Title: Author response: Structures of two aptamers with differing ligand specificity reveal ruggedness in the functional landscape of RNA
Abstract: Article Figures and data Abstract eLife digest Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract Two classes of riboswitches related to the ykkC guanidine-I riboswitch bind phosphoribosyl pyrophosphate (PRPP) and guanosine tetraphosphate (ppGpp). Here we report the co-crystal structure of the PRPP aptamer and its ligand. We also report the structure of the G96A point mutant that prefers ppGpp over PRPP with a dramatic 40,000-fold switch in specificity. The ends of the aptamer form a helix that is not present in the guanidine aptamer and is involved in the expression platform. In the mutant, the base of ppGpp replaces G96 in three-dimensional space. This disrupts the S-turn, which is a primary structural feature of the ykkC RNA motif. These dramatic differences in ligand specificity are achieved with minimal mutations. ykkC aptamers are therefore a prime example of an RNA fold with a rugged fitness landscape. The ease with which the ykkC aptamer acquires new specificity represents a striking case of evolvability in RNA. https://doi.org/10.7554/eLife.36381.001 eLife digest DNA’s iconic double helix has made it possibly the most widely recognized biological molecule. The closely related RNA, however, is less well known but just as vital. In contrast with DNA’s typical rigid structure, RNA is more flexible and can fold into a wide range of shapes; this allows RNA molecules to have many jobs. Some RNA molecules form structures called riboswitches. As the name suggests, these act as molecular switches that help cells to respond to the presence of important small molecules. When a riboswitch encounters the right molecule, it changes shape, which in turn changes how the cell behaves. It is very difficult, if not impossible, to predict how a riboswitch recognizes its preferred small molecule. To address this, scientists use a technique called X-ray crystallography to directly examine the riboswitch’s structure. Knappenberger, Reiss and Strobel have now determined the structures of two recently discovered riboswitches. The two switches detect molecules called PRPP and ppGpp, respectively. These riboswitches are structurally similar to one that binds to a very different type of chemical called guanidine. The aim was to understand how similar switches respond to different signals. The results reveal that a PRPP riboswitch could become a ppGpp riboswitch just by making a single change to the RNA sequence. Many scientists believe RNA preceded DNA and proteins in some of the earliest organisms on Earth. Understanding how RNAs have evolved and diversified could thus help to understand how early life developed. The results may also help to design synthetic riboswitches for a variety of uses. Since many riboswitches are unique to bacteria, this work could also contribute to the search for new antibiotics. https://doi.org/10.7554/eLife.36381.002 Introduction RNA has diverse functional capabilities, which has driven speculation that the first organisms may have been RNA-based (Breaker, 2012; Crick, 1968; Gilbert, 1986; Orgel, 2004; 1968; Strobel, 2001; Woese et al., 1966). For this hypothesis to be plausible, RNA must be adaptable; that is, capable of acquiring new functions through mutation. In the field of evolutionary biology, this trait is described as evolvability (Kirschner and Gerhart, 1998; Wagner and Altenberg, 1996). Evolvability is the propensity of a system to produce a mutated genotype that yields a beneficial phenotype under new selective pressures (Ancel and Fontana, 2000; Kirschner and Gerhart, 1998; Wagner, 2008; Wagner and Altenberg, 1996). Often, this occurs through mutation of an existing gene through divergent evolution. For example, bacterial β-lactamases demonstrate significant evolvability through mutations in the Ω-loop. This loop determines substrate specificity, but mutation or outright deletion of the loop does not dramatically affect the overall structure of the protein (Banerjee et al., 1998; Hujer et al., 2001; Kurokawa et al., 2000; Wachino et al., 2004). This locus of evolvability