Endothelin-1 expression is strongly repressed by AU-rich elements in the 3′-untranslated region of the gene

Abstract

The regulation of the synthesis of the endothelial-derived vasoconstrictor ET-1 (endothelin-1) is a complex process that occurs mainly at the mRNA level. Transcription of the gene accounts for an important part of the regulation of expression, as already described for different modulators such as the cytokine TGF-β (transforming growth factor-β). However, very little is known about mechanisms governing ET-1 expression at the post-transcriptional level. The aim of the present study was to investigate the regulation of the ET-1 expression at this level. Since the 3′-UTR (3′-untranslated region) of mRNAs commonly contains genetic determinants for the post-transcriptional control of gene expression, we focused on the potential role of the 3′-UTR of ET-1 mRNA. Experiments performed with luciferase reporter constructs containing the 3′-UTR showed that this region exerts a potent destabilizing effect. Deletional analyses allowed us to locate this activity within a region at positions 924–1127. Some (but not all) of the AREs (AU-rich elements) present in this region were found to be essential for this mRNA-destabilizing activity. We also present evidence that cytosolic proteins from endothelial cells interact specifically with these RNA elements, and that a close correlation exists between the ability of the AREs to destabilize ET-1 mRNA and the binding of proteins to these elements. Our results are compatible with the existence of a strong repressional control of ET-1 expression mediated by destabilization of the mRNA exerted through the interaction of specific cytosolic proteins with AREs present in the 3′-UTR of the gene.

INTRODUCTION

Endothelial cells play a key role in the local regulation of blood vessel tone via the release of vasoactive agents that control the contractility and proliferation of smooth muscle cells. Among the most important substances released by endothelial cells are the vasodilator gas NO (nitric oxide) and the vasoconstrictor peptide ET-1 (endothelin-1), which together locally regulate vascular tone and blood pressure [1–3].

ET-1 is synthesized as a precursor molecule of 212 amino acids, preproET-1. This precursor undergoes several proteolytic cleavage steps to yield the bioactive form of 21 residues [3,4]. Healthy vascular endothelial cells synthesize and secrete ET-1 constitutively. Most evidence suggests that synthesized ET-1 peptide does not accumulate in specific intracellular stores. Therefore modulators of its expression act by increasing or decreasing levels of ET-1 mRNA [5]. Molecular analysis of the sequences upstream of exon 1 in the human ET-1 gene have revealed the existence of putative regulatory elements, i.e. binding sites for nuclear transcription factors, which are important for the endothelial-specific regulation of the gene as well as for the response to a number of different signals [e.g. hypoxia, thrombin, angiotensin II and TGF-β (transforming growth factor-β), among others] [6–10]. Nevertheless, the effects of some modulators on ET-1 expression cannot be explained only by transcriptional regulation. Steady-state levels of ET-1 mRNA (as for any other gene) are controlled not only by the rate of synthesis but also through post-transcriptional mechanisms, such as mRNA degradation. Indeed, decay constants are important determinants of gene expression, as they govern the time course of change after induction or repression, i.e. the time required to reach a new steady-state level [11–13]. Little is known about potential regulatory mechanisms controlling ET-1 expression at this level. However, there is evidence to suggest that ET-1 expression is subject to strong control through posttranscriptional mechanisms. First, several authors have documented that the ET-1 transcript is very labile (half-life ∼1 h in endothelial tissues) [6]. This is a feature of post-transcriptionally regulated genes, and suggests that ET-1 mRNA is under very active turnover. In addition, many of the processes dealing with specific regulatory mechanisms at the post-transcriptional level take place within genetic elements present in the 3′-UTR (3′-untranslated region) of the genes [14]. The 1.1 kb 3′-UTR of the human ET-1 mRNA accounts for over 50% of the transcript length, and specific regions of this sequence are highly conserved among different species. In particular, the 3′-end of the human ET-1 3′-UTR contains a number of AREs (AU-rich elements) of the form AU_nA (with n ranging from 3 to 7). AREs have been demonstrated to play important regulatory roles in cytokine and proto-oncogene expression by influencing transcript half-life under basal conditions and in response to cellular activation [15–18]. Recently, an interesting study has shown that the 3′-UTR of the ET-1 gene may contribute to the expression of the gene by controlling transcript lability, although the characterization of the operating mechanism is still incomplete [19].

Here we show that the intrinsically labile nature of its mRNA exerts a strong repressional control on ET-1 expression. Genetic determinants present in the 3′-UTR of the gene, in particular some (but not all) of the AREs, are the target for this regulatory mechanism. We also present evidence that cytosolic proteins interact specifically with these RNA elements, and that a strict correspondence exists between the functionality of the AREs in down-regulating ET-1 mRNA expression and their binding to these proteins.

EXPERIMENTAL

Cell culture

BAEC (bovine aortic endothelial cells) were isolated from thoracic aortas as described previously [20]. Cells were maintained in RPMI 1640 (Invitrogen) supplemented with 10% (v/v) calf serum (Biowhittaker), 100 units/ml penicillin and 100 μg/ml streptomycin at 37 °C in a humidified atmosphere with 5% CO₂, and were used up to the ninth passage. Cells for RNA experiments or cytoplasmic extract preparation were seeded on to 10 cm-diameter culture plates, whereas for transfection they were cultured on 6-, 12- or 24-well plates depending on the particular experiment.

Tet (tetracycline)-inducible HEK 293 cells stably expressing an ET-1 or luciferase gene fused to fragments of the 3′-UTR of the human ET-1 gene were cultured in Dulbecco's modified Eagle's medium (Invitrogen) containing 4.5 g/l D-glucose, 10% (v/v) foetal calf serum, 200 mg/ml hygromycin B and 15 mg/ml blasticidin.

