A 3D matrix platform for the rapid generation of therapeutic anti-human carcinoma monoclonal antibodies

Significance

To select for novel monoclonal antibodies (mAbs) with anticancer activity, an experimental platform was established wherein human breast cancer cells were embedded in 3D collagen matrices and used as an immunogen to generate mAb libraries. Fifteen mAbs capable of inhibiting carcinoma cell growth in vitro were generated. A single function-blocking mAb was selected for further analysis and then validated as a potent inhibitor of carcinoma cell behavior in vivo. The target antigen was identified as the α₂ subunit of the α₂β₁ integrin, a major type I collagen-binding receptor whose expression was confirmed in tissues of patients with breast cancer. These findings describe a new discovery platform that allows for the rapid selection of function-blocking antibodies and identify α₂β₁ as a potential target in carcinomatous states.

Abstract

Efforts to develop unbiased screens for identifying novel function-blocking monoclonal antibodies (mAbs) in human carcinomatous states have been hampered by the limited ability to design in vitro models that recapitulate tumor cell behavior in vivo. Given that only invasive carcinoma cells gain permanent access to type I collagen-rich interstitial tissues, an experimental platform was established in which human breast cancer cells were embedded in 3D aldimine cross-linked collagen matrices and used as an immunogen to generate mAb libraries. In turn, cancer-cell–reactive antibodies were screened for their ability to block carcinoma cell proliferation within collagen hydrogels that mimic the in vivo environment. As a proof of principle, a single function-blocking mAb out of 15 identified was selected for further analysis and found to be capable of halting carcinoma cell proliferation, inducing apoptosis, and exerting global changes in gene expression in vitro. The ability of this mAb to block carcinoma cell proliferation and metastatic activity was confirmed in vivo, and the target antigen was identified by mass spectroscopy as the α₂ subunit of the α₂β₁ integrin, one of the major type I collagen-binding receptors in mammalian cells. Validating the ability of the in vitro model to predict patterns of antigen expression in the disease setting, immunohistochemical analyses of tissues from patients with breast cancer verified markedly increased expression of the α₂ subunit in vivo. These results not only highlight the utility of this discovery platform for rapidly selecting and characterizing function-blocking, anticancer mAbs in an unbiased fashion, but also identify α₂β₁ as a potential target in human carcinomatous states.

In mammalian systems, a specialized form of extracellular matrix (ECM), termed the basement membrane, normally separates epithelial cells from the underlying type I collagen-rich interstitial matrix (1, 2). Thus, in mature animals and under physiologic conditions, the epithelium does not establish stable physical contacts with interstitial tissues (1, 2). In contrast, in neoplastic states, transformed epithelial cells (i.e., carcinomas) dissolve the intervening basement membrane barrier and establish adhesive interactions with the newly exposed type I collagen fibrillar network (1–5). As carcinoma cells begin to infiltrate the interstitial matrix, they rapidly adapt themselves to their 3D environment and initiate the proliferative phenotypes that define tumor progression at both primary and metastatic sites (2, 6, 7). Indeed, emphasizing the importance of the tumor–ECM interface, carcinoma cells do not simply use the surrounding interstitial matrix as a passive substrate, but actively promote increased type I collagen deposition within the peritumoral microenvironment as a means of further enhancing invasive activity, local growth, and cancer stem cell formation (7–12).

Despite the importance of the carcinoma cell–type I collagen interface in vivo, therapeutic interventions that directly interfere with the specific cell–ECM interactions operating within this specialized tumor milieu have yet to be identified. Traditionally, new therapeutic agents are developed by identifying a preferred candidate and then generating a specific inhibitor for a targeted effector (13). In this regard, humanized monoclonal antibodies (mAbs) have been established as important players in the therapeutic armamentarium (13, 14). Strategies that allow for the rapid identification and validation of new targets remain problematic, however (13).

