|Home | About | Journals | Submit | Contact Us | Français|
The cGMP-dependent protein kinase (PKG) serves as an integral component of second messenger signaling in a number of biological contexts including cell differentiation, memory and vasodilation. PKG is homodimeric and large conformational changes accompany cGMP binding. However, the structure of PKG and the molecular mechanisms associated with protomer communication following cGMP-induced activation remain unknown. Here we report the 2.5 Å crystal structure of a regulatory domain construct (aa 78-355) containing both cGMP binding sites of PKG Iα. A distinct and segregated architecture with an extended central helix separates the two cGMP binding domains. Additionally, a previously uncharacterized helical domain (switch helix) promotes the formation of a hydrophobic interface between protomers. Mutational disruption of this interaction in full length PKG implicates the switch helix as a critical site of dimer communication in PKG biology. These results offer new structural insight into the mechanism of allosteric PKG activation.
As key members of the AGC family of protein kinases, all mammalian PKG isoforms (Iα, Iβ, II) share a similar domain configuration (Figure 1A). At their N-termini, coiled-coils promote a parallel homo-dimeric configuration, followed by an auto-inhibitory domain (AI) and two tandem cyclic nucleotide (cNT) binding sites, which cooperatively regulate C-terminal catalytic activity (for review, see Francis and Corbin, 1994; Hofmann et al., 2009; Pfeifer et al., 1999; Vaandrager et al., 2005). The current model of PKG Iα holoenzyme assembly and mechanism of activation stresses the importance of the AI domain (aa 64-78) within the N-terminal region (Heil et al., 1987; Wolfe et al., 1989). This flexible hinge segment has a pseudo-substrate sequence and is believed to auto-inhibit the C-terminal catalytic center in the enzyme's dormant state (Kennelly and Krebs, 1991; Landgraf and Hofmann, 1989; Ruth et al., 1991). Auto-phosphorylation at various N-terminal residues appears to weaken this intramolecular inhibitory interaction, however, only cGMP binding to both the A- and B-domains induces the conformational change necessary for full kinase activity (Aitken et al., 1984; Chu et al., 1997; Francis et al., 2002). Although it is believed that the dimer stays intact upon cGMP-induced activation, the overall conformational change is substantial (Alverdi et al., 2008; Richie-Jannetta et al., 2006; Zhao et al., 1997). A significant molecular elongation centered about both the N-terminal dimerization domain (D/D) as well as the C-terminal catalytic domain accompanies cGMP-induced activation. It is at this level of allosteric modulation that underscores the most distinct difference between PKG and the cAMP-dependent protein kinase (PKA). While the regulatory and catalytic subunits of PKA dissociate upon cyclic nucleotide-induced activation (Gill and Garren, 1971), PKG harbors these components on the same polypeptide chain (Sandberg et al., 1989; Takio et al., 1984; Wernet et al., 1989). Another notable difference for all cyclic nucleotide kinases is their distinct structural order of high and low affinity cyclic nucleotide binding. For PKG I, the high affinity cyclic nucleotide sites are closer to the N-termini, whereas for PKA I, PKA II and PKG II, the order of high and low nucleotide binding affinity sites is reversed (Reed et al., 1996). These structural differences suggest different means of activation by cyclic nucleotides and a comprehensive understanding of the mechanisms underlying cGMP-mediated PKG activation will invariantly require high-resolution biophysical studies.
To better understand the molecular details of PKG regulation, we solved the 2.5 Å crystal structure of a regulatory domain fragment of PKG Iα thought to be central for conveying allosteric modulation within the enzyme. The structure (amino acid residues 78-355) presents both tandem cGMP-binding domains and identifies a novel helical subdomain (switch helix, SW) that follows cGMP-binding site B (Figure 1B). The A-site is cAMP-bound and confirms a previously suggested disulfide bond between Cys117 and Cys195. The B-domain is not cNT-occupied, but rather stabilized by hydrophobic contacts with the SW from the other protein molecule in the asymmetric unit. This hydrophobic interaction highlights an unexpected natural interface, disruption of which alters the kinetic profile of full length PKG Iα. Therefore, the PKG78-355 structure expands our understanding of the landscape of PKG domain assembly and sheds new light on the molecular mechanism of holoenzyme formation and activation of PKG Iα.
