Structure and function of an RNase H domain at the heart of the spliceosome

Abstract

Precursor-messenger RNA (pre-mRNA) splicing encompasses two sequential transesterification reactions in distinct active sites of the spliceosome that are transiently established by the interplay of small nuclear (sn) RNAs and spliceosomal proteins. Protein Prp8 is an active site component but the molecular mechanisms, by which it might facilitate splicing catalysis, are unknown. We have determined crystal structures of corresponding portions of yeast and human Prp8 that interact with functional regions of the pre-mRNA, revealing a phylogenetically conserved RNase H fold, augmented by Prp8-specific elements. Comparisons to RNase H–substrate complexes suggested how an RNA encompassing a 5′-splice site (SS) could bind relative to Prp8 residues, which on mutation, suppress splice defects in pre-mRNAs and snRNAs. A truncated RNase H-like active centre lies next to a known contact region of the 5′SS and directed mutagenesis confirmed that this centre is a functional hotspot. These data suggest that Prp8 employs an RNase H domain to help assemble and stabilize the spliceosomal catalytic core, coordinate the activities of other splicing factors and possibly participate in chemical catalysis of splicing.

Introduction

Nuclear precursor-messenger RNA (pre-mRNA) splicing constitutes an essential step in the maturation of most eukaryotic primary transcripts, during which non-coding intervening sequences (introns) are removed and neighbouring coding regions (exons) are concatenated. Each round of splicing proceeds through two sequential transesterification reactions that are facilitated by the spliceosome, a dynamic molecular machine composed of five small nuclear RNAs (snRNAs) and more than 120 proteins (Will and Lührmann, 2006). In the first step, the 2′-hydroxyl of an invariant adenosine in the intronic branch point sequence (BPS) attacks the 5′SS (splice site), liberating the 5′-exon and an intron lariat-3′-exon. In the second step, the 3′-hydroxyl of the 5′-exon carries out a nucleophilic attack on the 3′SS, leading to exon ligation and excision of the branched intron (Moore et al, 1993).

Spliceosomes form only in the presence of a pre-mRNA substrate, on which they initially assemble as inactive particles that subsequently undergo catalytic activation (Brow, 2002). Spliceosome assembly and maturation are accompanied by profound changes in the snRNA–pre-mRNA network (Nilsen, 1998; Staley and Guthrie, 1998) and by the dynamic exchange of many proteins (Will and Lührmann, 2006; Bessonov et al, 2008). They proceed through landmark intermediates that are defined by the sequential incorporation and release of spliceosomal sn ribonucleoprotein particles (snRNPs; U1, U2, U4/U6 and U5). Initially, U1 and U2 snRNPs recognize, through base pairing, the 5′SS and BPS, respectively, forming complex A, which is then joined by the U4/U6–U5 tri-snRNP, generating complex B. Subsequently, the U1 and U4 snRNPs are expelled and the Prp19/Cdc5 multi-protein complex (the nineteen complex, NTC, in yeast) is stably integrated, generating the activated spliceosome (complex B^*). During activation, the U4/U6 duplex is unwound, U6 snRNA base pairs with U2 and forms a functionally important intramolecular stem loop (ISL) and the 5′SS is handed over from the 5′-end of U1 to a conserved ACAGAGA-box of U6.

Several lines of evidence suggest that the two transesterification reactions are chemically facilitated by the spliceosomal RNA components through a two metal ion mechanism (Steitz and Steitz, 1993; Sontheimer et al, 1997). Phosphorothioate suppression experiments (Yean et al, 2000) and structural studies (Huppler et al, 2002) had demonstrated that the U6 ISL positions a catalytically or structurally important metal ion. Furthermore, fragments of the U2 and U6 snRNAs alone can mediate a reaction that resembles the first step of splicing (Valadkhan and Manley, 2001). However, spliceosomal proteins are indispensable for the assembly and stabilization of a fully active snRNA–pre-mRNA network; for example, some 35 proteins are integral components of a salt-stable step-one spliceosomal core (Bessonov et al, 2008). Thus, the question remains whether and how any of these proteins might contribute directly to splicing catalysis (Collins and Guthrie, 2000).

Prp8 is the largest (2413 and 2335 residues in yeast and human) and most highly conserved spliceosomal protein (62% identity between yeast and human) and is considered a master regulator of the spliceosome (Collins and Guthrie, 2000; Grainger and Beggs, 2005). It can be physically crosslinked to U5 (Dix et al, 1998) and U6 snRNAs (Vidal et al, 1999), to a conserved GU dinucleotide at the 5′-end of an intron (Reyes et al, 1996, 1999), the BPS (MacMillan et al, 1994; McPheeters and Muhlenkamp, 2003) and the 3′SS (Teigelkamp et al, 1995a, 1995b). Furthermore, it interacts genetically with U4 and U6 snRNAs (Kuhn et al, 1999), both splice sites, the BPS, and a polypyrimidine (PPy) tract (Umen and Guthrie, 1995a, 1996; Collins and Guthrie, 1999; Siatecka et al, 1999) located between the BPS and the 3′SS. Prp8 is also thought to regulate the activities of key protein splice factors. It forms a salt-stable complex (Achsel et al, 1998) with the Brr2 helicase that is required for spliceosome catalytic activation (Laggerbauer et al, 1998; Raghunathan and Guthrie, 1998) and disassembly (Small et al, 2006) and with the Snu114 GTPase that regulates Brr2 activity (Small et al, 2006). To understand the molecular mechanisms underlying the functions of Prp8 in the spliceosome, we started to characterize its 3D structure.

