Expanding the Paradigm for Estrogen Receptor Binding and Transcriptional Activation

Estrogen receptor binds to a spectrum of nonconventional EREs to activate transcription, with HMGB1 facilitating greater activity in both.

Abstract

Estrogen receptor (ER) binds to a spectrum of functional estrogen response elements (ERE) within the human genome, including ERE half-sites (HERE), inverted and direct repeats. This has been confounding, because ER has been reported to bind weakly, if at all, to these sites in vitro. We show that ER binds strongly to these nonconventional EREs, and the binding is enhanced by the presence of high-mobility group protein B1 (HMGB1). Collectively, these and previous findings reinforce the notion of the plasticity of strong ER/ERE interactions, consistent with their broader range of observed binding specificity. In addition, transient transfection studies using luciferase reporter gene assays show that these EREs drive luciferase activity, and HMGB1 enhances transcriptional activity. Furthermore, HMGB1 gene expression knockdown results in a precipitous drop in luciferase activity, suggesting a prominent role for HMGB1 in activation of estrogen/ER-responsive genes. Therefore, these data advocate that the minimal target site for ER is a cHERE (consensus HERE) that occurs in many different contexts and that HMGB1 enhances both the binding affinity and transcriptional activity. This challenges the current paradigm for ER binding affinity and functional activity and suggests that the paradigm requires significant reevaluation and modification. These findings also suggest a possible mechanism for a cross talk between genes regulated by ER and class II nuclear receptors.

Estrogen receptor (ER) (NR3A1) is a member of a superfamily of related nuclear hormone receptors (NHR) that includes those activated by steroid hormones, thyroid hormone, retinoids, and vitamin D, in addition to orphan receptors that have similar structures but no identified ligand (1–3). Although there are six phylogenetic classifications of the 49 or more human NHR genes (4), one finds that ER, together with most of the nonsteroid receptors and the orphan estrogen-related receptors recognize the same consensus (half-site) sequence 5′-AGGTCA-3′ in DNA, whereas the other steroid hormone receptors recognize a different consensus sequence, 5′-AGAACA-3′ (5, 6). With the exception of the orphan receptors, which may not require a ligand, these receptors are modular, ligand-activated transcription factors that exhibit high selectivity and transcriptional activity that is regulated by a spectrum of cofactors. These cofactors may contribute to ER-activated transcription by mediating 1) its binding affinity and specificity, 2) its interaction with other regulatory factors and basal factors in the preinitiation complex, and/or 3) the active remodeling of nucleosome/chromatin structure (7).

The current paradigm for classification of NHR is based on their dimerization pattern and the nature of the response element they bind to. Class I receptors, such as ER and the other steroid hormone receptors, bind as homodimers, the class II (nonsteroid hormone) receptors bind predominantly as heterodimers, whereas orphan receptors bind as a monomer or dimer (1). The binding specificity for class I/II receptors is further predicated on a sequence of the 5-bp (or 6 bp) recognition half-site, the orientation of the half-sites [inverted repeat (IR) or direct repeat (DR)], and the number of nonspecific base pairs (the spacer) between the two half-sites (5, 8). The bipartite consensus response element for ER is the IR of two hexameric core half-site motifs, 5′-AGGTCA-3′ [consensus estrogen response element (ERE) half-site (HERE) (cHERE)] with a spacer of 3 bp [consensus ERE (cERE), 5′-AGGTCANNNTGACCT-3′], whereas class II receptors bind to this same cHERE in a DR arrangement, with specificity further determined by the number of base pairs in the DR. A perplexing finding is that although there are very few estrogen [17β-estradiol (E2)]-responsive genes that contain a simple palindromic cERE in their enhancer or promoter, an increasing number contain cHERE in a variety of contexts, including DR or a cERE, in which the spacer size differs from n = 3 (5, 9). Crystal structures in which the ER DNA-binding domain (DBD) (ERDBD) is bound to either a cERE or a non-cERE were the first to reveal how direct interactions may differ in two ERDBD/ERE structures and still lead to stable complexes (10, 11). In another crystal structure, the glucocorticoid receptor (GR)DBD binds to a glucocorticoid response element (GRE) with a 4-bp spacer [GRDBD/GRE4], and reveals a stable complex, in which one GR monomer interacts specifically to one GRE half-site (HGRE), whereas the other GR binds nonspecifically to the adjacent DNA (12, 13). Furthermore, numerous studies on promoters containing cHERE (14–19), in addition to the human genomic studies on ER binding, showed that the majority of ER binding sites did not contain an cERE but contain one or more cHEREs (9, 20, 21), further reinforcing this idea for ER targeting to cHEREs. On the other hand, studies have reported weak ER binding to cHERE and to DR that have relatively long spacers (14, 15, 22, 23).

High-mobility group protein B1 (HMGB1) is a ubiquitous, abundant, and highly conserved DNA-binding protein that interacts nonspecifically with DNA in the minor groove to produce enormous DNA flexure (24, 25). It has been shown to enhance the binding of a number of diverse transcription factors, including human TATA-binding protein, p53, p73, Hox, Oct1 and Oct2, Rel proteins, the Epstein-Barr viral activator, ZEBRA, and the steroid hormone receptors (26–28). HMGB1 was shown to only moderately enhance ER binding affinity to cERE (28–30), to facilitate cooperative binding on tandem EREs (28), and to increase the level of transcriptional activation in transient transfection studies, in which ERE were used to drive reporter gene expression (29, 31). HMGB1 has also been shown to greatly enhance the binding of ER to a single cHERE (28, 32) and to cEREns, in which n = 0, 1, 2, and 4. Although ER binding occurs to all the ERE in this cEREn series, the presence of HMGB1 increased the binding affinity to a level comparable, in all cases, with that observed for ER binding to cERE in the absence of HMGB1 (28, 33).

Because the presence of HMGB1 facilitated strong in vitro ER binding to a single cHERE and cEREn (n = 1–4), one aim of this study was to extend the ER binding studies to cHEREs in DRs, an everted repeat (EvRs) and a widely separated inverted repeat of cHEREs. The functional relevance of in vitro binding studies was then tested in in vivo transient transfection studies, in which the luciferase reporter activity was driven by a series of constructs, including a series of DRs of cEREs and cHEREs, in addition to cEREs with spacers, n = 0–4. Finally, HMGB1 knockdown (KD) studies are consistent with and further support the conclusion that HMGB1 is an important coactivator for E2/ER-responsive genes. This study, together with previous studies, provides compelling evidence for the importance of expanding the current paradigm for ER binding and transcriptional activation. These findings may also have relevance in the development of cancer, because overexpression of HMGB1 has been implicated in a number of human cancers (34, 35).

