Comparative analysis of integrins in vitro and in vivo in uveal and cutaneous melanomas.

Changes in integrin expression have been shown to be important for the growth and metastatic capacity of melanoma cells. In this study, we have examined the expression of alphav integrins by three uveal and four cutaneous malanoma lines. No lines expressed alphavbeta6 and only TXM13, a cutaneous line, expressed alphavbeta8. All lines expressed alphavbeta5 and alphavbeta3 (four out of four cutaneous, two out of three uveal) or avpl (OM431, an uveal line). Thus, OM431 is the second uveal melanoma we have described that expresses alphavbeta1 and this, we report again, functions as an alternative vitronectin/fibronectin receptor. Subcutaneous growth of cell lines in athymic mice correlated with an alphavbeta3-positive, alphavbeta1 -negative phenotype. Analysis of clinical material from cutaneous melanoma showed that although alphav expression was increased in 88% of metastases, this could not all be explained by up-regulation of alphavbeta3, with only 2 out of eight skin metastases expressing this heterodimer. Using antibody SZ.21, which as we report here works in archival material, only 1 out of 15 uveal metastases expressed detectable beta3. Thus, acquisition of alphavbeta3 expression, which has been implicated in cutaneous melanoma progression, may not be required for development of metastases from uveal melanoma or indeed for skin, as distinct from lymph node, metastases of cutaneous melanoma.

the integrin whose levels of expression have correlated most consistently with progression is the classical vitronectin receptor avP3 (Cheresh and Spiro, 1987). Thus, Albelda and colleagues (1990) noted that the ,3 subunit was only detected on VGP and metastases of cutaneous melanoma. This study almost certainly documented the appearance of avP3 as similar findings were described using an avP3-specific antibody in which it was noted that expression of this heterodimer was higher in cutaneous metastases than expression on less advanced tumours (Danen et al, 1994;Si and Hersey, 1994). These data suggest that av,3 may play an active role in the progression of cutaneous melanoma.
In vitro studies have supported this possibility. The av-deficient M21-L human melanoma cell line grew very poorly in nude mice compared with either the av-positive parental line or a line in which av expression was restored by transfection with a fulllength av cDNA (Felding-Habermann et al, 1992). Treatment of animals with the av-blocking antibody 17E6 inhibited the growth of the av-positive M21 melanoma cell line . Earlier studies by Boukerche and colleagues (1994) showed that the co-injection of antibody LYP18, which cross-reacts with both aIIb133 and avP3, inhibited tumour growth of the human melanoma cell line M3Dau (Boukerche et al, 1989). We have shown that the ability of human melanoma cell lines to form subcutaneous tumours in athymic nude mice correlated with levels of expression of avP3 (Marshall et al, 1991). However, av,3negative cutaneous melanoma cell lines have metastasized in nude mice (Boukerche et al, 1994;Danen et al, 1995), suggesting that expression of this heterodimer is not obligatory for malignant behaviour.  The av,3-integrin is not the only avp-heterodimer expressed by melanoma cells. We found that a uveal melanoma-derived cell line, which lacked av,3, expressed av,l, which functioned as a receptor for vitronectin, fibrinogen and fibronectin (Marshall et al, 1991). As there are no reagents that specifically recognize the xv,Bl-heterodimer, confirmation of avpl expression was by immunoprecipitation with antibodies to av followed by immunological analysis of the co-precipitated PI-sized subunit. Thus, the frequency of expression of av,l in either cutaneous or ocular melanoma is unknown.
Five different avf-heterodimers have been described to date: av[B1, avP3, av,B5, avP6 and av[8 (Hynes, 1992). The relative expression of these various heterodimers by cells derived from a single histological origin has not been studied in a systematic fashion. In the present study, we have examined integrin expression, with particular emphasis on av,-heterodimers, in a panel of cell lines derived from uveal and cutaneous melanomas and in clinical material from both types of cancer.