Results

A newly identified compact portion of Prp8 contains an RNase H domain expanded by unique elements

Apart from a putative RNA-recognition motif in the centre (Grainger and Beggs, 2005) and a Jab1/MPN domain at the very C terminus (Pena et al, 2007; Zhang et al, 2007), no known domains are discernable based on the Prp8 sequence. Guided by conservation patterns and disorder predictions, we expressed residues 1796–2092 of yeast (sc) Prp8 (scPrp8^1796−2092), located directly N-terminal of the Jab1/MPN domain. Limited proteolysis with chymotrypsin gave rise to a fragment, scPrp8^1827−2092, which could be crystallized directly or after recloning as scPrp8^1836−2092. We solved the structures of scPrp8^1836−2092 and scPrp8^1827−2092 at 2.0- and 2.1-Å resolution, respectively, and the structure of the corresponding human (hs) Prp8 protein (hsPrp8^1755−2016), obtained by a similar strategy, at 1.9-Å resolution (Table I). scPrp8^1836−2092 and hsPrp8^1755−2016 crystallized with one, scPrp8^1827−2092 with two protein molecules per crystallographic asymmetric unit.

Table 1.

Crystallographic data

Data collection
	ScPrp81836–2092	Inflection	Remote	scPrp81827–2092	hsPrp81755–2016
	Peak
Wavelength (Å)	0.97968	0.97987	0.91841	0.91841	0.91841
Temperature (K)	100	100	100	100	100
Space group	P31			P212121	P3221
Unit cell (a, b, c) (Å)	46.9, 46.9, 103.0			77.1, 84.2, 95.1	83.4, 83.4, 96.3
Resolution (Å)a	50.0–2.0 (2.07–2.00)	50.0–2.10 (2.18–2.10)	50.0–2.1 (2.18–2.10)	50.0–2.1 (2.18–2.10)	50.0–1.95 (2.02–1.95)
Reflectionsb
Unique	33 148 (3005)	29 331 (1906)	29 376 (1929)	36 691 (3623)	28 624 (2828)
Completeness (%)	96.8 (88.2)	98.8 (98.2)	99.1 (98.5)	99.6 (99.9)	99.9 (100.0)
Redundancy	1.9 (1.7)	2.0 (1.9)	2.0 (2.0)	3.6 (3.7)	7.9 (8.0)
I/σ(I)	16.8 (2.1)	18.3 (3.9)	12.6 (2.6)	10.4 (1.9)	31.6 (4.9)
Rsym(I)c	3.2 (24.6)	3.8 (14.0)	8.2 (22.0)	10.4 (56.2)	4.9 (29.2)
Phasing

Resolution (Å)	Se sites	CC/CCweakd	FOMe
50.0–2.5	8	35.3/25.0	0.42
Refinement
Resolution (Å)	30.0–2.0 (2.05–2.00)	30.0–2.1 (2.15–2.10)	30.0–1.95 (2.00–1.95)
Reflections
Number	16 963 (1175)	36 392 (2696)	28 264 (2035)
Completeness (%)	99.5 (93.1)	98.9 (99.9)	98.7 (98.8)
Test set (%)	5	5	5
Rworkf	17.7 (23.4)	24.0 (27.9)	20.5 (22.2)
Rfreef	24.5 (32.0)	29.8 (30.7)	24.8 (32.1)
ESU (Å)g	0.13	0.20	0.10
Protein atoms	2107	4217	2126
Water oxygens	269	259	315
Ions	—	—	2 Na+, 1 Cl−
Mean B-factors (Å2)
Wilson	21.4	35.3	31.4
Protein	20.2	36.4	37.5
Water	31.4	41.1	47.2
Ions	—	—	38.7
φ/ψ favoured/allowed/ outliersh (%)	98.4/1.6/0	95.1/3.7/1.2	96.9/3.1/0
r.m.s.d. geometry
Bond lengths (Å)	0.009	0.009	0.009
Bond angles (deg)	1.15	1.22	1.12
r.m.s.d. B-factors (Å2)
Main chain bonds	0.50	0.48	0.49
Main chain angles	0.87	0.85	0.91
Side chain bonds	1.70	1.28	1.45
Side chain angles	2.59	1.94	2.40
PDB ID	3E9O	3E9P	3E9L
aData for the highest resolution shell in parentheses.
bFriedel pairs for anomalous data sets of scPrp81836–2092 were not merged during processing but were merged for refinement against the peak data set.
cRsym(I)=∑hkl∑i∣Ii(hkl)−〈I(hkl)〉∣/∑hkl∑i∣Ii(hkl)∣; for n independent reflections and i observations of a given reflection; 〈I(hkl)〉—average intensity of the i observations.
dCC=[∑wEoEc∑w−∑wEo∑wEc]/{[∑wEo2∑w−(∑wEo)2] [∑wEc2∑w−(∑wEc)2]}1/2; w—weight (see http://shelx.uni-ac.gwdg.de/SHELX/shelx_de.pdf for full definitions).
eFOM—figure of merit=∣F(hkl)best∣/∣F(hkl)∣; F(hkl)best=∑αP(α)Fhkl(α)/∑αP(α).
fR=∑hkl∣∣Fobs∣−∣Fcalc∣∣/∑hkl∣Fobs∣; Rwork−hkl∉T; Rfree−hkl∈T; T—test set.
gESU—estimated overall coordinate error based on maximum likelihood.
hCalculated with MolProbity (http://molprobity.biochem.duke.edu/).