allows the protein to adapt to the selective pressures of novel antibiotics. The concept of evolvability has also been studied in RNA, including a notable paper by Draghi et al. (Draghi et al., 2010). This study found that adaptation rate is hastened when the build-up of some phenotypically neutral mutations occurs and the web of accessible phenotypes becomes broad. The speed at which an organism adapts is determined by this, as well as the ruggedness of the fitness landscape, which is related to the number of mutations required to reach a new fitness maximum. Variant riboswitches yield insight into the evolvability of RNA. Riboswitch variants are naturally occurring riboswitches with a conserved overall fold but altered ligand specificity. Examples include the guanine/adenine riboswitches and the cyclic-di-GMP/cyclic-GMP-AMP riboswitches (Kellenberger et al., 2015; Mandal et al., 2003; Mandal and Breaker, 2004; Ren et al., 2015; Serganov et al., 2004; Smith et al., 2009; Sudarsan et al., 2008). Bioinformatic and structural studies of the guanine/adenine and cyclic-di-GMP/cyclic-GMP-AMP aptamers showed that the altered specificity occurs simply by changing base pairing between the RNA and ligand. Recently, the ykkC RNA motif was identified as binding to multiple, chemically dissimilar ligands, which makes this specific scaffold a compelling target for structural studies of RNA evolvability (Sherlock et al., 2018a; 2018b; Nelson et al., 2017). The ykkC RNA was discovered in 2004 and its ligand(s) remained unknown for over a decade (Barrick et al., 2004; Nelson et al., 2017). In 2017, Nelson et al. published two pivotal discoveries regarding this motif: (1) the majority of these RNAs bind specifically to the guanidinium cation and (2) the ykkC riboswitch class can be divided into at least two subtypes. Subtype 1, which has approximately 1500 known examples, is the major class now known as the guanidine-I riboswitch. Subtype 2 was defined as all variants of this motif that do not recognize guanidine. The subtype 2 variants are overall quite similar to guanidine-I riboswitches. They retain the same overall fold, but possess a few characteristic differences at nucleotides crucial for guanidine binding. Notably, most of these differences are centered around a classic S-turn motif that forms the binding pocket of the guanidine-I riboswitch. A similar overall architecture with key differences in binding pocket nucleotides is a signature characteristic of a riboswitch variant (Weinberg et al., 2017). Variant ykkC RNAs are found upstream of a variety of genes, although two major groups are apparent. One major group regulates amino acid synthesis and transport genes, which are upregulated during the stringent response. The other regulates de novo purine biosynthesis, which produces purine nucleotides from smaller metabolites under conditions where intact nucleobases are not available (Sherlock et al., 2018a; 2018b; Ebbole and Zalkin, 1987; 1989). These riboswitches were designated as ykkC subtype 2a and 2b, respectively. When compared to guanidine riboswitches, subtypes 2a and 2b harbor systematic changes to residues directly involved in guanidine binding, which led to the suggestion that they may have different ligand specificity. For example, where guanidine riboswitches have a conserved adenosine residue (A46 in the guanidine-I structure solved by Reiss et al.), subtypes 2a and 2b have a pyrimidine (C49 in the present study) (Battaglia et al., 2017; Reiss et al., 2017). Sorting the entire ykkC class by the identity of this position alone results in a strikingly complete segregation of guanidine-related gene contexts from those that are incongruent with mitigation of guanidine toxicity (Nelson et al., 2017). Alignment of subtype 1, 2a, and 2b sequences also shows an extension of conservation at both the 5′ and 3′ ends of the 2a and 2b aptamer subtypes. These key differences in conserved residues and gene contexts suggested that these ykkC variants have altered ligand specificity while retaining the same overall architecture. Subtype 2a and 2b ykkC riboswitches do not retain the ligand specificity of their parent riboswitch. Using transcription termination