Polypyrimidine Tract Binding Protein Modulates Efficiency of Polyadenylation

Plasmid constructions.

A 526-nt fragment containing the cytomegalovirus (CMV) promoter (3) was isolated by PCR using the forward (CCACATGTTTACATAACTTACGGTAAATG) and reverse (GAGTACAGGCAAAAAGCAGAG) oligonucleotide primers (CMV fragment). The β-globin gene (described below) was isolated by PCR using the forward (CATGGTGCACCTGACTCCTGAGGAGAAGTC) and reverse (TCCAGATGCTCAAGGCCCTTCATAATATCC) oligonucleotide primers. These two PCR products were then blunt-end ligated, reamplified, and cut with ApaI (CMV-β-globin fragment).

The α-globin poly(A) sequence was isolated by PCR by using the forward (TTCTGCCTAAGGGCCCTCCTCCCCT) and reverse (CGACGTCTCTCCAGGAACCAGACTC) oligonucleotide primers. The product was digested with ApaI, ligated to the CMV-β-globin fragment, reamplified, cut with AflIII and XbaI, and ligated to pUC 18 cut with AflIII/XbaI (α). The C2 complement poly(A) sequence (36) was isolated using forward (TTGGCCTGTCCCCAGATTCCTTCCCT) and reverse (CAAGGCCAGCCCTACCTGGCC) oligonucleotide primers. The product was digested with ApaI, ligated to the CMV-β-globin fragment, reamplified, cut with AflIII and XbaI, and ligated to pUC 18 cut with AflIII/XbaI (C2). The β-globin gene and the β-globin poly(A) sequence have been described previously (14) (β). The CMV fragment together with the β-globin gene's exon 1 was amplified by PCR using the forward (CGTTACATAACTTACGGTAAATG) and the reverse (CTGCCCAGGGCCTCACCACCAACT) oligonucleotide primers (CMV-exon1 fragment). The β-globin gene's exon 2 was amplified by PCR using the forward (GCTGCTGGTGGTCTACCCTTGGACC) and the reverse (GGACTTCAAGAGTCCTAG) oligonucleotide primers (exon2 fragment). These two PCR products were then ligated and reamplified with the 3′ primer (CMV-exon1-exon2 fragment). The β-globin gene's exon 3 and the β-globin pA signal were isolated by PCR using the forward CTC (CTGGGCAACGTGCTGGTCTG) and the reverse (TTCGAACGTACGGACGTCCGT) primers. This PCR product was ligated to the CMV-exon1-exon2 fragment, reamplified, digested with AflIII/HindIII, and ligated to pUC 18 cut with AflIII/HindIII (β cDNA).

PTB plasmids were described previously (61). The Va plasmid, which expresses the adenovirus VA1 gene, was previously described (14).

Cell culture, transfection, and RNA isolation.

Subconfluent HeLa cells were transfected as previously described (18) using 3 μg of either α, β, C2, or cDNA plasmids, 1.5 μg of each PTB plasmid, and 0.5 μg of Va plasmid.

Cytoplasmic RNA was isolated as previously described (15); nuclear and total RNAs were isolated using the hot-phenol method (18). Pellets were resuspended in 40 to 90 μl of R loop buffer (2).

RNA analysis.

S1 nuclease analysis was performed as described previously (2), using the EcoRI-linearized (β) (β cDNA) or C2 construct end-filled with [α-³²P]dATP and the Bsp1201-linearized (α) construct end-filled with [α-³²P]dGTP as probes.

Nuclear run-on (NRO) analysis.

Nuclear isolation and run-on analysis were performed as previously described (19, 43). Probes BE, BE2, and B3 for the NRO analysis were described previously (43).

Control probes for M13 (background), Va, and HIS (adenovirus VA1 gene and mouse histone H4, respectively) have also been described (1).

Western blot analysis.

Cell lysates were prepared using RIPA buffer (12), boiled, and quantified using the Bradford assay, and 5 μg of each lysate was separated by sodium dodecyl sulfate (SDS)-polyacrylamide electrophoresis and transferred to a Hybond pure nitrocellulose membrane (Amersham). Membranes were immunoblotted using anti-PTB (1:2,000) and antiactin (1:2,000) antibodies by standard procedures (30).

