In vitro evolution of coenzyme-independent variants from the glmS ribozyme structural scaffold

Abstract

Uniquely among known natural ribozymes that cleave RNA sequence-specifically, the glmS ribozyme-riboswitch employs a small molecule, glucosamine-6-phosphate (GlcN6P) as a catalytic cofactor. In vitro selection was employed to search for coenzyme-independent variants of this ribozyme. In addition to shedding light on the catalytic mechanism of the ribozyme, such variants could resemble the evolutionary ancestors of the modern, GlcN6P-regulated ribozyme-riboswitch. A mutant pool was constructed such that the secondary structure elements, which define the triply-pseudoknotted global fold of the ribozyme, was preserved. A stringent selection scheme that relies on thiol-mercury affinity chromatography for separating active and inactive sequences ultimately yielded a triple mutant with a cleavage rate exceeding 3 min⁻¹ that only requires divalent cations for activity. Mutational analysis demonstrated that a point reversion of the variant toward the wild-type sequence was sufficient to partially restore GlcN6P-dependence, suggesting that coenzyme dependence can be readily be acquired by RNAs that adopt the glmS ribozyme fold. The methods employed to perform this selection experiment are described in detail in this review.

1. Introduction

1.1 Background

The glmS ribozyme-riboswitch is a catalytic, gene-regulatory mRNA domain widespread in Gram-positive bacteria. It is the only natural RNA known to be both a riboswitch and a ribozyme (reviewed in [1, 2]). Riboswitches are mRNA elements that regulate gene expression in cis in response to direct binding to a cognate ligand [3–6]. The glmS gene encodes the protein enzyme glucosamine 6-phosphate (GlcN6P) synthetase, and the ribozyme-riboswitch lies in its 5′-untranslated region. The self-cleavage activity of the ribozyme-riboswitch is activated by binding to GlcN6P, and self-scission leads to degradation of the mRNA [7, 8]. In this manner, the glmS ribozyme performs negative-feedback regulation of the intracellular levels of GlcN6P, an essential precursor for the synthesis of the bacterial cell wall. Because disruption of glmS ribozyme function is deleterious in vivo in the model organism Bacillus subtilis, this catalytic RNA is a candidate antibiotic target, and its molecular mechanism of action has been studied intensively[2].

Activation of the glmS ribozyme by GlcN6P could result from either of two distinct mechanisms. The small molecule could function as an allosteric effector whose binding allows the RNA to achieve a catalytically competent conformation. Alternatively, GlcN6P could be a coenzyme, providing functional groups essential for chemical catalysis to an otherwise incomplete ribozyme active site. Biochemical and crystallographic studies support the latter mechanism. The amine of GlcN6P is positioned to protonate the 5′-OH leaving group of the internal transestrification through which the ribozyme cleaves RNA, thus functioning as a general acid catalyst [9, 10]. Binding to the ribozyme lowers the pK_a of the amine group, thereby improving its catalytic efficiency [11–13]. Further arguing against an allosteric mechanism, the ribozyme has been shown readily to adopt its active conformation under physiological Mg²⁺ concentrations [14, 15], and its rigid, prefolded structure is not perturbed by binding of GlcN6P, or cleavage of the substrate strand [9, 16–19]. The catalytic mechanism of the glmS ribozyme, which relies on binding of a small molecule coenzyme, is unique among all known ribozymes that cleave RNA through internal transestrification, and it suggests the possible involvement of catalytic cofactors in a primordial RNA world [20].

1.2 In vitro selection of glmS ribozyme variants

Several studies have employed in vitro selection to isolate sequence variants of the glmS ribozyme [21, 22]. These studies aimed to investigate the coenzyme selectivity of the ribozyme, to discover variants that could efficiently employ molecules other than GlcN6P as catalytic cofactors, and to explore possible evolutionary origins of this gene-regulatory RNA. Because GlcN6P is a metabolite present in all kingdoms of life, it is difficult to selectively activate the glmS ribozyme in a heterologous context. Identification of a glmS ribozyme variant that employs a coenzyme other than GlcN6P, and which is not activated by GlcN6P itself, would provide a ribozyme and coenzyme set that is orthognal [23] to the natural ribozyme, and thus be useful for synthetic biology applications. Moreover, characterization of such an orthogonal system may also shed light on the molecular basis of coenzyme selectivity by the wild-type ribozyme-riboswitch.

