Active site metals mediate an oligomeric equilibrium in Plasmodium M17 aminopeptidases

M17 leucyl aminopeptidases are metal-dependent exopeptidases that rely on oligomerization to diversify their functional roles. The M17 aminopeptidases from Plasmodium falciparum (PfA-M17) and Plasmodium vivax (Pv-M17) function as catalytically active hexamers to generate free amino acids from human hemoglobin and are drug targets for the design of novel antimalarial agents. However, the molecular basis for oligomeric assembly is not fully understood. In this study, we found that the active site metal ions essential for catalytic activity have a secondary structural role mediating the formation of active hexamers. We found that PfA-M17 and Pv-M17 exist in a metal-dependent dynamic equilibrium between active hexameric species and smaller inactive species that can be controlled by manipulating the identity and concentration of metals available. Mutation of residues involved in metal ion binding impaired catalytic activity and the formation of active hexamers. Structural resolution of Pv-M17 by cryoelectron microscopy and X-ray crystallography together with solution studies revealed that PfA-M17 and Pv-M17 bind metal ions and substrates in a conserved fashion, although Pv-M17 forms the active hexamer more readily and processes substrates faster than PfA-M17. On the basis of these studies, we propose a dynamic equilibrium between monomer ↔ dimer ↔ tetramer ↔ hexamer, which becomes directional toward the large oligomeric states with the addition of metal ions. This sophisticated metal-dependent dynamic equilibrium may apply to other M17 aminopeptidases and underpin the moonlighting capabilities of this enzyme family.

M17 leucyl aminopeptidases are metal-dependent exopeptidases that rely on oligomerization to diversify their functional roles. The M17 aminopeptidases from Plasmodium falciparum (PfA-M17) and Plasmodium vivax (Pv-M17) function as catalytically active hexamers to generate free amino acids from human hemoglobin and are drug targets for the design of novel antimalarial agents. However, the molecular basis for oligomeric assembly is not fully understood. In this study, we found that the active site metal ions essential for catalytic activity have a secondary structural role mediating the formation of active hexamers. We found that PfA-M17 and Pv-M17 exist in a metal-dependent dynamic equilibrium between active hexameric species and smaller inactive species that can be controlled by manipulating the identity and concentration of metals available. Mutation of residues involved in metal ion binding impaired catalytic activity and the formation of active hexamers. Structural resolution of Pv-M17 by cryoelectron microscopy and X-ray crystallography together with solution studies revealed that PfA-M17 and Pv-M17 bind metal ions and substrates in a conserved fashion, although Pv-M17 forms the active hexamer more readily and processes substrates faster than PfA-M17. On the basis of these studies, we propose a dynamic equilibrium between monomer ↔ dimer ↔ tetramer ↔ hexamer, which becomes directional toward the large oligomeric states with the addition of metal ions. This sophisticated metal-dependent dynamic equilibrium may apply to other M17 aminopeptidases and underpin the moonlighting capabilities of this enzyme family.
The M17 family of metalloaminopeptidases, often referred to leucine aminopeptidases (LAPs), selectively cleave N-terminal amino acids from polypeptide substrates (1). LAPs have varying roles in numerous essential processes across all kingdoms of life, including regulation of immune responses in Solanceae (2,3), DNA recombination in Escherichia coli (4), free amino acid regulation in Toxoplasma gondii (5), and protection against oxidative stress in bovine cells (6). LAPs are routinely described as homohexamers, although variations in oligomeric states are observed in specific roles; monomers act as chaperones in tomatoes; catalytically active hexamers protect against oxidative stress; and multiple hexamers come together to aid in DNA recombination in E. coli (2,4,6,7).
LAPs have only ever been structurally resolved in the conserved homohexameric conformation. The catalytic Cterminal domains of each monomer are clustered in the center of the hexamer to form a buried catalytic core and are linked to the N-terminal domains by a central helical linker (8)(9)(10)(11). Each active site has capacity to coordinate at least two metal ions, which are essential for proteolytic activity (12,13). The active site metal ions are coordinated at two distinct positions, denoted site 1 and site 2 (11,12). Although zinc ions are the most common occupants of both sites, site 1, or the loosely bound site, can also coordinate Co 2+ , Mg 2+ , Mn 2+ , and Ca 2+ (14). Site 2, or the tightly bound site, is more limited and to date, has only been shown to only coordinate Zn 2+ and Co 2+ (12). Early studies suggested the two sites act independently to control enzyme function, with site 1 metal ions modulating rates of enzyme activity (k cat ) and site 2 metal ions controlling enzyme activation through manipulation of substrate affinity (K m ) (14,15). However, subsequent studies suggest the roles of site 1 and site 2 in enzymatic function may be more interdependent and complex than initially suggested (16). Despite their classification as zinc-metalloproteases, M17 aminopeptidases have been shown to have higher activity in the presence of cobalt or manganese ions than in the presence zinc ions (2,13,14,17,18).
In vivo localization studies indicate that PfA-M17 and Pv-M17 function within the Plasmodium cytosol and in vitro analysis of aminopeptidase activity show that the two enzymes are most active in mildly basic solutions, reflective of the cytosolic environment (pH 8.0) (19,20). Both PfA-M17 and Pv-M17 cleave N-terminal leucine residues effectively (19,20), and the complete PfA-M17 substrate selectivity profile indicates a preference for the hydrophobic amino acids leucine and tryptophan (26). Both aminopeptidases are present as hexamers in solution (19,20), and the PfA-M17 crystal structure forms the conserved M17 family hexameric assembly (11).
The two active site metal ions within M17 aminopeptidases have previously been only described in terms of their necessity for catalytic function. In this study, we discovered that the active site metal ions of PfA-M17 and Pv-M17 also play a structural role, operating as part of a previously undescribed mechanism of activity regulation for M17 aminopeptidases. We show that binding of active site metal ions mediates the association of inactive oligomers into active hexamers, and the dynamic equilibrium between those states could be manipulated with mutations designed to compromise active site metal binding. Structural characterization of the hexameric and tetrameric conformations using X-ray crystallography and cryoelectron microscopy, in combination with analytical ultracentrifugation (AUC) sedimentation velocity experiments, show that M17 aminopeptidases simultaneously adopt several oligomeric conformations and that the transition between oligomers (monomer dimer tetramer hexamer) is continuous, rapid, and directionally controlled by the metal ion environment.

