Activated Microglia in the Brain: Mitochondrial and Cell Membrane-Associated Targets for Positron Emission Tomography

The emission tomographic imaging of activated microglia in the brain moves into the focus of neuroscientific research with increasing recognition of contributions of early inflammatory processes to neurodegenerative, traumatic, cancerous and infectious diseases of the brain. Whereas the mitochondrial isoform of the 18 kDa translocator protein (TSPO1) has been the main cellular target for positron emission tomography (PET) of this type of cells for decades, alternative marker proteins in the plasma membrane of microglia challenge efforts in ligand development, recently. The present report includes PET approaches using the chemokine receptor CX3CR1 and the FR2 folate receptor in parallel to small molecule PET tracers available for in vivo visualization of the “classical” target TSPO1. It compares first and second generation of TSPO1 ligands as well as new compounds like the tetrahydrocarbazole [ 18 F]GE-180 and the quinazoline [ 11 C]ER176 presumed to reduce polymor-phism-related inter-subject variations, with allosteric ligands for the chemokine receptor CX3CR1 and with radio labelled folate conjugates targeting the folate “cargo” receptor FR1 and the FR2 receptor characteristic for anti-inflammatory M2 microglia.


Introduction
The imaging of microglia as a marker of inflammation in traumatic, degenerative, neuropsychiatric and infectious brain diseases has been already for years a surrogate method for the monitoring of cerebral disorders [1]. The resting/surveying microglia in healthy tissue (resident macrophages of the brain) communicates with adjacent neurons and glia cells [2]. Cross talk between microglia and neurons is presumed to play an important role for synaptic plasticity [3] [4] [5] [6] [7]. Microglia can be activated by diverse molecules released during pathological alterations and reacts adopting different activated phenotypes [2] [8] [9] [10] [11]. Recently, microglial activation is regarded as one of the initial factors turning early acute metabolic and structural alterations into chronic processes.
For immunological approaches, such polarization of the functionality of microglia corresponds to the actions of the subsets of T helper cells (CD4 cells), Th1 and Th2 lymphocytes releasing the respective Th1-type and Th2-type cytokines as pro-inflammatory or anti-inflammatory response [10] [15] [16].
Microglia accounts for 5% -10% of total glia cells of the human brain [10] [16] [17] and up to 20% in other mammalian species [18] [19] with lower percentage in the cortex and higher in corpus callosum. 30% -50% of cells in glioma are microglial cells; and glioma and astrocytoma are the most common brain tumors [20].
Investigations of typical microglial cells are limited in vitro, because the source of this cell population in the brain is, as widely accepted, the yolk sac releasing a wave of primitive macrophages as progenitor cells of microglia during the early embryonal development [21] [22] [23] [24]. Microglial population colonizes the whole CNS and maintains itself by local proliferation [23] [24].
A postnatal supplement of the resident cerebral microglial population by macrophages from the circulation is not very probably but also not completely out of discussion [24] [25]. As the normal functional state of microglia is accepted the resting (surveying) state [5].
Currently, a multitude of structural markers is known allowing immunohistological phenotyping of microglia of different shapes, functions and locations [26] in vitro or ex vivo. For in vivo imaging, one of the main targets investigated with PET or SPECT in microglial cells has been for many years mitochondrial TSPO1 (translocator protein 18 kDa), the former peripheral benzodiazepine binding sites (PBR) [27], overexpressed in activated microglia. The target TSPO1 is a protein of the outer mitochondrial membrane [31] [32], more precisely, it is located at the contact sites between the outer and inner membrane [33]. Functional tasks of TSPO1 have been presumed in porphyrine transport, protein import, cholesterol transport, steroid biosynthesis, ion transport, cell proliferation and differentiation [34] [35] [36]. TSPO is found, especially, in cells with steroid synthesis. The evolutionary conserved protein is constituent of species from bacteria to mammals upstairs [37]. A 3D-crystal structure of bacterial as well as of mammalian TSPO has been pub-

