Research Resource: Expression Profiling Reveals Unexpected Targets and Functions of the Human Steroid Receptor RNA Activator (SRA) Gene

Abstract

The human steroid receptor RNA activator (SRA) gene encodes both noncoding RNAs (ncRNAs) and protein-generating isoforms. In reporter assays, SRA ncRNA enhances nuclear receptor and myogenic differentiation 1 (MyoD)-mediated transcription but also participates in specific corepressor complexes, serving as a distinct scaffold. That SRA RNA levels might affect some biological functions, such as proliferation, apoptosis, steroidogenesis, and myogenesis, has been reported. However, the breadth of endogenous target genes that might be regulated by SRA RNAs remains largely unknown. To address this, we depleted SRA RNA in two human cancer cell lines with small interfering RNAs and then assayed for changes in gene expression by microarray analyses. The majority of significantly changed genes were reduced upon SRA knockdown, implicating SRA RNAs as endogenous coactivators. Unexpectedly, only a small subset of direct estrogen receptor-α target genes was affected in estradiol-treated MCF-7 cells. Eight bona fide SRA downstream target genes were identified (SLC2A3, SLC2A12, CCL20, TGFB2, DIO2, TMEM65, TBL1X, and TMPRSS2), representing entirely novel SRA targets, except for TMPRSS2. These data suggest unanticipated roles for SRA in glucose uptake, cellular signaling, T₃ hormone generation, and invasion/metastasis. SRA depletion in MDA-MB-231 cells reduced invasiveness and expression of some genes critical for this process. Consistent with the knockdown data, overexpressed SRA ncRNA coactivates certain target promoters and may enhance the activity of some coregulatory proteins. This study is a valuable resource because it represents the first genome-wide analysis of a mammalian RNA coregulator.

This study demonstrates that knockdown of SRA RNAs affects endogenous gene profiles in two human cancer cell lines and suggests new functions for SRA.

Only 1% of the human genome is estimated to encode protein (1), but about 93% may be transcribed (reviewed in Ref. 2). Originally thought of as junk DNA, these sequences are now known to encode functional noncoding RNAs (ncRNAs). An interesting class of these molecules consists of regulatory long ncRNAs (lncRNAs), which play various roles, including telomere maintenance, imprinting, and X-chromosome inactivation (reviewed in Refs. 1 and 3).

A newly asserted function for some of these lncRNAs is their ability to serve as transcriptional regulators (extensively reviewed in Refs. 3 and 4). These ncRNAs can act either in cis to regulate neighboring genes (e.g. during imprinting and X-inactivation, and RNAs generated from upstream promoter regions) or in trans. Conceptually similar to DNA-binding transcription factors and coregulatory proteins, lncRNAs can activate or repress their target genes as part of multiprotein complexes, and they can act at various points in the transcriptional process, from affecting basal machinery (e.g. Alu), transcription factor function (e.g. Evf-2), to being part of epigenetic regulatory complexes (e.g. imprinting and X-inactivation RNAs).

The steroid receptor RNA activator (SRA) was the first described mammalian trans-acting lncRNA, cloned in 1999 (5). It was shown to be a spliced, polyadenylated transcript that enhances liganded steroid receptor transcriptional activity on reporter genes and was present in steroid receptor coactivator (SRC)-1/SRC-2 coactivator-containing complexes (5,6). Five distinct stem-loop structures in SRA were identified as being important for coactivation (7). Since then, SRA has been reported to functionally interact with a larger variety of proteins, including other nuclear receptors (NRs) (e.g. RARγ) (8), myogenic differentiation 1 (MyoD) (9), and corepressors (10,11), leading to the proposal that this lncRNA acts as a scaffold for assembly/stability of coregulator complexes.

SRA makes direct contact with some of these proteins to execute transcriptional regulation (reviewed in Refs. 12 and 13). SRA binds the DEAD-box p68 and p72 RNA helicases, which act as coactivators of estrogen receptor-α (ERα) and MyoD by recruiting SRC-1/SRC-2 and Brg-1, respectively. SRA also interacts with thyroid receptor (TR) and orphan receptors Dax and steroidogenic factor-1 (14), leading to transcriptional activation of TR-dependent reporters and some steroidogenic genes. Two pseudouridine synthases (Pus1 and Pus3) bind SRA RNA and modify some of its uridines, leading to enhanced coactivation of NRs. SRA also binds the RNA recognition motifs of two NR corepressors (SHARP and SLIRP), resulting in their competition with coactivators for promoter recruitment. Although SRA’s coregulator role in NR- and MyoD-mediated transcription is documented, a remaining unanswered question is whether different SRA-containing complexes may affect other biological pathways and target genes.

SRA has some reported physiological functions, but identification of downstream target genes important for these phenotypes is still largely insufficient. SRA RNA is overexpressed in breast, uterine, and ovarian tumors (15,16) and affects the growth of certain hormone-dependent breast and prostate cancer cell lines (17,18). Recent publications also suggest that SRA may promote myogenesis (9) and steroidogenesis (14). Transgenic mice overexpressing human SRA in their mammary glands display increased epithelial hyperplasia, but without developing palpable tumors, due to compensating enhanced apoptosis (16). Interestingly, SRA also appeared to affect brown adipose tissue (BAT) function in these glands.

In this study, we address the breadth of endogenous target genes that might be regulated by SRA RNAs. To identify possible target genes, SRA RNA was depleted in two different human cell lines by transfection of small interfering RNAs (siRNAs), followed by genome-wide analysis using cDNA microarrays and subsequent reverse transcription-quantitative PCR (RT-qPCR) validation of select targets. Additionally, we tested the effect of SRA depletion on estradiol (E2)-regulated genes in MCF-7 breast cancer cells. SRA appears to play a bigger role in coactivation than corepression, because the majority of significantly changed genes were reduced. Unexpectedly, given SRA’s published role as an ERα coactivator, we found only a small subset of its direct targets being affected. Loss of SRA impacted the expression of bona fide target genes implicated in glucose uptake, cellular signaling, T₃ hormone generation, and invasion/metastasis. Consistent with the function of the latter set of genes, SRA depletion reduced invasiveness of MDA-MB-231 cells. Exogenously expressed SRA ncRNA coactivates certain target gene promoters in luciferase assays, and consistent with the genome-wide data, SRA may augment the intrinsic activity of some coregulatory proteins [e.g. peroxisome proliferator-activated receptor-γ (PPARγ) coactivator-1α (PGC-1α) and SRC-3].

Results

Strategy to identify possible target genes of SRA RNAs in HeLa and MCF-7 cells

SRA RNA has previously been shown to coactivate the glucocorticoid receptor (GR) and ERα in transient transfection assays (5). The primary SRA transcript can be alternatively spliced to generate protein-coding isoforms (19), but unlike SRA RNA, the function of SRA protein (SRAP) remains completely unclear. To identify transcripts that were altered upon decrease of SRA, we depleted SRA by transiently transfecting siRNAs into HeLa and MCF-7 cells, cells expressing GR and ERα, respectively, and analyzed the affected genes by cDNA microarray analysis (see Fig. 1 schematic). A secondary goal of this study was to examine what effect the loss of SRA RNA had on ERα target genes, so SRA knockdowns in MCF-7 cells were followed by either media addition without E2 (−E2) or media containing 10 nm E2 (+E2) being added to cells for 6 h (Fig. 1B). The overall experimental design was 2 × 3 [either nontargeting (NT) siRNAs or SRA-targeting siRNAs transfected into cells x three independent biological replicates] for a total of 18 RNA samples for subsequent cDNA synthesis and cRNA labeling. cRNAs were hybridized to Affymetrix (Santa Clara, CA) human genome U133A2.0 chips (Fig. 1), and microarray data were analyzed as detailed in Materials and Methods to obtain lists of probe sets that were significantly affected by loss of SRA [i.e. having a false discovery rate (FDR) ≤ 0.05]. Select genes from these lists were subsequently tested for confirmation by RT-qPCR (Fig. 1).

