Domain analysis of human U5 RNA. Cap trimethylation, protein binding, and spliceosome assembly.

We have analyzed the sequence requirements of the human U5 RNA during small nuclear ribonucleoprotein (snRNP) and spliceosome assembly. A collection of mutant derivatives of the human U5 RNA gene was constructed in a U1 expression vector and transiently transfected in mammalian cells. Using immunoprecipitation and affinity selection assays, the cap trimethylation, the binding of Sm proteins and of the U5 snRNP-specific protein p220, as well as the assembly of the U4/U5/U6 triple snRNP and of spliceosomes were determined. By mutational analysis we were able to assign distinct functions to several structural elements of the human U5 RNA. Efficient binding of the Sm proteins requires the 3′ stem-loop. Both the Sm protein-binding site and the 3′ stem-loop are necessary for the formation of the trimethyl guanosine cap, consistent with Sm protein binding being a prerequisite for cap trimethylation. Specific elements of the U5 RNA 5′ stem-loop contribute to efficient p220 association, in particular stem Ib. Interestingly, the highly conserved loop I appears to be a multifunctional element; in addition to its function in splice-site selection the 5′ loop is involved in binding of p220 and in the assembly of the U4/U5/U6 triple snRNP. In sum, this mutational analysis has identified four functional domains of the human U5 RNA.

We have analyzed the sequence requirements of the human U5 RNA during small nuclear ribonucleoprotein (snRNP) and spliceosome assembly. A collection of mutant derivatives of the human U5 RNA gene was constructed in a U1 expression vector and transiently transfected in mammalian cells. Using immunoprecipitation and affinity selection assays, the cap trimethylation, the binding of Sm proteins and of the U5 snRNP-specific protein p220, as well as the assembly of the U4/U5/U6 triple snRNP and of spliceosomes were determined. By mutational analysis we were able to assign distinct functions to several structural elements of the human U5 RNA. Efficient binding of the Sm proteins requires the 3 stem-loop. Both the Sm protein-binding site and the 3 stem-loop are necessary for the formation of the trimethyl guanosine cap, consistent with Sm protein binding being a prerequisite for cap trimethylation. Specific elements of the U5 RNA 5 stem-loop contribute to efficient p220 association, in particular stem Ib. Interestingly, the highly conserved loop I appears to be a multifunctional element; in addition to its function in splice-site selection the 5 loop is involved in binding of p220 and in the assembly of the U4/U5/U6 triple snRNP. In sum, this mutational analysis has identified four functional domains of the human U5 RNA.
Nuclear pre-mRNA splicing proceeds in a dynamic multicomponent RNA-protein complex, the spliceosome (for reviews, see Green, 1991;Krä mer, 1995;Moore et al., 1993;Nilsen, 1994;Sharp, 1994). The assembly of the spliceosome and the two steps of the pre-mRNA splicing reaction require a large number of essential cofactors, among them four small nuclear ribonucleoproteins (snRNPs), 1 the U1, U2, U4/U6, and U5 snRNPs (for a review, see Lü hrmann et al., 1990). Among these the U5 RNA is of central importance, since, together with the U6 RNA, it is thought to be closely associated with the catalytic center of the spliceosome and may participate directly in splicing catalysis (for a review, see Weiner, 1993). Evidence for this has been derived from genetic, biochemical, and cross-linking studies in the yeast and mammalian systems (see below).
While the U5 RNA secondary structure is conserved, primary sequence conservation of the U5 RNA is surprisingly low. Extensive sequence comparisons revealed that besides the highly conserved loop I with nine invariant positions, only the internal loop 2 and the Sm protein-binding site near the 3Ј end are conserved (Frank et al., 1994;Guthrie and Patterson, 1988). Deletion studies in yeast have defined a minimal functional U5 structure consisting of the 5Ј loop and stem Ic, the internal loop 2 with an adjacent closing stem of any sequence, and the Sm domain (Frank et al., 1994;see Fig. 2).
In the course of the splicing cycle, the U5 RNA becomes part of several dynamic RNA-protein complexes, the U5 core complex, the 20 S U5 snRNP, and the 25 S U4/U5/U6 triple snRNP Behrens and Lü hrmann, 1991); it is the triple snRNP form in which U5 is incorporated into the spliceosome.