TSPO as PET Target
Already, during the 70 ties and 80 ties of the last century was recognized the 18 kDa translocator protein (TSPO; originally classified as peripheral benzodiazepine receptor = PBR) [32] is not restricted to peripheral organs, but also present in CNS, glial and ependymal cells. Although the densities of TSPO in the brain is a magnitude below that in kidney, heart, testis, ovary and uterus its increased expression in pathophysiological conditions like multiple sclerosis (MS), amyotrophic lateral sclerosis (ALS) and Alzheimer's disease (AD) suggested a suitability as a biomarker of early stages of pathogenesis in inflammatory diseases or inflammation-related disorders [60] [61].
TSPO2 has been presumed also in nuclear membranes or close to nuclear membranes and in endoplasmic reticulum [35]. In contrast to TSPO1, it lacks in birds and mammals the isoquinoline-binding site [42].
For evaluation of potential structure-activity-relationships (SAR) of TSPO1, close relations can be important with further mitochondrial proteins involved in energy production and mitochondrial homeostasis as part of the mPTP complex Even if the role of mPTP is in part controversial [67] and not completely understood, it is presumed to be a key structure in several pathophysiological processes related to reperfusion injury, amyotrophic lateral sclerosis, Alzheimer's disease, alteration following traumatic injury and myopathies [67] [68].
Currently, the first reports on the 3D structure of TSPO1 [40] promise also new insight on the oligomeric structures of the protein and potential relationships to other structures of the mitochondrial membrane.
Some TSPO ligands are supposed to bind rather to the multi protein complex than to single sub-proteins forming it.
An interesting contribution to understanding of subcellular compartmentation of TSPO has been provided, recently, by Yasin et al. [72] with investigations in U118MG glioma cells. The first generation-TSPO ligand PK11195 is known to interact with TSPO1 by opening of the transition pore. The authors [72] demonstrated PK11195 prompts the cell to dislocate the binding places of the ligand from localizations widely spread within the cell (i.e. to mitochondria) to such located close to the nucleus [72]. This suggests a role of TSPO also in the promotion of interactions between mitochondria and nucleus by an influence on subcellular compartmentation of mitochondria.   than with unlabelled Ro5-4864 [78]. The change from benzodiazepine lead structure to isoquinoline and quinoline promised markedly higher specificities.
2) Non-benzodiazepine, first generation ligands The isoquinoline R-PK11195 [ Figure 2; ( [81]. However, yet relatively high non-specific binding and low signal-to-noise ratio, as well as high plasma protein binding and low brain permeability have been described as drawbacks also for the in vivo observations with the compound [73] [74]. PK14105 [ Figure 2, (9)], a structurally similar candidate suitable as 18 F-labelled tracer with promising properties, was early ranged in minor position, especially, because of somewhat weaker binding affinity [82] [83].
In spite of drawbacks in specificity, clinical investigations with [ 11 C]PK11195 revealed some interesting items on inflammatory processes in brain diseases.
Zhang [73] reviewed investigations using [ 11 C]PK11195 in frontotemporal dementia which is a neurodegenerative disease of people <65 years of age. The observations suggested accumulation in brain regions contralateral to the atrophic site of the brain and indicated early activations of microglia with potency to start neurodegenerative processes. Zhang [73] regarded the early activa-

