Structure and activity of tryptophan-rich TSPO proteins

See allHide authors and affiliations

Science  30 Jan 2015:
Vol. 347, Issue 6221, pp. 551-555
DOI: 10.1126/science.aaa1534

Structural clues to protein function

Translocator protein (TSPO) is a mitochondrial membrane protein thought to transport cholesterol and porphyrins. Its detailed function remains unclear, but interest in it is high because TSPO is involved in a variety of human diseases. Two papers now present crystal structures of bacterial TSPOs. Li et al. show that a mutant that mimics a human single polymorphism associated with psychiatric disorders has structural changes in a region implicated in cholesterol binding. Guo et al. suggest that TSPO may be more than a transporter. They show how it catalyzes the degradation of porphyrins, a function that could be important in protection against oxidative stress.

Science, this issue p. 555, p. 551


Translocator proteins (TSPOs) bind steroids and porphyrins, and they are implicated in many human diseases, for which they serve as biomarkers and therapeutic targets. TSPOs have tryptophan-rich sequences that are highly conserved from bacteria to mammals. Here we report crystal structures for Bacillus cereus TSPO (BcTSPO) down to 1.7 Å resolution, including a complex with the benzodiazepine-like inhibitor PK11195. We also describe BcTSPO-mediated protoporphyrin IX (PpIX) reactions, including catalytic degradation to a previously undescribed heme derivative. We used structure-inspired mutations to investigate reaction mechanisms, and we showed that TSPOs from Xenopus and man have similar PpIX-directed activities. Although TSPOs have been regarded as transporters, the catalytic activity in PpIX degradation suggests physiological importance for TSPOs in protection against oxidative stress.

The translocator protein (TSPO) is named for its putative roles in the transport of cholesterol, proteins, and porphyrins into mitochondria and elsewhere (1). TSPO was first identified as a peripheral-type benzodiazepine receptor in mammals (2)—attracting attention because of popular anxiolytic benzodiazepines such as diazepam (Valium)—and the tryptophan-rich sensory protein TspO of photosynthetic bacteria proved to be homologous (3). The sequences of TSPO proteins from diverse organisms are highly conserved in sequence (fig. S1) and also in function; notably, rat TSPO can substitute for the bacterial protein as a negative regulator of photosynthesis gene expression in Rhodobacter sphaeroides (4).

Benzodiazepine-like chemicals, such as PK11195, were found to bind to TSPO and regulate steroid biosynthesis (5), and porphyrins were identified as endogenous ligands of rat TSPO, with highest affinity for protoporphyrin IX (PpIX) (6). Many following studies have pursued TSPO interactions with cholesterol and PpIX (7, 8). Most recently, the seeming essentiality of TSPO for steroidogenesis (9) has been challenged by the finding that TSPO1−/− mice are viable and without defects in steroid hormone biosynthesis (10), and the presumed role in PpIX transport (11) has been amended by the discovery that bacterial TSPOs catalyze a photooxidative degradation of PpIX (12).

Although the underlying biochemical mechanisms for its activities are not fully understood, TSPO has elicited considerable interest for medicine. Roles for TSPO have been imputed in apoptosis (13), inflammation (14), HIV biosynthesis (15), cancer (16), Alzheimer’s disease (17), and cardiovascular diseases (18). TSPO is a therapeutic target (13, 1820) and has proven useful as a positron emission tomography–imaging biomarker (14, 17, 21, 22).

The translocator protein has been characterized in monomer, dimer, and higher oligomeric states (20). An electron microscopy structure for TSPO from R. sphaeroides (RsTSPO) revealed a dimer (23), and a nuclear magnetic resonance structure for a mouse TSPO showed a monomer comprising five transmembrane helices (24).

