Estrogen Coordinates Translation and Transcription, Revealing a Role for NRSF in Human Breast Cancer Cells

Abstract

Posttranscriptional regulation may enhance or inhibit estrogen transcriptional control to promote proliferation of breast cancer cells. To understand how transcriptome and translational responses coordinate to drive proliferation, we determined estrogen's global and specific effects on translation regulation by comparing the genome-wide profiles of total mRNA, polysome-associated mRNA, and monosome-associated mRNAs in MCF-7 cells after stimulation by 1 h of 10 nm 17β-estradiol (E2). We observe three significant, novel findings. 1) E2 regulates several transcripts and pathways at the translation level. 2) We find that polysome analysis has higher sensitivity than total RNA in detecting E2-regulated transcripts as exemplified by observing stronger E2-induced enrichment of E2 expression signatures in polysomes more than in total RNA. This increased sensitivity allowed the identification of the repression of neural restrictive silencing factor targets in polysome-associated RNA but not total RNA. NRSF activity was required for E2 stimulation of the cell cycle. 3) We observe that the initial translation state is already high for E2 up-regulated transcripts before E2 treatment and vice versa for E2 down-regulated transcripts. This suggests that the translation state anticipates potential E2-induced transcriptome levels. Together, these data suggest that E2 stimulates breast cancer cells by regulating translation using multiple mechanisms. In sum, we show that polysome profiling of E2 regulation of breast cancer cells provides novel insights into hormone action and can identify novel factors critical for breast cancer cell growth.

Genome-wide translational profiling characterizes novel modes of coordination between transcription and translation and identifies new factors mediating estrogen stimulation of breast cancer cells.

Estrogen influences many cell types and is a critical driver of many breast cancers. Estrogen stimulates 70% of breast cancers through the transcription factor estrogen receptor-α (ER). Expression profiling has provided many important insights into genes and networks controlling breast cancer proliferation (1,2,3,4). It has become clear that a complex network of genes and pathways is controlled by 17β-estradiol (E2) to promote breast cancer cell growth. Although transcriptional regulation is clearly central to E2 stimulation of proliferation, little is understood about the coordination with the myriad downstream posttranscriptional processes necessary to transduce these signals and synthesize proteins.

Estrogen stimulation of protein synthesis was first observed in the 1960s, leading to active discussion about whether estrogen directly stimulates transcription or translation (5,6,7). The discovery of ER solidified the idea that estrogen can directly regulate RNA synthesis (8,9). Since then, most efforts have focused on understanding transcriptional regulation and how this leads to proliferation of breast cancer cells (1,2,3,4). Complementing transcription, protein synthesis stimulation is vital for rapid cancer cell growth. The regulation of translation has emerged as a vital player in driving tumors (10), as exemplified by the role of mammalian target of rapamycin (mTOR) (11), the overexpression of initiation factors such as eukaryotic initiation factor 4E (eIF4E) (12,13,14), and the role of micro-RNAs in numerous cancers including ER⁺ breast tumors (15,16). Recently, miR-21 has been shown to be directly regulated by ER (17,18), providing a mechanism linking ER transcription control and translation regulation.

Most regulation of translation occurs during initiation as extensively reviewed (19). Cell cycle-dependent activation of translation initiation has been described (14,20). High expression levels of translation initiation factors such as eIF4E have transforming and strong proliferative properties, suggesting that these proteins may be important oncogenes (13,21). From these observations and the targeting of mTOR in breast cancer (22), translational control is emerging as a potential therapeutic target to regulate cancer cell growth. Homodirectional coordination between transcription and translation has been described in yeast (23) and makes intuitive sense to drive gene expression. The term potentiation was suggested by Preiss and co-workers (23) to describe this phenomenon.

Estrogen control of posttranscriptional regulation, including translation, has been reported for a number of transcripts (24,25,26). In Xenopus, polysomal ribonuclease 1 is a nuclease, activated by E2, that degrades polysome-associated transcripts (27). A second example is the E2-inducible protein vigilin, which binds 3′-untranslated regions to control pre-mRNA processing (28). These examples highlight how E2 action involves specific posttranscriptional events as well as transcriptional events. In cancer, E2 may reduce protein expression of ABC transporters posttranscriptionally to mediate drug sensitivity (29,30).

More recently, a new appreciation has emerged for the role of nongenomic, transcription-independent control of cell behavior by estrogens (31,32). Both genomic and nongenomic pathways are thought to contribute to breast cancer growth (33,34). Within 10 min of E2 treatment, rapid phosphorylation of translation machinery including some of the downstream targets of mTOR, ribosomal protein S6 kinase, and eukaroyotic translation initiation factor 4E binding protein in MCF-7 cells has been reported, suggesting a direct mechanism linking E2 with regulation of protein synthesis (35).

Translational profiling is one of the best ways to determine which fraction of the genome is actually made into proteins (36). This method determines the population of actively translated mRNAs by microarray analysis after sucrose gradient fractionation. Translational profiling has been applied to many model systems, from yeast (37,38) to cell lines (39,40), to uncover novel pathways and genes involved in stem cell differentiation (41) and apoptosis (40). Often, translational profiling is applied to systems where global translation is decreased, such as during apoptosis or other stress conditions (40,42,43,44). Other studies have examined specific translation regulation upon significant increase in protein synthesis such as during growth stimulation (23,41,45). All of these studies have found that transcripts loaded onto polysomes are likely translated (46). Specific putative internal ribosome entry sites such as polypyrimidine-rich elements (40) and 5′ terminal oligopyrmidine tract sequences (45) have been identified in mammalian transcripts as drivers of cap-independent translation. Specific regulation of translation by general translation factors, such as eIF4E, has been reported (45,47).

As a first step toward understanding how estrogens may control translation of specific transcripts and pathways in breast cancer cells, we probed RNA isolated from polysomes, monosomes, and total RNA on Affymetrix microarrays from MCF-7 breast cancer cells E2 stimulated for 1 h. We hypothesized that many key factors and pathways are translationally regulated by E2, complementing classic transcription control through estrogen receptors. The simplest expected mechanism is that some specific pathways and genes would show correlated regulation in both the transcriptome and translational responses. Such homodirectional changes have been termed potentiation, (23) and have been observed in a number of systems (48,49). Alternatively, transcripts may be stimulated translationally but not in the transcriptome. Consistent with this possibility, we observe pathways and genes likely critical for MCF-7 growth including ribosomal proteins, energy metabolism pathways, and the ERK pathway stimulated by E2 at the translation but not transcriptome level. Examination of E2-induced changes in polysome and total RNA revealed a second mechanism of apparent increased sensitivity of E2-regulated transcripts in polysomes, including known E2 expression signatures. We took advantage of this apparent amplification and observed that targets of neuron-restrictive silencer factor (NRSF)/RE-1 silencing transcription factor (REST) are down-regulated by E2 in polysomes more so than in total RNA. NRSF is a transcriptional repressor that suppresses neuronal-specific genes in nonneuron cells (50) and can regulate cancer cell proliferation (51). To test whether transcription factor networks detected in polysomes, but not total RNA, are important for E2 action, we knocked down NRSF and found that NRSF activity is important for E2 stimulation. Interestingly, these observations correlate with higher NRSF expression in relapsed ER⁺ breast tumors. Surprisingly, we observe a third mechanism coordinating the transcriptome and translational responses. The translation state of E2 up-regulated transcripts was already high before E2 treatment, anticipating the eventual expression changes. Together, these data highlight how transcription and translation coordinate at multiple levels and demonstrate how polysome analysis can be used to identify key regulators of steroid hormone action.

