A Conserved Protein Motif Is Required for Full Modulatory Activity of Negative Elongation Factor Subunits NELF-A and NELF-B in Modifying Glucocorticoid Receptor-regulated Gene Induction Properties*

Background: Understanding glucocorticoid receptor (GR)-regulated gene transcription requires detailed description of cofactor actions. Results: NELF subunits have new activities altering GR induction properties of exogenous and endogenous genes. Conclusion: NELF-A and NELF-B are decelerators functioning at two positions after GR and before/at the reporter gene site of action. Significance: Functional ordering of NELF-A and NELF-B relative to other factors in GR transactivation is determined. NELF-B is a BRCA1-interacting protein and subunit (with NELF-A, -C/D, and -E) of the human negative elongation factor (NELF) complex, which participates in RNA polymerase II pausing shortly after transcription initiation, especially for synchronized gene expression. We now report new activities of NELF-B and other NELF complex subunits, which are to attenuate glucocorticoid receptor (GR)-mediated gene induction, reduce the partial agonist activity of an antagonist, and increase the EC50 of an agonist during nonsynchronized expression of exogenous and endogenous reporters. Stable knockdown of endogenous NELF-B has the opposite effects on an exogenous gene. The GR ligand-binding domain suffices for these biological responses. ChIP assays reveal that NELF-B diminishes GR recruitment to promoter regions of two endogenous genes. Using a new competition assay, NELF-A and NELF-B are each shown to act independently as competitive decelerators at two steps after the site of GR action and before or at the site of reporter gene activity. A common motif in each NELF was identified that is required for full activity of both NELF-A and NELF-B. These studies allow us to position the actions of two new modulators of GR-regulated transactivation, NELF-A and NELF-B, relative to other factors in the overall gene induction sequence.

vators, corepressors, and comodulators). This behavior appears to be general in that all classical steroid receptors (androgen, estrogen, glucocorticoid, mineralocorticoid, and progesterone) and some nuclear receptors are sensitive to these factors (1)(2)(3). Furthermore, the ability of factors to adjust A max , EC 50 , and PAA is seen for both gene induction and gene repression (4 -7) and is not restricted to synthetic reporter genes but also occurs in a gene-selective manner with endogenous genes of primary human cells (i.e. peripheral blood mononuclear cells) (8). It thus is likely that alterations of the parameters of receptor-regulated gene expression are widespread and physiologically relevant.
Our understanding of how these parameters are modified is complicated by the fact that few landmarks in the sequence of steroid-regulated reactions have been determined other than steroid binding to its cognate receptor, direct or indirect binding of receptor-steroid complex to biologically active DNA sequences such as hormone response elements, and recruitment of various cofactors to increase or decrease gene transcription rates. Although the details of steroid hormone action are still poorly defined, most modulatory factors are thought to act shortly after receptor-steroid binding/tethering to DNA. However, any step can, under the appropriate conditions, influence the A max , EC 50 , and/or PAA (9,10). Not all factors will alter all three parameters (8,11). In fact, the ability to modulate A max and/or EC 50 , along with the direction of the modulation, conveys otherwise inaccessible mechanistic information (12)(13)(14). Consequently, we were attracted by reports that cofactor of BRCA1 (COBRA1 ϭ NELF-B) represses gene transcription by estrogen receptors (ERs) and androgen receptors (15)(16)(17) and binds glucocorticoid receptors (GRs) (16). NELF-B has been suggested to be a tumor suppressor (18) and an oncogene (19). NELF-B is also an integral subunit of the human negative elongation factor (NELF) complex (20,21) that is instrumental in RNA polymerase II pausing (22)(23)(24)(25)(26). Recent studies suggest that a major function of polymerase pausing is to enforce synchronized gene expression during differentiation and development (27). Chromatin immunoprecipitation experiments revealed that NELF knockdown can influence which transcriptional start site of a gene is used (28). These observations suggested that NELF-B could modify ER and androgen receptor actions at steps well downstream of receptor-steroid complex binding to hormone response elements.
The objectives of this study were to determine whether NELF-B interacts with glucocorticoid receptors (GRs) to repress transactivation and alter the A max , EC 50 , and PAA. To this end, we initially examined GR induction of a transiently transfected reporter gene, which would not be encumbered by mechanisms to ensure synchronized gene expression. Positive results with both synthetic and endogenous GR-regulated genes prompted us to investigate NELF-B effects on GR binding to enhancers. Our recently developed competition assay (13,14,29) was used to determine the kinetically determined mode and site of action of wild type and mutant NELF-B and NELF-A. We conclude that NELF-B and NELF-A are new GR cofactors that act independently, but additively, as a competitive decelerator (14), each at two different steps.

EXPERIMENTAL PROCEDURES
Unless otherwise indicated, all cell growth was at 37°C, and all other operations were performed at 0°C.
Plasmids-Renilla-TS reporter, rat GR (pSG5-GR), GREtk-LUC, pSG5-TIF2, and GAL/GR525C have been previously described (31). FR-LUC reporter is from Stratagene (La Jolla, CA). Human NELF-A (missing the first 11 residues), FLAG/ NELF-B (32), FLAG/NELF-C/D, and FLAG/NELF-E were from the University of Texas Health Science Center at San Antonio. FLAG/NELF-B (in pcDNA3) has the C-terminal 30 amino acids deleted and replaced by the C-terminal 52 amino acid of the neomycin gene. Full-length wild type (wt) NELF-B, also with an N-terminal FLAG tag, was constructed by removing the FLAG/ full-length wtNELF-B from IRES-COBRA with EcoR1/BamH1 digestion and then inserting it into the pcDNA3.1(Ϫ) vector (Invitrogen).
