INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The mammalian HMG-I, HMG-Y, and HMGI-C proteins are members of the HMGI(Y) family of nonhistone chromatin proteins that have been implicated in both positive and negative regulation of gene transcription in vivo (reviewed in references 14 and 96). The human HMG-I (11.5-kDa) and HMG-Y (10.4-kDa) proteins are produced by translation of alternatively spliced mRNAs coded for by the Hmgiy gene at chromosomal locus 6p21 (39), while the related, but distinct, HMGI-C protein (12 kDa) is coded for by the Hmgi-c gene at chromosomal locus 12q15 (15). Members of the HMGI(Y) family are often referred to as architectural transcription factors because of their ability to regulate gene activity through the recognition and alteration of the structure of DNA and chromatin substrates (14, 100). The HMGI(Y) proteins not only bind with high specificity to the narrow minor groove of A · T-rich, random sequence B-form DNA (97, 107) but also have the ability to bind with high affinity to bent or distorted DNA structures such as synthetic four-way junction DNAs (55, 56), supercoiled plasmid substrates (87), and nucleosome core particles (98, 100). The ability of the HMGI(Y) proteins to bind to a variety of DNA substrates is a consequence of the high degree of intrinsic flexibility of the proteins, particularly within their A · T hook DNA-binding motifs (97) whose structure in a cocomplex with a synthetic substrate has recently been determined (61). In addition to recognizing structure rather than sequence, the HMGI(Y) proteins also have the ability to bend, straighten, unwind, and supercoil DNA substrates (14, 33, 75, 87). It is this unusual combination of substrate-binding characteristics, combined with the ability to specifically interact with other transcription factors, that allows the HMGI(Y) proteins to function as gene-regulatory factors in vivo by participating in the formation of stereospecific multiprotein enhanceosome complexes (66, 122). Among the known transcription factors that interact with HMGI(Y) in vitro or in vivo are NF-B, ATF-2/c-Jun, Elf-1, Oct-2, Oct-6, SRF, NF-Y, PU-1, RAR, Sp1, NFAT, and others (14, 19, 24, 57, 58, 66, 86, 121). Their many potential protein partners provide the HMGI(Y) proteins with considerable flexibility in regulating the transcriptional activity of a large number of different genes in vivo (14), among which some, such as the intercellular adhesion molecule E-selectin (77), often exhibit aberrant expression patterns in human malignancies (6) whereas others, such as tumor necrosis factor beta (35) and beta interferon (IFN-) (83), are associated with angiogenesis, a necessary component of solid-tumor formation.

The level of expression of HMGI(Y) mRNAs and proteins is low (68, 69, 82) or undetectable (40, 44) in most differentiated or nonproliferating normal cells but is rapidly induced in response to growth-stimulatory factors such as serum (67, 69), transforming growth factor  (TGF-) (95), epidermal growth factor (EGF) (59), platelet-derived growth factor, and fibroblast growth factor (FGF) (74), as well as phorbol esters (22, 39), beta 1 interferon (IFN-1), and endotoxin (91). These observations, combined with the fact that deletion of the Hmgi-c gene results in the diminutive pygmy phenotype in mice (7), implicate the HMGI(Y) proteins in the control of cell proliferation. Consistent with this suggestion, in neoplastically transformed cells, as well as in embryonic cells that have not yet undergone overt differentiation, the constitutive level of HMGI(Y) gene products is often exceptionally high, with increasing concentrations being correlated with increasing degrees of malignancy or metastatic potential (14, 40, 44, 116). This correlation is so consistent and widespread in tumors that it has been suggested that elevated concentrations of HMGI(Y) proteins are diagnostic markers of both neoplastic transformation (43, 44) and increased metastatic potential (13, 117). For example, HMGI(Y) overexpression has been suggested to be a diagnostic indicator for human prostate tumors (118), thyroid neoplasia (18), uterine cervix cancers (4), and colorectal cancers (1), among others (reviewed in reference 116).

Of particular interest for the present study is the fact that increased expression of HMGI(Y) proteins has been correlated with the increased metastatic potential of both mouse (94, 95) and human (59, 79) mammary epithelial cancers. Importantly, HMGI(Y) overexpression has also recently been detected in primary human breast cancers by a combination of serial analysis of gene expression and array technologies (85, 92). Overexpression of full-length HMGI(Y) proteins induces anchorage-independent growth of both mouse breast epithelial cells (94) and Rat-1 fibroblast cells (132) in soft agarose, as does overexpression of either truncated or chimeric forms of the HMGI-C protein in murine fibroblasts (37). Furthermore, antisense-mediated suppression of HMGI(Y) protein synthesis suppresses retrovirus-induced neoplastic transformation of rat thryoid cells (8), as well as inducing apoptotic death in anaplastic human thyroid carcinoma cell lines but not in normal thyroid cells (102). Of additional interest is the fact that chromosomal rearrangements in which the A · T hook DNA-binding domains of the HMGI(Y) proteins are fused to ectopic peptide sequences to form chimeric oncogenic proteins are among the most common lesions in human cancer, being observed in a high percentage of benign mesenchymal neoplasias, including lipomas, leiomyomas, fibroadenomas, aggressive myxomas, pulmonary hamartomas, and endometrial polyps (reviewed in references 54 and 116). These previous studies, when considered together, present a compelling case for involvement of the HMGI(Y) proteins, and their A · T hook DNA-binding motifs, in both neoplastic transformation and metastatic tumor progression. Nevertheless, they did not directly and unambiguously demonstrate a causal in vivo relationship between the HMGI(Y) proteins and cancer in intact organisms. These previous studies also did not elucidate the molecular events underlying the biological consequences that overexpression or aberrant expression of HMGI(Y) proteins might have on processes such as oncogenic transformation, increased tumor metastatic potential, and aggressive neoplastic malignancy.

