3′-UTR sequences are required for inhibition of TNF-α mRNA splicing by2-AP

To examine whether TNF-α gene promoter or 5′-UTR sequences respond to 2-AP, cells were transfected with p5′CAT, in which 821 bp from the TNF-α gene, comprising 652 bp preceding the transcription start site, the 5′ UTR and the first six codons of the ORF are abutted to the CAT gene (p5′CAT, Fig. ​Fig.1A).1A). Expression of this hybrid mRNA proceeded unabated in the presence of 10 mm 2-AP in the cell culture (Fig. ​(Fig.2E),2E), a concentration that inhibited the expression of TNF-α mRNA from pTNF-α (Fig. ​(Fig.22A,D).

Splicing of TNF-α precursor transcripts bearing a truncated 3′ UTR, however, is no longer sensitive to 2-AP. In Figure ​Figure2F,2F, BHK-21 cells were transfected with pTNF-α(Δ3′UTR) which lacks the 573 terminal base pairs of the 792-bp TNF-α 3′ UTR and ends with a polyadenylation signal of the TNF-β gene (Fig. ​(Fig.1D).1D). Expression of precursor transcripts and mRNA transcribed from this truncated gene construct was abundant and as transient as for the complete gene (Fig. ​(Fig.2A),2A), yet expression of mRNA was no longer inhibited by 2-AP, nor was there any enhanced accumulation of precursor transcripts. Removal of certain TNF-α 3′-UTR sequences thus resulted in loss of splicing inhibition by 2-AP.

On the other hand, in cells transfected with p5′CAT(3′UTR-α) (Fig. ​(Fig.1B),1B), which carries the 3′-UTR sequences that were deleted in pTNF-α(Δ3′UTR), expression of the resulting hybrid CAT mRNA showed no sensitivity to 2-AP (data not shown). This result suggests that when fused to an intronless gene, such as the bacterial CAT gene, the TNF-α 3′ UTR does not mediate an inhibition of mRNA expression.

2-AP does not inhibit TNF-β gene expression atsplicing

The above results prompted us to examine a gene that does not respond to 2-AP by inhibition of splicing. We showed previously that in lymphoid cells, increasing doses of 2-AP led to a coordinate decline in TNF-β precursor transcripts and spliced mRNA, supporting an inhibition at transcription rather than at splicing (Jarrous et al. 1996). Accordingly, we examined the response of the TNF-β gene to 2-AP in transfected cells. The entire human TNF-β gene (pTNF-β, Fig. ​Fig.1E)1E) was transfected into BHK-21 cells and expression of TNF-β RNA was monitored with an antisense RNA probe covering adjacent segments of intron 3 and exon 4 (Fig. ​(Fig.3A).3A). Unspliced precursor transcripts (637-nucleotide protected fragment) were expressed transiently in large amounts, whereas TNF-β mRNA (571 nucleotides) was expressed at lower levels. This ratio attests to slow excision of intron 3 from human TNF-β precursor transcripts in BHK-21 cells. Inefficient excision of intron 3 was observed for the murine TNF-β gene in the lymphocytic cell line CTLL-2 (Weil et al. 1990). Addition of 2-AP did not elicit any increase in TNF-β precursor transcripts; instead, expression of both precursors and mRNA declined in this cell line.

Open in a separate window

Figure 3

TNF-α 3′-UTR sequences suffice to confer splicing control by 2-AP. BHK-21 cells were transfected with pTNF-β (A,B), pTNF-β(3′UTR-α) (C,D) or pTNF-α(3′UTR-β) DNA (E), and cotransfected with pSV₂CAT DNA. 2-AP was present from 12 (A,B) or 16 hr after transfection (C–E), and 20 μg/ml CHX from 16 hr (C,D). Cell viability remained constant. Total RNA was isolated at times shown and subjected to RNase protection analysis with ³²P-labeled antisense TNF-β intron 3/exon 4 RNA probe P to quantitate precursor transcripts (637-nucleotide band) and spliced RNA (571-nucleotide band) (A,C) or TNF-β exon 1/exon 2/exon 3/intron 3 probe P to quantitate precursor transcripts (324-, 275-, and 169-nucleotide bands) and spliced RNA (214-nucleotide band) (B,D). (M) Size marker. In E, TNF-α precursor transcripts (700-nucleotide band) and spliced RNA (341-nucleotide band) were analyzed as for Fig. ​Fig.2A;2A; the top autoradiograph shows a higher exposure of the 700-nucleotide band. CAT mRNA protects 250 nucleotides of probe.