Cross-linking and yeast genetic studies suggested that the U5 snRNP contributes to splice-site selection via both RNA-RNA and RNA-protein interactions. The U5 RNA interacts with the 5Ј and 3Ј splice sites; the U5-exon 1 interaction occurs already before the first step and helps to align the two exons during the second step of the reaction (Cortes et al., 1993;Newman and Norman, 1991;Newman and Norman, 1992;Sontheimer and Steitz, 1993;Wyatt et al., 1992). Studies of the U5 protein components have so far concentrated on the PRP8 protein from yeast and its mammalian homologue, p220. This largest known spliceosomal protein is splicing-essential (Jackson et al., 1988) and highly conserved (Anderson et al., 1989;Hodges et al., 1995). Recent studies suggest that this protein may act in conjunction with the U5 RNA. It was detected in close contact with both splice sites and the branch point region and could be cross-linked to both pre-mRNA, lariat intermediate, and excised lariat, indicating that the U5 association with the splicing substrate is established early and persists throughout the splicing reaction (Garcia-Blanco et al., 1990;MacMillan et al., 1994;Teigelkamp et al., 1995a;Teigelkamp et al., 1995b;Whittaker and Beggs, 1991;Umen and Guthrie, 1995). An additional role of PRP8 in 3Ј splice-site selection was suggested by Umen and Guthrie (1995).
Because of its dynamic interactions and the presence of so many protein components, the U5 snRNP may be the spliceosomal snRNP with the most complex structural organization. For binding of the Sm proteins two determinants were found to be important, the Sm binding site and internal loop 1 (IL1) (Jarmolowski and Mattaj, 1993). However, where specific proteins are bound, is less clear; there is only some evidence from protection experiments in mammalian extracts that the extended 5Ј stem-loop structure of U5 may be involved (Bach and Lü hrmann, 1991;Black and Pinto, 1989). Finally, what interactions contribute to the association between the U4/U6 and the U5 snRNPs is completely unknown.
We have undertaken a mutational analysis of the human U5 RNA, using transient transfection of mammalian cells and several assays to determine cap trimethylation, Sm and p220 protein binding, as well as U4/U5/U6 triple snRNP and spliceosome assembly. As a result, we were able to map distinct functional domains of the U5 RNA.
Mutagenesis-T7-HuU5 contains the human U5 sequence behind the T7 promoter. A fragment was constructed from two oligonucleotides carrying the T7 promoter and the entire U5 sequence between EcoRI and HindIII restriction sites; this fragment was cloned into pEMBL 8(ϩ). T7 transcription with HindIII-cut T7-HuU5 is expected to generate a U5 RNA with five extra nucleotides at the 5Ј end (GGGAG) and no or four extra nucleotides at the 3Ј end (AGCU).
The pGEM4 U1/U5 expression construct was generated by polymerase chain reaction (PCR) methods from three fragments. First, the upstream noncoding region of the U1 RNA gene was amplified, using pUC-U1 (Ach and Weiner, 1987) (a kind gift from A. Weiner, Yale University, New Haven) as a template and oligonucleotides U1ups-5Ј and U1ups-3Ј. Second, a U5/U1 hybrid fragment was amplified containing the U5 coding region and the U1 3Ј box, using T7-HuU5 (see above) and oligonucleotides T7U5-5Ј and U5/U1ds. As a marker a short linker sequence (consisting of the first four nucleotides of the U1 coding region and a SalI restriction site; see Fig. 1) was inserted between the transcriptional start site of U1 and the 5Ј end of the U5 coding region. A BamHI restriction site was introduced by mutating the 3Ј box. The amplified fragments were cut with EcoRI/SalI (upstream U1 fragment) and with SalI/HindIII (U5/U1 hybrid fragment) and ligated together into pGEM4.
The pGEM4 U2/U5 expression construct was generated by PCR methods in two steps. First, a cassette vector including the U2 upstream and downstream noncoding regions interrupted by the first four nucleotides of the U2 coding region and restriction sites for SalI, ClaI, and BamHI was generated from three fragments. The upstream and downstream noncoding regions of the U2 RNA gene were amplified, using pGEM4-U2 (Wu and Manley, 1989) (a gift from J. Manley, Columbia University, New York) as a template and oligonucleotide pairs SP6-primer/U2ups-3Ј and U2ds-5Ј/U2ds-3Ј. The amplified fragments were cut with EcoRI/ClaI (upstream U2 fragment) and with ClaI/ HindIII (downstream U2 fragment) and ligated together into pGEM4 cut with EcoRI and HindIII. Second, a U5 coding fragment flanked by SalI and BamHI restriction sites was amplified using T7-HuU5 (see above) and oligonucleotides T7U5-5Ј and T7U5-3Ј. This amplified fragment was cut with SalI and BamHI and ligated into the cassette vector.
