With the Integrin Avb3 (CD51/CD61): an Important Mechanism Mediating Melanoma Cell Adhesion to Activated Endothelium

Interaction of human Thy-1 (CD 90) with the integrin avb3 (CD51/CD61): an important mechanism mediating melanoma cell adhesion to activated endothelium

Anja Saalbach*,1, Anne Wetzel2, Uwe-Frithjof Haustein1, Michael Sticherling2, Jan CSimon 2 and Ulf Anderegg1

1Saxon Academy of Science, D-04107 Leipzig, Germany; 2Department of Dermatology, Venerology and Allergology, University of Leipzig, Stephanstr. 11, D-04103 Leipzig, Germany

The expression of the avb3 integrin (CD51/CD61) on into the blood stream mainly via small veins or the human melanoma cells has been shown to be associated sinusoidal neovasculature of their parent tumor (Orr most closely with tumor progression and metastases et al., 2000). This process involves the degradation of the formation in melanoma. Here, we demonstrated a specific extracellular matrix and of the basement membrane by interaction of the avb3integrinonmelanomacellswiththe cancer cell- as well as host cell-derived enzymes (Meyer human Thy-1, an inducible cell adhesion molecule expressed and Hart, 1998; Wandel et al., 2000, 2002). Many on the cell surface of activated endothelial cells (EC). studies document the correlation of metastasis and the The interaction was shown by the binding of purified Thy- expression of matrix-degrading enzymes (Muller et al., 1 protein to aVb3 transfected cells, to avb3-expressing 1991; Murray et al., 1996; Chambers and Matrisian, melanoma cells and to purified aVb3 integrin. Moreover, 1997; Yoshizaki et al., 1997). However, there is general melanoma cells adhere specifically to Thy-1 transfectants consensus that the arrest of circulating cancer cells is a via avb3 on melanoma cells showing the functional relevance prerequisite for their emigration from the blood stream of this interaction for cell adhesion. Finally, the importance and subsequent outgrowth into metastatic tumors. of the avb3/Thy-1 interaction for the adhesion of melanoma Specific interactions between cancer cells and endothe- cells to the activated endothelium was confirmed under lial cells (ECs) contribute to this process. Cell adhesion static and flow conditions by the inhibition of melanoma cell molecules of the selectin, integrin, cadherin and adhesion to and transmigration across activated EC by immunoglobulin family have been demonstrated to blocking the avb3/Thy-1 interaction. In conclusion, we have govern these adhesive interactions (Orr et al., 2000). identified a new pair of adhesion molecules Thy-1 and Several inducible adhesion molecules on ECs (e.g. avb3 mediating the interaction of melanoma cells and ICAM-1, VCAM-1, MadCAM-1) have been shown to activated EC. These data explain at least in part the high interact with cancer cells. Additionally, treatment of tumorigenicity of avb3-expressing melanoma cells and the experimental animals with inflammatory cytokines such association of avb3-positive melanoma cells with a high risk as IL-1 or TNF-a prior to injection of tumor cells results of metastasis and poor prognosis. in upregulation of inducible adhesion molecules and Oncogene (2005) 24, 4710–4720. doi:10.1038/sj.onc.1208559 subsequent increased formation of metastatic tumors Published online 16 May 2005 (Orosz et al., 1995; Anasagasti et al., 1997). The mechanisms of tumor cell adhesion to the endothelium Keywords: Thy-1; avb3; endothelial cell; adhesion; and the subsequent invasion into the surrounding tissue melanoma share similarities with the interactions occurring during leukocyte extravasation at sites of inflammation. Since the human Thy-1 (CD90) was recently characterized as an activation-associated cell adhesion molecule on ECs with impact on inflammatory process (Wetzel Introduction et al., 2004), we wished to explore the role of this molecule during tumor cell adhesion to activated Lymphatic and blood vessels provide anatomical ‘high- endothelium. ways’ for the traffic of cancer cells between primary Thy-1 is a well-known molecule which has been malignant tumors and their distant sites of metastasis. investigated for over three decades. It is familiar to Morphological studies suggest that cancer cells invade many researchers as a murine T-cell marker. However, Thy-1 shows a remarkable diversity among species (Vidal et al., 1990): In contrast to mouse, in humans *Correspondence: A Saalbach; E-mail: [email protected] Thy-1 is absent from T-cells and is only expressed by Received 4 October 2004; revised 6 January 2005; accepted 26 January fibroblasts, neuronal cells, blood stem cells and 2005; published online 16 May 2005 activated microvascular ECs (Williams et al., 1982; Thy1/avb3 mediates cell–cell adhesion A Saalbach et al 4711 et al a Crawford and Barton, 1986; Craig ., 1993; Saalbach CHO-vector b CHO-αvβ3 et al., 1999). In previous investigations we discovered a novel, up-to-now unrecognized function of human Thy- 1 serving as an adhesion molecule on activated human CHO-αvβ3 ECs (Saalbach et al., 2000; Wetzel et al., 2004). Our recent data prove that Thy-1, expressed on activated count human dermal microvascular EC(HDMEC),mediates albumin the binding to neutrophils and monocytes via the interaction with the leukocyte integrin Mac-1 (CD11b/ CD18) (Wetzel et al., 2004). This demonstrates that human Thy-1 – in addition to ICAM-1 – is an c M21 cells (αvβ3-positive) M21-L cells (αvβ3-negative) alternative ECreceptor for the leukocyte integrin aCD51 (αv) aCD51 (αv) Mac-1, serving an important role for the adhesion and transmigration of neutrophils and monocytes to sites of inflammation. Based on the findings that (i) Thy-1 is strongly expressed on peri- and intratumoral ECs in melanoma β β tissues, (ii) an unknown ligand of Thy-1 was detected on aCD61 ( 3) aCD61 ( 3) melanoma cells (Saalbach et al., 2002) and (iii) melanoma cells adhere to activated Thy-1-expressing ECs (Saalbach et al., 2002), we sought to identify the binding partner of the Thy-1 on human melanoma cells and to characterize the biological function of this interaction. purified Thy1 purified Thy1 In this study, we demonstrate that the specific interaction between human Thy-1 (CD90) on ECs and the integrin avb3 (CD51/CD61) on melanoma cells is of crucial relevance for the adhesion and transmigration of melanoma cells to activated, Thy-1-expressing microvascular ECs. Figure 1 Thy-1 binds to avb3 (CD51/CD61) expressed on the cell surface of avb3-transfected cells and melanoma cells. Binding of purified, biotinylated human Thy-1 or anti-CD51 and anti-CD61 antibodies was detected by streptavidin-phycoerythrin and goat- anti-mouse-Ig-FITC, respectively. Subsequently, cells were analysed by flow cytometry. (a) Purified, biotinylated human Thy-1 Results binds to the cell surface of avb3 (CD51/CD61) transfected CHO cells (black filled curve), whereas there is almost no binding to Human Thy-1 is a counterpart for the integrin vector-transfected CHO cells (gray filled curve) used as control. avb3 (CD51/CD61) The thin line represents the binding of biotinylated albumin to avb3-transfected CHO cells used as negative control. The figure To identify the avb3 integrin as a Thy-1-ligand, binding represents a typical experiment. (b) Binding of anti-avb3 antibody (LM609) to avb3-transfected CHO cells (gray curve) and the of purified, human Thy-1 to avb3-transfected CHO cells respective isotype control (white curve). (c) Antibodies against was studied. As shown in Figure 1a, purified, biotiny- CD51, CD61 and purified, biotinylated human Thy-1 bind to the lated human Thy-1 binds to avb3-transfected CHO- cell surface of M21 melanoma cells (gray filled curve). The selected cells, whereas almost no binding to vector-transfected variant M21L lacking av-expression was not stained by an anti- CHO control cells was detectable (Figure 1a). A CD51 antibody. An anti-CD61 antibody as well as the purified Thy-1 bound weakly to the cell surface of the M21L variant (gray biotinylated BSA was used as another negative control filled curve). The thin lines represent the binding of biotinylated to exclude unspecific binding of purified Thy-1 albumin as control for labeled Thy-1 and an IgG1 control antibody (Figure 1a). Both the av as well as the b3 chain of the as control for specific mAb, respectively. The figure represents a integrin were expressed on the cell surface of all typical experiment of four independent experiments transfected CHO cells, tested by binding of an antibody recognizing only the dimerized avb3 complex (Figure 1b). Since experiments shown in Figure 1 suggested an For further evidence, we compared the binding of an interaction of Thy-1 and avb3 integrin, the direct anti-CD51-(av) and an anti-CD61-(b3) antibody with interaction between the purified proteins Thy-1 and the binding of the purified labeled Thy-1-protein to avb3 was examined in a cell-free system. Indeed, a avb3-positive (M21) and a avb3-negative (M21L) binding of purified, labeled human Thy-1 to immobi- variant of the melanoma cell line M21. As shown in lized avb3 could be demonstrated (Figure 2). Binding of Figure 1c, purified Thy-1 and the anti-CD51 and anti- Thy-1 to immobilized BSA was used as blank to CD61 antibody displayed an identical binding pattern. determine the specific binding of Thy-1 to avb3. In contrast, purified Thy-1 bound only faintly to the Monoclonal antibodies against the av-chain (anti- avb3-negative M21L variant according to the avb3 CD51) as well as against the b3 chain (anti-CD61) deficiency of this variant (Figure 1c). inhibited specifically the binding of Thy-1 to the avb3