MATERIAL AND METHODS Cell lines and antibodies
The antibodies used in this study are detailed in Table 1. Six uveal melanoma lines were analysed: V(+)B2 and V(+)D9H are high avfl-expressing derivatives of VUP (Marshall et al, 1995)

Immunoprecipitation
The avo-heterodimers expressed by four uveal melanoma-derived cell lines, V(+)B2, OM431, OCM-I and SP6.5, were analysed by immunoprecipitation of surface-iodinated, NP40-detergent-lysed cell extracts as described previously (Marshall et al, 1991  were added in 50 gl volumes (1-2 x 104 cells per well) to quadruplicate wells. The plates were incubated for 60 min at 37°C before unattached cells were removed by gently flicking-out well contents and washing the plates twice by total immersion in a bath of PBS supplemented with calcium chloride (1 mM) and magnesium chloride (0.5 mM). The per cent adhesion was calculated from the residual radioactivity (c.p.m.) associated with the wells. Background (adhesion to BSA-coated wells) was usually < 2% of input and was subtracted from all results. In some experiments, extracellular matrix (ECM)-coated plates were placed on ice and 25 gl of anti-integrin antibodies were added to the wells before the addition of 25 ,ul volumes of twofold concentrated cells. After 10 min incubation on ice the plates were placed at 37°C for 60 min and the assay continued as described above.

Assessment of tumorigenicity
A total of 1-2 x 106 melanoma cells was injected subcutaneously into the right flank of athymic nude mice. Mice were monitored weekly for up to 12 months for the appearance of palpable tumours.

Immunohistochemical analysis of melanoma tissues
Fresh material from uveal, cutaneous and local nodal tissue was obtained at surgery, snap-frozen in liquid nitrogen and stored subsequently at -70°C. Tissue was examined from 21 benign cutaneous naevi (only one of which had histological features of atypia, the rest were dermal cellular naevi), nine cutaneous melanomas in radial growth phase, eight cutaneous melanomas that had entered vertical growth phase, eight cutaneous metastases and eight lymph node metastases. Material was obtained from 13 primary uveal ocular melanomas at enucleation. The uveal melanoma lesions varied histologically, being of both spindle and epithelioid type as well as a mixture of these cells. No frozen tissue from metastatic lesions was available but paraffin-embedded archival material representing 15 different metastases from six individual patients with uveal melanoma metastases was examined.
For the fresh tissue, cryostat sections (5 ,im) were taken on to poly-L-lysine coated slides, air-dried and stored at -20°C. Primary antibodies were applied to sections for 60 min at room temperature. After gentle washing in PBS, a standard peroxidase/anti-peroxidase technique was used according to the manufacturer's instructions (Vectastain Kit, Vector Laboratories, Peterborough, UK). Bound antibody was detected with 3-amino 9-ethyl carbazoyl (AEC), which gives rise to a red chromogen. For archival material, slides were dewaxed and endogenous peroxidase blocked with 0.05% hydrogen peroxide in methanol for 15 min. (For detection of P3, slides were placed into boiling 0.O1M sodium citrate buffer pH 6.0 in a pressure cooker and put under pressure for 2 min. The buffer was flushed away and slides washed in tap water.) Sections were blocked with 20% normal rabbit serum for 15 min before adding the primary antibodies 4B7 (anti-PI; undiluted supematant) or SZ.21 (anti-P3; 1:100 dilution in TBS) for 60 min. After washing in TBS, a standard avidin-biotin-chromogen method was used, and slides developed by diaminobenzidene to produce an insoluble brown end-product.

