| Nucleic Acids Research | Pages |
In vivo selection for intronic splicing signals from a randomized pool
Introduction
Materials And Methods
PCR cloning and in vivo selection
In vivo analysis
Statistical analysis
Results
Construction of viral DNAs containing a randomized PPyT
In vivo selection of competent PPyTs
Characterization of the PPyTs in replication-competent viruses
In vivo analysis of viruses containing selected PPyTs
PPyTs selected in vivo allow efficient viral replication
PPyTs selected in vivo affect splicing at either step 1 or step 2
Discussion
Acknowledgements
References
In vivo selection for intronic splicing signals from a randomized pool
ABSTRACT
INTRODUCTION
Splicing is the process by which intervening sequences are removed from precursor messenger RNAs (pre-mRNAs). The reaction takes place within a large complex called the spliceosome and requires the participation of five small nuclear ribonucleoprotein particles (snRNPs) as well as many non-snRNP proteins (for reviews see 1-5). Pre-mRNA splicing occurs by a two-step mechanism (Fig.
In vertebrates, cis-acting RNA signals that direct intron removal are only loosely conserved (6). RNA elements that participate include the 5[prime]-ss, 3[prime]-ss, BPS and a pyrimidine-rich stretch downstream of the BPS, the polypyrimidine tract (PPyT) (Fig.
The PPyT does not interact directly with snRNPs but, rather, is recognized by one or more protein splicing factors. Furthermore, this element appears to have multiple roles in the splicing reaction. Early in the reaction the splicing factor U2AF binds to the PPyT and promotes the association of U2 snRNP with the adjacent BPS (15,17). A late function for the PPyT is implied by the observation that certain mutations within this region allow the first step to proceed but inhibit the second step (18,19). For example, Smith et al. (19) have shown that insertion of a stem-loop into the PPyT of an efficiently spliced pre-mRNA results in a specific block after the first step. Sequence requirements of the PPyT have been examined by a number of groups. Mutational analysis has demonstrated that insertion of purines into the PPyT generally has a negative effect on splicing; long stretches of U residues activate splicing more than other combinations of pyrimidines and a continuous stretch of five U residues is important for efficient splicing (20-22).
Retroviruses employ balanced splicing, amongst other mechanisms, to express multiple proteins from a single primary transcript (23). The unspliced mRNA encodes the structural and enzymatic proteins while a spliced mRNA encodes the envelope glycoproteins. In addition to providing essential proteins, the unspliced message is also packaged into the virion to serve as the viral genome. Simple retroviruses, such as avian sarcoma virus (ASV), do not encode proteins that regulate splicing and, therefore, this incomplete or `balanced' splicing is mediated through interactions between the viral RNA and the host cell splicing machinery. We have found that this regulation can be manifested as a restriction or `partial block' at either step in the splicing pathway (18). The determinants include sub-optimal signals (the BPS and PPyT), which in turn are responsive to the ASV env exonic splicing enhancer (ESE) (18,24,25).
The requirement for balanced splicing imposes strong selective pressure on retroviral splicing signals. Because it is relatively easy to manipulate retroviral DNA, large numbers of test sequences can be examined for functionality by selection for replication competence. We have previously demonstrated that the ratio of spliced and unspliced retroviral mRNAs produced in ASV-infected cells can be dramatically affected by U->C transitions that are outside a five U stretch within the ASV env PPyT. Here we employ the retroviral system to further examine the requirements of the PPyT, starting from DNA populations that include all possible combinations of pyrimidine sequences of a defined length. The method we describe for examining the PPyT could be modified to examine other splicing elements within either the intron or exon.