PfA-M17 and Pv-M17 exhibit unique oligomerization behaviors
Recombinant PfA-M17 (residues 84-605) and Pv-M17 (residues 73-621) were successfully purified from a bacterial expression system using a two-step chromatography method. Both constructs contained an N-terminal deletion. Residues 1 to 83 were omitted from our original PfA-M17 construct to remove an asparagine-rich low complexity region after attempts to produce the full-length protein in E. coli were unsuccessful (19). Studies by Lee et al., 2010 (20) showed that Pv-M17 produced in E. coli was truncated at residue 73, which aligned to our original truncation for PfA-M17. Therefore, for consistency between the constructs, the Pv-M17 construct was designed with the same N-terminal truncation site despite the lack of an asparagine-rich low complexity region. The purified proteins were active in previously reported assay conditions (19,20) and had similar kinetic parameters to those previously reported (Table 1). Interestingly, the Pv-M17 enzyme is significantly faster than PfA-M17 in the presence of both Co 2+ and Mn 2+ , with k cat values at least 10x higher than PfA-M17 (Table 1). During purification, both Pv-M17 and PfA-M17 eluted from SEC as broad peaks suggestive of poorly separated oligomeric species. To further investigate the solution species of each protein, size exclusion chromatography multiangle light scattering (SEC-MALS) experiments with 100 μl samples at 3 mg/ml were used to dissect the oligomeric content of the protein samples. At the lower concentration and volume, PfA-M17 separated into two distinct populations at molar masses of 380 kDa and 70 kDa, which approximately correlated to a hexamer and monomer respectively (Fig. 1A). The peak area shows that PfA-M17 predominantly adopted the monomeric conformation with only a small proportion forming the hexamer. Pv-M17 also separated into two distinct populations, at a high molar mass (330-370 kDa) spanning the hexamer molecular weight and at a low molar mass (120 kDa), correlating to a dimer (Fig. 1B). These two populations were of approximately equal concentration and connected by a nonzero baseline, indicating dynamic movement between the two populations ( Fig. 1B, Gray bars). Catalytic activity of the separated PfA-M17 and Pv-M17 oligomeric states was almost completely limited to the hexameric populations ( Fig. 1, A-B). The small oligomeric species display some residual activity; however, this may be because of the small oligomers forming larger active species throughout the course of the experiment. More distinct separation of the Pv-M17 and PfA-M17 oligomeric states was achieved using AUC sedimentation velocity experiments. Higher resolution separation of Pv-M17 showed at least three distinct populations at standardized (s 20,w ) sedimentation coefficients of 6S, 9S, and 12S, with dynamic movement between these three populations (Fig. 1C). AUC analysis of PfA-M17 suggested the small molar mass peak observed in SEC-MALS experiments actually consists of two distinct populations at s 20,w of approximately 4S and 6S (Fig. 1C).
To create homogenous monomeric or stable small oligomeric species for use as a reference in AUC experiments, we sought to introduce point mutations to disrupt hexamerization. Investigation of our PfA-M17 structure identified that Trp525 and Tyr533 formed two sets of pi-pi stacking

M17 metal dependent dynamic equilibrium
interactions with corresponding residues from another chain within the hexamer. We hypothesized that these two residues were likely involved in stabilizing the hexameric conformation. We mutated both residues to Ala to eliminate these aromatic interactions between subunits. In solution, the PfA-M17 (W525A + Y533A) protein formed a single species that had a sedimentation coefficient of 3.7S (Fig. 1C), which we confirmed as monomeric using sedimentation equilibrium experiments (Fig. S1). Monomeric PfA-M17 (W525A + Y533A) was enzymatically inactive compared with wild type PfA-M17 in the same assay conditions (Fig. 1D). From this result, we could predict that the PfA-M17 WT sample in Figure 1C contained monomeric species at 4S and dimeric species at 6S, whereas Pv-M17 contained the dimeric species at 6S and tetrameric and hexameric species at 9S and 12S, respectively.

Oligomerization and activity are mediated by active site metal ions
With a relationship between the hexameric conformation and catalytic activity confirmed, we were interested in investigating factors that mediate the association of small inactive oligomers into active hexamers. M17 aminopeptidases are metal-dependent proteases that coordinate at least two metal ions per active site (11,12). Our analysis of purified PfA-M17 showed that it was predominantly monomeric in solution when purified from a bacterial expression system and that this monomer was catalytically inactive (Fig. 1, A and D). However, previous studies by us (11) and others have shown that addition of a divalent cation to assay buffers converts the inactive species to an active aminopeptidase. Therefore, we postulated that metal ions may also be playing an essential role in the formation of active hexamers. To investigate the influence of different metal ions on oligomerization and activity, both AUC sedimentation velocity experiments and catalytic activity assays were conducted in the presence of biologically relevant metal ions (Zn 2+ , Mg 2+ , Mn 2+ , and Co 2+ ). We observed unique oligomerization and activity responses to each metal ion tested and an overall behavioral difference between Pv-M17 and PfA-M17. PfA-M17 did not completely hexamerize in any of the metal ion conditions tested ( Fig. 2A), although Mn 2+ and Co 2+ did result in a shift toward the larger oligomeric states. Mg 2+ failed to induce any oligomerization of PfA-M17 given the AUC sedimentation profile remained similar to that of the no metal added sample. Pv-M17 associated into a single species in the presence of Mn 2+ and Co 2+ with sedimentation coefficients at 12S and 13.5S, respectively (Fig. 2B). Mg 2+ caused a more M17 metal dependent dynamic equilibrium defined separation of different oligomeric species although had minimal effect on overall oligomeric species present. Using AUC sedimentation equilibrium experiments, the Mn 2+induced 12S species was confirmed to have a molar mass of 356 kDa, consistent with the active hexamer (Fig. S2, Table S1). This also indicated that Co 2+ was possibly inducing Pv-M17 to form a soluble aggregate larger than the hexameric conformation. Zn 2+ was also tested as part of these experiments for both proteins; however, the proteins precipitated at low metal ion concentration.
Catalytic activity was also tested using the same panel of biologically relevant metal ions. Both PfA-M17 and Pv-M17 exhibited the highest activity in the presence of Mn 2+ and Co 2+ and negligible activity in Mg 2+ and Zn 2+ (Fig. 2C). When compared with AUC data, the metal ions that induced a complete or partial oligomeric shift toward the hexameric conformation were the same ions that corresponded with high catalytic activity levels. However, the presence of a large soluble aggregate was detrimental to catalytic activity, as evidenced by the Co 2+ -treated samples. Buffer conditions that produced relatively small proportions of the hexameric species resulted in comparatively low catalytic activity levels.
Given both PfA-M17 and Pv-M17 demonstrated a shift toward higher order oligomers coupled with high activity levels in the presence of Mn 2+ , we were interested to see if metal ion concentration, in conjunction with metal ion identity, contribute to formation of the active hexameric species. The AUC sedimentation velocity experiments were carried out Mn 2+ concentrations of 0 μM, 50 μM, and 1 mM. As metal ion concentration increased, we observed an increase in the  17) described a similar activity profile in increasing Zn 2+ concentrations and suggested the plateaus were because of saturation of the "loosely bound" and "tightly bound" metal-binding sites. By this model, Pv-M17 reached saturation of the loosely bound site at lower metal ion concentrations than PfA-M17, suggesting that Pv-M17 may coordinate metal ions in this position more readily.