Cogent arguments have been forwarded regarding the utility of phenotypic screens for the purpose of identifying new targets in an unbiased fashion (13, 15). Nevertheless, leveraging this approach requires the engineering of in vitro conditions that faithfully recapitulate carcinoma cell behavior in vivo, so that targets can be identified and their functional contribution assessed rapidly before in vivo testing. To this end, here we describe a novel screening platform wherein human carcinoma cells are cultured within aldimine cross-linked, 3D type I collagen hydrogels similar to those found at invasive sites in vivo (16), and the cancer cell–matrix composite used to generate a library of mAbs. In turn, the mAbs are then screened for their ability to suppress carcinoma cell proliferative responses under 3D growth conditions. Validating the utility of our in vitro approach, selected mAbs are shown to inhibit carcinoma cell proliferation and metastatic activity in xenograft models in vivo. Finally, using a combination of immunopurification, mass spectroscopy, and peptide mapping, the target antigens can be identified and their expression confirmed in human cancer tissues. Taken together, these findings not only establish a platform that allows for the rapid identification of function-blocking mAbs and their targets, but also provide key insights into regulation of the carcinoma cell–ECM interface within the in vivo setting.

Results

Characterization of Function-Blocking mAbs Directed Against MDA-MB-231 Carcinoma Cells.

MDA-MB-231 cells are a well-characterized, triple-negative breast carcinoma cell line whose gene expression profile closely recapitulates that found in human breast cancer tissues (17–19). Furthermore, in a manner similar to human carcinomas expanding in vivo, the cell line undergoes rapid proliferative and tissue-invasive responses when cultured within 3D type I collagen hydrogels in vitro (16, 20, 21). As such, MDA-MB-231 cells were embedded in covalently cross-linked networks of mouse type I collagen with an elastic modulus similar to that found in normal breast tissue, ∼150 Pa (11). After a 3-d culture period, the human carcinoma cell–mouse matrix composite was then recovered and used as an immunogen to generate a panel of ∼2,500 mAbs (Fig. S1). To identify MDA-MB-231–reactive clones, whole-cell–based ELISAs were then performed with ∼350 of the mAbs that scored positive in initial screens. Each of the reactive clones was then expanded, and the individual mAbs were tested for their ability to inhibit the proliferative responses of MDA-MB-231 cells in 3D culture (Fig. S1).

When embedded in 3D type I collagen hydrogels, MDA-MB-231 cells rapidly changed their morphology from a spherical to a bipolar, mesenchymal cell-like phenotype over the first 48 h before the initiation of proliferative responses (Fig. 1 A and B). Among the 15 mAbs displaying inhibitory activity in our initial screens, clone 4C3 was one of the most potent IgG1 class antibodies identified, displaying an ability to almost completely block MDA-MB-231 cell shape changes and proliferation in 3D collagen (Fig. 1 A and B). Moreover, inhibitory activity was not limited to a “preventative” protocol wherein mAb 4C3 was added at the start of the 3D culture period; addition of the inhibitory mAb at 4 d after the start of the culture period similarly inhibited carcinoma cell proliferation, with an IC₅₀ of ∼0.05 µg/mL (Fig. 1 B and C). Furthermore, 4C3 not only blocked MDA-MB-231 proliferative responses, but also initiated apoptosis in 3D culture as assessed by caspase 3 and 7 activation (Fig. 1D). In contrast, when cultured under standard 2D conditions atop tissue culture-treated plastic substrata or within 3D Matrigel, an ECM extract that neither recapitulates the structure of normal basement membranes structure nor that of the interstitial matrix (1, 2), mAb 4C3 exerted no inhibitory effects on MDA-MB-231 cell function (Fig. S2). Interestingly, the antiproliferative activity of mAb 4C3 is not confined to MDA-MB-231 carcinoma cells; similar inhibitory effects can be observed with human squamous cell carcinoma, ovarian carcinoma, and fibrosarcoma cell lines in 3D culture (Fig. S3).

Fig. 1.