Oncogene Thy1/avb3 mediates cell–cell adhesion A Saalbach et al 4712 120 CHO-vector CHO-Thy-1 100 3 (%)

β 80 v * α

40 * * 90% Thy-1 binding to Thy-1 20

IgG

anti-CD51 anti-CD61 Figure 3 Thy-1-transfected cells bind a mAb against Thy-1. CHO anti-CD51/CD61 cells were transfected with a pcDNA-vector or vector containing full-length Thy-1 cDNA (Accession number: M11749) from base Figure 2 Purified Thy-1 protein interacts with pure avb3 integrin 27 to 1519, resulting in the complete mature protein by electro- and this interaction is inhibited by mAb against the avb3 integrin in poration. Stably transfected cells were selected by adding a cell-free system. Binding of purified, biotin-labeled Thy-1 to blasticidin. Thy-1 transfectants (gray filled curve) expressed Thy- immobilized avb3 integrin or BSA was detected in the absence or 1 on their cell surface whereas vector-transfected control cells (gray presence of mAb to CD51 (av), CD61 (b3), in combination of anti- line) did not, as tested by flow cytometry using an anti-Thy-1 CD51 and anti-CD61 antibodies and/or IgG control antibody, antibody (mAb AS02). The figure show a typical histogram of four respectively. The binding of biotinylated Thy-1 was detected by independent experiments addition of alkaline phosphatase-conjugated streptavidin. Non- specific binding to BSA-coated wells was used as blank to calculate specific Thy-1-binding. Binding of biotin-labeled Thy-1 to avb3- coated wells in the presence of blocking antibodies against CD51 and CD61 was expressed as % compared to adhesion in the As shown in Figure 4a, melanoma cell lines expressing presence of an irrelevant control antibody which was set to 100%. both the av- as well as the b3-chain of integrins (Mewo, Data are represented as mean7s.d. of three independent experi- JPC298, P1, A375, M21) adhered clearly to Thy-1- ments measured in duplicate.nPo0.05 as compared to binding without blocking antibodies transfected cells, whereas only a weak adhesion to vector-transfected control cells was detectable. In contrast, cell lines expressing only the av-chain (SK- Mel-28, Bro, HT144) or lacking both chains (M21L) were not able to bind to Thy-1-transfected cells integrin up to 40–70%; by contrast, an IgG control (Figure 4a). Furthermore, binding of M21 as well as antibody had no effect (Figure 2). JPCmelanoma cells, but not of the avb3-deficient M21L cell line, to Thy-1-transfected CHO cells was significantly inhibited by about 70–80% following preincuba- Adhesion of melanoma cells to human Thy-1 is mediated tion with an antibody against the avb3 heterodimer by the avb3 integrin (clone LM609, Figure 4b). In fact, such an inhibition of To determine whether the Thy-1/avb3 interaction avb3 reduced the adhesion of avb3-positive melanoma is involved in melanoma cell adhesion to Thy- cells to the level of melanoma cells lacking avb3 on their 1-expressing cells, we established a CHO cell line stably cell surface (M21L). As yet another specificity control, transfected with human Thy-1. Full-length Thy-1 was incubation with an isotype control antibody as well transfected into CHO cells and stable transfectants were as antibodies detecting other cell surface antigens, for isolated. Using flow cytometry, we showed that >90% example, HLA-ABC, or other integrin chains (a3 of stably Thy-1-transfected CHO cells (CHO.Thy-1) (CD49c); a5 (CD49e)) did not affect the adhesion to expressed Thy-1 on their cell surface, whereas vector- Thy-1-transfected cells (Figure 4b). transfected control cells (CHO.vec) did not (Figure 3). To further examine the binding characteristics of the CHO-Thy-1-, CHO-vec-cells, the melanoma cell line avb3/Thy-1 interaction, we studied the impact of M21 expressing avb3 and its avb3-deficient variant stimulation of the affinity or avidity of avb3by M21L were used in cell adhesion assays (Figure 1c). preincubation of melanoma cells with Mn2 þ /Mg2 þ ions M21 melanoma cells adhered strongly to Thy-1 trans- or PMA, respectively. As shown in Figure 4c, divalent fectants consistently with the clear binding of purified cations had no significant effect on the Thy-1/avb3- Thy-1 to these cells (Figures 4a, 1c). The avb3-deficient mediated melanoma cell adhesion. In contrast, PMA M21L variant bound significantly less to CHO-Thy-1 clearly stimulated the Thy-1-mediated melanoma cell according to the faint binding of purified Thy-1 to adhesion (Figure 4c). Furthermore, an RGD peptide M21L (Figures 4a, 1c). Moreover, we tested several was able to significantly reduce the Thy-1-mediated melanoma cell lines differing in their avb3 expression. melanoma cell adhesion (Figure 4c).