RNA isolation and RNase protection analyses

Total RNA was isolated by guanidinium thiocyanate/phenol/chloroform extraction from confluent BAEC or HEK 293 cells treated as described in the Results section or in the Figure legends. To generate a luciferase cDNA plasmid for in vitro transcription, a 230 bp HinfI fragment of pGL3-Basic (positions 318–547) was cloned after blunting into the EcoRV site of pCR-Script, giving pCR-Luc-pGL3 as described [17]. A human ET-1 cDNA fragment was generated by PCR from reverse transcription products using the primers 5′-CCAAGGAGCTCCAGAAACAG-3′ (sense) and 5′-CTGTTGCCTTTGTGGGAAGT-3′ (antisense), corresponding to positions 379–398 and 625–644 respectively of the human ET-1 mRNA (GenBank accession no. XM_004293). The 266 bp PCR product was then TA-cloned into the pCR-II vector (Invitrogen) to give pCR-human ET-1. The plasmid pTRI-RNA-28 S containing a 115 bp cDNA fragment of the human 28 S rRNA gene was obtained from Ambion. To generate radiolabelled antisense probes, 0.5 μg portions of the correspond-ing linearized plasmids were in vitro transcribed using T7 or T3 RNA polymerases (Roche) and [α-³²P]UTP (NEN). To detect and quantify luciferase or human ET-1 mRNA levels, RNase protection experiments were performed as described [10,17]. For normalization, 28 S rRNA levels were also determined for each experiment. Densitometric analyses were performed with a FLA-3000 PhosphorImager (Fuji Photo Film Europe GmbH).

Construction of reporter plasmids and cell transfection

A human ET-1 cDNA fragment containing the 3′-UTR sequence of the gene was generated by digestion with EcoRI of the pUC13-full human ET-1 cDNA plasmid (kindly provided by Dr Kenneth Bloch, Cardiovascular Research Center, Massachusetts General Hospital, Charlestown, MA, U.S.A.). This fragment {partial 3′-UTR corresponding to positions 272–1127 plus a poly(A) tract of 22 nt, numbered relative to the first nucleotide of the 3′-UTR; GenBank accession no. J05008] [6]} was then cloned into the EcoRI site of pCR-Script, generating pCR-3′-UTR-human ET-1. This plasmid was used as the template for a PCR performed with primers UTR-2 and UTR-1 (Table 1) to generate the 3′-UTR fragment (272–1127) flanked by XbaI sites. The PCR product was digested by XbaI and cloned into the XbaI site of a luciferase reporter construct driven by a −650 bp fragment of the human ET-1 promoter (−650 bp-ppET-1 prom-luc) [10] to give ET-1 prom-luc-3′-UTR, named ‘full length’ throughout the text. Luciferase constructs containing deletion fragments of the 3′-UTR downstream of the reporter gene were obtained through the same cloning strategy, with the primers shown in Table 1. A similar set of constructs was generated with a luciferase reporter gene under the control of the Herpes simplex virus TK (thymidine kinase) promoter (TK prom-luc). This vector was constructed by cloning a BglII–HindIII TK promoter fragment from pRL-TK into the proper cloning sites of pGL3-Basic (Promega).

Table 1. Oligonucleotides used to generate constructs of the 3′-UTR of the human ET-1 mRNA, and (B) list of fragments generated.

(A) Nucleotides shown in bold carry specific sequences from the 3′-UTR. Oligonucleotides include an XbaI adapter to facilitate cloning (underlined). (B) The fragment name, the primer pair used to generate it and the fragment length in bp [including a poly(A) tract of 22 nt for those fragments including sequences up to the 3′-end] are shown.

(A)
Primer	Sequence	Target
UTR-1	GCTCTAGAAGGTCGACGGTATCGATAAG	Vector
UTR-2	GCTCTAGAATGAGTGGCTGCAGGAATTC	272–277
UTR-3	GCTCTAGAGACACAATGGTATAGGGTTG	810–829
UTR-4	GCTCTAGACAACCCTATACCATTGTGTC	829–810
UTR-5	GCTCTAGACCACATTTAATTATTGCCTCCC	609–630
UTR-6	GCTCTAGAGTACATCAAAATTCAATCCAACC	938–915
UTR-7	GCTCTAGAGAATTTTGATGTACTTATTTTTTTATAG	924–951
UTR-8	GCTCTAGACCCAAACAGATATTTAACATGG	1008–987
UTR-9	GCTCTAGACAGGCCATATTGGTCTATG	1009–1027
(B)
Fragment	Primer pair	Length (bp)
Full length (272–1127)	UTR-2/UTR-1	877
808–1127	UTR-3/UTR-1	341
609–1127	UTR-5/UTR-1	540
272–831	UTR-2/UTR-4	559
808–938	UTR-3/UTR-6	130
924–1008	UTR-7/UTR-8	84
1009–1127	UTR-9/UTR-1	140
924–1127	UTR-7/UTR-1	225

A series of −650 bp-ppET-1 promoter-driven luciferase reporters fused to fragment 924–1127 with specific mutations in ARE₃, ARE₄ and/or ARE₅ (substitutions of AUUUA to GGGCC) were generated by the synthesis of a pair of partially overlapping 54-mer oligonucleotides corresponding to the 85 nt fragment 924–1008. These oligonucleotides were annealed, single-strand-filled by Taq polymerase, and TA-cloned into the pCR-II vector to give the corresponding plasmids pCR-mutant 924–1008. The fragment 1009–1127 from pCR-II-1009–1127 (see below) was excised with XbaI and cloned into the SpeI site of the pCR-mutant 924–1008, to generate the plasmid pCR-mutant 924–1127. These constructs were then used as the template for a PCR reaction performed with primers UTR-7 and UTR-1 flanked with SpeI instead of XbaI sites. The PCR products were then digested with SpeI and cloned into the XbaI site of the luciferase reporter construct driven by a −650 bp fragment of the human ET-1 promoter, to finally yield the corresponding series of −650 bp-ppET-1 prom-luc-mutant 924–1127 plasmids with single mutations and also with combinations of these mutations. All constructs were verified by sequencing.