In vitro activity of mAb 4C3. (A) MDA-MB-231 cells were seeded in 3D collagen matrices in the absence or presence of mAb 4C3 (10 µg/mL). Cultures were evaluated by phase-contrast or confocal microscopy (red) at day 0 and day 4. (B) MDA-MB-231 cells were seeded in 3D collagen in 12-well plates (5 × 10⁴ cells/well) with mAb 4C3 (10 µg/mL) added at day 0 or day 4 (red arrow). At indicated times, cell numbers were determined by hemocytometry. Results are expressed as mean ± SEM (n = 3). *P < 0.05. (C) MDA-MB-231 proliferation was assessed by relative ATP levels after 48 h of treatment with indicated mAb 4C3 concentrations. Vehicle controls or control IgG1 mAb were without effect. Adhesion was assessed by allowing MDA-MB-231 cells to attach to collagen gels for 1 h, followed by staining with crystal violet. Results are expressed as mean ± SEM of three experiments. (D) Relative levels of caspase 3 and caspase 7 activity were determined for MDA-MB-231 cells embedded within 3D collagen gels for 72 h in the presence of the indicated concentrations of mAb 4C3 added 24 h before the assay. Vehicle control or control IgG1 were without effect. Results are expressed as mean ± SEM of three experiments. (E) Gene Ontology terms identifying cellular processes after mAb 4C3 (10 µg/mL) treatment of 3D-embedded MDA-MB-231 for 48 h. (F) Heat map of genes regulating cell cycle after mAb 4C3 treatment is shown.

mAb 4C3 Exerts Global Effects on the MDA-MB-231 Transcriptome.

In an effort to identify the potential signaling networks impacted by mAb 4C3, we cultured MDA-MB-231 cells in 3D type I collagen hydrogels in the presence of either a control IgG1 or mAb 4C3 for 48 h (i.e., before the initiation of proliferative responses), and RNA was harvested for gene expression profiling. Under these conditions, mAb 4C3 exerted global effects on gene expression, with almost 1,200 unique transcripts affected (i.e., 172 up-regulated and 1,004 down-regulated transcripts, respectively, using a 2.0-fold cutoff). Consistent with its effect on MDA-MB-231 proliferation, Gene Ontology analysis revealed that mAb 4C3 treatment elicited major alterations in cell cycle, regulation, RNA processing, and cell division-related programs (Fig. 1 E and F). Taken together, these results identify mAb 4C3 as a potent regulator of MDA-MB-231 cell function within the confines of a type I collagen-rich ECM.

mAb 4C3 Prevents Postextravasation Carcinoma Growth In Vivo.

In our in vitro model, embedded carcinoma cells are individually surrounded by a network of type I collagen fibrils, a scenario similar to that encountered when circulating tumor cells extravasate from vascular or lymphatic beds and enmesh themselves within the perivascular interstitial matrix (1–3, 6, 12). To examine the inhibitory potential of mAb 4C3 in a postextravasation program directly, we used a live embryonic chick xenograft model that faithfully recapitulates carcinoma cell behavior in mouse models (22, 23). As shown in Fig. 2A, the chick chorioallantoic membrane vasculature is readily visualized by confocal laser microscopy. Furthermore, using second harmonic generation to image type I collagen fibrils in situ (24) revealed that blood vessels were uniformly invested by a dense collagenous network (Fig. 2B).

Fig. 2.

Anticancer activity of mAb 4C3 in the chick xenograft model. (A) Vasculature of the chick CAM as visualized after GFP-isolectin B4 (green) infusion by confocal laser microscopy. (B) Perivascular interstitial collagen (blue) in the 11-d-old chick CAM as assessed by second harmonic generation. (C and D) RFP-labeled MDA-MB-231 cells and mAb 4C3 (0.8 mg/embryo), a vehicle control, or control IgG1 (0.8 mg/embryo) were introduced i.v. into the chicken embryos. After a 5-d incubation period, tissues were harvested and evaluated by florescent microscopy for the presence of MDA-MB-231 cells (orange) and blood vessels (green). Results are representative of three or more experiments performed. (E and F) Chicken embryos were inoculated i.v. with 2.5 × 10⁵ luciferase-expressing MDA-MB-231 cells at 5 d before harvest. Inhibitor, vehicle, or control IgG1 was coadministered with the carcinoma cells (day 0) or 24 h later (day 1). For imaging, eggs were injected i.v. with luciferin 10 min before retrieval of the lower CAM and imaged for bioluminescence and quantified. Results are expressed as mean ± SEM (n = 3). *P < 0.05.