MATERIALS AND METHODS
PCR cloning and in vivo selection
The method of cloning the randomized PPyT into the ASV genome is described in the text and Figure
In vivo analysis
Assays for the presence of reverse transcriptase (RT) activity in the supernatant of transfected cells were performed essentially as described (18). An aliquot (3 µl) of supernatant was mixed with poly(C) template, oligo(dG) primer and [[alpha]-32P]dGTP and incubated at 37°C for 90 min. The reactions were spotted onto DEAE-Dextran paper and washed three times with 1× SSC and twice with 95% ethanol. The filter was then dried and subjected to autoradiography. S1 nuclease analysis and primer extension were performed as described previously (18).
Statistical analysis
A test of binomial proportions was used to compare the frequency of U residues at each position in either a 10mer or 12mer to the overall fraction of U residues in the collection. The binomial used is an approximation to the hypergeometric distribution, suitable for large numbers.
Fisher's exact test (FET) was used to compare proportions of pyrimidines in several contexts. The number of U residues in the selected 10 nt PPyTs was compared with that in the selected 12 nt PPyTs using FET, as well as comparison of the proportion of U residues in one position of the sequences to the same proportion in the rest of the positions.
Two sample T tests were used to compare the longest runs of C residues (or U residues) in one set of sequences with those in another set. This test compared the mean maximum run length in one group with that in the other. The Wilcox test on two samples was also used, with adjustments for ties, for the same purpose.
A permutation test was constructed to compare the longest runs of C residues (or U residues) in any viral sequence set with the lengths of such runs if the letters in each sequence were placed randomly. First, the average length of such runs was determined for a sequence set. Next, each sequence in the set was randomly permuted, the longest sequence of C residues (or U residues) was found within each sequence and their overall average length determined. The average length of the permuted sequences was compared with that of the non-permuted sequences. If the average length of the permuted sequences equaled or exceeded the average length of the non-permuted sequences in 50 (of 1000) or fewer cases then we concluded that the average run lengths of C residues (or U residues) was longer than expected, with a P value [le]5%.
We estimated the number of sequences that would function given the presence of a duplicate as described previously (26).
RESULTS
Construction of viral DNAs containing a randomized PPyT
We have previously reported that transitions within the env PPyT of ASV severely affected viral replication and altered the balance of RNA splicing (18). To examine the contribution of each pyrimidine within the PPyT, we established an in vivo system that selects PPyTs from a randomized pool based upon their ability to support balanced splicing and concomitant replication. The first step (Fig.
Figure 1. Schematic diagram of the ASV pre-env mRNA transcript and the selection protocol employed. (A) The intron and exon structure of the env transcript is depicted. PPyT that was subjected to randomization is indicated by (Y)8,10,12. The viral genes, gag, pol, env and src are indicated below the genome and the locations of the splicing signals are listed above, including the 5[prime]-splice site (5[prime]-ss), branch point sequence (BPS), polypyrimidine tract (PPyT) and 3[prime]-splice site (3[prime]-ss). (B) A flow chart of the selection protocol is shown (see text for details). (C) Diagram of the two steps of the splicing reaction. Bacteria were transformed with the library of viral plasmids containing the randomized PPyTs and a sample of the primary transformants was counted to determine the number of individual clones that made up the starting pool. Taking into account the number of unique PPyTs of each length (28, 210 and 212), the likelihood that all possible combinations of pyrimidines within the PPyT exists in the starting pool can be calculated (Table 1). The results indicate that the starting pools should be sufficiently complex to test all PPyTs. Figure 2. Generation of randomized viruses through a PCR cloning strategy. Small horizontal arrows indicate primers; a region of one primer was randomized and the three different lengths of randomized sequences are indicated by (Y)8,10,12. This degenerate oligonucleotide contains unique sequences flanking the randomized region and was used to amplify a fragment of viral DNA corresponding to the downstream exon. A second PCR reaction was used to extend this fragment using sequences upstream of the randomized region. To reduce any potential bias introduced by PCR, the templates for each reaction did not contain sequences corresponding to the randomized region. Lastly, the 1 kb PCR product was ligated into the viral vector IS1 (see text) using unique restriction enzyme sites. To estimate the randomness of the three pools, transformants representing libraries of DNA clones from each length of PPyT were pooled and the plasmid DNAs extracted. Each pool of DNA was subjected to automated DNA sequencing and the results showed that the sequences flanking the PPyTs were unique, while the PPyT was randomized as expected (data not shown). Within the PPyT, the peak traces showed a qualitative bias towards C residues, although the degree of bias is unclear because comparison of fluorescence intensities is not quantitative. To estimate the degree of randomness of the 12 and 10 nt PPyTs more accurately, bacteria were transformed with an aliquot of DNA from these two pools, 16 and 9 individual clones were isolated, respectively, and the region of the PPyTs in each clone was sequenced. The average percent of U residues present in the PPyTs of these clones is shown in Table Table 1. Characterization of randomized libraries before selection
In vivo selection of competent PPyTs
In our selection protocol (Fig.