The greatest structural distinction between the two major cyclic nucleotide regulated protein kinases, PKA and PKG, is the fact that PKG maintains both its regulatory and catalytic elements on the same polypeptide chain (Gill et al., 1977), while PKA is divided into subunits (Gill and Garren, 1971). To better understand how these molecular differences contribute to the unique characteristics of each kinase, we crystallized the central portion of the regulatory domain of PKG Iα. While prokaryotic expression of PKG constructs containing the catalytic domain yields inactive, misfolded protein that is sequestered to inclusion bodies (Feil et al., 1993), previous attempts to crystallize full length PKG from mammalian systems have been unsuccessful as well. This is in part due to conformational heterogeneity stemming from mixed phosphorylation states at the N-terminus and a proteolytically exposed hinge at Arg77 (Aitken et al., 1984; Heil et al., 1987; Scholten et al., 2007). We hypothesized that exclusion of both the N-terminal D/D domain, as well as the C-terminal catalytic domain would give rise to a stable, monomeric fragment that is readily expressed in E. coli. Multiple regulatory fragments were designed to encompass both tandem cGMP-binding domains (PKG78-326, PKG78-341 and PKG78-355). The standard method for purification of recombinant PKG utilizes cyclic nucleotide affinity chromatography which saturates the allosteric binding sites through an elution with high concentrations of cNT (Dostmann et al., 1999). To prevent this artificial exposure of PKG protein to cNTs, PKG constructs were engineered with N-terminal hexa-histidine tags and purified using standard immobilized metal affinity chromatography (IMAC) methods. We made three PKG regulatory domain constructs (PKG78-326, PKG78-341 and PKG78-355). Despite high yielding, stable expression of all three constructs, only PKG78-355 yielded diffraction-quality crystals.
Crystals grew in the C2 space group with 2 molecules per asymmetric unit. Superimposition of Cα atoms from the two molecules sites gave an average root mean square deviation (rmsd) of 0.882. Initial phases were attained by molecular replacement using the CNB-A domain of PKA RIα (PDB ID, 1RGS) as a search model. The structure was refined to 2.5 Å with working and free-R factor values of 0.22 and 0.28, respectively (Table 1). Each protein molecule in the asymmetric unit contained 278 amino acids, 1 cAMP molecule and 2 phosphates.
The overall fold presents both cGMP-binding domains of the PKG Iα regulatory domain. cGMP-binding site A (residues 87-210) begins following a loop in the ‘hinge region’, for which we see clear backbone density, followed by cGMP-binding site B (residues 211-327) and a new helical subdomain we have termed switch helix (SW, residues 328-355) (Figure 1B, Supplement Figure 1A, secondary structural elements). Both cGMP-binding sites exhibit classic features of the conserved CNB fold (Figure 2A). A rigid eight-stranded β-barrel sandwiches the phosphate binding cassette (PBC) between β-strands 6 and 7 which serves as a docking site for cNTs. The β-barrel itself is flanked at the N-terminus by the αN helix, the 310 loop, and the αA helix (collectively termed the N3A motif) and C-terminally by the αB/αC helix. Surprisingly, the two CNB domains presented in this structure have been captured in different conformational states as is evident when comparing helical elements between A- and B-domains (Figure 2B). Superimposition of the two cGMP-binding yields an rmsd of 1.124. While the β-barrels superimpose well, there is variance in the helical subdomains.
