Structures of cGMP-dependent Protein Kinase in Malaria Parasites (2019) PDF

Structures of the cGMP-dependent protein kinase in malaria parasites reveal a unique structural relay mechanism for activation Majida El Bakkouria,b,1, Imène Kouidmib, Amy K. Wernimonta, Mehrnaz Amania, Ashley Hutchinsona, Peter Loppnaua, Jeong Joo Kimc, Christian Flueckd, John R. Walkera,2, Alma Se...

Structures of the cGMP-dependent protein kinase in malaria parasites reveal a unique structural relay mechanism for activation Majida El Bakkouria,b,1, Imène Kouidmib, Amy K. Wernimonta, Mehrnaz Amania, Ashley Hutchinsona, Peter Loppnaua, Jeong Joo Kimc, Christian Flueckd, John R. Walkera,2, Alma Seitovaa, Guillermo Senisterraa, Yoshito Kakiharae,3, Choel Kimc,f, Michael J. Blackmand,g, Charles Calmettesb,4, David A. Bakerd,4, and Raymond Huia,h a Structural Genomics Consortium, University of Toronto, Toronto, ON M5G 1L7, Canada; bINRS-Institut Armand-Frappier, Laval, QC H7V 1B7, Canada; c Department of Pharmacology, Baylor College of Medicine, Houston, TX 77030; dFaculty of Infectious and Tropical Diseases, London School of Hygiene & Tropical Medicine, London WC1E 7HT, United Kingdom; eDepartment of Biochemistry, University of Toronto, Toronto, ON M5S 1A1, Canada; fVerna and Marss McLean Department of Biochemistry and Molecular Biology, Baylor School of Medicine, Houston, TX 77030; gMalaria Biochemistry Laboratory, The Francis Crick Institute, London NW1 1AT, United Kingdom; and hToronto General Hospital Research Institute, Toronto, ON M5G 2C4, Canada Edited by James A. Wells, University of California, San Francisco, CA, and approved May 28, 2019 (received for review April 10, 2019) The cyclic guanosine-3′,5′-monophosphate (cGMP)-dependent pro- mobilization (7, 8). A phosphoproteomic study has linked Pf PKG tein kinase (PKG) was identified >25 y ago; however, efforts to activity to 107 phospho-sites on 69 different proteins in the P. fal- obtain a structure of the entire PKG enzyme or catalytic domain ciparum proteome, including some implicated in invasion and egress from any species have failed. In malaria parasites, cooperative acti- (9). Regulation of cGMP has also been associated with calcium flux vation of PKG triggers crucial developmental transitions throughout and egress in the related apicomplexan pathogens Toxoplasma and the complex life cycle. We have determined the cGMP-free crystal- Eimeria (10). These lines of evidence support PKG as a promising lographic structures of PKG from Plasmodium falciparum and Plas- target for antiparasitic drug discovery, as well as a gateway to a modium vivax, revealing how key structural components, including an N-terminal autoinhibitory segment (AIS), four predicted cyclic deeper understanding of parasite signaling. nucleotide-binding domains (CNBs), and a kinase domain (KD), are The cellular functions of parasite PKG are regulated by allo- arranged when the enzyme is inactive. The four CNBs and the KD steric and cooperative binding of cGMP (11), similarly to how are in a pentagonal configuration, with the AIS docked in the sub- mammalian PKG is activated. Allostery and cooperativity are strate site of the KD in a swapped-domain dimeric arrangement. We also the hallmarks of another eponymous member of the AGC show that although the protein is predominantly a monomer (the dimer is unlikely to be representative of the physiological form), the Significance binding of the AIS is necessary to keep Plasmodium PKG inactive. A major feature is a helix serving the dual role of the N-terminal helix Despite being one of the first protein kinases discovered, cyclic of the KD as well as the capping helix of the neighboring CNB. A guanosine-3′,5′-monophosphate (cGMP)-dependent protein ki- network of connecting helices between neighboring CNBs contrib- nase (PKG) has not been successfully crystallized previously, utes to maintaining the kinase in its inactive conformation. We pro- leaving many unanswered questions about its mechanism of ac- pose a scheme in which cooperative binding of cGMP, beginning at tivation. We report herein the structure of cGMP-free PKG from the CNB closest to the KD, transmits conformational changes around Plasmodium parasites, the causative agents of malaria, one of the the pentagonal molecule in a structural relay mechanism, enabling world’s deadliest infectious diseases. The structures, combined PKG to orchestrate rapid, highly regulated developmental switches with data from biochemical and biophysical experiments, provide in response to dynamic modulation of cGMP levels in the parasite. insight into a mechanism of activation that involves previously unpredicted interdomain communication via a structural relay cyclic GMP | signal transduction | malaria | Plasmodium | structure system. In addition to the full structure of PKG, our work con- tributes important functional information for a key antimalarial drug target. M alaria remains a serious global health problem, with close to 500,000 deaths and hundreds of millions of new infec- tions annually. Reports of prolonged parasite clearance times Author contributions: M.E.B., C.C., D.A.B., and R.H. designed research; M.E.B., A.K.W., M.A., A.H., P.L., C.F., A.S., Y.K., and C.C. performed research; R.H. contributed new re- and treatment failures using artemisinin combination therapies agents/analytic tools; M.E.B., I.K., A.K.W., J.J.K., J.R.W., G.S., Y.K., C.K., C.C., D.A.B., and (ACTs) are increasingly frequent in parts of Southeast Asia (1). R.H. analyzed data; and M.E.B., M.J.B., C.C., D.A.B., and R.H. wrote the paper. New targets to supply the next generation of antimalarial drugs The authors declare no conflict of interest. are being studied with urgency to tackle this growing trend, This article is a PNAS Direct Submission. particularly in anticipation of the spread of ACT resistance to This open access article is distributed under Creative Commons Attribution License 4.0 Africa. Among promising drug targets are protein kinases enco- (CC BY). ded by the genomes of the Plasmodium parasites responsible for Data deposition: The crystallography coordinates have been deposited in the National the disease (2). Previous work has demonstrated that one par- Center for Biotechnology Information’s Protein database, https://www.ncbi.nlm.nih.gov/ protein [PDB ID codes 5DYL (PvPKG), 5DYK (PfPKG), and 5DZC (PvPKG-AMPPNP)]. ticular kinase, known as cyclic guanosine-3′,5′-monophosphate 1 Present address: Paraza Pharma Inc., 2525 Marie-Curie, Montréal, QC H4S 2E1, Canada. (cGMP)-dependent protein kinase, or protein kinase G (PKG), has 2 Present address: Department of Biology, McMaster University, Hamilton, ON L8S 4L8, essential roles in multiple stages of the parasite life cycle. Selective Canada. pharmacological inhibition of Plasmodium falciparum PKG 3 Present address: Division of Dental Pharmacology, Niigata University Graduate School of (Pf PKG) blocks the egress of merozoites (3) and gametes (4) Medical and Dental Sciences, Niigata, Japan. from erythrocytes, as well as inhibits the motility of ookinetes in 4 To whom correspondence may be addressed. Email: [email protected] or the mosquito (4, 5) and invasion of hepatocytes by sporozoites [email protected]. (6). PKG orchestrates the progression of these key differentiation This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. events in Plasmodium via a complex system of second messenger 1073/pnas.1905558116/-/DCSupplemental. signaling, involving phosphoinositide metabolism and calcium Published online June 25, 2019. 14164–14173 | PNAS | July 9, 2019 | vol. 116 | no. 28 www.pnas.org/cgi/doi/10.1073/pnas.1905558116 subfamily of protein kinases, adenosine 3′,5′-cyclic mono- Plasmodium PKG Adopts a Pentagon-Shaped Architecture. We phosphate (cAMP)-dependent protein kinase (PKA). The reg- expressed and purified full-length recombinant Pf PKG and P. ulatory and catalytic subunits of PKA are distinct, separately vivax PKG (PvPKG). They crystallized as dimers in orthorhom- encoded proteins that form tetrameric complexes composed of bic and monoclinic space groups respectively, and both yielded two regulatory subunits and two catalytic subunits (R2C2) (12, 13). 