Crystal structure of Pistol, a class ofself-cleaving ribozymeLaura A. Nguyena, Jimin Wanga,1, and Thomas A. Steitza,b,c,1

aDepartment of Molecular Biochemistry and Biophysics, Yale University, New Haven, CT 06520; bHoward Hughes Medical Institute, Yale University, NewHaven, CT 06520; and cDepartment of Chemistry, Yale University, New Haven, CT 06520-8107

Edited by Jennifer A. Doudna, University of California, Berkeley, CA, and approved December 19, 2016 (received for review July 8, 2016)

Small self-cleaving ribozymes have been discovered in all evolution-ary domains of life. They can catalyze site-specific RNA cleavage, andas a result, they have relevance in gene regulation. Comparativegenomic analysis has led to the discovery of a new class of small self-cleaving ribozymes named Pistol. We report the crystal structure ofPistol at 2.97-Å resolution. Our results suggest that the Pistol ribo-zyme self-cleavage mechanism likely uses a guanine base in the ac-tive site pocket to carry out the phosphoester transfer reaction. Theguanine G40 is in close proximity to serve as the general base foractivating the nucleophile by deprotonating the 2′-hydroxyl to ini-tiate the reaction (phosphoester transfer). Furthermore, G40 can alsoestablish hydrogen bonding interactions with the nonbridging oxy-gen of the scissile phosphate. The proximity of G32 to the O5′ leav-ing group suggests that G32 may putatively serve as the generalacid. The RNA structure of Pistol also contains A-minor interactions,which seem to be important to maintain its tertiary structure andcompact fold. Our findings expand the repertoire of ribozyme struc-tures and highlight the conserved evolutionary mechanism used byribozymes for catalysis.

X-ray crystallography | ribozyme | self-cleavage |internal transesterification | A-minor interaction

The “RNA world” hypothesis speculates that RNA carriedout the majority of biochemical reactions before the evo-

lution of complex protein enzymes (1, 2). Ribozymes are non-coding RNA that carry out catalytic activities. Unlike proteinenzymes, only a handful of ribozymes have known biologicalfunctions. Their biological functions range from regulating geneexpression (e.g., riboswitches) and performing peptidyl-transferreactions (e.g., ribosome) to removing intron sequences in genes(e.g., self-splicing Group I intron ribozymes) (2–9). The biologicalfunctions and mechanism of these ribozymes have been discoveredthrough structural and biochemical studies.Currently, the classes of self-cleaving ribozymes consist of Ham-

merhead, Hairpin, Hepatitis Delta Virus ribozyme (HDV), VarkudSatellite (VS), glmS ribozyme, Twister, Twister sister, Hatchet, andPistol (10). These classes differ based on their size, structure, andcleavage mechanism. Known for their site-specific cleavage, ribo-zymes with defined biological function include the Hammerhead,VS, and HDV, which all participate in rolling circle replication,whereas the glmS ribozyme functions in controlling gene expression(11, 12). However, the biological function of a vast majority of thedifferent self-cleaving ribozymes remains to be explored.Through comparative genome analysis, there have been three

newly identified classes of self-cleaving ribozymes called Twistersister, Hatchet, and Pistol (3). Biochemical analysis reveals thatPistol can use a variety of divalent metal ions to carry out a com-plete site-specific, self-cleaving reaction, whereas utilization ofmonovalent cations results in modest cleavage rates (3, 10). Therate of Pistol self-cleavage has been estimated to be >10 min−1

under physiological conditions and >100 min−1 under optimalmagnesium and pH conditions (10). Pistol self-cleavage is via aninternal transesterification reaction, in which the substrate RNA ofPistol G10 2′-hydroxyl on the ribose makes a nucleophillic attack onthe adjacent 3′-phosphate (Fig. 1B) (3, 10). Self-cleaving ribozymes

can enhance the rates of the internal transesterification reaction byusing catalytic strategies, such as deprotonation of the 2′-hydroxylgroup and neutralizing the negative charge on the nonbridgingoxygen of the scissile phosphate or 5′-oxygen of the cleaved sub-strate (10, 13–15).We report the crystal structure of Pistol at 2.97-Å resolution. Our

structure reveals the nucleobases that are likely to be involved in theinternal transesterification reaction of Pistol. The structure validatesprior biochemical results of the Pistol self-cleavage mechanism andfurther elucidates additional mechanistic details that cannot be easilyaddressed with biochemical analysis (10). The structure shows thatPistol adopts an overall compact fold stabilized by the A-minor motifcommonly found in many RNA structures, which explains the highsequence conservation for the stretch of adenines found in Pistol.The overall fold and cleavage mechanism of Pistol shares similarfeatures with other self-cleaving ribozymes, such as the presence of apseudoknot fold and the proposed use of guanosine as a generalbase, highlighting their conserved evolutionary mechanism.