The two cGMP-binding sites are separated by an elongated B/C helix at the end of the A-domain (Figure 2A). This creates a dumbbell-like topology between the two CNB domains and resembles the conformation adopted by regulatory subunits of PKA when bound to the catalytic subunits in the holoenzyme conformation (Kim et al., 2007; Wu et al., 2007). A kink in the C-helix in the middle of this cleft is a likely point at which the two CNB domains undergo cGMP-mediated structural rearrangements. Residues in this kink are clearly solvent exposed (Figure 1B, Supplement Figure 1A) and proteolysis at Arg202 readily occurs in the absence of cGMP (Chu et al., 1997; Scholten et al., 2007). Incubation with cGMP induces conformational changes that prevent such proteolysis. Furthermore, cGMP binding increases solvent protection in this region (Alverdi et al., 2008) further suggesting that the A-domain C-helix may be the center of cGMP-induced conformational changes in the regulatory domain.
The extended conformation of PKG78-355 does not appear to allow direct communication between the A- and B-domains. This suggests that the structural determinants mediating cooperative binding of cGMP are not contributed by the core of the regulatory domain. In support of this finding, previous studies have demonstrated a loss of cooperativity upon deletion of regions outside the cGMP-binding domains (Dostmann et al., 1996; Heil et al., 1987). Moreover, in PKA holoenzyme structures of both type Iα (Kim et al., 2007) and IIα (Wu et al., 2007) the catalytic subunit docks to the cleft created by the extended topology of the CNB domains in the regulatory subunits. Residue Gln78, just C-terminal to the AI domain in PKG, is positioned in the center of the two cGMP-binding sites (Figure 1B), which is where the N-terminus of PKG interacts with the catalytic core of the enzyme via the AI domain (Heil et al., 1987; Hofmann et al., 2009). While the precise docking mechanism may differ in the PKG holoenzyme, the catalytic domain is likely to sit in this cleft and participate in crosstalk between the two cGMP-binding sites similar to the architectural arrangements observed in PKA.
The docking of cNTs to a CNB domain is made possible by the conserved and mobile motif of the PBC (Figure 3A) (Diller et al., 2001; McKay et al., 1982; Su et al., 1995). This short, 14-residue helix-turn contains residues that coordinate the ribose-phosphate moiety and allows for the preferential binding of cAMP versus cGMP. Invariantly, all CNB domains from cNT-regulated protein kinases have a Glu in the 3-position and an Arg in the 12-position, which hydrogen bonds with the 2′OH and equatorial oxygen, respectively. Unique to PKG is a Thr or Ser in the 13-position of the PBC, which has been predicted to provide specificity for cGMP by providing a hydrogen bond potential with both the 2-NH2 and apical oxygen of cGMP (Shabb and Corbin, 1992; Weber et al., 1989). PKA contains an Ala in the 13-position and mutation of this residue to a Thr produces a PKA mutant that no longer discriminates between cAMP and cGMP (Shabb et al., 1990).
To assess the occupancy of PBCs in the PKG78-355 structure, initial phases from the molecular replacement solution were used to generate a simulated annealing composite omit map. In order to minimize model bias, side chains for the invariant Glu, Arg and Thr residues were omitted in each PBC. A strong positive peak was observed in the A-domain which resembled a cyclic-3′,5′-nucleotide monophosphate containing a purine moiety in the syn configuration (Figure 3B), while the B-domain appeared cNT-free. While electron density at the 2′OH of the ribose and the 6-position of the purine was evident, we could not attribute the 6-position density to being either a keto/enol or amino group. However, there was no density extending from the 2-position where the 2-amino group of cGMP should reside. This was a clear indication that the cNT occupying the A-domain was syn-oriented cAMP. HPLC analysis and UV-spectroscopy confirmed the presence of cAMP and the absence of cGMP (see methods).
The primary interactions made with the cyclic nucleotide come from residues buried within the PBC. Specific contacts are made by Glu167, Arg176 and Thr177 (Figure 3C). Glu167 forms a hydrogen bond with the 2′OH of the ribose and Arg176 docks the equatorial oxygen of the phosphate. Thr177 hydrogen-bonds with the apical oxygen of the phosphate group with both side chain hydroxyl and backbone amino groups. Modeling of syn-oriented cGMP into the A-domain maintains these same contacts and shows that the 2-amino group is primed to interact with the Thr177 side chain hydroxyl. The Glu and Arg contacts are similar to those used by the cAMP-PKA interaction and the previously predicted interactions with Thr177 are confirmed (Shabb et al., 1991; Weber et al., 1989). A schematic summary of these specific interactions with cGMP modeled into the A-domain is presented in Figure 3E.