The mutant U5 derivatives (see Table I) were generated through PCR methods, using pGEM4 U1/U5 as a template and mutagenic oligonucleotides. The PCR fragments contain the coding regions of the mutant U5 derivatives flanked by SalI and BamHI restriction sites. After digestion with SalI and BamHI they were subcloned into pGEM4 U1/U5.
Immunoprecipitations-Nuclear and cytoplasmic extract were prepared from 293 cells 48 h after transfection as described by Lee et al. (1988). For anti-m 3 G immunoprecipitations 10 l of nuclear and 15 l of cytoplasmic extract were mixed, and RNA was isolated by proteinase K digestion and phenol/chloroform extraction. Immunoprecipitations were performed with a polyclonal rabbit anti-m 3 G antiserum prepared according to Bringmann et al. (1983); the expected specificity for m 3 Gcapped snRNAs was demonstrated (data not shown). RNA was isolated from the supernatant and pellet fractions and analyzed by primer extension. Anti-Sm immunoprecipitations were carried out with 30 l of nuclear extract according to Wolff and Bindereif (1993). RNA isolated from 20% of the total reaction and from the pellet was analyzed by primer extension. For anti-p220 immunoprecipitations a rabbit antiserum raised against a His-tagged fusion protein containing the Cterminal 50 kDa of human p220 was used (M. Moore, H. Luo, and G. Moreau, Brandeis University). 10 l of antiserum were incubated with 25 l of protein A-Sepharose beads (Pharmacia Biotech Inc.) in 700 l of PBS buffer (137 mM NaCl, 2.7 mM KCl, 12.2 mM Na 2 HPO 4 , 22 mM KH 2 PO 4 , pH 7.4) for 12 h at 4°C. The precoated beads were washed three times with 1 ml of IP100 buffer (50 mM Tris-HCl, pH 7.6, 100 mM NaCl, 2 mM MgCl 2 , 0.5 mM dithiothreitol, 0.05% Nonidet P-40). 40 l of nuclear extract were incubated for 2 h at 4°C with the beads resuspended in 250 l of IP100 buffer. The beads were washed three times with IP100 buffer, and immunoprecipitated RNAs were recovered by vortexing the beads for 5 min in 200 l of SBD buffer (100 mM Tris-HCl, pH 7.6, 150 mM NaCl, 300 mM NaAc, 10 mM EDTA, 1% SDS) and subsequent phenol/chloroform extraction. In addition, RNA from 8 l of nuclear extract (representing 20% of the total reaction) was prepared. RNA isolated from 20% of the total and from the pellet was analyzed by primer extension.
Affinity Selection of U4/U5/U6 Triple snRNPs-50 l of nuclear extract were preincubated for 15 min at 30°C under in vitro splicing conditions (3.2 mM MgCl 2 , 0.5 mM ATP, 20 mM creatine phosphate, 1.600 units/ml RNasin (Promega), and 40 g/ml tRNA). 20% of the reaction was used for the analysis of total RNA. The remainder was adjusted to 300 mM KCl and incubated with 1 g of biotinylated 2Ј-OMe RNA oligonucleotide U6e (complementary to nucleotides 82-101 of human U6 RNA (Wolff and Bindereif, 1992)). Streptavidin-agarose was preblocked (Barabino et al., 1989) and washed two times at 4°C, each for 10 min with 500 l of WB300 (20 mM HEPES, pH 8.0, 300 mM KCl, 0.5 mM dithiothreitol, 0.01% Nonidet P-40). The oligonucleotide-tagged snRNPs were bound to 25 l of preblocked streptavidin-agarose during a 2-h incubation at 4°C. The streptavidin-agarose beads were pelleted and washed three times with 500 l of WB300 each for 10 min at 4°C. Bound RNA was released by incubation in 200 l of PK buffer (100 mM Tris-HCl, pH 7.5, 12.5 mM EDTA, 150 mM NaCl, 1% SDS, and 0.5 mg/ml proteinase K) for 45 min at 55°C and subsequent phenol/chloroform extraction. Released RNA was analyzed by primer extension.
