Three-Dimensional Structure of TspO by ElectronCryomicroscopy of Helical Crystals

Vladimir M. Korkhov1,2,3,∗, Carsten Sachse1,2,4, Judith M. Short1, and Christopher G.Tate1,∗∗1Medical Research Council Laboratory of Molecular Biology, Hills Road, Cambridge CB2 0QH,UK.

SummaryThe 18 kDa TSPO protein is a polytopic mitochondrial outer membrane protein involved in a widerange of physiological functions and pathologies, including neurodegeneration and cancer. Thepharmacology of TSPO has been extensively studied, but little is known about its biochemistry,oligomeric state, and structure. We have expressed, purified, and characterized a homologousprotein, TspO from Rhodobacter sphaeroides, and reconstituted it as helical crystals. Usingelectron cryomicroscopy and single-particle helical reconstruction, we have determined a three-dimensional structure of TspO at 10 Å resolution. The structure suggests that monomeric TspOcomprises five transmembrane α helices that form a homodimer, which is consistent with thedimeric state observed in detergent solution. Furthermore, the arrangement of transmembranedomains of individual TspO subunits indicates a possibility of two substrate translocationpathways per dimer. The structure provides the first insight into the molecular architecture ofTSPO/PBR protein family that will serve as a framework for future studies.

Highlights► R. sphaeroides TspO reconstituted as helical tubular crystals ► Cryo-EM and image analysislead to structure of TspO at 10 Å resolution ► The structure confirms a dimeric fivetransmembrane domain architecture of TspO ► Two putative substrate translocation pathways perdimer are suggested

KeywordsPROTEINS; CELLBIO

© 2010 ELL & Excerpta Medica.This document may be redistributed and reused, subject to certain conditions.∗Corresponding author volodymyr.korkhov@mol.biol.ethz.ch. ∗∗Corresponding author cgt@mrc-lmb.cam.ac.uk.2These authors contributed equally to this work3Present address: Institute of Molecular Biology and Biophysics, Eidgenössische Technische Hochschule Hoenggerberg,Schafmattstrasse 20, HPK D11, 8046 Zurich, Switzerland4Present address: European Molecular Biology Laboratory, Meyerhofstrasse 1, 69117 Heidelberg, GermanyThis document was posted here by permission of the publisher. At the time of deposit, it included all changes made during peerreview, copyediting, and publishing. The U.S. National Library of Medicine is responsible for all links within the document and forincorporating any publisher-supplied amendments or retractions issued subsequently. The published journal article, guaranteed to besuch by Elsevier, is available for free, on ScienceDirect.

Sponsored document fromStructure (London, England :1993)

Published as: Structure. 2010 June 09; 18(6): 677–687.

Sponsored Docum

ent Sponsored D

ocument

Sponsored Docum

ent

IntroductionThe human 18 kDa translocator protein (TSPO), formerly known as the peripheralbenzodiazepine receptor (Papadopoulos et al., 2006a), is involved in a wide range ofphysiological functions and pathological conditions. TSPO is also among the key targets ofValium (diazepam), which is one of the most frequently prescribed drugs for anxietydisorders (Braestrup et al., 1977; Gavish et al., 1999). Among the established molecularfunctions of TSPO are translocation of cholesterol (Li and Papadopoulos, 1998; Li et al.,2001) and porphyrins (Taketani et al., 1994) across the mitochondrial outer membrane,where a large fraction of TSPO is normally localized within the cell. On a cellular level,TSPO has been shown to be involved in steroid biosynthesis (Lacapere and Papadopoulos,2003), cellular respiration (O'Hara et al., 2003), proliferation (Galiegue et al., 2004), andapoptosis (Maaser et al., 2001). In addition, biochemical and pharmacological data haveimplicated TSPO in a wide range of pathological conditions, including epilepsy (Veenmanet al., 2002), neurodegenerative diseases (Papadopoulos et al., 2006b), and cancer (Hanet al., 2003; Maaser et al., 2002; Weisinger et al., 2004). TSPO is also an establishedpositron emission tomography marker for pathologies of the central nervous system (Zhanget al., 2004). A wealth of pharmacological data on TSPO has been published and high-affinity ligands have been developed (Papadopoulos et al., 2006a; Scarf et al., 2009).However, although attempts have been made to study the purified and reconstituted TSPO(Delavoie et al., 2003; Garnier et al., 1994; Lacapere et al., 2001), little is known about itsthree-dimensional (3D) structure, oligomeric state, and mode of action, i.e., whether itoperates as a pump, a transporter, or a channel (Papadopoulos et al., 2006a). Bioinformaticpredictions and hydropathy analyses, confirmed by biochemical and biophysical evidence,have indicated that TSPO and its bacterial homologs probably contain five α-helicaltransmembrane domains that possibly form dimers or multimers (Bogan et al., 2007;Papadopoulos et al., 1994; Yeliseev and Kaplan, 2000).

Structural studies of eukaryotic membrane proteins are still difficult. Expressing, purifying,and stabilizing eukaryotic membrane proteins presents formidable obstacles that need to beovercome before their structures can be determined by X-ray crystallography, electroncryomicroscopy (cryo-EM), or nuclear magnetic resonance (NMR). To obtain insight intothe molecular structure of the proteins belonging to the TSPO family, we have thereforeused a bacterial homolog, tryptophan-rich sensory protein (TspO) from Rhodobactersphaeroides, as a structural and functional model of the human protein.

TspO from R. sphaeroides is involved in the control of photosynthetic gene expression andit has been proposed to act as an oxygen sensor (Yeliseev et al., 1997) that downregulatesthe photopigment genes in response to oxygen (Yeliseev and Kaplan, 1995, 1999; Zeng andKaplan, 2001). Its primary structure is very similar to that of the human TSPO, with 33.5%identity between the aligned amino acid sequences of R. sphaeroides TspO and humanTSPO (Figure 1), excluding the longer interhelical loops in TSPO (Yeliseev and Kaplan,1995); this level of homology is highly significant. Furthermore, bacterial TspO has beensuggested to have functional and structural similarities to the human protein (Yeliseev andKaplan, 2000), making it an ideal candidate for biochemical and structural work. Here wereport the structure of R. sphaeroides TspO at 10 Å resolution, which represents the firststructural data for a member of this family.

ResultsExpression and Purification of TspO in Escherichia coli

TspO from R. sphaeroides was expressed in E. coli strain BL21(DE3) (Figure 2A). Contraryto the reports of TspO localization in the outer membrane of R. sphaeroides, the

Korkhov et al. Page 2

Published as: Structure. 2010 June 09; 18(6): 677–687.

Sponsored Docum

ent Sponsored D

ocument

Sponsored Docum

ent

recombinantly expressed protein was inserted predominantly into the E. coli innermembrane (Figure 2B and 2C). This is unsurprising, given the predicted helical nature ofTspO and the fact that all known bacterial outer membrane proteins have been shown to beβ-barrel proteins, with the exception of Wza that contains a single C-terminaltransmembrane α helix (Dong et al., 2006). TspO was expressed with a C-terminal His6 tag,which allowed its purification in dodecylmaltoside (DDM) by Ni2+-affinity chromatographyand preparative-scale size exclusion chromatography. The purified TspO was monodisperseby size exclusion chromatography (Figure 2F) and was highly pure on Coomassie blue-stained SDS polyacrylamide gels (Figure 2A).

Detergent-Solubilized TspO Is DimericTspO purified in DDM was subjected to blue native PAGE analysis (Figures 2D and 2E).TspO formed two bands on the gel, with the predominant species appearing at an apparentmolecular weight of ∼80 kDa and a minor species at an apparent molecular weight of ∼160kDa. As these sizes include the mass of bound detergent, lipids, and Coomassie blue, it isnot possible to use these data directly to determine the oligomeric state of TspO. However,when TspO was incubated in 1% SDS, the apparent mass decreased, which is consistentwith the disruption of an oligomeric species by the harsh detergent. For comparison, EmrEwas used as a control because it is known to be a homodimeric membrane protein (Butleret al., 2004) of comparable size (15 kDa in its tagged form) to TspO (18 kDa); the mobilityof TspO in the blue native gels in the presence or absence of SDS is virtually identical toEmrE. All these data are consistent with TspO purified in DDM being a dimer. The variableappearance of a higher molecular weight band in the TspO sample was likely due to proteinaggregation during its concentration because the higher molecular weight fraction wasnegligible when the concentration of purified TspO was kept below 5 mg/ml (Figure 2E).This is consistent with size exclusion chromatography of TspO purified in DDM showing amonodisperse protein peak (Figure 2F) and is similar to the behavior of mouse TSPO uponanalysis by blue native PAGE (Rone et al., 2009).