Expression of integrins in melanoma cell lines determined by flow cytometry
Data from a series of individual experiments are summarized in Table 2. Expression of av and 31 was seen in all lines. Although the avp3-integrin was expressed by all cutaneous melanoma lines, only two out of four uveal melanoma lines expressed it (OCM1 and SP6.5). Of the eight lines detailed in Table 2, none expressed significant amounts of either avj6 or av8, except TXM13, which appeared to express avP8. Expression of av,B5 was relatively low on all cell lines although level of expression of this integrin may be variable.
Immunoprecipitation of uveal melanoma av-integrins SP6.5, OCM1 and OM431 uveal lines and, for comparison, the VUP-derived subline V + B2 were analysed by immunoprecipitation. Figure 1 shows that both SP6.5 and OCM-1 express av03 whereas V + B2 and OM431 do not. Immunoprecipitation with antibody to av (13C2) coprecipitated a (1-sized band from both OM431 and V(+)B2 but not from the SP6.5 or OCM-I cell lines. It appears that OM431 is the second uveal melanoma cell line that we have shown to lack avP3 but to express avil (Marshall et al, 1991). Figure 1 also confirms the flow cytometry data ( Table 2) that avP5 is expressed weakly by OM431, SP6.5, OCM-I and, as reported previously, V + B2 (Marshall, 1995).
British Journal of Cancer (1998) 77(4) The avil expressed by V(+)B2 binds to vitronectin and cooperates with a5p1 to bind to fibronectin (Marshall, 1995). To examine whether the avpl expressed by OM431 cells manifested a similar range of activities 51Cr-labelled cells were allowed to adhere to vitronectin or fibronectin in the presence or absence of various anti-integrin antibodies. For comparison, the uveal melanoma lines VUP (low av3l-expressing) and V(+)B2 (high av,Bl-expressing) as well as SP6.5 (avil-negative, av,B3-positive) were also studied. Figure 2 shows that, in the presence of a class-matched negative control antibody (14E2), all four cell lines bound to vitronectin, a binding which was reduced by > 80% in the presence of the av-blocking antibody, 17E6. The adhesion to vitronectin by SP6.5 appeared to be mediated principally by avP3and avp5-dependent mechanisms as shown by the inhibition by the antibodies LM609 (39.6% inhibition) and P3G2 (14.4% inhibition) (Figure 2). In contrast, the anti-av,3 antibody LM609 had no effect on the adhesion to vitronectin of the av,3-negative lines VUP, V(+)B2 and OM431.  The adhesion of OM431 to vitronectin was av-dependent as the presence of an av-blocking antibody (17E6) reduced adhesion by 86% ( Figure 2). The antibodies P3G2 (avP5-blocking) and P4C1O ((51-blocking) inhibited adhesion to vitronectin by 48.1% and 29.2% respectively ( Figure 2). These data show that OM431 binds to vitronectin via avP5and axv,1-dependent mechanisms, although it appears that the avP5 heterodimer may be the dominant vitronectin receptor.
Adhesion of V(+)B2 cells to vitronectin was inhibited by 17E6 (88.6%) and P4C1O (82.7%) but not by P3G2 (anti-av,5) ( Figure 2). However, the combination of P3G2 and P4C1O inhibited adhesion of V(+)B2 by 94.2%, suggesting that although binding of V(+)B2 to vitronectin is mediated predominantly via avpil the low level of avP5 expressed also functions as a vitronectin receptor. Adhesion to vitronectin by the low av1expressing parental line VUP was inhibited by 78.0% by antibody P3G2 (anti-avP5) and 14.1% by antibody P4C10 (anti-1), whereas the combination of P3G2 and P4C1O inhibited completely adhesion to vitronectin. These data appear to suggest that, in contrast to V(+)B2, avP5 is the major vitronectin receptor on the VUP cell line.
All of the uveal melanoma cell lines tested bound well to fibronectin, as illustrated in Figure 3. We have shown previously that V(+)B2 binds to fibronectin through the cooperative action of avpl and a5PI (Marshall, 1995). Adhesion of VUP and OM431 appeared to be via a similar mechanism. Thus, PID6 (anti-aS) when combined with 17E6 reduced adhesion to fibronectin of VUP and OM431 by 62.8% and 54.0% respectively. The inability of PiB5 (anti-a3) to affect adhesion to fibronectin, even when used in combination with 17E6, suggests that a3,B1 is not a major receptor for fibronectin in these unveal melanoma cell lines. Therefore, it appears that OM43 1, VUP and V(+)B2 adhere to fibronectin via an avfl/a5pl-dependent mechanism. The adherence of SP6.5 cells to fibronectin also appeared to utilize a combination of integrins ( Figure 3). Thus, the only single antibody to inhibit adhesion of SP6.5 to fibronectin significantly was 17E6 (29.1% inhibition). Combination of 17E6 30 .°20 to c 10) 10 (U with PiB5 or P1D6 caused a further inhibition of adhesion to 39.9% and 43.4% respectively. However, maximum inhibition (80.4%) of adhesion to fibronectin required the co-incubation of 17E6, PlD6 and P4C1O antibodies (Figure 3). Table 3 details the ability of six uveal and three cutaneous melanoma cell lines to form progressively growing subcutaneous xenografts in athymic nude mice. The VUP line and the two high avplexpressing derivatives V(+)B2 and V(+)D9H failed to form tumours. Two of eight mice inoculated with OM431 developed slow growing tumours, which reached 10 mm diameter after 210 and 330 days post inoculum. The remaining cell lines OCM-1, SP6.5, Mel 8, Mel 17 and XP44 were highly tumorigenic, forming tumours in 50-100% of animals (Table 3). Thus, the av,B3-positive uveal melanoma cell lines OCM-I and SP6.5 were more tumorigenic than the avil-positive uveal melanoma lines VUP, V(+)B2, V(+)D9H and OM43 1, which were either poorly or non-tumorigenic.