Affinity Selection of Splicing Complexes-Biotinylated MINX pre-mRNA (Zillmann et al., 1988) was synthesized by in vitro transcription (Wolff and Bindereif, 1992), using 30% of bio-11-UTP (Enzo). To assay the assembly of U5 RNA derivatives into splicing complexes, 30 l of nuclear extract prepared from transfected cells (Lee et al., 1988) was incubated with 150 ng of biotinylated MINX pre-mRNA for 30 min at 30°C under in vitro splicing conditions (3.2 mM MgCl 2 , 0.5 mM ATP, 20 mM creatine phosphate, 1.600 units/ml RNasin (Promega), and 40 g/ml tRNA). 20% of the reaction was used for the analysis of total RNA. The remainder was adjusted to 500 mM KCl and mixed with 1 volume of SB buffer (10 mM HEPES, pH 7.5, 500 mM KCl, 3 mM MgCl 2 , and 1 mM dithiothreitol). Streptavidin-agarose was blocked (Barabino et al., 1989) and washed two times, each for 10 min at 4°C with 500 l of NET-500 buffer (50 mM Tris-HCl, pH 8.0, 500 mM KCl, 0.01% Nonidet P-40). Splicing complexes were bound to 25 l of preblocked streptavidin-agarose beads during a 2-h incubation at 4°C. The streptavidinagarose beads were pelleted and washed three times, with 500 l of NET-500 buffer each for 10 min at 4°C. Bound RNA was released by incubation in 200 l of PK buffer (100 mM Tris-HCl, pH 7.5, 12.5 mM EDTA, 150 mM NaCl, 1% SDS, and 0.5 mg/ml proteinase K) for 45 min at 55°C and phenol/chloroform extraction. Released RNA was analyzed by primer extension.

Expression of Human U5 RNA and Mutant Derivatives in
Mammalian Cells-To express the human U5 RNA in mammalian cells, we constructed expression cassettes on the basis of the human U1 or U2 snRNA genes and inserted the sequence of the human U5 RNA (see "Experimental Procedures"). A 10-nucleotide marker sequence was added at the 5Ј end, which allowed us to distinguish the transfected U5 RNA from the endogenous U5 RNA by primer-extension analysis (for the structure of pGEM4 U1/5, see Fig. 1). The resulting plasmids, pGEM4 U1/5 and pGEM4 U2/U5, were transiently transfected into 293 cells; expression was measured by primer extension, using U5 RNA-specific oligonucleotides. Since after transient transfection the U5 RNA was consistently expressed at least 5-fold higher from the pGEM4 U1/U5 than from the pGEM4 U2/U5 construct (data not shown), all subsequent mutant derivatives were made on the basis of pGEM4 U1/U5. A series of mutant derivatives of pGEM4 U1/U5 was generated, in which individual sequence or structural elements were deleted or substituted, with the aim not to disturb the overall secondary structure ( Fig. 2 and Table I). Most of the mutations affect the extended 5Ј stem-loop of U5 (⌬5Ј loop; ⌬20 -57; ⌬IL1; ⌬IL2; sub-stem Ic; sub-5ЈIL2; sub-stem Ib; sub-3ЈIL1); in addition, the 3Ј stem-loop was deleted (⌬3Ј stem/loop), and a four-nucleotide substitution was introduced into the Sm site (sub-Sm).
Initially the expression of these U5 mutant derivatives during transient transfection in 293 cells was determined ( Fig. 3; summarized in Table II). Since a primer specific for the 3Ј stem-loop proved very inefficient in primer extension, three different primers had to be used to detect the expression of the U5 RNA derivatives, depending on the mutation (see three panels in Fig. 3); the efficiency of detection by primer extension varied considerably between the three primers (compare lanes WT in the three panels of Fig. 3) but was taken into account by simultaneously detecting the endogenous U5 RNA as an internal control. To control for primer specificity, RNA was also analyzed from untransfected cells. To compare transfection efficiencies, cotransfections were done with pE4-CAT. We conclude that the expression of the mutant derivatives ⌬5Ј loop, sub-stem Ic, and sub-stem Ib is similar to wild-type, that is in the order of 25% relative to the endogenous U5 RNA expression. The expression of sub-3ЈIL1 is reduced to ϳ50%; the other mutant U5 RNAs (⌬20 -57, ⌬IL2, sub-5ЈIL2, ⌬IL1, ⌬3Јstem/ loop, and sub-Sm) are expressed at very low levels (less than 10% of wild-type).