In the present study, we demonstrated that nontumorigenic human breast epithelial cells containing either a tetracycline-regulated HMG-I or HMG-Y transgene acquire the abilities to both grow anchorage independently in soft agar and form metastatic tumors when injected into nude mice only when the HMGI(Y) proteins are overexpressed. While overexpression of either the HMG-Y or HMG-I protein alone readily induced neoplastic transformation, as evidenced by the acquired ability of cells to grow in soft agar, somewhat surprisingly, only the HMG-Y isoform elicited efficient tumor formation and metastasis in vivo when expressing cells were injected into nude mice. Furthermore, expression of either antisense or dominant-negative HMGI(Y) constructs in tumor cells inhibited both their rate of proliferation and their ability to grow anchorage independently in soft agar. Analysis of the transcription profiles of cells employing cDNA arrays demonstrates that overexpression of HMGI(Y) proteins modulates the expression of a complex set of genes that are known to be involved in cell signaling, proliferation, tumor initiation, invasion, migration, induction of angiogenesis, and colonization. Importantly, immunohistochemical analyses of tumors formed in nude mice by cells overexpressing HMGI(Y) suggest that they have undergone an epithelial-mesenchymal transition (EMT) in vivo, an event consistent with the results obtained from gene transcription array analyses. These data represent direct experimental support for the hypothesis that overexpression of the HMGI(Y) proteins is causally associated with neoplastic transformation, metastatic progression, and mesenchymal transition of human breast epithelial cells in the context of an intact living organism.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Cell culture and selection of transgenic clones. The human breast epithelial cell line MCF-7 (catalog no. HTB-22) was obtained from the American Type Culture Collection (ATCC; Manassas, Va.) and was grown in Dulbecoco's modified Eagle's medium (DMEM) supplemented with 2 mM L-glutamine, 10 mM HEPES, 10% fetal calf serum (FCS), penicillin G sodium at 100 U/ml, and streptomycin sulfate at 100 µg/ml and maintained as recommended by the supplier. MCF-7/PKC cells (a generous gift of D. K. Ways, East Carolina University School of Medicine [128] were grown in the same medium as MCF-7 cells but supplemented with G418 at 100 µg/ml. The human epithelial cell lines Hs578Bst (ATCC catalog no. HTB-125) and Hs578T (ATCC catalog no. HTB-126), as well as the human HeLa cervical carcinoma cell line (ATCC catalog no. CCL-2), were also obtained from the ATCC and maintained as recommended by the supplier. The tetracycline-regulated M/tet, M/tet/Vec, M/tet/HA-I, and M/tet/HA-Y human breast epithelial cells were obtained as follows. The parental M/tet (MCF-7/Tet-off) cell line was purchased from Clontech, Palo Alto, Calif. (catalog no. C30071, lot no. 7080502). The M/Tet cell line was derived by stable transfection of MCF-7 cells with a tetracycline-controlled transgene coding for the tetracycline transactivator protein and exhibits normal slow-growth characteristics, is contact inhibited, and does not form tumors when injected into nude mice. The cell lines designated M/tet/HA-I or M/tet/HA-Y (e.g., M/tet/HA-IC7, M/tet/HA-YC21, etc.) are clonal derivatives of M/tet cells that have been stably transfected with a plasmid vector (Clontech) containing the tetracycline response element (pTRE) driving the expression of either a hemagglutinin (HA)-tagged HMG-I or HA-tagged HMG-Y cDNA transgene (69). The control M/tet/Vect cells are clonal derivatives of M/tet cells that have been stably transfected with the empty pTRE vector without any insert. M/tet cells were cultured in DMEM containing 10% FCS (tetracycline free), 2 mM L-glutamine, G418 at 100 µg/ml (for maintenance of selection of the gene for the tetracycline transactivator protein), penicillin G sodium at 100 U/ml, and streptomycin sulfate at 100 µg/ml. Selection and maintenance of the M/tet/Vect, M/tet/HA-I, and M/tet/HA-Y cell clones were done with the same medium used for the M/tet cells, except that it contained either hygromycin at 100 µg/ml for M/tet/Vect and M/tet/HA-I cells or zeocin at 50 µg/ml for M/tet/HA-Y cells. In addition, the M/tet/HA-I and M/tet/HA-Y clones were routinely maintained in tetracycline at 2 to 10 µg/ml (depending on the individual clone) to keep expression of the HA-I and HA-Y transgenes turned off.

Construction and use of HMGI(Y) plasmid expression vectors. Wild-type HMG-I and wild-type HMG-Y plasmid expression vectors were created by subcloning the PCR products of the coding regions of human HMG-I and HMG-Y cDNAs (69) into the pcDNA3.1/Zeo+ plasmid expression vector (Invitrogen) between the HindIII and BamHI sites of the polylinker. The construction and use of both the antisense expression vector pRcCMVIGMH (57, 99) and the dominant-negative expression vector HMG-I (cytomegalovirus-HMGImII,mIII) (58) have been described previously. HA-tagged HMG-I and HA-tagged HMG-Y expression vectors were created by fusing a synthetic DNA oligonucleotide encoding the peptide MYPYDVPDYASL from the influenza virus HA protein (i.e., HA tag) in frame to the N-terminal end of the HMG-I and HMG-Y cDNAs by standard PCR protocols (3). The PCR fragments containing either HA-tagged HMG-I or HA-tagged HMG-Y were then subcloned into the pTRE expression vector, and the sequences of the resulting expression constructs were confirmed by automated DNA sequencing prior to use.