Results

Hours of E2 stimulation significantly affects the expression of thousands of genes, whereas far fewer are observed to change significantly at 1 h. This suggests that many of the early responders are more likely to be directly regulated by ER. Here, we aimed to understand what the contribution of translational regulation may be to E2 stimulation of breast cancer cell proliferation. To probe E2 translational regulation, we compared transcript levels in polysomes and monosomes with total RNA between E2-treated and vehicle control MCF-7 cells (Fig. 1). MCF-7 cells were hormone deprived in charcoal-dextran-treated (CDT) serum for 3 d before addition of 10 nm E2. We chose these conditions to compare these data with the vast array of transcription and gene expression data (3,4,52,53,54). Polysomes and light subpolysome fractions, monosomes, were isolated as described in Materials and Methods by linear 10–45% (wt/vol) sucrose gradient fractionation (Fig. 1). We observe that 1 h of E2 stimulation induces no obvious, significant increase in polysome/monosome ratios in MCF-7 breast cancer cells (Fig. 1). Addition of the vehicle, 0.1% ethanol, does not significantly affect the polysome profile compared with untreated hormone-deprived MCF-7 cells (data not shown). Consistent with these observations, we typically isolate similar amounts of polysome RNA from 1 h E2-treated and vehicle-treated cells. The observed peaks are likely active polysomes, because their formation is puromycin dependent. Puromycin significantly reduces the appearance of polysomes by greater than 100-fold, suggesting that most of the mRNA cosedimenting with polysomes is likely associated with the polysomes (Supplemental Fig. 1, published on The Endocrine Society's Journals Online web site at http://mend.endojournals.org). [³⁵S]Methionine and cysteine labeling of total protein after 5 h of high 100 nm concentration of E2 suggests a modest 10% increase in protein levels, whereas slightly lower stimulation with 10 nm E2 is observed (data not shown). This is consistent with small increases in polysome/monosome ratios by 3 h (data not shown). Together, these observations suggest that as E2 stimulates hormone-deprived MCF-7 cells, resulting in only modest increases in global translation, indicating that any changes in the first hour of E2 stimulation are likely specific.

Figure 1.

E2 effect on polysomes and microarray analysis. A, Representative fractionation profile from 10–45% linear sucrose gradients from MCF-7 cells treated with 10 nm E2 or 0.1% ethanol (vehicle) for 1 h. The average ratio with the sd of triplicates of the P/M ratio in E2 vs. 0.1% ethanol (vehicle) is shown. Sucrose density gradients separated the mRNAs into the subpolysomal/monosome and polysome fractions. Sucrose fractionation was monitored at 254 nm. B, Scheme of the procedure of translational profiling. Unfractionated total RNA was isolated in each condition and hybridized on Affymetrix microarrays. Veh, Vehicle condition.

RNA was isolated from pooled polysome fractions by guanidine hydrochloride ethanol precipitation followed by purification through a QIAGEN RNeasy column. Biological triplicates were prepared and hybridized to Affymetrix Human Gene ST 1.0 arrays. Arrays were all of high quality and reproducibility, as determined by Affymetrix metrics and replication plots (Supplemental Fig. 2). As internal controls, we observe stimulation of well-known E2-regulated genes, including XBP1 (2-fold) and MYC (3-fold), in both total and polysome RNA (Supplemental Table 1).

To analyze the polysome data, we compare three values for each transcript: the signal in total RNA (T), the signal in the polysomes (P), and the signal in the monosomes (M). Total RNA measures the transcriptome and is indicative of the balance of transcription and degradation in all cellular compartments. Polysome RNA is indicative of transcripts engaged with ribosomes during the elongation phase. The light subpolysome fractions, monosomes, are dominated by interactions with preinitiation ribosomal complexes. The distribution between the polysome and monosome fraction (P/M) indicates the translation state in that condition (23). The polysome and total RNA (P/T) also can indicate the fraction of the RNA in the cell that is associating with polysomes and represents a second measure of translation state (45). In sum, there are two major approaches that we use to characterize how E2 may coordinate the transcriptome and translational responses. 1) Changes in the polysome fractions (P_E2/P_Veh) are sensitive to both translation and overall mRNA level changes. Comparing this ratio to changes in total RNA (T_E2/T_Veh) can identify candidate translationally regulated genes. 2) The translation state of each transcript defined as P/T or P/M indicates the fraction of the transcript cosedimenting with polysomes. Changes in these ratios would indicate translation regulation. Each analysis is sensitive to different dynamic range and noise, and thus, each analysis complements the other in understanding translation regulation and identifying gene expression regulators important for E2 action.

Because the polysome traces did not show significant differences when treated with E2, we did not expect to see changes in the global distribution of transcripts between the E2 and vehicle conditions. Figure 2 shows that the distribution of transcripts in both total vs. total and polysome vs. polysome comparisons is not globally shifted, suggesting that the observed differences are relatively specific in this background (Fig. 2, A and B). This also allows us not to use a scale factor to correct for significant mass differences in the E2 and vehicle conditions to identify specific translation state regulation.

Figure 2.

No significant global up-regulation is observed, suggesting specific changes. Graphical representations of the relationship between transcriptome and translational changes reveal no dramatic shift in the signal in polysome or total RNA after 1 h of 10 nm E2 treatment. A, Total RNA fold change compared with total RNA transcript signal shows no significant global shift in transcript expression levels. B, Polysome fold change compared with total RNA transcript signal indicates no significant global shift in polysome association. C, Color coding indicates whether the changes are in total RNA (orange), polysome RNA (red), both (light blue), or unchanged (dark blue) in E2 vs. vehicle. Only signals from RefSeq transcripts with average log intensity higher than 6 are included. Each point represents an individual transcript. D, qPCR analysis of the translation state correlates with microarray signals. Luciferase RNA was spiked in before purification for normalization. The y-axis indicates the log of the P/M ratios in the designated condition. The x-axis indicates the qPCR-determined P/M ratio. A linear relationship between the qPCR and array observations is indicated by least squares fitting of a line with high R² to the data. Veh, Vehicle.