Second Generation Compounds
The show lower non-specific binding than the first generation compounds [ Moreover, the inter-subject variations of these ligands are higher than for the first generation [72]. Responsible for these variations can be the rs6971 poly-   Figure 3, (14)]. They found the [ 18 F]FEPPA accumulation in AD patients with a two-compartment model significantly higher in grey matter than in white matter. The HABS identified genotypically showed higher binding than MABs in all ROIs [90].
Possibly, the low size of the groups allowed no further evaluation and differentiation of HAB, MAB and LAB participants. The authors classified [ 18 [88] whereas in the reports by Suridjan [89] [90] and Mabrouk [88] were included also people with severe cognitive impairment.
With this compound started a generation of non-azepine tri and tetracyclic compounds of TSPO ligands.
GE-180 was subjected to first clinical safety tests in healthy volunteers, recently [36] [97]. The racemate was analysed for the functional roles of its enantiomers, kinetics of uptake, elimination as well as metabolization [97]. The enantiomer S-GE-180 [ Figure 4, (19)] was shown to be the pharmacodynamically most suitable structure in the tricyclic carbazole class, with good metabolic stability in comparison to its R-enantiomer. With a log D of 2.95 at pH 7.4 the substance showed an appropriate lipophilicity in preclinical investigations. The authors recommended a reversible two-tissue-compartment model for the imaging of TSPO. A study using a one-compartment model as well as two-compartment models for this tracer in healthy volunteers was published in the same year by Feeney et al. [97]. Feeney et al. [97] used the cortical grey matter as pseudo-reference region based on observations by Owen et al. [98]. The latter described a negligible binding of [ 11 C]PBR28 in this region using the Lassen plot method with the purine derivative emapunil/XBD173 as TSPO inhibitor [99]. The results of these tests showed a relatively low uptake of [ 18   Ikawa et al. [102] reported that with the measurements with ER176, similarly like also for other tracers, differences in binding affinities related to genotypic subtypes were not revealed in control groups. One reason for the absence of dif-   In difference to other chemokine receptors and TSPO, is known only one endogenous CX3CR1 ligand.

CX3CR1 Chemokine Receptor
The CX3CR ligand FKN is a 373 a.a.r. membrane-bound polypeptide [49], which is anchored in the cell membrane and supports the scavenging of leukocytes. The extracellular domain can be separated by tumor necrosis factor α converting enzyme (TACE) and is then present as a globular, extracellular polypeptide. By this way FKN is supposed to act as signaling peptide with inflammatory and homeostatic action as well.
The search for potential small molecule ligands of CX3CR1 has been realized by docking studies in homology models using the cytomegalovirus-encoded chemokine receptor homolog US28 [50], however, to date with limited success.
Maciejewski-Lenoir et al. [111] demonstrated in rat cell cultures expression/ release of FKN by astrocytes and neurons but not by microglia cells.
The role of CX3C receptors in microglia is not completely clear, today, although it, without doubt, plays a role in the diapedesis of cells. The information about a toxic or protective role are similar ambivalent like for microglial cells [110].
Steen et al. [112] made the promiscuity of chemokine receptors for several peptide ligands responsible for the difficulties to develop effective small-molecule therapeutics targeting chemokine receptors. With regard to only one endogenous ligand, the CX3CR1 receptor provides better prerequisites than the most other chemokine receptors.

Small Molecule Ligands of CX3CR1
The use of microglial plasma membrane proteins as markers of activation or World Journal of Neuroscience proliferation earns not only attention because it could be applied to patients which are PET-negative due to TSPO polymorphisms.
A well accessible marker protein of the plasma membrane could allow also kinetics and dynamics of potential ligands with better feasibility for imaging purposes than an intracellular target like TSPO1.
The small molecule ligands reported as potential PET tracers for CX3CR1, to In 2015, Mease et al. [52] provided synthesis and first data of a 18 F-labelled PET tracer using AZD8797 as parent compound. FBTTP [ Figure 6; (29)] is the fluorine-18 labelled AZD8797 derivative [52] with a K i of 23 nM and 18fold selectivity  showed a 80fold selectivity in comparison with CXCR2 receptors at a K i (vs. [ 125 I]fractalkine) of 5.8 ± 0.4 nM.
Finally, there are several high-affinity small molecules with good selectivity to CX3CR1. The lipophilicity > 3.5 could be an obstacle for the use of these substances as PET tracers. However, the compounds are regarded as potentially useful drugs in multiple sclerosis and provide lead structures for development of potential PET tracers with better pharmacokinetic properties.