As has been described for other CNB domain structures, cNT docking to the PBC is stabilized by a capping mechanism that sandwiches the nucleotide base between hydrophobic surface on the β-barrel, and a hydrophobic cap that moves into place upon cNT-induced structural rearrangements of the protein (Berman et al., 2005; Das et al., 2007). In the PKG78-355 structure, the backside of the adenine ring is flanked by an array of hydrophobic contacts with no obvious hydrogen bond donor or acceptor potentials (Figures 3D). The identity of the specific capping residue is not disclosed, however, because the two CNB domains are in an extended, inactive conformation and capping occurs only after ligand-mediated conformational changes cause the two CNB domains to form a more compact structure.
A second shell of regulation is provided by the β2-β3 loop. In PKA this loop stabilizes the PBC arginine and provides allosteric communication of binding events in the PBC to the B-helix. In the A-domain of PKG Iα, Arg176 is coordinated by a number of conserved residues from the β2-β3 loop (Figures 3C). A hydrophobic interaction is made with Ile130, and the backbone carbonyl and amide of Arg176 forms hydrogen bonds with the Gly133 amide and Asp134 carbonyl, respectively. Furthermore, the guanidinium group of Arg176 interacts with the carbonyl of Leu138. Communication between the β2-β3 loop and the B-helix occurs via Arg193. The backbone amide of Arg193 is bridged by the carbonyl of Ser137. This is where PKG differs from PKA. The equivalent to Ser137 is Asp170 in the PKA RIα A-domain which hydrogen bonds not only with the backbone amide of the Arg in the B-helix, but also coordinates the guanidinium of the PBC Arg. In doing so, a direct link from the PBC to the B-helix hinge is made via the β2-β3 loop. In contrast, the PKG A-domain does not utilize this same direct allosteric mechanism. This may be a critical divergence in how these two cNT-regulated protein kinases communicate binding events in their PBCs with the rest of the molecule.
The presence of disulfide bonds in PKG Iα have been reported and an oxidation-induced mechanism of activation has been proposed as complementary mechanism to cyclic nucleotide-mediated regulation of kinase activity (Burgoyne et al., 2007; Landgraf et al., 1991). PKG Iα contains 11 cysteine residues (Takio et al., 1984), 5 of which have been suggested to contribute to oxidation-induced activation. Cys42, just C-terminal of the D/D domain forms an intermolecular disulfide bond with Cys42 from the opposing protomer in the holoenzyme assembly (Burgoyne et al., 2007). It has been suggested that H2O2-induced oxidation of PKG Iα promotes kinase activation via a bridging of these two cysteines. Additionally, exposure of PKG Iα to divalent cations with positive redox potentials promotes enzyme activation via disulfide bond formation between Cys117-Cys195 and/or Cys312-Cys518 (Landgraf et al., 1991). It was unclear, however, whether Cys117-Cys195 or Cys312-Cys518 was exclusively responsible for the observed oxidation-induced activation. Despite the observations that cysteine oxidation can lead to cGMP-independent activation of PKG Iα, no molecular mechanism of activation has been proposed.
The PKG78-355 structure reveals a disulfide bridge between Cys117 in the A-helix and Cys195 at the start of the B-helix (Figure 4A). Residues immediately C-terminal of these two cysteines are solvent-exposed (Supplement Figure 1A), while a hydrophobic sheath (Phe80, Ile114, Ile191, and Ile199) surrounds the disulfide bridge and provides solvent protection (Figure 4B). In support of the oxidation-induced activation of PKG Iα previously observed (Landgraf et al., 1991), the recruitment of Phe80 to the hydrophobic sheath provides a hypothesis as to how the enzyme might be activated in the absence of cGMP. The capping of Cys117-Cys195 by Phe80 serves to order the loop that precedes the N-helix of the A-domain (Figure 4A). As the AI domain resides just adjacent to Phe80, autoinhibition of the kinase may be relieved through a movement of this loop to the hydrophobic sheath. The reorganization of this region upon oxidation of Cys117-Cys195 may disrupt the interaction between the AI and catalytic center thereby releasing the kinase from a state of autoinhibition. Furthermore, the sulfhydryl group of Cys312 sits on β8 of the B-domain and points inward towards the center of the β-barrel (not shown). It seems unlikely that Cys312 is capable of forming a disulfide bridge with Cys518 from the activation loop in the catalytic domain. Our structure supports the Cys117-Cys195 disulfide bond as being involved in the metal-induced activation of PKG Iα.