Results and DiscussionStructure Determination. The Pistol RNA construct used for crys-tallization was derived from an extensive comparative genomicanalysis of the previously identified environmental sample 27,likely of bacterial origin (3, 16). The bimolecular Pistol RNAconstruct contains two RNA strands annealed together: one beingthe enzyme strand and the other being a substrate strand (Fig. 1)(3). The substrate strand contains the Pistol self-cleavage site,which is positioned between guanosine 10 (G10) and uridine 11(U11) (Fig. 1B). To trap the Pistol ribozyme in its precatalyticstate for crystallization, we generated a Pistol RNA substrate

Significance

Based on the “RNA world” theory, ribozymes likely carried outbiochemical reactions long before organisms evolved to useprotein enzymes as biocatalysts. The continued discovery of newstructures for small self-cleaving ribozymes has shed light onconserved mechanisms in evolution, such as acid–base catalysisfor self-cleavage reaction. Here, we present the crystal structureof a newly discovered class of self-cleaving ribozymes calledPistol and how it likely uses the phosphoester transfer mecha-nism for self-cleavage. The results presented here suggest thatPistol uses an evolutionarily conserved cleavage mechanism thatis like other self-cleaving ribozymes, such as Twister, Hammer-head, Hairpin, and Hepatitis Delta Virus ribozymes.

Author contributions: L.A.N. designed research; L.A.N. performed research; T.A.S. contrib-uted new reagents/analytic tools; L.A.N., J.W., and T.A.S. analyzed data; and L.A.N. and J.W.wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Data deposition: The crystallography, atomic coordinates, and structure factors have beendeposited in the Protein Data Bank, www.wwpdb.org (PDB ID code 5KTJ).1To whom correspondence may be addressed. Email: jimin.wang@yale.edu or thomas.steitz@yale.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1611191114/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1611191114 PNAS | January 31, 2017 | vol. 114 | no. 5 | 1021–1026

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Nov

embe

r 12

, 202

0

strand with a single-nucleotide mutation at the site of cleavage[from a guanosine to a deoxyguanosine (dG)]. With the loss of its2′-hydroxyl, G10 can no longer serve as a nucleophile for Pistol tocarry out its phosphoester transfer reaction.The Pistol crystals were grown as described in Materials and

Methods. Experimental phases were determined using the multi-wavelength anomalous dispersion (MAD) method. X-ray diffrac-tion datasets were collected at inflection wavelength, peakwavelength, and high-energy remote wavelength from Pistol crys-tals that had been soaked in samarium. The model was refined at2.97-Å resolution to a free R factor of 25.7% (Tables S1 and S2).The crystal form belongs to the I4122 space group (a = b = 85.19 Å,c = 256.54 Å, α = β = γ = 90°), with two molecules per asymmetricunit. Our experimental map shows the electron density for a por-tion of the active site along with two cobalt hexammine ions presentin the crystallization condition (Fig. S1A). A comparison of theexperimental map with our model phase map is also provided (Fig.S1 B and C).

Overview of the Pistol Structure. The Pistol structure has threehelical stems and one pseudoknot. The three stems contain Wat-son–Crick base pairs labeled P1–P3 (Fig. 1), which are connectedby three loops (Fig. 1A). The pseudoknot forms a stacking in-teraction with the P1 stem. The scissile phosphate that belongs tonucleotide U11 on the substrate strand is shown by the bluesphere, and the oxygen is shown in red (Fig. 1A). The secondarystructure of the Pistol RNA was generated based on our crystalstructure, which includes additional base interactions not pre-viously described (Fig. 1B) (3). The arrows in the secondarystructure indicate the direction of the strands from 5′ to 3′ anddepict where the helices and loops connect (Fig. 1B).We found that the coaxially stacked stem P1/pseudoknot helix

has a total of 11 bp, which is about one turn of a helix with aconformation that resembles an A-form RNA duplex (Fig. 1A).Notably, we observed that the pseudoknot contains 6 Watson–Crick bp instead of 5 bp as predicted previously (10). The pseu-doknot formation in our structure positions the active site of thePistol in close proximity with the P1 stem, which originally wasthought to be at a considerable distance away from the catalyticcenter (3, 10). Based on our structure, mutations that disruptWatson–Crick base pairs in the P1 stem or pseudoknot would likelydestabilize the pseudoknot formation and could change the overall

fold of Pistol necessary to form the active site (Figs. 1A and 2A). Inagreement with the biochemistry results, our findings illuminatehow two previously described mutations of nucleotides C2 and U3in the P1 stem of Pistol (Fig. 1B) yielded a ribozyme with reducedcleavage efficiency (10).Based on our structure, we report features of Pistol that were not

originally observed in prior secondary structural predictions. Onefeature is that the P1 and P2 stems are comprised of 5 Watson–-Crick bp instead of 4 bp (Fig. 1B) (10). The P2 stem has anadditional U-U wobble base pair between U29 of the enzymestrand and the U11 cleavage site on the substrate strand. Thepresence of this wobble U-U base pair may help to properlyorient the scissile phosphate for the cleavage reaction andcould also provide additional stabilization of the U11 base afterthe cleavage reaction (Fig. 1B).Another feature observed in our structure is that Loop 1 con-

