Lysyl Oxidase-related Protein-1 Promotes Tumor Fibrosis and Tumor Progression in Vivo¹

LO⁴ (protein-lysine 6-oxidase, EC 1.4.3.13) is a copper-dependent, secreted amine oxidase that oxidatively deaminates the ε-amino group of specific peptidyl lysine and hydroxylysine residues of collagen and of lysine in elastin. The resulting peptidyl aldehydes condense spontaneously with peptidyl aldehydes located in close proximity, or with unreacted ε-amino groups, to form inter- and intramolecular cross-linkages stabilizing the fibrous forms of collagen and elastin (1). However, this oxidative activity does not display high specificity because LO can also oxidize additional lysine-rich proteins such as histone-H1, as well as various lysine-rich synthetic peptides (2, 3). It was recently observed that LO can be translocated into cell nuclei and that it can regulate, by an as yet poorly understood mechanism, the expression of collagen-3A1, indicating that LOs may have additional functions (4, 5). LO is synthesized as an inactive proenzyme that is activated by the products of the BMP-1 gene (6).

Recently, several additional LO family members were identified. These new family members are characterized by the presence of a conserved LO-like domain containing conserved copper binding and catalytic domains at their COOH termini. These new LOL genes include LOL (7, 8), LOR-1 (or LOXL2; Ref. 9), LOR-2 (or LOXL3; Ref. 10), and LOXC (or LOXL4; Ref. 11). LOR-1, LOR-2, and LOXC differ with respect to LO and LOL in that they possess a much longer NH₂-terminal domain, suggesting that they constitute a distinct subclass of LOs, and that their functions may differ fundamentally from those of LOL and LO (12). LOR-1 was initially identified as a gene the expression of which is up-regulated in senescent fibroblasts and in adherent tumor-derived cells (9). It was also found to be highly expressed in reproductive tissues (13). It was demonstrated that LOL is also processed into its active form by bone morphogenetic protein-1, raising the possibility that the proteins encoded by the other family members may also be activated by proteolytic digestion after secretion.

The transition from a localized tumor to an invasive and metastatic tumor represents a landmark in the development of malignant disease because it is usually associated with a markedly worse prognosis. The understanding of the processes that govern this transition is therefore, of prime importance. It was recently observed that, in breast cancer, the transition from a localized to an invasive/metastatic tumor is associated in many cases with the formation of fibrotic foci and desmoplasia (the presence of unusually dense collagenous stroma) within the primary tumor (14, 15). There are also some indications that a similar correlation may exist in other types of cancers such as in colon cancer and in pancreatic cancer (16, 17). These observations represent apparent paradoxes at first glance because invasiveness has long been associated with the destruction of ECM by ECM-degrading enzymes like metalloproteases (18, 19) and heparanase (20). However, it is possible that the deposition of excess ECM may stimulate, in turn, expression of matrix-degrading enzymes that will contribute under certain circumstances to tumor invasion. In fact, there is some evidence that an increase in ECM deposition can, indeed, influence the production of ECM-degrading enzymes (21, 22).

Two LO family members, LO and LOL, were found to be expressed in areas of fibrogenesis in noninvasive in situ ductal breast carcinomas (23). During the preparation of this report, it was reported that metastatic breast cancer cell lines express LO, LOL, and LOR-1, whereas nonmetastatic breast cancer-derived cell lines, such as MCF-7 cells, do not, and that LO-expressing MCF-7 cells display increased invasiveness in in vitro invasiveness assays (24). These studies do not provide an explanation for observations that have noted a link between fibrosis and metastasis (14, 15). Here we provide evidence indicating that LOR-1 expression in nonmetastatic MCF-7 cells induces massive deposition of dense collagen fibers and the formation of numerous fibrotic foci in tumors that develop after the implantation of these cells in nude mice. These changes were accompanied by an increase in the invasiveness of the MCF-7 cells, although the LOR-1-expressing cells were still estrogen dependent.

Estrogen pellets (17β-estradiol, 0.72 mg/pellet, 60-day release) were from Innovative Research. The Masson’s trichrome staining kit was purchased from Bio-Optica Corp.; reverse transcriptase and G418 were from Life Technologies, Inc.; hygromycin B, tetracycline hydrochloride, and the reticulin staining kit, as well as the anti-collagen type I antibody (clone COL-1), were from Sigma. Restriction enzymes, and T4 ligase were from New England Biolabs. The bacterial expression vector pQE-30 and the nickel affinity column were obtained from Qiagen. [³²P]dATP was purchased from NEN. Monoclonal anti-cytokeratin-7 antibodies (CAM 5.2) coupled to FITC were acquired from Becton Dickinson. Anti-FITC antibodies conjugated to alkaline phosphatase were purchased from Roche. CAS blocking solution, antibody diluant reagent solution, citrate, and EDTA antigen retrieval buffers were purchased from Zymed.

