Purpose: The growth-related oncogene α (GROα) is a secreted interleukin-like molecule that interacts with the CXCR2 G-protein–coupled receptor. We found that the mRNA and protein products of GROα are more highly expressed in neoplastic than normal colon epithelium, and we studied potential mechanisms by which GROα may contribute to tumor initiation or growth.
Experimental Design: Cell lines that constitutively overexpress GROα were tested for growth rate, focus formation, and tumor formation in a xenograft model. GROα expression was determined by Affymetrix GeneChip (241 microdissected colon samples), real-time PCR (n = 32), and immunohistochemistry. Primary colon cancer samples were also employed to determine copy number changes and loss of heterozygosity related to the GROα and fibulin-1 genes.
Results: In cell cultures, GROα transfection transformed NIH 3T3 cells, whereas inhibition of GROα by inhibitory RNA was associated with apoptosis, decreased growth rate, and marked up-regulation of the matrix protein fibulin-1. Forced expression of GROα was associated with decreased expression of fibulin-1. Expression of GROα mRNA was higher in primary adenocarcinomas (n = 132), adenomas (n = 32), and metastases (n = 52) than in normal colon epithelium (P < 0.001). These results were confirmed by real-time PCR and by immunohistochemistry. Samples of primary and metastatic colon cancer showed underexpression of fibulin-1 when compared with normal samples. There were no consistent changes in gene copy number of GROα or fibulin-1, implying a transcriptional basis for these findings.
Conclusion: Elevated expression of GROα is frequent in adenocarcinoma of the colon and is associated with down-regulation of the matrix protein fibulin-1 in experimental models and in clinical samples. GROα overexpression abrogates contact inhibition in cell culture models, whereas inhibition of GROα expression is associated with apoptosis. Importantly, coexpression of fibulin-1 with GROα abrogates key aspects of the transformed phenotype, including tumor formation in a murine xenograft model. Targeting GRO proteins may provide new opportunities for treatment of colon cancer.
There is substantial information that the secreted interleukin-like molecule, growth-related oncogene α (GROα) functions as an oncogene in melanoma development (1, 2). Up-regulation of GROα expression at the mRNA and protein level is observed in melanoma, and evidence indicates that GROα plays a pivotal role in this neoplasm. We planned these experiments to examine the role of GROα in colorectal cancer and the mechanisms by which it might participate in colon cancer initiation (3).
Apart from a potential role as an oncogene, GROα has a variety of other putative functions, such as directing inflammatory cell migration, promoting angiogenesis, and participating in wound healing (4, 5). There is evidence that GROα induction upon HIV-1 infection stimulates HIV-1 replication in macrophages and T lymphocytes (6). During central nervous system development, GROα is involved in proliferation and migration of oligodendrocyte progenitors (7, 8). It may also be involved in the pathogenesis of Alzheimer's disease due to its role in inflammation (9).
GROα is a 73-amino-acid protein structurally and functionally related to its isoforms GROβ and GROγ (10) and to interleukin-8. All belong to the α chemokine superfamily (11). Alternative designations for this family are CXCL1 (GROα), CXCL2, CXCL3, and CXCL8 (GROβ, GROγ, and interleukin-8, respectively). They have a conserved CXC motif at the NH2 terminus but differ at the COOH terminus. GROβ and GROγ can also promote melanoma development by overexpression in immortalized melanocyte cells (12). The three different GRO chemokines differ in the binding affinity to their receptor CXCR2 (a G-protein–coupled receptor), with GROα bearing the highest affinity (13).
Transcription rate seems to be an important mechanism for regulating the activity of GROα. The GROα gene was first identified by its inducible expression in response to platelet-derived growth factor induction (14). GROα gene expression is induced by cytokines, including interleukin-1 (15) and tumor necrosis factor-α (16), probably through the nuclear factor κB signaling pathway (17, 18). Bacterial products, such as lipopolysaccharide (19), cell surface heparin sulfate proteoglycan (20), and Shiga toxins (21), may also increase GROα expression. Transcription factors HMGA1 and Sp1 up-regulate GROα gene expression by binding to the cis-elements within the GROα promoter region (22). Conversely, a human CUT homeodomain protein/CCAAT displacement protein can negatively regulate GROα gene expression by interacting with the GROα promoter (23).