tains three highly conserved adenosine nucleobases that form anA-minor interaction with the P1 stem (Figs. 1A and 4A). We foundthat these three adenosine nucleobases fix the P2 stem in theproper spatial geometry for the active site formation. The P2 stemis after Loop 1 and followed by Loop 2, which flanks the pseu-doknot along with Loop 3. Together, Loops 2 and 3 form the activesite of Pistol (Fig. 2A), which explains the high sequence conser-vation of nucleotides found in these loops. After the active site, theP3 stem forms 9 Watson–Crick bp and is positioned adjacent to theP2 stem and Loop 3. Finally, the sharp turn in the substrate strandbetween the P2 and P3 stems exposes the scissile phosphate on thesubstrate strand (Fig. 1A).

Active Site of Pistol and Catalytic Mechanism. The active site isformed by Loops 2 and 3 (Fig. 2A), which have high sequenceconservation. Loop 2 interconnects the P2 stem to the pseudoknot(Fig. 1A), and Loop 3 interconnects the pseudoknot to the P3stem (Fig. 1B). The pseudoknot interaction assists Loops 2 and 3in forming part of the active site geometry, thereby highlightingthe importance of the pseudoknot formation and its role in ca-talysis (Fig. 2A).The local environment surrounding the active site of Pistol re-

veals that the O6 functional group of the unpaired G42 base isinvolved in bifurcated hydrogen-bond (H-bond) interactions withthe NH2 group of dG10 and G40 (Fig. 2B). These interactionslikely help orient the three unpaired nucleobases within the activesite. Our structure of Pistol is a trapped precatalytic complexcontaining a noncleavable dG10 mutated base instead of a G10 inthe substrate strand. We have modeled a 2′OH group (red sphere)on the dG10 nucleotide (Fig. 2 B and D). In our structure, thedG10 nucleotide is stabilized within the active site of Pistol byforming a stable π-stacking interaction with the G40 base. Nucle-otide dG10 is further stabilized by the H bonds with the 2′OH ofG40 and a cobalt hexammine ion within its vicinity. Overall, theseinteractions may serve to place putatively important nucleobases inposition for catalysis.Based on the modeling of the 2′OH group on the dG10 (Fig.

2B), we propose that atom N1 of G40 is the likely candidate todeprotonate the 2′OH of G10 on the substrate strand. The N1atom of the G40 is within 2.6 Å of the modeled 2′OH of the dG10,and the 2′OH group is located at a 126° angle relative to the scissilephosphate and O5′. The deprotonation of the 2′OH of G10 willactivate the nucleophile to carry out the nucleophilic attack on thescissile phosphate. The N1 functional group of guanosine has a pKavalue of 9.2 (5, 17). For the G40 to serve as a general base todeprotonate the 2′OH of G10, the N1 functional group of G40needs to be deprotonated by ionization or tautomerization.The active site is further stabilized by nucleotide G33 located in

Loop 2, forming a base triple interaction with bases of G40 andC41 in Loop 3 (Fig. 2C and Fig. S1D). The base triple interactionhelps bring the bases of Loop 3 in range for catalysis. A cartoonrepresentation of the internal transesterification site-specific

Fig. 1. Overview of the Pistol ribozyme structure. (A) A standard view ofthe Pistol ribozyme is shown in the ribbon diagram. The structural domainsare colored as follows: P1 is in green, P2 is in blue, P3 is in gray, pseudoknot isin magenta, and three loops are in orange. The arrows indicate the 5′ to 3′direction. (B) A secondary structure of Pistol describes the interaction be-tween the enzyme strand from nucleotides 1–51 and the substrate strandnucleotides 1–15. The cleavage site of Pistol is between G10 and U11 as in-dicated by the arrow. Our substrate strand contains a noncleavable modifieddG base not depicted in this secondary structure. The base numberings arein black.

1022 | www.pnas.org/cgi/doi/10.1073/pnas.1611191114 Nguyen et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 12