Figure 1.

Design of global expression profiling of genes affected by SRA knockdown in HeLa and MCF-7 cells. A, Flow chart depicting gene expression assays of HeLa cervical carcinoma cells transiently transfected in triplicate with either a NT siRNA pool or a SRA-targeting Smartpool (25 nm) for 3 d before total RNA was isolated. SRA RNA knockdown was checked by RT-qPCR (Fig. 2A). Labeled cRNAs were hybridized to Affymetrix human genome (HG) U133A2.0 microarrays and processed, and then data were analyzed as detailed in Materials and Methods. RT-qPCR assays were performed for confirmation testing of select genes that were significantly different after microarray data analysis (FDR < 0.05) (Fig. 2B). B, Flow chart depicting gene expression assays of MCF-7 breast carcinoma cells (grown in charcoal-stripped, phenol red-free medium) transiently transfected in triplicate with either NT or a SRA-targeting Smartpool (50 nm). After 2.5 d knockdown, either fresh stripped medium or medium containing 10 nm E2 was added to the cells for 6 h, and then total RNA was isolated. SRA RNA knockdown was checked by RT-qPCR (Fig. 3A). The efficacy of the added E2 was checked by RT-qPCR assays of two well-known ERα target genes (TFF1 and GREB1, Fig. 3A). Labeled cRNAs were then hybridized to Affymetrix HG U133A2.0 microarrays and processed, and then data were analyzed as detailed in Materials and Methods. RT-qPCR assays were performed for confirmation testing of select genes that were significantly different after microarray data analysis (FDR < 0.05) (Figs. 3C and 4).

Efficient knockdown of SRA RNAs in HeLa and MCF-7 cells

A critical step in our experimental strategy was to obtain efficient knockdown of SRA RNAs before performing microarrays. We tested different siRNAs directed at SRA to determine the most efficient for knockdown. We decided to use pools of four siRNAs (Smartpools from Dharmacon, Lafayette, CO), because pooling siRNAs together can result in fewer off-target effects than using a single siRNA duplex at the same concentration (20). We found that transfecting HeLa and MCF-7 cells with a custom Smartpool (siRNA sequences shown in Supplemental Table 1, published on The Endocrine Society’s Journals Online web site at http://mend.endojournals.org) resulted in good knockdown of SRA RNA as compared with the negative NT control pool, as seen in multiplex conventional RT-PCR using β-actin mRNA as an internal control, and also Western blotting with a monoclonal antibody to SRAP showed that it was reduced in cells transfected in parallel (Supplemental Fig. 1; see Supplemental Materials and Methods and Supplemental References for more detail). This custom Smartpool gave much better knockdown of SRA RNA than an On-Target Plus Smartpool (data not shown).

Identification of differentially expressed genes in HeLa cells after SRA knockdown

Because siRNAs were found that resulted in efficient knockdown, HeLa cells were next transfected in triplicate with 25 nm NT or SRA siRNAs for 3 d, followed by RNA isolation for RT-qPCR to quantitate the percentage of SRA knockdown. Using β-actin mRNA as the normalizer, an approximately 95% depletion of SRA RNA was achieved (Fig. 2A), and microarrays from these triplicate RNA samples were next performed as described above. After data collection and statistical analysis, 4034 probe sets were different between the NT and SRA siRNA-transfected cells using a cutoff FDR of 0.05 or less with a fold change of 1.2 or more (see Supplemental Fig. 2A). A list of the top 30 probe sets (only 30 shown for brevity) with a more stringent cutoff (fold change ≥ 2; FDR < 0.001) is shown in Table 1. SRA itself was detected on our microarrays and was greatly reduced as expected. Because some siRNAs can induce an interferon response (21), we checked our microarray data for induction of six classical response genes (OAS1, OAS2, MX1, ISGF3γ, IFITM1, and IFNB1) by the SRA siRNA pool but did not see any significant differences, as compared with the NT pool, suggesting that changes we observed are not artifacts of this response. We also checked whether two well-characterized GR target genes, MT2A and SGK, were affected by SRA knockdown but did not see any significant differences. To further process this huge dataset, we tested by RT-qPCR whether the changes observed by microarray were valid. Our general approach was to pick genes with the lowest FDR and highest fold change, those observed with multiple probe sets (e.g. SLC2A3), and some representing both activated and repressed genes.

Figure 2.

Confirming differential gene expression in HeLa cells after SRA knockdown. A, Checking efficiency of SRA RNA knockdown by RT-qPCR. The efficiency of SRA depletion was about 95% using a Smartpool. Data are presented using the comparative Ct method, in which β-actin mRNA was used as the normalizer, and the normalized NT siRNA transfected cell data were set to unity. The mean ΔΔCt ± se of the fold change are shown. Differences between NT and SRA siRNA-treated cells were tested for significance using a Student’s t test: ***, P < 0.001; **, P < 0.01; *, P < 0.05. B, Confirming the effect of SRA knockdown on select genes identified by microarray analysis (Table 1) by RT-qPCR. Metabolic-related genes (SLC2A3, glucose transporter; INSIG1, cholesterol synthesis regulator; and ABCA1, cholesterol transporter) and TBL1X, encoding a subunit of a corepressor complex, were significantly reduced with SRA depletion mediated by the Smartpool, consistent with the majority of genes (see Supplemental Fig. 2B). However, a few genes were significantly increased upon SRA depletion, such as the chemokine CCL20 and the antiapoptotic factor IER3.

Table 1.

The top 30 differentially expressed Affymetrix Human Genome U133 Plus 2.0 Array probe sets in HeLa cells upon SRA knockdown

Gene symbol	Description	FDR	Geometric mean of intensities with SRA siRNA	Geometric mean of intensities with NT siRNA	Ratio of means (SRA/NT)
SRA1,	Steroid receptor RNA activator 1	<1 × 10−7	67.7	610.2	0.1
SRA1,	Steroid receptor RNA activator 1	<1 × 10−7	47.9	712.1	0.1
GLCCI1	Glucocorticoid-induced transcript 1	<1 × 10−7	130.7	334.4	0.4
TAF4	TAF4 RNA polymerase II, TATA box binding protein (TBP)-associated factor, 135 kDa	<1 × 10−7	17.1	74.2	0.2
CALR	Calreticulin	<1 × 10−7	361.9	998.7	0.4
SLC2A3,	Solute carrier family 2 (facilitated glucose transporter), member 3	<1 × 10−7	103.8	332.9	0.3
MKLN1	Muskelin 1, intracellular mediator containing kelch motifs	<1 × 10−7	20.9	59.8	0.3
CPM	Carboxypeptidase M	<1 × 10−7	105.0	239.1	0.4
CCL20,	Chemokine (C-C motif) ligand 20	0.0002	125.3	43.1	2.9
QKI	Quaking homolog, KH domain RNA binding (mouse)	0.0002	277.1	614.6	0.5
SLC2A3,	Solute carrier family 2 (facilitated glucose transporter), member 3	0.0002	151.9	325.0	0.5
TEGT	Testis enhanced gene transcript (BAX inhibitor 1)	0.0002	2246.3	4583.6	0.5
NAGA	N-acetylgalactosaminidase, α	0.0002	107.6	236.4	0.5
AOF1	Amine oxidase (flavin containing) domain 1	0.0002	39.0	87.4	0.4
LOC647979	Hypothetical protein LOC647979	0.0002	122.0	264.0	0.5
ATN1	Atrophin 1	0.0004	77.3	205.4	0.4
SLC2A3,	Solute carrier family 2 (facilitated glucose transporter), member 3	0.0004	158.3	327.7	0.5
SLC5A3	Solute carrier family 5 (inositol transporters), member 3	0.0004	37.9	93.6	0.4
CHD2	Chromodomain helicase DNA binding protein 2	0.0004	9.3	19.1	0.5
MAN2A1	Mannosidase, α, class 2A, member 1	0.0004	319.5	626.1	0.5
TBL1X	Transducin (β)-like 1X-linked	0.0004	120.4	265.8	0.5
LIFR	Leukemia inhibitory factor receptor α	0.0004	199.4	408.8	0.5
SFRS11	Splicing factor, arginine/serine-rich 11	0.0004	31.1	63.0	0.5
ABCB10	ATP-binding cassette, sub-family B (MDR/TAP), member 10	0.0005	96.1	191.1	0.5
TTC3	Tetratricopeptide repeat domain 3	0.0006	23.1	50.1	0.5
SLC5A3	Solute carrier family 5 (inositol transporters), member 3	0.0006	46.8	100.6	0.5
ABCA1	ATP-binding cassette, subfamily A (ABC1), member 1	0.0006	120.2	235.0	0.5
NCOA2	Nuclear receptor coactivator 2	0.0006	64.3	125.9	0.5
LIX1L	Lix1 homolog (mouse)-like	0.0007	66.8	133.6	0.5
IGF2BP2	IGF-II mRNA binding protein 2	0.0008	82.5	169.5	0.5
MAP3K7IP3	MAPK kinase kinase 7 interacting protein 3	0.0009	37.9	80.1	0.5
LASS6	LAG1 homolog, ceramide synthase 6	0.0009	30.5	66.6	0.5