The yeast and human Prp8 structures each comprise an ∼160-residue N-terminal α/β domain (NTD; encompassing strands β1–β6 and β1a/β1b, α-helices α1–α3 and 3₁₀-helices η1–η5) and an ∼95-residue C-terminal helical domain (CTD; encompassing α-helices α4–α7 and 3₁₀-helices η6 and η7), which are closely associated (ca. 2000 Ų of combined buried surface area; Figure 1A). The overall structures are reminiscent of a left-hand mitten (Figure 1A, inset), in which a central six-stranded mixed β-sheet and the surrounding α-helices of the NTD correspond to the palm, an extended β-hairpin of the NTD comprises the thumb and the α-helical CTD represents the fingers. The thumb and fingers frame a channel across the palm, the width of which at half height varies between a minimum of 16.5/24.0 Å in the hsPrp8^1755−2016 structure (distances between side chains/between Cα atoms of facing residues) and 22.2/29.0 Å in the scPrp8^1827−2092 structure (Figure 1B and C). The floor and the front opening of this channel are carpeted with extensive patches of positive surface potential (Figure 1B). The structures of scPrp8^1836−2092 and scPrp8^1827−2092 exhibit pairwise root-mean-square deviations (r.m.s.d.s) of 0.8–1.0 Å and align to hsPrp8^1755−2016 with r.m.s.d.s of 1.2–1.7 Å over ca. 230 Cα atoms, demonstrating that the structures of the yeast and human proteins are very similar. Superimposition of the independent structures showed that the palm regions are essentially invariant (Figure 1C). In contrast, the thumbs and the C-terminal portions of the fingers retain their intrinsic structures but adopt different relative orientations with respect to the palm, indicating that these elements are in principle mobile.

Figure 1.

Structure of an RNase H domain in Prp8. (A) Ribbon diagram of scPrp8^1827−2092 (left) and human RNase H1 (right; pdb ID 2QKK; Nowotny et al, 2007) in the same orientations. Top panels: the shape of scPrp8^1827−2092 resembles a mitten (inset) with palm, thumb and finger regions. Secondary structure elements and termini are labelled. Equivalent α-helices and β-strands in Prp8 and RNase H are blue and red, respectively; β-hairpin comprising the thumb (strands β1a and β1b)—magenta; η5 encompassing the peptide crosslinked to the 5′SS in human Prp8—cyan; C-terminal domain of Prp8—green. Acidic residues I–IV involved in metal ion binding in RNase H (the asparagine at position III is an aspartate in the wild-type enzyme) as well as D1853 and D1854 in Prp8, which may correspond to carboxylate I, are shown as sticks (carbon—yellow; oxygen—red; nitrogen—blue). Bottom panels: view on the RNase H-like NTD of scPrp8^1827−2092 (left) and hsRNase H1 (right) rotated 90° about the vertical axis as compared with the top panels. The CTD of scPrp8^1827−2092 was removed for clarity. In scPrp8^1827−2092, the C-terminal helix α3 of hsRNase H1 is divided into helix η5 and the perpendicular helix α3. (B) Electrostatic potential mapped to the surface of scPrp8^1827−2092. Blue—positive charge; red—negative charge. The position of the crosslinked peptide is circled (‘XL'). (C) Superimposition of scPrp8^1836−2092 (white/yellow) with two independent structure of scPrp8^1827−2092 (light grey/green and dark grey/red) and with hsPrp8^1755−2014 (black/blue). The thumbs, the peptides that crosslink to the 5′SS in the human system (‘XL peptide') and the C-terminal portions of the fingers are shown in colours. Thumbs and fingers are evidently flexibly hinged to the central palms.

Comparison of the present Prp8 structures to known structures in the Protein Data Bank (http://www.pdb.org) revealed a striking similarity of the palm region of the NTD to an RNase H fold (r.m.s.d. of 3.0 Å over 95 Cα atoms when compared with human RNase H1 (Nowotny et al, 2007); pdb entry 2QKK). RNase H-fold proteins exhibit a topologically identical and structurally very similar five-stranded mixed β-sheet and three corresponding α-helices surrounding the sheet (Figure 1A). The extended β-hairpin and the α-helical CTD are unique to the Prp8 protein, generating the mitten-like structure. The core of other RNase H-fold proteins can be likewise augmented by other elements (Supplementary Figure S1).

A 5′SS-contacting peptide has a variable sequence but a conserved structure

Residues 1894–1898 within the present portion of human Prp8 have been UV-crosslinked to the 5′SS in a trans-splicing system at the stage of U4/U6–U5 tri-snRNP addition and after establishment of the 5′SS-U6 ACAGAGA-box interaction (Reyes et al, 1996, 1999). The similarity of the protein to an RNase H fold suggests that its mode of RNA binding may resemble that of RNase H. To approximate how RNA may bind to this region of Prp8, we transplanted the RNA strand of an RNase H-DNA–RNA substrate complex (pdb ID 1ZBL) (Nowotny et al, 2005) on scPrp8^1827−2092 by superimposing the proteins. The nucleic acid came to lie on the central ridge between the thumb and the fingers (Figure 2A). At this position, the negatively charged sugar-phosphate backbone can favourably interact with the positively charged surfaces (Figure 1B).

Figure 2.

Functional interactions. (A) Diametrical views on the surface of scPrp8^1836−2092 with a modelled RNA mimicking a bound 5′SS region. The left orientation is identical to that in Figure 1A, top. 5′-exon—brown; 5′SS phosphate—black; intron—beige. The conserved GU at positions +1 and +2 of the intron are labelled. The surface patch corresponding to the peptide, which in hsPrp8 crosslinks to the 5′SS, is encircled. Positions that exhibit genetic interactions with the pre-mRNA or snRNAs are highlighted in colour. Interactions with 5′SS, 3′SS or BPS—blue; interactions with the PPy tract—purple (mutations at positions 1922 and 1946 always occur together with a mutation at 1834); interactions with u4-cs1—red. The region, in which mutations affect the physical interaction with Brr2, is in green. Residues changed in Prp8 mutants are listed. Conserved and invariant residues that lie in the region corresponding to the active site in RNase H enzymes (D1853, D1854, T1855, T1936 and R1937) are in gold. (B) RNA networks around the 5′SS shown on top of an outline of the present Prp8 domain. Before catalytic activation, U4 and U6 are base paired 3′ of the U6 ACAGAGA-box. U1 snRNA and U6 ACAGAGA-box interact at the 5′SS. In U4-cs1 (grey), an AAA sequence of U4 (underlined) is changed to UUG (lower case italics), which can form additional base pairs with U6 that involve the ACAGAGA-box as indicated.

