The 8p11-p12 genomic region is amplified in 15% of breast cancers and harbors several candidate oncogenes. However, functional evidence for a transforming role for these genes is lacking. We identified 21 genes from this region as potential oncogenes based on statistical association between copy number and expression. We further showed that three of these genes (LSM1, BAG4, and C8orf4) induce transformed phenotypes when overexpressed in MCF-10A cells, and overexpression of these genes in combination influences the growth factor independence phenotype and the ability of the cells to grow under anchorage-independent conditions. Thus, LSM1, BAG4, and C8orf4 are breast cancer oncogenes that can work in combination to influence the transformed phenotype in human mammary epithelial cells. (Cancer Res 2006; 66(24): 11632-42)
- breast cancer
One of the main molecular pathways in the natural history of human breast cancer development involves the focal amplification of distinct regions of the genome, resulting in copy number and expression changes of specific genes within these regions ( 1). Thus, gene amplification and overexpression is a major mechanism for oncogene activation in human solid cancers like breast cancers ( 1– 6). Although the genetic mechanisms that result in copy number increases occur over the entire genome, selection of cells with genomic alterations in specific regions occurs in a non-random fashion during the progression of breast cancer, supporting the hypothesis that these regions harbor dominantly acting oncogenes that play a causal role in cancer progression. The 17q21 genomic region is one of the most well studied regions of gene amplification in breast cancer because this amplicon harbors the ERBB2 oncogene, along with other important genes, such as GRB7 and TOPO2A ( 7– 11). Other important regions in breast cancer include the 8q24 amplicon, ( 12– 15), the 20q13 amplicon ( 16– 19), and the 11q12 amplicon ( 20– 25).
We have recently developed a novel panel of human breast cancer cell lines derived from several different molecular subclasses of human breast cancer ( 2, 26). Previous global genome analysis studies on these cell lines revealed that three cell lines in the panel (SUM-44, SUM-52, and SUM-225) harbor focal copy number increases in the 8p11-12 region ( 2, 3). Furthermore, past and recent studies on primary breast cancer specimens have shown that gene amplification occurs in this region in 10% to 15% of human breast cancers ( 14, 27– 31). Recent studies from our lab and others have suggested that the heretofore best candidate oncogene from this region (FGFR1) is not the only candidate oncogene in the region and, indeed, may be of marginal significance ( 32). Additionally, other candidate oncogenes have emerged based on statistical analysis of associated copy number changes and expression levels of genes in this region.
In the present studies, we sought to extend the correlative analysis of copy number and expression level of genes on the 8p11-p12 amplicon in human breast cancer specimens and cell lines by examining the amplicon in a panel of primary human breast cancers with copy number increases in this region. In addition, we used quantitative reverse transcription-PCR (RT-PCR) to determine the expression level of 53 genes from this region in breast cancer specimens compared with three different types of normal human mammary epithelial cells. To move beyond correlative approaches, we employed an alternative strategy aimed at identifying candidate oncogenes directly based on their ability to transform the immortalized human mammary epithelial cell line MCF-10A to specific growth factor independence. From these experiments, we have shown that LSM1, BAG4, and C8orf4(TC-1) are bona fide breast cancer oncogenes based on their copy number and expression status in human breast cancer and their ability to transform human mammary epithelial cells in vitro. In addition, these genes can cooperate to influence the expression of important altered growth phenotypes, which supports the hypothesis that common amplicons that occur in breast and other cancers harbor multiple oncogenes that can cooperate to influence the growth potential of cancer cells.
Materials and Methods
Cell lines and tumor specimens. The isolation and culture of the SUM series of HBC cell lines and MCF10A cells have been described in detail previously ( 2, 3). Human breast cancer specimens were obtained from Asterand, Inc. (Detroit, MI).