Several downstream signaling pathways are involved following the interaction of GROα with the CXCR2 receptor. These include the Ras-mitogen-activated protein kinase pathway, phosphatidylinositol 3-kinase pathway (9), nuclear factor κB pathway (18), and PAK signaling (24). Activation of these signaling pathways by GROα may lead to different GROα cellular responses and functions.
In this study, we confirmed that GROα mRNA expression is dysregulated in early and late stages of colorectal cancer. Evidence from both overexpression and inhibitory RNA (siRNA) experiments suggest that this dysregulation promotes tumor induction or propagation. Furthermore, we provide laboratory and clinical evidence that, in colorectal cancer, increased expression of GROα is associated with and is probably responsible for attenuated expression of the extracellular matrix proteins fibulin-1C and fibulin-1D in adjoining stroma. Abnormal expression of these fibulin-1 proteins has been associated with other kinds of cancers (25, 26), and we hypothesize that regulation of fibulin-1 proteins is an important role of GROα.
Materials and Methods
Cell lines and medium. NIH 3T3 and HCT-15 cell lines were grown in 10-cm2 tissue culture plates in DMEM and RPMI 1640, respectively, and supplemented with 10% fetal bovine serum. Cell stocks were purchased from the American Type Tissue Collection (Manassas, VA). Culture media were purchased from the American Type Tissue Collection or Life Technologies/Invitrogen (Carlsbad, CA) as was fetal bovine serum, G-418, Trypsin-EDTA, and streptomycin. Cells were maintained using routine cell culturing and subculturing protocols and incubated at 37°C and 5% CO2.
Construction of HCT-15 permanent GROα RNA interference cell line. Colon cancer cell line HCT-15 was stably transfected with GROα siRNA using a protocol supplied with the GeneSuppressor System (IMGENEX, San Diego, CA). The 21-nucleotide RNA interference target sequence was from the 235th base downstream of the start codon (AAGAATGGGCGGAAAGCTTGC). Several cell lines (SIMG-1, SIMG-4, and SIMG-5) were developed. The control cell line SIMG-C was transfected with vector alone.
Detection of DNA ladders. Approximately 5 × 105 cells (HCT-15, SIMG-C, SIMG-1, SIMG-4, and SIMG-5) were seeded in 10-cm2 cell culture plates. Four days later, cells were harvested, and DNA was extracted using the procedure supplied with Suicide-Track DNA ladder isolation kit (Oncogene Research Products, Cambridge, MA). The extracted DNA was subjected to agarose gel electrophoresis to detect formation of typical DNA ladders indicative of apoptosis.
Construction of GROα-overexpressing NIH 3T3 cell lines. GROα cDNA was obtained through PCR from a human cDNA library derived from normal colon epithelium (supplied by the Cooperative Human Tissue Network). PCR was primed with 5′-AGCTCTTCCGCTCCTCTCACA-3′and 5′-AGGGCCTCCTTCAGGAACA-3′. The 400-bp PCR product was cloned into pCR2.1-TOPO TA vector (Invitrogen, Carlsbad, CA). Three clones were isolated and each contained the correct GROα open reading frame. The above purified pCR2.1-TOPO containing GROα cDNA was cloned between the BamHI and XhoI sites of the 5.4-kb expression vector pcDNA3+ (Invitrogen). NIH 3T3 cells were stably transfected with this pcDNA3-GROα construct, and the resulting clones MB and MD were propagated under selection with G-418. A control cell line (PH) was derived from NIH 3T3 cells by transfection with pcDNA3+ with no insert.
All transfection procedures employed LipofectAMINE (Life Technologies/Invitrogen).
Construction of NIH 3T3 cell lines overexpressing both GROα and fibulin-1D. The fibulin-1D expression plasmid pPuro-Fibulin-1D was a gift from Dr. J McCormick (25). After confirming the sequence, it was transfected into MB cells. Control cells were established by transfecting MB with the empty vector pBABE. After confirming protein expression of both GROα and fibulin-1D expression, the resultant cell clones MBBABE, MBFBLN1(3), and MBFBLN1(6) were cultured in the presence of both puromycin and G-418.