We next asked whether we could use the scissile phosphate from the RNase H reaction as a landmark to assign the approximate positions of the 5′-exon and the intron around the 5′SS. Remarkably, considering this scissile phosphodiester bond as equivalent of the 5′-exon–intron junction, placed in the nucleotides, which would correspond to the conserved GU at positions +1 and +2 of the intron, directly next to the five amino-acid peptide that UV-crosslinks to this GU dinucleotide (Figure 2A) (Reyes et al, 1996, 1999). Thus, the simple model resulting from our structural comparison provides a facile explanation for the zero-length crosslink mapped between Prp8 and the 5′SS. In further agreement with this model, an 11-residue oligomer mimicking the 5′SS still specifically crosslinks to a recombinant portion of hsPrp8 comprising residues 1669–2034 (corresponding to scPrp8 residues 1741–2106; cited in Grainger and Beggs, 2005).

The peptide of hsPrp8 that crosslinks to the 5′-end of an intron and the corresponding yeast fragment designate the only portions of the present domains that significantly diverged in sequence (1894-QACLK-1898 and 1966-SAAMS-1970, respectively; Figure 3). Yet in both proteins, the peptides adopt an almost identical extended 3₁₀-helical structure (helix η5) at the centre of the pronounced channel (Figure 1C), suggesting that the unusual structure rather than the sequence of helix η5 may be functionally important.

Figure 3.

Multiple sequence alignment of Prp8 proteins. Darker background corresponds to higher conservation. Aligned orthologues are H.s.—Homo sapiens; N.c.—Neurospora crassa; Y.l.—Yarrowia lipolytica; C.g.—Candida glabrata; S.c.—Saccharomyces cerevisiae. Icons above/below the alignment indicate secondary structure elements of hsPrp8^1755−2016/scPrp8^1827−2092 as derived from the present crystal structures. Residues that exhibit genetic interactions with the 5′SS, 3′SS and BPS—blue; with u4-cs1—red; with the PPy tract—purple. Residues mutated in this work—yellow. Black triangles—conserved residues around the functional hotspot. The region of Prp8 crosslinked to the 5′SS and the region, in which mutations affect the interaction with Brr2, are boxed (cyan and green, respectively). Green residues in the Brr2-interacting region correspond to changes in the prp8-52 allele.

Prp8 mutants that suppress pre-mRNA defects map on both sides of the putative path of the 5′SS RNA

Numerous Prp8 mutants (prp8-D143 (K1864E), prp8-151 (N1869D), prp8-152 (N1869D/S1970R), prp8-153 (T1982A), prp8-154 (T1982A, SA1966/7AG), prp8-155 (T1982A, V1987A), prp8-162 (V1870N)) that suppress second-step defects (Query and Konarska, 2004) in the 5′SS, 3′SS or BPS (Umen and Guthrie, 1995a, 1996; Collins and Guthrie, 1999; Siatecka et al, 1999) map to the present Prp8 region. Residues 1966, 1967 and 1970 are part of the central peptide, the human equivalent of which crosslinks to the 5′SS (Figure 2A). The other mutants cluster in the flexible thumb and at its base (Figure 2A). According to our placement of RNA, the thumb is ideally positioned to directly interact with bound RNA. Similar hairpins that mediate RNA interactions are found in several ribosomal proteins, such as S5, S10, L4, L22 or L28 (Ban et al, 2000; Wimberly et al, 2000). However, unlike in Prp8, these modules are typically not stably structured off the ribosome (Worbs et al, 2000). These observations suggest that the above alleles exert an effect by affecting Prp8 regions that directly contact the 5′SS. As one possibility, the mutations may confer increased structural plasticity on the thumb and the central 5′SS-binding peptide, allowing them to accommodate deviant substrates.

Prp8 mutants that restore catalytic activation in the presence of a hyperstable U4-cs1/U6 duplex map to the flexible thumb

A variant U4 snRNA, U4-cs1, is thought to sequester the U6 ACAGAGA-box in base pairing and thereby inhibit catalytic activation (Kuhn et al, 1999) (Figure 2B). prp8-201 (T1861P) and other prp8-cat alleles (F1851L, V1860D/N, V1862A/D/Y and I1875T) suppress u4-cs1 phenotypes (Kuhn et al, 1999; Kuhn and Brow, 2000) and in their majority map at the base of the flexible thumb in direct vicinity of the positioned 5′SS RNA mimic (Figure 2A). On the basis of this observation, we suggest that the present Prp8 domain also interacts with U4/U6 and couples U4/U6 unwinding to U1-for-U6 strand exchange at the 5′SS (Figure 2B), consistent with a previous suggestion by Brow and coworkers (Kuhn et al, 1999).

Residues implicated in PPy tract recognition cluster on the Prp8 RNase H domain

Results from genetic suppression screens suggest that the present region of Prp8 is also involved in the recognition of the PPy tract (Umen and Guthrie, 1996). The corresponding alleles, prp8-101 and prp8-D102 to prp8-D107, exhibit E1960K/G or F1834L/S amino-acid exchanges; in prp8-D103 and prp8-D104, a change at position 1834 occurs together with a V1946A and R1922G mutation, respectively. The affected residues are all contained in the current Prp8 structures. Strikingly, although they are some 130 residues apart in the protein sequence, the primary sites, 1834 and 1960, cluster immediately next to each other opposite to the flexible thumb and below the finger region (Figure 2A). This observation suggests that pre-mRNA regions close to the BPS and 3′SS may come to lie at the front entrance of the pronounced channel next to the 5′SS RNA. Consistent with this interpretation, crosslinking of the G-1 residue at the 3′SS to scPrp8 is reduced in the E1960K mutant of prp8-101 (Umen and Guthrie, 1995b). Furthermore, binding of RNA at this location would be supported by the positive surface potential (Figure 1B).

