An ordered pathway of snRNP binding during mammalian pre-mRNA splicing complex assembly.

We have studied the assembly, composition and structure of splicing complexes using biotin-avidin affinity chromatography and RNase protection assays. We find that U1, U2, U4, U5 and U6 snRNPs associate with the pre-mRNA and are in the mature, functional complex. Association of U1 snRNP with the pre-mRNA is rapid and ATP independent; binding of all other snRNPs occurs subsequently and is ATP dependent. Efficient binding of U1 and U2 snRNPs requires a 5' splice site or a 3' splice site/branch point region, respectively. Both sequence elements are required for efficient U4, U5 and U6 snRNP binding. Mutant RNA substrates containing only a 5' splice site or a 3' splice site/branch point region are assembled into 'partial' splicing complexes, which contain a subset of these five snRNPs. RNase protection experiments indicate that in contrast to U1 and U2 snRNPs, U4, U5 and U6 snRNPs do not contact the pre-mRNA. Based upon the time course of snRNP binding and the composition of sucrose gradient fractionated splicing complexes we suggest an assembly pathway proceeding from a 20S (U1 snRNP only) through a 40S (U1 and U2 snRNPs) to the functional 60S splicing complex (U1, U2, U4, U5 and U6 snRNPs).     

Assembly of Snu114 into U5 snRNP requires Prp8 and a functional GTPase domain

Snu114 is a U5 snRNP protein essential for pre-mRNA splicing. Based on its homology with the ribosomal translocase EF-G, it is thought that GTP hydrolysis by Snu114 induces conformational rearrangements in the spliceosome. We recently identified allele-specific genetic interactions between SNU114 and genes encoding three other U5 snRNP components, Prp8 and two RNA-dependent ATPases, Prp28 and Brr2, required for destabilization of U1 and U4 snRNPs prior to catalysis. To shed more light onto the function of Snu114, we have now directly analyzed snRNP and spliceosome assembly in SNU114 mutant extracts. The Snu114-60 C-terminal truncation mutant, which is synthetically lethal with the ATPase mutants prp28-1 and brr2-1, assembles spliceosomes but subsequently blocks U4 snRNP release. Conversely, mutants in the GTPase domain fail to assemble U5 snRNPs. These mutations prevent the interaction of Snu114 with Prp8 as well as with U5 snRNA. Since Prp8 is thought to regulate the activity of the DEAD-box ATPases, this strategy of snRNP assembly could ensure that Prp8 activity is itself regulated by a GTP-dependent mechanism.     

Structure-probing of U1 snRNPs gradually depleted of the U1-specific proteins A, C and 70k. Evidence that A interacts differentially with developmentally regulated mouse U1 snRNA variants.

The interaction of the U1-specific proteins 70k, A and C with U1 snRNP was studied by depleting gradually U1 snRNPs of the U1-specific proteins by Mono-Q chromatography at elevated temperatures (20-37 degrees C). U1 snRNP species were obtained which were selectively depleted of either protein C, A, C and A, or of all three U1-specific proteins C, A and 70k while retaining the common proteins B' to G. These various types of U1 snRNP particles were used to study the differential accessibility of defined regions of U1 RNA towards nucleases V1 and S1 dependent on the U1 snRNP protein composition. The data indicate that in the U1 snRNP protein 70k interacts with stem/loop A and protein A with stem/loop B of U1 RNA. The presence or absence of protein C did not affect the nuclease digestion patterns of U1 RNA. Our results suggest further that the binding of protein A to the U1 snRNP particle should be independent of proteins 70k and C. Mouse cells contain two U1 RNA species, U1a and U1b, which differ in the structure of stem/loop B, with U1a exhibiting the same stem/loop B sequence as U1 RNA from HeLa cells. When we used Mono Q chromatography to investigate possible structural differences in the two types of U1 snRNPs, we observed that protein A was always preferentially lost from U1b snRNP as compared to U1a snRNPs. This indicates that one consequence of the structural difference between U1a and U1b is a lowering of the strength of binding of protein A to U1b snRNP. The possible functional significance of this finding is discussed with respect to the fact that U1b RNA is preferentially expressed in embryonal cells.     

Small nuclear ribonucleoprotein (RNP) U2 contains numerous additional proteins and has a bipartite RNP structure under splicing conditions.