The 4mtNELF-B was first generated using QuikChange II XL site-directed mutagenesis kit (Agilent Technologies) in the context of the chimeric NELF-B with the following primers: L283A/L287A forward: 5Ј-GAGGGCGCGGGAGGCCCAGG-GGTTTGCCGATGGCGTCAAG-3Ј and reverse: 5Ј-CTTGA-CGCCATCGGCAAACCCCTGGGCCTCCCGCGCCCTC-3Ј; K291A/K292A forward: 5Ј-GGGGTTTCTCGATGGCGT-CGCCGCCGGCCAGGAGCAGGTGCTGG-3Ј and reverse: 5Ј-CCAGCACCTGCTCCTGGCCGGCGGCGACGCCATC-GAGAAACCCC-3Ј. The above full-length wtNELF-B plasmid was digested with EcoR1/SbfI, and the larger of two fragments was used as the vector for ligation. The shorter fragment (1023 bp) was redigested with SacII to generate two fragments, of which the second shorter fragment 226 bp from EcoR1 to SacII was isolated. The chimeric 4mtNELF-B was digested by SacII/ SbfI to produce two DNA fragments. The shorter 797-bp species containing the four mutated amino acids was incubated with the 226-bp EcoR1/ScaII fragment and the above EcoRI/ SbfI vector fragment in a ratio of 1:1:3 overnight at 16°C to afford the desired plasmid.
Cell Culture, Transient Transfection, and Reporter Analysis-Monolayer cultures of U2OS, U2OS.rGR, COS-7, and CV-1 cells were grown as described previously (31,33). T47D cells were cultured in Dulbecco's modified minimum essential medium (DMEM) supplemented with 10% fetal bovine serum (FBS). NELF-B stable knockdown T47D cell line was described previously (15). Triplicate samples of cells were seeded into 24-well plates at 20,000 cells/well and transiently transfected the following day with luciferase reporter and DNA plasmids by using 0.7 l of Lipofectamine (Invitrogen) or FuGENE 6 (Roche Applied Science) per well according to the manufacturer's instructions. The total transfected DNA was adjusted to 300 ng/well of a 24-well plate with pBlueScriptII SKϩ (Stratagene). The molar amount of plasmids expressing different protein constructs was kept constant with added empty plasmid or plasmid expressing human serum albumin (31). Renilla-TS (10 ng/well of a 24-well plate) was included as an internal control. After transfection (32 h), cells were treated with medium containing appropriate hormone dilutions. The cells were lysed 20 h later and assayed for reporter gene activity using Dual-Luciferase assay reagents according to the manufacturer's instructions (Promega, Madison, WI). Luciferase activity was measured by an EG&G Berthold luminometer (Microlumat LB 96P). The data were normalized to Renilla null luciferase activity and expressed as a percentage of the maximal response with Dex before being plotted Ϯ S.E., unless otherwise noted.
Western Blotting-Cytosols for Western blots were prepared as described below for immunoprecipitation assays, probed with rabbit anti-GR, FLAG, NELF-B, NELF-A, and NELF-C/D antibodies (described above), and visualized by ECL detection reagents as described by the manufacturer (Amersham Biosciences).
Immunoprecipitation Assays-Immunoprecipitations were conducted as described previously (31) with the following modifications. The day before transfection, U2OS.rGR or COS-7 cells were seeded into 150-mm dishes at 2,000,000 cells/dish containing 20 ml of medium. On the next day, DNA was transfected. After 2 days of growth, cells were treated with ethanol, Dex, or RU486 and washed once with 20 ml of PBS 1 h later. The cells were lysed for 5 min at room temperature with 1.6 ml of CytoBuster protein extraction buffer (EMD Biosciences, La Jolla, CA) containing protease inhibitor cocktail (PIC, Roche Applied Science) and clarified by centrifugation at 13,000 ϫ g for 10 min at 4°C. Samples for sodium molybdate treatment were prepared with the modification that the cell lysates were adjusted to contain either 20 mM Na 2 MoO 4 or 20 mM Na 2 SO 4 immediately after rupturing the cells followed by treatment with ethanol, Dex, or RU486 and incubation at 20°C for 20 min. Aliquots (700 l) of supernatant were incubated on a roller drum (3 rpm) with 50 l of FLAG-M2 agarose (Sigma, 4°C for overnight) to immobilize the complexes with FLAG-tagged proteins. On the next day, the antibody complexes were centrifuged (1 min at 16,000 ϫ g at 4°C), washed four times with 1 ml of lysis buffer (50 mM Tris, 150 mM NaCl, 0.1% Nonidet P-40, AGCATGAAGCTCCCAGACAAC ACCCACAGGAGCCACGAATAA ␤-Actin GCGAGAAGATGACCCAGATCATG GTCACCGGAGTCCATCACGAT pH 7.5), extracted with 20 l of 2ϫ SDS loading buffer (95°C for 5 min), and separated by 10 -12% SDS-PAGE.
Chromatin Immunoprecipitation (ChIP) Assays-ChIP was performed as described (34) with minor modifications. Two days before harvest, U2OS.rGR or U2OS cells were seeded into 150-mm dishes at 3,000,000 cells/dish in DMEM with 10% FBS. On the next day, DNA (including GR plasmid with U2OS cells) was transfected. On the third day, after ligand Dex (100 nM) or Dex-Mes (1 M) treatment for 1 h, cells were cross-linked with 1 mM dimethyl 3,3Ј-dithiobispropionimidate in 1ϫ PBS at 37°C for 30 min. Then formaldehyde (37%) was added directly into the solution to a final concentration of 1% (room temperature for 10 min with shaking), which was followed by glycine to a final concentration of 0.125 M (room temperature for 5 min with shaking). Cells were washed with ice-cold PBS twice and harvested by scraping into 2 ml of ice-cold PBS with freshly added PIC. After centrifugation (at 2500 ϫ g for 10 min), the cell pellet was resuspended in 1 ml of cell lysis buffer (50 mM HEPES, 140 mM NaCl, 1 mM EDTA, 10% glycerol, 0.5% Nonidet P-40, 0.25% Triton X-100, freshly added PIC, pH 7.5) and incubated on ice for 10 min.
Chelex-100 (100 l of a 10% slurry) was added directly to the washed beads pellet and boiled for 10 min at 99°C to reverse cross-link. After centrifugation (13,000 ϫ g, 5 min, 4°C), the supernatant was transferred to a new tube. The 100-l input sample was precipitated by 10 l of 3 M sodium acetate (pH 5.2) and 250 l of 96% ethanol at Ϫ80°C for 30 min followed by centrifugation (20,000 ϫ g, 15 min, 4°C) and washing with 500 l of 70% ethanol. The input DNA pellet was reverse-crosslinked by adding 100 l of 10% Chelex-100 as described above. The 100-l input sample was purified with MinElute PCR purification kit (Qiagen) in 300 l of EB buffer (10 mM Tris-Cl, pH 8.5). The immunoprecipitated DNA was quantified by realtime PCR using the primers listed in Table 2.