We produced recombinant TSPO proteins from several bacterial and vertebrate organisms and found Bacillus cereus TSPO (BcTSPO) suitable for biochemical and structural characterization (see supplementary materials and methods). By size-exclusion chromatography, detergent-extracted BcTSPO exists in at least three different oligomeric states (fig. S2A). Apo BcTSPO crystallized from the monomer-dimer fraction (fig. S2, C and D) in two different lipidic cubic phase (LCP) conditions and also from the high-oligomer fraction as a detergent micelle; the BcTSPO-PK11195 complex crystallized as a dimer in LCP. A structure obtained by single-wavelength anomalous diffraction analysis from iodinated BcTSPO (fig. S2E) was the starting point for native structures that followed (tables S1 and S2). Refinements yielded structures at 1.7 Å (Fig. 1, A to E, and fig. S3) and 2.0 Å resolution for apo monomers in LCP, at 4.1 Å resolution for the apo dimer in detergent (Fig. 1F), and at 3.5 Å resolution for the dimeric ligand complex in LCP (Fig. 1G).

Fig. 1 Structures of BcTSPO.

(A and B) Ribbon drawing of TSPO from the type 1 apo structure, as viewed from within the membrane plane (A) and from the periplasm above (B). Coloring progresses spectrally from the N terminus (blue) to the C terminus (red). Side chains are shown for conserved tryptophan residues Trp31, Trp40, Trp51, and Trp138. (C and D) Electrostatic potential at the molecular surface of the apo monomer, as oriented in (A) and (B), respectively. Positive potentials are in degrees of blue; negatives are in red. Dotted planes show membrane boundaries computed by the Positioning of Proteins in Membranes (PPM) server. (E) Molecular surface as in (D) but colored by Consurf sequence conservation. Variable positions are in degrees of turquoise; conserved positions are in degrees of maroon. (F) Ribbon drawing of the apo dimer in detergent. Tryptophan residues and coloring are as in (A). (G) Ribbon drawing of the PK11195 dimer in LCP. The PK11195 moieties are drawn as pink sticks.

Each BcTSPO protomer has five transmembrane (TM) α helices organized in clockwise order (TM1-TM2-TM5-TM4-TM3), as viewed from the C-terminal end (Fig. 1, A and B, and fig. S1) where a short extra-membranous helix (α1,2) between TM1 and TM2 caps a C-terminal pocket (Fig. 1C). The electrostatic potential surface defines a transmembrane orientation (Fig. 1C) having its N-terminal end heavily positive, likely cytoplasmic in B. cereus, and its C-terminal end largely negative (Fig. 1D and fig. S4, A to D). A mapping of sequence conservation onto the molecular surface shows a conservative band between TM2 and TM5, a highly conserved pocket opening between TM1 and TM2, and features more conserved at the C-terminal end than on most of the exterior (Fig. 1E and fig. S4, E to H).

All four crystallized copies of BcTSPO are very similar: The two apo monomers have a root-mean-squared deviation (RMSD) of 0.64 Å for Cα positions, differing most in the TM1-TM2 loop, and protein conformations are essentially identical whether with or without the PK11195 ligand (Cα RMSDs between the 1.7 Å resolution apo monomer and subunits A and B of the PK11195 dimer are 0.51 and 0.59 Å, respectively). Dimeric associations in the detergent micelle and in LCP are also nearly the same (Fig. 1, F and G), although the apo micelle may be a swapped dimer. The associated chains are cleanly resolved in the liganded complex.

In comparing our structures of BcTSPO to that published for the mouse protein Mus musculus TSPO1 (MmTSPO1) (24), we found limited similarity apart from the topological order of helices. Comparing all 148 residues in common between the two models (fig. S1; 26.6% identity), the Cα RMSD equals 4.90 Å (PDB IDs 4RYQ versus 2MGY); when the comparison is restricted to the 54 positions that match within 3 Å, the RMSD value is still exceptionally high at 1.71 Å. TM2 and TM5 can be superimposed relatively well, but the other helices are rotated such that conserved residues that are inside in BcTSPO move outside in MmTSPO1 (fig. S5A). Having the sequence redisposed to the MmTSPO1 model also removes pocket-lining conservation from interior surfaces (fig. S5C). We also compared BcTSPO to recent structures of R. sphaeroides TSPO (RsTSPO) (25) and here found substantial similarity. On superimposition of those 126 of 148 Cα atoms within 3 Å of one another for BcTSPO versus RsTSPO Ala139→Thr139 (A139T) (26) (PDB IDs 4RYQ versus 4UC1), the RMSD value was 1.17 Å, with TM1 helices most disparate (fig. S5B). TM1 is the least conserved helix (fig. S1), and it is at the dimer interface of RsTSPO but not at that of BcTSPO (fig. S5D).

