TDAG51 is an ERK signaling target that opposes ERK-mediated HME16C mammary epithelial cell transformation

Introduction Signaling downstream of Ras is mediated by three major pathways, Raf/ERK, phosphatidylinositol 3 kinase (PI3K), and Ral guanine nucleotide exchange factor (RalGEF). Ras signal transduction pathways play an important role in breast cancer progression, as evidenced by the frequent over-expression of the Ras-activating epidermal growth factor receptors EGFR and ErbB2. Here we investigated which signal transduction pathways downstream of Ras contribute to EGFR-dependent transformation of telomerase-immortalized mammary epithelial cells HME16C. Furthermore, we examined whether a highly transcriptionally regulated ERK pathway target, PHLDA1 (TDAG51), suggested to be a tumor suppressor in breast cancer and melanoma, might modulate the transformation process. Methods Cellular transformation of human mammary epithelial cells by downstream Ras signal transduction pathways was examined using anchorage-independent growth assays in the presence and absence of EGFR inhibition. TDAG51 protein expression was down-regulated by interfering small hairpin RNA (shRNA), and the effects on cell proliferation and death were examined in Ras pathway-transformed breast epithelial cells. Results Activation of both the ERK and PI3K signaling pathways was sufficient to induce cellular transformation, which was accompanied by up-regulation of EGFR ligands, suggesting autocrine EGFR stimulation during the transformation process. Only activation of the ERK pathway was sufficient to transform cells in the presence of EGFR inhibition and was sufficient for tumorigenesis in xenografts. Up-regulation of the PHLDA1 gene product, TDAG51, was found to correlate with persistent ERK activation and anchorage-independent growth in the absence or presence of EGFR inhibition. Knockdown of this putative breast cancer tumor-suppressor gene resulted in increased ERK pathway activation and enhanced matrix-detached cellular proliferation of Ras/Raf transformed cells. Conclusion Our results suggest that multiple Ras signal transduction pathways contribute to mammary epithelial cell transformation, but that the ERK signaling pathway may be a crucial component downstream of EGFR activation during tumorigenesis. Furthermore, persistent activation of ERK signaling up-regulates TDAG51. This event serves as a negative regulator of both Erk activation as well as matrix-detached cellular proliferation and suggests that TDAG51 opposes ERK-mediated transformation in breast epithelial cells.


Background
The development and progression of breast cancer is the result of multiple genetic changes, which lead to complex alterations in signal transduction networks in breast cancer cells relative to their normal epithelial counterparts. Signaling differences between tumor and normal cells are reflected in altered gene expression patterns, a finding that has been widely investigated using various molecular techniques. Patterns of differential gene expression have been used for classification and prognostication of certain cancers and may be valuable for prospectively predicting responsiveness to therapeutic treatments [1]. In addition to serving as biomarkers, some differentially expressed gene products have been informative in defining the physiological differences between normal and cancer cells. In order to understand the impact of dysregulated signaling pathways upon gene expression and function, it is necessary to connect differential gene expression to upstream signaling pathways.
Ras activation is a common intermediary in signaling pathways initiated by a variety of cell surface receptors, and signaling pathways downstream of Ras have been implicated repeatedly in oncogenesis. Ras proteins are frequently mutated to an activated form in human cancers, particularly in tumors of the pancreas, colon, thyroid, and lung [2]. Ras mutations are rare in breast cancer, where the mutation rate is less than 5% [3]. However, Ras signaling is hyperactive in many primary breast tumors [4]. This is most likely due to the activation of growth factor receptors that activate Ras exchange factors, such as the epidermal growth factor receptors EGFR (ErbB1) and ErbB2 (HER2/ neu) and the colony stimulating factor-1 (CSF-1) receptor c-fms [5,6]. EGFR and ErbB2 receptor over-expression occurs in one quarter to one-half of all breast tumors, and this correlates with a significantly decreased disease-free and overall survival rates [5,7]. In addition, ligands for the EGFR such as transforming growth factor-a, amphiregulin, epiregulin, betacellulin, and heparin-binding EGF stimulate the receptor in primary breast tumors to enhance local growth and progression of the disease [8]. Finally, activating mutations in the PI3K pathway, an immediate downstream effector of Ras, occur in a subset of breast tumors [9,10]. Taken together, these findings suggest that breast cancer is a relevant model in which to study the biology of downstream Ras signal transduction.
The Ras oncogene binds numerous effectors that in turn activate a variety of signaling pathways. The most highly characterized of these are the Raf proteins, PI3K, and the Ral guanidine nucleotide exchange factors (RalGEFs), although Ras does recruit and activate other potentially transforming effectors [11]. The dissection of individual Ras signaling pathways is possible using effector domain mutants (EDMs) of Ras, for which single amino acid mutations in the effector binding domain allow the binding and activation of specific effectors, but not others [12]. These Ras EDMs, together with activated or dominantnegative versions of downstream effectors, have been used to characterize the pathways that contribute to the transformation of immortalized cells.
In mammary epithelial cells, both species-as well as cell line-dependent differences in Ras downstream signaling pathways have been found to induce tumorigenesis and/ or anchorage-independent growth, a measure of in vitro transformation. Studies using immortalized mouse EpH4 mammary epithelial cells have implicated Raf as well as PI3K pathways in supporting transformation and tumorigenesis [13]. For human immortalized mammary epithelial cells, Raf and PI3K clearly contribute to transformation, although each is usually not sufficient for tumor formation in animal models [14,15]. In fact, the immortalized human breast epithelial cell line HMLE required simultaneous activation of Raf, PI3K, and the RalGEF pathways for maximal anchorage-independent growth and tumorigenic transformation [14].
Dissecting the physiological consequences of individual Ras-mediated signaling pathways with respect to mammary epithelial transformation is of obvious interest. The ability of activated Ras and Raf to induce autocrine expression of epidermal-like growth factors has been implicated in the protection of MCF10A mammary epithelial cells from anoikis [16]. Using HMEC16C cells, a telomeraseimmortalized human mammary epithelial cell line, we have investigated the contribution of EGFR signaling to anchorage-independent growth initiated by Raf and additional signaling pathways downstream of Ras. We determined that ERK but not PI3K or RalGEF activation of HMEC16C cells supports anchorage independent proliferation independent of EGFR activation.
We performed a functional analysis of one gene in particular, TDAG51, whose expression is regulated by ERK through EGFR dependent and independent mechanisms. The loss of TDAG51 mRNA and protein has been correlated with breast adenocarcinoma and melanoma progression in clinical samples [17,18]. The importance of TDAG51 regulation on the transformed phenotype of Rasinfected cells was addressed using TDAG51-specific interfering small hairpin RNA (shRNA) to reduce TDAG51 levels. Consistent with a tumor suppressor role, loss of TDAG51 increased ERK-mediated cellular proliferation.