and in-line probing assays, Sherlock et al. found that neither subtype is responsive to guanidine (Sherlock et al., 2018a; 2018b). Instead, subtype 2a is responsive to guanosine tetra/pentaphosphate ((p)ppGpp, hereafter referred to as ppGpp), an alarmone that regulates the stringent response (Cashel and Gallant, 1969; Dalebroux and Swanson, 2012; Gaca et al., 2015). Subtype 2b is responsive to phosphoribosyl pyrophosphate (PRPP), a precursor in purine biosynthesis. Like the guanidine riboswitch, both function as ON switches. The consensus motifs for subtypes 2a and 2b are remarkably similar to each other, even relative to other ykkC RNAs (Sherlock et al., 2018a; 2018b). The most apparent difference is a highly-conserved guanosine (G96 in this study) in subtype 2b that is not conserved in subtype 2a. This residue is equivalent to G89 in the guanidine-I riboswitch and is a conserved part of its S-turn motif. Although bioinformatic data suggest that variation in G96 is central to the structural differences between subtype 2a and 2b riboswitches, its precise role in this context remains uncertain. Unlike guanidine, the biological roles of PRPP and ppGpp are both well-documented. PRPP is an activated form of ribose 5-phosphate, and a major macromolecular building block (Hove-Jensen et al., 2017). It is a central metabolite used in biosynthesis of purine and pyrimidine nucleotides, the amino acids histidine and tryptophan, nicotinamide adenine dinucleotide, thiamine diphosphate, flavins, and pterins (Hove-Jensen, 1988; Jiménez et al., 2008; White, 1996). The centrality of PRPP within metabolism makes it an appealing target for regulation. ppGpp is an alarmone that initiates the stringent response, a global reaction to nutrient starvation in bacteria (Cashel and Gallant, 1969; Cashel and Kalbacher, 1970; O'Farrell, 1978; Potrykus and Cashel, 2008). Amino acid starvation triggers synthesis of ppGpp and it binds to a variety of effector molecules to initiate sweeping changes in the cell’s transcriptional profile, including a reduction in tRNA and rRNA synthesis and an increase in transcription of amino acid biosynthesis genes (Cashel, 1970; Paul et al., 2005; Ryals et al., 1982; van Ooyen et al., 1976). Consistent with a role in the stringent response, the ppGpp riboswitch turns on transcription of amino acid biosynthesis and transport genes in response to alarmone binding. Although the tree topology is unknown, a common ancestral RNA likely diverged to recognize guanidine, PRPP, and ppGpp in spite of the chemical and structural diversity among these ligands. PRPP and ppGpp are more similar to each other than either is to guanidine, which reflects the greater similarity in their aptamers. While guanidine harbors a single delocalized positive charge, PRPP and ppGpp harbor multiple separate loci of negative charge. Guanidine is small and achiral with three-fold rotational symmetry, while PRPP and ppGpp are larger, chiral, asymmetric molecules. PRPP and ppGpp both contain ribose sugars and pyrophosphate moieties, but ppGpp has an entire guanine base that PRPP lacks. Bioinformatic evidence suggests that the 2a and 2b aptamers represent an especially concise solution to a central biophysical problem: biologically relevant switching entails recognition of a cognate ligand and rejection of structurally similar alternatives. We set out to determine how three RNA elements with a common scaffold could recognize such dissimilar ligands with high specificity. Central questions include how a polyanionic macromolecule differentially recognizes two distinct small polyanions, and how the presence or absence of the guanine base changes the RNA’s recognition strategy. To address these questions of molecular recognition by RNA, we report the near-atomic resolution structure of a native ykkC 2b riboswitch in complex with PRPP via X-ray crystallography. We also convert this construct into a ppGpp aptamer with a single G96A mutation and present the structure of the mutant bound to ppGpp. This structural and biochemical information reveals how the ykkC RNA differentiates between ppGpp and PRPP. This study showcases the functional plasticity of RNAs and the evolvability of RNA