Cell transfection and clone selection. Plasmid expression vectors were introduced into cells by employing either the DMRIE-C, the Lipofectamine, or the Lipofectamine Plus transfection reagent following the manufacturer's (Gibco BRL, Gaithersburg, Md.) instructions. Transfection efficiencies in transient-transfection experiments using antisense and dominant-negative HMGI(Y) expression vector constructs were routinely monitored by cotransfection of a cytomegalovirus-based plasmid vector expressing green fluorescent protein (GFP) at a DNA ratio of 1:1. The total viable cells and viable GFP-expressing cells were counted using a UV microscope. The percentage of GFP-expressing cells was then obtained and served as an estimate of the transfection efficiency. In many experiments, transfection efficiencies were also monitored by cotransfection of cells with a luciferase reporter plasmid and determination of luminescence activity in cell lysates using a luciferase assay kit (catalog no. E4030; Promega, Madison, Wis.) and a Top Count Luminometer (Packard Instruments).

Cell proliferation assays. Cells were trypsinized and suspended in DMEM with 10% FCS and counted using either a Coulter Counter (Coulter Corp., Hialeah, Fla.) or a hemacytometer. Cell viability was assessed using trypan blue exclusion (to obtain the percentage of viable cells). For growth curve determinations, 2 × 10⁴ or 4 × 10⁴ viable cells were initially seeded in 2 ml of medium per 35-mm diameter dish. In all experiment, cells in four or more separate culture dishes were counted on each day of a time course study in order to ensure statistical accuracy of the growth curves.

Acid extraction and Western blot analysis of total cellular HMGI(Y) proteins. Acid extraction of HMG proteins from cells with 5% perchloric acid was performed as previously described (99). The acid-soluble proteins, which include not only HMGI(Y) but the other HMG proteins and histone H1 as well, were electrophoretically separated on either a 4 to 20% gradient or a sodium dodecyl sulfate (SDS)-15 or 18% polyacrylamide gel electrophoresis (PAGE) minigel and then transferred to an Immobilon-P membrane (Millipore) by electrotransfer overnight. Immunodetection of the HMG-I (Y) proteins on the membrane was performed using a rabbit polyclonal anti-HMG-I antibody (MR-19) whose production and use have been described previously (99) and an enhanced-chemiluminescence detection procedure using the SuperSignal substrate supplied by Pierce Chemical Co., Rockford, Ill.

Immunoprecipitation and Western blot assays of HA-tagged proteins. After washing with phosphate-buffered saline, cells were directly lysed in their culture dishes (about 5 × 10⁶ cells/dish) by addition of 1 ml of radioimmunoprecipitation assay buffer (1× phosphate-buffered saline, 1% Nonidet P-40, 00.5% sodium deoxycholate, 0.1% SDS, phenylmethylsulfonyl fluoride and aprotinin [Sigma catalog no. A6279] at 30 µg/ml each, and sodium orthovanadate at 10 µl/ml). The lysate was passed through a 21-gauge needle to shear the DNA and incubated on ice for 60 min. Particulates were removed from the lysate by centrifugtion at 15,000 × g for 20 min at 4°C, and the supernatant containing the total soluble cellular proteins was collected. Approximately 50 µl of anti-HA tag mouse monoclonal antibody (12CA5; a generous gift of J. J. Chen [16]) was added to 100 µl of lysate, and the mixture was incubated on ice for 1 h. Protein A-agarose beads (50 µl; Sigma) were then added, and the mixture was incubated overnight at 4°C while rotating. The beads were collected by centrifugation, and the bound proteins were removed by washing the beads three times with radioimmunoprecipitation assay buffer. An equal volume of SDS-PAGE sample buffer was added to the eluate, and the mixture was boiled for 2 min. The proteins in the eluate were separated by electrophoresis on an SDS-15% PAGE minigel and then transferred to a nitrocellulose membrane using a semidry electrophoretic transfer cell (Trans-Blot SD; Bio-Rad). The enhanced-chemiluminescence detection procedure described above was then used to identify the membrane-bound, HA-tagged proteins.

Soft agar anchorage-independent growth assays. Seven milliliters of molten 0.6% Difco agar (44°C) in DMEM with 10% FCS and the desired antibiotic (such as G418, tetracycline, hygromycin, or zeocin, etc.) was poured into 60-mm diameter petri dishes and allowed to solidify for 30 min. The desired numbers of cells and the desired antibiotics were then layered over the hardened agar in 1.5 ml of molten (44°C) 0.33% low melting temperature Difco agar containing DMEM and 10% FCS, and the dishes were allowed to sit for 10 min before transfer to an incubator. Cultures were incubated at 37°C in 5% CO₂ with over 90% humidity for about 30 days. The colonies were stained using a 0.25-mg/ml solution of [2-(p-iodophenyl)-3-(p-nitropheyl)-5-phenyl tetrazolium chloride (Sigma) for 6 to 24 h at 37°C in a 5% CO₂ incubator. Colonies containing more than 50 cells were visually counted.

Analysis of gene expression profiles using cDNA arrays and quantitative RT-PCR. The commercial Atlas Human Cancer cDNA Expression Array kit (catalog no. 7742-1) supplied by Clontech was used to establish gene expression array profiles. The isolation of total cellular RNA, the preparation of cDNA probes, and the hybridization of the cDNA array membranes followed the protocols supplied by the manufacturer (Clontech manual PT3140-1, version PR89832). Following hybridization, the membranes were exposed to a PhosphorImager screen, the screen was developed, and the resulting hybridization images were analyzed and quantitatively assessed using ImageQuant software (Molecular Dynamics, Inc.).

