Substrate-dependent differences in U2AF requirement for splicing in adenovirus-infected cell extracts.

U2AF has been characterized as an essential splicing factor required for efficient recruitment of U2 small nuclear ribonucleoprotein to the 3'-splice site in a pre-mRNA. The U2AF65 subunit binds to the pyrimidine tract of the pre-mRNA, whereas the U2AF(35) subunit contacts the 3'-splice site AG. Here we show that U2AF35 appears to be completely dispensable for splicing in nuclear extracts prepared from adenovirus late-infected cells (Ad-NE). As a consequence, the viral IIIa and cellular IgM introns, which both have suboptimal 3'-splice sites and require U2AF35 for splicing in nuclear extracts from uninfected cells, are transformed to U2AF35-independent introns in Ad-NE. Furthermore, we present evidence that two parallel pathways of 3'-splice site recognition exist in Ad-NE. We show that the viral 52,55K intron, which has an extended pyrimidine tract, requires U2AF for activity in Ad-NE. In contrast, the IgM intron, which has a weak 3'-splice site sequence context, undergoes the first catalytic step of splicing in U2AF-depleted Ad-NE, suggesting that spliceosome assembly occurs through a novel U2AF-independent pathway in Ad-NE.

through a base pairing interaction with this sequence and U2 snRNA (reviewed in Ref. 4). Binding of U2 snRNP requires auxiliary splicing factors such as SF1/mBBP and U2AF (5). U2AF is a heterodimer consisting of a 65-and a 35-kDa subunit. The U2AF 65 subunit binds specifically to the pyrimidine tract through its RNA recognition domains (6), whereas U2AF 35 has been shown to make contact with the AG dinucleotide at the 3Ј-splice site (7)(8)(9).
In fact, introns appear to fall into one of two classes: AG-dependent introns and AG-independent introns (10). The AG-dependent introns typically have weak pyrimidine tracts, which make unstable interactions with U2AF 65 . In such introns, the U2AF 35 interaction with the 3Ј-splice site AG dinucleotide becomes important, by stabilizing the U2AF 65 interaction with the pyrimidine tract (11). In contrast, AG-independent introns do not require U2AF 35 for activity. In such introns, binding of U2AF 65 to the strong pyrimidine tract is usually sufficient to aid in the recruitment of U2 snRNP to the branchpoint sequence (7).
We are using the adenovirus major late region 1 (L1) as a model pre-mRNA to study the mechanisms controlling alternative splice site usage in adenovirus-infected cells. In the L1 unit, a common 5Ј-splice site can be joined to two alternative 3Ј-splice sites, resulting in the formation of the so-called 52,55K (proximal 3Ј-splice site) or IIIa (distal 3Ј-splice site) mRNAs. Early during virus infection, the 52,55K 3Ј-splice site is used exclusively, whereas the IIIa splice site becomes the preferred site in late virus-infected cells (reviewed in Ref. 12). Our previous work has demonstrated that regulated expression of the IIIa mRNA is, to a large extent, controlled by two sequence elements: a 49-nucleotide-long repressor element (the 3RE), which binds the SR family of splicing factors; and a 28-nucleotide-long virus infection-dependent splicing enhancer (the 3VDE).
Although we know much about the function of the 3RE (reviewed in Ref. 12), it appears clear that the 3VDE is the major element responsible for the enhanced splicing phenotype of the IIIa mRNA late during virus infection (13). The 3VDE is the minimal element required to confer an enhanced splicing phenotype to a heterologous transcript in nuclear extracts prepared from late virus-infected cells (Ad-NE). The 3VDE consists of the IIIa branch point sequence, pyrimidine tract, and AG dinucleotide. A limited mutational analysis of the 3VDE further revealed that the IIIa pyrimidine tract is a critical element required for the activated splicing of the IIIa pre-mRNA in Ad-NE (13, 14). However, much to our surprise, the increase in IIIa splicing did not correlate with an increased binding of U2AF 65 to the IIIa pyrimidine tract. Collectively, our results have suggested that the splicing enhancer activity of 3VDE may operate through a mechanism that might, in fact, be U2AF-independent (13, 15).