Prp8 contains an incomplete RNase H-like active site

Our discovery of an RNase H fold in the region of Prp8 investigated herein naturally led to the question whether an RNase H-like active centre is present in Prp8. Canonical RNase H enzymes contain four conserved carboxylates in their active sites, used to position two catalytic metal ions (Nowotny et al, 2005) (Figure 4A). Two aspartates (positions I and III; Figure 1A; Supplementary Figure S1) are found in the first β-strand and C-terminal of the fourth β-strand of the RNase H core (corresponding to strands β1 and β4, respectively, of the present Prp8 domain); two additional acidic residues (positions II and IV; Figure 1A; Supplementary Figure S1) originate from the first and last helix of the RNase H core domain (equivalent to Prp8 helices α1 and η5/α3, respectively). In Prp8, the invariant D1853/1781 and D1854/1782 (yeast/human numbering; Figure 3) are found directly C-terminal of strand β1 in topologically equivalent positions to the first aspartate (I) of RNase H (Figure 4A and B), the carboxylate of which bridges the two metal ions in the RNase H catalytic situation (Nowotny and Yang, 2006). We did not find equivalents in Prp8 of the other RNase H catalytic residues, which would be topologically and spatially conserved. There is no acidic residue C-terminal of strand β4. Although two glutamates (E1915/1843 and E1916/1844) originate from helix α1, they lie on the backside of the helix and extend away from D1853/1781 and D1854/1782. The C-terminal helix of RNase H enzymes is represented by two elements in Prp8, helix η5, which in human Prp8 crosslinks to the 5′SS, and helix α3, positioned at a right angle to η5 (Figure 1A). No acidic residue is found in helix η5. D1972/E1900 and D1976/1904 originate from helix α3 but are again remote from D1853/1781 and D1854/1782. Large conformational changes would be required to recruit any of these residues to a common active site centred on D1853/1781 or D1854/1782. Thus, the present Prp8 domain contains an incomplete set of metal-chelating residues or adopts a structure, in which these residues occupy non-productive spatial positions.

Figure 4.

Identification of a novel functional hotspot. (A) Active site of hsRNase H1 in complex with two divalent metal ions (A and B; green spheres) and a DNA–RNA substrate duplex (carbons—light/dark grey; phosphorus—violet; other colours as in Figure 1). The four active site carboxylates (I–IV) are shown as sticks and are colour-coded as before (the asparagine at position III is an aspartate in the wild-type enzyme). Landmark secondary structure elements are labelled. The view is rotated 80° about the vertical axis and 20° about the horizontal axis as compared with the view in Figure 1A, top. (B) Close-up of the region of scPrp8^1827−2092, which is spatially equivalent to the RNase H active site, in the same orientation as in (A). The invariant D1853, D1854, T1855, T1936 and R1937 are shown as sticks and are colour-coded as above. Other conserved residues are labelled. Dashed lines indicate hydrogen bonds and salt bridges. Selected secondary structure elements are labelled. D1853 occupies the same topological and spatial position as carboxylate I of hsRNase H1, the carboxylate of D1854 spatially coincides with the carboxylate II of hsRNase H1. (C) Failure of D1853A and R1937A mutants of scPrp8 to grow on 5-FOA plates. (D) Cell viability assay monitoring the effects of exchanging invariant Prp8 residues D1853, D1854, T1855, T1936 and R1937 as indicated. After selection of clones, the culture and serial dilutions were spotted and grown at the indicated temperatures for 2 days.

We tested whether divalent metal ions could bind to any of the Prp8 structures by soaking the crystals in solutions containing up to 20 mM Mg²⁺, Mn²⁺ or Zn²⁺. No metal ions were observed close to D1853/1781 or D1854/1782 (searching, for example, the anomalous difference Fourier maps of Mn²⁺- and Zn²⁺-soaked crystals). In one structure, a Mn²⁺ ion bound fortuitously between neighbouring molecules in the crystal lattice, remote from these aspartates. Although the failure to bind metal ions may be due to the lack of appropriately positioned carboxylates, any metal ion coordination would most likely be also strongly dependent on the presence of RNA, as observed earlier in other systems (Song et al, 2004).

A novel functional hotspot in Prp8

On transplanting the RNA strand from RNase H onto Prp8, the presumed 5′-exon–intron junction came to rest directly next to D1853/1781 and D1854/1782. These two aspartates are surrounded by additional invariant Prp8 residues, T1855/1783, T1936/1864 and R1937/1865 (Figures 3 and 4B). We investigated whether this cluster of residues was functionally important by mutagenesis. Single-residue mutants were brought into a genetic background, in which we could test whether they supported cell viability. All mutants were associated with dramatic growth defects (Figure 4C and D). D1853/1781 was lethal on conversion to alanine (Figure 4C) and produced both cold- and temperature-sensitive phenotypes on mutation to asparagine (Figure 4D). R1937/1865 was similarly lethal on mutation to alanine (Figure 4C) and conditionally lethal at low and high temperatures on change to a lysine (Figure 4D). D1854/1782, T1855/1783 and T1936/1864 were both cold- and temperature-sensitive on mutation to asparagine or alanines, respectively (Figure 4D). As some changes at each of the positions supported cell growth at moderate temperatures, we exclude a general folding problem as a source of the observed phenotypes. Remarkably, in a previous screen of 300 mutants covering the C-terminal three-quarters of the Prp8 sequence less than 10% (20 alleles) were found temperature sensitive and no cold-sensitive mutation was detected (Umen and Guthrie, 1996). Therefore, our findings mark a novel functional hotspot on Prp8 that coincides with a region spatially equivalent to an RNase H active centre.