Both protomers of the BcTSPO-PK11195 dimer showed omit-map density in the pocket opening between TM1 and TM2 (fig. S2F). When fitted by the ligand structure, the resulting model has the carbonyl-oxygen atom of PK11195 hydrogen bonded to indole-NH groups of both Trp51 and Trp138, and the Cl atom of the ligand is in van der Waals contact with Asn87 (Fig. 2A). PK11195 in BcTSPO also makes van der Waals contacts with residues Ser22, Tyr32, Pro42, Ile47, Phe55, Phe90, Ser91, Gln94, Cys107, Ala142, and Leu145. These contacts emanate from each of the transmembrane helices plus the TM1-TM2 loop. All among them except Ser22 are widely conserved residues. Because the apo and ligated BcTSPO proteins are essentially isostructural, the ligand-binding site is a highly conserved cavity in apo BcTSPO (Fig. 2B). This cavity is water-filled in our high-resolution apo structures (fig. S6A), and PpIX can be satisfactorily docked into the cavity in a particular orientation (fig. S6, B and C). The PK11195 conformation, ligand binding pose, and protein contacts are very different in the model for MmTSPO1 (fig. S5A).

Fig. 2 Features of the ligand binding site in BcTSPO.

(A) PK11195 in the binding pocket. The ligand is drawn as a stick figure with atom coloring of carbon (light gray), nitrogen (blue), and oxygen (red). Portions of a ribbon drawing of the TSPO protein structure are displayed faintly with coloring as in Fig. 1A. Side chains that contact PK11195 are drawn in stick representation and have carbon atoms colored as for the backbone. (B) Cross-sectioned views of the ligand binding pocket. The surface of an apo monomer, oriented as in Fig. 1A and sliced by a vertical plane perpendicular to the page, was opened as a book to show both sides of the invaginated pocket. The planar cut surfaces are gray, and exposed molecular surfaces and visible atomic structures are in ConSurf coloring as in Fig. 1E.

The side chain of Ala142 protrudes into the binding pocket such that its mutation would be expected to interfere with ligand binding. In accord, recent reports show that TSPO radioligands PBR28 (21) and [18F]-FEPPA (22) have different binding affinities for the predominant Ala147 and the minor Thr147 polymorphic variants of human TSPO1 (position 142 in BcTSPO).

In attempting to bind PpIX to BcTSPO, we noticed that initially red mixtures turned blue on standing out in the light for a few minutes, but not when dark. Thus, considering the reported photooxidative degradation of PpIX by Chlorobium tepidum TSPO (CtTSPO) (12), we characterized this activity of BcTSPO toward PpIX. As for CtTSPO, we observed a rapid decrease of the Soret absorbance at 405 nm when in the light (Fig. 3A) and a concomitant loss of the characteristic PpIX fluorescence at 629 and 697 nm (632 and 700 nm for BcTSPO-PpIX) when illuminated by ultraviolet (UV) light (27, 28). These changes were irreversible, and saturated PK11195 substantially inhibited the PpIX decay but did not entirely block it.

Fig. 3 Spectral analysis of bacterial TSPO-mediated activity in PpIX degradation and modulation.

