Binding of the ERα and ARNT1 AF2 domains to exon 21 of the SRC1 isoform SRC1e is essential for estrogen- and dioxin-related transcription

Nuclear receptors bind as homo- or heterodimers to specific recognition sites in the promoter regions of their target genes; for concomitant transcription, co-activator recruitment is essential for nuclear receptor function. Here we describe novel interactions of the SRC1 isoform SRC1e with ARNT1 and estrogen receptor α (ERα). SRC1e is a member of the nuclear receptor co-activator protein family (NCoA, SRC, p160), which interacts with and enhances the activity of nuclear receptors through their histone acetylase (HAT) activity (Spencer et al., 1997; Sterner and Berger, 2000); the activation domain 1 (AD1) of SRC1s binds to the C-terminal SRC1 interaction domain (SID) of the histone acetyltransferase p300, as well as its homologue, the cAMP response element binding protein (CREB)-binding protein (CBP) (Sheppard et al., 2001). p300 or CBP then links other transcriptional activators, basal transcription factors and HATs to the transcriptional machinery (Vo and Goodman, 2001). In hypoxic response, ARNT1, also named HIF-1β, together with the hypoxia-inducible factor-1α (HIF-1α) form the HIF-1 complex and recognize target genes through hypoxia responsive elements (HREs) in their promoter regions (Wenger, 2002). ARNT1 is also an obligate partner of the aryl hydrocarbon receptor (AhR) in response to halogenated aromatic hydrocarbons such as TCDD or polycyclic aromatic hydrocarbons (PAHs) such as 3-methylcholanthrene (3MC). Upon the uptake of these chemicals, the ligand-activated AhR–ARNT1 complex transcribes cytochrome P450s, UDP-glucuronosyltransferase and glutathione-S-transferase (GST) through xenobiotic responsive elements (XREs) within their promoter regions (Whitlock, 1999). In addition, ARNT1 is a potent co-activator of ERs because their ligand-binding domains (LBDs) interact after 17β-estradiol (E2) activation with the C-terminal transactivation domain (TAD) of ARNT1 (Brunnberg et al., 2003). It is commonly noted that ERs change their conformation after agonist binding of E2 to their LBD, leading to a helix 12 realignment with helices 3, 5–6 and 11 and thereby forming a lid on the LBD for the engulfed estrogen. Helix 12 contains a conserved activation function 2 (AF2) domain, which is in the E2-activated conformation the connective link to the LxxLL domains in the central nuclear receptor interaction domain (NID) of SRC1s (McInerney et al., 1998; Wu et al., 2005). Antagonist binding (e.g. that of raloxifen) prevents accurate helix 12 folding and the resulting incompetent conformation lacks SRC1 recruitment to the transcriptional machinery (Brzozowski et al., 1997). Here we describe human ARNT1.4, a novel splice variant of ARNT1.1, which is identical to ARNT1.1 except that it lacks exon 16. We found that exon 16 contains an unidentified AF2 domain that can bind to the C-terminal exon 21 of SRC1e. Exon 16 also contains two cyclin destruction boxes (D-boxes), one of which overlaps with the AF2 domain. The binding between exon 16 of ARNT1.1 and SRC1e exon 21 has no effect on the hypoxic response but is indispensable for the maximum TCDD response; moreover, the two D-boxes are essential for maintaining transcriptional upregulation. Interestingly, SRC1e exon 21 also binds to the AF2 domain of ERα in an estrogen-independent manner and further analysis revealed that this interaction is crucial for the estrogen response, without the involvement of ARNT1.1. An SRC1e splice variant lacking exon 21 [SRC1e(ΔC)] is reported in the human genome database without functional analysis. In summary, in this paper, we show that the transcriptional co-activator SRC1e occurs as splice variants with and without a specific binding site for the AF2 domains of two transcription factors ARNT1.1 and ERα, and this binding is essential for complete transcriptional activity. Moreover, ARNT1 also occurs as splice variants with (ARNT1.1) and without (ARNT1.4) the specific binding site for SRC1e. In addition, ARNT1.1 activation through SRC1e binding is highly dependent on two D-boxes, which are adjacent to its AF2 domain.

To obtain new splice variants of ARNT1, we amplified a human fetal brain cDNA library with ARNT1-specific primers. The resulting clones were digested with EcoRI and SacI. The restriction enzyme digestion pattern of ARNT1.1 exhibited fragments of 41, 86, 161, 945 and 1137 base pairs (bps), whereas that of ARNT1.3 exhibited fragments of 41, 86, 1061 and 1137 bps. An additional splice variant (ARNT 1.4) exhibited a fragment size pattern of 41, 86, 161, 945 and 1026 bps. The two smallest fragments appear as one band because of their low mass (Fig. 1A). After sequencing, we identified a new splice variant of ARNT1, hereby named ARNT1.4, which was identical to ARNT1.1 with exon 16 (aa 466–502) spliced out (Fig. 1B).