Two-factor Competition Assays-A full description is found in Refs. 13 and 29. Briefly, the effect of four concentrations of each of two factors (total ϭ 16 combinations) on the maximum induced activity (A max ) and EC 50 is determined from directly fitting a Michaelis-Menten curve to the average (n ϭ 3, S.D. is usually Ͻ10% for more than 95% of the triplicates) value of induced luciferase activity from transiently transfected reporter (GREtkLUC) with EtOH and three subsaturating concentrations of Dex (total samples ϭ 192). Graphs of 1/EC 50 and A max /EC 50 (and EC 50 /A max for a decelerator) versus the concentration of one cofactor were constructed at each of the concentrations of the second factor. A linear plot for EC 50 /A max versus factor 1 indicates that the plots in A max /EC 50 approach a zero asymptote at infinite factor 1, in which case factor 1 would be a competitive decelerator (C) before or at the concentrationlimiting step (CLS). A competitive decelerator acts similarly to a competitive inhibitor in enzyme kinetics. A downward curving, nonlinear plot that becomes linear after first subtracting an estimated nonzero asymptote from the A max /EC 50 values, and then plotting EC 50 /A max , indicates that factor 1 is a C after the CLS. The case of a nonlinear, upward curving plot is described in the text. In all cases, however, it is critical to make corrections if Western blots show that expression of the transiently transfected protein (at constant levels of either total cellular protein or an internal standard, such as ␤-actin) is nonlinear. This is because the interpretation of the graphs is predicated on the x axis being a linear scale. To determine the linear equivalent of expressed plasmid, the nonlinear plot of optical density versus ng of transfected plasmid is first fit to a Michaelis-Menten plot of The functional equivalent of the transfected plasmid that gives a linear optical density versus plasmid plot is then obtained from the formula of

Site (bp) Forward primer (5-3) Reverse primer (5-3)
IP6K3 TSS/pause (Ϫ27 to 160) The x axis value of the amount of plasmid in the various graphs is then this "corrected plasmid" value. These same Western blots are used to determine the relative amount of endogenous factor (in units of ng of factor plasmid) when dealing with an accelerator. It is not necessary to quantitate the relative amount of an endogenous factor that displays decelerator activity. If an experiment used n concentrations for each cofactor, then there would be a total of four to six graphs, each with n separate curves. The shape of the curves and how they change with the other cofactor are then compared with Table 1B of Ref. 13 to determine the kinetically defined mechanism of action and site of action, relative to each other and to the CLS. Many of the entries in Table 1B  Statistical Analysis-Unless otherwise noted, all experiments were performed in triplicate multiple times. Kaleida-Graph 3.5 (Synergy Software, Reading, PA) was used to determine a least-squares best fit (R 2 was almost always Ͼ0.95) of the experimental data to the theoretical dose-response curve, which is given by the equation derived from Michaelis-Menten kinetics of y ϭ (free steroid)/(free steroid ϩ dissociation constant (K d )) (where the concentration of total steroid is approximately equal to the concentration of free steroid because only a small portion is bound), to yield a single EC 50 value. The values of n independent experiments were then analyzed for statistical significance by the two-tailed Student's t test using InStat 2.03 for Macintosh (GraphPad Software, San Diego, CA). The Mann-Whitney test or the Alternate Welch t test is used when the difference between the S.D. values of two populations is statistically significant.

NELF-B Modulates GR Induction Parameters with Exogenous Reporter
Gene-Human osteosarcoma (U2OS) cells were transiently transfected with the synthetic reporter gene (GREtkLUC) and a GR-expressing plasmid plus varying concentrations of two different NELF-B plasmids: a chimera lacking the C-terminal 30 amino acids (Fig. 1A) and the full-length wild type NELF-B (Fig. 1B) (see "Experimental Procedures"). In each case, cells were induced with a range of steroid concentrations for 20 h as described under "Experimental Procedures." The low levels of endogenous GR in U2OS cells give a weak response to exogenous NELF-B that is greatly enhanced with added GR (Fig. 1A). In the presence of increased GR, the A max with saturating concentrations of the agonist Dex is dramatically decreased by NELF-B. At the same time, the -fold induction by Dex is lowered, the PAA of the antiglucocorticoid Dex-Mes is reduced, and the EC 50 of Dex is significantly increased. Similar results are obtained with lower amounts of all exogenous factors: the full-length NELF-B, GR, and GREtkLUC (Fig.  1B). This indicates that the C-terminal 30 amino acids of NELF-B are unnecessary for activity and that wt and chimeric NELF-B can be used interchangeably (see also Fig. 6 below). In all cases, the responses are directly dependent upon the concentration of added NELF-B. Thus, NELF-B is a very active modulator of GR induction parameters over a range of factor concentrations and has the hallmark of a corepressor, which is to reduce the A max (35).
To examine the effects of reduced levels of NELF-B, we used T47D cells in which NELF-B was decreased by stable transfection with short helical (sh) RNAs (18). T47D cells expressing a shEGFP oligonucleotide were used as a control. qRT-PCR revealed that the levels of NELF-B mRNA in the NELF-B knockdown (KD) cells are reduced by 80% relative to the control cells (Fig. 1C). Interestingly, transient transfection of NELF-B plasmid increases NELF-B mRNA in U2OS cells by a factor of 2 to a level that is only 40% of that in T47D cells expressing the shEGFP oligonucleotide (data not shown). Due to the low -fold induction by endogenous GR of the T47D cells, GR plasmid was cotransfected with GREtkLUC reporter. Under these conditions, decreased NELF-B causes a significant left-shift in the dose-response curve (Fig. 1D), consistent with added NELF-B moving the EC 50 to higher steroid concentrations in Fig. 1, A and B. At the same time, the A max goes down. However, the -fold induction increases in NELF-B KD versus control cells (Fig. 1E) because the basal level is much lower in the NELF-B KD cells (22.1 Ϯ 6.7 versus 59.6 Ϯ 7.8 (S.E., n ϭ 5)). The PAA for the antiglucocorticoid Dex-Mes is proportionately higher in the NELF-B KD cells with exogenous GR, but the increase is not statistically significant. We conclude that NELF-B has a consistent effect in two different cell lines regardless of whether the intracellular level of NELF-B is increased by transfected NELF-B plasmid or decreased by siRNA-mediated degradation of endogenous NELF-B mRNA.
