Expression of a Novel Factor in Human Breast Cancer Cells with Metastatic Potential¹

Recent experimental evidence implies that the ability of tumor cells to initiate growth after establishment in the secondary organ and to continue subsequent growth from micrometastases into macroscopic tumors is among the primary determinants of metastasis formation (1, 2). Early in the chain of tumor progression, metastases are limited in size and location (e.g., bone marrow micrometastasis) because the facility for metastatic growth has not yet been fully developed (3). The red bone marrow compartment still represents an important indicator organ of hematogenous micrometastatic spread in several types of carcinomas (4). However, disseminated tumor cells detected in the bone marrow of patients with early stages of solid cancers may simply represent irrelevant shed cells unless they hold biological properties that are crucial for establishment within secondary organs (5, 6).

Solid cancers metastasize in a selective manner to distant organs, and the organ specificity of the metastatic process is assumed to be governed by interactions between the malignant cells and local microenvironment factors (7, 8, 9, 10). The lack of biologically relevant model systems to study tumor-host interactions in metastasis, however, has limited insight into which molecular mechanisms are involved in the complex regulatory events of this fundamental aspect of cancer biology (11, 12).

The aim of this study was to identify cellular factors participating in the response of breast carcinoma cells to growth conditions offered by the target organ upon metastasis formation. Carcinoma cells in culture, originating from bone marrow micrometastases in a breast cancer patient who was clinically devoid of overt metastatic disease, were injected into the systemic circulation of immunodeficient rats, which subsequently developed metastases in the CNS.³ The carcinoma cells from the metastases were isolated by immunomagnetic selection and compared with the injected cells by means of differential display analysis, revealing a novel factor induced in the metastatic cells. We assume that this factor may mediate the growth response of tumor cells upon establishment in a secondary organ and, thereby, contribute to a cellular phenotype that is appropriate for cancer metastasis.

Micrometastatic cells were isolated by immunomagnetic selection of a bone marrow sample taken from a breast cancer patient who had invasive lobular carcinoma but who was clinically devoid of metastatic disease. The isolated carcinoma cells were established as the MA-11 cell line, as reported previously (13). The MA-11 cells were passaged in monolayer cultures for >100 passages before they were used in these animal experiments. We developed a novel experimental model for breast cancer metastasis to the CNS, using different routes of injection of MA-11 cells in athymic nude rats (Han: rnu/rnu Rowett), bred in the nude rodent facility at the Norwegian Radium Hospital (14). Cells in exponential growth phase were harvested, and suspensions were prepared for injection into either the LV (2.5 × 10⁶ cells in 200 μl per rat) or the CM (2.5 × 10⁶ cells in 30 μl per rat). The rats were sacrificed immediately upon symptoms of CNS disease. The care of animals and the experimental protocol were reviewed and approved by the National Animal Research Authority and carried out according to the European Convention for the Protection of Vertebrates Used for Scientific Purposes.

The affected tissues (spinal cords of rats injected in the LV and meninges of rats injected in the CM) were carefully removed and dissected. Tissue preparations from two to three equivalent rats were pooled and treated by mechanical dissociation to obtain cell suspensions (3.5 × 10⁶–5.0 × 10⁶ cells in 1 ml) for immunomagnetic cell separation. All solutions and cell preparations were kept on ice during the whole procedure to avoid nonspecific binding of immunobeads. A monoclonal antibody had previously been developed by immunizing nude mice (BALB/c nu/nu) with fresh tumor cells from an unclassified soft tissue sarcoma of high-grade malignancy (15). The antibody revealed a unique panhuman reactivity when tested by immunostaining on a wide range of human tumor cells, normal fibroblasts, and mononuclear cells from peripheral blood and bone marrow. No cell surface binding was observed on cell lines derived from rodents, and no reactivity was shown with normal tissue sections from mice, rats, or dogs.⁴ This antibody, purified from murine ascites, was conjugated to superparamagnetic sheep-antimouse IgG particles (Dynabeads SAM-450; Dynal A.S., Oslo, Norway). Antibody-coated beads were added at a ratio of 2:1 to total number of cells, and the suspensions were incubated for 30 min at 4°C on a rotating mixer. The cells were subsequently diluted with 1% human serum albumin in 0.9% NaCl solution and left in a magnet holder for 2 min. The supernatants, containing unbound cells, were decanted, and the remaining positive fractions (2 × 10⁵–3 × 10⁵ cells) were lysed in a hypertonic buffer containing 1% SDS (for RNA isolation when used in differential display analysis) or in TRIzol reagent (Life Technologies, Inc., Rockville, MD), of which three to four parallel samples were pooled for isolation of RNA for Northern blot analysis. The lysates were either used immediately or flash-frozen in liquid nitrogen and stored at −70°C. Similar lysates were also prepared of MA-11 cells in culture as well as of control tissues (spinal cords and meninges) from healthy rats (not injected with MA-11 cells).

