples were then embedded in Epon by conventional procedures, and 60-nm sections were cut using a Leica Ultracut UCT microtome (Vienna, Austria). Sections were stained with lead citrate and viewed using a transmission electron microscope (EM420; Philips, Eindhoven, The Netherlands). To quanti-tate ICAM-1 distribution, 26 randomly selected electron microscopy images of cells were counted. The distribution of 3378 gold particles was analysed. Apical and basolateral caveolae were defined as flask-shaped vesicles closer than 100 nm to the apical or basolateral membrane respectively. Arithmetic mean and standard deviation for each location from the 26 images were calculated using Microsoft Excel software.

To localize caveolin-1 by immuno-electron microscopy, HUVECs (1 × 105) were plated on 0.4-µm pore diameter filters (Falcon; Becton Dickinson, Erembodegum, Belgium) 24 h before TNF-α stimulation and ICAM-1 clustering using secondary antibody conjugated to 10 nm gold particles, as described above. Cells were fixed with 4% paraformaldehyde for 1 h, rinsed, labelled with anti-caveolin-1 antibod-ies followed by anti-rabbit IgG conjugated to 5 nm gold particles and fixed again for 1 h with 2.5% glutaraldehyde. Filters were rinsed in 0.1 M cacodylate buffer and immersed in a 2% tannic acid solution in cacodylate buffer for another hour. Each filter was divided into three strips for processing. Following rinsing, the filters were incubated with 1% osmium tetroxide in 0.1 M cacodylate containing 1.5% potassium ferrocyanide for 1 h. Filters were then stained with a 1% aqueous solution of uranyl acetate for 1 h and dehydrated by sequentially placing them in the following ethanol concentrations, each for 20 minutes: 50, 70, 90, 100% (three times). The final ethanol wash was replaced with propylene oxide. After three 15-min rinses in propylene oxide, the filters were placed in araldite resin (11 g Araldite resin, 10.5 g Dodecenyl sussinic anhydride, 120 uL Dibutyl Pthalate and 600 uL of DMP-30). After 6 h and 12 h this was changed. Following another 6 h the filters were placed in moulds at 60 °C overnight to polymerize. Blocks were sectioned using a Riechert Yung Ultracut ‘E’ microtome. Sections (90 nm) were collected on copper grids and observed under a Jeol 1200 transmission electron microscope. Images were grabbed with a Tietz video and image grabbing sytem. The images were processed using Adobe Photoshop.

siRNA knockdown experiments. All the siRNAs were obtained from Qiagen–Xeragon (Germantown, MD). Three siRNAs were selected with differ-ent efficiency to knockdown human caveolin-1. Cavlow targeted the sequence AACACCTCAACGATGACGTG (nucleotide 448–468 of caveolin-1 mRNA; NM_001753). Cavmed targeted the sequence AAGAGCTTCCTGATTGAGATT (nucleotides 681–701), and has been previously reported42. Cavhigh was the most efficient siRNA for caveolin-1 and was obtained from the validated library of Qiagen (Crawley, UK; cat. 1027110). siRNA for luciferase GL2 was used as a control and was obtained from the control library of Qiagen (cat. 1022070). HUVECs (passage 2) were plated into 6-well plates at 7.5 × 104 cells per well, 24 h before transfection. 15 to 20 µl of siRNA (20 µM stock) were premixed with 4 µl of Oligofectamine reagent (Invitrogen, Carlsbad, CA) following the manu-facturer’s instructions. Cells were transfected for 4 h at 37 °C in 1 ml of EBM-2 medium with growth supplements but no antibiotics or FBS. EBM-2 medium (1 ml ) with growth factors and 8% FBS was then added to each well and cells

were incubated overnight. A second transfection was performed 24 h after the first transfection under the same conditions. After the first transfection (48 h) cells were trypsinized and plated on fibronectin-coated coverslips for time-lapse microscopy and immunofluorescence analysis of T-lymphoblast transmigration or on transwells for transmigration assays (1 × 105 HUVECs and 1.5 × 105 T-lym-phoblasts per coverslip or transwell), and on 24 well dishes for lysis and western blot analysis of caveolin-1 expression (1 × 105 HUVECs per well). Seventy two hours after the first transfection, HUVECs were washed with starving medium and stimulated with 25 ng ml–1 of TNF-α in starving medium for 8 h before performing the different assays.