GR Ligand-binding Domain Binds NELF-B and Mediates NELF-B Modulatory Actions-We previously found that the GR ligand-binding domain (LBD) is sufficient to respond to the modulatory actions of the coactivator TIF2 and the comodulator Ubc9 (36 -38). Using transiently transfected GAL/GR chimera (GAL/GR525C, Fig. 1F) driving the synthetic GAL-regulated reporter (FRLuc) in the same knockdown T47D cells as in Fig. 1, D and E, reduced levels of NELF-B are seen to shift the dose-response curve to the left as compared with the EGFP KD cells, thereby reducing the EC 50 (Fig. 1, G and H). As with full-length GR, diminished NELF-B decreases the A max but increases the -fold induction while augmenting the PAA of the antiglucocorticoid Dex-Mes (Fig. 1H). The quantitative changes in all induction parameters with reduced NELF-B are very similar for GAL/GR525C (Fig. 1H) and full-length GR (Fig.  1E). We conclude that the GR LBD is sufficient to mediate most of the actions of NELF-B under these conditions.
Transiently transfected human GR and human NELF-B are reported to bind to each other (16). This was confirmed by showing that NELF-B binds to both ligand-free and -bound GAL/GR525C (Fig. 1I). Thus, the GR LBD is sufficient for bind-ing NELF-B. We next asked whether NELF-B binding to endogenous full-length GR required activation by examining the interactions of transiently transfected human FLAG-tagged NELF-B with endogenous GR in U2OS cells stably transfected with rat GR (U2OS.rGR cells) (39). To distinguish between unactivated and activated receptors, cell-free lysates were first treated with either sodium molybdate, to block activation of the receptor to the DNA-binding form (40), or the iso-electronic but inactive control, sodium sulfate. Vehicle, agonist steroid, or antiglucocorticoid was then added followed by brief warming to 20°C to activate the receptors. Finally, anti-FLAG antibody was added to immunoprecipitate the FLAG/NELF-B and asso-ciated proteins. When receptor activation to the DNA-binding form of the receptor is blocked by the addition of sodium molybdate, GR co-immunoprecipitation with FLAG/NELF-B is almost totally prevented as compared with samples containing sodium sulfate (Fig. 1J). The ability of ligand-free GR to be activated is consistent with earlier studies (41). Thus, we deduce that NELF-B binds to endogenously expressed GR but only after activation.
NELF-B Modifies the Induction Properties of Endogenous Genes-To determine the generality of the above observations, we asked whether NELF-B altered the GR induction parameters of four endogenous genes with low (basal) levels of expression in the absence of added steroid: GILZ (42), IP6K3 (IHPK3) (43), IGFBP1 (44), and LAD1 (45). The effect of NELF-B on GR induction was examined in U2OS cells because the extensive squelching that was usually seen with these four genes in U2OS.rGR cells prevents an accurate assessment of GR transactivation parameters (data not shown). In U2OS cells, NELF-B reduces the -fold induction of Dex and the PAA of Dex-Mes while increasing in EC 50 for IGFBP1 and IP6K3 (Fig. 2, A-C). The low -fold induction of GILZ and LAD1 with added NELF-B precluded an assessment of changes in EC 50 , but the -fold induction and PAA of Dex-Mes are each decreased (Fig. 2, D and E). Thus, the effects of NELF-B with synthetic reporters are recapitulated for endogenous genes in a manner that depends quantitatively upon the gene examined.
Binding of GR and NELF-B to IP6K3 Gene Elements-To obtain more information about how NELF-B modulates GR induction parameters, we asked whether NELF-B affected GR binding to the IP6K3 GRE. Guided by ChIP-seq data on GR regulation of gene expression in 3134 murine mammary adenocarcinoma cells, 6 in silico analysis revealed a putative GRE at ϩ665 in intron 1 of the human IP6K3 gene (see also 57). Fusion of this region upstream of tkLUC afforded a GR-inducible reporter ("IP6K3tkLUC") that is appropriately modulated by NELF-B in CV-1 cells (Fig. 3A), whereas mutation of the candidate GRE (ggaaCagcaTGtcct to ggaaTagcaACtcct; capital letters indicate affected nucleotides) eliminates GR inducibility (Fig. 3B). This argues that the IP6K3 GRE is in the middle of intron 1 and is active in different gene settings and cells.
The possible binding of GR and FLAG-tagged NELF-B to the TSS and to the usual location of paused polymerase at about ϩ50 bp (26,46), to the region 3Ј of the TSS/pause, and to the intronic GRE at ϩ665 was examined by ChIP assays in U2OS cells that had been transiently transfected with FLAG/NELF-B and GR and treated with Dex. FLAG/NELF-B was used because anti-NELF-B antibody was unable to detect any recruitment of NELF-B (data not shown). As indicated in Fig. 3, C-E, both strong GR binding to the GRE Ϯ steroid and weak binding of GR to the regions of the TSS and possible pause sites for RNA polymerase II are reduced by NELF-B. This response appears relatively specific because the signal at the GRE seen with IgG is much weaker and unaffected by NELF-B (Fig. 3C). Conversely, exogenous NELF-B produces a slightly greater increase in the binding of NELF-B to the GRE as compared with the TSS or pause regions (Fig. 3E). This is clearer when the data are plotted as increase in FLAG/NELF-B binding above the background of no added NELF-B (Fig. 3F). These results suggest that the modulatory effects of NELF-B on the GR induction parameters for IP6K3 are due to reduced association of GR with the GRE that may be accompanied by increased NELF-B binding.
A similar behavior was seen for GR and NELF-B recruitment to the GRE of the exogenous GREtkLUC reporter (Fig.  3, G and H). Thus, the parallel capacity of NELF-B to modulate the A max , EC 50 , and PAA for GR induction of exogenous GREtkLUC (Fig. 1A) and endogenous IP6K3 gene (Fig.  2C) is mirrored in the comparable lower binding of GR, and marginally higher NELF-B binding, to the GRE sequences of each target gene.