Results

In vitro binding studies

We have previously shown that the presence of 400 nm HMGB1 enhanced the binding affinity of both isoforms of ER to a cHERE, with a similar effect on binding to a series of cEREn, in which n (0–4 bp) is the number of base pairs in the spacer (28, 33). To extend this list of potentially effective nonconventional ERE (ncERE), a series of DR, an EvR and an IR with a long spacer (Table 1), was examined, using the vitellogenin B1 estrogen response unit (ERU) (28, 36) as the basic template from which the spacing and cHERE orientations were derived. Figure 1 shows a schematic representation of the five ER binding motifs examined in this context. These binding studies suggest that this series of elements, with different orientations and spacers between cHERE, can be generally subdivided into two groups, with a representative binding profile for each group. Figure 2A, i and ii, shows the binding profiles for ERα and ERβ, respectively, to DR15 (15-bp separation) in the absence and presence of HMGB1. Because of our previous findings that the presence of HMGB1 was required for strong ER binding to a single cHERE (28), we were surprised to find that, in the absence of HMGB1, ERα binds strongly to one of the half-sites of the DR15, with a dissociation constant (K_d) = 7 nm. No binding to the second cHERE was detectable up to 60 nm ER. On the other hand, the presence of HMGB1 significantly altered the nature of the binding profiles, with the ERα binding affinity to a single cHERE (C1) increased further, and this single complex is observed only at the very lowest ERα levels. HMGB1 facilitates strong ER binding to both of the spatially separated cHEREs (C2) at ERα levels as low as 6 nm, with the concomitant decrease in the single-site population (higher mobility band). As the accompanying percentage complex plot shows, the K_d value for binding at the initial site is 4 nm. ERα binding is highly cooperative to both cHEREs, with 100% of the two sites completely occupied at approximately 15 nm. It also includes a plot for the percentage of C1 and C2 complexes as a function of ER concentrations and reveals that in the presence of HMGB1, there is a limited concentration domain for the C1 complex and a sharp increase in the percentage of C2 as ER levels increase. We use R₅₀ (i.e. level of ER at which there is equal concentrations of C1 and C2) as a semiquantitative indicator of cooperativity, which in this case is 10 nm. The ERβ binding profile in Fig. 2A, ii, is generally comparable, with HMGB1 producing a much weaker cooperative effect. The binding profiles for ERα and ERβ to the EvR (EvR6) (data not shown) are similar to those for DR15 in that ER binding is limited to one cHERE in the absence of HMGB1, with binding to both cHERE occurring in the presence of HMGB1. ERβ again exhibits a much lower cooperativity in binding to the two cHEREs.

Table 1.

Oligonucleotides used in ER binding studies

Name	Sequence (5′ to 3′)
cERE	TGATGCCTCCCCAACCTGGTTGGGCAACCTAGGTCActgTGACCTTCTTAGTTGG
cHERE	TGATGCCTCCCCAACCTGGTTGGGCAACCTAGGTCActgGTTGGGTCTTAGTTGG
DR15	TGATGCCTCAGGTCActgGTTGGGCAACCTAGGTCActgGTTGGGTCTTAGTTGG
EvR6	TGATGCCTCCCCAACctgTGACCTCAACCTAGGTCActgGTTGGGTCTTAGTTGG
IR24	TGATGCCTCAGGTCActgGTTGGGCAACCTCCCAACctgTGACCTTCTTAGTTGG
DR20	TGATGCCTCAGGTCActgGTTGGGTTGGGCAACCTAGGTCActgGTTGGGTCTTAGTTGG
DR3	TGATGCCTCCCCAACCTGGTTGGGCAACCTAGGTCActgAGGTCATCTTAGTTGG

Fig. 1.

Schematic representation of ncEREs (DRs, EvR, and IR). The relative position and orientation of cHERE in the DNA constructs for DRs, an EvR (ER), and an IR. The spacing was derived from the positions of ERE2/ERE1 in the vitellogenin B1 ERU (28, 36, 73), with the original HEREs replaced with either cHERE or a sequence that ER does not bind to (28). The number after the designated orientation (e.g. DR15) is the number of base pairs between the cHEREs.

Fig. 2.

ER binding profile to cHEREs in DRs, an EvR, and a widely spaced IR. A, EMSA and equilibrium ERα (i) and ERβ (ii) binding profile to DR15 in the absence and presence of 400 nm HMGB1. ER binding at one half-site (C1) and two half-sites (C2) in DR15. The percentage of DNA fragment in each complex was obtained from EMSA profiles and plotted as a function of ER concentration. Complexes formed in the presence of HMGB1 are indicated by H. The data represent the average of points from three different titrations and were fitted to a smooth curve by SigmaPlot. A 100 pm probe of DR15 was incubated with 0 nm ERα (lane 1) and 2, 6, 11, 17, 24, 31, 38, 45, and 52 nm ERα, respectively, in the absence (lanes 2–10) and in the presence (lanes 11–19) of HMGB1. The identical conditions were used for ERβ binding studies in B. B, EMSA and equilibrium ERα (i) and ERβ (ii) binding profile to IR24 in the absence and presence of 400 nm HMGB1. The conditions and profiles are identical to those in A.

Fig. 2B, i and ii, shows the binding profiles for ERα and ERβ, respectively, to an IR of the cHEREs separated by 24 bp (IR24). ERα binds to a single half-site at low ER levels, but binding to the second cHERE is observed at higher ER levels, with equal levels of C1 and C2 observed at approximately 50 nm. However, the binding clearly occurs in a noncooperative manner. On the other hand, in the presence of HMGB1, ERα binding to one half-site occurs only at the very lowest ER, with binding at both half-sites occurring in a highly cooperative manner as ER levels increase. The percentage complexation plot also reveals this behavior. The binding profile for IR24 also more closely describes the behavior for DR20, the DR in which the cHEREs are separated by 20 bp. The profile is likewise similar to DR3, in which the cHEREs are separated by 3 bp, as is found in cERE, but that the cHEREs are oriented as a DR and not as an IR. The binding profile for ERβ is similar in that the C2 complex is clearly detectable in the absence of HMGB1, but to a less extent, and the binding is noncooperative. Cooperative binding is again facilitated by the presence of HMGB1.

Table 2 summarizes the K_d and R₅₀ values for ERα and ERβ binding to cHERE, in each of the five contexts, in the presence and absence of HMGB1. The constructs are grouped according to whether or not ER binds to the two cHERE sites in the absence of HMGB1. The cooperativity factor, R₅₀, generally shows that there is little or no binding cooperativity in the absence of HMGB1 (i.e. a high R₅₀ value), whereas HMGB1 facilitates cooperative binding as reflected in a significantly lower R₅₀ value.

