Structural basis for selective inhibition of human PKG Iα by the balanol-like compound N46

Activation of protein kinase G (PKG) Iα in nociceptive neurons induces long-term hyperexcitability that causes chronic pain. Recently, a derivative of the fungal metabolite balanol, N46, has been reported to inhibit PKG Iα with high potency and selectivity and attenuate thermal hyperalgesia and osteoarthritic pain. Here we determined co-crystal structures of the PKG Iα C-domain and cAMP-dependent protein kinase (PKA) Cα, each bound with N46, at 1.98 Å and 2.65 Å, respectively. N46 binds the active site with its external phenyl ring, specifically interacting with the glycine-rich loop and the αC helix. Phe-371 at the PKG Iα glycine-rich loop is oriented parallel to the phenyl ring of N46, forming a strong π-stacking interaction, whereas the analogous Phe-54 in PKA Cα rotates 30° and forms a weaker interaction. Structural comparison revealed that steric hindrance between the preceding Ser-53 and the propoxy group of the phenyl ring may explain the weaker interaction with PKA Cα. The analogous Gly-370 in PKG Iα, however, causes little steric hindrance with Phe-371. Moreover, Ile-406 on the αC helix forms a hydrophobic interaction with N46 whereas its counterpart in PKA, Thr-88, does not. Substituting these residues in PKG Iα with those in PKA Cα increases the IC50 values for N46, whereas replacing these residues in PKA Cα with those in PKG Iα reduces the IC50, consistent with our structural findings. In conclusion, our results explain the structural basis for N46-mediated selective inhibition of human PKG Iα and provide a starting point for structure-guided design of selective PKG Iα inhibitors.

the most prescribed medication class in the United States (2). The increasing prescription of opioid pain relievers is associated with a dramatic increase in opioid misuse, abuse, overdose, and opioid use disorder, contributing to a $504 billion economic cost in the United States in 2015 and more than 63,600 opioid overdose deaths in 2016 (2)(3)(4)(5)(6). Another major category of analgesics, COX inhibitors, has long-term cardiovascular side effects (7). Therefore, a new type of nonopioid-based pain reliever is in demand for effective pain management.
Reversible protein phosphorylation regulates all aspects of cell survival. Consequently, dysregulation of protein kinases is often involved in human diseases such as cancer (8), diabetes (9 -11), and chronic pain (12,13). More than 30 protein kinase inhibitors have been approved by the Food and Drug Administration in the past 23 years, with the majority of them targeting tyrosine kinases for cancer treatment (14).
Beyond its role as a central regulator of smooth muscle tone, cGMP-dependent protein kinase (PKG) 3 I␣ activation in nociceptive neurons results in long-term hyperexcitability that causes chronic pain (15,16). PKG I␣ is also a crucial modulator of cortical neuronal activity in pathological pain; thus, it represents a novel target for developing analgesic therapeutic agents (17). A recent study demonstrated that N46, a derivative of the fungal metabolite balanol, inhibits PKG I␣ with high potency and selectivity, resulting in attenuation of thermal hyperalgesia and osteoarthritic pain in rats (18).
PKG I␣ belongs to the AGC kinase family and consists of N-terminal regulatory (R) and C-terminal catalytic (C) domains ( Fig. 1A) (19,20). PKG I␣ shares a large degree of sequence similarity with cAMP-dependent protein kinase (PKA). In particular, the PKG I␣ C-domain shows 45% sequence identity with the PKA C␣, consistent with their similar structures. The C-domain includes small and large lobes that consist of mostly ␤ strands and ␣ helices, respectively. A highly acidic active site is formed between the two lobes that binds Mg 2ϩ , ATP, and substrates. In the absence of cGMP, the activity of PKG I␣ is negatively regulated by the interaction between the R-and C-domains (21,22).
Three classes of small-molecule PKG inhibitors have been widely used for functional studies of PKG (23, 24). The first class is the R-diastereomer of the phosphorothioate analogs of cGMP, including Rp-cGMPS (25). This compound binds the R-domain and stabilizes its inactive state without causing conformational changes required for activation (26). The second class consists of small molecules that compete with ATP by directly binding the active site within the C-domain. These reagents include H-89, balanol, and KT-5823 (27)(28)(29)(30)(31)(32). The third class includes peptide inhibitors that also bind the active site and prevent substrate binding. However, all of these inhibitors lack potency, specificity, and activity in vivo. For example, Rp-cGMPS is not potent (K i ϭ 20 M) and nonselectively inhibits other cyclic nucleotide effectors, such as phosphodiesterase and PKA (23). KT-5823 also inhibits other kinases and may not inhibit PKG in intact cells (33). Despite its high potency in vitro, DT-2 does not inhibit PKG in platelets or in rat mesangial cells (34). As mentioned, balanol is a potent inhibitor of PKG but also inhibits other serine and threonine kinases such as PKA, most PKC isoforms, and Ca 2ϩ -dependent protein kinase (30,35). To improve inhibitor selectivity for PKG I␣, a homology model of PKG I␣ docked with balanol was generated based on the crystal structure of the PKA C␣-balanol complex. Several amino acid differences near their binding pockets were identified, and balanol was modified to preferentially interact with PKG I␣-specific residues (18). In particular, the homology model showed that Thr-88 of PKA C␣ corresponds to Ile-406 in PKG I␣ (16). To exploit this difference, a propoxy group was added to the external phenyl ring (ring D) of the balanol derivatives to