Determination of cell number and viability. To evaluate cell number and viability, a 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium assay was done using the CellTiter 96 Aqueous One Solution Cell Proliferation Assay (Promega, Madison, WI). Cells were seeded in duplicate into 24-well plates at a density of 4 × 104/mL. Culture medium was replaced by 500 μL of fresh media after 48 hours. Each well received 100 μL of the 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium reagent, and the plate was incubated for 1 hour at 37°C in a humidified, 5% CO2 atmosphere. A 100-μL aliquot of the reaction mixture was transferred to a 96-well plate, and absorbance was measured at 490 nm.
Cell culture growth curves. To measure cell culture growth rate, the various lines PH, MB, MD, MBBABE, MBFBLN1(3), and MBFBLN1(6) cells were seeded at 50,000 per well in six-well plates and grown in DMEM containing 10% fetal bovine serum. G418 and puromycin were added as appropriate. Cells were counted periodically over 15 days with an improved Neubauer hemocytometer.
Western analysis. Total protein from the cell lysate of one 10-cm2 culture plate was obtained using 0.5 to 1 mL M-PER Mammalian Protein Extraction Reagent (Pierce Biotechnology, Rockford, IL) supplemented with protease inhibitors, including 0.1 μmol/L phenylmethylsulfonyl fluoride, 10 μg/mL leupeptin, and 10 μg/mL pepstatin. The cell lysate was centrifuged for 5 minutes, and the clear supernatant was taken and stored as total protein solution. Protein concentration was obtained by using the Bicinchoninic Acid Protein Assay Reagent kit (Pierce Biotechnology). Equal amounts of protein were separated on SDS-PAGE, and membrane transfer, primary antibody reaction, washing, and signal detection were carried out according to routine methods. To detect protein expression, rat anti-human GROα (U.S. Biological, Swampscott, MA) and mouse monoclonal anti-fibulin-1 (Santa Cruz Biotechnology, Santa Cruz, CA) were used. The fibulin-1 antibody recognizes amino acid residues 1 to 190, which is common to all isoforms of fibulin-1, including fibulin-C and fibulin-D. Therefore, Western blots employing this antibody to detect the human protein displays a series of bands, representing the various isoforms.
Tumor induction in nude mice. The ability of cell lines overexpressing GROα to produce xenograft tumors in nude mice (BALB/BALB, nu/nu) was evaluated. MB, MD, and PH (vector control) cells were harvested by scraping them off the culture plates into ice-cold PBS. Cells were injected s.c. into the flanks of three different groups of mice. Tumor formation was monitored for 17 days, at which time, the mice were sacrificed, and the tumors were surgically removed, measured, and submitted for microscopic analysis.
Subsequently, the ability of constitutively expressed fibulin-1D to suppress GROα-mediated tumor formation was evaluated. These experiments employed the previously described cell lines MBBABE (MB with empty pBABE expression vector), PHBABE (NIH 3T3 with two empty expression vectors, pBABE and pcDNA3), PHFBLN MBFBLN(3), and MBFBLN(6) (cell lines that constitutively express both GROα and fibulin-1). Based on preliminary titration experiments, 2 × 105 cells from each cell line were injected to the flanks of the nude mice, and the animals were observed for 5 months for tumor induction.
RNA extraction and purification. Total RNA was extracted from cultured cells using Trizol reagent (Invitrogen). RNA was further purified by an RNeasy purification column and reagents (Qiagen, Valencia, CA). RNA concentration was determined by recording UV absorption at 260 nm.
Expression profile analysis of colon tissue samples. To discern mRNA expression of GROα, GROβ, GROγ, and the receptor CXCR2 in neoplastic colon epithelium, mRNA was extracted from 241 mucosal samples from which adjacent connective tissue and muscle was manually microdissected under microscopic visualization. The remaining mucosal tissue was processed according to standard Affymetrix protocols, and the resulting labeled cRNA was hybridized to U133A GeneChips. In addition, the total RNA from the constitutively expressing GROα siRNA HCT-15 cell line (SIMG-1) and cells containing the empty vector (SIMG-C) was analyzed by the U133A GeneChip. MAS-3 software was used for obtaining raw data from the gene chips. Tissue samples included in this project were 24 normal colon samples, 32 adenoma tissue samples, 132 primary colon cancer tissue samples, and 52 metastasis samples (32 from the liver and 20 from the lung). Statistical testing was done with KaleidaGraph V 4.0 (Synergy Software). Significance was tested by Wilcoxon's test for paired data or ANOVA using Bonferroni all-pair comparison. Data are expressed as mean ± SE.