The top 30 probe sets, not counting SRA, with fold change of 2 or more and FDR less than 0.001 are shown. The entire dataset has been submitted to GEO.

^aProbe sets validated by RT-qPCR (by Student’s t test, P < 0.05).

^bProbe sets validated by RT-qPCR of cells depleted of SRA by siRNA SRA2 (Supplemental Fig. 3).

^cProbe sets not validated by RT-qPCR (P > 0.05).

From the set of validated genes affected by SRA depletion, we found that the glucose transporter 3 (GLUT3), SLC2A3, was significantly reduced (P < 0.05 by Student’s t test) as well as two genes whose products mediate cholesterol transport (ABCA1) and synthesis (INSIG-1, not in the top 30 probe sets) and a subunit of the SMRT corepressor, TBL1X (Fig. 2B). These effects are consistent with SRA’s published role as a coactivator, and after tabulation of the entire HeLa dataset for which the FDR of 0.05 or lower and fold change of 1.2 or higher was applied, we noted that the majority of affected genes fall into this category (Supplemental Fig. 2B). Interestingly, SLC2A3 mRNA is induced by the glucocorticoid dexamethasone (DEX) in human U20S cells (22), whereas ABCA1 is repressed by DEX in mouse liver (23).

Loss of SRA also led to an increased expression for a smaller group of genes. We confirmed the increased mRNA levels for two such genes, CCL20 and IER3 (Fig. 2B, IER3 was not in the top 30 probe sets). Interestingly, a GR regulatory element (GRE) localizes about 4.4 kb downstream of the transcriptional start site of CCL20, chemokine (C-C motif) ligand 20, and this GRE when fused to a luciferase reporter is activated by DEX in transfected cells (24). These data suggest that SRA may repress this gene by affecting GR, although future studies are needed to test this. IER3, immediate early response 3, is a known apoptosis effector protein and was tested because transgenic SRA mice display elevated proliferation and apoptosis in mammary epithelial cells (16).

To test whether another siRNA directed to SRA would affect the genes validated from the array data above, we transfected HeLa cells with 25 nm of a single SRA siRNA duplex, named SRA2 (sequence listed in Supplemental Table 1), that gave efficient knockdown in human prostate cancer cells (18). After RT-qPCR of SLC2A3, CCL20, TBL1X, and ABCA1, we confirmed that SRA knockdown by siSRA2 (∼71% effective) affected SLC2A3 and CCL20 significantly and similarly as with the above Smartpool (Supplemental Fig. 3). These data confirm that SLC2A3 and CCL20 are bona fide SRA targets.

Identification of differentially expressed genes in MCF-7 cells after SRA depletion

To further our study of potential SRA downstream target genes, similar knockdown-microarray experiments were performed in the human breast cancer line MCF-7 with or without E2 treatment to modulate ERα activity. After MCF-7 cells were transfected in triplicate with 50 nm NT or SRA siRNAs for roughly 3 d, total RNA was isolated for RT-qPCR and microarray analysis (see Fig. 1B). We confirmed that the knockdown by the SRA Smartpool was efficient (∼85%) in cells grown either in charcoal-stripped serum phenol red-free medium (−E2) or in cells treated with 10 nm E2 for 6 h (+E2) (Fig. 3A). Six hours of E2 treatment was chosen to minimize nongenomic effects. We also confirmed the efficacy of the added E2, because two well-known ERα target genes, TFF1 and GREB1 (25), were clearly induced (Fig. 3A). SRA depletion did not significantly affect either gene (P = 0.1 by Student’s t test), although TFF1 appeared increased, consistent with a previous study using an antisense oligodeoxynucleotide targeting SRA (26).

Figure 3.

Effect of SRA knockdown in MCF-7 cells on E2-regulated gene expression. A, Control for SRA knockdown and efficacy of E2 treatment (10 nm for 6 h) as assayed by RT-qPCR. The efficiency of SRA depletion was about 85% with or without E2. RT-qPCR confirmed induction of two ERα direct target genes, TFF1 and GREB1 (by Student’s t test, P = 0.003 for TTF1 and GREB1), but loss of SRA did not significantly affect their mRNA levels (P values indicated). Data are presented using the comparative Ct method, in which the normalized NT −E2 data were set to unity. The mean ΔΔCT ± se values of the fold change are shown. Differences between NT and SRA siRNA-treated cells were tested for significance using a Student’s t test: ***, P < 0.001; **, P < 0.01; *, P < 0.05. B, Comparison of E2-affected genes in the presence of NT (734) vs. SRA Smartpool siRNAs (780) reveals 394 genes in common and 386 genes specific to SRA depletion (see Venn diagram, statistical cutoff used was FDR < 0.05). Further analysis of these two sectors in terms of whether direct ERα target genes were affected by SRA depletion is presented in Supplemental Table 2. C, RT-qPCR assays of eight direct ERα target genes expression with or without E2 treatment. Seven genes are present in the 394 common sector of the Venn diagram in B, with only CYP4B1 being in the 386 sector specific to SRA siRNA (see Supplemental Table 2). These genes encode proteins with various functions, including apoptosis/angiogenesis/metastasis (THBS1), invasion/metastasis (TMPRSS3 and TMPRSS2), metabolism (ACOX2, NRIP1, and CYP4B1), and signaling (RET and RERG). P values are indicated in cases where siSRA did not significantly affect expression.

We then performed Affymetrix microarrays from NT or SRA siRNA-transfected cells with or without E2 treatment. A total of 340 differentially expressed probe sets were observed in the −E2 cells (at a FDR ≤ 0.05 and fold change ≥ 1.2) upon SRA RNA depletion, whereas 668 probe sets were found in +E2-treated cells (Supplemental Fig. 2A). A list of the top 35 or 30 probe sets (shown for brevity) with a more stringent cutoff of fold change higher than 1.5 or 1.6 for the −E2 or +E2 data are shown in Tables 2 and 3, respectively. Again, SRA was shown to be greatly reduced in both cases, and the SRA Smartpool did not induce expression of interferon response genes. We next analyzed the extent of E2-regulated genes that were affected by loss of SRA RNAs.

Table 2.