2.4-Å apo structures [Fig. 2A and SI Appendix, Fig. S1 and Table In contrast, the regulatory and catalytic domains of PKG S1; Protein Data Bank (PDB) ID codes: 5DYK and 5DYL]. The coexist within a single polypeptide encoded by a single gene. In two orthologs are 92% identical in sequence; furthermore, mammals, this assembly contains two cyclic nucleotide-binding alignment of their structures resulted in rmsds of 0.8 Å— (CNB) sites, compared with the four CNB domains in apicom- representing negligible differences (SI Appendix, Fig. S1). PvPKG plexan PKG isoforms (14). Even though PKA and PKG were also cocrystallized with adenylyl-imidodiphosphate (AMPPNP), both among the earliest kinases identified, understanding of how a nonhydrolyzable analog of adenosine triphosphate (ATP), the structural divergence of PKG from PKA affects cGMP- again as a dimer (PDB ID code 5DZC; 2.3 Å; SI Appendix, Table mediated allostery and cooperativity (15) in different organ- S1). In the ensuing description, unless stated otherwise, impor- isms is at best fragmentary. tant details are common to all of the structures. To ensure Structural biology is an essential tool in the study of protein clarity, important differences in residues are mentioned in the kinases. PKA, the first to be crystallized (16), is the archetypal text and highlighted in an alignment of the sequences of the two protein kinase in structural and mechanistic studies. There are orthologs (SI Appendix, Fig. S2). now a number of PKA structures revealing the standalone catalytic In the Plasmodium PKG crystal structures, each protomer in the and regulatory subunits, as well as the engaged R1C1 dimers and dimers can be described as four CNBs, which, together with the N R2C2 tetramers (12, 17, 18). In contrast, for both mammalian and lobe of the KD, form a pentagonal arrangement (Fig. 1 B and C), parasite PKG, only the structures of isolated regulatory domains with the KD C lobe locked in the center. CNB-A and -B make have been reported (19–26). A report disclosing the structure of direct contact with and constrain (and are constrained by) the C one of the cGMP-binding domains of Pf PKG (CNB-D) identified lobe of the KD from functionally essential movement (Fig. 1 B a triad of residues in its C-terminal helix that are essential for and C), resulting in a substructure of these three domains that MEDICAL SCIENCES regulation of enzyme activity (23). How this domain interacts with closely resembles the R-C heterodimer of PKA (Fig. 1D) (12). the remainder of the PKG structure and, more generally, how Interestingly, all three Plasmodium structures display a swapped Plasmodium PKG mediates cGMP signaling are unknown. domain arrangement, in which the N terminus of each protomer in Here, we describe the complete atomic structure of PKG from the dimer interacts with the active site of the other (Fig. 2A). two human malaria parasite species, P. falciparum and Plasmo- The N-Terminal Segment of Plasmodium PKG Has an Autoinhibitory dium vivax. To date this has not been achieved for PKG from any other organism. The resulting structures provide insights into Role. To study the significance of occupation of the Plasmodium how the CNBs interact with the kinase domain (KD) to keep it in PKG active site by the N-terminal residues, we generated a trun- a largely autoinhibited state and how the parasite maintains its cated form of Pf PKG which starts at position S30, and so lacks the cGMP signaling system in the off state. Combining our structural swapped N-terminal motif (Fig. 2A). Kinase activity assays (Fig. findings with biophysical and biochemical data allows us to 2B) showed that the truncated recombinant enzyme (Pf PKG– propose a structural relay model of cooperative activation of ΔAIS) had equivalent activity to the full-length protein in the presence of cGMP; however, only the truncated form demon- parasite PKG that may also have important implications for strated activity in the absence of cGMP (Fig. 2B). This suggested regulation of mammalian PKG. an inhibitory role for the N-terminal segment of Plasmodium PKG Results that is likely similar to that performed by an equivalent motif in PKA (18) and human PKG (28). Accordingly, we validated that the Plasmodium PKG Is a Member of a Distinct Class of PKG Enzymes. trans-binding of the N-terminal motif does not contribute to in- Mammalian PKGs are classified into types Iα, Iβ, and II (also hibition of the kinase (Fig. 2D), suggesting an intrasteric mode of known as PKG-Iα, -Iβ, and -II), all featuring a single polypeptide action as demonstrated by preincubation of the catalytically active comprising a regulatory domain of two CNBs fused to the N- N-terminally truncated form of Pf PKG (Pf PKG–ΔAIS) with an terminal flank of a single catalytic KD. Known to form inactive inactive full-length Pf PKG–E589A enzyme (ATP-binding-site dimeric holoenzymes when free of cGMP, they have an effective mutant). Furthermore, the activity curve of Pf PKG–ΔAIS is sig- regulatory stoichiometry (i.e., a CNB-to-KD ratio) of 4:2, similar moidal, with a Hill coefficient of 1.9 (Fig. 2C), indicative of the to that in a PKA R2C2 tetramer. In contrast, PKG enzymes of same homotropic positive cooperativity reported for full-length Plasmodium and other apicomplexan parasites have an extended PKG in Plasmodium and other type IIIα PKGs. N-terminal domain that features four CNBs (Fig. 1A) upstream The dimeric crystal structure was unexpected and required in- of the KD and are monomeric (27), resulting in an effective vestigation of its relevance, which was performed by using multi- regulatory stoichiometry of 4:1. There is thus significant struc- angle laser light scattering (full-length Pf PKG and PvPKG; SI tural divergence between Plasmodium and mammalian PKGs, Appendix, Fig. S3), analytical ultracentrifugation (full-length PfPKG prompting us to formally define kinases with more than two in- and PvPKG; full-length Pf PKG with AMPPNP; and Pf PKG–ΔAIS; tegrated cGMP-binding sites as type III or PKG-III. SI Appendix, Fig. S4), and immunoblot analysis of epitope-tagged A search of the National Center for Biotechnology Information native Pf PKG from parasite lysates (SI Appendix, Fig. S5). In all (NCBI) Protein database (https://www.ncbi.nlm.nih.gov/protein) led cases, the dimeric fraction ranged from very little (e.g., the strongest us to identify multiple subclasses of type III PKGs. In all apicom- dissociation constant was found to be ∼32 μM in area-under-the- plexan PKGs examined, one of the four CNBs is similar in sequence curve experiments, indicative of a very small dimer fraction) to to canonical cGMP-binding domains but lacks one or more critical undetectable. Collectively, the activity assays and biophysical char- residues (23). We refer to such a domain as a pseudo-CNB (pCNB) acterizations indicate that (i) cGMP is required for full activation of and the corresponding PKG as type IIIα. There are organisms, Pf PKG; (ii) the interaction between the N terminus of the protein including green algae such as Ostreococcus, in which PKG contains and the active site is required for complete autoinhibition (but not four cGMP-binding CNBs—that we call type PKG-IIIβ. Finally, for cooperativity), prompting us to name this region the auto- type IIIγ PKGs, such as some found in Paramecium and Tetrahymena inhibitory segment (AIS) for parasite PKG; and (iii) Plasmodium (both of which have multiple paralogues of PKG), contain three PKG is predominantly monomeric [consistent with what has been predicted CNBs. Type III PKGs are observed only in protist reported for the closely related Eimeria PKG (29)]—i.e., the genomes. In contrast, animals possess only types I and II PKGs. The swapped-domain dimer is unlikely to be representative of the architectures of these subtypes, along with those for types I and II as physiological form of the inactive or active protein. To date, our well as PKA, are shown for comparison in Fig. 1A. attempts to crystallize the AIS-truncated form of Plasmodium PKG, El Bakkouri et al. PNAS | July 9, 2019 | vol. 116 | no. 28 | 14165 the monomeric state of the full-length (or truncated) protein, or its overall autoinhibited state, Pf PKG–KD surprisingly exhibits many cGMP-bound configuration have been unsuccessful. structural features of an active kinase, including: (i) interaction between the catalytic lysine K570 and E589 (on helix αCK; Fig. Deviations of the KD of Plasmodium PKG from the PKA Catalytic 3A); (ii) positioning of the helix αCK toward the