In vitro transcription and in vitro cleavage analysis.

The α pA signal was amplified by PCR using the forward (AGATTTAGGTGACACTATACCTGGCCCACAAGTAT; containing an SP6 promoter sequence) and the reverse (CTGCAGAGAGGTCCTTGGTCTGAGACA) oligonucleotide primers. The β pA signal was amplified by PCR using the forward (AGATTTAGGTGACACTATACCTGGCCCACA AGTAT; containing an SP6 promoter sequence) and the reverse (GTTTGAACTAGCTCTTCATTTCTTTATG) oligonucleotide primers. These pA signals were in vitro transcribed, and cleavage reactions were carried out as previously described (36).

Ultraviolet (UV) cross-linking competition assays.

The ³²P-labeled RNA substrate was incubated with 10 ng of a CstF purified protein fraction (or a mixture of CPSF and CstF enriched fractions) and various amounts of rPTB (10, 50, 100, 250, or 500 ng) (36).

RNAi assays.

For RNAi, HeLa cells were plated to a density of 10,000 cells per well in a 24-well plate, and assays were performed as previously described (58) with minor modifications. On day 6, cells were transfected with 1 μg of each reporter and 0.5 μg of Va plasmid.

PTB overexpression reduces mRNA levels.

To investigate the potential effect of PTB on polyadenylation, a transient-transfection assay was employed using different β-globin gene constructs with transcription driven by the strong CMV promoter and one of three alternative pA signals: α-globin, β-globin, or C2 complement (α, β, or C2) (Fig. 1A). A further construct was created that lacked introns (β cDNA). The β and α constructs were cotransfected into HeLa cells with or without a PTB expressing plasmid (at a ratio of 2:1, β/α plasmid to PTB plasmid), and both nuclear and cytoplasmic RNA fractions were purified (Fig. 1B and C). Each transfection included a third plasmid containing the adenovirus Va gene. Since the Va gene is efficiently transcribed by RNA polymerase III (Pol III), its expression provides an internal transfection efficiency and loading control. The RNA samples were subjected to S1 nuclease mapping using either β or α 3′-end-labeled DNA probes, together with a Va-specific probe. Compared to the relatively constant Va signal (75 nt), both the β and α transcript 3′ signals showed a significant reduction (about threefold) in mRNA levels when the transfection included the PTB plasmid. This effect was particularly marked in the nuclear fractions. It should be noted that the α and β signals break up into several closely spaced bands which reflect the ability of S1 nuclease to partially degrade duplex structures near the 3′ terminus, especially when the sequence is AT rich. Overexpression of PTB had a similar effect on both weak pA (α) and strong pA (β) signals, suggesting a very efficient activity for PTB. The different efficiencies of the α and β pA signals have been previously described (17). Note that in Fig. 1, where the Va, β, and α S1 probes showed nearly equal specific activities, it is clear that the construct containing the weaker α pA resulted in significantly less cytoplasmic mRNA than the construct with the stronger β pA (compare Fig. 1B and C, lanes 3).

Control experiments.