Real-time PCR. To quantify GROα and fibulin-1 mRNA concentration (normalized by glyceraldehyde-3-phosphate dehydrogenase), the following primer sets and probes were synthesized (Integrated DNA Technologies, Coralville, IA). For GROα, the forward oligo sequence was from 207 to 224 nucleotides on the open reading frame, which was TCCGTGGCCACTGAACTG. The reverse primer was GTGGCTATGACTTCGGTTTG. The fluorescent-tagged probe was 5,6-FAM-CAGACCCTGCAGGGAATTCAC-TAMRA. For fibulin-1, oligos were designed according to the common region shared by fibulin-1C and fibulin-1D. The forward and reverse primer sequences were GGAGCAGTGCTGCCACAG and AGCACCTCTTCACAAATGTG, respectively. The fluorescent-tagged probe was 5,6-FAM-AGCCTGGCCAACGAGCAGGA-36-TAMRA. For glyceraldehyde-3-phosphate dehydrogenase, the two forward and reverse sequences were ACAACTTTGGTATCGTGGAAGG and CAGTAGAGGCAGGGATGATGTTC, respectively. The fluorescent probe was 5,6-FAM-ACCCAGAAGACTGTGGATGG-3BHQ. To make the target amplification primer-probe mix, 20 μL of 100 μmol/L probe, 40 μL forward prime (100 μmol/L), 40 μL reverse prime (100 μmol/L), and 700 μL RNase-free water were mixed together to yield concentration of 2.5 μmol/L probe and 5 μmol/L of each primer in the target mix.
Reverse transcription was done using the Taqman Reverse Transcription Reagents (Applied Biosystems, Foster City, CA). The reaction was set up and carried following the provided procedures. The cDNA was stored at −20°C.
Each 25 μL PCR reaction was carried out by adding 2.5 μL target primer-probe mix, 12.5 μL Taqman 2× PCR Master Mix (Applied Biosystems), and 10 μL of sample cDNA described above. The PCR and fluorescent signal detection was done on the ABI Prism 7000. To quantify mRNA levels, a standard curve was generated for each transcript by amplifying different amounts of control cDNA known to express the gene of interest.
Immunohistochemistry for GROα, fibulin-1, and CXCR2. Using the previously described antibodies for GROα and fibulin-1, tissue paraffin-embedded samples containing normal colon, colon adenoma, and adenocarcinoma were examined to ascertain the magnitude and cellular pattern of expression of these proteins. Antibody to the receptor CXCR2 was employed to detect expression of that protein in colon cancer samples (rabbit polyclonal to CXCR2; Abcam, Cambridge, MA.).
Determination of gene copy number and loss of heterozygosity. Twenty-five primary colon cancer samples containing no more than 20% contamination from infiltrating stroma were employed to determine copy number changes and loss of heterozygosity (LOH) following the manufacturer's instructions (Genechip Human Mapping 50K Array Xba 240; Affymetrix, Santa Clara, CA). Using the supplied software (Copy Number Analysis tool), the regions where GROα (4q13.3) and fibulin-1 (22q13.31) are located were analyzed for areas of LOH and copy number changes. When available, LOH and copy number were determined by probe sets within the gene of interest. When probes were unavailable within a gene, probe sets immediately either side of the gene were used, and the relevant values were averaged. Any region with a score of ≥5 was identified as an area of LOH. Copy number calls that varied >2 SDs from normal samples were considered as significantly gained or lost.
