Polyclonal rabbit antibodies against HSP90α were acquired from StressGen (Victoria, BC, Canada), and rabbit anti-STAT5B from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Mouse anti-phospho-histone 2A.X (Ser139) clone JBW301 and anti-GRB2 antibodies (Upstate, Charlottesville, USA) and mouse anti-GRB2 (BD Biosciences, Ontario, Canada) were also acquired. GRB2 was used as a loading control 42.

The HSP90A-luciferase reporter gene contains about 1.8 kb of promoter sequences (Xho I-Hind III) of the human HSP90A gene cloned into pLux F3 (KS89 α XL Lux). Expression constructs for the β-galactosidase gene, STAT5A and STAT5B 43, STAT1 and STAT3 (pME18S) and prolactin-receptor 44 have been described. The DNA encoding human HSP90A (2199 bp, accession number [GenBank:X15183]) was amplified by PCR from cDNA prepared from SKBR3 cells (SMART cDNA synthesis kit, Clontech Laboratories, Mountain View, CA, USA). The resulting DNA was cut with Bam HI and Not I (New England Biolabs, Ipswich, MA, USA) (adaptors on the primers), subcloned into pcDNA 3.1/Zeo(+) behind the constitutively active cytomegalovirus immediate early promoter (CMV) and the expected sequence verified.

SKBR3 cells, a human breast cancer cell line, were grown in Dulbecco's Modified Eagle Medium with L-glutamine and 10% fetal bovine serum. Undifferentiated HC11 cells, a mouse mammary epithelial cell line 45, were cultured in RPMI with fetal bovine serum and maintained in 0.01 μg/ml epidermal growth factor (EGF) and 5 μg/ml insulin. Confluent cells became competent to respond to lactogenic hormones after incubation in 0.01 μg/ml EGF for one to four days and then 5 μg/ml insulin and 1 × 10^-7 M dexamethasone for one day (option if treating with EGF for only one day). Differentiation was then induced (for three days if insulin and dexamethasone pre-treatment was used, or four days if not) with 1 × 10^-7 M dexamethasone, 5 μg/ml insulin and 5 μg/ml prolactin. Undifferentiated HC11 cells were transfected with HSP90A-pcDNA 3.1/Zeo(+) (HC11-HSP90α) or the empty vector pcDNA3.1/Zeo(+) (HC11-EV) using Lipofectamine 2000 (catalogue 11668-019, Invitrogen Corporation, Carlsbad, California, USA) according to the manufacturer's instructions. Transfected cells were selected as a pool with zeocin treatment and constitutive expression was verified by both northern and western blotting.

RNA was prepared (RNeasy, Clontech, Heidelberg, Germany) from SKBR3 cells seeded at 1 × 10⁶ and 2 × 10⁶ cells per 15 cm plate, starved of fetal bovine serum the following day for 16 hours and then treated with or without 5 μg/ml prolactin for 60 minutes in the presence of 10 nM cycloheximide. The subtraction hybridisation libraries were prepared using the SMART PCR cDNA Synthesis Kit and the PCR Select Subtractive Hybridization kit (Clontech, Heidelberg, Germany), and probed using the PCR Select Differential Screening kit (Clontech, Heidelberg, Germany). Positive clones were sequenced to obtain their identity.

Total RNA was extracted using peqGOLD Trifast (peqLab, Erlangen, Germany) and resolved on a formaldehyde agarose gel. The blot was blocked using ExpressHyb (Clontech, Heidelberg, Germany), and hybridised with radioactively labelled probes (Strip-EZ DNA, Ambion, AMS Biotechnology Ltd). A DNA probe encoding exon 3 of the human CIS gene 46 was amplified using the forward primer 5'-GCT GGT ATT GGG GTT CC-3' and the reverse primer 5'-TGA GGG CTC TGT ACA TGA AAG-3' The fragments were gel purified (Qiaquick Gel Extraction Kit, Qiagen GmbH, Hilden, Germany).

Electrophoretic mobility shift assays using a radioactively labelled fragment of the bovine β-casein promoter were performed as described 47. Essentially the double strand STAT responsive element from the β-casein promoter was prepared and radioactively labelled before incubation with protein extracts. The complexes were resolved on a non-reducing gel and autoradiographed.

Transfection using calcium chloride and luciferase assays using HeLa cells was performed as described 48 using overnight treatments with 5 μg/ml prolactin. A β-galactosidase gene was included to compare transfection efficiencies in individual experiments.