function from a single structural scaffold. Results The structure of the wild-type PRPP aptamer and a single point mutant ppGpp aptamer To understand the basis of ligand recognition by the PRPP riboswitch, we determined the crystal structure of the aptamer domain of the ykkC 2b riboswitch from Thermoanaerobacter mathranii at 2.5 Å resolution in the presence of its native ligand, PRPP (Supplementary file 1). PRPP is an activated metabolic intermediate. As a result, it is highly unstable. It degrades on a time course of minutes to hours via several mechanisms in the presence of divalent metal ions, acidic or basic pH, and/or elevated temperatures (Dennis et al., 2000; Hove-Jensen et al., 2017; Khorana et al., 1958; 1955; Meola et al., 2003; Remy et al., 1955). However, binding to the PRPP riboswitch aptamer domain protects PRPP on a time scale of hours to days (Figure 1—figure supplement 1). The stabilizing effect of the aptamer permitted crystals of the intact complex to be observed after two days. Once formed, unfrozen crystals disappeared after approximately five to ten days, underscoring the need for prompt crystallization and cryogenic preservation in this study. The structure was solved by molecular replacement using the guanidine-I aptamer as an initial model. After model building and refinement, the model fit the data with an Rwork of 0.216 and an Rfree of 0.253. Like its parent aptamer, the PRPP riboswitch contains two adjacent helical stacks (Figure 1). P3 forms a large portion of the binding pocket, and a conserved loop at the end of P3 docks into P1a. This allows conserved nucleotides from P1a to participate in ligand recognition. P1, P1a, P1b, and P2 together form a continuous coaxial stack adjacent to P3. However, unlike the guanidine aptamer, the PRPP aptamer has structured tails at the 5′ and 3′ ends that are not conserved in the guanidine riboswitch. The ends pair to form an additional short helix that we have termed P0, resulting in a four-way junction between P0, P1, P2, and P3. P0 coaxially stacks with P3 and extends the binding pocket for recognition of the larger PRPP ligand. The overall architecture of the PRPP aptamer reveals that it is a rather conservative adaptation of the guanidine aptamer with key differences that allow for PRPP recognition. Figure 1 with 2 supplements see all Download asset Open asset Overall structure of the PRPP riboswitch and its G96A mutant, which is a ppGpp aptamer. (A) Consensus sequence of the PRPP riboswitch, adapted from Sherlock et al. (Sherlock et al., 2018b). The secondary structure has been updated to show structural information gained from the present study. The sequence is depicted as in Sherlock et al. (see key). Nucleotides noted in blue are important bioinformatic differences between PRPP riboswitches and guanidine riboswitches. Base pair notation is as published previously (Leontis and Westhof, 2001). (B) Secondary structure of the PRPP riboswitch aptamer from T. mathranii. Nucleotides are colored by paired region. Paired regions are indicated in bold. Sequence numbering is indicated in gray. Nucleotides that directly contact PRPP are circled in red, and arrows indicate strand connectivity. (C) Crystal structure of the PRPP riboswitch. Chain A is shown. The RNA is depicted as a cartoon and PRPP is depicted as yellow spheres. Nucleotides are colored by paired region as in B. (D) Crystal structure of the G96A mutant. Chain A is shown. The RNA is depicted as a cartoon and ppGpp is depicted as green spheres. Nucleotides are colored by paired region as in B. https://doi.org/10.7554/eLife.36381.003 Figure 1—source data 1 Summary of fitted binding data without Bmax constraints. https://doi.org/10.7554/eLife.36381.006 Download elife-36381-fig1-data1-v2.csv Figure 1—source data 2 Raw binding data. https://doi.org/10.7554/eLife.36381.007 Download elife-36381-fig1-data2-v2.csv Although PRPP is unstable in solution, it has high occupancy in this crystal structure. PRPP is modeled with an occupancy of 1, and its B factors refined similarly to those of nearby residues. The quality of the fit between the electron density data and this model shows that a combination of protection by the riboswitch