Active site metal ion binding is essential for oligomerization
To confirm that the correlation between metal ion environment/concentration and hexamer formation was due to metal binding within the active site of the enzymes, we introduced point mutations designed to impair the ability of PfA-M17 and Pv-M17 to bind active site metal ions. We chose to alter the conserved Asp379 and Glu461 residues that show bidentate coordination of both metal ions to prevent or hinder the active site from functioning as a metal accepting pocket.
These residues were changed to Ala and Leu respectively, producing mutants PfA-M17 (D379A + E461L) and Pv-M17 (D395A and E477L), which we refer to herein as PfA-M17(AL) and Pv-M17(AL), respectively. The mutants were purified by the standard protocol and exhibited no obvious differences to wild-type enzymes(s) during purification. The catalytic efficiency of both PfA-M17(AL) and Pv-M17(AL) was approximately 10-fold lower than the respective wild-type enzyme in the presence of Co 2+ and approximately 1000-fold and 2000fold, respectively, in Mn 2+ (Table 1). In both PfA-M17 and Pv-M17, this extreme reduction in catalytic efficiency is a result of dramatically reduced substrate turnover rates (k cat ) ( Table 1).
Oligomerization of PfA-M17(AL) and Pv-M17(AL), compared with WT, was tested using AUC sedimentation velocity experiments in the presence and absence of 1 mM Mn 2+ . In the absence of Mn 2+ , PfA-M17(AL) sedimented as a very broad peak at 10S with a diminishing tail stretching to 4S and a second, much smaller peak at 4.5S (Fig. 3A), suggesting that the aminopeptidase was in a state of rapid and dynamic transition between low-order and high-order oligomers. When 1 mM Mn 2+ was added, the PfA-M17(AL) c(s) distribution became broader and absorbed the smaller peak observed earlier at 4.5S but remained largely unchanged with the addition of metal ions (Fig. 3A). Pv-M17(AL) showed similar oligomerization characteristics to PfA-M17(AL), adopting a broad curve that peaked at 11S with a diminishing tail that extended to 4S (Fig. 3B). Addition of 1 mM Mn 2+ caused the Pv-M17(AL) c(s) distribution to broaden slightly but as seen with PfA-M17(AL), the distribution remained largely unchanged (Fig. 3B). Distinct hexamerization is observed with addition of Mn 2+ to PfA-M17 WT and Pv-M17 WT (Fig. 3, A-B). In contrast, PfA-M17(AL) and Pv-M17(AL) both showed broad sedimentation profiles regardless of the presence of metal ions, suggesting that PfA-M17 (AL) and Pv-M17 (AL) fluctuate across a broad range of oligomeric species when metal binding is impeded. Furthermore, PfA-M17(AL) and Pv-M17(AL) adopted very similar oligomerization profiles when metal binding was impeded, unlike PfA-M17 WT and Pv-M17 WT which demonstrate distinct profiles when compared.
Cysteinyl-glycinase activity is mediated by a two-fold metaldependent mechanism M17 family aminopeptidases have been shown to digest the Cys-Gly dipeptide as part of the glutathione regulation pathway but only in the presence of manganese (6,7,27). We therefore investigated a secondary regulatory role of metal ions, in addition to influencing oligomerization. We characterized the cysteinyl-glycinase activity of PfA-M17 and Pv-M17 against the Cys-Gly dipeptide at pH ranging from 6.0 to 9.0 and in the presence of Zn 2+ , Mn 2+ , Co 2+ , Mg 2+ , and also with no metal ions added. PfA-M17 only showed catalysis of Cys-Gly in the presence of Mn 2+ at pH 8.0 (Fig. 4A). Pv-M17 also demonstrated activity in Mn 2+ at pH 7.0 and 8.0 (Fig. 4B). Pv-M17 maintains a faster rate of catalysis than PfA-M17; Pv-M17 reached absorbance levels in excess of 1.0 mAu while PfA-M17 reached absorbance levels of just 0.2 to 0.3 mAu in the same time frame. While PfA-M17 and Pv-M17 were both active against Leu-Mec in the presence of Co 2+ at pH 8.0 (Fig. 2C), neither exhibited catalysis of Cys-Gly in the same conditions despite the presence of the active hexameric conformation (Fig. 4, A-B). Despite the change in substrate from earlier characterization experiments, Pv-M17 maintained a faster catalysis rate than PfA-M17 (Fig. 4, A-B).
PfA-M17 and Pv-M17 unique behavior is not due to differences in active site structure Having continually observed differences in the metalresponsive behavior between PfA-M17 and Pv-M17, we were interested to investigate whether there were any structural differences in the active sites that would lead to disparities in metal binding and therefore in oligomerization and activity. The X-ray crystal structure of PfA-M17 was solved before this study, with all structures-both unliganded and in complex with various inhibitors (11,24,28,29)-showing the hexameric conformation that is characteristic of M17 family aminopeptidases. After failing to obtain high-resolution X-ray data of full length Pv-M17, we identified a unique 21 amino acid insertion rich in Gly and Ser (Gly = 42.8%, Ser = 33%) within the postulated N-terminal region (Fig. S3). The insertion which we have called the "Pv loop" was predicted to be a solvent-exposed flexible loop that may impede crystal packing and diminish data quality. To overcome this, the 25 amino acid Pv loop was excised from the N-terminal domain to produce the mutant Pv-M17Δ125-151.
Pv-M17Δ125-151 was successfully crystallized and a highresolution X-ray crystal structure solved to 2.6 Å (PDB ID: 6WVV) ( Table S2). The overall structure adopted the characteristic M17 hexameric conformation, composed of a "dimer of trimers" (Fig. 5A) with the six chains clustered within the hexamer core, with each catalytic domain connected to a solvent-exposed N-terminal domain (Fig. 5A). There were two complete hexamers in the asymmetric unit, with an overall RMSD over 2256 Cα atoms of 0.194 Å. Minor rigid body movement of the N-terminal domains relative to each other was evident upon overlay of each chain (Fig. S4A). Several solvent-exposed regions (Ser269-Glu271, Ala377-Glu381 in all chains, and Glu170-Asn172 and Lys194-Val196 in chains G-L) could not be modeled, suggesting these regions are disordered.
The Pv-M17Δ125-151 and PfA-M17 active site structures align closely, including the bound metal ions. However, while PfA-M17 crystals required zinc soaking to occupy both sites, Pv-M17Δ125-151 crystals did not require any metal ion treatment. The Pv-M17Δ125-151 maps showed clear electron density for two zinc ions in each active site at the same positions observed in the PfA-M17 structure (Fig. 5B,  Fig. S4C). Each of the metal ions was coordinated in a tetrahedral fashion by neighboring active site residues, and all residue coordination points were conserved with the PfA-M17 structure (Fig. 5B). PfA-M17, like most M17 aminopeptidases, has a carbonate ion located within the active site (Fig. S4C) (11). In Pv-M17Δ125-151, we observed a sulfate ion in the position of the carbonate (Fig. 5B). The presence of the sulfate ion in the Pv-M17Δ125-151 active site is likely a crystallization artefact (Fig. 5B), given the presence of sulfate ions in the crystallization buffer. Several M17 aminopeptidases have also reported the presence of an active site sulfate ion (30,31), although the vast majority of M17 aminopeptidases bind a carbonate ion at this position. Interactions between the sulfate ion and the zinc ion may contribute to stabilizing metal ion binding in the crystal structure, with atomic distances between the sulfate ion and Zn1 and Zn2 at lengths of 3.1 and 3.3 Å, respectively (Fig. S4B).
Kinetic characterization of Pv-M17Δ125-151 showed that the enzyme had a faster substrate turnover rate in comparison to the wild type enzyme, suggesting that the Pv loop has a role in substrate regulation (Table 1). Compared to the wild type enzyme, substrate affinity decreased in the presence of Co 2+ , while affinity remained largely consistent in the presence of Mn 2+ .The Pv loop is adjacent to a large channel through which substrates likely gain access to the buried active site (Fig. 5C). With excision of the Pv loop, the channel entrance is more exposed, allowing substrates to access the active site more readily and resulting in elevated substrate turnover speed. The channel entrance is flanked by a second loop, contributed by a neighboring chain, that has been described as a "substrate regulation loop" (Fig. 5C) (11). We further characterized the Pv-M17 structure using cryo-EM, which enabled us to determine the structure of the fulllength protein rather than the loop deletion mutant (PDB ID: 7K5K) (Table S3). Data were collected from samples prepared in the presence of 1 mM Mn 2+ to encourage formation of active hexamers and resulted in a map with global resolution of 2.8 Å (Fig. S5) that showed Pv-M17 in the conserved hexameric conformation (Fig. 6A). Map quality and resolution was the highest at the core of the hexamer and reduced toward the periphery of the hexamer. Each of the active sites had density for two metal ions, which were modeled as Mn 2+ ions, as well as a single carbonate ion. The carbonate ion electron density was planar in comparison to the Pv-M17Δ125-151 sulfate ion tetrahedral electron density observed at the same position (Fig. 6B). The position of residues involved in metalion coordination was conserved between the full length cryo-EM structure and the Pv-M17Δ125-151 crystal structure despite the difference in metal ion identity and inclusion of the carbonate ion. The Mn 2+ ions in the cryo-EM structure were slightly displaced when compared with the Zn 2+ ions in the crystal structure, although this may be because of the larger atomic radius of Mn 2+ in comparison to Zn 2+ . The carbonate ion bound in the cryo-EM structure active site was positioned 4.3 and 3.7 Å away from Mn1 and Mn2, respectively, distances that are likely beyond the physically feasible Mn-O bond length (Fig. S4D). This indicates that the occupation of both metal ion sites is not reliant on stabilization from the large sulfate ion as observed in the crystal structure.
While electron density for the Pv loop was poorer than the remainder of the hexamer, it could be modeled extending from the exterior of the hexamer into surrounding solvent. In its modeled position, the Pv loop is coiled and partially obstructs the substrate access channel (Fig. 6C). In solution, the Pv loop likely extends to occlude more of the channel and further Figure 6. The full length Pv-M17 cryo-EM structure also adopts the conserved M17 hexameric conformation and has a conserved active site conformation. A, 2.6 Å cryo-EM structure of full length Pv-M17. Cryo-EM structure also adopts the hexameric conformation, consisting of a front trimer depicted in shades of teal and labeled A, B, and C, and a back trimer depicted in shades of orange and labeled X, Y, and Z. B, Pv-M17 full length cryo-EM structure active site. Metal-coordinating residues are shown as yellow sticks, colored by atom. Mn1 and Mn2 are depicted as spheres. Mn1 is coordinated by the carboxyl group of Asp395, Asp475, and Glu477 and the amine group of Asp475. Mn2 is coordinated by the carboxyl group of Asp415, Asp395, Lys390, and Glu477. The active site additionally binds a carbonate ion that is beyond binding proximity to the two zinc ions. C, close up of N-terminal region with Pv loop of chain X (yellow) interacting with the substrate regulation loop of chain A (red) and occluding the active site substrate access channel.PDB ID: 7K5K.
restricts substrate access to the active site, accounting for the disparity in substrate turnover rates between wild type Pv-M17 and Pv-M17Δ125-151. The Pv loop interacts closely with the substrate regulation loop from the neighboring chain and together the two loops appear to have capacity to influence substrate entry and product egress from the active site. These two loops may function as substrate gatekeepers to the buried active site and operate in concert to regulate both substrate affinity and turnover speed (Fig. 6C).