The top two ARNT proteins in Fig. 1B are the known splice variants in human cells: ARNT1.3 in which aa 77–91 are spliced out and the 325 aa truncated ARNT1.2, which was first detected in estrogen-receptor-negative MDA-MB-231 cells (Wilson et al., 1997) and was later related to poor prognosis in breast cancer (Qin et al., 2001). To confirm the occurrence of ARNT1.4 in other cells, we subjected a poly-A cDNA of MCF7 cells to PCR with primers amplifying bases 874–1518 of ATNT1.1 and bases 829–1473 of ARNT1.3, which both produced a 645 bp fragment spanning exon 16. With the same primer, ARNT1.4 without exon 16 produced a 534 bp fragment. As shown in Fig. 1C, both 645 bp and 534 bp fragments appeared after the amplification of the cDNA of MCF7 cells. Taking an aliquot of this PCR reaction mixture, we performed another PCR with the same forward primers as before but with an ARNT1.4-specific reverse primer spanning the junctions of exon 15 at the 3′-end and exon 17 at the 5′-end, which are connected in ARNT1.4. The resulting ARNT1.4-specific 519 bp fragment was confirmed by sequencing after TA-vector cloning (Fig. 1D). Thus, these results show that ARNT1.4 also occurs in adult mammalian cells.

We initially examined whether ARNT1.1 and ARNT1.4 exhibited differences in response to hypoxia and performed luciferase reporter assays with a reporter plasmid containing three hypoxic response elements (3×HRE) in HeLa cells co-transfected with HIF-1α, HIF-2α or HIF-3α, respectively, and either ARNT1.1 or ARNT1.4. Both variants induced the highest luciferase signals with HIF-1α and an increase in luciferase activity with HIF-2α and HIF-3α that was approximately half of that observed with HIF-1α. There was no difference regarding the hypoxic response between ARNT1.1 and ARNT1.4 (Fig. 2A). Fig. 2B shows transcription levels of cytochrome P450, family 1, subfamily A, polypeptide 1 (CYP1A1) determined by a quantitative real-time PCR (qRT-PCR). ARNT1-deficient Hepa 1c1c7c4 cells (Numayama-Tsuruta et al., 1997) were transfected with empty pcDNA3.1 plasmids or pCDNA3.1 expressing ARNT1.1 or ARNT1.4. The cells were incubated for 18 hours after transfection and then treated for 20 hours with 10 nM TCDD in 0.1% dimethyl sulfoxide (DMSO). When compared with ARNT1.1, the ARNT1.4 variant lacking exon 16 exhibited a substantially lower response to TCDD. This result implies that ARNT1.1 exon 16 is a crucial regulatory element of PAH-related transcription, although a western blot analysis of ARNT protein levels with the same conditions as for the qRT-PCR revealed that there was no difference in the amount of ARNT.1.1 and ARNT1.4 protein (Fig. 2B, top).

To determine the reason for the response to TCDD caused by ARNT1.1 exon 16, we subjected the amino acid sequence to an analysis and identified an AF2 domain that is highly homologous to other transcription factors (Danielian et al., 1992). We also discovered two D-boxes neighboring the AF2 domain with the core sequence RxxLxxxxQ/N. Further computer analysis of these motifs showed a high similarity to a cyclin A D-box of the first motif and a cyclin B D-box of the second motif with the latter overlapping four amino acids with the AF2 domain (Fig. 2C).

Because exon 16 of ARNT1.1 contains an AF2 domain, we assumed that ARNT1.1 exon 16 might be a binding site for TCDD-related transcription co-factors and we performed a yeast two-hybrid screening by using ARNT1.1 exon 16 as the bait and a human brain cDNA library as the prey. Three clones were identified as the C-terminal aa 1317–1399 stretch of steroid receptor co-activator 1e (SRC1e; NCBI accession NM_147223) in the positive colonies (Fig. 3A). Next we analyzed binding between ARNT1.1 exon 16 and endogenous SRC1e in an immunoprecipitation (IP) assay in HeLa cells transfected with HA-ARNT1.1 exon 16, in which the binding to endogenous SRC1e was clearly confirmed (Fig. 3B). Because of this result, and the fact that a SRC1e splice variant without exon 21 [SRC1e(ΔC)] is published in the human genome database (nuclear receptor co-activator 1e, isoform CRA_c; NCBI accession EAX00748) without a functional analysis, we decided to restrict the tentative SRC1e-interacting aa sequence to exon 21 and investigated whether SRC1e exon 21 and ARNT.1.1 exon 16 showed binding. Therefore, we performed an IP assay with HeLa cell lysates after co-transfection of MYC-ARNT1.1 exon 16-GFP and HA–SRC1e(ΔC) or HA-SRC1e exon 21 comprised of aa 1356–1385 (Fig. 3A blue aa stretch in the C-terminus of AD2). As evidenced in Fig. 3C, after immunoprecipitation with HA antibodies, MYC-ARNT1.1 exon 16-GFP bound only to the HA-tagged SRC1e exon 21 but not to HA–SRC1e(ΔC). This result showed that ARNT1.1 exon 16 clearly binds to SRC1e exon 21. To confirm this binding, we transfected HA-tagged SRC1e, GFP-tagged ARNT1.1, HA-tagged SRC1e(ΔC) and GFP-tagged ARNT1.4 in HeLa cells. As shown in Fig. 4 after single transfection, ARNT1.1 and ARNT1.4 localized to the nucleus, whereas SRC1e and SRC1e(ΔC) were distributed mainly in the cytosol in a scattered manner, as described in the literature (Amazit et al., 2003). When co-transfected, SRC1e(ΔC) and ARNT1.4 maintained their distribution patterns, with SRC1e(ΔC) in the cytosol and ARNT1.4 in the nucleus. By contrast, upon co-transfection with ARNT1.1, SRC1e was almost completely shuttled into the nucleus and colocalized with ARNT1.1 in nuclear foci (Fig. 5).