Oncogene Thy1/avb3 mediates cell–cell adhesion A Saalbach et al 4713 Physiological role of the Thy-1/avb3 interaction in wished to determine the functional importance of this the adhesion and transmigration of melanoma cells interaction under more physiological conditions. First, we studied the functional relevance of the Thus far, the experiments described above demonstrated avb3/Thy-1 interaction in the adhesion of melanoma a specific interaction of Thy-1 and avb3 on melanoma cells to HDMEC, the physiological target in vivo that cells and its functional role in cell adhesion. Next we also express the naturally occurring panel of different adhesion molecules. To block Thy-1, we generated an mAb (B-C9) which a αvβ3-positive β3- or αvβ3-negative is capable of functionally blocking Thy-1-mediated cell– cell adhesion. The mAb B-C9 recognized Thy-1-transfected cells as well as Thy-1-expressing fibroblasts, 100 CHO.Thy-1 * whereas no binding to vector-transfected control cells * * CHO.vec was detectable (Figure 5a). Furthermore, we showed 80 that the mAb B-C9 recognizes specifically the Thy-1 protein when immunoprecipitated with the anti-Thy-1- * antibody AS02. Conversely, we also demonstrated that 60 the mAb B-C9 precipitates a protein which is recognized by the anti-Thy-1 antibody AS02 (Figure 5b). In both 40 * experiments, irrelevant control antibodies neither precipitated a protein which is detectable by one of these 20 antibodies nor were able to bind to the precipitated absolute number of adherent cells proteins. These results prove that mAb B-C9 recognizes the human Thy-1. Finally, in adhesion assays we 0 illustrated that the mAb B-C9 is able to inhibit the C 5 o 1 1 ro 8 4 L 7 w P 2 B l-2 4 1 JP 3 e M e T1 2 adhesion of melanoma cells to Thy-1-transfected cells to A M M H M SK- about 60% in comparison to an irrelevant isotype control antibody. Likewise, blocking avb3 with the anti- 120 b avb3 antibody (LM609) reduced the melanoma cell adhesion to a similar extent. This blocking effect is 100 specific, since both antibodies do not interfere with

* binding of melanoma cells to control cells in the absence 80 * of Thy-1 (CHO.vec; Figure 5c).

* # 40 # Figure 4 Adhesion of melanoma cells to Thy-1 transfectants is (CHO.Thy1-CHO.vec) # mediated by the avb3 integrin. (a) The adhesion of different % adhesion to CHO.Thy-1 20 melanoma cells to a confluent monolayer of Thy-1-transfected cells (black bar) and vector-transfected control cells (grey bar) was detected. Melanoma cells expressing avb3 adhered strongly and 0 specificially to Thy-1-transfected CHO cells. The lack of b3 chain G C 9e 9c 51 61 61 G 61 G 61 Ig B 4 4 D D D Ig D Ig D or avb3 on melanoma cells yields only very low basal binding to + -A D D C C C + C + C C A C C a a 1/ 21 1/ 1L 1/ JP L a +a + + 5 M 5 2 5 CHO.Thy-1 and CHO.vec control cells. Cells were biotin-labeled, aH + C C C D D M D + PC P JP JP C C C C J J +a +a +a added to CHO.Thy-1 and CHO.vec and incubated for 30 min at JP C 1 L P 2 21 371C. Unbound cells were removed, adherent cells were fixed and J M M visualized by incubation with CYt3-conjugated streptavidin. The c 500 experiments were performed in duplicate. Data represent the * 7 400 mean s.d. of three independent experiments (*Po0.05). (b)To demonstrate the specificity of the avb3/Thy-1 interaction, the 300 experiments were performed with two different melanoma cell lines 200 in the presence of an isotype control antibody and mAb against HLA-ABC, integrin a3 (CD49c), integrin a5 (CD49e), CD51, * CD61, and the dimerized avb3 (CD51/CD61). To calculate Thy-1- 100 specific adhesion of melanoma cells, binding to vector-transfected * cells was subtracted from those to Thy-1-transfected CHO cells for each condition. Adhesion of melanoma cells with the isotype * control antibody was set up to 100%. Data represent the mean7s.d. of three independent experiments. #Po0.05 vs M21 with IgG; *Po0.05 vs JPCwith IgG. ( c) Melanoma cells were (CHO.Thy1-CHO.vec)

% adhesion CHO.Thy-1 incubated with RGD peptide as indicated, 2 mM MnCl2,2mM MgCl2, or PMA (20 ng/ml) for 30 min at 371C. Cells were washed and the adhesion to Thy-1-transfected cells and vector-transfected 0 control cells was determined. To calculate Thy-1-speciﬁc adhesion - A µM µM µM l 2 l 2 of melanoma cells, binding to vector-transfected cells was 0 0 0 nC gC PM 1 5 10 M M subtracted from those to Thy-1-transfected CHO cells. Adhesion D D D G G G R R R of melanoma cells without any preincubation was set to 100%

Oncogene Thy1/avb3 mediates cell–cell adhesion A Saalbach et al 4714 a b anti-Thy-1 (AS02) IgG mAb BC9

mAb anti-Thy-1 (B-C9) marker 100 kDa 73 kDa 54 kDa CHO-vector 46 kDa 35 kDa

24 kDa CHO-Thy-1fibroblasts

16 kDa

10 kDa 123 4

c 100

80 IgG1 BC9 (anti-Thy-1)

20 absolut number of adherent cells absolut number

0 CHO.Thy-1 CHO.vec Figure 5 The mAb B-C9 recognizes the human Thy-1 and inhibits cell-cell adhesion. (a) The mAb binds to Thy-1-transfected CHO cells (gray filled curve) as well as fibroblasts (black filled curve) analysed by flow cytometry. In contrast, mAb B-C9 did not bind to vector-transfected CHO cells (gray line). An irrelevant IgG antibody (black line) did not bind to these cells. (b) An irrelevant IgG control antibody (lane 1) and the mAb B-C9 (lane 2) were coupled to goat-anti-mouse-magnetic beads. Beads were incubated with fibroblast extracts containing the Thy-1 protein. Bound proteins were eluted, loaded on a 12% polyacrylamide gel and blotted to a nitrocellulose membrane. Precipitated proteins were detected with the anti-Thy-1-antibody AS02 and a goat-anti-mouse antibody coupled to alkaline phosphatase. In opposite, the anti-Thy-1-antibody AS02 was coupled to goat-anti-mouse-magnetic beads (lanes 3 þ 4) and fibroblast extract was added. Precipitated proteins were detected with the mAb B-C9 (lane 4) or with an irrelevant control antibody (lane 3). (c) Adhesion of melanoma cells (JPC-298) to a confluent monolayer of Thy-1-transfected cells and vector-transfected control cells was detected after incubation with the mAb B-C9, an anti-avb3 antibody (LM609), or with an irrelevant control antibody, respectively. Melanoma cells were biotin-labeled and added for 30 min at 371Cto Thy-1-transfected cells and vector-transfected control cells. After incubation for 30 min at 371C, unbound cells were removed, adherent cells were fixed and visualized by incubation with CYt3-conjugated streptavidin. The adhesion of melanoma cells to Thy-1-transfected cells after incubation with an irrelevant IgG1 antibody was set to 100%. Data represent the mean7s.d. of at least three independent experiments. *Po0.05