Soluble protein extracts were prepared in a buffer containing 1% Nonidet-P-40, 50 mM Tris pH 7.5, 5 mM ethylene glycol tetraacetic acid and 200 mM sodium chloride, with freshly added protease and phosphatase inhibitors: 1 mM sodium vanadate, 20 μM phenylarsine oxide, 1 μg/ml leupeptin, 0.5 μg/ml aprotinin, 100 μM phenylmethylsulphonyl fluoride and 1 mM DTT. After protein concentrations were measured with the Bio-Rad Assay, 50 μg of each lysate was resolved by 15% SDS-PAGE and then transferred to Hybond-P PVDF transfer membrane (catalogue RPN303F, Amersham, GE Healthcare, Baie d'Urfé, Québec, Canada). The membrane was blocked in 5% non-fat milk in tris-buffered saline with 0.05% Tween 20 and incubated with 1 μg/ml of primary antibody mouse anti-phospho-histone H2A.X (Ser139), clone JBW301 (Upstate, Millipore, Billerica, MA, USA) followed with horseradish peroxidase (HRP)-conjugated goat anti-mouse secondary antibody. The signal was developed by solutions prepared with 250 mM luminal solution, 90 mM p-coumaric solution, 1 M Tris pH 8.5 and 30% hydrogen peroxide.

HC11 cells transfected with either the empty vector HC11-EV or with the HSP90α expression construct (HC11-HSP90α) were plated at 130,000 cells/well of a 96-well plate and differentiated as above, then starved by the absence of serum and hormones. SKBR3 cells were plated at 10,000 cells/well. The presence of mono- and oligo-nucleosomes in the cytoplasm were qualitatively measured using the Cell Death Detection Elisa Plus Kit (Roche, Mississauga, ON, Canada). Essentially, protein extracts were incubated with anti-DNA (HRP-coupled) and anti-histone (biotin coupled) antibodies, before incubation in streptavidin-coated 96-well plates. Colorimetric detection was performed at the absorbance wavelength of 405 to 490 nm.

To investigate the role of prolactin in breast cancer, we set out to identify prolactin responsive genes in the breast cancer cell line, SKBR3. We first examined the prolactin-based activation of STAT5 and the induction kinetics of previously identified STAT5-dependent genes. SKBR3 cells were treated with different doses of prolactin and the activated DNA-binding form of STAT5 was visualised in electrophoretic mobility shift assays (Figure 1a). The experiment shows that prolactin is able to activate STAT5 in SKBR3 cells in a dose-dependent manner. A STAT5 specific antibody was used to confirm the specificity of the protein-DNA complex (Figure 1a). Prolactin stimulation of 60 minutes resulted in activation and binding of STAT5 to DNA response elements present in the β-casein promoter. Of note is the lack of STAT5 activation in the absence of prolactin stimulation. SKBR3 cells have previously been shown to express the prolactin gene 49, but perhaps the endogenous levels of prolactin are not sufficient to induce activation of the JAK2-STAT5 pathway.

We next followed the kinetics of cytokine-inducible-SH2-containing protein (CIS) mRNA induction as a function of prolactin treatment of SKBR3 cells. The CIS gene 50 is a known target of the JAK2-STAT5 pathway activated by interleukin-2 or erythropoietin in lymphoid cells and is important for feedback inhibition 5152. The amount of CIS mRNA increased within 60 minutes of treatment and was further enhanced in the presence of cycloheximide, an inhibitor of protein synthesis (Figure 1b). The level of CIS mRNA reached a maximum at four hours and remained high for at least 18 hours. This confirmed that SKBR3 cells are appropriate to study the early induction of prolactin-JAK2-STAT5-regulated genes and that a time point of 60 minutes would result in the production of early target genes.

