Crystal structure of a eukaryotic group II intron lariat.

The formation of branched lariat RNA is an evolutionarily conserved feature of splicing reactions for both group II and spliceosomal introns. The lariat is important for the fidelity of 5' splice-site selection and consists of a 2'-5' phosphodiester bond between a bulged adenosine and the 5' end of the intron. To gain insight into this ubiquitous intramolecular linkage, we determined the crystal structure of a eukaryotic group IIB intron in the lariat form at 3.7 Å. This revealed that two tandem tetraloop-receptor interactions, η-η' and π-π', place domain VI in the core to position the lariat bond in the post-catalytic state. On the basis of structural and biochemical data, we propose that π-π' is a dynamic interaction that mediates the transition between the two steps of splicing, with η-η' serving an ancillary role. The structure also reveals a four-magnesium-ion cluster involved in both catalysis and positioning of the 5' end. Given the evolutionary relationship between group II and nuclear introns, it is likely that this active site configuration exists in the spliceosome as well.

Splicing of nuclear introns results in the formation of circular RNAs[1] having a branched lariat structure containing an unusual 2′-5′ phosphodiester bond[2,3]. This branched RNA product was also found in group II introns[4,5], which are self-splicing ribozymes. Defects in lariat formation result in aberrant splicing and human disease[6]. In higher eukaryotes, splicing of nuclear introns is catalyzed by a large ribonucleoprotein complex called the spliceosome, which is thought to share a common ancestor with group II introns[7,8].

Group II introns are catalytic RNAs with six structural domains (Extended Data Fig. 1) that splice via two transesterification reactions. In the first step of splicing, the 2′-OH of a bulged adenosine residue is the nucleophile that attacks the 5′ splice site to generate lariat RNA[4,5]. In the second step, the 3′-OH of the 5′ exon attacks the 3′ splice site to form ligated exons and excised intron lariat. The highly conserved domain V (DV) forms the group II intron active site by binding catalytic metal ions[9], and domain VI (DVI) contains the bulged adenosine used as the nucleophile in the first step of splicing[10].

Extended Data Figure 1

Secondary structure of P.li.LSUI2 intron crystallization construct. Tertiary interactions are indicated with Greek letters and domains are labeled with Roman numerals. Colouring of the individual domains is consistent with the overall view of the tertiary structure shown in Fig. 1.

Group II introns are divided into three structural classes: IIA, IIB, and IIC[11,12]. Historically, the two model systems used to study group II intron structure and function have been “canonical” eukaryotic IIB introns: P.li.LSUI2 from the brown algae Pylaiella littoralis[13] and aI5γ from the yeast Saccharomyces cerevisiae[4]. However, the only available crystal structure is of a IIC representative from the bacterium Oceanobacillus iheyensis (O. iheyensis)[9]. This idiosyncratic IIC intron class is the most primitive[14] and splices through hydrolysis to form linear intron[15]. In contrast, eukaryotic IIA and IIB introns form lariat, are evolutionarily later branching[14], and therefore more closely related to the spliceosome. We targeted the P.li.LSUI2 intron for structure determination since it contains a functional DVI that forms large amounts of lariat during splicing[13].

Overall structure

Here we present the structure of the P.li.LSUI2 intron in the post-catalytic lariat form with ligated exon product at 3.7 Å resolution (Extended Data Table 1a) solved using a Yb3+ derivative (Extended Data Fig. 2). This represents the first crystal structure of a 2′-5′ branched RNA molecule.