Purified TspO in Detergent Binds PorphyrinsTo establish the functional integrity of the expressed and purified TspO, we first performedin vivo and in vitro uptake and binding assays on either intact E. coli or membranepreparations, respectively, both containing overexpressed TspO. Although the results ofthese experiments were indicative of an ability of the expressed TspO to bind and transportporphyrins, there was very high background binding to control cells not expressing TspO,probably because of the hydrophobicity of porphyrins, precluding accurate determination ofsubstrate transport and binding affinity constants (data not shown). Therefore, intrinsictryptophan fluorescence quenching was used to monitor porphyrin binding to purified TspO(Figure 3). We tested the binding of three different porphyrin analogs to purified TspO,namely, protoporphyrin IX (PPIX), hemin, and coproporphyrin III (cPPIII). Titration ofTspO with increasing ligand concentrations revealed an efficient fluorescence quenching byall three analogs (Figures 3A–3C), with apparent half-maximal binding constants, Kapp, of8.6 ± 0.6 μM for PPIX, 10.0 ± 2.5 μM for hemin, and 17.9 ± 1.7 μM for cPPIII (the valuesare mean ± SD). All three porphyrin analogs interact with TspO with similar affinities,which suggests that TspO may have a broad range of porphyrin specificity in vivo. Thesedata confirmed that TspO was expressed and purified in a functional state. Moreover, theobserved values are comparable to the reported affinity of PPIX for mouse TSPO, with a Kiof 18.95 μM (Wendler et al., 2003). No fluorescence quenching was observed when thenontransported molecule D-aminolevulinic acid was added to TspO (Figure 3D).

Korkhov et al. Page 3

Published as: Structure. 2010 June 09; 18(6): 677–687.

Sponsored Docum

ent Sponsored D

ocument

Sponsored Docum

ent

TspO Reconstitution Yields Tubular Helical CrystalsPurified TspO was dialyzed in the presence of purified lipids in attempts to produce well-ordered 2D crystals for cryo-EM. A systematic screen of the reconstitution conditions withvarying lipids, lipid-to-protein ratios, buffer composition, temperature, and dialysis time wasperformed, which led to conditions where TspO formed narrow tubular crystals as observedby electron microscopy of negatively stained specimens (Figure 4A). TspO tubes wererapidly frozen on holey carbon grids by plunging them into liquid ethane and cryo-EMimages were collected under low-dose conditions at liquid nitrogen temperature. The ice-embedded tubes were 100–1000 nm long and 25 nm in diameter (Figure 4C). The tubesmake up 62% of lipid structures in the sample with another quarter belonging to a group ofspherical vesicles. The remaining 13% of lipid structures consisted of deformed ellipticalvesicles that may represent intermediate stages of partially assembled crystals (Figure 4C,inset).

Three-Dimensional Reconstruction and Molecular Architecture of TspO in the MembraneTo obtain 3D structural information of TspO, tubes were analyzed by cryo-EM and a 10 Åresolution map was determined using single-particle helical reconstruction. The micrographsof well-preserved tubes showed a clearly distinguishable stacked disk-like morphology(Figure 4D). From 36 micrographs, a total of 205 straight tubes were segmented into 7542overlapping image squares of an 108 nm dimension. Initially, the principal helicalorganization of the tubes could be derived from Fourier analysis of averaged tube images(see Figures S1B and S1C available online).

To ensure that all the processed tubes consisted of the same helical packing, we usedmultivariate statistical analysis to classify tube segments to carefully analyze their width andthe helical diffraction pattern. First, the width of all individual segments and their classaverages was determined to be 25 nm (Figure 4E). Second, two out of twelve classes gaverise to an interpretable set of three identical layer lines (Figure S1B). Subsequent indexingallowed the building of the helical lattice (Figures S1C and S1D; see ExperimentalProcedures). Hence, the tubes were composed of arrays of stacked rings each consisting of12 TspO dimers, each disk being 32.7 Å thick and related to the neighboring disk by arotation of 9.5° about the helical axis (Figure S1D). Subsequently, the derived helicalsymmetry parameters were refined using single-particle processing (Figures S1F and S1G)(Low et al., 2009). For the 3D image reconstruction, we used a recently developed singleparticle-based method that exploits helical symmetry (Sachse et al., 2007). To furtherinclude only those segments with a visible stacking order, we excluded two thirds of tubesegments that were devoid of the 32 Å layer line diffraction; 2610 segments, i.e., 20% of thetube images, were included in the final density map at 10 Å resolution (Figures 4D and 4F;Figure S3) (the statistics of the reconstruction are shown in Table 1).

The resulting EM model of TspO revealed a molecular arrangement in which a pair of TspOmonomers form a tightly associated symmetrical dimer of approximate dimensions 40 × 27Å in the membrane plane and about 40 Å perpendicular to the membrane (Figure 5). Thecross-sections through the final density map perpendicular to the lipid bilayer plane reveallow-density areas formed within the individual TspO monomers (Figures 5B–5F). Theselow-density areas are apparent also in sections parallel to the plane of the membrane(Figures 5H–5K), which show that five discrete density peaks comprise one TspO monomer(circled in Figures 5H–5K), consistent with the prediction from hydropathy plots that eachmonomer contains five transmembrane helices. Each TspO dimer in the helical crystal issurrounded by a low-density area corresponding to the annular lipids, which is evident fromFigures 5B–5F and 5H–5K. There are, of course, other possibilities for which a group offive density peaks in the dimer corresponds to the TspO monomer. Currently we favor the

Korkhov et al. Page 4

Published as: Structure. 2010 June 09; 18(6): 677–687.

Sponsored Docum

ent Sponsored D

ocument

Sponsored Docum

ent

interpretation depicted in Figures 5H–5K because of its simplicity and how well it conformsto generally accepted principles of membrane protein structure. In addition, there appears tobe no discernable density that could represent potential interhelical loops between theproposed monomers, which is also consistent with this interpretation.

Interpretation of the Density MapThe low-resolution density map contained a number of rod-shaped densities that we haveinterpreted as transmembrane α helices. We have fitted α-helical backbones manually intothese densities (Figures 6A–6E; Figure S4) either by modeling idealized α helices directlyinto the continuous density features or by connecting the strongest density elements withinthe map in a way that satisfied simple α-helical bundle geometry. At 10 Å resolution, it isnot possible to assign the TspO sequence to the positions in the 3D map, so each putativetransmembrane α helix is labeled a–e in one monomer and a′–e′ in the other monomer of theTspO dimer (Figure 6A). The interface between the two TspO monomers is formed byhelices a, b, a′, and b′ (Figures 6A and 6D) and it is relatively flat without any largeinvaginations. Helices b and b′ of the two parallel monomers are the main structuralelements that appear to be making intermonomer contacts throughout the lipid bilayer(Figures 6A and 6B). In contrast, a cleft is formed between helices d and e within themonomer, with helix e being the most highly tilted of all the helices (Figures 6C and 6E).Helix c forms contacts predominantly with helices b and d, although it may possibly formcontacts with helix a′ in the adjacent monomer.

In addition to the transmembrane densities that we have interpreted as α helices, there arelarge areas of density at the ends of helices c and e that appear on the outside of the tubes(Figure 6; Figure S4). This density appears to form the major contact between adjacentdimers in the crystalline lattice and, in conjunction with the absence of an equivalent densityon the opposite side of the membrane, it gives the dimer a distinctive wedge shape thatdefines the curvature of the tubes and contributes toward the overall helical crystalgeometry. It is unclear at 10 Å resolution what the densities at the ends of helices c and erepresent, but it is most likely to be the N terminus and/or C terminus, the latter still havingthe His6 tag attached. It seems less likely that the density represents a loop region betweenhelices d and e given the large distance between the ends of these helices, but the additionalmasses at the ends of helices a–c probably do represent interhelical loops. However,contributions to all these membrane surface densities from ordered lipid head groups, oreven perhaps porphyrin-like substrates copurified with the protein, cannot be whollyexcluded.