To test whether the mutant U5 RNAs possess correct 3Ј ends, Northern blot analysis was carried out (data not shown). Although the mutant U5 RNAs with very low expression could not be analyzed this way, the other U5 derivatives showed the expected length and should therefore carry correct 3Ј ends. The very low expression levels of certain mutant RNAs may be due to a variety of factors, including RNA stability, cap trimethylation, protein binding, localization, or 3Ј end formation. To avoid problems of low in vivo stability, an in vitro reconstitution system such as the one developed recently by Ségault et al. (1995) would provide a useful complementary approach. Although we were able to reconstitute U5 core snRNPs, no functional complementation of U5-depleted splicing extracts could be achieved under our conditions (data not shown).
U5 cap Trimethylation Requires the 3Ј-Terminal Domain-Previous studies have demonstrated for the U2 RNA that cap trimethylation depends on a functional Sm protein-binding site (Mattaj, 1986). Therefore, we tested whether U5 RNA mutations affected cap trimethylation. RNA was prepared from transfected cells and subjected to anti-trimethylguanosine (m 3 G) cap immunoprecipitation. To compare the trimethyl cap formation of the U5 derivatives, both the supernatant and the pellet fractions of each immunoprecipitation reaction were analyzed by primer extension; representative examples are shown in Fig. 4 (for a summary, see Table II). Between 50 and 80% of the endogenous U5 RNA were precipitated; the levels were similar for the transfected wild-type U5 RNA, as well as for the mutant derivatives ⌬5Ј loop, sub-stem Ic, sub-3ЈIL1, ⌬20 -57, ⌬IL2, sub-5ЈIL2, sub-stem Ib, and ⌬IL1 ( Fig. 4 and data not shown). In contrast, the sub-Sm mutant RNA was not detectably immunoprecipitated (Fig. 4, lanes sub Sm), demonstrating that an intact Sm site is essential for U5 cap trimethylation. Finally, mutant ⌬3Ј stem/loop was trimethyl-capped with a reduced efficiency of ϳ50% (lanes ⌬3Ј stem/loop), indicating that the 3Ј stem-loop of human U5 may be an additional sequence element required for efficient cap trimethylation.
In further experiments we found that Sm protein binding correlated with trimethyl capping. All the U5 mutant RNAs that were expressed at sufficient levels to be tested showed normal Sm protein binding (⌬5Ј loop, ⌬20 -57, sub-stem Ic, ⌬IL2, sub-5ЈIL2, sub-stem Ib, sub-3ЈIL1; the levels of the mutant U5 RNAs ⌬IL1 and sub-Sm were too low in this experiment); in contrast, deletion of the 3Ј terminal stem-loop (⌬3Ј stem/loop) resulted in reduced Sm protein binding (data not shown; summarized in Table II).
In sum, an intact Sm protein-binding site is essential for cap trimethylation. An additional important sequence element is the 3Ј stem-loop. No sequence requirements for cap trimethylation could be mapped within the extended 5Ј stem-loop structure.
Binding of the U5 snRNP-specific Protein p220 Depends on Elements of the 5Ј Stem-loop-The U5 snRNP contains at least seven specific protein components, among them the p220 protein, which is highly conserved and has been implicated in several essential functions in the splicing mechanism (for references, see Introduction). To obtain insight into functionally important domains in the U5 snRNP, we have examined the U5 mutant RNAs for their ability to associate with p220, using immunoprecipitation assays. Extracts were prepared from transfected cells; immunoprecipitations were done with an anti-p220 antiserum, and RNA was analyzed from the immunoprecipitates by primer extension (Fig. 5; for a summary of the PhosphorImager analysis, see Table II). The anti-p220 antiserum is a polyclonal rabbit serum specific for the C-terminal 50 kDa of human p220. 2 ; Under these conditions the efficiency of immunoprecipitating the endogenous U5 RNA was between 30 and 40%. In comparison to the endogenous U5 RNA, the immunoprecipitation efficiency of transfected wild-type U5 2 M. Moore, H. Luo, G. Moreau, unpublished data.
FIG. 1. Wild-type human U5 RNA expression construct pGEM4 U1/U5. White boxes represent the 5Ј and 3Ј flanking regions of the human U1 RNA gene; the hatched area represents the coding sequence of the human U5 RNA. The transcriptional start site is indicated by the arrow. The black box represents a marker sequence including the first four nucleotides of the U1 RNA coding sequence and a SalI restriction site. The BamHI restriction site at the end of the U5 coding sequence was introduced by mutation of the U1 3Ј box. The EcoRI/HindIII fragment was cloned into the vector pGEM4 (thin lines).