The mAb B-C9 was used to investigate the role of the Previous studies showed the functional role of Thy-1 Thy-1/avb3 interaction in melanoma cell adhesion to in the adhesion of melanoma cells under static condi- activated HDMEC. PMA treatment (48 h, 20 ng/ml) tions. Next, we further evaluate the impact of the avb3/ results in Thy-1 expression on 40–70% of cultured Thy-1 interaction for melanoma cell adhesion under HDMEC(Saalbach et al., 1999). Subsequently, mela- flow conditions simulating the physiological situation noma cell adhesion to activated, Thy-1-expressing during circulation and invasion of melanoma cells in HDMECwas determined in the presence of function- secondary tissues during metastasis formation. blocking antibodies to Thy-1 (mAb B-C9) as well as to Melanoma cells were directed over a confluent cell the avb3 integrin (LM609). As shown in Figure 6, both monolayer under a constant flow of 0.15 ml/min (5 dyn/ blocking of Thy-1 and avb3 results in a significant cm2). As shown in Figure 7a, avb3-positive melanoma decrease in melanoma cell adhesion to activated, Thy-1- cells (JPC) were also able to bind to Thy-1-transfected expressing HDMEC. Inhibition of avb3 reduced the cells under this conditions (2475 cells/microscopic field), adhesion of avb3-positive M21 melanoma cells to the whereas these melanoma cells adhered significantly level of avb3-deficient melanoma cells (M21L) similar to less to the vector-transfected control cells (1371 the experiments using Thy-1 transfectants (Figure 4b). cells/microscopic field). Thus, adhesion of avb3-expres- In contrast, the antibodies had no impact on the basal sing melanoma cells to Thy-1 transfectants is clearly adhesion of melanoma cells to unstimulated HDMEC enhanced compared to the control cells under static as (data not shown). well as under flow conditions (Figures 4a, 7).

Oncogene Thy1/avb3 mediates cell–cell adhesion A Saalbach et al 4715 160 a b CHO.vec+JPC HDMEC+PMA +JPC 140

120

100

80 # * # 60 * * 100% 100% # HDMEC+PMA+isotype antibody+JPC 40 CHO.Thy-1 +JPC absolut number of adherent cells absolut number 20

JPC M21 y-1) CD61 M21L

JPC+IgG1 JPC+aCD51JPC+a C9 (aTh JPC+aCD51/61 95% M21+aCD51/CD61 184% JPC+B-C9 (aThy-1) M21+B- M21L+aCD51/CD61 HDMEC+PMA+aThy-1 (BC9)+JPC Figure 6 Adhesion of melanoma cells to ECs is mediated partially by the interaction of Thy-1 with the avb3 integrin. HDMECwere stimulated with PMA to induce Thy-1 expression on 40–60% of the cells. Melanoma cells (JPC, M21 and M21L) were biotin-labeled and added to a confluent monolayer of stimulated HDMEC. After incubation for 30 min at 371C, unbound cells were removed, adherent cells were fixed and visualized by incubation with CYt3- conjugated streptavidin. For blocking the Thy-1/avb3 interaction, either melanoma cells were pre-incubated with indicated antibodies 41%41% against the avb3 integrin or the HDMECwere incubated with the mAb B-C9. The experiments were performed in duplicate and HDMEC+PMA+JPC (aCD51/CD61) the number of adherent cells was counted in 10 microscopic fields. The adhesion of melanoma cells to stimulated HDMECin the presence of an irrelevant control antibody (IgG) was set up to 100%. Data represent the mean7s.d. of at least three independent experiments. *Po0.05 vs JPC; #Po0.05 vs M21

29% Next, we were interested in the physiological role of the Thy-1/avb3 interaction in the adhesion of melanoma Figure 7 Melanoma cells adhere specificially via avb3 and Thy-1 transfectants and to activated HDMECunder flow conditions. cells to activated, Thy-1-expressing HDMECunder flow CHO.Thy-1, CHO.vec or stimulated HDMEC, respectively, were conditions. cultured on glass coverslips. FURA2/AM-labeled melanoma cells As shown in Figure 7b, melanoma cells (JPC) clearly were directed over the confluent cell monolayer in a heated adhered to activated HDMECunder flow conditions incubation flow chamber under a constant flow of 0.15 ml/min 2 (7871 cells/microscopic field). An isotype control anti- (5 dyn/cm ). Adherent melanoma cells were counted in 20 micro- 7 scopic fields. The adhesion to CHO.vec and PMA-activated body had no effect on this adhesion (74 14 cells/ HDMEC, respectively, was set to 100%. The adherent melanoma microscopic field). Blocking of Thy-1 as well as avb3 cells (red) and the underlying confluent cell monolayer (grey) were significantly inhibited (3172 and 2272 cells/micro- photographed. The figure represents typical images of three scopic field, respectively) the adhesion of melanoma independent experiments. (a) Adhesion of melanoma cells to Thy-1-transfected CHO.cells (CHO.Thy-1) and vector-transfected cells to activated HDMECabout 60–70% (Figure 7). CHO-cells (CHO.vec). (b) Adhesion of melanoma cells to activated Taken together, blocking of Thy-1 or avb3 reduce the HDMEC(70% Thy-1-positive) without preincubation or preincu- adhesion of melanoma cells to activated ECs about 50– bation with an isotype control antibody, with the anti-Thy-1 70% under static as well as under flow conditions antibody mAb BC9, or the anti-avb3 antibody (Figures 6, 7). Finally, the role of the Thy-1/avb3 interaction on the transmigration of melanoma cells across a confluent monolayer of PMA-activated, Thy-1-expressing function-blocking antibodies compared to an isotype HDMECwas examined. HDMECwere cultured on control antibody. Importantly, the transmigration of transwell inserts and Thy-1 expression was induced by melanoma cells (JPC) across a monolayer of activated PMA. HDMECwere extensively washed to remove Thy-1-positive HDMECwas inhibited about 50% by PMA before transmigration assay was performed. The blocking Thy-1 with the mAb BC9 as well as by functional relevance of Thy-1 and avb3 for transmigra- blocking avb3 on melanoma cells compared to an tion was studied by inhibition of Thy-1 or avb3by isotype control antibody (Figure 8). Thus, the specific