The top 35 differentially expressed Affymetrix Human Genome U133 Plus 2.0 Array probe sets in MCF-7 cells deprived of E2 upon SRA knockdown

Gene symbol	Description	FDR	Geometric mean of intensities with SRA siRNA	Geometric mean of intensities with NT siRNA	Ratio of means (SRA/NT)
SRA1,	Steroid receptor RNA activator 1	<1 × 10−7	90.6	779.6	0.1
SRA1,	Steroid receptor RNA activator 1	<1 × 10−7	134.7	656.5	0.2
DIO2,	Deiodinase, iodothyronine, type II	<1 × 10−7	454.0	1133.0	0.4
FLJ25076	Similar to CG4502-PA	<1 × 10−7	111.8	50.2	2.2
DIO2,	Deiodinase, iodothyronine, type II	<1 × 10−7	211.2	571.7	0.4
TMEM65,	Transmembrane protein 65	<1 × 10−7	290.9	575.0	0.5
THBS1	Thrombospondin 1	0.0005	122.3	256.1	0.5
ENTPD3	Ectonucleoside triphosphate diphosphohydrolase 3	0.0008	89.1	168.2	0.5
HEY1	Hairy/enhancer-of-split related with YRPW motif 1	0.0011	145.2	85.8	1.7
RDHE2	Epidermal retinal dehydrogenase 2	0.0011	94.5	52.1	1.8
TMPRSS3	Transmembrane protease, serine 3	0.0011	55.1	29.5	1.9
MME	Membrane metallo-endopeptidase	0.0015	234.6	144.3	1.6
AOX1	Aldehyde oxidase 1	0.0018	80.3	48.4	1.7
DIO2,	Deiodinase, iodothyronine, type II	0.0019	110.1	206.3	0.5
FILIP1L	Filamin A interacting protein 1-like	0.0020	93.9	170.5	0.6
C15orf48	Chromosome 15 open reading frame 48	0.0021	92.2	48.2	1.9
ACOX2	Acyl-coenzyme A oxidase 2, branched chain	0.0022	65.0	39.8	1.6
THBS1	Thrombospondin 1	0.0022	139.4	264.3	0.5
PCSK6	Proprotein convertase subtilisin/kexin type 6	0.0023	160.5	98.3	1.6
IGHα/IGHA1/IGHA2/LOC100126583	Ig heavy locus/Ig heavy constant α 1/Ig heavy constant α 2 (A2m marker)/hypothetical LOC100126583	0.0025	132.9	78.5	1.7
RET	Ret protooncogene	0.0030	135.9	82.5	1.6
PNKD	Paroxysmal nonkinesigenic dyskinesia	0.0030	638.1	381.4	1.7
ARHGDIB	ρ-GDP dissociation inhibitor (GDI) β	0.0030	81.5	50.1	1.6
CKMT1A/CKMT1B	Creatine kinase, mitochondrial 1B/creatine kinase, mitochondrial 1A	0.0030	331.6	542.3	0.6
RFX5	Regulatory factor X, 5 (influences HLA class II expression)	0.0032	422.0	664.4	0.6
SLC39A10	Solute carrier family 39 (zinc transporter), member 10	0.0035	487.0	759.3	0.6
TRIM2	Tripartite motif-containing 2	0.0035	117.2	182.2	0.6
SLC30A7	Solute carrier family 30 (zinc transporter), member 7	0.0035	65.0	102.4	0.6
GABRP	γ-Aminobutyric acid (GABA) A receptor, π	0.0035	42.3	78.5	0.5
ANKRD35	Ankyrin repeat domain 35	0.0035	142.3	241.1	0.6
SMAD2	SMAD family member 2	0.0036	136.7	214.0	0.6
UBL3	Ubiquitin-like 3	0.0039	759.5	1199.4	0.6
TIMP3	TIMP metallopeptidase inhibitor 3 (Sorsby fundus dystrophy, pseudoinflammatory)	0.0040	309.2	197.0	1.6
TIMP3	TIMP metallopeptidase inhibitor 3 (Sorsby fundus dystrophy, pseudoinflammatory)	0.0047	241.1	144.1	1.7
THBS1	Thrombospondin 1	0.0053	150.1	269.9	0.6
NPHP1	Nephronophthisis 1 (juvenile)	0.0053	47.3	30.6	1.5
TP53I11	Tumor protein p53 inducible protein 11	0.0069	123.3	206.5	0.6

The top 35 probe sets, not counting SRA, with fold change more than 1.5 and FDR less than 0.007 are shown. The entire dataset has been submitted to GEO.

^aProbe sets validated by RT-qPCR (by Student’s t test, P < 0.05).

^bProbe sets validated by RT-qPCR of cells depleted of SRA by siRNA SRA2 (Supplemental Fig. 5).

^cProbe sets not validated by RT-qPCR (P > 0.05).

Table 3.

The top 30 differentially expressed Affymetrix Human Genome U133 Plus 2.0 Array probe sets in MCF-7 cells treated with 10 nm E2 for 6 h upon SRA knockdown

Gene symbol	Description	FDR	Geometric mean of intensities with SRA siRNA	Geometric mean of intensities with NT siRNA	Ratio of means (SRA/NT)
SRA1,	Steroid receptor RNA activator 1	<1 × 10−7	89.3	807.3	0.1
SRA1,	Steroid receptor RNA activator 1	<1 × 10−7	111.9	706	0.2
DIO2,	Deiodinase, iodothyronine, type II	<1 × 10−7	189.8	533	0.4
DIO2,	Deiodinase, iodothyronine, type II	<1 × 10−7	427.1	1073	0.4
THBS1	Thrombospondin 1	<1 × 10−7	170.3	379.4	0.4
SNAI2	Snail homolog 2 (Drosophila)	0.0003	65.4	164.2	0.4
TMEM65,	Transmembrane protein 65	0.0003	35.5	81.3	0.4
TMEM65,	Transmembrane protein 65	0.0003	251.7	580.1	0.4
RFX5	Regulatory factor X, 5 (influences HLA class II expression)	0.0006	336.2	619.7	0.5
KLHL24	Kelch-like 24 (Drosophila)	0.0008	77.6	155.1	0.5
FILIP1L	Filamin A interacting protein 1-like	0.0008	68.1	143.5	0.5
GABRP	γ-Aminobutyric acid (GABA) A receptor, π	0.0008	37.1	76.5	0.5
DIO2,	Deiodinase, iodothyronine, type II	0.0008	115	224.3	0.5
FAM116A	Family with sequence similarity 116, member A	0.0011	355	662.6	0.5
CYP4B1	Cytochrome P450, family 4, subfamily B, polypeptide 1	0.0016	111.7	190.6	0.6
ACOX2	Acyl-coenzyme A oxidase 2, branched chain	0.0016	129.2	67.2	1.9
SLC2A12,	Solute carrier family 2 (facilitated glucose transporter), member 12	0.0016	65.2	135	0.5
NET1	Neuroepithelial cell transforming gene 1	0.0016	1601.6	2678	0.6
DST	Dystonin	0.0019	61.7	112.2	0.5
FILIP1L	Filamin A interacting protein 1-like	0.0019	86.3	156.9	0.6
CD44	CD44 molecule (Indian blood group)	0.002	169.4	291.6	0.6
INSL4	Insulin-like 4 (placenta)	0.0028	89.5	44	2.0
RAB21	RAB21, member RAS oncogene family	0.0031	64.5	118.4	0.5
DDEF1	Development and differentiation enhancing factor 1	0.0033	1235.2	2059.4	0.6
CRISP3	Cysteine-rich secretory protein 3	0.0033	26.4	46.8	0.6
CCL14/CCL15	Chemokine (C-C motif) ligand 14/chemokine (C-C motif) ligand 15	0.0034	57.8	33.2	1.7
STS	Steroid sulfatase (microsomal), isozyme S	0.0036	50.2	83.8	0.6
TP53INP1	Tumor protein p53 inducible nuclear protein 1	0.0036	1518.2	2821.5	0.5
FAM62B	Family with sequence similarity 62 (C2 domain containing) member B	0.0036	257.3	423.3	0.6
RRP15	rRNA processing 15 homolog (Saccharomyces cerevisiae)	0.0041	31.8	54.6	0.6
THBS1	Thrombospondin 1	0.0043	236.2	562.4	0.4
FLJ25076	Similar to CG4502-PA	0.0048	99.6	51.9	1.9

The top 30 probe sets, not counting SRA, with fold change more than 1.6 and FDR less than 0.005 are shown. The entire dataset has been submitted to GEO.

^aProbe sets validated by RT-qPCR (by Student’s t test, P < 0.05).

