nature structural & molecular biology Article https://doi.org/10.1038/s41594-023-00922-y Essential protein P116 extracts cholesterol and other indispensable lipids for Mycoplasmas Received: 7 September 2022 Lasse Sprankel1,7, David Vizarraga 2,7, Jesús Martín2, Sina Manger1, Jakob Meier-Credo 3, Marina Marcos4, Josep Julve 5, Noemi Rotllan5, Accepted: 6 January 2023 Margot P. Scheffer1, Joan Carles Escolà-Gil 5, Julian D. Langer 3,6, Published online: 13 February 2023 Jaume Piñol4, Ignacio Fita 2 & Achilleas S. Frangakis 1 Check for updates Mycoplasma pneumoniae, responsible for approximately 30% of community-acquired human pneumonia, needs to extract lipids from the host environment for survival and proliferation. Here, we report a comprehensive structural and functional analysis of the previously uncharacterized protein P116 (MPN_213). Single-particle cryo-electron microscopy of P116 reveals a homodimer presenting a previously unseen fold, forming a huge hydrophobic cavity, which is fully accessible to solvent. Lipidomics analysis shows that P116 specifically extracts lipids such as phosphatidylcholine, sphingomyelin and cholesterol. Structures of different conformational states reveal the mechanism by which lipids are extracted. This finding immediately suggests a way to control Mycoplasma infection by interfering with lipid uptake. Mycoplasma pneumoniae is a facultative intracellular human patho- lipid in the membranes, accounting for 35–50% of the total lipid frac- gen that causes community-acquired pneumonia that can result in tion6. Comprehensive studies on other cholesterol-utilizing bacteria severe systemic effects1. Unlike other respiratory pathogens, there is are largely lacking; the best characterized organism in this group is no approved vaccine against M. pneumoniae2. Mycoplasma species lack Mycobacterium tuberculosis, for which it has been proposed that an a cell wall and have the smallest known genomes3. M. pneumoniae, with ABC transporter homolog and other genes from the mce4 operon are a 816-kb genome, is a model organism for a minimal cell4. Many of the involved in cholesterol uptake9. M. tuberculosis uses cholesterol as a metabolic pathways that are required to synthesize essential products carbon source, enabling long-term infections with the bacteria10. It has are absent, which makes uptake by specialized mechanisms necessary. been shown that M. pneumoniae survive long-term in cholesterol-rich In fact, M. pneumoniae cannot synthesize several of the lipids that are atherosclerotic plaques11. For other clinically relevant bacteria that use important components of the cell membrane, such as sphingomyelin, cholesterol, like Borrelia burgdorferi or Helicobacter pylori, the uptake phosphatidylcholine and cholesterol5. Instead, it must take up lipids mechanism remains elusive12. To date, it is unclear how Mycoplasma from the host environment, and it adapts its membrane composition spp. uptake lipids from the environment. depending on the medium in vitro6–8. Cholesterol in particular, which In this work, we report the structural and functional characteri- is present in only a few prokaryotes, is essential for M. pneumoniae zation of P116. This protein was originally reported to contribute to cells and several other Mycoplasma species6. It is the most abundant host-cell adhesion. Furthermore, P116 is an essential protein for the 1 Buchmann Institute for Molecular Life Sciences and Institute of Biophysics, Goethe University Frankfurt, Frankfurt, Germany. 2Instituto de Biología Molecular de Barcelona (IBMB-CSIC), Parc Científic de Barcelona, Barcelona, Spain. 3Proteomics, Max Planck Institute of Biophysics, Frankfurt, Germany. 4 Institut de Biotecnologia i Biomedicina and Departament de Bioquímica i Biologia Molecular, Universitat Autònoma de Barcelona, Cerdanyola del Vallès, Spain. 5Institut de Recerca de l’Hospital de la Santa Creu i Sant Pau and CIBER de Diabetes y Enfermedades Metabólicas Asociadas (CIBERDEM), Barcelona, Spain. 6Proteomics, Max Planck Institute for Brain Research, Frankfurt, Germany. 7These authors contributed equally: Lasse Sprankel, David Vizarraga. e-mail: ifrcri@ibmb.csic.es; achilleas.frangakis@biophysik.org Nature Structural & Molecular Biology | Volume 30 | March 2023 | 321–329 321 Article https://doi.org/10.1038/s41594-023-00922-y a PhC TRITC HOECHST Overlap P1 pAb P116 pAb 1 µm 1 µm b Core domain 9 nm Dimerization interface 90° Side view Protrusion N-terminal domain N-terminal domain 7 nm Cavity Protrusion Top view 24 nm Fig. 1 | Localization of P116 in Mycoplasma pneumoniae cells and structure particles), 90° apart. The homodimer is held together by the dimerization of P116. a, Phase contrast (PhC) immunofluorescence microscopy images of interface (shown in pink). The core domains have four contiguous antiparallel M. pneumoniae cells using labeling with polyclonal antibodies (pAb) against helices (shown in blue) and a β-sheet with five antiparallel strands (shown in the ectodomains of adhesin P1 (top row; used as a reference) and P116 (bottom orange). The N-terminal domain is shown in green. The top view displays a huge row) (two separate experiments with independent samples were performed). cavity that is fully accessible to solvent. The cleft providing access to the cavity Labeling for P1 concentrates at the tip of the cell; for P116, it covers the whole spans the whole core domain. Each monomer also has a distinct protrusion surface homogenously. b, Two views of the cryoEM density map of the complete (shown in blue as part of the antiparallel α-helices). extracellular region of the P116 dimer at 3.3-Å resolution (from 1.3 million viability of M. pneumoniae cells and is strongly immunogenic, thus we describe the mechanism by which Mycoplasma spp. extract lipids making it a promising target for therapeutics13. Despite the essential from the environment and possibly also deposit them in their own role of P116, the M. pneumoniae genome contains only a single copy of membrane, thus explaining the essential role of P116 in the survival of mpn_213 (gene encoding P116), and, on average, only 34 copies of the M. pneumoniae cells. protein are present in M. pneumoniae14. By contrast, the most immu- nogenic protein, P1, is not essential, has multiple gene copies present Results in the genome15, and has a 20-fold-higher copy number14. To elucidate P116 is evenly distributed on the cell surface the role of P116, we first determined the structure of the ectodomain A construct predicted to span the whole ectodomain of P116 from by single-particle cryo-electron microscopy (cryoEM). To the best of M. pneumoniae (residues 30–957) was overexpressed in Escherichia our knowledge, this structure represents a previously uncharacterized coli and purified by His-tag affinity and gel filtration chromatography fold (with no matches in the Protein Data Bank) featuring a uniquely (Methods and Extended Data Fig. 1). Immunolabeling with both poly- large hydrophobic cavity that is fully accessible to solvent. Using mass clonal and monoclonal antibodies against this construct showed an spectrometry, we identified several different lipids (including choles- intense and uniform distribution of labeling across the whole surface of terol) bound to P116, some of which are essential, matching observed the M. pneumoniae cells (Fig. 1a), and adhesion and motility were unaf- densities in the hydrophobic cavity. On the basis of these findings, fected by the antibodies (Supplementary Table 1 and Supplementary Nature Structural & Molecular Biology | Volume 30 | March 2023 | 321–329 322 Article https://doi.org/10.1038/s41594-023-00922-y Table 1 | CryoEM data collection, refinement and validation statistics P116 P116 empty P116 refilled P116 + HDL (EMD-15274) (EMD-15275) (EMD-15276) (EMD-15277) Data collection and processing Microscope FEI Titan Krios FEI Titan Krios FEI Titan Krios FEI Titan Krios Detector Gatan K2 Summit Gatan K3 Summit Gatan K2 Summit Gatan K2 Summit Acquisition Software SerialEM 3.8 EPU 2.12 SerialEM 3.8 SerialEM 3.8 Magnification ×130,000 ×105,000 ×130,000 ×130,000 Voltage (kV) 300 300 300 300 Electron exposure (e−/Å2) 50 50 50 50 Defocus range (µm) −1 to −3.5 −1 to −3.5 −1 to −3.5 −1 to −3.5 Pixel size (Å) 1.05 0.831 1.05 1.05 Symmetry imposed C1 C1 C1 C1 Initial particle images 3,463,490 4,532,601 2,930,863 262,981 Final particle images 1,315,362 633,322 1,311,526 46,277 Map Resolution (Å) 3.3 4 3.5 8.6 FSC threshold 0.143 0.143 0.143 0.143 Map resolution range (Å) 2.5–5 3.5–6 3–7 7–12 Number of frames 34 50 29 30 Micrographs used 4,367 15,299 4,019 3,114 Processing software cryoSPARC v3.3.2 cryoSPARC v3.3.2 cryoSPARC v3.3.2 cryoSPARC v3.3.2 Motion correction cryoSPARC v3.3.2 cryoSPARC v3.3.2 cryoSPARC v3.3.2 cryoSPARC v3.3.2 CTF estimation cryoSPARC v3.3.2 cryoSPARC v3.3.2 cryoSPARC v3.3.2 cryoSPARC v3.3.2 Particle images after 2D classification 1,324,330 1,140,275 1,311,526 46,277 Map sharpening B factor −117 −105 −131 −490 P116 P116 empty Refinement (PDB 8A9A) (PDB 8A9B) Initial model used (PDB code) / 8A9A Model resolution (Å) 3.3 4 FSC threshold 0.143 0.143 Model resolution range (Å) 2.5–5.0 3.5–6.0 Map sharpening B factor (Å2) −94 −105 Model composition Non-hydrogen atoms 12,772 6,386 Protein residues 1,618 809 Ligands 0 0 B factors (Å2) Protein 51.22 35.30 Ligand 0 R.m.s. deviations Bond lengths (Å) 0.003 0.027 Bond angles (°) 0.651 