Synthesis and Effect of Conformationally Locked Carbocyclic Guanine Nucleotides on Dynamin

Guanine nucleotides can flip between a North and South conformation in the ribose moiety. To test the enzymatic activity of GTPases bound to nucleotides in the two conformations, we generated methanocarba guanine nucleotides in the North or South envelope conformations, i.e., (N)-GTP and (S)-GTP, respectively. With dynamin as a model system, we examined the effects of (N)-GTP and (S)-GTP on dynamin-mediated membrane constriction, an activity essential for endocytosis. Dynamin membrane constriction and fission activity are dependent on GTP binding and hydrolysis, but the effect of the conformational state of the GTP nucleotide on dynamin activity is not known. After reconstituting dynamin-mediated lipid tubulation and membrane constriction in vitro, we observed via cryo-electron microscopy (cryo-EM) that (N)-GTP, but not (S)-GTP, enables the constriction of dynamin-decorated lipid tubules. These findings suggest that the activity of dynamin is dependent on the conformational state of the GTP nucleotide. However, a survey of nucleotide ribose conformations associated with dynamin structures in nature shows almost exclusively the (S)-conformation. The explanation for this mismatch of (N) vs. (S) required for GTP analogues in a dynamin-mediated process will be addressed in future studies.


Introduction
One means of interrogating the mechanism of nucleotide-dependent processes is to use conformationally constrained bicyclic rings incorporated in the nucleotide analogues in place of ribose. Ribose may assume a range of ring twists, and if the protein-preferring conformation can be pre-established in the ribose-like ring, the potency and selectivity can be increased because of lowered energy barriers to binding. We have extensively explored the use of methanocarba (bicyclo[3.1.0]hexane) rings in this context [1][2][3][4]. By positioning the cyclopropyl ring that is fused to the cyclopentyl ring, either a North (N) or South (S) envelope conformation can be maintained. GTP-binding proteins and GTPases have widespread roles in receptor signaling and other biochemical processes in the cell. Although they are potentially useful probes of the biochemical role of guanine nucleotides, the (N) and (S) methanocarba analogues of GTP ((N)-GTP and (S)-GTP, respectively) have not previously been reported. We also introduce non-hydrolyzable modifications of the triphosphate moiety in the methanocarba-GTP series ((S)-GMPPCP), as was previously reported for a β,γ-methylene ATP analogue [4].
The effect of the novel methanocarba analogues of GTP ((N)-GTP and (S)-GTP) on GTPase activity was examined by using dynamin as a model system. Dynamin is a large GT-Pase that plays a role in numerous vesiculation events in the cell by wrapping around and pinching off budding vesicles. These membrane fission events include clathrin-mediated endocytosis, synaptic vesicle recycling, and intracellular vesical trafficking [5]. More recently, dynamin has been shown to bundle actin filaments prior to myoblast fusion, during muscle development [6]. The well-known membrane fission activity of dynamin is regulated by GTP binding and hydrolysis. Dynamin tubulates vesicles, constricts opposing lipid bilayers upon GTP binding (constricted state), and further constricts upon GTP hydrolysis (superconstricted state) that reaches the theoretical limit required for spontaneous membrane fission [7][8][9]. As the hydrolysis of GTP by dynamin is critical for its function, we examined how the conformational state of the GTP nucleotide affects dynamin membrane constriction and fission activity. The dynamin-bound GTP ribose ring appears to be in the (S) conformation according to a recent report [7].
In this study, we synthesized conformationally locked analogs: (N)-GTP, (S)-GTP, and (S)-GMPPCP. The effects of (N)-GTP and (S)-GTP on dynamin activity were probed by in vitro reconstitution of dynamin membrane constriction activity and imaging by cryoelectron microscopy (cryo-EM). Whereas dynamin tubulates liposomes and constricts the lipid tubules upon addition of GTP, only (N)-GTP, but not (S)-GTP, constricted lipid tubules. This suggests that (S)-GTP binding does not constrict lipid tubules or is markedly stalled in activity. We also examined dynamin structures in the protein databank and observed that they are predominantly in the (S)-conformation (stalled or stable state). This finding that (S)-conformation is stable (or stalled in activity) has implications for the structural biology of dynamin and other dynamin superfamily proteins to trap stable states of dynamin for X-ray crystallography and cryo-EM. The results also provide an avenue of investigation for future drug design and enzymology of GTPases.