Quantitative genomic PCR. Quantitative genomic PCR experiments used the Applied Biosystems Prism 7900HT Sequence Detection System. The ZNF703(FLJ14299), LSM1, FGFR1, C8orf4, and GAPDH Taqman probe and primer mix as well as the Taqman Universal PCR master mix were purchased from Applied Biosystems (Foster City, CA). The data were analyzed using Sequence Detector System v2.1 (Applied Biosystems) and Microsoft Excel software.
Genomic array comparative genomic hybridization. Genomic array comparative genomic hybridization (CGH) experiments were done using the Agilent 44K human genome CGH microarray chip (Agilent Technologies, Palo Alto, CA). The protocol used for this experiment was The Oligonucleotide Array-Based CGH for Genomic DNA Analysis (Agilent Technologies). For each array, female DNA (Promega, Madison, WI) was used as a reference sample and labeled with Cy-3. The biological samples of interest were each labeled with Cy-5.
Agilent's CGH Analytics software was used to calculate various measurement variables, including log 2 ratio of total integrated Cy-5 and Cy-3 intensities for each probe. Chromosomal aberrations were classified as amplification when the log 2 ratio was >0.33 and as loss when the ratio was less than −0.33. This number was determined based on the array data and our previous fluorescence in situ hybridization and Southern data in three cell lines.
Microfluidics-based quantitative RT-PCR. Predesigned Taqman probe and primer sets for 8p11-12 target genes and housekeeping genes were chosen, factory-loaded into the 384-well format, and spotted on a microfluidic card by the manufacturer (four replicates per assay). RNA was isolated from breast cancer specimens and cell lines as well as three different types of normal human mammary epithelial cells (normal breast tissue, MCF-10A cells, and HME cells) and converted into cDNA. Quantitative real-time PCR was done in an ABI Prism 7000 Sequence Detection System (Applied Biosystems). Gene expression values were calculated based on the ΔΔCt method. The expression levels of the housekeeping gene GUSB from each sample were used for normalization because this gene showed relatively constant expression across most of the analyzed samples.
Production of lentivirus and cell infection. Lentiviral expression constructs were created for candidate oncogenes from the 8p11-p12 amplicon, including LSM1, EIF4EBP1, C8orf4, PPAPDC1B(HTPAP), ZNF703, RAB11FIP1, BAG4, and FGFR1, using ViraPower Lentiviral Expression System (Invitrogen, Carlsbad, CA). Each construct was sequenced to ensure that the sequences and orientation are correct. Lentivirus was produced by cotransfecting the 293FT cell line with the pLenti expression construct and the optimized packaging mix (Invitrogen). MCF-10A cells were transduced with lentivirus, and gene expression level was detected using RT-PCR. For combination infections, virus of the same titer from two, three, or all eight genes was combined equally and used to infect MCF-10A human mammary epithelial cells. Control infections with pLenti-LacZ virus were done in parallel with other infections. Selection began 48 hours after infection in growth medium with 10 μg/mL blasticidin and without either insulin or epidermal growth factor (EGF). Upon confluence, selected cells were passaged and serially cultured.
Growth in soft agar. Soft agar assays were done as previously described ( 33). Briefly, dishes were coated with a 1:1 mix of the appropriate 2× medium for the cell line being studied and 1% Bactoagar. Cells were plated at 1 × 105 per well, fed thrice per week for 3 to 4 weeks, stained with 500 μg/mL p-iodonitrotetrazolium violet (Sigma, St. Louis, MO) overnight, and counted.
Immunoprecipitation and Western blots. Cell lysis and protein quantification were done as previously described ( 34). For immunoprecipitations, whole-cell lysate (1 mg) was incubated with 2 μg/mg IRS-1 antibody for 2 hours at 4°C or 2 μg/mg EGF receptor (EGFR) antibody for 1 hour on ice followed by incubation with protein A/G agarose beads (Sigma) for 1 hour at 4°C. Proteins were resolved on 7.5% polyacrylamide gels, transferred to polyvinylidene difluoride membranes, and probed for 1 hour at room temperature with 2 μg/mL IRS-1 (Upstate, Lake Placid, NY), p-Tyr (PY20; BioMol International, Plymouth Meeting, PA), or EGFR (Zymed, South San Francisco, CA) antibodies. Membranes were incubated for 1 hour at room temperature in peroxidase-labeled secondary antibody and developed in enhanced chemiluminescence (Pierce, Rockford, IL).