To test the quality of the microarray data, we sampled transcripts by quantitative PCR (qPCR). To perform qPCR, luciferase mRNA was added to each fraction to normalize for purification yields. We sampled 13 genes in each pooled fraction and observe high correlation (R² > 0.8; R², coefficient of determination which is square to the correlation coefficient) between the microarray and qPCR data, suggesting that the array data are accurate (Supplemental Table 1). To determine whether the microarray data accurately portray the translation state of transcripts, we compared the polysome to monosome ratio from the microarray to qPCR measurements for selected transcripts (Fig. 2D). This requires comparing RNA isolated from polysomes and monosomes, which originate from different yields of mass and volume. Figure 2D shows that although the absolute ratios differ between fractions for eight genes selected for significantly high and low translation states (P < 0.001, fold change (Fc) > 2) in the microarray data, the two measurements accurately determine the relatively high and low translation states for each transcript. It is not clear why the absolute ratios differ between the two measurements. Most reports compare microarray data to Northern blots without spike-ins and thus may not be accounting for variable purification yields from each fraction. These data indicate that relatively high and low translation states are accurately measured by two independent assays.

E2 regulates specific genes and pathways at the translation level

One mechanism for E2 regulation of translation is to control initiation, the major rate-limiting step in protein synthesis. Polysome cosedimentation indicates transcripts that have entered the elongation phase of the translation cycle. Exclusive changes in polysome RNA or total RNA would indicate translation or transcriptome regulation, respectively. We examined the correlation between E2 regulation in total and polysome RNA for the most significantly changing genes (Fig. 3A). We observed only a small number of significantly changing genes using thresholds of P < 0.05 and fold changes greater than 1.5, with a strong bias toward up-regulated genes in polysomes and total RNA (Supplemental Tables 2 and 3). Figure 3A shows that most transcripts are stimulated in both the total RNA and polysome RNA fractions, consistent with potentiation, where transcriptome and polysome changes are significantly correlated.

Figure 3.

E2 induces specific translational responses and transcriptome level changes. A, Significantly changing genes in the transcriptome and translational responses are plotted against each other to show that they are correlated. The y-axis indicates the log fold change in total RNA, transcriptome, and the x-axis is the log fold change in polysomes in the E2 and vehicle (Veh) conditions. The left graph compares transcripts with significant (P value <0.05; fold change >0.5) changes in the transcriptome. The right graph compares transcripts with significant changes (P value <0.05; fold change >0.5) in the polysomes. The Pearson correlation coefficient is indicated above each graph. B, GSEA expression analysis of a representative E2 expression signature (53) in the transcriptome and translational responses. The black lines demonstrate where each gene in the gene set falls within the 15,000 probe sets probing RefSeq transcripts ordered from left to right based on the E2 polysome/vehicle polysome ratio with gene 1 the most highly expressed in E2 polysomes. The green line represents the running enrichment score (ES) that becomes more negative as probe sets are identified toward the bottom of the list. Significant up-regulation is seen in both the transcriptome and polysomes. More significant bias of the Frasor ER expression signature is observed in the polysomes compared with the bias observed in the E2 Total vs. vehicle total, transcriptome comparison. More black lines are observed near the top of the rankings in E2 (left side) than in the vehicle (right side). C, Significant bias of the NRSF gene set is observed in the polysomes but not the total RNA because more black lines are near the bottom of the list in the E2 polysome vs. vehicle polysome comparison. The NRSF gene set is a group of genes with NRSF-binding motifs in their promoter regions. The NRSF gene set is fairly evenly distributed in the E2 total vs. vehicle total comparison.

We identified the genes with the highest and lowest translation states using a t test (P < 0.05) and fold change (>50% change) thresholds. The distribution of the translation states in E2- and vehicle-treated cells shows that genes with high translation states in the vehicle-treated cells are not stimulated to even higher translation states by E2 treatment (Supplemental Fig. 3A). On the other hand, transcripts with low translation states in the vehicle condition are modestly driven toward higher translation states at this early time point (Supplemental Fig. 3B). At longer times, we see a larger global shift into polysomes consistent with increased polysome/monosome ratios seen on the sucrose gradients (data not shown).

Because few genes are significantly changing, we sought to examine the overall trends in the data. To gain insight into the pathways that may be regulated at the transcriptome or translation levels, we compared E2 polysome vs. vehicle polysome and E2 total vs. vehicle total changes using gene set enrichment analysis (GSEA) (55). GSEA uses all the data to determine whether specific gene sets, derived from pathways or transcription factor networks, may be biased in one condition compared with another. GSEA can therefore identify regulated networks without arbitrary thresholds but, rather, by determining trends using all the microarray data. This approach has proven invaluable to identify and experimentally confirm the role of critical pathways in numerous systems (56,57). Table 1 summarizes some of the top pathways translationally affected by E2, i.e. in polysomes only, such as oxidative phosphorylation, or in total RNA only, such as the cAMP response element-binding protein (CREB) pathway, or in both, such as the MAPK pathway. E2 appears to stimulate energy metabolism and ribosomal proteins in polysomes but not total RNA. In sum, these observations suggest that E2 uses different mechanisms to stimulate genes required for growth and survival.

Table 1.

GSEA reveals stronger stimulation of known E2-regulated genes in polysomes

Pathway	E2 polysome vs. vehicle polysome		E2 total vs. vehicle total
	NES	P value	NES	P value
ER binding (genes within 10 kb)	2.06	0.000	1.61	0.000
Stossi E2 UP signature	2.35	0.000	1.74	0.000
Frasor E2 UP signature	3.04	0.000	2.00	0.000
GenMapp
Ribosomal proteins	2.87	0.000	0.68	0.975
Proteasome	2.07	0.000	−0.80	0.754
Oxidative phosphorylation	1.75	0.000	0.69	0.952
Glycolysis	1.65	0.023	0.51	1.000
DNA replication reactome	−0.83	0.738	1.42	0.034
G1 to S cell cycle reactome	0.87	0.723	1.20	0.168
ATP synthesis	1.17	0.261	0.51	0.985
RNA transcription reactome	1.44	0.060	1.59	0.016
Apoptosis	1.30	0.109	1.34	0.078
Biocarta
ERK pathway	1.80	0.003	1.25	0.163
WNT pathway	1.55	0.044	1.51	0.037
MAPK pathway	1.43	0.022	1.43	0.021
P38MAPK pathway	1.17	0.209	1.82	0.004
CREB pathway	0.99	0.462	1.58	0.025

NES higher than 1.4 suggests significant bias toward E2. Positive NES indicates bias toward E2, whereas negative values indicate vehicle bias. P values <0.1 are significant. Bold indicates possible translational regulation with significant E2 stimulation in the polysome level only. Italics indicates transcriptome-level E2 stimulation but not a translational response in the polysome comparisons. 

A second approach to examine translation regulation is to compare the translation state of pathways in the E2- and vehicle-treated cells. Pathways that show changes in both translation state and polysome levels, but not total RNA, are the most likely to be truly translationally regulated. Table 2 lists pathways significantly enriched in P/T or P/M measurements of the translation state. Pathways highlighted in bold or italics show different enrichments in the vehicle and E2 conditions. These include some of the same pathways observed to change in polysomes but not total RNA in Table 1, including the ERK pathway, whereas others change in total RNA but not polysomes such as the p38MAPK pathway. Similar to past studies (23), some enrichment is seen in one approach but not the other. This is likely due to experimental noise in each analysis. Thus, the most likely candidates to be translationally regulated are those showing significant responses in both approaches, such as the ERK pathway, which shows changes in polysome vs. polysome [normalized enrichment score (NES) = 1.80], but not total vs. total (NES = 1.25), and changes in translation state (E2 NES = 1.47 vs. vehicle NES = 1.28).