Transient transfection experiments were performed with BAEC or HEK 293 cells seeded on 12- or 6-well plates (60–70% confluency), using methods described previously [10]. Briefly, cells were transfected in OPTIMEM medium (Invitrogen) containing 0.5 μg of the luciferase plasmid, 0.1 μg of pRL-CMV (a Renilla luciferase under the control of the cytomegalovirus promoter; Promega), and 3 μl of Lipofectamine™ (Invitrogen), in accordance with the manufacturer's instructions. After 3–5 h, the DNA/lipofectamine-containing medium was removed and cells were incubated in complete medium (including serum and antibiotics) for 16 h. In some experiments, 5 ng/ml TGF-β (recombinant human TGF-β1, CHO-cell derived; R&D Systems) was included during this period. Then cells were lysed, and luciferase activity was measured and normalized using the Dual-Luciferase Reporter Assay system (Promega) using a TD-20/20 Luminometer (Turner Designs Instruments). In a parallel set of experiments, transfected cells were processed for RNA isolation as described above.

To estimate the half-lives of mRNAs in a Tet-inducible expression system, luciferase casettes fused to ET-1 3′-UTR fragments were cloned as KpnI–NotI fragments into the vector pcDNA5/FRT/TO (Invitrogen), to obtain the corresponding pcDNA5/FRT/TO-luc-3′-UTR plasmids. These constructs were then co-transfected with the Flp recombinase expression plasmid pOG44 into the Flp-In T-REx 293 cell line (Invitrogen). Cells of this line stably express the Tet repressor and contain a single integrated FRT (Flp recombination target) site. Flp recombinase expression from the pOG44 vector mediates insertion of luc-3′-UTR cassettes into the genome at the integrated FRT site through site-specific DNA recombination. After 48 h, transfected cells were selected for hygromycin B and blasticidin resistance by adding 200 mg/ml hygromycin B and 15 mg/ml blasticidin. Cell clones appeared after 10–15 days. Isogenic pooled clones were expanded and used for experiments of time course induction by the Tet analogue Dox (doxycycline) (Sigma). A similar strategy was used to generate stable transfectants with plasmids pcDNA5/FRT/TO-ET-1-Full length and pcDNA5/FRT/TO-ET-1-ΔUTR exhibiting Tet-inducible expression of ET-1 mRNAs with or without the 3′-UTR respectively.

Preparation of cytosolic extracts and in vitro degradation experiments

Cytosolic extracts were prepared according to a procedure already described [21]. Briefly, cell plates were washed once with ice-cold PBS and scraped into the same solution. Upon centrifugation they were resuspended in 10 mmol/l Hepes, pH 7.9, 10 mmol/l KCl, 0.1 mmol/l EDTA, 0.1 mmol/l EGTA, 1 mmol/l dithiothreitol, 0.5 mmol/l PMSF and 1 μg/ml each of leupeptin, aprotinin and pepstatin A (Sigma). After 15 min of incubation on ice, Nonidet P-40 was added to a final concentration of 0.5% and cells were vortexed for 10 s. Nuclei were then sedimented by centrifugation for 5 min at 14000 g, and the supernatant was stored at −80 °C as a crude cytosolic fraction. Protein content was determined with the BCA Protein Assay Reagent Kit (Pierce).

Degradation assays were performed as described [18]. Total RNA extracted from stable transfectants of 293 cells expressing ET-1 mRNAs with or without the 3′-UTR was used as the source of RNA. Samples of 20 μg of these RNAs were incubated at 30 °C with cytosolic extracts from BAEC (20 μg of protein) and, at each time point indicated, the RNA from these incubations was isolated by standard procedures and assessed by RNase protection experiments in order to detect and quantify human-derived ET-1 mRNA. The remaining (undegraded) ET-1 mRNA was quantified after normalization to 28 S rRNA levels as described above.

RNA binding experiments

PCR products corresponding to fragments of the 3′-UTR of the human ET-1 gene (wild type or mutant) were also cloned into the pCR-II vector (Invitrogen), which allows in vitro transcription of sense transcripts for RNA binding experiments by RNA-EMSAs (electrophoretic mobility shift assays). To generate radiolabelled 3′-UTR sense probes, 0.5 μg of DNA (linearized or PCR fragments) was transcribed in vitro as described above. Radiolabelled transcripts were analysed by urea-denaturing electrophoresis to estimate the yield and specific radioactivity. Incorporation of radioactivity in transcripts was usually >80%, and the specific radioactivity ranged from 0.2 to 0.5 μCi/pmol. Transcripts for competition were generated by the same procedure but using unlabelled UTP.

RNA binding experiments were carried out using 5–8 fmol of labelled RNA (approx. 60000 c.p.m.) and 4 μg of protein from cytosolic extracts in a final volume of 14 μl of binding buffer (50 mM NaCl, 1 mM MgCl₂, 0.5 mM EDTA, 5 mM dithiothreitol, 5% glycerol and 10 mM Tris, pH 7.4, containing 0.2 mg/ml tRNA as competitor for unspecific binding). After 15 min at room temperature, 6 units of RNase T1 (Roche) was added for 20 min to digest unbound RNA. RNA–protein complexes were resolved on a non-denaturing 5% (w/v) polyacrylamide gel in 0.25×Tris/borate/EDTA buffer and visualized by autoradiography. In competition experiments, cytosolic extracts were preincubated with the RNA competitor at the concentrations indicated in the Figure legend for 10 min prior to the addition of the radiolabelled probe.

RNase T1 mapping

Reaction mixtures contained radiolabelled RNA and cytosolic extracts in binding buffer as described above. After 15 min at room temperature, RNase T1 (6 units) was added and the reaction was continued for a further 20 min. The mixtures were diluted 1:6 with binding buffer and filtered through nitrocellulose (PROTRAN BA85; Schleicher & Schuell). After washing the nitrocellulose three times with 1 ml of binding buffer, bound RNA–protein complexes were extracted from the nitrocellulose filter with phenol/chloroform. The selected RNA was then ethanolprecipitated and resuspended in 0.3 M NaCl, 5 mM EDTA and 10 mM Tris/HCl, pH 7.4. Aliquots of this RNA were subjected to an additional RNase T1 digestion. Then, products were phenol-extracted and ethanol-precipitated. Fragments were resuspended in formamide/RNA loading buffer, analysed by 10% (w/v) polyacrylamide/urea gel electrophoresis and visualized by autoradiography [22].

