Skip to main content
NCBI home page
Nucleic Acids Research logoLink to Nucleic Acids Research
. 2025 Jul 22;53(14):gkaf702. doi: 10.1093/nar/gkaf702

Crystal structures reveal the distinct features of the 2′-dG-III riboswitch in the purine riboswitch family

Ke Chen 1,c, Wenjian Liao 2,3,c, Jiali Wang 4,c, Yangyi Ren 5, Zhizhong Lu 6,7, Xuemei Peng 8,9, Jia Wang 10,✉, Lin Huang 11,✉
PMCID: PMC12282948  PMID: 40694847

Abstract

Purine riboswitches, located in the 5′-untranslated regions of bacterial messenger RNA, regulate gene expression by sensing purines and their derivatives. A class of the guanine-I riboswitch variants was recently reported to be the third class of 2′-deoxyguanosine (2′-dG) riboswitch termed 2′-dG-III riboswitch. Here we present the crystal structures of the 2′-dG-III riboswitch bound with 2′-dG, guanosine, or guanine. Despite similarities in secondary and overall structures to other purine riboswitches, the 2′-dG-III riboswitch exhibits unique features in its loop–loop interaction, three-way junction, and ligand binding. The 2′-dG-III riboswitch exhibits a tuning fork-like structure with a unique six-tiered interaction within the three-way junction. The second, third, and fourth tiers form the ligand-binding pocket, with a consistent binding mode for the guanine moiety of all three ligands. Structural and biochemical analyses reveal detailed interactions between the 2′-dG-III riboswitch and different ligands, providing insights into its regulatory mechanisms in purine metabolism.

Graphical Abstract

Graphical Abstract.

Graphical Abstract

Open in a new tab

Introduction

Riboswitches are structured non-coding RNA elements commonly located in bacterial messenger RNAs′ (mRNAs′) 5′-untranslated regions (UTRs). These RNA motifs sense specific small-molecule ligands through their conserved three-dimensional architectures. Ligand binding induces structural rearrangements across the RNA scaffold, which modulate the accessibility of regulatory sequences (e.g. ribosome-binding sites or transcriptional terminators) to control downstream gene expression. This ligand-responsive conformational switching enables dynamic regulation of transcription elongation or translation initiation, effectively switching gene activity between activated (ON) or repressed (OFF) states [1–4]. More than 56 natural riboswitches have been characterized over the past two decades, the ligands they sense cover diverse biologically relevant compounds, including RNA precursors and derivatives, cofactors, elemental ions, signaling molecules, amino acids, sugars, and other metabolites [5–7].

The Bacillus subtilis xpt guanine-I riboswitch is one of the earliest identified riboswitches [8], and a riboswitch that senses adenine was identified from Vibrio vulnificus add mRNA [9]. The two riboswitches adopt nearly identical overall structures, including three helical stems (P1, P2, and P3), two hairpin loops (L2 and L3), and three junction-connecting segments that participated in forming the three-way junction (J1/2, J2/3, and J3/1) (Fig. 1A and B). Two helical stems are aligned in parallel by forming the extensive loop–loop interaction at their distal ends. In addition, the three-way junction located at the riboswitch core region forms a five-tiered interaction and contains a ligand-binding pocket. Swapping nucleotide 74 between cytosine (C) and uracil (U) in the two riboswitches reverses their ligand binding specificity, indicating that nucleotide 74 is one of the specificity determinants [10, 11].

Figure 1.

Figure 1.

Open in a new tab

Sequences and secondary structures of the five classes of purine riboswitches. (A) B. subtilis xpt guanine-I riboswitch (PDB ID: 1Y27). (B) V. vulnificus add adenine riboswitch (PDB ID: 1Y26). (C) Mesoplasma florum I-A 2′-dG-I riboswitch (PDB ID: 3SKI). (D) env1 2′-dG-II riboswitch (PDB ID: 6P2H). (E) Bacillus sp. VT712 2′-dG-III riboswitch (PDB ID: 8KEB). The consensus structures shared by the five classes of purine riboswitches are labeled as P (paired), J (junction), and L (loop). The nucleotides circled in red interact with the purine or deoxyribose moieties of ligands.

With the development of search strategies in structural bioinformatics, Breaker and colleagues have confirmed three classes of 2′-deoxyguanosine (2′-dG) riboswitches, 2′-dG-I [12], 2′-dG-II [13], and 2′-dG-III [14] in their research of the guanine-I riboswitch variants [15]. Together with the guanine-I riboswitch and adenine riboswitch, they are known as the purine riboswitch family [16]. The crystal structures of the M. florum I-A 2′-dG-I riboswitch and env1 2′-dG-II riboswitch have been determined, their overall structures are similar to the guanine-I and adenine riboswitches but also have their own characteristics. The 2′-dG-I riboswitch exhibits a lack of three nucleotides at loop L3, while two additional nucleotides are inserted into the J3/1 of the 2′-dG-II riboswitch. (Fig. 1C and D). Notably, both riboswitches form a six-tiered interaction within the three-way junction [17, 18].

2′-dG-III riboswitch is relatively rare and only regulates the purine nucleoside hydrolase genes that are not related to other purine riboswitches. In this work, we designed the Bacillus sp. VT712 2′-dG-III riboswitch candidate sequence, nuc-WT, as identified by Breaker et al., in order to obtain high-resolution crystal structures of the 2′-dG-III riboswitch bound with 2′-dG, guanosine, or guanine [14]. Although the secondary structure and overall structure are similar to other members of the purine riboswitch family, there are significant differences in the loop–loop interaction, three-way junction, and ligand binding (Fig. 1E). We have revealed the specific interactions between the 2′-dG-III riboswitch and three small molecules. Furthermore, biochemical analysis of several mutants also demonstrates that the nucleotides within the three-way junction play a crucial role in ligand binding.