Following transfection, the cultures were maintained for 10 days to allow any replication-competent viruses to spread. After this period, supernatants were collected and the virus was pelleted and analyzed as outlined in Figure
To test for the occurrence of multiple viruses in a single tissue culture plate, several plasmid DNAs from the final subcloning step were isolated and sequenced. Six transfected plates were analyzed in this manner and from three to six clones derived from the same transfection were analyzed. Results from two out of the six transfections indicated that more than one virus was amplified in the same tissue culture plate. In one case, two clones from the same culture differed by a single nucleotide in the PPyT region. A likely interpretation is that a virus introduced from the starting pool had acquired a second mutation during the 10 day passage in tissue culture. In the second case, the two clones differed in sequence significantly, suggesting that two independent viruses were amplified to significant levels in the same culture. These results show that more than one replication-competent virus could be selected in the transfection step under our conditions, but, as expected, this is a relatively rare event.
Characterization of the PPyTs in replication-competent viruses
The selection procedure outlined above was performed several times for each of the viral DNA pools. In total, ~150 100 mm plates were transfected and replication-competent virus appeared in ~80 plates. Sequences of the PPyTs from viruses selected for replication competence are listed in Table
Table 2. PPyTs selected in vivo Table 3. Characteristics of selected PpyTs
Figure 3. Occurrence of U and C stretches in PPyTs. PPyTs from the in vivo selected viruses and the unselected pools of DNA were examined for the presence of groups of nucleotides. (A) Groups of U residues are shown for the 10 and 12 nt randomizations. (B) Groups of C residues are shown for the 10 and 12 nt randomizations. Because the selection procedure is dependent upon viral replication and retroviruses have high mutation rates, it is expected that mutations outside the PPyT might be found in the virus populations at the end of the selection procedure. Additional mutations could also be introduced during the various PCR steps. To estimate the occurrence of mutations outside the PPyT we examined sequences immediately surrounding the PPyT as well as sequences upstream of the BPS and downstream of the 3[prime]-ss region. Analysis of the 10 and 12 nt tracts revealed that 8 out of the 85 viruses selected in vivo contained mutations outside the PPyT (Table 3). One secondary mutation that is not shown in Table 2 consists of a single base change in the pol gene, 51 nt upstream of the BPS (tract 46). This change introduces a stop codon into the integrase reading frame and the virus is predicted to express a protein that is truncated at the C-terminus by 16 amino acids. Apparently, this mutation does not significantly affect viral replication (see below). Together, the 10 and 12 nt selection experiments produced 81 tracts, of which 8 differed from the input material, suggesting that the rate of occurrence of secondary mutations is <10%. Two of the isolated tracts were identical (Table 2, tracts 56 and 57), which allows us to estimate the minimum number of 12 nt tracts that should be capable of supporting balanced splicing and viral growth at the 95% confidence level to be 466 (26). The maximum likelihood estimate is 2134, indicating that about half of all combinations of pyrimidines (4096) are likely to support balanced splicing in this system. Because no duplicate tracts were identified in the 14 PPyTs from in vivo selected viruses containing 10 nt PPyTs, no estimate can be made concerning the likely number of competent tracts that exist. Transfections using the 8 nt PPyT pool never produced replication-competent viruses containing tracts of 8 nt. Of the