Assay for monolayer growth and conditioned medium activity in insulin-independent and EGF-independent clones. For growth experiments, insulin-independent, EGF-independent, and control cells were seeded into six-well plates at 3.5 × 104 per well and grown in their appropriate medium. The EGF-independent clones were treated with 0.1 or 0.5 μmol/L of the EGFR inhibitor Iressa every 24 hours. Cell number was measured on day 9 using a Coulter counter.
Conditioned medium from subconfluent clones of MCF-10A cells infected with all combinations of two or three of LSM1, BAG4, and C8orf4 was collected. MCF-10A cells were seeded in six-well plates and grown in normal growth medium, medium without insulin (SFHE) or EGF (SFIH), or SFHE/SFIH medium supplemented 1:1 with conditioned medium from clones without insulin or EGF. Cell number was measured on day 9 using a Coulter counter.
Statistical methods. Kendall's tau was used to assess the statistical significance of the association between copy number and expression for each gene. Holm's stepdown procedure was used to adjust significance levels for the large number of estimates to reduce the likelihood of false positive results. We used P = 0.01 as a cutoff for a statistically significant association between copy number and expression.
Recently, we analyzed the 8p11-p12 genomic region for copy number and gene expression changes in three human breast cancer cell lines developed in our laboratory ( 32). From these experiments, we provided evidence that FGFR1 is not the only candidate breast cancer oncogene in this region. To provide further correlative and causal evidence for a transforming function of candidate genes from the region, we examined the 8p11-p12 region in 100 primary breast cancer specimens and cell lines at both the copy number and expression levels. We also analyzed the transforming potential of candidate oncogenes when overexpressed either individually or in specific combinations using a cell transformation assay based on the growth factor dependency of the human mammary epithelial cell line MCF-10A.
DNA copy number analysis in a panel of primary human breast cancers. One hundred unselected primary human breast cancers were screened for the presence of copy number increases in the 8p11-p12 region. As a preliminary screen for gene copy number increases in the 8p11-p12 region in these specimens, quantitative genomic PCR was done using primers for ZNF703, LSM1, FGFR1, and C8orf4, as these four genes are present within the amplified region as defined in the previous work. The results of this analysis showed that, of the 90 breast cancer specimens that yielded usable DNA, 24 tumors showed evidence of copy number increase (log 2 > 2) at one or more of the loci (see Supplementary Table S1). Twenty-two of these specimens yielded RNA of sufficient quality and quantity for expression analysis. In addition, we identified three more breast cancer specimens from our own bank and two additional breast cancer cell lines with copy number increases in the 8p11-p12 region and included these specimens in further studies.
To obtain a detailed map of the breast cancer specimens with putative copy number changes at 8p11-p12, the 22 breast cancer specimens identified in the preliminary screen were analyzed further by array CGH using the Agilent human genome CGH microarray chip. Representative array CGH profiles of three breast cancer specimens are shown in Fig. 1A , and copy number data for all 22 specimens analyzed are shown in Fig. 1B (see Supplementary Data for all array CGH profiles). The results of array CGH analysis confirmed and extended the results of previous experiments and showed the variations in amplicon structure within this region in different breast cancer specimens and cell lines. Interestingly, only 11 of the 22 specimens analyzed by array CGH were confirmed to have copy number increases in the 8p11-p12 region. Thus, in our panel of breast cancer specimens, the frequency of copy number increase in the 8p11-p12 region was ∼13% (11 of 90), which is consistent with estimates from other laboratories.
