In order to examine p68/p72 co-activator function we transfected COS-1 cells with an estrogen-responsive firefly luciferase reporter plasmid and combinations of full-length ERα or mutants lacking either AF1 or AF2 (Figure 1a), together with p68, p72 and the well-characterised steroid receptor co-activator SRC-1. In agreement with previous reports (Endoh et al., 1999; Watanabe et al., 2001), ERα activity was stimulated by co-transfection with p68 and p72 (Figure 1b); this co-activation was synergistically enhanced by SRC-1. As has also previously been described (Endoh et al., 1999), the RNA helicase activity of p68 is dispensable since co-transfection with a helicase-inactive p68 mutant did not prevent stimulation of ERα activity by p68 (Figure 1c), nor was co-activation of ERα blocked by a similar mutation in p72.

p68 was originally identified as an ERα co-activator following in vitro purification of proteins that interacted preferentially with ERα phosphorylated at ser-118 by MAPK (Endoh et al., 1999). In reporter assays using truncated ERα lacking the LBD/AF2, p68 and p72 stimulated AF1 activity (Fig. 1d). Whilst mutation of Ser-104 or Ser-106 did not prevent co-activation by p68 or by p72, mutation of Ser-118 gave little or no stimulation of ERα activity by p68 and p72, with mutation of all three residues completely preventing co-activation of AF1 by p68 and p72 altogether. Interestingly, however, both p68 and p72 stimulated the activity of an ERα deletion mutant lacking AF1 (Fig. 1e), suggesting that co-activation of ERα by p68/p72 involves the ERα DBD and/or AF2, in addition to AF1. Again this co-activation was synergistically enhanced by SRC-1.

Using GST-pulldowns, an in-vitro interaction between ERα and p68 was observed, the interaction being estrogen-independent (Fig. 2a). Interestingly, deletion mutants lacking AF1 or the AF2 interacted with p68, whilst only weak interactions were observed for AF1 or AF2, indicative of interaction of p68 and p72 with the ERα DBD. Similar results were obtained for p72, although interaction with the mutant lacking AF1 was weaker than for p68, indicating that the interaction of p68 and p72 with ERα is similar, at least in-vitro. The lack of interaction between ERα-AF1 and p68/p72 could be due to a requirement for AF1 phosphorylation. However, interaction of AF1-DBD (ERα-ΔAF2) and the AF1 region with p68 and p72 was unaffected by mutation of S118 or of S104/S106/S118 (Fig. 2b). Nor was interaction between these ERα mutants and p68 or p72 stimulated by in-vitro phosphorylation with ERK2 MAPK (Fig. 2c). Taken together, these findings show that regions in addition to the AF1 are important for the interaction between ERα and p68/p72 and indicate that the ERα DBD is likely to be important for interaction of ERα with p68 and p72.

Co-immunoprecipitation of nuclear extracts prepared from MCF-7 cells treated with 10nM estrogen for 6 hours showed that p68 and p72 co-immunoprecipitated with ERα in an estrogen-independent manner (Figure 2d). Furthermore, we found that while ERα co-immunoprecipitated both p68 and p72, p68 co-immunoprecipitated ERα efficiently, but only co-immunoprecipitated a very small amount of p72, implying that ERα-p68 and ERα-p72 complexes may be distinct within the cell.

To determine whether p68 and p72 are required for expression of endogenous estrogen-responsive genes, we knocked down expression of these proteins by transfection of specific siRNA molecules, which have previously been described (Bates et al., 2005). Knockdown of p72 resulted in significantly reduced pS2 expression at both the mRNA and protein levels (Figures 3a, e). For Cathepsin D (CTD), knockdown of p72 led to a significant decrease in protein expression although there was only a small, but reproducible, decrease in mRNA levels (Figures 3b, e). Surprisingly, p68 knockdown did not affect pS2 or CTD expression (Figures 3a, b, e). Knockdown of p68 and p72 was confirmed by quantitative RT-PCR and western blotting (Figures 3c, d). Interestingly, we still observed estrogen stimulation of pS2 and CTD expression following p72 knockdown, although the overall level is markedly reduced, suggesting that p68 or other ERα co-regulators can compensate for p72, at least in part.

Since p72 appears to be important for stimulation of estrogen-regulated gene expression we determined the effect of p68 or p72 knockdown on estrogen-stimulated growth of MCF-7 cells. Whilst knockdown of p68 had no significant effect on cell growth compared to a non-silencing siRNA (Figure 4a), knockdown of p72 (Figure 4b) significantly reduced cell growth. To test whether the effects of p72 were estrogen-specific, we treated these cells with 4-hydroxytamoxifen (4-OHT) alone or in combination with estrogen. We found that tamoxifen treatment does not further slow the growth of the cells, suggesting that the inhibition of MCF-7 cell growth following p72 knockdown is due to inhibition of ERα activity (Figure 4c). For all experiments, knockdown of p68 and p72 was confirmed by western blotting. Knockdown of p72 in ZR75-1 cells similarly inhibited estrogen-stimulated growth, indicating that p72 is also required for ER function in these cells (Supplementary Figure 1).