Oncogene Thy1/avb3 mediates cell–cell adhesion A Saalbach et al 4716 120 Since the interaction of melanoma cells with the endothelium and subsequent transmigration through the 100 ECbarrier is an important prerequisite for tumor cell metastasis, our ﬁndings present yet another, previously 80 unrecognized mechanism by which avb3 promotes melanoma progression. * * The human Thy-1 is an activation-associated cell 60 adhesion molecule on HDMEC(Saalbach et al., 1998, 1999, 2000). In situ, the expression of Thy-1 was 40 detected exclusively on ECthat show signs of cell

% of transmigrated cells % of transmigrated activation, inflammation or angiogenesis, including the 20 stromal environment of malignant melanomas (Saal- bach et al., 1998, 1999, 2002). In contrast, there was no 0 expression of Thy-1 on the cell surface of resting ECin 9) 61 healthy tissues (Saalbach et al., 1998, 1999; Wetzel et al., 2004). CD51/CD ontrol antibody nti-Thy1 (BC Leyton et al. (2001) identified the avb3 on murine MEC + a astrocytes as a Thy-1 ligand in mice. However, Thy-1 HD HDMEC + c HDMEC + a expression in humans and mice shows striking cross- Figure 8 Transmigration of melanoma cells to ECs is mediated species differences as described earlier (Saalbach et al., partially by the interaction of Thy-1 with the avb3 integrin. 2002). Therefore, data describing Thy-1 expression and HDMECwere stimulated with PMA to induce Thy-1 expression function cannot be easily transferred between species, on 40–60% of the cells. Subsequently, HDMECwere extensively washed to remove PMA. The transmigration of melanoma cells and complicate the validation of the physiological across a monolayer of PMA-activated Thy-1 expressing HDMEC importance of the Thy-1/avb3 interaction in in vivo was studied after blockade of Thy-1 or avb3 by the function- models. Furthermore, the functional relevance of the blocking anti-Thy-1 antibody BC9 or an anti-avb3 antibody, interaction of murine Thy-1 and avb3 is still unknown. respectively. Transmigration in the presence of an isotype control By contrast, in the present study, we identified for the antibody was used as control and set to 100%. The number of transmigrated cells was analysed by flow cytometry. Data are the first time the functional role of this interaction for the mean7s.d. of three independent experiments; *Poo0.05 adhesion of melanoma cells to ECs in humans. We demonstrated a specific interaction of the human Thy-1 with the avb3 integrin by the binding of purified Thy-1 protein to avb3-transfected cells, to avb3-expres- interaction between Thy-1 and avb3 contributes to both sing melanoma cells and to purified avb3 integrin. The the adhesion and transendothelial migration of mela- specificity of this interaction was demonstrated by the noma cells. inhibition with antibodies against the av- or b3-chain. The functional role of this interaction was demonstrated in cell–cell adhesion assays showing a strong and specific adhesion of different avb3-positive melanoma Discussion cell lines to Thy-1 transfectants. The inability of melanoma cells lacking the b3-chain to bind to Thy-1- The expression of the avb3 integrin (CD51/CD61) has transfected cells as well as the strong inhibition of the been shown to be associated most closely with tumor Thy-1/avb3 interaction by an anti-b3 antibody point to progression and metastasis formation in human malig- the importance of the b3-chain for this interaction. nant melanoma. It is presently the best molecular Moreover, blocking avb3 reduced the adhesion of marker correlating with the change from radial to avb3-expressing melanoma cells to the level of melano- vertical growth in primary melanoma (Johnson, 1999). ma cells lacking avb3. The specificity of the avb3 As this change is thought to be a hallmark of the blockade is also underlined by the fact that other development of a metastatic phenotype, avb3 expression antibodies against several cell surface molecule did not by melanoma cells is a reliable predictor of poor clinical influence the Thy-1-mediated melanoma cell adhesion as prognosis. Several molecular mechanisms may explain well, as that these antibodies did not interfere with the the tumor-promoting activity of avb3. For example, adhesion of avb3-negative melanoma cells. Further- expression of avb3 on ECs (EC) plays a pivotal role in more, an increase of the avidity of avb3 by PMA also tumor-induced angiogenesis (Varner and Cheresh, 1996; enhanced the Thy-1-mediated melanoma cell adhesion. Kumar et al., 2000). On the other hand, activation of These results prove that binding of melanoma cells to avb3 on melanoma cells prevents apoptosis, stimulates Thy1 is realized by the interaction with the avb3 integrin tumor growth and matrix invasion (Johnson, 1999). on melanoma cells. Here, we report for the first time that the avb3 Since avb3 interacts with an RGD motif and Thy-1 integrin on melanoma cells serves a functional role as contains an RLD peptide, a similar sequence, we tested a cell–cell-adhesion molecule mediating the binding of whether RGD peptides are able to interfere with the melanoma cells to the endothelium via the interaction avb3/Thy-1 interaction. The inhibition of the Thy-1- with the human Thy-1 (CD90). mediated adhesion of melanoma cells by RGD peptides