Culture of human epithelial cell lines
HME16C human mammary cells [19] were cultured in Clonetics Mammary Epithelial Basal Media (MEBM) with MEGM SingleQuot supplements (Cambrex), and HEK-HT human embryonic kidney epithelial cells [20] in DMEM (Invitrogen) plus 10% fetal bovine serum (Invitrogen). All cells were maintained at 37°C and 5% CO 2 . For induction of proteins from the tetracycline-inducible retroviral expression vector pLRT, 250 ng/mL of doxycycline (Sigma-Aldrich) was added to culture media.

Soft agar assays
Cells were trypsinized, neutralized, and 1.2 × 10 4 cells were plated per well in 0.36% bacto-agar (BD Biosciences), w/v in growth media, on a 0.6% bacto-agar support in 6-well culture plates. Prior to cell plating, 250 ng/ mL doxycycline, DMSO vehicle, or the anti-EGFR compound PD153035 (Calbiochem), was added to the agar, as appropriate. Appropriately-supplememented fresh media was added to wells every 4 days. Colonies greater than 100 μm in diameter were counted after 18 days using a 2× lens equipped with a graded grid to determine colony size.

Alamar Blue growth assays in ultra-low attachment tissue culture plates
Cells were trypsinized and plated in 500 μL appropriate growth media into 24-well ultra-low attachment plates (Corning) at 25,000 cells per well. At the times indicated, 50 μL of alamar blue (BioSource International) was added to each well and incubated for 6 hours under standard culture conditions. The amount of alamar blue reduction was quantified by measuring the fluorescence of each sample at 530 nm excitation/590 nm emission, corrected for the background fluorescence in wells containing media but no cells. The number of cells for initial plating and the alamar blue incubation times were determined by generating a standard curve of fluorescence versus cell number incubated with alamar blue for various times. Plating cells initially at 25,000/well insures that fluorescence measurements will occur on a linear part of the curve with a 6hour incubation time over the course of an experiment. It was determined that TDAG51-specific shRNA did not affect the redox potential of HME16C at various cell numbers. Statistical analysis was performed using 2-way analysis of variance (ANOVA) with Bonferroni post-tests using the GraphPad Prism software program (GraphPad Software, Inc.).

Tumorigenicity assays
Tumorigenicity assays were done essentially as described [23]. To induce gene expression, cells were treated with 250 ng/mL doxycycline three days prior to injection, and mice were fed with doxycycline-containing diet (200 mg/ kg, BioServ, Inc.) beginning four days prior to cell inoculation and continuing for the duration of the experiment.