The MCF-7 breast cancer cells were kindly given to us by Dr. Hadasa Degani (Weizmann Institute, Rehovot, Israel). The MDA-MB-435 breast cancer cell line was kindly provided by Dr. Israel Vlodavsky (Technion, Haifa, Israel). The MDA-MB-231 cells were given to us by Dr. Michael Klagsbrun (Harvard University, Boston, MA). These cell lines were routinely cultured in DMEM supplemented with gentamicin, amphotericin, glutamine, and 10% FCS. Human umbilical vein-derived endothelial cells were isolated and cultured as described previously (25). Tissue culture media, sera, and cell culture supplements were from Biological Industries, Kibbutz Beth-Haemek, Israel, or from Life Technologies, Inc. MCF-7 TetOff cells (Clontech), containing the tetracycline trans-activator (tTA) were grown in DMEM containing 10% Tet system-approved FCS (Clontech), in the presence of 100 μg/ml G418, 150 μg/ml hygromycin B, and 1 μg/ml tetracycline.

Empty vector-transfected MCF-7 cells or LOR-1-producing MCF-7 cells, derived from different clones, were cultured in DMEM supplemented with 10% FCS, 2 mm glutamine, and antibiotics. For proliferation assays, cells were plated in 24-well dishes at 20,000 cells/well. The cells were counted in a Coulter counter after 4 days.

Total RNA (4 μg) from human umbilical vein-derived endothelial cells (for LOR-1) or melanoma cells (for LOR-2) was reversed transcribed using moloney murine leukemia virus reverse transcriptase (Life Technologies, Inc.) as described previously (26). The cDNA was amplified using the expand long high-fidelity PCR system (Roche). The primers used for amplification of the LOR-1 cDNA were: 5′-CGCAAGCTTGGATCCGGGATGGAGAGGCCTCTGTGC (containing a HindIII restriction site) and 5′-CGCTCTAGAGGATCCTTACTGCGGGGACAGCTGGTTG (containing a XbaI restriction site). The primers used for amplification of the LOR-2 cDNA were: 5′-GCCATGCGACCTGTCAGTGTC and 5′-GGGCAGTGGCACTTAGAT.

The 2.3-kb cDNA fragment of LOR-1 was subcloned into the pGEM-T Easy vector (Promega) by T-A cloning. When sequenced, it contained three point mutations compared with the published sequence of LOR-1 (9). Two were silent mutations but one of these mutations changed amino acid 301 (counted from the first methionine) from serine to glycine. This amino acid change appeared in LOR-1 cDNA that was prepared from several completely independent RNA preparations and was, therefore, assumed to represent the correct sequence.

A cDNA fragment containing nucleotides 1641–2253 of LOR-1 was amplified using the expand high-fidelity PCR kit. The primers used were 5′-ACATGCATGCCCTGACCTGGTCCTCAATGC and 5′-CCCAAGCTTGGAACCACCTATGTGGCAGTT. The LOR-1 cDNA fragment was subcloned into the pGEM-T Easy vector (Promega) by T-A cloning. The 613-bp LOR-1 cDNA fragment was cut with SphI and HindIII and ligated into the bacterial expression vector pQE-30, which added a 6×His encoding tag to the 5′ of the fragment in-frame. The resulting plasmid was used to produce a recombinant, 6×His-tagged M_r 23,000 peptide. The peptide was purified from bacterial cell extracts using nickel-affinity chromatography and was further purified using SDS-PAGE. The gel was electroblotted onto nitrocellulose; the band containing the peptide was cut out, solubilized in DMSO, and used to immunize rabbits. Antiserum was affinity purified on protein-A Sepharose column followed by affinity purification on a column to which the recombinant peptide was coupled using a previously described method (27). The antibody was eluted from the column using 0.1 m glycine at pH 3. The anti-LOR-1 antibody did not cross-react with the related human LOR-2 as tested by Western blot analysis of conditioned medium and cell extracts that were taken from Chinese hamster ovary cells overexpressing human LOR-2 (data not shown; Ref. 10).