Statistical analysis

Experimental data were analysed using the unpaired Student's t test in the case of normal distribution of data, or non-parametric tests as appropriate. The P values obtained are indicated in the Figure legends when statistically significant.

Rates of decay of 3′-UTR mRNA species in Tet-regulated stably transfected HEK 293 cells were estimated by measuring luciferase activity before and after induction of cells with 1 μg/ml Dox for time periods ranging from 2 to 34 h. Luciferase induction kinetics for each cell line were fitted to an equation of the form:

where t and R_t (luciferase activity at a given t) are the variables, R₀ is the initial value, R_ss the value at the stationary state, and K_d the decay constant. K_d values for the R_ss that gave the best fit were chosen with the program FigP (Fig P Software Corporation) and are given as h⁻¹ (value±S.E.M.; ±95% confidence limits). These decay constants were used to calculate the luciferase protein half-life for these induction kinetics (t_1/2=0.693/K_d). As luciferase protein turnover was identical in all of the cell lines (3 h by cycloheximide experiments; results not shown), mRNA half-lives for the corresponding transcripts were estimated by subtracting this 3 h value from the protein half-life value [12,13].

RESULTS

The ET-1 3′-UTR controls the expression of the gene

Synthesis of ET-1 mRNA by vascular endothelial cells is constitutive, but potentially regulated by a variety of biological and pharmacological agents, which may act through transcriptional and post-transcriptional mechanisms. One of the most potent modulators of ET-1 expression is the cytokine TGF-β. We have recently shown that TGF-β increases transcription of the ET-1 gene by activating the Smad signalling pathway [10]. In the course of these experiments we observed that activation by TGF-β of the expression of a construct of the human ET-1 promoter (−650/+173) fused to a luciferase reporter differed from that of endogenous ET-1 mRNA; the promoter/reporter construct was slower to reach the plateau phase and showed a wider range of values. This behaviour suggested the existence of additional regulatory mechanisms for the control of ET-1 expression unrelated to its promoter sequence that modulate the response kinetics of a transcriptional activator such as TGF-β. As many post-transcriptional mechanisms (such as mRNA stability) are determined by cis-acting elements in the 3′-UTR, we decided to focus on the 3′-UTR of the human ET-1 gene.

In the present work, we have generated a luciferase reporter under the control of a −650 bp fragment of the human ET-1 promoter, with the 3′-UTR of the gene (272–1127 fragment, denoted as full length) cloned downstream of the reporter gene. We then performed transient transfection studies in BAEC and analysed the luciferase activity present in the cells. Under basal conditions, cells transfected with this full-length 3′-UTR construct showed significantly lower luciferase levels than cells transfected with a control plasmid, ΔUTR, that lacked the 3′-UTR (Figure 1A). Exposure of cells transfected with the full-length or ΔUTR construct to TGF-β evoked completely different responses. Whereas cells transfected with the full-length construct showed fast and saturating kinetics of low amplitude, luciferase activity from cells transfected with the ΔUTR construct was potently increased. However, the increment did not plateau during the experimental period. Similar results were found when we analysed luciferase mRNA levels by RNase protection assay (Figure 1B). These observations show that the deletion of the 3′-UTR results in a drastic change in the induction response of the reporter gene (higher constitutive activity, but lower sensitivity to rapid transcriptional changes). Thus the behaviour of the full-length construct resembles more closely that of the endogenous ET-1 gene [10]. We hypothesized that the 3′-UTR could act by exerting strong repressional control at the post-transcriptional level, probably mediated by a constitutive destabilization of the mRNA.

Figure 1. Time course of TGF-β-induced expression of luciferase in BAEC transfected with luciferase reporters driven by the human ET-1 promoter.

Full-length construct contained the 3′-UTR of the ET-1 gene downstream of the luciferase reporter, whereas ΔUTR lacked the 3′-UTR. After transfection, cells were either left under basal conditions or treated with 5 ng/ml TGF-β for different times. (A) Luciferase (Luc) activity was measured by luminometry. (B) mRNA expression was analysed in quantitative RNase protection experiments, as described in the Experimental section (upper panel, results of densitometric analysis of RNase protection gels; lower panel, representative gel showing luciferase and 28 S protected bands). Values are expressed as normalized units for luminometry, and as a percentage of the value for the full-length construct in the absence of TGF-β for RNA analysis after normalization with 28 S rRNA signal, and are means±S.D. (n=4). The positions of protected luciferase and 28 S mRNA are indicated, as is the lane with RNA from non-transfected cells (C).

Distal elements in the 3′-UTR mediate a constitutive destabilization of ET-1 mRNA

The 3′-UTR of the ET-1 mRNA contains elements that may mediate regulatory post-transcriptional mechanisms. In particular, the ET-1 3′-UTR mRNA includes a number of AREs, classical motifs found in mRNAs which have a short and finely-regulated half-life. In order to study whether the destabilizing effect can be assigned to specific regions of the 3′-UTR, we divided the full-length sequence into three smaller fragments (272–831, 808–1127 and 609–1127). AREs in the 3′-UTR of the human ET-1 mRNA are located predominantly in the distal portion, with three runs of AUUUA, one of AUUUUUA and one of AUUUUUUUA within a short region spanning positions 940–1035 (see Figure 5). A sparse AUUUA is also found at positions 614–618 (ARE₁). Subfragment 272–831 contains only ARE₁, 808–1127 contains the distal AREs, and 609–1127 extends upstream with respect to the latter and includes ARE₁. ET-1-promoter-driven luciferase reporters fused to these subfragments were transiently transfected into BAEC and luciferase activity measured by luminometry. Figure 2(A) shows that the 609–1127 and 808–1127 constructs mimicked the effect of the full-length 3′-UTR, whereas the 272–831 construct was similar to ΔUTR, with no destabilization observed. These results suggest that genetic elements involved in the mechanism map to the region 808–1127, which contains all of the AREs except for ARE₁. Essentially similar results were obtained with 3′-UTR constructs whose expression was under the control of the Herpes simplex virus TK promoter (Figure 2B), an observation which suggests that this effect does not depend on the promoter driving the expression, but only on the elements present in the 3′-UTR. In addition, the same pattern of luciferase activity was also found in transient transfection experiments performed in HEK 293 cells, showing that the mechanism is operative in both endothelial and non-endothelial cell types (results not shown).