Measurement of lymphocyte transmigration across HUVECs. Transmigration was analysed by time-lapse video microscopy. HUVECs were stimulated with TNF-α and incubated with T-lymphoblasts (1.5 T-cells per endothelial cell). Sequences of time-lapse images were collected on Axiovert inverted micro-scopes (Carl Zeiss) maintained at 37 °C and 5% CO2 through a ×20 objective, and projected onto KPM1E CCD cameras (Hitachi Denshi Ltd., Waltham Cross, UK). The acquisition of image data and synchronization of the illumination were controlled by Tempus Meteor software (Andor Technology, Belfast, UK) with a frame repetition rate of 15 s over 45 min. Transmigration was detected when phase-bright lymphocytes became phase-dark. Lymphocytes were tracked using Tempus Meteor software.

To quantify transcellular versus paracellular transendothelial migration T-lym-phoblasts were cocultured for between 12 and 15 min with TNF-α-stimulated HUVECs grown on glass coverslips. Medium was removed and cells were fixed with 4% paraformaldehyde at 37 °C. Coverslips were stained with phalloidin to visualize actin filaments, and with anti-caveolin-1 rabbit antibody and anti-ICAM-1 or anti-VE-cadherin monoclonal antibodies to identify endothelial cell perimeters. T-lymphoblasts undergoing transendothelial migration were easily distinguishable (phalloidin stained) by confocal microscopy as the leading edge is in contact with the coverslip whereas the rear uropod is still localized on the apical surface of the endothelial cell. VE–cadherin or ICAM-1 staining revealed whether the cell was undergoing paracellular or transcellular diapedesis.

Transwells were all precoated with fibronectin (10 µg ml–1) for 1 h at 37 °C. siRNA-treated HUVECs (1 × 105) were plated on 6.5-mm diameter transwells (3.5-µm pore size) (Costar, Corning, NY) for 20 h in EBM-2 medium with growth factors. HUVECs were rinsed with starving medium and stimulated for 7 h with TNF-α (25 ng ml–1 in starving medium). T-lymphoblasts (1.5 × 105) were added to each transwell and allowed to transmigrate for 150 min. Transmigrated T-lym-phoblasts in the bottom of the transwell were counted in a Casy Counter (Scharfe System, Reutlingen, Germany).

Note: Supplementary Information is available on the Nature Cell Biology website.

ACKNOWLEDGMENTSThis work was supported by the Ludwig Institute for Cancer Research and European Community contracts QLG1-CT-99-01036 and FP6–502935. J. Millán was supported by a Marie Curie fellowship (no. HPMF-CT-2000-01061) and British Heart Foundation intermediate fellowship (no. FS/04/006). We are grateful to the named donors for the gifts of plasmids and antibodies listed in the methods section, to E. Cernuda Morollon for providing T-lymphoblasts, and to members of the Ridley laboratory for helpful discussions. We thank Olympus for generously providing instrumentation and support to the Yale CINEMA lab.

COMPETING FINANCIAL INTERESTThe authors declare that they have no competing financial interest

Published online at http://www.nature.com/naturecellbiology/Reprints and permissions information is available online at http://npg.nature.com/reprintsandpermissions/

1. Butcher, E. C. Leukocyte–endothelial cell recognition: three (or more) steps to specificity and diversity. Cell 67, 1033–1036 (1991).

2. Johnson-Leger, C., Aurrand-Lions, M. & Imhof, B. A. The parting of the endothelium: miracle, or simply a junctional affair? J. Cell Sci. 113, 921–933 (2000).

3. Burns, A. R. et al. Neutrophil transendothelial migration is independent of tight junctions and occurs preferentially at tricellular corners. J. Immunol. 159, 2893–2903 (1997).

4. Burns, A. R. et al. Analysis of tight junctions during neutrophil transendothelial migra-tion. J. Cell Sci. 113, 45–57 (2000).

5. Oppenheimer-Marks, N. et al. Differential utilization of ICAM-1 and VCAM-1 during the adhesion and transendothelial migration of human T lymphocytes. J. Immunol. 147, 2913–2921 (1991).





