Length Increase of the Human α-Globin 3′-Untranslated Region Disrupts Stability of the Pre-mRNA but Not That of the Mature mRNA*

Polyadenylation increases the stability of mRNA molecules. By studying the effect of the length of 3′-untranslated region (UTR) on mRNA levels, we have found that α-globin pre-mRNA is stabilized by a mechanism that does not modulate the half-life of mature mRNA. The insertion of DNA fragments of various unrelated sequences into the 3′-UTR of the human α-globin gene strongly reduces mRNA abundance upon transfection into choriocarcinoma JEG-3 cells. We found an inverse relationship between mRNA levels and the length of the introduced fragments. In fact, mRNA levels as low as 1% were observed after inserting a 477-nucleotide (nt) fragment, whereas inserting a fragment of 86 nt at the same position had no effect on mRNA accumulation. DNA insertion induced no change in transcription rate or in half-life of mature mRNA. Semi-quantitative reverse transcription-polymerase chain reaction revealed that inserting a 477-nt fragment in the 3′-UTR resulted in decreased levels of nuclear pre-mRNA in proportion to that observed for mature mRNA. In contrast, the insertion of the 477-nt exogenous DNA in the last intron had no effect on mRNA levels despite the presence of intronic sequences in the pre-mRNA. This shows that the reduction of pre-mRNA level was not due to the insertion of putative ribonuclease cleavage sites or the insertion of a segment DNA that reduces the elongation efficiency. Taken together, our results strongly support the existence of a pre-mRNA stabilizing mechanism that can be disrupted by increasing the length of the 3′-UTR. The fact that the half-life of mature mRNA is not affected by DNA insertion is compatible with a pre-mRNA-specific stabilizing mechanism that acts specifically before polyadenylation.

Polyadenylation increases the stability of mRNA molecules. By studying the effect of the length of 3-untranslated region (UTR) on mRNA levels, we have found that ␣-globin pre-mRNA is stabilized by a mechanism that does not modulate the half-life of mature mRNA. The insertion of DNA fragments of various unrelated sequences into the 3-UTR of the human ␣-globin gene strongly reduces mRNA abundance upon transfection into choriocarcinoma JEG-3 cells. We found an inverse relationship between mRNA levels and the length of the introduced fragments. In fact, mRNA levels as low as 1% were observed after inserting a 477-nucleotide (nt) fragment, whereas inserting a fragment of 86 nt at the same position had no effect on mRNA accumulation. DNA insertion induced no change in transcription rate or in half-life of mature mRNA. Semi-quantitative reverse transcription-polymerase chain reaction revealed that inserting a 477-nt fragment in the 3-UTR resulted in decreased levels of nuclear pre-mRNA in proportion to that observed for mature mRNA. In contrast, the insertion of the 477-nt exogenous DNA in the last intron had no effect on mRNA levels despite the presence of intronic sequences in the pre-mRNA. This shows that the reduction of pre-mRNA level was not due to the insertion of putative ribonuclease cleavage sites or the insertion of a segment DNA that reduces the elongation efficiency. Taken together, our results strongly support the existence of a pre-mRNA stabilizing mechanism that can be disrupted by increasing the length of the 3-UTR. The fact that the half-life of mature mRNA is not affected by DNA insertion is compatible with a pre-mRNA-specific stabilizing mechanism that acts specifically before polyadenylation.
Before reaching cytoplasm, most vertebrate primary transcripts are processed at their 5Ј end by capping (1) and at their 3Ј end by polyadenylation (2), and they are spliced to remove introns (3,4). Pre-mRNA 3Ј-end processing is a two-step process that involves endonucleolytic cleavage and addition of 50 -250 adenosine residues. 3Ј-End formation requires the pres-ence of a polyadenylation signal (A(A/U)UAAA) (5) and a G/Urich sequence (6) localized upstream and downstream of the cleavage site on pre-mRNAs, respectively.
