Visualizing the ai5γ group IIB intron

Abstract

It has become apparent that much of cellular metabolism is controlled by large well-folded noncoding RNA molecules. In addition to crystallographic approaches, computational methods are needed for visualizing the 3D structure of large RNAs. Here, we modeled the molecular structure of the ai5γ group IIB intron from yeast using the crystal structure of a bacterial group IIC homolog. This was accomplished by adapting strategies for homology and de novo modeling, and creating a new computational tool for RNA refinement. The resulting model was validated experimentally using a combination of structure-guided mutagenesis and RNA structure probing. The model provides major insights into the mechanism and regulation of splicing, such as the position of the branch-site before and after the second step of splicing, and the location of subdomains that control target specificity, underscoring the feasibility of modeling large functional RNA molecules.

Figure 1.

Schematic secondary structure of ai5γIIB intron. Domains 1–6 are labeled in capital letters, subdomains are labeled in small letters. Exons are colored in magenta. Exon and intron binding sites are labeled as EBS and IBS, respectively. Regions highlighted in light blue are common in ai5γIIB and OiIIC introns (core region); they were built with homology modeling using OiIIC crystal structure as the template. Additional insertions specific to ai5γIIB (highlighted in orange) were built de novo and docked on to the homology model. D6 (highlighted in green), which is missing in all available crystal structures of OiIIC, was also modeled de novo. Long-range tertiary interactions are labeled in red.

Figure 2.

Overview of model building. Modeling was performed in three steps, in the first step the core structure of the intron was generated with homology modeling using crystal structure from IIC as the template. Next, all additional regions specific to the IIB intron were modeled using MC-Sym and docked onto the core structure. Finally, the model was refined using RCrane (see text) and AMBER.

Figure 3.

Model Validation. (A) Correlation plot between numbers of contacts of each nucleotide in the model with hydroxyl radical reactivities. The correlation coefficient (r = −0.61) is as good as that observed for crystal structures of other RNAs such as group I introns, RNase P and riboswitches (47). (B) Difference plot of hydroxyl radical reactivities between WT D135 and β-β′ mutant. Nucleotides with significant increases or decreases (greater than one unit of standard deviation) in reactivities are colored in red. (C) D1c2 and D1d2a stems are shown as a transparent surface representation colored in yellow. Regions with significant increase in hydroxyl reactivities are colored in red. The rest of the intron is colored in light blue. Most of the regions with significant increases in reactivities lie adjacent to D1c2 and D1d2a, indicating that mutation of the kissing-loop interaction between D1c2 and D1d2a has exposed the regions beneath them and that the model is in good agreement with the experimental data.

Figure 4.

Structural architecture of ai5γIIB. (A) Cartoon representation of the model depicting the overall architecture (also see Supplementary Movie S1 and Supplementary Figure S4). All domains and subdomains are color-coded as shown in the insert. Locations of three additional interactions, β-β′, μ-μ′ and EBS2-IBS2 are labeled and colored the same as domains involved in that particular interaction. D1c2 (shown in blue) and D1d2a (shown in raspberry) are the two additional subdomains involved in the β-β′ kissing-loop interaction. (B) 90° rotated view of the image shown in (A). (C) 180° rotated view of the image shown in (A).

Figure 5.

Location of the EBS2 binding site. The EBS2-IBS2 duplex is shown in orange, the EBS1-IBS1 duplex in shown in magenta. The linker between IBS1 and IBS2 is shown in red. The rest of the intron is shown as a transparent surface representation. The EBS2-IBS2 interaction is located next to the kissing-loop (β-β′) interaction that joins D1c2 (green) and D1d2a (blue), suggesting that function of the two motifs may be linked.

Figure 6.

Active versus silent form of D6: (A) D6 can exist in two different conformations: the active (green) and silent forms (magenta). (B) In the conformation needed for the first step of splicing (the active form) the 2′-OH group of branch-site adenosine (green) is positioned in the active site to initiate the first step of splicing. Active-site elements (catalytic triad, 2-nt bulge and J2\3 junction) are shown in blue, 5′-exon is shown in yellow, first nucleotide of the intron G1 is shown in purple. (C) In the conformation that forms after the first step of splicing, and which is required for the second step (the silent-form), the branch-site adenosine flips out and positions the 3′-exon in the active site and forms two new interactions (shown in red) IBS3-EBS3 and γ-γ′. Based on the model, toggling between the ‘active’ and ‘silent’ conformations of D6 involves a simple rotation of the domain rather than a large translational movement away from the active site.

Figure 7.

Comparison of the ai5γIIB model with the crystal structure of OiIIC. (A) Front and side views of the ai5γIIB model. Regions common to both introns are colored in light blue; D5 is colored in red. Long-range interactions that are not present in OiIIC are labeled in Greek letters. D1c2 is shown in purple, D1d2a is shown in magenta, D3a and D3b are shown in orange. EBS2-IBS2 helix is shown in green. For comparison with OiIIC, D6 is not shown in both views. (B) Front and side views of OiIIC crystal structure.