, 202

0

self-cleavage reaction is shown in Fig. 2E. The end product of thePistol self-cleavage reaction in this model will be a 5′ cleavageproduct containing a 2′,3′-cyclic phosphate and a 3′-cleaved sub-strate with a 5′OH (Fig. 2E) (10).We observed that the N2 atom of G40 in Loop 3 and a cobalt

hexammine ion are within H-bonding distance with the non-bridging oxygen of the scissile phosphate (Fig. S2), which maystabilize the charge of the oxygen during catalysis. Biochemicalstudies have shown that mutating G40 and C41, which are highlyconserved residues in Loop 3, renders Pistol incapable of un-dergoing self-cleavage (10).Previous biochemical analysis using phosphorothioate substitution

at the nonbridging oxygen of the scissile phosphate has exploredwhether divalent metal ions play a major role for inner-spherecontact with the nonbridging oxygen of the scissile phosphate during

catalysis. It is known that divalent manganese ions can bind sulfurwith the same affinity as oxygen. Assuming that a phosphorothioatesubstitution does not affect the folding of Pistol, this result wouldimply that the Pistol self-cleavage rate should not be reduced if di-valent metal ions were crucial for interacting with the nonbridgingoxygen of the scissile phosphate. Pistol can self-cleave 90% of itssubstrate under manganese conditions (10). Previous analyses haveshown that manganese failed to rescue the phosphorothioate sub-stitution at the nonbridging oxygen of the scissile phosphate, whichsuggests that there are no inner-sphere metal ion to nonbridgingoxygen contacts at the scissile phosphate (10). The absence of ametal ion having inner-sphere contact with the nonbridging oxygenof the scissile phosphate in our structure supports the biochemicalresults. Our structural data support that a nucleobase interacts witha nonbridging oxygen at the cleavage site.

Fig. 2. The Pistol active site. (A) An overview of the Pistol active site formation is made by Loops 2 and 3 (orange). The pseudoknot (magenta) assists withpositioning Loops 2 and 3 in close proximity to the substrate strand (gray) for active site formation. The scissile phosphate is the cyan sphere, and thenonbridging oxygens are the red spheres. For B–D, the oxygen atoms are red, and the amine groups are blue. (B) A detailed view of the Pistol active site withthe substrate strand in gray and the enzyme strand in orange. A 2′OH (red sphere) was modeled into the dG10 base on the substrate strand. The angleextending from the 2′OH (scissile phosphate) to the O5′ is 126°. A cobalt hexammine is near the active site and can form an H bond with the O6 atom of dG10(cyan sphere). (C) A base triple interaction between the nucleobases in Loops 2 and 3. (D) A detailed view of the Loop 2 nucleobases, and their interactionwith the Pistol substrate strand is shown in gray. A 2′OH (red sphere) was modeled in the dG10 base on the substrate strand. (E) A cartoon representation ofthe Pistol self-cleavage reaction. The proposed general base G40 N1 atom is in the deprotonated state, in which it abstracts a proton from the 2′OH of theG10. The 2′O nucleophile is activated and makes a nucleophilic attack on the scissile phosphate. The final cleaved product is a 5′ cleavage product containing a2′,3′-cyclic phosphate and a 3′-cleaved substrate with a 5′OH. The black dashed lines represent the H bonds, and the distances between the bonds are labelednext to the dashed lines. ‡Transition state.

Nguyen et al. PNAS | January 31, 2017 | vol. 114 | no. 5 | 1023

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Nov

embe

r 12

, 202

0

Furthermore, the presence of a sulfur group in place of a non-bridging oxygen may cause a clash with the N2 functional group ofG40 as observed in Fig. 2D, because sulfur has a larger atomicradius than oxygen (18). This sulfur atom might perturb the posi-tioning of G40 in the active site, which might explain why a phos-phorothioate substitution resulted in Pistol having no rescue inenzymatic activity in the presence of manganese or magnesium (10).Another rationale may be that the phosphorothioate substitutioncaused a change in the charge distribution on the scissile phosphate,therefore affecting the H-bonding network surrounding the scissilephosphate (18, 19). These factors still need additional investigationto establish whether the G40 interaction with the nonbridging ox-ygen of the scissile phosphate is crucial for Pistol activity.Additional analysis of Loop 2, which also forms the active site of

Pistol, reveals that the N7 atom of G32 is within H-bonding dis-tance of U30 2′OH. The G32 2′OH also forms an H bond with theG42 base from Loop 3, yet another interaction that may help tostabilize the positioning of the unpaired G32. Although the H-bonddistance of the G32 N2 atom is 4.3 Å from the O5′ and the posi-tioning of the base does not support perfect H-bond geometry, theG32 is the closest base to the O5′ leaving group (Fig. 2D).Therefore, we propose that G32 could potentially play a role inneutralizing the charge of the O5′ during catalysis.We have compared the active site of our independently built

experimental model with that of another Pistol ribozyme variantrecently published (20). The superposition of the active sites yieldsan rmsd = 0.494 Å (Fig. S3 A and C) (20), indicating that theoverall structure and active site are similar and within experimentalerror. The main difference between the two active site conforma-tions is that nucleotide 32 is a guanine in our structure, whereas it isan adenosine in the structure of Pistol Env.25 reported by Renet al. (20). It is known from comparative genomic analysis of Pistolthat the base at this position is highly conserved as a purine (3).Based on the recent biochemistry results, it has been suggested thatthe purine base at position 32 is the likely candidate to serve as thegeneral acid for donating a hydrogen to the leaving O5′. Muta-genesis studies have shown that the Pistol self-cleavage activity is100% in 10 min with an A32G mutation, whereas activity wasabolished when the A32 N3 atom was substituted by a carbon atom(20). Furthermore, the pKa of the A32 N3 atom was found to beshifted up by 1 pH unit to 4.7, which suggests that the A32 N3 atomhas an increased chance of being protonated, even under physio-logical pH (20). Our structure shows that the G32 N3 atom iswithin the same position as the A32 N3 atom, being positioned 4 Åaway from the O5′ leaving group of nucleotide U11 (Fig. 2D andFig. S3C). Additional studies will be necessary to verify whether theG32 N3 behaves similarly.Another peculiar aspect of the mutagenesis study showed that an