1.197 Validation MolProbity score 2.18 2.52 Clashscore 12 20 Poor rotamers (%) 0.14 0.71 Ramachandran plot Favored (%) 89.34 84 Allowed (%) 10.53 15 Disallowed (%) 0.13 1 Nature Structural & Molecular Biology | Volume 30 | March 2023 | 321–329 323 Article https://doi.org/10.1038/s41594-023-00922-y a N-terminal b domain Bridge helix N C Finger 2 N-terminal domain Finger 1 Dimerization Finger 4 Dimerization Finger 3 interface W681 Protrusion Core domain c Protrusion Hydrophobic cavity N-terminal domain N-terminal domain Dimerization interface Top view Protrusion –28 0 28 Fig. 2 | P116 structure and hydrophobic areas. a, Ribbon model of the P116 thumb. The dimerization helices (shown in pink) correspond to the wrist. b, The monomer, built from the density shown in Figure 1b and colored as in Figure 1. overall topology of P116. The N-terminal and core domains of P116 share a similar The overall shape of the structure corresponds to a left hand, with the four topology, which suggests that P116 might have been generated by duplication of antiparallel amphipathic α-helices representing fingers (shown in blue), and the an ancestor domain. Colors correspond to a. c, The hydrophobic map of the P116 bridge helix and β-sheet of five antiparallel strands representing the palm (shown homodimer shows that the cavity in the core domain is hydrophobic (amino acid in orange). The N-terminal domain, which is very flexible, corresponds to the hydrophobicity is colored according to the Kyte–Doolittle scale). Movies 1–3). This distribution contrasts with that of P1, an adhesion 1:4, compared with only dimers in the wild type (Extended Data Fig. 3b). protein that concentrates at the tip of the cell and whose inhibition The palm of the hand includes a long and well-defined central α-helix, has strong effects on adhesion and motility16,17. namely the bridge helix (residues 268–304), and a rigid β-sheet of five antiparallel strands that extends to the N-terminal domain (Fig. 2b). P116 has a previously uncharacterized fold with a The hand appears in a half-opened state with a large, elongated cleft lipid-accessible cavity across the whole core domain (Fig. 2c). The core domain forms a large The structure of P116 (30–957) was determined by single-particle cry- cavity that measures 62 Å from the proximal to the distal end and 38 Å oEM at 3.3-Å resolution (according to the gold-standard criterion of from the anterior to the posterior side. The cavity has a volume of Fourier shell correlation (FSC) = 0.143; Table 1 and Extended Data Fig. 2). ~18,000 Å3. The cavity is completely hydrophobic but is fully acces- It is an elongated homodimer of ~240 Å along its longest axis, which sible to the solvent (Fig. 2c and Supplementary Movie 6). In addition, adopts an arched shape with an arc diameter of ~20 nm (Fig. 1b and the core has two access points, one at the dorsal side and another at Supplementary Movies 4 and 5). Each monomer consists of two distinct the distal side (Fig. 3a). Using the DALI server, we found only very weak domains: the core domain (residues 246–867) and the amino-terminal structural relationships between P116 and all other experimentally domain (residues 60–245), situated distal to the dimer axis. The dimeri- determined protein structures in the Protein Data Bank, which shows zation interface, part of the core domain and proximal to the dimer axis that P116 has a unique fold. (Fig. 1b and Extended Data Fig. 3a,b), is very well resolved. By contrast, The N-terminal domain is compact and organized around a cluster the N-terminal domain has substantial hinge mobility with respect to of aromatic residues, at the center of which is the only tryptophan the core domain, evident by the poorer local resolution of the cryoEM residue of the domain (W121). The N-terminal and core domains of map (Extended Data Fig. 2), making model building difficult for the P116 superimpose for 126 equivalent residues (68% of the N-terminal most distal parts of the construct (see Methods and Extended Data domain), suggesting that P116 might have been generated by duplica- Fig. 3c). The homodimer displays substantial flexibility with many tion of an ancestor domain. The common secondary structural ele- vibrational modes, as illustrated by a complete vibrational analysis, ments in the N-terminal and core domains consist of a β-sheet and the showing a fluent transition between states (Extended Data Fig. 4). two helices preceding the β-sheet (Fig. 2b). The core domain is much The core domain resembles a half-opened left hand, with four larger than the N-terminal domain, mainly owing to two insertions contiguous antiparallel pairs of amphipathic α-helices correspond- containing 12 and 4 helices. ing to the four fingers and the N-terminal domain corresponding to For the inner part of the P116 core domain, the cryoEM maps show the thumb (Fig. 2a). The dimer interface, which corresponds to the prominent elongated densities (with a length of 10–19 Å and a width of wrist, is composed of helices with a conserved tryptophan residue 4 Å) that fill most of the hydrophobic areas (Fig. 3a and Supplementary (Trp681) that interacts tightly with the neighboring monomer. In the Movies 7 and 8). These elongated densities, which are unaccounted for variant with the W681A mutation, the rate of dimers to monomers is by the structure, cannot be explained by the protein residues missing in Nature Structural & Molecular Biology | Volume 30 | March 2023 | 321–329 324 Article https://doi.org/10.1038/s41594-023-00922-y a Dorsal Bridge helix core access 90° N-terminal domain 2 N-terminal 1 3 domain Distal Top view core access Side view Protrusion Densities fill the hydrophobic cavity b c Overlay between full and empty P116 Empty Filled 8Å 13 Å 13 Å 10 Å 8Å 12 Å 12 Å N-terminal domain Side view Front view Protrusion d Original P116 Refilled P116 Top view Top view Hydrophobic cavities facing Hydrophobic cavity Hydrophobic cavity rotated ~80° the same direction with respect to the other Fig. 3 | Purified P116 is filled with ligands and displays a large conformational compared with the full P116 (shown in cyan). Finger 1 has moved 8 Å sideways variation compared to empty P116. a, A cross-section through the core domain and towards the palm, finger 2 has moved 13 Å towards the palm and finger 3 has of original P116 exposes a series of elongated densities (shown in red), which moved 12 Å towards the palm. The cavity in the empty P116 is no longer sufficient cannot be accounted for by the structure. These densities are ~4-Å wide and to accommodate ligands. c, In the ribbon presentation, the conformation 10- to 19-Å long and are surrounded by highly conserved hydrophobic residues. differences between the empty and full P116 structures can be seen in the front The cross-section also reveals that the core domain can be accessed dorsally view. All four fingers have moved towards the palm (shown in orange) of the and distally. The side view of the core domain shows that the densities are hand (individual distances are indicated filled conformation in cyan, empty aligned to the bridge helix and away from the fingers (shown in red). Numbers conformation in purple). d, Two cryoEM classes reveal a wringing movement indicate individual fingers (finger 4 is not visible in this illustration). b, Overlay of P116. Comparison of the two density maps (superimposed with the ribbon between empty and full P116. The side view of the cross-section surface view of diagram of the structure) shows that the wringing movement of P116 allows for the empty, and full P116 (palm areas aligned) shows that the fingers of the empty the two hydrophobic cavities in the dimer to face almost opposite directions. The P116 (in purple) have come closer to the palm, massively reducing the cavity. top view on the left shows both cavities facing in one direction, and the top view The position of the fingers in the empty P116 (in purple) is markedly different on the right shows the cavities rotated ~80° to each other. the model. Instead, the mass excess of ~13 kDa, consistently measured smaller N-terminal domain, but the relative position of the domains by multiple angle light scattering (MALS) and mass spectrometry for does not closely match the experimental structure (Extended Data P116 in different preparations, could be explained by the presence of Fig. 6b). ligand molecules bound to P116 (Fig. 4a). Mass spectrometry analysis of the same samples from which the structure of P116 was determined Empty P116 cannot accommodate lipids (see Methods) showed the presence of several lipid species, predomi- To obtain empty P116 that is free of any bound ligands, we treated nantly phosphatidylethanolamine (PE) and phosphatidylglycerol (PG) the P116 samples with the detergent Triton X-100 (see below and lipids, as well as wax esters (Fig. 4b–d and Extended Data Fig. 5). Methods). Mass spectrometry confirmed a massive reduction of lipids in the sample (Fig. 4b). The structure of the empty P116 sample was P116 orthologs in other Mycoplasma spp solved by cryoEM at 4-Å resolution (Extended Data Fig. 7). Its over- P116 orthologs were found in at least eight other Mycoplasma species, all topology is almost identical to that of the original P116 sample; including M. genitalium and M. gallisepticum. The amino acids lining however, the core domain is constricted as a result of fingers 1, 2 and the hydrophobic