Here we extend these studies by examining the significance of the U2AF heterodimer for splicing in Ad-NE. Our results suggest that U2AF 35 is dispensable for splicing in Ad-NE. Thus, pre-mRNAs that are AG-dependent in HeLa-NE become AG-independent in Ad-NE. We further show that two parallel pathways of 3Ј-splice site recognition appear to function in adenovirus-infected cells. Consensus-type introns, containing long pyrimidine tracts, require U2AF 65 for activity, whereas at least one model transcript, the cellular IgM pre-mRNA, undergoes the first catalytic step of splicing in U2AF-depleted Ad-NE. These results suggest that splicing of certain suboptimal pre-mRNAs may be completely U2AF-independent in Ad-NE. Furthermore, the results indicate that a novel U2AF-independent pathway of spliceosome assembly exists in late adenovirusinfected cells and that this pathway appears to selectively favor splicing of weak 3Ј-splice sites, a feature typical for many of the adenoviral introns that are activated late during infection.
U2 snRNA Depletion-Oligonucleotide-directed RNase H cleavage of U2 snRNP in U2AF-depleted Ad-NE was done essentially as described previously (22). The oligonucleotides used were E15 directed against the 5Ј-end of U2 snRNA and R5S directed toward the 5S rRNA (23). Briefly, 100 pmol of respective oligonucleotide was pre-incubated in Ad(⌬U2AF) for 5 min at 30°C before initiation of splicing using the IgM or IgM-GA/C transcripts. The conditions for splicing were as described above.
Spliceosome Assembly-Standard splicing reactions were set up using Ad-NE or Ad(⌬U2AF) and the 32 P-labeled IgM-GA/C transcript. Reactions were incubated for 30 min at 30°C, mixed with heparin (final concentration, 0.5 g/l), and resolved on a 4% (60:1 acrylamide/bisacryalamide) native polyacrylamide gel, which was cast in a buffer containing 50 mM Tris-glycine and 5% glycerol. The gel was run in a cold room at 350 V for ϳ2 h in the Tris-glycine buffer lacking glycerol.
Western Blot-The extent of U2AF depletion was tested for each depleted extract preparation by Western blotting comparing it with serial dilutions of 10 l of HeLa-NE or Ad-NE (both with a total protein concentration of 10 mg/ml). Extracts were subjected to SDS-PAGE on a 12.5% gel under reducing conditions and transferred to a nitrocellulose membrane using a semidry transfer apparatus (Bio-Rad). U2AF was detected using a polyclonal anti-U2AF 35 antibody (11) or a monoclonal anti-U2AF 65 antibody (MC3) (24) and visualized by chemiluminescence according to the manufacturer's protocol (Pierce).

IIIa Splicing Becomes AG-independent in Nuclear Extract
Prepared from Adenovirus-infected HeLa Cells-Introns containing weak pyrimidine tracts are typically AG-dependent. In such pre-mRNAs, the 3Ј-splice site AG dinucleotide must be recognized by U2AF 35 prior to the first catalytic step of the splicing reaction. In contrast, removal of AG-independent introns requires the AG dinucleotide only during the second catalytic step. Therefore, AG-independent introns can still undergo the first catalytic step of splicing in transcripts where the 3Ј-splice site AG has been mutated to GA. Such interrupted splicing reactions accumulate the characteristic exon 1 and exon 2-intron lariat splicing intermediates (25).
The adenovirus IIIa 3Ј-splice site is considered to be a weak 3Ј-splice site because it has a short pyrimidine tract and is inefficiently spliced in HeLa-NE (26). According to the current model, this would suggest that the IIIa 3Ј-splice site should be classified as an AG-dependent intron. To test this hypothesis, we mutated the IIIa 3Ј-splice site from AG/A to GA/C (Fig. 1, IIIa and IIIa-GA/C). We also changed the first nucleotide of the second exon because previous work has shown that this nucleotide has a strong influence on U2AF 35 binding, with cytidine being less efficient in recruiting U2AF 35 compared with other nucleotides (7).
FIG. 1. Pre-mRNA splicing substrates used in this study. The sequence composition of splicing substrates at the 3Ј-splice site is shown, including branch sequence (underlined; branchpoint adenosine is in bold) pyrimidine tract (stretches containing two or more pyrimidines are highlighted), 3Ј-splice site (shown in bold), and the beginning of the 3Ј-exon (in capital letters).
