|Home | About | Journals | Submit | Contact Us | Français|
The glmS ribozyme-riboswitch is the first known natural catalytic RNA that employs a small molecule cofactor. Binding of glucosamine-6-phosphate (GlcN6P) uncovers the latent self-cleavage activity of the RNA, which adopts a catalytically competent conformation, which is nonetheless inactive, in the absence of GlcN6P. Structural and analog studies suggest that the amine of GlcN6P functions as a general acid-base catalyst, while its phosphate is important for binding affinity. However, the solution pKa of the amine is 8.06±0.05, which is not optimal for proton transfer. Here, we use Raman crystallography directly to determine the pKa’s of GlcN6P bound to the glmS ribozyme. Binding to the RNA lowers the pKa of the amine of GlcN6P to 7.26±0.09, and raises the pKa of its phosphate to 6.35±0.09. Remarkably, the pKa’s of these two functional groups are unchanged from their values for free GlcN6P (8.06±0.05 and 5.98±0.05, respectively) when GlcN6P binds to the catalytically inactive but structurally unperturbed G40A mutant of the ribozyme, thus implicating the ribozyme active site guanine in pKa tuning. This is the first demonstration that a ribozyme can tune the pKa of a small-molecule ligand. Moreover, the anionic glmS ribozyme in effect stabilizes the neutral amine of GlcN6P by lowering its pKa. This is unprecedented, and illustrates the chemical sophistication of ribozyme active sites.
In Gram-positive bacteria, the glmS ribozyme is part of the mRNA encoding the essential enzyme glucosamine-6-phosphate (GlcN6P) synthetase. Binding of GlcN6P to the ribozyme domain leads to self-cleavage, degradation of the mRNA, and negative-feedback regulation of GlcN6P synthesis.1–3 Three lines of evidence indicate that GlcN6P functions as a catalytic cofactor or coenzyme, rather than as an allosteric activator. First, crystallographic4 and biochemical5,6 analyses demonstrate that the RNA is pre-folded in the absence of GlcN6P, and does not undergo conformational changes as it binds GlcN6P and catalyzes RNA cleavage through internal transesterification. Indeed, the glmS ribozyme from the bacterium Thermoanarobacter tengcongensis binds GlcN6P and catalyzes RNA cleavage in the crystalline state.4 Second, GlcN6P binds in the active site of the ribozyme, with its amine in van der Waals contact with the scissile phosphate and within hydrogen bonding distance of the 5′-oxo leaving group of the reaction (Figure 1A).7,8 Consistent with a chemical role for the amine, glucose-6-phosphate (Glc6P, Figure S1), which differs from GlcN6P only in having a hydroxyl group in place of the amine, is not an activator.9 Glc6P competes with GlcN6P for binding to the RNA in solution,9 and crystal structures show that the two compounds bind in precisely the same manner.8 Third, several compounds that have vicinal amino and hydroxyl groups, such as glucosamine (GlcN, Figure S1), serinol, and L-serine, can weakly activate the glmS ribozyme.9 The stereochemical arrangement of amine and hydroxyl groups in these compounds is analogous to that of the anomeric hydroxyl and amine of GlcN6P. Importantly, the apparent pKa of the ribozyme-catalyzed reaction tracks the pKa of amine of the analog employed.9 Based on the foregoing structural and biochemical observations, it has been proposed that GlcN6P functions as a coenzyme and carries out general acid-base catalysis (Figure 1B).4
General acid-base catalysis is a strategy employed by other ribozymes that, without a coenzyme, carry out sequence-specific RNA cleavage through internal transesterification (reviewed in ref 10). Proton transfer is most efficient by functional groups with pKa’s near neutrality.11 Although RNA does not normally have such functional groups, the active sites of the hepatitis δ virus (HDV) and hairpin ribozymes have been demonstrated to perturb the ground-state pKa’s of nucleobases, bringing them closer to neutral pH. Raman microscopic measurements using ribozyme crystals indicate that the functionally essential active-site C75 of the HDV ribozyme12,13 has a pKa 6.4 (vs. 4.1 for CMP),14 and the catalytically critical active-site A38 of the hairpin ribozyme15–17 has a pKa of 5.5 (vs. 3.7 for AMP).18,19 These, and previous solution experiments (reviewed in ref 20), demonstrate that ribozyme active sites can tune the pKa’s of RNA residues to improve their catalytic activity. Because the glmS ribozyme is catalytically inactive in the absence of GlcN6P, its active site nucleobases are not expected to have perturbed ground-state pKa’s. However, it is possible that binding to the glmS ribozyme modulates the properties of the coenzyme GlcN6P, enhancing its catalytic capacity. The apparent pKa of the glmS ribozyme-catalyzed reaction has been variously estimated to be 7.8 or 6.9 by fitting of the pH-rate profile (refs 9,21). McCarthy et al. measured by potentiometric titration the pKa of the amine of free GlcN6P, and found it to be ~8.2 (ref 9). Thus, if the reaction pKa reflects acid-base catalysis by GlcN6P, the RNA might be lowering the pKa of the amine of the small molecule.9 Two molecular dynamics studies on the glmS ribozyme reached opposite conclusions in this regard. Xin and Hamelberg22 suggested that binding to the glmS ribozyme would decrease the pKa of the amine of GlcN6P to 6.2, whereas the pKa would increase to 8.4 when the small molecule binds to the catalytically inactive G40A mutant23 of the ribozyme. On the other hand, Banáš et al.24 concluded that the crystallographic structures of the ribozyme are not consistent with a significant shift in the pKa of the amine of the bound GlcN6P.