To confirm the functions of the three identified domains (D-box 1, D-box 2 and the AF2 domain) in exon 16, we generated several point mutations. According to previous reports, the AF2 domain was inactivated by converting the conserved methionine and glycine residues into alanines (Danielian et al., 1992), and the cyclin destruction boxes were inactivated by converting the conserved arginine residues into alanine (King et al., 1996). The effects of these various point mutations on the transcription of CYP1A1 upon TCDD treatment are shown in Fig. 6A. qRT-PCR analysis demonstrated that the effect of ARNT1.1 (lane 2) on CYP1A1 transcription was high (set to 100%), but that of ARNT1.4 was only 10.8% (lane 1). The AF2 domain without D-boxes could enhance the transcription of CYP1A1 by 17% (lane 3). Both D-boxes alone transcribed 37.8% CYP1A1 (lane 4), whereas an intact AF2 domain with one intact D-box induced transcription to 67.3% (D-box 1; lane 5) and 72.9% (D-box 2; lane 6) after 24 hours of TCDD treatment (Fig. 6A). A western blot with protein samples collected from cells treated the same way as described in Fig. 6A revealed that the protein expression pattern for CYP1A1 was the same as the transcription pattern in the RT-PCR analysis, though there was no visible difference in the protein levels of ARNT1.4 or ARNT1.1 and its mutants after TCDD treatment (Fig. 6B). The results indicate that the CYP1A1 expression level was not determined by the amount of available ARNT1, but by the D-boxes within ARNT1.1 exon 16 and to a lesser extent by the AF2 domain within ARNT1.1 exon 16. Surprisingly, ARNT1.1 mainly upregulates TCDD-related transcription through two D-boxes, in addition to the AF2 domain, without visible effects on its own protein level. We also compared CYP1A1 transcription in ARNT1-deficient Hepa 1c1c7c4 cells after transfection with wt ARNT1.1 and ARNT1.1 with mutated D-box 1 and D-box 2 (R→A) by investigating the time dependency of CYP1A1 transcription. As shown in Fig. 6C, the CYP1A1 transcription induced by wt ARNT1.1 and mutated ARNT1.1 with inactivated D-boxes was similar at 2 hours after TCDD induction, but with prolonged TCDD exposure, the transcriptional activity of mutated ARNT1.1 was gradually reduced compared with that of the wild-type ARNT1.1. These results revealed that the transcriptional upregulation through the AF2 domain in ARNT1.1 exon 16 essentially depends on the D-boxes in a time-dependent manner. To further analyze the degradation-dependent upregulation of ARNT activity, we transfected HeLa cells with equal amounts of SRC1e and wt ARNT1.1 or ARNT1.1 with inactivated D-boxes and incubated them with 1 μM 3MC, 1 μM 3MC plus 5 μM MG132 and 1 μM 3MC plus 10 μM MG132. As shown in Fig. 7A, application of the proteasome inhibitor MG132 enhanced the XRE-Luciferase signals to a higher level than 3MC alone. After 16 hours of incubation, the signals of wt ARNT1.1 and mutant ARNT1.1 shifted somewhat, but not to a significant level at that time point (Fig. 7A, left panel). Concomitant application of 1 μM 3MC plus 5 μM MG132 led to a rapid increase of Luciferase signals in cells transfected with wt ARNT1.1 with a peak at 8 hours and slow decrease until 16 hours after incubation start to the levels of mutant ARNT1.1 with inactivated D-boxes (Fig. 7A, middle panel). Incubation with the enhanced amount of 10 μM MG132 plus 1 μM 3MC showed equal signals for both over the entire time course (Fig. 7A, right panel). Fig. 7B shows western blot analyses of wt ARNT1.1 and the mutant ARNT1.1 protein levels after incubation for 8 hours without inducer, with 1 μM 3MC and 1 μM 3MC plus 5 μM MG132 (Fig. 7B left panel), as well as after incubation for 16 hours without inducer, with 1 μM 3MC or with 1 μM 3MC plus 10 μM MG132 (Fig. 7B right panel). There was no detectable change in the amounts of ARNT proteins after the indicated treatments and within the wild-type and mutant comparisons.