cavity are largely conserved (they are either identical 3 being closer to the palm by 8, 13 and 12 Å, respectively, and finger or have similar characteristics) (Extended Data Fig. 6a). Modeling the 4 moving 11 Å sideways to retain the distal core access to the cavity orthologs of P116 with AlphaFold18 resulted in all the models having a (Fig. 3b, Supplementary Movies 9 and 10 and Extended Data Fig. 8). similar tertiary structure, in which a large core domain is flanked by a These changes reduce the volume of the cavity from ∼18,000 Å3 to Nature Structural & Molecular Biology | Volume 30 | March 2023 | 321–329 325 Article https://doi.org/10.1038/s41594-023-00922-y ∼6,300 Å3. Consequently, the huge hydrophobic cavity reduces to cholesterol-containing HDL radiotracer was detected in the P116 two pockets that are large enough for lipids to pass through but that samples that had been incubated with HDL and then separated from appear unoccupied in the cryoEM density. A comparison of the filled it by purification (Methods and Extended Data Fig. 9), indicating a net and empty P116 structures shows that the original densities that were transfer of both cholesterol and cholesterol ester between HDL and unaccounted for create massive steric clashes in the constricted con- P116. The total absence of the most abundant HDL protein (APOA1), figuration, demonstrating that the cavity can no longer accommodate cross-checked by immune detection (Methods and Supplementary lipids (Supplementary Movie 11). In the empty P116, the dimeriza- Table 2), verified that no HDL remnants had contaminated the puri- tion interface is shifted towards the dorsal side of the molecule by fied P116 samples. 10 Å, resulting in a contraction that changes the arc diameter of the The highest rate of radiotracer transfer was achieved when [3H] dimer to ~10 nm and shifts the N-terminal domain towards the dimeri- cholesterol-containing HDLs were mixed with empty P116. [3H]cho- zation interface. lesterol was also transferred. Transfer of [3H]cholesterol esters to P116 would require a direct interaction between HDL and P116, as these Refilled P116 is identical to the purified P116 esters are buried in the core of the HDL particles (Table 2). Passive cho- We next refilled the empty P116 samples by incubating them either with lesterol transport from cellular membranes to HDL or from low-density fetal bovine serum (FBS) or with high-density lipoprotein (HDL) and lipoprotein to HDL has been reported20, but the concept that bacteria then re-purified them by affinity chromatography. Medium containing can actively extract cholesterol from HDL has not been previously FBS is a common growing broth for M. pneumoniae cultures, although characterized. lipoproteins, in particular HDL, are efficient substitutes for serum in We then conducted a detailed matrix-assisted laser desorption/ mycoplasma culture medium, likely because lipoproteins can provide ionization-time of flight (MALDI-TOF) and liquid chromatography the key lipids, in particular cholesterol, which are essential for myco- electrospray ionization coupled with tandem mass spectrometry plasma cells19. We solved the structure of the refilled P116 samples at (LC-ESI–MS/MS) analysis. We identified more than 500 lipid species in 3.5-Å resolution using cryoEM. The structure of the refilled monomer the samples and found striking differences between the original, empty P116 is practically identical to the structure of the original monomer and refilled P116 samples (Fig. 4b–d). In the original P116 sample, the P116 sample, including densities at the palm of the hand that can be predominant lipid species were PE, PG and wax esters. Wax esters are assigned to ligands. Mass spectrometry of the refilled samples shows not known to be required by M. pneumoniae, but they were part of the the clear presence of lipids (Fig. 4b). cultivation medium of the E. coli strain in which P116 was produced. Incorporation of many lipid species is in agreement with the fact that Conformational flexibility of P116 M. pneumoniae adapts its membrane composition to the available The structures of the original P116, empty P116 and refilled P116 samples lipid spectrum6–8. In the empty P116, we observed a substantial reduc- appear predominantly as homodimers. In all cases, the homodimer tion of lipids, with no specific lipid class enriched. In the P116 samples exhibits substantial flexibility. Most prominently, the arc diameter of refilled from FBS, we observed a clear accumulation of the essential the empty structure is approximately 10 nm smaller than that of the lipid classes phosphatidylcholine (PC) and sphingomyelin (SM), as original and refilled structures. In addition, a wringing motion is visible well as sterols and cholesterol (Fig. 4c,d and Supplementary Table 3). in the refilled structure: each monomer is twisted in the opposite direc- tion along the axis perpendicular to the dimer axis by ~80°, and bends P116 extracts specific lipid classes up to 20°, depending on its cargo (Fig. 3d, Supplementary Movie 12 and To analyze lipoprotein carryover in the FBS-refilled P116, we conducted Extended Data Fig. 4). In all P116 structures, the N-terminal domain is the an additional proteomics LC–MS/MS experiment (Supplementary most flexible. Within the core domain, temperature factors are higher at Table 4) using ultrasensitive, ion-mobility-assisted LC–MS/MS. In the fingertips, indicating movement of the antiparallel α-helices. When this experiment, we observed limited lipoprotein carryover into the the fingers approach the palm, this results in the core domain constrict- refilled sample. However, on the basis of peptide spectrum match ing and a clash with the densities therein (Supplementary Movie 11). (PSM) numbers and intensity values, we found P116 to be over 30-fold more abundant than the lipoproteins in the refilled sample. If the lipid P116 ligands include essential lipids spectrum in the FBS-refilled P116 sample originated from lipoprotein We next set out to characterize the possible ligands within P116. We carryover, we would expect a similar distribution of the lipid classes in first measured the rate of radioactivity transfer to P116 after incuba- both samples. In fact, we observed a specific enrichment of PC and SM tion with HDL particles containing either tritium-labeled cholesterol in the FBS-refilled sample, whereas TG, the most abundant lipid class ([3H]cholesterol) or tritium-labeled cholesteryl oleate as a representa- in the serum, was decreased and was barely detectable. Thus, although tive of cholesterol esters (Table 2). A substantial fraction of the [3H] P116 can extract a large range of lipids, it shows a preference for selected Fig. 4 | Analysis of the lipid spectrum and uptake of P116. a, MALDI-TOF mass A fatty acid ester; SE, sterol esters; PI, phosphatidylinositol; NAE, N-acyl spectrum of original P116 sample (linear mode, high mass range), showing a ethanolamines; LPE, lysophosphatidylethanolamine; LDGTS, lysodiacylglyceryl dominant peak at 105 kDa, corresponding to the singly charged full protein, as trimethylhomoserine; DGGA, diacylglyceryl glucuronide; CE, cholesteryl ester; well as the charged states two, three and four. a.u., arbitrary units. b, Stacked BMP, bismonoacylglycerophosphate; NAE, N-acyl ethanolamines; MGDG, MALDI-TOF mass spectra (reflector mode, low mass range) of the original monogalactosyldiacylglycerol; HBMP, hemibismonoacylglycerophosphate; purified P116 (purple, rear), empty P116 (black, middle) and refilled P116 DGTS, diacylglyceryl trimethylhomoserine. e, CryoEM analysis of empty P116 sample (orange, front) showing a change in the lipid distribution among the incubated with HDL shows that P116 binds HDLs between its N-terminal and samples. c,d, Hierarchical clustering of lipid compounds identified in positive core domains. P116 is attached to HDL through its distal core. Owing to the (c) and negative (d) ion mode lipidomics (LC–MS/MS) analyses (reproduced flexibility of P116 and the variability of HDL, only one subunit of P116 can be seen in three independent experiments), showing differential distributions of lipid at this threshold. The whole P116 can be seen in the individual class averages. compositions in original P116 (first column), empty P116 (second column), f, Schematic of the lipid uptake and conformational variations of P116 (here refilled P116 (third column) and serum (fourth column). The refilled P116 indicated by its structure anchored in the mycoplasma membrane. Linkers and shows a particular affinity to sterols and cholesterol specifically. All data were transmembrane domains not seen in the cryoEM structure are shown in purple). normalized to the mTIC of all identified compounds in each sample, and row-wise P116 starts in an empty, constricted state; incubation with HDL leads to each scaling was applied. PE, phosphatidylethanolamine; PG, phosphatidylglycerol; individual monomer filling up with approximately 20 lipids; and P116 changes to DG, diacylglycerol; PC, phosphatidylcholine; SM, sphingomyelin; TG, the open/filled state. We hypothesize that, through a wringing motion, lipids are triacylglycerol; FA, fatty acid; LPC, lysophosphatidylcholine; VAE, vitamin delivered into the mycoplasma membrane. Nature Structural & Molecular Biology | Volume 30 | March 2023 | 321–329 326 Article https://doi.org/10.1038/s41594-023-00922-y