Fractionation factors reveal hidden frustration in an ancient allosteric module

Protein kinase G (PKG) is a major receptor of the cyclic guanosine monophosphate (cGMP) second messenger. By binding to PKG, cGMP regulates intracellular signaling pathways that control a wide range of intracellular processes, such as cell differentiation, platelet activation, memory formation, and vasodilation.^1–5 Notably, the PKG signaling pathways are often distinct from those regulated by cyclic adenosine monophosphate (cAMP)-dependent proteins, such as protein kinase A (PKA).^6–12 A key element of the crosstalk between PKG and PKA signaling pathways is the selective activation of PKG by cGMP rather than cAMP, as both cGMP and cAMP can bind and cause conformational switches in cyclic-nucleotide-binding domains [(CNBs) within the proteins Fig. 1(a)]. CNBs typically include a β-subdomain, which forms a β-barrel containing the base-binding region (BBR) and phosphate-binding cassette (PBC) that bind cyclic nucleotides [Figs. 1(b)–1(d)], and an α-subdomain, comprising an N-terminal α-helix bundle (N3A) and C-terminal helices that rearrange in response to binding [Figs. 1(b)–1(d)].^13–15 

Previously, we assessed the mechanism of cGMP-versus-cAMP selectivity in PKG Iβ by performing comparative nuclear magnetic resonance (NMR)-based analyses of the apo, cAMP- and cGMP-bound forms of the C-terminal CNB (CNB-B) of PKG Iβ [i.e., “PKG Iβ(219–369)”; Figs. 1(a)–1(d)], using a combination of chemical-shift-based, backbone ¹⁵N relaxation (R₁, R₂, and HN-NOE, i.e. 1H,15N heteronuclear Overhauser effect), and hydrogen-exchange (H/D and H/H) analyses.¹³ This domain is adjacent to the catalytic domain of PKG Iβ [Fig. 1(a)], and is known to be a critical control unit for auto-inhibition and cGMP selectivity in PKG Iβ.^14,15 We found that cAMP demonstrates only partial agonism toward PKG Iβ CNB-B (compared to cGMP), and that this partial agonism cannot be explained by a simple reversal of the two-state inactive/active conformational equilibrium that rationalizes cGMP agonism, but is instead the result of cAMP-bound CNB-B sampling a distinct third conformational state.^13,16–19

We have also dissected the distinct contributions of the 2-NH₂ and 6-oxo guanine base substituents to the cGMP-versus-cAMP selectivity of PKG Iβ CNB-B.²⁰ In particular, the 2-NH₂ drives the anti-to-syn transition of the unbound cyclic nucleotide, while the 6-oxo is needed to fully engage the C-terminal lid of PKG Iβ CNB-B with the bound cyclic nucleotide.²⁰ However, only limited knowledge is currently available on how cyclic nucleotide binding allosterically rewires the network of hydrogen bonds in PKG Iβ CNB-B, and how such rewiring contributes to cGMP-versus-cAMP selectivity.^21,22 In the current work, we seek to address this gap.

Here, we extend the comparative analysis of apo, cAMP- and cGMP-bound PKG Iβ CNB-B by performing an assessment of backbone hydrogen-bond strengths within the protein via hydrogen–deuterium (H/D) fractionation factors (FFs).²¹ In this method, the protein is allowed to undergo hydrogen–deuterium (H/D) exchange in solutions with varying D₂O/H₂O mole ratios, whereby hydrogen bonds with the backbone amides of each amino acid residue influence the extent to which the backbone amides undergo H/D exchange upon equilibration.^21,23 In general, amino acid residues whose backbone amides are involved in strong hydrogen bonds prefer hydrogen over deuterium at equilibrium, and thus tend to exhibit FF values less than unity, while those involved in weak hydrogen bonds tend to exhibit FF values greater than unity, and an FF value of unity indicates a comparable distribution of deuterium and hydrogen between the backbone amide and bulk solvent.²¹ 

While fractionation factors have been used to examine catalytic proteins (i.e., enzymes),^21,24–26 to our knowledge, this method has not previously been attempted for cyclic-nucleotide-binding allosteric proteins such as PKG. Furthermore, due to its inherent differences from other methods in terms of the types of experimental data and calculations that are implemented, the FF-based analysis is a potentially useful complement to information obtained from other analyses, such as other NMR-based analyses,^13,27 or hydrogen bonding observed in PKG structures solved by neutron diffraction.^14,15 For instance, FFs could potentially identify some allosteric elements that have otherwise been previously overlooked by other methods.²⁸ Indeed, here we show that our FF-based analysis identifies not only allosteric features of PKG that are consistent with previous data, but also allosteric properties that were not previously reported, highlighting the FF-based analysis as a useful complement to other more conventional analyses of allostery.