Nude mouse tumor formation assays. Female BALB/c (nu/nu) mice 4 to 6 weeks old were purchased from The Jackson Laboratory, Bar Harbor, Maine. For tumor formation assays, approximately 5 × 10⁶ M/tet/Vect cells, HA-HMG-I-expressing cells, or HA-HMG-Y-expressing cells in serum-free medium were injected into either the mammary fat pad or the subcutaneous space under the skin of the nude mice. The mice were maintained in a germfree facility, and tumor incidence and size were observed at regular intervals and documented.

Histology and immunohistochemistry. Tumor nodules were fixed in 10% formalin and embedded in paraffin, and 4- to 6-µm sections were mounted onto slides and stained with hematoxylin and eosin by the histochemistry core facility of the Center for Reproductive Biology, Washington State University. Immunostaining of tissue sections that had been preserved in Histochoice fixative (Sigma) followed standard protocols. Antigen retrieval for vimentin detection was done by microwave heating in citrate buffer. Before adding horseradish peroxidase-conjugated reagents, endogenous peroxidase activity was blocked by a methanol-H₂O₂ solution. For immunolocalization studies, tissue sections were reacted with a primary antibody (either a monoclonal mouse antibody or a polyclonal rabbit antibody), washed, and then reacted with either biotinylated goat anti-mouse immunoglobulin G (IgG; Vector Laboratories; 1:200) or biotinylated goat anti-rabbit IgG (Bio-Rad; 1:1,000). Biotinylated antibodies were visualized by reaction with horseradish peroxidase-streptavidin reagent (Vector Laboratories) and diaminobenzidine staining, followed by counter staining of nuclei with Meyer's hematoxylin. The antibodies and dilutions used for tumor immunostaining were anti-HA tag mouse monoclonal antibody 12CA5 (1:50 dilution), anti-pan-cytokeratin (C11) mouse monoclonal antibody (1:50 dilution; Santa Cruz Biotechnology), and anti-vimentin (V9) mouse monoclonal antibody (1:50 dilution; Santa Cruz Biotechnology). Both the mouse monoclonal anti-human procollagen alpha type I antibody M-38 (1:50 dilution; developed by J. A. McDonald) and the mouse monoclonal anti-procollagen alpha type III antibody SP1.D8 (1:50 dilution; developed by H. Furthmayr) were obtained from the Developmental Studies Hybridoma Bank, a contract facility supported by the National Institute of Child Health and Human Development and maintained by the Department of Biological Sciences, University of Iowa, Iowa City. Rabbit polyclonal antibody 8691 (1:40 dilution), a general anti-human alpha 1 (XI) collagen antibody, and rabbit polyclonal antibody 1703 (1:40 dilution), specific for the v2 and vlav2 isoforms of human alpha 1 (XI) collagen associated with mesenchymal and not cartilaginous tissues, were both generous gifts from Julia Oxford, Oregon Health Sciences University.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Endogenous HMGI(Y) protein levels correlate with the tumorigenic potential of human mammary epithelial cells. In preliminary experiments, a number of different normal and malignant human breast epithelial cell types were examined for both the levels of endogenous HMGI(Y) proteins and the ability to respond to artificial manipulation of these levels by the introduction of antisense or dominant-negative HMGI(Y) expression vectors. Among those tested were two matched pairs of human mammary epithelial cell lines (lines Hs578Bst and Hs578T and lines MCF-7 and MCF-7/PKC). Each pair was originally derived from the same parent, but the individually lines exhibit markedly different tumorigenic phenotypes. The aneuploid Hs578T cell line originated from a human mammary epithelial carcinoma (46) and grows aggressively in soft agar, exhibits invasive properties in in vitro matrigel outgrowth assays, and causes tumors that extensively metastasize in immunodeficient nude mice (111). In contrast, the diploid Hs578Bst cell line is a nontumorigenic myoepithelial cell line established from normal breast tissue of the same patient (46) but, unlike Hs578T, does not grow in soft agar, is not invasive in matrigel assays, and does not cause tumors in nude mice. The human MCF-7 breast epithelial cell line is estrogen receptor positive and retains many of the biochemical and phenotypic characteristics of normal mammary epithelial cells (76). It has only a very limited ability to grow in soft agar (128), is noninvasive in matrigel assays, and does not form tumors in nude mice (111, 128). On the other hand, the MCF-7/PKC cell line is a derivative of MCF-7 cells that has been stably transformed with a transgenic cDNA that codes for the transforming oncogene protein kinase C (PKC). This PKC-over expressing cell line readily growns in soft agar in vitro, is invasive in matrigel assays, and readily produces aggressive tumors that metastasize when injected into nude mice (128).

View larger version (22K): [in this window] [in a new window]   FIG. 1.   Expression of endogenous HMGI(Y) proteins correlates with tumor phenotype. Equal amounts of total cellular protein (~5 µg) extracted from cells with different tumorigenic potentials were separated by electrophoresis on a 4 to 20% gradient polyacrylamide gel, and the proteins were transferred onto a nitrocellulose membrane. (A) Western blot obtained with rabbit MR19 polyclonal antibody against the HMGI(Y) proteins. Under the electrophoretic conditions employed, the HMG-I and HMG-Y isoforms of the proteins comigrate as a single band. Lanes: 1, HeLa cells; 2, MCF-7/PKC cells; 3, MCF-7 cells; 4, Hs578T cells; 5, Hs578Bst cells. (B) Histone H1 proteins on the same membrane stained with Coomassie brilliant blue. The H1 histones served as an internal standard for equal protein loading on the gel.