DiscussionSince its discovery as the “peripheral benzodiazepine receptor” in 1977 (Braestrup et al.,1977; Braestrup and Squires, 1977), TSPO has attracted considerable interest from both thepharmaceutical industry and academic research laboratories. High affinity drugs for thehuman homolog of the TSPO/PBR family have been developed for clinical applications andto elucidate its cellular functions, along with its involvement in physiological processes andpathophysiological conditions (Gavish et al., 1999; Papadopoulos et al., 2006a; Scarf et al.,2009). Indeed, much of what we have learnt about TSPO/PBR proteins was inferred fromindirect pharmacological evidence (Gavish et al., 1999; Papadopoulos et al., 2006a). Only afew studies have attempted to address the structure of TSPO (Murail et al., 2008) and mostof these have addressed the structural properties of drug binding sites as inferred fromstructure-activity assays (Anzini et al., 2001; Cinone et al., 2000). Consequently, up to nowmost of the questions regarding how the members of this protein family are organized on themolecular level remain unresolved. The low-resolution reconstruction presented here

Korkhov et al. Page 5

Published as: Structure. 2010 June 09; 18(6): 677–687.

Sponsored Docum

ent Sponsored D

ocument

Sponsored Docum

ent

provides a first glimpse of the architecture of TspO in its native environment, the lipidbilayer.

TspO from R. sphaeroides was purified to homogeneity in DDM in a functional dimericform, as suggested from the blue native PAGE analysis (Figure 2) and the cryo-EM structureof TspO (Figure 5). The observed dimeric form of TspO is consistent with mutagenesis data,which showed that the mutation W38C in TspO leads to the generation of covalent dimersdue to the formation of an intermolecular disulphide bond (Yeliseev and Kaplan, 2000).Here, we show that TspO is probably a dimer in the membrane based on the 3D structure ofTspO at 10 Å resolution determined from cryo-EM images of helical tubes. The repetitiveunit in the tubular crystals is a bundle of five α helices, which is consistent with the proposalthat the TspO monomer contains five transmembrane domains as predicted from hydropathyanalyses of the primary amino acid sequence and confirmed by epitope mappingexperiments (Joseph-Liauzun et al., 1998). The five α helices are packed into a ten helixbundle related by a two-fold symmetry axis perpendicular to the membrane. This dimer(Figure 6) is the most likely of all the possible dimers in the crystal structure to be ofphysiological relevance, due to the close proximity of the two monomers within thetransmembrane region and the relatively large interface between them. Surrounding thisdimer are regions in the crystal that show no strong density and are therefore predicted tocontain lipid, although strong density is present at the predicted membrane bilayer surfaces,particularly on the outside face of the tube. The origin of this density is unclear, but it mayrepresent ordered regions of either the N or C termini that mediate crystal contacts betweenneighboring TspO dimers. The densities for the transmembrane α helices are sometimesweaker than we might have anticipated, especially in the center of the lipid bilayer, as isseen for helices a, b, and c. These helices were modeled as continuous straight helicesthrough the membrane, which is possible if the discontinuities in the density have arisenbecause of the low resolution of the data or high helix mobility. Another possibility thatcannot be entirely discounted is that the helices contain regions that are non-helical as seenin many membrane protein structures (Gouaux and Mackinnon, 2005). Careful inspection ofthe density map with adjusted contour levels reveals presence of weak connecting density inthose areas where discontinuities are observed, lending support to our interpretation of themap (see Figure S4). However, higher resolution will be required to fully understand thestructure.

Presently, we cannot assign the amino acid sequence of TspO to particular α helices in ourlow-resolution model. In addition, there is insufficient unambiguous data to suggest theorder of helix-helix connectivity in TspO, which makes building a model of TspO from ourstructure more difficult; the model built of EmrE from the 7.5 Å resolution cryo-EMstructure was facilitated by the presence of clear connectivity between two of thetransmembrane α helices (Fleishman et al., 2006). However, it is tempting to suggest thattransmembrane segment 2 of TspO may be equivalent to the position of helix b in thereconstruction (Figure 6). This suggestion is based on the observation that residue W38 inthe loop near the edge of the transmembrane helix 2 of TspO forms covalent dimers whenmutated to cysteine (Yeliseev and Kaplan, 2000). Helices b and b′ in the reconstructionmake the closest intermonomer contacts and it is conceivable that a crosslink betweenresidues C38 of the two monomers may occur in this region.

The dimeric structure of TspO raises questions regarding the mechanism of transport ofsubstrates across the membrane, the stoichiometry of substrate and inhibitor binding andwhere these bind in the transporter. In most structures of solute transporters that contain sixor more α helices, there is one translocation pathway per polypeptide chain, regardless ofwhether the protein forms a monomer, trimer, or tetramer, although there are a fewexceptions, e.g., Sav1866 (Dawson and Locher, 2006). Another example of a transporter

Korkhov et al. Page 6

Published as: Structure. 2010 June 09; 18(6): 677–687.

Sponsored Docum

ent Sponsored D

ocument

Sponsored Docum

ent

where the binding site is formed at the interface between monomers is EmrE, which containsfour transmembrane α helices per monomer, which combine to form an asymmetrichomodimer; each dimer binds one molecule of substrate at the interface between the twomonomers (Korkhov and Tate, 2009; Ma and Chang, 2007; Ubarretxena-Belandia et al.,2003; Ubarretxena-Belandia and Tate, 2004). A small molecule binding site is typicallyformed from a crevice in a bundle of helices. In the structure of the TspO dimer (Figure 6)the most likely substrate binding pocket is, therefore, between helices d and e and between d′ and e′,which implies that there are two binding sites per dimer. This crevice also appears tobe relatively accessible from at least one leaflet of the lipid bilayer, which is characteristic ofmultidrug and lipid transporters like EmrE (Ubarretxena-Belandia et al., 2003), Sav1866(Dawson and Locher, 2006), and P-glycoprotein (Aller et al., 2009). It is also consistent withthe proposal that the TSPO family members can transport large hydrophobic compoundssuch as sterols and porphyrins (Papadopoulos et al., 2006a). The other possibility is that thebinding site is at the monomer-monomer interface, as seen in EmrE, which would imply thatone substrate molecule binds per dimer. However, this appears to be less likely because thisinterface in TspO appears relatively flat with the major interactions mediated by helix bfrom each monomer. It is, therefore, clearly desirable to determine the stoichiometry ofsubstrate binding to either TspO or to one of its homologs, as this would allow thedistinction between these two hypotheses. Similar data were important in defining thefunctional unit of EmrE as a homodimer (Butler et al., 2004), but this required an accurateradioligand binding assay, using a substrate that bound with high affinity, coupled to aminoacid analysis of the purified protein. The lack of high-affinity radioligands for TspOprecludes such a study in our work, although this may be possible for mammalian TSPO,which has a greater diversity of radiolabeled substrates and inhibitors, provided that TSPOcan be overexpressed and purified in a fully functional form.

The low-resolution cryo-EM structure of TspO we present here is suggestive of two bindingsites per dimer, which, in turn, allows for the possibility of cooperativity during substratetransport and potential effects of allosteric modulators. Given that human TSPO is 33.5%identical to R. sphaeroides TspO, it is likely that human TSPO will have a similararchitecture to bacterial TspO, implying that human TSPO could also have two substratetranslocation pathways and thus two potential drug binding sites in close proximity. Thus,the structure could explain the previously observed allosterism of ligand binding to TSPO/PBR proteins (Masonis and McCarthy, 1996) and the existence of several classes of drugbinding sites (Boujrad et al., 1994, 1996). Experimental determination of the oligomericstate of human TSPO will therefore be extremely informative. The structure of R.sphaeroides TspO at higher resolution will help to define further the pathway of substratetranslocation and will provide a valuable framework for future studies on this family ofproteins.

Experimental ProceduresReagents

Chemicals were purchased from Sigma-Aldrich. Hemin, PPIX, and cPPIII were purchasedfrom Frontier Scientific. DDM (Anagrade) was purchased from Anatrace. E. coli polar lipidswere purchased from Avanti Polar Lipids. TspO was isolated by PCR from plasmidpUI2701, generously provided by S. Kaplan, and cloned into plasmid pET29 (Novagen).

