Lost in translocation: The function of the 18 kD translocator protein

Abstract

Recent knock-out studies of the 18 kDa-translocator-protein (TSPO) report normal or latent phenotypes, raising doubts about the protein’s purported role in steroidogenesis and the validity of earlier data, notably in regard to the specificity and selectivity of TSPO-binding drugs.

Here, we summarize available protein structural data and based on these suggest regulatory functions of TSPO and its putative endogenous ligands.

We illustrate with examples the inherent limits of inference in loss-of-function studies and discuss the challenge in defining gene ‘functions’ that are emergent properties of complex interactions.

To the call to reduce the risk of late-stage failure in clinical trials through better characterization of target/off-target effects of TSPO-binding drugs, now fascilitated by the use of TSPO null-background animals, we add a call for a more differentiated understanding of the ill-defined term ‘neuroinflammation’ that currently underpins the argument for TSPO as a diagnostic or therapeutic target

Introduction

The evolutionary highly conserved 18 kDa translocator protein (TSPO, mTSPO), previously named peripheral benzodiazepine receptor (PBR), is an abundant, nuclear-encoded mitochondrial protein found in many organs but with particularly high constitutive expression in steroidogenic tissue as well as in tissue affected by an active pathological process, such as observed in acute or chronic brain disease. The scientific publication trends before and after the introduction of the new nomenclature for the TSPO/PBR (hereafter referred to TSPO according the new nomenclature) as well as the conceptual confluence of the TSPO’s “translocation” function with that of “neuroinflammation” has previously been reviewed by Liu et al., 2014 [1].

Significant efforts are being made to develop new drugs that exploit the diagnostic and therapeutic potential of TSPO. These efforts were predominantly based on the proposed role of TSPO in steroid biogenesis, and the dynamic increase of protein abundance in “neuro-inflammatory” disease conditions, including Alzheimer Disease, Multiple Sclerosis and anxiety disorders [2–4]. For diagnostic purposes, TSPO radioligands have been found to be generically useful as a biomarker of active disease or disease-related tissue remodeling [1]. From a therapeutic perspective, modulation of (neuro-)steroid production in microglia and subsequent protective effects on neurons mediated by steroids were the primary rationale to combat “neuro-inflammatory” disease with TSPO ligands.

Yet, recent studies of TSPO gene knock-out models [5–7] have not readily confirmed a role of the TSPO in steroid synthesis. While thus one important premise of the mechanism of action of TSPO targeting drugs has come to be questioned [8], the utility of TSPO null-background animals to ascertain the selectivity TSPO/PBR-binding drugs could be demonstrated. For example, in the GuwiyangWurra (‘FireMouse’) TSPO knock-out mouse model, in vivo radioligand imaging using micro-PET, receptor membrane binding and autoradiography confirm the selectivity of TSPO targeting drugs, such as the isoquinoline 1-(2-chlorophenyl)-N-methyl-N-(1-methylpropyl)-3-isoquinolinecarboxamide (PK 11195). This prototypical TSPO ligand is one of the original compounds that defined peripheral benzodiazepine binding sites, and was more recently used to achieve the stable TSPO conformation necessary to solve its three-dimensional structure.

Likewise, a chimeric tumor model in which a syngeneic TSPO -expressing brain tumor grows in the null-background of the GuwiyangWurra TSPO knock-out strain has been used to show the selectivity the newer TSPO-binding compounds, the imidazopyridines CLINDE and PBR11, as well as the specificity of an antibody against TSPO [6] (Fig. 1).

Figure 1.

While GuwiyangWurra (‘FireMouse’) Tspo^-/- knockout mice display apparently similar phenotypes compared to control animals, in vivo imaging with PET/CT using the radioligand [¹⁸F]PBR111 demonstrates the complete absence of any TSPO binding sites in the knockout (the small visible signals originate from gut and urinary bladder, i.e. the excretory pathways only; green circle: adrenal gland) (modified after [6]).

More detailed investigations on microglia from the GuwiyangWurra TSPO knock-out strain show changes in mitochondrial ATP production, a latent phenotype that at first sight does not seem to affect microglial activation. The possible implications of a role of TSPO in affecting mitochondrial energy metabolism are discussed below (see text boxes 2 and 3).