ATP-binding site Subunit Have Functional Implications. Similar to PKA, the KD of (Fig. 3A); (iii) an open (active) conformation of the activation Pf PKG adopts a classical bilobal structure flanked by an N-terminal loop (Fig. 3A); and (iv) intact hydrophobic spines (Fig. 3B). We helix (αAK) and a C terminus (Fig. 3A). Despite being locked in an further note that, compared with fully closed KDs (e.g., PDB ID Fig. 1. Plasmodium PKG displays a pentagonal architecture. (A) PKA is made up of two proteins: The regulatory subunit (R) contains CNB-A and -B and the catalytic subunit (C) includes an N-terminal helix (pink), the classical bilobal KD (yellow and green), and a C terminus (brown). Mammalian PKG-I and -II feature a PKA-like catalytic KD concatenated with a tandem of CNBs. In type III PKG, there are three or more CNB or CNB-like domains. PfPKG, an exemplar of type IIIα, contains four integrated CNBs, one of which does not bind cGMP. In contrast, type IIIβ PKG has four functional cGMP-binding sites, while type IIIγ members have exactly three CNBs. An inhibitory substrate sequence flanks the N terminus of the first CNB in each case. In PKA and mammalian PKG, there is an additional region of dimerization and docking motifs. The absence of this in type III PKGs suggests that they are likely monomeric. (B) Surface rendering of PfPKG colored as in A. PfPKG adopts a pentagonal architecture composed of the four CNB domains and the N lobe KD in the outer rim, and the C lobe KD locked in the middle. The CNB- A and -B domains make contact with the C lobe of the KD, similar to what is seen in all PKA heterodimers. CNB-D also makes contact with the C terminus (brown), whereas pCNB-C does not make direct contact with any part of the KD. CNB-D shows a unique arrangement not seen in PKA, making contacts with both the N lobe and the C terminus. (C) Structure of PfPKG drawn in cartoon representation. The long helix connecting CNB-A and -B is similar to that seen in PKA. We hypothesize that, inside the parasite, the AIS would occupy the KD’s active site to form a fully autoinhibited monomer. (D) Overlay of PfPKG and PKA heterodimer (PDB ID code 2QCS; in gray) depicts structural homologies. CNB-A, -B, and the KD form a substructure that closely resembles the heterodimer of PKA. 14166 | www.pnas.org/cgi/doi/10.1073/pnas.1905558116 El Bakkouri et al.  MEDICAL SCIENCES Fig. 2. Biochemical and biophysical assays. (A) The representation of the asymmetric unit from the PfPKG crystal illustrates a swapped domain dimer ar- rangement in which the N terminus of each protomer interacts with the active site of the other. The pseudosubstrate (red and gray surfaces) from the dimeric partner is shown binding (and inhibiting) the substrate site between the N and C lobes. These interactions contribute to autoinhibition of the protein. (B) The AIS inhibits PKG activity in vitro. In the absence of cGMP, the full-length protein exhibits negligible activity, whereas the truncated mutant protein (AIS removed) was partially active. Both samples were equally active in the presence of cGMP. wt, wild type. (C) Both the full-length and AIS-truncated constructs demonstrate homotropic positive cooperativity (Hill coefficient is 1.7 for the full-length sample and 1.9 for the truncated sample) when activated by addition of cGMP. The AIS is not required for full kinase activity under cGMP activation. (D) The AIS functions intrasterically to suppress kinase activity. The kinase activity of the catalytically active PfPKG–ΔAIS truncation mutant (1 μM) is monitored in the absence of cGMP with incremental concentrations of full-length PfPKG–E589A dead enzymes (1–8 μM). The AIS of the dead enzyme does not occupy/inhibit the active site of PfPKG–ΔAIS. (E) Using the dimeric structure to hypothesize a model for the autoinhibitory domain of monomeric PfPKG. The KD of one PfPKG-a is shown in yellow and green surface, along with its adjacent CNB-A domain aligned with the CNB-A of all four types of PKA (PDB ID codes 3FHI, 2QVC, 4DIN, and 4X6Q; all drawn in gray). The AIS of PfPKG-b (dark blue) overlays with the inhibitory segment of all four PKA structures (RIα in purple, RIIα in cyan, RIβ in pink, and RIIβ in yellow). In a monomer, we hypothesize that PfPKG-a would extend its own AIS (colored in red) in the same position (instead of in the KD of PfPKG-b). In all cases, a short helical overpass hovers above the long helix connecting the first two CNBs follows this inhibitory motif. The overpass of PfPKG-b is shown here to indicate how PfPKG-a might be different structurally as a monomer. It is also interesting that all of the structures are closely aligned from helix X:N onward, including Plasmodium PKG. This, along with all experimental data, supports the relevance of the rest of our crystal structure. El Bakkouri et al. PNAS | July 9, 2019 | vol. 116 | no. 28 | 14167 code 1ATP), the KD in all our structures can be described as par- pendix, Table S2). Rationalizing the significance of this putative tially open, with the AMPPNP-bound structure less open than the phosphothreonine requires additional experiments. nucleotide-free versions. The C terminus is a hallmark of AGC kinases (32) and an es- All three of our Plasmodium PKG structures feature an activa- sential cis-regulatory component. It is partially disordered in our tion loop that is both in its open conformation and unphosphory- Pf PKG structure, similar to apo PKA structures (e.g., PDB ID code lated—an uncommon combination previously observed in rho 1CTP). It is made up of four segments (Fig. 3C)—helix αI, the C- kinases and rho-associated kinases (30) such as ROCK and MRCK. lobe tether (CLT), the active-site tether (AST), and the N-lobe This is in contrast to PKA (and other members of the AGC sub- tether (NLT). In the CLT, the first proline in the PxxP motif in family and the majority of available S/T kinases crystallized with the PKA (and PKC)—known to play a role in interlobe movement—is activation loop in the open position), where T197 is phosphorylated conserved in Pf PKG (Fig. 3 C and D). In the AST, Pf PKG uses a (16). In recombinant Pf PKG, we found that T695 (homologous to motif starting with tyrosine (Fig. 3 C and D) to interact with the T197 in PKA) would become phosphorylated in the presence of ATP-binding pocket, similar to the role of the FDDY motif in cGMP, Mg2+, and ATP. The same modification has also been PKA, as well as in mammalian PKG (32). In the NLT, the hy- identified in a phosphoproteomics study (31). On the other hand, drophobic motif (HF) features WxxxF in place of FxxF (Fig. 3C), a we found that the recombinant protein maintained the same minor deviation conserved in all types of PKG. In our structure, activity profile (i.e., inactive without cGMP and fully active with W849 is disordered, but F853 engages a hydrophobic pocket at the cGMP) when T695 was mutated to either Ala or Gln (SI Ap- top of the N lobe (Fig. 3D), similar to what has been reported in Fig. 3. KD of PfPKG. (A) KD of PfPKG isolated with the rest of the protein hidden. The activation loop is high- lighted in blue in part to show that it is in its active conformation, but does not contain a phosphorylated site. The residue T695 is unphosphorylated, which is unusual for an active AGC kinase structure. The in- teraction between K570 and E589, the position of the helix αAK, and the conformation of the activation loop are all characteristic of an active KD. (B) Intact hydro- phobic spines are characteristic of an active KD. The C spine, shown here in yellow, comprises V555, A568, L626, I670, L671, L672, C725, and F729. It is broken in two sections because this crystal structure is missing ATP or a similar ligand. The R spine, made up of T593, L604, Y662, F683, and D718, is in red. (C) Similar to human PKA, the C terminus of PfPKG comprises helix αI, the CLT, the AST, and the NLT. (D) Two prolines (P813 and P814) anchor PfPKG’s CLT in a similar fashion as the PxxP (P314, F315, and P317) motif does in PKA. The FDDY motif in PKA is replaced by a single tyrosine (Y822) in PfPKG. The NLT in PfPKG comprises the motif WDIDF. In the structure, W849 is disordered, but F853 inserts itself in a hydrophobic pocket at the top of the N lobe of the KD. (E) The C tail of the KD (residues in brown) makes water-mediated contacts with CNB-D (cyan). (F) The main structural