In order to assess their significance in breast cancer pathogenesis, immunohistochemical staining of 233 ERα positive tumours was performed for p68 and p72 (Figure 5a). p68 and p72 staining data were obtained for 226 and 229 cases respectively, of which 67 were p68 positive, and 173 were p72 positive. These tumours had previously been immunostained for ERα phosphorylated at S118, as well as a series of other markers including Her2, PR (Sarwar et al., 2006) and AIB-1. We correlated expression of p68, p72 and these markers with other pathological and clinical data for each tumour, including histological grade, lymph node involvement, time to relapse and death (Table 1). There was no statistically significant association between p68 and p72 expression in this group of patients (p=0.397). p68 expression was associated with Her2 and AIB-1 positivity (p=0.001 and p<0.001, respectively), as well as with increasing tumour grade (p=0.044), all of which are markers of poor prognosis (Osborne et al., 2003). However, no association was identified between p68 expression and either relapse-free or overall survival (p=0.097 and p=0.200 respectively) in our patient cohort (Figure 5b, c). Conversely, p72 expression was associated with a favourable prognosis. Most strikingly, p72 positive tumours were associated with Her2 negativity (p=0.008), progesterone receptor positivity (p=0.037) and reduced likelihood of relapse (p=0.025) and cancer death (p=0.014). In agreement with this, Kaplan-Meier analysis demonstrated that p72 positive tumours are associated with a longer relapse-free period and overall survival (p=0.006 and p=0.016 respectively) (Figure 5d, e). However, multivariate analysis showed that p72 was not a significant predictor of relapse-free or overall survival (Table 2). Lymph node positivity, as well as Her2 and AIB1 expression were significant predictors of relapse-free survival (p<0.001, p=0.003 and p=0.024 respectively), whilst lymph node positivity, Her2 positivity and tumour grade predicted overall survival (p=0.001, p=0.017 and p=0.008 respectively).

Dysregulation of ERα-mediated inhibition of Her2 expression is one mechanism through which breast tumours may become resistant to anti-estrogen treatment. AIB-1 is thought to play a crucial role in this dysregulation by competing for ERα binding with Pax2 and reversing ERα-dependent repression of Her2 (Hurtado et al., 2008). We examined the effects of expression of p72 in AIB-1 positive tumours and its relation to Her2 expression. As has been previously described in a number of studies (Osborne et al., 2003), AIB-1 positivity in the tumours was associated with poorer prognosis (Figure 6a, b). While expression of p72 had little effect on the prognosis of patients with AIB-1 negative tumours, p72 expression in AIB-1 positive tumours was associated with an increased period of relapse-free and overall survival (Figure 6c, d, p=0.003 and p=0.043 respectively). Furthermore, the proportion of Her2 positive tumours was significantly lower in AIB-1 positive tumours that were also positive for p72 (p=0.005, Figure 6e). Expression of p68 did not show any association with prognosis in AIB-1 positive tumours, although the proportion of Her2 positive tumours was significantly increased in AIB-1 positive tumours expressing p68 (p=0.005), nor did p72 expression influence the prognosis of Her2 positive tumours (Supplementary Figures 2, 3).

MCF-7, ZR75-1 and COS-1 cells were routinely maintained in Dulbecco’s Modified Eagles Medium (DMEM) supplemented with 10% foetal calf serum (FCS), 2mM L-glutamine, 100 μg/ml streptomycin and 100 U/ml penicillin (all supplied by Invitrogen, Paisley, UK) in 5% CO₂ at 37°C. Prior to transfection, cells were placed for 72 hours in DMEM without phenol red (Invitrogen) supplemented with 5% double dextran-charcoal stripped FCS (First Link, Birmingham, UK), 2mM glutamine, 100 μg/ml streptomycin and 100 U/ml penicillin.

Plasmids expressing full length and deletion mutants of ERα have previously been described (Tora et al., 1989), as have the GST-ERα fusion proteins and the estrogen-responsive luciferase reporter gene (Lopez-Garcia et al., 2006). p68, p72 and the DEAD box mutants have also previously been described (Bates et al., 2005). The SRC-1 expression plasmid was a kind gift of Dr M. Parker. pERE3-TATA-luc has also previosuly been described (Thomas et al. 2008). A thymidine kinase promoter Renilla luciferase reporter plasmid (RLTK; Promega, UK) was used to control for transfection efficiency.

siRNA transfections were carried out using Lipofectamine RNAiMax (Invitrogen) by reverse transfection. siRNA oligos against p68, p72 and a non-silencing control have been previously described (Bates et al., 2005). 17ß-estradiol (Sigma, UK) was added to a final concentration of 10 nM, 56 hours following transfection, and cells were harvested after a further 16 hours for RNA or protein preparation. An equal volume of ethanol (vehicle) was added to the ‘no ligand’ controls.

COS-1 cells were transiently transfected with ERα or ERα deletion constructs, together with p68, p72, SRC1, together with pERE3-TAT-luc and pRL-TK reporter genes, as previously described (Thomas et al., 2008). Luciferase activities were determined using Dual-Glo reagents (Promega). Firefly luciferase levels were corrected for transfection efficiency using corresponding renilla luciferase levels. All experiments were independently repeated at least three times, and the data graphed as mean values, with error bars representing the standard errors of the mean.