For constitutive expression of LOR-1, the full-length LOR-1 cDNA was digested out of the pGEM-T easy vector (Promega) with HindIII and XbaI (which were incorporated into the primers used for the cloning of the LOR-1 cDNA) and ligated into the mammalian expression vector pcDNA3.1hygro (Invitrogen) to generate the expression vector pcDNA-LOR-1. Empty pCDNA3.1hygro plasmid or pcDNA-LOR-1 plasmid (10–20 μg) were stably transfected into MCF-7 cells using electroporation with a Bio-Rad gene pulser (960 μF, 0.28 V). Stable transfectants were selected using 300 μg/ml hygromycin B. Clones expressing recombinant LOR-1 were obtained in two consecutive stable transfections and screened for LOR-1 expression by using our anti-LOR-1 polyclonal antibodies. Conditioned medium was collected after 48 h from transfected cells and LOR-1 expression was monitored using Western blot analysis (Fig. 2 A). C6 glioma cells were transfected and screened for LOR-1 expression as described above.

For inducible expression of LOR-1, full-length LOR-1 cDNA was cloned into the pTET-Splice vector (Clontech), which enables an inducible expression under the control of tetracycline (Tet off system). The pTET-Splice plasmid DNA was digested with HindIII and SpeI and ligated with the 2.3-kb human LOR-1 cDNA that was cut out of the pCDNA–LOR-1 plasmid with HindIII and XbaI to yield the pTET-LOR-1 plasmid. This plasmid was cotransfected into MCF-7 TetOff cells with pTK-Hygro at a ratio of 20:1, respectively (i.e., pTET-LOR-1:pTK-Hygro). LOR-1-expressing cells were selected in medium containing 100 μg/ml G418, 150 μg/ml hygromycin B, and 1 μg/ml tetracycline. The best clone showing the highest induction levels in the absence of tetracycline and the lowest basal expression levels in the presence of tetracycline was designated MCF-7/Tet-LOR-1.

To examine the kinetics of LOR-1 posttranslational processing, MCF-7/Tet-LOR-1 cells were seeded in 24-well dishes and grown to confluence. The medium was exchanged to tetracycline-free and serum-free medium and aliquots of conditioned medium were analyzed by Western blot analysis using antibodies directed against the COOH-terminal domain of LOR-1.

Total RNA was extracted from cultured cells using Tri-Reagent (Molecular Research Center, Inc., Cincinnati, OH) according to the manufacturer’s instructions. Total RNA (15 μg) was loaded on a 1.2% agarose gel, and Northern blot analysis was carried out as described previously (28). LOR-1 and LOR-2 ³²P-labeled cDNA fragments (nucleotides 1–660 and 1061–1590, respectively) were used as probes.

Serum-free conditioned medium (40 μl) was separated on a 8% SDS-PAGE gel, and the proteins were electroblotted onto a nitrocellulose filters using semidry electroblotting. The filter was blocked for 1 h at room temperature with TBST buffer containing 10 mm Tris-HCl (pH-7.0), 0.15 m NaCl, and 0.3% Tween 20 supplemented with 10% low fat milk. The filter was incubated overnight at 4°C with affinity-purified rabbit anti-LOR-1 polyclonal antibody (1:2500) in TBST. It was subsequently washed three times in TBST and incubated with goat antirabbit IgG peroxidase-conjugated secondary antibodies for 1 h at room temperature. Bound antibody was visualized using the enhanced chemluminescence detection system.

Slow-release pellets containing 17β-estradiol (0.72 mg/pellet, 60-day release; Innovative Research) were preimplanted s.c. in female athymic nude mice (CD1), 6–8 weeks old. MCF-7 cells (10⁷ cells/mouse) were injected into the mammary fat pads, and tumor size was measured with a caliper once or twice a week. Mice were sacrificed 4 weeks after the injection of the MCF-7 cells. In other experiments, the tumors were excised when they reached a diameter of 0.8 cm. The primary tumor was removed, weighed, fixed in 10% buffered formalin, and embedded in paraffin.

In other experiments, the development of tumors from C6 glioma cells expressing recombinant LOR-1 was studied. Injections of empty vector-transfected cells or with LOR-1-transfected cells (2 × 10⁵ cells/mouse) were injected s.c. into the hind limb. Mice were sacrificed 3 weeks after the injection of the cells. The primary tumors were removed, fixed in 10% buffered formalin, and embedded in paraffin for analysis.