Considering the similarity between the AF2 domains of ERα and ARNT1.1, we hypothesized that ERα might also bind to SRC1 exon 21. After transfection of Myc-tagged ERα and Flag-tagged SRC1e exon 21 into HeLa cells, we harvested the cells after 24 hours and performed an immunoprecipitation assay with FLAG antibody beads. ERα could be detected in a western blot with Myc antibody, as a binding protein for SRC1e exon 21 (Fig. 8A). Fig. 8B represents an immunoprecipitation assay with endogenous ERα and Protein-G beads using MCF7 cells. The FLAG-tagged SRC1e exon 21 was detected by an antibody against FLAG in a western blot. Additionally, as shown in Fig. 8C, the binding of ERα to SRC1e exon 21 was dose dependent on exon 21 expression in HeLa cells, as determined by western blot analysis, clearly demonstrating that SRC1e exon 21 binds to ERα. For further measurements in a mammalian two-hybrid Gal4 assay, we generated ERα AF2 as a fusion protein with a Gal4 DNA-binding domain (Gal4-DBD) and SRC1e exon 21 as a fusion protein with a V16 transactivation domain. As seen in Fig. 8D, the ERα AF2 domain fused to the GAL4-DBD [ERα(AF2)–GAL4-DBD] alone enhanced the transcription rate as a result of endogenous activation, but co-transfected SRC1e exon 21 as a fusion protein with the VP16 activation domain doubled the luciferase signals of the ERα(AF2)–GAL4-DBD fusion protein alone in HeLa cells. Taken together, these results indicate that ERα AF2 binds to SRC1e exon 21, which is surprising because SRC1e exon 21 has never been reported to be an AF2 interaction site thus far.

Comparison of overexpression of single transcription factors with 0.3 μg or 0.6 μg transfected transcription factor plasmids revealed that only 0.6 μg ARNT1.1 could enhance TCDD-induced XRE-Luciferase reporter response to a higher level compared with the control, when overexpressed alone (Fig. 9, left). After co-transfection of MCF7 cells with 0.3 μg each ARNT1.1 and SRC1e, the relative XRE-Luciferase activity varied regarding enhancement of the TCDD and 3MC response. ARNT1.1 co-transfected with SRC1e led to a potent upregulation of XRE-Luciferase reporter signals, indicating that the activities of both factors are limited by the availability for the other factor during TCDD and 3MC response. ARNT1.4 co-transfected with SRC1e showed a higher response than co-transfected ARNT1.1 with SRC1e(ΔC), but both were significantly lower than ARNT1.1 co-transfected with SRC1e and significantly higher than ARNT1.4 co-transfected with SRC1e(ΔC), which showed the lowest enhancement. The same pattern as with TCDD also resulted with 3MC as inducer (Fig. 9, middle). A previous paper reports that ARNT1 is an enhancer of estrogen response through its C-terminal TAD binding to estrogen receptors (Brunnberg et al., 2003). To elucidate whether transcriptional differences in the splice variants regarding the TCDD response also occur in response to estrogen, we performed an estrogen responsive element (ERE)-reporter assay. As shown in Fig. 9, right, transfected ARNT1.1 and ARNT1.4 both enhanced ERE-Luciferase reporter response in MCF7 cells in a dose-dependent manner to the same level. Furthermore, we found that SRC1e distinctly enhanced estrogen response transcription, whereas SRC1e(ΔC) showed no enhancement. Also, in the co-transfection experiment, SRC1e with both ARNT1.1 and ARNT1.4 highly enhanced the estrogen response to the same levels, whereas SRC1e(ΔC) co-transfected with ARNT1.1 and ARNT1.4 yielded reduced signals (Fig. 9, right). This result indicates that SRC1e exon 21 plays an important role in estrogen response without the involvement of ARNT1.1 exon 16 and the splice variant SRC1e(ΔC) apparently has no function in estrogen response. To analyze the role of the binding between SRC1e exon 21 and the AF2 domain of ERα further, we overexpressed ERα in MCF7 cells with SRC1e, SRC1e(ΔC) as well as SRC1e plus SRC1e exon 21 and SRC1a exon 23 plasmids, respectively, and induced the cells with 10 nM E2, 10 nM E2 plus 1 μM tamoxifen (TAM) as well as with 1 μM TAM alone. Co-transfection of SRC1e with ERα strongly enhanced ERE-Luciferase signals upon E2 incubation. By contrast, SRC1e(ΔC) was a weak enhancer. Co-transfection of ERα plus SRC1e with SRC1e exon 21 and SRC1a exon 23 reduced the ERE-Luciferase signals, indicating that both exons are competitive inhibitors. Co-induction of 10 nM E2 and 1 μM TAM reduced the ERE-Luciferase signals, particularly of SRC1e, and 1 μM TAM alone reduced the Luciferase signals further (Fig. 10).