Binding of GR and NELF-B to IGFBP1 Gene Elements-Two GREs upstream of the TSS of the IGFBP1 gene have been reported (47). In silico scanning revealed the presence of a cluster of a possible tandem GRE and three half-sites in the first intron of the human IGFBP1 gene. When these GREs are fused upstream of the tkLUC plasmid, transiently transfected into CV-1 cells, and assayed for steroid inducibility in the presence of cotransfected GR plasmid, the intronic GRE is found to be much more responsive to the agonist Dex than the GREs in the 6 G. Hager, personal communication. promoter region (Fig. 3I). Mutation of the single palindromic GRE in intron 1 greatly decreases the activity with Dex. We conclude that the most potent GRE of the IGFBP1 gene is in intron 1.
ChIP assays could not detect any binding of NELF-B (in the form of FLAG/NELF-B) to various regions of the IGFBP1 gene in U2OS cells (data not shown). However, GR binding was significantly decreased by added NELF-B (Fig. 3, J and K), just as for the IP6K3 gene. The recruitment of GR to the intronic GREs, as well as the decrease in specific GR binding with added NELF-B, is much more robust than to either of the promoter GREs (IG1 and IG2). Similar results were seen in U2OS.rGR cells, with higher levels of GR recruitment resulting from the higher levels of stably transfected GR (data not shown). Thus, as compared with the promoter GREs, the behavior of GR binding to the intronic GREs more closely mirrors the modulatory properties of NELF-B in the gene activation assays of Fig. 2B.
All NELF Subunits Have the Same Modulatory Activity as NELF-B-The facts that NELF-B binds to endogenously expressed GRs after activation (Fig. 1J) and that the recruitment of NELF-B to biologically active GREs is slightly greater than to the potential pause regions (Fig. 3F) suggest that the modulatory activity of NELF-B on GR induction parameters in our system may be separate from the role of NELF-B in polymerase pausing as part of the NELF complex. This possibility is sup-ported both by the varying ratios of NELF proteins in several tissue culture cell lines (48) and by different tissue distributions of each of the NELF subunit mRNAs in assorted mouse tissues ( Fig. 4 and Table 3). All NELF subunits have the same effect on A max , EC 50 , and PAA in the transient transfection assay of Fig.  1, although some appear to be more active (Fig. 5, A and B). It has been reported that decreasing the levels of one NELF component often affects the levels of other NELF components (48). We do not see this when overexpressing NELF components in U2OS cells (Fig. 5C), nor does overexpressed NELF-B alter GR levels (Fig. 5D). These results suggest that each NELF is independently active because they cannot all be limiting for NELF complex action. In other words, if the individual NELF components act solely through the NELF complex, then the only way for them to affect A max , EC 50 , and PAA is through the formation of more complexes. This implies that under any given circumstance, only one of the components is limiting in that adding either of the other three would not increase the amount of NELF complex. Thus, the addition of only one of the NELF components will affect the A max , EC 50 , and PAA at any given time. Although the identity of this component can change as the relative concentrations of the components change, the addition of each of the four components to one set of concentrations, as in Fig. 5, A and B, cannot result in more complex formation.

Application of Competition Assay to Determine Actions of NELF-A and -B versus GR and GREtkLUC-
To further define the actions of NELF subunits, we employed a discriminating competition assay that compares the actions of two factors in a competition assay. This assay defines factor actions in kinetic terms of either an enzyme-kinetics activator (49,50), which we call an accelerator, or one of five types of enzyme-kinetics inhibitors, or decelerators, relative to the CLS (10,(12)(13)(14)29).
Given the sensitivity, and novel information provided by the competition assay, we wanted to reinvestigate the importance of the C-terminal residues that are missing in the chimeric NELF-B. Therefore, we first compared the properties of wtNELF-B with chimeric NELF-B in a competition assay with GREtkLUC. Briefly, several subsaturating concentrations of Dex were used to induce the GREtkLUC reporter in the presence of 16   four of NELF-B). Exact fits of GR induction of luciferase activity to a first order Hill plot yield the A max and EC 50 for each combination of GREtkLUC and NELF-B. Graphs (1/EC 50 , A max / EC 50 , and EC 50 /A max ) versus factor 1 with the four concentrations of factor 2 are plotted, and their characteristics as compared with those in Table 1B of Dougherty et al. (13). As previously reported for the chimeric NELF-B (14), graphs of 1/EC 50 versus GREtkLUC give horizontal plots that decrease in position with added NELF-B (Fig. 6A). Critically, as with the chimeric NELF-B, the graphs of EC 50 /A max versus wtNELF-B are again best fit by a power function with an exponent of n ϭ 2 (Fig. 6B). The expression of full-length NELF-B was found to be linear up to 8 ng of plasmid (data not shown). Therefore, no correction for nonlinear expression is needed to plot Fig. 6B. This means, as described previously (29), that NELF-B is a competitive decelerator acting at n ϭ 2 locations at or before the CLS. Furthermore, because two competitive decelerators cannot act at the same site according to the theory on which the assay is based (10,12,13), NELF-B must either act at the CLS and at another site before the CLS or act at two different sites before the CLS. Thus, the wtNELF-B and chimeric NELF-B have identical properties with GREtkLUC; GREtkLUC is an accelerator at the CLS, whereas both forms of NELF-B are competitive decelerators acting at two sites before or at the CLS.
We next examined the properties of wt and chimeric NELF-B in the competition assay with GR. The graph of A max /EC 50 ver-   Table 3. Similar results are seen on the BioGPS website. Adrenal G, adrenal gland; error bars represent S.D. of triplicates.

NELF-C AGGAGGAACAGGGTGGTTCT TGATGTCTGCAGGATTCAGG NELF-E GGACCCTGGAAGGGAAGTTA CCGAAGCCGATCACTAGAAG
sus GR is uniquely categorized (13) as linear plots that intersect at x ϭ y ϭ 0, with the slopes decreasing with added NELF-B (Fig. 6C). This plot identifies GR as an accelerator at or before the CLS and NELF-B as either a competitive decelerator (C) or a mixed decelerator before or at the CLS. This last possibility for NELF-B is eliminated by the graph of EC 50 /A max versus NELF-B (Fig. 6D), which displays the characteristic upward curving quadratic plot of a factor that is a competitive decelerator acting at n ϭ 2 locations at or before the CLS. Finally, from the linear 1/EC 50 versus GR plot, where the curves intersect at the origin with slopes that decrease with increasing chimeric NELF-B (data not shown) (13), we can further identify the actions of GR as an accelerator before both the CLS and the chimeric NELF-B. Assays with the full-length wild type NELF-B gave virtually identical results (Fig. 6, E and F). This is evident in plots of A max /EC 50 versus GR, which are linear and intersecting close to the origin (Fig. 6E versus Fig. 5D) and particularly in the EC 50 /A max versus NELF-B plots, which give the same upward curvature (Fig. 6F versus Fig. 5C). Thus, wtNELF-B and chimeric NELF-B remain indistinguishable in being a competitive decelerator acting at two sites before or at the CLS, whereas GR is an accelerator acting before the CLS. This further supports the conclusions from Fig. 1, A and B, that wt and chimeric NELF-B are interchangeable and that the C-terminal 30 residues of NELF-B are extraneous. We next determined the properties of NELF-A in a competition assay with GR. The results (Fig. 6, G and H) are nearly identical to those with NELF-B. Thus, NELF-A, like NELF-B, acts as a competitive decelerator at two positions before or at the CLS. At the same time, both sites of action of NELF-A are after GR, which is again an accelerator. However, these graphs give no information as to whether the two sites of action of NELF-A are the same or different from those of NELF-B.
