Structure and ion-release mechanism of PIB-4-type ATPases

Transition metals, such as zinc, are essential micronutrients in all organisms, but also highly toxic in excessive amounts. Heavy-metal transporting P-type (PIB) ATPases are crucial for homeostasis, conferring cellular detoxification and redistribution through transport of these ions across cellular membranes. No structural information is available for the PIB-4-ATPases, the subclass with the broadest cargo scope, and hence even their topology remains elusive. Here, we present structures and complementary functional analyses of an archetypal PIB-4-ATPase, sCoaT from Sulfitobacter sp. NAS14-1. The data disclose the architecture, devoid of classical so-called heavy-metal-binding domains (HMBDs), and provide fundamentally new insights into the mechanism and diversity of heavy-metal transporters. We reveal several novel P-type ATPase features, including a dual role in heavy-metal release and as an internal counter ion of an invariant histidine. We also establish that the turnover of PIB-ATPases is potassium independent, contrasting to many other P-type ATPases. Combined with new inhibitory compounds, our results open up for efforts in for example drug discovery, since PIB-4-ATPases function as virulence factors in many pathogens.


PIB-4-ATPases appear to have the widest scope of transported ions of the PIB-ATPases, 69
and it is possible that further sub-classification principles and sequence motifs will be 70 identified. Due to the broad ion transport range, they have been proposed to serve as 71 multifunctional emergency pumps that can be exploited under extreme environmental 72 stress to maintain heavy metal homeostasis (26) . 73 Hitherto, the available high-resolution structural information of full-length PIB-74 ATPases is limited to two structures each of ion-free conformations of the Cu + -75 transporting PIB-1-ATPase from Legionella pneumophila (LpCopA) (27,28) , and the Zn 2+ -76 transporting PIB-2-ATPase from Shigella sonnei (SsZntA) (28) . Thus, the principal 77 architecture of the PIB-4-ATPases remains debated, as sequence analyses have proposed 78 different topologies for the N-terminus: with or without i) the so-called called heavy 79 metal binding domains (HMBDs), and ii) the first two transmembrane helices, MA and 80 MB (16,(29)(30)(31) , which both are present in other PIB-ATPases (Supplementary Figure 1a). 81 These represent structural features that have been suggested to be important for ion-82 uptake and/or regulation in other PIB-ATPases (27,28,32,33) , raising questions if these 83 levels of protein control are absent or replaced in the PIB-4 group. In addition, despite a 84

Metal specificity 94
We employed the established PIB-4 model sCoaT (UniProt ID A3T2G5) to shed further 95 light on the structure and mechanism of the entire PIB-4-class. As the metal ion 96 specificity of the PIB-4-ATPases is known to be wide, the ATPase activity was assessed 97 in vitro in lipid-detergent solution, in the presence of a range of different heavy metals. 98 The protein exhibited clear Zn 2+ and Cd 2+ dependent ATPase activity, while Co 2+ only 99 stimulated ATP-hydrolysis at high ion concentrations. This is in partial agreement with 100 the ion range profile previously reported for sCoaT, as higher Co 2+ sensitivity has been 101 detected using a different functional assay and different experimental conditions (18) Page 4 (Supplementary Figure 2). However, it cannot be excluded that Co 2+ , rather than Zn 2+ , 103 is the preferred cargo in vivo as the relative intracellular availability of Co 2+ is more 104 than three orders of magnitude higher than that of Zn 2+ in certain bacterial cells (35) . 105 106

Structure determination 107
We determined structures of sCoaT in metal-free conditions supplemented with two 108 different phosphate analogues, BeF3and AlF4 -, respectively, which previously have 109 been exploited to stabilize E2 reaction intermediates of the transport cycle. The 110 structures were determined at 3.1 Å and 3.2 Å resolution, using molecular replacement 111 as phasing method and SsZntA as search model, and the final models yielded R/Rfree of 112 24.4/26.8 and 21.8/25.5 (Supplementary Table 1). The two crystal forms were 113 obtained using the HiLiDe method (crystallization in the presence of high 114 concentrations of detergent and lipids) (36) . Surprisingly however, the crystal packing 115 for both structures reveal only minor contacts between adjacent membrane-spanning 116 regions, which are critical for the crystals obtained of most other P-type ATPase 117 proteins (Supplementary Figure 3). Hence, some crystal forming interactions likely 118 take place through lipid-detergent molecules. To our knowledge, this is the first time 119 that type I crystals with unrestrained transmembrane domains are reported, but a 120 consequence is that peripheral parts of the membrane domain are less well-resolved 121 (Supplementary Figure 4). While this caused difficulty in modelling some 122 transmembrane helices, satisfying solutions were found with the aid of the software 123 ISOLDE (37) due to its use of AMBER forcefield which helped to maintain physical 124 sensibility in the lowest-resolution regions. 125 Page 6 Zn 2+ acquisition, an ability that is lost when exchanged with the N-terminal part of 171 ZiaA. From this it is clear that further studies are needed to shed light on the function 172 of the N-terminal region in PIB-ATPases, also in PIB-4-ATPases. 173 Associated, this raises questions also on the role of the above-mentioned MB' platform, 174 which has been proposed to serve as an interaction site for HMBDs in PIB-1-and PIB-2-175 ATPases, and for the Cu + -ATPases also as a docking site for metal delivering 176 chaperones (27,28,32,41) . As there are no known zinc/cadmium chaperones for PIB-4-177 ATPases, and because classical HMBDs thus appear to be missing in at least some 178 proteins of the group, the MB' function may need to be revisited. Alternatively, the N-179 terminus may have merely been maintained through evolution without conferring 180 functional benefits or disadvantages. 181 182