Figure 1B shows the copy number increases in the 8p11-p12 genomic region in all 22 breast cancer specimens that were analyzed by array CGH, as well as several breast cancer cell lines. Analysis of these results indicates the presence of a core region of gene amplification that spans from ZNF703 to the FGFR1 gene in all but one of the breast cancers. In addition, several breast cancers and cell lines contain this core region of amplification in addition to a more centromeric region of copy number increase that is often, but not always, contiguous with the core amplified domain. One breast cancer specimen in our panel (7391B1) showed evidence of copy number increase in the centromeric region that was independent of the core amplified domain. Thus, the results of these experiments indicate that the 8p11-p12 genomic region is frequently amplified in breast cancer and harbors a number of genes that are potentially important in breast cancer progression.
Expression of genes from the 8p11-p12 region in breast cancers and cell lines with gene copy number increases and controls. As shown previously, the region of copy number increase in the 8p11-p12 region spans ∼10 Mb and encompasses ∼53 known genes. To measure the expression level of these 53 genes in our panel of primary breast cancer specimens and breast cancer cell lines, both with and without gene amplification, we designed a microfluidics-based Taqman quantitative RT-PCR assay using primers specific for each of the 53 genes. RNA was isolated from 13 primary breast cancers and five breast cancer cell lines with an 8p11-p12 amplification, from 15 breast cancer specimens and two cell lines lacking the amplicon, and from three different types of normal human mammary epithelial cells. The results of this analysis are presented in Fig. 2A and show the expression level of each of the 53 genes in all 35 specimens relative to the expression levels obtained with normal breast tissue RNA. Examination of the data in Fig. 2A reveals a set of genes that are selectively overexpressed in the breast cancer specimens and cell lines with gene copy number increases in the 8p11-p12 region. Genes, such as LSM1, SPFH2, and BAG4, were both amplified and overexpressed in a significant fraction of the tumors analyzed, and expression levels for these genes were low when the genes were not amplified. Similar results were obtained for genes, such as IKBKB, POLB, and VDAC3, but the number of tumors with amplification and overexpression of these genes was smaller than those within the core amplified domain. In addition, genes, such as C8orf4, were overexpressed in breast cancer specimens independent of their copy number status.
To examine the relationship between copy number status and expression for all genes in the amplified region in all of the specimens in the panel, Kendall's tau, a measure of association, was used to assess the statistical significance of the association between copy number and expression for each gene. Holm's stepdown procedure was used to adjust significance levels for the large number of estimates, thereby reducing the likelihood of false-positive results. Figure 2B shows the expression level of a subset of genes from the region in breast cancer specimens both with and without copy number increases. Figure 2C shows the statistical relationship between copy number and expression for the same subset of genes for all of the breast cancer specimens and cell lines analyzed. The data show that the expression of some genes is tightly linked to their copy number status, whereas for some genes, there is no relationship between expression and copy number. Table 1 shows the statistical analysis of copy number and expression for all 53 genes in the region. Using P = 0.01 as a cutoff for a statistically significant association between copy number and expression, there are 21 genes in the 8p11-p12 region that can be considered as candidate breast cancer oncogenes. Included in this list are genes previously identified by us and others, such as LSM1, BAG4, and SPFH2. The list also includes genes such as POLB, VDAC3, LETM3, and EIF4EBP1 that have not been previously implicated in breast cancer development.