A32U mutation in Pistol exhibited a 50% cleavage activity, whereasan A32C abolished activity (20). This result raises an importantquestion on the role of nucleobase 32 in the Pistol O5′ leaving groupstabilization. Mutation of the nucleobase involved in general acid–base catalysis in other self-cleaving ribozymes causes significantlyreduced activity. For example, in HDV, VS, and Hairpin ribozymes,nucleobase mutation on the proposed general acid severely reducestheir activity (21–23). Additional experiments, such as the sub-stitution of the bridging O5′ atom of U11 to a 5′-phosphorothiolate,which serves as a better leaving group than oxygen and does notrequire protonation from the general acid, will be helpful to clarifythe role of A32U and A32C mutations in Pistol catalysis (21–24).The G32 N1 in our structure is about 5 Å away from the O5′

atom of U11. A comparison of the two Pistol structures indicatesthat a local conformational change needs to take place within theactive site to bring the G32 or A32 within closer range for the Hbond to form with the O5′ of U11 (20).The substrate used in this study contains a 2′-deoxy nucleotide

dG10 that is lacking the attacking 2′OH. The G10 to dG10 re-placement in our design does not seem to alter the folding of the

RNA structure, because the scissile phosphate observed between G10and U11 is in a splayed geometry similar to other small self-cleavingribozymes, including Twister, HDV, VS, and Hairpin (9, 12, 14, 25).Although the sugar puckering in dG10 might be different from G10,the modeled location of the 2′OH is a good approximation. When

Fig. 3. The proposed mechanism of self-cleavage across small self-cleavingribozymes. The following color schemes apply here. The H bonds between thenucleotide functional groups and the atoms are shown with black dashed lines(blue indicates amine, red indicates oxygen, and cyan indicates scissile phosphate,except in B). The substrate strands are colored gray, and the enzyme strands arecolored orange. (A) The Pistol active site is in the precatalytic state [Protein DataBank (PDB) ID code 5KTJ]. The G40 is the proposed general base, and the G32 iswithin the vicinity to form H bonds with the O5′. (B) The Hairpin active sitestructure was obtained from PDB ID code 1M5O, showing A − 1 as the adenosineupstream of the cleavage site and G + 1 as the downstream leaving group aftercleavage. Vanadate (cyan) was used in place of a scissile phosphate to trap thistransition state. Stabilization of the Hairpin transition state is accomplished byneutralizing the oxygen negative charge via H bonding with the nucleobases inthe active site (14). (C) A portion of the VS active site is shown (PDB ID code 4R4P)(12). G638 can serve as a general base in this self-cleavage reaction and stabilizethe charge of the scissile phosphate nonbridging oxygen.

1024 | www.pnas.org/cgi/doi/10.1073/pnas.1611191114 Nguyen et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 12

, 202

0

searching for a possible general base for activating the 2′-hydroxyl, theN1 atom of G40 is a good candidate to be the H-bond acceptor afterit is deprotonated by ionization or tautomerization.The cobalt hexammine and G40 in the active site form

H-bonding interactions with the nonbridging oxygen of the scissilephosphate (Fig. 2B and Fig. S4). Similarly, under biological con-ditions, this pocket will likely fit a fully hydrated magnesium ion,which can form similar H-bond interactions as seen with the cobalthexammine. The cobalt hexammine ion also forms an H-bondnetwork with the bases and phosphates surrounding the active site(Fig. S4). From our structure, we cannot conclude whether divalentmetals are directly involved in catalysis, but when divalent metalsare present, they may play a supporting role along with the G40nucleobase in stabilizing the charges on the nonbridging oxygens ofthe scissile phosphate. However, the presence of the divalent metalbinding sites shows their contribution to forming the stabile overallstructural fold of Pistol (Fig. S5).