lipid species (Fig. 4c,d and Supplementary Table 3). We conclude that interacting directly with HDL. The structure can be well fitted to the den- the lipid composition in the FBS-refilled P116 sample can be attributed sity map. Interestingly, the P116 region between the N-terminal domain predominantly to P116 itself and not to lipoprotein carryover. and the core interacts with HDL (Fig. 4f). Cryo-electron tomograms of whole M. pneumoniae cells indicate that this region faces away from the P116 binds at defined regions to HDL M. pneumoniae membrane and is thus accessible to vesicles and lipids. Next, we performed cryoEM on a sample containing empty P116 and This presents a possible explanation as to how P116 avoids extracting HDL. Of ~58,000 particles that were identified as HDL, ~25,000 were lipids from the M. pneumoniae membrane itself. However, the unam- attached to P116. The resulting density at a resolution of 9 Å shows P116 biguous identification of P116 on the M. pneumoniae membrane is a b MALDI spectrum of P116 MALDI spectra of low-molecular-weight compounds 4 [P116+H+] 104,978 Wax ester + [P116+2H ] 52,683 Intensity (a.u.) (× 104) 3 Intensity (a.u.) (× 105) 20,786 5 99,700 4 2 50,044 3 35,189 1 2 P116 1 P116 empty 0 P116 refilled 0 20,000 40,000 60,000 80,000 100,000 m/z 400 500 600 700 800 m/z Phospholipids c Positive ion mode lipidomics d Negative ion mode lipidomics P116 purified P116 empty P116 refilled Serum Lipid class P116 purified P116 empty P116 refilled Serum Lipid class Signal Sterols 1.5 Signal 1.5 TGs PEs, PGs 0 0 PCs, SMs PEs, PGs –1.5 –1.5 FAs Lipid class Lipid class Mixed lipids FAs VAE PG TG SE PE PI PC PG PE PCs, LPCs NAE PEs, PGs PC MGDG NAE LPC LPE LPC HBMP LDGTS FA DGGA DG DGTS DGs CE PCs, LPCs BMP e Characteristic class-averages f Schematic of lipid uptake Attachment to HDL, HDL particles alone membranes, et cetera HDL Constricted/ empty state 20 Å HDL a-HDL particles in complex with P116 HDL Acquisition of ~20 lipids P116 20 Å P116 3D reconstruction Mycoplasma cell P116 HDL Wringing/ delivery Open/ of lipids ? filled state 10 nm Nature Structural & Molecular Biology | Volume 30 | March 2023 | 321–329 327 Article https://doi.org/10.1038/s41594-023-00922-y Table 2 | Relative transfer of cholesterol from HDL to P116 species such as B. burgdorferi, may have similar, as yet undiscovered systems to regulate their cholesterol homeostasis. Whether P116 shares % of [3H] nmol, nmol functional similarities with other transfer proteins such as human cho- cholesterol cholesterol cholesterol transferred/mL transferred/ transferred/mg lesteryl ester transfer and phospholipid transfer proteins23,24 requires mL/hour P116a further investigation. However, the diversity and amount of lipids that P116 can bind appear to be unmatched by any other known prokaryotic HDL to empty P116 or eukaryotic lipid carrier. Interestingly, despite its broad lipid range, Free cholesterol 13.12 13.52 59.49 (6.3) P116 still shows a high specificity, largely enriching certain lipids (SM, Esterified cholesterol 6.98 7.22 31.75 (3.3) PC and cholesterol) while excluding others (TGs). This understanding HDL to original P116 of bacterial lipid uptake presents potential opportunities for treat- ment of mycoplasma infections and may for the first time2 enable the Free cholesterol 7.89 7.42 32.63 (3.4) development of a vaccine against M. pneumoniae. Esterified cholesterol 6.32 6.01 26.44 (2.8) a Numbers in parentheses are the estimated number of cholesterol molecules transferred per Online content P116 subunit (assuming a molecular weight of ~105 kDa for the construct). Any methods, additional references, Nature Portfolio reporting sum- maries, source data, extended data, supplementary information, challenging owing to the low copy number of P116 (ref. 14), and further acknowledgements, peer review information; details of author contri- experiments are required to better characterize the attachment of P116 butions and competing interests; and statements of data and code avail- with the M. pneumoniae membrane (Extended Data Fig. 10). ability are available at https://doi.org/10.1038/s41594-023-00922-y. Discussion References P116 is essential for the viability of the human pathogen M. pneumoniae4 1. Tsiodras, S., Kelesidis, I., Kelesidis, T., Stamboulis, E. & and is the target of a strong antigenic response in infected people21. Giamarellou, H. Central nervous system manifestations The P116 structure has a previously unseen fold with a uniquely large of Mycoplasma pneumoniae infections. J. Infect. 51, hydrophobic cavity filled with ligands. Mass spectrometry and radio- 343–354 (2005). activity transfer experiments confirm a lipid extraction from serum 2. Jiang, Z., Li, S., Zhu, C., Zhou, R. & Leung, P. H. M. Mycoplasma (FBS) and HDL. Further, the ligands have been identified as essential pneumoniae infections: pathogenesis and vaccine development. lipids for the survival of the cells. In fact, we found a high specificity Pathogens 10, 119 (2021). towards cholesterol, PC and SM, which are the most abundant mem- 3. Razin, S., Yogev, D. & Naot, Y. Molecular biology and brane lipids in M. pneumoniae8. Crosslinking mass spectrometry pathogenicity of mycoplasmas. Microbiol. Mol. Biol. Rev. 62, studies indicate one weak amino acid-pair interaction between P116 1094–1156 (1998). and MPN161 (a protein of unknown function)22. Thus, although the 4. Lluch-Senar, M. et al. Defining a minimal cell: essentiality of small involvement of other proteins in incorporating the extracted lipids ORFs and ncRNAs in a genome-reduced bacterium. Mol. Syst. into the Mycoplasma membrane cannot be excluded, it appears likely, Biol. 11, 780 (2015). given the observed conformational states upon lipid extraction, that 5. Gaspari, E. et al. Model-driven design allows growth of P116 is also responsible for incorporation, thus P116 is responsible for Mycoplasma pneumoniae on serum-free media. NPJ Syst. Biol. the complete uptake (Fig. 4f). Taken together, the P116 structure and Appl. 6, 33 (2020). our insights into different P116 conformations and the P116 complex 6. Razin, S., Argaman, M. & Avigan, J. Chemical composition formation with HDL reveal a mechanism by which Mycoplasma species of mycoplasma cells and membranes. J. Gen. Microbiol. 33, extract lipids from the environment and most likely incorporate them 477–487 (1963). into their own membrane. 7. Leon, O. & Panos, C. Long-chain fatty acid perturbations in The transition from a full to an empty P116 molecule involves a Mycoplasma pneumoniae. J. Bacteriol. 146, 1124–1134 (1981). ~70% volume reduction of the hydrophobic cavity in concert with a 8. Bittman, R. Mycoplasma membrane lipids. Chemical composition wringing motion of the core domains. During this wringing motion, in and transbilayer distribution. Subcell. Biochem. 20, 29–52 (1993). which the monomers are each twisted in the opposite direction around 9. Mohn, W. W. et al. The actinobacterial mce4 locus encodes a their long axis, the hydrophobic cavities face almost opposite direc- steroid transporter. J. Biol. Chem. 283, 35368–35374 (2008). tions. Because the N-terminal domain is near the C terminus, which 10. Pandey, A. K. & Sassetti, C. M. Mycobacterial persistence requires anchors the protein in the Mycoplasma membrane in vivo, the core is the utilization of host cholesterol. Proc. Natl Acad. Sci. USA 105, the domain that experiences the high flexibility seen in our data sets. 4376–4380 (2008). This flexibility enables an alternating wringing motion whereby one 11. Damy, S. B. et al. Mycoplasma pneumoniae and/or Chlamydophila monomer of the core domain faces the Mycoplasma membrane (that pneumoniae inoculation causing different aggravations in is, the monomer transferring lipids to the membrane) and the other cholesterol-induced atherosclerosis in apoE KO male mice. BMC monomer faces the environment (that is, the monomer extracting Microbiol. 9, 194 (2009). lipids from the environment). This wringing motion can be repeated 12. Huang, Z. & London, E. Cholesterol lipids and in a continuous manner. In this way, P116 could undergo a rolling move- cholesterol-containing lipid rafts in bacteria. Chem. Phys. Lipids ment on the Mycoplasma membrane, thus facilitating the transport of 199, 11–16 (2016). cholesterol and other essential lipids in an apparently simple way for 13. Svenstrup, H. F., Nielsen, P. K., Drasbek, M., Birkelund, S. & lipid transporters (Fig. 4f). Christiansen, G. Adhesion and inhibition assay of Mycoplasma Mycoplasma species have a minimal genome and are capable genitalium and M. pneumoniae by immunofluorescence of incorporating many different lipids into their membranes6,7. The microscopy. J. Med. Microbiol. 51, 361–373 (2002). lipid-binding versatility shown by P116 enables a single molecu- 14. Yus, E. et al. Impact of genome reduction on bacterial metabolism lar system to cope with the transport of diverse lipids required by and its regulation. Science 326, 1263–1268 (2009). Mycoplasma. Although only Mycoplasma shares genes with sequences 15. McGowin, C. L. & Totten, P. A. The unique microbiology and similar to that of p116, other microorganisms that require uptake of molecular pathogenesis of Mycoplasma genitalium. J. Infect. Dis. lipids from the environment, including clinically relevant bacterial 216, S382–S388 (2017). Nature Structural & Molecular Biology | Volume 30 | March 2023 | 321–329 328 Article https://doi.org/10.1038/s41594-023-00922-y 16. Krause, D. C., Leith, D. K., Wilson, R. M. & Baseman, J. B. 24. Zhang, M. et al. Structural basis of the lipid transfer mechanism of Identification of Mycoplasma pneumoniae proteins associated with phospholipid transfer protein (PLTP). Biochim. Biophys. Acta 1863, hemadsorption and virulence. Infect. Immun. 