GST-tagged proteins were expressed in Rosetta BL21 E. coli cells and purified on glutathione beads (GE Healthcare) as previously described (Bates et al., 2005). p68 and p72 were translated and ³⁵S-labelled using the TNT in-vitro transcription/translation kit (Promega). The GST pull-down of in-vitro-translated protein was carried out as described by (Hsieh et al., 1999). Where GST-tagged proteins were phosphorylated, the following modifications were made to the protocol. Following binding of GST lysates, the beads were washed and resuspended in 2.5 bed volumes of 1xMAPK buffer (New England Biolabs), containing 200μM ATP. 0.5μl per 30μl of buffer of activated ERK2 (New England Biolabs) was added to the beads, which were then incubated at 30°C for 30 minutes. Following this, the beads were washed and the protocol proceeded with incubation with ³⁵S-labelled proteins as previously described (Hsieh et al., 1999).

Nuclear extracts were prepared as described previously (Bates et al., 2005) from MCF-7 cells grown in estrogen-depleted media (DMEM containing 5% dextran-coated charcoal-stripped FCS) for 72 hours with the addition of 10nM 17β-estradiol or an equivalent volume of ethanol for 6 hours, prior to extract preparation. Immunoprecipitations were carried out using antibodies against ERα (6F11; Novocastra, Newcastle-upon-Tyne, UK) and p68 (PAb204; Millipore, USA) as described previously (Bates et al., 2005).

Western blot analysis was carried out as described previously (Bates et al., 2005) using antibodies against p68 (PAb204 and 2907 (rabbit polyclonal raised against the C-terminal 15 amino acids of p68)); p72 (DRD-K12 (rabbit polyclonal raised against amino acids 12-26 of p72)); ERα (HC20; Santa Cruz Biotechnologies, CA, USA); ERα-phospho-S118 (Cell Signalling Technology, MA, USA); pS2 (FL-84, Santa Cruz Biotechnologies); Cathepsin D (CTD-19; Abcam, Cambridge, UK); and actin (A2066; Sigma-Aldrich, Poole, UK). Appropriate HRP-conjugated anti-mouse and anti-rabbit secondary antibodies for immunoblotting were purchased from DAKO (Ely, UK).

Cells were transfected in 6-well plates with siRNA, as described above, except with the addition of 10nM 17ß-estradiol (E2), 100nM 4-hydroxytamoxifen (4-OHT, Sigma-Aldrich), a combination of both ligands, or an equivalent volume of ethanol at the time of transfection. Transfections were carried out in triplicate. 96 hours after transfection, cells were trypsinised and counted using a haemocytometer. Following counting, cells were washed twice in PBS and lysed in RIPA buffer. Lysates were sonicated and cleared by centrifugation at 14,000 rpm for 15 mins. Protein concentrations were measured by Bradford assay (Sigma-Aldrich), equilibrated and analysed by Western blotting.

Total RNA was extracted from cells using the RNeasy kit (Qiagen) according to the manufacturer’s protocol. 1μg of RNA was treated with RQ1 RNase-free DNase (Promega) and reverse-transcribed using M-MLV reverse transcriptase (Invitrogen) according to the manufacturer’s instructions. Quantitative (Taqman) RT-PCR was carried out using the Stratagene MX3005-P instrument (Agilent). Taqman gene expression assays (Applied Biosystems) were purchased for each gene (Supplementary Table 2). β-actin was used as a control and fold changes and differences between samples were calculated using the ΔΔCt method in Microsoft Excel.

Two hundred and thirty three patients with primary invasive breast cancer, who had undergone surgery at Charing Cross Hospital between 1981 and 2003 were selected on the basis of the availability of clinical details at presentation and follow-up, including time to relapse, time to death, ER and PR status (Sarwar et al., 2006). ER and PR status was confirmed using ER (VP-E613; Vector Laboratories), PR (MU328-UC; Biogenex, UK) antibodies, and Her2 status was determined, as described (Jiang et al., 2007). The clinico-pathological characteristics of the patient cohort are shown in Table 1. Immunohistochemical staining and scoring was carried out as previously described (Sarwar et al., 2006). Immunostaining and scoring for AIB1 was carried out as described (Hurtado et al., 2008). Staining for p68 and p72 was carried out using the antibodies PAb204 (Stevenson et al., 1998) and DRD-K12 respectively. p68 and p72 were scored positive if greater than 10% of tumour cell nuclei were stained. This study fulfilled the Institutional Ethics Review Board’s guidelines for the use of stored tissues samples.

All statistical analyses were carried out using SPSS v14.0. Comparisons of cell counts were carried out by t-test. Immunohistochemical scores for p68 and p72 were compared with clinico-pathological features using the Pearson chi-squared test. Survival analyses were carried out using the Kaplan-Meier survival function and the Mantel-Cox Log-rank test. Multivariate analysis was carried out using the Cox Proportional Hazards model.