four viruses that were isolated, two tracts were 10 nt long, one was 11 and one 12. We also note that a significantly larger amount of DNA was needed to initiate viral infection (1 µg/60 mm plate) and that the percentage of plates that produced viruses was also reduced compared with the 10 and 12 nt tracts (~10% versus ~50%). These results suggest that 8 nt is below the minimal distance between the BPS and 3[prime]-ss. Because the 10 nt tracts were the shortest isolated, we conclude that the minimal length of the PPyT in this system is 10 nt and the corresponding minimal distance between the BPS and 3[prime]-ss is 16 nt. Table Published reports demonstrated the importance of a five U stretch in the PPyT and thus we analyzed the starting pools of DNA and the selected viruses for such stretches (Fig. Identical comparisons were made for C stretches in Figure Each position in the tracts was also examined for U or C occurrence. Comparisons of 10 nt tracts from selected viruses with the starting pools of DNA revealed a preference for U at the second position downstream of the BPS (P = 0.028). The viruses containing 12 nt PPyTs demonstrated a preference for U at the first position downstream of the BPS (P = 0.013).
A
B
In vivo analysis of viruses containing selected PPyTs
Because the selection procedure produced a wide variety of PPyTs, we decided to examine a representative subset more closely. A total of 10 different PPyTs were chosen based upon their U content and the 3[prime]-ss regions of these isolates were re-inserted into the original infectious viral DNA clone. Three of these reconstructed viral DNAs either failed to produce virus or acquired additional mutations within the PPyT during further passage (data not shown). This suggests that ~30% of the selected viruses are unstable or contain additional mutations outside the 3[prime]-ss region that allowed them to grow in tissue culture. We have not investigated further the viruses that failed to replicate.
Sequences in the 3[prime]-ss region of the seven reconstructed clones that produced virus are shown in Figure
Figure 4. Description of parental and in vivo selected 3[prime]-ss regions examined for functional capability. The BPS, PPyT and 3[prime]-ss are listed for the viruses that we examined in more detail after selection. The sequences of two previously characterized viruses are listed, wild-type and IS1. The viruses derived from the selection procedure are listed in two groups; one group contains PPyTs of 10 nt in length, the other 12 nt. The number of U residues within each tract is listed.
PPyTs selected in vivo allow efficient viral replication
Figure
Previously we have shown a correlation between balanced splicing and viral growth. To examine the balance of splicing in the in vivo selected viruses, we isolated RNA from infected cells after the 10 day passage shown in Figure
Figure 5. Growth rate and splicing pattern of selected viruses in vivo. (A) The appearance of RT activity in the supernatant of transfected cells was monitored as described (Materials and Methods). `Mock' indicates cells transfected with vector DNA containing no viral sequences. (B) Total cellular RNA from chronically infected cells was hybridized with a DNA probe spanning the 5[prime]-ss and then subjected to S1 nuclease digestion. The bands corresponding to spliced and unspliced RNA were quantitated by phosphorimaging and the percentage of each species is listed below the lanes. The `mock' lane indicates RNA isolated from a mock transfection.
PPyTs selected in vivo affect splicing at either step 1 or step 2
The control of balanced env splicing in the ASV wild-type and variants has been shown to involve either step 1 or step 2 of the splicing reaction (18). To examine potential differential effects on the two steps by the selected PPyTs, we assayed for the presence of lariat-exon 2 intermediates in vivo by primer extension using RNA extracted from infected cells (Fig.