Small nuclear (sn) ribonucleoprotein (RNP) U2 functions in the splicing of mRNA by recognizing the branch site of the unspliced pre-mRNA. When HeLa nuclear splicing extracts are centrifuged on glycerol gradients, U2 snRNPs sediment at either 12S (under high salt concentration conditions) or 17S (under low salt concentration conditions). We isolated the 17S U2 snRNPs from splicing extracts under nondenaturing conditions by using centrifugation and immunoaffinity chromatography and examined their structure by electron microscope. In addition to common proteins B', B, D1, D2, D3, E, F, and G and U2-specific proteins A' and B", which are present in the 12S U2 snRNP, at least nine previously unidentified proteins with apparent molecular masses of 35, 53, 60, 66, 92, 110, 120, 150, and 160 kDa bound to the 17S U2 snRNP. The latter proteins dissociate from the U2 snRNP at salt concentrations above 200 mM, yielding the 12S U2 snRNP particle. Under the electron microscope, the 17S U2 snRNPs exhibited a bipartite appearance, with two main globular domains connected by a short filamentous structure that is sensitive to RNase. These findings suggest that the additional globular domain, which is absent from 12S U2 snRNPs, contains some of the 17S U2-specific proteins. The 5' end of the RNA in the U2 snRNP is more exposed for reaction with RNase H and with chemical probes when the U2 snRNP is in the 17S form than when it is in the 12S form. Removal of the 5' end of this RNA reduces the snRNP's Svedberg value from 17S to 12S. Along with the peculiar morphology of the 17S snRNP, these data indicate that most of the 17S U2-specific proteins are bound to the 5' half of the U2 snRNA.     

U5 small nuclear ribonucleoprotein: RNA structure analysis and ATP-dependent interaction with U4/U6.

To understand how the U5 small nuclear ribonucleoprotein (snRNP) interacts with other spliceosome components, its structure and binding to the U4/U6 snRNP were analyzed. The interaction of the U5 snRNP with the U4/U6 snRNP was studied by separating the snRNPs in HeLa cell nuclear extracts on glycerol gradients. A complex running at 25S and containing U4, U5, and U6 but not U1 or U2 snRNAs was identified. In contrast to results with native gel electrophoresis to separate snRNPs, this U4/U5/U6 snRNP complex requires ATP to assemble from the individual snRNPs. The structure of the U5 RNA within the U5 snRNP and the U4/5/6 snRNP complexes was then compared. Oligonucleotide-targeted RNase H digestion identified one RNA sequence in the U5 snRNP capable of base pairing to other nucleic acid sequences. Chemical modification experiments identified this sequence as well as two other U5 RNA sequences as accessible to modification within the U5 RNP. One of these regions is a large loop in the U5 RNA secondary structure whose sequence is conserved from Saccharomyces cerevisiae to humans. Interestingly, no differences in modification of free U5 snRNP as compared to U5 in the U4/U5/U6 snRNP complex were observed, suggesting that recognition of specific RNA sequences in the U5 snRNP is not required for U4/U5/U6 snRNP assembly.     

Evidence for three distinct D proteins, which react differentially with anti-Sm autoantibodies, in the cores of the major snRNPs U1, U2, U4/U6 and U5.

Electrophoresis of the mixture of proteins from purified snRNPs U1, U2, U4/U6 and U5 on SDS-polyacrylamide gels that had been allowed to polymerise in the presence of high TEMED concentrations have revealed the presence of proteins in the snRNPs that previously had eluded detection. The most striking case is that of protein D, heretofore generally observed as a single broad band; in high-TEMED gels, this splits into three clearly-separated bands, identified as three distinct proteins. We have denoted these proteins D1 (16 kDa), D2 (16.5 kDa) and D3 (18 kDa). Chemical and immunological studies have shown that D1 is identical with the common snRNP protein D, whose structure was recently resolved by cDNA cloning (Rokeach et al. (1988), Proc. Natl. Acad. Sci. USA, 85, 4832-4836) and that D2 and D3 are clearly distinct from D1 and very probably from each other. In addition to D1, proteins D2 and D3 are present in purified U1, U2, U4/U6 and U5 snRNPs isolated from HeLa cells, so these also belong to the group of common snRNP proteins. They are also found in snRNPs isolated from mouse cells, indicating that the role of these proteins in the structure and/or function of UsnRNPs has been conserved in evolution. Interestingly, patients with systemic lupus erythematosus produce populations of anti-Sm autoantibodies that react differentially with the D proteins; some recognise all of them and others only a subset. The high-TEMED gels allow improved resolution not only of the D proteins, but also of some of the U5-specific proteins contained in 20S U5 snRNPs, in particular the 15-kDa protein. In addition, under these conditions, the common G protein, previously observed as a single band, appears as a doublet. Whether the additional band represents a distinct common snRNP protein or a post-translationally modified version of G is not yet known.     