Figure 5. Further mapping of destabilizing elements in the distal portion of the ET-1 3′-UTR.

(A) Luciferase (Luc) reporters driven by the human ET-1 promoter fused to subfragments of the 3′-end of the ET-1 mRNA 3′-UTR were transfected into BAEC and activity measured. Values are expressed as a percentage of luciferase activity of ΔUTR after normalization with Renilla control plasmid (n=6; *P<0.01 compared with ΔUTR value). (B) Sequence of the 3′-end (positions 808–1127) within the 3′-UTR of human ET-1 mRNA. AREs shown schematically in (A) are underlined (ARE₁ is present at positions 614–618).

Figure 2. Luciferase activity of ET-1 3′-UTR-derived constructs.

Luciferase cassettes driven by either the human ET-1 promoter (A; ET-1 prom) or the viral Herpes simplex TK promoter (B; TK prom), and fused to the indicated fragments of the 3′-UTR of ET-1 mRNA, were transfected into BAEC and reporter expression was measured. Values are expressed as percentage of luciferase (Luc) activity for the corresponding full-length construct after normalization with Renilla control plasmid (n=6; *P<0.01 compared with the corresponding full-length value).

Classical methods for estimating mRNA half-life values use transcriptional inhibitors such as actinomycin D or 5,6-dichloro-1-β-ribofuranosyl benzimidazole. However, these drugs are not the best choice when analysing the regulation of ARE function, as they clearly interfere with the mRNA decay machinery [23,24]. Approaches involving metabolic pulse-labelling of the RNA pool followed by chase and detection of specific mRNA molecules are less problematic methodologies, although very tedious. A method based on the use of the Tet-regulatory promoter system has been described as a good alternative to address mechanisms involving post-transcriptional regulation [25]. In the Tet-responsive promoter system, a hybrid transactivator specifically stimulates the transcription of promoters containing Tet operator (tetO) sequences in a Tet-dependent way. Two versions exist, Teton and Tet-off, distinguished by whether Tet evokes a conformational change in the transactivator that turns the promoter on or off respectively. Previous studies have described the use of the Tet-off strategy to accurately determine mRNA half-lives by blocking constitutive transcription of the Tet-regulated promoter with Tet and monitoring mRNA decay thereafter [25–27].

We have developed a new variant of the method, based on the Tet-on system, that exploits the availability of the Flp-In T-REx 293 cell line. This cell line combines the features of the Tet-inducible promoters (constitutive expression of the Tet transactivator) with a DNA recombination system based on the site-specific recombinase Flp to facilitate integration of the gene of interest into a specific site in the genome. By means of this methodology, we have generated stable isogenic transfectants of HEK 293 cells that exhibit Tet-inducible expression of luciferase reporters fused to ET-1 3′-UTR fragments. Addition of the Tet analogue Dox to the cultures induces luciferase expression, whereas Dox removal results in gene silencing. Preliminary experiments with these cell lines showed that, for half-life parameters to be accurately measured, it was better to analyse induction of the gene after Dox addition, rather than repression after its removal, probably due to uneven reagent elimination from cells. The time course for the concentration of a gene product to attain a new steady state is related to the decay constant or half-life, irrespective of whether the concentration is increasing or decreasing. Therefore decay constants can be estimated from Dox induction kinetics as shown in Figure 3(A). These values correspond to luciferase protein induction (apparent protein half-life), from which an estimation of mRNA half-life can be calculated by considering that relative delays between induction of mRNAs and the corresponding proteins are related to the half-life of the protein (approx. 3 h for all of these cell lines; results not shown) (Figure 3B). RNase protection determinations of luciferase mRNA confirmed these observations (results not shown), but calculations of the parameters were performed using the luciferase activity data, as with these we were able to process more time points per cell clone. Careful analysis of these parameters showed clearly that mRNA half-lives for the ΔUTR and 272–831 constructs were double those of the full-length, 808–1127 and 609–1127 constructs. Stable transfectants of 293 cells exhibiting Tet-inducible expression of ET-1 mRNAs with or without the 3′-UTR fragment were also generated using the same procedure, and essentially similar results were found compared with the luciferase cell lines, therefore validating the reporter system (results not shown).

Figure 3. Estimation of luciferase–3′-UTR mRNA half-lives in Tet-inducible HEK 293 cells.

Stable isogenic transfectants of HEK 293 cells exhibiting Tet-inducible expression of luciferase reporters fused to ET-1 3′-UTR fragments were generated as described in the Experimental section. (A) Time course of luciferase (Luc) induction after addition of Dox to the cultures. Activity is expressed as normalized units (n=4; normalization with total protein determination). Induction kinetics were fitted to an exponential function to estimate decay constants as shown in (B). These decay constants allow calculation of the corresponding apparent half-lives of luciferase protein and, after subtracting the 3 h value from these values, of the mRNA half-lives (see Experimental and Results sections for details).

The difference in half-life values is due to an intrinsic instability of mRNAs containing the 3′-UTR, as observed in degradation experiments (Figure 4). In this assay, cytosolic extracts from BAEC were incubated with total RNA isolated from 293 cells exhibiting expression of ET-1 mRNAs with or without the 3′-UTR fragment after induction with Dox. These cell clones constituted a source of mature, fully processed ET-1 mRNAs that differed only in the presence of the 3′-UTR fragment. RNase protection experiments performed after a second RNA isolation step revealed that ET-1 mRNA including the 3′-UTR displayed accelerated degradation compared with the same transcript without it. Therefore the presence of the 3′-UTR results in a more unstable mRNA species, capable of strongly modulating the kinetics of gene induction upon transcriptional activation by TGF-β in BAEC or Dox in stable 293 cells. Elements involved in this effect are contained in the 808–1127 region.

