This study has identified genes whose differential expression may contribute to ErbB2-dependent transformation and which define common and specific signalling events induced through EGFR and ErbB3 receptor-containing complexes. Although we and others have previously examined ErbB2-dependent gene expression changes in the same cell model, and find overlap in the genes identified [37, 38], to the best of our knowledge, this is the first study to simultaneously investigate long-term ErbB2- and GF-dependent gene expression using ligands that activate specific ErbB receptor complexes in the same cell system. A number of gene expression changes were further validated using qRT-PCR and we report a good correlation between the datasets, indicating the robustness of the microarray protocol employed.
There were significantly more HRG-responsive genes than EGF-responsive genes and in many cases the HRG response was elevated in the ErbB2-overexpressing cells. This is likely to be a consequence of the higher expression of ErbB2 and ErbB3 in these cells  and the preferred heterodimerzation of these receptors [3–6], which would act to augment the response to HRG. We do not think that ErbB4 (also a HRGβ1 receptor) plays a major role in orchestrating signalling events in this cell system, since it appears to be expressed at very low levels, if at all, in these cell lines (data not shown). Although HRG-induced expression was generally of a lower magnitude than for EGF, it was often sustained compared to EGF, consistent with our previous finding that HRG-dependent mitogenic signalling is sustained in these cells . Such temporal differences may be connected with differential rates of receptor or signal down-regulation, but also highlight the fact that the two growth factors initiate diverse responses which are likely to be relevant in vivo. Genes induced robustly by HRG (ZNF236, ZFP36L1, ZFP36L2, MADH4, TRIO, HMGCR, SLC16A1, SLPI, GYS1, SFRS5, CTNND1, LCAT, LYN, STAT1, KRT15, C20orf16 and FN1) are likely candidates for regulation by the PI3K/Akt pathway which is potently activated by HRG through ErbB2-ErbB3 heterodimers [7, 8]. Since HRG expression itself correlates with tumourigenicity and metastasis in breast cancer cells lines [39, 40], it will be interesting to assess whether the induction of these genes is affected by chemical inhibition of the PI3K pathway, or whether such inhibitors would make clinically useful therapeutics for breast cancer treatment.
Notably, a group of EGF-specific genes (e.g. AREG, S100A2 and CTSC) were induced exclusively in the HB4a cells, potentially through EGFR homodimers which predominate in these cells . One of these genes, AREG, is a ligand of EGFR itself, suggesting that EGF could drive autocrine signalling to enhance EGF-specific responses. Members of the MT family were also potently induced by EGF. Whilst induction of MT1 expression by EGF has been shown in rat hepatocytes , this is the first report of MT1 (and MT3) induction by EGF in human epithelial cells. Since the altered expression of MT family members has been implicated in neoplasia and drug resistance [42, 43], it will be interesting to investigate whether MT expression is linked to deregulated GF signalling in cancer.
Many of the identified genes have been previously implicated in tumour progression, found to be aberrantly expressed in different tumour types and/or to be linked with poor prognosis, hyper-proliferation, cell survival or tumour invasiveness. Our findings suggest that dysregulated ErbB signalling can account for changes in the expression of these genes, and may thus contribute to the establishment and progression of ErbB2-overexpressing breast tumours. For example, of the genes induced by both GFs and augmented by ErbB2, the proto-oncogenic transcription factor MYC has been associated with many forms of cancer often indicating poor prognosis . Importantly, patient survival was significantly reduced in breast cancers where MYC and ErbB2 are co-amplified . The MYC-induced glycoprotein EMP1 was also similarly regulated and whilst its function is unknown, it has reported tumourigenic activity  and was identified as a marker of gefinitib-resistance in xenograft models . Thus, one possible scenario that warrants further investigation is that EMP1 acts in concert with MYC to promote ErbB2-dependent proliferation and drug resistance. A pattern of ErbB2-augmented GF-induction was also observed for other genes known to be involved in proliferation, autocrine signalling and anti-apoptosis (e.g. ATF4, FOSL1, IER3, MAP2K1/MEK1, MAP2K3/MEK3, PDGF, TNFAIP3, VEGF) and it is possible that these changes contribute to the reported hyper-proliferative phenotype of these ErbB2-overexpressing cells [31, 32]. Induction of the pro-angiogenic factor VEGF is particularly relevant to tumour progression and confirms previous data [48, 49]. Notably, VEGF expression was shown to depend upon ATF4 expression under certain conditions  and we hypothesize that such a regulatory circuit exists in these cells, whereby ErbB2-augmented GF signalling would promote VEGF expression through up-regulation of ATF4. The induction of some genes was perhaps surprising given their reported functions. GADD45A, SFN and the dual-specificity phosphatases DUSP1/MKP1 and DUSP5 were induced by GF treatment and are involved in genotoxic stress-induced growth arrest , p53-dependent negative regulation of G2/M progression  and down-regulation of MAPK signalling, respectively . We propose that these may be negative feedback mechanisms adapted to self-regulate proliferative signalling.