Material and methods

RNA preparation

The 2′-dG-III riboswitch and mutants were transcribed by T7 RNA polymerase in vitro. The strategies employed to design sequences for crystallizing the 2′-dG-III riboswitch are outlined in Supplementary Fig. S1. DNA templates required for transcription were synthesized by polymerase chain reaction (PCR) using Primer 1 and Primer 2, following the methods described by Horton et al. [19]. DNA templates have a T7 polymerase promoter and the starting nucleotides were replaced by GGA in order to facilitate transcription. All primer sequences are given in Supplementary Table S1. RNA was purified by electrophoresis in a polyacrylamide gel containing 7 M urea after 6 h of in vitro transcription at 37°C. Targeted RNA was observed through ultraviolet light, then excised and electroeluted in 0.5 × Tris-Borate-EDTA (TBE) buffer for 6 h at 200 V. Finally, the obtained RNA was precipitated with isopropanol, washed with 75% ethanol, and dissolved in double-distilled water to prepare 10 μg/μl solution for crystallization [20].

Crystallization and X-ray diffraction

A solution containing 10 μg/μl RNA was heated to 95°C for denaturation under the conditions of 5 mM HEPES (pH 7.5), 100 mM KCl, and 5 mM MgCl2 for 1 min; 5 mM ligand (2′-dG, guanosine, and guanine) was added to the solution after the RNA was refolded at room temperature for 10 min. Crystallization was performed by mixing 0.2 μl of the RNA–ligand complex with 0.2 μl of reservoir solution using sitting drop vapor diffusion at 18°C. Subsequently, 0.8 μl of the RNA–ligand complex was mixed with 0.8 μl of reservoir solution to optimize the crystal using the hanging drop vapor diffusion method. Crystals of the 2′-dG-III riboswitch bound with 2′-dG, guanosine, or guanine grew from 0.08 M sodium chloride, 0.01 M potassium chloride, 0.01 M magnesium chloride hexahydrate, 0.04 M sodium cacodylate trihydrate pH 6.4, 30% v/v (±)-2-methyl-2,4-pentanediol (MPD), 0.012 M spermine tetra-hydrochloride and 0.1 M glycine. Crystals can be directly frozen by mounting in nylon loops and plunging into liquid nitrogen due to the inclusion of 30% MPD in growth conditions. X-ray diffraction data were collected on beamline BL02U1, and BL10U2 at the Shanghai Synchrotron Radiation Facility (SSRF) and the National Facility for Protein Science in Shanghai (NF-PSS), and processed with XIA2 or XDS [21, 22]. The resolution cutoff for the data was determined by examining CC1/2 and the density map [23].

Structure determination and refinement

Using the xpt-pbuX guanine riboswitch (PDB ID: 4FE5) as an initial search model [11], the structure of the 2′-dG-III riboswitch bound with 2′-dG was determined by molecular replacement using PHASER [24]. Likewise, the structure determined above was used as an initial search model to determine the structure of the 2′-dG-III riboswitch bound with guanosine or guanine. Models were refined manually using Coot and performed several rounds of adjustment and optimization using Coot [25], phenix.refine [26], and PDB_REDO [27]. Model geometry and the fit to electron-density maps were monitored with MOLPROBITY [28] and the validation tools in Coot. Simulated annealing omit maps were calculated by composite omit maps in the PHENIX suite using the annealing method. Atomic coordinates and structure factor amplitudes have been deposited with the PDB with accession code as listed in Supplementary Table S2 [29].

Isothermal titration calorimetry

Binding affinities of the RNA–ligand complex were measured by MicroCal PEAQ-ITC calorimeter. The sequences of the 2′-dG-III riboswitch and mutants are given in Supplementary Table S1; 0.1 mM RNA was dialyzed overnight at 4°C in Isothermal titration calorimetry (ITC) buffer containing 50 mM K-HEPES (pH 7.5), 100 mM KCl, and 20 mM MgCl2. Ligand was dissolved in ITC buffer at room temperature, with a concentration of approximately 1.0 mM. Experiments were performed three times with a reference power of 10 μcal/s, titrating ligand into RNA in the sample cell by 16–18 continuous injections of ∼2 μl each [17, 18, 30]. The C value corresponding to a satisfactory fitting curve is between 10 and 100. The apparent dissociation constant (KD) was determined by fitting data to a single-site binding model using MicroCal PEAQ-ITC Analysis Software.

Results

Crystallization of the 2′-dG-III riboswitch

Based on the conservative distribution properties of other purine riboswitches [8, 9, 12, 13], we modified the nucleotide and length of stem P1 of the candidate sequence nuc-WT, and designed stem P1 with lengths of 7, 9, and 11 bp. Through experiments, it was found that the nuc-9bp sequence can grow relatively satisfying crystals. Subsequently, G·U base pairs at three different positions were designed for the nuc-9bp sequence to improve the quality of crystals [31], and finally, high-quality crystals were obtained through the nuc-9bp-M3 sequence. Sequences and design methods are shown in Supplementary Fig. S1.

We successfully determined the crystal structures of the 2′-dG-III riboswitch bound with 2′-dG, guanosine, or guanine (Fig. 2A), each of which crystallized in space group C121. The crystal structure formed by the 2′-dG-III riboswitch and 2′-dG diffracted to a resolution of 2.20 Å.

Figure 2.

Figure 2.

Open in a new tab

The overall structure of the 2′-dG-III riboswitch. (A) The chemical structures of the 2′-dG-III riboswitch ligands; 2′-dG, guanosine, and guanine. (B) Secondary structure model of the Bacillus sp. VT712 2′-dG-III riboswitch. L2–L3 interaction is shown by full lines. (C) Scheme of the structure observed in the crystal structure of the riboswitch bound with 2′-dG. Base interaction notation is derived from Leontis and Westhof. The dashed line represents a single hydrogen bonding interaction. (D) Cartoon representation of the crystal structure of the 2′-dG-III riboswitch bound with 2′-dG (magenta). (E) Structural superimposition of the 2′-dG-III riboswitch bound with different ligands. The structures are: 2′-dG-III riboswitch bound with 2′-dG (PDB ID: 8KEB); 2′-dG-III riboswitch bound with guanosine (PDB ID: 8KHH); 2′-dG-III riboswitch bound with guanine (PDB ID: 8KED).

