Full-Length NAD+-I Riboswitches Bind a Single Cofactor but Cannot Discriminate against Adenosine Triphosphate

Bacterial riboswitches are structured RNAs that bind small metabolites to control downstream gene expression. Two riboswitch classes have been reported to sense nicotinamide adenine dinucleotide (NAD+), which plays a key redox role in cellular metabolism. The NAD+-I (class I) riboswitch stands out because it comprises two homologous, tandemly arranged domains. However, previous studies examined the isolated domains rather than the full-length riboswitch. Crystallography and ligand binding analyses led to the hypothesis that each domain senses NAD+ but with disparate equilibrium binding constants (KD) of 127 μM (domain I) and 3.4 mM (domain II). Here, we analyzed individual domains and the full-length riboswitch by isothermal titration calorimetry to quantify the cofactor affinity and specificity. Domain I senses NAD+ with a KD of 24.6 ± 8.4 μM but with a reduced ligand-to-receptor stoichiometry, consistent with nonproductive domain self-association observed by gel-filtration chromatography; domain II revealed no detectable binding. By contrast, the full-length riboswitch binds a single NAD+ with a KD of 31.5 ± 1.5 μM; dinucleotides NADH and AP2-ribavirin also bind with one-to-one stoichiometry. Unexpectedly, the full-length riboswitch also binds a single ATP equivalent (KD = 11.0 ± 3.5 μM). The affinity trend of the full-length riboswitch is ADP = ATP > NAD+ = AP2-ribavirin > NADH. Although our results support riboswitch sensing of a single NAD+ at concentrations significantly below the intracellular levels of this cofactor, our findings do not support the level of specificity expected for a riboswitch that exclusively senses NAD+. Gene regulatory implications and future challenges are discussed.

Inclined A-minor adenines from J3a-3b and J3b-3a (aptamer2) interact with helix P1 (aptamer1).Interactions include: A136, which forms a type I-like interaction with the G8-C64 pair; A135, which forms a type II interaction with the A7-U65 pair; A118, which forms a type II interaction with the A7-U65 pair; A119, which forms a type III-like interaction with G6-C66; and A120, which forms a type 0-like interaction with U4-A67 Here and elsewhere, putative hydrogen bonds are depicted as dashed lines.(B) Interdomain interface of the hairpin ribozyme (PDB entry 2oue).A canonical G+1-C25 base pair connects the two domains at the site of the S-turn, which extrudes nearby bases to make room for the cross-domain interaction.(C) A ribose zipper motif supports interdomain contacts between the loop A and B domain of the hairpin ribozyme.The view is rotated ~90° about the y-axis from panel B.

Figure S1 .
Figure S1.NAD + -I riboswitch secondary structure predictions based on calculations using RNAstructure.(A) Lowest energy structure for the domain I 54-mer predicted by the Duplex algorithm reveals a dimer formed by intermolecular WC pairing (ΔG = -53.8kcal mol -1 ).(Β) Lowest energy structure for the domain I 54-mer predicted by the Bifold algorithm reveals a dimer formed by intra-and intermolecular WC pairing (ΔG = -59.1 kcal mol -1 ).(C) Lowest energy structure for the domain I 54-mer based on "predict secondary structure" reveals a hairpin formed by intramolecular WC pairing (ΔG = -25.8kcal mol -1 ).The hairpin is less stable than intermolecular dimers in A & B. (D) Lowest free energy structure for a split domain I construct comprising 28-mer and 22-mer strands.The results were identical using Duplex and BiFold.The intermolecular WC pairing produced a ΔG of -28.5 kcal mol -1 .Misfolding into intramolecular hairpins by the individual 28-mer and 22-mer was relatively weak, yielding ΔG values of -10.7 kcal mol -1 and -10.3 kcal mol -1 .(E) Lowest energy structure for the domain II 50-mer predicted as described in A (ΔG = -39.5 kcal mol -1 ).(F) Lowest energy structure for the domain I 50-mer predicted as described in B (ΔG = -38.8kcal mol -1 ).(G) Lowest energy structure for the domain II 50-mer as described in C (ΔG = -17.8kcal mol - 1 ).The hairpin is predicted to be less stable than intermolecular dimers in E & F. (H) Lowest free energy structure for a split domain II construct comprising 28-mer and 20-mer strands.The results were identical using Duplex and BiFold.The intermolecular WC pairing produced a ΔG of -22.5 kcal mol -1 .Misfolding into intramolecular hairpins by the individual 28-mer and 20-mer was very weak, yielding ΔG values of -5.8 kcal mol -1 and -1.8 kcal mol -1 .

