Nucleic Acids Research

RNA secondary structure is not a determinant for polyadenylation in the FBP1 gene

The TA-rich region located in the 3[prime] end of the FBP1 gene is an important efficiency element for polyadenylation in both orientations in vivo, although activity in the reverse orientation could not be detected in vitro using yeast whole-cell extract and exogenously added RNA substrate (data not shown). To further analyze essential features of this region, we made several modifications to the TA repeat in its natural context in the 397 fragment. First, point mutations were designed to abolish the ability of the (UA)₁₄ sequence to fold into an RNA hairpin structure (Fig. 2A). To predict the structure that the RNA of this region would adopt, we used the MFOLD program on the sequence between positions 1125 and 1164 (Fig. 2B). It is important to note that the hairpin shown here is also one of the most significant structural motifs when the entire sequence contained in the 397 or 144 fragments was folded with the same program (data not shown).

Because of the presence of UG repeats alternating with UA repeats adjacent to the (UA)₁₄ element and the stability of the G:U pairs, the RNA in the forward orientation forms a longer hairpin than the one predicted for the reverse orientation. In the forward orientation, a single mutation (SM) at position 1146 (A/U change, Fig. 2A) opens up the loop of the hairpin, but causes a minimal change in the free energy (-7.8 kcal/mol compared with -8.4 kcal/mol of the wild-type RNA, Fig. 2B). When a double mutation (DM, Fig. 2A) at positions 1146 (A/U change) and 1149 (U/A change) is constructed, new base pairs are formed to give a longer stem, producing an even lower free energy (-10.2 kcal/mol).

In the reverse orientation, the wild-type sequence forms a predicted RNA hairpin containing only the (UA)₁₄ element. The SM mutation introduces a central unpairing which modifies the perfect symmetry of the hairpin (Fig. 2B), causing a significant reduction in the free energy (from -6.4 kcal/mol in the wild-type sequence to -2.1 kcal/mol). The double mutation (DM, Fig. 2B) produces a complete reorganization of the whole region, giving only a very short and unstable hairpin in the UA repeat region (-0.8 kcal/mol).

A

B

Figure 3. In vivo polyadenylation of pVPA plasmids carrying nested deletions of the 397 fragment. (A) 5[prime] deletions. Total RNA was isolated from yeast cells carrying the plasmid indicated above each lane (F, forward orientation; R, reverse orientation). Samples were analyzed on a 2% agarose-formaldehyde gel. The blot was hybridized with probe from ACT1 sequences, as shown. The polyadenylation efficiency (expressed as the percentage of truncated transcripts) is shown at the bottom of each lane, and was determined as an average from three different experiments, with the standard deviation error being in all cases <3%. ACT1, R, and T denote the chromosomal ACT1, readthrough and truncated mRNAs, respectively. I indicates the intermediate band discussed in the text. pME729 is the original plasmid without insert. (B) 3[prime] deletions. Experimental conditions and symbols as in (A).

These structures predict that if the global RNA structure is responsible for processing, the point mutations would not affect polyadenylation efficiency in the forward orientation, but they would progressively and adversely affect it in the reverse orientation. However, there was no loss of processing efficiency, with wild-type levels of truncated transcripts in the reverse orientation of the fragment (Fig. 2C) or in the forward orientation (Table 1). These results argue against a role for the RNA hairpin in processing of transcripts in the reverse orientation. They also suggest that a perfect TA repeat in both parts of the (TA)₁₄ stretch is not necessary for 3[prime] efficient processing in either orientation, and that some TA pairs are redundant.