The overall structure of the 2′-dG-III riboswitch

In general, the secondary and overall structures are similar to other purine riboswitches, adopting a tuning fork-like conformation (Fig. 2B and C). Stem P1 serves as the handle of the tuning fork, while stems P2 and P3 are anchored at the distal ends through the loop–loop interaction between hairpin loops L2 and L3, thus forming two parallel prongs of the tuning fork. The three-way junction between the handle and prongs connects three helical stems and forms a ligand-binding pocket in its central region (Fig. 2D).

When we superimposed the structures of the 2′-dG-III riboswitch bound with different molecules, the differences in root mean square deviation (RMSD) values were: 1.241 Å between 2′-dG and guanosine, 0.320 Å between 2′-dG and guanine, and 1.234 Å between guanosine and guanine (Fig. 2E).

The L2–L3 interaction in the 2′-dG-III riboswitch

Hairpin loops L2 and L3 in the 2′-dG-III riboswitch both contain seven nucleotides and are anchored together through the extensive loop–loop interaction formed by five base pairs (Fig. 3A and B). G37-C59 and G36-C60 are standard cis-Watson–Crick base pairs, while the remaining three are noncanonical. These noncanonical interactions include the trans-Watson–Crick/Hoogsteen pairing between A32 and A65, the trans-Watson–Crick/Hoogsteen pairing between U33 and A64, and the trans-Hoogsteen/Watson–Crick pairing between A34 and A63. Further alignment amongst these base pairs results in the formation of the A32·A65·C59-G37 quadruple interactions and the U33·A64·C60-G36 quadruple interactions, which effectively stabilizes the L2–L3 interaction (Fig. 3C and D). In addition, N6 of A63 donates a single hydrogen bond to O2′ of U33 (Fig. 3E). As for the other bases, G31 is stacked between A32 and the U30·U38 base pair. U35 orients outwards from the stacked array of L2–L3 interaction. A61 and C62 are located at the apex of the L2–L3 interaction scaffold, with A61 stacked on A34.

Figure 3.

Figure 3.

Open in a new tab

The L2–L3 interaction in the 2′-dG-III riboswitch. (A) Image of the loop–loop interaction between L2 and L3 in the 2′-dG-III riboswitch. (B) Detailed view of the L2–L3 interaction in the 2′-dG-III riboswitch. (C, D) The two quadruple base interactions, the A32·A65·C59-G37 (C) and the U33·A64·C60-G36 (D). (E) Direct interaction between U33 and A34·A63 base pair.

The three-way junction and 2′-dG-binding pocket in the 2′-dG-III riboswitch

The three-way junction of the 2′-dG-III riboswitch is formed by the apex of stem P1, the bases of stems P2 and P3, and the three junction-connecting segments, J1/2, J2/3, and J3/1. The only ligand-binding pocket is located at the center within the three-way junction (Fig. 4A).

Figure 4.

Figure 4.

Open in a new tab

The three-way junction and 2′-dG binding pocket in the 2′-dG-III riboswitch. (A) Detailed view of the tertiary structure of the three-way junction and 2′-dG binding pocket. The lower image is obtained by rotating the upper image 120° clockwise. (B) Scheme of the six-tiered interaction within the three-way junction. T denotes the tier number. (C) Binding of ligands to the 2′-dG-III riboswitch measured through ITC in 20 mM MgCl2 condition. The concentrations of the 2′-dG-III riboswitch, 2′-dG, and guanosine are 0.1, 1.0, and 1.0 mM, respectively.

A series of nucleotide interactions within the three-way junction can be divided into six tiers from base to apex (T1–T6) according to their stacked planes, and the nucleotides in the second, third, and fourth tiers (T2–T4) form the binding pocket for 2′-dG (Fig. 4B). The first tier (T1) comprises an A47·(G75·U19) triple, and N6 of A47 donates a single hydrogen bond to O2′ of G75. The second tier (T2) consists of an A48·(U74-A20) triple, forming the cis-Watson–Crick/Sugar pairing between A48 and U74. Furthermore, the 5′-OH of 2′-dG deoxyribose forms two hydrogen bonds with N3 and O2′ of A20. The furan oxygen (O4′) accepts a single hydrogen bond from N6 of A48, and the 3′-OH donates a single hydrogen bond to N7 of A48. In the third tier (T3), 2′-dG is recognized by forming a U49·(2′-dG-C73) triple. C73 forms a cis-Watson–Crick base pair with the guanine moiety of 2′-dG, while U49 and the guanine moiety are aligned through the trans-Watson–Crick/Sugar pairing. A50-U21 is a cis-Watson–Crick base pair in the fourth tier (T4), O2′ of U21 forms two hydrogen bonds with O6 and N7 of the guanine moiety. As previously described, the binding of 2′-dG to the 2′-dG-III riboswitch involves eleven hydrogen bonding interactions formed by five nucleotides. The A22·(G45-C51) triple constitutes the fifth tier (T5), wherein A22 interacts with G45 through the trans-Watson–Crick/Sugar pairing.

Lastly, the sixth tier (T6) is a specific structure that distinguishes the 2′-dG-III riboswitch from the other two classes of 2′-dG riboswitches. The Hoogsteen edge of A52 pairs with the Watson–Crick edge of U23 in a cis conformation, located between the U53·U71 base pair and A72.

We identified five key nucleotides (A20, U21, A48, U49, and C73) crucial for 2′-dG binding in the 2′-dG-III riboswitch. To explore the influence of a single point mutation on the binding ability of the 2′-dG-III riboswitch, we performed single point mutations on A48 in the second tier (T2), U49, and C73 in the third tier (T3). Because U21 forms pairing with A50 in the fourth tier (T4), and A20 pairs with U74 in the second tier (T2), mutations are likely to disrupt the overall structure of the 2′-dG-III riboswitch. Therefore, A20 and U21 were excluded from the experiments. In addition, for the first (T1) and fifth (T5) tiers not involved above, we selected A47 and A22 for experiments, respectively. We also selected C46 and A72, which are involved in the formation of three-way junction, but not within the six-tiered interaction. Moreover, we performed a co-mutation on the A52·U23 base pair located at the sixth tier (T6) within the three-way junction.