To identify additional genes regulated by prolactin in SKBR3 cells, we prepared subtraction hybridisation libraries. Based on the above observations, SKBR3 cells were treated for 60 minutes with 5 μg/ml prolactin and the RNA was used for preparation of the subtractive hybridisation libraries. Cycloheximide was added to cell preparations, both untreated and treated with prolactin in order to avoid identification of secondary targets. We prepared subtraction libraries for the genes differentially expressed on prolactin treatment (forward) and for genes expressed in the absence of prolactin (reverse). The two libraries were used in the differential screen for prolactin-regulated genes. About 1200 gene fragments were screened from the forward library and 770 from the reverse library. Genes were screened using Southern blotting in batches of 100 genes using the forward and reverse libraries as probes. The genes with the most intense differential signal on each blot, as measured using a Biorad phosphoimager (Bio-Rad, Munich, Germany), were selected for sequencing. Seventy-two positive clones selected on the basis of high expression in the forward (Table 1) or reverse libraries (Table 2) were sequenced and identified. Some genes were observed more than once (Tables 1 and 2), also indicating high differential expression. Of the 51 genes represented, 19 genes with the highest expression levels were rescreened in dot-blots using both the unsubtracted probes (cDNA) and subtracted probes and the expression patterns of 16 of the 19 genes were confirmed (86%) (Tables 1 and 2).

Although there were no pre-existing large studies of prolactin-regulated genes in breast cancer cells, we compared our results with other prolactin-related studies. Of note, 11 of the genes we identified overlapped with those previously identified as downstream targets of the prolactin pathway. In one study, prolactin gene targets were identified by gene array using regenerated mammary glands from prolactin-receptor-/- and cyclin D-/- mammary epithelial cell transplants. Cyclin D1 was thought to represent a secondary target of prolactin, and therefore genes identified in this screen would represent effectors downstream of prolactin and upstream of cyclin D1 18. Although the transplants were nontransformed cells, unlike SKBR3 cells, we noted some similarities. Genes that were identical in this study and our report include the genes encoding the mouse ferritin heavy chain gene, HSPs 70, 71 and 84 (HSP90B). We also found overlap with prolactin-regulated genes identified in the rat Nb2-11c lymphoma cell line 53, including HSP70 and HSP86 (HSP90A). Genes found to be similar (either functionally related or different subunits of a complex) between these two studies and ours include Sec 23 18 and Sec 22 53 (similar to Sec 24 in our study), elongation factor 2 1853 (similar to elongation factor 1 alpha), lactate dehydrogenase 1 A chain 18 (similar to lactate dehydrogenase H chain), myosin heavy chain (similar to myosin regulatory light chain), T-complex protein 1 e and h subunits 53 (similar to T-complex b subunit). The degree of overlap with the prolactin-regulated genes of these two studies is comparable with the level previously described between other prolactin target gene studies 19. The diversity of the gene lists found in each study is thought to derive from the differences between experimental conditions, cell types and methods.

The gene HSP90A was identified five times in the initial group selected for sequencing; and based on its high representation, differential expression and its function in cancer cells, was used for further analysis.

We used northern blotting analysis to independently verify the induction of the HSP90A gene in SKBR3 cells by prolactin. We plated cells at two different confluences (about 40% and 60% confluent on the day of stimulation) as was performed for the preparation of the library. Both populations responded to 60 minutes of prolactin treatment with the increase in HSP90A mRNA, but the dose response of HSP90A induction in the two-cell population was distinguishable. Cells at the higher confluence exhibited a maximal response at 1.5 μg/ml, whereas cell at lower confluence required 5 μg/ml for maximal induction (Figure 2a). Higher concentrations of prolactin reduce the maximal response. The observed lower level of total RNA in high confluence cells treated with 10 μg/ml prolactin, and potentially the inhibitory presence of CIS, may explain the reduction in signal observed in higher confluence cells at 10 μg/ml. Another alternative is that the high concentration of prolactin induced a refractory state of prolactin signal transduction 54. These results confirm that the HSP90A gene is a prolactin-regulated gene in the human mammary carcinoma cell line, SKBR3.

We then investigated whether there was an increase in the encoded protein HSP90α in prolactin-treated SKBR3 cells. SKBR3 cells were cultured in the absence or presence of prolactin and protein extracts were resolved by SDS-PAGE. Western blotting using antibodies directed against HSP90α indicated that there is a two-fold increase in the amount of HSP90α protein in prolactin treated cells (Figure 2b). Prolactin therefore induces both the expression of HSP90A mRNA and increases HSP90α protein in breast cancer cells.