Formalin-fixed, paraffin-embedded tissues were cut into serial sections of 5 μm each and were used for immunohistochemistry. Sections were deparaffinized by heating to 60°C for 1 h, washed twice with xylene for 5 min, and rehydrated by consecutive washes in 100, 95, and 70% ethanol, followed by a 5-min wash in water. Endogenous peroxidase activity was inhibited by a 15-min incubation with 3% hydrogen peroxide in methanol, followed by consecutive washes with water and PBS. The sections were then antigen retrieved by heating them twice for 10 min in a microwave oven to 90°C in citrate antigen retrieval buffer (for anticytokeratin antibodies) or in EDTA antigen retrieval buffer (Zymed; for anti-LOR-1 antibodies). Blocking of tissue sections was subsequently done using the CAS blocking solution. After blocking, the sections were incubated with affinity-purified anti-LOR-1 antibody (at a dilution of 1:30 to 1:50 in antibody diluant solution), or with monoclonal FITC-conjugated antihuman cytokeratin-7 antibodies for 1.5 h at room temperature. The sections were washed three times with TBST, and the detection of bound antibodies was done using anti-FITC alkaline phosphatase-conjugated antibodies (Roche) at a 1:200 dilution for cytokeratin-7 detection, or with the DAKO Envision detection system (horseradish peroxidase-conjugated) for LOR-1 detection. Bound antibodies were detected either by using 5-Bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium chloride (Roche) for the visualization of bound anticytokeratin antibodies or by using 3-amino-9-ethylcarbazole solution (DAKO) for the visualization of bound anti-LOR-1 antibodies. The slides were subsequently counterstained by hematoxylin and photographed. In control experiments, the primary antibodies were omitted.

Formalin-fixed, paraffin-embedded tissues were sectioned, deparaffinized, and rehydrated as described above. H&E staining was performed as described previously (29). Masson’s trichrome staining of fibrous collagens and reticulin staining of collagen type 3 were performed according to the manufacturer’s instructions. Sirius red staining of fibrous collagen was performed as described previously (30).

To estimate the amounts of collagen type 1 in cultured cells the cells were seeded in 48-well plates and grown to subconfluence. They were washed twice with PBS and fixed with 4% paraformaldehyde at room temperature for 15 min. After fixation, cells were washed again twice with PBS and were permeabilized using 0.25% Triton X-100 for 15 min. The cells were blocked in PBS containing 1% BSA for 1 h and then washed twice with PBS. They were incubated for 90 min at room temperature with an anti-collagen type-1 antibody diluted in PBS containing 1% BSA (1:1000). After extensive washes with PBS, bound antibodies were detected using the DAKO Envision detection system.

Desmoplasia and formation of fibrotic foci in breast cancer tumors is associated with the transition from a localized, relatively benign tumor to an invasive/metastatic tumor (14, 15). LOs contribute to the deposition of collagen by covalently cross-linking collagen monomers (1). To find out whether expression of LOs is associated with the invasive/metastatic phenotype we screened several human breast cancer derived cell types for the expression of LOs. Northern blot analysis revealed that the LOR-1 gene is expressed in the malignant, estrogen-independent MDA-MB-231 and MDA-MB-435 cells (31) but not in estrogen-dependent nonmetastatic MCF-7 cells (Fig. 1,A, panel A). LOR-2 on the other hand was expressed only in MDA-MB-435 cells but not in the MDA-MB-231 cells nor in MCF-7 cells (Fig. 1 A, panel B). These experiments prompted us to investigate in more detail the possible role of LOR-1 in the progression of breast carcinoma tumors.