Extended Data Table 1a

	P.li.LSUI2 native	DV Triad Mutant	G79A Mutant	Yb3+		Tl+
Data collection				Peak	Inflection
Space group	C2221	C2221	C222	C2221	C2221	C2221
Cell
a, b, c (Å)	163.7, 255.4, 136.8	161.9, 264.5, 137.5	211.7, 457.2, 179.5	164.6, 257.7, 138.0		161.2, 257.7, 136.3
α, β, γ (°)	90, 90, 90	90, 90, 90	90, 90, 90	90, 90, 90	90, 90, 90	90, 90, 90
Wavelength (eV)	11218.5	12662	8949.5	8949.5	8946.5	12657.5
Resolution (Å)	150.0-3.68 (3.74-3.68)	50.0-7.25 (7.37-7.25)	50.0-9.75 (9.92-9.75)	50.0-4.65 (4.73-4.65)	50.0-4.49 (4.57-4.49)	50.0-4.48 (4.56-4.48)
Rsym	14.9 (>100)	9.5 (77.9)	8.6 (97.0)	8.8 (73.9)	9.2 (72.8)	8.8 (74.9)
I/σI	6.4 (0.6)	13.2 (1.7)	14.3 (1.8)	16.1 (2.1)	9.2 (1.2)	21.8 (2.2)
Completeness	99.9 (99.9)	94.6 (82.5)	96.3 (96.9)	99.9 (99.6)	97.9 (88.7)	99.3 (99.6)
Redundancy	6.8 (3.8)	5.7 (4.7)	6.5 (6.0)	4.4 (3.7)	3.6 (2.8)	7.8 (6.4)
CC*	(0.743)
Refinement
Resolution (Å)	81.8-3.68
No. reflections	31107
Rwork/Rfree	23.9/27.4
No. atoms	13979
RNA	13471
Ligand/ion	393
Water	115
B-factors
RNA	201.6
Ligand/ion	181.5
Water	177.0
R.m.s.
Bond lengths	0.017
Bond angles	1.483

Highest resolution shell is shown in parenthesis.

Extended Data Figure 2

The Yb-MAD experimental, density modified map of the portion of DV containing the catalytic triad contoured at 1.8σ.

Reflecting the ability of IIB introns to form lariat, there are a multitude of unique tertiary interactions in the P.li.LSUI2 intron compared to the O. iheyensis structure (Fig. 1). These newly visualized contacts include EBS2-IBS2 (Extended Data Fig. 3), μ-μ’, ε-ε’ and the canonical form of κ-κ’. Unlike the O. iheyensis structure, domains II and III interact with multiple domains through long-range interactions to stabilize the overall fold of the P.li.LSUI2 intron. We can now visualize the location of DVI within the intron structure (Fig. 1; Extended Data Fig. 4), as well as the 2′-5′ lariat linkage between the first residue and the bulged adenosine.

Figure 1

A comparison of the tertiary structures of O. iheyensis and P.li.LSUI2 group II introns. P.li.LSUI2 has a significantly larger and more complex structure with a correspondingly greater number of unique RNA tertiary contacts. Domain VI of P.li.LSUI2 is highlighted within the box.

Extended Data Figure 3

The path of the 5′ exon through the intron structure. The EBS1-IBS1 and EBS2-IBS2 interactions position the 5′ exon. They do not form a continuous binding interface with the presence of a highly distorted backbone at the junction between these two motifs. As a result, the helical axes of the EBS1-IBS1 and EBS2-IBS2 pairings are positioned ~90° relative to each other. The EBS3-IBS3 interaction places the 3′ exon in the active site.

Extended Data Figure 4

Overall tertiary structure of the P.li.LSUI2 intron. Individual domains and subdomains are depicted in different colors. Domain names are labeled with Roman numerals. a and b show different rotations of the intron structure.

Newly visualized tertiary interactions

One of the most highly conserved tertiary contacts in group II introns is the κ-κ’ interaction between the base of the catalytic DV stem and domain I (DI)[16]. The conserved κ sequence GAA, nucleotide A171 from near the κ region, and residues from a GUAAC pentaloop in DIII converge to form a pentuple adenosine base stack (underlined residues) that inserts into the minor groove at the base of DV; rigidly placing the active site into the DI scaffold (Fig. 2a; Extended Data Fig. 5a).

Figure 2

Tertiary interactions in a IIB intron. a, The κ loop (grey) and DIII (yellow) converge to form an extended base stack involving five adenosine residues inserting into the minor groove of DV (red). b, G106 and C107 form Watson-Crick pairs with C4 and G3 from the 5′ end to form the ε-ε’ interaction that positions the 5′ end in the active site. c, J1/2 (green) interacts with the 5′ end (A6 residue in green), DIII (yellow), and J2/3 (blue). d, DIII (yellow) acts as a brace on the surface of the intron and forms three tertiary interactions (boxed) with DI (green), DII (blue) and DV (red).

Extended Data Figure 5

Companion to Fig. 2 showing the location of the individual tertiary interactions relative to the overall structure.

The ε-ε’ interaction is critical for catalysis, with disruption through mutagenesis resulting in complete loss of splicing activity[17]. This interaction consists of nucleotides G106 and C107 pairing with C4 and G3 from the 5′ end of the intron (Fig. 2b; Extended Data Fig. 5b). The end result of these contacts is the formation of five conserved bases stacking in the following order (from bottom to top): A573, U2, G5, C4, and G3. This serves to structure the 5′ end of the intron.