RNA 3Ј-end formation and splicing can occur independently in vitro (7)(8)(9). However, when introns were removed from genes, very low levels of mRNA accumulated upon transfection of culture cells (10 -15) and in transgenic mice (16,17). In fact, the presence of introns in pre-mRNAs increases levels of mature mRNA up to 50-fold by enhancing 3Ј-end formation (15, 18 -22). It is now generally accepted that this effect depends on the splice acceptor site of the last intron (15,18,19,(23)(24)(25)(26)(27)(28). Little or no effect on 3Ј-end formation was observed by deleting the donor site of the last intron indicating that the intron removal per se is not essential in 3Ј-end formation. This conclusion was reinforced by the observation that pre-mRNAs transcribed from two human ␤-globin genes defective in splicing were cleaved and polyadenylated as the normal ␤-globin pre-mRNA (29). A favored model to explain the role of the acceptor site is that spliceosome assembly at the 3Ј splice site of the last intron may facilitate cleavage and polyadenylation processes (18,19,25,26). We reasoned that if the enhancing effect of the intron on 3Ј-end formation involves an interaction between elements present in the two regions, then the length between the last intron and the polyadenylation site should influence the efficiency of 3Ј-end formation and, hence, the steady state levels of mRNA.
To study the interaction between a cis-acting element of the polyadenylation region and other element(s) located further upstream, we have increased the length of the terminal exon of the ␣-globin gene, more precisely, the length of its 3Ј-untranslated region (UTR), 1 and analyzed mRNA steady state level. This was performed by inserting DNA fragments of various length and sequences. Co-transfection of the resulting clones with the parental plasmid into a choriocarcinoma cell line, JEG-3, that does not express the endogenous globin genes revealed that the length of the 3Ј-UTR strongly influences the level of ␣-globin mRNA. We succeeded in linking this effect to the disruption of pre-mRNA stability with no change in mature mRNA decay. fragment of the plasmid pUC␣ (30) at the EcoRI site of pSV-SPORT-1-⌬, downstream of the SV40 early promoter that encompasses the 72-bp direct repeats and the 21-bp GC-rich sequences. The pUC␣ plasmid contains the 1.5-kilobase pair PstI-PstI fragment of the ␣ 1 -globin gene. The HinfI-HinfI fragment contained the ␣ 1 -globin gene sequence from the 18th base upstream of the initiation codon to the 101st base downstream of the poly(A) signal and 209 bases of pUC19. All of the clones presented were confirmed by DNA sequencing.
Substitution of the 3Ј-UTR of the ␣-globin gene by that of SV40 present in vector sequences was performed by deletion. SV-␣GL was digested by BstXI and NotI to remove the 3Ј-UTR, a part of exon 2, and the entire intron 2 and exon 3 of ␣-globin. The resulting fragment was ligated to a BstXI-and NotI-digested PCR product containing the ␣-globin sequence from BstXI to the stop codon inclusively, followed by a NotI site. The 5Ј primer sequence containing the BstXI site of ␣-globin was 5Ј-CTGACCAACGCCGTGGCG-3Ј. The 3Ј primer sequence containing the NotI site joined to ␣-globin sequences that begin at the stop codon and extend into the open reading frame (ORF) was 5Ј-GGGGCG-GCCGCTTAACGGTATTTGGAG-3Ј. The resulting DNA was named SV-␣GL-⌬3ЈUTR-A (Fig. 4A). The 3Ј-UTR of this clone contains no ␣-globin sequences and starts at the first nucleotide of the NotI sequences and extends over the poly(A) signal present in the vector sequences. SV␣GL was also digested by BalI and NotI, blunted, and ligated in the presence (SV-␣GL-⌬3ЈUTR-B/3␤-477-S) or in the absence (SV-␣GL-⌬3ЈUTR-B) of the 3␤-HSD 477-bp fragment (Fig. 4A). The 3Ј-UTR of these two clones contains the first 16 bases of the ␣-globin 3Ј-UTR. PCRs were performed with the Pwo enzyme (Roche Molecular Biochemicals), which has proofreading activity.