Since the PTB-mediated reduction in mRNA levels could result from a range of different mechanisms, a series of controls was devised to rule out effects other than the inhibition of 3′ end processing. As shown in Fig. 2A, Western blotting indicates that PTB transfection significantly increases PTB protein levels. Considering that transfection efficiency approaches 50%, this suggests that the transfected HeLa cell population is overexpressing PTB by approximately fourfold. Furthermore, mutation of the PTB-expressing plasmid by introducing a frame-shift mutation near the N terminus of the gene shows a loss of the PTB inhibitory effect on β mRNA levels (using the same transfection and RNA mapping analysis as shown in Fig. 1B), demonstrating that PTB protein expression is required for the mRNA reduction effect. This result also excludes the possibility that there is competition for transcription factors between the CMV promoters of the cotransfected plasmids (Fig. 2B). In view of PTB's known effect on splicing, it was further possible that the reduction in mRNA levels was caused by failure to splice efficiently. However, transfection of the β cDNA construct with each of the different PTB isoforms (PTB1, -2 and -4) still showed a marked reduction in mRNA levels (with all three PTB isoforms) as shown by S1 mapping using the 3′ β probe (Fig. 2C, β pA band). Note that the expression levels are much lower with this cDNA gene construct, since intron sequences are required for efficient transcription (19) and export of mRNA (47). The 60-nt band in the lane lacking PTB cotransfection (lane 1) may reflect a cryptic splicing event, since a potential 5′SS sequence is present at this location. Interestingly, this product is also repressed by PTB overexpression. As a final control, we tested whether PTB overexpression might be affecting overall transcription levels. Nuclear run-on analysis was performed on the β construct with or without cotransfected PTB (Fig. 2D). Quantitative comparison of β nascent transcription with or without PTB cotransfection revealed no significant change compared to the cotransfection control Va transcript, which is transcribed by Pol III instead of Pol II. Note that the overall NRO signals are twofold lower in the β plus PTB sample than with β alone, as judged by the endogenous histone gene signal. This may reflect lower RNA recovery in this experiment. Overall, these data indicate that the threefold reduction in mRNA levels caused by PTB overexpression cannot be attributed to effects on transcription.

PTB overexpression inhibits mRNA 3′ end cleavage.

Many of the mechanisms that might account for the observed reduction in mRNA levels caused by PTB overexpression are ruled out in the above experiments (Fig. 2). We therefore directly tested the possible effect of PTB overexpression on mRNA 3′ end processing both in vivo and in vitro. For in vivo 3′-end analysis, a specific 3′-end S1 probe was designed that can distinguish between polyadenylated transcripts (pA transcripts) and uncleaved (read-through) transcripts extending beyond the normal pA site. The ratio of pA to read-through bands gives a measure of the efficiency of mRNA 3′ end processing. As shown in Fig. 3, cytoplasmic RNA from β transfections in the absence of cotransfected PTB gave virtually no read-through signal, indicating the high efficiency of the β-globin pA site. However, when any of the three PTB isoforms was cotransfected with β, a threefold reduction in the β pA band was observed with a commensurate fivefold increase in the read-through signal. These results strongly suggest that the reductions in mRNA levels mediated by PTB overexpression are the result of impaired mRNA 3′ end processing.

To obtain direct evidence for PTB inhibition of mRNA 3′ end processing, in vitro 3′ processing reactions were carried out using synthetic ³²P-labeled β or α pA site RNA substrates, synthesized by SP6 RNA polymerase (Fig. 4, input band). 3′ processing reactions were performed using nuclear extracts under standard conditions that allow 3′ processing but not polyadenylation to occur (36). Either recombinant PTB (0.25 μM) or an equivalent weight of bovine serum albumin (BSA) was added to the reactions as indicated. Since the amount of PTB in the unsupplemented nuclear extracts was measured at ∼0.2 μM, the levels of PTB were increased about twofold. For both the β and α substrates, the 5′ and 3′ products of cleavage are clearly visible (lane 1), though for β the 5′ product is a doublet band due to heterogeneity in the actual site of cleavage. Importantly, PTB but not BSA addition caused a clear threefold reduction in the level of 3′ end cleavage. These results demonstrate that increasing the levels of PTB twofold causes a severe inhibition of the 3′ cleavage reaction.

PTB competes with CstF₆₄ binding to α and β pA signals.