To determine the expression patterns of LOR-1 in normal breast and in breast carcinoma tumors differing in their malignancy, we screened sections derived from human tumors for expression of the LOR-1 protein using anti-LOR-1 antibodies. The expression of LOR-1 in normal human breast was confined to the epithelial cells of the ducts (Fig. 1,B, panel A). Expression of LOR-1 was also detected in intraductal noninvasive carcinoma cells (ductal carcinoma in situ) in the three tumors that were examined (Fig. 1,B, panel B). Periductal invasive low-grade (grade 1) breast carcinoma cells form duct-like structures in the periductal stroma. The tumorigenic cells located in these duct-like structures did not express LOR-1 in most of the tumors that were examined (8 of 11; Fig. 1,B, panel C, black arrows). In contrast, intraductal tumor cells that have not migrated out of the ducts presumably still respond to the signals provided by the duct environment and express LOR-1 (Fig. 1,B, panel C, white arrow). However, a decisive majority (13 of 16 tumors) of highly malignant grade-3 invasive ductal breast carcinomas, a more aggressive and desmoplastic type of tumor that is characterized by a 3-fold higher incidence of death, as compared with grade-1 breast carcinoma, and a 5-fold higher incidence of recurrence after surgery (32, 33) produce large amounts of LOR-1, even though these cells are not associated with ducts anymore (Fig. 1 B, panel D). Thus, a significant association exists between LOR-1 expression and tumor grade (P = 0.012, as determined by two-tailed Fisher’s exact test). We hypothesize that highly malignant breast carcinoma cells are able to express LOR-1, even though the malignant cells have lost positional signals that are required by less malignant cells so as to enable LOR-1 production.

To determine whether LOR-1 expression contributes to the progression of breast cancer tumors to the invasive/metastatic phenotype, we transfected noninvasive MCF-7 cells with an expression plasmid containing the full-length human LOR-1 cDNA. Several clones of cells that expressed large amounts of LOR-1 were isolated. Two such clones of cells, clones 12 and 24, were used in additional experiments. The conditioned medium also contained shorter forms of LOR-1 as revealed by Western blot analysis using antibodies directed against the LOR-1 c-terminal domain. These shorter forms may be generated by proteolytic processing in analogy to other members of the LO family (Refs. 6 and 34; Fig. 2,A, panel A). Clone 12 cells expressed larger amounts of LOR-1 as compared with clone 24 cells (Fig. 2,A, panel A). To make sure that the low molecular weight forms are produced as a result of posttranslational processing, we also expressed LOR-1 in MCF-7 cells under the control of a tetracycline-inducible promoter (35). It can be seen that once the tetracycline inhibition is removed, the cells start to produce full-length LOR-1, which is then converted into a shorter, M_r 70,000 COOH-terminal-containing form (Fig. 2 A, panel B). It is not yet clear whether all of these forms are enzymatically active.

Parental MCF-7 cells and empty plasmid-transfected MCF-7 cells, as well as cells of the two LOR-1-producing MCF-7 clones, were injected into the mammary fat pads of female nude mice and allowed to form tumors. The parental MCF-7 cells as well as all of the clones of cells that we derived from the parental cells formed tumors only in the presence of s.c. estrogen slow-release pellets (data not shown; Ref. 36). Interestingly, the rate at which tumors containing LOR-1-expressing cells developed was inhibited as compared with the rate of development of tumors derived from parental (not shown) or empty vector-transfected MCF-7 cells (Fig. 2 B, panel A and B). The significance of the differences in the sizes of the tumors was determined using a two-tailed Student’s t test on day 25. The differences in the average size of tumors that developed either from clone-12 cells or from clone-24 cells in comparison with the average size of tumors derived from control cells were significant (P < 0.05 for clone-24 and P < 0.001 for clone-12). To determine whether the decreased rate of tumor growth was caused by a slower rate of cell proliferation, we also compared the proliferation rate of empty vector-transfected MCF-7 cells with that of clone 12 and clone 24 cells. However, we could not detect significant differences in their rates of proliferation (data not shown).

To understand better why tumors expressing LOR-1 develop slowly, we examined H&E-stained sections of control and LOR-1 producing tumors. Interestingly, tumors that develop from clone 12 cells (Fig. 3,A, panel B) or from clone 24 cells (not shown) contain many necrotic areas, whereas tumors derived from parental (not shown) or empty expression vector-transfected cells (Fig. 3,A, panel A) contain very few necrotic areas. Additionally, the LOR-1-expressing tumors also contained extensive fibrotic areas that contained mainly host-derived cells such as fibroblasts rather than MCF-7 cells. The host-derived cells were easily distinguishable from the tumor cells because they did not react with an antibody directed against human cytokeratin 7, a marker of breast epithelial cells (Fig. 3,A, panel C) and were only counterstained with hematoxylin (Fig. 3 A, panel C; Refs. 37 and 38).