Conversely, the down-regulation of genes with anti-proliferative functions identifies mechanisms by which increased ErbB2 signalling may promote proliferation and survival. Examples include the multiple ISGs that were identified and IGFBP3. G1P2/ISG15 was the most down-regulated gene in the dataset. Like ubiquitin, G1P2 is conjugated to proteins in a process called ISGylation which appears to modulate protein activity during the immune response and signalling . The other ISGs were UBE2L6 (the proposed E2 enzyme for ISGylation ), IFIT1, IFITM1, IFITM2, OAS1 and ISGF3G/p48/IRF9. Notably, ISGF3G is a component of a transcription factor complex that with STAT1 and STAT2 controls type I IFN-mediated induction of ISGs containing interferon-stimulated regulatory elements (ISREs) . The lowered expression of ISGF3G could thus account for the down-regulation of the other ISGs in the ErbB2-overexpressing cells, as suggested by our previous work . Whilst the ISGs were induced by IFN treatment in the HMLECs, induction of ISGF3G (particularly with IFNγ) was blocked by GF co-treatment, revealing a possible cross-talk between the IFN and ErbB signalling pathways (data not shown). Although preliminary, our data suggested an inverse correlation between ErbB2 and ISG expression, supporting a role for repressed basal ISG expression in the pathogenesis of ErbB2-dependent breast cancer.
IGFBP3 mRNA and protein expression were both markedly lower in the ErbB2-overexpressing cells, whilst mRNA levels were decreased by GF treatment, particularly in the parental cells. Given IGFBP3's putative role as a negative regulator of IGF1 signalling , its anti-proliferative role  and the negative correlation between serum IGFBP3 levels and cancer risk [58–60], we investigated a possible link between its expression and IGF1 signalling. We found that IGF1-mediated ERK and Akt activation and proliferation were increased in the ErbB2-overexpressing cells and that the signalling effect was reversed by siRNA-mediated knockdown of ErbB2. The mechanism by which this occurs is unclear, although does not involve altered IGF1R expression, and may be mediated through interaction between ErbB receptors and IGF1R as previously reported in other cell models [61–63]. ErbB2 may also down-regulate IGFBP3 expression to promote IGF1 signalling. We propose that ErbB2-dependent suppression of IGFBP3 expression is a long-term adaptive response and would be the reason why IGFBP3 protein levels were not affected by transient ErbB2 knockdown. We speculate that this may be due to IGFBP3 promoter methylation, as previously reported for other cancers [64, 65]. In the C3.6 cells, IGFBP3 expression is suppressed, allowing maximal IGF1 signalling through ErbB2-IGF1R interaction [61–63]. Knocking down ErbB2 in these cells therefore does not affect IGFBP3 levels, but abrogates IFG1 signalling. In HB4a cells, IGF1 signalling is restricted by normal IGFBP3 expression with knockdown of IGFBP3 enhancing basal ERK1/2 and Akt activation, thus supporting its role as a negative regulator of proliferation and survival. Although reduced IGFBP3 expression did not affect acute IGF1 triggering, our data partly support findings in primary and immortalized human esophageal cells, where EGF-mediated down-regulation of IGFBP3 was shown to determine cellular response to IGF1 . However, this effect may be mediated by the as yet unknown IGF1-independent actions of IGFBP3 (reviewed in [67, 68])
The observed increases in invasiveness and anchorage-independent growth of ErbB2-overexpressing SKBR3 cells following knockdown of IGFBP3 supports a role for IGFBP3 as a negative regulator of cellular transformation in breast cancer and we propose that its down-regulation is a mechanism whereby ErbB2 promotes tumour cell growth through increased IGF1-dependent proliferation, survival and invasion. Indeed, a requirement for IGF1 in EGF-mediated cell cycle progression has been shown in primary murine mammary epithelial cells . Whilst an attractive model, other studies report that IGFBP3 can potentiate EGF-stimulated proliferation in MCF10A cells  and that IGFBP3 expression is associated with growth stimulation of T47D human breast cancer cells . These differences may be explained by cell type-specific effects and are possibly dependent upon the extent of interaction with the ErbB receptor system . Future experiments should explore the effects of overexpressing IGFBP3 on IGF1 signalling, proliferation, survival and invasion and to investigate the level of IGFBP3 promoter methylation in this cell system.
