Tamoxifen is the most commonly used drug to treat breast cancer and acts by blocking ERα (oestrogen receptor α) signalling. Although highly effective, its usefulness is limited by the development of resistance. Given this, strategies that limit resistance by sensitizing cells to tamoxifen may be of use in the clinic. To gain insight into how this might be achieved, we used chemical and genetic screens to identify targets and small-molecule inhibitors that cause tamoxifen sensitization. A high-throughput genetic screen, using an RNA interference library targeting 779 kinases and related proteins, identified the PDK1 (phosphoinositide-dependent kinase 1) signalling pathway as a strong determinant of sensitivity to multiple ERα antagonists, including tamoxifen. A chemical screen using existing drugs and known kinase inhibitors also identified inhibitors of the PDK1 pathway, including triciribine and tetrandrine. Aside from identifying novel agents and targets for tamoxifen sensitization, this approach also provides evidence that performing chemical and genetic screens in parallel may be useful.
- chemical screen
- phosphoinositide-dependent kinase 1 (PDK1)
- RNA interference screen
Approx. 70% of breast tumours express ERα (oestrogen receptor α) and, of these, most are fully dependent on oestrogen signalling for their growth and can therefore be treated with anti-oestrogen therapies. Tamoxifen, the most commonly used anti-oestrogen therapy, has been shown to be of significant benefit in the treatment of ERα-positive breast cancer  and works by blocking ERα signalling in the breast. This ERα antagonism leads to the inhibition of cell-cycle progression and arrest at the G1 checkpoint [2,3]. Despite its effectiveness, the usefulness of tamoxifen is limited by the development of drug resistance, which occurs in all patients with metastatic disease and in approx. 40% of those with early-onset breast cancer who are treated with this agent as an adjuvant to surgery [4,5].
Considerable efforts have been made to identify the determinants of tamoxifen resistance and sensitivity. Primarily, this has been performed on a gene-by-gene basis and although this has been informative, a complementary approach is to perform high-throughput simultaneous analysis of many genes and pathways. This can be achieved genetically by the use of approaches such as RNAi (RNA interference) screening , which relies on the ability of siRNA (short interfering RNA) to silence gene function in a sequence-specific fashion. We have used such a high-throughput RNAi screen to identify CDK (cyclin-dependent kinase) 10 as a novel determinant of tamoxifen resistance . In addition to genetic screens, a separate approach is to use libraries of chemical inhibitors to interrogate cellular phenotypes, such as tamoxifen response. Given that chemical inhibition of proteins may sometimes produce differing phenotypes to those achieved by gene silencing or knockout , the use of chemical screening has the potential to identify effects not achievable by genetic screening. Moreover, if the chemical libraries used include drugs in clinical use and with known pharmacokinetic and safety profiles, these screens offer the possibility of identifying effects that can be rapidly evaluated in Phase II clinical trials, eliminating much of the toxicological and pharmacokinetic assessments required for novel compounds . Additionally, the targets of existing drugs are often established, potentially allowing the mechanism of action to be determined more easily.
Given the relative benefits of genetic and chemical screens, in the present study we reasoned that parallel RNAi and inhibitor screens might be a useful approach to identify the targets and inhibitors that cause sensitization to tamoxifen. This proved to be the case, and a number of novel targets were identified. Notably, multiple components of the PDK1 (phosphoinositide-dependent kinase 1) pathway were identified by the RNAi screen. Results from the chemical screen validated the results of the RNAi screen, therefore providing proof of the principle that genetic and chemical screens may indeed have complementary roles in target discovery.
Cell lines, compounds and siRNA
MCF7 cells were obtained from the A.T.C.C. and were maintained in Phenol-Red-free RPMI 1640 medium (Invitrogen) supplemented with 10% (v/v) dextran-charcoal-treated FCS (fetal calf serum), 1 nM oestradiol, insulin, glutamine and antibiotics. Oestradiol and 4-OH-tamoxifen were obtained from Sigma, ICI 182780 was obtained from Tocris Bioscience, triciribine was obtained from BIOMOL International and Akti-1,2 was obtained from Calbiochem. MCF7 cells were transfected with SMARTpool siRNAs using Dharmafect 3 transfection reagent according to the manufacturer's instructions (Dharmacon). The kinase siRNA library (siARRAY: targeting 779 known and putative human kinase genes, including protein kinases, lipid kinases, adaptors and related genes) was obtained in ten 96-well plates from Dharmacon. Each well in the library contained a SMARTpool of four distinct siRNA species targeting different sequences of the target transcript. The compound library encompassed the Prestwick Chemical Library® (Prestwick Chemicals) and the Screen-Well™ library (BIOMOL International).