As a control for an AG-independent substrate, we used the adenovirus major late first intron (AdML), which is often used as a prototypical AG-independent intron (7,16,17). The major late intron contains a very strong pyrimidine tract ( Fig. 1, AdML). For the same reason as described above, we changed the AdML 3Ј-splice site from AG/A to GA/C (Fig. 1, AdML and AdML-GA/C). All four substrates were tested in in vitro splicing reactions using HeLa-NE or Ad-NE.
The experimental conditions were selected such that the IIIa and AdML wild type substrates should be spliced with a reasonable efficiency in both HeLa-NE and Ad-NE (Fig. 2, A and  B, lanes 1 and 2). As has been shown for HeLa-NE (7,16,17), and as one would have expected for Ad-NE, the mutated AdML-GA/C substrate could still undergo the first catalytic step of splicing in both types of extracts (Fig. 2B, lanes 3 and 4).
In contrast, mutating the AG dinucleotide at the IIIa 3Јsplice site abolished all splicing activity in HeLa-NE ( Fig. 2A,  compare lanes 1 and 3), suggesting that IIIa, as predicted because of its weak sequence context (26), is an AG-dependent substrate. Interestingly, incubation of the IIIa-GA/C transcript in Ad-NE resulted in a very efficient accumulation of the first exon and the exon 2-lariat splicing intermediate ( Fig. 2A, lane  4). This result was surprising because it suggests that the IIIa pre-mRNA was converted from an AG-dependent intron in HeLa-NE ( Fig. 2A, lanes 1 and 3) to an AG-independent intron in Ad-NE (Fig. 1A, lane 4). It is noteworthy that the accumulation of splicing intermediates did not result from a general increase of the splicing activity in Ad-NE, as shown by the reactions using the wild type AdML substrate (Fig. 2B, compare lanes 1 and 2). In some experiments, splicing of the IIIa-GA/C transcript produced faint bands migrating at the position of the authentic spliced mRNA and one band migrating slightly faster ( Fig. 2A, lane 4). Reverse transcription-PCR analysis showed that the faster migrating species was a minor aberrant product resulting from the use of a cryptic AG dinucleotide located 10 nucleotides downstream of the authentic IIIa 3Јsplice site (data not shown). No cDNA product corresponding to an authentic splicing event was observed with the IIIa-GA/C transcript, suggesting that the slower migrating species most likely was a breakdown product.
To investigate whether the weak pyrimidine tract of IIIa (26) was the critical element in the IIIa pre-mRNA responsible for the AG-dependent splicing in HeLa-NE, we replaced 3VDE with the branch site and pyrimidine tract from the rabbit ␤-globin second intron. In addition, the globin 3Ј-splice site was mutated from AG/A to GA/C. As shown in Fig. 2C, splicing intermediates were visible in both HeLa-NE and Ad-NE with the IIIa(-3VDE)GA/C transcript, indicating, as one would expect, that a strong pyrimidine tract can convert the IIIa transcript into an AG-independent substrate also in HeLa-NE.
IgM Becomes an AG-independent Substrate in Nuclear Extract Prepared from Adenovirus-infected Cells-Hypothetically, the AG-independent splicing of the IIIa pre-mRNA in Ad-NE might be part of a viral strategy to enhance splicing of introns with a weak 3Ј-splice site context, a feature typical for many of the late viral 3Ј-splice sites. To test this hypothesis, we used the well-characterized immunoglobulin pre-mRNA substrate in our assay system (17,27,28) (Fig. 1, IgM and IgM-GA/C). This pre-mRNA has been characterized as an AG-dependent substrate, which requires binding of U2AF 35 to its 3Ј-splice site for activity (17,28). As expected from previous results, splicing of the IgM transcript in HeLa-NE was abolished when the 3Јsplice site AG was mutated to GA (Fig. 3, lane 3). In contrast, incubating the IgM-GA/C transcript in Ad-NE resulted in an efficient accumulation of exon 1 and the exon 2-intron lariat splicing intermediates characteristic of the first catalytic step of splicing (Fig. 3, lane 4). We conclude that the IgM pre-mRNA, like the IIIa pre-mRNA, was converted to an AGindependent substrate in Ad-NE. Collectively, our results suggest that the cellular splicing machinery may have undergone a major shift in activity during virus infection, a shift that is not specific to viral transcripts.