Comparison of Self-Cleavage Mechanisms Across Small, Self-CleavingRibozymes. Our structure suggests that Pistol likely uses two strate-gies for its self-cleavage mechanism. One requires the activation ofthe nucleophile via deprotonation of the 2′-hydroxyl group, and theother requires neutralization of the negative charge on the non-bridging oxygen of the scissile phosphate, important in stabilizing thetransition state (15). For comparison, two self-cleaving ribozymestructures, named Hairpin and VS ribozymes, were selected, becausethey were shown to likely use a guanosine as a general base for theself-cleavage reaction (Fig. 3) (12, 14, 22, 26). Other self-cleavingribozyme examples that have guanosine as the proposed generalbase not shown here include Twister and Hammerhead (27). In theactive site of Pistol, nucleobase G40 serves as the potential “generalbase” to carry out the deprotonation of the substrate strand G10 2′OH (the 2′OH is modeled as a red sphere in Fig. 3A). The nucle-obase G32 in our structure is positioned closest to the O5′ atom ofthe U11 nucleobase. A local conformation change will need to takeplace to bring the G32 within H-bond distance to the O5′ atom (20).To compare our Pistol structure with the Hairpin ribozyme active

site (Fig. 3B), the Hairpin was trapped in a “transition state” using avanadate in place of the scissile phosphate (14). The structural datasuggest that the nucleobase G8 can serve as the putative generalbase in the Hairpin, which potentially acts to deprotonate the 2′OHof the A − 1 ribose during the Hairpin self-cleavage (Fig. 3B) (14).Furthermore, an additional H bond from nucleobase G8 shows thatit could interact with the nonbridging oxygen of the scissile phos-phate. The structure indicates that the transition state stabilizationof Hairpin uses mainly nucleobase functional groups to carry out

Hairpin catalysis (14). Experimental data suggest that, before par-ticipating in stabilizing the transition state, G8 may serve as thegeneral base to execute Hairpin ribozyme catalysis (21).Based on experimental and structural data, the VS ribozyme

structure suggests that the self-cleavage mechanism likely usesnucleobase G638 (Fig. 3C) to serve as the general base in catalysis(Fig. 3C) (12). Similarly, this G638 base can also H bond with thenonbridging oxygen of the scissile phosphate. It seems that Pistol,Hairpin, and VS ribozyme may rely on the guanosine aminefunctional groups to enhance their self-cleavage rates throughdeprotonating the 2′OH group to activate the nucleophile for self-cleavage and stabilizing the charge on the nonbridging oxygen ofthe scissile phosphate through H bond. Additional studies, such assubstituting N1 for a carbon atom, will be necessary to confirm thatthe Pistol N1 atom of G40 is crucial in serving as a general base.Our structure provides evidence that small, self-cleaving ribozymesmay have evolved to use guanosine for similar catalytic strategies toenhance phosphoester transfer.

A-Minor Interaction and the Extensive Network of H Bonds. In additionto elucidating the self-cleavage mechanism of Pistol, we also ob-served features in the structure of Pistol. We found seven consec-utive base interactions between nucleotides of Loop 1 and theminor groove of the P1 stem (Fig. 4A). This triple-stranded struc-ture forms extensive A-minor interactions in our Pistol structure.The A-minor interactions with the P1 stem are involved in stabi-lizing the P2 stem in a specific conformation through extensiveH-bond formation with bases in the P1 stem (Fig. 4 and Fig. S3).The A-minor interactions involve a stretch of adenosines in ei-

ther a loop or a stem of RNA forming interactions with the minorgrove of a neighboring helix through an H-bond network. A-minorinteractions have been observed in various RNA and ribozymestructures and serve mainly in stabilizing tertiary interactions be-tween helices and loops (28–31). Similarly, the stretch of highlyconserved adenosines in Loop 1 of Pistol helps position the scissilephosphate adjacent to the P2 stem, which is important for catalysis.Nucleobases A19 and A20 form extensive bifurcated H-bond in-teractions with bases of nucleotides U3, C4, and A14 in the P1 stem(Fig. S3A). Nucleobase A21 displays similar A-minor H-bondingpatterns (Fig. S3B) that have previously been described in the 50Sribosomal subunit, where adenosines form different types of in-teractions with Watson–Crick base pairs in the same plane (29).Without these A-minor interactions, the strand that forms the P2stem will likely be more flexible, which may alter the Pistol catalysisrate or hinder the self-cleavage reaction, because the P2 stem isadjacent to the cleavage site (Fig. 2B).

Fig. 4. An overview of Loop 1 A-minor interaction. (A) The A-minor interaction of Loop 1 (orange) and the P1 stem (green) is depicted in this triple-strandedPistol structure. The conserved adenosine bases are labeled in black. The seven bases in Loop 1 form extensive contact with the P1 minor groove. (B) Adifferent view of the Loop 1 forming A-minor interaction with the P1 stem. This A-minor interaction from Loop 1 makes a tight turn to connect to the P2 stem(blue). The scissile phosphate (cyan sphere) and the nonbridging oxygen (red spheres) are adjacent to the P2 stem. A portion of the P3 stem (gray) is next tothe P2 stem. The base numberings are in black.