Protein Expression and PurificationTspO was cloned into the NdeI and XhoI sites restriction sites of plasmid pET29 in E. colistrain BL21(DE3) for expression from the T7 promoter. Briefly, cells transformed with thepET29-TspO plasmid were grown to an OD600 of 2.0–2.5 in 2xTY medium and induced

Korkhov et al. Page 7

Published as: Structure. 2010 June 09; 18(6): 677–687.

Sponsored Docum

ent Sponsored D

ocument

Sponsored Docum

ent

with 0.5 mM IPTG for 3 hr at 32°C. The cells were harvested by centrifugation (15 min,5000 × g, 4°C). The pellets were resuspended in buffer A on ice [20 mM Tris (pH 7.5) and200 mM NaCl] and the cells broken open using a continuous flow cell disruptor. After abrief spin at low speed (30 min, 10,000 × g, 4°C), the cell membranes were collected using aTi45 rotor in an ultracentrifuge (2 hr, 170,000 × g, 4°C), resuspended in buffer A, flashfrozen in liquid nitrogen, and stored at −80°C. For purification, the membranes weresolubilized in buffer A supplemented with 2% DDM, and the insoluble material wasremoved by ultracentrifugation as before. The protein was purified using a His6 tag onNiNTA-agarose (QIAGEN), according to the manufacturer's protocol. The protein elutedfrom the Ni-affinity column was separated from the aggregates using gel filtration in 20 mMTris (pH 7.5), 30 mM NaCl, and 0.025% DDM.

Inner and Outer Membrane SeparationTo separate the inner and outer membranes of E. coli, a two-step sucrose gradientfractionation protocol was used. The membrane preparations from TspO-expressing cellswere mixed with sucrose (20% final concentration) and layered onto a pre-formed 70%–60% two-step sucrose gradient containing equivalent amounts of buffer and salt in acentrifuge tube (final volume ratio for the solutions with different sucrose concentrations of1:1:1). Following centrifugation (16 hr, 260,000 × g, 4°C) the bands corresponding to innermembranes (golden layer between the 20% and 60% sucrose layers) and outer membranes(white layer between the 60% and 70% sucrose layers) were collected, diluted with bufferA, and centrifuged (2 hr, 310,000 × g, 4°C). The final pellet was resuspended in buffer Aand used for SDS-PAGE and western blot analysis, according to standard procedures. ForTspO-His6 detection, an HRP-conjugated His tag antibody (Novagen) was used, followingthe vendor's instructions.

Blue Native PAGETo determine the oligomeric state of the protein in detergent micelles, blue native PAGEwas performed (Schagger and von Jagow, 1991). The 4%–16% gradient Novex Bis-Tris gels(Invitrogen) and the commercially available cathode and anode buffers for blue nativePAGE (Invitrogen) were used. Protein samples in DDM were mixed with the loading bufferand loaded on a gel (5–10 μg protein/lane). Electrophoresis (performed at 4°C for 4–6 hr)and gel destaining were performed according to the vendor's instructions.

Analytical Size Exclusion ChromatographyA purified TspO sample, concentrated to 1 mg/ml, was loaded on a Superdex 200 10/300GL column, pre-equilibrated with buffer composed of 20 mM Tris (pH 7.5), 200 mM NaCl,and 0.025% DDM. The UV absorbance at 280 nm was monitored during chromatographyand plotted against time (in minutes).

Binding AssaysTo assess the interactions between purified TspO and porphyrins, intrinsic tryptophan-fluorescence quenching was used. In brief, purified TspO in DDM at a final concentration of2.5 μM was titrated with increasing amounts of heme in the desired concentration range.Each titration point was followed by a spectrum measurement (excitation at 295 nm,emission at 300–400 nm) and detection of tryptophan fluorescence. Arbitrary fluorescenceunits at 340 nm were plotted against ligand concentration to evaluate the fluorescencequenching. The apparent dissociation constants, Kapp, were estimated using a simple one-site binding model in GraphPad Prism.

Korkhov et al. Page 8

Published as: Structure. 2010 June 09; 18(6): 677–687.

Sponsored Docum

ent Sponsored D

ocument

Sponsored Docum

ent

Crystallization of TspO TubesReconstitution of the purified TspO into helical tubular crystals was performed in a setupsimilar to that used previously for EmrE (Korkhov and Tate, 2008; Tate et al., 2001). Theprotein preparations at final concentrations of 0.5–1 mg/ml were incubated with E. coli polarlipid extract at lipid/protein ratios of 0.2–0.6 in Slide-A-Lyzer cassettes (Pierce) submergedin 100 ml beakers filled with dialysis buffer [20 mM Tris (pH 7.5), 100 mM NaCl, and 2mM EDTA] for 3 days, with daily exchange of dialysis buffer. The tubular crystals obtainedthis way were used directly for cryo-EM.

Electron MicroscopyFor freezing and cryo-EM, tubes of TspO at a concentration of 0.5 mg/ml in reconstitutionbuffer were pipetted on to a glow-discharged holey carbon grid (Agar Scientific). Thesolution was blotted from the opposite side of the grid with a strip of filter paper and the gridwas plunged into a reservoir of liquid ethane. Grids frozen this way were then cryo-transferred into storage boxes. Standard cryo-transfer procedures were then used with aGatan 626 cryo-stage for electron microscopy. The tubes were analyzed with a Tecnai F20electron microscope operating at an accelerating voltage of 200 kV, using standard low-doseimaging conditions (10–15 e/Å−2). Images were acquired on film, using flood-beamillumination at a magnification of 50,000 with an underfocus between 0.7 and 2.3 μm. Forfurther analysis, 36 micrographs were digitized using a home-built KZA scanner (Hendersonet al., 2007) with a 6 m step size and processed as described below. For the handdetermination, TspO tubes were negatively stained using a 0.75% uranyl formate solution(Ohi et al., 2004); we recorded images of these tubes on an FEI Tecnai T12 microscope at amagnification of 42,000 using a 2 × 2k CCD camera. The final pixel size on the specimen of3.25 Å/pixel was calibrated by measuring the (23 Å)−1 layer line of 18 tobacco mosaic virus(TMV) reference specimens.

Image Processing and 3D Image ReconstructionThe structure of TspO lipid tubes was determined using a recently developed single particle-based helical reconstruction algorithm (Sachse et al., 2007). Initially, the method was basedon iterative helical real-space reconstruction (Egelman, 2007) but has been significantlyextended by applying a full correction of the contrast-transfer function (CTF) of themicroscope, imposing restraints on the alignment and a new symmetrization procedure(Sachse et al., 2007). This improved single-particle reconstruction procedure imposes helicalsymmetry in real space and thus requires precise knowledge of parameters of helicalsymmetry. Therefore, we derived initial parameters of helical symmetry from image analysisof the helical diffraction pattern of averaged class sums (Crowther et al., 1996) and furtherrefined it using the single-particle method described below.

Fourier-Bessel Analysis of Diffraction Pattern and Determination of Helical SymmetryParameters

First, we inspected the helical diffraction pattern on single tubes. For this analysis, TspOlipid tubules were boxed, padded, and floated using Ximdisp (Smith, 1999) and wereFourier transformed. Selected tubes possessed three layer lines at (102 Å)−1, (48 Å)−1, and(32 Å)−1. Because of the limited signal present in single tube images, we classified thesegmented tube stack using multivariate statistical analysis and further analyzed two well-ordered classes with apparent stacking striations from a total of twelve classes (Figure S1B).The remaining classes did not show clear stacking or moiré features owing to inherentcrystal disorder and insufficient separation of different azimuthal or out-of-plane views. Theamplitudes and phases of the best two classes could be readily analyzed and the helicallattice was indexed. The (32 Å)−1 layer line exhibits a maximum on the meridian and has,

Korkhov et al. Page 9

Published as: Structure. 2010 June 09; 18(6): 677–687.