Table 2.

Transcriptome or translation level analysis of specific pathways by GSEA

Pathway	E2 polysome vs. E2 total		Vehicle polysome vs. vehicle total		E2 polysome vs. E2 monosome		Vehicle polysome vs. vehicle monosome
	NES	P value	NES	P value	NES	P value	NES	P value
Genes within 10 kb ER binding site	1.03	0.421	1.00	0.503	1.33	0.003	1.39	0.001
Stossi E2 UP signature	−1.05	0.375	−1.32	0.068	1.06	0.359	1.02	0.441
Frasor E2 UP signature	1.49	0.041	1.50	0.041	1.98	0.000	1.81	0.000
GenMapp
Ribosomal proteins	−2.48	0.000	−2.96	0.000	−3.69	0.000	−3.77	0.000
Proteasome	1.73	0.011	1.11	0.331	2.02	0.000	2.08	0.000
Oxidative phosphorylation	1.15	0.219	1.15	0.211	1.92	0.000	1.92	0.000
Glycolysis	−0.99	0.441	−0.96	0.525	1.16	0.239	1.17	0.228
DNA replication reactome	1.18	0.183	1.45	0.045	1.67	0.002	1.45	0.044
G1 to S cell cycle reactome	1.29	0.080	1.27	0.130	1.56	0.002	1.33	0.068
ATP synthesis	0.80	0.715	0.95	0.514	2.04	0.004	1.83	0.002
RNA transcription reactome	1.95	0.000	1.43	0.043	1.93	0.000	1.66	0.008
Apoptosis	1.41	0.044	1.41	0.048	1.25	0.108	1.49	0.020
Biocarta
ERK pathway	0.57	0.971	0.77	0.8	1.47	0.046	1.28	0.143
WNT pathway	−0.87	0.644	1.10	0.346	1.29	0.146	1.24	0.175
MAPK pathway	1.88	0.000	1.44	0.011	1.82	0.000	1.62	0.004
P38MAPK pathway	1.64	0.005	1.56	0.091	1.79	0.002	1.69	0.004
CREB pathway	1.49	0.044	1.53	0.015	1.90	0.000	1.87	0.002

NES higher than 1.4 suggests significant bias toward E2. Positive NES indicates bias toward E2, whereas negative values indicate vehicle bias. P values <0.1 are significant. Changes in translation state in between conditions indicate possible translation regulation of the indicated pathway. Bold indicates translationally up-regulated by E2, and italics indicates translational down-regulation by E2. 

E2-regulated genes are more enriched in polysomes than in total RNA

Correlated changes in both the transcriptome and translation state is a phenomenon termed potentiation (23). We anticipated that many pathways and expression signatures stimulated by E2 would show both increased transcriptome expression and translation states, consistent with potentiation. This is the case for some pathways, such as the MAPK pathway, where upon E2 treatment, the translation state increases (E2 NES = 1.82 and vehicle NES = 1.62), transcriptome levels increase (NES = 1.43), and polysome levels increase (NES = 1.43). The data in Tables 1 and 2 also suggest that some pathways are more enriched in polysome comparisons than in total RNA. Most striking is the observed stronger enrichment of known E2 expression signatures (4,53) in polysomes compared with transcriptome level changes. This is shown in Fig. 3B, where more of the genes in the Frasor E2 expression signature are biased toward E2 in polysomes than in total RNA. To test whether ER-regulated genes are also enriched in polysomes, we created gene sets for all genes with transcription start sites within 10 kb of chromatin immunoprecipitation observed binding regions (3,58). Stronger biases are observed in E2 polysome vs. vehicle polysome compared with E2 total vs. vehicle total (Table 1). These observations suggest that increased expression of transcripts previously associated with E2 action have large fold changes in polysomes from two possible mechanisms. 1) Even before total mRNA levels of ER-bound genes are observed to increase, these transcripts are translationally stimulated in the polysomes. This model implies that the cells may be specifically stimulating ER genes translationally before responding to E2 genomic action. 2) A more likely explanation may be that detection of mRNA level changes is more sensitive in polysome RNA than total RNA, perhaps because small changes in total RNA are amplified at the translational level, enhancing E2-induced transcriptome changes. Polysome mRNA may also represent a higher level of purification of mature mRNA, without confounding effects of nuclear fractions as recently suggested by examining cytoplasmic vs. nuclear RNA (59). In sum, E2-stimulated genes are observed with higher sensitivity in polysomes compared with total RNA after 1 h of E2 treatment.

Enhanced sensitivity of gene expression regulation in polysomes reveals NRSF

We hypothesized that the higher apparent sensitivity observed for E2 expression signatures and ER-bound genes in polysomes could be useful to identify novel networks that may mediate E2 action. Therefore, we investigated whether enrichment of gene sets with common transcription factor-binding promoter motifs is observed in total and/or polysome RNA fractions. Transcription factors are often key regulators controlling cell fate, and significant precedent exists for identifying them using GSEA (56). GSEA reveals a number of gene sets with common transcription factor motifs in promoters enriched in polysome RNA but not total RNA (Table 3). E2F1-regulated genes are enriched in polysomes, but not total RNA, at 1 h. E2F1 has been observed to be enriched in E2-stimulated MCF-7 cells from a long time-course experiment (60). Genes regulated by the aryl hydrocarbon receptor are also enriched in E2 polysome compared with vehicle polysome RNA. Aryl hydrocarbon receptor has previously been associated with E2 signaling (61). These data suggest that polysome profiling has increased sensitivity in detecting significant changes in specific gene expression programs known to be important for E2 action. Together, these findings suggest that even at 1 h, E2 is rapidly stimulating many of the specific gene expression programs critical for E2 action and that these are more readily detected in polysome RNA than in total RNA.

Table 3.

GSEA reveals enrichment of transcription factor programs

Transcription factor	E2 Polysome vs. vehicle polysome		E2 Total vs. vehicle total		E2 polysome vs. E2 total		Vehicle polysome vs. vehicle total		E2 polysome vs. E2 monosome		Vehicle polysome vs. vehicle monosome
	NES	P value	NES	P value	NES	P value	NES	P value	NES	P value	NES	P value
NRSF	−1.68	0.000	−1.18	0.170	−2.32	0.000	−2.02	0.000	−2.32	0.000	−2.20	0.000
E2F1	1.53	0.000	1.10	0.211	1.65	0.000	1.64	0.001	1.76	0.000	1.86	0.000
AHR	1.69	0.000	0.89	0.699	1.36	0.023	1.48	0.008	1.31	0.031	1.28	0.051

Gene sets are defined by common sequence motifs in promoter regions. NES higher than 1.4 suggests significant bias toward E2. Positive NES indicates bias toward E2, whereas negative values indicate bias toward vehicle. P values <0.1 are significant. 