Other DNA clones were prepared as follows. The first step in the construction of clone SV-␣GL ⌬in1 was performed by digestion of SV-␣GL by ClaI and BstXI to remove all ␣-globin sequences upstream of the BstXI site and a part of the vector sequences. This fragment was ligated to a ClaI-BstXI-digested PCR fragment produced from SV-␣GL. The 5Ј primer contained vector sequences encompassing the ClaI site (5Ј-ACAGCATCGATGCTGTGG-3Ј), whereas the 3Ј primer was composed of a BstXI site, followed by ␣-globin sequences encompassing the first codon (5Ј-CGCCCACGGCGTTGGTCAGCACCATGGTGGGTTC-3Ј). In the resulting clone, the third codon was joined to the 68th codon, and the sequence remained in frame (Fig. 9A). The clone SV-␣GL ⌬in2 was prepared by digestion of SV-␣GL by BstXI and NotI to remove the ␣-globin sequence downstream of the BstXI site and part of the downstream vector sequences. The resulting fragment was ligated to a PCR product digested by BstXI and NotI. The 5Ј primer was composed of a BstXI site, followed by the ␣-globin sequence starting two nucleotides before the stop codon and extending into the 3Ј-UTR (5Ј-GGGC-CAACGCCGTGGGTTAAGCTGGAGCCTCGG-3Ј). The 3Ј primer encompassed the NotI site of the vector (5Ј-GGGAGCGGCCGCCGAC-TAGTG-3Ј). In the resulting clone, the 71st codon was followed by a new glycine residue and the stop codon (Fig. 9A). The clone SV-␣GL ⌬SmaI-NarI was produced by circularization of a DNA fragment prepared by a total NarI digestion, followed by a partial SmaI digestion of SV-␣GL to remove the 340-bp fragment containing exon 2 and part of introns 1 and 2. The 3␤-HSD-1-PvuII-EcoRI fragment of 477 bp was inserted into the BalI site of these clones to produce clones SV-␣GL ⌬in1/477, SV-␣GL ⌬in2/477, and SV-␣GL ⌬SmaI-NarI/477. SV-␣GL-cDNA was constructed as follows. First, human blood was collected and centrifuged to remove serum and the white interface. The erythrocyte-enriched pellet was subjected to RNA extraction as described below, and first strand cDNA synthesis was performed using a Ready-To-Go T-Primed First-Strand Kit (Amersham Pharmacia Biotech). Then, an aliquot of this reaction was subjected to PCR using a 5Ј primer composed of a PstI restriction site, followed by sequences present in SV-␣GL, starting immediately downstream the PstI site of the Multiple Cloning Site (MCS) and ending at the 18th base of the ␣-globin gene (5Ј-GGGCTGCAGGTACCGGTCCGGAATTACTCAGAGAGAAC-CCACC-3Ј) and a 3Ј primer composed of ␣-globin 3Ј-UTR sequences localized downstream the BalI site (5Ј-TTCAAAGACCACGGGGGTAC-3Ј). The resulting PCR fragment was then digested with PstI and BalI. Second, SV-␣GL was digested by BalI and then partially digested by PstI. The resulting molecules were ligated to the PCR product. Molecular screening permitted isolation of the SV-␣GL-cDNA clone that contains exactly the same sequences as SV-␣GL but without introns. The 3␤-HSD-1 fragment of 477 nt was then inserted into the BalI restriction site to obtain SV-␣GL-cDNA/477.
DNA extractions were performed as described by Sambrook et al. (34) with two rounds of purification on cesium chloride (CsCl) gradients. Three different DNA preparations of each of the SV-␣GL, SV-␣GL/3␤/ 370-S, and SV-␣GL/3␤/477-S plasmids were used for transfections into JEG-3 cells without variation in their relative levels of expression.