35, 809–817 (1982). 1082–1094 (2018). 17. Baseman, J. B., Cole, R. M., Krause, D. C. & Leith, D. K. Molecular basis for cytadsorption of Mycoplasma pneumoniae. J. Bacteriol. Publisher’s note Springer Nature remains neutral with regard to 151, 1514–1522 (1982). jurisdictional claims in published maps and institutional affiliations. 18. Jumper, J. et al. Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589 (2021). Open Access This article is licensed under a Creative Commons 19. Washburn, L. R. & Somerson, N. L. Lipoproteins as substitutes Attribution 4.0 International License, which permits use, sharing, for serum in Mycoplasma culture medium. J. Clin. Microbiol. 10, adaptation, distribution and reproduction in any medium or format, 586–589 (1979). as long as you give appropriate credit to the original author(s) and the 20. Phillips, M. C. Molecular mechanisms of cellular cholesterol source, provide a link to the Creative Commons license, and indicate efflux. J. Biol. Chem. 289, 24020–24029 (2014). if changes were made. The images or other third party material in this 21. Drasbek, M., Nielsen, P. K., Persson, K., Birkelund, S. & article are included in the article’s Creative Commons license, unless Christiansen, G. Immune response to Mycoplasma pneumoniae P1 indicated otherwise in a credit line to the material. If material is not and P116 in patients with atypical pneumonia analyzed by ELISA. included in the article’s Creative Commons license and your intended BMC Microbiol. 4, 7 (2004). use is not permitted by statutory regulation or exceeds the permitted 22. O’Reilly, F. J. et al. In-cell architecture of an actively transcribing- use, you will need to obtain permission directly from the copyright translating expressome. Science 369, 554–557 (2020). holder. To view a copy of this license, visit http://creativecommons. 23. Zhang, L. et al. Structural basis of transfer between lipoproteins org/licenses/by/4.0/. by cholesteryl ester transfer protein. Nat. Chem. Biol. 8, 342–349 (2012). © The Author(s) 2023 Nature Structural & Molecular Biology | Volume 30 | March 2023 | 321–329 329 Article https://doi.org/10.1038/s41594-023-00922-y Methods radiotracers, as appropriate, and then incubated for 16 hours in a The research complies with all relevant ethical regulations. The experi- 37 °C bath26. The labeled HDLs (both 3H-cholesterol-containing and 3 mental procedures to immunize mice and obtain monoclonal antibod- H-cholesteryl oleate-containing HDLs) were re-isolated by gradient ies were approved by the Ethics Committee on Animal and Human density ultracentrifugation at 1.063–1.210 g/mL and dialyzed against Experimentation from the Universitat Autonòma de Barcelona under PBS through gel filtration chromatography. Specific activities of 3 the document CEEAH 1002R3R2R. H-cholesterol-containing and 3H-cholesteryl oleate-containing HDLs were 1,221 and 185 counts per minute (cpm)/nmol, respectively. The Bacterial strains, tissue cultures and growth conditions cholesterol transfer to P116 (1 g/L) was measured after adding either The M. pneumoniae M129 strain was grown in cell culture flasks contain- [3H] free cholesterol-containing or [3H]cholesteryl oleate-containing ing SP4 medium and incubated at 37 °C and 5% CO2. Surface-attached HDL (0.5 g/L of APOA1) and incubation for 2 hours at 37 °C. HDL and mycoplasmas were collected using a cell scraper and resuspended in P116 were separated by a HisTrap HP affinity and size-exclusion columns SP4 medium. To grow mycoplasma cells on IBIDI eight-well chamber (Extended Data Fig. 10). The radioactivity associated with each P116 slides, each well was seeded with about 1 × 105 colony-forming units and and HDL fraction was measured through liquid scintillation counting. incubated for 12–24 hours in 200 μL SP4 supplemented with 3% gelatin. The percentage of [3H]cholesterol transferred per mL was determined NSI myeloma cells25 were grown in RPMI 1640 medium supple- for each condition. The specific activities for each radiotracer were mented with 10% fetal bovine serum (FBS) and 50 μg mL−1 gentamicin used to calculate the amount of free cholesterol and cholesteryl ester (complete RPMI). Hybridomas were selected in complete RPMI sup- transferred from HDL to P116. Total cholesterol levels in the HDL frac- plemented with HAT medium and BM-Condimed (Sigma Aldrich). tion were determined enzymatically by using a commercial kit adapted for a COBAS 6000 autoanalyzer (ref 03039773190, Roche Diagnostics). Cloning, expression and purification of P116 constructs Human APOA1 levels were determined in both the HDL and purified Regions corresponding to the mpn_213 gene from M. pneumoniae P116 fractions by an assay (ref 3032566122, Roche Diagnostics) that were amplified from synthetic clones using different primers for each used anti-APOA1 antibodies that react with the antigen in the sample construct: P116F30 and P116R957 for P116(30–957); P116F13 and P116R957 to form antigen–antibody complexes, which, after agglutination, for P116(13–957); P116F212 and P116R862 for P116(212–862); and P116W681 were measured turbidimetrically in the COBAS 6000 autoanalyzer to generate variant P116 W681A. PCR fragments were cloned into the (Supplementary Table 2). expression vector pOPINE (gift from R. Owens; plasmid no. 26043, Addgene) to generate constructs, with a carboxy-terminal His-tag. Size-exclusion chromatography and multi-angle light Recombinant proteins were obtained after expression at 22 °C in B834 scattering (DE3) cells (Merck), upon induction with 0.6 mM IPTG at an optical Molecular weights were measured from P116 samples using a Superose density at 600 nm (OD600) of 0.8. Cells were collected and lysed by 6 10/300 GL (GE Healthcare) column in a Prominence liquid chromatog- French press in binding buffer (20 mM TRIS-HCl pH: 7.4, 40 mM imi- raphy system (Shimadzu) connected to a DAWN HELEOS II multi-angle dazole and 150 mM NaCl) and centrifuged at 49,000g at 4 °C. Superna- light scattering (MALS) detector and an Optilab T-REX refractive index tant was loaded onto a HisTrap 5 ml column (GE Healthcare) that was (dRI) detector (Wyatt Technology). ASTRA 7 software (Wyatt Technol- pre-equilibrated in binding buffer and elution buffer (20 mM TRIS-HCl ogy) was used for data processing and analysis. An increment of the pH: 7.4, 400 mM imidazole and 150 mM NaCl). Soluble aliquots were specific refractive index in relation to concentration changes (dn/dc) pooled and loaded onto a Superdex 200 GL 10/300 column (GE Health- of 0.185 mL/g (typical of proteins) was assumed for calculations. care) in a protein buffer (20 mM TRIS-HCl pH 7.4 and 150 mM NaCl). To obtain empty P116, 2.6% Triton X-100 was added to the protein Matrix-assisted laser desorption/ionization-mass sample and incubated for 1.5 hours at room temperature. Subsequent spectrometry purification followed the same methodology described above, but also All samples were mixed in a 1:1 ratio with sDHB (Super-DHB, Bruker) included a wash step with the binding buffer plus 1.3% Triton X-100, matrix solution (50 mg mL in 50% acetonitrile (ACN), 50% water, and followed by extensive washing with at least 20 column volumes of 0.1% trifluoroacetic acid). Subsequently, 1-μL aliquots of the mixture wash buffer (20 mM TRIS-HCl pH: 7.4, 20 mM imidazole) before elut- were deposited on a BigAnchor MALDI target (Bruker) and allowed to ing the samples from the column. P116 was concentrated with Vivaspin dry and crystallize at ambient conditions. 500 centrifugal concentrators (10,000 MWCO PES, Sartorius) to a final MS spectra were acquired on a rapifleX MALDI-TOF/TOF (Bruker, concentration of >0.5 mg/mL. Germany) in the mass range of 20,000–120,000 m/z in linear posi- To refill P116 with lipids, the empty protein was incubated with tive mode for intact protein measurements and in the mass range of approximately 1 mL FBS per mg P116 for 2 hours at 30 °C while still 100–1,600 m/z in reflector positive mode for lipid measurements. The bound on the column. After extensive washing with at least 40 column Compass 2.0 (Bruker) software suite was used for spectra acquisition volumes of wash buffer, elution and concentration were performed as and processing. described above. Lipidomics analysis (LC–MS/MS) HDL isolation and determination of cholesterol transfer rate Lipid samples, with an equivalent of 10 µg of protein, were extracted Human HDL (density 1.063–1.210 g/mL) was isolated from plasma of using a modified MTBE/Methanol extraction protocol27 and submitted healthy donors through sequential gradient density ultracentrifuga- to LC–MS/MS analysis using a nanoElute (Brukrt) system, equipped tion, using potassium bromide for density adjustment, at 100,000g for with C18 analytical column (15 cm × 75 µm, particle size: 1.9 µm 24 hours with an analytical fixed-angle rotor (50.3, Beckman Coulter). (PepSep)), coupled to a timsTOF Pro 2 mass spectrometer (Bruker). The amount of cholesterol and apolipoprotein A1 were determined Samples were loaded directly onto the analytical