GROα and GROγ is highly expressed in colorectal cancer tissues. Expression profiles using the U133A GeneChip from 241 microdissected colon tissue samples were reviewed for GROα mRNA expression. GROα expression averaged 4.6-fold higher in primary adenocarcinomas (n = 132) than in normal colon epithelium (P < 0.001). Expression of GROα in adenomas (n = 32) was also elevated relative to normal (P < 0.001), and in metastatic samples (32 liver and 20 lung), expression of GROα was intermediate between normal colon mucosa and primary colon cancer (Fig. 1A). The array data were confirmed by real-time PCR in paired normal and cancer samples from 16 patients (Fig. 1B). When normalized to glyceraldehyde-3-phosphate dehydrogenase, the expression of GROα was 0.64 ± 0.07 in the normal tissue and 3.2 ± 0.64 in the cancer samples (P < 0.001). To confirm that overexpression is related to neoplastic epithelial cells rather than inflammatory cells or non-epithelial stroma, we did immunohistochemistry staining for GROα, which established that GROα was expressed by neoplastic colon epithelial cells rather than inflammatory cells (Fig. 2A).
We also examined the mRNA expression of the GROβ and GROγ genes because their protein products each serve as ligands for the CXCR2 receptor. Although expression of GROβ did not vary between normal and neoplastic samples (P = 0.69), expression of GROγ expression averaged 8.0-fold higher in primary adenocarcinomas (n = 132) than in normal colon epithelium (P < 0.001). Expression of GROγ in adenomas (n = 32) and metastases (32 liver and 20 lung) was also elevated relative to normal (8.8-fold and 4.9-fold, respectively; P < 0.001 for all comparisons; Supplementary Fig. SA).
Expression of mRNA to the CXCR2 receptor was reduced in adenoma, primary colon cancer, and metastasis (P = 0.0017). The mean mRNA expression of this receptor was 23.1 ± 4.1 in normal epithelium, 14.6 ± 1.2 in the primary cancer, and 10.3 ± 1.4 in metastasis to lung and liver (Supplementary Fig. SB). Furthermore, immunostaining of tissue slices from cancer specimens for CXCR2 was variable and did not disclose the typical apical membrane pattern of staining that is observed for this receptor. This may imply that the target cell for GRO secretion by the neoplastic colon epithelial cell is not the colonocyte itself but a neighboring cell.
Growth of colorectal cancer cell line HCT-15 is inhibited by GROα siRNA. To evaluate whether expression of GROα is important to growth and viability of neoplastic colon cell lines, expression of GROα mRNA was knocked down in HCT-15 colon cancer cells. Three stable HCT-15 GROα siRNA expressing cell lines (SIMG-1, SIMG-4, and SIMG-5) were constructed. Reverse transcription-PCR documented that expression of GROα mRNA relative to glyceraldehyde-3-phosphate dehydrogenase following knock down was reduced to ∼50% to 60% of the level of expression in the control cell. The siRNA cell lines had a reduced rate of proliferation relative to the parental line or the control line (SIMG-C), which received a vector with no siRNA coding insert (Fig. 3A). Expression of siRNA to GROα was also associated with apoptosis in all three GROα siRNA cell lines (Fig. 3B), consistent with the known role of the related chemokine (interleukin-8) in suppressing apoptosis in some types of cells (27, 28). The apoptotic index increased from ∼10% in parental and control cells to 70% in cells receiving the GROα targeting vector.
Fibulin-1 expression is increased in HCT-15 cells transfected with siRNA to GROα. To understand the molecular mechanism of GROα function in oncogenesis, gene expression of cell lines SIMG-C (control cells) and SIMG-1 (containing siRNA to GROα) was monitored using the Affymetrix U133 A GeneChip. mRNA expression of numerous genes varied by >4-fold (data on file). Prominently, expression of fibulin-1C and fibulin-1D increased 12.1- and 14.9-fold, respectively (see Supplementary Table S1). Consistent with this change in mRNA expression, Western analysis indicated that protein levels of fibulin-1 were higher in SIMG-1 and SIMG-4 than in SIMG-C and HCT-15 (Fig. 3A, bottom). Fibulin-1 is represented by several bands because the antibody recognized multiple isoforms in humans.