Figure 6. Detection of lariat-exon 2 intermediates in vivo. Total cellular RNA from infected cells was examined by primer extension assay using a labeled primer that hybridizes to intronic sequences as indicated in the diagrams to the right. `Mock' indicates RNA from mock transfected cells. RNAs from cultures infected with viruses containing in vivo selected PPyTs were analyzed in a similar manner. Some showed a pattern similar to the wild-type while others showed accumulation of lariat-exon 2 intermediates. Both of these types of splicing regulation are observed with the 10 and 12 nt tracts tested. We note that the relative amount of lariat-exon 2 detected varies among viruses. For example, 12Y-7U (lane 10) reproducibly shows greater accumulation of lariat-exon 2 intermediate relative to unspliced than the IS1 parental virus. Conversely, other viruses show anintermediate level of lariat-exon 2 intermediates (less than IS1, but more than wild-type). The significance of this variation is unclear.
DISCUSSION
In this report we describe the use of a retroviral system to selectively amplify sequences that support balanced splicing from a pool of randomized sequences. This system selects for splice sites that allow weak but functional splicing. As weak splice sites are also a prerequisite for alternative splicing of cellular pre-mRNAs, the results of our studies are applicable to regulation of host pre-mRNA splicing. The experiments described here focus on analysis of the PPyT. The PPyT at the env 3[prime]-ss in ASV was altered in length, randomized with respect to C and U residues and replication-competent isolates were obtained. Previously, we found that this PPyT was sensitive to subtle changes; addition of two pyrimidines activated splicing and abrogated viral growth, whereas a single transition within this extended PPyT reduced the amount of spliced product and restored viral growth (18). Thus, the results from our randomization experiments which show that many different PPyTs can support viral growth were unexpected. The diversity of the PPyTs of selected viruses may result from a number of participating factors. Several proteins are predicted to interact with the PPyT and the binding preferences of these proteins is likely to be degenerate (28-32). The wide variety of PPyTs selected in our experiments could reflect different balances in such interactions. Another source of PPyT diversity may result from the fact that different residues in the tract contribute to different steps of the splicing reaction. For example, we have shown that the PPyT can play a role in either the first or second step of the splicing reaction (18).
A close examination of the functional PPyTs selected in vivo revealed some trends. For example, shorter PPyTs contained more U residues on average than the longer ones. In vitro experiments have demonstrated that increasing the U content of a PPyT will increase usage of the corresponding 3[prime]-ss. This has also been addressed in vivo through analysis of two competing 3[prime]-splice sites (33). It was noted that the proximal PPyT was shorter and contained fewer U residues than the distal PPyT. Interchanging the two PPyTs or increasing the length and U content of the proximal PPyT resulted in its activation. In our system, the competition is a splice/no splice choice. In addition, all of the selected PPyTs produced similar ratios of spliced and unspliced products, which allows us to make comparisons between different lengths of PPyTs with all else being equal.
Because we were unable to isolate viruses containing shorter PPyTs, we conclude that the minimal length of a PPyT in our system is 10 nt. One interpretation of this result is that when the BPS and 3[prime]-ss are in close proximity, the splicing machinery is unable to identify them and the splice site is skipped. The 10 nt tract corresponds to a distance between the branch point adenosine and the 3[prime]-ss of 16 nt. Reed (22) has suggested the minimal distance between BPS and 3[prime]-ss to be 16 nt based upon sequence comparisons. Our results provide confirmation of this suggestion.
Comparison of the stretches of U or C residues in the PPyTs of viruses selected in vivo reveals that the 10 nt tracts contained more long U stretches than the starting pool and the 12 nt tracts contained longer C stretches. We conclude that the selective pressure with respect to the shorter tracts was different from that on the longer tracts. One interpretation of the data is that the selection in the 10 nt tracts was based upon the presence of longer U stretches and the selection in the 12 nt tracts was for longer C stretches. Because the selection requires maintenance of balanced splicing, the presence of a higher U content in the 10 nt tracts may suggest that recruitment of positive factors can compensate for the short distance between the BPS and 3[prime]-ss. The effect of longer C stretches within the PPyT has not been reported previously and the role is unclear.