Transforming activity of candidate oncogenes and identification of oncogene interactions in the 8p11-p12 amplicon. Given the number of plausible candidate genes identified based on their expression status when amplified, it is important to develop experimental approaches that allow for direct analysis of the transforming function of candidate oncogenes and can detect potential gene interactions within an amplicon. Therefore, we designed a cell transformation strategy that makes use of the growth factor dependency of the MCF-10A human mammary epithelial cell line ( Fig. 3A ). In previous work, we showed the ability of bona fide breast cancer oncogenes to transform MCF-10A cells to growth factor independence, and we have used transformation to growth factor independence to design an expression cloning strategy that can detect the transforming function of oncogenes from a defined library of genes ( 33, 35– 38). To begin to analyze the transforming function of genes in the 8p11-p12 amplicon, we focused on the genes in the core region of gene amplification as defined in these and previous studies. We chose eight genes from this region based on their statistical association between copy number increase and expression level, or based on previous associations with transforming function when overexpressed, and prepared lentiviral expression constructs for each gene. Each lentiviral vector was packaged, titered, and combined before infecting midconfluent cultures of MCF-10A cells. Forty-eight hours after infection, MCF-10A cells were switched to media devoid of either insulin, EGF, or both insulin and EGF and cultured for 2 to 3 weeks. Colonies that emerged in growth factor–free media were propagated continuously in the same media, and those that exhibited continuous growth factor–independent proliferation were characterized further. To determine which genes from the library were expressed in growth factor–independent clones, RT-PCR analysis was done using primers specific for the library genes and for the vector.
Table 2 shows the results of experiments aimed at detecting the acquisition of insulin-independent growth capacity in library-infected cells. As shown in the table, three genes (C8orf4, BAG4, or LSM1) were consistently present in the insulin-independent clones isolated. Furthermore, one or more of these three genes were detected in every insulin-independent clone recovered. By contrast, the remaining five genes from the library (ZNF703, RAB11FIP1, FGFR1, PPAPDC1B, and EIF4EBP1) were only detected in 4 or 5 of the 20 clones examined, and each of those clones also expressed either C8orf4, BAG4, or LSM1 (see Supplementary Table S2 for insert data on each clone). These results suggested that C8orf4, BAG4, and/or LSM1 were primarily responsible for the acquisition of insulin-independent growth of MCF-10A cells. To confirm these results, MCF-10A cells were infected with C8orf4, BAG4, or LSM1 lentiviral vectors individually, and as shown in Table 2B, all three genes did, indeed, yield clones that could grow in the absence of insulin. Similarly, infection of MCF-10A cells with each of the other five constructs individually never gave rise to any insulin- or EGF-independent clones. Thus, three of the eight genes examined could induce insulin-independent growth, but none of the genes by themselves resulted in EGF-independent proliferation.
Next, MCF-10A cells were infected with the lentiviral library and selected for growth in EGF-free medium. In this experiment, many EGF-independent colonies emerged. Table 2 shows that the EGF-independent colonies that emerged contained predominantly the same three genes that were detected in the insulin-independent colonies. However, all EGF-independent colonies contained two or more of the genes previously shown to induce insulin-independent growth when expressed alone (see Supplementary Table S2 for inserts in individual clones). In addition, the remaining five genes in the library were rarely present in the EGF-independent clones and, when present, were always in clones that expressed at least two of the previously implicated genes. Taken with the previous results, which indicated that C8orf4, BAG4, or LSM1 can induce insulin-independent but not EGF-independent growth when expressed alone, these results strongly suggested that the same three genes can act in combination to transform cells to EGF independence. To confirm this observation, MCF-10A cells were infected with lentiviral vectors for LSM1, BAG4, and C8orf4 in all combinations of two or three and selected directly for growth in EGF-free media. The results of this experiment are summarized in Table 2B and confirmed the results of the library experiments. In this study, any combination of LSM1, BAG4, and C8orf4 resulted in cells that could grow in EGF-free media. Interestingly, however, none of the gene combinations resulted in cell growth in the absence of both insulin and EGF.
Characteristics of oncogene-transformed MCF-10A cells. To characterize further the transformed phenotypes exhibited by MCF-10A cells expressing LSM1, BAG4, and/or C8orf4, we examined soft agar growth and determined the requirement for growth factor receptor activation in factor-independent proliferation. Figure 3B and C shows that, whereas parental MCF-10A cells have no capacity for growth in soft agar, both insulin-independent and EGF-independent transformants formed numerous colonies in agar. In this assay, the highest level of agar colony-forming efficiency was observed in the EGF-independent cells expressing all three oncogenes. These results confirm that cells selected based on growth factor independence exhibit other transformed phenotypes in vitro, and that oncogene interaction plays a role in expression of these phenotypes.