N46 specificity for PKG I␣
selectively interact with Ile-406 of PKG I␣. Although one such compound, N46, was reported to have a high selectivity and potency for PKG I␣ over PKA C␣, the exact molecular basis of its improved affinity and specificity is unknown.

Results and discussion
Several crystal structures have been solved for mammalian PKG I, but these are of various fragments of the R-domains (36 -39). Because N46 directly targets the C-domain of PKG I␣, we first obtained an isolated C-domain that is fully active. To understand the molecular basis of N46's high selectivity for human PKG I␣, we determined co-crystal structures of N46 bound to the human PKG I␣ C-domain and human PKA C␣ at 1.98 Å and 2.65 Å, respectively ( Fig. 1 and Fig. S1 and Table S1). The PKG I␣ C-N46 complex was crystallized in the P4 2 space group with one molecule in the asymmetric unit. The molecule shows clear electron density for the bound N46 and the C-domain used for crystallization, excluding the first 10 residues at the N terminus (Fig. 1B). The PKA C␣-N46 complex was crystallized in the P3 1 21 space group with one molecule in the asymmetric unit ( Fig. 1C and Fig. S1). The final model shows clear density for the C␣ subunit except for the first 10 residues. Unlike previous PKA C␣ structures, the N-terminal ␣A helix disengages from the catalytic core because of unusual crystal packing interactions (Fig. S2). The ␣A helix of a neighboring symmetry mate occupies the equivalent position seen in previous structures and provides the same set of interactions with the catalytic core.
The overall structure of the PKG I␣ C-N46 complex is similar to the AMP-PNP-bound structure. 4 It shows a closed conformation with the fully ordered glycine-rich loop and C-terminal tail (Fig. 1B). N46 binds to a pocket that extends from the hinge region to the inner surface of the ␣C helix and spans ϳ20 Å ( Fig. 2A). The pocket can be divided into three subsites according to the interaction between the PKG I␣ C-domain and AMP-PNP: the adenine, the ribose, and the extended triphosphate subsites. N46 binds to all three subsites in the extended active site of the PKG I␣ C-domain ( Fig. 2A).
The B-ring (pyrrolidine ring), which connects the A-ring to the C-ring, interacts with the acidic ribose subsite directly and indirectly through water molecules (Fig. 2B). The ribose subsite consists of the hinge and activation loop residues. The side chains of Glu-445 at the hinge and Asp-502 at the activation loop form hydrogen bonds with the amine groups on either side. Two water molecules bridge the interaction with N46 at this subsite. These water molecules are located adjacent to the amide connecting the B-ring to the A-ring, bridging them to the side chains of Glu-445 and Asp-502 through hydrogen bonds.
The C-ring (phenyl ring) interacts with ␤1 and the glycinerich loop through van der Waals (VDW) contacts (Fig. 2B). In particular, Val-368, Gly-369, and Gly-370 are within 3.4 -3.8 Å from the C-ring, providing VDW interactions. Because these interactions are through backbone atoms, this region does not provide any PKG-selective contacts.
The D-ring (external phenyl ring) with the propoxy and methoxy groups provides two interactions that are PKG-specific and may explain its high selectivity for PKG I␣ over PKA C␣ (Fig. 2B). In designing N46, the propoxy group was added to the phenyl ring to provide a preferential interaction with Ile-406 of PKG I␣ over PKA C␣, which has a threonine (Thr-88) at the analogous position ( Fig. S3) (18). However, the structure shows that the methoxy group points toward the side chain of Ile-406 instead, whereas the propoxy group points toward the