Box 2. Studying endogenous ligands as a window into TSPO function.

The biological purpose of heme, porphyrin or cholesterol binding to TSPO remains largely undefined. Many intracellular metabolites are well known to feedback on homeostatic mechanisms of metabolic control, for example on transcription through chromatin modifiers [46, 47] or on enzyme function through allosteric effects [48, 49]. Studying the functional relationship of heme binding to TSPO may yield valuable insights into physiological processes that rely on TSPO.

Heme is known to act in a classical feedback loop that limits its own biosynthesis through binding of the nuclear receptor REV-ERBα, a key regulator of cellular energy metabolism[50]. Levels of free heme have to be tightly regulated based on energy substrate flux, as excess heme can cause oxidative stress and damage cells when reacting with molecular oxygen. Here, we propose that binding of heme to TSPO is part of a larger feedback network that regulates mitochondrial energy metabolism and heme homeostasis, and that complements the nuclear signaling of heme through REV-ERBα (Figure 3). Heme and porphyrin catabolism have been proposed as the anciently conserved function of TSPO [51]. Whether the recently reported regulation of ATP production by TSPO is an independent function of TSPO, or is connected to heme signaling is currently unclear. Defining the role of cholesterol and heme as feedback molecules that regulate mitochondrial metabolism through TSPO may guide the way to a more dynamic and context-dependent understanding of TSPO biology.

Box 3. An alternative metabolic function of TSPO/PBR?

Recent in vivo studies in three different species suggest a link between TSPO and the regulation of mitochondrial energy metabolism. Combining the strengths of several animal models will help to test this emerging concept.

	Tspo-/- knockout mice are viable and fertile without an obvious phenotype. However, microglia cells derived from Tspo-/- mice show a lower baseline of mitochondrial respiration and reduced ATP production [6]. Thorough metabolic phenotyping of Tspo-/- mice will determine a role of TSPO in mitochondrial energy metabolism. Furthermore, Tspofl/fl mice allow the cell-type specific deletion of TSPO in tissues of interest. Caution has to be applied with respect to the specificity of Cre driver strains as well as mosaic or incomplete knockdown effects [52, 53].
	Zebrafish are amenable for in vivo drug screening. A recent screen has identified the synthetic TSPO ligands PK 11195 and Ro5-4864 as enhancer of a fasting response, and to protect mice against glucose intolerance and hepatic steatosis [54]. Targeted disruption of the tspo gene, the CRAC domain or single nucleotides that are important for synthetic or endogenous ligand binding will yield insights into gene function and pharmacology.
	Drosophila are amenable to dietary, genetic and pharmacological interventions. A recent study shows that loss of dTSPO extends life span, is protective against neurodegeneration and impairs mitochondrial cellular respiration [55]. An important advantage of Drosophila as a research tool is that a genome-wide gene knockdown repository is available, which will allows one epistasis experiments and the dissection of gene networks that are associated with metabolic functions of TSPO.

Structure of the 18 kDa translocator protein

Early structural predictions identified that TSPO contains five transmembrane helices on the basis of its primary sequence [9]. Topological analysis further showed that the C-terminus of TSPO is exposed to the cytoplasm such that two loops are found on each side of the membrane [10]. According to hydropathy analysis the longest loop is present between transmembrane helices (TM) TM1 and TM2 and is located in the cytoplasm. An overall helical structure of TSPO was supported by circular dichroism of TSPO from mouse in combination with nuclear magnetic resonance (NMR) studies on peptide fragments corresponding to the five putative transmembrane regions [11, 12]. Evidence for transmembrane helices was also provided by electron microscopy of the tryptophan-rich sensory protein (TspO) from Rhodobacter sphaeroides (RsTspO) [13]. Two RsTspO molecules each comprising five transmembrane helices were arranged into a homodimeric arrangement, in agreement with the pronounced tendency of RsTspO to assemble into dimers [13, 14]. Based on the observed dimeric arrangement a substrate translocation pathway was suggested.