Splicing of remaining exons 1–3 of the transfected TNF-β gene was followed next by an antisense RNA probe that overlaps TNF-β exons 1–3 and part of intron 3 (Fig. ​(Fig.3B).3B). Primary transcripts protect 169 nucleotides whereas partially spliced ones protect fragments of 169, 324, and 275 nucleotides. A 214-nucleotide fragment became prominent by 24 hr. This fragment is protected by spliced RNA molecules in which exons 2, 3, and 4 are joined. These RNA molecules still retain intron 1 but have an intact ORF because translation starts within exon 2 (Nedwin et al. 1985). These RNA species were also detected in lymphoid cells (Jarrous et al. 1996). Fully spliced mRNA (which protects a 263-nucleotide fragment), although seen in lymphoid cells (Jarrous et al. 1996), was not detected in BHK-21 cells (Fig. ​(Fig.3B).3B). Transfection of pTNF-β yielded transient expression of precursor transcripts whose decline was accompanied by the appearance of spliced RNA (214-nucleotide band). Again, formation of spliced TNF-β RNA was relatively slow and the yield was low. Addition of 2-AP led to a decrease in spliced RNA (lane 3 vs. 6) as well as in precursor transcripts (lane 2 vs. 5). This response, seen in Figure ​Figure3,3, A and B, stands in marked contrast to that of the TNF-α gene (Fig. ​(Fig.2),2), an indication that 2-AP failed to inhibit expression of a transfected TNF-β gene at mRNA splicing.

2-AP inhibits splicing of TNF-β precursor transcripts carrying TNF-α 3′ UTRsequences

In Figure ​Figure3,3, C and D, transfection was with pTNF-β(3′UTR-α) (Fig. ​(Fig.1F),1F), in which the TNF-β 3′ UTR is truncated 160 bp downstream from the stop codon and joined to a 823-bp TNF-α gene fragment comprising 573 terminal base pairs of the 3′ UTR, the polyadenylation site, and downstream sequences. Addition of 2-AP led to a shift from spliced RNA (214-nucleotide band) to precursor transcripts, which increased strongly (Fig. ​(Fig.3D,3D, lanes 1–4 vs. 5–7). 2-AP failed to block transcription of the chimeric TNF-β(3′UTR-α) gene, inhibiting instead the excision of TNF-β introns 2 and 3 required for expression of spliced RNA. These findings are corroborated by use of the complementary TNF-β intron 3/exon 4 probe (Fig. ​(Fig.3C).3C). Again, addition of 2-AP led to a strong increase in precursor transcripts and a concomitant decline in spliced RNA, resulting in a pronounced rise in their ratio (lanes 1–4 vs. 5–7). In contrast, in cells transfected with pTNF-β(Δ3′UTR) (Fig. ​(Fig.1G),1G), lacking the TNF-α insert of pTNF-β(3′UTR-α), expression of spliced TNF-α RNA was insensitive to 2-AP (F. Osman and R. Kaempfer, unpubl.).

Addition of CHX led, in the pTNF-β(3′UTR-α)-transfected cells, to enhanced expression of both precursor transcripts and spliced RNA (Fig. ​(Fig.3,3, C and D, lane 2 vs. 8). Expression of spliced RNA was shifted to earlier times, followed by a decline in precursor transcripts and, more slowly, in spliced RNA (lanes 8–9). Here, too, generation of spliced RNA was inhibited by 2-AP (lane 8 vs. 10), whereas precursor transcripts increased (lane 9 vs. 11), the ratio of precursor transcripts to spliced mRNA increasing strongly. We conclude that 2-AP inhibits splicing of TNF-β precursor transcripts carrying a TNF-α 3′ UTR, and that this inhibitory effect does not require protein synthesis.