(A) UV-visible spectra of WT BcTSPO with PpIX before and after reaction. AU, absorbance units. (B) Comparison of the spectrum for the bilindigin degradation product as extracted from the postreaction spectrum in (A), with spectra from biliverdin and phycocyanobilin. (C and D) Fluorescence analysis of WT BcTSPO activity toward PpIX. Specta after indicated light exposures are shown in (C), and time courses of fluorescence at the 632-nm ground-state peak are tracked in (D). Exposure time is measured in light pulses, where each pulse comprised 50 s light plus 10 s read-out in the dark. Time courses are compared for PpIX in the indicated associations. (E and F) Fluorescence analysis of BcTSPO W138F activity toward PpIX. Spectra and time courses are as described for (C) and (D). (G and H) Fluorescence analysis of BcTSPO W51F activity toward PpIX. Spectra and time courses are as described for (C) and (D). Error bars in (D), (F), and (H) indicate SD.

The post-illumination absorbance spectrum of BcTSPO-PpIX is like that reported from CtTSPO-PpIX (12), and the spectrum that we extracted for the photo-degradation product shows features like those of spectra from biliverdin and phycocyanobilin, but with absorption peaks further blue-shifted in series (Fig. 3B). By analogy with biliverdin and for its indigo color, we call this product bilindigin. Because the photooxidation of PpIX yields oxidized vinyl groups, predominantly generating formyl groups when in aqueous micelles (29), and the modifications that distinguish phycocyanobilin from biliverdin are also at vinyl positions, we contemplate a formyl analog of biliverdin as a plausible product candidate (fig. S7). Thus, TSPO might catalyze both the vinyl-to-formyl oxidation and also the oxidative cleavage of PpIX at the methene bridge between vinyl-bearing pyrrole rings.

A mechanistic clue for TSPO-mediated degradation of PpIX comes from the observation that the W→F mutation of a conserved tryptophan was said to abrogate PpIX degradation by CtTSPO without affecting PpIX binding (12). In our BcTSPO-PpIX model (fig. S6C), both the corresponding residue, Trp138, and another conserved tryptophan, Trp51, are poised for potential interactions with the scissile methene bridge and with the vinyl groups. To examine the possible roles of these residues in PpIX degradation, we compared the reaction for the wild type with reactions for the W138F, W51F, W51F/W138F variants, and we also tested A142T because of its relation to the polymorphic variation at the analogous position (A/T147) in Homo sapiens TSPO1 (HsTSPO1). We used the sensitive PpIX fluorescence spectrum (27, 28) to monitor the changes.

The prominent fluorescence features of free PpIX were shifted slightly and quenched appreciably (~30%) upon interaction with wild-type (WT) BcTSPO, and these features disappeared within minutes during light exposure (Fig. 3, C and D). Preincubation with saturated PK11195 slowed the decay of PpIX greatly (Fig. 3D) but did not block it entirely. As for CtTSPO, the W138F mutant of BcTSPO greatly reduced the catalytic degradation of PpIX compared with the WT protein (Fig. 3, E and F); however, in this case, the degradation was not abolished altogether. For the W51F mutant, we saw no evidence of the product formed by WT and W138F BcTSPO. Instead, we observed a rich fluorescence spectrum (Fig. 3, G and H) with new peaks growing at 649 and 673 nm as the primary peak at 632 nm decreased. We attribute the new peaks to secondary excited states, probably due to incipient photooxidation (29). These states are intrinsic to PpIX (fig. S8, A and B) but stabilized by TSPO (Fig. 3G and fig. S8C) and reversed on PpIX dissociation (fig. S8, D and E). We conclude that Trp51 is essential for PpIX cleavage and that Trp138 participates in the reaction.

Neither the W51F/W138F double mutant nor the A142T showed any PpIX cleavage, but each generated reversible excitation at 673 nm as the primary 632-nm fluorescence decreased (Table 1 and fig. S9, A to D). To test for the greater generality of TSPO-mediated PpIX reactions, we produced various eukaryotic TSPO homologs. Among these we have been able to extract and purify three vertebrate proteins: Xenopus tropicalis TSPO (XtTSPO), HsTSPO1 A147T, and HsTSPO2. From the sequences (fig. S1) considered in light of our analysis of BcTSPO properties, we expected that XtTSPO should degrade PpIX similarly to WT BcTSPO but that the other two would show aberrant behavior. Indeed, XtTSPO degraded PpIX, as monitored by our fluorescence assay, whereas both HsTSPO1 (Thr147) and HsTSPO2 (Thr145) showed growth of 673-nm fluorescence at the expense of 632-nm decay but no irreversible degradation (Table 1 and fig. S9, E to J). Because the latter behavior is as for BcTSPO A142T, we infer that having threonine at position 142 precludes cleavage-appropriate binding of PpIX in each of them. Protein instability complicated our analysis of reaction rates for the vertebrate TSPOs.