N46 specificity for PKG I␣
glycine-rich loop, each providing hydrophobic interactions. Additionally, the D-ring, along with the carbonyl group that connects the D-ring to the C-ring, docks to the tip of the glycine-rich loop through hydrogen bonds and VDW interactions. The interconnecting carbonyl group hydrogen bonds with the backbone amide of Phe-371 and uniquely forms a lone pairinteraction with its side chain. The D-ring and the side chain of Phe-371 are off-centered, and they interact through a parallel displaced interaction.
The overall interactions between the PKA C␣ subunit and N46 are similar to those in the PKG I␣ C-N46 complex because most of the contact residues are highly conserved between the two kinases (Fig. 3A). However, the structure shows differences that may explain a higher IC 50 value for the PKA C␣ subunit.
The A-ring binds the adenine subsite, and the interactions in this region are essentially the same as in PKG I␣. These include hydrogen bonds between the A-ring and the backbone atoms of Glu-121 and Val-123 at the hinge and VDW contacts with a hydrophobic pocket consisting of Leu-49, Val-57, Ala-70, Val-104, Met-120, Leu-173, and Phe-327 (Fig. 3B). Tyr-122 at the hinge region provides an additional hydrophobic contact unseen in PKG I␣ because Tyr-122 replaces Ala-440 of PKG I␣. Although the B-ring similarly docks onto the ribose subsite, its amine group interacts only with the hinge residue Glu-127 through a hydrogen bond, not with the activation loop residue Asp-184 (Fig. 3B). Unlike Asp-502 of PKG I, which forms a hydrogen bond with N46 (Fig. 2B), the side chain of Asp-184 points away and no longer interacts with N46 in PKA. The C-ring similarly docks to ␤1 and the glycine rich loop and interacts with the backbone atoms of Thr-51, Gly-52, and Ser-53.
The D-ring interacts less strongly with PKA C␣ compared with PKG I␣ because of two PKA-specific residues, Phe-54 and Thr-88 (Fig. 3B). The structure shows that the side chain of Phe-54 at the tip of the glycine-rich loop rotates ϳ30°and provides a weaker T-shaped interaction with the D-ring. Because of this rotation, the interconnecting carbonyl no longer forms a lone pair-interaction with the aromatic Phe-54. In addition, the side chain of Thr-88 of the ␣C helix is smaller and less hydrophobic than that of Ile-406 of PKG I␣, thus providing a much weaker hydrophobic interaction with the methoxy group (3.7 Å) (Fig. 3B). The structural alignment with the PKG I␣ C-N46 complex suggests that a steric clash between the side chain of the preceding residue, Ser-53, and the propoxy moiety causes the rotation of the Phe-54 side chain. As seen in Fig. S3, N46 moves away slightly from the active site because of the steric clash. This allows more room between the D-ring and the glycine-rich loop, causing the rotation of the Phe-54 side chain.
The reported inhibition constant of balanol for PKA C␣ is 1.6 nM, whereas N46 inhibits PKA with an IC 50 of 1.0 M, showing an over 600-fold increase (18). Comparing the PKA C␣-N46 complex with the PKA C␣-balanol complex reveals that this reduction is mostly due to loss of hydrogen bonds (Fig. 4). The PKA C␣-balanol complex shows 12 nonsolvent mediated hydrogen bonds and large numbers of VDW interactions between the extended active site and balanol. The PKA C␣-N46 complex shows that, although the most of the VDW contacts are preserved, N46 forms only 6 direct hydrogen bonds because of the modifications on the C and D rings.
Substituting the phenol of balanol (Fig. 4A, ring a) with the indazole ring of N46 (Fig. 4B, ring A) does not reduce the number of hydrogen bonds and VDW contacts with the adenine subsite. In the PKA-balanol complex, the phenol forms hydrogen bonds with the same backbone atoms of Glu-121 and Val-123 at the hinge region (Fig. 4A) as the indazole. However, replacing a more puckered azepane ring of balanol (Fig. 4A, ring  b) with a less puckered pyrrolidine of N46 (Fig. 4B, ring B) results in one additional hydrogen bond at the ribose subsite. The puckered azepane ring interacts mainly with a conserved catalytic loop residue, Glu-170, through its backbone (Fig. 4A). In the PKA-N46 complex, the less puckered pyrrolidine ring brings its amine group within a hydrogen bonding distance of the Glu-127 side chain, forming a new hydrogen bond (Fig. 4B).
Removing two hydroxyl groups from the c-ring of balanol (Fig. 4A) disrupts all four hydrogen bonds with the triphosphate subsite. In the PKA C␣balanol complex, two hydroxyl groups on the c-ring interact with Gly-55, Lys-72, and Asp-184