deviation between PKA and PKG is the N-terminal helix, as seen in the super- position of the catalytic subunit of PKA from mouse (PDB ID code 1ATP; αAK in gray) and the KD of PfPKG (αAK in beige). (G) In PfPKG-KD, helix αAK interacts with both lobes of the KD as well as the C terminus, with a number of interactions maintaining the KD in an open conformation and the overall protein in its auto- inhibited state. These interactions include a π-bond between H524 on helix αAK and R809 on the C tail, with the latter secured by salt bridges with P599 in the N lobe and D673 and D675 in the C-lobe. 14168 | www.pnas.org/cgi/doi/10.1073/pnas.1905558116 El Bakkouri et al. PKA (32). In addition, the C tail makes interactions with the N- teractions mutually constrain the lower part of the KD and the first terminal helix αAK and water-mediated contacts with CNB-D cGMP-binding domains similarly to what has been observed in the (Fig. 3E), the former of which is described below. PKA heterodimer (12, 18, 37) [e.g., the R194C–D241R–D267R triad The N-terminal helix, αAK in PKG, is the only major structural in the PKAC/PKAR–RIa heterodimer (18) is replaced by element to deviate in position noticeably from its PKA coun- R692:F164:N190 in Pf PKG]. To explore the significance of these terpart (Fig. 3F). As shown in Fig. 3G, αAK in Pf PKG makes interactions, we generated mutants of Pf PKG and investigated their multiple contacts with the N and the C lobes, some of which activity (SI Appendix, Table S2). The resulting recombinant proteins involve R809 as well as the C terminus of Pf PKG–KD, and none of mutants with K157, F164, N190, or Y694 modified to alanine were of which have been observed in PKA. The contact between the either unstable or inactive. On the other hand, mutants with two termini features a salt bridge between R528 and D597 as H128 and R692 similarly substituted remained stable and active. well as a π-bond between H524 and R809. Substitution with al- In PKA, site 4 is the region of interaction between domain B anine of H524, R528, or R809 (Pf PKG–H524A, –R528A, (specifically helix αB) and the KD (specifically helices αH and and –R809A; SI Appendix, Table S2) resulted in polypeptides αI). This site has been fully described in holoenzymes comprising that were either unstable (i.e., prone to precipitate) or inactive, PKAR–RIa (12) and –RIIa (37). Intriguingly, there is negligible establishing the significance of these interactions. contact in this area observed in the Plasmodium PKG structures. This is likely due to the fact that this contact, required in PKA to Structures of the CNB Domains of PfPKG Explain Their Distinct cGMP- secure the position of CNB-B, is obviated by the pentagonal Binding Properties. The Plasmodium PKG structures contain four single polypeptide arrangement in Plasmodium PKG. distinct CNBs. All four share the reported canonical fold of CNB In addition to PKA-like interactions, there are parasite-specific domains (33, 34), including the N3A bundle (helices αN and αA, interdomain contacts observed in the Plasmodium PKG structures, with a 310 loop in between) in the N terminus, an eight-stranded which are represented in the right half of the pentagon (consisting β-barrel, followed by a C-terminal hinge made up of helices αB of pCNB-C, CNB-D, and the KD). Notably, αAK is not only the and αC (Fig. 4A). In the middle of the β-barrel of CNB-A, -B, and -D N-terminal helix of the KD, but also the C-terminal αC helix of is a 24-residue-long phosphate-binding cassette (PBC) featur- CNB-D. In this conjoined arrangement, αAK engages CNB-D, MEDICAL SCIENCES ing the universally conserved glutamate and arginine residues, as particularly its PBC in multiple contacts. This includes a π-bond well as a short flexible helix, sometimes referred to as the B′-helix between R528 and Y417 that is supported by a salt bridge between (35). Their interactions with cGMP were determined by iso- Q532 and R528 (SI Appendix, Fig. S7C). In addition, a salt bridge thermal titration calorimetry (ITC), revealing 14, 17, and 0.17 μM between D533 and R484 (of the PBC of CNB-D) also contributes binding affinities for the CNB-A, -B, and -D when expressed as to the interaction between αAK and the PBC. Not surprisingly, a standalone recombinant protein samples, respectively (Fig. 4B). mutant form of PfPKG with R484 replaced by alanine was enzy- The third CNB, pCNB-C, stands out because a hydrophobic matically inactive (SI Appendix, Table S2). These interactions, network consisting of Y363, H373, F371, and F359 occludes the along with other contacts between CNB-D and the C terminus, cGMP-binding pocket. Accordingly, we were not able to detect constitute site 5, which is not found in PKA. any binding activity toward cGMP due to a degenerate binding Interestingly, in Plasmodium PKG, the junction of CNB-A and -B site (Fig. 4B). Furthermore, D361 and P370 take the places of conserves the long connecting helix observed in PKA (Fig. 1 C and D). glutamine and arginine, respectively, which are universally con- However, between CNB-B and pCNB-C, as well as between pCNB-C served in determined cAMP- and cGMP-binding CNBs (35). and CNB-D (Fig. 1C), this helix is twisted, which results in two helices: Our crystal structures and binding affinities strongly support the C-terminal helix (αCB) of the first CNB and the starting helix of previous suggestions that only three of the four CNB domains in the second CNB (αN). In this arrangement, the helix αN makes Plasmodium PKG bind cyclic nucleotides (36). contact with the preceding CNB, including interactions with the PBC. CNB-D stands out from the other regulatory domains in a These connecting helices not only constrain neighboring domains, but different way; its C-terminal helices (αBD and αCD) are closer to are likely also the mediator of interdomain communication, propa- their cGMP-capping positions than in the other CNB domains, gating movement in one domain to the other. To further investigate as confirmed by comparison with existing structures of cGMP- the cGMP allosteric regulation transmitted along connecting helices, and cAMP-binding CNBs (SI Appendix, Fig. S6). This domain we attempted solving the cocrystal structures of other isolated CNB also has the highest affinity for cGMP by two orders of magni- domains bound to cGMP, including CNB-A and -B. The PfCNB-A tude, compared with CNB-A and -B (Fig. 4B), such that this domain bound to cGMP yielded high-quality crystals diffracting to regulatory unit is likely the first domain in Pf PKG to become 1.65-Å resolution (PDB ID code 5E16) and illustrates conformational occupied as the cellular concentration of cGMP rises. This is changes imposed by the repositioning of the helix αBA to lock the consistent with a model proposed by Kim et al. (23). cGMP in place within the binding pocket (Fig. 4E). This conforma- tional change is accompanied by a larger displacement of the con- Interdomain Contacts Provide Insight into Regulation of Plasmodium necting helix αCA (42° rotation) that is anticipated to redraw the PKG by cGMP. To study the interaction between different domains neighboring interaction between CNB-A and KD in the full-length in the Plasmodium PKG pentagon, we divided the structure into protein (Fig. 4F). Especially, we denote the probable rearrangement two overlapping halves. The left half consists of CNB-A, -B, and of key interactions upon cGMP activation, particularly the K157–Y694 the KD. As shown in Fig. 1D, this trio of domains is arranged in a π-bond interaction (site 2) between the connecting helix αCA and the way that is highly similar to the PKAR:PKAC heterodimer. activation loop in the immediate vicinity of the kinase active site (Fig. Closer examination reveals that some of PKA’s interdomain 4 E and F and SI Appendix, Table S2). This clearly suggests a mech- contacts, previously cataloged into four sites (12, 18), are also anism by which conformational modifications induced by binding of conserved in Plasmodium PKG, as described immediately below. cGMP to one CNB could be relayed to adjacent domains. In site 1, the basic AIS in PfPKG is docked against the activation loop of the KD in a manner reported for PKA, with key lysines Discussion (K15 and K16) replacing homologous arginines (R94 and R95 in In mechanistic studies of protein kinases, obtaining their inactive PKAR–RIa) in