To determine whether transcription factor networks enriched in translational responses are important for E2 stimulation, we tested the function of NRSF. NRSF-regulated genes are repressed in E2 polysomes, compared with vehicle polysomes, whereas no significant change is observed in total RNA (Fig. 4). Moreover, NRSF-regulated genes have lower (NES = −2.32) P/M translation states in E2 compared with vehicle (NES = −2.20). To our knowledge, NRSF, a transcription repressor of neuron-specific genes in nonneuronal tissue (50), has not previously been associated with E2 action, although it has been suggested to be a tumor suppressor in breast cancer (62,63). However, NRSF has also been shown to be oncogenic in some contexts (as reviewed in Ref. 64). NRSF can mediate glucocorticoid receptor action (65), providing precedent for direct connections to nuclear receptors.

Figure 4.

NRSF enhances E2 stimulation of the cell cycle over 24 h. A, Representative Western blot with anti-NRSF antibody shows significant knockdown. NRSF runs at approximately 180 kDa. qPCR RNA level knockdown is typically about 8-fold. B, Representative propidium iodine staining DNA content FACS shows that NRSF knockdown inhibits E2 stimulation of the cell cycle. MCF-7 cells were hormone deprived in phenol red-free CDT serum for 3 d before 10 nm E2 treatment. Almost 90% of the cells became G1 growth arrested. The percentage of S-phase cells observed after E2 treatment is reduced when NRSF is knocked down before E2 stimulation. The acquired FACS data were modeled with ModFit LT software. EtOH, Ethanol. C, Significantly fewer S-phase cells are observed when NRSF is knocked down, as indicated by the black lines comparing the two conditions. *, P < 0.05. Error bars are sd from triplicate experiments indicating that the NRSF effect is reproducible.

NRSF is required for E2 stimulation of ER⁺ breast cancer cells

The repression of transcripts with NRSF promoter motifs in polysomes predicts that NRSF activity may be crucial for E2 stimulation of the cell cycle in MCF-7 cells. To determine whether NRSF expression is required for E2 action, we knocked down NRSF with small interfering RNAs (siRNAs) in hormone-deprived MCF-7 cells and tested for effects on E2 stimulation of the cell cycle by DNA content fluorescence-activated cell sorting (FACS). Figure 4A shows that significant knockdown of NRSF is observed at the protein level in typical experiments. The DNA content cell cycle data suggest that the cells were synchronized upon hormone removal with approximately 90% of the cells in G1 similar to previous reports (66). Reduced NRSF expression reproducibly inhibited E2 stimulation of the cell cycle, compared with control scrambled siRNA sequences (Fig. 5B). NRSF knockdown did not significantly affect the growth arrest induced by hormone deprivation but inhibited the E2 response. NRSF also affects E2 action in a second ER⁺ cell line, T-47D, suggesting that NRSF's importance in mediating the cell cycle is not MCF-7 specific (Supplemental Fig. 4). We tested whether NRSF affects MCF-7 growth in rich medium and observed that knockdown of NRSF causes a modest 30% decrease in the number of viable cells after 3 d (data not shown). Because NRSF is typically thought to be a transcriptional repressor, these observations are consistent with a model of decreased expression of target genes being important for ER⁺ breast cancer cell growth.

Figure 5.

Translation state anticipates the direction of transcriptome changes upon E2 stimulation. A, Clustering of significantly changing genes from Affymetrix U133 microarray expression data (3) classifies genes into six major patterns of gene expression over a 12-h time course. The number of transcripts in each cluster is indicated in parentheses. Shown in each plot is the averaged expression behavior of all transcripts in the indicated cluster, with error bars indicating the deviation from the mean for the entire cluster. Expression of each transcript was obtained by subtracting the mean of all points at all times from each expression value in the original time series and dividing by the sd. B, The distribution of fraction of transcript associating with polysomes (P/T), translation state, for each cluster is shown as histograms. The y-axis indicates the translation state, whereas the x-axis is the probability density ranging from 0–1.5. The left side of each panel is the translation state distribution for all the genes in gray, whereas the right side of each panel, in black, is the translation state distribution for the genes in the cluster. The P value testing the difference between the distributions of translation states for the cluster compared with all E2-regulated transcripts is shown below each panel along with the median (med) translation state for the transcripts in each cluster. E2 up-regulated genes, clusters 1 and 2, are strongly biased toward higher translation states, whereas E2-down-regulated genes, clusters 5 and 6, are biased toward lower translation states. C, The distribution of the P/M translation state of the transcripts in each cluster is shown. The P value, testing the difference between the distributions of translation states for the cluster compared with all E2-regulated transcripts, is shown below the density distribution, along with the median translation state for the transcripts in each cluster. Similar patterns are seen for both the P/M and P/T ratios in each cluster with stronger biases seen for the P/T distributions.

NRSF may mediate both E2 and tamoxifen (TAM) action. We tested whether NRSF affects TAM activity using a 4′,6-diamidino-2-phenylindole (DAPI) assay to detect apoptotic cells (Supplemental Fig. 5). Only a modest number of apoptotic cells are induced by 2 μm TAM as previously reported (67). When NRSF is knocked down, an increase in the number of apoptotic nuclei is observed compared with negative siRNA controls with and without TAM (Supplemental Fig. 5). These data indicate that NRSF induces MCF-7 cell death with no significant synergistic, but perhaps additive, effects on TAM sensitivity in hormone-deprived conditions. Because TAM has mild agonist activities, this could be complicating the balance of effects observed under these conditions. Additional experiments testing for TAM and NRSF effects on caspase-3, caspase-7, and caspase-9 activity did not reveal any significant effects (data not shown). These data suggest that NRSF is important for E2 stimulation of ER⁺ breast cancer cells but may not be critical for TAM-induced cell death.

These in vitro data suggest a possible role for NRSF in controlling growth of ER⁺ breast cancer cells. To test whether NRSF may be important for breast tumor progression and patient prognosis, we examined NRSF expression in breast tumors using Oncomine (68). For all breast carcinomas, NRSF expression is significantly higher for patients who did not survive 5 yr compared with those who did (Supplemental Fig. 6A). In particular, NRSF is more highly expressed in primary ER⁺ tumors in relapsed patients compared with those with no recurrent disease with modest significance (Supplemental Fig. 6B). Interestingly, NRSF is more highly expressed in ER⁻ breast carcinomas compared with ER⁺ breast carcinomas (Supplemental Fig. 6C). Because ER⁻ carcinomas are typically more aggressive and lethal than ER⁺ tumors, higher NRSF expression may be contributing to faster growth. NRSF expression is also significantly higher in breast carcinomas compared with normal breast tissue (Supplemental Fig. 6D). In sum, higher NRSF expression in poor-prognosis breast carcinomas is consistent with our findings that NRSF is necessary for maximal E2 stimulation of ER⁺ breast cancer cells. These data suggest that additional investigation of NRSF is warranted to understand the role of the NRSF network in breast cancer.