Table 2.

Summary of ER binding to DRs, EvR6, and IR24 in the presence and absence of HMGB1

DNA	ERα				ERβ
	w/o HMGB1		w HMGB1		w/o HMGB1		w HMGB1
	Kd	R50	Kd	R50	Kd	R50	Kd	R50
DR15	7.0		4.2	10	8.0		7.0	14
EvR6	7.2		7.0	14	8.5		8.5	28
IR24	12	53	3.0	5.0	16	80	9.0	13
DR20	10	70	4.5	8.0	23	68	6.0	18
DR3	11	65	4.8	11	11	65	4.0	23

R₅₀ is the concentration of ER at which there are equal levels of C1 and C2. The K_d values in all tabulations are a mean of three independent EMSA experiments with a sd of 15% or less for all determinations. w, With; w/o, without.

These data show that ER can bind very strongly to multiple cHEREs (DRs, EvRs, and widely spaced IRs) in vitro, both in the absence and presence of HMGB1. However, in the presence of HMGB1, ERα binds to multiple cHEREs and exhibits a pronounced cooperativity in its binding behavior, whereas the cooperative behavior is significantly less for ERβ.

Transient transfection and luciferase reporter activity

With the collective findings on in vitro ER binding to potential EREs (28, 33), it was of interest to determine whether other ncEREs that do not fit into the current ER binding paradigm, and are effectively identical or comparable with naturally occurring EREs, could serve as functional EREs. Accordingly, we asked 1) whether these elements could serve as EREs for ER-activated transcription, 2) the relative activity of each ncERE compared with the cERE, 3) the extent to which exogenous, overexpressed HMGB1 increased the transcriptional activity, and 4) whether there was a correlation between the in vitro ER binding affinity for the ERE and their functional activity in estrogen/ER-dependent activated transcription. The ncEREs are shown in Tables 3–5 and include a single cERE, or multiple (two or three) DRs of cEREs in tandem, with a parallel series of cHEREs as DRs. In addition, because of our previous ER binding studies with cEREns (n = 0–4), these elements were also included to make the study more inclusive (28, 33).

Table 3.

Oligonucleotides used in making EREs in luciferase reporter assay: cEREs

Insert	Strand	Sequence (5′ to 3′)
1cERE	Coding	TCGAGAGGTCACTGTGACCTA
	Noncoding	GATCTAGGTCACAGTGACCTC
2cERE	Coding	TCGAGAGGTCACTGTGACCTAGATCCGCAGGTCACTGTGACCTA
	Noncoding	GATCTAGGTCACAGTGACCTGCGGATCTAGGTCACAGTGACCTC
3cERE	Coding	TCGAGATCTAGGTCACAGTGACCTGCGGATCCGCAGGTCACTGTGACCTAGATCCGCAGGTCACTGTGACCTA
	Noncoding	GATCTAGGTCACAGTGACCTGCGGATCTAGGTCACAGTGACCTGCGGATCCGCAGGTCACTGTGACCTAGATC

Table 4.

Oligonucleotides used in making EREs in luciferase reporter assay: cHEREs

Insert	Strand	Sequence (5′ to 3′)
1cHERE	Coding	TCGAGAGGTCACTGGTTGGGA
	Noncoding	GATCTCCCAACCAGTGACCTC
2cHERE	Coding	TCGAGAGGTCACTGGTTGGGAGATCCGCAGGTCACTGGTTGGGA
	Noncoding	GATCTCCCAACCAGTGACCTGCGGATCTCCCAACCAGTGACCTC
3cHERE	Coding	TCGAGAGGTCACAGGTTGGGGCGGATCCGCAGGTCACTGGTTGGGAGATCCGCAGGTCACTGGTTGGGA
	Noncoding	GATCTCCCAACCAGTGACCTGCGGATCTCCCAACCAGTGACCTGCGGATCCGCCCCAACCTGTGACCTC

Table 5.

Oligonucleotides used in making EREs in luciferase reporter assay: spacer variants

Insert	Strand	Sequence (5′ to 3′)	Spacer Size
cERE0	Coding	TCGAGAGGTCATGACCTA	0
	Noncoding	GATCTAGGTCATGACCTC
cERE1	Coding	TCGAGAGGTCACTGACCTA	1
	Noncoding	GATCTAGGTCAGTGACCTC
cERE2	Coding	TCGAGAGGTCACGTGACCTA	2
	Noncoding	GATCTAGGTCACGTGACCTC
cERE3	Coding	TCGAGAGGTCACTGTGACCTA	3
	Noncoding	GATCTAGGTCACAGTGACCTC
cERE4	Coding	TCGAGAGGTCACTAGTGACCTA	4
	Noncoding	GATCTAGGTCACTAGTGACCTC

In initial transient transfections, different levels of pCMVflag:hERα expression vector were introduced to define the optimum level for luciferase expression. The luciferase reporter was driven by 3cERE in the ER-negative U2OS cell line (Supplemental Fig. 1A, published on The Endocrine Society's Journals Online website at http://mend.endojournals.org) treated with 10 nm E2. Maximum activity was observed at 5–10 ng vector, which is comparable with those reported previously (29, 37, 38).

With the ER level established in the presence of 10 nm E2, experiments in the presence and absence of E2 and ERα (Supplemental Fig. 1B) showed that the luciferase activity was both E2 and ER dependent. Supplemental Fig. 1C shows the relative activity as a function of increasing pHMGB1, whereas Supplemental Fig. 1D summarizes the results of a Western blot analysis for total HMGB1 in the cells and shows HMGB1 protein levels increasing with increasing levels of transfected pHMGB1 vector. As a result of these findings, all subsequent experiments were carried out at 5 ng pCMVflag:hERα and 1 μg of pHMGB1 or with 1 μg of pBlueScript used in lieu of pHMGB1.

In the first series of reporter constructs, expression was driven by one, two, or three cEREs in tandem, serving as either a simple cERE or an ERU. Figure 3 summaries 1) the relative increase in luciferase activity as the number of cERE increases, 2) the increase in the activity in the presence of exogenous HMGB1, 3) the effect of HMGB1 on luciferase activity for each element, and 4) the level of synergy in the absence and presence of HMGB1. Table 6 collectively summaries both the binding and the transcription data.

Fig. 3.