Synthesis of Conformationally Locked Carbocyclic Guanine Nucleotides
We prepared three methanocarba-GTP analogues, i.e., 4a (MRS4589, also referred to as (S)-GTP), 4b (MRS4591, also referred to as (S)-GMPPCP), and 4d (MRS4590, also referred to as (N)-GTP), for use in studying the conformational requirements of dynamin-GTPase (Scheme 1), but a fourth planned analogue, 4c, could not be synthesized. The 1 H-NMR, 31 P-NMR, and HRMS data are presented in the Supplementary Materials. The key intermediate 2 was quickly accessed in moderate yields by reacting compound 1 [2,10] with 2-amino-4,6-dichloropyrimidin-N 5 -formamide under microwave conditions [11]. The treatment of this important intermediate 2 with aqueous Trifluoroacetic acid hydrolyzed the purine moiety and removed both protecting groups on the pseudo-ribose ring to give (S)-methanocarba-guanosine derivative 3 ((S)-mc-G). The conformationally locked guanosine was phosphorylated by using phosphorus oxychloride and a sequential addition of pyrophosphate gave (S)-mc-GTP 4a. Unfortunately, the trimetaphosphate method, using salicylyl phosphorus (III) chloride, did not give 4a [12]. Compound 4a seemed to be unstable in neutral aqueous conditions or pure water/pure water, but it was stable in pH 7-8 triethylammonium acetate/ bicarbonate buffers. A similar approach to prepare 4b,c did not materialize, and the reactions to synthesize (S)-mc-GMP [7] resulted in low yields. Hence, an alternate route was employed to access compound 7. Sodium hydroxidemediated hydrolysis gave a 2,3-isopropylidine protected guanosine analogue 5, which was phosphorylated by using phosphoramidite chemistry to give 6. A subsequent acidic hydrolysis using DOWEX-H + resin gave the desired (S)-mc-GMP in excellent yields. Reacting the monophosphate 7 with tributylammonium salt of medronic acid in the presence of 1,2-dimethyl-N-tosylimidazolium triflate (TsDMIm-OTf) [13] gave β,γ-methylene-(S)mc-GTP 4b. It is perplexing that this method failed to give 4a,c but was able to give (N)-mc-GTP 4d [3]. Moreover, activation of the phosphate group of (S)-mc-GMP by using carbonyldiimidazole (CDI) and reacting with imidopyrophosphate did not give 4c.

Effect of Conformationally Locked Carbocyclic Guanine Nucleotides on Dynamin Activity
We applied cryo-EM to investigate the effects of conformationally locked carbocyclic guanine nucleotides on dynamin activity. Consistent with previous reports, we observed via cryo-EM that dynamin alone tubulates liposomes and assembles as a helical polymer on lipid tubules with an outer diameter of ~50 nm ( Figure 1A) [7][8][9]14,15]. Moreover, upon dynamin binding and hydrolysis of commercially available GTP, dynamin constricts the lipid tubule to an outer diameter of ~40 to 36 nm and results in fission ( Figure 1B). Interestingly, we observed different effects of (N)-GTP and (S)-GTP on dynamin constriction and fission activity. Whereas the activity of dynamin upon binding to (N)-GTP was similar to the effect of GTP binding, binding to (S)-GTP was similar to the dynamin activity in the absence of nucleotide ( Figure 1C,D).

Effect of Conformationally Locked Carbocyclic Guanine Nucleotides on Dynamin Activity
We applied cryo-EM to investigate the effects of conformationally locked carbocyclic guanine nucleotides on dynamin activity. Consistent with previous reports, we observed via cryo-EM that dynamin alone tubulates liposomes and assembles as a helical polymer on lipid tubules with an outer diameter of~50 nm ( Figure 1A) [7][8][9]14,15]. Moreover, upon dynamin binding and hydrolysis of commercially available GTP, dynamin constricts the lipid tubule to an outer diameter of~40 to 36 nm and results in fission ( Figure 1B). Interestingly, we observed different effects of (N)-GTP and (S)-GTP on dynamin constriction and fission activity. Whereas the activity of dynamin upon binding to (N)-GTP was similar to the effect of GTP binding, binding to (S)-GTP was similar to the dynamin activity in the absence of nucleotide ( Figure 1C,D).