Discussion

U2, U5 and U6 snRNAs and a single protein, Prp8, have been implicated as active site components of the spliceosome by the evolutionary conservation of regions that physically or genetically interact with the functional sites on the pre-mRNA (Collins and Guthrie, 2000). Here, we have investigated one such conserved region of yeast and human Prp8 by a combination of crystal structure analysis and site-directed mutagenesis. We find that this portion of Prp8 adopts an RNase H fold with additional elements specific to Prp8 orthologues. This observation motivated us to derive an approximate 5′SS-RNA-binding model from comparison with the structure of an RNase H–substrate complex, which was consistent with the surface topology and charge distribution of the Prp8 domain. Strikingly, on this crude model, biochemical and genetic interaction data quite naturally ‘fall into place'. Numerous Prp8 mutants that genetically interact with the pre-mRNA, snRNAs or other protein splice factors mapped close to each other and around a peptide that can be UV-crosslinked to the 5′SS on the structure of the Prp8 RNase H domain (Figure 2A), strongly suggesting that the genetic interactions represent direct physical contacts. Our results provide a structural basis for expected Prp8 activities (Kuhn et al, 1999; Collins and Guthrie, 2000), suggesting how the protein mediates and proofreads RNA transactions on the way to a functional catalytic core (5′SS-U1 to 5′SS-U6 strand exchange, U4/U6 unwinding) and how it stabilizes RNA networks during catalysis (5′SS-U6 ACAGAGA-box, apposition of 5′SS and BPS). While this paper was under review, a paper describing a structure consistent with ours of a very similar fragment of yeast Prp8 appeared in press (Yang et al, 2008).

An RNase H domain as an ideal device to mediate nucleic acid transactions at the 5′SS on the way to the activated spliceosome

Genetic and physical interaction studies show that the Prp8 RNase H domain is involved in RNA rearrangements at the 5′SS during catalytic activation. At that time, the 5′SS-U1 snRNA duplex is unwound and the 5′SS-U6 ACAGAGA-box duplex is concomitantly formed. The RNase H domains of different enzymes are tuned to interact with diversely structured nucleic acids, such as DNA–RNA hybrids in the case of bona fide RNase H's (Nowotny et al, 2005), RNA duplexes in the case of Argonaute proteins (Song et al, 2004) or DNA duplexes in the case of retroviral integrases, DNA transposases or Holliday junction resolvases (Rice and Baker, 2001). Furthermore, the latter domains handle more than one nucleic acid strand or duplex at the same time as is also required during formation of the spliceosomal active sites and during splicing catalysis. Therefore, an RNase H domain seems to be the ideal choice for a device that needs to control coordinated structural rearrangements involving multiple RNA strands and duplexes (Figure 2B).

Coordination of activities and events during catalytic activation

Apart from Prp8, the activity of the DEAD-box protein Prp28 is also required for strand exchange at the 5′SS (Staley and Guthrie, 1999). Interestingly, the prp8-201 allele (mapped at the base of the thumb of the Prp8 RNase H domain) and the prp8-501 allele (I1825K and L1835F; coinciding with the PPy tract recognition region) interact genetically with the prp28 locus (cited in Grainger and Beggs, 2005). In the framework of a dysfunctional U1-C protein, Prp28 is dispensable for catalytic activation (Chen et al, 2001) but becomes essential again in the presence of prp8-201 or prp8-501. Thus, the Prp8-201 or Prp8-501 mutant proteins may interfere with U1-5′SS unwinding or with pairing of the U6 ACAGAGA-box to the 5′SS. It is possible, for example, that they stabilize the U1-5′SS duplex or inhibit positioning of the 5′SS.

The above genetic interactions suggest that Prp28 and the Prp8 RNase H domain exert an effect in close physical neighbourhood during catalytic activation. Indeed, a 5′SS RNA can be crosslinked to a position in Prp28 close to its ATP-binding site in the same spliceosomal assembly stage (Ismaili et al, 2001), in which the 5′SS GU dinucleotide crosslinks to the Prp8 RNase H domain (Reyes et al, 1996). Taken together, these observations show that the RNase H domain of Prp8 is involved in the recognition of the 5′SS early during the splicing cycle when it is still base-paired with U1 snRNA, and that it closely cooperates with Prp28 to provide the trans-helicase activity required for hand-over of the 5′SS from U1 to U6 during catalytic activation. In yeast Prp8, regions different from the RNase H domain can be crosslinked to the 5′SS (Turner et al, 2006). Thus, it is possible that the 5′SS binds only transiently at the present domain during early assembly stages of the spliceosome and is then transferred to other domains of Prp8.

The mapping of prp8-cat alleles (that suppress u4-cs1 phenotypes) at the base of the thumb, close to the 5′SS interaction channel, indicates that the present domain of Prp8 also exerts an effect to couple U4/U6 unwinding to the rearrangements at the 5′SS, a previously postulated function of Prp8 (Kuhn et al, 1999). The exact nature of this coupling is not clear. Positioning of U4/U6 appropriately for unwinding could be affected by binding of the U6 ACAGAGA-box by the present Prp8 domain, which in turn may be dependent on correct positioning of the 5′SS RNA. In this picture, the prp8-cat suppressors may destabilize the non-natural U4-cs1/U6 duplex to allow catalytic activation.