N46 specificity for PKG I␣
through four hydrogen bonds. In major contrast, the C-ring of N46 (Fig. 4B) no longer binds these residues and interacts with the glycine-rich loop through VDW contacts.
Last, substituting a carboxyl group and a hydroxyl group on the d-ring of balanol (Fig. 4A) with a bulky and hydrophobic propoxy group and a fluorine atom, respectively (Fig. 4B, ring D of N46), significantly weakened the interaction with the glycine-rich loop and the ␣C helix. In the PKA-balanol complex, the carboxyl group on the d-ring forms strong hydrogen bonds with both the side chain and backbone of Ser-53 at the glycinerich loop, whereas the 3-hydroxyl group binds the side chains of Glu-91 and Lys-72 through hydrogen bonds. Additionally, the d-ring is oriented parallel to the side chain of Phe-54, allowing a parallel -stacking interaction between them as well as a lone pairinteraction between the carbonyl group and Phe-54. None of these interactions are preserved in the PKA-N46 complex, although a new hydrogen bond forms between the propoxy group and the backbone amide of Ser-53.
We noticed that the side chain of Phe-54 remains parallel to the d-ring when bound to balanol and rotates when bound to N46 (35). The balanol-bound PKA structure shows that this is because balanol binds deeper into the pocket, allowing a parallel -stacking interaction with Phe-54 (Fig. S4A). In contrast, N46 cannot bind as deeply because of its bulky methoxy group, resulting in enough space between the D-ring and Phe-54, which allows Phe-54 to rotate to provide VDW contact with the D-ring (Fig. S4B).
To test the molecular basis for N46's PKG I␣-selective inhibition over PKA, we mutated the unique contact residues in PKG I␣ to those in PKA and vice versa. Specifically, for PKG I␣, we mutated Gly-370 and Ile-406 to the corresponding residues in PKA C␣ (i.e. G370S and I406T). We also mutated these two PKA C␣ residues into the corresponding PKG I␣ residues (S53G and T88I). For PKG I␣, we generated two single mutants (G370S and I406T) and a double mutant (G370S/I406T). For PKA C␣, we only generated a double mutant (S53G/T88I). We then measured IC 50 values using in vitro kinase assays (Fig. 5). N46 showed an IC 50 of 43 nM for WT PKG I␣, whereas it inhibited PKA C␣ with an IC 50 of 1030 nM, showing an ϳ24-fold difference in selectivity. The PKG I␣ single mutants were inhibited with higher IC 50 values of 90 nM and 142 nM for G370S and I406T, respectively. The double mutant PKG I␣ showed an IC 50 value of 301 nM, demonstrating a synergistic effect of the two mutations. In contrast, the PKA C␣ double mutant showed an IC 50 of 552 nM, which is almost half of that seen in WT C␣. The higher IC 50 values seen in the PKG I␣ mutants and the lower value of the PKA C␣ double mutant compared with their respective WT are consistent with our structural findings.
Despite lack of data on inhibition constants of N46 for other kinases, our model of a PKC␣ isoform (PDB code 3IW4) docked with N46 suggests that N46 is a poor inhibitor for the PKC␣ isoform ( Fig. S5) (40). The model shows that the tip of the glycine-rich loop curls in toward the active site and clashes with the C-ring. In particular, Phe-350 at the glycine-rich loop occupies the part of the pocket the C-ring binds, suggesting that N46 would interact poorly with PKC␣. Consistent with the model, Sung et al. (18) reported that, at 0.75 M of N46, PKC␦ had 68% residual activity, whereas PKG I␣ was completely inhibited with 0% residual activity .
Our structural and biochemical data suggest new strategies for generating N46 derivatives with higher selectivity for PKG I␣ over PKA C␣. Amino acid sequences at the hinge region and ␤7 that make up the left edge and the base of the adenine pocket are different in PKG I␣ compared with PKA C␣. PKG I␣ has Ala-Cys-Leu (residues 440 -442) at the hinge, whereas PKA has Tyr-Val-Pro (residues 122-124) (Fig. S6). This causes PKG I␣ to have a wider adenine pocket compared with the PKA C␣ subunit (Fig. 6). Additionally, at the base of the adenine pocket, PKG I␣ has an isoleucine (Ile-491 at ␤7) replacing a leucine (Leu-173) of PKA C␣, providing a slightly deeper pocket. Thus, to improve selectivity for PKG I␣, bulkier heterocyclic rings could be engineered in N46 to fill this unique pocket. Also, a reactive group can be placed here to covalently link to Cys-441 because PKA lacks a cysteine residue at the analogous position (Fig. S6). During the initial design of N46, the propoxy group was added to increase its interaction with Ile-406 at the ␣C helix. However, our structures revealed that this group points in an opposite direction (toward the glycine-rich loop) and interacts with Gly-370 instead. Thus, it may be possible to add an additional ethyl or propyl group here to improve interaction with PKG I␣. This modification should cause steric hindrance with Ser-53 of PKA at the glycine-rich loop while providing additional nonpolar interactions with Gly-370 in PKG I␣. In conclusion, our structural and biochemical data in part explain N46's selectivity for PKG I␣ and provide a starting point for structure-guided design of selective PKG I␣ inhibitors.