Structural characterization of the small Pv-M17 oligomers
To gain further insight into the metal-dependent oligomerization mechanism and pathway, we sought to also structurally investigate the small oligomers of Pv-M17 that self-associate to form the active hexamer that has been resolved as part of this study (Fig. 5A). For this, we used cryo-EM, where we could control the environmental metal ion conditions and therefore the dynamic equilibrium and oligomeric species present.
In the presence of 100 mM EDTA, Pv-M17 formed a single species corresponding to a tetramer when analyzed using SEC (Fig. S6). We used cryo-EM to investigate the structure of this tetramer. Despite the apparent homogeneity observed on SEC, the EDTA-treated samples were incredibly difficult to optimize for freezing conditions that resulted in good particle distribution across the grid. In total, 56,679 particles were picked from grids, and 16,127 were used to generate a density map with a global resolution of 8.8 Å (Fig. 7A, Fig. S5). Molecular dynamic flexible fitting (MDFF) using various combinations of tetramer derived from the hexamer atomic model were used to produce the final model of Pv-M17 in the absence of metal (Fig. 7A). After MDFF, there were some regions of the map where the model fit poorly, which may be a result of the low map resolution or unanticipated restructuring of the tetramer subunits and interactions. The best fit tetramer consisted of a dimer from each side of the hexamer, in a "dimer of dimers" conformation ( Fig. 7A; A and C from front trimer; X and Y from back trimer). The two dimers interact via the C terminal at the base of the structure and are in an open conformation at the top (Fig. 7B). Although the map resolution was not sufficient to achieve an accurate atomic model, the backbone indicated that the C-terminal domains likely interact through an aromatic tetrad consisting of Trp541 and Trp542 with the four Trp residues forming two T-shaped pi-pi interactions (Fig. 7B). The Pv loop (Ala125-Ala151) from chains C and Y extend to interact with the neighboring chain, with the Pv loop from chain C interacting with chain Y and vice versa. (Fig. 7C). Chains C and Y interact with chains A and X through a secondary C-terminal interface that is in close proximity to the active sites of each chain (Fig. 7D). A disordered loop from chains C and Y (Ser396-Met412) extends into the C-terminal domain of chains A and X respectively and comes into close proximity with the active sites of A and X. Simultaneously, the loop pulls Asp395 and Asp415 away from the active sites of C and Y suggesting that the active sites are likely in a nonfunctional conformation.