Although expression of TNF-β mRNA from both pTNF-β and pTNF-β(3′UTR-α) is sensitive to 2-AP, the response to 2-AP is shifted from an apparent inhibition of transcription for pTNF-β to inhibition of splicing for pTNF-β(3′UTR-α). Replacement of part of the TNF-β 3′ UTR with TNF-α 3′-UTR sequences renders splicing sensitive to 2-AP at multiple sites, shown in Figure ​Figure3,3, C and D, for joining of TNF-β exons 2, 3, and 4. The observation that 2-AP fails to inhibit expression of mRNA from p5′CAT(3′UTR-α), yet inhibits mRNA expression from pTNF-β(3′UTR-α) at splicing, emphasizes a requirement for exon/intron junctions in addition to TNF-α 3′-UTR sequences. However, there is no requirement for TNF-α exons or introns per se, because pTNF-β(3′UTR-α) contains only TNF-β exon/intron junctions. A cis-acting element that renders splicing sensitive to inhibition by 2-AP thus is located within the TNF-α 3′ UTR.

Splicing of TNF-α precursor transcripts carrying TNF-β 3′-UTR sequences is insensitive to2-AP

The results obtained above led us to predict that splicing of precursor transcripts encoded by the reciprocal chimeric gene, pTNF-α(3′UTR-β) (Fig. ​(Fig.1H),1H), should proceed unabated in the presence of 2-AP. In pTNF-α(3′UTR-β), the TNF-α gene was truncated 219 bp into the 3′ UTR and joined to a 572-bp TNF-β gene fragment comprising the 3′-terminal 389 bp of the 628-bp 3′ UTR, the polyadenylation site and downstream sequences. Expression of spliced mRNA (341-nucleotide band) from pTNF-α(3′UTR-β) failed to decline when 2-AP was present, nor did precursor transcripts (700 nucleotides) increase (Fig. ​(Fig.3E).3E). Splicing of TNF-α mRNA carrying TNF-β 3′-UTR sequences, thus, is insensitive to inhibition by 2-AP, again showing that a 2-APRE resides specifically in the TNF-α 3′ UTR.

Activation of PKR by RNA deriving from the TNF-α 3′UTR

Because 2-AP is a known inhibitor of PKR, we asked whether the TNF-α 3′ UTR harbors an RNA element able to activate this kinase. T7 RNA transcripts were generated from various fragments of TNF-α 3′-UTR DNA (Fig. ​(Fig.4A).4A). To eliminate any contaminating dsRNA, these transcripts were purified by gel electrophoresis followed by CF-11 cellulose chromatography, a treatment shown previously to eliminate dsRNA contaminations (Circle et al. 1997). At low RNA concentration (0.05 ng/μl), RNA derived from the 104-bp EcoRI–PstI fragment (3′UTR–αEP) activated PKR even more strongly than dsRNA, as judged by phosphorylation of the PKR (68 kD) and eIF2α (38 kD) bands in the ribosome fraction of rabbit reticulocyte lysate (Fig. ​(Fig.4B,D).4B,D). Assignment of these bands was verified by Western blot analysis with anti-PKR and anti-eIF2 antibodies (data not shown). Like dsRNA, 3′UTR–αEP RNA induced phosphorylation of PKR and eIF2α at low, but not high, RNA concentration (Fig. ​(Fig.4D).4D). The NcoI (EN) transcript required a higher RNA concentration than 3′UTR–αEP for phosphorylation of the 68-kD band, whereas RNAs derived from still longer fragments or other regions of the TNF-α 3′ UTR did not induce extensive PKR phosphorylation (Fig. ​(Fig.4B).4B). Apparently, the longer 3′-UTR fragments are less efficient than 3′UTR–αEP RNA in supporting PKR phosphorylation in vitro as a result of hindrance of the activating RNA domain. As an inducer of autophosphorylation of PKR, which is a primary indicator of the activation of this kinase, 3′UTR–αEP RNA stood out with the highest molar specific activity (Fig. ​(Fig.4C).4C). Thus, the EP domain in the TNF-α 3′ UTR, located well upstream of the AU-rich element (Fig. ​(Fig.4A),4A), encodes an RNA that, at low concentrations, strongly activates PKR in vitro.