Microarray analysis
Total RNA was extracted using Trizol reagent (Invitrogen) after treatment of cells with 250 ng/mL doxycycline for 72 hours to induce gene expression, or with doxycycline and 0.25 μM EGFR receptor kinase inhibitor PD153035 as indicated. Twenty micrograms of RNA were used for cDNA generation, and cDNA labeled with Cy3 or Cy5 monofunctional reactive dye to amino allyl-modified dUTP incorporated into cDNA using the FairPlay Microarray labeling kit (Stratagene). Labeled cDNA was hybridized to long-oligo (70-mer) cDNA microarrays from the NCI/CCR Microarray Center, NCI, Frederick, MD, according to standard protocols [25]. Hybridized arrays were analyzed using a GenePix 4000B array scanner and Gene-Pix Pro 4.0 software (Axon Instruments, Molecular Devices Corporation). Data from GenePix Pro 4.0 was uploaded to the microarray database at the NCI/CCR Microarray Center website [25] for further analysis. Signal intensities of microarray features were calculated by sub-tracting the median local background from the median signal intensity. Features were considered for analysis if the signal intensity was greater than one standard deviation above background with at least a 2:1 signal-to-background ratio. Signal intensities for an entire microarray were normalized to the 50% percentile median value. After filtering and normalization, the Cy3 and Cy5 values were expressed as a ratio to indicate the fold up-or downregulation. Two independent experiments for each comparison were performed, with a dye switch for each experiment, therefore yielding four separate data sets. For determining gene expression changes greater than or less than 2-fold, data sets were filtered for genes containing at least two significant values out of four array sets. Prior to filtering, all data points were analyzed using statistical analysis of microarray (SAM) data and a resultant gene set was chosen at a delta value of 0.4 that limited the false discovery rate (number of predicted false positive values/ number of significant genes) for each analysis to less than 1%. Minimal information about a microarray experiment (MIAME)-compliant microarray data has been deposited with the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus, accession number GSE8916, available at [26].

Real-time RT-PCR analysis
cDNA was synthesized from RNA obtained for microarray analysis using the SuperScript™ III First-Strand Synthesis System for RT-PCR (Invitrogen). Quantification of relative cDNA levels for each gene was accomplished using the Platinum ® SYBR ® Green qPCR Supermix UDG real-time RT-PCR kit (Invitrogen) and a Rotor-Gene3000™ thermocycler with Rotor-Gene 5.0.37 software (Corbett Research) that calculates relative PCR synthesis rates by comparative quantification. The specificity of product synthesis was verified by melting curve analysis by the Rotor-Gene 5.0.37 software, and by running of real-time PCR products on 2% agarose gels to verify product size and rule out primer-dimer contribution to calculated values. The sequence-specific primers used in real-time RT-PCR can be found in Additional file 2.

Cell cycle analysis
Cells grown under anchorage-independent conditions were spun down, washed once in sterile PBS, and suspended in a 50% mixture of PBS and ACCUMAX cell detachment solution (Chemicon International) for 10 minutes at 25°C to dissociate cell clumps. After dissociation, cells were washed once in cold PBS and then fixed with 70% ethanol. Fixed cells were treated with RNAse A for 20 minutes at 37°C, and nuclear DNA was stained with 50 μg/mL propidium iodide at 4°C. Cell cycle profiles were generated using a FACSCalibur flow cytometer (BD Biosciences) and modeling cell cycle phases with the cell cycle option of FlowJo flow cytometry software (Tree Star, Inc.) after gating for viable, single cells. Statistical analysis was performed using 2-way ANOVA with Bonferroni post tests using the GraphPad Prism software program (GraphPad Software, Inc.).

EdU cell proliferation assay
Measurement of cell proliferation by 5-ethynyl-2'-deoxyuridine (EdU) incorporation was measured using the Click-iT EdU cell proliferation Assay Kit for Flow Cytometry (Invitrogen) according to the manufacturer's protocol. Briefly, cells were plated at 1 × 10 6 cells per well in 6-well ultra-low attachment plates or 5 × 10 5 cells per dish in 60 mm tissue culture dishes for attached control cells. Attached cells were treated with either DMSO or 10 μM EdU for 4 hours, and cells grown in low attachment plates were treated with DMSO or 10 μM EdU for 24 hours. Using FlowJo flow cytometry software (Tree Star, Inc.), DMSO-treated control cells were used to determine the threshold above which cells could be considered EdU positive, and the fraction of cells incorporating EdU was then determined for experimental samples.

Cytotoxicity assay
The release of lactose dehydrogenase (LDH) into cell culture supernatant was measured using the cytotoxicity detection kit LDH (Roche Molecular Biochemicals) according to the manufacturer's instructions.