It is thought that complex biological functions arise step-wise, with each step providing a selective advantage [24]. Thus, in the case of the glmS ribozyme-riboswich, it is conceivable that an ancestral molecule existed that was either a ribozyme or a riboswitch, but not both [22]. Phylogenetic analyses have thus far failed to uncover possible molecular ancestors of the glmS ribozyme-riboswitch, but because none of the other known natural ribozymes that cleave RNA through internal transesterification employ a coenzyme, and because other riboswitches that bind sugar derivatives are unknown, a “ribozyme-first” scenario is more likely. That is, the molecular ancestor of this gene-regulatory RNA could have been an RNA-cleaving ribozyme that only subsequently acquired dependence on GlcN6P, and thus the ability to regulate cellular levels of this metabolite.

We recently employed in vitro selection to search for coenzyme-independent variants of the wild-type glmS ribozyme (hereafter glmS^WT) [22]. In order to preserve the overall structure of the ribozyme, only nucleotides in the catalytic core that do not participate in forming the characteristic triply-pseudoknotted glmS ribozyme fold were mutagenized. The resulting library of sequence variants was subjected to in vitro selection, allowing us to identify an RNA with three adenosine mutations (hereafter glmS^AAA) that is active in cleaving RNA in the absence of GlcN6P. This mutant RNA achieved cleavage rates as high as 3 min⁻¹ in the presence of the cations Ca²⁺ or Mg²⁺, but its activity was not enhanced by GlcN6P. SAXS and crystallographic analyses confirmed that glmS^AAA adopts the same fold as the wild-type ribozyme. The three adenosine mutations surround the scissile phosphate and abrogate RNA functional groups that coordinate GlcN6P in the wild-type. Biochemical characterization of point mutants of glmS^AAA demonstrated that a single nucleotide reversion to the wild-type sequence is sufficient to confer GlcN6P-dependence (neither glmS^AAA nor the various revertants have yet been tested in vivo). Thus, in addition to pinpointing the molecular requirements for GlcN6P utilization by the glmS ribozyme, this work revealed a possible path for the evolution of the bacterial ribozyme-riboswitch from an unregulated RNA-cleaving ribozyme that only requires divalent cations for full activity. Because the wild-type glmS ribozyme is very active (close to 100 min⁻¹) and can employ adventitious coenzymes (such as the buffer Tris) in the absence of GlcN6P [11], stringent conditions were needed for in vitro selection. The methods employed for the selection of glmS^AAA could be useful for other in vitro ribozyme evolution experiments and are detailed in this article.

2. Methodology

2.1 Mutagenized pool construction

A mutagenized DNA pool based on glmS^WT was designed to retain sequences forming the helical elements (secondary structure) of the RNA (Figure 1A). Nucleotides from the core of the ribozyme, along with those from loops and bulges (43 positions in total, Note 1), were each mutagenized by 30% relative to glmS^WT (Figure 1A, [22, 25]). Flanking the pool sequences are constant regions. The 5′ constant region contains a bacteriophage T7 RNA polymerase promoter, followed by a 40-nt leader (“substrate”) sequence (optimal length, Note 2). Following transcription, this 40-nt RNA leader can be cleaved in cis if its downstream sequence encodes an active ribozyme. The 3′ constant region contains a primer-binding site for reverse transcription and PCR. For our selection, the 205-nt single-stranded DNA pool was synthesized at the one-micromole scale using standard cyanoethyl phosphoramidite chemistry. For the synthesis, a molar ratio of 0.28:0.23:0.27:0.22 (dA:dG:dC:dT) was used to compensate for differences in coupling efficiencies [26, 27]. Following synthesis, ~60 random isolates from the DNA pool were sequenced (Note 3) to provide an estimate of the actual sequence diversity and mutation rate per position.

Figure 1. In vitro selection of a coenzyme-independent glmS ribozyme variant, glmS^AAA.

(A) Template design for the starting mutagenized pool based on the tertiary structure [9] of glmS^WT. Nucleotides in black upper-case were kept constant, and those in red lower-case were mutagenized by 30% relative to glmS^WT. The green arrow indicates position of the scissile phosphate of glmS^WT. Positions 49, 51 and 65 (dotted boxes) were found to be mutated to adenosines (cyan) in glmS^AAA. (B) In vitro selection scheme employing a thiol-mercury gel (APM-PAGE). Black dotted line, glmS^WT cleavage site; LD, leader sequence; T7, T7 RNA polymerase promoter sequence.

