In the LacIv-fos cell system, the control of FBJ/R v-fos expression is dependent on the presence of isopropyl-b-D-thiogalactopryanoside (IPTG) in the cell culture medium 11. In the presence of 5 mM IPTG, LacIv-fos cells did not express v-Fos protein detectable by Western blot analysis (Figure 1a). When IPTG was washed away, LacIv-fos cells expressed v-Fos protein with peak expression detected at 72 hours following removal of IPTG and progressive loss of v-Fos protein levels upon re-addition of IPTG. In addition, cells were morphologically transformed within a 72-hour period. Transformation was indicated by an overall change in cell shape that led to a more rounded and light refractory morphology as well as dramatic cytoskeletal alterations (Figure 1b). When IPTG was added back to these transformed cells, v-Fos expression was again repressed and the cells returned to their original morphology within a 72-hour reversion period.

The ability to control both v-Fos expression and morphological transformation in vitro provides an opportunity to investigate global gene expression changes relative to the temporal v-Fos cellular transformation and reversion process. RNA was extracted from LacIv-fos cells treated with 5 mM IPTG at time zero (t = 0); from cells at 24, 48 and 72 hours following the removal of IPTG (transforming); and from cells at 24, 48 and 72 hours after the re-addition of 5 mM IPTG (reverting). Previous studies have demonstrated that the addition and removal of IPTG does not itself induce persistent changes in gene expression in the parental 208F rat fibroblast cell line 911. RNA samples were processed and analyzed by hybridization to Affymetrix rat U34A GeneChip microarrays.

Self-organizing map (SOM) analysis is a powerful tool that can be used to categorize gene expression data into groups that share common temporal expression profiles 151617. To identify patterns of gene expression, we created an SOM by using the LacIv-fos microarray data obtained at the seven time points throughout v-Fos transformation and reversion (Figure 2a). The data were mapped into 18 groups that provided the greatest pattern distinction among groups as well as the best pattern consistency within groups. Several assumptions were made on the basis of simple visualization of the map. For example, the expression of genes grouped into pattern 1 peaked 24 hours following removal of IPTG, as well as at 24 hours following re-addition of IPTG. This group likely represents genes whose expression was induced within 24 hours following culture media change rather than genes whose expression coincided with v-Fos-mediated transformation. Indeed, genes represented by pattern 1 included serum-responsive genes such as c-Jun and cyclin D1. However, other patterns included genes that were regulated in a manner more consistent with conditional v-Fos transformation and reversion. For example, pattern 8 included genes that were dramatically upregulated during the 3-day transformation process, yet return to baseline levels during the three-day reversion process (Figure 2b). This group included two probe sets specific for FBR v-fos included on the Affymetrix U34A GeneChip. These probe sets give a reliable measurement of FBJ/R v-fos expression and they did not cross-hybridize with endogenous c-fos transcripts 11. Conversely, group 11 represents genes that were conditionally downregulated specifically during conditional v-Fos transformation (Figure 2c). Plots of the normalized signal values for each probe set included in these groups further confirmed the reliability of the classifications of gene expression profiles within the SOM (Figure 2d-e).

Patterns within the SOM that are consistent with v-Fos transformation and reversion were selected on the basis of the representative expression pattern of each SOM group. For candidate genes whose expression was upregulated, the standard deviation of the mean signal values at a minimum of two of the three time points in the absence of IPTG (transforming) had to be greater than that of the mean signal values at time zero (i.e., the time prior to removal of IPTG from the culture medium). Also, the standard deviation of the mean signal values at a minimum of two of the three final time points in the presence of IPTG (reverting) had to be below that of the mean signal value of the final time point during transformation (i.e., 72 hours after removal of IPTG from the culture medium). For candidate genes whose expression was downregulated, the standard deviation of the mean signal values at a minimum of two of the three transforming time points had to be lower than that of the mean signal values at time zero, and the standard deviation of the mean signal values at a minimum of two of the three time points during reversion had to be higher than that of the final time point during transformation. On the basis of these criteria, patterns 7, 8, 14 and 15 (transforming) and patterns 5, 11, 12, 17 and 18 (reverting) were selected for further analysis. Each of these groups consisted of a large number of genes (Figure 2a). These groups likely included not only genes functionally involved in cellular transformation, but also genes that share similar expression profiles during conditional transformation and reversion yet have no role in the process or maintenance of cellular transformation itself. Interestingly, both patterns 5 and 7 included genes with higher signal-ratio values at the end of morphological reversion (t = 72 hours, reverting) than prior to v-fos induction (t = 0). This difference in transcript levels at the beginning of the time course relative to the final time point may be related to a requirement for a higher level of expression of particular genes during the reversion of morphological transformation than during maintenance of the non-transformed state. Alternatively, this may reflect a type of over-compensation in that expression levels within particular expression pattern groups and transcript levels may not completely return to steady-state within the three-day reversion period.