Fibulin-1 expression is abrogated by forced expression of GROα in NIH 3T3 cells. These results in the HCT-15-derived colon cancer cell lines suggested that GROα expression might suppress expression of fibulin-1. To test this hypothesis, NIH 3T3 fibroblasts were transfected with a vector containing a human cDNA for GROα. Stable cell lines were selected by passage in G-418, and GROα protein expression was confirmed by Western analysis (Fig. 4A). Fibroblasts that received control vector with no insert (cell line PH) displayed normal growth characteristics. However, cell lines with forced expression of GROα (lines MB and MD) displayed characteristics of transformed cells, including focus formation (data not shown) and loss of contact inhibition when grown in culture (Fig. 4B). Furthermore, s.c. injection of MB or MD cells into the flank of nude mice resulted in tumor growth within 17 days, whereas the control PH cells did not induce tumor growth after 60 days. These tumors had the histologic appearance of a fibrosarcoma (Fig. 4C).
Next, fibulin-1 expression was examined in MB and MD cells by Western analysis. Consistent with the hypothesis that GROα negatively regulates fibulin-1 expression, fibulin-1 expression was suppressed in MB and MD cells (Fig. 4D), complementing the results of the GROα siRNA experiments.
To learn whether down-regulation of fibulin-1 was important to the transformation of NIH 3T3 cells by GROα, cell lines that overexpressed both GROα and fibulin-1D were created by transfecting MD cells with a fibulin-1D construct (Fig. 4E, inset). These resulting neomycin/puromycin–resistant stable cell lines MBFBLN(3) and MBFBLN(6) grew more slowly than the MB cells, in which only GROα was overexpressed (data not shown). To test whether forced expression of fibulin-1D could delay or abrogate GROα cellular transformation, 2 × 105 MBFBLN(3) and MBFBLN(6) cells were injected in the flanks of nude mice, with injection of MB cells transfected with empty BABE vector served as the control. As expected, control MB cells induced tumor growth after 4 weeks (Fig. 4E). MBFBLN(3) and MBFBLN(6) cells did not produce tumors over the 5-month observation period.
Fibulin-1C and fibulin-1D expression is reduced in human colorectal cancer. To extend these observations to human colorectal cancer, Affymetrix gene expression profiles from the same 241 colon samples were reviewed for expression of fibulin-1C and fibulin-1D mRNA. Expression of both isoforms was significantly reduced in primary colon cancer, polyps, and liver or lung metastasis relative to normal tissue (P < 0.0001 for each comparison, ANOVA with Bonferroni correction; Fig. 4F). The array data were confirmed by real-time PCR in paired normal and cancer samples from 16 patients. To extend the information about fibulin-1 to the tissue and protein level, immunohistochemistry images were obtained after probing colon cancer tissue samples with anti-GROα and anti-Fiblin-1. These images revealed elevated GROα expression and attenuated stromal fibulin-1 protein expression in colon cancer samples. Representative images are shown in Fig. 2.
Single nucleotide polymorphism analysis does not reveal consistent changes in GROα or fibulin-1 at the DNA level. To evaluate the possible role of GROα gene amplification or chromosomal duplication in producing elevated GROα expression, single nucleotide polymorphism array analysis was done on 25 primary cancer samples. This analysis revealed frequent LOH in chromosomal regions that contain the GROα gene (7 samples from a total of 26). Interestingly, LOH was not always accompanied by a loss in copy number: five samples were copy number neutral of the seven samples showing LOH, implying gene conversion events were also occurring. Furthermore, LOH in regions containing the fibulin-1 gene showed LOH in three samples and three different samples showed copy number changes (two decreases and one increase). These data imply that the down-regulation of fibulin-1 mRNA expression is due to a decrease in transcriptional activity and not to a decrease in copy number. Interestingly, tumor samples that showed LOH in regions containing the GROα gene showed no difference in mRNA expression when compared with those samples that showed no LOH. This observation suggests that the regulatory mechanism that is common to most clinical samples and that leads to the marked increase in GROα expression is not related to the infrequent and inconsistent genetic changes seen at the DNA level. Furthermore, sequencing of the GROα cDNA in several samples with elevated expression and LOH revealed no mutations of the coding region.
GROα is regarded as a proinflammatory, chemotactic molecule that affects leukocyte migration (29) and stimulates cell proliferation (14). It also functions in wound repair and inflammation, attracts neutrophils, and promotes angiogenesis (4, 5). The current work aims to define the role of GROα overexpression in colorectal cancer and the involvement of the subsequent down-regulation of the extracellular matrix protein fibulin-1.