Sponsored Docum

ent Sponsored D

ocument

Sponsored Docum

ent

therefore, a Bessel order of 0. The reciprocal radius of the (102 Å)−1 and (48 Å)−1 layer linemaxima and the presence of equal phases at the layer line peaks mirrored along the meridianindicate a Bessel order of 12 with opposite signs, respectively (Figure S1C). Hence, thederived lattice consists of helically related stacked rings, each containing a 12-fold rotationalsymmetry (Figure S1D). This configuration resulted in helical symmetry parameters of 9.6°and 32.4 Å. Because of the possible ambiguity in Bessel order determination, we also testedsingle-particle helical reconstructions with 10, 11, 13, and 14 start helical lattices, but noneof these reconstructions were meaningful (data not shown). In contrast, the diffractionpattern and side views of the 12 start helical lattice model showed maximum correlationwith the diffraction of the experimental image data and their class averages (data notshown). In addition, we also performed a classical Fourier-Bessel 3D reconstruction usingthree selected tubes with good diffraction (Crowther et al., 1996). The resulting 30 Åresolution map is comparable with the low-pass-filtered version of the single-particlereconstruction (data not shown).

Refinement of Helical Symmetry ParametersOptimization of helical symmetry parameters started at a helical rise of 32.4 Å and a helicalrotation of 9.6°. The grid searches covered 25 combinations ranging from 32.0–32.8 Å inhelical rise to 9.2–10.0° in helical rotation (Figures S1F and S1G). For each symmetrycombination, four cycles was sufficient to produce a stable 3D reconstruction with each setof fixed symmetry parameters. We evaluated the cross-correlation between the powerspectra sum of in-plane rotated segments and the simulated power spectrum of the 3Dreconstruction of the particular combination of helical symmetry parameters. The optimalcombination of helical symmetry parameters was determined by a grid search subdividedinto a finer grid until the cross-correlation converged. In order to minimize the influence ofnoise and resolution-dependent amplitude scaling, we correlated resolution ringsindividually between 100 Å and 16 Å and used their average as a final measure(Figure S1G). Finally, a helical rise of 32.67 Å with a rotation of 9.53° produced maximumcorrelation between the sum of experimental power spectra and the simulated symmetry-imposed power spectrum of the 3D reconstruction (Figures S1D and S1E).

Single-Particle Helical Reconstruction of TspO at 10 Å ResolutionTspO tubules of a width of ∼250 Å were segmented into overlapping images of 108 × 108nm using Boxer from the EMAN software suite (Ludtke et al., 1999). We used a step size of32 Å that directly corresponds to the repeat of the stacked rings. The CTF of the electronmicrographs was determined using CTFFIND and subsequently refined by CTFTILT toyield segment-specific CTF parameters (Mindell and Grigorieff, 2003). Further imageprocessing was performed using SPIDER software (Frank et al., 1996). For CTF correction,each segment was convoluted by the corresponding CTF and then subjected to iterativecycles of projection matching. Finally, angles around the helical axis were sampled in 2°intervals and out-of-plane tilt of the tubules was taken into account ±12° at 2° increments.Using the determined alignment parameters, each image segment was inserted 12 times intothe 3D image reconstruction corresponding to the 12-fold rotational symmetry present in thering. Asymmetric units related by the helical symmetry were already included as separateimages on the stack due to the 32 Å step size used during the segmentation procedure. Inaddition to the helical lattice, we exploited the two-fold symmetry present within the TspOdimer. During projection matching, we found that there was no preferred polarity among thesegments of a single tubule. Therefore, we aligned each segment twice with an in-planerotation differing by 180° and included them in the 3D reconstruction. Finally, the volumewas corrected by division of the average 3D value of CTF2 in addition to a Wiener filterconstant of 1% of the maximum amplitude. In order to minimize noise, helical symmetrywas also imposed on the final 3D volume after each cycle. We excluded segments whose

Korkhov et al. Page 10

Published as: Structure. 2010 June 09; 18(6): 677–687.

Sponsored Docum

ent Sponsored D

ocument

Sponsored Docum

ent

(32 Å)−1 ring correlation of its power spectrum with the sum of all in-plane rotated powerspectra was below 0.45. Because images were taken at low underfocus between 0.7 and 2.3μm, the ring correlation at 32 Å is only minimally affected by CTF variation. Excludedsegments with low ring cross-correlation showed weak layer-line diffraction. We includedonly 35%, or 2610, of the 7542 segments in the final 3D image reconstruction of 72 × 72 ×72 nm in size. The resolution of the final map was estimated using Fourier shell correlationcriteria at 0.5/0.143 cut-offs yielding 10.2/7.8 Å, respectively (Figure S3). It should be notedthat the numerical resolution cutoff appeared sensitive to the tightness of the applied mask.In order to avoid overestimation of the resolution, we dilated the structural mask by anadditional 10 Å. The resolution estimate of 10.2 Å corresponds to the detail expected from a10 Å resolution map. More image processing statistics are displayed in Table 1. The finalvolume was corrected for amplitude decay with a B factor of −400 Å2 while applying afigure-of-merit weighting scheme (Rosenthal and Henderson, 2003).

Hand DeterminationWe determined the hand of TspO tubes using one-sided negative staining and helicaldiffraction analysis. Tubes from negatively stained grids showed significant asymmetry intheir layer lines between adjacent quadrants in their power spectrum due to variation of stainthickness across the grid. Frequently, stain covers only the side of the particle facing thecarbon (Klug, 1965; Klug and Finch, 1965). In the case of TMV, an asymmetric layer linewas observed at 69 Å. According to the TMV reference structure (Finch, 1972), the (69Å)−1 layer line corresponds to a Bessel order of 16 with left-handed helicity (Figure S2B).As expected, when imaging the specimen with the carbon side facing the electron gun, thestriations present on the near side give rise to a stronger layer line on the right-hand side ofthe upper quadrant of the power spectrum (Figure S2C). In order to exclude asymmetriceffects arising from different azimuthal views, we performed a control simulation, whichshows that the (69 Å)−1 layer line is not subject to intensity variation during rotation aroundthe helical axis or around the out-of-plane tilt axis (data not shown). The analysis of one-sided TspO tubules yields a (102 Å)−1 layer line, which shows a stronger reflection on theright-hand side of the upper quadrant, thus it also possesses left-handed helicity (FiguresS2E and S2F). Therefore the (48 Å)−1 layer line carries the opposite sign according to thehelical lattice. This asymmetry in the layer line is less obvious than for the (102 Å)−1 layerline. In order to increase the signal of the layer lines and compensate for long-range bending,overlapping segments of the tubes were excised using BOXER (Ludtke et al., 1999), rotatedin the image plane, and Fourier transformed, and their power spectra were averaged. Byanalogy, we performed these tests for the structure of TspO and came to the sameconclusion that the observed asymmetry in the (102 Å)−1 layer line is not due to differentazimuthal or out-of-plane views. We also performed rotary shadowing analysis and tiltingexperiments to try and determine the hand of the structure independently; results for bothmethods were inconclusive. First, no lattice features were revealed by the smooth-surfacedhelix in the platinum shadow. Second, we tested tilting the helix perpendicular to the helicalaxis (Finch, 1972), but, due to the relatively long helical pitch, insufficient asymmetry wasshown by either the left or right side of the projections.

Accession NumbersThe reconstructed TspO map has been deposited at the EMDB databank with accessionnumber EMD-1698.

Supplemental InformationRefer to Web version on PubMed Central for supplementary material.

Korkhov et al. Page 11

Published as: Structure. 2010 June 09; 18(6): 677–687.

Sponsored Docum

ent Sponsored D

ocument

Sponsored Docum

ent

Supplemental InformationRefer to Web version on PubMed Central for supplementary material.

AcknowledgmentsWe would like to thank both R. Henderson and N. Unwin for helpful comments on image processing andmanuscript preparation. The work was supported by core funding from the Medical Research Council and by an ECFP7 grant for the EDICT consortium (HEALTH-201924). V.M.K. was supported by a Federation of EuropeanBiochemical Societies Long-Term Fellowship and C.S. was the recipient of a European Molecular BiologyOrganization Long-Term Fellowship. The authors declare that they have no conflict of interest.