Open in a separate window

Figure 4

The TNF-α 3′ UTR harbors an RNA element that activates PKR. TNF-α 3′ UTR DNA constructs (A) were used to generate T7 RNA transcripts that at the concentrations shown were incubated with rabbit reticulocyte ribosomal fraction and [γ-³²P]ATP; dsRNA [poly(I–C)] served as control (B). An autoradiograph of the protein gel (B) was quantitated for each TNF RNA at 0.05 ng/μl, to yield molar-specific phosphorylation activity for PKR (C). In D, dsRNA or purified 3′UTR–αEP T7 transcript was incubated and analyzed as in B, with 2-AP present at the indicated concentrations (in mm).

Phosphorylation of PKR and eIF2α elicited by 3′UTR–αEP RNA transcripts in vitro was inhibited progressively by 1 and 5 mm of 2-AP (Fig. ​(Fig.4D).4D). Activation of PKR by 3′UTR–αEP RNA was less sensitive to 1 mm 2-AP when compared with the activation by dsRNA, another indication that the former RNA is also a potent activator of PKR. These results show that 3′UTR–αEP RNA activates PKR in a 2-AP-sensitive manner.

Structure of 3′UTR–αEPRNA

The structure of 3′UTR–αEP RNA transcript was analyzed by T1, U2, and V1 RNase sensitivity mapping (Fig. ​(Fig.5).5). 3′UTR–αEP RNA forms a stable, 5′-proximal 48-nucleotide stem–loop containing 17 bp (ΔG=−59 kJ at 30°C). As calculated by RNADraw (http://rnadraw.base8.se/) and mfold algorithms (Zuker 1989), this stem–loop structure persists in the longer EP-containing RNA fragments shown in Figure ​Figure4A.4A.

Open in a separate window

Figure 5

Nuclease sensitivity mapping of TNF-α 3′UTR–αEP RNA. 5′-End-labeled 3′UTR–αEP RNA was digested with T1, U2, or V1 nuclease directly to assay structure (str) [(c) without nuclease] and, for T1 and U2, also after denaturation at 50°C in 7 m urea (seq). The nucleotide ladder was generated by alkaline hydrolysis (OH). An autoradiograph of the sequencing gel is shown. Stem and loop regions relate to secondary structure at right, showing sites of nuclease attack, on the basis of multiple analyses. GGGC is from plasmid. Phylogenetic conservation of sequences is shown for Homo sapiens (human), Sus scrofa (wild boar), Oryctolagus cuniculus (rabbit), Bos taurus (bull), and Capra hircus (goat).

The 3′UTR-αEP domain is the 2-APRE of the TNF-αgene

To test the concept that the EP domain in the TNF-α 3′ UTR, able to activate PKR in vitro (Fig. ​(Fig.4),4), mediates the sensitivity of TNF-α mRNA splicing to 2-AP (Fig. ​(Fig.2A,D),2A,D), we abutted it to the TNF-α(Δ3′UTR) and TNF-β genes, neither of which show sensitivity to 2-AP at mRNA splicing (Figs. ​(Figs.22 and ​and3).3). In the constructs thus generated, pTNF-α(3′UTR–αEP) (Fig. ​(Fig.1I)1I) and pTNF-β(3′UTR–αEP) (Fig. ​(Fig.1J),1J), the EP domain is followed by polyadenylation signal sequences of the TNF-β gene.

BHK-21 cells transfected with pTNF-α(3′UTR–αEP) showed transient expression of precursor transcripts and spliced mRNA (Fig. ​(Fig.6A,6A, lanes 1–4). Addition of 2-AP at 20 hr elicited a progressive rise in precursor transcripts up to 36-fold at 53 hr, coupled with a decline in mRNA (lanes 5–8). Likewise, in cells transfected with pTNF-β(3′UTR–αEP), addition of 2-AP led to a pronounced shift from spliced mRNA to precursor transcripts, which increased seven-fold (Fig. ​(Fig.7A).7A). In pTNF-α(3′UTR–αEP)-transfected cells, unspliced precursor transcripts accumulating in the presence of 2-AP could be chased into mRNA by washing out 2-AP in the presence of actinomycin D (Fig. ​(Fig.6C).6C). Thus, introduction of the 3′UTR–αEP fragment suffices for a gain of function, rendering splicing of precursor transcripts sensitive to inhibition by 2-AP, whether these arise from TNF-α or TNF–β gene sequences. We conclude that the 3′UTR–αEP domain is a functional 2-APRE.