column with enzymatically and by an immunoturbidimetric assay, respectively, twice the sample pick-up volume with buffer A. Lipids were separated using commercial kits adapted for a COBAS 6000 autoanalyzer (Roche on the analytical column at 60 °C with a flow rate of 400 nL/minute, Diagnostics, Rotkreuz, Switzerland). Radiolabeled HDLs were pre- with the following gradient: 1% B for 1 minute, 1 to 30% B in 2 minute, pared in the following way: 10 μCi of either [1,2-3H(N)] free choles- 30 to 51% B in 4 minutes, 51 to 61% B in 5 minutes, 61 to 70% B in 5 min- terol or [1,2-3H(N)]cholesteryl oleate (Perkin Elmer) were mixed with utes, 70 to 99% B in 5 minutes and constant 99% B for 13 minutes. This absolute ethanol, and the solvent was dried under a stream of N2. HDL was followed by column re-equilibration with buffer A (ACN/water (0.5 mL, 2.25 g/L of ApoA1) was added to the tubes containing the (60/40, vol/vol) with 10 mM ammonium formate and 0.1% FA) and Nature Structural & Molecular Biology Article https://doi.org/10.1038/s41594-023-00922-y buffer B (2-propanol/ACN (90/10, vol/vol) with 10 mM ammonium equipped with a GIF Quantum S.E. post-column energy filter in zero loss formate and 0.1% FA). peak mode and a K2 Summit detector (Gatan). For P116, P116 refilled Lipids eluting from the column were ionized online using a cap- and P116 with HDL, a total of 4,376, 4,019 and 3,114 micrographs with tive spray ion source and were analyzed in two replicates for positive 34, 29 and 30 frames per micrograph and a frame time of 0.2 sec- and negative mode using DDA-PASEF with a ramp time of 100 ms and onds were collected. The camera was operated in dose-fractionation 3 PASEF-MS/MS events. Spectra were acquired over the mass range from counting mode with a dose rate of ~8 electrons per Å2 s−1, resulting in 50–1,550 m/z and a mobility window from 0.55–1.95 Vs/cm2. a total dose of 50 electrons per Å2 s−1. Defocus values ranged from –1 Raw data were converted into ibf files and analyzed using the to –3.5 µm. MS-DIAL lipidomics pipeline (version 4.9 (ref. 28)) with default pro- For P116 empty, dose-fractionated movies were collected using cessing parameters for timsTOF data. Identified lipids were aligned EPU 2.12 (Thermo Scientific) at a nominal magnification of ×105,000 to a pooled control sample and filtered by blank abundances (sample (0.831 Å per pixel) in nanoprobe EFTEM mode at 300 kV with a Titan intensity/blank intensity >5). Intensities were normalized by mTIC, Krios G2 electron microscope (Thermo Scientific), equipped with a exported and further analyzed in R using the lipidr package29. BioQuantum-K3 imaging filter (Gatan), operated in zero loss peak mode with 20 eV energy slit width. In total,15,299 micrographs with Proteomics analysis (LC–MS/MS) 50 frames per micrograph and frame time of 0.052 seconds were col- Protein samples were reduced with TCEP and cysteines alkylated with lected. The K3 camera was operated in counting mode with a dose rate IAA (Thermo Fisher). Subsequent proteolytic digests were performed of ~16 electrons per A2 s−1, resulting in a total dose of 50 electrons per using S-TRAPs (Protifi), according to the manufacturer’s instructions. Å2 s−1. Defocus values ranged from −0.8 to −3.5 µm. Peptides were further desalted and purified on Isolute C18 SPE car- CryoSPARC v3.2 (ref. 32) was used to process the cryoEM data, tridges (Biotage, Sweden) and dried in an Eppendorf concentrator unless stated otherwise. Beam-induced motion correction and CTF (Eppendorf). estimation were performed using CryoSPARC’s own implementation. After solubilization in 0.1% formic acid (FA) in ACN/water Particles were initially clicked with the Blob picker using a particle (95/5, vol/vol), samples were subjected to LC–MS/MS analysis on a diameter of 200–300 Å. Particles were then subjected to unsuper- nanoElute (Bruker) system, equipped with C18 analytical column vised 2D classification. For the final processing, the generated 2D (15 cm × 75 µm, particle size: 1.9 µm (PepSep)) coupled to a timsTOF averages were taken as templates for the automated particle picking; Pro 2 mass spectrometer (Bruker). for the processing of P116 with HDL, no template picking was per- Samples were loaded directly onto the analytical column with formed. In total, 3,463,490, 4,532,601 particles, 2,930,863 particles twice the sample pick-up volume with buffer A. Peptides were separated and 262,981 particles were picked and extracted with a binned box on the analytical column at 60 °C with a flow rate of 500 nL/minute, size of 256 pixels for P116, P116 empty, P116 refilled and P116 with HDL, with the following gradient: 2 to 35% B in 17.8 minutes, 35 to 95% B in respectively. False-positive picks were removed by two rounds of unsu- 0.5 minutes and constant 90% B for 2.4 minutes with buffer A (0.1% FA pervised two-dimensional classification. The remaining 1,324,330 par- in water) and buffer B (0.1% FA in acetonitrile). ticles (P116), 1,140,275 particles (P116 empty), 1,311,526 particles (P116 Peptides eluting from the column were ionized online using a cap- refilled) and 46,277 particles (P116 with HDL) were used to generate an tive spray ion-source and analyzed in DDA-PASEF mode with a cycle time ab initio reconstruction with three classes followed by a subsequent of 100 ms and 4 PASEF-MS/MS events. Spectra were acquired over the heterogeneous refinement with three classes. For the final processing, mass range of 100–1,700 m/z and a mobility window of 0.85–1.3 Vs/cm2 1,315,362 particles (P116), 633,332 particles (P116 empty), 1,311,526 par- Data analysis was performed in FragPipe 18 using MSFragger 3.5 ticles (P116 refilled) and 46,277 particles (P116 with HDL) were used. For for database searches30. Raw files were recalibrated, search param- the remaining particles, the beam-induced specimen movement was eters automatically optimized and searched against the combined corrected locally. Uniprot reference proteomes for M. pneumoniae, E. coli and Bos taurus The CTF was refined per group on the fly within the non-uniform (UP000000808, UP000000625, UP000162055; obtained 2022-06-23). refinement. The obtained global resolution of the homodimer was 3.3 Å The database search space was restricted to tryptic peptides with (P116), 4 Å (P116 empty), and 3.5 Å (P116 refilled) (Extended Data Figs. 2 a length of 7–50 amino acids, allowing for up to two missed cleavages and 8 and Table 1). To analyze the sample with regard to its flexibility, and with a minimum of one unique peptide per protein group. Car- the particles were subjected to the 3D variability analysis of cryoSPARC, bamidomethylation of cysteine was set as a fixed modification and which was used to display the continuous movements of the protein. oxidation of methionine, as well as N-terminal acetylation, were set as variable modifications. Percolator was used to estimate the number of Cryo-electron tomography of M. pneumoniae false positive identifications, and the results were filtered for a strict M. pneumoniae M129 cells of an adherently growing culture were target false discovery rate (FDR) < 0.01. scraped in a final volume of 1 mL of SP4 medium and washed three times in PBS. This solution was mixed with fiducial markers (Protein A Single-particle cryoEM conjugated to 5 nm colloidal gold: Cell biology department, University For single-particle cryoEM, a 3.5-µL drop of purified P116 (100–400 µg/mL Medical Center Utrecht). From this stock, a 3.5-µL drop was applied to in 20 mM Tris, pH 7.4 buffer or 600 µg/mL in 20 mM Tris, 2 mM CHAPSO, a (45 s) glow-discharged R1.2/1.3 C-flat grid (Electron Microscopy Sci- pH 7.4 buffer) or P116 mixed with HDL (250 µg/mL P116 and 1116 µg/mL ence), and plunge-frozen in liquid ethane (Vitrobot Mark IV, Thermo HDL in 20 mM Tris, pH 7.4 buffer) was applied to a 45 s glow-discharged Scientific) at 100% relative humidity, 4 °C, and a nominal blot force of R1.2/1.3 C-flat grid (Electron Microscopy Science), and plunge-frozen –1, with a wait and blotting time of 10 seconds. in liquid ethane (Vitrobot Mark IV, Thermo Scientific) at 100% relative Tilt-series were recorded using SerialEM v3.8 (ref. 31) at a nominal humidity, 4 °C, a nominal blot force of –3, and a wait time of 0.5 seconds, magnification of ×105,000 (1.3 Å per pixel) in nanoprobe EFTEM mode with a blotting time of 12 s. Before freezing, Whatman 595 filter papers at 300 kV with a Titan Krios (Thermo Scientific) electron microscope were incubated for 1 hour in the Vitrobot chamber at 100% relative equipped with a GIF Quantum S.E. post-column energy filter in zero humidity and 4 °C. loss peak mode and a K2 Summit detector (Gatan). The total dose per Dose-fractionated movies of P116, P116 refilled and P116 mixed tomogram was 120 e−/Å2, and the tilt series covered an angular range with HDL were collected with SerialEM v3.8 (ref. 31) at a nominal mag- from −60° to 60° with an angular increment of 3° and a defocus set at nification of ×130,000 (1.05 Å per pixel) in nanoprobe EFTEM mode −3 µm. Tomograms were reconstructed by super-sampling SART33 with at 300 kV with a Titan Krios (Thermo Scientific) electron microscope a 3D CTF correction34. Nature Structural & Molecular Biology Article https://doi.org/10.1038/s41594-023-00922-y P116 model building and refinement examination. Cells were observed by phase contrast and epifluores- The initial tracing of the core domain was