Luciferase activity for DRs of cEREs in the absence and presence of overexpression of HMGB1; 1 μg of p1cERE, p2cERE, and p3cERE were transfected into 4 × 10⁵ U2OS cells along with 5 ng of pERα and 1 ng of pGL4.70 hRLuc, with and without 1000 ng of pHMGB1 expression vector. Six hours after transfection, the cells were treated with 10 nm E2 for 24 h, the cells were harvested, and the cell extracts were assayed for luciferase activity. Data are displayed as the ratio of RLU of firefly and Renilla luciferase activity. These data are the mean ± sd of four separate determinations. Tabulations at the top of the plots indicate (i) the relative increase in luciferase activity from one cERE to two or three cEREs without pHMGB1, (ii) the same relative increase in this series in the presence of pHMGB1, (iii) the effect of pHMGB1on each element in the series, and (iv) the fold synergy on going from 1cERE to 2cERE and 3cERE, both in the presence and absence of pHMGB1.

Table 6.

ERα binding and transcriptional activation for single and tandem cEREs or cHEREs in the presence and absence of HMGB1

DNA	Kd (nm)		Effect	Transcriptional activity		Effect
	−HMGB1	+HMGB1		−HMGB1	+HMGB1
1cERE	7.4	5.1	1.4	18 ± 1.17	75 ± 4.98	4.2
2cERE	2	1	2.0	72 ± 6.37	321 ± 12.51	4.5
3cERE				117 ± 7.07	367 ± 11.09	3.1
1cHERE	80	15	5.3	5 ± 2.30	18 ± 5.86	3.6
2cHERE (DR17)	7	4	1.8	15 ± 3.34	55 ± 9.28	3.7
3cHERE (DR19/DR17)				42 ± 2.28	100 ± 10.58	2.4

Figure 3 shows that, in the absence of exogenous HMGB1, although a single cERE exhibits a relatively low activity, the luciferase activity increases sharply and progressively for two cERE (4.0-fold) and three cERE (6.5-fold). If transcriptional synergy for any multiple cERE can be expressed by their activity level divided by the number of cERE in the response element, relative to that for the single cERE (37), the synergistic effect of multiple cERE on luciferase activity can be evaluated. The synergy exhibited by two and three cERE is very high, being approximately 2.0 and 2.2, respectively. These data clearly indicate that 1) going from a single ERE to an ERU with increasing numbers of ERE, luciferase activity is progressively increased; and 2) there is a strong, but comparable, synergic effect for the multiple tandem cERE. This latter effect agrees generally with previous findings on multiple cEREs, except that no cooperativity or synergy was observed previously for two tandem repeats (37).

In the presence of exogenous HMGB1, Fig. 3 shows that the luciferase activity is greatly enhanced in the series. Within each response element, the activity is significantly increased, with the increased effect being a 4.2-, 4.5-, and 3.1-fold effect for one, two, and three cEREs, respectively. Using the same definition for synergy, the synergistic effect for the two and three cEREs are 2.1 and 1.7, respectively. Thus, for this series of cEREs, HMGB1 exerts an enormous effect on enhancing luciferase expression. Although there is a significant synergistic effect with multiple elements, the effect is not enhanced by the presence of HMGB1. HMGB1 increases the ERα binding affinity and likewise increases the level of transcriptional activation. Nevertheless, the effect of HMGB1 on enhancing transcriptional activation is much greater than its effect on enhancing binding affinity, indicating that HMGB1 plays additional roles in transcription or that other factors are involved. Table 6 collectively summarizes both the binding and the transcription data.

Figure 4 shows the luciferase activity that is driven by a single or multiple cHEREs in DRs. First of all, and as might be expected from previous studies with imperfect EREs (5, 38), the level of activated expression derived from a single cHERE is lower than that for a single cERE. Likewise, the level of activities for the DR containing cHEREs is also significantly less than that for the comparable cEREs. However, the activity for the DR with two or three cHEREs increases by 3.0- and 8.4-fold, respectively, over that for the single cHERE. The transcriptional synergic effect is 1.5 and 2.8 for two and three cHEREs, respectively. Similar to what was observed with the cERE series, the presence of HMGB1 exerts a very significant effect for each ERE, increasing the activity by 3.6-, 3.7-, and 2.4-fold for one, two, and three cHEREs, respectively. The presence of HMGB1, although increasing the transcriptional activity significantly, does not enhance synergy. Finally, it is of note that the effect of multiple cHEREs on transcriptional activation by ERα is comparable with, or greater than, that for many nonpalindromic EREs (5).

Fig. 4.

Luciferase activity for DRs of cHEREs in the absence and presence of overexpression of HMGB1; 1 μg of p1cHERE, p2cHERE or p3cHERE was transfected into 4 × 10⁵ U2OS cells along with 5 ng of pERα and 1 ng of pGL4.70 hRLuc with and without 1000 ng pHMGB1 expression vector. Experimental conditions are as outlined in Fig. 3. These data are the mean ± sd of four separate determinations. The numbers displayed on top of the figure for relative increase in activity (i) without pHMGB1, (ii) with pHMGB1, (iii) the overall effect of HMGB1 overexpression, and (iv) the fold synergy as outlined in Fig. 3.

Previous findings have shown that the binding affinity of ER to cEREn (n = 0–4) depends on n, with K_d values varying in the range of 10–80 nm. However, the presence of HMGB1 enhances the binding affinity to all cEREns (33), with the HMGB1 effect being generally inversely related to the ER binding affinity, such that ERα and ERβ bind to all these elements with comparable K_d values (5–15 nm) and in the range of the ER binding affinity observed to the cERE in the absence of HMGB1 (cERE3, K_d = 7.4 nm). With these results in hand, the influence of the spacer size in cEREn and HMGB1 on the relative luciferase activity is shown in Fig. 5, with the K_d values and transcriptional activities for these elements summarized in Table 7. The activity for cEREn for n = 1 and 4 are relatively low, being comparable with that for the cHERE. Surprisingly, the activity for cERE2 is greater than that for a single cERE (n = 3), whereas the activity for cERE0 is intermediate in value. In the absence of HMGB1, the order of increasing ERα binding affinity is: n = 1 < 2, 4 < 0, 3, whereas the order of the transcriptional activity is: n = 1, 4 < 0, 3 < 2, with n = 2 having an anomalous value in both series. Except for cERE1, for which HMGB1 had no effect on transcription, HMGB1 enhanced the expression driven by all other response elements, from 2.2- to 3.6-fold. The order of luciferase activity remained generally the same, irrespective of the absence or presence of the exogenous HMGB1. In marked contrast to the effect of HMGB1 on producing comparable levels of ER binding affinity (33), clearly this does not translate and lead to comparable levels of luciferase activity for these elements. Interestingly, the level of transactivation stimulated by HMGB1 on cERE0 is nearly as great as for cERE3, whereas the level of transactivation for cERE2 was comparable with that for the cERE (n = 3). In summary, the collective findings indicate that cERE and cHERE, the DRs of these elements, in addition to the elements in the cEREn (n = 0–4) series, confer significant E2/ER-specific responsiveness to luciferase expression in vivo. This therefore suggests that these elements can serve as functional ERE within human cells.