Discussion
Self-assembly into high order oligomeric conformations is a common mechanism for controlling protein function (32). This form of regulation is involved in countless biological systems; hemoglobin must form α2β2 tetramers to transport Tetramer model was fit to calculated map using molecular dynamic flexible fitting (MDFF). The tetramer adopts a "dimer of dimers" conformation that associate via the C terminal of chains C and Y. The teal dimer is composed of chains A and C from the "front trimer" of the hexameric conformation, and the orange dimer is composed of chains X and Y from the "back trimer" of the hexameric conformation. B, proposed aromatic tetrad mediating the C-terminal interface between dimers. Tetrad consists of Trp541 and Trp542 from chain C and Y, each residue forming T-shaped interactions with two others. C, bottom view of tetramer; the flexible N-terminal loop (residues 125-151) of chains C and Y extend to associate with the neighboring chain and may help to stabilize the C-terminal interface. D, short C-terminal loops from chains C and Y (residues 396-412) extend into chains A and X, coming into close contact with the active sites of chains A and X.
M17 metal dependent dynamic equilibrium oxygen; actin forms long oligomers to facilitate cellular movements (33); porphobilinogen synthase must form homooctamers to facilitate production of building blocks required to form chlorophyll and vitamin B12 (34). More specifically, many metalloenzymes show a propensity to self-associate into large oligomers; M42 (TET aminopeptidase) (35)(36)(37) and M18 (aspartyl aminopeptidase) family aminopeptidases both form tetrahedral dodecamers, whereas M12 (carboxyl peptidase) family aminopeptidases form tetramers (38). The M17 aminopeptidases are one such family that undergo oligomerization to form homo-hexamers, a process that is essential for catalytic activation of the enzyme.