RH70, a bidirectional RNA helicase,co-purifies with U1snRNP

In this report, I describe the co-purification of a novel 70-kDa RNA helicase (RH70) and U1snRNP through six column steps. Peptide sequence analysis by mass spectrometry and Edman degradation revealed that RH70 is the previously reported DDX17. Biochemical characterization of RH70, obtained by partial separation from U1snRNP, yielded the following results. (a) RH70 mediates the unwinding of duplex RNA but not DNA in an ATP-dependent manner. (b) Both the RNA-dependent ATPase and RNA helicase activities of RH70 are highly specific for ATP, exhibiting an apparent K(m) of 0.5 mm. (c) RH70 catalyzes the unwinding of duplex RNA containing single-stranded regions at either the 5'- or 3'-end. Its association with U1snRNP and ATP specificity suggest a role for RH70 in pre-mRNA splicing, in particular, at the early stages of the splicing reaction involving U1snRNP.     

The HeLa 200 kDa U5 snRNP-specific protein and its homologue in Saccharomyces cerevisiae are members of the DEXH-box protein family of putative RNA helicases.

The primary structure of the 200 kDa protein of purified HeLa U5 snRNPs (U5-200kD) was characterized by cloning and sequencing of its cDNA. In order to confirm that U5-200kD is distinct from U5-220kD we demonstrate by protein sequencing that the human U5-specific 220 kDa protein is homologous to the yeast U5-specific protein Prp8p. A 246 kDa protein (Snu246p) homologous to U5-200kD was identified in Saccharomyces cerevisiae. Both proteins contain two conserved domains characteristic of the DEXH-box protein family of putative RNA helicases and RNA-stimulated ATPases. Antibodies raised against fusion proteins produced from fragments of the cloned mammalian cDNA interact specifically with the HeLa U5-200kD protein on Western blots and co-immunoprecipitate U5 snRNA and to a lesser extent U4 and U6 snRNAs from HeLa snRNPs. Similarly, U4, U5 and U6 snRNAs can be co-immunoprecipitated from yeast splicing extracts containing an HA-tagged derivative of Snu246p with HA-tag specific antibodies. U5-200kD and Snu246p are thus the first putative RNA helicases shown to be intrinsic components of snRNPs. Disruption of the SNU246 gene in yeast is lethal and leads to a splicing defect in vivo, indicating that the protein is essential for splicing. Anti-U5-200kD antibodies specifically block the second step of mammalian splicing in vitro, demonstrating for the first time that a DEXH-box protein is involved in mammalian splicing. We propose that U5-200kD and Snu246p promote one or more conformational changes in the dynamic network of RNA-RNA interactions in the spliceosome.     

The association of the U1-specific 70K and C proteins with U1 snRNPs is mediated in part by common U snRNP proteins.

The U1 small nuclear ribonucleoprotein particle (snRNP)-specific 70K and A proteins are known to bind directly to stem-loops of the U1 snRNA, whereas the U1-C protein does not bind to naked U1 snRNA, but depends on other U1 snRNP protein components for its association. Focusing on the U1-70K and U1-C proteins, protein-protein interactions contributing to the association of these particle-specific proteins with the U1 snRNP were studied. Immunoprecipitation of complexes formed after incubation of naked U1 snRNA or purified U1 snRNPs lacking their specific proteins (core U1 snRNP) with in vitro translated U1-C protein, revealed that both common snRNP proteins and the U1-70K protein are required for the association of U1-C with the U1 snRNP. Binding studies with various in vitro translated U1-70K mutants demonstrated that the U1-70K N-terminal domain is necessary and sufficient for the interaction of U1-C with core U1 snRNPs. Surprisingly, several N-terminal fragments of the U1-70K protein, which lacked the U1-70K RNP-80 motif and did not bind naked U1 RNA, associated stably with core U1 snRNPs. This suggests that a new U1-70K binding site is generated upon association of common U1 snRNP proteins with U1 RNA. The interaction between the N-terminal domain of U1-70K and the core RNP domain was specific for the U1 snRNP; stable binding was not observed with core U2 or U5 snRNPs, suggesting essential structural differences among snRNP core domains. Evidence for direct protein-protein interactions between U1-specific proteins and common snRNP proteins was supported by chemical crosslinking experiments using purified U1 snRNPs. Individual crosslinks between the U1-70K and the common D2 or B'/B protein, as well as between U1-C and B'/B, were detected. A model for the assembly of U1 snRNP is presented in which the complex of common proteins on the RNA backbone functions as a platform for the association of the U1-specific proteins.     