As such, fluorescent-tagged MDA-MB-231 cells were injected into the host vasculature of 11-d-old immuno-incompetent chicken embryos in tandem with a control IgG1 or mAb 4C3, and postextravasation growth was monitored. After a 6-d culture period in vivo, extravasated MDA-MB-231 cells initiated proliferative activity in close association with the chick vasculature (Fig. 2C). In contrast, in the presence of mAb 4C3, MDA-MB-231 cell proliferation was markedly inhibited, where tumor colony formation was readily monitored by both visual inspection and quantification of luminescent signals using luciferase-tagged carcinoma cells (Fig. 2 C–F).

To rule out the possibility that 4C3 blocks proliferative responses by interfering with MDA-MB-231 extravasation itself, we injected carcinoma cells into the chick vasculature, and after a 24-h period that allows for the completion of extravasation (23), introduced mAb 4C3 intravascularly. Even under these conditions, mAb 4C3 exerted potent inhibitory effects, equivalent to those obtained when the antibody was introduced at the start of the in vivo assay (Fig. 2F).

Antimetastatic Activity of mAb 4C3 in a Mouse Xenograft Model.

Unlike humans, whose mammary tissues are dominated by type I collagen, the mouse mammary gland contains only small amounts of type I collagen that is largely confined to periductal regions, thus rendering mouse xenograft orthotopic models less useful for analyzing carcinoma cell–type I collagen matrix interactions (25). Alternatively, the organic matrix of mouse bone, like that of humans, is composed largely of type I collagen (17, 26–28). Furthermore, bone is a frequent site of breast cancer metastatic activity in human disease (17). As such, after intracardiac injection, we assessed the ability of luciferase-tagged MDA-MB-231 to generate bone metastatic lesions in nude mouse recipients in the presence of control IgG1 or mAb 4C3 by in vivo imaging as well as micro-CT analyses over a 28-d assay period.

When treated with twice-weekly dosages of 10 mg/kg of the control mAb, MDA-MB-231 cells generated large tumors in the mandible, hindlimb, and spine of inoculated mice (Fig. 3 A and B). In contrast, in the mAb 4C3-treated group, carcinoma growth in the mandible and hindlimb was impaired, with significant inhibitory effects recorded in vertebral metastases where bone-erosive lesions were readily observed in micro-CT scans of the control antibody-treated group (Fig. 3 A–C). Whereas ∼50% of the control antibody-treated mice required euthanization due to spinal cord compression and resulting limb paralysis, fewer than 20% of the mAb 4C3-treated mice were similarly affected, consistent with the ability of mAb 4C3 to block the progression of bony metastases (Fig. 3D).

Fig. 3.

MDA-MB-231 bone metastasis model. (A and B) Luciferase-expressing MDA-MB-231 cells (1 × 10⁵) were injected into the left ventricle of nude mice with 10 mg/kg of a control IgG1 or mAb 4C3 twice weekly for 4 wk, and tumor progression was monitored by bioluminescent imaging. Representative images shown in A were obtained at 3 wk postinjection. Results are expressed as mean ± SEM of control mAb-treated (n = 9) and mAb 4C3-treated (n = 8) mice. *P < 0.05. (C) At termination, bioluminescent imaging of spinal fields (black arrow) was assessed with guided X-ray analysis of affected areas in the vertebral column (white arrows), evaluated by microCT. (D) Effect of control IgG1 versus mAb 4C3 on the development of hindlimb paralysis during the 4-wk treatment period. Results are expressed as mean ± SEM of control IgG1-treated (n = 38) and mAb 4C3-treated (n = 29) mice. *P < 0.05.

Identification of the mAb 4C3 Target Antigen and Its Expression in Human Breast Cancer Bone Metastases.

To identify the target antigen recognized by mAb 4C3, we applied whole-cell lysates of MDA-MB-231 cells to immunoaffinity columns constructed using the purified antibody as the capturing agent. After antigen recovery, a major bond of ∼150 kDa was isolated and submitted for mass spectrometry analysis after trypsin fragmentation (Fig. 4A). Bioinformatic analysis of the generated fragments identified the target antigen as the integrin subunit, alpha 2 (α₂) (29, 30). Immunoprecipitation of MDA-MB-231 lysates with mAb 4C3, followed by immunoblotting with an independent anti-α₂ antibody, further confirmed the target antigen as the α₂ integrin subunit (Fig. 4B).