Metal-dependent oligomerization: a multitiered control mechanism for catalytic activity
Our results indicate that the M17 aminopeptidase oligomerization pathway and formation of the active hexamer is mediated by the acquisition and binding of metal ions within the active site. A similar metal-mediated oligomerization mechanism has previously been described to control the formation of catalytically active M42 dodecamers and porphobilinogen synthase octamers (34)(35)(36)(37). Further to controlling the process of M17 hexamerization, both the identity and concentration of metal ions available dictate the extent to which oligomerization occurs. Mn 2+ and Co 2+ induced oligomerization of PfA-M17 and Pv-M17, whereas Mg 2+ had little influence, and Zn 2+ caused irreversible protein aggregation. As formation of the hexamer is intrinsically linked to catalytic activity, it was expected that maximal catalytic activity would be observed in the presence of Mn 2+ and Co 2+ , which we saw for both PfA-M17 and Pv-M17. Similarly, increasing Mn 2+ concentrations increased formation of hexamer, accompanied by an increase in catalytic activity rates. That we could disrupt this oligomerization/activity relationship by mutating active site residues (PfA-M17(AL)] & Pv-M17(AL)) confirmed that active site metal binding is central to the oligomerization process. To investigate the biological relevance of this metaldependent oligomerization mechanism, further characterization of the smaller oligomeric states is required in addition to analysis of the native proteins isolated from Plasmodium parasites. The metal-dependent control of oligomeric states may hypothetically serve numerous purposes; to prevent unwanted proteolytic damage to cells, to prevent formation of inactive aggregates, or to dictate the role of M17 within an organism. Plasmodium in particular are known to have fluctuations in metal ion availability through the life cycle (39), meaning this equilibrium could potentially act as a biological control mechanism or a metal-dependent "activity switch".
In addition to the metal-dependent oligomerization mechanism controlling activity via formation of the active hexamer, M17 aminopeptidases have a secondary metal-dependent regulation mechanism. The catalysis of the Cys-Gly dipeptide as part of the glutathione breakdown pathway is a postulated role for several bacterial and plant M17 aminopeptidases and involves formation of the active hexamer to carry out the hydrolysis mechanism (40,41). However, Cys-Gly catalysis by PfA-M17 and Pv-M17 only occurred in the presence of Mn 2+ and notably did not occur in the presence of other metal ions known to induce hexamerization (e.g., Co 2+ ). Cappiello et al. (6) suggested any cations other than Mn 2+ would interact with the sulfhydryl ion of the Cys residue in such a way that the dipeptide would adopt an unhydrolyzable pose within the active site. Therefore, for Cys-Gly catalysis to occur, M17 aminopeptidases must adopt the hexameric conformation and bind the dipeptide substrate in a hydrolyzable pose; events that are both reliant on active site metal binding.
Differences in PfA-M17 and Pv-M17 oligomerization behavior is not dictated by active site structure Throughout this study, we continually observed unexpected differences in metal-regulated oligomerization and activity behaviors between PfA-M17 and Pv-M17. When the 2.6 Å Pv-M17 crystal structure was solved and the active site compared with that of PfA-M17, we found that the active site structure was highly conserved and held no indication that metal ions would bind differently between the two structures. The conserved affinity of PfA-M17 and Pv-M17 for Leu-Mec reiterates the high degree of active site conservation and indicates that substrate binding is also conserved between the two homologs (PfA-M17 K m = 15.0 μM, Pv-M17 K m = 20.1 μM). Although Pv-M17 and PfA-M17 metal ion and substrate binding were conserved, Pv-M17 exhibited both a faster substrate turnover rate and a greater capacity to form the active hexameric conformation than PfA-M17. These different catalytic and oligomeric behaviors are likely because of structural differences at a location distinct from the active site or eventuate from differences in the oligomerization pathway. While the catalytic domains of PfA-M17 and Pv-M17 have 90% sequence identity, the N-terminal domains are more divergent with only 53% sequence identity (Fig. S3). The Ntermini domains are largely responsible for the intramolecular interactions that likely help to establish and maintain the hexameric conformation. In particular, the N-terminal interface between chains at the "top" of the hexamer, where the chains from opposite sides of the hexamer interact, is considerably more positively charged in Pv-M17 than PfA-M17. This is largely because of the string of positively charged residues His-Lys-Lys (residues 230-232) in place of the PfA-M17 Leu-Ser-Lys. The positively charged regions from each of the N terminally associating chains sandwich a negatively charged region that, together, contribute to stabilizing this Nterminal interaction. This unique Pv-M17 feature is likely just one of many subtle structural differences between the PfA-M17 and Pv-M17 that culminate in the differences we observe in oligomerization behavior. The potential role of the N-terminal domains in modulating the oligomeric state of M17 aminopeptidase is caveated by the use of PfA-M17 and Pv-M17 constructs with N-terminal truncations of 83 and 72 residues, respectively. Several attempts to produce M17 aminopeptidases (PfA-M17 and tomato LAP-A) in E. coli have proven unsuccessful (2,19) and only succeeded upon truncation of their N-terminus. The LAP-A aminopeptidase was shown to undergo enzymatic processing in vivo, by way of M17 metal dependent dynamic equilibrium cleavage of a 5 kDa region from the N-terminal domain, that resulted in the active protein. Isolation of M17 aminopeptidases from Plasmodium parasites would clarify if the N-terminal regions undergo this same type of preprocessing, and the ultimate role these regions play in oligomerization.

Metal ions drive the M17 dynamic equilibrium toward large oligomeric states
Metal-dependent oligomerization is a key process in dictating the role and function of M17 aminopeptidases; however, the process of how M17 aminopeptidases selfassociate from monomer to hexamer had not been investigated before this study. Our results suggest that in the presence of low concentrations of metal-ion, monomeric M17 aminopeptidase can self-associate to form a dimer, via aromatic interactions at the C-terminal domain interface (Fig. 8). Interruption of these aromatic interactions results in the M17 enzyme being locked into a monomeric species, regardless of the metal ion environment. As the environmental metal ion concentration increases, formation of the tetrameric species is observed. This is achieved through association of two free monomers with the existing dimer via C-terminal interactions at a secondary C-terminal interface different to that involved in the dimerization process. Finally, in the presence of sufficient cation, the final two monomers join to form the active hexamer (Fig. 8). Based on the MDFF simulated model, the tetramer must undergo slight structural rearrangements to accommodate these final two monomers, which interact with the tetramer species via both C-terminal and N-terminal interactions. This final hexamerization process was disrupted when metal ion binding was compromised, as PfA-M17 (AL) and Pv-M17 (AL) oligomerized to a tetramer or pentamer but failed to form the hexameric species. This suggests that unimpeded metal binding in the active site is essential for tetramers to undergo the structural rearrangements necessary to form the active hexamer.
We have seen from both AUC and SEC-MALS that Pv-M17 and PfA-M17 (WT) simultaneously adopt a series of oligomeric species at any one time, indicating that the formation or breakdown of hexamers does not occur via a regimented stepwise fashion with all aminopeptidases adopting the same oligomeric species before progressing to the next. In fact, this would likely be impossible as our oligomerization model suggests the larger oligomeric states are formed by interaction of high-order and low-order oligomers. Rather, PfA-M17 and Pv-M17 exist in a rapid and continuous equilibrium between different oligomeric states that can be controlled by modifying environmental metal ion conditions.
Beyond their application as antimalarial drug targets (22,23,25), M17 aminopeptidases show potential for use in agricultural (42,43) and environmental industries (44). The ability to adopt numerous oligomeric conformations with distinct functionalities underpins the diversity of potential M17 aminopeptidase applications. The metal-dependent hexamerization mechanism together with the metal-dependent substrate positioning results in a highly regulated and multitiered system with capacity for fine control over catalytic activity. This kind of mechanism lends itself to manipulation of the metal ion environment to dictate enzyme efficiency, substrate specificity, and control oligomer-specific functions for pharmaceutical, agricultural, and industrial purposes.