To find out whether tumors derived from LOR-1-expressing MCF-7 cells contain increased concentrations of collagen, we stained tumor sections with Masson’s trichrome stain, which reacts mainly with collagen type 1 producing an azure coloring (39). Tumors that developed from parental or empty vector-transfected MCF-7 cells contained limited amounts of collagen that was scattered between the tumor cells and did not contain any fibrotic foci (Fig. 3,A, panel D, arrows). In contrast, tumors that developed from clone 12 or clone 24 LOR-1-producing MCF-7 cells contained much larger amounts of dense collagen fibers between the tumor cells, indicating that LOR-1 induces accumulation and deposition of collagen in the tumors (Fig. 3,A, panel E, arrows). Furthermore, LOR-1 expression induced the formation of numerous fibrotic foci that appeared choked with collagen fibers (Fig. 3,A, panel F, white arrow). We have also stained tumor sections derived from empty vector transfected MCF-7 cells and from clone 12 and clone 24 cells with Sirius red, a dye which stains collagen type-1 primarily and, to a much lesser extent, collagen type-3 (30). This method also revealed a large increase in collagen deposition in tumors derived from LOR-1-expressing MCF-7 cells as compared with tumors derived from empty vector-transfected MCF-7 cells (data not shown). Blood vessels within tumors that develop from vector-transfected MCF-7 cells contain thin fluffy collagen fibers around blood vessels (Fig. 3,A, panel G, white arrow). In contrast, blood vessels found in tumors derived from LOR-1-producing MCF-7 cells were sheathed by a dense layer of collagen underneath the endothelial cell layer, as revealed by Masson’s trichrome staining (Fig. 3,A, panel H, black arrows). In some of the vessels, there was also a second, more distal sheath of thick collagen fibers around the vessels, which was indicative of perivascular fibrosis (Fig. 3 A, panel H, white arrow).

These experiments indicate that LOR-1 enhances, on its own, the accumulation and deposition of collagen in vivo. To find out whether LOR-1 can promote collagen accumulation in other types of tumors, we expressed LOR-1 in C6-glioma cells. C6-glioma cells do not express fibrous collagens, and the tumors that they form on implantation in mice are not fibrotic (40). When we implanted C6-glioma cells transfected with empty vector in nude mice, the cells formed, as expected, tumors that contained insignificant amounts of collagen. In contrast, the tumors that developed from C6-glioma cells expressing recombinant LOR-1 contained large deposits of collagen fibers as revealed by Masson’s trichrome staining, especially at the invasive front of the tumors (data not shown). It therefore follows, that the induction of collagen deposition by LOR-1 is a general property that is not specific to MCF-7 cells or to breast cancer. The tumors that developed from either LOR-1-producing MCF-7 cells or from LOR-1-producing C6-glioma cells also contained much higher concentrations of collagen-type 3 fibers (Fig. 3,B, panels B and D) in comparison with tumors that developed from empty vector-transfected cells (Fig. 3,B, panels A and C), as revealed by reticulin staining (41). In addition, the collagen-type 3-containing fibers in the tumors that developed from the LOR-1-expressing cells also appeared to be much thicker than collagen-type 3 fibers in control tumors (Fig. 3 B). These results indicate that LOR-1 can affect simultaneously the deposition of several types of collagen. The source of the deposited collagen is unclear at this point. However, it is likely to be produced by the cells of the host mice because LOR-1-expressing MCF-7 cells that are grown in culture do not produce significant amounts of collagen (data not shown).

It was reported that the appearance of fibrotic foci in breast cancer tumors correlates with their degree of invasiveness (15). Because LOR-1-producing MCF-7 cells form tumors rich in fibrotic areas, we also asked whether the LOR-1-expressing cells that form these tumors acquire invasive properties. Tumors that developed after the implantation of empty vector-transfected MCF-7 cells in nude mice were surrounded by a thick pseudocapsule. The border between the tumor and the pseudocapsule in these tumors was sharply defined, and the MCF-7 cells did not migrate out of the tumors into the pseudocapsules (Fig. 4, A and B). We could not detect tumorigenic cells within blood vessels located adjacent to these control tumors, nor were there tumorigenic cells observed in the perineural space of neighboring nerves and between muscle located near the tumors (Fig. 4, A and B). In contrast, MCF-7 cells expressing recombinant LOR-1 were able to invade the pseudocapsules of the tumors that developed after their implantation in nude mice (Fig. 4,C, arrows). The LOR-1-expressing tumor cells infiltrated the pseudocapsules of these tumors and invaded adjacent tissues. Tumor cells invading muscles adjacent to the tumor were identified using antibodies directed against human cytokeratin-7 (Fig. 4,D, dark blue stain, black arrows) as well as with an antibody directed against LOR-1 (Fig. 4,E, red stain, white arrows). In addition, we have also observed tumor cells invading the vascular system (Fig. 4, F and G, arrows) as well as the perineural space of adjacent nerves (Fig. 4 H, arrow). These last characteristics are hallmarks of metastatic tumor cells (42, 43). These observations provide evidence indicating that the production of LOR-1 by breast cancer tumor cells contributes to the transition from the noninvasive stage of breast cancer to the invasive/metastatic stage.