Table 1 Reactions of PpIX with TSPO proteins.

Down arrows denote the decrease with time in intensity of photostimulated fluorescence from the first excited state. Up arrows denote the increase with time of fluorescence from secondary excited states, presumably reached by oxidative photochemistry processes.

View this table:

What is the chemical basis for TSPO-mediated PpIX degradation? The intrinsic photochemistry of PpIX involves the generation of singlet oxygen 1O2 and other reactive oxygen species (ROS) upon photostimulation (2730), and it is evident from our observations that excited states of PpIX are stabilized when bound to TSPOs. Additionally, ROS interactions with tryptophan residues can generate tryptophan radicals, which are essential for the oxidation of aromatic substrates by lignin peroxidases (31). We suggest that such tryptophan radicals at Trp51 and Trp138 (BcTSPO numbering) may be responsible for TSPO-mediated oxidation of PpIX. Radical interactions may work at a distance, and conserved Trp31 and Trp40 may also play a role (fig. S10). Although ROS come from light in our in vitro experiments, ROS-generating cellular processes may be responsible in vivo.

What physiological role might PpIX degradation serve? Porphyrins such as PpIX are tightly controlled because they can generate toxic ROS, especially when excited by light. An incisive analysis of the role of TSPO in such control came in studies of oxidative stress in the moss Physcomitrella patens (32). Mitochondrial PpTSPO1, which has catalytic residues the same as in BcTSPO (fig. S1), was shown to be induced by oxidative stress from the inhibition of mitochondrial respiration, and knockout lines showed an accumulation of PpIX and ROS. In other studies, porphyrins exerted toxic effects on liver in the dark, generating hydroxyl radicals (33). Biliverdin is an oxyradical scavenger and is cytoprotective (34). We suggest that TSPO-mediated degradation of PpIX to bilindigin both reduces ROS generation and also promotes ROS neutralization.

Supplementary Materials

Materials and Methods

Figs. S1 to S10

Tables S1 and S2

References (3553)

References and Notes

  1. Single-letter abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.
  2. Acknowledgments: We thank members of the Hendrickson laboratory, especially J. Lidestri for LCP instrumentation, O. Clarke for thought-provoking discussion, K. Hu from Stuyvesant High School for help in collecting fluorescence data, and O. Clarke and J. Qin for help in crystallographic computing. We also thank J. Schwanof and R. Abramowitz at National Synchrotron Light Source (NSLS) beamlines X4A and X4C for help in measuring diffraction data, M. L. Hackert of the University of Texas at Austin for insightful suggestions related to phycocyanobilin, A. Palmer of Columbia University for advice on spectroscopic interpretation, and S. Ferguson-Miller of Michigan State University for sharing coordinates ahead of publication. This work was supported in part by NIH grant GM095315 and GM107462 to W.A.H. X4 beamlines were supported by the New York Structural Biology Center at the NSLS of Brookhaven National Laboratory, a U.S. Department of Energy facility. The crystallographic data reported here are deposited in the Protein Data Bank with identification codes listed in table S2. Y.G. and W.A.H. designed research, analyzed data, and wrote the paper; Y.G. performed experiments; E.K. and B.R. performed bioinformatics analyses; B.K. and R.B. performed expression tests; R.C.K. and R.B. overexpressed eukaryotic proteins; Q.L. participated in diffraction analysis; and C.G. participated in the design of activity assays.
View Abstract

Navigate This Article