An internal deletion of the (TA)₁₄ region gives a reduction in both polyadenylation and transcription termination efficiencies in the reverse orientation

We have shown previously (12) that reduction of the (TA)₁₄ element to a simple TATATA hexanucleotide still gives wild-type levels of processing in the forward orientation, but complete removal of the (TA)₁₄ sequence greatly decreased the polyadenylation efficiency. To determine the effect of reduction of the TA content of this element on 3[prime]-end formation in the reverse orientation, we tested the internal deletions shown in Figure 2A. The D1 fragment carries a deletion of 10 bp inside the (TA)₁₄ region, which reduced the TA pairs from 13 to nine, the D2 fragment conserved just three TA pairs and D3 carries a complete deletion of the TA-rich sequence. When we analyzed these deletions in the reverse orientation, we found the results to be similar to those obtained for the forward orientation. Reduction of the TA content did not significantly affect the processing efficiency (89% WT, 86% D1 and 80% D2, Fig. 2D), in agreement with the idea that some of the TA pairs are redundant. With the D3 fragment, the level of truncated species dropped to 30%, confirming a major role for this sequence in the reverse direction. In this orientation, the decrease in truncated species is accompanied by an increase in the readthrough transcripts. The ratio of the total amount of RNA transcribed from the ACT1 promoter in the pVPA plasmids to that from the chromosomal ACT1 gene is similar for all four plasmids in this experiment (3.5 for the WT fragment, 3.3 for D1, 3.2 for D2 and 2.5 for D3), indicating that termination has also been affected. These results contrast with what we found previously for the D3 deletion in the forward orientation, which was defective for polyadenylation but competent for termination (12). Thus, in the reverse orientation, transcription termination as well as polyadenylation has been disrupted by removal of the TATATA sequence.

When the D2 deletion was made in the 3[prime]-144 fragment, reduction of the TA repeat had a more dramatic effect than the same deletion in the longer 397 fragment, reducing the polyadenylation efficiency from 65 to 36% (Table 1 and Fig. 2D). This confirms the presence of additional activating sequences in the region between 1198 and 1445 of fragment 397, which can compensate for the partial disruption of the TA repeat in the D2 deletion. With the D3 deletion, the efficiency dropped further to 12% and, as in the case of the 397 fragment, there is a corresponding increase in the readthrough transcripts.

Overlap exists between the YLR376 promoter and the FBP1 terminator

Run-on experiments, carried out previously, have indicated that transcription occurs through the whole 397 fragment in the direction of the FBP1 gene but not in the opposite direction (12). The region between the coding sequences of the FBP1 and YLR376c genes contains just 191 bp, and the distance between the FBP1 poly(A) site and the YLR376c start codon is only 50 nucleotides, suggesting that elements important for 3[prime]-end formation and transcription initiation might overlap.

To study the expression of the FBP1 and YLR376c genes, we first analyzed mRNAs derived from the chromosomal copies of these two genes in the strain CJM88. Because the expression of the FBP1 gene is subjected to catabolite repression, we prepared RNA samples from this strain grown in either repressing or derepressing conditions. RNAs were analyzed by northern blots, using as a probe a 2 kb fragment corresponding to the FBP1 and YLR376c region (Fig. 4A). As expected, an FBP1 mRNA of 1.3 kb is detected only after derepression. Another transcript of 0.95 kb is detected under both culture conditions, and is the size predicted for the YLR376c mRNA. It is worth noting that the part of the filter corresponding to the YLR376c transcript is overexposed in this figure. In fact, according to our calculations, the levels of YLR376c transcript are about 10 times lower, relative to those of the FBP1 transcript, under derepressed conditions of this gene. Neither transcript is observed in the RG1-5d strain, which lacks most of both ORFs (Fig. 4A).AB

A

B

Figure 4. Analysis of the transcription of the YLR376c gene. (A) Analysis of the transcription of the FBP1 and YLR376c genes in the CJM88 and RG1-5d strains in conditions for repression (R) and derepression (D) of the FBP1 gene. Total RNA was isolated and samples were analyzed by a 1.8% agarose-formaldehyde gel. A 2 kb fragment which hybridizes to both genes was used as probe. The lower panel shows an ethidium bromide stain of the rRNA in these samples. (B) Analysis of the transcription of the YLR376c gene in the strains carrying the pVPA plasmids. Total RNA was isolated. Samples were analyzed by a 2% agarose-formaldehyde gel. A probe of 142 bp corresponding to the coding region of this gene was used. Y indicates the band corresponding to the transcription of this gene, and R denotes the readthrough transcript.