E2 coordination of transcription and translation

The translation state may be reflecting mass action driving transcripts into polysomes such that higher translation state and expression are correlated. To determine whether the translation state of these clusters is correlated with the expression level, we compared the translation state to the total RNA expression levels as defined by the hybridization score (Supplemental Figs. 7 and 8). Up-regulated genes are slightly biased toward lower initial expression levels compared with down-regulated genes. This may reflect that stimulated transcripts are selected to go up upon E2 stimulation and thus may start from a lower position than down-regulated transcripts. In any case, transcripts with a higher translation state are not biased toward higher expression levels as may have been expected. In fact, E2-stimulated transcripts have lower expression levels at the zero time point compared with repressed transcripts, further suggesting that the initial translation state is independent of the expression level. These findings suggest that the translation state of each transcript is independent of expression levels in these conditions.

After 1 h of E2 stimulation, there is not a significant increase in global translation as indicated by comparing E2 polysome vs. vehicle polysome transcripts (Fig. 2) and by the distribution of translation states in each condition (Supplemental Fig. 3). When examining the translation state of E2-stimulated transcripts, we noticed that the translation state was higher in the vehicle control for E2-stimulated transcripts than for E2-repressed transcripts. We considered RefSeq genes measured in Affymetrix U133 plus 2.0 microarray time-course data from Carroll et al. (3) and compared the time patterns to the initial translation state observed in the vehicle control. A sampling of transcripts by qPCR suggested that 0.1% ethanol and untreated MCF-7 cells are very similar (data not shown). We thus equate vehicle-treated cells with untreated cells as equivalent to the cell state before the addition of E2. Upon E2 stimulation of MCF-7 cells, hundreds of genes are stimulated, whereas others are down-regulated, as shown in Fig. 5A. Significantly changing transcripts were selected by ANOVA with P values <0.1 and fold changes more than 1.5-fold and clustered into six distinct groups (Fig. 5A). We compared the initial translation state at the zero time point to the time-course behavior of E2-regulated genes. Figure 5B shows that stimulated transcripts (clusters 1 and 2) have high initial translation states in the absence of E2. On the other hand, transcripts that will be down-regulated by E2 in the transcriptome (clusters 5 and 6) have lower initial translation states. These observations suggest that the inherent translation state of transcripts predicts the direction that E2 will drive expression in total RNA. Comparing the polysome to total (P/T) ratio indicates the fraction of the cellular RNA engaged with active ribosomes. Recent data suggest that total RNA is significantly different from cytoplasmic RNA (59). An alternative indicator of the translation state is the P/M ratio. Similar trends are observed, albeit with slightly more moderate biases as indicated by the higher, but still significant P values, as shown in Fig. 5C. The P/M ratio reflects how each transcript fractionates in the sucrose gradient and is a refined metric of translation state and indicates the ratio of transcripts that are before or after initiation of translation. The difference between the P/M and P/T distributions may indicate that P/T is sensitive to all processes that can affect polysome and total mRNA levels including presence in the cytoplasm, whereas P/M is more specific for translation state. Therefore, the reduced bias seen in P/M compared with P/T may indicate the importance of other steps such as nuclear export in controlling mRNA levels. In sum, these data suggest that under these cell conditions, E2 increases the overall expression of some genes by increasing their transcriptome levels, leading to higher levels in both polysomes and monosomes. On the other hand, transcripts down-regulated in the transcriptome have low initial translation states yet shift toward higher translation states, suggesting that transcriptome and translational responses may be anti-correlated.

We expected genes stimulated by E2 in total RNA would show higher translation states after E2 stimulation, reflecting potentiation. To determine how E2 may affect the global translation state, we compared the translation state before and after E2 treatment. Stimulated transcripts in clusters 1 and 2 show a very modest decrease in translation state, whereas significantly down-regulated transcripts in clusters 5 and 6 show a modest increase in translation state (Supplemental Figs. 9 and 10). Similar patterns are observed for both P/T and P/M definitions of translation state. The up-regulated transcripts are already shifted to a much higher translation state before E2 treatment and therefore may not be able to have a further significant increase (Supplemental Fig. 9, clusters 1 and 2). On the other hand, the down-regulated transcripts start at a lower translation state and then have the potential to be driven into a higher translation state (Supplemental Fig. 9, clusters 5 and 6). Similar trends are seen for all transcripts that are in high and low translation states (Supplemental Fig. 11). Therefore, although total RNA levels may be decreasing for the transcripts in clusters 5 and 6, the overall expression may not be down-regulated because the translation state is increasing, counteracting the transcription effects. These observations suggest that the translation state is not correlated with expression level but is an independent indicator of gene expression, consistent with the well-known low correlation between total RNA and protein expression levels (69).

To understand why MCF-7 cells appear to be trying to translate E2-stimulated transcripts before the addition of E2, we investigated whether there is functional organization to these observed patterns. It is possible that mRNAs primed for E2 stimulation with higher translation states are transcripts that MCF-7 cells likely need to survive when hormone deprived and growth arrested. We examined gene ontology (GO) enrichment using GOStat (70) to compare the genes in each cluster to all 3273 E2-regulated genes measured on both array experiments. Up-regulated genes in clusters 1 and 2, with higher translation states before the addition of E2, are enriched for genes involved in cell growth and survival including the cell cycle, ribosome biogenesis, and DNA replication (Supplemental Table 4). Only modest enrichment of vacuole, lysosome, and membrane-associated transcripts in clusters 5 and 6 is observed. Clusters 3 and 4 do not show any significant enrichment. In sum, these observations suggest that the translation state predicts or anticipates the subsequent direction of expression upon E2 treatment.

Discussion

Our findings suggest that E2 uses multiple mechanisms to coordinate transcription and translation to stimulate ER⁺ breast cancer cell growth. They also illustrate the value of genomic analysis of polysome profiling to identify novel factors and networks important in the regulation of breast cancer cell proliferation. Together, these findings highlight the importance of understanding translation regulation as a key contributor to hormonal regulation of proliferation in breast cancer.

Our approach focuses on changes occurring within 1 h of E2 stimulation, allowing detection of several E2-regulated genes and programs at the expression and translation levels. This early time point increases the probability that these observations are more directly controlled by E2. The mechanisms leading to translation control by E2 are not yet clear. This could be regulated by either rapid transcriptional control of key regulators such as MYC, which promotes ribosome biogenesis and translation (71). It could also be through rapid induction of signaling pathways including mTOR, whose targets are reported to be rapidly phosphorylated within 10 min in MCF-7 cells, likely leading to increased polysome association and stimulation of some transcripts (35). Global initiation factors appear to regulate specific transcripts as recently suggested for eIF4E, which stimulates ribosomal proteins and other transcripts critical for cell growth more quickly than others, suggesting an element of specificity (45). E2 increases polysome association for transcripts involved in energy metabolism and the ERK pathway without significantly affecting transcriptome levels, suggesting that these pathways are regulated translationally.