and a vast molar excess of ligand permitted a high degree of aptamer saturation when data were collected. PRPP is a potentially challenging ligand for RNA to recognize; it has three negatively charged phosphate groups and lacks a moiety resembling a nucleobase. PRPP is known to interact with two divalent metal ions per molecule in solution. The 5-phosphate associates weakly with one metal and the pyrophosphate moiety more strongly coordinates a second metal (Thompson et al., 1978). In the current model, these two metals are present in the complex with the riboswitch (Figure 2). One metal (M1) associates with the 5-phosphate, and the second metal (M2) associates with the pyrophosphate. Both metals form contacts bridging PRPP and the RNA aptamer. A third metal ion, M3, forms a water-mediated coordination to the 5-phosphate. The same water molecule also coordinates M1. The three phosphate groups are major elements of recognition via interactions with nucleobase amines and divalent metal ions. Figure 2 Download asset Open asset The binding pocket of the PRPP riboswitch. (A) Crystal structure of the ligand-binding site in chain A. Relative to Figure 1, the structure is rotated 180° about the y axis. PRPP is depicted as sticks and colored by element with purple carbons. Nucleotides are depicted as blue sticks. Metal ions are depicted as gray spheres. Individual nucleotides and metals are labeled. An FO–FC map contoured at 2.5 σ is shown as a gray mesh. The map was calculated using an otherwise complete model lacking PRPP, M1, and M2. (B) Ligand interaction map. The map is colored essentially as in A. All RNA and metal contacts to PRPP are shown. Dashed black lines indicate hydrogen bonds. Solid black lines indicate coordination to a metal ion. Brackets indicate interactions shown in individual panels of Figure 3. https://doi.org/10.7554/eLife.36381.008 Figure 3 with 1 supplement see all Download asset Open asset Notable contacts to the PRPP ligand in chain A. PRPP is depicted as sticks and colored by element with purple carbons. Nucleotides are depicted as sticks and colored by element with blue carbons. Individual nucleotides and metals are labeled. Dashed black lines indicate hydrogen bonds. A dashed green line shows the lone pair-π interaction between A103 and G105. Solid black lines indicate coordination to a metal ion. Relative to Figure 1, the structure is rotated 180° about the y axis. Panel A is additionally rotated nearly 90° about the x axis. Panel D is rotated approximately 45° about the x axis in the opposite direction. (A) Contacts among the 5-phosphate of PRPP, residues G48, C49, C77, and metal ions M1 and M3. (B) Hydrogen bonds between the ribose of PRPP and residues C77, G96, and G104. (C) Coordination of metal M2 by PRPP and residue G6. (D) Recognition of the pyrophosphate group of PRPP by residues A5, G6, A101, and G105. https://doi.org/10.7554/eLife.36381.009 This construct crystallizes in the presence of BaCl2, so both Ba2+ and Mg2+ are present in the crystallization condition. M1 and M3 are modeled as Ba2+ due to the appearance of large positive peaks in the electron density map when they are modeled as Mg2+. M2 is modeled as Mg2+, but exhibits coordination distances higher than expected for this species (Figure 3—figure supplement 1). The aptamer binds PRPP with nearly equal affinity in the presence of either Ba2+ or Mg2+ alone (2.0 ± 0.4 and 2.0 ± 0.3 µM, respectively). Given that both metals support binding, we expect that there may be partial occupancy of these two species that cannot be resolved at this resolution. The 5-phosphate of PRPP experiences recognition by a metal ion and the amino groups of conserved nucleotides (Figure 3A). The N1 and N2 of G48 form hydrogen bonds with two phosphate oxygens, while the N4 of C78 hydrogen bonds to the third non-bridging phosphate oxygen. The 5-phosphate also coordinates M1, which is held in place by coordination interactions with a non-bridging phosphate oxygen of C77 and the O2 of C49. The residue equivalent to C49 is conserved as an adenosine in the guanidine-I riboswitch but is a pyrimidine in PRPP and ppGpp riboswitches, and the identity of residue 49 was used as a marker to