To identify only the most promising candidate genes, we compared these datasets with gene expression profiles of the parental 208F cells and 208F cells transformed by stable expression of either c-Fos (CMVc-fos cells) or v-Fos (FBJ/R cells). Cells transformed by stable expression of c-Fos or v-Fos exhibit a large number of differentially expressed genes. For example, when differential expression was defined as a 2-fold increase or decrease based on a comparison with expression in 208F cells and when probe sets for expressed sequence tags (ESTs) were excluded, 70 probe sets were scored as upregulated and 104 were scored as downregulated in both CMVc-fos and FBJ/R cells 11. However, eliminating genes whose expression changes are consistent in both stably and conditionally transformed cell model systems can reduce the level of nonspecific transcriptional variation. Of the 174 differentially expressed probe sets identified in comparisons of both CMVc-fos and FBJ/R cells to 208F cells, 63 (36%) were also categorized into LacIv-fos SOM patterns consistent with increased or decreased expression during conditional v-Fos transformation. Therefore, the majority of gene expression changes associated with stably Fos-transformed cell lines were not recapitulated during conditional v-Fos transformation and reversion. These findings emphasize the value of combining analyses of stable and conditional cellular transformation systems.

To address the reproducibility of gene expression changes during conditional v-Fos transformation, we performed microarray analysis using RNA samples isolated from LacIv-fos cells at time points during a second LacIv-fos transformation and reversion time course experiment. To maintain stringent biological replication of the experiment, these seven time points were treated as an independent experiment rather than as expression values that could be averaged between experiments. Expression patterns of genes differentially expressed in both CMVc-fos and FBJ/R cells relative to 208F cells, as well as conditionally regulated within the LacIv-fos SOM obtained from the first experiment, were compared to expression patterns obtained from the independent LacIv-fos transformation and reversion time course experiment. Of the 63 probe sets representing differentially expression in stably Fos-transformed cells and conditional regulation in the LacIv-fos SOM, 40 (64%) demonstrated a reproducible temporal profile of gene expression in the independent LacIv-fos transformation and reversion time course experiment (Additional file 1 and Figure 3). In addition, a previous study indicated by Northern blot analysis that representative genes (i.e. TGF-β 3 and CAIII) had nearly identical patterns of gene expression when compared to microarray analysis 11. Therefore, the combination of gene expression analysis of stably transformed cells with the analysis of gene expression in an experimentally controlled conditional transformation system is a useful approach to reduce complicating transcriptional variation inherent in any single cell model system.

Unlike c-fos, v-fos contains several deletions and point mutations that affect its oncogenic potential 18. Although sustained expression of c-fos is capable of inducing cellular transformation and tumorigenesis, v-fos is a much more potent oncogene 7. The increased tumorigenic potential of v-Fos suggests that it directly or indirectly influences the expression of genes not affected by c-Fos. Consistent with this hypothesis, gene expression profile analyses to identify genes differentially expressed by a 2-fold margin in FBJ/R cells, but not in CMVc-fos cells, revealed a dramatic increase in the number of transcriptionally upregulated genes in FBJ/R relative to CMVc-fos cells: 337 probe sets were scored as specifically upregulated in FBJ/R cells, whereas only 70 probe sets common to CMVc-fos and FBJ/R cells were considered as increased. In contrast, 61 probe sets were scored as downregulated specifically in FBJ/R cells but 104 probe sets in both CMVc-fos and FBJ/R cells were considered as downregulated 11. These FBJ/R cell-specific changes in gene expression were compared with the LacIv-fos SOM and the results of the independent LacIv-fos transformation and reversion time course as described previously. These comparisons revealed an additional 38 upregulated and 29 downregulated probe sets that are differentially regulated in v-fos-transformed cells, but not in c-fos transformed cells (Tables 1 and 2, Figure 4).

The identification of conditionally regulated genes within this system provides potentially unique opportunities to investigate precise transcriptional regulatory mechanisms of endogenous genes within an experimentally amenable cellular context. Therefore, we assembled these datasets into a web-accessible database equipped with various query functions so that users can mine the data to address questions relevant to their own specific interests (see methods). The initial interface of the database is the LacIv-fos SOM (Figure 2a). Users can select SOM patterns of interest to obtain a list of all probe sets mapped within that common pattern of gene expression. Each probe set is linked to the raw Affymetrix data for both independent LacIv-fos time course experiments. This arrangement provides an opportunity for convenient confirmation of consistent regulation between the independent time course experiments. In addition, users can query the database for information about specific genes by using Genbank accession numbers or by using genes name as keywords. Keyword queries may include the qualifier functions "and" or "or". These queries will provide a list of all probe sets matching the supplied criteria, and each entry is again linked to the raw Affymetrix data obtained from both independent LacIv-fos time course experiments.