We have previously reported a high correlation between mRNA and protein expression for a subset of genes in these cell lines , and a previous proteomic study found reduced expression of GSTP1, PRDX5 and USP14 and increased expression of KRT13, ALDH1A3 and NME1 in the C3.6 cells , in agreement with the mRNA data presented here. In the present study, the mRNA expression of several targets (MYC, CLDN4, S100A6, ZYX, PHB, MAP2K1, NME1, AGR2, PKM2, IGFBP3, ISGF3G, G1P2 and ANXA2) correlated with altered protein expression, signifying that these changes are likely to be functionally relevant. However, correlation between protein and mRNA expression was not apparent for some targets in response to the GF treatments. For example, the repression of IGFBP3 mRNA by GF treatment was not confirmed at the protein level and neither was induction of DUSP1 or SFN. This suggests that the IGFBP3 protein may be relatively stable over the time course of the assay or that the DUSP1 and SFN mRNAs are not translated. Such post-transcriptional regulatory mechanisms are likely to be important, and whilst some mRNA changes appear to be redundant, they may be relevant in other circumstances, for example, during development, differentiation or stress.
A relatively large group of genes involved in regulating the cytoskeleton, cell adhesion and motility were identified. Whilst various patterns of gene expression were apparent, genes up-regulated to a greater degree by either GF in the ErbB2-overexpressing cells (ZYX, VIM, VCL, TAGLN, VIL2, PDLIM1, ITGA2, ITGA3, PLAT, PLAUR, SERPINE1 and ANXA2) are perhaps the most interesting, since they may promote the ErbB2-mediated anchorage-independent growth and reduced cellular adhesion previously observed in this cell model system [31, 38]. Notably, some of these genes are members of the plasminogen activator system and have been implicated in tumour progression and invasiveness through proteolysis of the extracellular matrix. Indeed, increased levels of PLAUR and SERPINE1 have been associated with poor prognosis in breast cancer patients [73, 74]. Our data thus implicates ErbB2-mediated signalling in the regulation of the plasminogen activator system, as well as cell adhesion-related events.
Finally, a number of genes of unknown or poorly-defined function were identified and several were validated. These include BCAR3, CPNE3, CSRP1, HPCAL1, LCP1, MGC10471, NME1, SMAP, ZFP36L1, ZFP36L2 and ZNF236, which were differentially regulated by GF in an ErbB2-dependent manner and AGR2, LOC402057, NPC2, PSCA, S100P and SERF2, which were differentially expressed in an ErbB2-dependent manner. Our data reveals that the expression of these genes can be regulated by ErbB receptor signalling and thus implicates them as possible biomarkers and effectors of ErbB2-dependent tumourigenesis. Indeed, AGR2, LCP1 and S100P overexpression have been previously correlated with breast cancer progression [75–77], and we now link the aberrant expression of these genes with ErbB2 expression.