Due to guanine’s low solubility in water, we could not prepare a suitable guanine solution for ITC experiments. Instead, we measured the binding affinities of 2′-dG and guanosine to the riboswitch in 20 mM MgCl2, obtaining KD values of 1.98 ± 0.07 μM and 4.31 ± 0.29 μM, respectively. The 2′-dG-III riboswitch exhibits a higher affinity for 2′-dG compared to guanosine, consistent with the in-line probing results from Breaker et al. [14]. Subsequently, we measured the binding affinities of 2′-dG to various 2′-dG-III riboswitch mutants. The results exhibited that A48G, U49C, and C73U, which directly interacted with 2′-dG, all lost the binding ability. The single point mutation of A47U also significantly decreased the affinity, with a KD value of only 22.60 ± 1.47 μM, and A22G almost caused the complete loss of affinity (Supplementary Fig. S2). As for C46U and A72G, the affinity of the former decreased to 15.00 ± 0.53 μM, while the latter exhibited minimal influence (with a KD value of 1.41 μM). Lastly, the A52G·U23C co-mutation also entirely disrupted the binding of 2′-dG to riboswitch (Fig. 4C). Detailed thermodynamic parameters are shown in Supplementary Table S3.

The guanosine or guanine-binding pocket in the 2′-dG-III riboswitch

We determined the 2.30 Å crystal structure of the 2′-dG-III riboswitch bound with guanosine, which closely matches the 2′-dG-bound structure, with an RMSD of 1.241 Å (Fig. 2E). The structure contains a guanosine-binding pocket formed by nucleotides in the second, third, and fourth tiers (T2–T4) within the three-way junction (Supplementary Fig. S3A). The interaction between the 2′-dG-III riboswitch and guanosine is highly consistent with those of 2′-dG. The difference is that the 5′-OH of guanosine ribose only forms a single hydrogen bond with A20, while the 2′-OH donates a single hydrogen bond to O4 of U21 (Supplementary Fig. S3B).

The crystal structure of the 2′-dG-III riboswitch bound with guanine has a resolution of 2.10 Å, the highest among the three structures. Its overall structure is nearly identical to the structure of 2′-dG binding, with an RMSD value of 0.320 Å (Fig. 2E). The interaction between the 2′-dG-III riboswitch and the guanine moiety is consistent with those of 2′-dG and guanosine (Supplementary Fig. S3C and D).

Structural comparison with the other two classes of 2′-dG riboswitches

Analysis of the 2′-dG-III riboswitch structure revealed that despite its high similarity to the other two classes of 2′-dG riboswitches, several notable differences exist. The first prominent difference is that the L2–L3 interactions among the three classes of 2′-dG riboswitches are inconsistent, except for the formation of two G-C base pairs in all. The 2′-dG-I riboswitch forms fewer hydrogen bonds compared to the 2′-dG-II and 2′-dG-III riboswitches, as the L3 contains only 4 nucleotides. A71 from L3 inserts into L2, strengthening the L2–L3 interaction. O2′ of A71 donates a single hydrogen bond to N1 of A40, while OP2 accepts a single hydrogen bond from N3 of U41. As for the 2′-dG-II riboswitch, the interaction between L2 and L3 is extremely similar to the 2′-dG-III riboswitch, but N2 of G63 donates an additional hydrogen bond to O4 of U64 (equivalent to A61 and C62 in the 2′-dG-III riboswitch) (Fig. 5A and B).

Figure 5.

Figure 5.

Open in a new tab

Comparison of the L2–L3 interactions in the three classes of 2′-dG riboswitches. (A) Image of the loop–loop interactions between L2 and L3 in the three classes of 2′-dG riboswitches. (B) Detailed view of the L2–L3 interactions in the three classes of 2′-dG riboswitches. PDB IDs are provided for each structure.

The three classes of 2′-dG riboswitches also exhibit distinct differences in the three-way junction and ligand binding. Firstly, the lengths of J3/1 in the 2′-dG-I and 2′-dG-III riboswitches are 2 nt, while in the 2′-dG-II riboswitch is 4 nt. Secondly, all 2′-dG riboswitches form a six-tiered interaction within the three-way junction, sharing similarity in the first five tiers but differing in the sixth. Compared to the cis-Hoogsteen/Watson–Crick pairing between A52 and U23 in the 2′-dG-III riboswitch, N2 of G33 donates two hydrogen bonds to O6 of G34 and O4 of U52 to form a G33·(G34·U52) triple in the 2′-dG-I riboswitch. The insertion of two nucleotides in J3/1 of the 2′-dG-II riboswitch promotes the formation of new interactions. Specifically, A24 forms a cis-Hoogsteen/Watson–Crick base pair with U75 in the sixth tier (T6), and A77 interacts with A52 through the cis-Watson–Crick/Sugar pairing, forming an A77·(A52-U22) triple in the fourth tier (T4) (Fig. 6A). Lastly, the interactions formed by the binding of 2′-dG to the three classes of 2′-dG riboswitches differ significantly. In the 2′-dG-I riboswitch, the nucleotides involved in recognizing the guanine moiety are C31, C58, and C80; in the 2′-dG-II riboswitch, they are U22, C51, and C78; and in the 2′-dG-III riboswitch, they are U21, U49, and C73. Moreover, the position of U49 in the 2′-dG-III riboswitch is shifted relative to C58 in the 2′-dG-I riboswitch and C51 in the 2′-dG-II riboswitch, leading to differences in hydrogen bond formation sites (Fig. 6B). The number and type of nucleotides involved in recognizing the deoxyribose moiety differ among the three classes of 2′-dG riboswitches: A30, C56, and C57 in the 2′-dG-I riboswitch; G21, C47, U49, and C50 in the 2′-dG-II riboswitch; and A20 and A48 in the 2′-dG-III riboswitch. Due to the different conformation of 2′-dG deoxyribose, the three classes of 2′-dG riboswitches can be differentiated based on the number and binding sites of hydrogen bonds. The 2′-dG-III riboswitch forms four hydrogen bonds with the deoxyribose moiety through two nucleotides, while the 2′-dG-I riboswitch forms three hydrogen bonds through three nucleotides, and the 2′-dG-II riboswitch forms five hydrogen bonds through four nucleotides (Fig. 6C).