RNA Extraction and Hybridization-Cellular RNA was extracted with Tri Reagent (39, 40) from pools of four dishes. RNAs were purified on CsCl gradients (41). The RNA samples were glyoxalized, electrophoresed, and gels transferred to Nytran Plus membranes (Schleicher & Schuell). Filters were hybridized at 42°C and washed under high stringency conditions (33). Autoradiographs were quantified by densitometric scanning (Amersham Pharmacia Biotech RAS image analyzer system). The ␣-globin DNA probe was prepared using the 1119-bp HinfI-HinfI fragment of the ␣-globin gene. Clones with deletions in the ORF and those with a DNA insertion into the BalI site were also hybridized with a BalI-PstI 183-bp probe encompassing the 3Ј-UTR but lacking the open reading frame (ORF). Their levels of expression relative to SV-␣GL were the same as for the 1119-bp probe (data not shown). The ␥-actin cDNA probe was composed of a 2-kilobase pair cDNA fragment (42). All of the probes were labeled by random priming (43).
Pre-mRNA Analysis by Reverse Transcription-PCR (RT-PCR)-After transfection of JEG-3 cells with plasmids SV-␣GL-⌬3ЈUTR-B and/or SV-␣GL-⌬3ЈUTR-B/3␤-477-S, nuclear RNA was extracted with Tri Reagent from isolated nuclei prepared as described above, and purified on CsCl gradients (41). After DNase I treatment and purification, each 5-g RNA sample was subjected to RT using the First-Strand cDNA Synthesis Kit (Amersham Pharmacia Biotech) and a primer starting at the 174th nt and ending at the 157th nt downstream from the SV40 polyadenylation signal used by these recombinant ␣-globin genes (5Ј-CAGTGAGCGAGGAAGCGG-3Ј). Only 2% of the RT reactions were used in PCR reactions in the presence of the same 3Ј primer and the following 5Ј primer corresponding to sequences within the intron 2 of the ␣-globin gene, 5Ј-GGAGCGATCTGGGTCGAG-3Ј. PCRs were performed with the Expand High Fidelity PCR System (Roche Molecular Biochemicals) as follow: 5 min at 94°C; 10 min at 72°C; 28 cycles (1 min at 94°C; 1 min 15 s at 55°C; 1 min 45 s at 72°C); 10 min at 72°C.

RESULTS
An ␣-globin DNA, beginning in the 5Ј-UTR sequences and extending over the polyadenylation site, was placed under the control of the SV40 early promoter (Fig. 1A). Except when indicated, all plasmid constructs were derived from this parental clone, named SV-␣GL.
To study specifically the effect of the length of the 3Ј-UTR on ␣-globin mRNA levels, transfections were performed in nonerythroid JEG-3 cells to eliminate possible effects of putative tissue-specific elements. The endogenous ␣-globin gene contains a CpG island extending both upstream and downstream from the transcription start site (44). As it has been suggested that CpG island DNA may increase transcription specifically upon integration in the genome, we performed transient transfections.
Two overlapping fragments originating from the type 1 3␤-HSD cDNA clone were inserted in the unique BalI restriction site of the ␣-globin gene localized 14 bases downstream of the stop codon (Fig. 1A). The resulting clones, containing an insertion of DNA of 477 bp (SV-␣GL/3␤-477-S) or of 370 bp (SV-␣GL/ 3␤-370-S) in sense orientation, were each co-transfected with SV-␣GL into JEG-3 cells. Their relative transient expression levels were analyzed by Northern blot. Although the insertion of the 370-bp 3␤-HSD-1 fragment reduced the level of mRNA to 15% that of parental ␣-globin mRNA, the effect of the 477-bp 3␤-HSD-1 fragment was more pronounced, leading to an mRNA level 1% that of ␣-globin mRNA (Fig. 1B). Similar results were observed in COS-7 cells (data not shown) indicating that the effect of DNA insertion is not specific to choriocarcinoma cells.