A RT I C L E S

6. Schenkel, A. R., Mamdouh, Z. & Muller, W. A. Locomotion of monocytes on endothelium is a critical step during extravasation. Nature Immunol. 5, 393–400 (2004).

7. Muller, W. A. Leukocyte–endothelial-cell interactions in leukocyte transmigration and the inflammatory response. Trends Immunol. 24, 327–334 (2003).

8. Feng, D. et al. Neutrophils emigrate from venules by a transendothelial cell pathway in response to FMLP. J. Exp. Med. 187, 903–915 (1998).

9. Carman, C. V. & Springer, T. A. A transmigratory cup in leukocyte diapedesis both through individual vascular endothelial cells and between them. J. Cell Biol. 167, 377–388 (2004).

10. Dvorak, A. M. & Feng, D. The vesiculo-vacuolar organelle (VVO). A new endothelial cell permeability organelle. J. Histochem. Cytochem. 49, 419–432 (2001).

11. Mundy, D. I. et al. Dual control of caveolar membrane traffic by microtubules and the actin cytoskeleton. J. Cell Sci. 115, 4327–4339 (2002).

12. Gratton, J. P., Bernatchez, P. & Sessa, W. C. Caveolae and caveolins in the cardiovascular system. Circ. Res. 94, 1408–1417 (2004).

13. Carman, C. V., Jun, C. D., Salas, A. & Springer, T. A. Endothelial cells proactively form microvilli-like membrane projections upon intercellular adhesion molecule 1 engage-ment of leukocyte LFA-1. J. Immunol. 171, 6135–6144 (2003).

14. Barreiro, O. et al. Dynamic interaction of VCAM-1 and ICAM-1 with moesin and ezrin in a novel endothelial docking structure for adherent leukocytes. J. Cell Biol. 157, 1233–1245 (2002).

15. Thompson, P. W., Randi, A. M. & Ridley, A. J. Intercellular adhesion molecule (ICAM)-1, but not ICAM-2, activates RhoA and stimulates c-fos and rhoA transcription in endothelial cells. J. Immunol. 169, 1007–1013 (2002).

16. Etienne-Manneville, S. et al. ICAM-1-coupled cytoskeletal rearrangements and transen-dothelial lymphocyte migration involve intracellular calcium signaling in brain endothe-lial cell lines. J. Immunol. 165, 3375–3383 (2000).

17. Kiemer, A. K. et al. Inhibition of p38 MAPK activation via induction of MKP-1: atrial natriuretic peptide reduces TNF-α-induced actin polymerization and endothelial perme-ability. Circ. Res. 90, 874–881 (2002).

18. Feng, D., Nagy, J. A., Dvorak, H. F. & Dvorak, A. M. Ultrastructural studies define soluble macromolecular, particulate, and cellular transendothelial cell pathways in venules, lymphatic vessels, and tumor-associated microvessels in man and animals. Microsc. Res. Tech. 57, 289–326 (2002).

19. Toomre, D. & Manstein, D. J. Lighting up the cell surface with evanescent wave micro-scopy. Trends Cell Biol. 11, 298–303 (2001).

20. Steyer, J. A. & Almers, W. A real-time view of life within 100 nm of the plasma mem-brane. Nature Rev. Mol. Cell Biol. 2, 268–275 (2001).

21. Vasile, E., Simionescu, M. & Simionescu, N. Visualization of the binding, endocytosis, and transcytosis of low-density lipoprotein in the arterial endothelium in situ. J. Cell Biol. 96, 1677–1689 (1983).

22. Pelkmans, L., Puntener, D. & Helenius, A. Local actin polymerization and dynamin recruitment in SV40-induced internalization of caveolae. Science 296, 535–539 (2002).

23. Kvietys, P. R. & Sandig, M. Neutrophil diapedesis: paracellular or transcellular? News Physiol. Sci. 16, 15–19 (2001).

24. Engelhardt, B. & Wolburg, H. Transendothelial migration of leukocytes: through the front door or around the side of the house? Eur. J. Immunol. 34, 2955–2963 (2004).