The cNT-dependent structural dynamics of CNB domains are well established (Berman et al., 2005; Kornev et al., 2008; Rehmann et al., 2007). While the β-barrel does not appear to undergo significant conformational changes upon cNT binding, the α-helical sub-domain moves In and Out relative to the barrel (Table 2). Ligand association with the PBC initiates these structural changes as several residues make specific interactions with the nucleotide and close the PBC. This structural change is communicated through hydrophobic residues to the N3A motif and B/C helix. As a result, an extended B/C helix forms a hinge and closes inward towards the β-barrel. This rearrangement is accompanied by an outward shift of the N3A motif. In PKA, these cNT-induced conformational changes bring the two CNB domains closer together, which is then stabilized by the association of a hydrophobic capping residue with the nucleotide base. Although there is sequence and spatial variability as to the identity of the cap that secures the nucleotide base to the β-barrel, this allosteric mechanism is present in all CNB domain structures described to date.
The two CNB domains presented here are in a mixed, hybrid configuration, with helical elements in both apo and cNT bound orientations (Table 2). As only the high affinity A-domain is cNT bound, the A-domain PBC is closed, relative to that of the B-domain (Figure 2B). However, the subsequent conformational changes inherent to CNB domains are not observed. An overlay of the CNB domains from PKG and PKA indicate that the overall fold is conserved, but positional differences in the helical subdomain evince the hybrid nature of this PKG structure (Figures 5A, B). The B/C helix of the A-domain is clearly extended and the N3A motif is In, representative of a CNB domain in the apo, unliganded state rather than a cNT-bound conformation (Kornev et al., 2008) (Figure 5A). Despite the A-domain PBC being occupied by cAMP, allosteric communication of this binding event with the rest of the domain is severed. Interestingly, the Cys117-Cys195 disulfide bond provides a major structural determinant for this stable transition state. The covalent bridging of the A-helix to the B-helix prevents the B/C helix from closing upon cNT binding to the PBC. In turn, the N3A motif cannot move out from the β-barrel (Figure 5A, arrow). This disulfide bond therefore uncouples communication of allosteric events in the A-domain from being transmitted to the B-domain.
The B-domain of this PKG structure is similarly locked in a hybrid conformation. This CNB domain is cNT free, thus the unbound PBC is Open and maintains an extended B/C helix (Figure 5B). However, the N3A motif is also Out, an orientation reminiscent of a cNT-bound state (Kornev et al., 2008). The B-domain N3A motif is stabilized in the Out position by a set of hydrophobic residues originating from the C-terminal end of the SW in the other molecule in the asymmetric unit (Figure 5A). The two symmetry mates are related by non-crystallographic symmetry and this interaction promotes a previously uncharacterized docking interface between PKG Iα protomers.