interacting with the P+1 loop on the KD (SI Ap- and active structures are important landmarks representing two pendix, Fig. S7A). Sites 2 and 3 (SI Appendix, Fig. S7B) are the main key states in a regulatory or signaling system. Using a combination interface where CNB-A and -B meet the activation loop, the of structural biology with biochemical and biophysical assays, our P+1 loop, and the C lobe of the KD. Key contacts include π-bonds study has established that cGMP-free Plasmodium PKG is a mo- between K157 and Y694 as well as F164 and R692 (the latter sup- nomeric protein held in an autoinhibited state by four main fea- ported by a hydrogen bond between N190 and R692). Furthermore, tures: (i) intrasteric regulation effected by the cis-binding of the there is a hydrophobic cluster involving the substrate-binding loop AIS in the substrate site; (ii) immobilization of the two lobes by (L696), the C lobe (F747), and CNB-A (I127 and H128). These in- the CNBs using a number of interdomain contacts; (iii) interaction El Bakkouri et al. PNAS | July 9, 2019 | vol. 116 | no. 28 | 14169 Fig. 4. The CNBs of PfPKG. (A) Surface rendering of the four PfPKG CNB domains, shown with their cGMP-binding pocket in gray surfaces. A cGMP molecule is manually docked in their putative binding sites for illustration. The third CNB, pCNB-C, is similar overall to the other CNBs, except for the cGMP-binding pocket, which is occluded due to a network of hydrophobic residues. (B) ITC binding curves for the CNB-A, -AB, -C, and -D constructs to cGMP. Upper and Lower display the ITC titration curves and the binding isotherms, respectively. As the CNB-B domain did not yield soluble recombinant protein, its binding activity was tested within a CNB-AB tandem construct. The binding affinity for the CNB-A, -B, and -D domains is 1.4, 1.7, and 0.17 μM, respectively; the CNB-C domain does not bind to cGMP. (C) Alignment of the CNB-D in our apo PfPKG structure (in color) with the cGMP-bound costructure (PDB ID code 4OFG; in gray) shows that the helix αCD un- dergoes noticeable displacement when cGMP is engaged. The triad (R484, Q532, and D533) key to activation of the enzyme is also displaced to (R484′, Q532′, and D533′). (D) Orthogonal representation of C. (E) Alignment of the CNB-A in our apo PfPKG structure (in color) with the cGMP-bound costructure (PDB ID code 5E16; in gray) shows the rearrangement of helices αBA and αCA in response to cGMP binding. Upon cGMP activation, the helix αCA undergoes a large rotation of 42°, but does not contribute to cGMP capping as observed in PKA and PfCNB-D; instead, the cGMP ligand is stabilized by interactions with the backbone of helix αBA. (F) Alternative representation of E that includes the CNB-A neighboring domains of PfPKG (apo form). The cartoon illustration depicts rearrangements of the connecting helix αCA relative to the CNB-B and kinase activation loop, with the disruption of the important K154–Y694 π-bond interaction. 14170 | www.pnas.org/cgi/doi/10.1073/pnas.1905558116 El Bakkouri et al. between the N-terminal helix and the C terminus of the KD in a keeping all these domains and subdomains in a mutually con- mutually locking arrangement; and (iv) arrangement of connecting straining arrangement. When cGMP binds CNB-D, at least some helices between CNB-B and pCNB-C, pCNB-C and CNB-D, and, of these contacts are expected to be abrogated or rearranged, most significantly, the conjoined helix between CNB-D and the likely resulting in changes in αAK. This hypothesis is confirmed by KD. To the best of our knowledge, the last two features are not a study in which Kim et al. (23) identified the formation of a only distinct from PKA, but entirely unique among kinases to date. capping triad (R484, Q532, and D533) in CNB-D when cGMP enters its pocket; a comparison of our full-length structure and Intrasteric Regulation Is Mediated by an N-Terminal AIS. Auto- their cGMP-bound structure (PDB ID code 4OFG) showed that, inhibition in mammalian PKA and PKG involves interactions in order for Q532 and D533 to form this triad and for R484 to between the regulatory and catalytic domains (18, 38). The engage cGMP, helix αAK has to undergo noticeable displacement proposed conformational changes mediated by cGMP binding (Fig. 4 C and D). To overcome stability issues of PKG in the are thought to release PKG from its autoinhibition and allow it presence of cGMP, we produced isolated CNBs to capture cGMP- to become active. Although the autoinhibitory domains of PKA dependent conformational changes in other CNBs. As such, we and PKG do not have a universally conserved sequence, they all report the cocrystallized PfCNB-A structure bound to cGMP, share a consensus small residue (most commonly glycine, re- which illustrates the reorganization of helix αCA upon cGMP ac- placeable by alanine, serine, or valine; underlined in SI Appendix, tivation (Fig. 4E). The conformational alteration of this connect- Fig. S7A) in the so-called P0 position, where a phosphorylation ing helix is expected to make a significant contribution to kinase target—namely, serine or threonine—might be found in a activity, as it makes important interactions with the kinase acti- pseudosubstrate peptide (e.g., the pseudosubstrate bound in the vation loop (K157–Y694 π-bond stabilization; Fig. 4F and SI Ap- AIS site in one of the first PKA structures). Alignments of se- pendix, Table S2). The torsion of the connecting helix αCA is quences and structures indicate that this residue (Ala18 in reminiscent of cAMP-dependent activation by PKA, which uses a Pf PKG) is conserved in the AIS of apicomplexan PKGs. This similar mechanism to destabilize the extended αCA helix to un- adds weight to the notion that the Plasmodium AIS performs an couple CNB-A and -B before dissociation of the complex (12). autoinhibitory function similar to the inhibitory segment (IS) in Despite a lack of phosphorylation, the activation loop is in its MEDICAL SCIENCES PKA (18) and AIS in mammalian PKG, particularly in view of open (active) conformation. This is presumably necessary to allow the observable increase in basal activity (i.e., in the absence of docking of the AIS. In kinase structures (including PKA) where cGMP) of the truncated version of Pf PKG missing this motif. the activation loop is phosphorylated, T197 is extended by the Although the domain-swapped dimer in our crystal structures addition of phosphate such that it can engage the C-alpha helix, in suggests the intriguing possibility of an intermolecular regulatory part to keep the activation loop in the stable conformation. In our system (in which each protomer extends N-terminal residues to structures, the function of the phosphate is instead replaced by block the substrate-binding site of its dimeric partner), the to- interaction with CNB-A (SI Appendix, Fig. S7B). tality of our biochemical and biophysical data indicates cGMP- free (i.e., inactive) Plasmodium PKG to be predominantly a A Structural Relay Model for Activation of Plasmodium PKG. In monomer. Accordingly, we were not able to detect trans- Plasmodium parasites, stringent timing is required for cGMP- inhibition between a full-length and a truncated form of mediated events, including egress of merozoites (41) and gam- Pf PKG, as illustrated in Fig. 2D. This suggests that the dimeric etes (4) from erythrocytes. A key trigger in the regulation of this interaction observed in our crystal structures is likely a crystal- timing is cooperative binding of cGMP and activation of PKG— lographic artifact. Attempts to crystallize Plasmodium PKG in a a property confirmed by the sigmoidal shape of the activity monomeric form have been unsuccessful, allowing us only to curves of our samples and a Hill coefficient of 1.9 observed in speculate on how the AIS might mediate autoinhibition in mo- our experiments. Whereas cooperativity in PKA is conferred by nomeric PKG-III. We observe that in all available PKA holo- the additional constraints imposed in the tetrameric holoenzyme enzyme structures, as well as in our Plasmodium PKG structures, (13), our data indicate that, in Plasmodium PKG, and likely a short helix follows the inhibitory region and passes over the PKG-IIIα in general, it is mediated by a network of connecting long helix, connecting the first two CNBs. In comparing the two helices which enables the