2.2 Synthesis of double-stranded DNA template pool

The single-stranded DNA pool (~5 nanomoles, ~3 × 10¹⁵ unique sequences) was converted to double-stranded and amplified in a 5 mL PCR reaction containing 20 mM Tris-HCl pH 8.4, 50 mM KCl, 0.2 mM of each dNTP, 1.5 mM MgCl₂, 2.5 μM of 5′ selection DNA primer (5′ – TTC TAA TAC GAC TCA CTA TAG GAA ACC ATC ATA ACA GTT CTT GCT ACA AAT CAT TTA TCA GGG CCT GGA CTT AAA GCC GCG AGG – 3′), 2.5 μM 3′ reverse selection primer (5′ – TTC CTG CCC GGA CTG – 3′), 0.03 U/μL Taq DNA polymerase (Invitrogen), using the following cycling program in a thermal cycler with a heated lid: (1) 72°C for 5 minutes (3′-primer extension); (2) 94°C for 4 minutes (denaturation); (3) 55°C for 5 minutes (annealing); (4) 72°C for 8 minutes (primer extension). Steps (2) to (4) were repeated for 4 cycles, which should produce 8 copies of each unique sequence in the starting population (2ⁿ⁻¹, where n is the number of PCR cycles). The resulting DNA pool was purified by extraction with one volume of 50% (v/v) Tris-saturated phenol in chloroform, and concentrated by precipitation by addition of 2.5 volumes of ethanol.

2.3 Transcription and purification of RNA pools for selection

2.3.1 First selection round

Approximately 20 nanomoles of DNA were transcribed into RNA (50 mM HEPES-KCl pH 7.5, 25 mM MgCl₂, 5 mM of each NTPs, 5 U/μL of T7 RNA polymerase, Figure 1B), in the presence of 200 μM of a 33-nt DNA oligonucleotide (5′ - GCT TTA AGT CCA GGC GCT GAT AAA TGA TTT GTA – 3′) that hybridizes to the RNA region encoding the potential cleavage site. This functions to inhibit ribozyme cleavage and retain full-length RNA constructs during the transcription process where millimolar concentrations of Mg²⁺ as well as other molecules that may serve as adventitious cofactors are present (Figure 3). Transcription was terminated after an incubation of 2 hrs at 37°C by adding one volume of loading buffer (90% formamide, 25 mM EDTA), and purified by electrophoresis through a denaturing 8% PAGE. The 185-nt RNA was visualized by UV shadowing, excised, and eluted overnight by electro-elution or passively with water. The eluted RNA was brought up to 100 mM NaCl and precipitated with 2.5 volumes of chilled ethanol.

Figure 3. Self-cleavage of glmS^WT is blocked in the presence of a cleavage inhibitor DNA.

Autoradiogram of a PAGE analysis of the transcription of body-labeled cis-cleaving glmS^WT in the presence or absence of a 33-nt DNA oligonucleotide complementary to sequences flanking the cleavage site.

2.3.2 Subsequent selection rounds

To increase the stringency of the selection, sulfur-mercury affinity PAGE was employed starting from the second selection round (Figure 1B). In each round, sulfur-mercury affinity PAGE was performed twice; first for purification of a 5′ thiol-tagged RNA pool, and then for the selection of active ribozyme sequences. Sulfur-mercury affinity PAGE exploits the strong interaction between thiol-tagged RNAs and mercury-containing compounds. N-acryloylaminophenyl-mercuric acetate (APM) can be used to incorporate mercury into polyacrylamide gels during polymerization (Figure 2, [22, 27]). Pool RNAs were 5′ thiol-tagged by including 6-thioguanosine monophosphate (^6SGMP) during transcription (one of the tautomeric forms of ^6SGMP has a free thiol). RNAs synthesized in this manner can have 6-thioguanosine only at their 5′ termini, since the monophosphate form of the nucleotide cannot serve as a substrate for transcription elongation. 100 μL reactions contained 1 nmol DNA template, 8 mM ^6SGMP, and 2 mM NTPs (reduced from 2–5 mM in normal conditions; Note 4; [28, 29]). The reduced NTP concentration enhances incorporation of ^6SGMP during transcription initiation. Following the reaction, the thiol-tagged RNAs were purified through an 8% denaturing PAGE containing 2 μg APM per mL of gel solution (otherwise comprised of acrylamide: bisacrylamide (29:1), urea, TBE buffer, and water). RNAs with a 5′ ^6SGMP have a reduced electrophoretic mobility in the APM gel compared to unlabelled RNAs [27, 30]. Gel slices containing the labeled RNAs were recovered by elution overnight in 1 mM dithiothreitol (DTT), which competes for binding to mercury, helping release the thiol-tagged RNAs from the gel.