Remarkably, the PKG78-355 crystal structure forms a symmetry-related dimer through the formation of an interface between the SW and the opposing B-domain (Figure 6A, B). It is well established that PKG assembles into parallel homodimers, an assembly mediated by leucine zipper motifs at the immediate N-termini. The structure presented here provides the first evidence for intermolecular communication at sites distal to the classical N-terminal D/D domain. The SW extends from the B-domain and residues at the C-terminus of each SW interact with an open hydrophobic network in the B-domain of the opposing protomer (Figure 7A, B). Hydrophobic ‘knobs’ at the end of the SW (residues 350-354) stabilize the open N3A motif in the B-domain of the neighboring protomer. Side chains from Phe350, Phe351, and Leu354 fill the void created by an extensive hydrophobic ‘nest’ (Figure 7A, B). The eight residues that comprise the nest (Phe221, Leu224, Leu232, Trp288, Gln295, Phe320, Ile324, Leu327) are noncontiguous and recruited from the entire B-domain (Supplement Figure 1A). This knob-nest interface is further strengthened by Asn353 the only polar residue at the end of the SW, which forms a hydrogen-bond with the backbone carbonyl of Thr220 in the 310 loop of the B-domain (Figure 6A, B). Formation of this dimeric assembly protects 2,740 Å2 of surface area, and has a free energy (ΔG) of 15.8 kcal/mol required for dissociation, an indication that this arrangement is thermodynamically stable in solution. Native PAGE analysis of PKG78-355 in solution displays a small population of dimeric protein (Supplement Figure 2A). Similarly, the migration of crystalline PKG78-355 is consistent with a dimer (Supplement Figure 2B).
Our initial DxMS studies on PKG78-355 offered further validation of the significance of this knob-nest interface. Analysis of peptide fragments from the SW indicated that the C-terminal residues containing the hydrophobic knobs and Asn353 had a slower rate of deuterium exchange compared to residues in the more solvent-exposed region of this helix. This finding suggested that the very C-terminus of the SW is protected in solution and that the knob-nest interaction may serve as the focal point of interchain communication between PKG protomers. Interestingly, the knob residues appear to be unique to PKG I isoforms, as PKG II has a large amino acid insertion at the site of the SW (see sequence alignment, Supplement Figure 1A,). However, hydrophobic residues at the equivalent nest positions are conserved in both PKG I and II isoforms.
The functional relevance of the knob-nest interface was probed by Alanine scanning mutagenesis of the SW knob residues in full length PKG Iα containing both the N-terminal D/D and the catalytic domains. Extracts from HEK293 cells expressing wild type and mutant PKG Iα were examined for phosphoryl transfer activity. Removal of the specific hydrogen bond provided by Asn353 (N353A) resulted in a significantly decreased activation constant (Figure 7C, Table 3). The entire knob-nest interface was disrupted by a quadruple mutation wherein all hydrophobic knob residues in addition to the specific Asn353 were substituted for alanine (F350A, F351A, N353A, L354A). Neutralization of these interactions further reduced the activation constant greater than 4-fold (Figure 7C, Table 3). Additionally, this mutant displayed a loss of cooperativity (nH). While we were unable to calculate absolute basal and Vmax values, the relative changes in KA, combined with the observed decrease in Hill coefficient, illustrates the importance of the knob-nest interface in maintaining the kinetic fidelity of PKG Iα. In the context of full length PKG Iα, the SW seems to act as a tether for the catalytic domain, disruption of which causes the kinase to be more easily activated.
Direct communication between regulatory and catalytic elements is not without precedent in AGC kinases. In PKA, there are at least four major sites of contact between the R and C subunits, with CNB domain A providing the largest docking surface for the C subunit (Boettcher et al., 2011; Kim et al., 2007). Likewise, in PKC βII, the conserved NFD motif in the catalytic domain is clamped by the diacylglycerol-binding C1B domain until fully activated (Leonard et al., 2011). While these interactions highlight the diversity of interdomain complementarity among AGC kinase family members, they also demonstrate clear commonalities in their mechanisms of regulation.
This crystal structure provides the first atomic view of PKG and provides a new platform for understanding the allosteric regulation of the holoenzyme complex. The overall fold of PKG78-355 is surprising, in that a previously uncharacterized allosteric interface promotes a novel means of communication between PKG78-355 protomers wherein the catalytic domain can be tethered between inactive and active states. The crossing of switch helices between protomers (Figure 6B) suggests that the catalytic domain from one protomer is regulated in part by the regulatory domain of the other protomer. A complete understanding of PKG structural dynamics will undoubtedly require additional studies, however this new structure presents crucial details into the long sought after mechanism of allosteric PKG activation.