binding of cGMP at CNB-D to facili- PKG molecules in the dimer with the PKA structures (Fig. 2E), tate binding at CNB-B and -A and ultimately removes all in- we hypothesize that, as a monomer, this overpass would shift hibitory constraints from the KD. Specifically, we propose the over from position A to position B, thus enabling Plasmodium following mechanism of cooperative activation (a sequence that PKG (and possibly other PKG-III enzymes in general) to me- may also represent four potential stable or metastable states of diate intrasteric inhibition in the same manner as PKA. Pf PKG). (i) When cGMP and ATP concentrations are low, all three functional CNBs and the ATP-binding site are ligand-free. Interdomain Contacts Regulate the Function of Plasmodium PKG. The substrate-binding site of the KD is occupied by the AIS. This There are three types of interdomain interactions in Plasmodium is the totally autoinhibited state of Pf PKG. (ii) As the level of PKG: (i) contacts between the KD, CNB-A, and -B, which are cGMP rises, the domain with the highest affinity, CNB-D, mostly conserved in PKA; (ii) contacts between CNB-D and KD becomes occupied first, resulting in a series of conformational (including and particularly the participation of the two termini), changes that involves displacement of αAK. This releases the N which are previously unrecognized and provide insight into how lobe of the KD from some of its constraints. Pf PKG is only ba- cGMP activation may take place (see below); and (iii) arrangement sally active in this state because the AIS continues to inhibit the of connecting helices between all neighboring domains except for KD, and the C lobe remains constrained by the first two CNBs. CNB-A and -B, which suggests a mechanism in interdomain Pf PKG (with ATP) with cGMP-bound CNB-D may be a struc- communication in kinases. When some of these interactions were turally metastable state, one that is primed to engage cGMP at abrogated by cGMP binding at one or more CNBs or by muta- CNB-B. (iii) The movement initiated at CNB-D is propagated to genesis, we observed significant protein aggregation, indicative of pCNB-C via the connecting helix between the domains. In return, their essential role in maintaining the structural integrity and this propels movement of the helix shared by CNB-B and pCNB- consequently the off state of Pf PKG (SI Appendix, Table S2). C, allowing it to be recruited by CNB-B as the capping helix as In PKA, the N-terminal helix may be phosphorylated and cGMP moves in, thus initiating bending of the helical bridge be- myristoylated, with both posttranslational modifications affecting tween CNB-A and -B in a manner seen in PKA and enabling the helical structure of αAK and the protein’s localization to binding of cGMP to CNB-A. (iv) Binding of cGMP in CNB-A membranes (39). These sites are missing in Plasmodium PKG and disrupts interactions between this domain and the KD, including in mammalian type I PKG [type II PKG has been reported to be the hydrophobic stack. This can release the AIS from its auto- myristoylated (40)]. Instead, αAK in Plasmodium PKG is a contact inhibitory position and the C lobe from all constraints, and free interface for the two lobes of the KD, the C terminus and CNB-D, the KD, which is already in its active conformation. El Bakkouri et al. PNAS | July 9, 2019 | vol. 116 | no. 28 | 14171  Collectively, our structural-relay model of cooperative activa- 20.4% and 21.2%, respectively. The PvPKG–AMPPNP structure was de- tion proposes that cGMP-mediated activation of Plasmodium termined by refining the previously determined PvPKG apo structure against PKG involves a series of conformational changes, initiated by the data acquired from the isomorphous complex crystals. Refinement was binding to CNB-D, which are propagated around the pentagonal carried out by using Buster (50) and REFMAC (49) combined with iterative molecule in a relay-like manner, increasing the affinity for cGMP manual model building using the molecular graphics program Coot (47) to a at CNB-B and, ultimately, CNB-A. Binding at three sites leads to final R factor of 22.1%. The PfCNBD–cGMP structure was solved by using a fully active KD in a yet-to-be-determined structural arrange- Phaser for molecular replacement using the CNB-D domain coordinates from ment of the active full-length protein. PfPKG as a search model; the structure was refined by using REFMAC to a final R factor of 19.6%. The geometry of the final models was checked by Materials and Methods using MolProbity (51) for reasonable clash scores and no Ramachandran outliers. Crystallographic details and refinement statistics are summarized in Protein Expression and Purification. Synthetic DNA for PfPKG and PvPKG from SI Appendix, Table S2. All of the illustrations in this work were generated by Genscript (sequences are in SI Appendix) was amplified by PCR and subcloned using MacPymol (Schrödinger, LLC). The coordinates have been deposited in into the pFBOH-MHL vector—a baculovirus expression vector with an N-terminal the NCBI protein structure database [https://www.ncbi.nlm.nih.gov/protein; Hexa-Histag followed by a TEV cleavage site (https://www.thesgc.org/reagents/ vectors). Individual PfCNB domains (CNB-A, -AB, -C, and -D) were subcloned into PDB ID codes 5DYL (PvPKG), 5DYK (PfPKG), and 5DZC (PvPKG–AMPPNP)]. the pET15-MHL vector—a bacterial expression vector with an N-terminal Hexa- Histag followed by a TEV cleavage site. PfPKG mutants were cloned by using a Enzymatic Assays. Kinase activity was characterized by using an NADH/ATPase described method (42) and expressed and purified as described above. coupled assay (52). The reactions were performed at 25 °C in a 384-well plate For protein expression, recombinant viral DNA was generated by transformation by using the Synergy 4 plate reader (Biotek). The reaction mix typically of the resulting plasmid into DH10Bac Escherichia coli competent cells. This was in contained the enzyme at a concentration of 250 nM, 500 μM Peptide 7 turn transferred into Sf9 insect cells by using Cellfectin transfection reagent (Life (RRRAPSFYAK), 150 μM NADH, 300 μM phosphoenolpyruvate, 1 mM ATP, a Technologies, Inc.). The insect cells were grown in HyQ SFX insect serum-free me- lactate dehydrogenase/pyruvate kinase mix from Sigma (3 units/mL), 20 mM dium (Thermo Fisher Scientific, Inc.), infected with 15 mL of P3 viral stocks per 0.8 L Hepes (pH 7.5), 30 mM NaCl, 10 mM MgCl2, 1 mM CaCl2, 1 mM dithiothreitol, of suspension cell culture, and incubated at 27 °C by using a platform shaker at and 0.01% Tween 20. The reaction was started by adding cGMP (0–200 μM) 100 rpm. The cells were collected at 60- to 72-h postinfection, washed once with and monitored for 1 h by using the rate of NADH absorbance decrease at phosphate-buffered saline (PBS), and transferred to buffer A (25 mM Hepes, pH 7.5, 340 nm, which is proportional to the rate of ATP hydrolysis. 500 mM NaCl, 5% glycerol, and 5 mM imidazole). Individual PfCNB domains were produced in E. coli BL21 (DE3) grown in Terrific Broth medium supplemented with ITC. The binding constant and thermodynamic parameters of cGMP binding to ampicillin (100 μg/mL) at 37 °C and induced for overexpression at 18 °C with 0.5 mM the CNB domains of PfPKG were assessed by using a nano-isothermal titration isopropyl β-D-1-thiogalactopyranoside by using a LEX bubbling system. Cells calorimeter (TA Instruments). Experiments with PfCNB-A, -AB, -C, and -D were were incubated for 1 h at room temperature in the presence of 10% 3-[(3- performed at 25 °C. The sample cell was filled with 169 μL of purified protein cholamidopropyl)dimethylammonio]-1-propanesulfonate followed by 5 min samples prepared at a concentration of 60 μM in buffer A [25 mM Hepes, pH of sonication by using a dual-horn sonicator (pulse length set to 15 s on and 7.5, 300 mM NaCl, and 0.5 mM Tris(2-carboxyethyl)phosphine (TCEP)] and 20 s off). The total cell lysate was centrifuged at 16,000 × g for 1 h at 4 °C, stirred constantly at 350 rpm. The syringe was filled with 50 μL of cGMP at a and soluble fractions were collected. Protein samples were purified by using concentration of 0.7 mM in buffer A and titrated into the sample cell by using affinity chromatography (nickel beads; Sigma). The His-tagged PKG con- 3-μL injections at 180-s intervals. The net binding