U2AF Is Required for Splicing of the Strong 52,55K Intron in Nuclear Extracts Prepared from Infected Cells-We have previously shown that introns with a weak 3Ј-splice site context are activated in Ad-NE, whereas introns with a strong pyrimidine content are slightly repressed (14). The results presented here suggest that splicing of both types of introns becomes U2AF 35 -independent in Ad-NE. This finding prompted us to investigate whether the U2AF heterodimer was necessary for splicing in Ad-NE. For this experiment, U2AF was depleted from HeLa-NE and Ad-NE by oligo(dT)-cellulose chromatography (6,29). This protocol has previously been shown to remove both U2AF subunits almost to completion (21). The bound U2AF was subsequently eluted from the column and used for reconstitution experiments. As shown in Fig. 4A, the depletion protocol was effective and removed U2AF 65 and U2AF 35 to undetectable levels in the NE(⌬U2AF) and Ad(⌬U2AF) extracts, respectively (i.e. far below 1% of the initial amount; see Fig. 4B for a longer exposure).
In the first set of experiments, we tested whether U2AF was necessary for splicing of an intron containing a consensus type of 3Ј-splice site. For this experiment, we choose the 52,55K pre-mRNA, which has an extended polypyrimidine tract ( Fig.  1) that binds U2AF 65 efficiently (14). As shown in Fig. 5, depletion of U2AF from either HeLa-NE or Ad-NE abolished 52,55K splicing completely (Fig. 5, lanes 2 and 6). Re-addition of U2AF eluted from the oligo(dT) column restored 52,55K pre-mRNA splicing in both the depleted HeLa-NE and Ad-NE (Fig. 5, lanes 3 and 7). Taken together, these results are in agreement with the conclusion that splicing of an intron containing a strong polypyrimidine tract (Fig. 1) requires U2AF for activity in both virus-infected and uninfected extracts. However, we found one interesting difference: supplementing HeLa-NE depleted of U2AF with a recombinant U2AF 65 protein did not restore 52,55K splicing. In contrast, the U2AF 65 protein was able to restore 52,55K splicing in U2AF-depleted Ad-NE (Fig. 5, lane 8). This finding adds further support to our conclusion that 3Ј-splice site recognition is altered in Ad-NE.
The Weak IgM Intron Is Spliced in U2AF-depeleted Ad-NE-We have previously observed an inverse correlation between the efficiency of IIIa splicing and the stable recruitment of U2AF 65 to the weak IIIa pyrimidine tract (13, 14). This result may suggest that splicing of introns with a weak sequence context, as in the viral IIIa and also the cellular IgM pre-mRNAs, might occur in the absence of U2AF in Ad-NE. To test this hypothesis, our preferred transcript would have been the IIIa pre-mRNA. However, for unknown reasons, we have not been able to recover IIIa splicing in oligo(dT)-depleted extracts by any combination of reconstitution experiments (see Supplementary Fig. 1). Much additional work is needed to uncover what IIIa-specific factor(s) is lost/inactivated during the oligo(dT) fractionation of nuclear extracts.
Because reconstitution of IIIa splicing was not possible in the depleted extracts, we continued our analysis using the IgM pre-mRNA. The IgM intron resembles a typical late viral intron in that it has a 3Ј-splice site with a poor pyrimidine tract (Fig.  1). As shown in Fig. 6A, depletion of U2AF from HeLa-NE resulted in a complete loss of IgM splicing (lane 2). However, IgM splicing was fully restored by addition of the eluted NE-U2AF to the depleted NE(⌬U2AF) extract (Fig. 6A, lane 3). In contrast, in splicing reactions supplemented with the IgM-GA/C mutant transcript, no splicing intermediates were observed even in the presence of native U2AF eluted from the oligo(dT) column (Fig. 6A, lanes 5-8). As reported previously (17), and consistent with our results, this finding indicates that IgM splicing is AG-dependent and requires the U2AF heterodimer for activity in HeLa-NE.
Interestingly, incubation of the IgM pre-mRNA in Ad(⌬U2AF) resulted in a significant accumulation of the free exon 1 splicing intermediate (Fig. 6B, lane 2). This result is noteworthy because it suggests that neither U2AF 35 nor U2AF 65 is required for the first catalytic step of IgM splicing in Ad-NE. Supplementing the reactions with purified U2AF fully restored IgM splicing (Fig. 6B, lane 3). This result indicates that U2AF is capable of stimulating IgM splicing in the presence of a functional 3Ј-splice site, even if it appears not to be essential for the first catalytic step of splicing in Ad-NE (Fig.  3B, lane 4).