^bProbe sets validated by RT-qPCR of cells depleted of SRA by siRNA SRA2 (Supplemental Fig. 5).

^cProbe sets not validated by RT-qPCR (P > 0.05).

Effect of SRA knockdown on direct ERα target genes

Because TFF and GREB1 were not significantly affected by SRA depletion (Fig. 3A), this raised the question of how many ERα direct target genes are indeed affected by loss of SRA. To address this, we first determined the number of genes significantly affected (either increased or reduced at a FDR of <0.05) upon E2 treatment in cells transfected with either NT or SRA siRNAs. This analysis is depicted as a Venn diagram in Fig. 3B. In the presence of the NT siRNAs, 734 genes (not probe sets) were E2 affected, whereas 780 genes were E2 affected when the SRA Smartpool was transfected. MCF-7 cells have been extensively studied in terms of E2-regulated and ERα direct genes by combined microarray/chromatin immunoprecipitation analyses. When the sectors of the Venn diagram in Fig. 3B were compared with three of these recent studies (25,27,28), we discovered that the common sector of 394 genes contained 108 direct ERα targets, whereas the SRA siRNA-specific sector of 386 genes had only 26 (Supplemental Table 2). This analysis implies that SRA affects only a subset of the total direct targets of ERα. To further clarify this, we then categorized these genes in terms of whether SRA knockdown had any effect upon E2 treatment. Roughly 85% (22 of 26) genes in the SRA siRNA-specific sector and about 88% (95 of 108) genes in the common sector displayed no significant difference at FDR less than 0.05 (Supplemental Table 2), and RT-qPCR assays confirmed these findings for 10 genes (Fig. 3, A and C). The combined analyses suggest that SRA RNAs affect only a small subset of ERα direct target genes in MCF-7 cells.

Gene ontology (GO) analysis of SRA-affected transcripts

From the above analyses, only a subset of genes regulated by estrogen signaling was affected by SRA depletion. However, SRA RNAs may regulate other genes and pathways. To classify the biological functions of all putative SRA target genes, we did GO analysis of the HeLa and MCF-7 genes affected by SRA knockdown using the following cutoffs: a FDR of less than 0.05 for the microarrays and L-statistics (LS) and Kolmogorov-Smirnov (KS) permutation P values for the GO grouping at less than 0.005 (see Supplemental Table 3). From the analysis of the HeLa dataset, SRA RNAs may play a role in protein targeting to membranes and transporting metal ions and lipids/cholesterol as well as affecting fatty acid biosynthesis. The −E2 MCF-7 dataset suggests that SRA may regulate select genes involved in cell motility/locomotion, cAMP metabolism, and interestingly, heart contraction (e.g. see TGFB2 below). Finally, GO analysis of the +E2 MCF-7 data suggests possible SRA function in thyroid hormone metabolism (e.g. see DIO2 below) and, again, heart contraction (Supplemental Table 3).

Validated genes affected by the loss of SRA in MCF-7 cells

RT-qPCR assays were conducted to validate select MCF-7 genes, either those with known ERα binding sites (Fig. 3C) or those lacking published sites (Fig. 4), in the absence or presence of 10 nm added E2 (data summarized in Supplemental Fig. 2A). From the complete set of genes reduced by SRA Smartpool depletion in E2-deprived cells (Table 2 lists some), we found the following validated genes: deiodinase, iodothyronine, type II (DIO2); TMEM65, a transmembrane protein of unknown function; thrombospondin 1 (THBS1); TMPRSS3; TMPRSS2; NRIP1 (also called RIP140); SLC2A12 (another GLUT called GLUT12); caveolin 1 (CAV1); cytochrome P450, family 4, subfamily B, polypeptide 1 (CYP4B1); and TBL1X, ABCA1, SLC2A3, and INSIG1 as described above. Only genes that we analyzed further in this study will be described below. DIO2 encodes an enzyme that converts inactive T₄ to its active T₃ form that is the ligand for TR, and THBS1 encodes a glycoprotein that affects apoptosis, angiogenesis, and metastasis (29). TMPRSS2 encodes a serine protease, with a likely role in invasion and metastasis. CAV1 encodes a plasma membrane-bound protein that affects cellular growth and promotes invasion (30). These same genes were reduced in MCF-7 cells upon E2 treatment (Figs. 3C and 4; Table 3 lists some). Similar to the HeLa cell dataset, this class of genes reduced with the loss of SRA was predominant upon E2 treatment (Supplemental Fig. 2B), again reflecting SRA functionality as a coactivator.

Figure 4.

Testing the effect of SRA depletion on select non-ERα direct target genes in MCF-7 cells. Cells were treated with or without 10 nm E2 for 6 h. RT-qPCR data are plotted as in Fig. 3 with differences between NT and SRA siRNA-treated cells tested for significance using a Student’s t test: ***, P < 0.001; **, P < 0.01; *, P < 0.05. P value is indicated where siSRA did not significantly affect expression. A, GLUTs SLC2A3 and SLC2A12 are reduced with SRA depletion. SLC2A3 was not seen as significantly affected in the MCF-7 microarray data but was tested given the data in Fig. 2. B, TIMP3, a gene implicated in invasion/metastasis, is increased only under estrogen deprivation. C, TBL1X, a subunit of a corepressor complex, is reduced when SRA levels drop. D, TMEM65, a gene encoding a transmembrane protein of unknown function, is reduced with SRA knockdown. E, TGFB2, a cytokine with many cellular functions, is reduced with SRA depletion. F, CAV1, an inhibitor of cell growth and promoter of metastasis, is reduced with loss of SRA. G, Three genes encoding proteins with known metabolic functions are reduced with SRA knockdown. DIO2 generates active T₃ hormone, whereas INSIG1 and ABCA1 play roles in cholesterol homeostasis.

We validated four genes as being up-regulated by SRA depletion but dependent of whether E2 was added to the MCF-7 cells. Analysis of the complete set of genes affected by SRA depletion revealed that only with E2 deprivation were more genes up-regulated (Supplemental Fig. 2B). Thirteen of these up-regulated −E2 genes are shown in Table 2, and we validated the increased expression for three of them (the two signaling molecules RET and RERG, Fig. 3C) and the matrix metalloproteinase (MMP) inhibitor TIMP3 (Fig. 4B). ACOX2 was the only gene significantly up-regulated upon SRA depletion and E2 treatment that we confirmed (Fig. 3C).

To test whether SRA depletion mediated by 50 nm of the Smartpool affects more than the RNA level of validated MCF-7 cell targets, we tested the effect of SRA knockdown on THBS1 and DIO2 protein levels by Western blotting. Interestingly, both THBS1 and DIO2 protein levels were clearly reduced by SRA knockdown (confirmed by SRAP Western) when MCF-7 cells were E2 deprived but without much effect when 10 nm E2 was added to the cells for 6 h (Supplemental Fig. 4; see Supplemental Materials and Methods for more detail). Perhaps E2 affects translation or stability of these proteins, given that their mRNA levels are significantly reduced upon E2 treatment (Figs. 3C and 4G), but this is a subject for future investigations.

We also tested what effect transfection of 50 nm of the single SRA siRNA duplex SRA2 had on genes validated from the above array data (16 were tested). After RT-qPCR assays were performed, we confirmed that SRA knockdown by siSRA2 (∼81% effective, Supplemental Fig. 5A) did not significantly affect the E2-induced TFF1 level (Supplemental Fig. 5B) but significantly affected the following genes similarly as the Smartpool: SLC2A3, SLC2A12, TMEM65, DIO2, TMPRSS2, TBL1X, and TGFB2 (Supplemental Fig. 5C). These data confirm these genes are bona fide SRA targets.