Fig. 5.

Luciferase activity for variant spacer cEREns in the presence and absence of overexpression of HMGB1; 1 μg of pcERE0, pcERE1, pcERE2, pcERE3, or pcERE4 was transfected into 4X10⁵ U2OS cells along with 5 ng of pERα and 1 ng of pGL4.70 hRLuc vector. Experimental conditions are as outlined in Fig. 3. These data are the mean ± sd of four separate determinations. The numbers displayed on top of the figure indicate the relative increase in activity (i) without HMGB1, (ii) with HMGB1, and (iii) the effect of HMGB1 as outlined in Fig. 3.

Table 7.

ERα binding and transcriptional activation for single EREs in the presence and absence of HMGB1

DNA	Kd(33) (nm)		Effect	Transcriptional activity		Effect
	−HMGB1	+HMGB1		−HMGB1	+HMGB1
cERE0	10	5.5	1.8	14 ± 2.13	46 ± 7.25	3.3
cERE1	80	15	5.3	7 ± 1.18	5 ± 1.22	(−1.4)
cERE2	25	7.3	3.4	38 ± 4.26	85 ± 8.62	2.2
cERE3	7.4	5.1	1.4	27 ± 2.16	98 ± 8.68	3.6
cERE4	25	12	2.1	7 ± 0.25	23 ± 4.55	3.3

Small interfering RNA (siRNA) KD of HMGB1 gene expression

The finding that overexpression of HMGB1 significantly enhances the transcriptional activity of virtually all the EREs in these series suggested that HMGB1 plays a prominent role in the regulation of transactivation for estrogen-responsive genes. To determine the extent to which endogenous HMGB1 was necessary for optimum transactivation of estrogen-responsive genes, the expression of the HMGB1 gene in the U2OS cells was KD by siRNA methodology. The housekeeping gene, human glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as an internal control or “calibrator” to normalize levels of luciferase activity in the presence of additional levels of the exogenous HMGB1. Figure 6A shows the effect that 200 and 400 nm siRNA HMGB1 oligos bring to bear on the luciferase activity driven by 3cERE. At both 200 and 400 nm siRNA, the luciferase activity was sharply reduced to less than 20%, compared with the control cocktail that produced no KD in expression. The level of luciferase activity was effectively unaffected by 200 nm nontargeting siRNA cocktail (data not shown). Figure 6, B and C, show that under these conditions, the level of cellular HMGB1 protein was reduced to about 70% of control. Based on these findings, these data clearly demonstrate that HMGB1 acts as an important coactivator and directly enhances estrogen-responsive gene expression.

Fig. 6.

Effect of siRNA KD of endogenous HMGB1 expression on E2/ER-mediated luciferase activity. A, Zero, 200, and 400 nm siRNA cocktail for human HMGB1 gene (or 100 nm nontargeting siRNA cocktail) (data not shown) (Dharmacon, Inc.), 5 ng of pCMVflag:hERα, 1 μg of p3cERE-Luc reporter vector, and 1 ng of pGL4.70 hRLuc control reporter vector were cotransfected into 4 × 10⁵ U2OS cells. Six hours after transfection, the cells were treated with 10 nm E2 for 24 h, the cells were harvested, and the cell extracts were assayed for luciferase activity. Data are displayed as the ratio of RLU of firefly to Renilla luciferase activity. These data are the mean ± sd of four separate determinations. Luciferase activity in the presence of 100 nm nontargeting siRNA was comparable with that for zero siRNA cocktail. B, Western blotting showing HMGB1 and GAPDH protein in cells transfected with siRNA, mock control, and untransfected cells. C, A quantitative representation of cellular levels of HMGB1 from the Western blotting calculated by quantifying band intensities of the HMGB1 and GAPDH using Odyssey version 3.0. The data are represented as the ratio of HMGB1 to GAPDH normalized to the mock control. These data are the mean ± sem of six separate determinations.

Discussion

In vitro binding studies

The conventional model for ER binding maintains that the ER dimer binds to a consensus or pseudopalindromic ERE, in which each ER monomer binds in the major groove of each half-site, with a 3-bp spacer size being important for dimer stability and to restrict target specificity, the binding interaction and ER/ERE complexation (1, 2). However, an increasing number of studies show that ER activates transcription from promoters containing nonconventional elements (ncERE) (5, 14, 17, 19, 23, 39, 40). In addition, genomic studies provide overwhelming evidence that ER binds to a spectrum of recognition sites, the majority of which do not contain cERE but do contain one or more cHERE sequences (9, 19). Numerous reviews have also drawn attention to the presence of ncEREs (5, 39, 40). However, a confounding question has been how to reconcile the experimental findings for diverse ER response elements within the confines of this current paradigm. This and previous studies with ncEREs provide substantial experimental support that highlights the enormous plasticity, not only of the ER binding to ERE but also extending to the ability to activate transcription. These results are in sharp contrast to parallel binding studies using ERDBD, which reveal a lack of plasticity in binding to ncEREns, with spacers other than n = 3 (11). Collectively, these data reinforce the notion that the minimal recognition element for ER binding is a HERE (5′-GGTCA-3′; cHERE) in a variety of contexts. Furthermore, in the presence of the transcriptional coactivator protein, HMGB1, ER binding, and transcriptional activity is significantly enhanced at these ncEREs that include one or more cHERE. The structural nature of the ER/ERE complexes can be expected to be very ERE dependent, as shown by protease digestion profiles for different ER/EREn complexes (33, 38).

We find that the ER dimer binds individually and with high affinity to a cHERE in a DR, IR, or EvR (Table 2) in the absence of HMGB1. In the cases in which ER binds to the second half-site, it binds noncooperatively. In contrast to this, the presence of HMGB1 increases the binding affinity for the initial binding and additional binding to the second site occurs in a cooperative manner. This is generally true for both ERα and ERβ. It is interesting to also note that although the individual half-sites are separated by different “spacer” lengths, have different orientations, and have different relative rotational orientations to each other, there are no distinctive differences in the binding profiles.

ER binding affinity and transcriptional activity

In light of conflicting previous reports (5), an additional objective was to determine whether the level of in vivo transcriptional activities for the different series of EREs were simply related to their relative in vitro binding activity. Tables 6 and 7 summarize these findings.