Further analysis demonstrated a preference for U at the second position downstream of the BPS for the 10 nt tract and at the first position downstream of the BPS for the 12 nt tract. The meaning of these preferences is not clear, but it may indicate a requirement for U residues at the 5[prime]-end of the PPyT. This has been suggested by others through in vitro mutagenesis studies and our data may provide in vivo confirmation of these results (20).
In earlier studies, we described the control of balanced splicing in ASV as a partial block in the splicing reaction. This block allows some unspliced transcripts to escape splicing to the cytoplasm, but sufficient RNA is spliced to allow appropriate production of env mRNA (18,24). We have shown that a block may occur at two different stages within the splicing pathway, either before step 1 or during step 2. The step at which the block occurs in infected cells is reflected in the presence or absence of lariat-exon 2 intermediates; presence of these intermediates is taken as indicative of a block at step 2, while absence indicates a block at step 1. When we analyzed a subset of the PPyTs from selected viruses for the presence of lariat-exon 2 intermediates, we find that some display this intermediate in vivo, while others do not. Because these viruses differ primarily in the composition of pyrimidines within their PPyTs, this supports a role for the PPyT in the second, as well as the first, step of splicing.
Previously, we have shown that in our system, in vivo selected suppressor mutations in the BPS and PPyT affect in vitro splicing as well as UV crosslinking of two splicing factors, SAP49 and U2AF65 (34). Similar analysis was carried out with the selected PPyTs described here. Preliminary results indicate that the crosslinking patterns of these two factors are also affected when the various selected PPyTs are compared (data not shown).
Other groups have used in vivo selection procedures to examine elements important for splicing. Studies by Chen and Chasin examined exon skipping in the dihydrofolate reductase gene (35). This negative selection procedure uncovered inactivating mutations in the PPyT, BPS and 3[prime]-ss. Coulter et al. (36) examined exonic sequences which could activate splicing, starting with a pool of DNA that was randomized at a short stretch within the downstream exon. Through iterative rounds of transfection, amplification of spliced product and reconstruction of unspliced template, a number of sequences that promote splicing at the upstream intron were selected. This procedure is limited to examination of sequences within the exon. The retroviral system described here is capable of identifying weak but functional splicing elements that reside within an intron. With minor modifications, this same retroviral system procedure can be used to examine any element important for splicing.
ACKNOWLEDGEMENTS
We would like to thank Warren Kruger and John Taylor for critical comments on the manuscript. We also thank the Fox Chase Cancer Center Fannie E. Ripple Foundation Biotechnology Core Facility for preparation of DNA primers. This work was supported by National Institutes of Health grants CA71515, AI40385 and AI40721, institutional grant CA06927 from the National Institutes of Health and also by an appropriation from the Commonwealth of Pennsylvania. The contents of this manuscript are solely the responsibility of the authors and do not necessarily represent the official views of the National Cancer Institute or any other sponsoring organization.
REFERENCES
This page is run by Oxford University Press, Great Clarendon Street, Oxford OX2 6DP, as part of the OUP Journals
Comments and feedback: www-admin{at}oup.co.uk
Last modification: 19 Sep 1998
Copyright©Oxford University Press, 1998.
CiteULike 
Connotea 
Del.icio.us What's this?
This Article 
Abstract
Print PDF (243K)
Alert me when this article is cited
Alert me if a correction is posted
Services
Email this article to a friend
Similar articles in this journal
Similar articles in ISI Web of Science
Similar articles in PubMed
Alert me to new issues of the journal
Add to My Personal Archive
Download to citation manager
Request Permissions
Commercial Re-use Guidelines
for Open Access NAR Content
Google Scholar
Articles by Bouck, J.
Articles by Katz, R. A.
Search for Related Content
PubMed
PubMed Citation
Articles by Bouck, J.
Articles by Katz, R. A.
Social Bookmarking
What's this?