A uniformly ¹⁵N-labeled PKG Iβ(219–369) construct was expressed in Escherichia coli and purified as described previously.¹³ To ensure a consistent protein concentration, a single stock solution of PKG Iβ(219–369) was used to prepare all of the samples for each state of the construct. All NMR samples contained ∼75 µM PKG Iβ(219–369), 50 mM Tris buffer (pH 7.0), 100 mM NaCl, 1 mM DTT, and 0.02%(w/v) NaN₃. For the cGMP- and cAMP-bound PKG Iβ(219–369) samples, 2 mM of each cyclic nucleotide was added to achieve saturated ligand binding, thus minimizing the influence of differing ligand binding affinities on the results of the analysis. To achieve varying concentrations of D₂O (i.e., 5%, 20%, 40%, and 60% by volume) in the samples, 50 mM Tris buffer (pH 7.0) with 100 mM NaCl, 1 mM DTT, and 0.02%(w/v) NaN₃ was prepared in D₂O, and added to each sample in a sufficient quantity to achieve the desired D₂O concentration.²¹ A maximum D₂O concentration of 60% was used for reasons outlined by Li et al. (2015).²¹ To ensure that the H/D exchange within the samples reached equilibrium, the samples were incubated at room temperature for 24 h (as prescribed by Li et al., 2015)²¹ and stored at 4 °C¹³ for at least an additional week before the NMR spectra were acquired.

A series of two-dimensional (¹H, ¹⁵N) heteronuclear single-quantum coherence (HSQC) NMR spectra were acquired for apo, cGMP- and cAMP-bound samples of PKG Iβ(219–369) containing 5%, 20%, 40%, and 60% D₂O [Fig. 1(e)]. All spectra were acquired at 306 K on a Bruker Avance 700-MHz NMR spectrometer equipped with a 5 mm TCI cryoprobe, with 16 scans and a recycle delay of 3.0 s. The spectra were processed with NMRPipe²⁹ and analyzed using Sparky³⁰ to determine the peak intensities, and peak assignments for the spectra were obtained from standard three-dimensional triple-resonance NMR spectra as described previously.¹³ 

The results from the linear least-squares fitting were then filtered to remove φ values with poor linear fits (i.e., fits with R² < 0.70), as well as φ values exceeding 2.0 (which are affected by rapid exchange and T₁ value variations), since these φ values are considered to be less reliable.²¹ Notably, the residues whose φ values exhibited poor linear fits were confined predominantly to the inner β-strands of the PKG construct’s β-subdomain, and may thus be a result of the slower H/D exchange observed previously for residues in these regions due to the structural rigidity of this part of the protein.¹³ The results were further filtered by removing data for residues that exhibited an HSQC peak overlap, visible H/H-exchange spectrum peaks,¹³ and/or HN-NOE values <0.5,¹³ as these residues may exhibit artifacts in their HSQC peak intensities that could affect the calculated fractionation factors.

As an initial assessment of the FFs [Figs. 1(e) and 1(f) and 2; Fig. S1], we compared the results for the apo state of PKG Iβ CNB-B [Fig. 3(a)] to those from our previous analyses¹³ to assess whether the new analysis successfully captured key trends in the structural dynamics of PKG Iβ CNB-B. Indeed, FF values less than 0.5 are confined primarily to the inner β-strands of the β-subdomain [Fig. 3(a)], which include some of the most protected residues in PKG Iβ CNB-B according to previous H/D protection-factors data.¹³ For example, several residues of the β3 (i.e., residues 274–277) and β8 (i.e., residues 327–330) strands exhibit FF values significantly less than 0.5 [Fig. 3(a)], suggesting strong backbone hydrogen bonding within these regions. On the contrary, the β4-5 loop (i.e., residues 286–294) of the β-subdomain displays a tendency for FF values greater than unity [Fig. 3(a)], suggesting weak backbone hydrogen bonding within this region, which is consistent with the marked flexibility observed in the β4-5 loop from previous HN-NOE and H/D protection-factors data.¹³ 