Antibodies against the following epitopes were used: p21Cip1 (556430; BD Biosciences), phospho-Akt Ser473 (4058; Cell Signaling Technology), Akt (9272; Cell Signaling Technology) and β-tubulin (T4026; Sigma). All secondary antibodies used for Western blot analysis were HRP (horseradish peroxidase) conjugated.
siRNA screening method
The RNAi screening method used has been described previously . Briefly, MCF7 cells plated in 96-well plates were transfected after 24 h with siRNA (final concentration of 100 nM) using Dharmafect 3 (Dharmacon) following the manufacturer's instructions. After transfection for 24 h, cells were trypsinized and divided into four identical replica plates. After 48 h transfection, two replica plates were treated with 50 nM 4-OH-tamoxifen in medium and the two other replica plates were treated with 0.05% ethanol vehicle in medium. Medium containing 4-OH-tamoxifen or vehicle was replenished after 48 and 96 h incubation, and cell viability was assessed after 7 days of 4-OH-tamoxifen exposure using the CellTiter-Glo Luminescent Cell Viability Assay (Promega) following the manufacturer's instructions. The luminescence reading for each well on a plate was expressed relative to the median luminescence value of all of the wells on the plate. The screen was performed in duplicate.
For each transfection, the effect on cell growth and tamoxifen sensitivity were calculated. The effect of each individual siRNA SMARTpool on cell growth alone was calculated by dividing the mean luminescence in the two replica wells treated with ethanol vehicle by the mean luminescence of the replica wells transfected with siControl (control siRNA), and is expressed as a percentage as follows:
The sensitivity to 4-OH-tamoxifen for each siRNA SMARTpool was assessed by calculating the surviving fraction following 4-OH-tamoxifen treatment as follows:
The surviving fractions were centred on the median surviving fraction of all 80 SMARTpools from one 96-well plate transfection, and the results from all ten siRNA plates were combined and the results expressed as a Z-score . For the Z-score, the S.D. of the screen was estimated from the S.D. of all siControl wells.
Validation of gene silencing by siRNA
Validation of RNAi gene silencing was measured by RT-qPCR (real-time quantitative PCR). MCF7 cells were transfected with siRNA, and RNA was extracted 48 h later with TRIzol® and phenol/chloroform extraction, followed by propan-2-ol precipitation. cDNA was synthesized using the SuperScript III First Strand Synthesis System for RT-PCR (reverse transcription-PCR) (Invitrogen) with oligo dT following the manufacturer's instructions. Assay-on-Demand primer/probe sets were purchased from Applied Biosystems, details of which are available on request. RT-qPCR was performed on the 7900HT Fast Real-Time PCR System (Applied Biosystems), with GAPDH (glyceraldehyde-3-phosphate dehydrogenase) used as an endogenous control. Standard curves were calculated for all reactions with serial dilutions of siControl-transfected cells to calculate reaction efficiency. Gene expression was calculated relative to the expression of the endogenous control gene (GAPDH) and adjusted relative to the expression in siControl-transfected cells.
Small molecule screening
MCF7 cells were plated in 96-well plates. After 24 h, cells were treated with 10 μM BIOMOL and Prestwick-compound libraries, with one-half of the plates treated in combination with 50 nM 4-OH-tamoxifen in medium and the other half of the plates treated in combination with 0.05% ethanol vehicle in medium. Medium containing compounds, 4-OH-tamoxifen or vehicle, was replenished after 48 and 96 h, and cell viability was assessed after 7 days of 4-OH-tamoxifen exposure using the CellTiter-Glo Luminescent Cell Viability Assay (Promega) following the manufacturer's instructions. The luminescence reading for each well of a plate was expressed relative to the median luminescence value for all wells on the plate. The screen was performed in duplicate. For each replicate, the effect on cell growth and tamoxifen sensitivity were calculated. The effect of each individual compound on cell growth alone was calculated by dividing the luminescence in the well treated with the ethanol vehicle by the luminescence of the replica wells treated with DMSO control, and expressed as a percentage as follows:
The sensitivity to 4-OH-tamoxifen for each compound was assessed by calculating the surviving fraction following 4-OH-tamoxifen treatment as follows:
The surviving fractions were centred on the median surviving fraction of all 80 compounds from each plate and results are expressed as a Z-score . For the Z-score, the S.D. of the screen was estimated from the S.D. of all DMSO control wells.