Structures in transition states of dephosphorylation 183
The classical view of P-type ATPases is that the E2P state is outward-open and that the 184 following transition state of dephosphorylation, E2.Pi, is occluded, and that these 185 conformations can be stabilized using the phosphate analogues employed here for 186 structure determination, BeF3and AlF4 -, respectively. Furthermore, distinct ion release 187 pathways have been proposed among PIB-ATPases Notably, analogous highly similar BeF3 --and AlF4-stablized structures have recently 205 also been observed for the Ca 2+ -specific P-type ATPase from Listeria monocytogenes 206 (LMCA1) (43) . It was proposed that LMCA1 pre-organizes for dephosphorylation 207 already in a late E2P state (E2P*, stabilized by BeF3 -), in accordance with its rapid 208 dephosphorylation. Favoured occlusion and activation of dephosphorylation directly 209 upon ion-release may also be the case for sCoaT, and consequently the E2-BeF3 -210 structure captured here may represents a late (or quasi) E2P state (E2P*). The high affinity binding site in PIB-4 ATPases has previously been suggested to be 236 formed of residues from the conserved SPC-(starting from S325) and HEGxT (from 237 H657) -motifs of M4 and M6, based on X-ray absorption spectroscopy (XAS) and 238 mutagenesis studies (18,44) . An outstanding remaining question is however how the ion 239 then is discharged to the extracellular site? Remarkably, E658 of M6 is pointing away 240 from the ion-binding region around the SPC motif in the sCoaT structures (Figure 2f  241 and Supplementary Figure 4a). We anticipate that E658 rotates away from its ion-242 binding configuration in the E1P to E2P transition, thereby assisting to lower the cargo-243 affinity to permit release via the M4, M5 and M6 cavity (Figure 2f). The conserved 244 E120 of M2 (sometimes replaced with an aspartate in PIB-4-ATPases) is located along 245 this exit pathway, and it overlays with the conserved E202 in SsZntA (Figure 2g), 246 which has been suggested to serve as a transient metal ligand, stimulating substrate 247 release from the CPC motif of PIB-2-ATPases (28) . We propose a similar role for E120 in 248  proposed an ion-binding stoichiometry of one (18,22,26,44) , however no information is 267 available regarding the presence or absence of counter transport. 268 In the E2-BeF3 -sCoaT structure, we identify a tight configuration of HEGxT-motif 269 H657, being sandwiched between the SPC residues, distinct from the M5 lysine -M6 270 aspartate interaction observed in PIB-2-ATPases (Figure 2h-i). Despite the packing issues of the generated crystals, clear electron-density is visible for H657, indicating a 272 rigid conformation (Figure 2h). Moreover, activity measurements of an alanine 273 substitution of H657 demonstrate that it is crucial for function (Figure 2a). In light of 274 these findings and an earlier report suggesting that a mutation of the equivalent of H657 275 in MtCtpD leaves the ion affinity unaffected (44) , we suggest this histidine serves as an 276 internal counter-ion, similarly as for the invariant lysine in SsZntA, perhaps preventing 277 back-transfer of released ions and for charge stabilization, however we cannot exclude 278 that H657 is also part of the high affinity binding site in sCoaT. 279 The rigid conformation observed for H657 in the E2-BeF3structure is also observed in 280 the E2-AlF4structure (Supplementary Figure 4b). In contrast, for SsZntA the 281 interaction between K693 and D714 is only detected in the E2.Pi state. Thus, the 282 interaction pattern is consistent with the idea that sCoaT pre-organizes for 283 dephosphorylation already in the (late) E2P state, with the associated occlusion and 284 internal counter-ion interaction taking place earlier than for SsZntA. 285