Prp8 has also been implicated in coordinating the activities of other proteins that mediate spliceosome assembly and remodelling steps, including those of the DEXD/H-box protein Brr2 that is required for U4/U6 unwinding (Laggerbauer et al, 1998; Raghunathan and Guthrie, 1998). Interestingly, substitutions in region 2033–2067 of Prp8 (e.g. Y2037H and I2051T in the prp8-52 allele), which encompasses a long α-helix and the preceding loop on the backside of the finger region (Figure 2A), enhance the physical interaction with Brr2 in yeast two-hybrid and pull-down assays (van Nues and Beggs, 2001). Thus, apparently the Prp8 RNase H domain also coordinates RNA structural transitions by either recruiting the RNA remodelling enzyme Brr2 (directly or indirectly by interacting with a Brr2-binding site), by allosterically influencing its activities or by competing for its substrates. We have shown earlier that the Jab1/MPN domain of Prp8, which lies immediately C-terminal of the RNase H domain, also serves as a binding module for Brr2 (Pena et al, 2007). In the light of their close proximity, it appears feasible that Brr2 jointly recognizes epitopes on both of these Prp8 domains.

The Prp8 RNase H domain may also organize pre-mRNA regions close to the 3′SS

Fe-BABE-directed hydroxyl radical probing has demonstrated that the 5′-end of U2 snRNA is in close proximity of pre-mRNA functional regions and of U1 snRNP in early spliceosomal complexes (Dönmez et al, 2007). U2 snRNA base pairs with the BPS of the intron close to the PPy tract. The close proximity of U1, U2 and the functional pre-mRNA regions they bind would favour concomitant recognition of the 5′SS-U1 snRNA duplex and regions close to the U2-BPS duplex by Prp8, consistent with the mapping of PPy tract recognition mutants close to the 5′SS-binding region on our structures (Figure 2A). As the PPy tract marks the approximate position of the BPS, which attacks the 5′SS in the first step, we suggest that the Prp8 RNase H domain is involved in correct positioning of the two pre-mRNA regions that react in the first step of splicing. In activated spliceosomes, portions of the U6 ISL also approach the U2 BP-binding sequence and the 5′SS (Rhode et al, 2006). These data place the Prp8 RNase H domain in close proximity of the U6 ISL that is expected to constitute another active site component. Thus, the U6 ISL and the Prp8 RNase H domain may cooperate to achieve splicing catalysis.

The Prp8 RNase H domain facilitates structural transitions between two or more competitive conformations of the spliceosome

In analogy to the ribosome, the spliceosome may toggle between competing pre-step-one, step-one and step-two conformations, where the first and third states may resemble each other (Query and Konarska, 2004). According to this model, alleles that suppress step-two defects destabilize a step-one conformation (or interaction network), thus favouring adoption of the step-two conformation. Step-one mutants can be rationalized by a negative effect on a conformation (or interaction network) of the second step or of a pre-step-one stage. Significantly, both step-one (prp8-101/D102 and prp8-D103/4/5/6/7) and step-two alleles (suppressors of all other pre-mRNA defects) map to the present structures. Combining our data with the toggling model, the step-one suppressor mutations could impair recognition of 3′-portions of the intron by the Prp8 RNase H domain in a pre-step-one stage (affecting assembly) or in the second step (affecting conversion of step one to step two or catalysis of step two). The step-two suppressor alleles could exert an effect by destabilizing Prp8–5′SS interactions required for the first step (affecting stabilization of a step-one situation or catalysis of step one). Thus, the Prp8 RNase H domain provides a platform that may exert an effect during catalytic activation and throughout both catalytic steps.

Possible roles of the Prp8 RNase H domain in the catalysis of pre-mRNA splicing

Pre-mRNA splicing chemically recapitulates self-splicing by group II introns and several snRNA elements resemble catalytic portions of group II introns. A recent crystal structure of a hydrolytic group IIC intron in the post-catalytic state (Toor et al, 2008) revealed that the U6-like ISL binds two metal ions 3.9 Å apart, in agreement with a two-metal-ion mechanism mediated by that site, which may also apply to the spliceosome. However, details of the spliceosomal catalytic strategies may have diverged from the RNA-based mechanisms of group II introns (Michel and Ferat, 1995; Collins and Guthrie, 2000). Most obviously, catalysis by the spliceosome depends on many protein splice factors but their precise roles are normally not known. Our results show how an RNase H domain of the Prp8 protein is involved in the correct preparation and positioning of substrates during pre-mRNA splicing. As positioning is a crucial catalytic function, this finding alone attributes to Prp8 a direct role in splicing catalysis.

Similar to group II introns, RNase H-fold enzymes employ a two-metal-ion mechanism that invariably leads to retention of the phosphate group on the 3′-portion of the product and a free 3′-hydroxyl on the 5′-portion, as observed in both steps of splicing. Our mutational analysis has identified candidate residues on the RNase H domain in Prp8 (Figure 4B), through which the protein might even chemically facilitate splicing catalysis. For example, Prp8 may contribute D1853/1781 or D1854/1782 (yeast/human numbering; equivalent to the RNase H carboxylate I) to a composite RNA–protein active site, in which they could assist RNA residues in coordinating a catalytic metal ion.

Previous data indicated a particular importance for the recognition of the base moiety at position +2 of the intron (Reyes et al, 1996). Residues in the functional hotspot and the neighbouring Q1907/1835, which is conserved and recognizes substrate bases in other RNase H-fold enzymes (Supplementary Figure S2), are appropriately positioned to mediate such interactions (Figure 4B). Residues in this region may also recognize backbone functionalities. They could sequester the 2′-hydroxyl group of a scissile phosphate, which in an irregular RNA structure, such as around the 5′SS in the first step, could otherwise function as a nucleophile and lead to aberrant 5′-OH and 2′–3′-cyclic phosphate products. However, none of the 2′-OH groups is absolutely required for trans-splicing of a 5′SS oligo (Konarska, 1998) and the 2′-OH at the 5′SS is not essential for the first step of splicing (Moore and Sharp, 1992), in general agreement with the conditional phenotypes we observe with many mutants. It is also conceivable that R1937/1865, which is notably conserved in resolvases and transposases (Supplementary Figure S2), could carry out functions normally associated with the catalytic metal ions (activation of the nucleophile, stabilization of the transition state or stabilization of the negative charge on the leaving group) in one of the steps. An analogous situation has been observed in EcoRV, in which a lysine cooperates with metal ions (Horton et al, 1998).