To analyze the YLR376c promoter, we used some of the pVPA plasmids employed previously for the polyadenylation competition assays. RNAs from strains carrying these plasmids were analyzed by northern blots, using a probe complementary to the YLR376c ORF sequence contained in the insert. This probe will detect a transcript (Y, 0.6 kb) initiating from the YLR376c promoter, as well as the readthrough transcript (R, 1 kb) originating from the ACT1 promoter (Fig. 4B). The ACT1 promoter gives a level of transcription into the insert that is very similar to that of the FBP1 promoter in derepressed conditions (47). In spite of a high level of transcription into the FBP1 3[prime] region (Fig. 3A), the YLR376c transcript derived from the pVPA397 plasmid was easily detectable (Fig. 4B, lane 1). The level of readthrough transcript is low but distinguishable. Deletion of 42 bp from the 5[prime] end of the insert (the pVPA355 plasmid) causes a complete disappearance of the YLR376c transcript, indicating that an element required for initiation at the YLR376c promoter has been affected. In this construct, the polyadenylation signal of the FBP1 gene has not been damaged (12), as can be inferred from the small amount of readthrough transcript and wild-type levels of truncated transcript (Table 1). Successive deletions increase the amount of readthrough because they affect the efficiency of FBP1 polyadenylation signal (12) and, obviously, do not permit YLR376c transcription. In the D2 internal deletion (Fig. 4B, lane 5) the YLR376c expression is not affected, but it drops significantly in the D3 deletion (lane 6), suggesting that the TA repeat is also important for function of the YLR376c promoter. Assuming that poly (A) tails in mature mRNAs are usually 70-80 bases long in yeast and that the YLR376c transcripts end at the ADH1 trap, we calculate from the size of the Y transcript that YLR376c transcription begins in a region shortly downstream of the (TA)₁₄ element and probably overlaps with precursor transcripts from the FBP1 promoter.

DISCUSSION

In our previous study, we found that the 3[prime] end of the FBP1 gene is unusual because of the novel organization of its polyadenylation signal elements and the uncoupling between 3[prime]-end formation and transcription termination (12). We also showed that polyadenylation and transcription termination are efficient in forward and in reverse orientations. In this report, we analyze the poly(A) signal in reverse orientation and show that sequences necessary for transcription of the neighboring gene are found upstream of the FBP1 poly(A) site.

The existence of polyadenylation signals working in both orientations is not a general case in yeast (30). The situation found in the CYC1 (34) and ADH1 (17) genes could be due to overlap with the signal of another gene transcribed convergently. In the case of the FBP1, the adjacent gene (YLR376c) transcribes in the same direction. While this suggests no physiological necessity for a polyadenylation site in the reverse direction, the fragment in this orientation does function as a strong polyadenylation signal in vivo (although not in vitro), and it even has a higher efficiency than the 3[prime] flank of the TRP1, TRP4 and ARO4 genes tested in a similar system (17). Thus, analysis of such a signal can provide further insight into the requirements for yeast polyadenylation.

Table 1 shows a comparison between the efficiency of polyadenylation in the forward (12) and reverse orientations of several fragments from the 3[prime] end of the FBP1 gene. By deletion analysis, we found that at least five regions were important for optimal efficiency of the flipped FBP1 3[prime] flank. One is the cluster of cryptic poly(A) sites activated by the presence of upstream elements. However, the cellular machinery is not very demanding about poly(A) sites, since processing of transcripts with this cluster deleted is still very high. A similar retention of efficient polyadenylation, with utilization of new sites, was found in the forward FBP1 orientation (12) and has also been reported for the CYC1 (10) and GAL7 (11) genes, when natural poly(A) sites were deleted.