Figure 2. Chemical basis of thiol-mercury affinity PAGE.

(A) Structure of N-acryloylaminophenyl-mercuric acetate (APM) incorporated into the polyacrylamide gel matrix. The mercury derived from APM interacts strongly with thiol-containing molecules, enabling APM-PAGE separation between inactive RNAs that bear an intact 5′-terminal 6-thioguanosine and active RNAs that have cleaved their leader sequence.

2.4 Selection of active ribozyme sequences

2.4.1 Size selection in the first round

In the initial round, ~5 nanomoles of transcribed RNAs (at 5 μM) were incubated overnight at room temperature in 25 mM MgCl₂ (50 mM HEPES, 200 mM KCl, pH 7.5) together with 40 μM of a cleavage inhibitor DNA (5′ - AAA TGA TTT GTA GCA AGA ACT GTT ATG ATG G – 3′), which hybridizes to the first 31 nucleotides of the RNA leader sequence. This DNA oligonucleotide was added in excess to the RNAs to prevent catalytic scission within the 5′ leader sequence, i.e., to enforce cleavage at the wild-type cleavage site. Following incubation, the reaction was stopped by the addition of 25 mM EDTA, desalted with water using a centrifugal microfilter (Amicon, 30 kDa nominal molecular weight cut-off), and electrophoresed through a 8% denaturing PAGE. Markers of unrelated RNAs with lengths 185-nt, 145-nt, and 40-nt were loaded in an adjacent lane to indicate the position of the active ribozymes (145-nt) after cis cleavage of the leader sequence at the wild-type locus. To enhance stringency, subsequent selection steps employed mercury-thiol affinity electrophoresis.

2.4.2 Thiol-mercury affinity PAGE selection in subsequent rounds

In subsequent rounds, purified RNA pool sequences with a 5′- thiol tag were incubated in Mg²⁺ buffer with a DNA that inhibits cleavage within the leader sequence, as described, and purified through a second APM-PAGE (Figure 1B and Figure 2). Active ribozymes with their leader sequence removed would migrate faster than the inactive RNAs that retained their 5′-thiol tag. The combined effect of the difference in length and the presence or absence of the 5′- ^6SGMP tag between the inactive and active RNA sequences results in a large difference in electrophoretic mobility. As a size marker, a wild-type glmS ribozyme was incubated in parallel in 25 mM MgCl₂ and 200 μM GlcN6P to induce self-cleavage, and electrophoresed in an adjacent lane on the APM gel (see Note 5). In the later selection rounds, the selection pressure was increased to promote the isolation of the fastest ribozymes by reducing the incubation time. In our selection, we halved the incubation time in each round starting from round 5 and continued for 4 additional rounds.

2.5 Amplification of active sequences

2.5.1 DNase I digestion and reverse transcription

Following passive overnight elution (into 1 mM DTT in the case of APM gels), RNAs were ethanol precipitated, and treated with 1 U/μL of DNase I (Roche) at 37°C for 30 minutes (400 mM Tris-HCl, 100 mM NaCl, 60 mM MgCl₂, 10 mM CaCl₂, pH 7.9). To inactivate the nuclease, the reaction was incubated at 75 °C for 20 minutes, followed by extraction with one volume of phenol and chloroform, and precipitation with 2.5 volumes of chilled ethanol. These steps ensure that DNAs present during the transcription and selection steps are removed and not enriched in the subsequent PCR amplification. Following DNase treatment, the active RNAs were reverse-transcribed to generate single-stranded cDNAs (50 mM Tris-HCl, 75 mM KCl, 3 mM MgCl₂, 10 mM DTT, 560 μM of each dNTP, dNTP, pH 7.5, 10 U/μL Superscript II (Invitrogen), incubation at 48°C for 1 hr). The reaction was stopped by heating the sample at 90 °C for 20 minutes after adjusting to 100 mM KOH, in order to degrade the RNAs and denature the enzyme. This was followed by neutralization by adjusting to 250 mM Tris-HCl, pH 8.5.