Generation of H-Ras-and Rlf-CAAX-expressing HME16C cell lines
Retroviral vectors coding for amino terminal HA-tagged activated H-Ras V12 , the H-Ras V12 effector domain mutants (EDM) H-Ras V12G37 , H-Ras V12S35 , and H-Ras V12C40 , and the constitutively activated version of a RalGEF, Rlf-CAAX, were used to infect telomerase immortalized HME16C human mammary epithelial cells. Anti-Ras and anti-HA western blotting demonstrated approximately equal levels of ectopic Ras expression among Ras-infected cells, with slightly lower levels in HME16C Ras V12 -infected cells relative to EDM-infected cells ( Figure 1A). Analysis of activated, GTP-bound Ral A demonstrated highly elevated levels of activated Ral A only in Rlf-CAAX-expressing cells and not in Ras-infected cells grown under standard culture conditions ( Figure 1B). To assess activation of the ERK pathway, anti-phospho Erk western blotting was performed and showed significantly elevated Erk phosphoprotein in Ras V12 -and Ras V12S35 -infected HME16C cells relative to control pLRT-infected cells ( Figure 1A). Elevated levels of phosphorylated Erk were also observed in Ras V12G37 -and Ras V12C40 -infected cells, although at a much lower level than that found in Ras V12 -and Ras V12S35infected cells. To assess activation of the PI3K signaling pathway, anti-phosphoAkt (Ser473) western blotting was performed to detect activated, phosphorylated Akt. In cells that were serum starved for 24 hours and matrix detached for 6 hours, elevated levels of phosphorylated Akt were observed in Ras V12 -, Ras V12C40 -, and Ras V12G37infected HME16C, with highest levels present in Ras V12infected cells ( Figure 1C).

Anchorage-independent growth of mammary epithelial cell lines
To assess transformation by different Ras signaling pathways, anchorage-independent growth assays were performed in soft agar and in ultra-low attachment tissue culture plates. Ras-and Ras EDM-infected HME16C cells formed significantly more soft agar colonies >100 μm in diameter than pLRT vector-infected cells (Figure 2A). The Ras V12 -infected cells formed large colonies, many exceeding 1000 μm, although the total number exceeding 100 μm was typically less than that for the Ras EDM ( Figure   2B). Among the Ras EDM-infected cells, the Ras V12S35 -infected cells formed the largest colonies. These were similar to, but smaller than, the Ras V12 -infected cells. For colonies above 100 μm in diameter, Ras V12C40 -infected cells were the most efficient at colony formation, despite the smaller mean size of colonies. Rlf-CAAX-infected cells formed slightly more colonies above 100 μm than vectortransfected control cells, but these were significantly smaller than those formed by Ras-and Ras EDM-infected cells.
When grown under anchorage-independent conditions in ultra-low attachment plates, the accumulation of cells for the various infected cell lines roughly paralleled the total cell masses seen in soft agar growth assays ( Figure 2C). The Ras V12 -and Ras V12 EDM-expressing cells grew well, while the growth of the Rlf-CAAX-expressing cells was significantly less. The HME16C cells maintained viability but did not increase in number. To compare our results to others, we verified the function of pLRT vector driven Ras EDM mutants and Rlf-CAAX in HEK-HT cells, which previously have been reported to form colonies in soft agar upon expression of H-Ras V12G37 and Rlf-CAAX [27]. In our hands, expression of H-Ras V12 , H-Ras V12G37 , and Rlf-CAAX from the pLRT vector induced efficient soft agar colony growth, and H-Ras V12S35 and H-Ras V12C40 did not (not shown), identical to previously reported results [27]. Expression of exogenous H-Ras and Rlf-CAAX, and activation of endogenous RalA, in this cell line was similar to that observed in HME16C cells (not shown). Therefore, expression of Rlf-CAAX and subsequent RalA activation did not appear to be sufficient to induce anchorage-independent growth in the HME16C mammary epithelial cell line in contrast to HEK-HT and various other immortalized human cell types [27].

Tumorigenesis of HME16C cell lines in nude mice
Anchorage-independent growth often predicts the ability of cells to grow as xenografted tumors in immuno-compromised mice. Tumor formation was assessed following subcutaneous inoculation. The Ras V12 -infected HME16C cells formed rapidly growing, fluid-filled tumors with an average latency of 4 weeks and a mean tumor volume of 808.8 mm 3 at 6 weeks (Table 1). Approximately one-half of the tumors were aspirated prior to sacrifice, and a serosanguinous fluid was observed, on average accounting for roughly one-third of the measured tumor volume. Histological analysis of H&E stained tumor sections revealed poorly differentiated spindle-shaped tumor cells with prominent squamous cell differentiation and extracellular keratin deposition (not shown). Tumors also contained a strong inflammatory component. Of the other cell lines tested, only the Ras V12S35 -infected HME16C cells formed Anchorage-independent growth of HME16C cell lines palpable tumors in 50% of injected animals with an average latency of approximately 12 weeks and a mean tumor volume of 109.0 mm 3 at 16 weeks, considerably smaller than Ras V12 -expressing tumors (Table 1). Cells within Ras V12S35 -infected tumors resembled the histology of Ras V12 tumor cells but with less keratin deposition and without the formation of fluid-filled spaces (not shown). Empty vector-, Ras V12G37 -, Ras V12C40 -, and Rlf-CAAXinfected cells failed to form palpable tumors four months after injection. The metastatic potential of Ras V12 -and Ras V12G37 -expressing cell lines was tested by tail-vein injection in nude mice, but no metastatic lesions were observed by histological analysis in lungs, liver, spleen, or kidneys at 16 weeks post injection (not shown).