Expression and purification of the hPKG I␣ C-domain
The sequence encoding the human PKG I␣ C-domain (327-671) was cloned into the pBlueBacHis2A vector. The vector was modified to put a tobacco etch virus (TEV) protease site just before the PKG coding sequence. The protein was expressed in High Five cells. The cells were grown at 28°C and infected at a multiplicity of infection of 3.0 for 32 h. All cells were lysed in buffer A (25 mM Tris (pH 7.5), 500 mM NaCl, and 1 mM ␤-mercaptoethanol) with a Constant Systems TS cell disrupter (Daventry Northants, UK) and cleared via ultracentrifugation. The supernatant was loaded onto a Bio-Rad Nuvia nickel affinity column, washed with buffer A and eluted with buffer A containing 300 mM imidazole. The His tag was removed by incubating the sample with TEV protease at 4°C overnight. TEV was removed from the protein sample by performing a second nickel affinity chromatography and collecting the flow-through fractions. The sample was further purified by anion exchange chromatography (Mono Q 10/100 GL, GE Healthcare) in buffer B (25 mM Tris (pH 7.5) and 1 mM ␤-mercaptoethanol) with and without 1 M sodium chloride. This was followed by size exclusion chromatography (Hiload 16/60 Superdex 75, GE Healthcare) in buffer C (25 mM Tris (pH 7.5), 150 mM sodium chloride, and 1 mM tris(2-carboxyethyl)phosphine (TCEP)).