We also investigated the response of the HSP90A gene in mouse mammary epithelial HC11 cells. HC11 cells exist in an undifferentiated state until competent cells are stimulated appropriately with lactogenic hormones, dexamethasone, insulin and prolactin, to differentiate and produce milk proteins 55. We compared the levels of HSP90A mRNA in HC11 cells in their undifferentiated state and after a one-hour treatment of competent cells with lactogenic hormones, including prolactin (Figure 2c), simulating the time point used in SKBR3 cells. Lactogenic hormone induction resulted in a rapid four-fold increase of HSP90A mRNA in HC11 cells, as quantified by phosphoimager analysis of northern blots using actin as a loading control. We also observed up to a two-fold increase in HSP90α protein, peaking at 48 hours of lactogenic hormone induction of HC11 cells (Figure 2d). This demonstrated that the HSP90A mRNA and protein are elevated during early mammary epithelial cell differentiation in response to lactogenic hormones.

Inspection of the human HSP90A gene upstream sequence indicated the presence of at least two potential STAT-binding DNA elements that could bind STAT1, STAT3 or STAT5 56 (nucleotides 1611-1603 and 1177–1185, [GenBank:U25822]). Although STAT1 and STAT3 have been reported to respond to prolactin, the prolactin signal is mainly conferred through the activation of STAT5A and STAT5B, two highly homologous members of the STAT family 57. In order to investigate the prolactin responsiveness of the HSP90A gene promoter, we conducted reporter assays using a gene construct containing about 1.8 kb of the human HSP90A upstream regulatory sequence fused to a luciferase reporter gene. We transfected HeLa cells with expression vectors for the long form of the human prolactin-receptor, the HSP90A-luciferase reporter construct and various STATs.

Prolactin activation of cells transfected with STAT5B, or STAT5A and STAT5B together, caused over a four-fold or an over two-fold increase in luciferase activity, respectively, when compared with cells without exogenous STAT5 expression (Figure 3a). As the induction in the presence of exogenous STAT5A alone was not statistically significant, the significance of the STAT5A/5B result may be due to the presence of STAT5B. The smaller effect of the combination of STAT5A/5B over STAT5B alone may be due to the sequestering of STAT5B through the formation of heterodimers. We also performed reporter assays in both COS-7 as well as SKBR3 cells and obtained similar results with respect to the preferential transcription of the reporter by STAT5B rather than STAT5A (data not shown). In HeLa cells, a small effect of STAT1 and no effect of STAT3 on luciferase activity were observed (Figure 3b). STAT5B is the predominant form of STAT5 in breast tumour cell lines including SKBR3 58 and most likely contributes to the elevated expression of HSP90α in breast cancer cells. These reporter assays confirm that HSP90A is a prolactin-STAT5 regulated target gene.

Following the line of reasoning that prolactin acts as a survival factor in breast cancer cells 5960616263 and that STAT5 promotes survival in normal mammary epithelial cells 6465, we tested whether the addition of prolactin to starved, untransformed mammary epithelial HC11 cells would rescue differentiated cells from apoptosis. HC11 cells were induced to differentiate after they reached confluence by the addition of the lactogenic hormones prolactin, dexamethasone and insulin. Differentiated HC11 cells were then starved of serum and lactogenic hormones with individual hormones returned as indicated for 72 hours, followed by analysis of mono- and oligo-nucleosomes as an indicator of apoptosis (Figure 4). Serum and hormone withdrawal of HC11 cells is known to induce apoptosis 66. Each of the lactogenic hormones, including prolactin, greatly protects HC11 cells equally well from apoptosis when serum and other lactogenic hormones are removed.

We reasoned that HSP90α would have important functions downstream of prolactin not only in cancerous breast cells, but also in untransformed mammary epithelial cells, such as HC11. For this purpose, we created HC11 cell lines that either constitutively expressed the gene for human HSP90α (HC11-HSP90α) or carried the empty vector (HC11-EV). There is a two-fold increase in HSP90α in the HC11-HSP90α line. Given that we showed that prolactin is a survival factor in HC11 cells, we then investigated whether the HC11-HSP90α cells were susceptible to apoptosis induced by the removal of prolactin and other survival factors. We used two independent methods.

First, we used an antibody against phosphorylated-histone 2A.X as a marker of the apoptotic DNA damage 67 that occurs in response to starvation. As expected, there was little to no indication of phosphorylated-histone 2A.X in the undifferentiated, competent or differentiated cells, which are cultured in the presence of serum and hormones (Figure 5a). Differentiated cells were then starved of serum and all lactogenic hormones for up to 48 hours to induce apoptosis. The HC11-HSP90α cell lines were more sensitive to starvation than the parental HC11 cells or cells expressing the empty vector. After 24 hours of serum and hormone withdrawal, phospho-histone 2A.X is easily detected by western blot in the HC11-HSP90α cells, but not as easily in the control HC11 cells (Figure 5b) or cells carrying the empty vector (HC11-EV) (not shown as the response was similar).