In order to quantify the effects of conformationally locked carbocyclic guanine nucleotides on dynamin activity, we measured the diameter of dynamin-decorated lipid tubules in the presence or absence of various nucleotides. Briefly, we obtained cryo-EM micrographs of dynamin lipid tubules, similar to Figure 1, and boxed out segments of the tubes across many micrographs to obtain 5000 segments for each treatment: no nucleotide, GTP, (N)-GTP, and (S)-GTP. For each treatment, all the 5000 segments were overlaid in Fiji software ( Figure 2A). Moreover, for each treatment, the 5000 segments were classified by 2D classification in RELION into 20 classes (Supplementary Figure S1). Measurements of the outer diameters of the classes enabled us to compare the diameter of dynamin-decorated lipid tubules after various treatments ( Figure 2B). We observed that treatments of GTP and (N)-GTP constricted dynamin tubes, while (S)-GTP failed to constrict, similar to the condition in the absence of GTP ( Figure 2B). In order to quantify the effects of conformationally locked carbocyclic guanine nucleotides on dynamin activity, we measured the diameter of dynamin-decorated lipid tubules in the presence or absence of various nucleotides. Briefly, we obtained cryo-EM micrographs of dynamin lipid tubules, similar to Figure 1, and boxed out segments of the tubes across many micrographs to obtain 5000 segments for each treatment: no nucleotide, GTP, (N)-GTP, and (S)-GTP. For each treatment, all the 5000 segments were overlaid in Fiji software ( Figure 2A). Moreover, for each treatment, the 5000 segments were classified by 2D classification in RELION into 20 classes (Supplementary Figure S1). Measurements of the outer diameters of the classes enabled us to compare the diameter of dynamin-decorated lipid tubules after various treatments ( Figure 2B). We observed that treatments of GTP and (N)-GTP constricted dynamin tubes, while (S)-GTP failed to constrict, similar to the condition in the absence of GTP ( Figure 2B). We determined that (N)-GTP, but not (S)-GTP, enables dynamin-mediated membrane constriction. The North and South conformers of GTP ((N)-GTP and (S)-GTP) can be used with other GTPases in future studies, and it will be interesting to determine if this is a common phenomenon that (S)-GTP is not active. We hypothesize that (S)-GTP is inactive because the activity of GTPases requires transition through the North conformation.
Another explanation could be a defect of (S)-GTP in binding the GTPase, but this is likely not the case, as most structures of dynamin-related structures in the protein databank (Table 1) are bound to nucleotides in the South conformation ( Figure 3). In fact, an ensemble of structures of the dynamin superfamily has been subjected to structural investigation to check the puckering propensity of nucleoside/nucleotide-like compounds bound to these targets. We computed the puckering parameters and the distribution of the pseudorotation phase angle (P), and of the degree of deformation from the plane (ν max ) is reported in Figure 3 and Supplementary Figure S2. Most of the nucleotide analogues fall in the South hemisphere of the polar plot, with the exception of GDP bound to the human OPA1 that assumes a N-E conformation, and of GDP and ADP bound to a bacterial dynaminlike protein (Nostoc punctiforme) and EH domain-containing protein 4 (Mus musculus), respectively, which are flat. Given the propensity of atomic structures of GTPases bind to GTP and its derivatives with a ribose (S)-conformation, we also hypothesize that (S)-GTP stabilizes the GTPases to enable their structural determination. Going forward, the new analogues will provide a powerful tool to explore the effects of methanocarba analogues (N)-GTP and (S)-GTP on a wide range of GTPases and to determine if the (S)-GTP conformation is predominantly inactive. We determined that (N)-GTP, but not (S)-GTP, enables dynamin-mediated membrane constriction. The North and South conformers of GTP ((N)-GTP and (S)-GTP) can be used with other GTPases in future studies, and it will be interesting to determine if this is a common phenomenon that (S)-GTP is not active. We hypothesize that (S)-GTP is inactive because the activity of GTPases requires transition through the North conformation.