Expression of integrins by cutaneous and uveal melanoma tissues
Cryostat sections of cutaneous and uveal melanoma tumour tissues were analysed by immunohistochemistry for the expression of a2, a3, a4, aS, a6, av, av,B3 and av,B5 (Table 4).
The major integrin subunits expressed in primary uveal melanoma were a3 and av, which were present on 13 out of 13 samples. The integrin av,5 was detected on 11 out of 12 tumours and was possibly the major av,B-heterodimer present as av,3 was not found on any of the 13 tumours analysed. Analysis of 15 uveal melanoma metastases showed that only one of the tumours was 03positive. An internal positive control was often present on these sections as blood vessels stained positively for 03 (data not shown).
In contrast, expression of , 1 was detected in 8 out of 15 uveal melanoma metastases in this small series of archival material.  In cutaneous melanoma, the expression of a3 and av was not detectable on benign lesions but was expressed on almost all of the metastases. The a4and a5-subunits were absent on primary cutaneous lesions but were present on seven out of nine and five out of nine of lymph node metastases, respectively, but on only one out of eight skin metastases (Table 4). Expression of avP3 was also confined to metastases being detected on five out of nine lymph node and two out of eight skin metastases. In contrast, expression of avP5 was higher on the primary lesion (six out of ten naevi, four out of seven VGP) compared with metastases (one out of eight skin metastases); 10 out of 17 vs 1 out of 8 (P . 0.04, Fisher's exact test).