Figure 6.

Figure 6.

Open in a new tab

Comparison of the six-tiered interactions within the three-way junctions in the three classes of 2′-dG riboswitches. (A) Scheme of the six-tiered interactions within the three-way junctions in the three classes of 2′-dG riboswitches. (B) The interactions between the three classes of 2′-dG riboswitches and 2′-dG guanine moiety. The electron density shows the simulated annealing omit map contoured at 1.5σ. (C) The interactions between the three classes of 2′-dG riboswitches and 2′-dG deoxyribose moiety.

Structural comparison with other purine riboswitches

Comparative analysis of the structure of the three classes of purine riboswitches—adenine riboswitch, guanine-I riboswitch, and 2′-dG riboswitch—reveals that the major differences are evident in the L2–L3 interaction, three-way junction, and ligand binding. As for the L2–L3 interaction, in the adenine riboswitch, N7 of A35 accepts a hydrogen bond from N6 of A64, while A34 and A63 at the equivalent positions in the 2′-dG-III riboswitch form a trans-Hoogsteen/Watson–Crick base pair. The guanine-I riboswitch forms an additional hydrogen bond compared to the 2′-dG-III riboswitch, with N2 of G62 donating a single hydrogen bond to O4 of U63 (equivalent to A61 and C62 in the 2′-dG-III riboswitch) (Supplementary Fig. S4A and B).

In terms of the three-way junction, both the adenine and guanine-I riboswitches form a five-tiered interaction. In the first tier (T1) of the adenine riboswitch, U49 interacts with A76 through the cis-Watson–Crick/Sugar pairing, which is also observed in the guanine-I riboswitch (Supplementary Fig. S5A). Regarding the ligand binding, the nucleotides responsible for recognizing the purine moiety vary across the riboswitches: U22, U51, and U74 in the adenine riboswitch; U22, U51, and C74 in the guanine-I riboswitch; and U21, U49, and C73 in the 2′-dG-III riboswitch. In the guanine-I riboswitch and 2′-dG-III riboswitch, C74 and C73, respectively, act as specificity determinants, forming Watson–Crick base pairs with the guanine moiety of ligands, while U74 serves the same role in the adenine riboswitch. Furthermore, O2′ of U22 in the adenine and guanine-I riboswitches accepts only a single hydrogen bond from N7 of ligands, contrasting with the formation of two hydrogen bonds in the 2′-dG-III riboswitch. Lastly, the bases interacting with the sugar edge of ligands in the three classes of purine riboswitches are uracils, but U51 in the adenine riboswitch differs from those in the other two classes, leading to altered binding sites for hydrogen bonds (Supplementary Fig. S5B).

Discussion

In this work, we present the high-resolution crystal structures of the 2′-dG-III riboswitch bound with three molecules 2′-dG, guanosine, or guanine, respectively. The three structures crystallize in the same space group and are highly superimposable. All the structures contain a nearly identical ligand-binding pocket within the three-way junction, demonstrating the high tolerance and definite specificity of the 2′-dG-III riboswitch to different ligands.

G·U is the most common non-Watson–Crick base pair in RNA structure, which plays a critical role in RNA folding and crystallization due to the unique chemical, thermodynamic, and conformational properties [31–36]. Notably, our research group recently published a paper titled “A general strategy for engineering GU base pairs to facilitate RNA crystallization”, which reported eight successful examples of engineering G·U base pairs [31]. In this work, we introduce a G·U base pair in the stem region by substituting cytosine with uracil or adenine with guanine to facilitate the crystallization of the 2′-dG-III riboswitch. Ultimately, the designed G57·U67 base pair in stem P3 achieved a remarkable increase in resolution from 3.01 Å to 2.20 Å, which is one of the successful cases demonstrated in our aforementioned study.

The secondary structure of the 2′-dG-III riboswitch basically corresponds with that reported by Breaker et al. [14], and the overall structure is similar to other purine riboswitches [10, 11, 17, 18]. Stems P1, P2, and P3 are connected through the three-way junction, the interaction between L2 and L3 promotes the parallel alignment of stems P2 and P3, thus forming a tuning fork-like folding of the tertiary structure. The L2–L3 interaction formed by the highly conserved nucleotides is relatively consistent with other purine riboswitches except the 2′-dG-I riboswitch, indicating the importance of L2–L3 interaction in stabilizing the overall structure. In the ligand-binding pocket, the deviation in pairing between U49 and the guanine moiety is also observed in the guanine-I riboswitch, which explains why guanine can bind to the 2′-dG-III riboswitch with high affinity [14]. We believe that the ability of the 2′-dG-III riboswitch to bind with ligands with similar chemical properties may have evolved during the regulation of purine metabolism and corresponding physiological reactions because these regulatory processes cannot be easily achieved by sensing a single compound among the numerous metabolites [18, 37].