The first point is that the binding affinity for the formation of the initial complex, C1, is increased on going from a single cERE to two cEREs. This is in accord with previous findings for multiple elements (5, 41), and we also find this to be true for the cHEREs. In parallel with this, the transcriptional activity increases as the number of EREs in the promoter is increased. Generally, HMGB1 enhances luciferase activity much greater than its effect on ERα binding affinity, with a significant synergic effect in both the cERE and cHERE series, although the effect on cHERE is smaller than for the cERE. Clearly, multiple cEREs or cHEREs enhance ER binding and facilitate a greater transcriptional activation. There are very few documented ERE series in which a correlation of binding affinity and transcriptional activation has been shown, suggesting that many of these simple elements reside in complex promoters, in which multiple cellular factors participate to enhance transcriptional activity (5, 42).

ER binding affinity and transcriptional activity on single elements

The collection of ER binding affinities for the series of single EREs (cERE, cHERE, and cEREn) (28, 33) reveals that, with the exception of cERE2 in this series, there is a general correlation that shows that as the binding affinity increases, the level of transcription also increases (Tables 6 and 7). Curiously, however, although the K_d value for cEREn, for n = 2 and 4, were essentially the same, the transcriptional activity differ by more than 5-fold. Also, although the K_d values for cEREn, n = 1 and 4, were significantly different (80 vs. 25 nm, respectively), these elements drive luciferase expression to effectively the same level.

The effect of HMGB1 on ERα binding and its effect on transcriptional synergy on this same series show the same order (cERE3 < cERE0 < cERE4 < cHERE, ERE1), indicating a general correlation between binding affinity and the level of transcription, again with the exception of cERE2. Interestingly, cERE1 is the only element for which HMGB1 increases the ER binding affinity by more than 5-fold, yet exhibits no apparent effect on the level of transcriptional activation. It is also of note that although the presence of HMGB1 on the series of cEREns produced similar K_d values, a similar effect was not observed for the corresponding transcriptional activities, because they varied by a factor of 20.

Although HMGB1 is highly expressed in virtually all cell types (43), the finding that E2/ER-responsive gene expression is increased with further expression of exogenous HMGB1 indicates that, at least for transcriptional activation by E2/ER, HMGB1 is limiting in U2OS cells. This is consistent with previous findings in HeLa cells (31). It has been reported that overexpression of HMGB1, and its possible effect on gene expression, appears to be associated with a number of cancers and metastases (34).

KD of HMGB1 expression

The KD of endogenous HMGB1 gene expression in U2OS cells reflects the importance of HMGB1 in the regulation of E2/ER-activated transcription. The HMGB1 KD resulted in a sharp decrease in HMGB1 mRNA, with a decrease of approximately 30% in HMGB1 protein levels. This led to a drop in luciferase activity of over 80% and supports the contention that HMGB1 plays a significant role in the control of estrogen-responsive gene expression. Interestingly, in some studies, HMGB2 protein has been shown to exhibit an effect comparable with HMGB1 in effecting steroid hormone receptor binding and transcriptional activity (28, 44). However, HMGB2 was not KD in our studies, and the protein level determined by Western blotting was specific for HMGB1, indicating that the precipitous drop in luciferase activity in the U2OS cells was a direct result of HMGB1 KD.

How does ER bind to cHERE?

The current model for ER dimer interaction with the classical cERE was suggested from binding studies and supported by the crystal structure for ERDBD interacting with cERE. The latter showed that each ER subunit of the dimer inserts its P box (helix) into the major groove of the cHERE, with the 3-bp spacer affording an optimum spacing for D box interactions; thus providing stability in the dimer interface and specificity for the interaction (10, 11).

The data from our lab and others (5, 9) have shown that ERα binds to cHERE in a variety of contexts and stimulates transcriptional activation. One working model for the ER interaction with a single cHERE may be supported from the crystal structure of the (similar) GRDBD binding to an altered response element. The structure showed the complex of GRDBD bound to DNA with the two consensus HGRE (cHGRE) in an IR but with a separation of 4 bp instead of the usual three (12, 13) As a result, one GRDBD was bound to a cHGRE with the recognition helix in the P box inserted into the major groove, whereas the other GRDBD in the dimer is forced out of register with the cHGRE and bound nonspecifically to an adjacent sequence that exhibits no resemblance to the consensus half-site. This structure suggests that a sequence-specific interaction with one half-site may be sufficient in many cases for a binding interaction. In addition to this, wild-type ER contains the major dimerization function in the E domain that further stabilizes dimer binding and provides additional flexibility, with both effects providing sufficient stability for these interactions. This model may well be a mechanism that occurs in the binding to the DRs, the EvR, and the IR with a long spacer, in which allosteric interactions from binding at one site influences binding at other half-sites.

Furthermore, the ERα homodimer was shown to bind to the steroidogenic factor 1 binding element (TCAAGGTCATC) (32, 45). The deoxyribonuclease (DNase) I footprinting showed protection only within the single core element, suggesting that ERα was anchored by only direct contact with the single half-site (46). DNase I protection and ExoIII digestion of ERα/cERE and ERα/cHERE were distinctly different in the presence of HMGB1, in which the ERα/cHERE exhibited strong protection only at the cHERE, with significantly less protection over the adjacent sequence (28). It has also been reported that HMGB1 is required for binding to the cHERE (32). These observations are further supported by the general findings that the majority of EREs contain one or more consensus HEREs, in which the additional HERE sequence may have an altered sequence, with the altered sequences generally exhibiting a much less-defined DNase I footprint (41, 47, 48).

How does HMGB 1 enhance ER binding and facilitate a cooperative interaction between multiple ERE?

HMGB1 is a multifunctional, architectural protein that clearly plays a role in transcriptional activation (27, 28, 44, 49, 50). Perhaps the most notable physical characteristics of the HMGB1 protein is its nonspecific binding in the minor groove to DNA to facilitate enormous bending of the DNA, leading to an increase in the dynamic flexure in the DNA (24, 25, 51). In this regard, the increased level of flexure may be expected to permit ER to more effectively sample a greater conformational space, capture a favorably bent, low energy state, which would reduce the energetics of ER/cERE bending and binding, thus facilitating complexation. In addition, because ER binding itself bends DNA (52, 53), even in the absence of HMGB1, ER binding can itself serve as an allosteric effector on DNA (47, 48), as observed in our findings.

An alternate hypothesis for PR binding, which may be extended to ER, proposes that HMGB1 interacts directly with the C-terminal extension (CTE) in the DBD to alter the repressive conformation of the CTE and facilitate its enhanced binding affinity (32, 54). It was further shown that the intercalating residues that give HMGB1 much of its bending ability were not essential for the binding enhancement. This model may likewise apply to ER/cERE binding (32), and if so, it provides an additional, attractive strategy to enhance binding affinity.

How does HMGB1 enhance transcriptional activity?