SRA RNA level affects invasiveness of MDA-MB-231 cells

Because SRA Smartpool-mediated depletion in MCF-7 cells resulted in reduced THBS1, CAV1, TMPRSS2, and TMPRSS3 expression, genes associated with invasion/metastasis (29,30), these data imply SRA might affect cell invasion. To test this hypothesis, we used MDA-MB-231 breast cancer cells, instead of MCF-7, because they are much more invasive. MDA-MB-231 cells were transfected with either NT or SRA Smartpool siRNAs for 3 d and then were plated on Matrigel-covered inserts in a two-chamber, transwell system (see Materials and Methods), with the remaining cells harvested for RNA isolation for RT-qPCR assays. SRA RNAs were efficiently depleted (∼89% as compared with the NT siRNAs, Fig. 5A), correlating with significantly reduced cell invasion through Matrigel (Fig. 5B). To begin to understand a mechanism for this effect, we tested the above genes (and others) for expression changes upon loss of SRA. CAV1 and TMPRSS2 mRNAs were significantly reduced when SRA was depleted (Fig. 5C). At least 12 MMPs are known to promote MDA-MB-231 cell invasion (31), and we screened some of these genes for reduced expression. Whereas MMP-2, MMP-7, MMP-13, and MMP-14 were not affected by SRA depletion, we found significant reductions in MMP-1 and MMP-9 mRNAs (Fig. 5C and data not shown).

Figure 5.

Effect of SRA depletion on MDA-MB-231 cell invasion and invasion-related genes. A, Checking efficiency of SRA RNA knockdown by RT-qPCR. The efficiency of SRA depletion mediated by the Smartpool was about 89%. Data are presented using the comparative Ct method as described in Fig. 2. The mean ΔΔCt ± se of the fold change are shown. Differences between NT and SRA siRNA-treated cells were tested for significance using a Student’s t test: ***, P < 0.001. B, Transfection of MDA-MB-231 cells with the SRA Smartpool reduces invasion through Matrigel. Differences between NT and SRA siRNA-treated cells were tested for significance using a Student’s t test: *, P = 0.015; n = 4. C, SRA depletion in MDA-MB-231 cells reduces the expression of some genes that affect cell invasion. RT-qPCR data for four genes are presented as in A. Differences between NT and SRA siRNA-treated cells were tested for significance using a Student’s t test: **, P < 0.01; *, P < 0.05. CAV1 and TMPRSS2 were significantly reduced in MCF-7 cells upon SRA knockdown (Figs. 4F and 3C, respectively), whereas two MMPs (MMP-1 and MMP-9) were not significantly affected in the MCF-7 microarrays.

Noncoding SRA coactivates some target gene promoters

The above described effects mediated by SRA knockdown on various genes could involve both direct and indirect mechanisms and could be mediated by either SRA RNA or SRAP. To address the latter, we tested the effect of exogenously expressed SRA RNA that is noncoding and incapable of productive translation or SRA RNA that encodes protein (see Materials and Methods) on a few potential SRA target gene promoters, representing genes affecting different biological processes. If the protein is the functional effector, then only transfection with pSCT-SRAP would be able to exert an effect, whereas if SRA RNA is the effector, either vector would show an effect. After transfection and luciferase assays were performed, it was apparent that pSCT-SRA (ncSRA) and pSCT-SRAP coactivated an approximately 6.9-kb DIO2 promoter the strongest (3.7–3.8 times), followed by an approximately 2.1-kb THBS1 promoter (2.6–3.2 times) as compared with the pGL3-basic control (1.3–1.4 times) (Fig. 6A). Under these conditions, an approximately 1.23-kb SLC2A3 promoter was similarly affected as the pGL3-basic control (data not shown). However, when less SLC2A3 promoter (clone 11 or 17, see Materials and Methods) was transfected with more pSCT-SRA (ncSRA) or SRAP, we observed that both coactivated the promoter (Fig. 6B). These data suggest that SRA RNA, regardless of its ability to make protein, regulates these genes at the promoter level.

Figure 6.

Exogenously expressed ncSRA coactivates human DIO2, THBS1, and SLC2A3 gene promoters in transiently transfected HeLa cells. A, One microgram of luciferase reporter (pGL3-basic, THBS1, or DIO2) was cotransfected with 0.5 μg of an empty vector (pSCT), ncSRA (pSCT-SRA), or pSCT-SRAP (encoding both RNA and protein). After more than 40 h, whole-cell extracts were made and luciferase assays done. Relative light units (RLU) were normalized to protein amount (in micrograms) as determined by Bradford assays. pGL3-basic served as a negative control, because all promoters were cloned into this reporter. B, Titration of ncSRA results in enhanced SLC2A3 promoter activity. Two different SLC2A3 promoter clones (11 and 17, see Materials and Methods for more details) were created in pGL3-basic. Two hundred fifty nanograms of pGL3-derived luciferase reporter were cotransfected with increasing amounts (333–1000 ng) of ncSRA (pSCT-SRA) or pSCT-SRAP, and the total pSCT-1 plasmid amount was held constant at 1000 ng by adding empty vector, pSCT, where needed. Two hundred fifty nanograms of pGL3-basic served as filler in this experiment to achieve 1.5 μg total DNA.

The mechanism used by ncSRA to coactivate the above promoters is unknown. As a first step in addressing this, we performed an in silico analysis of the seven bona fide SRA target genes that were validated by both SRA Smartpool and SRA2 siRNA knockdowns in MCF-7 cells (Supplemental Fig. 5C), assaying approximately 1 kb of their promoters for overrepresented transcription factor binding site motifs using the program Pscan (32). From this analysis, the SRA targets might be regulated by zinc-finger (e.g. Gfi, Klf4, RREB1, Roaz, MZF1, and Gata1), forkhead (e.g. Foxq1), and bZIP (e.g. Cebpa) transcription factors as well as other NRs (RORA, AR, and PPARG-RXRA) (Supplemental Table 4). Because only androgen receptor (AR) has been reported to be coactivated by SRA RNA (5), the in silico analysis suggests additional DNA-binding factors that SRA might modulate.

Noncoding SRA may enhance the activity of some coregulatory proteins

SRA might alternatively affect the activity of a coregulator necessary for transcription of the above target genes. We first tested whether ncSRA might augment SRC protein activities, because SRA coimmunoprecipitates with SRC-1 and SRC-2 (5,6), by cotransfecting a Gal4-dependent reporter with Gal4-coregulator fusions (SRC-1, -2, or -3) with or without ncSRA. Luciferase assays were performed using an empty Gal4 DNA-binding domain (DBD) vector (pBind) as a negative control. Interestingly, we found that SRA coactivated only Gal4-SRC-3 (Fig. 7A). We additionally tested whether ncSRA might affect PGC-1α and cAMP response element-binding protein-binding protein (CBP) activities, given these proteins function with SRC-1 (33) and that PGC-1α has an RNA-binding domain at its C terminus (34). Although exogenous ncSRA had little effect on Gal4-CBP activity (data not shown), ncSRA substantially enhanced Gal4-PGC-1α activity (Fig. 7A) but not its protein level (Fig. 7B).

Figure 7.

Effect of exogenously expressed ncSRA on Gal4-PGC-1α and Gal4-SRC activities and on Gal4-PGC-1α protein level in HeLa cells, under the transfection conditions employed for luciferase assays. A, ncSRA enhances Gal4-PGC-1α and Gal4-SRC-3 activities but not Gal4 DBD (pBind), Gal4-SRC-1e, or Gal4-SRC-2. Seven hundred fifty nanograms of pG5-Luc, 25 ng Gal4 fusion vector, and 750 ng pSCT or ncSRA (pSCT-SRA) were cotransfected. Luciferase assays were performed as described in Materials and Methods. B, pSCT-SRA transfection does not increase Gal4-PGC-1α protein level under luciferase assay conditions. Forty micrograms of protein of transfected HeLa cell extract were run on an SDS-PAGE gel, followed by electrotransfer to polyvinylidene difluoride. Gal4-PGC-1α (calculated molecular mass ∼109 kDa) was detected as described in Materials and Methods. Molecular mass standards are listed. Duplicate transfected lysates were analyzed. Image contrast and brightness were modified uniformly throughout the entire anti-Gal4 DBD Western panel.