25. Makinen, T. et al. Isolated lymphatic endothelial cells transduce growth, survival and migratory signals via the VEGF-C/D receptor VEGFR-3. EMBO J. 20, 4762–4773 (2001).

26. Sotgia, F. et al. Intracellular retention of glycosylphosphatidyl inositol-linked proteins in caveolin-deficient cells. Mol. Cell Biol. 22, 3905–3926 (2002).27. Yang, L. et al. ICAM-1 regulates neutrophil adhesion and transcellular migration of

TNF-I activated vascular endothelium under flow. Blood 106, 584–592 (2005).28. Wojciak-Stothard, B., Williams, L. & Ridley, A. J. Monocyte adhesion and spreading

on human endothelial cells is dependent on Rho-regulated receptor clustering. J. Cell Biol. 145, 1293–1307 (1999).

29. Heiska, L. et al. Association of ezrin with intercellular adhesion molecule-1 and -2 (ICAM-1 and ICAM-2). Regulation by phosphatidylinositol 4, 5-bisphosphate. J. Biol. Chem. 273, 21893–21900 (1998).

30. Stahlhut, M. & van Deurs, B. Identification of filamin as a novel ligand for caveolin-1: evidence for the organization of caveolin-1-associated membrane domains by the actin cytoskeleton. Mol. Biol. Cell 11, 325–337 (2000).

31. Parton, R. G. Caveolae — from ultrastructure to molecular mechanisms. Nature Rev. Mol. Cell Biol. 4, 162–167 (2003).

32. Simionescu, M., Gafencu, A. & Antohe, F. Transcytosis of plasma macromolecules in endothelial cells: a cell biological survey. Microsc. Res. Tech. 57, 269–288 (2002).

33. Pelkmans, L. et al. Genome-wide analysis of human kinases in clathrin- and caveolae/raft-mediated endocytosis. Nature 436, 78–86 (2005).

34. Fielding, C. J. Caveolae and signaling. Curr. Opin. Lipidol. 12, 281–287 (2001).35. Wary, K. K., Mariotti, A., Zurzolo, C. & Giancotti, F. G. A requirement for caveolin-1 and

associated kinase Fyn in integrin signaling and anchorage-dependent cell growth. Cell 94, 625–634 (1998).

36. Tilghman, R. W. & Hoover, R. L. The Src-cortactin pathway is required for clustering of E-selectin and ICAM-1 in endothelial cells. FASEB J. 16, 1257–1259 (2002).

37. Michaely, P. A., Mineo, C., Ying, Y. S. & Anderson, R. G. Polarized distribution of endogenous Rac1 and RhoA at the cell surface. J. Biol. Chem. 274, 21430–21436 (1999).

38. Etienne, S. et al. ICAM-1 signaling pathways associated with Rho activation in micro-vascular brain endothelial cells. J. Immunol. 161, 5755–5761 (1998).

39. Millan, J. et al. Lipid rafts mediate biosynthetic transport to the T lymphocyte uropod subdomain and are necessary for uropod integrity and function. Blood 99, 978–984 (2002).

40. Toomre, D. et al. Fusion of constitutive membrane traffic with the cell surface observed by evanescent wave microscopy. J. Cell Biol. 149, 33–40 (2000).

41. Simionescu, N. & Simionescu, M. Galloylglucoses of low molecular weight as mordant in electron microscopy. I. Procedure, and evidence for mordanting effect. J. Cell Biol. 70, 608–621 (1976).

42. Beardsley, A. et al. Loss of caveolin-1 polarity impedes endothelial cell polarization and directional movement. J. Biol. Chem. 280, 3541–3547 (2005).




S U P P L E M E N TA RY I N F O R M AT I O N

WWW.NATURE.COM/NATURECELLBIOLOGY 1

Figure S1 Clustering of surface receptors on HUVECs and HDMVECs. (a-c) Localization of antibody-crosslinked ICAM-1, VCAM-1 and E-selectin in HUVECs. (a) VCAM-1 and E-selectin (green) were clustered on confluent TNF-α-stimulated HUVECs by incubation at 37oC with specific antibodies followed by FITC-labelled secondary antibodies. Cells were stained with TRITC-phalloidin to visualize F-actin. (b) ICAM-1 (red) was clustered on confluent TNF-α-stimulated HUVECs by incubation at 37oC with anti-ICAM-1 antibodies followed by TRITC-labelled secondary antibodies as in Fig. 1. Cells were stained with anti-β-catenin antibody to visualize adherens junctions (green) or FITC-conjugated anti-β-tubulin to detect microtubules. (c) CD59 and transferrin (Tf) do not colocalise with clustered ICAM-1.