Open in a separate window

Figure 6

The 2-APRE renders splicing of TNF-α mRNA dependent on activation of PKR. BHK-21 cells were transfected with pTNF-α(3′UTR–αEP) or pTNF-α(Δ3′UTR) and where indicated, cotransfected with pPKRΔ6 or pPKR. pSV₂CAT DNA was cotransfected in each case. 2-AP was added at 18 (C) or 20 hr (A,B) after transfection, where shown. In C, 2-AP was removed (w) and actinomycin D added as for Fig. ​Fig.2D.2D. Total RNA was analyzed as for Fig. ​Fig.2A2A to quantitate TNF-α precursor transcripts (700 nucleotides) and spliced RNA (341 nucleotides); in C and D, the top autoradiograph shows a higher exposure of the 700-nucleotide band. CAT mRNA protects 250 nucleotides of probe.

Open in a separate window

Figure 7

The TNF-α 2-APRE renders splicing of TNF-β mRNA dependent on activation of PKR and does not affect translation efficiency of TNF-α mRNA. BHK-21 cells were transfected with pTNF-β(3′UTR–αEP) (A,B) or pTNF-β (C) and, where indicated, cotransfected with pPKRΔ6. pSV₂CAT DNA was cotransfected in each case. In A, 2-AP was added at 20 hr after transfection. Total RNA was analyzed as for Fig. ​Fig.3A3A to quantitate TNF-β precursor transcripts (637 nucleotides) and spliced RNA (571 nucleotides). (M) Size marker. CAT mRNA protects 250 nucleotides of probe. (D) BHK-21 cells were transfected with pTNF-α, pTNF-α(Δ3′UTR), or pTNF-α(3′UTR–αEP). pSV₂CAT DNA was cotransfected in each case. The ratios of TNF-α protein secreted into the medium and spliced TNF-α RNA normalized to CAT mRNA were determined. The bar graph shows mean and s.e. for four independent experiments.

The function of the 2-APRE in mRNA splicing involves activation ofPKR

On one hand, RNA transcribed from the 3′UTR–αEP domain is able to activate PKR in vitro; on the other hand, this domain renders mRNA splicing sensitive to inhibition by the PKR inhibitor, 2-AP. These results raised the possibility that PKR functions as a trans-acting factor that regulates mRNA splicing in precursor transcripts that carry the 2-APRE. To obtain direct evidence for this concept, we cotransfected cells with TNF gene constructs and a vector expressing either wild-type PKR or PKRΔ6, a transdominant-negative mutant that prevents the activation of wild-type PKR (Koromilas et al. 1992). If PKRΔ6 were to block the activation of endogenous PKR, it should cause inhibition of mRNA splicing in a manner resembling 2-AP. As seen in Figure ​Figure6A6A for pTNF-α(3′UTR–αEP) and in Figure ​Figure7B7B for pTNF-β(3′UTR–αEP), the effect of PKRΔ6 cotransfection was strikingly similar to that of addition of 2-AP, shifting the pattern of RNA molecules from spliced to unspliced forms, with precursor transcripts increasing strongly. In Figure ​Figure6A,6A, the effect was particularly pronounced at later times when PKRΔ6 expression should become more prominent (lanes 1–4 vs. 9–12), but even at the early times a reduction in spliced RNA was evident. This shift to unspliced RNA forms was not observed for the TNF-α(Δ3′UTR) gene (Fig. ​(Fig.6B)6B) nor for the TNF-β gene (Fig. ​(Fig.7C),7C), and thus requires the 2-APRE. For the TNF-β gene, total expression of RNA was reduced by PKRΔ6, reflecting the results with 2-AP (Fig. ​(Fig.33A,B).

Conversely, overexpression of wild-type PKR enhanced mRNA splicing in a 2-APRE-dependent manner (Fig. ​(Fig.6D).6D). In this experiment, precursor transcripts were relatively more abundant than in Figure ​Figure6A.6A. Cotransfection with PKR led to a shift from unspliced RNA into mRNA encoded by pTNF-α(3′UTR–αEP), the ratio of mRNA over precursor transcripts increasing up to 20-fold by 45 hr. Such enhancement was absent when pTNF-α(Δ3′UTR) was used instead.