Fig. 4.

Identification of the mAb 4C3 target antigen. (A) mAb 4C3 immunocaptured a 150-kDa band from lysates of MDA-MB-231 cells as detected by SDS/PAGE/silver staining. Mass spectrometry sequencing of the band identified the protein as the α2 integrin subunit. (B) MDA-MB-231 lysates were immunoprecipitated with mAb 4C3 and immunoblotted with a second antibody directed against the α2 integrin subunit. (C) Peptide mapping of the mAb 4C3-binding sites in an overlapping series of peptides (each 10-aa long) that span the α2 integrin subunit. Asterisk indicates decapeptide epitope localized within the α-I domain. (D) Schematic illustrating the putative α-I domain elements recognized by mAb 4C3 and 8F10 (labeled “C” and “F”, respectively, within the red circle). Peptide 67 (mAb 4C3 peak) lies within structural element “F,” whereas peptide 44 (mAb 8F10 peak; Fig. S4) is located within β sheet “C” of the α2 chain as described. The three colored chains (green, yellow, and blue) represent a portion of the type I collagen triple helix. [Model adapted from ref. 29 with permission from Elsevier.]

Consistent with the fact that α₂ integrin subunit only forms heterodimeric complexes with the β1 integrin to generate the dominant mammalian type I collagen receptor, α₂β₁ peptide mapping of mAb 4C3 interactions with the α₂ subunit identified a major epitope lying within the α-I domain of the integrin, the dominant type I collagen recognition site of the α₂β₁ heterodimer (29, 30) (Fig. 4C). As expected from its collagen-binding properties, mAb 4C3 inhibited MDA-MB-231 adhesive interactions with type I collagen (Fig. 1C). Interestingly, a second, inhibitory α₂ integrin-reactive mAb that was identified independently in our screen (mAb 8F10) also bound to a distinct, but overlapping, epitope located within the α-I domain (Fig. S4).

Given these results, along with earlier studies demonstrating the ability of MDA-MB-231 cells to form α₂β₁-dependent adhesive interactions with bone matrices in vitro (26–28), we sought to determine whether our in vitro model accurately predicts α₂ integrin expression patterns found in type I collagen-rich metastatic lesions recovered from human breast cancer patients. For this, bone biopsy specimens were obtained from a series of seven patients with metastatic disease and immunostained for α₂ expression. Validating the results of our in vitro and xenograft models, all seven patients expressed α₂ in breast cancer cells in bone metastatic sites, with both carcinoma cells and surrounding vascular endothelial cells scoring positive in blinded analyses (Fig. 5 and Fig. S5). Given that archived biopsy material was available from the original primary breast cancer site in a subset of three of these patients, and the fact that type I collagen levels are distinctly higher in human breast tissue than in mouse mammary gland (25), we assessed α₂β₁ staining in these samples as well. Interestingly, distinct α₂ integrin expression was likewise detected in breast carcinoma cells in each of these patients (with weaker staining localized to normal myoepithelial cells), including tumor microemboli found within lymphatic vessels (Fig. S6).

Fig. 5.

α2 integrin expression in breast carcinoma bone metastases. Biopsy specimens of breast carcinoma bone metastasis were immunostained for α2 integrin expression in a series of seven patient samples (three shown here; the others shown in Fig. S3). Asterisks indicate bone tissue.