performed manually with cence in an Eclipse TE 2000-E inverted microscope (Nikon). Phase Coot35. It contained numerous gaps and ambiguities that were slowly contrast images, 4′,6-diamidino-2-phenylindole (DAPI, excitation polished by alternating cycles of refinement using the ‘Real Space’ 387/11 nm, emission 447/60 nm) and Texas Red (excitation 560/20 nm, protocol in the program Phenix36,37 and manual reinterpretation and emission 593/40 nm) epifluorescence images were captured with an rebuilding with Coot. The tracing and assignment of specific residues Orca Fusion camera (Hamamatsu) controlled by NIS-Elements BR in the N-terminal domain were very difficult owing to the low local software (Nikon). resolution of the map for this domain, and only a partial interpreta- tion was achieved. Using Robetta and AlphaFold18, we obtained dif- Time-lapse microcinematography ferent predictions of the N-terminal domain structure using different The effect of anti-P116 antibodies and anti-P1 polyclonal serum on parts of the sequence. The highest ranked predictions, selected using mycoplasma cell adhesion was investigated by time-lapse cinematogra- the partial experimental structure already available, were obtained phy of M. pneumoniae cells growing on IBIDI eight-well chamber slides. with AlphaFold for residues 81–245, which allowed us to complete the Before observation, medium was replaced with PBS containing 10% FBS building of the N-terminal domain according to the cryoEM map. The and 3% gelatin prewarmed at 37 °C. A similar medium has been used to root-mean-square deviation between the AlphaFold prediction and test the effect of P1 antibodies on mycoplasma adhesion and gliding the experimental model was 2.6 Å for 104 (63%) structurally equiva- motility41. After incubation for 10 minutes at 37 °C and 5% CO2, the slide lent residues. Some residues at the N end of the N-terminal domain was placed in a Nikon Eclipse TE 2000-E inverted microscope equipped were difficult to identify and were represented as alanines in the final with a Microscope Cage Incubation System (Okolab) at 37 °C. Images model. The whole P116 model was then refined using Phenix, and were captured at 0.5-second intervals, for a total observation time of the final refined structure was deposited in the EMDB under code 10 minutes. After the first 60 seconds of observation, the different EMD-15274 (Table 1). antibodies were dispensed directly into the wells. The frequencies of motile cells and detached cells before the addition of antibodies were Polyclonal and monoclonal antibody generation calculated from the images collected between 0 and 60 seconds of Two BALB/C mice were serially immunized with four intraperitoneal observation. The frequencies of motile cells and detached cells after injections, each containing 150 μg of recombinant P116 ectodomain the addition of antibodies were calculated from the images collected (residues 30–957) in 200 μL of PBS with no adjuvants. The last injection in the last minute of observation. was delivered 4 days before splenectomy. Isolated B lymphocytes from the immunized mice were fused to NSI myeloma cells25 to obtain stable Reporting summary hybridoma cell lines producing monoclonal antibodies38. Supernatants Further information on research design is available in the Nature Port- from hybridoma cell lines derived from single fused cells were first folio Reporting Summary linked to this article. investigated by indirect ELISA screening against the recombinant P116 ectodomain. Positive clones were also tested by western blot against Data availability protein profiles from M. pneumoniae cell lysates and by immuno- Cryo-electron microscopy densities of the original P116 density map fluorescence using whole, non-permeabilized M. pneumoniae cells (3.3-Å resolution), the empty P116 (4-Å resolution) and the refilled (see below). Only those clones with supernatants revealing a single P116 (3.5-Å resolution) have been deposited in the EMDB under the 116-kDa band in protein profiles and also exhibiting a consistent fluo- accession codes EMD-15274, EMD-15275 and EMD-15276, respectively. rescent staining of M. pneumoniae cells were selected and used in this Model coordinates of original and empty P116 have been deposited work. Polyclonal sera were obtained by cardiac puncture of properly in the PDB under the accession codes 8A9A and 8A9B, respectively. euthanized mice just before splenectomy and titred using serial dilu- The mass spectrometry proteomics data have been deposited to the tions of the antigen. The titer of each polyclonal serum was determined ProteomeXchange Consortium (http://proteomecentral.proteomex- as the half-maximal inhibitory concentration (IC50) value from four change.org) via the PRIDE partner repository42 with the dataset identi- parameter logistic plots and was found to be approximately 1/4,000 fier PXD037758. Source data are provided with this paper. for both sera. Polyclonal anti-P1 antibodies were obtained by immuniz- ing two BALB/C mice with recombinant P1 proteins39, respectively, as References described above. The titers obtained for polyclonal anti-P1 antibodies 25. Köhler, G. & Milstein, C. Continuous cultures of fused cells were approximately 1/2,500 and 1/3,000, respectively. secreting antibody of predefined specificity. 1975. J. Immunol. 174, 2453–2455 (2005). Immunofluorescence microscopy 26. Cedó, L. et al. LDL receptor regulates the reverse transport of The immunofluorescence staining of mycoplasma cells on chamber macrophage-derived unesterified cholesterol via concerted slides was similar to previously described40, with several modifica- action of the HDL–LDL axis: insight from mouse models. Circ. Res. tions. Cells were washed with PBS containing 0.02% Tween 20 (PBS-T) 127, 778–792 (2020). prewarmed at 37 °C, and each well was fixed with 200 μL of 3% para- 27. Vasilopoulou, C. G. et al. Trapped ion mobility spectrometry and formaldehyde (wt/vol) and 0.1% glutaraldehyde. Cells were washed PASEF enable in-depth lipidomics from minimal sample amounts. three times with PBS-T, and slides were immediately treated with 3% Nat. Commun. 11, 331 (2020). BSA in PBS-T (blocking solution) for 30 minutes. The blocking solu- 28. Tsugawa, H. et al. A lipidome atlas in MS-DIAL 4. Nat. Biotechnol. tion was removed, and each well was incubated for 1 hour with 100 μL 38, 1159–1163 (2020). of the primary antibodies diluted in blocking solution. For P116 and 29. Ahmed Mohamed, A. M. & Molendijk, J. Lipidr (Bioconductor, P1 polyclonal sera, we used a 1/2,000 dilution; a 1/10 dilution was used 2019). for monoclonal antibodies from hybridoma supernatants. Wells 30. Kong, A. T., Leprevost, F. V., Avtonomov, D. M., Mellacheruvu, D. & were washed three times with PBS-T and incubated for 1 hour with a Nesvizhskii, A. I. MSFragger: ultrafast and comprehensive peptide 1/2,000 dilution of a goat anti-mouse Alexa 555 secondary antibody identification in mass spectrometry-based proteomics. Nat. (Invitrogen) in blocking solution. Wells were then washed three times Methods 14, 513–520 (2017). with PBS-T and incubated for 20 minutes with 100 μL of a solution 31. Mastronarde, D. N. Automated electron microscope tomography of Hoechst 33342 10 μg/μL in PBS-T. Wells were finally washed once using robust prediction of specimen movements. J. Struct. Biol. with PBS-T and replenished with 100 μL of PBS before microscopic 152, 36–51 (2005). Nature Structural & Molecular Biology Article https://doi.org/10.1038/s41594-023-00922-y 32. Punjani, A., Rubinstein, J. L., Fleet, D. J. & Brubaker, M. A. Forschungsgemeinschaft (FR 1653/14-1 for MS and, FR 1653/6-3 for cryoSPARC: algorithms for rapid unsupervised cryo-EM structure LS) and the Research Training Group iMOL (GRK 2566/1 for SM). determination. Nat. Methods 14, 290–296 (2017). 33. Kunz, M. & Frangakis, A. S. Super-sampling SART with ordered Author contributions subsets. J. Struct. Biol. 188, 107–115 (2014). L. S.: Planned and carried out the single-particle analysis; solved the 34. Kunz, M. & Frangakis, A. S. Three-dimensional CTF correction P116 structure in the original, emptied and refilled state; proposed improves the resolution of electron tomograms. J. Struct. Biol. the mechanism of cholesterol uptake on the basis of the structural 197, 114–122 (2017). data; solved the structure of P116 with HDL. D. V.: initiated the project; 35. Emsley, P. & Cowtan, K. Coot: model-building tools for prepared of the P116 gene synthesis; cloned two constructs, 13–957 molecular graphics. Acta Crystallogr. Sect. D Biol. Crystallogr. 60, and 30–657; did expression and purification tests; adapted the best 2126–2132 (2004). conditions for its stabilization; did expression and purification of 36. Headd, J. J. et al. Use of knowledge-based restraints in mutant W681A and analysis by MALS. The order of L. S. and D. V. in phenix.refine to improve macromolecular refinement at the author list was decided randomly. Both authors are entitled to low resolution. Acta Crystallogr. Sect. D Biol. Crystallogr. 68, use their names first when citing this work. D. V., L. S., I. F., A. S. F.: 381–390 (2012). model building of the original, emptied and refilled P116 protein. D. V. 37. Afonine, P. V. et al. Towards automated crystallographic structure and J. M.: emptying protocol; on-site assistance in the experimental refinement with phenix.refine. Acta Crystallogr. Sect. D Biol. preparation of the HDL-P116 interaction and the uptake of radioactive Crystallogr. 