We used multiple approaches to examine translation regulation including two measures of the translation state. The P/T and P/M ratios indicate the fraction of each transcript in the cell associating with polysomes. The P/T ratio can be influenced by transcription, stability, and subcellular location. This metric indicates how the expression level in the transcriptome compares to transcript levels associating with polysomes. The P/M ratio is a more direct measure of translation state because it indicates how each transcript migrates in the sucrose gradient in preinitiation (monosomes) and postinitiation (polysome) complexes. Similar global trends are seen for both analyses as shown in Fig. 4. Stronger biases are seen in the P/T comparison, perhaps because the total RNA is starting at lower or higher expression values (Supplemental Fig. 7), which affects the P/T ratio. This likely reflects how transcriptome levels are uncoupled from translational responses because the polysome levels do not necessarily follow the transcriptome levels. The genes that appear to be E2 up-regulated in total RNA appear to have a combination of biases including low starting expression compared with down-regulated genes (Supplemental Fig. 7) and high translation states compared with down-regulated genes (Fig. 4). Thus, translational profiling may be a general approach to identify E2-regulated transcripts, even those that may be directly transcriptionally controlled by ER. Proving ER's role in driving the observed amplification of predicted ER gene sets (Table 1 and Fig. 3B) would require additional experiments such as translational profiling to test the role of ER specifically.

A second general approach to determine how translational responses and transcriptome levels are being regulated by E2 is to compare changes in polysomes to changes in the total RNA as highlighted in Table 1. The polysome fold change vs. total RNA fold change and the translation state approach each are subject to different experimental noise and caveats. We therefore present both, similar to past studies (23). Candidate translationally regulated pathways are those showing significant regulation in both approaches such as the ERK pathway. These data strongly suggest that specific E2 translational regulation is a key component of E2's ability to drive breast cancer cell growth.

Polysome analysis reveals apparent increased sensitivity to detect E2-regulated gene sets compared with total RNA. Transcriptional mechanisms have been shown to influence posttranscriptional processes through control of poly(A) tail lengths, alternative 3′ processing, and formation of specific messenger ribonucleoprotein that can all influence and enhance translational regulation (72,73). Some E2-stimulated genes, including many potentially directly regulated by ER, are likely transcriptionally stimulated yet are detectable more significantly in polysomes than in total RNA at an early 1-h time point. This type of enhanced expression in polysomes compared with total RNA has been reported in yeast, where coordinated transcriptome and translation was found to be part of a transcription factor's regulation of gene expression in the stress response (48). Thus, not only do these transcripts show correlated up-regulation at the transcriptome and translation levels, but they also appear to amplify the degree of change in polysomes more significantly compared with other genes in the transcriptome. Although the mechanism of polysome but not total RNA enrichment is not clear, previously known key regulators of E2 action are more significantly enriched in polysome RNA compared with total RNA (Tables 1 and 3). This increased sensitivity may be indicative of an amplification of small transcriptome changes. Additional time points may provide further insight into the coordination between transcriptome and translational responses. At later times, more global increases in translation state may be observed. This would suggest that translational responses and transcriptome changes respond to E2 with different kinetics. Based on the observations at the 1-h time point, it seems likely that a variety of translation state responses would also be observed at later time points. Another possibility is an amplification model where small changes in total RNA levels lead to much larger changes in polysomes for some transcripts. Although true for some pathways such as glycolysis, this is not always the case, as seen for transcripts coding for the CREB pathway. This amplification model highlights how transcription and translation can be sensitive to E2-stimulated gene expression to different degrees at each time point.

We can take advantage of this increased sensitivity in polysomes to detect novel networks regulated by E2. We tested the role of NRSF in E2 action and found it to be necessary for maximal E2 stimulation of the cell cycle. We suspect that NRSF was not previously seen in gene expression studies because it fell below the sensitivity of total RNA comparisons but was detectable in polysomes. NRSF has previously been implicated as both a tumor suppressor and oncogene in tumors (51). Here, we find NRSF to be promoting E2-driven growth. NRSF has been found to be a tumor suppressor in rich-media conditions in ER⁻ and ER⁺ cell lines (63) as well as in a RNA interference screen (62). Our results differ for ER⁺ breast cancer cells, perhaps because we focus on the response to E2 under hormone-reduced conditions where E2's effects are isolated. Moreover, specific genetic backgrounds are known to lead to NRSF having oncogenic or a tumor suppressor activity (50,51). These in vitro findings correlate with NRSF mRNA expression levels in breast tumors. As a putative transcriptional repressor, higher NRSF would lead to suppression of the expression of many neuron-specific genes. Further investigation of the NRSF gene expression program in breast cancer would address these apparent discrepancies linking neuroendocrine systems, cancer biology, and estrogen action. NRSF may be developed as a breast cancer biomarker that could improve breast cancer treatment and diagnosis. Investigating the NRSF network may also identify key novel genes important for breast cancer growth and E2 action.

One of the most striking observations is that E2-regulated genes are biased toward higher translation states primed before the addition of E2. Transcripts that will be stimulated by E2 are biased toward polysomes, whereas those that will be repressed are biased away from polysomes. These data suggest that the translation state of many transcripts anticipates the eventual expression changes induced by E2. This coordination may be mediated by sequence elements in either the 5′- or 3′-untranslated region; however, we have yet to identify a specific sequence enriched in the data. Gene ontology enrichment analysis suggests that many of the E2-stimulated, high-translation-state transcripts include cell cycle regulators and DNA replication machinery likely required for MCF-7 cells to reenter the cell cycle. Preliminary data suggest that at longer E2 treatment times where many genes are up-regulated by E2, no significant changes in translation state are observed, consistent with the translation state already being very high for most of these mRNAs (data not shown). Even under these cell conditions, the common expectation would have been that E2 would stimulate both the expression and translation state. Instead, the translation state of most E2-stimulated transcripts is already very high before the addition of E2. This could be because the cells are primed for the addition of a mitogen like E2 in an effort to survive during hormone deprivation. Alternatively, this observed priming of the translation state may be due to incomplete E2 deprivation during the 3 d of CDT treatment. The translation state could be very sensitive to low-level E2 signaling leading to the observed high translation states before the addition of E2. It is known that longer estrogen deprivation can further reduce E2 signaling. We used these short-term deprivation conditions to be able to compare with the vast array of previous genomic studies that have focused on transcription control including expression and chromatin immunoprecipitation experiments (3,4,52,53,54). These short-term deprivation experiments suffer from incomplete removal of estrogen signaling, which requires longer, 2-wk hormone withdrawal. However, long-term deprivation may lead to complications from selecting cell populations that may be resistant to E2 (74). We do not know whether longer hormone deprivation may reduce the translation state, which would allow E2 to then stimulate the translation state of E2 transcriptionally regulated genes. These same E2-stimulated genes appear to have lower translation states in log-phase MCF-7 cells grown in 10% fetal bovine serum (FBS) with phenol red medium compared with 3-d CDT-treated, hormone-deprived cells (data not shown). This suggests that the whole balance of transcription and translation for E2-regulated genes is significantly different in hormone-deprived, growth-arrested cells compared with log-phase growing MCF-7 cells. It may also imply that the translation state is sensitive in unknown ways to a combination of E2 and other growth factor signaling driving MCF-7 growth in rich media. These findings are currently being further investigated to understand the relationship between E2 signaling and translational responses.