Four other regions had additive effects on polyadenylation efficiency. The first two were far upstream (almost 200 nucleotides) of the poly(A) site cluster. The furthest contained a match to a TAG...TAGT motif important for the function of many other yeast poly(A) sites. The second, a 17 bp region removed in deletion 3[prime]-144, does not correspond to any reported motifs. Another region involved in polyadenylation efficiency is identified when additional sequence is removed to create 3[prime]-94 fragment. This deletion has eliminated three TA pairs of an almost perfect (TA)₁₄ repeat located 11 bases upstream of the first poly(A), and also a TACATA, which could function as an efficiency element (10). Finally, as might be expected, more severe encroachment into the (TA)₁₄ stretch from either end (as in constructs 5[prime]-307 and 3[prime]-84) further decreased 3[prime]-end processing. These deletions also demonstrated that, in this context, a single TA tract of 10-12 bases, despite its reputation as a potent signal, was not directing efficient polyadenylation. Precise excision of just the (TA)₁₄ stretch caused a dramatic drop in polyadenylation efficiency, and this defect was most pronounced when the two far upstream regions were missing. In summary, the (TA)₁₄ repeat contributed the most to polyadenylation in the reverse orientation, but additional upstream regions were essential for full function. This result is different from the one for the TRP4 gene. For this, 3[prime]-end formation in the reverse orientation depended only on one TAGT motif (28). These results support the emerging model that many yeast polyadenylation signals are complex mosaics of efficiency elements of varying strength whose sequence specifications we do not fully understand. Even less is known about how the polyadenylation machinery recognizes such diverse and plentiful elements in choosing a poly(A) site.

An interesting question is whether the intrinsic symmetry of a TA repeat signal in a natural poly(A) site would always lead to function in both orientations, as happens in genes like MRP2 (48) or FBP1 (12 and this study). If this is the case, it is possible that the putative secondary structure adopted by these sequences is involved in the process of 3[prime]-end formation, as has been suggested for other genes in several organisms (8,15,28,49-51). The RNA structure of the HIV-1 enhancer is a critical determinant of in vitro poly(A) site recognition by the cleavage and polyadenylation specificity factor in mammalian cells (52), and an RNA hairpin is involved in processing of the histone messengers in higher eukaryotes (49). However, in the case of the reverse orientation of the FBP1 gene, point mutations in the (TA)₁₄ region which create a much shorter and less stable RNA hairpin did not affect the processing efficiency. With an internal deletion that conserves only the last three TA pairs, there was no loss of efficiency in the forward direction, and only a slight decrease in the reverse orientation. Thus, a TA hexanucleotide which is not able to form a stable RNA hairpin is sufficient if other close sequences are not removed. Further analysis is required, but it seems likely that secondary structure does not play much of a role in in vivo yeast polyadenylation, and the processing factors are probably recognizing primary sequence in the mRNA rather than a structural motif.

Another interesting aspect of mRNA 3[prime]-end formation in yeast is the relationship between polyadenylation and termination. In higher eukaryotes, the efficiency of transcription termination by RNA pol II depends on the presence of functional polyadenylation signals as well as on the strength of a transcription pause site downstream of the poly(A) site (1). Recent findings of the association of polyadenylation factors with RNA pol II transcriptional complexes (53) have lent credence to a model in which binding of these factors to nascent polyadenylation signals makes the polymerase susceptible to termination at the downstream pause site. In S.cerevisiae, a coupling between mRNA maturation and termination of transcription has been established for the CYC1 gene (27) and deletion of the sequence in this gene required for cleavage and polyadenylation has also been shown to abolish transcriptional termination in vivo (25,26). However, some mutations in the ADH2 gene affect the polyadenylation signal but not in vitro termination (54). Removal of the (TA)₁₄ stretch in the FBP1 gene did not affect transcriptional termination in forward orientation, indicating that in this example, the elements required for transcription termination are separable or do not exactly overlap with those required for polyadenylation (12). In contrast, as described in this report, deletions in the FBP1 3[prime] flank, which were detrimental to polyadenylation in the reverse orientation, always resulted in a corresponding increase in the level of readthrough transcripts, even for a complete removal of the (TA)₁₄ stretch. This observation indicates that transcription termination was also adversely affected by these mutations, and leads to the conclusion that the elements which specify transcription termination are not the same in the forward and reverse orientations of the FBP1 sequence. These results may not seem as puzzling if the following interpretation is considered. As we proposed previously (12), partial processing complexes may assemble on forward orientation FBP1 precursor lacking the TA repeats and be sufficient to promote transcription termination but not polyadenylation. The FBP1 sequence in the reverse orientation is not a natural polyadenylation signal, i.e. one optimized through evolution, and in addition may not have a good pause site associated with it. As we have shown here, the repeats of the TATATA consensus on this precursor are the major determinants of polyadenylation efficiency and, by induction, polyadenylation complex assembly. In this case, their removal would be expected to affect both processes.