DNA for bovine PKG Iα encoding amino acids Q78-K355 was amplified by PCR using the primers 5′-CGGGATCCATGCAGGCATTCCGGAAGTTC-3′ (sense) and 5′-GGAATTCCTACTACTTCAGGTTGGCGAAG-3′ (antisense). The PCR product was digested with BamHI and EcoRI and ligated into pRSET-A (Invitrogen) in frame with the N-terminal 6×-His tag. The plasmid was transformed into E. coli BL21 and 5 ml starter cultures in LB were allowed to grow to OD600=0.6 at 37°C, 300RPM in the presence of 50 μg/mL Ampicillin. 1L of Overnight Express (Novagen) media was inoculated 1:1000 in 4L-baffled flasks with the mid-log starter prep and grown at 25°C, 300 RPM for 16-24 hours. Bacteria were harvested by centrifugation at 5K RPM for 10 minutes at 4°C, and pellets were stored at -80°C. Pellets were resuspended in 5ml/gram of 50mM MES pH=6.8, 100mM NaCl, 5mM MgCl2, 3mM TCEP, 5% glycerol (Buffer A) plus protease inhibitors. Cells were lysed at 1200 psi using a french pressure cell. Homogenates were clarified via centrifugation at 15K RPM for 90 minutes at 4°C and recombinant PKG Iα78-355 was purified from the supernatants on a Profinia protein purification system (Biorad) using the native IMAC protocol. Eluted protein was subjected to 3 rounds of dialysis in 2L of Buffer A at 4°C. A secondary purification step was performed on an AKTAprime FPLC system by passing Ni++-purified protein over a HiLoad 16/60 Superdex 75 (GE Healthcare) gel filtration column and collecting 1 mL fractions in Buffer A. Protein homogeneity was assessed via SDS-PAGE and Coomassie staining. Desired fractions were pooled and concentrated using 10K MWCO centrifugal concentrators (Sartorius). Typical yields were 100 mg purified protein per liter of media.
Sparse matrix kits from Hampton Research were used to screen initial crystal growth conditions. Crystals used for data collection were grown via hanging drop vapor diffusion in 2.2M (NH4)2SO4, 100mM Tris pH=8.0, 0.2% MPD at a protein concentration of 25-35 mg/mL at 20°C. Crystals were cryoprotected for 10-30 minutes in 2M Li2SO4, 100mM Tris pH=8.0, 0.2% MPD and flash frozen in liquid N2. 2.5 Å diffraction data was collected at beamline 8.2.1 at the Advanced Light Source, Lawrence Berkeley National Laboratory. HKL2000 was used for data processing and scaling (Minor, 1997). Phases were generated with the program Phaser (McCoy et al., 2007) using the PKA regulatory domain (PDB ID 1RGS) as a search model. Crystals grew in the C2 space group with two molecules per asymmetric unit. The PKG Iα78-355 model was manually built into electron density using the programs TURBO-FRODO (A. Roussel, 1991) and Coot (Emsley et al.). CNS (Brunger et al., 1998) was used for structure refinement. Publication-quality images were generated using Pymol (Schrodinger). Interface and assembly measurements of symmetry mates were calculated using the protein interfaces, surfaces and assemblies service PISA at European Bioinformatics Institute (Krissinel and Henrick, 2007).