Discussion

Recent interest has focused on designing unbiased phenotypic screens wherein the identification of function-blocking effects precede efforts to dissect the underlying molecular mechanisms that give rise to the desired outcomes (13, 15, 31, 32). With increasing evidence that cell behavior in 3D culture systems more faithfully recapitulates in vivo function, greater emphasis has been placed on developing improved in vitro models for screening purposes, including the use of basement membrane-like gels, pepsin extracts of dermal collagen, and synthetic hydrogels (1, 2, 6, 33–35). The degree to which any of these constructs recapitulate the structure or function of the native ECM deposited in vivo remains controversial, however (1, 2, 16, 34). In carcinomatous states, neoplastic cells at both primary and metastatic sites are known to interface a network of covalently cross-linked type I collagen fibrils whose physical properties modulate tumor phenotypes (2, 3, 5–12, 16). Consequently, we elected to use type I collagen hydrogels that are naturally cross-linked by lysyl oxidase-derived aldimine bonds (16) to promote carcinoma cells to express a more in vivo-like display of surface antigens that could serve both as an immunogen for mAb production and as a physical platform for functional screening.

With this experimental approach, ∼5% of the generated cell-reactive mAbs displayed inhibitory effects. mAb 4C3 was selected for additional analysis based on its inhibitory activity in our in vitro screen, and further characterized as a proof-of-principle prototype to determine (i) whether function-blocking activity detected initially in vitro can be extended into in vivo settings, (ii) whether the mAb-reactive antigen can be identified, and (iii) whether target antigens discovered using human carcinoma cell–type I collagen composites can faithfully predict in vivo patterns of expression in patient samples. As described above, mAb 4C3 successfully inhibited the perivascular proliferation of extravasated MDA-MB-231 cells within the type I collagen-rich interstitial matrix of the chicken embryo, a model xenograft system in which cancer cell behaviors, including invasion, proliferation, and metastasis, recapitulate those observed in mouse xenograft models (22, 23).

We were initially surprised to find that mAb 4C3 did not affect primary tumor growth after orthotopic injection of MDA-MB-231 cells into the mouse mammary gland (Fig. S7). However, assessment of the local ECM environment at the site of carcinoma inoculation confirmed earlier studies describing the paucity of type I collagen in the mouse mammary fat pad (25). As an alternative, we chose mouse skeletal tissues as a surrogate tissue for assessing mAb 4C3-mediated effects on MDA-MB-231 proliferation, given its rich type I collagen content and the predilection of human breast carcinomas to metastasize to this organ system (17, 26–28). Although the effects of mAb 4C3 on carcinoma growth within the mandible and hindlimb supported mAb-mediated inhibitory effects, the high variability of this in vivo model did not allow these trends to reach statistical significance (Fig. 3). More importantly, however, MDA-MB-231 proliferation in the vertebral column was almost completely inhibited, with significant effects on the development of paralysis-associated morbidity. Further studies are needed to assess the ability of mAb 4C3 to affect tumor growth within preestablished metastases, either as a single agent therapeutic or in combination with other interventions. Nevertheless, these findings provided sufficient impetus to warrant identification of the mAb 4C3 target antigen.

After immunoaffinity purification and mass spectroscopy, the mAb 4C3 target antigen was identified as the α₂ integrin subunit, whose only known partner, the β₁ integrin chain, forms a heterodimeric complex that serves as a major type I collagen-binding receptor (29, 30). Peptide mapping characterized the mAb 4C3 epitope within the α-I domain of the α₂ integrin, a metal ion-dependent adhesion site that is responsible for ligand recognition and binding (29, 30). Although these results complement a number of reports documenting important roles for α₂β₁ in mediating cancer cell–type I collagen interactions in vitro, ranging from proliferation and invasion to epithelial–mesenchymal transition and cancer stem cell formation (36–48), the function of the α₂ integrin in neoplastic states in the in vivo setting is less clear. Recently, Ramirez et al. (49) concluded that α₂β₁ serves as a metastasis suppressor in mouse models as well as human cancers. Using α₂ integrin-null mice that were bred into a mouse mammary tumor virus-Neu transgenic line, they demonstrated that despite the complete absence of α₂β₁, tumor initiation was only marginally affected, whereas lung metastatic activity was actually enhanced (49). In this mouse model, however, all tissues are rendered α₂ integrin-deficient throughout embryonic and postnatal development, and thus the MMTV-Neu oncogene is expressed, by necessity, in α₂ integrin-null mammary epithelial cells, where potential effects of the integrin on tumor transformation and progression are difficult to define (i.e., as opposed to deleting the α₂ integrin in committed carcinoma cells). Indeed, in contrast to these findings, targeting α₂β₁ with either function-blocking antibodies or shRNA-based strategies has been reported to block metastatic activity in a number of animal model systems (50–53). Likewise, in a second in vivo model of cancer progression using α₂-null mice bred into a K14-HPV16 transgenic line, squamous carcinoma cell proliferation and metastatic activities were decreased in the absence of the α₂ integrin (54).