2.5.2 PCR amplification and enrichment of active sequences

The reverse-transcribed cDNAs were amplified in two sequential PCR reactions. The first was performed under standard conditions in 100 μL volume using the 5′ internal primer, which corresponds to the first 20-nt of the RNA selection pool minus the 40-nt leader sequence (5′-GCG CCT GGA CTT AAA GCC GC -3′), and the 3′ selection primer. Amplification using the 5′ internal primer in this first PCR step is essential in the selection as it avoids amplifying 185-nt cDNAs generated from reverse-transcription of full-length, inactive RNAs. The 145-nt PCR products, after ~12 cycles of PCR (see Note 6), were analyzed and purified using a 1.5% agarose gel and recovered using an agarose gel purification kit (Qiagen).

In the second PCR reaction (500 μL), ~0.1 nmols of DNA recovered from the first PCR amplification were used as template and amplified using the 5′ selection and 3′ reverse-selection primers. This PCR reaction regenerates the T7 promoter and the 40-nt substrate sequences, both of which are needed for synthesis of full-length RNAs for the next selection round (Figure 1B). Starting from round 4, the leader sequence of the 5′ selection primer was changed every two rounds to avoid enrichment of ribozymes that are specific to particular leader sequences.

The PCR amplification step can also be used to increase sequence diversity, thereby enhancing the likelihood of finding active ribozymes that are close in sequence space to RNAs abundant in the selection pool. This technique (mutagenic PCR) uses skewed dNTP and metal ion concentrations (20 mM Tris-HCl pH 8.4, 50 mM KCl, 7 mM MgCl₂, 0.5 mM MnCl₂, 0.2 mM dATPs, 0.2 mM dGTPs, 1 mM dCTPs, 1 mM dTTPs) and a prolonged PCR cycling program that make the Taq DNA polymerase more prone to adding incorrect nucleotides [31]. A typical 30-cycle mutagenic PCR will introduce mutations at a frequency of ~0.7% per position.

2.6 Sequence analysis of the selection

Starting from round 4, the amplified DNA pool at the end of each round was cloned and ~30 isolates were sequenced as described (see Note 3). Sequences were aligned using the Clustal algorithm from MegAlign (DNASTAR) and compared with the glmS^WT sequence. Duplicate sequences, or isolates that shared recognizable sequence features in the mutagenized regions, were not observed until round 6. At the end of the final selection round, ~70 isolates were sequenced. Of these, 26 retained the core secondary structure that characterizes the wild-type glmS ribozyme. Alignment of the sequences of these 26 isolates revealed 3 highly conserved mutations. The glmS^AAA variant containing just these 3 mutations was synthesized and further characterized.

3. Concluding remarks

Pioneering experiments with the Tetrahymena group I intron demonstrated that one RNA, or close sequence variants, can catalyze a number of biochemically distinct reactions such as self-splicing [32], RNA cleavage [33], aminoacyl ester hydrolysis [34], template-dependent oligonucleotide polymerization [35] etc., Nonetheless, for many of those reactions, the ribozymes derived from the Tetrahymena intron exhibited only modest catalytic activity. Advances in the structural elucidation of RNAs with compact, stable folds [36] provide an opportunity to mutagenize ribozymes rationally, in ways that are likely to preserve their overall structural integrity, and to subject them to in vitro selection for new biochemical activities. Our discovery that three point mutations suffice to convert the GlcN6P-dependent glmS ribozyme into a highly active and strictly divalent-cation-dependent catalytic RNA that nonetheless retains the overall three-dimensional structure of the wild-type suggests that only a small number of nucleotides can be responsible for the specific reactivity of a given RNA (see also Ref. [22, 37]). Thus, in vitro selection applied to stable RNA scaffolds is likely to result in the discovery of families of closely related RNAs that catalyze several biochemically distinct transformations.

Abbreviations

APM

N-acryloylaminophenylmercuric acetate

dNTPs

deoxyribonucleotide triphosphates

DTT

dithiothreitol

GlcN6P

glucosamine-6-phosphate

glmS^WT

wild-type glmS ribozyme

^6SGMP

6-thioguanosine monophosphate

glmS^AAA

triple adenosine-mutant glmS ribozyme

nt

nucleotides

NTPs

ribonucleotide triphosphates

PAGE

polyacrylamide gel electrophoresis