Furthermore, binding sites and structural properties are validated through the 2′-dG-III riboswitch mutants. The results of ITC experiments reveal that the A48G mutation probably disrupts the hydrogen bonding interaction with the 2′-dG deoxyribose moiety, while the U49C and C73U mutations both disrupt the hydrogen bonding interaction with the 2′-dG guanine moiety. A47 forms a single hydrogen bond with G75 to stabilize the first tier (T1) through A47·(G75·U19) triple, while A22 forms a trans-Watson–Crick/Sugar base pair with G45 to stabilize the fifth tier (T5) through A22·(G45-C51) triple. Both A47U and A22G may disrupt the original base triple, weakening the structural stability of tiers (T1 and T5), and consequently affecting the overall stacking structure of the six-tiered interaction and the conformation of the ligand-binding pocket. Based on the determined structure, C46 does not interact with other nucleotides, and its pyrimidine moiety also orient to the outside of the ligand-binding pocket. After replacing cytosine with uracil, it may orient to the inside and occupy part of the ligand-binding pocket space, thus interfering the binding of small molecule. The purine moiety of A72 is inserted into the inside of the ligand-binding pocket in the original structure. The overall conformation may still maintain high similarity after replacing adenine with guanine, so the affinity is almost the same as before. Intriguingly, the noncanonical A52·U23 base pair located at the apex of the three-way junction may also play a key role in maintaining the conformation of the ligand-binding pocket. Therefore, the A52G and U23C co-mutation is possible to interfere with the internal conformation of ligand-binding pocket and three-way junction, leading to the loss of ligand binding ability. Supplementary Table S4 compares ligand affinities across purine riboswitches. The Guanine-I riboswitch binds ligands with high affinity despite fewer hydrogen bonds than 2′-dG riboswitches, suggesting that hydrogen bond number is not the sole determinant of affinity.

Riboswitches perform gene regulatory functions through the conformational change in the expression platform driven by the ligand binding to the aptamer domain. Our research illustrates how the 2′-dG-III riboswitch recognizes different small molecules in ligand-binding pocket, providing a more comprehensive understanding of the regulatory mechanisms. The downstream gene regulated by 2′-dG-III riboswitch encodes purine nucleoside hydrolase. When nucleosides accumulate excessively, they act as ligands and bind to riboswitch to activate the expression of downstream genes and produce nucleoside hydrolase, degrade nucleosides to produce free nucleobases, and thus play a role in regulating purine metabolism [14]. However, the coupling process of ligand binding and gene regulation cannot be observed owing to the absence of an expression platform in the crystal structure. The 2′-dG-III riboswitch shows a high degree of conservation according to the results of structural comparison and ITC experiments, whether the phenomenon is related to changing the conformation in the expression platform?

To date, researchers pay more attention to the biochemical functions of riboswitches, as the structural and ligand binding properties enable them to broaden a wide range of applications through appropriate modifications. It has been proved that a natural adenine riboswitch can be repurposed into a well-folded fluorogenic aptamer, which exhibits improved cellular fluorescence efficiency [38, 39]. Riboswitches are also likely to be promising targets for novel antibiotics based on their widespread presence in various bacteria, influencing bacterial homeostasis, development, pathogenicity, and antibiotic resistance at the transcriptional or translational level [40–44]. In conclusion, how to perform directed evolution of riboswitches and ultimately apply them in vivo still remains a challenge.

Supplementary Material

gkaf702_Supplemental_File
gkaf702_supplemental_file.pdf (3MB, pdf)

Acknowledgements

The authors thank the staff from the BL18U1 and BL19U1 beamlines at the National Facility for Protein Science in Shanghai (NFPS), as well as the staff of the BL02U1 (https://cstr.cn/31124.02.SSRF.BL02U1) and BL10U2 (https://cstr.cn/31124.02.SSRF.BL10U2) beamlines at the Shanghai Synchrotron Radiation Facility (SSRF) for assistance in X-ray data collection. We also acknowledge the Drug Discovery Platform of Guangzhou Laboratory for conducting ITC experiments and thank Tianyong Tu for sample preparation support.

Author contributions: Ke Chen (Conceptualization, Data curation, Writing—original draft, Writing—review & editing), Wenjian Liao (Conceptualization, Data curation, Writing—original draft), Jiali Wang (Methodology, Writing—review & editing), Yangyi Ren (Methodology, Visualization), Zhizhong Lu (Software, Visualization), Xuemei Peng (Methodology, Funding acquisition), Jia Wang (Conceptualization, Project administration, Supervision, Funding acquisition), Lin Huang (Conceptualization, Project administration, Supervision, Funding acquisition). All authors have read and approved the final manuscript.

Contributor Information

Ke Chen, Guangdong Provincial Key Laboratory of Malignant Tumor Epigenetics and Gene Regulation, Guangdong-Hong Kong Joint Laboratory for RNA Medicine, Medical Research Center, Sun Yat-Sen Memorial Hospital, Sun Yat-Sen University, Guangzhou 510120, China.

Wenjian Liao, Guangdong Provincial Key Laboratory of Malignant Tumor Epigenetics and Gene Regulation, Guangdong-Hong Kong Joint Laboratory for RNA Medicine, Medical Research Center, Sun Yat-Sen Memorial Hospital, Sun Yat-Sen University, Guangzhou 510120, China; Department of Urology, Sun Yat-Sen Memorial Hospital, Sun Yat-Sen University, Guangzhou 510120, China.

Jiali Wang, Guangdong Provincial Key Laboratory of Malignant Tumor Epigenetics and Gene Regulation, Guangdong-Hong Kong Joint Laboratory for RNA Medicine, Medical Research Center, Sun Yat-Sen Memorial Hospital, Sun Yat-Sen University, Guangzhou 510120, China.

Yangyi Ren, Guangdong Provincial Key Laboratory of Malignant Tumor Epigenetics and Gene Regulation, Guangdong-Hong Kong Joint Laboratory for RNA Medicine, Medical Research Center, Sun Yat-Sen Memorial Hospital, Sun Yat-Sen University, Guangzhou 510120, China.

Zhizhong Lu, Guangdong Provincial Key Laboratory of Malignant Tumor Epigenetics and Gene Regulation, Guangdong-Hong Kong Joint Laboratory for RNA Medicine, Medical Research Center, Sun Yat-Sen Memorial Hospital, Sun Yat-Sen University, Guangzhou 510120, China; School of Life Sciences and Biopharmaceutics, Guangdong Pharmaceutical University, Guangzhou 510120, China.

Xuemei Peng, Guangdong Provincial Key Laboratory of Malignant Tumor Epigenetics and Gene Regulation, Guangdong-Hong Kong Joint Laboratory for RNA Medicine, Medical Research Center, Sun Yat-Sen Memorial Hospital, Sun Yat-Sen University, Guangzhou 510120, China; Comprehensive Department, Sun Yat-Sen Memorial Hospital, Sun Yat-Sen University, Guangzhou 510120, China.