The conserved GUAA linker connecting domains I and II (J1/2) adopts an unusual backbone configuration that interacts with the 5′ end, the junction between domains II and III (J2/3), and DIII (Fig. 2c; Extended Data Fig. 5c). The 5′ end and J2/3 directly interact with the active site through J1/2 positioning these regions to stabilize the core. This new long range contact (designated as ρ-ρ’) consists of two adenosines from J1/2 docking into the basal stem of DIII (Fig. 2c).

We can now correlate the function of DIII as a catalytic effector in group II introns. DIII interacts with the intron core through the aforementioned GUAAC pentaloop, which docks into the base of DV (Fig. 2d). This μ-μ’ interaction[18] serves to buttress the opposite side of DV from where catalysis takes place. Furthermore, a GAAA tetraloop from DIII interacts with the base of the DII stem (designated as τ-τ’) to provide additional reinforcement. Therefore, DIII functions as an external brace located on the outside surface to stabilize the entirety of the structure. This important role is consistent with deletion or mutagenesis of this domain rendering the intron unstable and not competent for efficient catalysis[19].

DII positions DVI in the active site

In the O. iheyensis structure, DII was drastically truncated to a small stem loop structure, and in many previous biochemical studies of the aI5γ intron, DII was similarly shortened to study the first step of splicing. We can now visualize the intact DII substructure and find that it serves as a central hub for four different tetraloop receptor interactions (Fig. 3a). DII makes contacts with domains I, III, and VI to organize a large portion of the intron structure.

Figure 3

The position of DVI within the intron structure. a, DII forms an inverted Y-shaped structure that engages in four distinct tetraloop-receptor interactions with the IC stem (orange), DIII (yellow), and DVI (purple). DVI interacts with DII via two tandem tetraloop-receptor interactions, η-η’ and π-π’. b, In vitro self-splicing assays of the P.li.LSUI2 intron. The wild-type (WT) intron efficiently catalyzes both steps of splicing and forms intron lariat and ligated exons. Mutagenesis of either η-η’ (Δη’)[20] or π-π’ (Δπ) inhibits the second step, resulting in the accumulation of lariat-3′ exon and 5′ exon. A combination of both mutations (Δη’Δπ) nearly blocks the second step of splicing with predominantly lariat-3′ exon present. c, Fo–Fc density for the 2′-5′ lariat phosphodiester bond contoured at 3σ. The nucleobase of A615 is disordered and not visualized. This map was calculated using a model deleted for A615, G1, and U2 (shown in stick format) to avoid model bias. The 5′ (G1) and 3′ (A620) ends form a non-canonical base pair.

DII has a Y-shaped RNA secondary structure with two stems D2a and D2b (Extended Data Fig. 1) coaxially stacking on top of each other such that a tetraloop receptor from D2a and a GCAA tetraloop from D2b are facing the same side (Fig. 3a). This combination provides a binding interface for DVI, which contains both a GAGA tetraloop and a tetraloop receptor. Therefore, DVI is tightly placed in the core of the intron via two tandem tetraloop receptor interactions with DII. The interaction between D2a and DVI is known as the η-η’ contact[20] and we are designating the newly discovered interaction between D2b and DVI as π-π’. The π-π’ interaction is especially interesting due to its proximity to the bulged adenosine residue A615, which is the nucleophile for the first step of splicing. The π tetraloop interacts with nucleotides directly adjacent to the bulged adenosine (Fig. 3a) and therefore is likely to have important effects upon the positioning of this nucleotide within the active site.

Mutagenesis of the GNRA tetraloops to UUCG was done to test the effects of these two interactions on splicing. Disrupting either contact significantly inhibited the second step of splicing, leading to an accumulation of lariat-3′ exon and 5′ exon (Fig. 3b; Extended Data Fig. 6). However, disrupting both interactions simultaneously resulted in a near complete block of the second step. This indicates that η-η’ and π-π’ are synergistic interactions essential for the transition to the second step of splicing and likely function through moving the 3′ splice site (which is attached to DVI) into the active site.