ReferencesAller S.G. Yu J. Ward A. Weng Y. Chittaboina S. Zhuo R. Harrell P.M. Trinh Y.T. Zhang Q. Urbatsch

I.L. Chang G. Structure of P-glycoprotein reveals a molecular basis for poly-specific drug binding.Science 2009;323:1718–1722. [PubMed: 19325113]

Anzini M. Cappelli A. Vomero S. Seeber M. Menziani M.C. Langer T. Hagen B. Manzoni C.Bourguignon J.J. Mapping and fitting the peripheral benzodiazepine receptor binding site bycarboxamide derivatives. Comparison of different approaches to quantitative ligand-receptorinteraction modeling. J. Med. Chem. 2001;44:1134–1150. [PubMed: 11312914]

Bogan R.L. Davis T.L. Niswender G.D. Peripheral-type benzodiazepine receptor (PBR) aggregationand absence of steroidogenic acute regulatory protein (StAR)/PBR association in the mitochondrialmembrane as determined by bioluminescence resonance energy transfer (BRET). J. SteroidBiochem. Mol. Biol. 2007;104:61–67. [PubMed: 17197174]

Boujrad N. Gaillard J.L. Garnier M. Papadopoulos V. Acute action of choriogonadotropin on Leydigtumor cells: induction of a higher affinity benzodiazepine-binding site related to steroidbiosynthesis. Endocrinology 1994;135:1576–1583. [PubMed: 7925120]

Boujrad N. Vidic B. Papadopoulos V. Acute action of choriogonadotropin on Leydig tumor cells:changes in the topography of the mitochondrial peripheral-type benzodiazepine receptor.Endocrinology 1996;137:5727–5730. [PubMed: 8940407]

Braestrup C. Squires R.F. Specific benzodiazepine receptors in rat brain characterized by high-affinity(3H)diazepam binding. Proc. Natl. Acad. Sci. USA 1977;74:3805–3809. [PubMed: 20632]

Braestrup C. Albrechtsen R. Squires R.F. High densities of benzodiazepine receptors in human corticalareas. Nature 1977;269:702–704. [PubMed: 22814]

Butler P.J. Ubarretxena-Belandia I. Warne T. Tate C.G. The Escherichia coli multidrug transporterEmrE is a dimer in the detergent-solubilised state. J. Mol. Biol. 2004;340:797–808. [PubMed:15223321]

Cinone N. Hotje H.D. Carotti A. Development of a unique 3D interaction model of endogenous andsynthetic peripheral benzodiazepine receptor ligands. J. Comput. Aided Mol. Des. 2000;14:753–768. [PubMed: 11131968]

Crowther R.A. Henderson R. Smith J.M. MRC image processing programs. J. Struct. Biol.1996;116:9–16. [PubMed: 8742717]

Dawson R.J. Locher K.P. Structure of a bacterial multidrug ABC transporter. Nature 2006;443:180–185. [PubMed: 16943773]

Delavoie F. Li H. Hardwick M. Robert J.C. Giatzakis C. Peranzi G. Yao Z.X. Maccario J. LacapereJ.J. Papadopoulos V. In vivo and in vitro peripheral-type benzodiazepine receptor polymerization:functional significance in drug ligand and cholesterol binding. Biochemistry 2003;42:4506–4519.[PubMed: 12693947]

Dong C. Beis K. Nesper J. Brunkan-Lamontagne A.L. Clarke B.R. Whitfield C. Naismith J.H. Wza thetranslocon for E. coli capsular polysaccharides defines a new class of membrane protein. Nature2006;444:226–229. [PubMed: 17086202]

Egelman E.H. The iterative helical real space reconstruction method: surmounting the problems posedby real polymers. J. Struct. Biol. 2007;157:83–94. [PubMed: 16919474]

Finch J.T. The hand of the helix of tobacco virus. J. Mol. Biol. 1972;66:291–294. [PubMed: 5038455]

Korkhov et al. Page 12

Published as: Structure. 2010 June 09; 18(6): 677–687.

Sponsored Docum

ent Sponsored D

ocument

Sponsored Docum

ent

Fleishman S.J. Harrington S.E. Enosh A. Halperin D. Tate C.G. Ben-Tal N. Quasi-symmetry in thecryo-EM structure of EmrE provides the key to modeling its transmembrane domain. J. Mol. Biol.2006;364:54–67. [PubMed: 17005200]

Frank J. Radermacher M. Penczek P. Zhu J. Li Y. Ladjadj M. Leith A. SPIDER and WEB: processingand visualization of images in 3D electron microscopy and related fields. J. Struct. Biol.1996;116:190–199. [PubMed: 8742743]

Galiegue S. Casellas P. Kramar A. Tinel N. Simony-Lafontaine J. Immunohistochemical assessment ofthe peripheral benzodiazepine receptor in breast cancer and its relationship with survival. Clin.Cancer Res. 2004;10:2058–2064. [PubMed: 15041726]

Garnier M. Dimchev A.B. Boujrad N. Price J.M. Musto N.A. Papadopoulos V. In vitro reconstitutionof a functional peripheral-type benzodiazepine receptor from mouse Leydig tumor cells. Mol.Pharmacol. 1994;45:201–211. [PubMed: 8114671]

Gavish M. Bachman I. Shoukrun R. Katz Y. Veenman L. Weisinger G. Weizman A. Enigma of theperipheral benzodiazepine receptor. Pharmacol. Rev. 1999;51:629–650. [PubMed: 10581326]

Gouaux E. Mackinnon R. Principles of selective ion transport in channels and pumps. Science2005;310:1461–1465. [PubMed: 16322449]

Han Z. Slack R.S. Li W. Papadopoulos V. Expression of peripheral benzodiazepine receptor (PBR) inhuman tumors: relationship to breast, colorectal, and prostate tumor progression. J. Recept. SignalTransduct. Res. 2003;23:225–238. [PubMed: 14626449]

Henderson R. Cattermole D. McMullan G. Scotcher S. Fordham M. Amos W.B. Faruqi A.R.Digitisation of electron microscope films: six useful tests applied to three film scanners.Ultramicroscopy 2007;107:73–80. [PubMed: 16872749]

Joseph-Liauzun E. Delmas P. Shire D. Ferrara P. Topological analysis of the peripheralbenzodiazepine receptor in yeast mitochondrial membranes supports a five-transmembranestructure. J. Biol. Chem. 1998;273:2146–2152. [PubMed: 9442055]

Klug A. Structure of viruses of the papilloma-polyoma type. II. Comments on other work. J. Mol. Biol.1965;11:424–431. [PubMed: 14290354]

Klug A. Finch J.T. Structure of viruses of the papilloma-polyoma type. I. Human wart virus. J. Mol.Biol. 1965;11:403–423. [PubMed: 14290353]

Korkhov V.M. Tate C.G. Electron crystallography reveals plasticity within the drug binding site of thesmall multidrug transporter EmrE. J. Mol. Biol. 2008;377:1094–1103. [PubMed: 18295794]

Korkhov V.M. Tate C.G. An emerging consensus for the structure of EmrE. Acta Crystallogr. D Biol.Crystallogr. 2009;65:186–192. [PubMed: 19171974]

Lacapere J.J. Delavoie F. Li H. Peranzi G. Maccario J. Papadopoulos V. Vidic B. Structural andfunctional study of reconstituted peripheral benzodiazepine receptor. Biochem. Biophys. Res.Commun. 2001;284:536–541. [PubMed: 11394915]

Lacapere J.J. Papadopoulos V. Peripheral-type benzodiazepine receptor: structure and function of acholesterol-binding protein in steroid and bile acid biosynthesis. Steroids 2003;68:569–585.[PubMed: 12957662]

Li H. Papadopoulos V. Peripheral-type benzodiazepine receptor function in cholesterol transport.Identification of a putative cholesterol recognition/interaction amino acid sequence and consensuspattern. Endocrinology 1998;139:4991–4997. [PubMed: 9832438]

Li H. Yao Z. Degenhardt B. Teper G. Papadopoulos V. Cholesterol binding at the cholesterolrecognition/interaction amino acid consensus (CRAC) of the peripheral-type benzodiazepinereceptor and inhibition of steroidogenesis by an HIV TAT-CRAC peptide. Proc. Natl. Acad. Sci.USA 2001;98:1267–1272. [PubMed: 11158628]

Low H.H. Sachse C. Amos L.A. Lowe J. Structure of a bacterial dynamin-like protein lipid tubeprovides a mechanism for assembly and membrane curving. Cell 2009;139:1342–1352. [PubMed:20064379]

Ludtke S.J. Baldwin P.R. Chiu W. EMAN: semiautomated software for high-resolution single-particlereconstructions. J. Struct. Biol. 1999;128:82–97. [PubMed: 10600563]

Ma C. Chang G. Structure of the multidrug resistance efflux transporter EmrE from Escherichia coli.Proc. Natl. Acad. Sci. USA 2007;104:3668. [PubMed: 17360700]

Korkhov et al. Page 13

Published as: Structure. 2010 June 09; 18(6): 677–687.