Next, growth factor–independent cells were characterized for the activation of signaling receptors that could mediate the growth factor independence phenotype. Figure 4A shows that all of the EGF-independent clones isolated expressed constitutively tyrosine phosphorylated EGFR. By contrast, control MCF-10A cells expressed tyrosine phosphorylated EGFR only in the presence of exogenous EGF. In addition, exposure of these EGF-independent cells to the EGFR tyrosine kinase inhibitor Iressa (gefitinib) resulted in complete growth inhibition, indicating the necessity of constitutive EGFR activation for the growth of these transformed cells ( Fig. 4B, left). The presence of constitutively activated EGFR in the EGF-independent cells suggests the presence of an autocrine factor that is driving growth in the absence of exogenous EGF. To examine this question, conditioned medium was collected from the EGF-independent MCF-10A variants and tested for its ability to substitute for exogenous EGF in parental MCF-10A cells. The data in Fig. 4B (right) show that conditioned medium from EGF-independent cells has substantial EGF-like growth factor activity that is likely responsible for EGFR activation and growth in EGF-free media. Furthermore, the addition of Iressa to conditioned medium from EGF-independent cells prevented MCF-10A growth, suggesting that its EGF-like growth factor activity depends on EGFR activation ( Fig. 4B, right).
In contrast to the results obtained with EGF-independent cells, MCF-10A transformants growing in insulin-free media showed no signs of constitutive activation of IRS-1, the main signaling molecule activated by the IGF-I receptor ( Fig. 4C). However, low-level insulin-replacing activity was detected in conditioned media derived from these cells, suggesting the presence of a non-IGF family growth factor present in conditioned medium that may play a role in the insulin-independent growth of the MCF-10A transformants ( Fig. 4D).
Oncogene expression related to clinical variables of the breast cancer panel. The data (Supplementary Table S3) show the demographic and clinical variables of the 90 breast cancers in our original panel separated by the presence or absence of copy number increases at 8p11-p12. Consistent with results from other laboratories, there were no clear clinical features that were associated with gene amplification in this region. Because we do not have outcome data associated with our panel of breast cancer specimens, we were not able to analyze the role of amplification to variables of disease progression. However, in the recent work of Gelsi-Boyer et al. ( 31), there was a statistically significant association of decreased metastasis-free survival and the presence of the 8p11-p12 amplicon.
In summary, our results suggest that there are as many as 21 genes in the 53 gene region that are overexpressed in breast cancer when their copy numbers are increased. The large number of candidate oncogenes identified using statistical approaches illustrates the importance of using functional assays to identify the true transforming oncogenes. In that regard, our results provide the first evidence for cooperation among the LSM1, BAG4, and C8orf4 oncogenes that alters the transformed phenotypes of mammary epithelial cells.
The 8p11-p12 chromosomal region has been the subject of significant interest, particularly in human breast cancer where focal copy number increases occur in ∼15% of cases ( 27, 31, 39). Thus, it is highly likely that one or more breast cancer oncogenes reside in this genomic region. In our original experiments using three breast cancer cell lines developed in our laboratory, we provided evidence that the previously suspected candidate oncogene (FGFR1) was not the only oncogene candidate and, indeed, may be among the least interesting genes from this region, at least in breast cancer. In addition, we provided evidence for a role of LSM1, ZNF703, RAB11FIP1, PPAPDC1B, C8orf4, and TACC1 in breast cancer development ( 32).
Based on our previous findings, we set out to validate and extend these observations using a panel of 100 primary human breast cancer specimens. The results obtained in these experiments now point to several genes that are very likely to play important roles as bona fide breast cancer oncogenes. In addition, we now provide some of the first functional evidence for a transforming role for these genes, when overexpressed either alone, or in specific combinations.