Antisense HMGI(Y) expression inhibits cell proliferation and anchorage-independent growth. To further explore the role of HMGI(Y) in tumorigenesis, transfection experiments were performed in order to modulate the amounts of these proteins in various epithelial cell lines in order to determine whether there is a mechanistic connection between the endogenous level of HMGI(Y) proteins and cell proliferation rates or other phenotypic characteristics. Previous studies have demonstrated that the endogenous levels of HMGI(Y) proteins can be dramatically reduced in cells by the introduction of antisense expression vectors (8, 102) and that both antisense and dominant-negative expression vectors can significantly inhibit the in vivo transcription of genes regulated by the HMGI(Y) proteins (57, 58, 121). Figure 2 shows the results of experiments in which either the antisense pRcCMVIGMH (57) vector or, as a control, the empty parental pRcCMV vector without any insert was individually transfected into MCF-7, Hs578T, and HeLa cells. The transfectants were selected with G418 for a week, and pools of the G418-resistant cells from each cell line were then subcultured and their relative growth rates were determined by cell counting. As is evident from Fig. 2, expression of antisense HMGI(Y) inhibits the growth of MCF-7 cells (panel A) and Hs578T cells (panel B) but not HeLa cells (panel C). Control experiments with cotransfected reporter genes as internal reference standards demonstrated that this difference in the degree of growth inhibition in the different cell lines as a result of antisense HMGI(Y) expression is not due to the differences in transfection efficiency between the different cell lines (data not shown). The question therefore arises of why growth of the MCF-7 and Hs578T cells is significantly inhibited by antisense HMGI(Y) expression but growth of HeLa cells is not. A likely answer lies in the relative levels of endogenous HMGI(Y) proteins found in these different cell lines (Fig. 1). As shown in Fig. 2D, when the levels of HMGI(Y) proteins in the transfected cells were determined by Western blot analysis, it was seen that the antisense vector reduced the amount of proteins remaining in the MCF-7 and Hs578T cells (lanes 3 and 5, respectively) to a far greater extent than it did in the HeLa cells (lane 7). Thus, it is reasonable to propose that the reason why the growth rate of HeLa cells is unaffected by the antisense expression vector is that it is unable to reduce the very high concentrations of endogenous HMGI(Y) proteins found in HeLa cells to a level that would slow cell proliferation. Nevertheless, these transfection results clearly indicate that antisense HMGI(Y) expression impacts the growth rate of cells with low-to-moderate levels of endogenous HMGI(Y) proteins. Additional transfection experiments also demonstrated that antisense expression inhibited the anchorage-independent growth of some lines of mammary epithelial tumor cells in soft agar (data not shown). Taken together, the results from these two independent lines of antisense experiments indicate that reduction of endogenous HMG-I(Y) levels in cells can lead to a decrease in their proliferation rates and a reduction of the ability of some tumor cells to grow anchorage independently.

View larger version (34K): [in this window] [in a new window]   FIG. 2.   Expression of antisense HMG-inhibits tumor cell proliferation. MCF-7 (A), Hs578T (B), and HeLa (C) cells were transfected with 6 µg of either the empty pRcCMV expression vector () or the pRcCMV-IGMH (antisense HMG-I) vector (). The transfectants were selected with G418 at 150 µg/ml in culture medium for 7 days, and then the viable cells were subjected to a cell growth rate assay. Data were collected from three independent experiments with four dishes for each time point per experiment. (D) Western blots of HMGI(Y) proteins in MCF-7 cells (lanes 2 and 3), Hs578T cells (lanes 4 and 5), and HeLa cells (lanes 6 and 7) transfected with either the empty control vector (lanes 2, 4, and 6) or the antisense HMG-I vector (lanes 3, 5, and 7). Approximately equal amounts of total acid-soluble proteins were loaded into all of the lanes, and the Western blots were probed with a rabbit polyclonal antibody (MR19) against the HMGI(Y) proteins. On these gels, the HMG-I and HMG-Y proteins comigrated as a single band.

Overexpression of exogenous HMGI(Y) proteins promotes anchorage-independent growth in soft agar. A series of preliminary transfection experiments demonstrated that transfection of MCF-7 cells with a wild-type HMG-I or HMG-Y vector whose transcriptional expression was driven by a highly efficient constitutive promoter resulted in a marked increase in the total amount of HMGI(Y) proteins in the recipient cells and also resulted in a significant increase in their ability to grow anchorage independently in soft agar (data not shown). There is, however, an intrinsic problem that interferes with an unambiguous interpretation of these and other, similar studies (37, 94, 132). Namely, owing to the fact that HMGI(Y) proteins can influence the transcriptional activity of the host cell's endogenous Hmgiy gene promoter (unpublished observations), it is difficult, if not impossible, to distinguish between the amount of HMGI(Y) proteins in the transfected cells contributed by expression of the introduced transgene and the amount of these proteins contributed by expression of the endogenous host cell gene.

View larger version (33K): [in this window] [in a new window]   FIG. 3.   Identification of HA-HMG-I- and HA-HMG-Y-expressing clones using immunoprecipitation and Western blots. Total cellular proteins were isolated from 5 × 106 cells from individual transgenic MCF7 clones treated without (ON) or with (OFF) tetracycline at 10 µg/ml, and the extracts were immunoprecipitated using the 12CA5 monoclonal antibody directed against the HA tag peptide of the transgenic proteins. The precipitated proteins were separated by SDS-15% PAGE and then transferred onto a nitrocellulose membrane. Western blotting of the membrane was performed using the MR19 antibody against the HMGI(Y) proteins. Under these electrophoresis conditions, the HMG-I and HMG-Y isoform proteins comigate on the gel. Individual transgenic clones expressing either the HMG-I (HA-I-Cs, HA-I-7C, and HA-IC14) or the HMG-Y (HA-Y-C10, HA-Y-C19, and HA-Y-C21) isoform protein are indicated. The +/ designation indicates that tetracycline was added (+) to expressing ON cells () several days prior to protein extraction in order to inhibit transgene expression.