Autocrine EGFR signaling is required for Ras V12G37 -and Ras V12C40 -mediated, but not Ras V12S35 -mediated, HME16C cell anchorage-independent growth
EGFR signaling is frequently altered in breast cancer, where EGFR and ErbB2 over-expression are common events. cDNA microarray and real-time RT-PCR analysis of HME16C Ras V12 -and Ras EDM-infected cells (Additional file 1) revealed increased levels of mRNAs for EGFR ligands, including epiregulin, amphiregulin, and TGFa (Table 2). In addition, increased levels of phospho-Erk were unexpectedly seen in Ras V12G37 -and Ras V12C40infected cells, possibly the result of autocrine activation of endogenous EGFR by secreted EGFR ligands. The presence of EGFR was established using Western blots (not shown). Therefore, we sought to determine if autocrine signaling by the EGFR was required for the anchorage-independent growth of Ras-or Ras EDM-infected cells. Ras V12 -and Ras EDM-infected HME16C cells were grown in soft agar in the presence or absence of the EGFR-specific inhibitors, PD153035 and PD168393. Complete inhibition of colony formation for Ras V12G37 -infected and Ras V12C40infected cells was observed at 0.25 μM PD153035, a concentration that specifically inhibits EGFR [28], whereas both the Ras V12 -and Ras V12S35 -infected cells formed colonies efficiently at this same concentration of inhibitor ( Figure 3A and 3B). Identical results were found for the EGFR-specific inhibitor PD168393 used at 0.1 μM, a concentration that specifically inhibits EGFR and Her-2 receptors [29](not shown). Similarly, treatment of cells grown in ultra-low-attachment plates also demonstrated that EGFR inhibition substantially inhibited growth of Ras V12G37 -and Ras V12C40 -expressing cells relative to that of Ras V12 -and Ras V12S35 -expressing HME16C ( Figure 2C). Western blotting of cellular lysates from cells treated with 0.25 μM PD153035 showed that high levels of phosphorylated Erk were maintained only in Ras V12 -and Ras V12S35infected cells, but were significantly reduced in Ras V12G37or Ras V12C40 -infected cells treated with the inhibitor (Figure 3C), while phoshorylated Akt was minimally affected (not shown). Anchorage-independent growth therefore correlated with maintenance of high Erk activity in HME16C cells. Consistent with this observation, inhibition of MEK, and therefore ERK signaling, using the MEKspecific inhibitor PD98059 at 10 μM, significantly inhibited soft agar colony formation by all cell lines (not shown).

Microarray analysis of gene expression changes in Ras V12 -, Ras V12G37 -, Ras V12S35 -, and Ras V12C40 -infected HME16C cells
Activation of the Ras oncogene is accompanied by the stimulation of multiple signal transduction pathways leading to the activation or repression of numerous transcription factors as well as changes in mRNA translation and stability, and thus, the modulation of gene expression. To determine which gene expression changes accompany the transformation of HME16C human epithelial cells by activated Ras, we examined our transformed  To evaluate the effect of EGFR inhibition on gene expression, Ras V12 -, Ras V12S35 -, and Ras V12G37 -infected cells were induced with doxycycline and subsequently incubated either in the presence or absence of 0.25 μM PD153035, and microarray analysis comparisons were made to vehicle-treated pLRT-infected cells. Almost all Ras and Ras EDM-induced upregulated transcriptional responses were blocked by pharmacological inhibition of EGFR, consistent with previous reports for inhibition of Raf-regulated transcription [30]. Our analysis identified PHLDA1 as an up-regulated gene in both vehicle-treated and PD153035treated Ras V12 and Ras V12S35 cells, although the relative fold increase was reduced following EGFR inhibition. By comparison, PHLDA1 was down-regulated in PD153035treated Ras V12G37 relative to vehicle-treated cells (not shown). Therefore, PHLDA1 represents a Raf/ERK-respon- sive gene whose expression parallels EGFR-independent HME16C mammary epithelial cell transformation.