The appearance of fibrotic foci in breast carcinoma tumors had been linked to high malignancy, metastasis, and poor prognosis (14, 15). We have shown here for the first time that overexpression of LOR-1, a recently identified LO family member, in MCF-7 breast cancer cells and in C6-glioma cells induces the deposition of large amounts of dense collagen fibers (desmoplasia) in the tumors that develop after the implantation of these cells in nude mice. These experiments demonstrate that the expression of a LO family member suffices to induce collagen accumulation and fibrosis in vivo. Furthermore, we have shown that LOR-1 expression enhances the invasiveness of the MCF-7 cells, and that LOR-1 is likely to contribute to the emergence of metastasis because the LOR-1-expressing MCF-7 cells also display characteristics that are usually associated with metastatic cells, such as the ability to invade the vascular system. Taken together, these observations argue for a role of LOs in the progression of breast cancer and show for the first time that the expression of a single protein can induce simultaneously fibrosis and invasiveness, two processes that had been previously observed to be linked in the progression of spontaneous breast cancer tumors (14, 15).

We have found that cells of well-differentiated grade-1 ductal carcinomas that have migrated out of the ducts do not usually produce LOR-1. In contrast, malignant grade-3 ductal carcinoma cells located outside of the ducts do produce large amounts of LOR-1. We hypothesize that the ducts produce a signal that induces LOR-1 production in the epithelial cells of the ducts, and that this signal is absent in the periductal stroma. If our hypothesis is correct, than the well-differentiated grade-1 ductal carcinoma cells that are located in the periductal stroma lose their ability to express LOR-1 because the signal that induces LOR-1 production is missing, whereas adjacent tumor cells located within the ducts still express LOR-1 (Fig. 1 B, panel C). The basement membrane is known to provide signals that are essential for the proper differentiation and function of the epithelial cells lining the ducts (44, 45). It is, therefore, conceivable that such a localized signal will be provided by the basement membrane. We hypothesize that, once the carcinoma cells lose contact with the duct environment, their LOR-1-producing ability is tightly linked to their degree of malignancy. Thus, the highly malignant cells of grade-3 ductal breast carcinomas seem to express LOR-1 in the vast majority of tumors regardless of their location, whereas the less malignant cells of grade-1 ductal breast carcinoma fail to produce LOR-1 in most tumors once they lose contact with the duct environment. Interestingly, it was previously observed that highly malignant grade-3 carcinomas usually contain fibrotic foci, whereas the less malignant grade-1 carcinomas display a lower incidence of fibrotic foci (46).

The differentiated epithelial cells located within the ducts and in the lobules of the breast normally produce LOR-1, as do tumorigenic cells of in situ ductal carcinomas. We could not detect LOR-1 expression in the stromal cells surrounding the ducts. This expression pattern differs dramatically from the reported expression patterns of LO and LOL, which are not expressed by the tumor cells of ductal carcinomas in situ but are expressed by stromal cells surrounding the ducts (23, 47). Interestingly, production of LOR-1 by the normal epithelial cells of the duct or by cells of in situ ductal carcinomas is not accompanied by fibrosis (data not shown), nor are these cells invasive. We hypothesize that LOR-1-induced fibrosis occurs only when LOR-1-producing cells come into close contact with stromal collagen-producing cells. Likewise, invasiveness may also be potentiated by LOR-1 indirectly, as a consequence of the induction of fibrosis, thus linking the two processes. Alternatively, it is possible that invasiveness is induced by LOR-1 using a separate mechanism, and that normal, LOR-producing cells possess inhibitory mechanisms that hinder LOR-1-induced invasiveness. However, because a link has been found between the presence of fibrotic foci in spontaneous breast cancer tumors and poor prognosis (14, 15), it is more likely that these two processes are somehow interdependent. This would seem, at first glance, to be a paradox, because many studies have established a link between the production of ECM degrading proteases and tumor invasiveness (18, 19, 20). How then can an enzyme that apparently induces the deposition of ECM proteins and is, therefore, expected to inhibit tumor invasiveness, enhance invasiveness instead? One possible explanation is that the production of a great excess of ECM proteins can in turn trigger expression of degrading enzymes that will then overcome local inhibitory effects of the deposited ECM. Alternatively, the fibrosis-inducing activity and the invasiveness-enhancing activities of LOR-1 may be induced by LOR-1 simultaneously and independently.