Oncogene Thy1/avb3 mediates cell–cell adhesion A Saalbach et al 4717 suggest that avb3 could bind to the RLD motif in the In conclusion, we have identified a new pair of Thy-1. Thus, our results also demonstrate that con- adhesion molecules Thy-1 and avb3 mediating the ceivable therapeutic strategies based on the inhibition of interaction of melanoma cells and activated EC. These avb3 (humanized antibodies, small molecule inhibitors, data explain at least in part the high tumorigenicity or RGD-based peptides) would also affect melanoma of avb3-expressing melanoma cells and the association cell adhesion to Thy-1-expressing cells, including acti- of avb3-positive melanoma cells with a high risk of vated ECs or fibroblasts. metastasis and poor prognosis. Finally, the relevance of the avb3/Thy-1 interaction for the adhesion of melanoma cells to ECs was investigated. The significant inhibition (about 50–60%) Materials and methods of the binding of melanoma cells to activated ECby blocking the avb3/Thy-1 interaction demonstrates the Monoclonal antibodies (mAb) importance of this mechanism for melanoma cell The mAbs against the avb3 integrin and the IgG1 isotype adhesion to ECs. control were purchased from Chemicon (anti-CD51, anti-avb3 Our findings that also under a constant shear flow (clone LM609), Hofheim, Germany) and Immunotech (anti- (5 dyn/cm2) melanoma cells strongly adhere to Thy-1 CD61 Hamburg, Deutschland). Phycoerythrin (PE)- and transfectants and that melanoma cell adhesion to CYt3-labeled streptavidin, dichlorotriazinyl-amino-fluore- activated, Thy-1-expressing ECs is significantly inhibited scein (DTAF)-labeled goat-anti-mouse were purchased from by blocking the avb3/Thy-1 interaction underlines the Jackson Immunoresearch (USA). Monoclonal antibody AS02 importance of this interaction for the attachment of (anti-Thy1) was used as hybridoma cell culture supernatant as melanoma cells to ECs as a prerequisite for their described previously (Saalbach et al., 1998). The RGD peptide (GRGDSPK ) was purchased from Sigma (Deisenhofen, transmigration into the tissue. Moreover, the Thy-1/ Germany). avb3 interaction is also involved in the transmigration of melanoma cells since blocking Thy-1 and avb3 resulted Cell lines and culture conditions in a significant inhibition of the migration of melanoma cells across a monolayer of activated, Thy-1-expressing The melanoma cell lines JPC298 (Aubert et al., 1993), A375 ECs. Taken together, our data provide strong evidence (Giard et al., 1973), Mewo (Bean, 1975), P1 (kindly provided that Thy-1 expressed on ECs is involved in the adhesion by Dr Dummer, Zurich), SK-Mel28 (Carey et al., 1976), HT144 (Ahrens et al., 2001), Bro (Lockshin et al., 1985) were and transmigration of melanoma cells via the interaction maintained in RPMI 1640 with stable glutamine supplemented with avb3 integrin, thereby promoting the invasion of with penicillin (400 U/ml), streptomycin (50 mg/ml), ascorbic melanoma cells into secondary tissues and subsequent acid (50 mg/ml) and 10% fetal calf serum (Seromeds, metastasis formation. In accordance, avb3-expressing Biochrom KG; Berlin, Germany), in a humidified atmosphere melanoma cells exhibit a high tumorigenicity in mice. of 5% CO2 at 371C. In contrast, a melanoma cell line (M21L) lacking The M21 melanoma cell line and its stable variant lacking the avb3 expression showed a dramatically reduced av integrin (M21L) were kindly provided by Dr D Cheresh tumorigenicity when transplanted into nude mice (The Scripps Institute, La Jolla, CA, USA) and cultured in (Cheresh and Spiro, 1987; Felding-Habermann et al., DMEM with 10% FCS and 1% penicillin/streptomycin at 1 1992). 37 C,5%CO2. The M21 cell line was passaged after reaching confluence using 0.05% trypsin, 0.02% EDTA, whereas M21L Until now, the best known ligands of the avb3 grows in suspension. integrin are matrix proteins such as vitronectin, fibro- The avb3-, vector-transfected CHO cells and CHO cells were nectin, laminin, collagen and osteopontin (Smith and kindly provided by Professor Gawaz (Herzzentrum Munich, Cheresh, 1990; Felding-Habermann and Cheresh, 1993; Germany). Cells were cultured in DMEM with 10% FCS and Byzova et al., 2000). In this respect, our finding that 1% penicillin/streptomycin at 371C,5%CO2, and passaged Thy-1 and avb3 interact is important since it demon- after reaching confluence using 0.05% trypsin, 0.02% EDTA. strates that the avb3 does not only interact with matrix HDMEC were prepared from human foreskin. The epi- components but also with a cell surface receptor on EC dermis was removed after 24 h incubation with dispase (2.5 resulting in cell–cell adhesion. The identification of the mg/ml; Roche Applied Science, Mannheim, Germany). The Thy-1 on activated ECas a avb3 ligand is in good dermis was firmly pressed with a scalpel to release ECs. The EGM-MV Bulletkit (Promocell, Heidelberg, Germany) accordance with data showing that the integrin avb3 was used for culture according to the original protocol. recognizes the human neuronal cell adhesion molecule To assess the quality of cell preparation (contamination L1 and platelet ECadhesion molecule-1 expressed on with fibroblasts), CD31 expression (anti-CD31, Beckmann- EC(Buckley et al., 1996; Ebeling et al., 1996; Felding- Coulter, Krefeld, Germany) was tested by flow cytometry as Habermann et al., 1997). Both interactions were described (Saalbach et al., 2002). Only preparations with assumed to be involved in leukocyte extravasation to >90% CD31-positive cells were used. To induce Thy-1 sites of inflammation, tissue injury or infection (Piali expression, HDMECwere stimulated with 20 ng/ml phorbo- et al., 1995; Duczmal et al., 1997; Weerasinghe et al., lester (PMA) for 24 h (phorbol myristate acetate; Invitrogen, 1998). Voura et al. (2001) also demonstrated the Carlsbad, CA, USA). involvement of the avb3 interaction with L1 in the transmigration of melanoma cells. Thus, avb3 integrin Generation of stably Thy-1-transfected CHO cells mediates both cell adhesion to extracellular matrix as Full-length Thy-1 cDNA (accession number: M11749) from well as cell–cell adhesion. base 27–1519 resulting in the complete mature protein was