Differential display analysis was performed essentially as reported previously (16). mRNA was extracted with the use of biotinylated anchored primers (5′ -T₁₁VN-3′) and Dynabeads M-280 Streptavidin (Dynal A.S.). First strand cDNA synthesis was subsequently carried out using Moloney murine leukemia virus reverse transcriptase (Life Technologies, Inc.). The cDNA generated was amplified using Taq DNA polymerase (Boehringer Mannheim, Mannheim, Germany), [³⁵S]dATPα (Amersham Pharmacia Biotech, Uppsala, Sweden) as label, and the anchored primers in conjunction with six different upstream arbitrary primers (AP1–6; random pentamers). Thirty-five cycles of PCRs were performed, each consisting of 30 s at 94°C, 60 s at 34°C, and 120 s at 72°C. The PCR products were resolved on standard denaturing sequencing gels that were subsequently dried and subjected to autoradiography. Unique PCR fragments were identified and excised from the gels, and the eluted products were reamplified using the original set of primers and the same thermal cycling condition. The reamplified products were subsequently used as probes to verify differential expression by Northern blot hybridization. The differentially expressed PCR products were cloned into the pCRII vector (Invitrogen, Groningen, the Netherlands) for sequencing. The full-length com1 cDNA was subsequently determined by 5′-rapid amplification (Marathon; Clontech, Hampshire, United Kingdom) of cDNA generated from nonlactating mammary gland mRNA.

Alignment analysis (17) using the predicted protein sequence of com1 revealed a rat homologue, p8, of which the COOH terminus displays some slight similarity with the key helices of homeodomain-type HTH DNA-binding proteins (18). Hence, com1 was aligned to p8 as well as to four different HTHs: hom (19), hnf (20), pue (21), and fia (22). Helices in the known HTH structures were predicted according to the definition given by Wintjens and Rooman (23) and helical conformations in com1 and p8 by the program Prelude (24). The complete DNA-binding unit, comprising three consecutive α-helices (H_R-2, H_R-1, and H_R), was analyzed. The third α-helix along the sequence is usually the H_R (recognition helix) that penetrates into the major DNA groove, whereas the orientation of H_R-2 (the first α-helix) determines structural subfamilies possessing different protein-DNA interaction modes (23). Turn motifs referred to as αBABBBα and α GBBα (α denotes an α-helical stretch, whereas A, B, and G denote residues with a backbone torsion angle in the domain typical of right-handed helices, extended β-type structures, or left-handed helical conformations, respectively; details given in Ref. 23) were also determined.

The MA-11, MT-1, MDA-MB-231, and MDA-MB-435 breast cancer cell lines were grown in RPMI 1640 (Life Technologies, Inc.) supplemented with 10% FCS (Life Technologies, Inc.) and glutamine (2.0 mm; Life Technologies, Inc.). The MCF-7 cells were kept in MEM containing phenol red (Life Technologies, Inc.) supplemented with 5% FCS and glutamine. The MCF7/LCC1 and MCF7/LCC2 cells were carried in MEM without phenol red (Life Technologies, Inc.) supplemented with glutamine in addition to 5% steroid-depleted FCS, as described previously (25). The cultures were kept at 37°C in a humidified 5% CO₂ atmosphere, and the cells were harvested in exponential growth phase.

Total RNA was extracted and analyzed by Northern blot technique. The probes were labeled with [α-³²P]dCTP (Amersham Pharmacia Biotech) by random priming technique, and standard Church hybridization conditions were used. Finally, to evaluate the amounts of RNA loaded, we hybridized the filters to kinase-labeled oligonucleotides specific for human 18S rRNA or a conserved sequence of 28S rRNA.

Suspensions of MA-11 cells were injected into the systemic circulation of athymic nude rats, and after ∼40 days, the rats injected in the LV developed hind leg paralysis caused by spinal cord metastasis (Fig. 1,a). As a control for cellular growth in the CNS, cells were injected directly into the subarachnoid space, and after ∼25 days, the rats injected CM showed signs of high intracranial pressure (usually hemiparesis), possibly as a result of massive meningeal tumor growth around infratentorial structures (Fig. 1,b). At higher magnification of these sections, mitotic figures are seen (Fig. 1, c and d), indicating actively proliferating tumors. Small tumors, presumably nonsymptomatic, were found also in the brain parenchyma of the rats injected LV (data not shown).