HUVECs were treated and ICAM-1 was clustered crosslinked as in (b) and stained using anti-CD59 specific antibodies (green). For Tf localization TNF-α-stimulated HUVECs were preloaded with 100 µg/ml of human holo-transferrin during antibody-mediated ICAM-1 crosslinking. Bars, 10 µm. (d) ICAM-1, transferrin receptor (TfR) or VEGFR3 were crosslinked (X-) on confluent TNF-α-stimulated HDMVECs by incubation at 4oC for 45 min with specific mouse monoclonal antibodies followed by FITC-labelled secondary antibodies for 60 min at 37oC (merged in green). Cells were stained with TRITC-phalloidin to visualize F-actin (merged in blue) and with specific antibodies for caveolin-1 (merged in red). Each panel shows a single projection of a z-stack of between 6 and 8 confocal sections. Bars, 20 µm

© 2006 Nature Publishing Group


2 WWW.NATURE.COM/NATURECELLBIOLOGY

Figure S2 ICAM-1 is internalized at caveolin-enriched domains near cell edges. (a) Schematic representation of the protocol used to discriminate surface ICAM-1 from the total ICAM-1 pool. TNF-α-stimulated HUVECs were incubated for 45 min at 37oC with biotinylated mouse anti-ICAM-1 antibody followed by TRITC-streptavidin for 90 min at 37oC. Cells were then incubated at 4 °C with FITC- or Cy5-coupled anti-mouse secondary antibody for 15 min, fixed and permeabilized. (b) In the upper panels, internalized ICAM-1 is labelled only with TRITC-streptavidin (arrowhead), whereas surface ICAM-1 is labelled with both TRITC-streptavidin (red) and Cy5 anti-mouse antibody (green). Further staining of permeabilized cells with the same secondary antibody (anti-mouse-Cy5x2) was performed (lower panels) to demonstrate that biotinylation and/or streptavidin incubation does not impair secondary antibody binding to the anti-ICAM-1 antibody. F-actin was detected with FITC-phalloidin (shown in blue in merged images). (c) Further staining of permeabilized cells with anti-caveolin-1 antibody (green). Regions outlined with red boxes (upper panel in merged image, c) are shown at 3-fold higher magnification in the lower two sets of panels. Discontinuous lines indicate cell perimeters, arrowheads in (c) show caveolin-1 enriched

domains colocalizing with internal ICAM-1. (d-h) Antibody-clustered ICAM-1 and VCAM-1 are recruited to caveolin-1-enriched detergent-resistant lipid rafts. HUVEC monolayers were treated with TNF-α for either 15 h (d,e,g,h) or 4 h (f). ICAM-1 (d), VCAM-1 (e) and E-selectin (e and f) were clustered with antibodies for the indicated times as described in Fig. 1. Cells were lysed and fractionated using selective solubilisation with β-octyl-glucoside. Equivalent volumes of each fraction were separated by SDS-PAGE and the indicated proteins detected by western blotting. (S) TX100-soluble fraction; (R) TX100-resistant /βΟG-sensitive lipid rafts; (P) pellet. (g, h)) Cell treated with (+) or without (-) crosslinking antibodies to ICAM-1, VCAM-1 or E-selectin for 30 min were lysed, fractionated by sucrose density gradient centrifugation and analysed by western blotting. (P), pellet; (S) TX100-soluble fraction; (DRM), detergent-resistant membranes. ICAM-1 and VCAM-1 are recruited to insoluble lipid rafts in both methods, a smaller fraction of E-selectin was recruited using sucrose density gradient centrifugation (h) as previously described 45 but remained mostly soluble using the β-octyl-glucoside method (e,f).