TDAG51 expression is up-regulated by Ras signaling in a ERK-dependent manner, and is associated with EGFRindependent transformation
The PHLDA1 gene is of interest as it has been suggested to be a tumor suppressor in breast adenocarcinoma and melanoma [17,18]. We further analyzed the signal dependent expression of the PHLDA1 gene and its protein product, TDAG51. Microarray analysis identified the PHLDA1 gene as being dramatically up-regulated in Ras V12 -and Ras EDM-infected cells to levels that correlated with the level of ERK activation and the extent of anchorage-independent growth (Additional file 1). Western blotting confirmed that TDAG51 was also upregulated in a similar manner ( Figure 4A). The PHLDA1 gene was elevated in PD153035-treated Ras V12 -and Ras V12S35infected cells but was significantly dependent upon EGFR tyrosine kinase activity for upregulation in Ras V12G37 -and Ras V12C40 -infected cells (not shown), and the expression of the encoded TDAG51 protein approximately paralleled PHLDA1 RNA expression ( Figure 4B). As shown in Figure  3C, EGFR inhibition significantly reduced ERK signaling in Ras V12G37 -and Ras V12C40 -infected cells without affecting Ras V12 -and Ras V12S35 -infected cells. To confirm that TDAG51 up-regulation was induced specifically by ERK activation, we treated pLRT-, Ras V12 -, and Ras V12S35infected cells with the MEK-specific inhibitor PD98059. PD98059 used at 20 μM appears to be specific for MEK1 as it does not nonspecifically inhibit a variety of other protein kinases that have been assayed. [31]. TDAG51 up-regulation was attenuated by MEK inhibition ( Figure 5). TDAG51 therefore represents an ERK-inducible gene whose up-regulation in HME16C is correlated with an EGFR-independent, ERK-mediated transformation.

TDAG51 up-regulation opposes ERK-mediated HME16C transformation
To analyze the role of TDAG51 in ERK-dependent growth, we reduced TDAG51 expression in Ras V12 -and Ras V12S35infected cells to a level comparable to that in non-transformed vector-infected control cells using stably expressed TDAG51-specific shRNA ( Figure 6A). Cell proliferation of attached cells grown on tissue culture plastic was unaffected by lowered TDAG51 protein levels (not shown). However, cell growth under anchorage-independent conditions in ultra-low attachment plates was significantly enhanced by TDAG51 knock-down in Ras V12S35 -infected cells ( Figure 6B). Likewise, results with Ras V12 -infected cells stably infected with TDAG51-targeting shRNA also showed enhanced growth relative to vector-infected control cells, although to a lesser extent than that seen with Ras V12S35 -infected cells ( Figure 6C). This suggests that Ras signaling pathways other than ERK compensate partially for the negative growth effects of TDAG51, or that Ras V12infected cells are already close to maximally transformed under anchorage-independent growth conditions.

Loss of TDAG51 expression in transformed cells stimulates both cell cycle progression and apoptosis
To characterize the effect of TDAG51 on cell proliferation under anchorage-independent conditions, cell cycle analysis and cell proliferation assays for Ras V12S35 -and Ras V12infected cells were performed. Both Ras V12S35 -and Ras V12 -TDAG51 shRNA-expressing cells demonstrated an increased S-phase fraction versus pLVTHM vector control cells at various time points during anchorage-independent growth ( Figure 7A). In concordance with these results, Ras V12S35 and Ras V12 cells expressing the TDAG51-specific shRNA showed enhanced incorporation of 5-ethynyl-2'deoxyuridine (EdU) in cell proliferation assays, indicating a higher rate of DNA synthesis in cells with reduced TDAG51 protein ( Figure 7C).
Since cell growth under anchorage-independent conditions is a balance between cell proliferation and cell death, we sought to evaluate the effect upon cellular death of TDAG51 knock-down. We used an assay of cellular cytotoxicity that measures the release of the lactose deydrogenase enzyme, LDH. LDH release was increased by TDAG51 shRNA-expressing Ras V12S35 cells relative to pLVTHM-infected cells at various time points after the initiation of matrix-detached growth (Figure 8). The difference in LDH release for TDAG51 shRNA-expressing Ras V12 cells was minimal and rarely approached statistical significance. Sub-G1 peaks indicative of dead cells were sometimes seen with cell cycle analysis at late time points, but varied from experiment to experiment. Therefore, for Ras V12S35 -infected cells, the differences in cell growth after TDAG51 reduction under anchorage-independent conditions resulted from an enhanced rate of cellular proliferation that exceeded a concomitant increase in cell death.

Reduction of TDAG51 in transformed cells enhances proximal ERK signaling
Reducing TDAG51 protein levels in ERK-driven cellular transformation enhanced cell growth under anchorageindependent, but not attached, conditions. To test whether TDAG51 might affect proximal ERK signaling, we examined the activation status of Erk in cells expressing TDAG51-specific shRNA. Interestingly, the levels of phosphorylated Erk were enhanced when TDAG51 protein levels were reduced in Ras V12S35 and Ras V12 cells grown under anchorage-independent, but not attached, conditions ( Figure 9). The fact that the activation status of Erk was unchanged in cells grown under attached conditions suggests that reducing TDAG51 expression had no selective effect with regard to ERK activation in these cells. Rather, the enhanced activation of Erk was specific to anchorageindependent growth conditions.