Figure S2 .
Figure S2.Electrospray ionization mass spectrometry analysis of NAD + , ADP, NADH, AP2Ribavirin and ATP.(A) ADP showed a peak at 426.01 corresponding to the expected m/z for ADP in negative ion mode.(B) NAD + showed a peak at 664.11 corresponding to the expected m/z for ADP in positive ion mode.(C) NADH showed a peak at m/z= 331.55 corresponding to the expected doubly charged species in negative ion mode.(D) AP2-ribavirin showed a peak at m/z 325.54 corresponding to the expected doubly charged species in negative ion mode.(E) ATP showed a peak at m/z = 508.00corresponding to the protonated species in positive ion mode.

Figure S3 .
Figure S3.ITC simulations under low affinity conditions and tests of various two-site binding models.(A) Simulation using MicroCal PEAQITC Analysis Software indicating the ability to detect 1 site binding by domain II, although accurate fitting would not be possible.We assumed a KD value of 1 mM with 1 mM ligand in the syringe and 50 μM riboswitch in the cell.The reference power was 6 μcal s -1 ; the total heat was 7.6 μcal; feedback was high.(B) Simulation of domain II binding using conditions described in panel A, except that the KD was set to 3.4 mM.(C) The simulated sequential-site binding model (solid line) for the full-length riboswitch showing the fit for a cell concentration of 30 μM and a syringe concentration of 630 μM, which was used to match experimental conditions (Supporting FigureS3a).The simulation assumes two high-affinity sites (KD1 of 30 μM and KD2 of 50 μM).Simulations were conducted using SEDPHAT.Experimental inputs were from NAD + titration into the full-length Aai nadA riboswitch (upper thermogram, closed circles).(D) The simulated sequential-site binding model derived using conditions described in panel C.The simulation assumes one high affinity site (KD1 of 20 μM) and one low affinity site (KD2 of 3.4 mM).

Figure S4 .
Figure S4.Representative ITC injection and titrations of ligand into ITC buffer.(A) Replicate thermogram and isotherm fit for ADP injection into the full-length Aai nadA riboswitch.(B).ADP in the syringe was injected into ITC buffer to measure the heat of dilution.The ligand concentration matched that of experiments reported in the main text.(C).NAD + in the syringe was injected into ITC buffer as described in panel B. (D) NADH in the syringe was injected into ITC buffer as described in panel B.

Figure S5 .
Figure S5.Representative size-exclusion chromatography (SEC) analysis of various nadA riboswitch hairpin constructs.(A) Molecular weight standards and calibration curve based on folded RNAs.Here and elsewhere the colored symbols correspond to elution profiles provided.The standards are indicated by colored circles and were normalized based on absorbance.(B) SEC profiles for the Aca domain I hairpin.Comparison of the first-peak elution volume (blue triangle) to the linear trend set by calibration standards shows that the sample does not fall on the curve as a monomer.(C) SEC profiles for the Aca domain II hairpin.Comparison of the first-peak elution volume (brown triangle) to the linear trend set by calibration standards shows that the sample does not fall on the curve as a monomer.

Figure S6 .
Figure S6.Size-exclusion chromatography of the full-length Aai riboswitch in the absence of NAD + .

Figure S7 .
Figure S7.Interdomain junction interfaces of nadA riboswitch domain I. (A) A distinct interface in the Aai riboswitch (PDB entry 7d7w).Here and elsewhere, crystallographically related subunits are labeled domain I (red) and domain I´ (green).(B) Another distinct interface from the Ckv nadA riboswitch (PDB entry 6tf1).

Figure S8 .
Figure S8.Representative ITC thermogram of ligand titration into the full-length Aai riboswitch in K + and Na + containing buffers.(A) ITC thermogram and isotherm fit for NAD + titration into the full-length Aai nadA riboswitch in KCl-containing ITC buffer.(B) Replicate thermogram and isotherm fit for NAD + titration into the full-length Aai nadA riboswitch in presence of NaCl-containing ITC buffer.

Figure S9 .
Figure S9.Intradomain interactions found in two-domain riboswitches and ribozymes.(A) Representative interaptamer interface of the tandem glycine riboswitch from Fusobacterium nucleatum (PDB entry 3p49).Inclined A-minor adenines from J3a-3b and J3b-3a (aptamer2) interact with helix P1 (aptamer1).Interactions include: A136, which forms a type I-like interaction with the G8-C64 pair; A135, which forms a type II interaction with the A7-U65 pair; A118, which forms a type II interaction with the A7-U65 pair; A119, which forms a type III-like interaction with G6-C66; and A120, which forms a type 0-like interaction with U4-A67 Here and elsewhere, putative hydrogen bonds are depicted as dashed lines.(B) Interdomain interface of the hairpin ribozyme (PDB entry 2oue).A canonical G+1-C25 base pair connects the two domains at the site of the S-turn, which extrudes nearby bases to make room for the cross-domain interaction.(C) A ribose zipper motif supports interdomain contacts between the loop A and B domain of the hairpin ribozyme.The view is rotated ~90° about the y-axis from panel B.