Metastatic MA-11 cells isolated by immunomagnetic separation of the tumor-affected tissues were examined by light microscopy, and the selected cell fractions were evaluated for host cell contamination (i.e., presence of cells with fewer than five immunobeads bound to their surface). This revealed a highly enriched CM population, whereas the fraction isolated from the intramedullary spinal cord tumors (LV population) contained ∼20% cells defined as contaminants (data not shown).

To identify site-specific gene expression, we compared the metastatic cells (selected LV population) with the CNS growth control (selected CM population) as well as to the injected cells (MA-11 cells in culture) by means of differential display analysis. Differentially expressed PCR fragments were considered to represent true metastatic properties if they were present in the LV population and absent in the CM population and MA-11 cells in culture and they were not concomitantly present in rat spinal cord (i.e., not evidently representing contamination from host tissue). Of the candidate fragments, differential expression was confirmed by Northern blot hybridization for two bands. The first, appearing to be the Grb2-associated binder-1 (26, 27), showed ∼2-fold higher levels in the LV population compared with the CM population, which further showed another ∼2-fold up-regulation relative to the MA-11 cells in culture (data not shown). The second, a novel “candidate of metastasis” (com1; Fig. 2), revealed ∼3-fold higher expression in the LV population compared with MA-11 cells in culture and ∼2-fold increase compared with the CNS growth control (CM population), as well as a complete lack of expression by the host tissues (Fig. 3).

The com1 PCR fragment matched expressed sequence tags found in several sequence databases. Because these expressed sequence tags revealed some slight differences in their sequences, the missing 5′ end of com1 cDNA was amplified from cDNA generated from nonlactating mammary gland mRNA, which was considered to represent a “common lineage” phenotype of breast gland-derived neoplasms. The 5′ fragment contained the complete coding sequence, a finding consistent with the nature of PCR products identified by differential display analysis, which generates 3′-nontranslated expressed sequence tags.

The full-length com1 cDNA sequence encoding an 82 amino acid peptide was determined (Fig. 4) and appeared to be the human homologue of the rat p8 cDNA (18). Protein segment analysis (Fig. 5) indicated that com1 shows ∼70% sequence identity with the rat p8 protein and lower but still significant similarity with H_R-2, H_R-1, and H_R (the three key helices) of four types of HTH DNA-binding domains: hom (19), hnf (20), pue (21), and fia (22).

To examine whether com1 expression may distinguish breast cancer cells with metastatic potential from cells without, we analyzed a panel of human cell lines (Fig. 6). Expression of com1 mRNA was observed only in the cell lines that are experimentally metastatic. The phenotypes of the tested cell lines were depicted by mRNA analysis of the ER and the estrogen-responsive factor pS2 (28, 29). The estrogen-responsive MCF-7 and derivative cell lines were pS2 positive, with high constitutive expression of pS2 in the MCF7/LCC1 and MCF7/LCC2 cells, as reported previously (25, 30). Expression of aberrant pS2 mRNA species (≠ 0.6 kb) by the ER negative MDA-MB-435 cell line is, to our knowledge, an original observation.

Current opinion suggests that metastasis formation primarily is a result of the ability of disseminated tumor cells to initiate and continue growth in the target organ (1, 2, 10, 12). We have applied a novel approach, comparing the phenotypes of tumor cells originating from bone marrow in a breast cancer patient who was clinically devoid of overt metastatic disease and the cell population isolated from experimental metastases formed by these tumor cells, to identify properties that might prevail in the metastatic cells. com1, identified as up-regulated in the metastases, presumably represents an HTH DNA-binding protein and may participate in the response of breast carcinoma cells to growth conditions offered by the target organ.

We based the criteria for identification of candidate genes partly upon the concept of organ specificity of the metastatic process (7, 8, 9, 10) by defining the CM population (representing cellular growth in CNS after direct inoculation) as biological control for the LV population (representing spontaneous metastatic establishment in CNS). However, less stringent selection criteria would also include gene expression specific for the CM population because meningeal tumor growth may represent a biologically relevant situation by itself. Clinically, meningeal malignant infiltration most often results from direct extension of carcinoma cells with a predilection for tumor sedimentation at the base of the brain (31), which precisely corresponds to the experimental case of rats injected CM.

Differential expression of com1 was confirmed, defined as at least a 2-fold difference in levels of mRNA expression between the LV population and the control CM population. A complete lack of expression by the host tissues was observed, clearly indicating that com1 mRNA expression was specific for MA-11 cells in this experimental model. Furthermore, the distinct com1 expression by the cell populations defined as CM and LV argues against severe contamination of rat cells in the immunomagnetic cell preparations.

The predicted 82 amino acid peptide encoded by com1 cDNA presumably represents a DNA-binding protein of HTH type, although it is unclear to which HTH subfamily, if any, it belongs. Protein structure of com1 was predicted by comparison with four types of HTH domains of known structure: hom, hnf, pue, and fia (19, 20, 21, 22).