Additional invariant residues (N1856/1784, R1859/1787, K1910/1838 and K1938/1866) surround the functional hotspot (Figure 4B) and are suitably positioned for RNA binding. Furthermore, the invariant F1965/1893 rests on another conserved phenylalanine, F1851/1779, and fixes the N terminus of the η5 helix. Extended 3₁₀-helices are rarely found in proteins and may interconvert with α-helices (Topol et al, 2001). Such transitions at η5 could be important for the remodelling around the 5′SS during spliceosome assembly and catalysis. For example, structural transitions on RNA binding could expose F1965/1893 and allow it to stack with nucleobases. Interestingly, F1865L is the only prp8-cat mutation that maps to the present domain outside of the thumb. A leucine at position 1851 provides a smaller platform for F1965, which could in turn more easily rearrange to interact with an RNA ligand.

Taken together, our results point to various mechanisms by which the spliceosome may function as a veritable ribonucleoprotein enzyme (‘RNPzyme').

Truncation of the Prp8 RNase active site may protect the pre-mRNA during assembly

The activation of substrates in the two transesterification reactions of pre-mRNA splicing has to be timed with exquisite precision. An RNase H domain that handles functional pre-mRNA regions during assembly and catalysis poses the danger of activating a water molecule that could initiate an inappropriate hydrolytic attack on the pre-mRNA. Thus, mechanisms must be at work, which inactivate the Prp8 RNase H domain as a hydrolytic enzyme. In the conformation seen in the present structures, the Prp8 RNase-like active site is incomplete and does not support metal ion binding on its own. It is conceivable that conformational changes on substrate binding could lead to rearrangements that bring about a complete active site. Perhaps more likely, the contraction of multiple RNA elements on the surface of the present Prp8 domain in proximity of the 5′SS could lead to complementation of the truncated protein active site by RNA elements. Similar mechanisms have been observed before in proteins that were originally presumed to be pseudoenzymes (Mukherjee et al, 2008). The expected molecular crowding may additionally help to exclude water molecules from the active site environment.

Materials and methods

Details of the cloning, expression, purification and the crystallographic procedures are given in the Supplementary data.

Limited proteolysis

For analytical digestion, 10 μg of scPrp8^1796−2092 was mixed with increasing amounts of chymotrypsin (Sigma-Aldrich) and incubated for 1 h on ice. The reactions were stopped by the addition of SDS–PAGE loading buffer and incubation at 95°C. A stable band resulting from treatment with chymotrypsin was analysed by tryptic peptide mass fingerprinting (Griffin et al, 1995) and indicated scPrp8^1827−2092 as a stable subfragment of scPrp8^1796−2092.

For preparative digestion of scPrp8^1796−2092 with chymotrypsin, 3.5 mg of scPrp8^1796−2092 was mixed with 0.075 mg chymotrypsin and incubated for 19 h at room temperature. The reaction was stopped by the addition of phenylmethylsulphonylfluoride (PMSF) to 1 mM and dialysis against 10 mM Tris–HCl, pH 7.5, 150 mM NaCl, 1 mM DTT and 5% glycerol. The digested protein was concentrated to 7 mg/ml.

For preparative digestion of hsPrp8^1747−2016 with chymotrypsin, 10 mg of the protein was mixed with 0.4 mg of chymotrypsin and incubated for 2 h at 16°C. The reaction was stopped by the addition of PMSF to 0.1 mM and the buffer was exchanged to 10 mM Tris–HCl, pH 7.5, 150 mM NaCl, 1 mM DTT, 5% glycerol on a NAP-25 column (GE Healthcare). A stable band that resulted from the proteolysis was analysed by tryptic peptide mass fingerprinting and identified as hsPrp8^1755−2016. The truncated protein was concentrated to 37 mg/ml.

Site-directed mutagenesis and plasmid shuffling

The desired substitutions were introduced into the PRP8 gene of plasmid pJU186 (Umen and Guthrie, 1995a) (wild-type PRP8 on plasmid pSE362 (CEN, HIS) was a kind gift of Catherine Collins and Christine Guthrie, University of California, San Francisco, USA) by the QuikChange site-directed mutagenesis strategy (Stratagene) and verified by sequencing. Wild-type and mutant plasmids were transformed into yeast strain JDY8.06 (kindly provided by Richard Grainger and Jean Beggs, University of Edinburgh, UK), containing wild-type PRP8 on a counter-selectable URA3-marked plasmid (Brown and Beggs, 1992). Before plasmid shuffling (Boeke et al, 1987), cells were selected at 25°C in a medium lacking histidine. Transformants were streaked once on medium lacking histidine and grown at 25°C and patches were streaked three times on 5-fluoroorotic acid (5-FOA) plates to select for cells lacking the URA3 plasmid. Cells that survived on 5-FOA plates were streaked on rich medium, and their growth phenotypes were analysed by incubating cells (ca. 5 × 10⁴ cells) and serial eight-fold dilutions at 37, 30, 25 and 16°C for 1–2 days. Cold- and temperature-sensitive strains failed to grow on −Ura plates at any of the temperatures and, in addition, did not grow in rich medium at 16 or 37°C, respectively, but grew at 25 or 30°C.

Supplementary Material

Supplementary Figures S1 and S2