Experimental procedures Molecular techniques
The gene encoding recombinant wild-type Pv-M17 (residues 73-621) with an in-frame C-terminal His 6 tag was chemically synthesized by DNA 2.0 using codons optimized for gene expression in E. coli. Pv-M17 was cloned into a pJ404 expression vector, which also encodes for ampicillin resistance. Pv-M17 mutants were generated by PCR mutagenesis. Metal binding mutant PvM17-D395A was constructed and used as a template to produce double mutants, D395A E477L, utilizing primers outlined in Table S4. The Pv-M17 N-terminal loop deletion mutant (PvM17Δ125-151) was constructed using primer pairs amplifying outward from the region of interest (Table S4). The PfA-M17 protein was produced from constructs described previously (11), and mutants PfA-M17(AL) and PfA-M17(W525A + Y533A) were synthesized by Figure 8. Schematic of the metal-dependent dynamic equilibrium of M17 aminopeptidase oligomeric species . As metal ion concentration increases, the M17 aminopeptidase dynamic equilibrium shifts toward generation of active hexameric species. In low metal ion concentration, monomers selfassociate to form dimers. As metal ion concentration increases, monomers associate with the existing oligomers to form tetramers and then hexamers.
GenScript. Wild-type and mutant sequences were confirmed using Sanger sequencing.

Production and purification of Pv-M17 and PfA-M17
Recombinant PfA-M17 and Pv-M17 genes were expressed in E. coli BL21 DE3 cells. Cells were grown in autoinduction media to force overexpression of the target proteins. E. coli cells were lysed by sonication, centrifuged, and the soluble supernatant was pooled. Proteins were purified via a two-step chromatography process involving nickel-affinity chromatography via the encoded C-terminal hexa-histidine tag (HisTrap Ni 2+ -NTA column; GE Healthcare Life Sciences) and sizeexclusion chromatography (Superdex S200 10/300 GL column; GE Healthcare Life Sciences) as previously described (11,45). Proteins were stored in 50 mM HEPES pH 8.0, 0.3 M NaCl at −80 C until use.

Enzyme kinetics
Aminopeptidase enzyme assays were carried out in white 384-well plates (Axygen) in a final volume of 50 μl. Aminopeptidase activity was determined by continually measuring the liberation of fluorogenic leaving group, 7-amido-4-methylcoumarin (NHMec) from the commercially available substrate, Leucine-7-amido-4-methylcoumarin hydroxide (referred to in this study as Leu-Mec) (Sigma-Aldrich), using a FLUOStar Optima plate reader (BMG Labtech) with excitation and emission wavelengths of 355 nm and 460 nm, respectively. Emitted fluorescence was continuously recorded and quantified in fluorescence units. Data were analyzed using PRISM GraphPad 8 software. Metal ion and pH dependence were measured in 100 mM bis-Tris (pH 6.0, 7.0) or Tris (pH 8.0, 9.0), with 0.2 mM or 1.0 mM MnCl 2 , MgCl 2 , ZnCl 2 , and CoCl 2 . 150 nM PfA-M17 or Pv-M17 was incubated with metal ions for 10 min at 37 C before addition of 10 μM substrate and activity measured for 1 h, or until steady-state was achieved. Kinetic assays were carried out in 100 mM Tris pH 8.0 buffer supplemented with 1 mM CoCl 2 or MnCl 2 . PfA-M17 and Pv-M17 concentrations were constant at 150 nM. Substrate concentration ranged from 0 μM to 500 μM. Substrate was incubated for 10 min at 37 C before addition of protein and activity measured for 1 h.
Hydrolysis of the cysteinyl-glycine dipeptide substrate (referred to as Cys-Gly in this study) was measured using a colorimetric method detecting liberation of free cysteine from the Cys-Gly dipeptide, using a protocol adapted from (6). Reactions were carried out in 1.5 ml microcentrifuge tubes, and reactions volumes were 100 μl. 250 nM of PfA-M17/Pv-M17 was incubated with 2 mM Cys-Gly at 37 C for 15 min in varying metal ions at 0.2 mM (ZnCl 2 , MnCl 2 , CoCl 2 , NiCl 2 , MgCl 2 , Metal free) and pH conditions (bis-Tris pH 6.0, bis-Tris pH 7.0, Tris-Cl pH 8.0, Tris-Cl pH 9.0). Reactions were stopped by the addition of 100 μl 100% acetic acid. 100 μl of a ninhydrin reagent solution containing 50 mg ninhydrin dissolved in 2 ml 4 M HCl/acetic acid (2:3) was added, and tubes were lightly mixed and centrifuged using a benchtop microcentrifuge. Samples were heated at 100 C for 10 min, then cooled on ice before 100 μl was transferred to a 96-well plate and absorbance measured at 560 nm (Clariostar, BMG LabTech).

Analytical size-exclusion chromatography
A 100 μl sample was loaded on to a Superdex S200 Increase 10/300 GL gel-filtration column (GE Healthcare Life Sciences) preequilibrated with 50 mM HEPES pH 8.0, 0.3 M NaCl using a Fast Protein Liquid Chromatography system (ÄKTA Purifier, GE Healthcare Life Sciences). Before loading, concentrated samples were centrifuged at 13,000 rpm, 20 min, 4 C to remove any residual debris or aggregate. A flow-rate of 0.5 ml/ min was applied for a minimum of 25 ml and 0.5 ml fractions collected (Äkta Frac-920, GE Healthcare Life Sciences). Size Exclusion Chromatography Multi-Angle Light Scattering was carried out in identical conditions to analytical SEC experiments (UFLC, Shimadzu; Dawn Helios-II, Wyatt; Optilab T-rEX, Wyatt), and data were analyzed using Astra 6. To test fraction activity, 25 μl of each fraction (concentration not normalized) was pipetted into a white 384-well plate and incubated at 37 C for 10 min. Following incubation, 10 μM Leu-Mec substrate diluted in 100 mM Tris pH 8.0, 1.0 mM CoCl 2 was added, and activity monitored for 60 min. Resultant activity across the 60-min time frame was expressed as a fluorescence units/min and plotted against corresponding fraction elution volume using Prism GraphPad 8.