A possible function for transcription termination signals downstream of poly(A) sites is to prevent RNA polymerase from entering the neighboring gene. In this respect, our results indicate that in a short intergenic region of just 191 bp, the elements specifying 3[prime]-end formation in the FBP1 gene and those needed for initiation of transcription of the YLR376c gene overlap. We found that one promoter element required for YLR376c transcription is located upstream of the positioning and efficiency elements for FBP1 polyadenylation, and a second is found in the (TA)₁₄ stretch.

The location of a transcription terminator element within a yeast promoter has been shown in the case of the URA3 promoter (32), although in this case this terminator has no physiological role. In other systems, like Ty transposons in yeast (29) or mammalian virus (55,56), overlapping of terminators and promoters has also been found. The extreme compactness of the yeast genome (35) suggests that overlapping between gene ends may be a common event for this organism. A statistical analysis of intergenic distances shows that ORFs with divergent promoters are, on average, 618 bp while convergent terminators are 326 bp apart (35). This has been interpreted as 309 and 163 bp being the average size for promoters and terminators, respectively (35,36). Although these sizes fit well with experimentally determined lengths for both (25,57), they are only averages of a wide range of different sizes. A close inspection of chromosome III, for instance, reveals that many ORFs are much closer than those average values but some others are much farther. Thus, in some ORFs intergenic distances represent true non-genic regions, but for many other ORFs, the short intergenic distances mean that either the sizes of promoter and terminator are shorter than those values or they overlap. We favor the second possibility because it is difficult to conceive that short distances, such as 191 bp in the case of FBP1, could accommodate separately two promoters, two terminators or a promoter-terminator combination. The case of ARO4 and HIS7, analyzed by Springer et al. (36), in which 417 bp separate both genes, could be an example of spatially separated, non-overlapping elements. However, the case of FBP1/YLR376c, in which promoter and terminator clearly overlap is probably more frequent. A statistical analysis of the intergenic distances in chromosome III shows that 58 of the 94 promoter-terminator combinations (62%) are shorter than 472 bp (163 + 309). This raises the question of how the two transcriptional machineries cope with these kinds of situations. In natural situations, it appears that high levels of transcriptional interference have been avoided. However, low levels of interference are still possible and even tolerated. We have observed that in at least one case of overlapping terminators, that of YIL161w/POT1, there is only 56 bp between the stop codons and a small amount of interference can be detected (S.Puig, J.E.Pérez-Ortín and E.Matallana, in preparation). Whether this is also the case for promoter-terminator combinations, such as that of FBP1/YLR376c, will require further study.

ACKNOWLEDGEMENTS

We thank Dr M. Kessler for his suggestions and for his support with the in vitro experiments, and Dr E. Matallana for critical reading of the manuscript. We also thank Drs G. Braus,J. M. Gancedo and S. Emr for the gift of plasmids and yeast strains. We thank Servei de Bioinformàtica de la Universitat de València for technical support. This work was funded by PB91-0329 grant from DGICYT (Spanish Ministry for Education and Science) to J.E.P.-O., 1160 grant from the Universitat de València to M.d.O. and 5R01 GM4 1752 grant from the National Institute of Health of USA to C.M.