To establish definitively whether the glmS ribozyme modulates the pKa’s of its small-molecule coenzyme, we employed Raman crystallography25 and directly measured the ionization state of the amine and phosphate groups of GlcN6P bound to ribozyme single crystals. The crystals are comprised of a trans-acting form of the T. tengcongensis ribozyme with a single 2′-deoxyribose substitution at the cleavage site, which traps the RNA in the pre-cleavage state. Previously, these crystals yielded the most accurate structure of the RNA to date (1.7 Å resolution). That structure was indistinguishable from the structure determined using crystals of an all-ribose ribozyme construct that was catalytically active in the crystalline state.4,8
Difference spectra were measured as a function of pH by subtracting the Raman spectrum of the glmS crystal from that of the same crystal containing stoichiometric amounts of active-site bound GlcN6P, following “soak in” of the ligand. The resulting spectrum of bound GlcN6P contained marker bands that reported on the ionization states of the ligand (Figure 2).26,27 The peak at 980 cm−1 is a phosphate stretching mode, present only for the -O-PO32- form of the phosphate; and the feature near 546 cm−1 is a skeletal mode of the glucose ring that is present only for the -NH3+ form of the amine. At each pH, the intensities of these bands relative to that of the mode at 1462 cm−1 was plotted against pH. The latter band is an internal intensity standard, due to C-H deformations, that does not vary with pH (see experimental procedures and Figures S2, S3 and S4 in Supporting Information). GlcN6P exists in rapidly interconverting α and β anomeric forms in solution.21 Crystal structures indicate that the ribozyme-bound GlcN6P is exclusively in the α conformation.4,8 Consistent with this, we see low intensity peaks in the solution Raman spectra that are absent from the Raman spectra of the glmS-ribozyme bound GlcN6P. We use the 546 cm−1 peak in the Raman spectra of free and bound GlcN6P to arrive at the pKa of the amino group. Since this is assigned to a vibrationally coupled mode involving the ring and the -NH3+ groups that would be sensitive to conformation of the ring, it is very likely that we are titrating the α anomer in both the solution and crystalline states.
Previously, we discovered that a G40A mutant glmS ribozyme is catalytically inactive even in the presence of saturating concentrations of GlcN6P. The 2.7 Å resolution crystal structure of the mutant shows that the RNA is in nearly identical conformation to that of wild-type. Remarkably, the experimental electron density maps indicate that GlcN6P binds to the G40A mutant in the same location as it does to the wild-type glmS ribozyme. Thus, although GlcN6P is indispensable for catalytic activity of the wild-type ribozyme, presence of guanine at position 40 is also absolutely required.23 To determine if this catalytically essential guanine residue might exert its effect, at least in part, through GlcN6P, we also carried out Raman pH titrations of GlcN6P using crystals of the G40A mutant glmS ribozyme. These experiments show that the pKa of the amine of the ribozyme-bound GlcN6P differs substantially between the wild-type and G40A mutant ribozymes (Figure 3A). The pKa of the phosphate of GlcN6P also differs between the wild-type and mutant ribozymes, but to a smaller extent (Figure 3B).
The pKa values for the phosphate and amino groups were derived using a two state model and equation (1).
where IRH+ and IR represent the intensities of the Raman marker bands of protonated and deprotonated amine or phosphate groups. IIS is the intensity of the internal standard band, which is insensitive to pH change. The pKa’s for the amino and phosphate groups of GlcN6P bound to wild-type and G40A glmS ribozymes are compared in Table 1 to those for free aqueous GlcN6P (details of measurements on free GlcN6P are given in Supporting Information, Figures S2, S5, and Table S1). The pKa values for the GlcN6P amine and phosphate groups for the G40A mutant-bound ligand are very similar to those seen for the free ligand in aqueous solution. However, when in complex with the wild-type ribozyme, the pKa of the amine of GlcN6P is lowered by 0.8 units to near neutrality, while the pKa of its phosphate is raised by 0.37 units.