All deuterium exchange reactions were performed on ice, in a 4°C cold room. Exchange reactions were initiated by adding 6 μl buffered D2O to 2 μl purified PKG Iα78-355 at 7 mg/ml. At the appropriate time points exchange was quenched by adding 12 μl 1.6M GuHCl, 0.8% formic acid. In addition, non-deuterated samples were prepared by incubating the protein in buffered H2O. Fully deuterated samples were prepared by incubating the protein in D20 buffer containing 1% formic acid overnight at room temperature. The samples were frozen on dry ice and stored at -80°C until analysis by mass spectrometry. Samples were manually thawed on wet ice and immediately analyzed by LC-MS. Procedures for pepsin digestion for DXMS have been described previously (Burns-Hamuro et al., 2005; Hamuro et al., 2004; Pantazatos et al., 2004; Spraggon et al., 2004). Briefly, the samples were passed through an immobilized pepsin column and the protease-generated peptides were collected on a C18 HPLC column. The peptides were eluted from the C18 column and the effluent was directed to a Thermo Finnigan LCQ electrospray ion trap type mass spectrometer with data acquisition in either MS1 profile mode or data-dependent MS2 mode. The pepsingenerated peptides from the MS/MS data sets were identified using SEQUEST (Thermo Finnigan Inc.), followed by analysis using customized DXMS data reduction software (Sierra Analytics Inc., Modesto, CA). Corrections for back exchange were made through measurement of loss of deuterium from fully deuterated samples. Deuterium incorporation for each peptide was calculated using the methods of Zhang and Smith (Zhang and Smith, 1993):
Where m(P), m(N), and m(F) are the centroid value of the partially deuterated, nondeuterated, and fully deuterated peptide, respectively. The experiments were performed twice, and the reported results are the average of the two experiments.
To confirm the copurification of cAMP, purified PKG78-355 was incubated on ice in a 1:1 ratio of acetonitrile. The extract was clarified via centrifugation at 4°C and the supernatant was subjected to UV-spectroscopy. The extract had a single peak with a λmax of 258 nm and lacked the characteristic cGMP shoulder. The same extract was subjected to HPLC analysis using a Merck Hitachi HPLC system comprising a L-2130 pump, a L-2400 UV detector, a L-2350 column oven and a L-2200 autosampler, and data was processed with EZChrom Elite evaluation software (3.2.1). Isocratic runs were performed at 30° C on a RP-18 reversed phase silica column (250 × 4 mm, ODS A,YMC) with 2.5% isopropanol and 25 mM triethylammonium formate buffer, (pH 6.9) with a flow of 1 ml/min at 255 nm.
WT PKG Iα was cloned into pcDNA 3.1 using BamHI and EcoRI. Mutations were made using QuickChange site-directed mutagenesis (Agilent Technologies, Santa Clara, CA) per manufacturer's recommendations. HEK293 cells grown in 10cm dishes were transfected for 5 hours using Metafectene (Biontex, San Diego, CA) at a ratio of 20 μg DNA per 60 μL lipid. 60 hours after transfection cells were harvested by scraping into PBS, 2mM benzamidine-HCl, 200 μM EDTA. Protein concentration was determined by the Bradford method and stocks were stored in 50% glycerol at -20°C.
Determination of activation constants was performed using a [γ-32P]-ATP transfer assay as previously reported (Ruth et al., 1991; Tegge et al., 1995). Briefly, 0.5 μg of wild-type or mutant protein from PKG-transfected HEK293 cells was incubated in buffer with various concentration of cGMP in the presence of [γ-32P]-ATP and the PKG-specific peptide substrate TQAKRKKSLAMA (Dostmann et al., 1999). Aliquots, spotted on P81 Whatman paper were subjected to scintillation counting. All experiments were performed in the presence of the PKA-specific inhibitor, PKI5-24 (70 nM) to suppress endogenous PKA activity. Mock-transfected cells did not show any activity.
This presents the first crystal structure of tandem cGMP binding domains.
A novel subdomain promotes unexpected protomer-protomer communication in PKG 78-355.
The two cGMP-binding domains remain extended despite occupation of the A-domain.
A C117-C195 disulfide bond uncouples communication between the two CNB domains.
We thank Nico Villanueva, Karl Zahn and Brian Eckenroth of the UVM Center for X-ray Crystallography for their expertise. HPLC analysis was graciously performed by Hans-G. Genieser and Frank Schwede of BIOLOG Life Science Institute, Bremen, Germany. This work was supported by NIH grants HL68891 (BWO, CJM, CKN, WRD), GM34921 (JW, APK, SST), CA099835, CA118595, AI076961, AI081982, AI2008031, GM020501, GM066170, NS070899 GM093325 and RR029388 (VLW), CA124517 (DEC), and by the Totman Trust for Medical Research (CKN, WRD).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.