Sponsored Docum

ent Sponsored D

ocument

Sponsored Docum

ent

Maaser K. Grabowski P. Sutter A.P. Hopfner M. Foss H.D. Stein H. Berger G. Gavish M. Zeitz M.Scherubl H. Overexpression of the peripheral benzodiazepine receptor is a relevant prognosticfactor in stage III colorectal cancer. Clin. Cancer Res. 2002;8:3205–3209. [PubMed: 12374690]

Maaser K. Hopfner M. Jansen A. Weisinger G. Gavish M. Kozikowski A.P. Weizman A. Carayon P.Riecken E.O. Zeitz M. Scherubl H. Specific ligands of the peripheral benzodiazepine receptorinduce apoptosis and cell cycle arrest in human colorectal cancer cells. Br. J. Cancer2001;85:1771–1780. [PubMed: 11742501]

Masonis A.E. McCarthy M.P. Direct interactions of androgenic/anabolic steroids with the peripheralbenzodiazepine receptor in rat brain: implications for the psychological and physiologicalmanifestations of androgenic/anabolic steroid abuse. J. Steroid Biochem. Mol. Biol. 1996;58:551–555. [PubMed: 8918981]

Mindell J.A. Grigorieff N. Accurate determination of local defocus and specimen tilt in electronmicroscopy. J. Struct. Biol. 2003;142:334–347. [PubMed: 12781660]

Murail S. Robert J.C. Coic Y.M. Neumann J.M. Ostuni M.A. Yao Z.X. Papadopoulos V. Jamin N.Lacapere J.J. Secondary and tertiary structures of the transmembrane domains of the translocatorprotein TSPO determined by NMR. Stabilization of the TSPO tertiary fold upon ligand binding.Biochim. Biophys. Acta 2008;1778:1375–1381. [PubMed: 18420025]

O'Hara M.F. Nibbio B.J. Craig R.C. Nemeth K.R. Charlap J.H. Knudsen T.B. Mitochondrialbenzodiazepine receptors regulate oxygen homeostasis in the early mouse embryo. Reprod.Toxicol. 2003;17:365–375. [PubMed: 12849846]

Ohi M. Li Y. Cheng Y. Walz T. Negative staining and image classification - powerful tools in modernelectron microscopy. Biol. Proced. Online 2004;6:23–34. [PubMed: 15103397]

Papadopoulos V. Boujrad N. Ikonomovic M.D. Ferrara P. Vidic B. Topography of the Leydig cellmitochondrial peripheral-type benzodiazepine receptor. Mol. Cell. Endocrinol. 1994;104:R5–R9.[PubMed: 7821699]

Papadopoulos V. Baraldi M. Guilarte T.R. Knudsen T.B. Lacapere J.J. Lindemann P. Norenberg M.D.Nutt D. Weizman A. Zhang M.R. Gavish M. Translocator protein (18kDa): new nomenclature forthe peripheral-type benzodiazepine receptor based on its structure and molecular function. TrendsPharmacol. Sci. 2006;27:402–409. [PubMed: 16822554]

Papadopoulos V. Lecanu L. Brown R.C. Han Z. Yao Z.X. Peripheral-type benzodiazepine receptor inneurosteroid biosynthesis, neuropathology and neurological disorders. Neuroscience2006;138:749–756. [PubMed: 16338086]

Rone M.B. Liu J. Blonder J. Ye X. Veenstra T.D. Young J.C. Papadopoulos V. Targeting and insertionof the cholesterol-binding translocator protein into the outer mitochondrial membrane.Biochemistry 2009;48:6909–6920. [PubMed: 19552401]

Rosenthal P.B. Henderson R. Optimal determination of particle orientation, absolute hand, andcontrast loss in single-particle electron cryomicroscopy. J. Mol. Biol. 2003;333:721–745.[PubMed: 14568533]

Sachse C. Chen J.Z. Coureux P.D. Stroupe M.E. Fandrich M. Grigorieff N. High-resolution electronmicroscopy of helical specimens: a fresh look at tobacco mosaic virus. J. Mol. Biol.2007;371:812–835. [PubMed: 17585939]

Scarf A.M. Ittner L.M. Kassiou M. The translocator protein (18 kDa): central nervous system diseaseand drug design. J. Med. Chem. 2009;52:581–592. [PubMed: 19133775]

Schagger H. von Jagow G. Blue native electrophoresis for isolation of membrane protein complexes inenzymatically active form. Anal. Biochem. 1991;199:223–231. [PubMed: 1812789]

Smith J.M. Ximdisp—A visualization tool to aid structure determination from electron microscopeimages. J. Struct. Biol. 1999;125:223–228. [PubMed: 10222278]

Taketani S. Kohno H. Okuda M. Furukawa T. Tokunaga R. Induction of peripheral-typebenzodiazepine receptors during differentiation of mouse erythroleukemia cells. A possibleinvolvement of these receptors in heme biosynthesis. J. Biol. Chem. 1994;269:7527–7531.[PubMed: 8125973]

Tate C.G. Kunji E.R. Lebendiker M. Schuldiner S. The projection structure of EmrE, a proton-linkedmultidrug transporter from Escherichia coli, at 7 A resolution. EMBO J. 2001;20:77–81. [PubMed:11226157]

Korkhov et al. Page 14

Published as: Structure. 2010 June 09; 18(6): 677–687.

Sponsored Docum

ent Sponsored D

ocument

Sponsored Docum

ent

Ubarretxena-Belandia I. Tate C.G. New insights into the structure and oligomeric state of the bacterialmultidrug transporter EmrE: an unusual asymmetric homo-dimer. FEBS Lett. 2004;564:234–238.[PubMed: 15111102]

Ubarretxena-Belandia I. Baldwin J.M. Schuldiner S. Tate C.G. Three-dimensional structure of thebacterial multidrug transporter EmrE shows it is an asymmetric homodimer. EMBO J.2003;22:6175–6181. [PubMed: 14633977]

Veenman L. Leschiner S. Spanier I. Weisinger G. Weizman A. Gavish M. PK 11195 attenuates kainicacid-induced seizures and alterations in peripheral-type benzodiazepine receptor (PBR) proteincomponents in the rat brain. J. Neurochem. 2002;80:917–927. [PubMed: 11948256]

Weisinger G. Kelly-Hershkovitz E. Veenman L. Spanier I. Leschiner S. Gavish M. Peripheralbenzodiazepine receptor antisense knockout increases tumorigenicity of MA-10 Leydig cellsin vivo and in vitro. Biochemistry 2004;43:12315–12321. [PubMed: 15379570]

Wendler G. Lindemann P. Lacapere J.J. Papadopoulos V. Protoporphyrin IX binding and transport byrecombinant mouse PBR. Biochem. Biophys. Res. Commun. 2003;311:847–852. [PubMed:14623258]

Yeliseev A.A. Kaplan S. A sensory transducer homologous to the mammalian peripheral-typebenzodiazepine receptor regulates photosynthetic membrane complex formation in Rhodobactersphaeroides 2.4.1. J. Biol. Chem. 1995;270:21167–21175. [PubMed: 7673149]

Yeliseev A.A. Kaplan S. A novel mechanism for the regulation of photosynthesis gene expression bythe TspO outer membrane protein of Rhodobacter sphaeroides 2.4.1. J. Biol. Chem.1999;274:21234–21243. [PubMed: 10409680]

Yeliseev A.A. Kaplan S. TspO of rhodobacter sphaeroides. A structural and functional model for themammalian peripheral benzodiazepine receptor. J. Biol. Chem. 2000;275:5657–5667. [PubMed:10681549]

Yeliseev A.A. Krueger K.E. Kaplan S. A mammalian mitochondrial drug receptor functions as abacterial “oxygen” sensor. Proc. Natl. Acad. Sci. USA 1997;94:5101–5106. [PubMed: 9144197]

Zeng X. Kaplan S. TspO as a modulator of the repressor/antirepressor (PpsR/AppA) regulatory systemin Rhodobacter sphaeroides 2.4.1. J. Bacteriol. 2001;183:6355–6364. [PubMed: 11591680]

Zhang M.R. Maeda J. Ogawa M. Noguchi J. Ito T. Yoshida Y. Okauchi T. Obayashi S. Suhara T.Suzuki K. Development of a new radioligand, N-(5-fluoro-2-phenoxyphenyl)-N-(2-[18F]fluoroethyl-5-methoxybenzyl)acetamide, for pet imaging of peripheral benzodiazepinereceptor in primate brain. J. Med. Chem. 2004;47:2228–2235. [PubMed: 15084121]

Korkhov et al. Page 15

Published as: Structure. 2010 June 09; 18(6): 677–687.