Despite the presence of five putative transmembrane regions, TSPO appears to have an unstable tertiary fold. This is supported by low signal dispersion in NMR spectra of TSPO in detergent micelles [11, 15] and a less well defined electron density for some of the TM helices of RsTspO in tubular helical crystals [13]. In contrast, binding of the synthetic radioligand (PK 11195) stabilized the structure of TSPO and increased the helical content [11, 15]. Stabilization of the TSPO conformation by (R)-PK 11195 then made it possible to solve the three-dimensional structure of the TSPO-PK 11195 complex solubilized in dodecyl phosphocholine micelles [15]. The structure revealed a bundle of five TM helices that are arranged around the ligand PK 11195 (Fig. 2). Due to the presence of several proline residues most of the helices have kinks. When viewed from the cytosol, the five TM helices are arranged in the clockwise order TM1-TM2-TM5-TM4-TM3. NMR-based dynamics measurements further showed that the five TM helices are rigidly formed, while residues 1-4 and 160-168 of TSPO remain flexible and can potentially change their conformation upon interaction with endogeneous interaction partners. The ligand binding pocket comprises many conserved residues including an alanine at position 147, which is substituted by threonine in case of the rs6971 polymorphism [16]. Although the affinity of PK 11195 is similar for both polymorphs of TSPO, several synthetic second generation ligands have a reduced affinity when a threonine residue is present at position 147. The PK 11195-binding pocket is further closed by the TM1-TM2 loop which folds into a short helix and thus assumes a stable conformation (Fig. 2). Several small molecules were shown to compete with PK11195 for binding to TSPO (for a review see [4]), highlighting the importance of the hydrophobic pocket and the TM1-TM2 loop for recognition of small molecules. In complex with PK11195 the five TM helices tightly pack together, such that no channel in the interior of a TSPO monomer was detected [15]. On the other hand, molecular dynamics simulations suggested that a single TSPO molecule could accommodate a cholesterol molecule in an interior cavity when PK11195 is not present [17].

Figure 2.

Three-dimensional structure of mouse TSPO (PDB code: 2MGY; [15]). The TSPO structure is formed by five transmembrane helices (TM) in complex with PK 11195 (shown in brown). The topology of the five transmembrane helices is TM1-TM2-TM5-TM4-TM3 when viewed from the cytosol [15]. The highly positively charged C-terminus (top right), which contains the cholesterol binding site, is flexible and points into the cytosol.

Interaction of TSPO with cholesterol

After successful recombinant production of mouse TSPO and reconstitution into liposomes, TSPO was shown to bind cholesterol with nanomolar affinity [18]. The specific nature of the cholesterol-TSPO interaction was further supported by site-directed mutagenesis [19]; i.e. amino acids Y153 and R156 were identified to be critical for the interaction with cholesterol. Y153 and R156 are located in a sequence motif at the C-terminal end of transmembrane helix 5, which was denoted as cholesterol recognition amino acid sequence (residues A147 to S159) [19]. Residues A147 to S159 are helical in the three-dimensional structure of TSPO in complex with (R)-PK 11195 (Fig. 2) [15]. A cholesterol molecule can be docked there such that R156 may interact with the sterol hydroxyl group of cholesterol [20]. In the three-dimensional structure of the TSPO-PK 11195 complex, the side chains of Y153 and R156 point to the hydrophobic environment of the membrane, suggesting that cholesterol might bind from the outside to the TSPO structure. For a transport mechanism, the high affinity of the cholesterol-TSPO interaction would require additional, currently unknown mechanisms to release cholesterol from the protein.

Mammalian TSPO belongs to the family of tryptophan-rich sensory proteins (TspO), which is also found in bacteria. For example RsTspO is important for photosynthetic gene expression and might act as an oxygen sensor [21]. Bacterial TspO and mammalian TSPO share sequence similarity with residues in the TM1 region being least conserved [22]. Consistent with these sequence differences, bacterial TSPO and mammalian TSPO differ in important aspects. While mammalian TSPO is functional as a monomer and is predominantly monomeric when reconstituted into detergent micelles, bacterial TspO has a strong tendency to dimerize [13]. Moreover, the affinity of RsTspO for PK 11195 is about 1000-fold lower when compared to the mammalian protein [14]. In addition, the cholesterol recognition sequence of the mammalian protein is not conserved in bacterial TspO, consistent with the absence of cholesterol from bacterial membranes. This highlights the need to obtain further structural insight into mammalian TSPO as many of the functional questions cannot be addressed by studies with bacterial TspO proteins.