Comparison of Actions of NELF-B and NELF-A in Competition with TIF2-To gain further mechanistic information about NELF-A and NELF-B actions, and to determine whether there are any differences between NELF-A and NELF-B, we conducted competition assays with other factors. First, we asked whether NELF-B can reverse the effects of the accelerator TIF2 (14,29). As shown in Fig. 7A, increasing amounts of NELF-B effectively counteract the actions of TIF2. We therefore asked whether NELF-B and NELF-A are equally capable of reversing the accelerator effects of TIF2. With NELF-B, the positive linear slopes of 1/EC 50 versus TIF2 (Fig. 7B) and A max /EC 50 versus TIF2 (Fig. 7C) intersecting at a point of y Ϸ 0 and x less than the amount of endogenous TIF2 (as determined by Western blots (data not shown) and expressed in terms of ng of TIF2 plasmid that give the same amount of TIF2 protein), combined with the decreasing slopes with increasing NELF-B, are diagnostic of TIF2 being an accelerator after the CLS and NELF-B being a decelerator before or at the CLS (13). Finally, as seen above for GR and NELF-B in Fig. 6, B, D, and F, the plot of EC 50 /A max versus NELF-B (Fig. 7D) is decidedly nonlinear. This argues that NELF-B is again acting as a competitive decelerator at two sites, before or at the CLS.
We next examined the competition of NELF-A with TIF2. As with NELF-B, the plots of 1/EC 50 and A max /EC 50 (data not shown) versus TIF2 give positive slopes that decrease with added NELF-A and intersect at values of TIF2 that are more negative than the calculated point of zero endogenous TIF2. Importantly, the plot of EC 50 /A max is again best fit by a quadratic function (Fig. 7E). Therefore, NELF-A, like NELF-B, exerts its effects as a competitive decelerator of TIF2 actions by acting at two distinct steps before or at the CLS. Furthermore, with both NELF-A and NELF-B, TIF2 is an accelerator that functions after the CLS.

Actions of NELF-A and NELF-B Are Additive-
The above similar actions of NELF-A and NELF-B with TIF2 cannot determine whether the two NELFs act individually or via a common complex. To answer this question, we asked whether the effects of two NELF subunits are additive by competing NELF-A with NELF-B. If the effects are additive, then the two NELFs must be completely interchangeable and act independently. Additivity also precludes them from exclusively acting in a common complex because both cannot be limiting simultaneously, i.e. only one of either NELF-A or NELF-B can affect A max , EC 50 , and PAA at any given time if they only act through the NELF complex. Plots of EC 50 /A max versus each factor display qualitatively identical properties, with the best fits being with quadratic curves (Fig. 7F and data not shown). As described above, such plots are diagnostic of both factors acting as a competitor at two sites before or at the CLS. Importantly, in each case, the addition of the second factor increases the responses of the first factor to give higher positioned curves in all graphs. This behavior is consistent with the two factors acting additively. As further support for this conclusion, a plot of EC 50 /A max versus the combined total of functionally active NELF-A and NELF-B was constructed. To do so, a comparison of the amount (in ng) of each transfected NELF plasmid by itself that is required for half-maximal reduction of A max /EC 50 and 1/EC 50 (data not shown) revealed that 8.6 Ϯ 1.4-fold (S.E., n ϭ 6) more activity is expressed per ng of NELF-B plasmid (corrected for nonlinear expression) than from the NELF-A plasmid. Therefore, the total amount of active NELF proteins in each of the 16 combi-nations is calculated as ng of NELF-A plasmid plus 8.6 times the ng of corrected NELF-B plasmid. The graph of EC 50 /A max versus the combined amount of active NELFs is nicely fit by a single upward curving quadratic plot (Fig. 7G). Together, the graphs of Fig. 7, F and G, strongly argue that the combined effects of NELF-A and NELF-B under these conditions are additive, with FIGURE 7. NELF-B and NELF-A function additively as competitive decelerators at two steps with different cofactors. Competition assays were conducted as in Fig. 6 with GR, GREtkLUC, NELF-A, full-length wild type NELF-B, and/or TIF2. A, NELF-B reverses effects of TIF2 on GR induction parameters. The average -fold changes in GR induction parameters from 5 independent experiments (Ϯ S.E.) in U2OS cells transiently transfected with GR Ϯ TIF2 plasmid and the indicated amounts (ng) of NELF-B plasmid are plotted in the same manner as Fig. 1A. *, p Ͻ 0.05, **, p Ͻ 0.005, ***, p Ͻ 0.0005 versus GR plus TIF2. B-E, graphical analysis of competition of NELF-B or NELF-A, with TIF2 for modulation of GR induction parameters. Experiments were performed with the indicated amounts of NELF-B (after correction for nonlinear expression) or NELF-A and TIF2 (after correction for nonlinear expression) and then analyzed as above in Fig. 6. The "a versus b plots" for A max /EC 50 versus TIF2 graphs such as in panel C (see "Experimental Procedures"), which yield an unbiased determination of a common intersection point, gave average values for the x and y axis coordinates of the intersection point as Ϫ12.8 Ϯ 6.4 and 0.73 Ϯ 1.20 (S.D., n ϭ 4), respectively. Western blots determined that endogenous TIF2 protein is equivalent to 4.1 ng of TIF2 plasmid, so Ϫ4.1 ng TIF2 plasmid is where there is no TIF2 in the cells, and the intersection point of the A max /EC 50 versus TIF2 graphs is less than this. F and G, competition assay with NELF-A and NELF-B. A graph of EC 50 /A max versus corrected (cor'd) amount of NELF-B with indicated amounts of NELF-A (F) was constructed as for panel E. The graph of EC 50 /A max versus the combined amounts of NELF-A and NELF-B (G) was compiled as described under "Experimental Procedures." Similar results were seen in 2 additional independent experiments. both factors acting independently as competitive decelerators at two sites before or at the CLS.