Moreover, when PKR was cotransfected, the amount of mRNA expressed by pTNF-α(3′UTR–αEP) was significantly greater than that expressed by pTNF-α(Δ3′UTR), despite higher levels of unspliced precursor transcripts in cells transfected by the latter construct (Fig. ​(Fig.6D).6D). This result shows that the ability to respond to PKR at the splicing step is mediated by the 2-APRE.

The 2-APRE renders mRNA splicing dependent on activation of PKR, which can be abrogated by 2-AP or expression of PKRΔ6 and enhanced by expression of wild-type PKR. Our results reveal that activation of PKR is required for splicing of mRNA when precursor transcripts contain the 2-APRE, and that an increase in the level of PKR enhances their splicing efficiency.

Physiological role of PKR insplicing

Our results indicate that mRNA splicing follows a novel path for genes containing a 2-APRE, which are thereby placed under the control of PKR, exemplified here by the human TNF-α gene, as well as the TNF genes in pTNF-α(3′UTR–αEP) and pTNF-β(3′UTR–αEP), as distinct from the canonical path used by genes that lack a 2-APRE, including the human TNF–β gene and the mutant TNF-α gene in pTNF-α(Δ3′UTR) and pTNF-α(3′UTR-β).

The presence of the 2-APRE not only renders splicing of precursor transcripts dependent on PKR but also allows it to proceed with higher efficiency in response to increased expression of PKR. Thus, even though transfected BHK-21 cells express constitutive levels of PKR that suffice for splicing of 2-APRE-containing mRNA, such levels limit the extent of splicing. For a gene containing a 2-APRE, modulation of splicing efficiency by PKR may thus be expected in conditions in which PKR is induced, as during an inflammatory response.

Our findings lead to the prediction that physiological signals that cause activation of PKR should promote splicing of TNF-α precursor transcripts. Conversely, signals causing an inhibition of PKR should act to reduce their splicing. In this context, T-cell activation by anti-CD3 antibodies triggers splicing of pre-existing TNF-α precursor transcripts, yielding an immediate production of TNF-α protein (Yang et al. 1998).

TNF-α exhibits antiviral activity (Wong and Goeddel 1986) and induces PKR gene expression (Yeung et al. 1996). PKR has a direct role in TNF-α-induced apoptosis (Yeung et al. 1996; Der et al. 1997) through phosphorylation of eIF2α and inhibition of translation (Srivastava et al. 1998). Induction of PKR by TNF-α or other signals may thus create a positive feedback loop for TNF-α mRNA expression.

The nature of the splicing controlelement

Analysis of a 104-nucleotide RNA transcribed from the human TNF-α 2-APRE by nuclease sensitivity mapping reveals a stable stem–loop containing 17 bp (including 10 G:C pairs). This structure is conserved in human, porcine, rabbit, bovine, and goat TNF-α genes. Phylogenetic conservation is most pronounced in the upper stem and 6-nucleotide loop, which thus are likely to be involved in PKR-mediated regulation of splicing. This folding is preserved in longer, 2-APRE-containing human TNF-α 3′-UTR RNA fragments. Moreover, the location of the 2-APRE within the human TNF-α 3′ UTR, halfway between the stop codon and the AU-rich motif and ~200 nucleotides upstream of the latter, is conserved in the porcine, rabbit, and bovine genes.

We did not find significant sequence homology of the 2-APRE with adenovirus VA RNA, which binds to and inhibits PKR (Clarke and Mathews 1995), nor with HIV-1 TAR (Maitra et al. 1994) or human α-tropomyosin 3′ UTR (Davis and Watson 1996), which reportedly activate PKR.

Moreover, no similar structure is found in the human TNF-β 3′ UTR. We have shown that splicing of human TNF-β precursor transcripts is inherently sluggish, possibly rendering additional regulation at this step less essential. TNF-α and TNF-β genes differ also in their transcriptional control and mRNA stability (English et al. 1991). Expression of TNF-α mRNA in PBMC occurs earlier and to higher levels but also is more transient than that of TNF-β mRNA. Tighter control of TNF-α gene expression, through a dependence of mRNA splicing on activation of PKR, may also reflect the wider range of biological activities of TNF-α.

Splicing of TNF-β precursor transcripts carrying an inverted TNF-α 3′-UTR sequence is as sensitive to inhibition by 2-AP as splicing of TNF-β(3′UTR-α) transcripts (F. Osman, N. Jarrous, and R. Kaempfer, unpubl.), supporting the concept that structure, rather than sequence, of the 2-APRE is important for its function in splicing control.