Based on our statistical analysis of the correlation between copy number increase and expression levels of each of the 53 genes in the 8p11-p12 region compared with three different sources of human mammary cell RNA, we identified 21 genes that were overexpressed in association with copy number increase at the P < 0.01 level (see Table 1). Several of these genes had been implicated as oncogenes in our previous experiments, including LSM1, RAB11FIP1, PPAPDC1B, and EIF4EBP1. By contrast, other genes that we suspected could play a role as oncogenes in the cell lines did not make our list of candidate genes in the present studies, including TACC1 and C8orf4. Recent data from other laboratories also are not consistent with an oncogenic role for TACC1 in breast cancer and suggest it to be a candidate tumor suppressor gene from this genomic region ( 40). C8orf4, by contrast, seems to be overexpressed in a high proportion of breast cancers as well as thyroid and gastric cancer, regardless of copy number status ( 41, 42).
Recently, two groups reported results of copy number and gene expression analysis of the 8p11-p12 region in breast cancer. Gelsi-Boyer et al. ( 31) reported an extensive molecular cytogenetic analysis of 37 breast cancer cell lines and 62 primary breast cancers. They also analyzed the relationship between copy number and gene expression in 17 breast cancers. Garcia et al. ( 30) did a more focused analysis of 33 primary breast cancers and concentrated their studies on a 1-Mb core region of the amplicon common to most of the specimens in their panel. Thus, there are now three recent studies on the relationship between gene amplification and gene expression in over 50 primary breast cancers that have genomic alterations in this region. Several genes have been identified by all three groups as probable oncogenes based on statistical association between copy number and gene expression. These genes include SPFH2, BRF2, RAB11FIP1, LSM1, and PPAPDC1B. It is worth noting that Prentis et al. in their studies on the chromosomal rearrangements that occur near the NRG1 locus also found that SPHF2 was amplified and overexpressed in breast cancers with the 8p11-p12 amplicon ( 43). Several genes are common between our list of candidate genes and those identified by Gelsi-Boyer et al., including PROSC, DDHD2, WHSC1L1, FGFR1, TM2D2(BLP1), and AP3M2. Finally, two genes were commonly identified by us and by Garcia et al, including ASH2L and BAG4. In addition to the genes commonly identified by our group and others, eight genes reached statistical significance in our study that were not mentioned in the studies of Garcia et al. or Gelsi-Boyer et al., including FUT10, C8orf41(FLJ23263), EIF4EBP1, LETM2, AGPAT6, POLB, VDAC3, and HOOK3 (Supplementary Table S4).
The above discussion shows that statistical analysis of copy number and expression changes for individual genes within an amplicon can point to several genes that meet the criteria as candidate oncogenes. However, the number of genes identified and the variability in candidate oncogene lists obtained from different laboratories point to a clear need to validate candidate oncogenes based on transforming function. Accordingly, we tested the transforming activity of eight candidate oncogenes from the 8p11-p12 region using well-established methods in our laboratory.
The genes we chose to examine (LSM1, BAG4, RAB11FIP1, PPAPDC1B, ZNF703, EIF4EBP1, FGFR1, and C8orf4) were chosen based on several criteria. Most were chosen based on their statistical association between copy number and expression in our study, or on historical significance to the amplicon (e.g., FGFR1). C8orf4 was included because of prior evidence from our laboratory, which indicated that C8orf4 overexpression plays an important role in the transformed phenotype of the SUM-52 cell line. The results obtained from these functional studies confirmed and extended our results with C8orf4 and showed its transforming potential towards MCF-10A cells. In addition, we found that both LSM1 and BAG4 can act individually to induce an IGF-independent phenotype and anchorage-independent growth capacity. In addition, by using an expression cloning strategy in which MCF-10A cells were infected with the entire mini-library of expression vectors, we were able to identify combinations of genes that induced an altered growth phenotype that was not induced by any of the genes when overexpressed individually. In previous studies from our laboratory, we have shown that the acquisition of EGF-independent growth potential is an indicator of highly transformed cells. In addition, the breast cancer cell lines that harbor the 8p11-p12 amplicon were originally isolated based on the ability to proliferate in growth factor–deficient media. Thus, growth factor independence is a hallmark of these breast cancer cell lines. Consistent with those previous findings, we showed that the three genes that were found to induce an insulin-independent growth phenotype when overexpressed individually could combine to render cells EGF independent. Thus, our first functional studies of candidate breast cancer oncogenes from the 8p11-p12 region provide strong evidence that LSM1, BAG4, and C8orf4 are breast cancer oncogenes that have transforming function when overexpressed in human mammary epithelial cells.