View larger version (116K): [in this window] [in a new window]   FIG. 4.   Anchorage-independent growth of MCF-7/HA-I and MCF-7/HA-Y clones. Assays for growth in soft agar of MCF-7/HA-Y cells (A and B) and MCF-7/HA-I cells (C and D) were performed as described in Materials and Methods. The agar medium in the plates shown in panels A and C contained tetracycline at 10 µg/ml (i.e., transgene expression was off), whereas those in panels B and D were without tetracycline (i.e., transgene expression was on).

Array and RT-PCR analyses of gene transcription profiles in cells overexpressing HMG-I and HMG-Y. To identify the tumor-promoting genes that are altered by HMG-I and HMG-Y overexpression in vivo, Atlas Human Cancer cDNA Expression Arrays (Clontech Inc.) were used to generate tumor gene expression profiles. The array was a carefully selected collection of cDNA fragments arranged on nylon membranes for rapid assessment of the differential gene expression of 588 tumor-associated genes. These well-characterized genes broadly represented many crucial cellular signaling pathways and other complex biological functions that have been associated at the molecular level with changes underlying tumor progression and metastasis.

View larger version (92K): [in this window] [in a new window]   FIG. 5.   Array analysis of gene expression in transgenic cells over- expressing HA-HMG-Y. Total RNA was isolated from control off M/tet/Vect cells not expressing HMG-Y and on HA-Y-C21 cells expressing high levels of HA-HMG-Y protein and reverse transcribed into cDNA with 32P labeling. The two cDNA probes were separately hybridized with identical Atlas Array nylon membranes (Clontech) on which 588 genes associated with human cancer were spotted in duplicate. The membranes were then developed by exposure to a phosphorimaging screen, and the profiles of the two RNA populations were quantitatively compared. (A) One small section of the large array membrane hybridized with a probe from control off cells. (B) A corresponding region of a duplicate membrane hybridized with a probe from HA-HMG-Y on cells. The small diagonal arrows indicate corresponding positions on the two membrnaes where significant differences in spot intensities can be observed.

View larger version (58K): [in this window] [in a new window]   FIG. 6.   RT-PCR analysis of gene expression profiles regulated by HMG-I and HMG-Y indicates that the two protein isoforms do not have identical functions in vivo. The DNA-free RNAs from two HMG-Y-expressing clones (M8AC10 and M8AC19), two HMG-I-expressing clones (M7CC7 and M7CCs), and control M/tet/Vect cells were subjected to semiquantitative RT-PCRs as described in Materials and Methods. The PCR products were electrophoretically separated by SDS-15% nondenaturing PAGE. The results of this analysis demonstrate that while both HMG-I and HMG-Y upregulate the expression of integrin 1 and MMP13, HMG-Y, but not HMG-I, upregulates Notch-4 and Jagged-1. (A) Co-RT-PCR of integrin 1 and HPRT messages from different transgenic cell clones. Lanes: 1, M/Tet/Vec; 2, M8AC10; 3, M8AC19; 4, M8AC21; 5, M7CC7; 6, M7CCs. (B) Co-RT-PCR of MMP13 and HPRT messages from different transgenic cell clones. Lanes: 1, M7CCs; 2, M7CC7; 3, M8AC21; 4, M8AC19; 5, M8AC10; 6, M/Tet/Vec. (C) Co-RT-PCR of Notch-4 and HPRT messages from different transgenic cell clones. Lanes: 1, M/Tet/Vec; 2, M8AC10; 3, M8AC19; 4, M8AC21; 5, M7CC7; 6, M7CCs. (D) Co-RT-PCR of Jagged-1 and HPRT messages from different transgenic cell clones. Lanes: 1, M/Tet/Vec; 2, M8AC10; 3, M8AC19; 4, M8AC21; 5, M7CC7; 6, M7CCs.

Genes regulated by HMG-Y overexpression. The 588 cDNAs present on the membrane array represent 13 different categories of genes that have been reported to play key roles in different biological processes of tumor progression and metastasis. These categories include the following: A, cell cycle and growth regulators (79 genes); B, intermediate filament markers (19 genes); C, apoptosis regulators (69 genes); D, oncogenes and tumor suppressors (29 genes); E, DNA damage response, repair, and recombination genes (33 genes); F, cell fate and development regulators (23 genes); G, receptors (42 genes); H, cell adhesion, motility, and invasion genes (82 genes); I, angiogenesis regulators (16 genes); J, invasion regulators (39 genes); K, Rho family small GTPases and their regulators (21 genes); L, cell-cell interaction genes (38 genes); M, growth factors and cytokines (98 genes).

View larger version (55K): [in this window] [in a new window]   FIG. 7.   Diversity of genes that are regulated by overexpressed HMG-Y proteins. Transcription profiles determined by cDNA array analyses indicate that overexpression of HMG-Y protein either up- or downregulates a broad range of genes thought to be involved with various aspects of neoplastic transformation, tumor progression, and metastasis in human cells. The graph depicts the percentages of genes in various groups or categories of cancer-related genes present on the cDNA array membrane. Conservatively, we consider only those genes whose expression is either up- or downregulated by a factor of 4 or more to be the most important ones modulated in vivo by HMG-Y overexpression. At the bottom is a list of the various categories of genes represented on the cDNA array membrane. For example, in group G, HMG-Y upregulates ~12% and downregulates ~10% of the receptor genes present on the membrane.

