Recent studies have shown that the expression of αvβ3 promotes spontaneous metastasis of breast cancer to bone (Takayama et al., 2005; Sloan et al., 2006) and the functional state of integrin αvβ3 is critical in many facets of this process. For example, constitutive activation of αvβ3 on breast cancer cells (Felding-Habermann et al., 2001) promotes hematogenous dissemination of breast cancer by facilitating tumor cell arrest via interaction with platelets (Felding-Habermann et al., 2001). Previous studies have focused on expression and functional status of tumor-specific αvβ3 in the hematogenous spread of cancers through observing bone homing potential. However, the role of tumor-specific αvβ3 in processes such as cancer progression within bone and tumor induced bone remodeling has been overlooked.

This study was designed to establish the biological significance of the expression and functional state of prostate cancer-specific αvβ3 on tumor growth and tumor-induced remodeling of bone. The preparation of cells used in this study have been described previously (De et al., 2005). In short, we have generated LNCaP C4–2 prostate cancer cells that have lost β3 expression and used them to re-express (1) β3 WT, which can form heterodimers with αv to form αvβ3 in a relaxed, activatable state (αvβ3 WT cells); (2) mutant β3 D723R, which locks αvβ3 in the activated state (Hughes et al., 1996) thus bypassing inside-out signaling (αvβ3 constitutively active (CA) cells); or (3) mutant S752P, which locks αvβ3 in the relaxed state (Chen et al., 1992) resulting in non-active integrin (αvβ3 inactive cells). Flow cytometry indicated that expression levels of αvβ3 in cells re-expressing β3 integrin, whether WT, CA or inactive, were similar; however, fibrinogen binding was threefold higher for αvβ3 CA cells compared to αvβ3 WT cells (Supplementary Table 1). We have previously shown the activating nature of the β3 D723R mutation by illustrating clustering characteristics of αvβ3 CA in unstimulated C4–2 cells in suspension (De et al., 2005). We further confirmed the nature of this mutation by demonstrating constitutive phosphorylation of focal adhesion kinase following ligand binding (Supplementary Figure 1), an effect previously noted to be the result of constitutive integrin activation (Li et al., 2005). Thus, functional differences in αvβ3 between WT, CA and inactive cells are critical in this model, not expression changes among these groups.

Microcomputed tomography (microCT) was used to monitor tumor-induced changes in bone structural indices over time. To obtain baseline bone structural indices, microCT analysis was performed 1-day post-intratibial implantation of prostate cancer cells. Contralateral tibiae were injected with vehicle to account for surgically induced remodeling. We found a wide range among multiple bone structural indices across the sample population of mice (n=19) upon initial scanning of the volume of interest. For example, the trabecular thickness in μm (range=45.25–76.69, mean=60.17±7.05 (s.d.)), trabecular spacing in μm (range=46.95–130.96, mean=86.85±20.55 (s.d.)), trabecular number per mm (range=9.64–16.97, mean=12.45±1.90 (s.d.)), and trabecular bone surface area in mm² (range=3.04–8.59, mean=6.02±1.34 (s.d.)) within individual tibiae varied drastically; however, no significant difference between the tibiae (right vs left) of individual animals was found. To account for these inherent differences in bone structural indices, we expressed our data as percent differences in vehicle and tumor injected tibiae over time. Representative three-dimensional microCT images of tumor injected tibia from each group are shown in Figure 1 and changes in bone structural indices are in Table 1. As shown in Figure 2a, intratibial injection of αvβ3 (—) tumor cells resulted in a bone fraction (% bone in the volume of interest) loss of ~14% whereas the three other groups (αvβ3 WT, αvβ3 inactive and αvβ3 CA) displayed a moderate bone gain (15–28%). In addition, a 50% decrease in the number of trabeculae was observed in tibiae injected with αvβ3 (—) cells, whereas a 35% increase in the number of trabeculae was observed in tibiae injected with αvβ3 WT cells compared to mock-injected limbs (Figure 2b). Although prostate cancer cells expressing dis-regulated αvβ3 demonstrated modest elevations in trabecular number accompanied with reductions in trabecular spacing (Table 1), bone gains were not significant. Significant differences in trabecular density were present when intergroup comparisons were made for change in spacing (C4–2 — PBS control) of C4–2 αvβ3 (—) injected tibia vs that of αvβ3 WT or inactive injected tibiae, but not when comparisons between C4–2 and control injected tibiae between animals within the same group were considered. This may be due to differences in tumor growth within injected tibiae as discussed below. Taken together, our data indicate that prostate cancer cells lacking αvβ3 integrin expression promote bone loss, whereas the presence of tumor-specific αvβ3 promotes bone deposition, thus illustrating a prominent role for tumor-specific αvβ3 integrin in the development of osteoblastic bone lesions.