A split luciferase-based reporter for detection of a cellular macromolecular complex

The spliceosome is a highly dynamic macromolecular ribonucleoprotein (RNP) machine that catalyzes pre-mRNA splicing by assembling U1, U2, U4, U5, and U6 small nuclear RNPs (snRNPs). To process large numbers of introns with a limited number of snRNPs, synthesis and recycling of snRNPs must be maintained within an appropriate range to avoid their shortage. However, the mechanism that maintains cellular snRNP levels is unknown. Molecules that modulate cellular snRNP levels may help to define this mechanism but are not available. Therefore, the goal of the current study was to develop a reporter for snRNP levels using split luciferase based on proteomic analysis of snRNPs. We constructed an expression library of a luciferase fragment fused to core components of U5 snRNP and used it to isolate pre-mRNA processing factor 6 (PRPF6) and small nuclear ribonucleoprotein 40 kDa (U5-40K) that specifically reconstitute luciferase activity in the U5 snRNP complex. Here we show that this reporter detects the effects of small molecules on the levels of the U5 snRNP reporter protein complex. Our approach provides an alternative assay to discover small molecules targeting a macromolecular complex when the structure of the complex is not precisely identified.     

The human U4/U6 snRNP contains 60 and 90kD proteins that are structurally homologous to the yeast splicing factors Prp4p and Prp3p.

Immunoaffinity-purified human 25S [U4/U6.U5] tri-snRNPs harbor a set of polypeptides, termed the tri-snRNP proteins, that are not present in Mono Q-purified 20S U5 snRNPs or 10S U4/U6 snRNPs and that are important for tri-snRNP complex formation (Behrens SE, Lührmann R, 1991, Genes & Dev 5:1439-1452). Biochemical and immunological characterization of HeLa [U4/U6.U5] tri-snRNPs led to the identification of two novel proteins with molecular weights of 61 and 63kD that are distinct from the previously described 15.5, 20, 27, 60, and 90kD tri-snRNP proteins. For the initial characterization of tri-snRNP proteins that interact directly with U4/U6 snRNPs, immunoaffinity chromatography with an antibody directed against the 60kD protein was performed. We demonstrate that the 60 and 90kD tri-snRNP proteins specifically associate with the U4/U6 snRNP at salt concentrations where the tri-snRNP complex has dissociated. The primary structures of the 60kD and 90kD proteins were determined by cloning and sequencing their respective cDNAs. The U4/U6-60kD protein possesses a C-terminal WD domain that contains seven WD repeats and thus belongs to the WD-protein family, whose best-characterized members include the Gbeta subunits of heterotrimeric G proteins. A database homology search revealed a significant degree of overall homology (57.8% similarity, 33.9% identity) between the human 60kD protein and the Saccharomyces cerevisiae U4/U6 snRNP protein Prp4p. Two additional, previously undetected WD repeats (with seven in total) were also identified in Prp4p, consistent with the possibility that 60kD/Prp4p, like beta-transducin, may adopt a propeller-like structure. The U4/U6-90kD protein was shown to exhibit significant homology, particularly in its C-terminal half, with the S. cerevisiae splicing factor Prp3p, which also associates with the yeast U4/U6 snRNP. Interestingly, U4/U6-90kD shares short regions of homology with E. coli RNase III, including a region encompassing its double-stranded RNA binding domain. Based on their structural similarity with essential splicing factors in yeast, the human U4/U6-60kD and 90kD proteins are likely also to play important roles in the mammalian splicing process.     