In addition to its effect on enhancing ER/ERE binding as outlined above, evidence suggests that HMGB1 appears to play multiple functions when it comes to transcription. HMGB1 binds to a collection of so-called “HMGB1-sensitive” transcription factors, including human TATA-binding protein and TATA box binding protein-associated factor II230, both key component of TFIID TATA box binding protein in association with TAFs and essential for initiation in RNA polymerase II transcription (28, 55, 56). As a result of HMGB1 interactions with these components in the preinitiation complex and regulatory factors, HMGB1 may also contribute to transcriptional activation by serving as one of many conduits (coactivators) to assist the molecular communications from regulatory complexes, such as ER/ERE or other steroid hormone receptor/DNA complexes, to the transcriptional machinery. HMGB1 has also been shown to enhance the kinetics of the ATP-dependent chromatin remodeling complexes, chromatin accessibility complex (57). In addition, HMGB2, and presumably HMGB1, can serve as a coactivator that enhances SII and p300 functions in the transcriptional elongation process (58). In light of these multiple effects on transcription, HMGB1 can be expected to exert a multitude of variable and context-dependent effects in the transcription process, which would imply that, in many contexts, a simple correlation between ER binding affinity and transcriptional activation may not be very common.

Is ERα involved in functional cross talk between NHR?

DRs with different spacings serve as response elements for class II nuclear receptors. These receptors and ER, in contrast to the other steroid hormone receptors, bind to the common canonical HERE, 5′-AGGTAC-3′ (half-site), that occurs in DRs. We show that ER binds strongly to DRs and can drive significant levels of transcriptional activity and that HMGB1 enhances both these characteristics. This clearly shows that ER is extremely malleable and exhibits significant plasticity in its interactions. A notable characteristic of class II nuclear receptors is that, in addition to their DBD interaction in the major groove of the half-site, the CTE interacts in the minor groove of DNA to enhance their binding (1, 6, 59). Roemer et al. (54) have proposed that the CTE in ER and PR normally exhibit a repressive conformation that inhibits strong binding. However, they report that HMGB1 interaction with the CTE facilitates the CTE interaction in the minor groove of DNA, leading to stronger binding. This parallel in the HMGB1/ER binding characteristics, with those of class II NR, which are not assisted by HMGB1, supports the notion for potential competitive binding of ER with the class II NR. It has also been shown that the CTE is essential for ER binding to HERE (32). With respect to functional cross talk, it is well documented that the thyroid hormone receptor and ERα can both bind to the common recognition site, cERE0, and the cross talk between the two lead to differential gene expression (33, 60–62). In further support of this, numerous reports suggest that ERα and retinoic acid receptor participate in cross talk in breast cancer cells, with the evidence indicating that one mechanism for this “receptor cross talk” is competitive or differential receptor binding to a common response element (63–65). Because HMGB1 is implicated in and overexpressed in a number of cancers (34), its effect may further modulate or disrupt communication pathways in the affected cells. From these collective findings, we propose that the binding of ER to half-sites in DRs may provide a mechanistic route to a cross talk with class II receptors and perhaps some orphan receptors, such as estrogen-related receptor (45).

In summary, these findings for ER binding to ncERE, together with previous studies on E2-responsive genes and genomic-wide studies revealing ER binding sites in the human genome (5, 9), provide compelling evidence and highlight a growing recognition for 1) the significance of ER binding to and transcriptional activation associated with ncERE in E2-responsive genes, 2) the fundamental role that HMGB1 plays as a coactivator in enhancing both ERE binding and transactivation, and 3) a possible mechanism for cross talk between ER and other nonsteroid NHR that target single or multiple HEREs [5′-(A)GGTCA-3′]. Collectively, this suggests that the current paradigm for ER binding to EREs is far too restrictive and does not adequately account for its binding characteristics and its impact on transcriptional activation. Therefore, the current paradigm should be reexamined and broadened to be far more inclusive.

Materials and Methods

Proteins

Bacculovirus-expressed ERα and ERβ were purchased from Panvera Corp. (Madison, WI). The ER binding activity was determined to be greater than 90% by standard EMSA titrations with cold oligonucleotides (66). HMGB1 protein was isolated and purified from calf thymus as described previously (67) and was more than 95% pure as gauged by SDS-PAGE.

Oligonucleotides

The 55-bp oligonucleotides used in the binding study were obtained from Integrated DNA Technologies (Coralville, IA) and are listed in Table 1. The Xenopus laevis vitellogenin B1 ERU, a natural template that contains two non-cERE in tandem, with a separation of 20 bp center-to-center, was used as the basic template for making all the DRs, the EvR, and the IRs (36). With the spacing retained, one half-site in each ERE was changed to a cHERE, whereas the other half-site was modified to provide no detectable binding (28, 68); thus producing either a DR, an EvR, or an IR containing a long spacer region. In addition, the spacing between the cHERE in DR15 was increased by 5 bp to make the DR20 construct to determine whether the spacing influenced ER binding.

Electrophoretic mobility shift assay

EMSA were performed as generally outlined previously (28). The DNA fragments were ³²P-end-labeled and incubated with ER alone or both ER and HMGB1 proteins in ER binding buffer [80 mm KCl, 10% glycerol, 15 mm Tris-HCl (pH 7.9), 0.2 mm EDTA, 0.4 mm dithiothreitol, and 100 ng/μl BSA] with final concentration of poly(deoxyinosinic-deoxycytidylic) acid at 0.7 ng/μl. The proteins were incubated with buffer for 10 min at 4 C, and then the DNA reaction mixture was added, followed by a 15-min incubation at 4 C. All samples were electrophoresed in 0.35× Tris-borate, EDTA buffer at 200 V for 90 min in 4 or 5% nondenaturing polyacrylamide gels at 4 C. After electrophoresis, the gels were dried and exposed to x-ray film at −80 C. The K_d values were obtained from the titration data of 100 pm DNA over a range of ER concentrations, with equilibrium established after 15 min at 4 C. Experiments in which HMGB1 proteins were present were carried out with 400 nm HMGB1 protein. The band intensities for the complex and free DNA were measured by exposing the dried gel to a PhosphorImager screen, which was scanned using the Molecular Dynamics (Sunnyvale, CA) PhosphorImager system. The ImageQuant software program (Molecular Dynamics) was used to measure the band intensities, and the K_d was determined as previously outlined (28). The percentage of complex was plotted vs. the concentration of ER to generate the binding curves. The best fit of the data was derived (using SigmaPlot for PC) from three independent determinations.