GROα protein is highly expressed by melanoma cell lines and melanomas but not in normal melanocytes (30). The involvement of GROα in melanoma tumorigenesis has been well established and reviewed (1, 31). GROα expression is also increased in other malignancies, such as adrenal cancer (32), squamous cell carcinoma (33), and prostate cancer (34). The current report shows that this chemokine is overexpressed and plays an important role in colorectal cancer.
In melanoma, overexpression of GROα up-regulates M-ras expression at both the mRNA and protein levels, and it is likely that ras activation is required for MGSA/GROα–mediated transformation in melanocytes (30), through a ras/mitogen-activated protein kinase kinase kinase 1/p38/nuclear factor κB–mediated cascade (18). The roles of activating k-ras mutations in colon cancer are well established (35, 36); however, we found no difference in GROα or fibulin-1 mRNA expression levels in clinical samples containing these mutations compared with their wild-type counterparts (data not shown).
Using a combination of Affymetrix GeneChip experiments and real-time PCR, we found GROα overexpression at the mRNA level in microdissected adenomas, primary cancers, and metastases. A subsequent increase in expression of GROα at the protein level was confirmed by immunohistochemistry in tissue microarrays. Of particular significance, the immunohistochemistry results indicate that the GROα protein is secreted by the neoplastic colon epithelial cells, rather than by nonresident inflammatory cells. A role for GROα in the initiation of colon cancer is given further indirect support by the observation that GROα expression is elevated in the intestinal mucosa of individuals with ulcerative colitis, a disease that is a known risk factor for colon cancer (37, 38).
Forced expression of GROα in rodent fibroblasts was associated with transformation of the cells (focus formation, loss of contact inhibition, and tumor formation in a xenograft model). The resulting cells overexpressed GROα and produced fibrosarcoma when injected s.c. into athymic mice. Expression profiles of these transformed fibroblasts revealed that expression of fibulin-1C and fibulin-1D was markedly reduced compared with the parent cells, consistent with our findings in clinical colorectal samples.
Fibulin-1 is a member of a family of glycoproteins found in extracellular matrix and blood (39). There are six reported members of the fibulin family, which contain a series of epidermal growth factor–like modules coupled to a COOH-terminal fibulin module. Fibulin-1 has at least four splice variants (A-D), which vary in the COOH-terminal fibulin-type molecule. It can self-associate and bind to extracellular matrix proteins, including fibronectin, laminin, and to the coagulation protein fibrinogen (40), and is found in basement membrane and in loose connective tissues associated with matrix fibers (41, 42). Fibulins probably form intramolecular bridges that stabilize supramolecular extracellular matrix structures (39). In addition to its clear structural role, studies in a tumor model and in Caenorhabditis elegans also suggest that fibulin-1 has a role in controlling directed cell migration (43).
A role for down-regulation of fibulin in tumorigenesis was initially suggested because tumors formed by s.c. injection of carcinogen-transformed MSU 1.1 cells into athymic mice display marked down-regulation of fibulin-1D mRNA (44). Furthermore, 80% of malignant cell lines derived from human fibrosarcomas show attenuated expression of this mRNA. In contrast, studies conducted in breast and ovarian cancers have shown an increase in fibulin-1 protein expression when compared with control tissue (45, 46). These differences are likely to be due to tissue-specific factors and possibly hormonal regulation. Although it has been suggested that an increase in the ratio between fibulin-1C and fibulin-1D is important, where fibulin-1C acts as an oncogene and fibulin-1D acts as a tumor suppressor (47), we found a decrease in both fibulin-1C and fibulin-1D isoforms associated with GROα overexpression and in human clinical samples.
Forced expression of fibulin-1D resulted in extended latency in tumor formation in athymic mice and reduced the ability of these cells to form colonies in soft agar (25). These results are consistent with our xenograft model in which forced expression of GROα suppresses expression of fibulin-1C and fibulin-1D and promotes fibrosarcoma formation. Conversely, we report that restoring expression of fibulin-1D prevents tumor formation. Furthermore, ectopic expression of fibulin-1D also inhibits cellular transformation by the papillomavirus E6 gene (26).