data were fitted by using the structs were eluted with buffer A containing 250 mM imidazole. The eluted NanoAnalyze Software (TA Instruments) to calculate the binding parameters. protein was then subjected to size-exclusion chromatography using an ÄKTAxpress system equipped with a Superdex 200 26/60 column on a system Analytical Ultracentrifugation. Sedimentation equilibrium analytical ultra- (GE Healthcare) preequilibrated in buffer A. centrifugation experiments were performed by using an Optima XL-A ana- lytical ultracentrifuge (Beckman). Protein samples at concentrations of 0.4, Crystallography. PfPKG, PvPKG, and PvPKG–AMPPNP were set up in sitting- 0.8, and 1.2 mg/mL were prepared in a buffer containing 25 mM Hepes drop vapor-diffusion experiments at room temperature (18 °C). All proteins (pH 7.5), 150 mM NaCl, and 1 mM TCEP. The samples were spun for 36 h at 6,000; yielded quality-diffraction crystals in a concentration range of 10–15 mg/mL. 8,000; and 10,000 rpm by using an An-60 Ti analytical rotor until equilibrium PfPKG was crystallized in the C2221 space group in 0.4 M L-proline, 10% was reached at 4 °C. Absorbance was monitored at 280 nm. The partial poly(ethylene glycol) (PEG) 3350, 0.1 M Hepes (pH 7.8), and 15% ethylene specific volume, solvent density, and solvent viscosity were estimated by glycol. PvPKG was crystallized in the C2 space group in 10% PEG 5000 mono- using the Sednterp program (University of New Hampshire; server located at methyl ether, 5% tascimate, 0.1 M Hepes, 15 mM spermidine (pH 7.0), and http://rasmb.org/sednterp/). Data analysis was done with the Origin MicroCal 25% glycerol. PvPKG–AMPPNP was crystallized in the C2 space group in 18.2% XL-A/CL-I Data Analysis Software Package (Version 4.0). PEG 3350, 0.1 M Hepes (pH 7.0), 0.1 M succinate, 2 mM MgCl2, and 5 mM AMPPNP. PfCNB-D was cocrystallized in the C2221 space group with 1 mM Multiangle Light Scattering. The molecular size of purified PfPKG was measured cGMP in 25% PEG 33350, 0.2 M NaCl, and 0.1 M Hepes (pH 7.5). at 25 °C by using a Viscotek Tetra detection system equipped with detectors for For PfPKG (apo form) and PvPKG–AMPPNP (cocrystal), data were collected static light scattering, UV, and refractive index (Malvern Instruments) connected at beamline 19ID of the Argonne National Laboratory’s Advanced Photon downstream of a size-exclusion chromatography system with a Superdex 200 HR Source (http://www.sbc.anl.gov/index.html) and processed by using HKL- 10/30 column mounted. The column was equilibrated in Equilibrium Buffer 3000 (43). For PvPKG (apo form), data were collected at beamline 23ID (25 mM Hepes, 150 mM NaCl, and 0.5 mM TCEP). Protein samples and the bovine and processed by using XDS (44). Finally, the PfCNBD–cGMP (cocrystal) data serum albumin standard were dialyzed overnight against Equilibration Buffer. were collected on a home-source X-ray diffractometer (Rigaku FR-E Super- Protein was diluted to a concentration of 6 mg/mL. A volume of 100 μL of both Bright rotating anode generator) and processed by using HKL3000 (43). The was sequentially injected by using an auto-sampler into the chromatography PvPKG structure was solved by using the molecular-replacement pipeline program BALBES (45). One KD in the asymmetric unit was found by using the system at a flow rate of 0.2 mL/min. Molar weight determination was performed program, with additional weak density in the lattice clearly showing addi- by using the Omnisec software (Malvern Instruments). tional beta sheets and alpha helices. A C-α backbone trace was made into the weak density. The models for the four CNBs were created by using P. falciparum Culture and Preparation of Soluble Protein Extracts. Transgenic P. FFAS03 (46). In an iterative process, the CNB models were manually driven falciparum clone 3D7/PfPKG-HA-3A (27) was grown in A+ erythrocytes (Na- into the backbone traces, and then real-space rigid-body refinement was tional Blood Transfusion Service) according to standard procedures (53) and used in the program Coot (47) to more accurately generate the coordinates. synchronized by multiple sorbitol treatments (54). Late schizont-stage parasites After each CNB was located, a round of REFMAC (48) was run, resulting in were released from red blood cells by using 0.15% saponin in PBS and washed improved electron-density maps that enabled placement of the next do- twice in PBS. The parasite pellet was resuspended in hypotonic lysis buffer (5 mM main. Loops and side-chain placements not matching the electron density Tris·HCl, pH 8.0) and frozen at −80 °C. All buffers were supplemented with were removed before refinement, and the overall model was built manually protease inhibitors (cOmplete protease inhibitor mixture; Roche). The lysate was by using Coot. The PfPKG structure was solved by using Phaser for molecular thawed and centrifuged at 16,000 × g for 15 min. The supernatant containing replacement and the PvPKG coordinates as a search model. Both PfPKG and the soluble protein fraction was analyzed by native polyacrylamide gel electro- PvPKG structures were refined by using REFMAC (49) to final R factors of phoresis (PAGE) and immunoblotting. 14172 | www.pnas.org/cgi/doi/10.1073/pnas.1905558116 El Bakkouri et al. Native Gel Electrophoresis and Immunoblotting. Protein samples were mixed Ontario Ministry of Economic Development and Innovation, Pfizer, Takeda, and with 2× native sample buffer (62.5 mM Tris·HCl, pH 6.8, 25% glycerol, and the Wellcome Trust. This work was also supported by Wellcome Trust Grants 1% Bromophenol blue ± DTT) or sodium dodecyl sulfate/PAGE sample 092809/Z/10/Z, 106240/Z/14/Z and 106239/Z/14/A (to D.A.B. and M.J.B.). This buffer + DTT and resolved on an 8% Tris·HCl (pH 8.8) polyacrylamide gel run work was also supported by the Francis Crick Institute (M.J.B.) , which receives in Tris·glycine buffer (pH 8.3) for 3 h at 150 V in an ice/water bath or on an its core funding from Cancer Research UK Grant FC001043; UK Medical Re- 8% Tris.HCl (pH 8.8) polyacrylamide gel run in Tris.glycine SDS buffer (pH search Council Grant FC001043; and Wellcome Trust Grant FC001043. C.C. is 8.3), respectively. Proteins were transferred to nitrocellulose and the PfPKG- the recipient of Fonds de Recherche Québec–Santé (FRQS) Research Scholar HA fusion protein visualized with anti-hemagglutinin antibody (clone 3F10, Junior 1 Career Award 251848, supported by Natural Sciences and Engineering Roche; diluted to 1:3,000) followed by horseradish peroxidase-conjugated Research Council of Canada (NSERC) Discovery Grant RGPIN-2017-06091; Fonds anti-rat (catalog no. SC-2006, Santa Cruz; at 1:6,000). The blot was reacted de Recherche Québec–Nature et Technologies (FRQNT) Grant 2019-NC-253753; with enhanced chemiluminescence plus substrate (Pierce) and exposed to X- as well as with instrumentation and infrastructure support provided by the ray film. Armand–Frappier Foundation. This research used resources of the Canadian Light Source at Beamline 08ID-1, which is supported by the Canada Foundation ACKNOWLEDGMENTS. The Structural Genomics Consortium is a registered for Innovation, Natural Sciences and Engineering Research Council of Canada, charity (no. 1097737) that receives funds from AbbVie, Boehringer Ingelheim, and Canadian Institutes of Health Research; and of the Advanced Photon the Canada Foundation for Innovation, the Canadian Institutes for Health Re- Source, a US Department of Energy (DOE) Office of Science User Facility oper- search, Genome Canada through the Ontario Genomics Institute (OGI-055), ated for the DOE Office of Science by Argonne National Laboratory under GlaxoSmithKline, Janssen, Lilly Canada, the Novartis Research Foundation, the Contract DE-AC02-06CH11357. 1. R. M. Fairhurst, Understanding artemisinin-resistant malaria: What a difference a year 27. C. S. Hopp, P. W. Bowyer, D. A. Baker, The role of cGMP signalling in regulating life makes. Curr. Opin. Infect. Dis. 28, 417–425 (2015). cycle progression of Plasmodium. Microbes Infect. 14, 831–837 (2012). 2. C. Doerig, O. Billker, D. Pratt, J. Endicott, Protein kinases as targets for antimalarial 28. V. Alverdi et al., cGMP-binding prepares PKG for substrate binding by disclosing the intervention: Kinomics, structure-based design, transmission-blockade, and targeting C-terminal domain. J. Mol. Biol. 375, 1380–1393 (2008). host cell enzymes. Biochim. Biophys. Acta 1754, 132–150 (2005). 