To test this hypothesis further, we analyzed the splicing phenotype of the IgM-GA/C transcript in Ad(⌬U2AF). As shown in Fig. 6B (lane 6), the exon 1 splicing intermediate accumulated essentially with the same efficiency as in the non-depleted Ad-NE (lane 5). Interestingly, the IgM splicing activity did not increase in reactions supplemented with isolated U2AF (Fig. 6B, lane 8). This result is important because it supports the conclusion that U2AF is not required for the first catalytic step of IgM splicing in Ad-NE. Furthermore, this result makes it unlikely that the first step of splicing observed with the IgM-GA/C pre-mRNA is caused by residual amounts of U2AF in the depleted extract. It is noteworthy that the lariat exon 2 intermediate diminished significantly in Ad(⌬U2AF)  6B). The most likely explanation is that U2AF-depleted extracts have a higher debranching activity. However, it is important to note that low levels of the lariat exon 2 intermediate are detectable in Ad(⌬U2AF) in a splicing-dependent manner (see below and Supplementary Fig. 2).
Requirement of a Functional U2 snRNP for IgM Splicing in U2AF-depleted Ad-NE-The finding that IgM splicing appears to be U2AF-independent in Ad-NE was intriguing. Therefore, we tried to find evidence that the accumulation of the band with a size of the IgM exon 1 in Ad(⌬U2AF) extracts was due to splicing and was not a breakdown product that for unfortunate reasons had the same size as exon 1. To test this, we used oligonucleotide-directed cleavage of U2 snRNA to determine whether accumulation of the first exon band was dependent on the catalytic activity of the spliceosome. As shown in Fig. 7B, pre-incubation of Ad(⌬U2AF) with oligonucleotide E15, directed against the 5Ј-end of U2 snRNA, effectively abolished the accumulation of the first exon intermediate (lane 4) using the IgM-GA/C pre-mRNA as a substrate. In contrast, pre-incubation of the depleted extract with the control oligonucleotide R5S, directed against 5S RNA, did not have negative effects on the accumulation of the first exon (Fig. 7B, lane 5). To further show that the band was exon 1 and not an unspecific fragment, we used transcript IgM-GA/C-ext (Fig. 7A), which has a first exon that is 24 nucleotides longer than exon 1 in the IgM-GA/C transcript. As shown in Fig. 7C, the size of exon 1 was shifted to a new position in both Ad-NE (lane 1) and Ad(⌬U2AF) extracts (lane 2). The increase in size of exon 1 was, as predicted, around 25 nucleotides. Oligonucleotide-directed depletion of U2 snRNA in Ad(⌬U2AF) resulted in loss in the accumulation of exon 1 (Fig. 7C, lane 3), supporting the conclusion that the IgM-GA/C transcript undergoes the first catalytic step of splicing in Ad(⌬U2AF). Further support for the conclusion that the exon 1 intermediate (Fig. 7, B and C) is generated by splicing can be seen by looking at the accumulation of the lariat exon 2 intermediate in Ad(⌬U2AF). As noted above, U2AFdepleted extracts appear to have a higher debranching activity compared with Ad-NE. However, weak bands corresponding to the lariat exon 2 intermediate can be seen in both Fig. 7B (lanes  3-5) and Fig. 7C (lanes 2-4) (for a longer exposure, see Supplementary Fig. 2). Interestingly, pretreatment of Ad(⌬U2AF) with the E15 oligonucleotide abolished accumulation of both exon 1 and the lariat exon 2 intermediate (Fig. 7B, lane 4; Fig.  7C, lane 3), whereas a control oligonucleotide had no effect on the accumulation of the splicing intermediates (Fig. 7B, lane 5;  Fig. 7C, lane 4).