As a further assessment of the FFs, we checked the results for the cyclic-nucleotide-bound states of PKG Iβ CNB-B [Figs. 3(b)–3(d); Fig. S1] against those from our previous analyses. In particular, the C-terminal half of the αB-helix demonstrates a tendency for stabilization of backbone hydrogen bonding in the cGMP-bound state relative to the cAMP-bound state [Figs. 3(d) and 4], in agreement with the previous H/D protection-factors data.¹³ In addition, a visible stabilization of backbone hydrogen bonding is observed for the β5 strand in the cGMP-bound state relative to both the apo and cAMP-bound states [Figs. 3(b)–3(d)], which is likely due to stabilization via previously reported cGMP-selective interactions between the guanine base and β5-strand residue R297.^13–15 In fact, the formation of a strong hydrogen bond between cGMP and the R297 guanidinium moiety has been reported to contribute to high cGMP binding selectivity in the CNB-B. This hydrogen-bond is not seen with bound cAMP, leading to a weaker affinity for cAMP binding.¹³ In general, the β-strands demonstrate an overall preference for hydrogen-bond strengthening, rather than weakening, upon cyclic nucleotide binding [Figs. 3(b) and 3(c)], as expected based on the previous H/D protection-factors data.¹³ Finally, the β2-3 loop exhibits a tendency for stabilization of backbone hydrogen bonding in the cGMP-bound state relative to both the cAMP-bound and apo states [Figs. 3(b)–3(d)], in agreement with the previous HN-NOE and H/D protection-factors data.¹³ 

In addition to the cGMP-bound vs cAMP-bound and cyclic-nucleotide-bound vs apo PKG Iβ CNB-B differences described above, additional trends were observed from the FF data that were seemingly non-intuitive and had previously remained elusive to other methods.^13–15 First of all, the β-subdomain demonstrates an overall tendency for a mix of backbone hydrogen-bond strengthening and weakening in the cGMP-bound state relative to the cAMP-bound state [Fig. 3(d)]. For instance, unlike the β2-3 loop, the adjacent phosphate-binding cassette [PBC; Figs. 1(b)–1(d)] exhibits a weakening of hydrogen bonding in the cGMP-bound state relative to both the cAMP-bound and apo states [Figs. 3(b)–3(d)], contrary to the expected trend of stabilization upon binding of the endogenous allosteric effector for PKG. Second, although the N3A region [Figs. 1(b)–1(d)] is far removed from the cyclic-nucleotide-binding site, and seemingly demonstrates a preference for hydrogen-bond strengthening in the cyclic-nucleotide-bound states relative to the apo state [Figs. 3(b) and 3(c)], it exhibits a visible tendency for hydrogen-bond weakening in the cGMP-bound state relative to the cAMP-bound state [Figs. 3(d) and 4], in contrast to the αB-helix and contrary to expectations.

While seemingly contrary to expectations, the hydrogen-bond weakening observed for the PBC may be associated with the increased ps–ns dynamics that arise in the middle of the PBC upon cyclic-nucleotide binding, as observed from the previous HN-NOE data.¹³ Meanwhile, the changes in hydrogen bonding observed for the N3A region suggest that hydrogen-bond changes can occur even upon rigid-body movement, such as the movement of the N3A relative to the β-subdomain that occurs upon cyclic-nucleotide binding [Figs. 1(b)–1(d)].^13–15 More notably, it is possible that the weakening of hydrogen bonds observed in these regions may serve to balance the effects of the strengthening of hydrogen bonds observed in other parts of the protein, thus contributing to the reversibility of the allosteric inactive-to-active conformational shift that occurs within the protein upon cyclic-nucleotide binding [Figs. 1(b)–1(d)].^13–15 

As an additional assessment of backbone hydrogen-bond strengths within apo, cAMP- and cGMP-bound PKG Iβ(219–369), we computed backbone hydrogen-bond free-energy differences from the respective amino acid residue fractionation factors [Fig. S2], in accordance with the methodology reported by Cao and Bowie.³⁶ Overall, the computed free-energy differences reflected trends of hydrogen-bond stabilization/destabilization that were very similar to those observed from fractionation factor differences [Figs. 3(b)–3(d)], thus reaffirming our observations of hydrogen-bond stabilization/destabilization based on fractionation factor differences.