distinguish between these two variants (Sherlock et al., 2018a; 2018b). The O6 of G48 coordinates M3, but M3 is too distant from the 5-phosphate to be directly coordinated by it. The ribose moiety of PRPP also makes extensive interactions with the RNA aptamer (Figure 3B). The sugar edge of G96 forms hydrogen bonds with the 2- and 3-hydroxyl groups. The N4 of C77 donates a hydrogen bond to the ribose oxygen, and the N1 group of G104 donates a hydrogen bond to the 2-hydroxyl group. These three residues are all highly conserved in the consensus sequence of this aptamer. At 2.5 Å resolution, conclusive determination of the sugar pucker is not possible, but a C2-endo pucker is the most likely conformation in this complex and it fits the electron density data well. This conformation avoids a steric clash between the 2-hydroxyl and the β-phosphate and allows the 3-hydroxyl to coordinate M2. This conformation is also consistent with previously reported structures of PRPP in complex with macromolecules (Evans et al., 2014; González-Segura et al., 2007; Héroux et al., 2000). The P0 region of the aptamer extends below P3 and permits a suite of interactions with the pyrophosphate group of PRPP (Figure 3C–D). The β-phosphate of PRPP is more extensively recognized than the α-phosphate. The O6 of G6 coordinates M2, which in turn forms several interactions with the pyrophosphate group (Figure 3C). The N6 group of the weakly conserved A101 (>75% conserved as a purine) contacts a non-bridging oxygen of the α-phosphate (Figure 3D). The N6 group of A5 and the N1 groups of G6 and G105 make direct contacts with non-bridging oxygens of the β-phosphate. An abrupt deformation in the local backbone conformation positions A103 under G105, allowing a lone pair-π interaction to form between the O6 atom of G105 and the six-membered ring of A103 (Chawla et al., 2017; Egli and Sarkhel, 2007; Ran and Hobza, 2009; Sarkhel and Desiraju, 2003; Singh and Das, 2015). The present results show that the PRPP aptamer recognizes its ligand through a shifted and extended helical ligand-binding region, allowing for the retention of bound metal ions and extensive hydrogen bond donation to phosphate groups. The intracellular PRPP concentration in bacteria is estimated to be in the millimolar range (Hove-Jensen et al., 2017; Jendresen et al., 2011; Jensen et al., 1979; Nygaard and Smith, 1993; Saxild and Nygaard, 1991; Schneider and Gourse, 2004; Yaginuma et al., 2015). However, enzymes and protein regulatory elements that sense PRPP concentrations in bacteria typically have micromolar dissociation (Kd) or Michaelis (KM) constants (Bera et al., 2003; Hove-Jensen et al., 2017; Jørgensen et al., 2008). Sherlock and colleagues recently found that the T50 (the ligand concentration that produces half-maximal effect) of a PRPP riboswitch in transcription termination assays is 90 μM (Sherlock et al., 2018b). We determined the Kd of the riboswitch aptamer domain for PRPP (Table 1, see also Figure 1—figure supplement 2A) by equilibrium dialysis using radiolabeled [β-33P]-PRPP. This assay yields a Kd of 2.0 ± 0.3 μM. There are two notable differences between the present experimental system and that employed by Sherlock et al. First and most importantly, the present study examines binding affinity in an isolated aptamer domain, while Sherlock et al. focused on the ability of the full riboswitch to terminate transcription. The full system is governed by the kinetics of ligand association and RNA folding, while the present experimental system only measures the thermodynamics of ligand binding. Also, in this study, [β-33P]-PRPP was used in trace quantities and the amount of intact PRPP remaining in each sample was carefully measured to deconvolute the counts obtained from intact PRPP and the counts obtained from breakdown products. Sherlock et al. used unlabeled PRPP and could not quantify the extent of degradation, likely resulting in some underestimation of PRPP’s ability to terminate transcription. The present data show that the affinity of the complex is at least of low micromolar affinity, placing it well within the range observed for complexes of PRPP with protein elements (Bera et al., 2003; Jørgensen et