Conceptual issues

While, here, we cannot fully explore the broader issue of how specific functions are attributed to specific genes [23, 24] in order to convert “data into knowledge and knowledge into understanding” [25], it is important to realize that the controversy surrounding earlier reports on the embryonic lethal out-come of a TSPO knock-out and the subsequent observations to the contrary often contains implicit assumptions, that impact on experimental design and data interpretation or can even impede progress in a field [26].

What does it mean to be old and non-essential?

It is a popular assumption that ancient, evolutionary highly conserved genes should naturally be vital and ‘essential’ in the regulation of well-definable, fundamental physiological process, hence one should expect embryonic lethal outcomes or at least an obvious phenotype from the loss of such a gene, as has been reported for TSPO[27].

However, phylostratigraphic analysis of the human genome suggests that ancient genes predominate among the disease-associated, but “non-essential”, genes [28, 29]. At the same time, there are indications that disease genes might be more frequently involved in stress response pathways, i.e. complex multi-pathway processes for which it might currently be difficult to assign a function in simple physiological terms [30].

Thus, while the evolutionary dynamics behind the preservation of the ancient TSPO is likely to be more complex than presently understood [31], an importance as a disease-modifying gene, regulated in cellular stress or conditions of pathological challenge or adaptation should be considered.

Current evidence in humans is sparse but intriguing, in as far as the known rs6971 polymorphism in the TSPO gene [32] may relate to complex mental health phenotypes, notably to different metabolic responses as seen in differences in weight gain under anti-psychotic treatment [33].

The part, the whole and the emergence of function

Loss-of-function studies, such as gene knockouts or through use of modulators with a single site of action, are classical attempts at explanatory reductionism. However, the assumption, for example, that an unimpaired knockout phenotype automatically refutes the importance of the removed gene for its posited function in the normal wild-type is not borne out empirically. Text box 1 discusses some examples of normal or only moderately altered in vivo-phenotypes in laboratory mice lacking proteins thought to be essential for mitochondrial function.

Box 1. Recent data from Tspo^-/- knockout mice challenge an old dogma – or not?

Recent publications of viable Tspo^-/- knockout mice by independent groups challenge the long standing assumption that TSPO is critical for life: Tspo^-/- mice develop normally and live a healthy laboratory existence. A physiologically relevant role of TSPO in cholesterol transport, steroid biosynthesis or mitochondrial permeability transition pore regulation could not be found [5, 6, 36]. These startling findings call for a thorough re-investigation of previously postulated TSPO functions. However, caution is warranted to omit “established” functions prematurely. Many examples exist of broadly-expressed proteins that biochemically appear to critically impact on mitochondrial function in cell culture settings, and that present at first sight with surprisingly mild in vivo phenotypes in laboratory mice. For example, mice deficient for the mitochondrial NAD⁺-dependent deacetylase SIRT3 are viable, fertile and are, for most of their lives, indistinguishable from wildtype littermates, despite de-acetylating and functionally modifying many enzymes of beta-oxidation and cellular oxidative damage response pathways. However, under cold exposure Sirt3^-/- mice rapidly lose the capacity to maintain the body temperature in a healthy range due to defects in β-oxidation, a defect that would likely render them incompatible with life in the wild [37]. Besides cold-intolerance, Sirt3 mice develop phenotypes during aging, such as hearing-loss or hepatocellular cancer [38, 39].