Oncogene Thy1/avb3 mediates cell–cell adhesion A Saalbach et al 4718 cloned into pcDNA6 vector containing a blasticidin resistance Immunoprecipitation gene (Invitrogen). CHO cells were detached using 0.05% The anti-Thy-1 mAb as well as an isotype control IgG trypsin, 0.02% EDTA and transfected with 1 mg plasmid antibody (10 mg) were bound to 1 Â 108 goat-anti-mouse IgG- containing Thy-1 or with the vector alone by electroporation DYNABEADSR M450 (Dynal/ Oslo, Norway) overnight at (300 V, 1500 mF; Equibio (Peqlab, Erlangen, Germany)). Cells 41C, respectively. After washing with PBS, these beads were were immediately transferred into cell culture and blasticidin incubated with crude extracts of 5 Â 106 fibroblasts for 16 h at (7.5 mg/ml, Invitrogen) was added after 48 h to select stably 41C. Proteins bound to magnetic beads were eluted with SDS- transfected cells. Transfectants expressing Thy-1 were identi- loading buffer after extensive washing with PBS/0.3% Tween fied by flow cytometry using an anti-Thy-1-antibody (mAb 20. The isolated proteins were separated by SDS–PAGE and AS02; Saalbach et al., 1998). blotted onto nitrocellulose membranes. The anti-Thy-1 mAbs or an isotype control IgG antibody followed by a secondary Generation of an anti-Thy-1-antibody blocking Thy-1-dependent goat-anti-mouse antibody conjugated with alkaline phospha- cell adhesion tase were used to detect the isolated proteins as described previously (Saalbach et al., 1996). Balb/c mice were immunized intraperitoneally with 1 Â 107 cells of Thy-1-transfected CHO cells (Thy-1-CHO) 9 and 3 weeks before booster immunization. Spleen cells were fused In vitro ligand–receptor interactions with X63-Ag.8.653 myeloma cells using polyethyleneglycol MaxiSorb plates (Nunc, Wiesbaden, Germany) were coated according to the principle of Ko¨ hler and Milstein (1975). with 200 ng/well avb3 integrin (Chemicon) or BSA in TBS Hybridoma supernatants were screened for antibodies that 2mM CaCl2/0.1 mM MgCl2 overnight. Plates were washed bind to Thy-1-CHO but do not recognize vector-transfected three times with TBS and blocked with TBS/1% BSA for CHO cells. One hybridoma clone of interest (patent pending) 30 min at room temperature. Purified, biotinylated Thy-1 was selected, recloned and used as hybridoma cell culture (1 mg/well) was added to immobilized avb3 as well as to BSA- supernatant in subsequent investigations. coated control wells. Plates were washed with TBS containing 0.1% BSA, 0.05% Tween-20, 2 mM CaCl2 and 0.1 mM MgCl2 after 2 h at room temperature. Thy-1 binding to immobilized Protein purification of human Thy-1 avb3 as well as to BSA as negative control was performed in Thy-1 was purified from cell extracts of human dermal the absence or presence of mAb to CD51, CD61 (10 mg/ml), fibroblasts by immunoaffinity-chromatography using mAb respectively. The binding of biotinylated Thy-1 was detected AS02 as described previously (Saalbach et al., 1998). by addition of alkaline phosphatase-conjugated streptavidin Briefly, the Thy-1-specific mAb AS02 was immobilized (Sigma). Subsequently, para-nitrophenyl-phosphate (pNPP, on cyanogen bromide-activated sepharose (Amersham- 1 mg/ml) was used to generate color reaction that was Biosciences, Freiburg, Germany) according to the manufac- measured at 405 nm. Nonspecific binding to BSA-coated wells turer’s protocol. Human dermal fibroblasts were solubilized was used as blank to calculate specific Thy-1-binding. in 20 mM phosphate buffer with 0.5% Triton X-100, 1 mM phenylmethylsulfonyl-fluoride, 25 mM EDTA, 10 mg/ml leupep- Static cell adhesion assays tin and aprotinin for 30 min and were cleared by centrifu- gation at 10 000 g; 10 min at 41C. The resulting supernatant Cell adhesion to Thy-1-expressing cells Vector-transfected was incubated with the cyanogen bromide-activated sepha- CHO cells, Thy-1 transfected CHO cells or stimulated rose coupled with mAb AS02 in batch overnight at 41C. HDMECwere seeded on chamber slides (NUNC,Wiesbaden, After extensive washing with PBS, the bound proteins were Germany) and cultured for 24 h to obtain a confluent cell D eluted with 0.2 M triethylamine, pH 11.5 and neutralized monolayer. Melanoma cells were biotinylated with -bioti- with 10 mM phosphate buffer, pH 6.0. The eluted fractions noyl-e-aminocaproic-acid-hydroxysuccinimide-ester (Roche Biochemicals, Mannheim, Germany) according to the manu- containing purified Thy-1 were pooled and concentrated 5 by ultrafiltration (Amicon, Witten, Germany). Subsequently, facturer’s protocol, and 5 Â 10 of the respective cells were purified Thy-1 was labeled with D-biotinoyl-e-aminocaproic added to each well of chamber slides. After incubation for 1 acid-hydroxysuccinimide ester (Roche Biochemicals, Man- 30 min at 37 C, unbound cells were removed by several washes nheim, Germany) according to the manufacturer’s protocol. with PBS. Adherent cells were fixed with ice-cold methanol for 10 min at À201C, air-dried and blocked by incubation with PBS/1% BSA. Subsequently, adherent biotinylated cells were Flow cytometry analysis visualized by incubation with CYt3-conjugated streptavidin (Beckman Coulter, 1 : 500 in PBS/1% BSA) for 30 min at room Adherent cells were harvested and washed twice in PBS and temperature and finally washed twice with PBS. In blocking about 2 Â 105 cells were pelleted in 96-well plates. Subse- experiments, cells were preincubated with mAb against CD51, quently, cells were incubated with 1 mg of the respective mAbs CD61, Thy-1 or with IgG1 as control for 30 min, and, (mAb against CD51, CD61, Thy-1 or IgG1 isotype control). subsequently the adhesion assay was performed as described After washing twice with PBS/Gelafusal (Serumwerke above. The number of adherent cells was counted in 10 Bernburg, Germany), cells were incubated with DTAF- microscopic fields. labeled goat-anti-mouse secondary antibody (diluted 1 : 200 in PBS). Cell adhesion assays under flow conditions Alternatively, cells were incubated with biotinylated purified Thy-1 or biotinylated bovine serum albumin (BSA) as negative To detect melanoma cell adhesion under flow conditions, control for 60 min at 41Cfollowing PE-labeled streptavidin CHO.Thy-1, CHO.vec or PMA-stimulated HDMEC were (1 : 100 in PBS). cultured on glass coverslips coated with 0.1% gelatine. The Finally, cells were washed twice with PBS, and antibody or adherent cells were maintained in RPMI 1640 medium without Thy-1 binding to cells was analysed by flow cytometry (Epics serum in a heated incubation chamber (PTS GmbH, Germany) XL, Beckman Coulter, Krefeld, Germany). at 371C. Melanoma cells were labeled for 30 min at 371Cin

Oncogene Thy1/avb3 mediates cell–cell adhesion A Saalbach et al 4719 RPMI 1640 without serum containing 1 mM FURA2/AM cells were incubated with a function-blocking antibody (1-(2-(5-carboxyoxoazol-2-yl)-6-aminobenzofuran-5-oxy)-2-(20- against the dimer of avb3 (LM609). Melanoma cells were amino-50-methylphenoxy)-ethane-N,N,NX0,N0,-tetraacetate, penta- labeled with carboxyfluorescein diacetate succinimidyl ester acetoxymethyl-ester, Molecular Probes, USA), washed twice (CFDA; Molecular Probes) according to the manufacturer’s and resuspended in RPMI 1640 without serum (2 Â 105 cells/ protocol. After extensive washing of the HDMECmono- ml). Labeled melanoma cells were directed over the confluent layer, a melanoma cell suspension (2 Â 105 in 500 ml RPMI cell monolayer of CHO.Thy-1, CHO.vec or stimulated without serum) was given to the upper compartment of the HDMECin a flow chamber under a constant flow of transwell culture inserts, whereas 500 ml RPMI/10% FCS 0.15 ml/min (5 dyn/cm2) for 2 min. The monolayer was was given to the lower compartment as chemoattractance. washed with RMPI 1640 for 3 min under the same flow The transmigration was performed for 16 h at 371C,5%CO2. conditions to remove unbound cells. Finally, adherent cells Subsequently, the transwell culture inserts were removed, were counted in 20 microscopic fields ( Â 20 magnification) cells migrated into the lower compartment were resus- and representative images were recorded using a TILL- pended and fluorescein-labeled cells were counted by flow Photonics digital video imaging system consisting of a cytometry. TILL-imago CCD camera and a Polychrome II monochro- mator (TILL-Photonics, Germany) connected to a IX-71 inverted microscope (Olympus, Germany) as described pre- Statistical analysis viously by Averbeck et al. (2004). The Fura-labeled melanoma Statistical analysis was performed using the Wilcoxon t-test; cells were recorded at 340 nm/ 380 nm excitation and 510 nm P-values of o0.05 were regarded as significant. emission as well as the confluent cell monolayer was detected in transmission images (800 nm). Three independent experiments were performed. Acknowledgements The work was supported by Deutsche Forschungsge- meinschaft (SA-863-02; Bonn, Germany) and the Saxon Transendothelial migration assay Academy of Science (Leipzig, Germany). We thank HDMECwere seeded into transwell culture inserts (pore size Dr M Averbeck for the help and assistance of the realization 8 mm; Falcont, BD biosciences, Heidelberg, Germany). After of the adhesion assays under flow conditions. We thank reaching a confluent cell monolayer, cells were stimulated with Professor M Gawaz (Herzzentrum, Munich, Germany) for PMA (20 ng/ml) for 24 h to induce Thy-1 expression. To block providing us the avb3-transfected CHO cells. The M21 Thy-1, cells were incubated with the function-blocking anti- melanoma cell line and its variant lacking the av integrin Thy-1-antibody BC9 or with an isotype control antibody for (M21L) were kindly provided by Dr D Cheresh (The Scripps 30 min at 371C. For blocking avb3 on melanoma cells, these Institute, La Jolla, CA, USA).