Plasmid constructs

The firefly luciferase reporter plasmid was adapted from pGL2–3cERE-TATA-Inr-Luc (69), which contained three tandem cEREs to drive luciferase reporter expression. The series of reporter plasmids was constructed by excising this original response element by XhoI and BglII digestions and ligation of the other ERE that contained Xho and BglII linkers into the plasmid. The resultant plasmids were transfected into JM109 cells, and the transformed colonies were screened for the plasmids with the ERE insert using colony PCR. The sequences for a forward primer and a reverse primer are: 5′-AGCTCTTACGCGTGCTAGCT-3′ and 5′-TTACCAACAGTACCGGAATGC-3′, respectively. The PCR products were then analyzed for the DNA fragments containing ERE inserts on a 6% polyacrylamide gel. The constructs included a series consisting of a single cERE, two or three tandem repeats of cERE, a similar series of cHEREs, and a singular cEREn with spacer DNA are shown in Tables 3–5. The sequence of the constructs were verified by DNA sequencing. The cERE closest to the TATA-box is positioned at 38 bp away from the central nucleotide of the ERE spacer DNA to the center of the TATA-box, which is 3.6 helical turns from center-to-center from the TATA-box. This arrangement was reported to produce optimum activity (70). To construct the cHERE series, the HERE proximal to the TATA-box was modified to nonspecific sequence to provide no detectable ER binding (28, 68) while maintaining the 38-bp distance between the center of the spacer DNA to the center of the TATA-box.

Cell maintenance and transient transfections

The ER-negative human osteosarcoma (U2OS) cell line (71) was maintained in minimum Eagle's medium (MEM) (GIBCO, Freiburg, Germany) containing 15% heat-inactivated fetal calf serum, and cells were transferred to phenol red-free MEM supplemented with 5% charcoal dextran-treated calf serum (Transfection B medium) 2 d before transfection as described (72). Transient cotransfection experiments were carried out with luciferase reporter vectors. Cells (4 × 10⁵) were plated in each well of a 24-well plate and maintained in Transfection B medium for 18 h. Transfections were carried out using Lipofectamine LTX/PLUS reagent (Invitrogen, Carlsbad, CA) with 5 ng of the human ERα expression vector pCMV5-hERα (38, 73), 1 ng of the Renilla luciferase vector (pGL4.70 hRLuc) (Promega, Madison, WI), 1 μg of pHMGB1 or the empty vector, pBlueScript (29), and 1 μg of firefly luciferase reporter construct. After incubation with the Lipofectamine LTX/PLUS reagent/DNA, cells were maintained in Transfection B medium containing ethanol vehicle or 10 nm E2 for 24 h. Dual luciferase reporter assay (Promega) was carried out according to the manufacturer's protocol, and the luciferase activity was normalized by taking the ratio of firefly luciferase relative light units (RLU) to Renilla luciferase RLU.

siRNA cotransfection

The mixture of specifically designed siRNA oligonucleotides (SMART) cocktail of siRNA against human GAPDH and HMGB1, as well as (ON-TARGET plus) Nontargeting siRNA cocktail, were purchased from Dharmacon, Inc. (Lafayette, CO). The sequences for the siRNA for HMGB1 were derived from sequence analysis and specifically designed to target the 3′-untranslated region of the HMGB1 transcript. Importantly, these siRNA do not target the HMGB2 transcript. A Basic Local Alignment Search Tool search of the two genes further showed that there was a perfect match in the 3′-untranslated region in HMGB1 and none for HMGB2. Preliminary experiments to suggest the range of siRNA needed to significantly KD specific gene expression were determined by transfecting 50, 100, 200, and 400 nm siRNA SMART pool for GAPDH in 4 × 10⁵ U2OS cells. Based on these findings that showed that greater than 90% KD accomplished at 200 nm siRNA for GAPDH, 200 and 400 nm siRNA SMART pool of HMGB1 (or 100 nm nontargeting siRNA cocktail) was transfected into 4 × 10⁵ U2OS cells along with 1 μg luciferase reporter construct, 5 ng of ERα expression vector, and 1 ng of Renilla luciferase reporter vector using Lipofectamine 2000 (Invitrogen) as outlined in manufacturer's protocol. After incubation with the Lipofectamine 2000/DNA/siRNA, cells were maintained in Transfection B medium containing 10 nm E2 for 24 h. Dual luciferase reporter assay (Promega) was carried out according to the manufacturer's protocol, and the luciferase activity was normalized by taking the ratio of firefly luciferase RLU and Renilla luciferase RLU.

siRNA KD of cellular HMGB1 protein

One million U2OS cells per well were seeded in a six-well plate in DMEM (GIBCO) + 10% fetal bovine serum. The day after, the cells were either untreated or treated with Lipofectamine 2000 only (mock control), 200 nm nontargeting siRNA pool, and 200 or 400 nm SMART pool siRNA for HMGB1 using Opti-MEM (Invitrogen). After 6.5 h of transfection, the media were changed to DMEM + 10% fetal bovine serum, and cells were incubated at 37 C for 24 h. Cells were harvested using Hunter's lysis buffer (25 mm HEPES, 150 mm NaCl, 1.5 mm MgCl₂, 1 mm EGTA, 10 mm sodium pyrophosphate, 10 mm NaF, 0.1 mm sodium orthovanadate, 1% sodium deoxycholate 1% Triton X, 0.1% sodium dodecyl sulfate, and 10% glycerol) containing protease inhibitors (1 mm phenylmethylsulfonylfluoride, 10 μg/ml leupeptin, and 10 μg/ml aprotinin), which were added just before cell harvest. The total protein was quantified using bicinchoninic acid assay using BSA as a standard reference. Fifteen micrograms of total protein were separated electrophoretically in a gradient gel (4–12% Bis-Tris) (Invitrogen), and proteins were transferred to nitrocellulose membrane. HMGB1 was probed with 1 μg/ml of rabbit polyclonal anti-HMGB1 (Upstate, Lake Placid, NY). The immunoblot for HMGB1 was detected by using goat antirabbit antibody conjugated with fluorescent dye (IR Dye 800CW, Odyssey; LI-COR, Lincoln, NE) and visualized in LI-COR Odyssey Infrared Imaging System.

Real-time PCR

Real-time PCR was performed using SYBR Green PCR Master Mix (QIAGEN, Valencia, CA) and the ABI Prism 7500 series (Applied Biosystems, Foster City, CA). Quantitative PCR reactions were performed under conditions standardized for each primer. Primers (Real Time Primers, LLC, Elkins Park, PA) for quantitative PCR for GAPDH were: 5′-GAG TCA ACG GAT TTG GTC GT-3′ and 5′-TTG ATT TTG GAG GGA TCT CG-3′, whereas those for HMGB1 were 5′-CTC TGA GTA TCG CCC AAA AA-3′ and 5′-TTT TCA GCC TTG ACA ACT CC-3′.

NURSA Molecule Pages:

• Nuclear Receptors: ER-α | ER-β;

• Coregulators: HMG-1.