Evidence from complementation assays in vitro that U5 snRNP is required for both steps of mRNA splicing.

We have established an in vitro complementation system that has allowed us to investigate the role of individual purified snRNPs in the splicing of pre-mRNA molecules. For the preparation of snRNP-depleted nuclear extracts we have first removed the majority of endogenous snRNPs from the nuclear extracts by one passage over an anti-m3G column and then degraded the remaining snRNPs with micrococcal nuclease. The mixture of snRNPs U1, U2, U4/U6 and U5, obtained by anti-m3G immuno-affinity chromatography, was functionally active and able to restore the splicing of snRNP-depleted nuclear extracts. Mono-Q chromatography was used for further fractionation of the snRNPs U1-U6. This produced three fractions that were highly enriched in snRNPs U1 and U2, U5 and U4/U6 respectively. Conditions were found where addition of the [U1, U2] and the U4/U6 snRNP fractions to the snRNP-depleted nuclear extracts gave rise to the formation of splice intermediates in the absence of any 3' cleavage/exon 1-exon 2 product formation. Only when purified 20S U5 snRNPs were added did both steps of the splicing reaction occur efficiently. Our data suggest that U5 snRNP is absolutely required for the second step of splicing and is needed further for efficient initiation of the splicing reaction. The requirement for U5 snRNPs for splicing was corroborated by glycerol gradient sedimentation analysis of the respective reconstituted pre-mRNP complexes. Stable and efficient formation of 50-60S spliceosomes was observed only in the presence of all snRNPs.     

Biochemical and genetic analyses of the U5, U6, and U4/U6 x U5 small nuclear ribonucleoproteins from Saccharomyces cerevisiae.

We have purified the yeast U5 and U6 pre-mRNA splicing small nuclear ribonucleoproteins (snRNPs) by affinity chromatography and analyzed the associated polypeptides by mass spectrometry. The yeast U5 snRNP is composed of the two variants of U5 snRNA, six U5-specific proteins and the 7 proteins of the canonical Sm core. The U6 snRNP is composed of the U6 snRNA, Prp24, and the 7 Sm-Like (LSM) proteins. Surprisingly, the yeast DEAD-box helicase-like protein Prp28 is stably associated with the U5 snRNP, yet is absent from the purified U4/U6 x U5 snRNP. A novel yeast U5 and four novel yeast U4/U6 x U5 snRNP polypeptides were characterized by genetic and biochemical means to demonstrate their involvement in the pre-mRNA splicing reaction. We also show that, unlike the human tri-snRNP, the yeast tri-snRNP dissociated upon addition of ATP or dATP.     

Evidence for nuclear factors involved in recognition of 5'' splice sites.

To investigate soluble factors involved in pre-messenger RNA splicing we have fractionated nuclear extract by simple centrifugation to produce a supernatant pellet pair. Factors larger than 15S including U2, U4, U5, and U6 snRNPs fractionate with the pellet; U1 snRNPs distribute equally in pellet and supernatant. Each fraction is individually incompetent for splicing and spliceosome assembly; mixing restores wild type activity and assembly. The pellet fraction directs an aberrant assembly pathway in which proper 3', but improper 5' splice site recognition occurs. Complexes formed with the pellet fraction are distinguishable from wild-type complexes using native gel electrophoresis. Pellet complexes contain U1 snRNP antigens and their formation requires ATP, U1 snRNPs, U2 snRNPs, and sequences at the 3' end of the intron - properties shared with the initial steps of normal assembly and directed by sequences at the 3' end of the intron. In contrast, pellet complex assembly shows no dependence on the presence of a 5' splice junction within precursor RNA. Furthermore, binding of factors to the 5' splice junction is deficient in pellet assemblies. Thus, the pellet lacks a factor required for proper recognition of 5' splice sites. This factor can be supplied by the supernatant. Complementation occurs when supernatant U1 RNA is destroyed, suggesting that the supernatant factor recognizing 5' splice sites is not U1 snRNPs.     