This is the first report on the interactions between ARNT1.1 and ERα AF2 with the C-terminal SRC1e exon 21, and each has a crucial function for enhancing TCDD- and estrogen-related transcription. Although the AhR–ARNT1.1 complex resembles a nuclear receptor, the location of the AF2 domain and the LBD is unusual because in nuclear receptors, both domains are located on one transcription factor as a functional unit (Warnmark et al., 2003), whereas in the AhR–ARNT1.1 complex, they are divided on AhR and ARNT1.1. Furthermore, we found that ARNT1.1 exon16 contains two D-boxes (Fig. 2C). In western blot and qRT-PCR analyses we evaluated the effect of these two D-boxes on protein expression and transcription activity of ARNT1.1 by comparing wt ARNT1.1 with deletion mutants of the AF2 domain, as well as the D-boxes. Surprisingly, except for the upregulated transcription of CYP1A1 – which relies on the AF2 domain – in response to TCDD, the two D-boxes are the main factors that determine transcription; however, they have no visible effect on the stability of the ARNT1 protein. (Fig. 6A,B). A time-course experiment revealed when D-boxes were inactivated, the transcriptional activity of this mutant gradually decreased over a period of 30 hours compared with the wild-type ARNT1.1 with intact D-boxes (Fig. 6C).