Other evidence to support that fibulin-1 acts as a tumor suppressor (48) has come from studies showing that forced expression of fibulin-1D inhibits the motility of breast carcinoma cells on fibronectin (47). In preliminary studies, we have observed that expression of GROα impairs assembly of fibronectin matrix (data not shown).
Our work also extends some of these experimental results to clinical samples. Immunohistologic examination clearly shows that abnormal colon epithelial cells are the origin of the robust expression of GROα in neoplastic samples. Furthermore, fibulin-1 protein expression along the glandular basement membrane was clearly and significantly suppressed in colon adenocarcinoma. Abnormal expression of GROα and fibulin-1 seems to be a relatively early event in the development of colon cancer because GROα expression was significantly elevated and fibulin-1 expression significantly reduced even in the adenomas that we sampled. Indeed, in the GeneChip experiments, expression of fibulin-1C and fibulin-1D in the polyps was even lower than in the primary carcinoma samples (P < 0.001). However, this could simply represent different proportions of non-epithelial constituents in the adenoma and the primary cancers.
Single nucleotide polymorphism arrays conducted on primary colon cancers showed no marked change in copy number to explain the up-regulation of GROα. Likewise, there was no significant loss of gene regions encompassing fibulin-1. Furthermore, there was no significant difference in expression levels in samples that showed LOH compared with those that did not. These findings are consistent with the transcriptional regulation of GROα.
Regulation of GROα expression is complex and has not been completely elucidated. Exposure of HT-29 cells to tumor necrosis factor-α results in augmented expression of GROα (49, 50). Shiga toxins (Stxs) also induce GROα mRNA and protein in intestinal epithelial cells, perhaps by stabilizing GROα mRNA (21), and several other proinflammatory signals have been shown to increase GROα mRNA stability (19). On the other hand, transcriptional regulation is also important in GROα mRNA expression, involving cis-acting elements interacting with factors, such as nuclear factor κB, Sp1, and an immediate upstream element that interacts with the CCAAT displacement protein (23). Whether mRNA stabilization or transcriptional derepression is involved in the up-regulation of GROα in colon cancer is not known and is the subject of ongoing experiments. Although we did not investigate the role of the closely related gene GROγ in colon cancer, the observation that it is also significantly dysregulated implies either a common mechanism of regulation with GROα or a cancer-promoting effect (or both). By elucidating the mechanism by which GROα promotes colon cancer, therefore, it will be possible to gain a sharper understanding of the role of the other CXCR2 ligands in this process.
In summary, in both human samples and an experimental fibrosarcoma model, elevated expression of GROα is associated with down-regulation of fibulin-1C and fibulin-1D mRNA and protein expression. Histologic examination indicates that in adenocarcinoma samples, the surrounding stoma is depleted of fibulin-1. GROα is a secreted protein, and it seems likely that secreted GROα from malignant colonocytes diffuses to the surrounding stroma, inhibiting expression of fibulin-1. It remains to be clarified whether this effect is mediated by the CXCR2 receptor. Fibulin-1 is known to participate in the organization of the basement membrane and other ECM structures and also to suppress the motility of breast carcinoma and fibrosarcoma cells. We hypothesize, therefore, that abrogation of fibulin-1 expression facilitates the ability of neoplastic cells to permeate the basement membrane and to spread contiguously and through metastasis. Furthermore, we showed that inhibition of GROα expression induced apoptosis in HCT-15 cell lines (Fig. 3B), suggesting the hypothesis that an additional role of GROα in facilitating tumor formation is to inhibit apoptosis.
Increased GROα expression by neoplastic colon epithelial cells is a regular feature of adenocarcinoma of the colon and is associated with down-regulation of fibulin-1. The resulting inhibition of apoptosis and depletion of fibulin-1 could facilitate progression and spread of disease. Therefore, targeting these proteins may provide new opportunities for treatment of colon cancer.
Grant support: NIH/National Cancer Institute grant P01-CA65930.
Note: Supplementary data for this article are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/).
Affymetrix experiments were done by the Shared Gene Expression Resource of the Cancer Institute of New Jersey and supported by NIH/National Cancer Institute grant P30-CA072720.
We thank Dr. J. Bertino for his review of the article.