Effect of the Sequence and the Length of Exogenous DNA Inserts on ␣-Globin mRNA Levels-To define the properties of putative inhibitory inserts, we first studied the same DNA fragments inserted in both orientations. Both orientations of the 477-and the 370-bp 3␤-HSD-1 segments decreased mRNA levels to the same extent (Fig. 1C), showing that the effect of DNA insertion is orientation-independent. To determine whether the sequence of the DNA insert is important, a 295-bp cDNA segment originating from the ORF of the 17␤-HSD-1 cDNA was cloned into the same BalI restriction site of SV-␣GL. The sequence identity between the 3␤-HSD-1 3Ј-UTR and 17␤-HSD-1 cDNA fragments was less than 5%. Co-transfections with SV-␣GL revealed that the 17␤-HSD-1 fragment reduced mRNA levels to approximately that of the 370-bp 3␤-HSD-1 insert (Fig. 1D). Thus, the inhibitory effects of DNA insertion are independent of the sequence of the DNA insert.
To determine whether the length of the DNA insert correlates with mRNA levels, we introduced two DNA fragments of 286 and 86 bp in both orientations into the BalI site of the ␣-globin gene. The four clones were each co-transfected with SV-␣GL in JEG-3 cells. Clones containing the 477-or the 370-bp 3␤-HSD-1 fragment in both orientations were included as controls. An example of a Northern blot with clones containing an insertion in the sense orientation is shown in Fig. 2A. The 86-bp 3␤-HSD-1 fragment had no effect on mRNA levels in both sense and antisense orientations (Fig. 2B). This observation demonstrates that DNA insertion does not destroy any putative ␣-globin regulatory element encompassing the BalI site. Thus, the insertion of DNA per se does not explain the above observations. Moreover, a negative correlation was found between the relative expression of plasmids and the length of the DNA inserts (Fig. 2C). Therefore, the effects of DNA inserts on mRNA levels depend of the length of these inserted fragments.
The Effect of Exogenous DNA Insert Is Function of Its Localization in the ␣-Globin Sequences-We also examined whether exogenous DNA alters mRNA levels when introduced into the last intron. The SV-␣GL/3␤/INTRON clone was constructed in which the 477-bp 3␤-HSD-1 fragment was inserted into the first half of the second intron of the ␣-globin gene (Fig. 3A). Messenger RNA from SV-␣GL/3␤/INTRON and SV-␣GL had similar lengths, indicating that RNA splicing occurred correctly, albeit the presence of an insert into intron 2 (Fig. 3B). No difference in mRNA levels was observed between the clones after readjustment with ␥-actin mRNA for the amount of RNA loaded per lane. These results demonstrate that the strong effect observed on ␣-globin mRNA levels by insertion of DNA fragments depends on its localization in the ␣-globin sequence. Moreover, because the presence of the 477-nt insert in the transfected plasmid does not necessarily correlate with low steady state mRNA levels, the decrease observed with SV-␣GL/ 3␤-477-S and -A compared with the parental plasmid is not an artifact caused by unequal transport of individual plasmid during transfection.
␣-Globin-specific Regulatory Element Downstream from the Site of DNA Insertion-The fact that the length of DNA inserts determine the extent of the effect suggests that an interaction between two hypothetical cis-acting elements localized on each side of the BalI site is disrupted by DNA insertion. The next experiment was designed to determine whether an ␣-globinspecific regulatory element may be found downstream from the site of insertion. Sequences downstream from the stop codon of SV-␣GL (Fig. 4A, SV-␣GL-⌬3ЈUTR-A) were deleted by joining the stop codon directly to the vector sequence, which contains its own polyadenylation signal originating from SV40. The replacement of the ␣-globin 3Ј-UTR by the 3Ј-UTR of the vector in both JEG-3 and COS-7 cells had no or only a weak effect on mRNA accumulation (Fig. 4B). Thus, in JEG-3 and COS-7 cells, no ␣-globin-specific regulatory element can be mapped within the 3Ј-UTR by substitution.