RNA was reduced to about half, possibly resulting from the additional marker sequence at the 5Ј end of U5 (lanes WT). The strongest effect of the mutations came from deleting the internal loop 2 (⌬IL2) and substituting stem Ib (sub-stem Ib). In comparison to the transfected wild-type U5 RNA, the other mutations reduced p220 binding to intermediate levels (⌬5Ј loop, ⌬20 -57, sub-ЈIL2, sub-Sm) or had no effect (sub-stem Ic, sub-3ЈIL1, ⌬3'stem/loop); due to inefficient expression, the ⌬IL1 mutant could not be analyzed.
This mutational analysis demonstrates that several discrete elements of the 5Ј stem-loop contribute to efficient p220 association. Specifically, stem Ib is essential, and other elements contribute to efficient p220 binding.
Mutational Analysis of U4/U5/U6 Triple snRNP and Spliceosome Assembly-In order to analyze the assembly of U5 mutant RNAs into U4/U5/U6 triple snRNPs, a new assay based upon biotin-streptavidin selection was developed (see "Experimental Procedures"). U4/U5/U6 triple snRNPs were selected in extracts from transfected cells, using a biotinylated 2Ј-OMe RNA oligonucleotide specific for the 3Ј-terminal sequence of U6 RNA (U6-e) and streptavidin-agarose. Because of the instability of the triple-snRNP at high ionic strength (Cheng and Abelson, 1987;Lossky et al., 1987;Wolff and Bindereif, 1992) and because of nonspecific interactions at low ionic strength it was important to carry out the selections at an intermediate salt concentration (300 mM KCl). Under these conditions endogenous U4/U5/U6 triple snRNPs could be affinity-selected with an efficiency of between 10 and 15%; the selection was specific by the following criteria. First, it was dependent on the addition of biotinylated 2Ј-OMe RNA oligonucleotide; second, it was sensitive to heparin treatment (Wolff and Bindereif, 1992); third, if the extract was first fractionated on a glycerol gradient and the individual gradient fractions then used for affinity selections, U5 could be detected only in the 25 S region (where the U4/U5/U6 triple snRNP is concentrated) but not in the 20 S region (where U5 snRNPs fractionate) (data not shown).

FIG. 2. Secondary structure model of the human U5 RNA (A) and schematic representation of the U5 RNA mutant derivatives (B).
A, the Sm binding site of the U5 RNA (underlined) and the marker sequence at the 5Ј end (boxed) are indicated within the secondary structure model (Baserga and Steitz, 1993). B, substituted sequence elements are highlighted by thick lines or the respective sequence. The Sm binding site and the marker sequence are indicated as boxes. All mutant derivatives were cloned into the expression construct pGEM4 U1/U5 (for details, see Table I and "Experimental Procedures").  3. Expression of U5 RNA derivatives in 293 cells. Cells were transfected with wild-type U5 expression construct (WT) or mutant derivatives (as indicated). As an internal control for the transfection efficiency pE4cat was cotransfected. RNA was prepared 48 h after transfection and analyzed by primer extension (for the primers used, see "Experimental Procedures"). In control reactions RNA from untransfected cells was analyzed (control). The mobilities of the primers (P) and the extension products from the endogenous (e-U5) and transfected (t-U5) U5 RNAs are indicated.
Triple snRNP selections were performed with extracts from transfected cells, and the content of endogenous and transfected U5 RNAs was determined by primer-extension assays and PhosphorImager analysis, using 20% of the total reaction (lanes T) and the entire affinity-selected material (lanes P) (Fig. 6; summarized in Table II). A control showed that the transfected wild-type U5 RNA was assembled and selected at a similar efficiency as the endogenous U5 RNA (lanes WT); in the absence of oligonucleotide no U5 RNA could be detected in the selected material (lanes control). The sub-stem Ic derivative did not significantly differ from the endogenous wild-type U5 RNA in its efficiency of triple snRNP assembly; the U5 mutant RNAs ⌬5Ј loop and sub-3ЈIL1, however, were reduced to approximately half of the wild-type level. Due to their low abundance and the relatively low signals in the selected material, the other U5 derivatives could be not analyzed in this manner.