To gain additional support for the hypothesis that IgM splicing is U2AF-independent in Ad-NE, we analyzed spliceosome assembly in Ad(⌬U2AF). As shown in Fig. 7D, the IgM-GA/C transcript was very efficient in forming the pre-spliceosomal A complex in Ad-NE (lane 1). Slower migrating species, which most likely correspond to the spliceosomal B and C complexes, can also be seen. The accumulation of all the complexes is dependent on spliceosome assembly because depletion of U2 snRNA in Ad-NE abolishes complex formation (data not shown). Interestingly, the A complex was also formed in Ad(⌬U2AF) (Fig. 7D, lanes 2 and 3). We show the results from two independent depleted extracts because we have noted that the efficiency of spliceosomal complex assembly in the U2AFdepleted extracts varies between different batches of extracts, suggesting that complexes becomes less stable in the absence of U2AF. However, we note that the pre-spliceosomal A complex and also small amounts of complexes that most likely correspond to spliceosomal complexes B (Fig. 7D, lanes 2 and 3) and C (Fig. 7D, lane 3) can be discerned in the depleted extracts (Fig. 7D, lanes 2 and 3). Collectively, all these results strongly support the hypothesis that splicing of the IgM intron in Ad-NE is, indeed, U2AF-independent.
U2AF from Ad-NE Is Fully Functional-The results presented here suggest that splicing in Ad-NE shows a transcriptspecific variability in the requirement for U2AF. This raises the question of whether U2AF in Ad-NE is fully functional, or whether it has become functionally inactivated or distorted in activity during infection. It should be noted that the steadystate level of both U2AF subunits is essentially the same in HeLa-NE and Ad-NE (Fig. 4). Thus, U2AF is not subjected to a virus-induced degradation.
To address the question of whether U2AF is post-translationally inactivated during virus infection, we compared the capacity of NE-U2AF and Ad-U2AF to activate IgM splicing in the two types of depleted extracts (NE(⌬U2AF) and Ad(⌬U2AF)). As shown in Fig. 6, both U2AF fractions showed an essentially indistinguishable capacity to activate splicing. Thus, NE-U2AF and Ad-U2AF were efficiently activating IgM wild type splicing in both NE(⌬U2AF) and Ad(⌬U2AF) (Fig. 6,  A and B, lanes 3 and 4). Also, NE-U2AF and Ad-U2AF were both unable to activate IgM-GA/C splicing in NE(⌬U2AF) (Fig.  6A, lanes 7 and 8). Similarly, NE-U2AF and Ad-U2AF were both unable to stimulate the first catalytic step of IgM-GA/C splicing in Ad(⌬U2AF) (Fig. 6B, lanes 7 and 8). Collectively, these results suggest that U2AF is not functionally inactivated as a general splicing factor during an adenovirus infection. This conclusion is further supported by the observation that both HeLa-U2AF and Ad-U2AF are able to activate splicing of the 52,55K pre-mRNA that contains a consensus type of 3Јsplice site (Fig. 5). DISCUSSION Previous results have suggested that the host cell RNA splicing machinery undergoes a major change in activity during a lytic adenovirus infection (reviewed in Ref. 12). Thus, splicing of introns with consensus-like 3Ј-splice site signals is repressed, whereas splicing of pre-mRNAs with weak non-consensus-like 3Ј-splice site signals, like the IIIa pre-mRNA, is activated in Ad-NE (14). In agreement with this hypothesis, we have shown that the enhanced splicing of the IIIa pre-mRNA in Ad-NE does not correlate with a more efficient recruitment of U2AF 65 to the IIIa pyrimidine tract (13). In fact, we have observed the opposite, a reduced binding of U2AF in Ad-NE. This and other observations have led us to speculate that splicing of certain suboptimal pre-mRNAs may operate without efficient U2AF recruitment and potentially may be U2AF-independent (13, 15).
Here, we have further investigated the significance of U2AF for splicing in Ad-NE. Although we have only used a few substrate RNAs, the evidence that is accumulating appears to provide strong support for the hypothesis that 3Ј-splice site recognition differs substantially in Ad-NE compared with extracts prepared from uninfected cells. In fact, our results suggest that two parallel pathways of 3Ј-splice site recognition exist in late adenovirus-infected cells. The normal U2AF-dependent recognition of the pyrimidine tract is functional, as demonstrated by the fact that 52,55K splicing requires U2AF for activity in Ad-NE (Fig. 5). This conclusion is further supported by the observation that U2AF isolated from Ad-NE is fully functional, and it is as active as U2AF isolated from uninfected HeLa cells in reconstituting IgM (Fig. 6) and 52,55K splicing ( Supplementary Fig. 1) in U2AF-depleted infected or uninfected extracts. Thus, it seems unlikely that U2AF is a target for a virus-induced post-translational inactivation, as is the case for the SR family of splicing factors (30). It appears likely that pre-mRNAs with consensus-type introns continue to use the U2AF pathway in Ad-NE.