Prp5 bridges U1 and U2 snRNPs and enables stable U2 snRNP association with intron RNA

Communication between U1 and U2 snRNPs is critical during pre-spliceosome assembly; yet, direct connections have not been observed. To investigate this assembly step, we focused on Prp5, an RNA-dependent ATPase of the DExD/H family. We identified homologs of Saccharomyces cerevisiae Prp5 in humans (hPrp5) and Schizosaccharomyces pombe (SpPrp5), and investigated their interactions and function. Depletion and reconstitution of SpPrp5 from extracts demonstrate that ATP binding and hydrolysis by Prp5 are required for pre-spliceosome complex A formation. hPrp5 and SpPrp5 are each physically associated with both U1 and U2 snRNPs; Prp5 contains distinct U1- and U2-interacting domains that are required for pre-spliceosome assembly; and, we observe a Prp5-associated U1/U2 complex in S. pombe. Together, these data are consistent with Prp5 being a bridge between U1 and U2 snRNPs at the time of pre-spliceosome formation.     

Reconstituted mammalian U4/U6 snRNP complements splicing: a mutational analysis.

We have developed an in vitro complementation assay to analyse the functions of U6 small nuclear RNA (snRNA) in splicing and in the assembly of small nuclear ribonucleoproteins (snRNPs) and spliceosomes. U6-specific, biotinylated 2'-OMe RNA oligonucleotides were used to deplete nuclear extract of the U4/U6 snRNP and to affinity purify functional U4 snRNP. The addition of affinity purified U4 snRNP together with U6 RNA efficiently restored splicing activity, spliceosome assembly and U4/U5/U6 multi-snRNP formation in the U4/U6-depleted extract. Through a mutational analysis we have obtained evidence for multiple sequence elements of U6 RNA functioning during U4/U5/U6 multi-snRNP formation, spliceosome assembly and splicing. Surprisingly, the entire 5' terminal domain of U6 RNA is dispensable for splicing function. In contrast, two regions in the central and 3' terminal domain are required for the assembly of a functional U4/U5/U6 multi-snRNP. Another sequence in the 3' terminal domain plays an essential role in spliceosome assembly; a model is strongly supported whereby base pairing between this sequence and U2 RNA plays an important role during assembly of a functional spliceosome.     

A general approach for identification of RNA-protein cross-linking sites within native human spliceosomal small nuclear ribonucleoproteins (snRNPs). Analysis of RNA-protein contacts in native U1 and U4/U6.U5 snRNPs

We describe a novel approach to identify RNA-protein cross-linking sites within native small nuclear ribonucleoprotein (snRNP) particles from HeLa cells. It combines immunoprecipitation of the UV-irradiated particles under semi-denaturing conditions with primer extension analysis of the cross-linked RNA moiety. In a feasibility study, we initially identified the exact cross-linking sites of the U1 70-kDa (70K) protein in stem-loop I of U1 small nuclear RNA (snRNA) within purified U1 snRNPs and then confirmed the results by a large-scale preparation that allowed N-terminal sequencing and matrix-assisted laser desorption ionization mass spectrometry of purified cross-linked peptide-oligonucleotide complexes. We identified Tyr(112) and Leu(175) within the RNA-binding domain of the U1 70K protein to be cross-linked to G(28) and U(30) in stem-loop I, respectively. We further applied our immunoprecipitation approach to HeLa U5 snRNP, as part of purified 25 S U4/U6.U5 tri-snRNPs. Cross-linking sites between the U5-specific 220-kDa protein (human homologue of Prp8p) and the U5 snRNA were located at multiple nucleotides within the highly conserved loop 1 and at one site in internal loop 1 of U5 snRNA. The cross-linking of four adjacent nucleotides indicates an extended interaction surface between loop 1 and the 220-kDa protein. In summary, our approach provides a rapid method for identification of RNA-protein contact sites within native snRNP particles as well as other ribonucleoprotein particles.     

Protein-RNA interactions in 20S U5 snRNPs.

The interaction of the U5-specific polypeptides with U5 snRNA was investigated by comparison of the differential accessibility towards nucleases and dimethylsulfate of defined regions of U5 snRNA in purified 20S and 10S U5 snRNPs. While 20S U5 snRNPs contain eight U5-specific proteins in addition to the common proteins, the 10S U5 snRNPs contain only the latter proteins. The results indicate that only the central part of stem/loop I of U5 snRNA including internal loops IL2 and IL2', contains binding sites for U5-specific proteins, suggesting that several U5-specific proteins may be bound to U5 snRNP via protein-protein interactions. Moreover, they show that the core polypeptides do not interact with stem/loop I.