Figure 4. Time course for the degradation of full-length and ΔUTR ET-1 mRNAs by cytoplasmic fractions of BAEC.

In vitro degradation assays were performed with the incubation of 20 μg of protein from cytoplasmic extracts from BAEC with 20 μg of total RNA from stable transfectants of 293 cells expressing ET-1 mRNAs with or without the 3′-UTR. At the indicated time points, reactions were stopped and remaining (undegraded) ET-1 mRNA was quantified by RNase protection experiments as described in the Experimental section. (A) Representative gel showing ET-1 and 28 S protected bands. (B) Densitometric analysis after normalization with 28 S rRNA signal. Similar results were obtained in two independent experiments. The human ET-1 riboprobe used in these experiments recognizes a 266 nt fragment within the coding sequence; therefore it detects both full-length and ΔUTR ET-1.

Binding of specific proteins to AUUUA repeats within positions 954–986 mediates the destabilizing effect of the ET-1 mRNA 3′-UTR

The distal portion of the 3′-UTR contains most of its AREs in a region extending for approx. 100 nt. We investigated whether these repeats are involved in the destabilization of mRNA. For this purpose, we dissected the 808–1127 fragment into three smaller subfragments (808–938, 924–1008 and 1009–1127), as displayed in Figure 5(A). Transient transfection of BAEC with luciferase reporters fused to these subfragments revealed that none of them was able to promote significant destabilization, although a clear tendency to show reduced reporter expression was observed with constructs 924–1008 and 1009–1127; however, this was not significant. As AREs are contained in both of these subfragments (Figure 5B), these results may indicate that both fragments are necessary for the mechanism to operate. Therefore we generated an additional luciferase reporter, 924–1127, that included both. Figure 5(A) shows that this subfragment was indeed able to reproduce the decrease in luciferase expression observed with the 808–1127 and full-length constructs.

We then investigated whether specific cellular proteins interact with the 3′-UTR of ET-1 mRNA. For this we performed RNA-EMSAs with radiolabelled probes corresponding to different fragments of the 3′-UTR of the gene. To visualize RNA–protein complexes, RNA sequences not directly involved in the binding were digested by incubation with RNase T1. An initial broad screening with probes 272–831 and 808–1127 showed that prominent RNA–protein bands were obtained with the latter probe (Figure 6A). Analysis of this 808–1127 probe revealed that most of the protein binding occurred within the 924–1008 fragment, or with probes containing this sequence, such as the minimal construct active in functional experiments, 924–1127. Confirming the specificity of these interactions, binding of cellular proteins to the radiolabelled 924–1127 probe was competed with excess unlabelled probe, but not with unlabelled 272–831 sequence used as a control (Figure 6B). To further confirm the involvement of the AREs in the binding, the ability of synthetic poly(U) (polyuridylic acid) to compete the formation of RNA–protein complexes was analysed by RNA-EMSA. As shown in Figure 6(C), small quantities of poly(U) (2 or 20 ng) were able to displace the binding, whereas equal or higher amounts of an unrelated RNA molecule (tRNA from baker's yeast; 20–2000 ng) had no effect. Taken together, these results indicate that cellular proteins interact with the 3′-UTR mRNA through A-rich elements (most probably AREs) contained in the 924–1127 region.

Figure 6. Interaction of RNA probes from the 3′-UTR of human ET-1 mRNA with cytoplasmic proteins extracted from BAEC.

(A) Radiolabelled, in vitro-transcribed RNA fragments corresponding to the indicated subfragments employed in the functional experiments shown in Figures 2 and 5 were incubated in the absence (−) or presence (+) of cytoplasmic extracts from BAEC. After binding, RNA–protein complexes were treated with RNase T1 and resolved by non–denaturing gel electrophoresis as described in the Experimental section. (B) Specific binding of cytosolic proteins to probe 924–1127 was specifically displaced by a molar excess (1, 10 and 100 fold) of unlabelled fragment 924–1127, but not by 272–831. (C) Effects of increasing amounts of unlabelled poly(U) (0.2, 2 and 20 ng) or unlabelled tRNA (20, 200 and 2000 ng) on specific binding to probe 924–1127. The positions of the detected protein complexes (black arrowhead) and of the free probes (white arrowhead) are indicated. Autoradiograms shown correspond to representative RNA-EMSAs performed at least twice with two independent protein preparations.

To locate more precisely the minimum binding site required for interaction with cellular proteins, we performed RNase T1 mapping experiments. Figure 7(A) shows that proteins from cytosolic extracts protected a fragment of approx. 40 nt from ribonuclease attack (lane 2). No fragments were observed in the absence of added protein (lane 1). Subsequent RNase T1 digestion of the isolated 40 nt fragment did not yield any shorter product (Figure 7B, lane 5). Theoretical end-point RNase T1 digestion of the 924–1127 probe gave a product of 40 nt as the longest fragment (corresponding to RNase T1 cutting after G-951 and G-991). The RNase T1-protected fragment can therefore be assigned to sequence 952–991, which includes ARE₃, ARE₄ and ARE₅. To confirm the participation of these elements, specific mutations (AUUUA to GGGCC) were introduced and protein binding was analysed by EMSA. As shown in Figure 8(A), modification of any of the AREs present in the 952–991 fragment resulted in altered binding capacity. Compared with the RNA–protein complex observed with the wild type, the single mutant radiolabelled probes gave diffuse bands that were significantly less intense, an indication of a drastic lost of affinity for the interaction. Figure 8(A) also shows that the simultaneous introduction of these mutations in the triple mutant radiolabelled probe completely blocked the binding of cytosolic proteins. We also investigated whether a correspondence exists between functional experiments and RNA–protein binding assays. For this, we analysed the capacity to promote mRNA destabilization of the 924–1127 constructs mutated in ARE₃, ARE₄ or ARE₅, and also the triple mutant. Figure 8(B) shows that none of the mutant constructs was capable of conferring the destabilizing phenotype, in clear contrast with the wild type. Therefore these results indicate that the destabilizing activity lies in the AREs contained in fragment 924–1127, and that the presence of intact ARE₃, ARE₄ and ARE₅ is indispensable for both the binding of cytosolic proteins and also the mRNA destabilization.