Does insertion of DNA also decrease the abundance of mRNA when all sequences positioned downstream from the site of insertion in the 3Ј-UTR are replaced by heterologous 3Ј-UTR sequences? To answer this question, plasmids SV-␣GL-⌬3ЈUTR-B and SV-␣GL-⌬3ЈUTR-B/3␤-477-S were constructed (Fig. 4A). Results indicate that DNA insertion into the BalI site had similar effects on mRNA levels in both the ␣-globin and the vector 3Ј-UTR contexts (Fig. 4C). The only common features downstream from the BalI site between the two 3Ј-UTR might be the cis-acting elements involved in 3Ј-end formation. For this reason, if any cis-acting regulatory element positioned downstream from the site of insertion is disrupted by DNA insertion, this element might be included within sequences involved in 3Ј-end formation.
DNA Insertion and Promoter Activity-We verified whether insertion of DNA in the 3Ј-UTR also reduces expression of the ␣-globin gene under the control of its promoter. The 3␤-HSD-1 477-bp fragment was inserted in both orientations at the BalI site of the plasmid pUC␣, which contains the ␣-globin promoter and gene (30). Co-transfections in JEG-3 cells revealed that insertion of DNA in both orientations strongly reduced the expression of the ␣-globin gene under the control of its promoter (Fig. 5). We conclude that insertion of DNA had similar effects both in homologous and heterologous promoter contexts.
Nuclear runoff experiments were performed using clones SV-␣GL-⌬3ЈUTR-B and SV-␣GL-⌬3ЈUTR-B/3␤-477-S. Although these two clones showed strong differences in mRNA accumulation (Fig. 4C), they produced signals of similar intensity in runoff assay (Fig. 6). This result indicates that insertion of DNA does not alter the rate of transcription.
DNA Inserts Do Not Affect ␣-Globin Mature mRNA Halflife-Is mature mRNA decay affected by insertion of DNA into the BalI site? We used two different approaches to examine this possibility. The first is based on the fact that two mRNAs with different half-lives, but transcribed from the same promoter, give a ratio close to 1.0 soon after transfection, but this ratio changes over time until the mRNAs eventually reach steady state levels. Each of the two clones, SV-␣GL/3␤-477-S and -370-S, were co-transfected with SV-␣GL, and their mRNA levels were analyzed at different times. We observed no variation in SV-␣GL/3␤-370-S/SV-␣GL or in SV-␣GL/3␤-477-S/SV-␣GL mRNA ratios over time (Fig. 7A). In fact, the ratios were independent of the establishment of mRNA steady state levels. Second, we did a time course experiment in the presence of actinomycin D (ActD) after co-transfection of SV-␣GL/3␤-370-S and SV-␣GL. SV-␣GL/3␤-370 and SV-␣GL mRNAs had very close t1 ⁄2 values, as evidenced by the stable mRNA ratio over time (Fig. 7B). These two observations indicate that the addition of an insert DNA into the BalI site of the ␣-globin gene did not alter mRNA decay.
Levels of Nuclear Pre-mRNA-We next planned an experiment to show whether the effect of DNA insertion occurs before RNA processing. RT-PCR was performed on nuclear RNA isolated from JEG-3 cells transfected with SV-␣GL-⌬3ЈUTR-B and/or SV-␣GL-⌬3ЈUTR-B/3␤-477-S. To amplify specifically pre-mRNA, the 5Ј primer was located in the last intron, and the 3Ј primer was positioned 157 bases downstream from the polyadenylation signal (Fig. 8A). The presence of the 477-nt insert into the BalI site of the 3Ј-UTR strongly reduced levels of nuclear pre-mRNA (Fig. 8B). It should be noted that when plasmid DNA was added in a negative RNA sample before the DNase I reaction that preceded the RT-PCR, no amplification product was observed (data not shown). Therefore, the addition of an insert DNA into the BalI site resulted in a proportional decrease of pre-mRNA and mRNA levels, indicating that the effect of DNA insertion occurs before RNA processing.