68, 352–367 (2012). cholesterol (free and esterified). J. J., N. R., J. C. E.-G.: assessed uptake 38. Guasch, A. et al. Structure of P46, an immunodominant surface of radioactive cholesterol (free and esterified). S. M.: planned and protein from Mycoplasma hyopneumoniae: interaction with a carried out the single-particle sample preparation for the empty monoclonal antibody. Acta Crystallogr. Sect. D Struct. Biol. 76, and refilled samples and the sample mixed with HDL. J. P. and 418–427 (2020). M. M.: obtained hybridomas and monoclonal antibodies against P116 39. Vizarraga, D. et al. Immunodominant proteins P1 and P40/P90 protein; immunolocalization of P116 by epifluorescence microscopy, from human pathogen Mycoplasma pneumoniae. Nat. Commun. time-lapse microcinematography. M. P. S.: Advised on single-particle 11, 5188 (2020). experiments and structure determination procedures. J. D. L. and 40. Nakane, D., Kenri, T., Matsuo, L. & Miyata, M. Systematic structural J. M.-C.: Mass spectrometry analyses of all samples prepared for analyses of attachment organelle in Mycoplasma pneumoniae. single-particle analysis; lipidomics and proteomics analyses. I. F.: PLoS Pathog. 11, e1005299 (2015). Designed and supervised research. A. S. F.: Designed and supervised 41. Seto, S., Kenri, T., Tomiyama, T. & Miyata, M. Involvement of research. I. F. and A. S. F.: Wrote the manuscript, with contributions P1 adhesin in gliding motility of Mycoplasma pneumoniae as from all authors. revealed by the inhibitory effects of antibody under optimized gliding conditions. J. Bacteriol. 187, 1875–1877 (2005). Competing interests 42. Perez-Riverol, Y. et al. The PRIDE database and related tools and The authors declare no competing interests. resources in 2019: improving support for quantification data. Nucleic Acids Res. 47, D442–D450 (2019). Additional information Extended data is available for this paper at Acknowledgements https://doi.org/10.1038/s41594-023-00922-y. We thank M. R. Vabulas for valuable discussions and comments. We also thank the Frankfurt Center for Electron Microscopy for Supplementary information The online version contains supplementary measurement time. We thank the Central Electron Microscopy material available at https://doi.org/10.1038/s41594-023-00922-y. Facility at the MPI of Biophysics in Frankfurt, which enabled us to collect the empty P116 dataset, particularly S. Welsch who Correspondence and requests for materials should be addressed to assisted during the data collection. We thank L. Company and Ignacio Fita or Achilleas S. Frangakis. I. Fernández-Vidal for their support during MALS and mass spectroscopy measurements, A. Iborra (Servei de Cultius Cellulars, Peer review information Nature Structural & Molecular Biology thanks Anticossos Citometria, UAB) for his assistance with immunizing the anonymous reviewers for their contribution to the peer review of mice, D. Santos for his assistance in the radioactivity experiment this work. Primary Handling Editors: Katarzyna Ciazynska and Florian and R. Pérez-Luque and D. Aparicio for their constant support Ullrich, in collaboration with the Nature Structural & Molecular Biology and discussions. J. P. was funded by grants BIO2017-84166-R and team. Peer reviewer reports are available. PID2021-125632OB-C22 from the ministerio de Ciencia, Innovación y Universidades (MICINN, Spain). I. F. was funded by MICINN-Spain Reprints and permissions information is available at grant PID2021-125632OB-C21. A. S. F. was supported by the Deutsche www.nature.com/reprints. Nature Structural & Molecular Biology Article https://doi.org/10.1038/s41594-023-00922-y Extended Data Fig. 1 | P116 constructs. Overview of the different constructs (30–957, 30–818, 246–845) used for expression. For expression purposes a His-tag (KHHHHH) was added at the C-terminus. For the structural analysis by cryoEM the construct from 30-957 was used. Nature Structural & Molecular Biology Article https://doi.org/10.1038/s41594-023-00922-y Extended Data Fig. 2 | Overview of cryoEM processing of P116. (a) cryoEM density map (C1 symmetry) of P116 ranging from 2.5 to 5 Å. (d) Schematic Representative Micrograph and obtained 2D classes used for template picking. processing overview for the P116 construct. The processing for P116 empty, P116 (b) Fourier shell correlation of P116 reports a final resolution of 3.3 Å according to refilled and P116 + HDL was carried out in a similar manner. the 0.143 cut-off criteria. (c) Local resolution map at different thresholds of the Nature Structural & Molecular Biology Article https://doi.org/10.1038/s41594-023-00922-y Extended Data Fig. 3 | Dimerization interface and model building of P116. a very large MW (molecular weight) range, and the clear decrease in size with Detail of the dimerization zone with the two monomers and the TRP-681 residue respect to the WT due to the now predominance of the monomeric state. (c) of both chains contacting the opposite one. (b) SEC-MALS of P116 W681A. The Exemplary superimpositions of the cryoEM density and the PDB model for the greatest polydispersity of this sample is observed, which oscillates between core and N-terminal domain. Nature Structural & Molecular Biology Article https://doi.org/10.1038/s41594-023-00922-y Extended Data Fig. 4 | Flexibility of P116. CryoEM classes display different between the individual vibrational modes. Here ‘eigenvector 3’ (EV3) is plotted flexibility modes. Individual reconstructions are shown on the left column. Their against ‘eigenvector 4’ (EV4) that display the largest resemblance to the radius variability is shown in the middle column. Plotting of the individual eigenvector change and ‘eigenvector 4’ (EV4) is plotted against ‘eigenvector 5’ (EV5) that weights shows, that no clear distribution is visible and a smooth transition display the largest resemblance to the twist. Nature Structural & Molecular Biology Article https://doi.org/10.1038/s41594-023-00922-y Extended Data Fig. 5 | MALDI-TOF of P116, P116 empty and P116 refilled. Individual low molecular weight MALDI-MS spectra of P116, P116-empty and P116-refilled as displayed in in Fig. 4d. Nature Structural & Molecular Biology Article https://doi.org/10.1038/s41594-023-00922-y Extended Data Fig. 6 | See next page for caption. Nature Structural & Molecular Biology Article https://doi.org/10.1038/s41594-023-00922-y Extended Data Fig. 6 | P116 in other Mycoplasma species. (a) Conserved other amino acids. (b) Predictions of P116 from different Mycoplasma spp. in (minimum 55 %) amino acids pointing inwards the hydrophobic cavity. Black comparison with the structure resolved with cryoEM. The N-terminal and the indicates amino acids with a hydrophobic character. Red indicates amino acids core domain have been predicted independently and subsequently fitted in with a hydrophilic character. Blue indicates amino acids with a negative side the cryoEM density map of P116. The coloring of the models is identical to the chain. Purple indicates amino acids with a positive side chain. Green indicates Figs. 1–4. Nature Structural & Molecular Biology Article https://doi.org/10.1038/s41594-023-00922-y Extended Data Fig. 7 | Fourier shell correlations of P116 empty, P116 refilled and P116 + HDL. (a) Final reported resolution of P116 empty at 4 Å according to the 0.143 cut-off criteria. (b) Final reported resolution of P116 refilled at 3.5 Å according to the 0.143 cut-off criteria. (c) Final reported resolution of P116 + HDL at 9 Å according to the 0.143 cut-off criteria. Nature Structural & Molecular Biology Article https://doi.org/10.1038/s41594-023-00922-y Extended Data Fig. 8 | Conformational change between P116 and P116 empty. come closer to the core domain, massively reducing the available volume. Their Superimposition of one subunit of P116 with one subunit of P116 empty. The new position is markedly different compared to the full P116 (shown in light movement of the fingers 1-4 is indicated by dashed lines. Side view of the ribbon blue). Finger 1 moved 8 Å sideways and towards the core, finger 2 has moved 13 Å representation of the empty and full P116 shows that the fingers (in purple) have towards the core and Finger 3 has moved 12 Å towards the core. Nature Structural & Molecular Biology Article https://doi.org/10.1038/s41594-023-00922-y Extended Data Fig. 9 | HisTraps and Size-exclusion running profiles of the purification. (a) Running profile of the purification of the original P116 construct. (b) Running profile of the separation between P116 and HDL. The green slope indicates the increase of the imidazole concentration during the purification. Nature Structural & Molecular Biology Article https://doi.org/10.1038/s41594-023-00922-y Extended Data Fig. 10 | Cryo-electron tomogram of a M. pneumoniae cell. tomograms were recorded) (b) Gallery of different possible P116 proteins from (a) Representative 10 nm thick slice of a tomographic reconstruction of an several tomographic reconstructions depicting P116 attached to the membrane individual M. pneumoniae cell depicting a possible P116 protein on the surface (P116 molecules could be observed from three independent preparations). of the M. pneumoniae membrane (One slice from a tomogram is shown – 10 Tomograms were acquired with a defocus of −3 µm. Nature Structural & Molecular Biology