Function of the 2-APRE

Although the 2-APRE requires the presence of introns to regulate splicing, there is no requirement for TNF-α introns per se. An inhibition of splicing at multiple sites, mediated by the 2-APRE, is also observed when all splice junctions and introns are of TNF-β origin.

Our finding that splicing of human TNF-α precursor transcripts is regulated by a 3′-UTR element implies that this process takes place only once transcription has progressed well into the last exon, beyond the 2-APRE. Although pre-mRNA splicing is mainly a cotranscriptional event, in which factors supporting transcription, splicing, and 3′-end processing of precursor transcripts interact as functional complexes and colocalize within the nucleus (Neugebauer and Roth 1997), splicing may occur either during transcription or post-transcriptionally, depending on the position of the intron (Bauren and Wieslander 1994; Misteli et al. 1997). The small size of the TNF introns and genes may in this case favor a predominantly post-transcriptional splicing mechanism. In resting lymphoid cells, accumulation of unspliced TNF-α precursor transcripts has been observed, which are processed into mRNA only on stimulation (Yang et al. 1998).

When 2-AP was present, precursor transcripts accumulated at the expense of mRNA. This accumulation was pronounced, reaching >10-fold in Figure ​Figure2D2D and >30-fold in Figure ​Figure6A.6A. A marked increase in precursor transcripts, coupled with a decline in spliced mRNA, was seen also when transdominant-negative mutant PKRΔ6 was expressed (Figs. ​(Figs.6A6A and ​and7B).7B). These reciprocal changes in levels of precursor transcripts and mRNA are indicative of a splicing block and could not be explained by effects on transcription. Precursor transcripts are either processed or degraded and, hence, are unstable intermediates that do not accumulate to the levels of mRNA. The possibility that the 2-APRE would render precursor transcripts more stable in the presence of 2-AP (or of PKRΔ6), yet mRNA less stable, is not supported by our RNA chase experiments. Accumulated precursor transcripts remained unstable in the presence of 2-AP, yet they could be chased into spliced mRNA once 2-AP was removed. In the chase experiments, a pronounced conversion of precursor transcripts into spliced mRNA was observed on the removal of 2-AP (Figs. ​(Figs.2D2D and ​and6C).6C). Moreover, TNF-α precursor transcripts that accumulated in the nucleus in the presence of 2-AP were not transported to the cytoplasm (Jarrous et al. 1996). Located well upstream of the AU-rich motif, the 2-APRE is in a region not known to be involved in destabilization of TNF-α mRNA. Deletion of 3′-UTR sequences, including the 2-APRE as well as the AU-rich motif, did not alter the kinetics of mRNA expression when 2-AP was absent (Fig. ​(Fig.2A,F).2A,F). In lymphoid cells, 2-AP inhibited splicing of the endogenous TNF-α mRNA but did not affect its stability (Jarrous et al. 1996). Thus, the primary action of 2-AP is to block splicing.

A 3′-proximal structure in yeast pre-rRNA initiates processing of upstream rRNA species, with the stem essential for function (Allmang and Tollervey 1998). Removal of transcribed spacers from rRNA and excision of introns from precursor transcripts may be similarly controlled through remote elements. The 2-APRE may represent the first example of a class of ordered cis-acting RNA elements involved in mRNA-splicing control.

Cell transfection

Monolayers of BHK-21 cells were seeded at a density of 1×10⁶ to 2×10⁶ cells/8-cm Petri dish and grown overnight; 4 hr before cotransfection, culture medium was replaced. A mixture of 10 μg of test DNA construct, 4 μg of salmon sperm DNA, and 2 μg pSV₂CAT DNA was permeated into cells by calcium phosphate–DNA coprecipitation (Jarrous et al. 1996). After 12–18 hr, culture medium was replaced with fresh medium prewarmed to 37°C. For Figures ​Figures2E,2E, ​E,6,6, and ​and7,7, LipofectAmine (Life Technologies) was used with 1 μg of each DNA for transfection. Efficiency of transfection was >72%, as judged by fluorescence-activated sorting of cells transfected with pCMV–EGFP expressing green fluorescent protein. 2-AP (Sigma) was added at 10 mm unless otherwise indicated.