Figure 7. RNase T1 mapping of the RNA–protein complex.

(A) The radiolabelled probe 924–1127 was incubated in the absence (−) or presence (+) of cytosolic extracts. Then, complexes were digested with RNase T1, and mixtures from lanes 1–2 were filtered through nitrocellulose; material from lane 3 was processed without filtering. Bound RNA fragments were extracted from the filter and analysed by urea-denaturing gel electrophoresis. (B) Aliquots from samples run in lanes 1–3 were subjected to an additional step of digestion with RNase T1, then RNA fragments were purified and subsequently analysed by gel electrophoresis to yield the bands shown in lanes 4–6 respectively. Final digestion products are indicated by arrows, together with their sizes. The sequence of the protected 952–991 fragment is also shown. Experiments were performed twice with two independent protein preparations and gave essentially the same results.

Figure 8. Contribution of AREs within the 952–991 fragment to mRNA destabilization.

(A) Binding of cytosolic proteins to radiolabelled wild-type 924–1127 probe and probes containing mutant AREs. The autoradiogram shown corresponds to a representative RNA-EMSA performed twice with two independent protein preparations. (B) The sequences of wild-type and mutant (mARE₃, mARE₄, mARE₅ and triple mutant) probes are shown. ET-1 promoter-driven luciferase (Luc) reporters fused to wild-type 924–1127 and the probe containing mutant AREs were transfected into BAEC, and activity was measured by luminometry. Results for ΔUTR and construct 808–1127 are shown for comparison. Values are expressed as the percentage of luciferase activity for ΔUTR after normalization with Renilla control plasmid (n=3; *P<0.01 compared with ΔUTR value).

TGF-β is an essential modulator of the endothelin system. We have shown a fundamental role of this cytokine in transcriptional activation of the human ET-1 promoter [10]. Nevertheless, TGF-β has also been described to modulate ET-1 levels by increasing the expression of ECE-1 (endothelin-converting enzyme-1), the enzyme responsible for the cleavage of big ET-1 to give mature 21-amino-acid ET-1. This effect seems to involve the specific stabilization of ECE-1 mRNA through increased expression of ECE-1 mRNA 3′-UTR-binding proteins [28]. We therefore assessed the effect of TGF-β treatment of BAEC on the contribution of AREs to ET-1 regulation. Figure 9(A) shows that binding of cytosolic proteins to fragment 924–1127 was not modified by incubation with TGF-β. Additionally, we analysed its effect on ET-1-promoter-driven luciferase reporters without the 3′-UTR (ΔUTR) or with fragment 924–1127. As shown in Figure 9(B), TGF-β strongly induced luciferase expression from both constructs. However, although mRNA destabilization of a reporter-containing fragment 924–1127 resulted in reduced basal and induced luciferase activities compared with ΔUTR, the level of induction was approximately the same (10.78- and 12.00-fold for ΔUTR and 924–1127 respectively). This suggests that TGF-β is probably acting mainly through transcriptionally mediated ET-1 promoter activation, without the participation of elements contained in the 3′-UTR of the gene, although these elements may contribute to regulation of the magnitude of the response.

Figure 9. Analysis of the effect of TGF-β on the contribution of AREs to regulation of ET-1.

(A) Radiolabelled 924–1127 fragment was used as a probe for the binding of cytosolic proteins isolated from BAEC either left under basal conditions or treated with 5 ng/ml TGF-β for 4 or 24 h. Tha label ‘probe’ means that no protein was added. The autoradiogram shown corresponds to a representative RNA-EMSA performed twice with two independent protein preparations. (B) BAEC transfected with ΔUTR or with fragment 924–1127 were challenged with TGF-β for 16 h, and luciferase (Luc) activity was measured by luminometry. Values are expressed as a percentage of luciferase activity for ΔUTR in the absence of TGF-β after normalization with Renilla control plasmid (n=4). Numbers over the bars indicate the level of induction (fold).

DISCUSSION

The data presented here demonstrate that ET-1 expression is subject to strong repressional control that takes place at the level of mRNA stability and is mediated by the binding of specific cytosolic proteins to AREs present in the 3′-UTR of ET-1 mRNA. Loss of this control mechanism (by deletion or mutation of these RNA elements) disrupts the regulation of ET-1 gene expression, resulting in higher basal levels and delayed kinetics of equilibration after induction, as shown here for the effect of TGF-β, a transcriptional activator of expression of the gene [10]. This intrinsic mRNA instability confers on the ET-1 gene the capacity to respond rapidly to changes in transcriptional activity. This is important in the context of vascular pathophysiology, as the endothelium functions as a sensor of physical and chemical stimuli, and needs to be able to activate specific responses, such as fine regulation of ET-1 biosynthesis.

ET-1 expression is restricted to a number of cells or tissues, i.e. smooth muscle cells, cardiomyocytes, glomerular mesangium and, particularly, endothelial cells, the most relevant loci for the peptide's biosynthesis [5]. It is thus likely that specific mechanisms will exist that drive ET-1 expression exclusively in these cell types. However, ET-1 mRNA instability was observed in endothelial and non-endothelial cells. We speculate that transcriptional mechanisms account for cell-type-restricted ET-1 expression, with mRNA stability being a more general type of regulation that is probably utilized by different cell types to confer high mRNA turnover on a number of genes.