LSM1 has been previously implicated as a transforming oncogene in pancreas cancer ( 44), and more recent work has suggested that alterations in mRNA stability that occur when LSM1 is overexpressed play a mechanistic role in its transforming function. BAG4 has not been previously implicated as a transforming oncogene; however, this protein has been implicated in the radiation resistance of certain cancer cell lines, and its overexpression can prevent anoikis induced by blocking integrin signaling in normal epithelial cells ( 45– 48). C8orf4 is an interesting oncogene shown to be overexpressed in >90% of thyroid cancers ( 41, 42) and suggested to interact with Chibby, a negative regulator of WNT/β-catenin signaling ( 49, 50).
Perhaps as interesting as the genes that exhibited transforming function when overexpressed in vitro were some of the genes that were negative in this assay. First, FGFR1 overexpression did not result in expression of altered growth phenotypes, consistent with the results reported in our previous study. In contrast, previous work from other laboratories has shown a role for FGFR1 in prostate cancer progression ( 51, 52); in the regulation of transformation, proliferation, and migration of mouse mammary epithelial cells ( 53, 54); and in myeloproliferative disorders caused by FGFR1 gene translocations ( 55). Although our results are not consistent with an oncogenic role for FGFR1 in breast cancers with an amplified 8p11-p12 region or in our model of growth factor independence, it is possible that FGFR1 may affect cancer progression in a way not specifically illustrated by our experiments and therefore merits continued investigation. In addition, RAB11FIP1 and PPAPDC1B, which have been consistently implicated as candidate oncogenes based on their statistical association between amplification and expression and their biological function, had no transforming activity in our biological assays. Interestingly, PPAPDC1B has been recently implicated as a possible metastasis suppressor in hepatocellular carcinoma by showing that PPAPDC1B overexpression decreased invasion and metastasis with no effect on growth of the primary tumor ( 56).
In summary, the results reported here extend the previous studies by our laboratory and others, which implicate a number of key genes as transforming breast cancer oncogenes from the 8p11-p12 region. Our data not only implicate LSM1, BAG4, and C8orf4 as bona fide oncogenes but have shown the potential for oncogene interactions within an amplicon in human breast cancer cells. Future studies will examine all 21 candidate oncogenes for transforming function, both singly and in combination, to examine how these genes interact with overexpressed genes from other amplicons within the same tumor specimen. This approach will ultimately result in the development of oncogene signatures that are likely to have important predictive power both for the natural history of disease progression and for predicting the best targeted therapeutic strategies.
Grant support: NIH grants RO1 CA100724 and RO1 CA70354 (S.P. Ethier) and Department of Defense Breast Cancer Program grant DAMD17-03-1-0459 (Z.Q. Yang).
The array comparative genomic hybridization work was facilitated by the Microarray and Bioinformatics Core Facility of the Wayne State University Environmental Health Sciences Center (National Institute of Environmental Health Sciences grant P30 ES06639).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Sonja Markwart, Joe Washburn, and James MacDonald for technical assistance on the quantitative genomic PCR and microfluidic card quantitative RT-PCR analysis.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
Z.Q. Yang and K.L. Streicher contributed equally to this work.
- Received August 9, 2006.
- Revision received October 6, 2006.
- Accepted October 16, 2006.
- ©2006 American Association for Cancer Research.