More interestingly, our results provide evidence for a second alternative assembly pathway in Ad-NE, which appears to have a relaxed U2AF requirement and, on certain transcripts, may be U2AF-independent. For example, we show that the AG dinucleotide, which appears to be a signature for the requirement of the U2AF 35 subunit in 3Ј-splice site definition (7)(8)(9), is dispensable for the first catalytic step of splicing in Ad-NE. A virus-induced shift to AG-independent splicing was observed on two substrates, the viral IIIa ( Fig. 2A) and the cellular IgM ( Fig. 3) pre-mRNAs, which both contain 3Ј-splice sites with a suboptimal sequence context (Fig. 1) and therefore are AG-dependent in HeLa-NE ( Figs. 2A and 3). As expected, splicing of substrates containing strong pyrimidine tracts, which are already AG-independent in HeLa-NE, were also AG-independent in Ad-NE (Fig. 2, B and C). Secondly, we show that the first catalytic step of IgM splicing occurs in Ad-NE depleted of U2AF (Fig. 6B). This observation was interesting because depletion of U2AF from HeLa-NE completely abolished IgM splicing ( Fig.  6A; see also Ref. 17). The step 1 splicing activity of the IgM-GA/C mutant transcript was not stimulated by re-addition of a fraction highly enriched in U2AF to the depleted Ad-NE, a result that argues strongly against the hypothesis that trace amounts of U2AF in the depleted Ad-NE are responsible for the efficient first step of splicing in the depleted Ad-NE. Thirdly, we show that the viral 52,55K pre-mRNA, which requires U2AF for activity in both HeLa-NE and Ad-NE (Fig. 5), is also activated by a recombinant U2AF 65 protein. However, this activation is only observed in Ad-NE (Fig. 5). Taken together, these results provide strong support for the hypothesis that 3Ј-splice site recognition differs significantly in uninfected versus adenovirus-infected cells. Thus, a parallel, potentially U2AF-independent spliceosome assembly pathway appears to exist in adenovirus-infected cells. This pathway appears to favor splicing of introns with a weak 3Ј-splice site context, a feature typical for many of the adenoviral introns that are activated at late times of infection (14).
To the best of our knowledge, there are only two examples of U2AF-independent splicing described in the literature. First, MUD2, which is the yeast homolog of U2AF 65 (31), is not an essential gene (32). Second, splicing of a chimeric major late/ ␤-globin hybrid transcript has been reported to take place in a U2AF-depleted extract supplemented with high amounts of the SR protein SC35 (33). Although formally possible, it appears FIG. 7. Requirement of U2 snRNA for IgM splicing in U2AF-depleted Ad-NE. A, schematic structure of the pre-mRNAs used. Radiolabeled IgM-GA/C (B) or IgM-GA/C-ext (C) pre-mRNAs were spliced in Ad-NE (Ad) or Ad(⌬U2AF) pretreated with an oligonucleotide directed against the 5Ј-end of U2 snRNA (U2) or a control oligonucleotide directed against 5S (5S). In B, lane 1 shows a control reaction in which the IgM wild type transcript was spliced in Ad-NE. Products were resolved by gel electrophoresis and visualized by autoradiography. The positions of the splicing substrate, product, and splicing intermediates are indicated to the left of the panels. The size of the DNA marker fragments (lane M) is shown in C. D, spliceosome assembly in U2AFdepleted Ad-NE. The IgM-GA/C transcript was incubated under splicing conditions in Ad-NE or Ad(⌬U2AF). After a 30-min incubation at 30°C, spliceosomal complex formation was analyzed by native gel electrophoresis.

U2AF-independent Splicing
unlikely that an SR protein takes over the function of U2AF in adenovirus-infected cells because available data suggest that SR proteins are functionally inactivated as splicing factors late during an adenovirus infection through a virus-induced dephosphorylation (30,34). In conclusion, our data support the hypothesis that an alternative 3Ј-splice site binding factor of viral or cellular origin may replace the classical U2AF heterodimer in spliceosome assembly on certain transcripts in Ad-NE. Despite extensive efforts, we have not been able to biochemically purify the elusive component that appears to substitute for U2AF in spliceosome assembly on the non-consensus type of 3Ј-splice sites in Ad-NE. However, this work is clearly at the center of our current interest.