Extended Data Figure 6

Splicing assays for the DVI mutants showing the proportion of branched product. Blue and orange bars correspond to lariat-3′ exon and lariat, respectively. The Δη’ and Δπ mutants accumulate large amounts of lariat-3′ exon, thus indicating a second step splicing defect. The Δη’Δπ double mutant is almost completely blocked before the second step. The A620C mutant shows 2.4-fold greater accumulation of lariat-3′ exon compared to the WT intron indicating that the interaction between G1 and A620 is important for the second step. In the yeast aI5γ intron, G1 instead interacts with the penultimate residue[44], indicating a certain degree of flexibility for this pairing. There is evidence for a similar interaction between the termini of nuclear introns[45] involving nearby (but not exactly equivalent) residues, which also has a significant effect upon the second step of splicing. Therefore, the 5′ and 3′ ends of nuclear introns may have a similar arrangement within the spliceosome.

We can now visualize the overall fold of DVI embedded within a catalytically active intron (Fig. 3a). DVI contains a purine-rich internal loop, which forms non-canonical pairings that induce a slight bend in the helix, allowing DVI to simultaneously form both the η-η’ and π-π’ contacts with DII. These interactions position the ribose sugar of the bulged A615 directly under the 5′ end of the intron. The adenosine in this 2′-5′ lariat linkage is highly constrained, since it is connected to phosphates on three different sides of the nucleotide (Extended Data Fig. 7a). The nucleobase component of the A615 residue is disordered (Fig. 3c; Extended Data Fig. 7b), which is consistent with it not playing a role in the later stages of splicing. In the current post-catalytic state, the lariat phosphodiester bond is located ~20 Å from the active site. Therefore, it has undergone a large-scale movement away from the catalytic core after the first step of splicing. In close proximity to the lariat bond, the 5′ and 3′ ends of the intron interact with each other through G1 forming a non-canonical base pair with A620 (Fig. 3c), which is important for the second step (Extended Data Fig. 6).

Extended Data Figure 7

a, The lariat 2′-5′ phosphodiester bond in wall-eyed stereo format. b, Stereo version of Fig. 3c. See Fig. 3c legend for details. Fo-Fc density for the lariat bond contoured at 3σ.

Active site metal ion configuration

Mg2+ ions are an absolute requirement for the catalysis of RNA splicing. To identify active site metal ions, soaks were performed using the anomalous scatterer Yb3+, which exhibits the same octahedral coordination geometry as Mg2+ and preferentially binds to sites containing highly coordinated magnesium ions[9,21]. This revealed four large anomalous peaks in the ribozyme core (Fig. 4a). Two of these peaks (M1 and M2) are also found in O. iheyensis[9] and are embedded within DV to coordinate to the junction phosphate between the ligated exons, while the other two peaks (M3 and M4) are coordinated to the 5′ end of the intron.

Figure 4

The core of the of P.li.LSUI2 intron. a, Yb3+ anomalous map (orange mesh) contoured at 12σ reveals four large peaks, which correspond to highly-coordinated magnesium ions (orange spheres). b, Comparison of the catalytic triplexes found in the O. iheyensis and P.li.LSUI2 introns. O. iheyensis contains a continuous triple helix spanning between the DV catalytic triad (CGC) and J2/3. In P.li.LSUI2, the J2/3 residue A422 is disengaged from the catalytic triad (AGC) and forms a base triple with J4/5 and J5/6 nucleotides (U549 and U584, respectively).

M3 is located in a binding pocket formed by the highly conserved 5′ end (Extended Data Fig. 8a), which has the sequence GUGCG. Nucleotides in this region exhibit a highly contorted backbone configuration that wraps around all sides of M3 due to the ε-ε’ interaction involving residues G3 and C4 (Fig. 2b). Therefore, we postulate that the primary function of ε-ε’ is to order a crucial metal-binding platform which structures the 5′ end of the intron.

Extended Data Figure 8

a and b depict the RNA ligands surrounding metals M3 and M4, respectively. Yb3+ anomalous map contoured at 9σ. c, Yb3+ anomalous map for WT contoured at 9σ. d, Compared with the WT intron, the Yb3+ anomalous map for the ι G79A mutant (contoured at 4σ) is lacking the peaks corresponding to M3 and M4, even at a lower contour level. e, Tl+ was used as a probe for monovalent ions in the RNA structure[46]. The Tl+ anomalous map (purple mesh contoured at 5.5σ ) revealed a strong peak located 3.8 Å from M1 that coordinates to the nucleobase of J2/3 residue G421 and the backbone of DV nucleotide G550. This sodium ion Na1 (purple sphere) is significantly closer to M1 than the equivalent K+ ion found in O. iheyensis[25]. Otherwise, this monovalent ion binding site is relatively conserved between these two introns.