References

Ahrens T, Assmann V, Fieber C, Termeer C, Herrlich P, Felding-Habermann B and Cheresh D-A. (1993). Curr. Opin. Hofmann M and Simon JC. (2001). J. Invest. Dermatol., 116, Cell Biol., 5, 864–868. 93–101. Felding-Habermann B, Mueller B-M, Romerdahl C-A and Anasagasti MJ, Olaso E and Calvo F. (1997). J. Natl. Cancer Cheresh D-A. (1992). J. Clin. Invest., 89, 2018–2022. Inst., 89, 645–651. Felding-Habermann B, Silletti S and Mei F. (1997). J. Cell. Aubert C, Rouge F, Reillaudou M and Metge P. (1993). Int. J. Biol., 139, 1567–1581. Cancer, 54, 784–792. Giard DJ, Aaronson SA, Todaro GJ, Arnstein P, Kersey Averbeck M, Braun T, Pfeifer G, Sleeman J, Dudda J, Martin JH, Dosik H and Parks WP. (1973). J. Natl. Cancer Inst., SF, Kremer B, Aktories K, Simon JCand Termeer C. 51, 1417–1423. (2004). Eur. J. Immunol., 34 (10), 2708–2719. Johnson J-P. (1999). Cancer Metastasis Rev., 18, 345–357. Bean MA, Bloom BR, Heberman RB, Old LJ, Oettgen HF, Ko¨ hler G and Milstein C. (1975). Nature, 256, 495–497. Klein G and Terry WD. (1975). Cancer Res., 35, 2902–2913. Kumar CC, Armstrong L, Yin Z, Malkowski M, Maxwell E, Buckley CD, Doyonnas R and Newton JP. (1996). J. Cell Sci., Ling H, Yaremko B, Liu M, Varner J, Smith EM, Neustadt 109, 437–445. B and Nechuta T. (2000). Adv. Exp. Med. Biol., 476, Byzova T-V, Kim W, Midura R-J and Plow E-F. (2000). Exp. 169–180. Cell Res., 254, 299–308. Leyton L, Schneider P and Labra CV. (2001). Curr. Biol., 11, Carey TE, Takahashi T, Resnick LA, Oettgen HF and Old LJ. 1028–1038. (1976). Proc. Natl. Acad. Sci. USA,, 73, 3278–3282. Lockshin A, Giovanella BC, Vardeman DM, Mendoza JT, Chambers A-F and Matrisian L-A. (1997). J. Natl. Cancer Quian C, Kozielski T and Stehlin Jr JS. (1985). J. Natl. Inst., 89, 1260–1270. Cancer Inst., 74, 899–903. Cheresh D-A and Spiro R-C. (1987). J. Biol. Chem., 262, Meyer T and Hart I-R. (1998). Eur. J. Cancer, 34, 214–221. 17703–17711. Muller D, Breathnach R, Engelmann A, Millon R and Craig W, Kay R, Cutler R-L and Lansdorp P-M. (1993). Bronner G. (1991). Int. J. Cancer, 48, 550–556. J. Exp. Med., 177, 1331. Murray G-I, Duncan M-E, O’Neil P, McKay J-A, Melvin Crawford J-M and Barton R-W. (1986). Lab. Invest., 54, W-T and Fothergrill J-E. (1996). Nat. Med., 2, 461–462. 122–135. Orosz P, Kruger A, Hubbe M, Ruschoff J, Von Hoegen P and Duczmal A, Schollhammer S, Katich S, Ebeling O, Schwartz- Mannel D-N. (1995). Int. J. Cancer, 60, 867–871. Albiez R and Altevogt P. (1997). Biochem. Biophys. Res. Orr F-W, Wang H-H, Lafrenie R-M, Scherbarth S and Nance Commun., 232, 236–239. D-M. (2000). J. Pathol.,, 190, 310–329. Ebeling O, Duczmal A and Aigner S. (1996). Eur. J. Immunol., Piali L, Hammel P, Uherek C, Bachmann F, Gisler R-H, 26, 2508–2516. Dunon D and Imhof B-A. (1995). J. Cell Biol., 130, 451–460.

Oncogene Thy1/avb3 mediates cell–cell adhesion A Saalbach et al 4720 Saalbach A, Anderegg U, Bruns M, Schnabel E, Herrmann Vidal M, Morris R, Grosveld F and Spanopoulou E. (1990). K and Haustein U-F. (1996). J. Invest. Dermatol., 106, EMBO J., 9, 833–840. 1314–1319. Voura E-B, Ramjeesingh R-A, Montgomery A-M-P and Siu Saalbach A, Haustein U-F and Anderegg U. (2000). J. Invest. C-H. (2001). Mol. Biol. Cell, 12, 2699–2710. Dermatol., 115, 882. Wandel E, Grahoff A, Mittag M, Haustein U-F and Saalbach A, Hildebrandt G, Haustein U-F and Anderegg U. Saalbach A. (2000). Exp. Dermatol.,, 9, 34–41. (2002). Microvasc. Res., 264, 86. Wandel E, Raschke A and Hildebrandt G. (2002). Arch. Saalbach A, Kraft R, Herrmann K, Haustein U-F Dermatol. Res., 293, 601–608. and Anderegg U. (1998). Arch. Dermatol. Res., 290, Weerasinghe D, McHugh K-P, Ross F-P, Brown E-J, Gisler 360–366. R-H and Imhof B-A. (1998). J. Cell Biol., 142, 595–607. Saalbach A, Wetzig T, Haustein U-F and Anderegg U. (1999). Wetzel A, Chavakis T, Preissner KT, Sticherling M, Haustein Cell. Tissue Res., 298, 307. UF, Anderegg U and Saalbach A. (2004). J. Immunol., 172, Smith J-W and Cheresh D-A. (1990). J. Biol. Chem., 265, 3850–3859. 2168–2172. Williams AF. (1982). Biosci. Rep., 2, 277–287 (review). Varner JA and Cheresh D-A. (1996). Adv. Oncol., Yoshizaki T, Sato H and Maruyama Y. (1997). Cancer, 79, 69–87. 139–144.

Oncogene