al., 2008). Table 1 Dissociation constants for PRPP and ppGpp binding to the wild type and G96A T. mathranii aptamers with calculated fold specificity changes. https://doi.org/10.7554/eLife.36381.011 Dissociation constants for WT and G96A binding to PRPP and ppGppConstructKd for PRPPKd for ppGppFold specificity for PRPP over ppGppEstimated magnitude of overall specificity switchWild type2.0 ± 0.3 μM91 ± 3 μM46~40,000G96A1600 ± 200 μM1.8 ± 0.1 μM~0.001 In parallel with structural inquiries into the PRPP riboswitch, crystallization of native ppGpp aptamers was pursued. However, crystallization was unsuccessful with the subset of ppGpp aptamers tested. Considering the evident versatility of the ykkC motif and the overt similarity between the consensus sequences of ykkC RNA subtypes 2a and 2b, a specificity switch of the PRPP aptamer to a ppGpp aptamer was pursued via mutation as an alternative strategy. Close examination of the consensus motifs of the PRPP and ppGpp riboswitch aptamers revealed that the ppGpp aptamer consensus sequence was almost entirely a subset of the PRPP aptamer consensus sequence, with the PRPP aptamer generally having more stringent requirements than the ppGpp aptamer. The most salient difference between the two consensus sequences is at position 96. In the PRPP aptamer, this position is >97% conserved as a guanosine, but this conservation is lost in the ppGpp aptamer. In the ppGpp aptamer, the lack of conservation in this region complicates the process of sequence alignment. However, it appears that this nucleotide is not always present and, when it is, it appears to be conserved as A, C or U, but not G (Sherlock et al., 2018a). The dramatic difference in conservation at this site suggested that it may be critical for differential recognition of PRPP and ppGpp. We mutated position 96 in the T. mathranii PRPP aptamer from guanosine to adenosine, generating the G96A mutant. The wild-type aptamer shows low affinity for ppGpp (Kd = 91 ± 3 μM) and 46-fold greater affinity for PRPP (Kd = 2.0 ± 0.3 μM) (Table 1). Conversely, the G96A mutant binds ppGpp with an affinity equivalent to that of wild-type for PRPP (Kd = 1.8 ± 0.1 μM), but PRPP binding is abolished in the mutant up to 400 μM RNA (estimated Kd = 1600 ± 200 μM). The G96A mutant has approximately 900-fold higher affinity for ppGpp than PRPP. The G96A mutation thus strikingly resulted in approximately a 40,000-fold switch in ligand specificity from PRPP to ppGpp. The mutant’s affinity for ppGpp is well within the range of native aptamers tested (data not shown). Co-crystal structure of the generated ppGpp aptamer and its ligand Having shown that the G96A mutant is a ppGpp aptamer, we solved its crystal structure in the presence of ppGpp to 3.1 Å resolution. The crystallization conditions that reproducibly gave rise to co-crystals of the wild-type PRPP aptamer did not yield comparable results for co-crystals of the G96A mutant. However, the G96A mutant was found to crystallize in a separate condition that also produced crystals of the wild-type aptamer. The crystallization reagent used for G96A lacks barium, which was the most abundant divalent metal ion in the wild type crystallization condition. Potassium chloride, sodium chloride, and magnesium chloride were present in the crystallization drops. K+ and Mg2+ ions are observed in the mutant crystal structure. The best mutant crystal diffracted to a resolution of 3.1 Å and its structure was solved by molecular replacement using chain A of the PRPP riboswitch as an initial model. The asymmetric unit contained four aptamer molecules. Molecular replacement and refinement revealed robust density for the electron-dense pyrophosphate groups of ppGpp as well as its guanine base. In the initial solution and throughout refinement, the quality of the electron density was worse in chain D compared to chains A-C. The model of chain D is consistent with that of chains A-C, but is excluded from discussion in the text. Overall, the architecture of the G96A mutant is very similar to that of the wild-type aptamer (Figure 1D). Notably, the 2FO−FC map generated directly by molecular replacement showed no electron density in the former location