Discussion

This study represents the first genome-wide screen for possible target genes of a mammalian transcriptional RNA coregulator (summarized in Supplemental Fig. 2). Given that previous studies on SRA RNA showed coactivation of steroid receptors (5), we were surprised that loss of SRA affected only a small subset of E2-regulated genes in MCF-7 cells. Instead, our data show that SRA is a more general coregulator. Also, from analyses of the microarrays, new, unexpected functions for SRA RNAs are suggested. First, SRA likely modulates some distinct metabolic processes, such as glucose uptake and T₃ hormone synthesis. Second, SRA RNA levels may affect some discrete signaling pathways, such as those mediated by the chemokine CCL20 and cytokine TGFB2. Finally, altered SRA RNA levels may affect breast cancer invasion/metastasis, because loss of SRA in MDA-MB-231 cells led to lowered invasiveness.

Previous studies have reported only a handful of possible SRA downstream target genes, which is why this genome-wide study is essential. Besides three SRA target genes identified in muscle (9) and two (14) in adrenal cells, not much was known concerning SRA targets before our study. SRA depletion reduced the expression of the AR target gene TMPRSS2, a gene often found in certain translocations in prostate cancer, although not affecting two other well-known targets (18). That loss of SRA does not affect all AR target genes is consistent with our finding that SRA appears to regulate only a subset of ERα direct targets. Our data further suggest that SRA coactivates TMPRSS2 in breast cancer cells. Whether loss of SRA affects invasiveness of prostate cancer cells remains to be tested, as well as direct demonstration that reduced TMPRSS2 affects invasion. Cooper et al. (17) screened 57 genes using a breast cancer and ER RT-qPCR array for an effect when SRA intron-1 retention was increased in ERα-positive T5 breast cancer cells. Consistent with our findings in MCF-7 cells, they found only five genes significantly affected; however, only one (THBS1) was found in common with our study. THBS1 is an angiogenesis/cell growth inhibitor, and consistent with this function, altered SRA RNA levels have been reported to affect cell viability (17) and apoptosis (16).

Heart contraction was suggested by GO analyses to be affected by SRA RNA levels, because two genes (TGFB2, TGF-β2; PRKCA, protein kinase Cα) were significantly reduced by loss of SRA. TGFB2 was further validated as a bona fide SRA target gene. A recent clinical study reported the human SRA1 gene to be one of three cosegregating genes (others being HBEGF and IK) on chromosome 5 impacting cardiomyopathy and that a morpholino directed to the zebrafish (Danio rerio) SRA RNA ortholog led to myocardial contractile dysfunction in injected embryos (35). In our microarrays, loss of SRA did not significantly affect IK gene expression, whereas HBEGF was affected only in HeLa cells (1.5-fold up, FDR = 0.04). These data provide support that endogenous SRA RNAs are trans-acting, not cis-acting, regulators. Whether SRA misregulation of TGFB2 may contribute to heart disease is subject to future studies.

SRA ncRNA may play an additional role in modulating TR signaling. Before this study, it was known that TR directly binds SRA RNA, and overexpressed SRA coactivates TR-dependent reporter genes (36). We now show that the SRA RNA level modulates the expression of DIO2 (seen at the endogenous mRNA and protein level and on a transfected promoter). As previously noted, DIO2 is important for T₃ synthesis, the ligand that activates TR. Interestingly, transgenic mice expressing human SRA RNA display increased BAT in their mammary glands (16). SRA regulation of DIO2 may play a role in this phenotype, given that T₃ and DIO2 are important for BAT formation (37).

Perhaps the most interesting of SRA downstream target genes identified in this study are the GLUTs SLC2A3 and SLC2A12. They are members of a large family of GLUTs and SLC2A3 has one of the lower Michaelis-Menten constants for glucose (38). SLC2A12 has only recently been cloned and found overexpressed in breast and prostate cancers, but its role in tumorigenesis is unknown (38). Invasive cancer cells often overexpress GLUT1 and -3 (SLC2A1 and SLC2A3), because they rely predominantly on glycolysis to meet their increased energy demands, and uptake of glucose becomes critical (39), correlating with poor prognosis and decreased survival (38). Interestingly, SRA Smartpool-mediated depletion in MDA-MB-231 cells significantly reduced the SLC2A3 mRNA level but did not affect SLC2A1 mRNA (Supplemental Fig. 6), leading to the hypothesis that SRA may affect cell invasiveness by up-regulating SLC2A3 transcription for increased glucose uptake. Besides DEX inducing SLC2A3 expression, progesterone and cAMP have also been reported to increase SLC2A3 mRNA levels (40). SRA does not coactivate cAMP response element-binding protein (CREB) but does enhance GR and progesterone receptor activity (5). It remains to be seen whether SRA is critical for DEX and progesterone-induced SLC2A3 expression and whether this may affect cell invasion.

In summary, our unbiased genome-wide microarray study in human cervical and breast cancer cell lines has provided unanticipated evidence for a much wider physiological role for SRA than just simply coactivating steroid hormone-mediated transcription. We identified eight bona fide (and many other putative) target genes of SRA and provide evidence that SRA ncRNA can regulate the DIO2, THBS1, and SLC2A3 genes at the promoter level. Our findings that SRA positively regulates genes involved in growth/apoptosis and invasion/metastasis reveal possible molecular mechanisms for the observed phenotypes exhibited by transgenic mice overexpressing SRA (16) and decreased MDA-MB-231 cell invasion upon SRA depletion. New roles for SRA in modulating glucose uptake, cellular signaling, and T₃ synthesis are suggested by this study. In sum, the identified downstream SRA targets represent a new resource to guide future experiments on this novel RNA coregulator.

Materials and Methods

Human cell lines

HeLa (cervical, GR-positive), MCF-7 (breast, ERα-positive), and MDA-MB-231 (breast, ERα-negative) cancer cell lines were from the American Type Culture Collection (Manassas, VA) or the Baylor College of Medicine (BCM) Tissue Culture Core. For routine growth, they were cultured in complete medium (high-glucose DMEM with 10% fetal bovine serum and 1% penicillin and streptomycin) in a 37 C, 5% CO₂ incubator.

Plasmids and hormones

pSCT and pSCT-SRA have been described (5). pSCT-SRAP was constructed by removing the first intron of SRA left in pSCT-SRA by BamHI digestion, followed by ligation to a BamHI-digested RT-PCR product (made from HeLa cell cDNA with Pfu DNA polymerase; Stratagene, La Jolla, CA) containing the start codons of the human SRA1 gene (41). Primers for the PCR were forward 5′-tccccgggatccggaaatgacgcgctgc-3′ and reverse 5′-ccaggagaagtctctgatgc-3′, with the BamHI site underlined. There is an internal BamHI site in SRA that was used for the 3′ end. This expression vector produces SRAP in transfected HeLa cells, as detected with a monoclonal antibody (described in Ref. 5), whereas no expression was seen with pSCT-SRA (data not shown). pGL3-basic was from Promega (Madison, WI). pGL3-SLC2A3 clones 11 and 17 were made by ligating the PCR product of human genomic DNA (Novagen, Madison, WI) amplified with Platinum Taq High-Fidelity polymerase (Invitrogen, Carlsbad, CA) and the primers forward 5′-aaacggggtacccttgagactagcagaaagtg-3′ and reverse 5′-tgtccccatcgctgtaatc-3′ (with Acc65I site underlined; an internal BglII site in the SLC2A3 promoter at −30 relative to the translational start site was used for the 3′ end; similar primers were previously used for cloning the Glut3 promoter) (40) after digestion with Acc65I and BglII into similarly digested pGL3-basic. All constructs were validated by sequence analysis. Clones 11 and 17 are different subclones from the same PCR and contain 1229 and 1234 bp, respectively, of SLC2A3 promoter sequence including the BglII site, due to different lengths of poly(A) tracts in the promoter region. Other human promoters cloned into pGL3-basic were kind gifts and have been published [−2033/+66 THBS1 from Dr. Olga Stenina, Cleveland Clinic Foundation (42) and an ∼6.9-kb DIO2 from Dr. Balazs Gereben, Institute of Experimental Medicine, Budapest, Hungary (43)]. pBIND and the Gal4-luciferase reporter (pG5-luc) were from Promega. pBIND-SRC1e, pBIND-SRC-2, pBIND-SRC-3, and pBIND-mouse CBP-HA have been described (44). pCMX-Gal4-PGC-1α was a kind gift of Dr. Bruce Spiegelman, Harvard Medical School (Boston, MA) (33). Water-soluble E2 was from Sigma Chemical Co. (St. Louis, MO).