Tibiae were also examined histologically to determine tumor incidence (Figure 3). Cells lacking αvβ3 expression as well as those expressing αvβ3 CA or αvβ3 inactive exhibited severely impaired tumor-forming potential in bone. Overall, these observations are consistent with the findings of others. Lack of αvβ3 activity via expression as an inactive conformer results in severely impaired ligand binding and signal internalization (Chen et al., 1992), and reduces tumor-forming potential in bone (Pecheur et al., 2002). Also, our observation of impaired tumor-forming potential of CA αvβ3 expressing prostate cancer cells is in agreement with several other studies regarding the deleterious effects of CA β3 integrins on long-term tumor cell survival (Kanamori et al., 2004) and integrin function in vivo (Ruiz et al., 2001). In contrast to cells expressing dis-regulated αvβ3, C4–2 αvβ3 WT cells (Figure 3a) displaced normal marrow cells of the tibial metaphysis to fill 75–90% of the bone cavity, penetrated the cortex and invaded adjacent soft tissue. Immunohistochemical detection of prostate-specific membrane antigen (PSMA) verified the presence of C4–2 αvβ3 WT cells in the marrow space (Figure 3a). PSMA expression levels of the C4–2 lines used in our experiments were all similar as determined by immunoblotting (not shown). Despite the undetectable presence of tumor in αvβ3 inactive and CA cell-injected tibiae in the majority of animals, it is possible that a brief growth period affords protection from bone destruction as tibiae injected with dis-regulated αvβ3 cells showed resistance to bone loss compared to tibiae injected with αvβ3 (—) cells as illustrated in Table 1 and Figure 2. In sum, expression of fully functional αvβ3 WT integrin but not its dis-regulated conformers enable tumor progression in bone. The presence of αvβ3 WT on prostate cancer cells also promoted tumorigenicity and growth in the subcutis (incidence 12/12; mean=0.71±0.14 g (s.d.); P<0.01 compared to αvβ3 (—) tumors), a distinctly different environment from bone, while loss of αvβ3 (incidence: 3/5; mean=0.12±0.03 g (s.d.)), or expression of αvβ3 CA (incidence: 8/12; mean=0.11±0.06 g (s.d.); P=0.75 compared to αvβ3 (—) tumors) or αvβ3 inactive (incidence: 8/12; mean=0.18±0.07g (s.d.); P=0.20 compared to αvβ3 (—) tumors) reduced tumorigenicity. Representative tumors are shown in Figure 3b. Although our in vitro findings indicate that proliferation among the C4–2 variants used in our study are similar (not shown), we cannot conclude that the cell lines have similar growth rates in vivo. Thus, functional regulation of prostate cancer αvβ3 appears to determine tumorigenicity and is especially important in tissues rich in αvβ3 ligands such as bone.

The abundance of αvβ3 ligands in bone may explain the attraction of both prostate and breast carcinomas to this tissue. We found that the activation status of αvβ3 plays a role in the adhesion and migration to individual bone matrix proteins and total bone extracts. Cells expressing constitutively active αvβ3 promoted robust cellular adhesion to vitronectin, a reference ligand for αvβ3 and major extracellular component of mature bone (Supplementary Table 2), whereas αvβ3 WT cells required stimulation to achieve the same effect. In addition, migration to vitronectin was dependent upon the activation state of αvβ3. In accordance with a previous study (Byzova et al., 2000), migration to bone sialoprotein (BSP) occurred in a concentration-dependent manner for αvβ3 WT and CA cells, whereas migration of αvβ3 (—) and αvβ3-inactive cells was negligible (Supplementary Table 2). In the presence of blocking antibody to αvβ3 (LM609), all cell types displayed similar levels of migration, indicating that αvβ3 is the specific integrin responsible for the observed migration. Migration of tumor cells to secreted protein acidic and rich in cysteine (SPARC) mirrored that of cells to BSP with SPARC promoting migration of αvβ3 WT and CA cells but not αvβ3 (—) cells. Pretreatment of αvβ3 WT and CA cells with LM609 resulted in decreased migration by approximately 70% indicating specificity. Little effect was noted in αvβ3 (—) and αvβ3-inactive cells. To ensure that our results were not owing to the use of isolated bone-derived proteins and thus do not reflect the milieu of bone constituents that prostate cancer cells would encounter upon entrance to the skeletal compartment, we confirmed our adhesion and migration results using total bone extract. Results from adhesion and migration experiments utilizing total bone extract mirrored and support those where isolated bone-derived proteins were used.

It is possible that αvβ3-inactive cells, which fail to transmit survival signals as a result of inefficient ligand binding (Chen et al., 1992) and express low levels of growth factors such as VEGF as a result of reduced αvβ3 ligand affinity (De et al., 2005), leads to a diminished capacity for long-term survival in bone (Kitagawa et al., 2005). Conversely, αvβ3 WT cells, which readily colonize bone, efficiently bind and migrate to ligand following αvβ3 activation. Interestingly, our adhesion and migration data indicated that αvβ3 CA cells should thrive within bone; however, αvβ3 CA cells exhibited a severely diminished capacity for sustained in vivo bone colonization. Our previous studies indicated that bone metastatic prostate cancer, in comparison to normal matched prostate tissue, expressed dramatically elevated levels of activated αvβ3 (De et al., 2003). This activation, however, was due to a feedback mechanism wherein VEGF signaling initiates inside-out αvβ3 activation leading to enhanced ligand-binding affinity. A constitutively active mutant of αvβ3 would bypass the need for activation and could not dynamically regulate cell adhesion, a condition, which may be deleterious for long term in vivo persistence in bone (Schwartz and Ginsberg, 2002). It is possible that constitutively active αvβ3, in the context of abundant matrix bound ligands such as that in bone, may lead to excessive, unregulated outside-in signaling that could be detrimental to tumor growth. Thus, expression of inactive αvβ3 by prostate cancer cells may inhibit sustained bone colonization and proliferation in bone tissue, whereas expression of constitutively active αvβ3 by prostate cancer cells may promote metastasis and initial engraftment within bone but inhibit subsequent growth.

Supplement