A Common Motif Is Involved in NELF-A and NELF-B Actions-
The binding of NELF-B to GR (Fig. 1, I and J), the weak recruitment of NELF-B to the GRE of the IP6K3 gene (Fig. 3, E and F), and the additive activity of NELF-A and NELF-B (Fig. 7, F and G) suggest that both NELFs have activities independent of the NELF complex. Interestingly, an in silico screen of all four NELF subunits revealed a common motif of $XXXLXGX-L$GX$$, where $ ϭ charged amino acid (Fig. 8A). Hydropathy plots of the surrounding region in each NELF suggested that this "NELF motif" is part of an ␣-helical structure, with local hydrophobic and nonbulky groups at opposite sides and neighboring charged residues (Fig. 8B). This motif does not contain any phosphorylatable residues and thus is not a target of Cdk9 or other kinases. No mutants of this region in any NELF were found in the literature. Interestingly, however, this region of NELF-A was found to be dispensable for NELF complex formation, binding to RNA polymerase II, and activity (21). Furthermore, the region containing this sequence is highly conserved in each NELF from humans to zebrafish (and to Drosophila for NELF-A and NELF-B) (data not shown), which suggests a bio- logical function. To test this hypothesis, we made a quadruple alanine mutant in NELF-A and NELF-B by converting $XXX-LXGXL$GX$$ to $XXXAXGXA$GXAA (bold letters indicate changed amino acid residues) and testing the activity of the mutant versus wt protein in the above competition assays. These experiments were corrected for differences in protein expression as detected by Western blots. Specifically, wtNELF-A is expressed 2.45 times more efficiently than 4mtNELF-A, whereas 4mtNELF-B is produced 2.09-fold better than wtNELF-B (data not shown).
Mutation of NELF-A (4mtNELF-A) does not alter the ability of either GR to act as an accelerator before the CLS or GREtk-LUC to be an accelerator at the CLS (data not shown). What has changed is that the curvilinear quadratic fit for EC 50 /A max versus wtNELF-A with increasing GR (Fig. 8C) is now a linear fit with 4mtNELF-A (Fig. 8D). Similarly, in competition assays with GREtkLUC, the four-amino acid mutation of the NELF motif converts the normal curvilinear plot of EC 50 /A max versus wtNELF-A (Fig. 8E) to a linear plot (Fig. 8F). These results mean that the mutations have inactivated the ability of wtNELF-A to act at one of the two sites before or at the CLS and that the 4mtNELF-A now acts as a competitive decelerator at just one site before or at the CLS. Furthermore, these mutations have not simply disabled the wtNELF-A because the 4mtNELF-A still exhibits partial wild type activity.
With NELF-B, the effect of mutating the NELF-motif is less dramatic but nonetheless significant. Graphs of EC 50 /A max versus both wt and 4mtNELF-B, with increasing GR or GREtk-LUC, still display upward curvature that is fit by quadratic plots, but the curvature is less with 4mtNELF-B (data not shown). This was quantitated by determining the amount of NELF-B that is required for a 50% reduction in the A max /EC 50 versus NELF plots (Fig. 8, G and H), each of which eventually goes to zero. After correcting for more efficient expression of 4mtNELF-B, the amount of plasmid needed for 50% decrease is raised from 1.68 Ϯ 0.13 ng of plasmid with wtNELF-B to 2.76 Ϯ 0.25 ng of plasmid with 4mtNELF-B (S.E., n ϭ 16, p ϭ 0.0007). Similarly, in competition assays with GREtkLUC, 1.68 Ϯ 0.12 ng of wtNELF-B is sufficient to inhibit the A max /EC 50 by 50%, whereas 2.37 Ϯ 0.24 ng of 4mtNELF-B (after adjustment for higher expression; S.E., n ϭ 12, paired p ϭ 0.0007) is needed. These results argue that the mutated NELF domain in NELF-B causes a decrease in the activity of at least one site of NELF-B action. Collectively, an intact NELF domain is required for full activity of both NELF-A and NELF-B.

DISCUSSION
NELF-A, -B, -C/D, and -E are best known for their role as subunits of the NELF complex, which is involved in the pausing of RNA polymerase II. We now find that each subunit has the additional ability to increase the EC 50 and decrease the A max of GR-mediated gene transactivation. The properties of NELF-A, and especially NELF-B, were extensively investigated. NELF-B is equally active with exogenous and endogenous genes. NELF-B both reduces GR recruitment to the GREs of target genes and may weakly bind to GREs in a manner that depends upon the endogenous gene. A target of NELF-B is the GR LBD, as witnessed by the binding of NELF-B to the GR LBD and by the ability of a GAL/GR-LBD chimera to respond similarly to the full-length GR either with added NELF-B or in cells in which endogenous NELF-B had been knocked down by NELF-B shRNA. Competition assays of NELF-B or NELF-A with GR, GREtkLUC reporter, or TIF2 reveal that both NELFs each act as competitive decelerators at two sites before or at the CLS. However, the action of both NELF-A and NELF-B at one site is attenuated by the mutation of an evolutionarily conserved motif that is shared by all four NELF proteins.