We then investigated whether the amount of HSP90α differs between the differentiated and starved cells. When cells are starved for 24 hours, the amount of HSP90α in control cells decreases slightly, but remains more constant in the cells constitutively expressing HSP90A, HC11-HSP90α cells (Figure 5c). We can not be certain that the effect on survival we observed (shown in Figure 5b) is due to the differential protein levels between the differentiated and starved cells in the two cell lines, or to the overall two-fold elevated levels of HSP90α in the HC11-HSP90α line. The greater amounts of phospho-histone 2A.X at 24 hours of starvation indicate that constitutive expression of the gene encoding HSP90α increases levels of the marker for DNA damage, phospho-histone 2A.X, and may sensitise the cells to apoptosis.

To confirm that the phosphorylation of histone 2A.X represented events that occur during apoptosis, we also qualitatively assayed the mono- and oligo-nucleosomes generated due to apoptotic DNA nuclease activation. As both HC11-HSP90α cell lines behaved similarly, we used only line two for further characterisation. The HC11 cell lines were treated as shown in Figure 5b, and equal amounts of the cytoplasmic extracts were assessed for the presence of mono- and oligo-nucleosomes. HC11-HSP90α cells clearly had higher levels of nucleosomes than control cells HC11-EV (empty vector) after 24 and 48 hours of starvation, indicating greater levels of apoptosis (Figure 5d). These results support that although prolactin acts as a survival factor in the absence of prolactin or other survival factors, HC11 cells constitutively expressing HSP90α are sensitised to starvation-induced apoptosis.

We used the HSP90 inhibitor 17-AAG to confirm our results. First, we confirmed that HSP90α promotes survival in the presence of prolactin and serum. Differentiated HC11-HSP90α (Figure 6a) or HC11-EV cells (Figure 6b) were untreated or treated with 1 μg/mL 17-AAG either in differentiation medium (dexamethasone, insulin and prolactin plus serum) or starvation medium (no serum or hormones) for a total of 24 hours before measuring apoptosis. Consistent with the role of HSP90α promoting survival, inhibition of HSP90 by 17-AAG induced apoptosis in differentiated HC11-HSP90α (Figure 6a) and HC11-EV cells (Figure 6b) in the presence of survival factors such as prolactin and serum. HSP90α also promotes survival in the control HC11-EV cell line in starvation medium, as demonstrated by the increase in apoptosis with the addition of the inhibitor 17-AAG (Figure 6b).

Second, we confirmed that constitutive expression of HSP90α promotes apoptosis in starved cells. Consistent with the effect of constitutive HSP90α expression on the sensitisation of HC11 cells, 17-AAG reduced the starvation-induced apoptosis (Figure 6a). Overall, starvation enhanced the general level of apoptosis in differentiated cells of both cell lines, although the effects of the HSP90 inhibitor were different. Together this confirms our initial observations and indicates that HSP90α functions to promote survival in differentiated cells in the presence of survival factors, but that constitutive expression sensitises these immortal mammary epithelial HC11 cells to starvation-induced apoptosis.

Increased expression of HSP90α has been reported in breast cancer, including SKBR3 cells 35, and cytotoxicity has also been reported for the use of HSP90α inhibitors in breast cancer cells, including SKBR3 686970. To test the role of HSP90α in SKBR3 cells, we assessed oligo-nucleasome formation in the presence and absence of 17-AAG and in the presence or absence of serum. HSP90α promotes survival, in the presence of serum, as indicated by the increase in apoptosis after treatment with 17-AAG (Figure 7). Interestingly, overall there was less apoptosis in the absence of serum and the amount of apoptosis was independent of 17-AAG. In general, cancer cells are known to be resistant to apoptotic stimuli, but the role of HSP90α in the absence of serum seems to be minimal. HSP90α promotes survival of SKBR3 cells in the presence of serum.