All cells were maintained in DMEM supplemented with 10% fetal calf serum, L-glutamine and penicillin/streptomycin at 37°C in the presence of 5% CO₂. Establishment of the stably transformed cell lines, FBJ/R and CMVc-fos, was described previously 9. Establishment of the LacIv-fos conditional cell line and the conditions required for the transformation/reversion time course were described previously 11. Briefly, for the v-fos transformation and reversion time course experiment, total RNA was extracted from LacIv-fos cells cultured in the presence of 5 mM IPTG at time zero; 24, 48 and 72 hours after removal of IPTG (transforming); and 24, 48 and 72 hours after the re-addition of 5 mM IPTG (reverting). Total RNA was extracted from cells by using Tripure reagent according to the manufacturer's instructions (Roche). The integrity of all RNA samples was verified by using an Agilent 2100 Bioanalyzer.

Expression of Fos protein during transformation and reversion was evaluated by Western blotting analysis using an antibody directed against Fos 7. Whole-cell lysates were prepared in lysis buffer (20 mM Tris pH 7.5, 100 mM NaCl, 0.5% NP-40, protease inhibitor cocktail [Roche]) and separated on a 12% SDS-polyacrylamide gel. For immunofluorescence microscopy, cells were fixed in 4% (w/v) formaldehyde. Fixed cells were incubated with primary mouse monoclonal anti-vinculin antibody (Sigma; dilution of 1:200) and phalloidin-congjugated to Alexa-568 (Molecular Probes). An anti-mouse secondary antibody (dilution of 1:1000) conjugated to Alexa-488 (Molecular Probes) was used to detect vinculin.

RNA samples were processed for hybridization without amplification and hybridized to Affymetrix Rat Genome U34A GeneChips that include probe sets representing approximately 7,000 annotated genes and 1,000 EST clones 34. Labelling and hybridization were preformed as described 35. GeneChips were scanned using a laser confocal scanner (Agilent Technologies) and images were analyzed using Affymetrix Microarray Suite v.5.0. Datasets were standardized by global scaling of the average fluorescent intensities of all probe sets to a constant target value of 500 for all arrays. Quality control parameters for each hybridization were within MIAME compliant specifications 36. A variance filter was applied to the dataset to remove data from probe sets representing genes that were not expressed throughout the time course and to standardize expression values for genes whose expression was scored absent at particular time points 37. Data that were derived from probe sets that reported absent change calls at all seven time points were removed from the analysis. To standardize expression values of genes whose expression was scored as absent at fewer than seven time points, we converted the signal values that corresponded to absent change calls to a value of 1. Genes scored as differentially expressed in CMVc-fos cells and FBJ/R cells relative to 208F cells had a signal log ratio ≥ 1 (increased) or ≤ -1 (decreased), and a change p-value < 0.001 (increased) or > 0.99 (decreased). For visualization of specific profiles of gene expression during the LacIv-fos conditional transformation and reversion time course, signal values obtained at each LacIv-fos time point were plotted by using Microsoft Excel.

Microarray data that were obtained from the LacIv-fos time course experiments and passed the variance filter were grouped into relative expression pattern bins by the self-organizing map (SOM) program in the GeneMaths software package (version 1.5; Applied Maths, Austin, TX). The SOM analysis was performed by using a 6 × 3 node format to allow optimal representation of gene expression patterns in a reasonably small number of independent bins.

The SOM searchable database has been implemented on an Open Source MySQL 4.0.14 relational database management system. The database has a web interface at http://www.stjuderesearch.org/v-fosSOM/. The web application to query and manage the database is driven with a server-side scripting technology JSP. The entries of this database are generated from the tab-delimited output files of the SOM analysis. Data obtained from the SOM analysis can be queried by three approaches. Users can click on a graphical pattern key to view a list of all genes whose expression profiles match the particular self-organization pattern. Each entry in the list contains two parameters associated with the respective gene probe, i.e., the Affymetrix probe set ID number and a description of the gene. The detailed Affymetrix analysis data for a probe set can be obtained by clicking the link of the respective probe set ID number. The corresponding Affymetrix analysis page provides information about the number of oligonucleotide pairs, the signal magnitude, the detection status, and the detection p value at seven different time points derived from 2 independent experiments. Another search method is to query via the GenBank accession number. The query by accession number is based on the fact that each Affymetrix probe set ID number in our current experiments corresponds to a gene accession number. If an accession number in a query matches with any part within a probe set ID number, the entry containing the probe set ID number, its SOM pattern, and gene description is displayed. Hyperlinks are provided for both probe set ID numbers and SOM pattern. By following the link for a probe set ID number, the user can view the detailed Affymetrix analysis data for a selected gene probe; by following the link for an SOM pattern, the user can view the detailed Affymetrix analysis data for all the gene probes within the selected pattern. Also, a keyword-based text search is included in the database. A search for keywords in the paragraph of the gene description field is conducted in a case-insensitive manner. The search-by-keyword method is facilitated by the use of Boolean operations. The search can narrow or broaden quickly the results of the search by combining two or more keyword(s) with only one type of Boolean (and / or) operation.