A more potent A-domain modulatory site 287
A conserved K + -site, which cross-links between the A-and P-domains in E2 states and 288 thereby allosterically stimulates the E2P to E2 process (52, 53) , has been suggested to be 289 present also in PIB-ATPases (52) . However, our new E2 structures and available 290 structures of PIB-1-and PIB-2-ATPases suggest that the A-/P-domain linker is maintained 291 without K + in PIB-ATPases, and instead is established directly between R273/D601 in 292 sCoaT as also supported by potassium titration experiments monitoring sCoaT ATPase 293 activity (Figure 3a-d). Nevertheless, the point-of-interaction appears critical for PIB-294 ATPases, as functional characterization of R273A, D601A and D601K result in a 295 marked reduction of turn-over (Figure 2a). This differs from similar mutations of 296 classical P-type ATPases, where only minor effects are observed (52,53) . Furthermore, 297 substitution of D601 with glutamate suggests that even the A-/P-domain distance is 298 critical (Figure 2a). It is possible that PIB-ATPases are more reliant on this particularly 299 tight, ion-independent stabilization, as the A-M1/A-domain linker is absent, and 300 because many other P-type ATPases also have a complementary A-/P-domain 301 interactions. Thus, our data indicate that this regulation is a general feature of many P-302 type ATPase classes, yet featuring unique properties for PIB-ATPases. 303 304 New metal-transport blockers 306 PIB-2-and PIB-4-ATPases serve as virulence factors and are critical for the disease caused 307 by many microbial pathogens, as underscored by the frequent presence of several 308 redundant genes (54)(55)(56)(57) . In this light and because these P-type ATPases are missing in 309 humans, they represent putative targets for novel antibiotics. The shared mechanistic 310 principles identified here suggest that compounds can be identified that inhibit both PIB-311 groups, for example directed against the common release pathway, thereby increasing 312 efficacy. Indeed, screening of a 20,000-substance library using a complementary in 313 vitro assay, uncover several leads that abrogate function of sCoaT and SsZntA ( Figure  314 3e-f, data only shown for sCoaT). Furthermore, initial tests of two of these suggest they 315 have a potent effect against mycobacteria, which previously have been shown to be PIB-  Table 1). For 397 the E2-AlF4structure, initial phases were obtained by the molecular replacement (MR) 398 method using software PHASER (59) of the Phenix package (60) , and using the AlF4 --399 stabilized structure of SsZntA (PDB ID: 4UMW) as a search model. The E2-BeF3 -400 structure was solved using the generated E2-AlF4structure as a MR model. Both 401 crystal forms display poor crystal packing between the membrane domains 402 (Supplementary Figure 3), deteriorating the quality of the electron density maps in 403 these regions (Supplementary Figure 4). In this light, model building of the membrane 404 domains were executed with particular prudence, taking into consideration the 405 connectivity to the well-resolved soluble domains, distinct structural features as well as 406 sequence and structure conservation patterns. Examples of such include the conserved 407 resolution, the SPC motif that twists the M4 helix and the conserved and functionally 409 important well-resolved residue H657 that assisted assigning nearby residues. 410 Initial manual model building was performed primarily using COOT (61) . ISOLDE (62) in 411 ChimeraX (63) was employed for model building and analysis, and was critical for 412 obtainment of the final models with reasonable chemical restraints and low clash score. 413 In particular, ISOLDE's interactive register shifting tool was instrumental in 414 determining the register of the most weakly resolved TM helices. Secondary structure 415 restraints were applied in some flexible regions, also taking into consideration 416 homology to sCoaT and other models. 417 During final refinements with phenix.refine (64) the geometry was restrained in torsion 418 space to ISOLDE's output. Molprobity was exploited for structure validation (65) . The 419 final models are lacking the first 40 residues only, which is shorter than a classical 420 MBD of 67 amino acids. All structural figures were generated using Pymol The two crystal structures, E2-AlF4and E2-BeF3 -, were inserted into a DOPC (1,2-499 dioleoyl-sn-glycero-3-phosphocholine) membrane patch using the CHARMM-GUI 500 membrane builder (69) . The membrane positions were predicted by the Orientations of 501 Proteins in Membranes (OPM) server (70) . During the simulation equilibration phase, 502 position restraints were gradually released from the water and lipids for a total of 30 ns 503 followed by 500 ns non-restrained production runs. Each protein state was simulated in 504 independent repeat simulations starting from a different set of initial velocities, adding 505 up to a sampling total of 500 ns x 4. A Nose-Hoover temperature coupling (71) was 506 applied using a reference temperature of 310 K. A Parrinello-Rahman pressure 507 coupling (72) was applied with a reference pressure of 1 bar and compressibility of 4.5e-508 5 bar -1 in a semi-isotropic environment. The TIP3P water model was used and the 509 system contained 0.15 M NaCl. The E2-AlF4system was composed of 256 lipids and 510 29429 water molecules while E2- BeF3  ATPase assay in lipid-detergent solution with targeted residues in sequential order. The 579 wild-type (WT) specific activity using the employed experimental conditions in the 580 presence of 50 μM metal is 1.00 ± 0.01 μmol mg -1 min -1 with Zn 2+ and 2.80 ± 0.06 μmol 581 mg -1 min -1 with Cd 2+ , comparable to the activity previously measured for PIB-4-582 ATPases. For biological averages and s.d. see Supplementary Figure 2c D601K forms) as well as SsZntA (wild-type), using protein samples purified in the 608 absence of K + and Na + (see Methods). The mean + s.d. of technical replicates is shown 609 (n=3). KCl leaves the function of sCoat and SsZntA essentially unaffected in the 610 presence of Zn 2+ (cyan) or Cd 2+ (gray). The equivalent form of sCoaT D601K has 611 previously been exploited to demonstrate K + -dependence in the Na,K-ATPase (53) . 612 Collectively, these data suggest that the P-/A-domain site regulation is K + -independent 613 in PIB-ATPases, in contrast to classical P-type ATPases.