Independent of studies in mouse models, recent studies of human breast cancer and prostate cancer samples have indicated that α₂ mRNA expression levels can decrease as a function of increased metastatic burden and decreased survival (49). However, at the protein level, α₂β₁ is readily detected at both primary and metastatic sites in a variety of cancers, including breast (as described herein) and prostate cancer (52, 55, 56). Although it may be reasonable to conclude that high levels of α₂β₁ can potentially retard motile responses by promoting adhesion, lower levels of the integrin nevertheless may be required to support the cell–ECM interactions most conducive to invasion and growth. Nevertheless, it is unlikely that all carcinomas will prove equally dependent on α₂β₁, considering that other collagen-binding adhesion molecules, including α₁β_1, α₁₀β_1, α₁₁β₁, and discoidin receptors, have been described previously (30). As such, it should be stressed that the intent of using carcinoma cell–type I collagen composites as an antigen for mAb production is not to simply identify collagen-binding ligands, but rather to generate mAbs that interfere with cancer cell behavior in an environment similar to that encountered in vivo. Indeed, our preliminary studies indicate that most of the function-blocking mAbs identified in screens preformed to date target not type I collagen receptors, but rather surface molecules with as-yet uncharacterized mechanisms of action.

After applying the above-outlined strategy to identify function-blocking mouse antibodies, these reagents could be leveraged to generate humanized mAbs (14). From a therapeutic perspective, the broad distribution of the α₂β₁ integrin in normal tissues, as well as its ability to ligate other ECM proteins [e.g., type IV collagen, laminin, type XXIII collagen (30, 57)], might raise concerns regarding potential toxicities associated with targeting strategies. However, it is noteworthy that α₂-null mice are viable and fertile, and that α₂-integrin–deficient human patients who present only with mild bleeding diatheses have been identified (58–61). Interestingly, small-molecule α₂β₁ inhibitors have been developed as potential antithrombotic agents (62, 63), but our preliminary studies indicate that these agents are not as effective as mAb 4C3, at least in terms of interfering with MDA-MB-231–type I collagen adhesive interactions (Fig. S8). Thus, it remains possible that mAb 4C3 exerts unique effects on carcinoma cell function that might not be recapitulated by small-molecule inhibitors or α₂ integrin silencing. Finally, although our studies have emphasized potential roles for α₂β₁ in neoplastic states, we note that the integrin also has been implicated in fibrosis, inflammation, platelet-mediated thrombosis, and angiogenesis, raising the possibility that similar targeting strategies might be applied in other diseases as well (64–70).

The experimental approach outlined herein allows for the rapid identification of new target antigens in an unbiased fashion, as well as the isolation of murine mAbs suitable for humanization. Although our initial studies used a well-studied human breast carcinoma cell line, our approach is similarly amenable to the use of primary carcinoma cells or cancer stem cells. Indeed, we recently used primary human glioblastoma cancer stem cells to generate mAb libraries that were also found to exert inhibitory effects with target identification in process. As such, we anticipate that our phenotypic screening stratagem, using either human cancer cell lines, primary cancer cells, or even cancer cell–stromal cell composites (71), as well as more complex ECM-supplemented hydrogels to better recapitulate the anticipated changes that occur in connective tissue composition during tumor progression (12, 72), will prove valuable in the search for new targets and therapeutics in neoplastic states.

Methods

MDA-MB-231 cells (American Type Culture Collection) were embedded in mouse type I collagen hydrogels (21). The cell–matrix composite was then inoculated into 6-wk-old BALB/c female mice for immunization. MDA-MB-231–reactive mAbs were isolated and screened for antiproliferative activity in 3D collagen constructs (21). All procedures are described in more detail in SI Methods.