Cell viability assays to measure drug sensitivity
MCF7 cells were transfected with siRNA using Dharmafect 3 following the manufacturer's instructions, divided 24 h after transfection into 96-well plates and exposed to various doses of drugs from 48 h post-transfection. Cell viability was assessed using the CellTiter-Glo Luminescent Cell Viability Assay (Promega) at 9 days post-transfection, and the surviving fraction for each dose of drug was assessed by dividing the luminescence value of the drug-treated wells by the luminescence value of the vehicle-treated wells.
Median effect/CI (combination index) analysis  was used to determine the synergy of a combination of exposures to inhibitors. Cell cultures were treated with each agent individually at its IC50 concentration and at fixed multiples (two and four times) and fractions (one-half, one-quarter and one-eighth) of the IC50 concentrations. The inhibitors were also combined in these same dose-fixed ratios to determine the CI. Synergy was defined as any CI value below 1. Experiments were performed in triplicate, and each experiment yielded CI values for each survival fraction. Experiments that yielded an average CI value of less than 1 were repeated at least three times to allow for determination of confidence intervals for the CI values obtained.
MCF7 cells were transfected with siRNA using Dharmafect 3 following the manufacturer's instructions. After 24 h transfection, the medium was refreshed with no oestradiol, 1 nM oestradiol, or 1 nM oestradiol with 100 nM 4-OH-tamoxifen. Protein lysates were prepared after 48 h using RIPA lysis buffer [50 nM Tris/HCl (pH 8.0), 150 mM NaCl, 0.1% SDS, 0.1% DOC (sodium deoxycholate), 1% Triton X-100, 50 mM NaF, 1 mM sodium vanadate and protease inhibitors]. Total cell lysate (100 μg) was loaded on to pre-fabricated Bis/Tris gels (4–12% gels) (Invitrogen) with a full-range rainbow molecular-mass marker (GE Healthcare) as a size reference, and resolved by electrophoresis. Proteins were transferred on to nitrocellulose membranes (Bio-Rad), blocked and probed with primary antibodies (1:1000 dilution) in 1× TBS-T (Tris-buffered saline with 0.1% Tween 20) with 5% (w/v) BSA overnight at 4 °C. Incubation with secondary antibodies (1:5000 dilution) in 1× TBS-T with 5% (w/v) non-fat dried skimmed milk powder was performed for 1 h at room temperature (20 °C). Protein bands were visualized using ECL (enhanced chemiluminescence) (GE Healthcare) and MR or XAR film (Kodak).
MCF7 cells were transfected with siRNA using Dharmafect 3 following the manufacturer's instructions. After 24 h, the medium was refreshed with 1 nM oestradiol, or 1 nM oestradiol with 100 nM 4-OH-tamoxifen. After 48 h, the cells were fixed in 70% (v/v) ice-cold ethanol and stained with 4% (w/v) propidium iodide and 10% (w/v) RNase A in PBS. The sample readout was performed using the FACSCalibur system (Becton Dickinson) and the results were analysed using CellQuest Pro (Becton Dickinson).
We reported recently the results of a high-throughput RNAi screen to identify the determinants of tamoxifen sensitivity . In the present study, we performed a screen which targeted 779 kinases and related proteins by RNAi and measured changes in the cellular response to tamoxifen, as assessed by cell viability. We elected to screen this protein family as kinases represent targets that are amenable to inhibitor development. Furthermore, we reasoned that this approach would increase the likelihood that agents developed already could be used to inhibit targets identified in our screen. In brief, the ERα-positive, tamoxifen-sensitive, human breast cancer cell line, MCF7, was transfected with siRNA in a 96-well-plate format and then divided 24 h after transfection into four replica plates. After transfection for 48 h, two replica plates were treated with 4-OH-tamoxifen (the active metabolite of tamoxifen) and two plates were treated with vehicle (Figure 1A). The screen was repeated in duplicate and a comparison of results from each screen revealed this approach to be highly reproducible. Results from the duplicate screens were combined in the final analysis (Figure 1B).