An example of a disease relevant gene whose deficiency in mice does not phenocopy the human pathophysiology is PTEN-induced putative kinase 1 (PINK1). PINK1 mutations are the second most common cause of autosomal-recessive Parkinson’s Disease(PD) [40]. PINK1 is thought to play a major role in pathways of mitochondrial quality control and metabolism, and often serves as a paradigmatic example of a primary mitochondrial defect that leads to neurodegenerative disease. However, the loss-of-function phenotypes in mice are mild, include defects in mitochondrial bioenergetics in the striatum in the young, and are followed by a broader deficit in the aged brain. Only subtle morphological changes of mitochondrial shape could be observed. Importantly, drosophila, but not mice, show prominent defects in mitochondrial dynamics, bioenergetics, loss of dopaminergic motor neurons and PD-like motor deficits [41–43]. With hindsight, the history of PINK1 research shows similarities to where TSPO research is now; i.e. much of the initial work hypotheses had been generated in immortalized cell lines, and the field then was challenged by the difficulty to assign PINK1 functions in the face of moderate phenotypes in mice[44].

In addition to differences between species, the discrepancy between an important biochemical function in vitro and the in vivo phenotype is caused by a high capacity of mitochondria to adapt and sustain tissue functionality in response to cellular stress [45]. TSPO functions in steroid production or other unknown functions may become phenotypically relevant when mitochondria lose the capacity to compensate during aging or in response to other stressors. Refining assays, and to modulate TSPO function(s) in mice and model organisms will be key to detect and explain phenotypes that may surface during stress conditions or aging.

A more fundamental reason for the limits of inference in gene knock-out studies is the fact that simple explanatory reductionism ignores that ‘functions’ generally emerge from more complex interaction between different proteins. In other words and with specific reference to the Tspo^-/- knockout models, for as long as it is not established whether and how the absence of TSPO results in adaptive changes, it is difficult to rule out that at least some the TSPO emergent functions in normal animals include regulatory effects on steroid synthesis. In the context of emergent properties of microglia, one of the cell types characterized by high abundance of inducible TSPO, Svahn et al. (2014) [34] have made the case that emergence at its core means that ‘the interaction is an entity of the system’ and thus cannot be meaningfully reduced into other entities. As complement to the now ongoing loss-of-function studies, gain-of-function studies should be of value, i.e. an approach where the interactions are not removed but enhanced or altered.

Finally, observations in Guwiyang Wurra Tspo^−/− mice after peripheral nerve injury [6] have revealed no obvious failure in the activation of perineuronal microglia typically found around the somata of the injured neurons and determined by the up-regulation of microglial CD11b (Mac -1 α, Mβ2 integrin, also known as the C3 complement receptor), a highly sensitive and commonly used method to detect microglial ‘activation’ and from that postulate the presence of ‘neuroinflammation’. Thus, the state change of microglia (‘activation’) does not appear to critically depend on the actions of the TSPO. It remains to be established whether the absence of the TSPO impairs the further transformation of microglia into inflammatory macrophages as the altered mitochondrial function in isolated Tspo^−/− microglia might suggest [6].

Notwithstanding the need for more detailed experimentation, it is becoming clear that similar to the term ‘translocation’, the terms ‘activation’ of microglia and ‘neuroinflammation’ have moved away from their earlier pragmatic definitions, usually confined to a specific context, to have become ill-defined, lumping terms to be covaried with yet more similarly complex phenomena, such as pain or certain behavioural states.

Outlook

The future research agenda for the TSPO needs to clarify terminologies, such as “translocation” and “neuroinflammation” [1, 35], and make use of the evolving conceptual development and tools of systems biology.

The basis for this, however, is the availability and sharing of tools and protocols for laboratory or clinical experimentation that allow independent replication, such as TSPO binding compounds, antibodies and others with comprehensively validated specificity and selectivity.

Highlights.

1. The protein structure of TSPO is commensurate with a role in cholesterol transport and cell metabolism

2. Non-essentiality despite evolutionary conservation of gene does not imply an absence of important regulatory influence, and may be relevant as a disease- or disease-modifying gene.

3. TSPO null-background animals are useful in addressing questions about the target selectivity and off-target effects TSPO-binding drugs prior to clinical trials

Outstanding questions.

• What is the structure and dynamics of mammalian TSPO in the absence of a ligand?

• Is TSPO able to adopt its conformation to different environments, oligomerization states and interaction partners?