Another explanation could be a defect of (S)-GTP in binding the GTPase, but this is likely not the case, as most structures of dynamin-related structures in the protein databank (Table 1) are bound to nucleotides in the South conformation ( Figure 3). In fact, an ensemble of structures of the dynamin superfamily has been subjected to structural investigation to check the puckering propensity of nucleoside/nucleotide-like compounds bound to these targets. We computed the puckering parameters and the distribution of the pseudorotation phase angle (P), and of the degree of deformation from the plane (νmax) is reported in Figure 3 and Supplementary Figure S2. Most of the nucleotide analogues fall in the South hemisphere of the polar plot, with the exception of GDP bound to the human OPA1 that assumes a N-E conformation, and of GDP and ADP bound to a bacterial dynamin-like protein (Nostoc punctiforme) and EH domain-containing protein 4 (Mus musculus), respectively, which are flat. Given the propensity of atomic structures of GTPases bind to GTP and its derivatives with a ribose (S)-conformation, we also hypothesize that (S)-GTP stabilizes the GTPases to enable their structural determination. Going forward, the new analogues will provide a powerful tool to explore the effects of methanocarba analogues (N)-GTP and (S)-GTP on a wide range of GTPases and to determine if the (S)-GTP conformation is predominantly inactive. In conclusion, we synthesized the analogues of GTP containing a rigid ribose substitution, i.e., the methanocarba ring in both (N) and (S) fixed conformations. These correspond to two isomeric placements of a bicyclo[3.1.0]hexane ring system in place of the furanose ring. We used these analogues to interrogate the conformational requirements of a GTP-dependent dynamin-mediated process, i.e., membrane constriction and fission activity. Cryo-EM images of constricted dynamin-decorated lipid tubules show that (N)-GTP, but not (S)-GTP, enables this process. However, an automated survey of naturally occurring nucleotide ribose conformations in experimental structures of dynamin-related proteins shows almost exclusively the (S)-conformation. The explanation for this mismatch of conformations associated with dynamin-mediated processes, (N) in biochemical studies and (S) in experimental structures, will be addressed in future studies. We suggest that these new synthetic methanocarba-GTP analogues, including non-hydrolyzable modifications of the triphosphate moiety, here applied only to the dynamin system, will have broad applications as biochemical tools for studies of the role and conformation of guanine nucleotides in biochemistry [16].  Table 1. A red circle indicates two compounds (ADP) not bound to the dynamin-related binding site (in both cases ADP is bound to a myosin fusion protein), so they are not of interest.   Table 1. A red circle indicates two compounds (ADP) not bound to the dynamin-related binding site (in both cases ADP is bound to a myosin fusion protein), so they are not of interest. In conclusion, we synthesized the analogues of GTP containing a rigid ribose substitution, i.e., the methanocarba ring in both (N) and (S) fixed conformations. These correspond to two isomeric placements of a bicyclo[3.1.0]hexane ring system in place of the furanose ring. We used these analogues to interrogate the conformational requirements of a GTPdependent dynamin-mediated process, i.e., membrane constriction and fission activity. Cryo-EM images of constricted dynamin-decorated lipid tubules show that (N)-GTP, but not (S)-GTP, enables this process. However, an automated survey of naturally occurring nucleotide ribose conformations in experimental structures of dynamin-related proteins shows almost exclusively the (S)-conformation. The explanation for this mismatch of conformations associated with dynamin-mediated processes, (N) in biochemical studies and (S) in experimental structures, will be addressed in future studies. We suggest that these new synthetic methanocarba-GTP analogues, including non-hydrolyzable modifications of the triphosphate moiety, here applied only to the dynamin system, will have broad applications as biochemical tools for studies of the role and conformation of guanine nucleotides in biochemistry [16].

Computation of Puckering Parameters
The computation of puckering parameters was performed by using an in-house script as a modified version of the reported script [17]. We used as input a list of UniProt [18] codes of protein of interests, as reported in the second column of Table 1. The script automatically retrieves the PDB and CIF files deposited in the Protein Data Bank [19] for each protein of interest. The SDF files of all the ligands in complex with the proteins (all chains) were downloaded, as well, and nucleoside/tide-like compounds were identified on the basis of SMARTS notation, as reported before [17]. The puckering parameters for these compounds were automatically computed and plotted on a polar plot reporting the pseudorotation phase angle (P) and the degree of deformation from the plane (ν max ) on the angular and radial coordinates, respectively. The script was written in python3, using the modules requests and urllib for interaction with PDB and UniProt websites, rdkit [20] for SMARTS [21] substructure search and dihedral computation, numpy [22] and math for angles conversion and puckering computation, and matplotlib [23] for plotting purposes.