Several lines of evidence suggest that the above new properties of NELF-A and NELF-B are independent of the NELF complex and result instead from their actions either as individual proteins or in complexes with other proteins that are unique for each NELF. First, the ability of exogenous NELF-A and NELF-B to produce additive effects (Fig. 7, F and G) is incompatible with both factors acting through a common complex because both NELF-A and NELF-B cannot be limiting at the same time, i.e. both factors affecting A max , EC 50 , and PAA at the same time. Likewise, the ability of each NELF subunit to evoke similar changes in GR transactivation properties (Fig. 5, A and B) is incompatible with action via a common complex because all four factors cannot be simultaneously limiting. Second, a NELF domain has been identified in all four NELF proteins. Surprisingly, this domain is irrelevant for NELF-A activity in the NELF complex (21). However, mutation of this domain alters the activity of NELF-A in our assays, which strongly indicates that an activity independent of the NELF complex has been affected. Third, several cell lines are reported to contain unequal amounts of the four proteins (48), and all of the NELF subunit mRNAs are present at different levels in many tissues (Fig. 4). In Drosophila, ChIP-chip assays revealed that 80% of the binding sites of NELF-B and NELF-E overlap but 20% do not (26). In both cases, nonequivalent abundances or binding of NELF proteins mean that some NELF components are in excess and could have NELF complex-independent activities. Fourth, both NELF-A and NELF-B are found to act at two steps. Thus, to the extent that the NELF complex is thought to cause polymerase pausing at a single step (51,52), at least one step of NELF-A and NELF-B action would be different from that of the NELF complex. Fifth, recent results suggest that polymerase pausing is restricted to those genes requiring synchronized expression during development and differentiation (27), which is not the case for our transiently transfected reporter gene system. The NELF complex has also been proposed to alter chromatin organization (53,54). However, this activity is again unlikely with the exogenous GREtkLUC reporter in our studies. Transfected genes are thought to possess little if any organized chromatin structure (55). Furthermore, recent ChIP-Seq experiments indicate that 80 -95% of endogenous GR-regulated genes already have an open chromatin conformation (56 -58), thereby minimizing the need for chromatin reorganizations. Finally, each constituent of the NELF complex is listed on the protein interaction database BioGRID to interact with 8 -32 proteins, many of which are unique to each specific NELF subunit. It is likely that several of these complexes evoke biological responses. To summarize, our results indicate that NELF-A and NELF-B have inhibitory actions on steroid-mediated gene expression that are independent of the NELF complex. This does not preclude that NELF-A and NELF-B are still involved in the NELF complex; instead, it says that the NELF complex does not play a quantitative role in controlling the amount of GR-induced gene product in the present system.
We have used the recently developed competition assay (13,14) to identify both NELF-A and NELF-B as each being competitive decelerators at two kinetically defined steps at or before the CLS, which is the equivalent in a steady-state system of the rate-limiting step in enzyme kinetics. This assay does not require any prior knowledge of the mechanism of action of the competing factors (29). In fact, any mechanism by which the factor influences GR activity will be detected. In this manner, this assay is a high-throughput assay to detect GR activity-modulating mechanisms. These assays tell us that both NELFs act after the site of GR action and before the site of TIF2 action, which is after the CLS, although the precise molecular mechanisms remain unknown. These results also confirm our earlier findings that GREtkLUC is an accelerator at the CLS (14) and that TIF2 is an accelerator after the CLS with both endogenous and exogenous genes (12,13). Thus, we can construct the following sequence of cofactor action that applies for GR-regulated transactivation in this system and, we propose, in several other systems.

GR Ͻ NELF-A and NELF-B Յ GREtkLUC Ͻ TIF2 (Eq. 3)
With NELF-B, at least one step results in a decrease of GR binding to the GRE.
The competition assay positions the kinetically defined mode of factor activity relative to a benchmark step (the CLS) but is silent regarding where or when factor binds. Therefore, we cannot yet incorporate the NELF-B-induced reduction of GR binding to the GRE, as revealed by ChIP assays, into the above kinetically defined sequence of NELF actions. It should also be realized that the step at which a factor binds is not always equivalent to when that factor exerts its biological effect. For example, the number of genomic binding sites for GRs is ϳ10-fold greater than the number of induced genes (59,60), there is no temporal correlation between GR binding to enhancer regions and the transcriptional response (61), p300 is recruited to the androgen-responsive elements of the TMPRSS2 and FKBP5 genes but is required for androgen induction of only TMPRSS2 (62), and paused RNA polymerase II is bound well before it is active in transcriptional elongation.
Co-immunoprecipitation and ChIP assays suggest that NELF-B associates with the LBD of activated GRs in a manner that decreases the binding GR-steroid complexes to those GREs controlling gene expression (Figs. 1 and 3). This contrasts with the behavior of ER (15). NELF-B binds to the ER LBD and decreases ER-mediated gene transactivation, just as seen for GR. However, the recruitment of ER to the promoters of ERregulated genes is not decreased by excess NELF-B. Thus, the major actions of NELF-B with GR and ER appear different, with NELF-B having a greater effect on promoter-bound ERs while reducing either the amount of GR that binds to the promoter region or the strength of GRE-bound receptors.
Interestingly, the decrease in GR recruitment to the GRE in the presence of added NELF-B is similar with exogenous and endogenous reporter genes (Fig. 3, D and G). It is generally believed that transiently transfected genes are not good models for receptor-regulated gene expression because they do not possess the same level of chromatin organization (55,63). However, the similar effects of factor on gene induction and GR recruitment to endogenous and exogenous genes both in the present study and in earlier studies (29) suggest that situations exist where the two systems may be studied interchangeably.
In conclusion, NELF-A and NELF-B are new cofactors for GRs that possess additive activities for altering the A max and EC 50 of GR-regulated transactivation of exogenous and endogenous genes in different cells in addition to their involvement in the NELF complex. These activities rely, in part, on the integrity of an evolutionarily conserved motif shared not only by NELF-A and NELF-B but also by NELF-C/D and NELF-E. Evidence is presented that at least some these activities are independent of the NELF complex composed of all four proteins. Because NELF-C/D and NELF-E display the same modulatory activity as NELF-A and NELF-B with GR-induced transcription, we propose that this NELF motif is intimately involved in the individual actions of each protein. Competition assays have identified the kinetically defined mode and site of action of NELF-A and NELF-B as competitive decelerators at two steps relative to several factors during GR transactivation. Such dual site action has also been found for another GR cofactor, PA1 (29), and thus may be more prevalent than currently appreciated. Further studies are required to identify the precise steps that NELF-A and NELF-B are inhibiting (which may be different), to determine whether NELF-C/D and NELF-E also act at two steps before or at the CLS and if so whether they are the same or different from those affected by NELF-A and NELF-B, and to further define the role of the NELF motif in these new actions of each NELF protein. Such studies of this novel family of cofactors should significantly increase our molecular understanding of GR-regulated gene transcription and open new windows for modulating GR induction properties.