Overexpression of HMG-Y promotes tumor formation and metastasis in nude mice. The tumorigenicity and metastatic potential of MCF-7 cells overexpressing tetracycline-regulated exogenous HMG-I and HMG-Y were evaluated by injection of cells expressing these HA-tagged proteins from a number of different clones into immunodeficient BALB/c (nu/nu) nude mice. When 5 × 10⁶ cells were injected into either the mammary fat pad or the subcutaneous space, the HMG-Y-expressing clones (HA-Y-C19 and HA-Y-C21) formed tumors. No mice injected with HA-I-C7 clones expressing the HMG-I protein or control clones expressing only the empty M/tet parental vector developed tumors within a 3-month observation period, as summarized in Table 3. When inoculated subcutaneously into the backs of mice, HA-Y-C19 cells formed solid primary tumor nodules with local invasion of the surrounding tissues (data not shown). Morphological examination indicated that these tumors were well circumscribed and bordered by what appeared to be a network of thin, fibrous tissue. Histological examination of the HA-HMG-Y-induced primary tumors revealed that they were composed of solid masses of tightly packed carcinoma cells with a highly necrotic center. No metastases of cells to other areas of the body were observed when cells were injected subcutaneously (unpublished observations).

View this table: [in this window] [in a new window]   TABLE 3.   Overexpression of HMG-Y proteins leads to tumor growth in immunodeficient nu/nu nude mice

View larger version (85K): [in this window] [in a new window]   FIG. 8.   MCF-7 cells overexpressing HMG-Y protein form tumors in nude mice. Metastatic tumors developed in the mesentery membranes and in the lining of the peritoneal cavity when the mammary fat pads of nude mice were injected with approximately 5 × 106 cells (in this case, clone HA-Y-21) expressing the HA-HMG-Y protein. Many of the metastatic tumor nodules were seen in mesenteric membranes within and lining the peritoneal cavity (A), as well as in the greater omentum and other mesenteries supporting the small intestines (B) and in the mesenteries supporting and lining the colon (C).

View larger version (131K): [in this window] [in a new window]   FIG. 9.   Histological analysis of sections of metastatic tumors formed in nude mice overexpressing HA-HMG-Y protein. (A to F) Hematoxylin-and-eosin staining. (G to L) Immunohistochemical analysis with specific antibodies and counterstaining with Meyer's hematoxylin. (A) Paraffin section of a secondary metastatic carcinoma with a cobblestone-like epithelial morphology (stained blue) formed by HA-Y-C21 cells located near a blood capillary shown in the upper part of the picture. Magnification, ×100. (B) Section of a secondary metastatic carcinoma formed by clone HA-Y-C21 cells showing that the tumor (blue) has invaded the striated muscle of the diaphragm (pink with striations) in close proximity to blood capillaries filled with red cells. Magnification, 100×. (C) Region of a metastatic tumor showing the coexistence of collage-rich and non-collagen-rich areas within the tumor mass. In this section, an island of cobblestone-like carcinoma cells containing little extracellular collagen (lower left) is surrounded by areas of more elongated, spindle-like cells with a dense extracellular collagen matrix (stained pink). Magnification, 40×. (D to F) Regions of carcinosarcoma-like cells found in three independently derived HA-Y-C21-induced tumors. Magnification, 60×. (G) Carcinoma region reacted with a control nonimmune mouse IgG antibody. Magnification, 100×. (H) Carcinoma region reacted with mouse monoclonal IgG (12CA5) specifically directed against the HA tag peptide. Magnification, 100×. Note the prominent nuclear staining (arrows), although cytoplasmic staining is also evident. (I) Carcinosarcoma-like region reacted with anti-HA tag mouse monoclonal IgG. Magnification, 40×. (J) Carcinosarcoma-like region reacted with mouse monoclonal IgG (M-38) directed against human fibroblast procollagen type I (which has no cross-reactivity with mouse collagens). Magnification, 40×. (K) Disorganized region of tumor reacted with a polyclonal rabbit antibody (1703) directed against the v2 and vlav2 isoforms of human alpha 1 (XI) collagen associated with mesenchymal, but not cartilagenous, tissues. Magnification, ×60. (L) Mixed carcinoma and carcinosarcoma-like regions reacted with mouse monoclonal IgG (V9) directed against human vimentin (no reaction with vimentin of mouse origin or other intermediate filaments. Magnification, ×40.

HMG-Y overexpression induces EMT in tumor cells. Immunohistochemical staining of tumor sections was performed using a number of monoclonal and polyclonal antibodies directed against specific proteins to confirm that the tumors formed in nude mice had, indeed, originated from the transgenic MCF-7 cells and to further investigate the possibility that these cells had undergone an EMT. The antibodies used for these studies were selected on the basis of the gene products indicated by cDNA array analysis to be produced in the transgenic cells (Table 2) and on the basis of the histological appearance of the tumors themselves (Fig. 9A to F). For example, reaction of tumor sections with a monoclonal antibody directed against the HA tag peptide show that the transgenic human HMG-Y protein was localized primarily in the nuclei (but with detectable cytoplasmic staining) of cells in both regions of carcinoma morphology (Fig. 9H) and regions of carcinosarcoma morphology (Fig. 9I). These results unambiguously demonstrate that the tumors originated from the injected human MCF-7 cells overexpressing the HMG-Y protein, a fact also confirmed by the ability to reisolate tetracycline-regulated transgenic cell lines from the tumors by growth in selective medium (data not shown).

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

HMGI(Y) proteins are causal agents in tumor progression. As recounted in the introduction, published reports of human clinical studies, as well as numerous studies of animal and cell culture model systems, have firmly established that aberrant expression or overexpression of members of the HMGI(Y) gene family positively correlates with neoplastic transformation and metastatic progression of a wide variety of tumors. Nevertheless, the results reported here represent the first direct demonstration that overexpression of the HMGI(Y) proteins is causally involved in modulating the expression of genes involved in all stages of tumor progression from transformation to anchorage-independent growth in vitro, to the formation of primary and metastatic tumors in nude mice, and the induction of in EMT in vivo. The current data unambiguously confirm the long-held suspicion the Hmgiy is a bone fide proto-oncogene and also provide insights into probable molecular events underlying its in vivo mode of action.