In our initial report, we focused upon RNAi effects that caused resistance to tamoxifen and established that CDK10 silencing could cause resistance, an effect probably explained by an up-regulation of MAPK (mitogen-activated protein kinase) signalling . In the present study, we used the same screen results to identify tamoxifen-sensitization effects. The most sensitizing siRNA SMARTpools are listed in Table 1. As a complementary approach to RNAi identification of sensitization effects, we also screened a library of 1200 drugs and kinase inhibitors to identify tamoxifen sensitizers. This chemical screen involved treating MCF7 cells with either 4-OH-tamoxifen or vehicle (ethanol) in combination with the compounds for 7 days before cell viability was assessed. The screen was repeated in duplicate and a comparison of results from each screen revealed this approach to be highly reproducible. Results from the duplicate screens were combined in the final analysis (Figure 1C). Of the compounds which caused sensitization to tamoxifen, the ten most significant compounds are listed in Table 2. As a mark of the validity of the chemical screen, tamoxifen itself was identified as causing a sensitizing effect.
By cross-referencing the results of the RNAi and compound screens, we identified the PDK1 intracellular signalling pathway  as an important target for tamoxifen sensitization. Seven of the twenty most sensitizing hits identified from the RNAi screen were components of the PDK1 signalling pathway, as depicted in Figure 2(A). In addition, all of the top ten most sensitizing hits identified from the compound screen, with the exception of betulin, significantly inhibited the PDK1 signalling pathway, as demonstrated by inhibition of phosphorylation of the well-characterized PDK1 substrate Akt (Figure 2B). The most potent inhibitor of PDK1 signalling identified from the compound screen, triciribine, was studied further, along with PDK1 siRNA. We confirmed that PDK1 siRNA, as well as triciribine, inhibited Akt activation, as measured by phosphorylation (Figures 2C and 2D). Furthermore, we confirmed that targeting the PDK1 pathway could cause sensitization to tamoxifen. This was achieved by performing cell survival assays with differing combinations of tamoxifen and an Akt inhibitor (Akti-1,2) that is structurally unrelated to triciribine. To assess true synergy, the survival results were analysed using the median effect model  that estimates the CIs of two compounds. CIs of less than 1 are indicative of synergy and CIs greater than one indicate antagonistic activity . Tamoxifen and Akti-1,2 combinations resulted in a mean CI of 0.41 (Figure 2E), indicating that inhibition of the PDK1 signalling pathway was synergistic with tamoxifen.
In the high-throughput RNAi screen, we used pools of siRNA (SMARTpools) each consisting of four individual siRNAs targeting one gene. To assess the possibility that the tamoxifen-sensitization effect caused by PDK1 silencing in the high-throughput screen was the result of off-target effects , we confirmed the gene silencing of PDK1 by each individual siRNA contained within the individual SMARTpool (Figure 3A). We also demonstrated that each of these siRNAs were able to elicit tamoxifen sensitization. In contrast with the single dose of tamoxifen used in the parallel screens, we assessed the effect of individual PDK1 siRNAs over a range of tamoxifen doses (Figure 3B). This analysis indicated that the PDK1 siRNA effect observed in the screen was unlikely to be the result of an off-target siRNA effect. We also validated the effect of triciribine over a range of tamoxifen doses (Figure 3C). This demonstrated that triciribine increased the sensitivity to tamoxifen 7-fold (DMSO SF50=8.5 nM and triciribine SF50=1.2 nM) (where SF50 is the surviving fraction 50, the dose at which only 50% of cells survive). Finally, we confirmed tamoxifen–triciribine synergy by use of the median effect model . Triciribine and tamoxifen gave a mean CI of 0.39, indicative of synergy.
Oestrogen signalling can be targeted by pharmacological agents other than tamoxifen and we assessed the possibility that the tamoxifen-sensitization effects observed with triciribine and PDK1 silencing were also applicable to other forms of endocrine therapy. All four siRNAs targeting PDK1 sensitized cells to ICI 182780 (fulvestrant), a drug that induces ERα degradation  (Figure 4A). Furthermore, combinations of triciribine and ICI 182780 also showed clear synergy (Figure 4B). The action of aromatase inhibitors may be modelled by oestrogen deprivation  and either PDK1 siRNA or triciribine-sensitized cells to oestrogen deprivation (Figures 4C and 4D). Taken together with the results observed using tamoxifen and ICI 182780, this suggested that suppression of the PDK1 pathway sensitizes cells to inhibition of ERα signalling generally, rather than to the effects of tamoxifen specifically.