A second affinity selection assay was designed to quantitate the assembly of mutant U5 RNAs into spliceosomes (see "Experimental Procedures"). Splicing-active miniextracts were prepared from transfected cells, an in vitro splicing reaction was carried out with biotinylated MINX pre-mRNA, and pre-mRNA complexes were selected under stringent conditions with streptavidin-agarose. Finally, the content of endogenous U5 RNA and transfected mutant U5 RNAs in the spliceosomes was analyzed by primer extension and PhosphorImager analysis, using 20% of the total reaction (lanes T) and the affinityselected material (lanes P) (Fig. 7). Initial controls ensured that spliceosome selection required the addition of ATP (data not shown) and biotinylated pre-mRNA (lane control). Depending on the splicing extract, between 3 and 15% of the total U5 RNA present in the extract could be selected in the form of spliceosomes.
Wild-type U5 RNA and the mutant derivative sub-stem Ic were found in splicing complexes at similar levels as the endogenous U5 RNA. In contrast, mutants ⌬5Ј loop and sub-3ЈIL1 were significantly reduced in their spliceosome assembly ( Fig.  7 and data not shown; for a summary of the PhosphorImager analysis, see Table II). Most likely this is due to the effects of these mutations on U4/U5/U6 triple snRNP assembly (see Fig.  6). Unfortunately, the other U5 mutant RNAs were not expressed at levels sufficient for this type of analysis. DISCUSSION We have used transient expression assays and a mutational analysis of the human U5 RNA to study the domain structure of this spliceosomal RNA, which has been implicated in several functions of the splicing mechanism. Several structural elements of human U5 RNA were found to be necessary for efficient expression: both internal loops (IL1, IL2), the Sm protein-binding site, and the 3Ј stem-loop. Our mutational analysis demonstrated that these elements have important functions during U5 snRNP biogenesis and assembly into the spliceosome.
The three major conclusions of our study are summarized and discussed in the following. 1) Several structural elements of the extended 5Ј stem-loop of U5 RNA, in particular stem Ib, contribute to efficient p220 binding. 2) The 5Ј loop partici- The following characteristics of all U5 snRNA mutant derivatives are listed: expression level, m 3 G cap formation, Sm protein binding, p220 protein binding, U4/U5/U6 triple-snRNP, and spliceosome assembly. The expression levels are given relative to the wild-type construct pGEM4 U1/U5 (WT). The efficiency of m 3 G cap formation, Sm, and p220 protein binding as well as U4/U5/U6 triple snRNP and spliceosome assembly of mutant derivatives are compared with those of endogenous U5 snRNA and indicated as ϩϩ (60 -100%), ϩ (10 -60%), ϩ/Ϫ (Ͻ10%, but above background level) and Ϫ(background level). p220 protein binding, U4/U5/U6 triple snRNP and spliceosome assembly were quantitated by PhosphorImager analysis. ND, not determined. pates not only in splice site selection, as earlier studies have established, but is also necessary for efficient p220 binding and triple snRNP assembly. 3) Both the Sm site and the 3Ј stem-loop are important for Sm protein binding and cap trimethylation.
Earlier chemical modification and protection studies had shown that U5 snRNP proteins are bound within the extended 5Ј stem-loop of the U5 RNA (Bach and Lü hrmann, 1991;Black and Pinto, 1989). We have used anti-p220 immunoprecipitation assays to analyze a major U5 snRNP component, the p220 protein. We note that our approach cannot distinguish between a direct U5 RNA/p220 contact and an indirect, protein-mediated association. Interestingly, efficient p220 association with the U5 snRNP requires several elements of the 5Ј stem-loop. Considering the complexity of the U5 snRNP protein composition, it is perhaps not so surprising that, with the exception of sub-stem Ib, mutations of single sequence elements result only in moderate effects on triple snRNP assembly. This likely reflects the involvement of multiple protein-protein interaction in U5 snRNP assembly. Our finding that the stem Ib is essential FIG. 5. Analysis of p220 binding of U5 RNA derivatives by anti-p220 immunoprecipitation. RNA prepared from transfected cells was immunoprecipitated and analyzed by primer extension (T, 20% of the total reaction; P, immunoprecipitated pellet fraction). In control reactions RNA from cells transfected with the wild-type construct (WT) was used, or the immunoprecipitation was done with nonimmune human serum (control). Indicated are the positions of the primer (P) and the extension products from the endogenous (e-U5) and transfected (t-U5) U5 RNAs. On the panels to the right, two different exposures are shown in order to visualize the extension products from both the endogenous (e-U5) and transfected (t-U5) U5 RNAs.