The H_R of hom is relatively similar to the corresponding region of com1 and rat p8, although the typical homeodomain pattern WFQNRR (underlined residues are strictly invariant among homeodomains; Refs. 32 and 33) is only partially conserved. com1 displays the alternative pattern KLQNSE, whereas rat p8 presents the pattern KFQNSE. Another characteristic of homeodomains, the type αBABBBα turn between the H_R-2 and H_R-1 (represented by the pattern NRxxT, where x denotes any residue; Ref. 23), however, is nonexistent in com1 and rat p8. The type αGBBα turn between the H_R-1 and H_R (embodied by the residues CLT in hom or GIN in fia), typical for classical HTHs (23), is also absent.

The H_R of com1 and rat p8 is also similar to that of hnf, an HTH domain structurally resembling histone H5. The conserved pattern is +hQNSxR+ (where + denotes positively charged residue and h denotes a hydrophobic residue). In pue, a winged HTH motif with a small β-sheet located between H_R-2 and H_R-1, the latter helix shows similarity with the corresponding region of com1 and rat p8. The conserved sequence pattern is ++EAxA. In fia, an HTH belonging to the toxin repressor subfamily, the similarity to com1 and rat p8 is essentially located in H_R-2, represented by the pattern DxxDLYxL.

The cultured MA-11 cells were com1 positive, suggesting a requirement of constitutive com1 expression in carcinoma cells from which metastatic tumors originate. Hence, it is conceivable that com1 expression may distinguish breast cancer cells with facility to develop a fully metastatic potential from cells without. This notion was supported by com1 expression observed exclusively in breast cancer cell lines that are experimentally metastatic. The human MT-1 cell line (34) forms bone-eroding metastases in the spine upon LV injection in rats (14). The MDA-MB-435 cell line cultured in our laboratory gives rise to extensive metastasis when inoculated into rodents, whereas our MDA-MB-231 cell line does not metastasize, whatever routes of injection tested.⁴ The MCF-7 cell line and the derivative sublines MCF7/LCC1 and MCF7/LCC2 reflect the phenotypes of carcinoma cells observed during clinical progression of breast cancer. The parental cells are highly responsive to estrogens but poorly metastatic in animal models (35). The estrogen-independent MCF7/LCC1 cells possess a metastatic phenotype but are still ER positive (25, 35). The MCF7/LCC2 cells have retained the phenotype of the MCF7/LCC1 cells and are also tamoxifen resistant (30).

The Grb2-associated binder-1, identified as an intracellular effector molecule in tyrosine kinase signaling (26, 27), is shown to mediate nerve growth factor-dependent inhibition of apoptosis in promotion of cell survival (36). Our data support the concept that responses mediated by receptor tyrosine kinases are crucial for establishment of CNS metastasis (37), irrespective of the mechanism of such tumor growth (by direct extension as for meningeal metastasis or by hematogenous dissemination when parenchymal metastases are formed; Refs. 31 and 38).

Metastasis in the CNS represents a severe complication for patients with breast cancer. After clinical diagnosis of such advanced cancer disease, survival is limited (39). Breast cancer metastasis to the CNS or other distant organs may appear after extended periods of apparent freedom from disease (3, 39), a phenomenon that is conceptually due to a somewhat loosely defined proliferative reactivation of disseminated tumor cells (1, 2, 10). This may also explain why dormant micrometastases escape conventional therapies that are directed specifically against actively dividing tumor cells (40).

In conclusion, this study has identified and partially characterized a novel factor, termed com1, expressed in breast carcinoma cells forming metastatic tumors. com1 presumably represents a DNA-binding protein of HTH type, which might participate in intracellular signaling mediating growth responses of the tumor cells upon establishment in a secondary organ. Future studies of com1 function will include the identification of growth factors that induce com1 expression as well as presumably com1-responsive genes to map the complete regulatory pathway involved in this process.

We acknowledge Dr. S. Kaul at University of Heidelberg (Heidelberg, Germany) for originally establishing the MA-11 cell line, H. K. Hø ifødt at the Norwegian Radium Hospital (Oslo, Norway) for performing immunomagnetic preparation of target cells, Dr. R. Wintjens at Institut Pasteur de Lille (Paris, France) and G. Iurcu at Université Libre de Bruxelles (Brussels, Belgium) for valuable comments to the protein segment analysis, Dr. A. J. Wong at Jefferson Cancer Institute (Philadelphia, PA) for providing a plasmid for the Grb2-associated binder-1, Dr. M. V. Govindan at Université Laval de Quebec (Montreal, Quebec, Canada) for providing an ER cDNA, and Dr. P. Chambon at Institut de Chimie Biologique (Strasbourg, France) for providing a pS2 cDNA.