Analytical ultracentrifugation
AUC experiments were conducted using a Beckman Coulter Optima Analytical Ultracentrifuge at a temperature of 20 C. For sedimentation velocity experiments, samples were loaded at a concentration of 0.5 mg/ml. Sample buffer (50 mM HEPES pH 8.0, 0.3 M NaCl) was supplemented with 50 μM MnCl 2 , CoCl 2 , MgCl 2 , ZnCl 2 , or 1 mM MnCl 2 as appropriate. 380 μl of sample and 400 μl of reference solution (sample buffer) were loaded into a conventional double sector quartz cell and mounted in a Beckman 4-hole An-60 or 8-hole An-50 Ti rotor. Samples were centrifuged at a rotor speed of 40,000 rpm, and the data were collected continuously at a single wavelength, most commonly at 280 nm. Solvent density and viscosity as well as estimates of the partial specific volume (0.742928 ml/g for PfA-M17, 0.739998 ml/g for Pv-M17 at 20 C) were computed using the program SEDNTERP (46). Sedimentation velocity data were fitted to continuous size [c(s)] distribution model using the program SEDFIT (47,48). Data were plotted using Prism GraphPad 8.
For sedimentation equilibrium experiments, reference and sample sectors were loaded with 140 μl reference and 100 μl sample (8.3 μM and 4.2 μM) plus 20 μl FC-43 oil in the sample sector. After initial scans at 3000 rpm to determine optimum wavelength and radial range for the experiment, samples were centrifuged at 12,000 and 18,000 rpm at 20 C. Data at each speed were collected at multiple wavelengths every hour until sedimentation equilibrium was attained (24 h). Sedimentation equilibrium data were analyzed with SEDPHAT (49) using the single species analysis model. X-ray crystallography of Pv-M17 N-terminal mutant N-terminal loop deletion Pv-M17Δ125-151 crystallization conditions were identified using sparse-matrix crystal screening. Crystals were observed in 0.1 M HEPES pH 6.5, 15% PEG 3350 and 0.15 M (NH 4 ) 2 SO 4 at protein concentrations of 3 and 6 mg/ ml and reservoir: protein ratio of 1:1. Crystals were mostly singular and rectangular prism in morphology. Data were collected at 100 K at the Australian Synchrotron MX2 beamline 31D1 (50). Diffraction data were collected to 2.6 Å. Diffraction images were processed using X-ray detector software and Pointless (51,52). 5% of the reflections were set aside for calculation of R free to compare against refined data. The Pv-M17Δ125-151 structure was solved by molecular replacement using a search model constructed from the PfA-M17 crystal structure (RCSB ID 3KQZ) in Chainsaw (53). A molecular replacement search using Phaser (54,55) identified 12 copies of the search model in the asymmetric unit, arranged into two independent hexamers. Model building proceeded from the initial Phaser output using PHENIX refinement (56). For initial refinement rounds in PHENIX, noncrystallographic symmetry (NCS) maps were produced to assist in initial model building. Between rounds of refinement, the structure was visualized using COOT (57,58), and the structure adjusted according to 2Fo-Fc and Fo-Fc electron density maps. After several iterations of refinement and model building, NCS maps were disabled as an option in PHENIX, and the stereochemistry and ADP weight optimized. On occasion, an NCS map was calculated in COOT to assist in modeling of flexible regions and placement of ligands. The structure was validated with Molprobity (59) and figures generated using PyMOL, version 1.8.2.3.

Cryoelectron microscopy of Pv-M17 oligomeric states
For the single particle cryo-EM analysis of the hexameric and smaller oligomeric conformations, Pv-M17 was concentrated to 2 mg/ml and preincubated in the presence of 1 mM MnCl 2 or 100 mM EDTA, respectively. 4 μl of sample was applied to Quantifoil Cu R 1.2/1.3 grids (Quantifoil Microtools GmbH) which had undergone a plasma discharge and were vitrified in liquid ethane using a Vitrobot MkIV plunge freezer (Thermo Fisher Scientific), with the sample chamber set to 100% humidity and 4 C. Grids were stored in liquid nitrogen until data collection. Data were collected on a FEI Titan Krios (Thermo Fisher Scientific) operating at 300 kV fitted with a Quantum energy filter (Gatan) and K2 (Gatan) Direct Electron Detector. The data were collecting using a pixel size of 1.06 Å and with a total average electron dose per image of 63 e/Å 2 for the hexameric species and 105 e/Å 2 for the tetrameric species.
Micrograph movies were imported into RELION (v3.0, v3.1) (60) and motion corrected using MotionCor2 and CTF estimation performed using CTFFIND4.1 (61). Particles were picked using the autopicking function in RELION using a Laplacian-of-Gaussian approach. Extracted particles were analyzed using RELION and cryoSPARC (62), and reference free 2D classification was performed. 2D Classes that corresponded to the expected Pv-M17 molecule were selected, and an ab initio 3D model was generated in cryoSPARC. This model was used as an initial model for further Euler angler refinement in RELION.
Final 3D maps were further refined using RELION using the Gold-standard approach to a final resolution of 2.8 Å. Maps were inspected using UCSF Chimera (63). For the hexameric structure, the Pv-M17Δ125-151 crystal structure determined as part of this study was used as an atomic model template and fitted to the cryo-EM electron density map. Adjustments to the template model were carried out using Coot and the final hexamer model refined using the real-space refinement tool in PHENIX. A tetramer based on the hexamer model was used as the starting model for the low oligomeric weight map. The starting model was docked into target map using Chimera "Fit in Map" function. MDFF simulations were used to produce a reasonable starting atomic model and was performed using NAMD (v2.13) software package (64). Final model refinements were performed by real space refinement as implemented in the PHENIX software package (56) followed by manual model curation in COOT (57,58). See Table S3 for structure statistics.

Data availability
The coordinates for the X-ray crystal structure of Pv-M17Δ125-151 can be found at PDB ID 6WVV and the cryo-EM Pv-M17 7K5K.pdb.