M4 coordinates to conserved IIB intron residues A6, C7, and A341 to further stabilize the 5′ end (Extended Data Fig. 8b). These residues are in close spatial proximity to the recently proposed ι motif (Extended Data Fig. 1) that is known to promote lariat formation and is predicted to serve as a receptor for positioning DVI[22]. An ι G79A point mutant was crystallized, revealing strong signals for M1 and M2; however, M3 and M4 were no longer visible (Extended Data Fig. 8c, d). Splicing assays of this mutant also show a negative effect on the first step of splicing (Extended Data Table 1b). Furthermore, the ε-ε’ interaction, which forms the metal-binding platform for M3 and M4, persists throughout group II intron catalysis[17]. Taken together, the data suggests that M3 and M4 participate in the first step of splicing by positioning the 5′ splice site in the active site to present the scissile phosphate to M1, M2, and the bulged adenosine. However, it is possible that M3 exists only in the post-catalytic state to stabilize the repositioned lariat bond. Given the rarity of highly coordinated magnesium sites in large RNAs[23], the existence of four such metal ions in close proximity in the active site of P.li.LSUI2 is striking. In addition, we observe a conserved monovalent ion near the M1/M2 catalytic center (Extended Data Fig. 8e).

Extended Data Table 1b

	kfast (min−1)	kslow (min−1)
WT	9.51 ± 1.09	0.136 ± 0.0246
G79A	2.97 ± 1.57	0.0767 ± 0.00942

Catalytic triplex rearrangement

In the O. iheyensis structure, J2/3 and residues from DV form a catalytic triplex structure essential for the catalysis of splicing and is also found in the spliceosome[24]. In the O. iheyensis triplex, J2/3 residues G288 and C289 (analogous to G421 and A422 in P.li.LSUI2) form base triples with the first two nucleotides of the catalytic triad. Unexpectedly, we observe in the P.li.LSUI2 structure that one of the J2/3 residues is completely disengaged from the catalytic triad. Specifically, A422 has moved away from the triad and stacks directly underneath the base of the DV helix to form a base triple with residues from the J4/5 and J5/6 linkers (Fig. 4b). The possibility of J2/3 participating in conformational changes has been previously postulated[25]. However, we can now see the specific nature of this “switch” with the disengagement of J2/3 from the catalytic triad into an alternate configuration. The linkers between domains are highly conserved for each subclass of group II introns and based on the P.li.LSUI2 structure, we hypothesize that they are dynamic and modulate splicing.

π-π’ is a dynamic interaction

DVI is proposed to engage in large-scale conformational changes between the two steps of splicing[20,22]. To test this model, a catalytic triad mutant (AGC→GAU) inactive for splicing was crystallized and solved at 7 Å (Extended Data Fig. 9). Strong electron density was observed for the η-η’ interaction between DII and DVI, indicating that this contact persists throughout both steps of splicing and that there is no large-scale change in DVI position. Therefore, a new model is required to explain how DVI mediates the transition between the two steps of splicing.

Extended Data Figure 9

2Fo-Fc density for DVI in the pre-catalytic structure contoured at 1σ. The η-η’ interaction persists throughout the splicing reaction and is visible in the pre-catalytic state. The weaker density for the central region of DVI suggests a partially disordered, dynamic region with possible helical remodeling in the conserved internal loop during splicing. The general pattern of side-by-side packing of domains II and VI persists between the two steps. Catalytic triad mutation consisted of an AGC→GAU substitution.

During the first step, the bulged adenosine must be in close proximity to M1, M2, and the 5′ splice site in order to engage in nucleophilic attack. However, π-π’ places the bulged adenosine ~20 Å from the active site in the post-catalytic structure. Mutagenesis of π-π’ also has no significant effect on the first step of splicing (Fig. 3b), and is unlikely to be engaged for lariat formation. Furthermore, DII is likely to remain largely stationary during catalysis due to the strong anchoring effect of multiple tetraloop receptor interactions with this domain. Based on these observations, we propose that π-π’ is a dynamic interaction that toggles DVI between two different states to mediate the transition between the first and second steps of splicing (Fig. 5). In the first step, the bulged adenosine is engaged in the active site for nucleophilic attack at the 5′ splice site. At this stage, π-π’ exists in the “off” state where DII is disengaged from the base of DVI. The DVI helix would also presumably exist in a relaxed conformation due to the lack of constraint provided by π-π’. Following lariat formation, DVI likely engages in remodeling of its central internal loop adjacent to A615, causing helical compression to turn “on” the π-π’ interaction; thus sequestering the bulged adenosine away from the active site. A second possible model is that the base pairs between the G6 sequence (residues 588 to 593) and a pyrimidine-rich tract (612 to 614; 616 to 618) at the proximal side of the DVI stem rearrange to reposition the lariat phosphate and engage π-π’. Both models serve to empty the active site of the 5′ end and allow entry of the 3′ splice site, which is directly attached to the end of DVI. In fact, the primary function of the lariat may be to covalently attach to the 5′ end to provide an attachment point for this pulling action.