Nguyen et al. PNAS | January 31, 2017 | vol. 114 | no. 5 | 1025

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Nov

embe

r 12

, 202

0

ConclusionWe report that Pistol uses A-minor interactions to stabilize tertiarycontacts between the stems and loops as observed in other ribo-zymes. Based on our structure, we propose two catalytic strategiesthat Pistol likely uses during self-cleavage. The first includes depro-tonation of the G10 2′OH using the proposed G40 as the generalbase, and the second is neutralizing the charge of the nonbridgingoxygen from the scissile phosphate. These two mechanisms havebeen speculated in other small self-cleaving ribozyme structures, suchas VS and Hairpin ribozymes. We observed that the scissile phos-phate nonbridging oxygen interacts with the Watson–Crick edge ofnucleobase G40 through H-bond interactions with the N1 and N2functional groups. The G40 may stabilize the nonbridging oxygen ofthe scissile phosphate during the phosphoester transfer by H bond.Additional studies will be needed to determine whether this in-teraction is crucial for the catalytic activity of Pistol. The use of theguanosine functional groups in the active site via deprotonation ofthe 2′OH group and neutralization of the charge on the nonbridgingoxygen from the scissile phosphate during catalysis may be a dualstrategy of Pistol for enhancing self-cleavage rates.

Materials and MethodsThe bimolecular Pistol ribozyme construct was generated using Pistol environ-mental sequence number 27 obtained from previous reports (3). Details of theconstructs and crystallizationmethods are provided in SI Materials andMethods.

X-ray diffraction data collection was performed at the Advanced PhotonSource in the Argonne National Laboratory at beamline 24ID-C. MAD datasetswere collected at three different wavelengths for Samarium derivative.Samarium-derivative datasets were collected at wavelengths of 1.85 Å forinflection, 1.84 Å for peak, and 1.02 Å for high-energy remote. The datawere processed and scaled using the XDS software in the I4122 spacegroup (32).

Five datasets from a single samarium chloride-soaked crystal were used fordetermining experimental phases. The high-energy remote datasets that gavethe best resolution at 2.96 Åweremerged (Tables S1 and S2). Using the HKL2Minterface containing the SHELX suite, we determined the heavy-atom sub-structure from our MAD experiments with SHELXC (33). The location of theheavy atoms and the first initial experimental phase map were calculated withSHELXD and SHELXE, respectively (34). From this map, we were able to buildour model using COOT (35). After model building, Refmac from the CCP4i suitewas use for refinement (36–38). The final cross-validated Rfree after modelrefinement was 25.7% (Tables S1 and S2).

ACKNOWLEDGMENTS.We thank Prof. Ronald Breaker for his critique of andexpert advice on this manuscript and Dr. Matthieu Gagnon for his helpfulcritiques of this manuscript. We also thank Drs. Kim, Lomakin, Lin, andPolikanov for discussions regarding this project. We thank Joseph Watsonfor assisting with the RNA purification. Staffs at the Argonne NationalLaboratory (Northeastern Collaborative Access Team 24ID-C) have been ex-tremely helpful in facilitating X-ray data collection. This work was supportedby the Howard Hughes Medical Institute. L.A.N. was supported, in part, byNIH Ruth L. Kirschstein National Research Service Award Grant 5 F32AI114095-02.

1. Benner SA, Ellington AD, Tauer A (1989) Modern metabolism as a palimpsest of theRNA world. Proc Natl Acad Sci USA 86(18):7054–7058.

2. Steitz TA, Moore PB (2003) RNA, the first macromolecular catalyst: The ribosome is aribozyme. Trends Biochem Sci 28(8):411–418.

3. Weinberg Z, et al. (2015) New classes of self-cleaving ribozymes revealed by com-parative genomics analysis. Nat Chem Biol 11(8):606–610.

4. Roth A, et al. (2014) A widespread self-cleaving ribozyme class is revealed by bio-informatics. Nat Chem Biol 10(1):56–60.

5. Ferré-D’Amaré AR, Scott WG (2010) Small self-cleaving ribozymes. Cold Spring HarbPerspect Biol 2(10):a003574.

6. Adams PL, Stahley MR, Kosek AB, Wang J, Strobel SA (2004) Crystal structure of a self-splicing group I intron with both exons. Nature 430(6995):45–50.

7. Winkler WC, Nahvi A, Roth A, Collins JA, Breaker RR (2004) Control of gene expressionby a natural metabolite-responsive ribozyme. Nature 428(6980):281–286.

8. Cate JH, et al. (1996) Crystal structure of a group I ribozyme domain: Principles of RNApacking. Science 273(5282):1678–1685.

9. Ferré-D’Amaré AR, Zhou K, Doudna JA (1998) Crystal structure of a hepatitis deltavirus ribozyme. Nature 395(6702):567–574.

10. Harris KA, Lünse CE, Li S, Brewer KI, Breaker RR (2015) Biochemical analysis of pistolself-cleaving ribozymes. RNA 21(11):1852–1858.