␣-Globin Regulatory Elements in the ORF and the Introns-We next examined the possibility that the effect of DNA insertion may be related to an ␣-globin-specific regulatory element localized in the ORF or in an intron. Messenger RNA levels were only slightly lowered by deletion of DNA regions encompassing the intron 1 (SV-␣GL ⌬in1) or intron 2 (SV-␣GL ⌬in2) of the ␣-globin gene (Fig. 9, A and B). When the 477-nt fragment was inserted in the BalI site of these two clones, a strong decrease in mRNA accumulation was observed (Fig. 9C). The segment SmaI-NarI was then removed from SV-␣GL. This deletion overlaps the deletion of clones SV-␣GL ⌬in1 and SV-␣GL ⌬in2. Again, the addition of the 477-nt insert into the BalI site strongly reduced levels of mRNA accumulation (Fig. 9C). Therefore, we failed to associate the effect of DNA insertion to the presence of any single ␣-globin specific sequence between the 10th nt of the ORF to the stop codon.
Because common features exist between introns as donor and acceptor sites, it is possible that the effect of the DNA insertion into the BalI site persists as long as one intron is present in the gene. To study this possibility, both introns were removed (clone SV-␣GL-cDNA, Fig. 9A). The resulting clone had mRNA levels similar to that observed after the insertion of the 477-nt fragment into the BalI restriction site of SV-␣GL (Fig. 9D). However, the addition of the 477-nt insert into the BalI site of SV-␣GL-cDNA also strongly reduced levels of mRNA accumulation (Fig. 9D). Therefore, the effect of DNA insertion is independent of the presence of intron sequences in the gene. DISCUSSION We have clearly demonstrated that increasing the length of the 3Ј-UTR by insertion of DNA segments had a strong negative effect on ␣-globin mRNA level. In our model, neither the nucleotide sequence of the insert nor that of the 3Ј-UTR itself, excluding features common to all 3Ј-UTR, contribute to the observed effect. The only characteristic of the inserts that was related to the level of mRNA was their length.
We first postulated that increasing the length of the ␣-globin 3Ј-UTR might disrupt an interaction between the acceptor site of the last intron and the polyadenylation site, thereby reduce polyadenylation efficiency and levels of mRNA. It is now clear that the strong decrease in mRNA levels observed by increasing the length of the 3Ј-UTR was not due to the disruption of such a mechanism because the effect of the length of the 3Ј-UTR was observed in the absence of intronic sequences as well (clone SV-␣GL-cDNA). In fact, insertion of DNA into the BalI site did not change the rate of transcription but decreased levels of pre-mRNA and mature mRNA proportionally. Therefore, our results strongly suggest that the ␣-globin pre-mRNA was actively protected from degradation in the nucleus and  were co-transfected with SV-␣GL in JEG-3 cells. t 0 was fixed at 20 h after the addition of the calcium-phosphate precipitates. Northern blots were probed with ␣-globin DNA. B, plasmids SV-␣GL/3␤-370-S and SV-␣GL were co-transfected in JEG-3 cells. Cells were trypsinized, pooled, and re-plated after transfection to eliminate inter-variations. The next day, ActD (5 g/ml) was added for 45 min, and RNA was extracted at different times. The t 0 corresponds to the end of the ActD treatment. Samples were processed as described in A. open box corresponds to 3Ј-UTR, thin line to intron 2, and wavy line corresponds to RNA sequences present in pre-mRNA but removed by endonucleolytic cleavage during the polyadenylation process. Arrows show position of the oligonucleotides used in RT-PCR. RT reaction was performed using 5 g of nuclear RNA and the specific 3Ј primer, and semi-quantitative PCR conditions were obtained using only 2% of RT reactions. B, analysis of RT-PCR end products by Southern blot using the 1119-bp HinfI-HinfI ␣-globin probe. Because sequences localized downstream from the BalI site of these clones are not ␣-globin sequences, the ␣-globin probe hybridized specifically to a segment of the same length for the two RT-PCR products, which extend from the 5Ј primer to the BalI site. Fragments amplified from plasmids SV-␣GL-⌬3ЈUTR-B and SV-␣GL-⌬3ЈUTR-B/3␤-477-S were of 574 and 1051 bp, respectively. JEG-3 cells were transfected with SV-␣GL-⌬3ЈUTR-B (lanes 1-3) and/or SV-␣GL-⌬3ЈUTR-B/3␤-477-S (lanes 1, 4 and 5). The relative intensity of specific signals are the same than those observed with ethidium bromide staining. Relative levels of SV-␣GL-⌬ 3ЈUTR-B and SV-␣GL-⌬3ЈUTR-B/3␤-477-S mRNAs were similar to these observed in Fig. 4 as determined by Northern blot analysis of the corresponding cytoplasmic RNA (data not shown).