Figure 5

Model for DVI as the conformational switch for splicing. π-π’ mediates the transition between the two steps of catalysis. See text for details.

Evolutionary implications

This bulged adenosine of DVI is analogous to the branch site adenosine in spliceosomal introns. The branch site sequence UACUAAC (nucleophilic adenosine underlined) pairs to the U2 snRNA to extrude the adenosine from the helix[26] as in DVI. Due to the mechanistic and evolutionary similarities with group II introns, it is likely that the branch site adenosine residue and the 5′ end of nuclear introns will adopt a similar spatial arrangement relative to the catalytic core of the spliceosome. We predict that the spliceosomal branch site adenosine will be sequestered after the first step of splicing with an RNA and/or protein contact analogous to the π-π’ interaction in P.li.LSUI2. In fact, an interaction observed between a region just downstream of the branch site and the U5 snRNA loop may be the spliceosomal counterpart of π-π’[27].

The P.li.LSUI2 structure provides a rationale for the phylogenetic conservation of the 5′ end. Group II introns have the consensus sequence GUGYG (Y=U or C)[11], which is similar to the GUAYG and GURAG (R=G or A) sequences found at the 5′ ends of yeast and mammalian introns, respectively[28]. In the spliceosome, the U6 snRNA pairs with the 5′ end[29] in an analogous manner to the ε-ε’ interaction in group II introns. Therefore, these conserved sequences likely form a similar metal-binding platform in spliceosomal introns, with the 5′ end forming a highly distorted backbone to coordinate magnesium ions that orient the splice site in the core.

In regards to the evolutionary rationale for the 2′-5′ phosphodiester bond, it is known that the lariat is required for full reversibility of group II intron splicing[30]. These reverse splicing reactions are the mechanism through which group II introns are able to invade DNA[31] and disperse throughout genomes. It is likely that the lariat bond pre-organizes the core of the intron structure to facilitate this reversibility. Reverse splicing has also been proposed as a vehicle for the prolific expansion of nuclear introns[32], which comprise ~25% of mammalian genomes. There is biochemical support for this hypothesis in that the spliceosome has recently been shown to catalyze reverse splicing reactions[33]. The lariat is likely to play a similar role in spliceosomal introns by allowing reverse splicing to occur, thus accounting for its phylogenetic conservation across the kingdoms by enabling “selfish introns” to replicate. Therefore, the P.li.LSUI2 crystal structure has provided the first glimpse of the branched lariat linkage that was likely crucial for intron proliferation in eukaryotes.

Methods

Cloning and preparation of Pylaiella littoralis LSU I2 intron RNA

DNA was synthesized (Genscript) corresponding to the second intron interrupting the large ribosomal subunit (LSU) in the mitochondria of the brown alga Pylaiella littoralis (P.li.LSUI2). The crystallization construct contains a 15-nt 5′ exon and a 5-nt 3′ exon followed by a Hind III restriction site. The DIV open reading frame (ORF) was removed from the P.li.LSU2 intron and replaced with a UUCG stem tetraloop. This was cloned into the EcoRV site of pUC57. Plasmid was linearized by HindIII digestion prior to in vitro transcription. Non-conserved loops in domain 1, 2 and 4 were changed from the wild type sequence. Most significantly, changing the sequence UAUUUAUA to UCGACAUAAGG in the ID2 stem loop improved both crystallization rate and diffraction. The final construct retained wild-type splicing activity. Transcription was performed overnight at 37°C using T7 RNA polymerase in 25 mM MgCl2, 2 mM spermidine, 5 mM DTT, 40 mM Tris-HCl pH 7.5, 0.05% Triton X-100, 2.5 mM of each NTP, and thermostable inorganic pyrophosphatase (New England Biolabs). CaCl2 was added to a final concentration of 1.2 mM, treated with DNase I for 45 minutes, followed by proteinase K digestion for 1 hour. The intron reacted to completion during in vitro transcription and was subjected to a native purification procedure previously used in the structure determination of the O. iheyensis intron[9]. Spliced intron RNA was repeatedly washed with 10 mM MgCl2 and 5 mM sodium cacodylate pH 6.5 and concentrated to 10 mg/ml using a 100 kDa molecular weight cut-off Amicon Ultra-15 column.