FIG. 6. U4/U5/U6 triple-snRNP formation of U5 RNA derivatives. Nuclear extract prepared from transfected cells was analyzed by affinity selection, using a biotinylated anti-U6 2Ј-OMe RNA oligonucleotide and streptavidin-agarose. RNA prepared from 20% of the total reaction (lanes T) and the affinity-selected RNA (lanes P) were analyzed by primer extension. In control reactions, the affinity selection was carried out in the absence of 2Ј-OMe RNA oligonucleotide (control) or with extract from cells transfected with the wild-type U1/U5 construct (WT). The positions of the primer (P) and the extension products from the endogenous (e-U5) and transfected (t-U5) U5 RNAs are indicated.
FIG. 7. Spliceosome assembly of U5 RNA derivatives. Nuclear extract was prepared from transfected cells and incubated for 30 min with biotinylated MINX pre-mRNA under in vitro splicing conditions. Splicing complexes were selected with streptavidin-agarose. RNA prepared from 20% of the total reaction (lanes T) and the affinity-selected RNA (lanes P) were analyzed by primer extension. In control reactions, the affinity selection was carried out with unbiotinylated pre-mRNA (control) or with extract from cells transfected with the wild-type U1/U5 construct (WT). The positions of the primer (P) and the extension products from the endogenous (e-U5) and transfected (t-U5) U5 RNAs are indicated. for p220 binding points to a potential difference between the sequence requirements of the mammalian and yeast U5 snRNP assembly (compare with Frank et al., 1994). Surprisingly, the extensive deletion in the mutant RNA U5 ⌬20 -57 still allowed p220 association, although at a reduced level, whereas deleting only the internal loop 2 abolished p220 binding. Several explanations of this apparent discrepancy are possible, for example incorrect folding of mutant RNA ⌬IL2.
Interestingly, substituting the 5Ј loop had a significant effect on p220 binding, suggesting a protein binding function of this highly conserved U5 element in addition to its role in splice-site selection. This result is consistent with an earlier study of Bach and Lü hrmann (1991) who demonstrated that in the 20 S U5 snRNP the 5Ј loop is less accessible to enzymatic digestion than in the U5 core particle. On the other hand, Black and Pinto (1989) have shown that the 5Ј loop is accessible to chemical modifications in the U4/U5/U6 triple snRNP. Therefore the 5Ј loop of U5 RNA may exist in different conformations, whether U5 occurs as free RNA, in the U5 snRNP, or in the U4/U5/U6 triple snRNP.
Both the internal loop 1, at least its 3Ј part, and the 5Ј loop are also necessary for efficient U4/U5/U6 triple snRNP and spliceosome assembly. This probably reflects the requirement of binding of the human p220 protein for triple snRNP and spliceosome assembly, as earlier demonstrated for the yeast homologue, PRP8 (1992). The effects of two mutations, ⌬5Ј loop and sub-3ЈIL1, on splicing complex formation and U4/U5/U6 triple snRNP assembly correlate, suggesting that a correctly assembled U4/U5/U6 triple snRNP is a prerequisite for spliceosome formation.
Finally, our mutational analysis showed that the Sm binding site and the 3Ј stem-loop are important elements for cap trimethylation. As first demonstrated for the U2 RNA (Mattaj, 1986), binding of the Sm proteins is a prerequisite for cap trimethylation. Our data show that this conclusion holds also in the case of the human U5 RNA. A difference appears to exist between the mammalian and yeast U5 snRNP assembly in that the human U5 RNA requires, in addition to the Sm site, the 3Ј stem-loop for efficient Sm protein binding (this study), whereas in yeast the 3Ј stem-loop is dispensable (Frank et al., 1994).
In sum, this mutational analysis defined four functional domains of human U5 RNA. First, the 3Ј domain, including the Sm binding site and the 3Ј stem-loop, is necessary for Sm protein binding and cap trimethylation. Second, IL1 is involved in Sm protein binding (Jarmolowski and Mattaj, 1993) and U4/U5/U6 triple snRNP assembly (this study). Third, stem Ib is essential for the binding of specific proteins (p220) and appears to function in combination with other structural elements of the extended 5Ј stem-loop of U5 RNA. Finally, the 5Ј loop is a multifunctional element, involved not only in splice-site selection, but also in p220 binding.