Jia Wang, College of Pharmacy, Shenzhen Technology University, Shenzhen 518118, China.

Lin Huang, Guangdong Provincial Key Laboratory of Malignant Tumor Epigenetics and Gene Regulation, Guangdong-Hong Kong Joint Laboratory for RNA Medicine, Medical Research Center, Sun Yat-Sen Memorial Hospital, Sun Yat-Sen University, Guangzhou 510120, China.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

None declared.

Funding

We acknowledge the financial support of the National Natural Science Foundation of China (32400021 to X.P., 32001639 to J.W., and 32171191 to L.H.); Shenzhen Bay Scholars Program to L.H.; the Guangdong Science and Technology Department (2024A1515012594, 2023B1212060013, 2020B1212030004) to L.H.; the Key Realm R&D Program of Guangdong Province (2022B0202050003) to J.W.

Data availability

The coordinates and structure factors of all the reported crystal structures have been deposited in the PDB under accession numbers 2′-dG-III riboswitches bound with 2′-dG (8KEB), 2′-dG-III riboswitches bound with guanine (8KED), 2′-dG-III riboswitches bound with guanosine (8KHH).

References

  • 1. Nahvi  A, Sudarsan  N, Ebert  MS  et al.  Genetic control by a metabolite binding mRNA. Chem Biol. 2002; 9:1043–9. 10.1016/S1074-5521(02)00224-7. [DOI] [PubMed] [Google Scholar]
  • 2. Vitreschak  AG, Rodionov  DA, Mironov  AA  et al.  Riboswitches: the oldest mechanism for the regulation of gene expression?. Trends Genet. 2004; 20:44–50. 10.1016/j.tig.2003.11.008. [DOI] [PubMed] [Google Scholar]
  • 3. Roth  A, Breaker  RR  The structural and functional diversity of metabolite-binding riboswitches. Annu Rev Biochem. 2009; 78:305–34. 10.1146/annurev.biochem.78.070507.135656. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4. Serganov  A, Nudler  E  A decade of riboswitches. Cell. 2013; 152:17–24. 10.1016/j.cell.2012.12.024. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. Breaker  RR  The biochemical landscape of riboswitch ligands. Biochemistry. 2022; 61:137–49. 10.1021/acs.biochem.1c00765. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6. Kavita  K, Breaker  RR  Discovering riboswitches: the past and the future. Trends Biochem Sci. 2023; 48:119–41. 10.1016/j.tibs.2022.08.009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7. Bu  F, Lin  X, Liao  W  et al.  Ribocentre-switch: a database of riboswitches. Nucleic Acids Res. 2024; 52:D265–72. 10.1093/nar/gkad891. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8. Mandal  M, Boese  B, Barrick  JE  et al.  Riboswitches control fundamental biochemical pathways in Bacillus subtilis and other bacteria. Cell. 2003; 113:577–86. 10.1016/S0092-8674(03)00391-X. [DOI] [PubMed] [Google Scholar]
  • 9. Mandal  M, Breaker  RR  Adenine riboswitches and gene activation by disruption of a transcription terminator. Nat Struct Mol Biol. 2004; 11:29–35. 10.1038/nsmb710. [DOI] [PubMed] [Google Scholar]
  • 10. Serganov  A, Yuan  YR, Pikovskaya  O  et al.  Structural basis for discriminative regulation of gene expression by adenine- and guanine-sensing mRNAs. Chem Biol. 2004; 11:1729–41. 10.1016/j.chembiol.2004.11.018. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11. Batey  RT, Gilbert  SD, Montange  RK  Structure of a natural guanine-responsive riboswitch complexed with the metabolite hypoxanthine. Nature. 2004; 432:411–5. 10.1038/nature03037. [DOI] [PubMed] [Google Scholar]
  • 12. Kim  JN, Roth  A, Breaker  RR  Guanine riboswitch variants from Mesoplasmaflorum selectively recognize 2'-deoxyguanosine. Proc Natl Acad Sci USA. 2007; 104:16092–7. 10.1073/pnas.0705884104. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13. Weinberg  Z, Nelson  JW, Lunse  CE  et al.  Bioinformatic analysis of riboswitch structures uncovers variant classes with altered ligand specificity. Proc Natl Acad Sci USA. 2017; 114:E2077–85. 10.1073/pnas.1619581114. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14. Hamal  Dhakal S, Panchapakesan  SSS, Slattery  P  et al.  Variants of the guanine riboswitch class exhibit altered ligand specificities for xanthine, guanine, or 2'-deoxyguanosine. Proc Natl Acad Sci USA. 2022; 119:e2120246119. 10.1073/pnas.2120246119. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15. Nawrocki  EP, Burge  SW, Bateman  A  et al.  Rfam 12.0: updates to the RNA families database. Nucleic Acids Res. 2015; 43:D130–7. 10.1093/nar/gku1063. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16. Porter  EB, Marcano-Velazquez  JG, Batey  RT  The purine riboswitch as a model system for exploring RNA biology and chemistry. Biochim Biophys Acta. 2014; 1839:919–30. 10.1016/j.bbagrm.2014.02.014. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17. Pikovskaya  O, Polonskaia  A, Patel  DJ  et al.  Structural principles of nucleoside selectivity in a 2'-deoxyguanosine riboswitch. Nat Chem Biol. 2011; 7:748–55. 10.1038/nchembio.631. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. Matyjasik  MM, Batey  RT  Structural basis for 2'-deoxyguanosine recognition by the 2'-dG-II class of riboswitches. Nucleic Acids Res. 2019; 47:10931–41. 10.1093/nar/gkz839. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19. Horton  RM, Hunt  HD, Ho  SN  et al.  Engineering hybrid genes without the use of restriction enzymes: gene splicing by overlap extension. Gene. 1989; 77:61–8. 10.1016/0378-1119(89)90359-4. [DOI] [PubMed] [Google Scholar]
  • 20. Peng  X, Liao  W, Lin  X  et al.  Crystal structures of the NAD+-II riboswitch reveal two distinct ligand-binding pockets. Nucleic Acids Res. 2023; 51:2904–14. 10.1093/nar/gkad102. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21. Winter  G  Xia2: an expert system for macromolecular crystallography data reduction. J Appl Crystallogr. 2010; 43:186–90. 10.1107/S0021889809045701. [DOI] [Google Scholar]