Crystallization

The native crystals were grown in sitting drops by vapor diffusion at 30°C. Equal volumes of RNA (10 mg/ml) were mixed with 0.4 mM spermine, 21% 2-Methyl-2,4-pentanediol (MPD), 175 mM magnesium acetate tetrahydrate, and 90 mM MES monohydrate (pH 5.6). Rod-like crystals appeared within two days, and grew to a maximum size of 50×50×900 μm. Crystals were gradually exchanged into 21% MPD, 100 mM magnesium acetate tetrahydrate, 50 mM MES monohydrate (pH 5.6), 3 mM iridium hexammine, 0.5 mM spermine, and 100 mM NaCl; followed by flash freezing in liquid nitrogen. Iridium hexammine was used as an additive for the native crystals as it reduced mosaicity. Pre-catalytic intron RNA was obtained by mutating the AGC catalytic triad of DV to GAU. The mutation maintained stem pairing but completely inhibited splicing activity. Crystallization of this mutant was done by microseeding under conditions described above. The ι G79A mutant crystallized in 0.2 M ammonium acetate, 10 mM calcium chloride, 50 mM sodium cacodylate pH 6.5, and 10% w/v polyethylene glycol 4000. Resulting crystals were cryoprotected through gradual transfer to the crystallization solution supplemented with 30% ethylene glycol.

Structure determination

The crystal structure of the P.li.LSUI2 intron was solved at 3.7 Å resolution using multi-wavelength anomalous dispersion (MAD) with crystals soaked in 0.5 mM ytterbium (III) chloride. Yb3+ soaks were performed in 21% MPD, 100 mM magnesium acetate tetrahydrate, 50 mM MES monohydrate (pH 5.6), 0.5 mM spermine, 100 mM NaCl, and 0.5 mM YbCl3 for 3 hours at room temperature. Tl+ soaks were performed in the same manner with 10 mM thallium acetate. X-ray data sets were collected at NE-CAT’s 24-ID-C beamline at the Advanced Photon Source (Argonne National Laboratory, Argonne, Illinois). Data was processed using HKL-2000[34], heavy metal sites were identified with SHELXD[35] and phasing done using Phenix[36] and SHELXE[37]. RNA nucleotides were modeled using COOT[38] and the RCrane plugin[39]. The phylogenetically predicted secondary structure (Extended Data Fig. 1) guided modeling into the electron density. Structure refinement was done using Buster[40], Phenix[36], DEN[41], and Phenix.Erasser[42]. The 2′-5′ phosphodiester bond was restrained using a cif restraint file in phenix.refine. All software was compiled by SBGrid[43].

In vitro self-splicing assays

The construct used for the in vitro self-splicing assays contained wild-type P.li.LSUI2 sequence with DIV ORF removed and a 250-nt 5′ exon and 75-nt 3′ exon. This was cloned into the pUC57 plasmid. Plasmid was linearized using HindIII and used for in vitro transcription with T7 RNA polymerase. Radiolabeled transcripts were prepared as above using 10 μCi [α-32P] UTP (3000 Ci/mmole), 0.5 mM UTP, 1 mM other NTPs, and 10mM MgCl2. Transcripts were gel purified on a 4% polyacrylamide (19:1)/8 M urea gel, RNA was recovered by diffusion into 300 mM NaCl, 0.01% SDS, 1 mM EDTA. Self-splicing experiments were performed for 30 min at 45°C in a splicing buffer containing 10 mM MgCl2, 1M NH4Cl, 40 mM Tris-HCl (pH 7.5), and 0.02% SDS. Reactions were stopped by addition of EDTA to a final concentration of 20 mM. Splicing products were resolved using a denaturing 4% polyacrylamide (19:1)/8 M urea gels. Rate constants for the WT and G79A mutant were derived from curves fit to a biphasic exponential equation. All splicing assays were done in triplicate.