Binding to the wild-type glmS ribozyme, but not to the G40A mutant, modulates the pKa’s of both the amine and the phosphate of GlcN6P. Moreover, both pKa’s shift toward neutrality, in principle improving the ability of each functional group to function as a general acid-base catalyst. Either could account for the experimentally determined reaction pKa of 6.9 or 7.8 (refs 9,21). From the structural standpoint, however, only the amine is likely to participate in catalysis. While the amine of the bound GlcN6P is 3.1 Å from the 5′-oxo leaving group of transesterification reaction, and therefore well placed to function as a general acid catalyst, the phosphate is ~12 Å from the 2′-OH nucleophile and ~9 Å from the 5′-oxo leaving group, making it unlikely to participate in catalytic proton transfer either as a general base or acid, respectively (Figure 1A).4 This is consistent with analog studies, which imply that the amine is directly involved in catalysis, while the phosphate probably only contributes to the binding energy, since GlcN, which has the amine but lacks the phosphate, is an activator (albeit weaker than GlcN6P), while Glc6P, which has the phosphate but lacks the amine, is not an activator.9 In this regard, it is noteworthy that the pKa shift of the amine of GlcN6P is twice as large as that of its phosphate (Table 1).
To our knowledge, this is the first report of the tuning of a catalytically relevant property of a small molecule by a ribozyme. This discovery extends the previously documented ability of ribozyme active sites to tune the pKa’s of RNA functional groups to improve their catalytic activity.11 In addition, this is the first example of the lowering of the pKa of an ionizable catalytic group by RNA. The active sites of the HDV and the hairpin ribozymes raise the pKa’s of RNA functional groups.14,18,19,28 It is, in principle, easy to see how the strongly negatively charged environment of the interior of an intricately folded RNA could stabilize the cationic forms of nucleobases such as adenine and cytosine, and thereby raise their pKa’s. How the glmS ribozyme preferentially stabilizes the neutral amine form of GlcN6P over its ammonium form is not immediately apparent, but our experiments indicate that the nucleobase at position 40 (which is at least 7 Å away from the amine nitrogen of GlcN6P) plays a critical role. (While this paper was under review, Davis et al. independently reported NMR studies suggesting that the wild-type glmS ribozyme from Bacillus anthracis also perturbs the pKa of its cofactor, but those authors did not examine the role of G40.21) Although it is possible that ribozyme-bound metal ions could contribute to lowering the pKa of GlcN6P, there are no crystallographically ordered cations directly bound to either the amine of GlcN6P or the scissile phosphate. Moreover, the structure of the G40A mutant, including the location of the tightly bound metal ions, is indistinguishable from that of the wild-type ribozyme. This points to the nucleobase of residue 40, rather than bound metal ions, as playing a key role in perturbing the pKa of the bound GlcN6P.
Several catalytic RNAs, such as the hammerhead, hairpin and VS ribozymes have active site guanine residues that could function as general acid-base catalysts (reviewed in ref 10). Because the unperturbed pKa of GMP is ~10, the mechanisms available to RNA to lower the pKa’s of such active site functional groups are of considerable biological importance. The present work establishes the glmS ribozyme as a model system with which to dissect the mutual modulation of the chemical properties of an RNA and its small molecule ligand.
We thank M. Lau and H. Xiao for help in preparation of glmS crystals. This work was supported in part by grants from the National Institutes of Health (GM84976 to D.J.K., GM63576 to A.R.F., and GM84120 to P.R.C.), the W.M. Keck Foundation (A.R.F.), and by the intramural program of the National Heart, Lung and Blood Institute (A.R.F.)
Supporting Information Available. Figure S1 shows the structures of GlcN6P, GlcN and Glc6P. Figures S2, S3 and S4 show the Raman spectra of aqueous GlcN6P, GlcN and Glc6P, respectively, in the pH range from 5 to 10. Fig. S5 plots the intensity of the 546 cm−1 skeletal ring mode originating from the -NH3+ group relative to the internal intensity standard – the C-H deformation at 1462 cm−1, and the intensity of the band at 980 cm−1, due to the -PO32− form of the phosphate group, relative to the internal intensity standard, of free GlcN6P. Table S1 provides the results of data analysis using an equation with non-unity Hill coefficient. Details are also provided of the Raman experimental protocols. This information is available free of charge via the Internet at http://pubs.acs.org/