It seemed plausible that the capacity of PTB to inhibit the 3′ cleavage reaction might result from competitive binding of PTB and CstF to the α and β DSEs, since both of these sequences are pyrimidine rich. This would be analogous to the competitive binding of U2AF and PTB at the 3′SS pyrimidine tract as described above. To investigate this possibility, the competitive binding of α and β pA site RNAs to CstF and PTB was monitored by UV cross-linking analysis (Fig. 5). Each RNA was incubated with purified CstF fractions followed by the addition of increasing amounts of recombinant PTB. Both the α and β RNAs gave a single cross-linking protein band of 64 kDa with CstF alone, corresponding in size to CstF₆₄, the subunit known to directly bind the DSE of the pA signal (25, 52, 53, 60). However, as PTB was titrated into the binding reactions, a lower-mass (57 kDa) band corresponding to PTB became apparent. Furthermore, as the PTB concentration was increased, the higher CstF₆₄ band appeared to be displaced. Interestingly, with the α RNA, complete displacement of CstF occurred at 250 ng of PTB, while for β RNA the larger amount of 500 ng was required to displace CstF₆₄. This probably reflects the relative binding strength of these two pA sites for CstF, since α has a short GU-rich DSE with a lower affinity for CstF₆₄ while β has a longer GU- and U-rich DSE with a higher affinity for CstF₆₄. These data provide a clear molecular explanation for the inhibitory effect of PTB on the mRNA 3′-end cleavage reaction. As with PTB competing for the binding of U2AF at 3′SS pyrimidine tracts, so too PTB can block the binding of CstF to pA site DSEs.

Reducing PTB levels by RNAi has a selective effect on pA site recognition.

The advent of RNAi technology has allowed the possibility of reducing expression of specific proteins by transfecting tissue culture cells with short duplex RNAs corresponding in sequence to target mRNAs (16, 27). This technique has recently been applied to down-regulate PTB expression and resulted in the alteration of alternative splicing profiles for the FGF-R2 gene, known to be regulated by PTB (58). We wished to apply this same procedure to study PTB regulation of mRNA 3′-end cleavage as demonstrated in these studies. Two different small interfering RNAs (siRNAs), known to partially or more completely oblate PTB expression, were utilized (Fig. 6 and 7). As indicated by Western blot analysis, PTB levels were differentially reduced following HeLa cell transient transfection with these two PTB-specific siRNAs compared to results with an unrelated siRNA preparation (Fig. 6A, upper panel). As an internal control the level of actin expression was also measured and remained unchanged, indicating the specificity of the siRNAs. As shown in Fig. 6B and C, siRNA-induced knockdown of PTB had no effect on the levels of mRNA expressed from either the β or α construct. In contrast, levels of mRNA from the C2 construct were found to be highly sensitive to PTB knockdown (Fig. 7A). The human C2 pA signal (as described above) is unusual in possessing a USE but no DSE. Previous in vitro studies indicated that PTB is required for the positive activity of the C2 USE on mRNA 3′ end processing (36). Consistent with these studies, the C2 pA site (unlike α and β globin) does show sensitivity to the PTB-specific siRNAs. As shown in Fig. 7A, a stepwise decrease in C2 mRNA is clearly evident, with C2-transfected HeLa cells previously transfected with the two PTB-specific siRNAs. Furthermore, this effect is not seen with HeLa cells transfected by the unrelated siRNA. Using a C2 S1 3′-end probe, a 290-nt band corresponding to the C2 mRNA is detected, and this signal diminishes (up to threefold) compared to the Va control signal with the PTB-specific-siRNA-treated cells.

Although PTB is shown here to act as a positive activator of the C2 pA signal, it is also known that CstF is required for the recognition and polyadenylation of the C2 pA site (36). Consistent with this fact is the observation that overexpression of PTB blocks mRNA expression from the C2 gene construct, just as we observed for the α and β gene constructs (Fig. 7A). Thus, HeLa cells transfected with the C2 construct showed reduced 3′-end signals when cotransfected with a PTB expression plasmid. To verify that PTB was able to bind the C2 USE in competition with CstF₆₄, we performed UV cross-linking experiments with a synthetic C2 pA signal RNA. The pA signal was first incubated with CstF₆₄, and then increasing amounts of recombinant PTB were added to the reaction. Figure 7B shows that PTB was able to efficiently compete with CstF₆₄ for binding to the C2 USE. This result confirms that PTB can act as a negative regulator of both DSE-dependent and USE-dependent pA sites by blocking the binding of CstF₆₄ and thereby inhibiting 3′-end processing.