Generally TCDD and MG132 co-induction is performed to measure AhR-related activity, and most researchers pre-incubate cells with MG123 (Davarinos and Pollenz, 1999; Wormke et al., 2003; Pollenz and Buggy, 2006) or use only short time periods of 30 minutes to 5 hours of co-incubation (Roberts and Whitelaw, 1999; Ma and Baldwin, 2000). After application of proteasome inhibitors, the AhR pool is filled because of inhibition of its physiological degradation, and the TCDD response then rises to very high levels. In our measurements, after co-induction with 1 μM 3MC and 5 μM MG132 wt ARNT1.1 showed accelerated transcription until 8 hours after the start of induction. This might reflect the high availability of undegraded AhR, but could also reflect remaining proteasome activity for wt ARNT1.1 degradation, because this pattern was not visible for ARNT1.1 with inactivated D-boxes co-induced with 1 μM 3MC and 10 μM Mg132 (Fig. 7A). Dioxin-activated AhR is known to become unstable because after the AhR–ARNT1 complex attaches to the dioxin-responsive elements (DREs), AhR ubiquitylates itself and is degraded during each round of transcription (Ohtake et al., 2007). The half-life of AhR changes from 28 hours without a ligand to 3 hours during TCDD exposure (Ma and Baldwin, 2000). The remaining ARNT1.1 is recycled and serves with newly liganded AhR for multiple rounds of transcription. Because there are ten DREs within the 1000 bp upstream area of the mouse Cyp1a1 promoter (Sun et al., 2004), transcription must be rapid to allow the subsequent transcription factor complexes to become active. We suggest that, when an unproductive transcriptional pre-initiation complex assembled, ARNT1.1 is degraded by proteasomal recognition through ubiquitylation of the anaphase promoting complex (APC), thereby leaving a cleared promoter for the next round of transcription. An explanation as to why degradation of ARNT1.1 during response to PAH is not detectable (Fig. 2B, Fig. 6B, Fig. 7B), as demonstrated by previous measurements as well (Ma and Baldwin, 2000), is that the liganded AhR recruits only 15% of the ARNT1.1 pool (Pollenz et al., 1994; Pollenz, 1996). In addition, only a part of the ARNT1.1 involved in the PAH response might be degraded, and most of the molecules are recycled for several rounds of transcription, leading to the masking of losses by the undegraded bulk of ARNT1.1 (Lipford and Deshaies, 2003). Transcription factor activation combined with degradation has been reported for other transcription factors (Salghetti et al., 2000), but the destruction elements in these factors were described as acidic amino acids. The ARNT1.1 instead contains D-boxes similar to cyclin A and cyclin B D-boxes. The activity of the APC as the E3 ubiquitin ligase for proteins, which contain cyclin D-boxes, is not restricted to cell division because ubiquitylation also occurs during the entire G1 phase to control cyclins left from the M-phase and to prevent their accumulation during the G1 phase, thus inhibiting unregulated cell division (Lukas and Bartek, 2004). Taken together, our results imply that the activity of the AhR–ARNT1 complex regarding ARNT1 is regulated by two mechanisms. One is the binding of the ARNT1.1 exon 16 AF2 domain to exon 21 of SRC1e and the other one is the integrity of the two D-boxes in ARNT1.1 exon 16. In our experiments, the binding between SRC1e exon 21 and the ERα AF2 domain occurs without estrogen. Interaction between unliganded ER and SRC1 has been mentioned in previous papers (Tremblay et al., 1997; Llopis et al., 2000; Stenoien et al., 2001; Bai and Giguere, 2003; Koterba and Rowan, 2006; Sharp et al., 2006). More detailed reports regarding hormone-independent binding between AF2s in unliganded nuclear receptors and SRC1 have been published mainly for estrogen-related receptor alpha (ERRα) and ERRβ, both members of the orphan receptor family. These interact constitutively with SRC1 and glucocorticoid receptor interacting protein 1 (GRIP1), another member of the p160 family, as demonstrated in ERE assays (Xie et al., 1999; Zhang and Teng, 2000). SRC1e exon 21 does not contain an LxxLL motif, but AF2 binding to motifs other than LxxLL have been reported. The androgen receptor is regulated by the N-terminal–C-terminal (N/C) interactions of FxxLF, LKDIL and WxxLF with its AF2 domain (He et al., 2000; Burd et al., 2005; Askew et al., 2007). Another paper also described ER AF2 domain interactions with a motif other than LxxLL (Lee et al., 2005). It has been noted that comparing the two isoforms SRC1e and SRC1a for their transcriptional activation of ERs and thyroid receptors that SRC1a is a weaker enhancer of transcriptional upregulation than SRC1e. They are identical in aa 1–1385; however, SRC1a has a unique 56 aa sequence at the C-terminus that contains an additional LxxLL motif, and lacks the 14 most C-terminal aa of SRC1e (Fig. 3A) (Hayashi et al., 1997). Because SRC1a was detected as a transcription co-activator of nuclear receptors, the C-terminus (aa 1241–1441) was used as an inhibitor for nuclear receptor transcription (Onate et al., 1995) and named activation domain 2 (AD2). The inhibitory effect was on the one hand attributed to the binding competition of the additional LxxLL motif in the C-terminus of SRC1a, which blocked the estrogen-activated AF2 domain against binding to the central mainly LxxLL motif 2 (Mak et al., 1999) of SRC1, when expressed as a truncated C-terminal SRC1a fragment. On the other hand, a previous paper reported that deletion of the C-terminal LxxLL motif of SRC1a had no effect on transcriptional activity related to estrogen response (Kalkhoven et al., 1998). SRC1e exon 21 is located at the C-terminus of AD2 between aa 1355 and 1386 (Fig. 3A, blue), and we suggest that the estrogen-independent interaction of ERα AF2 with SRC1e exon 21 is a major transcription-enhancing mechanism in the AD2 of SRC1e. This is demonstrated by the fact that SRC1e(ΔC), whether transfected alone or co-transfected with both ARNTs, did not upregulate the luciferase signal during E2 induction, in contrast to the full-length SRC1e, which showed a robust E2 response (Fig. 9). Also in the ERα co-transfection experiments, the SRC1e(ΔC) response with E2 induction was the same as that of SRC1e with E2 plus TAM and even lower than the competitive inhibitory effects of SRC1e exon 21 and SRC1e exon 23 on SRC1e activity with E2 induction (Fig. 10). The ER conformation is reported to be essential for the binding of ligands (Gee and Katzenellenbogen, 2001) and in the unliganded state, heatshock protein 90 (Hsp90) maintains the ER in a conformation with high ligand-binding affinity (Fliss et al., 2000). Because of the ligand-independent mode of the interaction between SRC1e exon 21 and ERα AF2, it might be involved in ligand-binding affinity, as reported for AF2-dependent affinity for ligand binding in the vitamin D3 receptor (VDR) (Nayeri et al., 1996). As shown in Fig. 9, left and middle panels, ARNT and SRC1e when overexpressed alone did not enhance transcription to a maximum level and both factors depend on each other, especially in response to TCDD and 3MC. ARNT1.1 co-transfected with SRC1e(ΔC) showed a higher XRE-Luciferase reporter response than ARNT1.4 co-transfected with SRC1e(ΔC), which might be explained by the active D-boxes in ARNT1.1 in that context. In addition, the TAD of AhR binds to SRC1e aa 896–1200 (Kumar and Perdew, 1999), which also might explain the residual upregulation without SRC1e(ΔC) and ARNT1.1 exon 16 binding (Fig. 9). In addition, ARNT1.4 co-transfected with SRC1e showed a higher response than ARNT1.1 co-transfected with SRC1e(ΔC) after TCDD as well as 3MC induction. When SRC1e is co-transfected with ARNT1.4 into MCF7 cells, SRC1e exon 21 seems to react with another transcription-enhancing partner. Whether it is ERα, which would be a reasonable candidate according to our present findings, remains to be elucidated, but this is supported by a previous paper that noted that the liganded AhR–ARNT1 complex recruits unliganded ERα to the XRE in the promoter region of CYP1A1, thereby enhancing its transcription (Matthews et al., 2005).

The lack of a difference between ARNT1.1- and ARNT1.4-induced transcription enhancement indicated that ARNT1.1 exon 16 is not a transcription-determining factor during the estrogen response, which is in agreement with a previous report that found that the ARNT1 C-terminal 171 aa (aa 619–789) interact with the LBD of ERs (Brunnberg et al., 2003), excluding a direct interaction of ARNT1.1 exon 16, because it is located on aa 465–502. The enhanced estrogen response of co-transfected full-length SRC1e with both ARNT1.1 and ARNT1.4 to similar levels also indicates that exon 16 of ARNT1.1 is not involved in the estrogen response.