Hybridization probes

For RNase protection analysis, DNA was subcloned in pBS (Stratagene) under the T3 or T7 promoter and transcribed with [α-³²P]UTP to generate labeled antisense RNA transcripts. A TNF-α RNA probe (map, Fig. ​Fig.2A)2A) was generated from a 700-bp Sau3AI–Sau3AI fragment consisting of 10 bp of intron 2, exon 3, intron 3, and part of exon 4. A TNF-β probe was generated from a 637-bp Bsu36I–EcoRI DNA fragment, containing part of intron 3 and 571 bp of exon 4 (map, Fig. ​Fig.3A).3A). A second TNF-β RNA probe (map, Fig. ​Fig.3B)3B) was generated from a 324-bp EcoRI–Bsu36I DNA fragment, containing part of exon 1, exon 2, exon 3, and a portion of intron 3, derived from pSV–LT, provided by Dr. Rao Movva (Biogen, Geneva, Switzerland). A probe of 250 nucleotides, generated from a HindIII–EcoRI fragment of the coding region of the CAT gene (Gorman et al. 1982), was used to quantitate CAT mRNA. Intactness of RNA probes was assessed by PAGE (Jarrous et al. 1996).

Ribonuclease protection analysis

Total RNA was isolated with guanidinium isothiocyanate/CsCl. RNase protection analysis with RNases A and T1 was performed with genomic TNF-α and TNF-β riboprobes and CAT or actin probes as described (Jarrous et al. 1996). Protected RNA fragments were separated by electrophoresis in 5% polyacrylamide/8 m urea gels. The size marker in Figure ​Figure33 was generated by MspI digestion of pGEM-3 DNA and filling in with [α-³²P]dCTP (Jarrous and Kaempfer 1994); the size marker in Figure ​Figure77 is a 100-bp DNA ladder (Fermentas) that was dephosphorylated and 5′ end labeled with [γ-³²P]ATP. These markers were also used to determine fragment sizes in Figures ​Figures2,2, ​,3,3, and ​and6.6. Autoradiographs were scanned and band intensity was quantitated using NIH Image 1.61 software.

Activation of PKR by 3′-UTRtranscripts

An 823-bp EcoRI–EcoRI TNF-α gene fragment comprised of 573 3′-terminal base pairs of the 3′-UTR, the polyadenylation site, and downstream sequences from pTNF-α was inserted into pBS under the T7 promoter. To obtain 3′-UTR sense transcripts, the DNA was digested with PstI (EP), NcoI (EN), AflIII (EA), SspI (ES), or XmnI (EX) and transcribed in vitro. A 616-bp NcoI–EcoRI TNF-α gene fragment comprised of 366 3′-terminal base pairs of the 3′ UTR, the polyadenylation site and downstream sequences from pTNF-α was inserted into pBS under the T7 promoter. To obtain 3′-UTR sense transcripts, the DNA was digested with AflIII (NA) or XmnI (NX) and transcribed in vitro. The EP RNA transcript was purified twice by gel electrophoresis, followed by chromatography on CF-11 cellulose, washing with ethanol, and eluting with water as described (Circle et al. 1997). Phosphorylation of PKR and the eIF2α chain was assayed by incubating poly(I–C) or 3′-UTR transcript with the ribosome fraction from rabbit reticulocyte lysate in the presence of [γ-³²P]ATP for 20 min at 30°C as described (Rosen et al. 1981) and subjecting the reaction mixture to electrophoresis in 10% polyacryamide gels containing sodium dodecyl sulfate. Band intensity was quantitated with NIH Image 1.61 software.

Nuclease sensitivity mapping

RNA (60 pmole), dephosphorylated with calf alkaline phosphatase and 5′ end labeled with [γ-³²P]ATP and T4 polynucleotide kinase, was purified on a 6% polyacrylamide gel in 8 m urea before its digestion for 20 min at 30°C with 1 unit of RNase T1 (Worthington) or RNase U2, or 0.15 unit of V1 nuclease (Pharmacia). The mixture was made 9 m in urea, cooled on liquid nitrogen, and separated on an 8% polyacrylamide sequencing gel.

The publication costs of this article were defrayed in part by payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 USC section 1734 solely to indicate this fact.