We identified a number of genes in our screen of breast cancer SKBR3 cells whose expression either increased or decreased in response to prolactin. Some of the genes have been previously associated with cancer progression (T complex protein-1 beta 71, tetratricopeptide repeat protein-4 72), cancer survival (Bax inhibitor 1 73, mitofusion 2 7475), heterogeneous nuclear riboprotein A1 76), drug resistance (T complex protein-1, HSP70 77), or cancer cell migration (Hs1 binding protein 7879). Prolactin signalling has been implicated in each of these phenomena in breast cancer cells though its role in drug resistance has not yet been thoroughly examined.

In this study, we also identified HSP90A as a prolactin-induced STAT5-activated target gene. HSP90α is a molecular chaperone of a large number of proteins involved in critical signal transduction pathways. The role of HSP90α downstream of prolactin helps explain the multiple effects of prolactin in normal cells and emphasises the significant contribution of prolactin-JAK2-STAT5 signal transduction to breast cancer.

HSP90 in cancer cells is present in an active form, in a multi-chaperone complex with high ATPase activity, in contrast to the HSP90 in normal cells, which is in an inactive, uncomplexed form. It is thought that these differences account for the high affinity of cancer-associated HSP90α for the inhibitory ATP mimetic drugs such as 17-AAG 38. Chemotherapeutic drugs, such as 17-AAG, inhibit HSP90 and usually result in the degradation of HSP90 client proteins 80. There are multiple client proteins of HSP90α, including steroid hormone receptors such as oestrogen receptor, protein kinases, cell cycle proteins and transcription factors that are essential targets in cancerous cell growth, survival, immortalisation, angiogenesis and metastasis 81. The prolactin-mediated induction of HSP90α implicates prolactin in the acquisition or maintenance of each of those cancer-related traits.

We also determined that prolactin, dexamethasone and insulin, each act as survival factors in differentiated mammary epithelial HC11 cells. Insulin 6682 and the glucocorticoid receptor have previously been identified as survival factors 8384. There is existing evidence showing a survival role for prolactin in breast cancer cells 5960616263, and for STAT5 in normal mammary epithelial cells 6465. The survival function of prolactin is partially due to the prolactin-mediated activation of AKT/protein kinase B 8586. AKT/protein kinase B is a survival factor and a regulator of mammary gland involution, as transgenic mice expressing constitutively active AKT/protein kinase B in the mammary gland showed delayed involution and a delayed onset of apoptosis 8788. AKT/protein kinase B is also a client protein of HSP90 8990.

The prolactin-JAK2-STAT5 target gene, HSP90α, can contribute to survival in the presence of prolactin, but sensitises untransformed mammary epithelial HC11 cells to starvation-induced apoptosis when constitutively expressed. We confirmed these results with the use of 17-AAG.

There is evidence for the pro-apoptotic function of HSP90α in other cell types 919293. We can also hypothesise that the pro-apoptotic function is due to the action of one of the client proteins either stabilised under these conditions, such as mutant p53, or disengagement from one of its client proteins such as AKT/protein kinase B. Although it is known that HSP90 stabilises mutant p53 forms 94, many of these mutant forms contribute to cellular immortalisation and transformation. The mutant forms of p53 in HC11 cells are thought to contribute to their immortalisation 95, but it is not known if, under certain conditions, mutant p53 could contribute to the sensitisation of cells to starvation-induced apoptosis as does wild-type p53 96. The fact that HSP90 can stabilise mutant p53 forms that can contribute to immortalisation and transformation emphasises a contribution of prolactin to these functions, as one of its upstream inducers.

We propose that the switch from survival to apoptosis in untransformed cells involves the loss of survival factors and the availability of HSP90α. In contrast, the switch is absent in SKBR3 breast cancer cells, which do not respond to 17-AAG during serum starvation. HSP90 is important for proliferation and survival, as SKBR3 cells have been shown to respond to 17-AAG by a reduction in proliferation 6870 and an increase in apoptosis 6869. This latter observation is consistent with our results in this report. Possible mechanisms involved in loss of HSP90-mediated survival after 17-AAG treatment include the loss of AKT/protein kinase B 97 or ERBB2 699899.

Together with our results, this indicates that in addition to a role in survival, HSP90α also has a pro-apoptotic role that may be cell-type specific, specific to the hormone milieu in the environment or specific to the cellular state of transformation and complement of tumour-suppressor proteins. We hypothesise that HSP90α together with prolactin-mediated events support survival in differentiated or cancerous cells, whereas HSP90α alone may sensitise differentiated mammary cells to wild-type p53-independent apoptosis depending on the cellular context.