Expression and purification of hPKA C␣
The pET15b plasmid encoding human PKA C␣ was transformed into BL21 (DE3) Escherichia coli cells. The cells were grown at 37°C until OD 600 ϭ 1.0 was reached. The expression was induced by 0.5 mM isopropyl ␤-D-thiogalactopyranoside at 18°C for 18 h. The cells were then lysed by the Constant Systems TS cell disruptor in buffer A. The lysate was then cleared by ultracentrifugation and membrane filtration. The supernatant was applied onto a GE His-Trap column for nickel affinity purification. The protein was eluted by buffer A containing 300 mM imidazole. The His tag was removed by incubating the protein with TEV protease at 4°C overnight, followed by a second nickel affinity chromatography. The protein was then further

N46 specificity for PKG I␣
purified by anion exchange chromatography (anion exchange chromatography, Mono Q 10/100 GL, GE Healthcare) in buffer D (25 mM potassium phosphate (pH 7.0) and 1 mM ␤-mercaptoethanol) with and without 1 M sodium chloride. This was followed by size exclusion chromatography (Hiload 16/60 Superdex 75, GE Healthcare) in buffer C.

Crystallization and structure determination
To obtain crystals of the PKG I␣ C-domain-N46 complex, 14 mg ml Ϫ1 of the PKG I␣ C-domain was incubated with 1 mM N46 for 30 min at room temperature. Crystals were obtained by mixing 1 l of the C-domain-N46 complex solution with 1 l of well solution (24% w/v PEG 1500 and 20% v/v glycerol) and 0.2 l of additive (30% w/v trimethylamine N-oxide dihydrate) at 22°C. To obtain crystals of the PKA C␣-N46 complex, 12 mg ml Ϫ1 of PKA C␣ was incubated with 1 mM of N46 for 30 min at room temperature. Crystals were obtained by mixing 0.2 l of the C-domain-N46 complex solution with 16% (w/v) PEG 8000, 0.04 M potassium phosphate (monobasic), and 20% (v/v) glycerol. PKG I␣ C-domain and PKA C␣ crystals were cryoprotected by paratone, and diffraction images were collected at the Advanced Light Source (Berkeley, CA). Data were processed using CCP4.iMosflm (41). The structures of the PKG I␣ C-domain-N46 and PKA C␣-N46 complexes were determined by Phaser-MR using AMP-PNP-bound PKG I␣ C-domain (PDB code 6BG2) and balanol-bound PKA C␣ (PDB code 1BX6) as molecular replacement probes (42). Both final structures were manually built using Coot and refined using Phenix.Refine (43,44). The figures were generated using PyMOL (Schrödinger, LLC).

In vitro kinase assays
FLAG-tagged WT and mutant PKG I␣ proteins were purified from transiently transfected 293T cells as described previously (45). PKA C␣ WT and its mutant were purified as described above. The purified kinases were diluted in kinase dilution buffer (10 mM potassium phosphate (pH 7.0), 1 mM EDTA, 35 mM ␤-mercaptoethanol, and 0.1% BSA) so that the reactions produced ϳ10 5 counts per reaction (corresponding to about 36 pmol phosphate incorporation). Reactions were initiated by adding 10 l of diluted kinase to 5 l of 3ϫ kinase reaction mixture (120 mM HEPES (pH 7.4), 1.56 mg/ml Kemptide, 30 mM MgCl 2 , 300 M ATP, 360 Ci/ml [␥-32 P]ATP and 30 M cGMP) containing variable amounts of the N46 inhibitor diluted in DMSO (control assays contained DMSO alone). Reactions were run for 1.5 min at 30°C and stopped by spotting on P81 phosphocellulose paper. Unincorporated [␥-32 P]ATP was removed by washing P81 paper 4 ϫ 2 liters in 0.45% O-phosphoric acid. 32 P incorporation was measured by liquid scintillation counting. The data were analyzed using GraphPad Prism 7.