Sponsored Docum

ent Sponsored D

ocument

Sponsored Docum

ent

Figure 1.Comparison of Sequence and Topology of the Human and Bacterial TspO Homologues(A) Alignment between the amino acid sequences of human TSPO (hTSPO) and TspO fromR. sphaeroides (rsTspO). Bars underneath the sequences represent hydrophobic regions thatare predicted by hydropathy analysis to form transmembrane domains.(B) Cartoon of the predicted topology of R. sphaeroides TspO based on hydropathy analysisand the “positive inside” rule.

Korkhov et al. Page 16

Published as: Structure. 2010 June 09; 18(6): 677–687.

Sponsored Docum

ent Sponsored D

ocument

Sponsored Docum

ent

Figure 2.Dimeric TspO Is Expressed in the E. coli Inner Membrane(A) TspO protein expression was confirmed by Coomassie blue staining of the SDS-PAGEgel. TspO was enriched in a small-scale batch NiNTA purification procedure in DDM (laneTspO) and compared with control cells (lane pET). DDM-purified TspO is shown forcomparison (lane TspO).(B and C) Fractionation of the TspO-expressing BL21(DE3) cell membranes. Lane OMcorresponds to the outer membrane; lanes IM1–IM3 correspond to the inner membrane band(IM2) and the flanking fractions IM1 and IM3 that also contained inner membrane material.The gel in (B) was stained with Coomassie blue, whereas (C) is a western blot of anidentical gel probed with an anti-poly His antibody to identify the position of recombinantTspO.(D and E) Blue native PAGE of purified TspO and purified EmrE. The asterisk indicates themonomeric species formed upon addition of 1% SDS for 10 min prior to the addition of theloading dye (+SDS). As a control, a standard dimeric membrane protein of a comparablesize (15.2 kDa including the C-terminal tag), EmrE, is shown. Two and three asterisksindicate dimeric and multimeric protein species, respectively. If TspO is concentrated tobelow 5 mg/ml, the multimeric band is negligible (E).(F) Size exclusion chromatography profile of TspO in DDM.

Korkhov et al. Page 17

Published as: Structure. 2010 June 09; 18(6): 677–687.

Sponsored Docum

ent Sponsored D

ocument

Sponsored Docum

ent

Figure 3.Binding of Substrates to TspOTspO purified in DDM was titrated with increasing amounts of PPIX (A), hemin (B), andcPPIII (C) and the tryptophan fluorescence was measured at 340 nm. For comparison, themeasurements using the nontransported molecule D-aminolevulinic acid (ALA) is shown in(D). The tryptophan fluorescence emitted at 340 nm was plotted against the concentration ofthe quenching reagent to obtain a dose-response curve. The insets in each graph show theraw spectra in the 300 to 400 nm range. The data shown are the representative dose-dependence profiles of the tryptophan fluorescence quenching by the indicated substances;

Korkhov et al. Page 18

Published as: Structure. 2010 June 09; 18(6): 677–687.

Sponsored Docum

ent Sponsored D

ocument

Sponsored Docum

ent

the number of experiments for PPIX, hemin, and cPPIII were three, two, and two,respectively.

Korkhov et al. Page 19

Published as: Structure. 2010 June 09; 18(6): 677–687.

Sponsored Docum

ent Sponsored D

ocument

Sponsored Docum

ent

Figure 4.Cryo-EM Analysis of the TspO Helical Crystals(A) An image of negatively stained TspO tubes and small vesicles formed on TspOreconstitution; the scale bar corresponds to 100 nm.(B) Cryo-EM of vitrified TspO tubes on a holey carbon grid. The stacked disk arrangementof the well-preserved tubes is evident from unprocessed images; the scale bar corresponds to50 nm.(C) Scatter plot of length and width of lipid structures present in the ice-embedded sample.The 25 nm wide tubes constitute two thirds of the sample and vesicles make up the

Korkhov et al. Page 20

Published as: Structure. 2010 June 09; 18(6): 677–687.

Sponsored Docum

ent Sponsored D

ocument

Sponsored Docum

ent

remainder as shown in the pie chart (top inset). The distribution of tube lengths ranged from100 to 1000 nm (bottom inset).(D) Comparison of an unprocessed cryo-EM image of a tube segment (left) with an averageof the aligned image segments of the same azimuthal views (center) and the matchedprojection from the 3D image reconstruction, which was convoluted by its correspondingCTF (right). The top and bottom row differ in their azimuthal view by a rotation of 16°about the tube axis.(E) Width profiles correspond to the images displayed above (left and center). The pixelrows perpendicular to the tube axis were added to yield an averaged width profile of thetubes, which shows that all included segments contain tubes with a width of 25 nm (left).Averaged projection classes give rise to a less noisy profile (center).(F) Three-dimensional surface presentation of a TspO segment reconstructed at 10 Åresolution.

Korkhov et al. Page 21

Published as: Structure. 2010 June 09; 18(6): 677–687.

Sponsored Docum

ent Sponsored D

ocument

Sponsored Docum

ent

Figure 5.Grayscale Density Slices of the TspO 3D Reconstruction(A) A slice of density through the helical tube with the inset depicting the orientation of the4.8 Å thick slices depicted in (B)–(K).(B–G) TspO viewed parallel to the membrane plane; slices of density were madeperpendicular to the helical axis of the tube, at various distances from the center of the dimer(D), including −9.6 Å (B), −4.8 Å (C), 4.8 Å (E), 9.6 Å (F), and 14.4 Å (G).(H–K) The TspO dimer viewed perpendicular to the membrane plane; radial slices ofdensity were taken through the membrane plane at positions −7.2 Å (H), −2.4 Å (I), +2.4 Å(J), and +7.2 Å (K) with respect to the center of the bilayer, with + being toward the externalsurface of the tube. The density interpreted as one molecule of TspO has been circled.

Korkhov et al. Page 22

Published as: Structure. 2010 June 09; 18(6): 677–687.

Sponsored Docum

ent Sponsored D

ocument

Sponsored Docum

ent

Figure 6.Interpretation of the Experimental Density Map(A) A view perpendicular to the membrane plane of the TpsO dimer with α helices fittedinto the density map by eye. Red and blue α helices represent individual monomers of TspOand they are labeled arbitrarily a–e and a′–e′.(B) View parallel to the membrane plane of the TspO dimer and after a 40° rotation to showthe highly tilted helices e and e′.(C) View perpendicular to the membrane plane.(D) TspO monomer viewed parallel to the membrane plane, from the perspective of thedimer interface formed by helices a and b and a′ and b′.(E) TspO monomer viewed from the lipid bilayer. The experimental density maps in all thepanels was contoured at 1.5 σ.

Korkhov et al. Page 23

Published as: Structure. 2010 June 09; 18(6): 677–687.

Sponsored Docum

ent Sponsored D

ocument

Sponsored Docum

ent

Sponsored Docum

ent Sponsored D

ocument

Sponsored Docum

ent

Korkhov et al. Page 24

Table 1

Statistics of Helical Reconstruction

Image Processing Statistics

Resolution FSC at 0.5/0.143 (Å) 10.2/7.8

Total length of nonoverlapping segments (nm) 38,894

Number of tubes 205

Number of segments/finally included 7,542/2,610

Segment size (nm) 108

Size of 3D reconstruction (nm) 72

Segment step size (nm) 3.2

Helical rise (Å) 32.67

Helical rotation (°) 9.53

Pixel size on the specimen (Å) 1.20

Number of asymmetric units 62,640

Published as: Structure. 2010 June 09; 18(6): 677–687.