Discussion
Ras is a common signaling node for various cell surface receptors that contribute to epithelial cell transformation [14]. In this study, we used the hTERT-immortalized human mammary epithelial cell line HME16C to examine which Ras signaling pathways are sufficient for transformation and to identify transcriptional targets downstream of those pathways that might modulate this phenotype. Transduction of HME16C with pathway-discriminating Ras effector domain mutants demonstrated that multiple downstream Ras signal transduction pathways contribute to anchorage-independent growth including Raf-, Ral-GEF-, and PI3K-mediated signaling. Transformation of HME16C by the Ras V12G37 effector domain mutant but not activated Rlf-CAAX suggest that Ras V12G37 -binding effectors other than RalGEF contribute to mammary epithelial transformation.
Microarray analyses of Ras V12 and Ras effector domain mutant transduced cells demonstrated a common upregulation of EGFR ligands among transformed cell lines. This suggested that autocrine EGFR ligand secretion was an important component of Ras-mediated cellular transformation. Following blockade of EGFR signaling with the EGFR-specific inhibitor PD153035, the only pathway downstream of Ras that promoted anchorage-independ-ent growth was Raf/ERK, suggesting that Raf activation was able to substitute for EGFR activity in this cell line. In contrast, previous studies with MCF10A cells demonstrated that EGFR tyrosine kinase activity was necessary to inhibit anoikis upon matrix detachment, even in cells expressing activated Raf [16]. By contrast, under matrix detached conditions, the parental HME16C cells are nonproliferative, but do not actively undergo anoikis. The studies here measure a proliferative response of Raf/ERK activation in the absence of a strong apoptotic component. In addition, blockade of MEK activity with the MEKspecific inhibitor PD98059 prevented the transformation of Ras V12 -and Ras V12S35 -expressing cells. This, taken together with the finding that Ras V12 -and Ras V12S35 -transduced HME16C cells were able to form tumors in nude mice, indicates that the Raf-MEK-Erk axis plays a crucial role in mediating transformation and tumorigenesis in this model. However, it should be noted that Ras-transformed HME16C, as well as the similar HMLER [32], give rise to squamous metaplasias and not adenocarcinomas. It has been hypothesized that HMLE human mammary epithelial cells represent a distinct precursor population from those mammary epithelial cells that give rise to glandular adenocarcinomas, the predominant form of breast cancer [32].
The significantly reduced tumorigenicity of Ras V12S35 -as compared to Ras V12 -expressing cells emphasizes the importance of cooperation between Ras signal transduction pathways for a fully transformed phenotype. Following EGFR inhibition, Ras V12C40 -and Ras V12G37 -transduced lines lost anchorage-independent growth, suggesting that non-ERK signaling pathways contribute to cellular transformation through an EGFR dependent mechanism. Indeed, one consequence of EGFR activation is ERK pathway activation and may explain the role of EGFR in cooperating with Ras V12G37 -and Ras V12C40 -mediated cell signaling to promote anchorage-independent growth.
Microarray analyses identified up-regulation of the PHLDA1 gene product as being correlated with ERKmediated cellular transformation. Likewise, the protein product of the PHLDA1 gene, TDAG51, displayed an identical expression pattern. Expression of PHLDA1 mRNA and TDAG51 protein in breast cancer has recently been described [17]. Using TDAG51 immunohistochemistry of tissue microarrays, 699 individual primary invasive breast tumor specimens were examined; loss of TDAG51 was found to correlate with a poorer disease-free and overall survival rate in multivariate analysis. Similar results have been found in a small series of melanomas, where TDAG51 immunoreactivity was found to decrease during the progression of melanocytic nevi to primary melanomas and finally to metastatic melanoma [18]. These studies suggest that TDAG51 has a suppressive TDAG51 expression is induced by ERK pathway activation The PHLDA1 gene is a member of the pleckstrin homology-related domain family that includes Ipl/Tssc and Tih [33]. PHDLA1 has been described as an immediate early gene with transcriptional activation resulting from engagement of receptors such as the FGF and IGF tyrosine kinase receptors and the T-cell receptor [34][35][36]. Strong TDAG51 induction by Ras V12S35 transformation of mammary epithelial cells suggests that ERK activation may at least partially explain TDAG induction in these previous reports. In vitro studies have suggested a role for TDAG51 in the control of cellular proliferation and in the induction of apoptosis in response to a variety of stresses including proteotoxic cellular stresses such as lung cancer cell responses to chemotherapy [35,[37][38][39].
Overexpression of TDAG51 has been shown to reduce cell proliferation and induce cell death in a variety of cell types including T cells, neuronal, endothelial, melanoma, and cervical carcinoma cell types [34,35,38]. By contrast, TDAG51 functioned as an anti-apoptotic factor downstream of insulin-like growth factor receptor (IGFR) signaling as TDAG51 was required to protect NIH3T3 cells from apoptosis upon IGF-1 withdrawal [36].
Reduction of TDAG51 levels in Ras V12 and Ras V12S35 cells enhanced cell proliferation under anchorage-independent conditions. This suggests that TDAG51 expression limits cellular proliferation. In addition to enhancing cellular proliferation, reduction of TDAG51 protein levels also increased the absolute amount of cell death in transformed Ras V12S35 cells under these same conditions. Therefore, TDAG51 loss, in the context of oncogene activation, may indirectly promote cell death as a consequence of enhanced cell cycling. However, the overall increased cell numbers in anchorage-independent conditions showed that enhanced cellular proliferation exceeded enhanced cell death.
Interestingly, a decrease in TDAG51 expression during Ras-mediated cellular transformation promoted the  growth of cells under anchorage-independent conditions but did not affect the growth of attached cells. This suggests that TDAG51 might act in conjunction with both cellular stress, in this case matrix detachment, and a proliferative signal, in this case oncogenic activation.