siRNA transient transfections

HeLa, MCF-7, and MDA-MB-231 cells (∼3 × 10⁵) were plated in six-well dishes in complete medium or charcoal-stripped medium (high-glucose, phenol red-free DMEM, 5% charcoal-stripped fetal calf serum, and 1% penicillin and streptomycin) and then transfected in triplicate with a custom SRA Smartpool (Dharmacon, Lafayette, CO) (see Supplemental Table 1) or a nontargeting (NT) siRNA pool (Dharmacon; D-001810-10) at final concentrations of 25 nm for HeLa cells and 50 nm for MCF-7 and MDA-MB-231 cells with Trans-IT-TKO (Mirus Corp., Madison, WI), according to the manufacturer’s protocol, for roughly 3d. E2 (10 nm) in stripped medium or medium alone (as the −E2 control) was added to MCF-7 cells after 2.5 d for 6 h.

cDNA microarrays

Total RNAs were isolated with Trizol (Invitrogen) from HeLa or MCF-7 cells transfected with siRNA pools and treated with or without 10 nm E2 for 6 h, representing three independent biological replicates. RNAs were further purified by RNeasy columns (QIAGEN, Valencia, CA).

The BCM Microarray Core Facility conducted sample quality checks using a NanoDrop ND-1000 spectrophotometer and Agilent 2100 Bioanalyzer before their labeling of cRNA and hybridizations (using the Affymetrix Genechip microarray system). Briefly, RNAs were labeled using the standard Affymetrix T7 oligo(deoxythymidine) primer protocol. Five micrograms of total RNA were reverse transcribed to produce double-stranded cDNA. The cDNA product was used as a template for an in vitro transcription reaction, producing biotin-labeled cRNA. The labeled cRNA was quantified using the NanoDrop spectrophotometer, and 15 μg of the labeled cRNA were fragmented and rechecked for concentration. A hybridization cocktail containing Affymetrix spike-in controls and labeled cRNA was loaded onto a GeneChip human genome U133 Plus 2.0 Array (containing 54,675 probe sets representing 33,765 genes). The array was hybridized overnight at 45 C with rotation at 60 rpm and then washed and stained with a streptavidin, R-phycoerythrin conjugate stain. Signal amplification was performed using biotinylated antistreptavidin. The stained array was scanned on an Affymetrix GeneChip Scanner 3000. The images were analyzed and quality control (QC) metrics recorded using Affymetrix GCOS software version 1.4.

Microarray data analysis

We used the following software packages for data QC and statistical analysis: Affymetrix Expression Console (www. affymetrix.com) for data QC (all three triplicates of the three experiments showed good quality, data not shown), BRB Array Tools (http://linus.nci.nih.gov/BRB-ArrayTools.html) (45) for statistical analysis, dChip (www.dchip.org) (46) for clustering pictures, and ArrayTrack (http://www.fda.gov/ScienceResearch/BioinformaticsTools/Arraytrack/default.htm) for Venn diagrams. Expressions were estimated with BRB Array Tools using the Robust Multichip Average (RMA) method (47). We used for the analysis approximately 40,000 probe sets present in at least 25–30% of all arrays for each of the cell lines. Because our sample size was equal to three for each of the groups, we performed t tests with the random variance model, which is designed for small sample size experiments (48). The method of Benjamini and Hochberg (49) was used for estimation of FDR. The cutoffs for differentially expressed genes were FDR of 0.05 or less and fold change of 1.2 or greater for the entire datasets (Tables 1–3 show a list of the top probe sets at a higher fold change and lower FDR cutoff for brevity). We used the Gene Set Comparison Tool from BRB Array Tools to find GO categories that are differentially expressed between NT and SRA siRNA-transfected groups (Supplemental Table 3). This tool uses the functional scoring method (50), based on predefined gene sets, representing a more powerful method of identifying differentially expressed gene classes than performing overrepresentation analysis of differentially expressed gene lists.

Microarray data described in this study comply with the MIAME standard, and the data have been submitted to the NCBI Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo) with accession no. GSE20081.

RT-qPCR

One microgram of total RNA was reverse-transcribed with random primers and Superscript II (Invitrogen) in 20-μl reactions. These reactions were diluted 1:10 with nuclease-free water (Ambion, Austin, TX). Duplicate qPCR contained 2 μl diluted first-strand cDNA, 900 nm intron-spanning primers specific for a particular mRNA (see Supplemental Table 1), 100 nm FAM-TAMRA-labeled Universal Roche (Indianapolis, IN) probe, and 1× TaqMan Universal PCR Master Mix (Applied Biosystems, Foster City, CA). Reactions (20 μl) were run in 96-well optical plates (Applied Biosystems). Average threshold cycle (Ct) values of β-actin (ACTB) mRNA (the chosen normalizer, because its level is not sensitive to E2 or SRA depletion) were subtracted from corresponding average Ct values of a target mRNA to obtain ΔCt values. A ΔΔCt value was then calculated by subtracting the ΔCt value from the NT siRNA-treated cells (for MCF-7 cells, this additionally was without E2). Relative RNA level was then expressed as 2^−ΔΔCt. Error bars represent the se of the fold change, as described (51). Differences between NT and SRA siRNA-treated cells were tested for significance using an unpaired two-tailed Student’s t test (in Microsoft Excel). P < 0.05 was considered significant.

Cell invasion assay

MDA-MB-231 cell invasion was assayed in a two-chamber transwell system containing an 8-μm cell culture insert precoated with growth factor-reduced Matrigel (BD Biosciences, San Jose, CA). A total of 100,000 cells, resuspended in serum-free medium, was added to the insert and incubated in the bottom chamber filled with 10% fetal bovine serum-containing medium. After 47 h, cells on the upper surface of the transwell were removed using cotton swabs, and the cells migrated to the bottom of the membrane were fixed with 4% paraformaldehyde in PBS (USB Corporation, Cleveland, OH), stained with 0.25% crystal violet, and counted under a microscope at ×100 magnification.

Luciferase reporter assays

HeLa cells grown in complete medium in 12-well plates were transiently transfected using Superfect (QIAGEN) as described (7), so that all transfections had the same final amount of DNA (1.5 μg) per well. All transfections were balanced with the corresponding empty vector pSCT. Amounts of expression vectors cotransfected with luciferase reporters is described in the Fig. 6 and 7 legends. Cells were harvested more than 40 h after transfection in 200 μl Cell Culture Lysis Reagent (CCLR) lysis buffer (Promega), and luciferase activity was determined relative to protein level (as assayed by Bradford assays using BSA-generated standard curves) using a luciferase reporter assay (Promega) and a Berthold 96-well plate reader.

Detection of Gal4-PGC-1α in transfected HeLa cells

Roughly 40 μg protein lysates made in the above CCLR lysis buffer were run on a 4–15% gradient SDS-PAGE gel (Bio-Rad, Hercules, CA), followed by electrotransfer to polyvinylidene difluoride (Bio-Rad). Gal4-PGC-1α was detected using a monoclonal antibody directed to the Gal4 DBD (RK5C1; Santa Cruz Biotechnology, Santa Cruz, CA), followed by incubation with sheep antimouse-horseradish peroxidase conjugate (GE Healthcare, Piscataway, NJ). β-Actin was used as a loading control and was also detected with a monoclonal antibody (A5441; Sigma).