We have not been able to detect collagen production by cultured control MCF-7 cells nor could we detect collagen production by the LOR-1-expressing MCF-7 cells grown in conventional cell culture (data not shown). It, therefore, seems likely that the collagen seen in tumors developing from LOR-1-producing MCF-7 cells or from LOR-1-producing C6-glioma cells originates from stromal host cells. It was shown that LO can activate the collagen-3A promoter (5), and it is, therefore, possible that LOR-1 can function as a signaling molecule to induce collagen synthesis in stromal host cells by a paracrine mechanism.

Interestingly, the development of tumors from LOR-1-expressing MCF-7 cells was inhibited as compared with tumors developing from cells that do not express LOR-1. LO was shown to behave as a tumor suppressor that inhibits the expression of ras (48, 49). It is possible that LOR-1 expression affects tumor development similarly, although the high-level expression of LOR-1 in grade-3 ductal carcinoma and the lack of an effect on the in vitro proliferation of the LOR-1-expressing MCF-7 cells seems to conflict with this hypothesis. These tumors also contained abundant necrotic areas. This observation is puzzling in that usually tumors that expand quickly are more likely to develop necrotic areas, because angiogenesis cannot catch up with the rapid expansion of the tumor (50). In the case of the LOR-1-expressing tumors, it is possible that the collagen deposits inhibit angiogenesis and lead to the formation of necrotic areas. Indeed, the necrotic areas were usually observed near the centers of fibrotic foci (see Fig. 3 A, panels B and F). In addition, we have also seen that tumors containing LOR-1-producing cells contained blood vessels that were surrounded by a thick sheath of collagen fibers. Such coating may also contribute to the inhibition of tumor angiogenesis and favor the formation of necrotic areas. Another contributor to the formation of necrotic areas in LOR-1-producing tumors may be hydrogen peroxide, a toxic side product of LO activity (51). It was shown that hydrogen peroxide can cause apoptosis and death of MCF-7 cells (52, 53). Hydrogen peroxide concentrations would probably reach their highest concentrations in the areas in which the largest concentrations of collagen can be found, because, in such areas, the activity of LOR-1 would be expected to be maximal, and the necrotic areas in the tumors were indeed located usually at the center of fibrotic foci.

The recent discovery of five distinct enzymes that contain a LO-like catalytic domain raises questions regarding their substrate specificity and biological roles. The collagen types that were deposited as a result of the expression of LOR-1 included both collagen type 1 and collagen type 3. LO, LOL, and LOX-C are known to oxidize lysine residues on collagen type-1 (34, 54, 55). LOR-1 can also oxidize lysine residues on collagen type 1.⁵ It is, therefore, likely that these enzymes overlap in their biological functions. It is, thus, likely that additional enzymes belonging to this family besides LOR-1 contribute to the induction of fibrosis and to tumor progression in breast cancer, as well as in other malignancies in which there is some evidence linking abnormal fibrosis with tumor progression (16, 17). This possibility is supported by observations that indicate that metastatic breast cancer-derived MDA-MB-435 cells express both LOR-1 and LOR-2 (Fig. 1), whereas MDA-MB-231 cells were found to express LO in addition to LOR-1 (24). It is, therefore, likely that the coexpression of several LO family members may produce synergistic effects and, thus, enhance fibrosis and invasiveness more potently than in cases in which only one LO family member is produced.

To conclude, our experiments indicate that LOR-1 is a potent inducer of tumor fibrosis and an enhancer of tumor invasiveness. Furthermore, the expression of LOR-1 in spontaneous tumors is associated with higher tumor grade and malignancy. It remains to be seen whether it would be possible to inhibit the progression of breast cancer tumors using LOR-1 inhibitors.

We thank Sharon Soueid for excellent technical assistance. We thank Assaf Gilad for his help with immunohistochemistry and for helpful discussions.