Reduction of TDAG51 expression in Ras V12 and Ras V12S35 -infected cells and their anchorage-independent growth
Other studies also have implicated TDAG51 in functioning during cellular stress and survival, particularly endoplasmic reticulum stress associated with the unfolded protein response [37,38]. A mechanism of action for TDAG51 is not known. The preliminary finding that TDAG51 binds to proteins involved in protein translation has been used to suggest that TDAG51 may play a role in regulating protein synthesis in response to proteotoxic stress [40].
Reduction of TDAG51 expression resulted in an increase in Erk activation in cells grown under anchorage-independent conditions. How TDAG51 expression might suppress ERK signaling is not known, but appears to represent a negative feedback pathway that directly or indirectly limits ERK activation. This is not likely to be due to an inhibition of ERK protein synthesis by TDAG51, since Erk2 levels were unaffected by reduced TDAG51. However, another important component of ERK activation is the dual specific ERK phosphatases, a highly regulated Cell proliferation is increased in Ras-transformed cell lines with reduced TDAG51 expression class of proteins, whose relative level of activity may be affected by TDAG51 protein levels. Understanding the mechanisms by which TDAG51 regulates ERK pathway activation and the balance between cellular proliferation and apoptosis of transformed cells represents a future challenge. Finally, TDAG51 acts in a suppressive manner during matrix-detached growth of HME16C cells. Taken together with the identification of TDAG51 as a stressinduced gene in a variety of cell lines and a growth inhibitor in melanoma cell lines, it is reasonable to suggest that loss of TDAG51 may act to promote progression of breast cancer through an intrinsic growth regulatory mechanism.

Conclusion
Expression of activated Ras effector domain mutants that bind Raf, PI3K, or RalGEF are sufficient to induce the anchorage-independent growth of the human mammary epithelial cell line HME16C and are associated with upregulation of EGFR ligands. However, only the ERK pathway is capable of supporting transformation in the absence of EGFR signaling and of supporting tumorigenesis in nude mice. Up-regulation of TDAG51 occurs during Ras-mediated transformation in an ERK-dependent fashion, but opposes ERK-mediated transformation by suppressing ERK signaling and decreasing cellular proliferation under matrix-detached conditions. Therefore, in this model of mammary epithelial cell transformation, TDAG51 acts as a growth inhibitor of ERK-driven proliferation and may help explain why loss of TDAG51 expression has been found to correlate with progression in human breast cancer and melanoma.

Abbreviations
EGFR: epidermal growth factor receptor; ErbB2: epidermal growth factor receptor family member B2; Erk: extracellular regulated kinase; PHDLA1: pleckstrin homology domain-like, family A, member 1; PI3K: phsphatidyl LDH release by Ras V12S35 -or Ras V12 -expressing HME16C infected with pLVTHM vector or anti-TDAG51 shRNA 36-1 ERK pathway activation in Ras V12 and Ras V12S35 cells infected with pLVTHM or anti-TDAG51 shRNA 36-1 Figure 9 ERK pathway activation in Ras V12 and Ras V12S35 cells infected with pLVTHM or anti-TDAG51 shRNA 36-1. Ras V12 or Ras V12S35 cells infected with either pLVTHM vector control or anti-TDAG51 shRNA 36-1 were grown for 72 hours attached to tissue culture plastic (attached) or in ultralow attachment plates (anchorage-independent) in the presence of 250 ng/mL doxycycline. ERK pathway activation was measured by determination of phosphorylated Erk1/2 levels by western blotting using an anti-phosphorylated Erk mAb (phospho-Erk). The blot was stripped and subsequently probed with an anti-Erk2 mAb (Erk2).