WWW.NATURE.COM/NATURECELLBIOLOGY 3

Figure S3 ICAM-1 translocation does not require intercellular junctions and occurs close to focal adhesions. (a) TIRFM/EPI time-lapse microscopy of antibody-clustered ICAM-1 in subconfluent cells. Cells were clustered and analyzed as in Fig. 2. Note the shape of ICAM-1 clusters in contact with the basal membrane after 45 min of croslinking. At the top left corner they

form three converging straight lines whereas at the bottom right clusters are organized into circular structures that are internalised (b) Time-lapse confocal microscopy of a basal section showing partial co-localization of clustered ICAM-1 (red) with paxillin-GFP (green). Bar, 20 µm.

Figure S4 Decrease of transcellular TEM correlates with low levels of caveolin-1. Relative total TEM (TEM) and transcellular TEM were plotted against relative levels of caveolin-1, determined by western blotting for each

siRNA transfection from the 4 different experiments in Fig. 6. Black rounded dots in the graphs represent values from siRNA control (x4) taken as 100%.



4 WWW.NATURE.COM/NATURECELLBIOLOGY

SUPPLEMENTARY METHODSDetergent extraction procedures. Selective solubilisation with β-octyl-glucoside. TNF-α-stimulated HUVECs were lysed for 20 min in TST (25 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.2% Triton X-100, 10 µg ml-1 each leupeptin and aprotinin, 1 mM PMSF) at 4oC. Lysates were centrifuged at 14,000 g for 3 min at 4oC and the supernatant (S) and pellet separated. Soluble proteins, solubilised membrane proteins and lipid raft domains not big enough to pellet by conventional centrifu-gation were separated in the S fraction. Pellets were then solubilised in 1 ml of TST plus 60 mM β-octyl-glucoside (OG) for 20 min at 37oC. Homogenised pellet was then centrifuged at 3x104 g for 3 min at room temperature and supernatant (R) (OG-sensitive lipid rafts) and nuclei and cytoskeleton pellet (P) were separated. Equivalent volumes of P, S and R fractions were separated by SDS-PAGE, transferred to PVDF membranes and proteins detected by western blotting. Sucrose density gradient centrifugation. TNF-α-stimulated HUVECs were lysed for 20 min in 1 ml TST at 4oC. Low-density detergent resistant membranes (DRMs) were isolated by standard procedures 41. The pellet (P), soluble fractions (S) (1-2) and insoluble fractions (DRMs) (30-5% interphase) were pooled separately. Equivalent volumes of each fraction were separated by SDS-PAGE and the distribution of proteins detected by western blotting.

Movie 1 Dynamics of ICAM-1 movement to the cell pole. HUVECs were transfected with pEYFP-actin, stimulated with ΤΝF-α and ΙCΑΜ-1 was clustered with antibodies as described in Fig. 2 and Materials and Methods. Images of clustered ICAM-1 (red) and actin-YFP (green) were recorded at 10-sec intervals with a timelapse EPI microscope. The video is displayed at 30 frames/sec.Movie 2 ICAM-1 segregation to the cell periphery is accompanied by translocation to the basal membrane. ΗUVECs stimulated with ΤΝF-α and ICAM-1 was clustered as described in Fig. 2 and materials and Methods. ICAM-1 movement was followed by simultaneous TIRFM/EPI timelapse microscopy at 37oC. TIR (green) and EPI (red) images were recorded at 10-sec intervals. The video is displayed at 8 frames/sec.Movie 3 ICAM-1 segregation to the cell periphery does nor require intercellular junctions ICAM-1 was clustered on subconfluent HUVECs as described in Fig. 2 and Suppl. Fig. 3a. ICAM-1 movement was followed by simultaneous TIRFM/EPI timelapse microscopy at 37oC. TIR (green) and EPI (red) images were recorded at 10-sec intervals. The video is displayed at 8 frames/sec.Movie 4 Dynamics of ICAM-1 and caveolin-1 movement following ICAM-1 clustering. TNF-α-stimulated HUVECs expressing caveolin-1-YFP were analysed by EPI time-lapse microscopy to follow the movement of caveolin-1-YFP (green) and antibody-clustered ICAM-1 (red), as described in Fig. 2. Images were recorded at 10-sec intervals and the video is displayed at 5 frames/sec.


lymphocyte transcellular migration occurs through

Documents