Decay signals are often located in the 3′-UTRs of mRNAs [14], and this is also the case for the ET-1 gene. Within the 1.1 kb of the 3′-UTR we have identified a short sequence comprising positions 952–991 that is crucial for the destabilizing effect. AREs are found in the 3′-UTRs of many mRNAs encoding protooncogenes, nuclear transcription factors and cytokines, and are involved in the regulation of gene expression through stability/instability mechanisms [15,16]. The 3′-UTR of ET-1 mRNA contains six AREs, three of which are localized within the 952–991 fragment. Site-directed mutagenesis of these AREs (ARE₃, ARE₄ and/or ARE₅) impaired the mRNA destabilization effect. We also show here that cytosolic proteins from endothelial cells interact specifically with probes of the 3′-UTR, and that this binding activity targets the AREs within RNA segment 952–991. Interestingly, this RNA fragment, the minimal binding site identified by RNA binding assays or RNase T1 mapping, is shorter than the minimal fragment required for the destabilizing effect (924–1127). The most plausible explanation for this apparent contradiction is that protection from digestion by RNase T1 (used in both assays) might occur only where the protein–RNA interaction is tight. Other parts of the RNA molecule involved in binding may therefore be accessible to the action of the enzyme. The fact that specific mutations altering the AREs within the 952–991 region abolished both binding and the functional effect indicates that the two events are in fact specifically related to these elements.

During the preparation of the present paper, a very interesting study by Mawji et al. appeared [19]. These authors examined the fundamental role of the 3′-UTR of the ET-1 gene in mediating control of expression by mRNA stability mechanisms [19]. Although our work agrees with that study completely with regard to the importance of this post-transcriptional mechanism, diferent sequences are identified as being involved in the effect. Mawji et al. [19] showed that AREs within segment 924–1127 are essential, but found that, for these elements to be functional in destabilizing mRNA, they have to work in conjunction with the proximal 312–570 fragment. This is in clear contrast with our finding that the distal 924–1127 fragment was sufficient to promote RNA destabilization. At present it is difficult to reconcile these differences. One possible explanation may lie in the fact that the two studies analysed 3′-UTR segments comprising slightly different fragments. At the 5′-end, Mawji et al. [19] included in their study the upstream sequence 1–272, not analysed in the present report. Nevertheless, it is unlikely that this would have made a difference, as these authors showed that this fragment was not involved in the mechanism. On the other hand, at the distal end of the 3′-UTR, Mawji et al. employed a fragment including sequence up to position 1103, whereas our construct comprised the end of the mRNA (up to position 1127), and was extended to include a short poly(A) tract of 22 nucleotides. Although the contribution of this additional fragment is not obvious, it has been described that some AU-rich-binding proteins also interact with the poly(A) tail and/or with poly(A)-binding proteins [29,30]. The presence of the distal portion plus the poly(A)₂₂ tract may promote additional binding/functional properties. Additionally, we completed our functional studies with a comprehensive RNA binding analysis, using the cytosolic fraction of endothelial cells. The two sets of results coincide in demonstrating the fundamental role of the AREs (ARE₃, ARE₄ and ARE₅) within the distal probe, without the involvement of any other region of the RNA molecule.

An essential regulator of the endothelin system is the cytokine TGF-β. It has been shown that modulation by TGF-β takes place at different levels. We have described the molecular mechanism by which TGF-β increases the transcription of the ET-1 gene by activation of the Smad signalling pathway and participation of specific TGF-β-regulated sites within the promoter [10]. TGF-β is also able to up-regulate ET-1 peptide levels by increased expression of ECE-1 in a model of hepatic wound healing, as described by Shao et al. [28]. Interestingly, this effect occurs at the level of mRNA stabilization through increased expression of certain, as yet unidentified, proteins capable of binding to a C-rich region of the form ACAC within the proximal end of the 3′-UTR of ECE-1 mRNA. This 3′-UTR also contains some AREs that are not involved in this TGF-β-mediated regulatory mechanism. Therefore it seems that TGF-β does not contribute to the regulation of AREs. Our results also agree with that concept, as they show that the treatment of endothelial cells with TGF-β altered neither the binding of proteins to the 3′-UTR of ET-1 mRNA nor the level of induction of ET-1 promoter-driven luciferase constructs ΔUTR or 924–1127, the minimal competent to mediate destabilization. However, the 3′-UTR contributes to modulate the maximal response to TGF-β in absolute terms.

Several RNA-binding proteins that interact specifically with AREs have been characterized. These include AUF1 [ARE/poly-(U)-binding/degradation factor; heterogeneous nuclear ribonucleoprotein D], the ELAV family (HuR, HuB, HuC, HuD), tristetraprolin, TIA/TIAR, Hsp70 and many others. However, current understanding of how these proteins influence mRNA stability is far from clear [16]. We are currently establishing protocols to identify the endothelial cytosolic proteins that interact specifically with the 3′-UTR of ET-1 mRNA. Precise identification of these proteins will allow us to gain insight into this mechanism of control of ET-1 levels.

Although much attention has been directed towards the role of transcription in the activation or repression of genes important in vascular pathophysiology, post-transcriptional events, particularly the stability of specific mRNAs, are increasingly recognized as important determinants of gene expression. In vitro examples for this are, among others, the β-adrenergic receptors and the endothelial and inducible nitric oxide synthases [14,17]. Experiments conducted in animal models have also shown that genetic defects in the regulatory elements of the 3′-UTR of a number of genes are responsible for abnormal levels of expression that lead to pathogenic phenotypes. For example, elimination of the AREs contained in the 3′-UTR of TNF-α (tumour necrosis factor-α) mRNA results in deregulation of TNF-α production, which in turn causes the development of chronic inflammatory diseases in transgenic mice [31,32]. It is tempting to speculate about the existence of genetic polymorphisms in the sequences of the AREs involved in the regulation of ET-1 gene that may result eventually in increased ET-1 levels associated with disease entities such as hypertension or fibrosis. Further investigation will be required to confirm this hypothesis. In conclusion, we have shown that the ET-1 gene provides an elegant example of how transcriptional and post-transcriptional mechanisms work together to dynamically regulate the expression of this vasoactive peptide, and hence of vascular tone.