29. S. P. Salowe, J. Wiltsie, P. A. Liberator, R. G. Donald, The role of a parasite-specific 3. H. M. Taylor et al., The malaria parasite cyclic GMP-dependent protein kinase plays a allosteric site in the distinctive activation behavior of Eimeria tenella cGMP- central role in blood-stage schizogony. Eukaryot. Cell 9, 37–45 (2010). dependent protein kinase. Biochemistry 41, 4385–4391 (2002). MEDICAL SCIENCES 4. L. McRobert et al., Gametogenesis in malaria parasites is mediated by the cGMP- 30. M. Jacobs et al., The structure of dimeric ROCK I reveals the mechanism for ligand dependent protein kinase. PLoS Biol. 6, e139 (2008). selectivity. J. Biol. Chem. 281, 260–268 (2006). 5. R. W. Moon et al., A cyclic GMP signalling module that regulates gliding motility in a 31. L. Solyakov et al., Global kinomic and phospho-proteomic analyses of the human malaria parasite. PLoS Pathog. 5, e1000599 (2009). malaria parasite Plasmodium falciparum. Nat. Commun. 2, 565 (2011). 6. K. Govindasamy et al., Invasion of hepatocytes by Plasmodium sporozoites requires 32. N. Kannan, N. Haste, S. S. Taylor, A. F. Neuwald, The hallmark of AGC kinase func- cGMP-dependent protein kinase and calcium dependent protein kinase 4. Mol. tional divergence is its C-terminal tail, a cis-acting regulatory module. Proc. Natl. Microbiol. 102, 349–363 (2016). Acad. Sci. U.S.A. 104, 1272–1277 (2007). 7. M. Brochet et al., Phosphoinositide metabolism links cGMP-dependent protein kinase 33. Y. Su et al., Regulatory subunit of protein kinase A: Structure of deletion mutant with G to essential Ca2+ signals at key decision points in the life cycle of malaria parasites. cAMP binding domains. Science 269, 807–813 (1995). PLoS Biol. 12, e1001806 (2014). 34. G. Y. Huang et al., Structural basis for cyclic-nucleotide selectivity and cGMP-selective 8. H. Fang et al., Epistasis studies reveal redundancy among calcium-dependent protein activation of PKG I. Structure 22, 116–124 (2014). kinases in motility and invasion of malaria parasites. Nat. Commun. 9, 4248 (2018). 35. A. P. Kornev, S. S. Taylor, L. F. Ten Eyck, A generalized allosteric mechanism for cis- 9. M. M. Alam et al., Phosphoproteomics reveals malaria parasite Protein Kinase G as a regulated cyclic nucleotide binding domains. PLOS Comput. Biol. 4, e1000056 (2008). signalling hub regulating egress and invasion. Nat. Commun. 6, 7285 (2015). 36. W. Deng, A. Parbhu-Patel, D. J. Meyer, D. A. Baker, The role of two novel regulatory 10. H. I. Wiersma et al., A role for coccidian cGMP-dependent protein kinase in motility sites in the activation of the cGMP-dependent protein kinase from Plasmodium fal- ciparum. Biochem. J. 374, 559–565 (2003). and invasion. Int. J. Parasitol. 34, 369–380 (2004). 37. J. Wu, S. H. Brown, S. von Daake, S. S. Taylor, PKA type IIalpha holoenzyme reveals a 11. C. A. Diaz et al., Characterization of Plasmodium falciparum cGMP-dependent protein kinase combinatorial strategy for isoform diversity. Science 318, 274–279 (2007). (PfPKG): Antiparasitic activity of a PKG inhibitor. Mol. Biochem. Parasitol. 146, 78–88 (2006). 38. S. H. Francis et al., Arginine 75 in the pseudosubstrate sequence of type Ibeta cGMP- 12. C. Kim, C. Y. Cheng, S. A. Saldanha, S. S. Taylor, PKA-I holoenzyme structure reveals a dependent protein kinase is critical for autoinhibition, although autophosphorylated mechanism for cAMP-dependent activation. Cell 130, 1032–1043 (2007). serine 63 is outside this sequence. J. Biol. Chem. 271, 20748–20755 (1996). 13. P. Zhang et al., Structure and allostery of the PKA RIIβ tetrameric holoenzyme. Science 39. C. Breitenlechner et al., The typically disordered N-terminus of PKA can fold as a helix 335, 712–716 (2012). and project the myristoylation site into solution. Biochemistry 43, 7743–7749 (2004). 14. D. A. Baker, W. Deng, Cyclic GMP-dependent protein kinases in protozoa. Front. 40. F. Hofmann, D. Bernhard, R. Lukowski, P. Weinmeister, cGMP regulated protein ki- Biosci. 10, 1229–1238 (2005). nases (cGK). Handb. Exp. Pharmacol. 191, 137–162 (2009). 15. R. W. McCune, G. N. Gill, Positive cooperativity in guanosine 3′:5′-monophosphate 41. C. R. Collins et al., Malaria parasite cGMP-dependent protein kinase regulates blood stage binding to guanosine 3′:5′-monophosphate-dependent protein kinase. J. Biol. Chem. merozoite secretory organelle discharge and egress. PLoS Pathog. 9, e1003344 (2013). 254, 5083–5091 (1979). 42. D. Neculai et al., Structure of LIMP-2 provides functional insights with implications for 16. D. R. Knighton et al., Crystal structure of the catalytic subunit of cyclic adenosine SR-BI and CD36. Nature 504, 172–176 (2013). monophosphate-dependent protein kinase. Science 253, 407–414 (1991). 43. W. Minor, M. Cymborowski, Z. Otwinowski, M. Chruszcz, HKL-3000: The integration 17. D. R. Knighton et al., Structural basis of the intrasteric regulation of myosin light of data reduction and structure solution—from diffraction images to an initial model chain kinases. Science 258, 130–135 (1992). in minutes. Acta Crystallogr. D Biol. Crystallogr. 62, 859–866 (2006). 18. C. Kim, N. H. Xuong, S. S. Taylor, Crystal structure of a complex between the catalytic 44. W. Kabsch, XDS. Acta Crystallogr. D Biol. Crystallogr. 66, 125–132 (2010). and regulatory (RIalpha) subunits of PKA. Science 307, 690–696 (2005). 45. F. Long, A. A. Vagin, P. Young, G. N. Murshudov, BALBES: A molecular-replacement 19. G. Y. Huang et al., Neutron diffraction reveals hydrogen bonds critical for cGMP- pipeline. Acta Crystallogr. D Biol. Crystallogr. 64, 125–132 (2008). selective activation: Insights for cGMP-dependent protein kinase agonist design. 46. L. Jaroszewski, L. Rychlewski, Z. Li, W. Li, A. Godzik, FFAS03: A server for profile– Biochemistry 53, 6725–6727 (2014). profile sequence alignments. Nucleic Acids Res. 33, W284–W288 (2005). 20. G. Y. Huang et al., Structural basis for cyclic-nucleotide selectivity and cGMP-selective 47. P. Emsley, K. Cowtan, Coot: Model-building tools for molecular graphics. Acta Crys- activation of PKG I. Structure 22, 116–124 (2014). tallogr. D Biol. Crystallogr. 60, 2126–2132 (2004). 21. J. J. Kim et al., Co-crystal structures of PKG Iβ (92-227) with cGMP and cAMP reveal the 48. M. D. Winn, G. N. Murshudov, M. Z. Papiz, Macromolecular TLS refinement in RE- molecular details of cyclic-nucleotide binding. PLoS One 6, e18413 (2011). FMAC at moderate resolutions. Methods Enzymol. 374, 300–321 (2003). 22. T. M. Moon, B. W. Osborne, W. R. Dostmann, The switch helix: A putative combina- 49. G. N. Murshudov, A. A. Vagin, E. J. Dodson, Refinement of macromolecular structures by torial relay for interprotomer communication in cGMP-dependent protein kinase. the maximum-likelihood method. Acta Crystallogr. D Biol. Crystallogr. 53, 240–255 (1997). Biochim. Biophys. Acta 1834, 1346–1351 (2013). 50. G. Bricogne et al., Buster (Version 2.10.0, Global Phasing Ltd., Cambridge, UK, 2011). 23. J. J. Kim et al., Crystal structures of the carboxyl cGMP binding domain of the Plas- 51. I. W. Davis et al., MolProbity: All-atom contacts and structure validation for proteins modium falciparum cGMP-dependent protein kinase reveal a novel capping triad and nucleic acids. Nucleic Acids Res. 35, W375–W383 (2007). crucial for merozoite egress. PLoS Pathog. 11, e1004639 (2015). 52. K. Kiianitsa, J. A. Solinger, W. D. Heyer, NADH-coupled microplate photometric assay 24. J. C. Campbell et al., Structural basis of cyclic nucleotide selectivity in cGMP- for kinetic studies of ATP-hydrolyzing enzymes with low and high specific activities. dependent protein kinase II. J. Biol. Chem. 291, 5623–5633 (2016). Anal. Biochem. 321, 266–271 (2003). 25. J. J. Kim et al., Crystal structure of PKG I:cGMP complex reveals a cGMP-mediated dimeric 53. W. Trager, J. B. Jensen, Human malaria parasites in continuous culture. Science 193, interface that facilitates cGMP-induced activation. Structure 24, 710–720 (2016). 673–675 (1976). 26. B. W. Osborne et al., Crystal structure of cGMP-dependent protein kinase reveals 54. C. Lambros, J. P. Vanderberg, Synchronization of Plasmodium falciparum erythrocytic novel site of interchain communication. Structure 19, 1317–1327 (2011). stages in culture. J. Parasitol. 65, 418–420 (1979). El Bakkouri et al. PNAS | July 9, 2019 | vol. 116 | no. 28 | 14173