• What is the structural basis for the influence of the rs6971 polymorphism on the ligand interaction of TSPO?

• Does TSPO form functional complexes with other proteins?

• Would a fundamental role in mitochondrial energy metabolism also explain effects on steroid metabolism?

• What is the consequence of decreased ATP production for microglial proliferation and “neuroinflammatory” function of activated microglia cells?

• Which TSPO compounds have off-target effects or are off-target effects by other drugs targeted at different receptors explained by action via the TSPO?

• Which TSPO binding molecules, peptides or antibodies are specific and selective enough for robust experimentation and clinical trials?

• What are the minimum standards for regarding a TSPO ligands as a validated target?

Figure 3. Heme is a feedback molecule that regulates mitochondrial energy metabolism and heme biosynthesis in a classical feedback loop.

A) Intracellular concentrations of heme must be tightly regulated to sustain oxidative metabolism in mitochondria during fasting, while limiting heme-mediated oxidative damage. Heme is an endogenous ligand of REV-ERBα, a nuclear receptor that acts as a transcriptional suppressor of gluconeogenesis, lipid metabolism (not depicted) and heme biosynthesis, and that is a critical component of the circadian clock core machinery that adapts cellular metabolism to circadian rhythms[56, 57]. The biological role of the binding of heme to TSPO is currently unclear, although a role in porphyrin degradation to reduce mitochondrial reactive oxygen species (ROS) production has been proposed [51]. Here, we speculate that TSPO plays a critical role in energy state-dependent heme metabolism, and that binding of heme to TSPO could constitute a nuclear transcription-independent, signaling pathway to directly modulate mitochondrial metabolism. In this model, similar to REV-ERBα being a heme sensor for transcriptional control of energy balance in the nucleus, TSPO binds free heme or porphyrins. Based on this interaction, mitochondrial ATP production is reduced through unkown mechanisms, while TSPO-catalyzed degradation of heme and porphyrins reduces ROS toxicity. B) When the demand for heme increases during fasting, the transcriptional co-regulator PGC1α induces expression of Alas1, which encodes the rate-limiting enzyme of heme biosynthesis, as well as that the gluconeogenic genes Pck1 and G6pc in the liver [58, 59]. In addition to the heme - REV-ERBα -nuclear axis, low levels of free heme may activate TSPO and release ATP production in mitochondria. Thus, the binding of heme to TSPO may integrate mitochondrial metabolism with nuclear gene transcription to regulate cellular energy balance and to reduce oxidative damage. Genetic lack of TSPO in mouse microglia or drosophila leads to decreased ATP production [6, 55]. Furthermore, treatment of fasted mice or zebrafish with synthetic TSPO ligands enhances the activation of a fasting gene signature in the liver, including increased expression of Alas1, Pck1 and G6pc in the liver [54]. Dashed arrows indicate hypothetical links, whereas continuous lines represent experimentally established connections. Pgc1a, peroxisome proliferator-activated receptor γ coactivator 1 alpha; G6pc, glucose-6-phosphatase; Pck1, phosphoenolpyruvate-carboxykinase; Alas1, Delta-aminolevulinate synthase 1.

Glossary

Bioenergetics

Cellular processes that are implicated in the transformation of energy; most commonly refers to ATP-generation by carbon breakdown and oxidative phosphyralation in mitochondria.

NMR spectroscopy

a spectroscopic method based on nuclear magnetic resonance that allows determination of the structure and dynamics of proteins

PK 11195

the synthetic radioligand 1-(2-chlorophenyl)-N-methyl-N-(1-methylpropyl)-3-isoquinolinecarboxamide, which binds mammalian TSPO with nanomolar affinity

Protein dynamics

Proteins are not rigid entities but are flexible and can respond to different external stimuli by changing their conformation

Radioligand

Small synthetic molecules that are used for positron emission tomography studies

TSPO/PBR

the mammalian 18 kDa translocator protein, which was previously identified as a binding site for benzodiazepines.

TspO

family of bacterial tryptophan-rich sensory proteins

TM

Region of a protein that crosses the hydrophobic environment of a membrane