To understand the role of PDK1 in the response to tamoxifen, we examined whether the induction of cell-cycle arrest at the G1 checkpoint caused by tamoxifen  could be enhanced by PDK1 silencing. In the absence of tamoxifen, transfection of PDK1 siRNA did not grossly alter the cell-cycle profile of cells compared with cells transfected with a siControl [compare the percentage of cells in each cell-cycle phase after treatment with siRNA PDK1(G0/G1=54.4%, S=6.0%, G2/M=14.8%) and siControl (G0/G1=58.5%, S=4.2, G2/M=15.7%); Figure 4E). Upon the addition of tamoxifen, the G0/G1 proportion of cells was modesty increased in siControl-transfected cells (from 58.5% to 63.7%), consistent with arrest at the G1 checkpoint (Figures 4E and 4F). In PDK1 siRNA-transfected cells, this tamoxifen-induced increase in G0/G1 cells was more profound, rising from 54.4% before tamoxifen treatment to 84.7% after treatment, suggesting that PDK1 silencing exacerbated tamoxifen-induced G1 arrest. This increase in G1 arrest was also mirrored by a decrease in cells in the G2/M-phase of the cell cycle (Figures 4E and 4F). Tamoxifen is thought to induce cell cycle arrest by causing the up-regulation of CDK inhibitor proteins, including p21Cip1 [2,18]. p21Cip1 protein levels are also known to be down-regulated in an Akt phosphorylation-dependent mechanism [19,20]. Consistent with these observations, we found that p21Cip1 expression was significantly enhanced after tamoxifen treatment in PDK1-silenced cells (Figure 4G), suggesting a mechanism by which the down-regulation of PDK1 could enhance the G1 arrest caused by tamoxifen.
In addition to triciribine, the chemical screen also identified sertraline, tetrandrine and methotrimeprazine maleate as non-toxic putative tamoxifen sensitizers (toxicity was defined as cell viability of less than 0.5). To confirm the observations of the highthroughput screen and to establish the magnitude of the effect, dose–response curves were performed over a range of 4-OH-tamoxifen concentrations. All of the compounds significantly increased the sensitivity of cells to tamoxifen (see Supplementary Figure S1 at http://www.BiochemJ.org/bj/417/bj4170361add.htm), confirming the results from the tamoxifen-sensitization drug screen. The most potent tamoxifen sensitizer was tetrandrine.
Tetrandrine [(1β)-6,6′,7,12-teramethoxy-2,2′-dimethyl-berbaman] is derived from the root of Stephenia tetrandra and has been used for treatment of cardiovascular diseases in traditional Chinese medicine for centuries . Tetrandrine has also been used in both animal and human studies, showing little toxicity alongside both anti-inflammatory and anti-neoplastic activity [21,22]. To confirm the observations of the high-throughput screen, we performed dose–response curves and demonstrated that tetrandrine did indeed sensitize cells to tamoxifen, causing a 9.9-fold increase in sensitivity (DMSO SF50=14.8 nM, 10 μM tetrandrine SF50=1.46 nM; Figure 5A). As for triciribine, we used the median effect model to assess tamoxifen–tetradrine synergy, which, although moderate, was confirmed (mean CI of 0.72; Figure 5B). It has been demonstrated recently that tetrandrine is able to induce G1 arrest by inhibiting Akt  and we also observed moderate Akt inhibition by tetrandrine (Figure 2B). Given this, and in light of our results from triciribine treatment and PDK1 silencing, we assessed whether tetrandrine was able to exacerbate tamoxifen-induced G1 arrest. Examination of cell-cycle profiles supported this hypothesis (Figures 5C and 5D). In cells treated with the tetrandrine vehicle DMSO, tamoxifen induced a modest G1 arrest (proportion of cells in G1-phase rises from 32.9% to 37.1%). Conversely, tamoxifen plus tetrandrine produced a more pronounced G1 arrest (proportion of cells in G1-phase rises from 38.4% to 54.6%).