In vivo, the HMG-I and HMG-Y isoform proteins are not functionally equivalent. One of the most unexpected findings in this study was that transgenic cells overexpressing the HMG-Y isoform protein were much more effective in inducing both primary and metastatic tumors when injected into nude mice than were transgenic cells overexpressing the HMG-I isoform protein (Table 3). In less than 2 months following injection, over half (four of seven) of the nude mice injected with cells overexpressing the HMG-Y protein had developed tumors whereas during this same time period none of the five mice injected with cells overexpressing HMG-I developed tumors. Only much later (after about 4 months) did one of the mice injected with cells overexpressing the HMG-I protein develop small primary tumors (data not shown). Thus, HMG-Y is far more efficient in promoting tumor progression in vivo than HMG-I even though these isoform proteins differ only by an internal deletion of 11 amino acids in the former. These results were initially puzzling, since expression of both isoform proteins efficiently induced anchorage-independent growth in soft agar (Fig. 4; Table 1). Nevertheless, they were consistent with an earlier report demonstating that the tumor promoter 12-O-tetradecanoylphorbol acetate preferentially induces HMG-Y protein expression in transformation-sensitive, but not in transformation-resistant, mouse JB6 epithelial cells in culture (22). The cumulative data therefore strongly imply that, in addition to sharing many common biological functions, the HMG-I and HMG-Y proteins also exhibit important differences in their in vivo functions. This view is further supported by the finding that in vivo, the HMG-Y protein contains many more secondary biochemical modifications than the HMG-I isoform and that these modifications differentially affect the substrate-binding properties of the two proteins (5). A potential difference in function between the two isoforms is likewise supported by the results of the cDNA expression profiles and RT-PCR experiments reported earlier (Fig. 5 and 6; Table 2). The data from these experiments demonstrate that the HMG-I and HMG-Y proteins do, indeed, regulate the transcriptional expression of a large constellation of common genes but also indicate that, in addition, each protein can regulate its own independent set of genes. For example, both proteins upregulate the expression of genes such as those which encode the p38 mitogen-activated protein kinase (MAPK), collagen type XVIII, integrin 1, MMP13, and TIMP (tissue inhibitor of MMP3), but on the other hand, only HMG-Y upregulates genes such as those for Notch-4, Jagged-1, and Wnt-10B (Fig. 6) and only HMG-I downregulates the genes for Wnt-10B and Wnt-8B (data not shown). It is reasonable to suspect that the constellation of genes regulated by both the HMG-I and HMG-Y proteins is responsible for the anchorage-independent growth in soft agar of cells overexpressing these proteins. On the other hand, it is likely that some combination of the individual genes that are either up-or downregulated by the HMG-Y or HMG-I isoform protein is responsible for the much greater in vivo tumorigenic and metastatic potential of the HMG-Y protein. It should be stressed that the cDNA array analyses reported here sampled the transcriptional activity of only a very limited number of cellular genes. To fully address the issue of the extent to which the HMG-I and HMG-Y isoform proteins have distinct biological functions in tumor progression, a much more comprehensive transcriptional expression analysis employing high-density oligonucleotide array technology is necessary.

Genes modulated by HMGI(Y) overexpression. The list of genes differentially regulated by HMG-Y overexpression in Table 2 demonstrates a broad range of functional activity of the HMG-Y architectural transcription factor. During mouse development, very high expression of HMGI(Y) proteins has been detected in all embryonic tissues up to 8.5 days of gestation, a period during which the most critical events of organogenesis start to take place, and its expression continues at lower levels to the end of gestation (17, 78). Based on these patterns of embryonic expression, the suggestion has been made that the HMGI(Y) proteins are not only involved with cell proliferation but also are likely to be involved with the establishment of various cell types during organogenesis. This suggestion is supported by the present data, which demonstrate that overexpression of HMG-Y significantly upregulates genes known to be involved in development, such as those for Frizzled 5 (161-fold), Wnt-13 (14.6-fold), Wnt-10B (4.1-fold), as well as Jagged-1 and its receptor, Notch-4 (7.5- and 4.1-fold, respectively). HMG-Y also upregulates a number of genes involved in both cell cycle regulation and signal transduction. These include the genes for CLK-1 (14.6-fold), cdc25A (98.6-fold), cdc25B (9.1-fold), cyclin C (5.1-fold), JNK2 (7.4-fold), and p38 MAPK (5.7-fold) and others. Each of these genes has been demonstrated not only to be involved in either the control of cell cycle or signaling processes but also to participate in neoplastic transformation and/or tumor progression in various systems. For example, the genes coding for the protein phosphatase enzymes cdc25A and cdc25B are well known as G₁/S transition and G₂/M transition regulators, respectively. In addition, they are all proto-oncogenes whose overexpression has been implicated in many types of tumors, including breast cancers (60, 135).

HMG-Y modulates expression of EMT genes. Although studies analyzing the molecular characteristics of both mesenchymal and epithelial cells are still at an early stage, a group of genes has already been identified that is both characteristic of these two different cell types and influential in EMT (reviewed in reference 9). Analysis of the cDNA profile of epithelial cells overexpressing HMG-Y protein indicates that transcriptional expression of several EMT marker genes is modulated in HMG-Y overexpressing cells (Table 2). Among the group of upregulated genes are type I, III, and IV collagens, vimentin, certain cytokeratins, a number of matrix metalloproteases, basic FGF receptor 1, FGF7, and SF (scatter factor-hepatocyte growth factor) protein. Amoung the group of downregulated genes are those coding for the EGF receptor and TGF- receptor III, as well as those coding for both the Met and FGFR2b proteins.