  • 22. Kabsch  W  XDS. Acta Crystallogr D Biol Crystallogr. 2010; 66:125–32. 10.1107/S0907444909047337. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23. Karplus  PA, Diederichs  K  Linking crystallographic model and data quality. Science. 2012; 336:1030–3. 10.1126/science.1218231. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24. McCoy  AJ, Grosse-Kunstleve  RW, Adams  PD  et al.  Phaser crystallographic software. J Appl Crystallogr. 2007; 40:658–74. 10.1107/S0021889807021206. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25. Emsley  P, Lohkamp  B, Scott  WG  et al.  Features and development of Coot. Acta Crystallogr D Biol Crystallogr. 2010; 66:486–501. 10.1107/S0907444910007493. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26. Adams  PD, Afonine  PV, Bunkóczi  G  et al.  PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr D Biol Crystallogr. 2010; 66:213–21. 10.1107/S0907444909052925. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27. Joosten  RP, Long  F, Murshudov  GN  et al.  The PDB_REDO server for macromolecular structure model optimization. IUCrJ. 2014; 1:213–20. 10.1107/S2052252514009324. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28. Williams  CJ, Headd  JJ, Moriarty  NW  et al.  MolProbity: more and better reference data for improved all-atom structure validation. Protein Sci. 2018; 27:293–315. 10.1002/pro.3330. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29. Berman  HM, Westbrook  J, Feng  Z  et al.  The Protein Data Bank. Nucleic Acids Res. 2000; 28:235–42. 10.1093/nar/28.1.235. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30. Edwards  AL, Batey  RT  A structural basis for the recognition of 2'-deoxyguanosine by the purine riboswitch. J Mol Biol. 2009; 385:938–48. 10.1016/j.jmb.2008.10.074. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31. Ren  Y, Lin  X, Liao  W  et al.  A general strategy for engineering GU base pairs to facilitate RNA crystallization. Nucleic Acids Res. 2025; 53:gkae1218. 10.1093/nar/gkae1218. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32. Masquida  B, Westhof  E  On the wobble GoU and related pairs. RNA. 2000; 6:9–15. 10.1017/S1355838200992082. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33. Varani  G, McClain  WH  The G x U wobble base pair. A fundamental building block of RNA structure crucial to RNA function in diverse biological systems. EMBO Rep. 2000; 1:18–23. 10.1093/embo-reports/kvd001. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34. Leontis  NB, Stombaugh  J, Westhof  E  The non-Watson–Crick base pairs and their associated isostericity matrices. Nucleic Acids Res. 2002; 30:3497–531. 10.1093/nar/gkf481. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35. Westhof  E, Yusupov  M, Yusupova  G  The multiple flavors of GoU pairs in RNA. J Mol Recognit. 2019; 32:e2782. 10.1002/jmr.2782. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36. Ruszkowska  A, Zheng  YY, Mao  S  et al.  Structural insights into the 5'UG/3'GU wobble tandem in complex with Ba(2+) cation. Front Mol Biosci. 2021; 8:762786. 10.3389/fmolb.2021.762786. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37. Batey  RT  Structure and mechanism of purine-binding riboswitches. Quart Rev Biophys. 2012; 45:345–81. 10.1017/S0033583512000078. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38. Dey  SK, Filonov  GS, Olarerin-George  AO  et al.  Repurposing an adenine riboswitch into a fluorogenic imaging and sensing tag. Nat Chem Biol. 2022; 18:180–90. 10.1038/s41589-021-00925-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39. Truong  L, Kooshapur  H, Dey  SK  et al.  The fluorescent aptamer Squash extensively repurposes the adenine riboswitch fold. Nat Chem Biol. 2022; 18:191–8. 10.1038/s41589-021-00931-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40. Sudarsan  N, Cohen-Chalamish  S, Nakamura  S  et al.  Thiamine pyrophosphate riboswitches are targets for the antimicrobial compound pyrithiamine. Chem Biol. 2005; 12:1325–35. 10.1016/j.chembiol.2005.10.007. [DOI] [PubMed] [Google Scholar]
  • 41. Serganov  A, Polonskaia  A, Phan  AT  et al.  Structural basis for gene regulation by a thiamine pyrophosphate-sensing riboswitch. Nature. 2006; 441:1167–71. 10.1038/nature04740. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42. Edwards  TE, Ferre-D’Amare  AR  Crystal structures of the thi-box riboswitch bound to thiamine pyrophosphate analogs reveal adaptive RNA-small molecule recognition. Structure. 2006; 14:1459–68. 10.1016/j.str.2006.07.008. [DOI] [PubMed] [Google Scholar]
  • 43. Lee  ER, Blount  KF, Breaker  RR  Roseoflavin is a natural antibacterial compound that binds to FMN riboswitches and regulates gene expression. RNA Biol. 2009; 6:187–94. 10.4161/rna.6.2.7727. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44. Chakradhar  S  Bringing RNA into the fold: small molecules find new targets in RNA to combat disease. Nat Med. 2017; 23:532–4. 10.1038/nm0517-532. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

gkaf702_Supplemental_File
gkaf702_supplemental_file.pdf (3MB, pdf)

Data Availability Statement

The coordinates and structure factors of all the reported crystal structures have been deposited in the PDB under accession numbers 2′-dG-III riboswitches bound with 2′-dG (8KEB), 2′-dG-III riboswitches bound with guanine (8KED), 2′-dG-III riboswitches bound with guanosine (8KHH).

Articles from Nucleic Acids Research are provided here courtesy of Oxford University Press

RESOURCES