In summary, we discovered an ARNT1.1 D-box-regulated AF2 domain, which interacts with SRC1e exon 21. SRC1e exon 21 also serves as an estrogen-related binding site for the ERα AF2 domain. Both interactions are enhancers of TCDD or estrogen-related transcriptional responses, each in an independent manner. The occurrence of splice variants with and without each domain might differentiate them for specific transcription mechanisms, and the splice variant SRC1e(ΔC) and ARNT1.4 might be isoforms that are involved in transcriptional processes other than TCDD and estrogen responses, extending the role of transcription factors or co-factors from cell-specific availability (Spiegelman and Heinrich, 2004) to transcription-specific splice variants.

HeLa and MCF7 cells were grown in Dulbecco's modified Eagle's medium (Sigma) containing 10% heat-inactivated fetal bovine serum (Hyclone), 50 U/ml penicillin G, and 50 μg/ml streptomycin (Invitrogen) in a 5% CO₂ atmosphere. Hepa-1c1c7c4 mouse hepatoma cells, kindly provided by Keiki Nohara (National Institute for Environmental Studies, Tsukuba, Japan) were cultured in minimum essential alpha medium without nucleosides (Invitrogen) containing 10% heat-inactivated fetal bovine serum (Hyclone), 50 U/ml penicillin G, and 50 μg/ml streptomycin (Invitrogen) in a 5% CO₂ atmosphere. The cells were seeded at a density of 5×10⁴ per ml of the medium; the next day, they were transiently transfected using the calcium phosphate method with 8 μg total plasmid DNA/well (six-well tissue culture plates) or 20 μg total plasmid DNA per 60 mm tissue culture dishes, or alternatively with Lipofectamine 2000 (Invitrogen) following the manufacturer's instructions.

N-terminal Myc-, FLAG- or HA-tagged pcDNA3.1 (+) vectors were constructed in our laboratory; Myc-tagged human HIF-1α was purchased from Nevus Biologicals. Human HIF-2α and HIF-3α were gifts from Thomas Kietzman (University of Kaiserslautern, Kaiserslautern, Germany) and Shuntaro Hara, (Showa University, Tokyo, Japan), respectively. SRC1e was a gift from Yoshitaka Hayashi (Hayashi et al., 1997). Full-length cDNAs encoding ARNT1.1, ARNT1.3 and ARNT1.4 were amplified from a human fetal brain cDNA library (Takara) with ARNT1-specific primers, using the Expand High Fidelity PCR System (Roche Applied Science). The ARNTs were subcloned into pcDNA3.1 vectors. Point mutations were introduced into pcDNA3.1-ARNT1.1 with the Transformer Site-Directed Mutagenesis Kit (Clontech) according to the manufacturer's manual. Unless mentioned otherwise, constructs were cloned into pCDNA3.1. SRC1e exon 21 and SRC1a exon 23 were cloned from human DNA, and ERα was derived from MCF-7 cDNA using the Expand High Fidelity PCR System (Roche Applied Science). For the mammalian two-hybrid experiments, constructs were created according to the manufacturer's instructions.

MCF7 and HeLa cells were transfected with 0.2 μg pGL4.26/Luc2-minP vectors containing three estrogen response elements (EREs) or four xenobiotic response elements (XREs) and 0.6 μg indicated plasmids in single transfection experiments as well as 0.3 μg indicated ARNT and 0.3 μg indicated SRC1e in co-transfection experiments. Cells were treated with 10 nM 17β-estradiol (E2) and/or 1 μM tamoxifen (TAM) solved in a total of 0.1% ethanol, 1 μM 3-methylcholanthrene (3MC), 10 nM 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) as well as indicated amounts of MG132, all dissolved in a final volume of 0.1% DMSO and the maximum signals were set to 100% with the other values accounted accordingly. For hypoxia measurements, 0.8 μg DNA, including 3×HRE-VEGF-luciferase (pGL3-HRE-VEGF-Luciferase) firefly luciferase reporter vector was transfected into HeLa cells. Hypoxic treatment of HeLa cells was performed in an N₂-O₂-CO₂ incubator (ESPEC) in an atmosphere of 2% O₂ and 5% CO₂ at 37°C for 24 hours. Luciferase activity was determined with the dual Luciferase Assay System (Promega). As a reference plasmid to normalize transfection efficiency, Renilla RLSV40-luciferase (Promega) was co-transfected in all experiments. All transfections were performed using Lipofectamine 2000 (Invitrogen). Results are expressed as means for at least three independent experiments.

The yeast two-hybrid screenings were carried out using MATCHMAKER GAL4 Two-Hybrid System 3 (Clontech) according to the manufacturer's protocol. For bait construction, a cDNA fragment encoding ARNT1 exon 16, aa 466–502, was subcloned into the pGBKT7 GAL4 DNA-binding domain (DBD) vector. pGBKT7-ARNT1 exon16 was transformed into the yeast host strain AH109 (MATα strain). The transformed AH109 yeast strain was mated with a Matchmaker pre-transformed cDNA library of human brain (Clontech). Positive clones were selected by complementation of auxotrophic growth on medium lacking tryptophan, leucine, histidine and adenine, as well as by β-galactosidase signals using o-nitrophenyl β-D-galactopyranoside as a substrate.