In the present study, we have performed chemical and genetic screens which reveal the potential for targeting the PDK1 pathway as a means to sensitize cells to tamoxifen. PDK1 is activated by PI3K (phosphoinositide 3-kinase) and serves as a master regulator of the AGC (protein kinase A/protein kinase G/protein kinase C) family of protein kinases . Interestingly, PDK1 is highly expressed in many human cancer cell lines  and breast tumours , suggesting a role for PDK1 in breast cancer tumorigenesis; however, few studies have evaluated PDK1 as a potential target for cancer therapy. Significantly, both of the present compound and RNAi screens identified components of the PDK1 pathway as targets for tamoxifen sensitization.
Consistent with this, previous studies have shown that inhibition of the PDK1 activator, PI3K, can cause sensitization to endocrine therapies in both in vitro and in vivo models and that activation of the PDK1 substrate, Akt, causes resistance to tamoxifen in breast cancer models and tumours [26,27]. However, the PI3K inhibitors used, such as wortmannin or LY294002, have poor clinical potential and are unlikely to be developed for clinical use . Wortmannin has a short half-life  and LY294002 causes dermatitis in animal models . Furthermore, both act non-specifically on other PI3K family members, resulting in considerable toxicity . For these reasons, neither compound has been used in clinical trials, despite their anti-tumour activities both in vitro and in vivo . The identification of triciribine from the compound screen is therefore significant, as triciribine has shown promise in in vitro models as a potential anticancer drug previously  and preliminary clinical trials demonstrated that low concentrations of triciribine were non-toxic [34,35], suggesting that combining low concentrations of triciribine with tamoxifen, fulvestrant or aromatase inhibitors in the clinic may be a relatively non-toxic approach for delaying the development of drug resistance. In addition to triciribine, tetrandrine was also identified as a potent sensitizer to tamoxifen. Tetrandrine has been used in animal and human studies with little toxicity and has shown both anti-inflammatory and anti-neoplastic activity [21,22]. Previous studies have also identified tetrandrine as a sensitizer to a range of cancer therapies, including doxorubicin , vincristine and ionizing radiation , although the mechanisms underlying these effects remain unknown. Importantly, this is the first study to identify tetrandrine as a sensitizer to tamoxifen.
The present study verifies the importance of PI3K signalling in the modulation of tamoxifen sensitivity and is the first to identify PDK1 as a target for tamoxifen sensitization. We discovered that silencing of PDK1 increases sensitivity to multiple endocrine therapies, including tamoxifen, fulvestrant and oestrogen deprivation. The observation that inhibition of PDK1 causes sensitization to a range of oestrogen-signalling inhibitors suggests that the combination of PDK1 inhibitors with endocrine therapy merits investigation in the clinic. Although not currently in clinical use, PDK1-specific inhibitors have been developed recently and may be suitable for this purpose in the future .
In the present study, we provide an example of how parallel RNAi and chemical screens performed in mammalian cells may complement each other . Given the complexity of biochemical pathways and the number of protein–protein interactions now described for each of the proteins within a cell, deconvoluting the mechanism by which one RNAi screen hit controls a phenotype can be difficult. To this end, we have demonstrated that, by the use of a parallel chemical screen, it is possible to simultaneously validate results from an RNAi screen and suggest potential mechanisms by which RNAi hits determine a given cellular phenotype. It is also possible that RNAi screens could be used to partially deconvolute hits from parallel compound screens . Although in vitro small molecule screens are generally used to identify compounds that inhibit a validated protein target, the inhibitors identified may perform poorly in living cells. One solution is to perform small-molecule inhibitor screens in cells, but this approach is also limited, as it requires the cellular targets of inhibitors to be identified, which can be challenging. By performing parallel RNAi and compound screens, small molecules and gene-specific siRNA reagents can be identified that cause similar cellular phenotypes, thus simplifying target identification. This proof-of-principle method suggests that a combination RNAi/chemical approach in human cell lines could streamline the development of small molecules into drugs by improving compound target deconvolution .
This work was supported by Breakthrough Breast Cancer [grant number BBC 024]. E. I. was the recipient of a studentship from the Tertiary Education Commission of New Zealand.
Abbreviations: CDK, cyclin-dependent kinase; CI, combination index; ERα, oestrogen receptor α; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; PDK1, phosphoinositide-dependent kinase 1; PI3K, phosphoinositide 3-kinase; RNAi, RNA interference; RT-qPCR, real-time quantitative PCR; siRNA, short interfering RNA; SF50 etc., surviving fraction 50 etc.; siControl, control siRNA; TBS-T, Tris-buffered saline with 0.1% Tween 20
- © The Authors Journal compilation © 2009 Biochemical Society