that the insertion of DNA into the BalI site, but not within intron 2, had disrupted this mechanism of stabilization. The decrease in mRNA amount was not due to insertion of putative ribonuclease sites present on the exogenous DNA insert, as demonstrated by the insertion of the same DNA segment within intron 2 with no effect on ␣-globin mRNA level. Interestingly, stabilization of ␣-globin pre-mRNA is achieved through a mechanism that seems to be inactive in mature polyadenylated mRNA, as witnessed by the absence of effect of DNA insertion on mature mRNA half-lives.
The negative correlation between RNA abundance and the length of DNA inserts strongly suggest the disruption of an interaction between two elements positioned on each side of the BalI restriction site used for DNA insertion. Substitution of 3Ј-UTR sequences downstream from the site of insertion did not change the effect of DNA insertion. Therefore, if a downstream cis-acting element is actually related to the effect of DNA insertion, it might be positioned within sequences related to polyadenylation. The upstream cis-acting element(s) that is aimed at interacting with this 3Ј element is unknown, but our results allow the conclusion that it is not localized within introns and that at least one copy of it might be localized between the beginning of exon 3 and the BalI site. This arises from the observation that no effect on mRNA abundance was observed when the 477-nt segment was introduced within the intron 2. However, the effect of DNA insertion in the BalI site was conserved even though all sequences localized between the fourth codon and the stop codon were sequentially deleted. Therefore, our results appear to be in agreement with reiterated 5Ј-cis-acting elements.
Increasing the length of the ␣-globin 3Ј-UTR resulted in destabilization of ␣-globin pre-mRNA but not mature mRNA, and this was true in both homologous and heterologous 3Ј-UTR contexts. Considering also the fact that ␣-globin is not normally expressed in JEG-3 and COS-7 cells, it is tempting to suggest that pre-mRNA stabilization could be a general mechanism of eucaryotic pre-mRNAs. Further investigation will help to elucidate this point, but some cases of modulation of gene expression, probably through the control of pre-mRNA decay, have yet been reported. For example, GM-CSF pre-mRNA accumulation was observed following treatment of murine B-cells with interleukin 1␣, whereas decreased GM-CSF expression observed after treatment with interleukin 4 should be due to intranuclear destabilization of GM-CSF pre-mRNA (45). In another case, it was reported that decreased peptidylglycine ␣-amidating monooxygenase expression after 17␤-estradiol treatment might be largely due to intranuclear destabilization of the primary transcript (46). Therefore, pre-mRNA decay seems to be controlled at least for these genes.
We have demonstrated that ␣-globin pre-mRNA is stabilized and that this stabilization is responsible for an increase in ␣-globin mRNA of 100-fold or more. In addition, our results strongly suggest that this phenomenon is regulated by a mechanism that necessitates cooperation between two cis-acting elements localized each side of the BalI site. It is of considerable interest that this mechanism of RNA stabilization is specific to pre-mRNA, which would appear to represent a putative level of gene regulation.