Total RNA was isolated from Hepa 1c1c7c4 cells (plated on 60 mm tissue dishes) using Isogen (Nippon Gene). First-strand DNA was synthesized using a SuperScript First-Strand Synthesis System for reverse transcription PCR (Invitrogen). Real-time quantitative PCR was performed by TaqMan PCR using a 7000 Sequence Detection System with Assay on Demand probes (Mm 00487218 m1 and Mm 00607939 s1, Applied Biosystems). A standard curve for serial dilutions of murine β-actin was generated, and the relative standard curve method (Applied Biosystems) was used to calculate the levels of CYP1A1 expression. Results are expressed as means for at least three independent experiments. In all TCDD induction experiments, the maximum CYP1A1 transcription was set to 100%, and other values have been amended accordingly.

HeLa cells were plated on 18 mm micro coverglasses (Matsunami) and transfected with 2 μg of the appropriate plasmids, using the calcium phosphate method or Lipofectamine 2000 reagent. After 5 hours, the cells were washed three times with Dulbecco's modified Eagle's medium containing 10% fetal bovine serum and then incubated in a 5% CO₂ atmosphere chamber. After 24 hours, the cells were fixed in 3.7% paraformaldehyde in phosphate-buffered saline (PBS) for 10 minutes and washed with PBS containing 0.1% Nonidet P-40. For immunostaining, fixed cells were incubated with HA antibodies (12CA5 hybridomas, MBL) for 1 hour, washed, and then incubated with Cyanine 3 (Cy3)-conjugated secondary antibodies for 30 minutes. The cells were then washed and mounted in Vectashield (Vector Laboratories) mounting medium containing 4,6-diamidino-2-phenylindole (DAPI). Fluorescence images were visualized using an Olympus 1×70 inverted system microscope equipped with a charge-coupled device. The resulting images were analyzed using Metamorph computer software.

Immunoprecipitation assays were performed using a modified protocol from Sigma and Active Motif. Transfected HeLa cells were grown in a 21% O₂, 5% CO₂ atmosphere chamber and lysed with RIPA buffer containing 1 mM PMSF, 0.1 M DTT and protease inhibitor cocktail (Roche Applied Science); the whole cell lysate was then centrifuged at 15,000 r.p.m. for 45 minutes. The supernatant was incubated with 60 μl anti-HA-conjugated agarose (Sigma), normal mouse IgG-agarose (Santa Cruz Biotechnology), mouse anti-Flag M2-agarose (Sigma) or human ERα mouse monoclonal antibodies (MA1-310, Thermo) in combination with protein G Sepharose beads and incubated overnight at 4°C. After incubation, the resulting precipitate was washed four times with RIPA buffer and was then boiled at 95°C in 2× SDS sample buffer for western blot analysis. Total protein was separated on 10% SDS-polyacrylamide (SDS-PAGE) gels and then transferred to polyvinylidene difluoride membranes. The membranes were incubated overnight at 4°C in blocking buffer, followed by exposure to primary antibodies for 1 hour. Polyclonal anti-SRC1 antibodies were purchased from Abcam. Anti-Myc antibodies and anti-HA antibodies were purified from poly-Myc and 12CA5 hybridomas, respectively (Sigma, MBL). Antibodies against ERα (Thermo) and FLAG were purchased from Sigma. The membranes were incubated with horseradish-peroxidase-conjugated anti-mouse or anti-rabbit IgG (Pierce) for 1 hour and washed six times with PBS containing 0.1% Tween 20. Immunoprecipitate analysis was done using a SuperSignal West Dura chemiluminescent detection system (Pierce).

For mammalian two-hybrid assays, 8×10⁵ HeLa cells/ml were seeded and the next morning transfected with pGal4-Luc reporter vector along with pFN11-Gal4 DNA binding domain vector constructs and FN10 VP 16 activation domain vector constructs as indicated in the text in a 1:1:1 ratio using Lipofectamine 2000 (Invitrogen). For controls, each binding and activation construct was co-transfected with the control vector without the fusion protein according to the manufacturer's recommendation (Promega). After 48 hours of transfection, the cells were harvested and the luciferase activity was determined with the dual Luciferase Assay System (Promega). The pFN11A vector contains a SV40-promoter-driven Renilla luciferase, which serves as a control to adjust for transfection differences. Results are expressed as means of at least three independent experiments. The maximum value has been set to 100% and the other values were amended accordingly.

Results are expressed as the mean ± s.d. Comparisons using two-sample (unpaired) Student's t-tests were performed with Prism version 5 (Graph-Pad Software Inc., San Diego, CA) to identify significant group differences. *P<0.05, **P<0.01 and ***P<0.001.

We thank Keiki Nohara for providing the Hepa1c1c7 c4 mouse hepatoma cells and K. Uchida for her technical assistance. We also thank D. Wolf for his revision of our manuscript.