11. Doherty EA, Doudna JA (2000) Ribozyme structures and mechanisms. Annu RevBiochem 69:597–615.

12. Suslov NB, et al. (2015) Crystal structure of the Varkud satellite ribozyme. Nat ChemBiol 11(11):840–846.

13. Emilsson GM, Nakamura S, Roth A, Breaker RR (2003) Ribozyme speed limits. RNA9(8):907–918.

14. Rupert PB, Massey AP, Sigurdsson ST, Ferré-D’Amaré AR (2002) Transition state sta-bilization by a catalytic RNA. Science 298(5597):1421–1424.

15. Breaker RR, et al. (2003) A common speed limit for RNA-cleaving ribozymes anddeoxyribozymes. RNA 9(8):949–957.

16. Weinberg Z, Perreault J, Meyer MM, Breaker RR (2009) Exceptional structured non-coding RNAs revealed by bacterial metagenome analysis. Nature 462(7273):656–659.

17. Singh V, Fedeles BI, Essigmann JM (2015) Role of tautomerism in RNA biochemistry.RNA 21(1):1–13.

18. Frey PA, Sammons RD (1985) Bond order and charge localization in nucleosidephosphorothioates. Science 228(4699):541–545.

19. Lupták A, Doudna JA (2004) Distinct sites of phosphorothioate substitution interfere withfolding and splicing of the Anabaena group I intron. Nucleic Acids Res 32(7):2272–2280.

20. Ren A, et al. (2016) Pistol ribozyme adopts a pseudoknot fold facilitating site-specificin-line cleavage. Nat Chem Biol 12(9):702–708.

21. Kath-Schorr S, et al. (2012) General acid-base catalysis mediated by nucleobases in thehairpin ribozyme. J Am Chem Soc 134(40):16717–16724.

22. Wilson TJ, et al. (2010) Nucleobase-mediated general acid-base catalysis in the Varkudsatellite ribozyme. Proc Natl Acad Sci USA 107(26):11751–11756.

23. Das SR, Piccirilli JA (2005) General acid catalysis by the hepatitis delta virus ribozyme.Nat Chem Biol 1(1):45–52.

24. Lilley DM (2011) Catalysis by the nucleolytic ribozymes. Biochem Soc Trans 39(2):641–646.

25. Eiler D, Wang J, Steitz TA (2014) Structural basis for the fast self-cleavage reactioncatalyzed by the twister ribozyme. Proc Natl Acad Sci USA 111(36):13028–13033.

26. Schmidt S, et al. (1996) Base and sugar requirements for RNA cleavage of essentialnucleoside residues in internal loop B of the hairpin ribozyme: Implications for sec-ondary structure. Nucleic Acids Res 24(4):573–581.

27. Jimenez RM, Polanco JA, Lupták A (2015) Chemistry and biology of self-cleaving ri-bozymes. Trends Biochem Sci 40(11):648–661.

28. Mitton-Fry RM, DeGregorio SJ, Wang J, Steitz TA, Steitz JA (2010) Poly(A) tail recognitionby a viral RNA element through assembly of a triple helix. Science 330(6008):1244–1247.

29. Nissen P, Ippolito JA, Ban N, Moore PB, Steitz TA (2001) RNA tertiary interactions in thelarge ribosomal subunit: The A-minor motif. Proc Natl Acad Sci USA 98(9):4899–4903.

30. Brown JA, et al. (2014) Structural insights into the stabilization of MALAT1 noncodingRNA by a bipartite triple helix. Nat Struct Mol Biol 21(7):633–640.

31. Martick M, Scott WG (2006) Tertiary contacts distant from the active site prime a ri-bozyme for catalysis. Cell 126(2):309–320.

32. Kabsch W (2010) Xds. Acta Crystallogr D Biol Crystallogr 66(Pt 2):125–132.33. Sheldrick GM (2008) A short history of SHELX. Acta Crystallogr A 64(Pt 1):112–122.34. Pape T, Schneider TR (2004) HKL2MAP: A graphical user interface for macromolecular

phasing with SHELX programs. J Appl Crystallogr 37:843–844.35. Emsley P, Cowtan K (2004) Coot: Model-building tools for molecular graphics. Acta

Crystallogr D Biol Crystallogr 60(Pt 12 Pt 1):2126–2132.36. Winn MD, et al. (2011) Overview of the CCP4 suite and current developments. Acta

Crystallogr D Biol Crystallogr 67(Pt 4):235–242.37. Potterton E, Briggs P, Turkenburg M, Dodson E (2003) A graphical user interface to

the CCP4 program suite. Acta Crystallogr D Biol Crystallogr 59(Pt 7):1131–1137.38. Skubak P, Murshudov GN, Pannu NS (2004) Direct incorporation of experimental

phase information in model refinement. Acta Crystallogr D Biol Crystallogr 60(Pt 12Pt 1):2196–2201.

1026 | www.pnas.org/cgi/doi/10.1073/pnas.1611191114 Nguyen et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 12

, 202

0