DISCUSSION
For malignant cells to metastasize they must decrease their attachment to neighbouring cells. In addition, as maximum motility requires intermediate adhesiveness (Palacek et al, 1997) they may also require reduced adhesion to underlying ECM proteins. This may partly explain why development of breast and colorectal cancer is often associated with reduced or aberrant expression of a2, a3 and a6 (for references see Gui et al, 1997). However, ligation of integrins to the ECM can generate survival signals (reviewed by Meredith and Schwartz, 1996) and, thus, increased expression or de novo expression of specific integrins could also promote cancer. Cutaneous melanoma is an example of a cancer in which tumour progression correlates with a net gain in several integrins, most notably av,B3 (Albelda et al, 1990;Danen et al, 1994;Si and Hersey, 1994;Natali et al, 1997). Although 10% of melanoma occurs in the eye, the majority in the uvea (Shields and Shields, 1992), very little has been documented on the integrins expressed by these tumours. We have therefore compared the expression of integrins by uveal vs cutaneous melanoma cell lines and tissues.
Using flow cytometry (Table 2) and immunoprecipitation ( Figure 1) we now show that OM43 1 is the second uveal melanoma that lacks av,3 but expresses av,B1 a vitronectin/fibronectin receptor (Figures 2 and 3). However, expression of xv,31 is not universal for all uveal lines as it was not detected in SP6.5 or OCM-1, which instead express avP3 (Figure 1).
Analysis of the tumorigenicity of the cell lines (Table 3) revealed that the avP3-expressing lines, regardless of uveal or cutaneous origin, were highly tumorigenic, forming xenografts 50-100% of inoculated animals. In contrast the av,B1-positive lines were either poorly or non-tumorigenic. In addition, using flow cytometry we have measured expression of av,5, avP6 and avP8 on the nine cutaneous melanoma cell lines already examined for av,B3 expression (Marshall et al, 1991). Together with the data reported here, we have found that formation of subcutaneous xenografts by 17 human (cutaneous and uveal) melanoma lines correlates with an avP3positive, av,B1-negative phenotype. Thus, our data may suggest that loss of av Il by the VUP and OM431 lines may promote xenograft formation. We have found no correlation between tumorigenicity and expression of avP35 or avP8 (avP6 was not expressed by melanoma cell lines; Marshall and Hart, 1996).
In a recent study, Natali and colleagues (1997) failed to detect any av-integrins on eight uveal melanomas. However, our analysis of uveal melanoma clinical material confirmed a previous report (ten Berge et al, 1993) that primary uveal melanomas appear to be avP3 negative, av,B5 positive. However, these workers also showed that two of three metastases expressed av,B3, which we did not observe in our own series. Using the antibodies SZ.21 (anti-53) and 4B7 (anti-p1), which as reported here work on paraffin-embedded material, only 1 out of 15 uveal melanoma metastases were 33-positive, whereas 8 out of 15 were P 1 positive.
Thus, unlike cutaneous melanoma, we found no positive correlation between expression of avP3 and uveal melanoma metastases. We detected integrin av[3 on five of nine lymph node and two of eight skin metastases (Table 4), although it should be noted that Natali et al (1997) did not note a difference in expression of this integrin between these types of metastases. In addition unlike previous reports (Albelda et al, 1990;Si and Hersey, 1994), only one out of the eight vertical growth phase lesions was found to be av,3 positive.
Expression of avP5 was higher in the primary lesions (six out of ten naevi, four out of seven VGP) than on the metastases (one out of six skin metastases) in agreement with the data of Danen and colleagues (1995). However, although most metastases from cutaneous melanoma had increased levels of av integrins, this was not always accounted for by a commensurate increase in either avP5 or av,3 (Table 4), suggesting that non-,3 av-integrins were up-regulated.
Like others, we found that a3,B1 (Natali et al, 1993), a4p1 (Schadendorf et al, 1993), and ac51 (Danen et al, 1994) show an increased expression on more advanced stages of cutaneous melanoma; particularly on the metastases (Table 4). It may be significant that expression of a4 and aS was detected on seven out of nine and five out of nine lymph node metastases, respectively, but only on one out of eight skin metastases (Table 4). These data could suggest that expression of these integrins may increase the propensity of melanoma cells to colonize lymph nodes partly, perhaps, by using a4p1 to adhere to VCAM-1 (Mould et al, 1994).
The observation by several groups (Albelda et al, 1990;Danen et al, 1994;Si and Hersey, 1994;Natali et al, 1997) and ourselves that av,3 expression is increased in the later stages of cutaneous melanoma is consistent with this heterodimer having an active role in malignancy. Several functions have been ascribed to avP3 that may contribute to such a mechanism. Thus, it has been reported that avP3 may cause retention of melanoma cells in lymph nodes through binding to lymph node vitronectin (Nip et al, 1992), whereas ligation of avP3 has resulted in increased expression of the metalloproteinase MMP2 (72 kDa type IV collagenase) (Seftor et al, 1992). Recently, Brooks and colleagues (1996) have reported that av,B3 bound to, and thus located, MMP2 at the surface of invasive cells. Moreover, avP3, which is not normally a receptor for interstitial collagen, binds to denatured (for example collagenase-digested) collagen type I and in doing so may provide survival signals to melanoma cells (Montgomery et al, 1994). Thus, in addition to its role as a major adhesive and migratory integrin (Marshall and Hart, 1996), av,3 may have other functions during melanoma development.
In conclusion, an avP3-positive, av,1-negative phenotype is associated with the capacity of cutaneous or uveal melanoma cell lines to form xenografts in nude mice. However, in clinical material, although av,B3 was expressed by > 50% nodal metastases, the majority of uveal melanoma metastases and cutaneous melanoma skin metastases lacked detectable avP3, suggesting that expression of this integrin may not be a prerequisite for formation of either of these melanoma lesions.