The classic model of transcription and translation coordination, potentiation, predicts that E2 would stimulate both the transcriptome and translation state. Our findings suggest that under these experimental conditions, used for the majority of expression profiling and transcription analysis of E2 action, E2-stimulated transcripts already have a high translation state before the addition of E2. Thus, increased expression for transcriptionally up-regulated transcriptions appears to be driven by mass action where increased transcriptome levels lead to higher levels in polysomes and not a change in translation initiation rates for these mRNAs.

In sum, we demonstrate the power of translation profiling to gain insight into the coordination of transcription and translation as a new approach to identify factors important for steroid hormone action. Together, these findings also suggest that translation control may be particularly sensitive to cell conditions and E2 signaling. Translational profiling can reveal novel E2-regulated genes and pathways. We believe that higher sensitivity in the polysomes and amplification of gene expression changes in the polysomes can help identify novel factors important for E2 signaling as exemplified by NRSF. Our data reveal novel mechanisms of E2 control of breast cancer cell growth, including specific translation regulation and surprising mechanisms to coordinate transcription and translation regulation of gene expression. Further investigation to understand how different estrogens and antiestrogens may coordinate transcription and translation may uncover novel mechanisms important to understand the full gene expression picture and regulators of cell state that may be relevant to breast cancer in vivo.

Materials and Methods

Antibodies, Western blots, and cell culture

MCF-7 cells were cultured in DMEM (Mediatech, Manassas, VA) with 10% FBS (HyClone, Logan, UT), 100 U/ml penicillin, and 100 mg/ml streptomycin (GIBCO, Carlsbad, CA). Cells were hormone deprived in phenol red-free DMEM with 5% CDT FBS (Omega Scientfic, Tarzana, CA), 100 U/ml penicillin, and 100 mg/ml streptomycin. Cells were treated with E2 (Sigma-Aldrich, St. Louis, MO) or 4-hydroxytamoxifen (Sigma-Aldrich), anti-NRSF antibody [Millipore Corp., Billerica, MA (07-579)], anti-β-tubulin antibody [Abcam Inc., Cambridge, MA (ab6046)], and antirabbit IgG and horseradish peroxidase (Cell Signaling Technology, Inc., Danvers, MA).

Determination of protein synthesis rates

A total 100 μCi EasyTag EXPRESS³⁵S protein labeling mix (PerkinElmer, Waltham, MA) was added to hormone-deprived MCF-7 cells. Protein lysate was trichloroacetic acid/acetone precipitated, resuspended in guanidinium-HCl and added to scintillation fluid for counting. Counts were normalized to microliters of extract, and percentage of trichloroacetic acid incorporation was determined by dividing the trichloroacetic acid precipitate counts per minute per milliliter extract by total counts per minute per milliliter extract.

Sucrose gradient density centrifugation and RNA analysis

To prepare polysomes, MCF-7 cells were hormone deprived for 3 d before addition of 10 nm E2. Cycloheximide (CHX) (100 μg/ml) was added for 3 min, medium was removed, and the cells were washed with PBS with 100 μg/ml CHX. Lysis buffer [15 mm Tris-HCl (pH 7.4), 10 mm MgCl₂, 300 mm KCl, 100 μg/ml CHX, and 1% Triton X-100] was added. Equivalent OD₂₆₀ units of total lysate was separated on a 10–45% continuous, linear sucrose gradient prepared with lysis buffer without Triton X-100, prepared with a Gradient Master (BioComp Instruments, Fredericton, New Brunswick, Canada), by ultracentrifugation in a SW40 Ti rotor (Beckman Coulter, Brea, CA) for 3 h at 36,000 rpm at 4 C. Polysome profiles were collected with a Brandel gradient fractionator (Gaithersburg, MD) and determined by monitoring RNA absorbance at 254 nm. Fractions 3–6 were pooled to represent the subpolysomes, and fractions 7–13 were pooled to represent the polysomes. For puromycin experiments, 1 mm puromycin (Sigma-Aldrich) was added to the cells for 5 min and to the lysis buffer for 15 min at 37 C while mixing before ultracentrifugation.

Microarrays and analysis

RNA was purified by precipitation with guanidinium-HCl and ethanol followed by an RNeasy column (QIAGEN, Valencia, CA) according to manufacturer's instructions including treatment with deoxyribonuclease I. RNA quality was determined on an Agilent 2100 Bioanalyzer (Agilent Technologies, Inc., Santa Clara, CA). All RNA processing, array hybridizations, and scanning were performed by the Brown University Center for Genomics and Proteomics core facility. RNA was prepared for hybridization using Affymetrix standard protocols and applied to Human Gene 1.0 ST microarrays (Affymetrix, Santa Clara, CA). We quantile normalized all 18 arrays and then used the Affymetrix Probe Logarithmic Intensity Error (PLIER) algorithm to generate signal estimates for all RefSeq genes. To select actual signal, we discarded those transcripts belonging to the lower quartile of their respective datasets. Analysis of significantly changing genes was determined using R (http://www.r-project.org/). Gene attributes were downloaded from UCSC Genome Browser (75). Complete microarray data have been deposited at the National Center for Biotechnology Information's Gene Expression Omnibus (accession no. GSE18592).

To identify gene expression patterns in the U133 time-course data, we applied a correlation analysis to separate transcripts into groups with similar expression profiles as previously reported (76). Briefly, each probe set expression profile was normalized by subtracting its mean and dividing by its sd. All normalized expression values of all probe sets in the class in question were then pooled together, and the mean and sd was computed at each time point.

Quantitative real-time PCR

Equal amounts of RNA from polysomes or total RNA were reverse transcribed using Superscript III and random hexamers (Invitrogen, Carlsbad, CA). Resulting cDNA was renormalized using Quant-iT PicoGreen (Invitrogen) before mixing with primers and Power SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA). Reactions were performed in an Applied Biosystems 7900HT Fast Real-Time PCR system. The fold change for total vs. total RNA was calculated, using spiked-in luciferase mRNA for normalization. To control for differential masses and purification yields from polysomes, luciferase mRNA was spiked in as a normalization control. Primer sequences are list in Supplemental Table 5.

siRNA knockdown, cell cycle, and apoptosis assays

Cells were plated in phenol red-free medium containing 5% CDT FBS at 450,000 and 10,000 cells per well in six- or 96-well plates, respectively. Silencer siRNAs targeting NRSF and control were purchased from Ambion/ABI (Austin, TX). Three siRNAs (siRNA ID 115696, 115695, and 107778) were pooled, and 75 nm total concentration was transfected using DharmaFECT 1 (Dharmacon, Inc., Lafayette, CO) according to the manufacturer's instructions. After 48 h, cells were treated with E2 or vehicle (0.1% ethanol). Cell cycle was determined by propidium iodine staining, using a BD FACSCalibur flow cytometer (BD, Franklin Lakes, NJ). FACS data were modeled using ModFIT LT software (Verity Software House, Topsham, ME). For the DAPI assay, cells were serum starved in phenol red medium with no FBS for 24 h before treatment with tamoxifen for 24 h. Cells were fixed in paraformaldehyde, stained with DAPI, and imaged with a Zeiss Axiovert 200M fluorescence microscope.