It has previously been observed that cGMP binding enables formation of strong protein–ligand hydrogen bonds at select residues, which has been used to explain the mechanism of PKG CNB-B selectivity for cGMP.¹⁵ In contrast, the balancing of hydrogen bond strengths reported here for the CNB-B:cGMP complex provides fine-tuned interactions that lead to a more optimized mode of allosteric activation.³⁷ This balancing of strengthened and weakened hydrogen bonds could thus represent a further facet of cGMP selectivity that was previously overlooked. Additionally, hydrogen-bond weakening upon cGMP binding helps avoid protein over-rigidification that may otherwise trap the protein in its active state by excessively slowing cGMP dissociation and subsequent phosphodiesterase-catalyzed cGMP hydrolysis, which are required for cGMP signal termination.³⁸ Furthermore, the identification of this previously overlooked feature of PKG Iβ CNB-B by our FF-based analysis highlights the potential of this approach as a complement to information obtained from other analyses.

In the current work, we extended our previous comparative analysis of apo, cAMP- and cGMP-bound PKG Iβ CNB-B¹³ by performing an assessment of protein backbone hydrogen-bond strengths via residue-specific H/D fractionation factors (FFs).²¹ While previously applied to catalytic proteins (i.e., enzymes), this method, to our knowledge, had not previously been applied to cyclic-nucleotide-binding allosteric proteins such as PKG. The FF-based analysis identified a number of trends in PKG Iβ CNB-B dynamics that are consistent with previous observations from independent analyses,^13–15 including the observation of structural rigidity within the CNB-B β-subdomain, and the stabilization of the αB-helix C-terminus and β2-3 loop structural elements of CNB-B upon cGMP binding (Fig. 3).

Interestingly, the FF-based analysis also identified select allosteric elements that had previously been overlooked by other methods, thus highlighting its potential as a useful complement to information obtained from other methods.²⁷ In particular, the PBC and N3A regions both exhibit a weakening of hydrogen bonding in the cGMP-bound state relative to the cAMP-bound state, contrary to the expected trend of stabilization, and in general, the β-subdomain demonstrates a mix of backbone hydrogen-bond strengthening and weakening. Notably, the changes in hydrogen bonding observed for the N3A region suggest that hydrogen-bond changes can occur even for rigid-body motions, such as the movement of the N3A relative to the β-subdomain that occurs upon cyclic-nucleotide binding [Figs. 1(b)–1(d)].^13–15 In addition, the weakening of hydrogen bonds observed in these regions may serve to balance the effects of strengthening of hydrogen bonds observed in other parts of the protein, thus avoiding over-rigidification that may otherwise trap the protein in its active state, and contributing to the reversibility of the allosteric inactive-to-active conformational shift [Figs. 1(b)–1(d)].^{13–15,23,28,37,38} This phenomenon points to the presence of inherent frustration within the hydrogen-bond networks, whereby not all hydrogen bonds can be simultaneously stabilized, a concept that may be applicable to other allosteric domains as well.

See the supplementary material for supplementary figures. Figure S1 shows the amino acid residue H/D fractionation factors for cAMP-bound and cGMP-bound PKG Iβ(219–369); Fig. S2 shows the backbone hydrogen-bond free-energy differences computed from the amino acid residue H/D fractionation factors for apo, cAMP- and cGMP-bound PKG Iβ(219–369).

We would like to express our appreciation to Andrew Tu, Dr. M. Akimoto, Dr. R. Ahmed, J. Huang, and Karla Martinez Pomier for their helpful discussions. This work was supported by Canadian Institutes of Health Research Grant No. 389522 (to G.M.) and Natural Sciences and Engineering Research Council of Canada Grant No. RGPIN-2019-05990 (to G.M.).

The authors have no conflicts to disclose.

The manuscript was written by B.V.S., L.D.L. and G.M. B.V.S. and G.M. conceived the project. B.V.S. acquired the data. B.V.S. and L.D.L. analyzed the data. All authors have given approval to the final version of the manuscript.

Bryan VanSchouwen: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Project administration (equal); Validation (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Leonardo Della Libera: Formal analysis (equal); Validation (equal); Writing – review & editing (equal). Giuseppe Melacini: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Methodology (equal); Project administration (equal); Resources (equal); Supervision (equal); Validation (equal); Writing – review & editing (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.