Folate Receptors
Antifolates play a key role in therapies of cancer and rheumatic diseases since many decades. Folic acid is an essential part of the carbon-one metabolism. Because it is not synthesized in mammalian cells, it has to be obtained as Vitamin B9 dietary compound.
Four membrane proteins have been described to be involved into the uptake of folate [ Figure 7, (31)] and of 5-methyltetrahydrofolate (5-MTHF) occurring in the liver after intestinal uptake of folic acid [115].
The proton-coupled folate transporter (PCFT) [116], highly expressed in small intestinum, is responsible for uptake of folate into the epithelium at low pH. have been described [120], which should be of relevance when folic acid decreases to nanomolar concentrations in blood serum [118]. Homologies between the receptor isoforms are high with only small differences in the pharmacodynamically relevant structures. However, FR3 is a soluble receptor and, recently, a role of FR1 was described as a transcriptionally relevant protein, which may interact with nuclear structures following internalization [121] [122].
Blockage of the cerebral folate receptors and of the uptake across the blood brain barrier is known to cause marked folic acid deficiency in the brain [123] resulting in epileptic symptoms, which can be, potentially, reversed by folic acid application.
FR1, is localized in epithelial cells, kidney, lung and also in the choroid plexus.
It is the most widely distributed subtype of the folate receptors. However, its World Journal of Neuroscience  density is low, with exception of high expression in some tumors [117]. In general, folate receptor expression is typical for cells with functions in embryonic development [58].
FR2 is expressed in the cells of the hematopoetic system, in placenta, in spleen and thymus [117] [120] and has been mentioned as overexpressed with some specificity in anti-inflammatory macrophages/microglia [124] [125]. World Journal of Neuroscience In consequence, folate receptors advanced to candidates for use in targeted therapy and imaging approaches, for instance, as a cargo system for folate conjugates with cytotoxic substances in the treatment of tumors [54] [55] [56] [57] [58] and to potential markers of glial activation in inflammatory diseases.

Potential FR2 Ligands for PET Tomographic Imaging
To date, most efforts for the development of ligands at FR2 receptors suitable for radio-imaging have been reported on SPECT tracers [126]. One report potentially interesting also for the analysis of cerebral microglia has been provided by Jager et al. [126]  FR2 enhancement is involved in the pathogenesis of arthritis but also of many other inflammatory diseases e.g. Crohn disease, diabetes, lupus erythematodes, glomerulonephritis [117] [127].
Overexpression of FR2 in the prefrontal cortex of rats was demonstrated, recently, in a model of acute and chronic stress [48] in parallel with other M2 microglial markers.
Brain uptake of folic acid, 5MTFH and methotrexate was first suggested by the results of a study by Spector and Lorenzo 1975 [128]. Since this time, it is known choroid plexus binds the major part of folic acid achieving its cells.
Spector and Lorenzo [128] observed 5MTHF can also leave the cells of the choroid plexus in direction of the cerebrospinal fluid. Release of folate from FR1 protein is possible by interaction with the PCFT [129]. The key role of FR1 receptor became clear also from investigations on autoimmune diseases due to antibody blockade of FR1. FR ligands/substrates for PET are available, similar to SPECT tracers, predominantly, as diagnostics for tumors [55] [56] [130] where, however, especially the FR1 receptor plays a key role, while FR2 has been proposed rather as a target for imaging in arthritis and rheumatic diseases [54] [127]. Gent et al. [54] [54].
Synthesis of folate conjugates using diverse prosthetic groups was the method of choice, already years ago, because direct fluorination of folic acid was not possible.
Bettio et al. [130] synthesized a folate tracer using 4-fluorobenzylamine (FBA) as prosthetic group with region selective linkage to the α-and to the γ-carbonyl group of folic acid [ Figure 7, (31) [136] which is regarded with a K i of 1.8 nM as a promising PET tracer for tumors. If folate-based tracer can be efficient also in monitoring microglial activation in the brain with intact blood brain barrier remains to be elucidated.

Summary and Conclusions
Visualization of microglial activities as a key factor in pathogenesis and progres-  [137] is increasingly included in the development of new in vivo-strategies of brain imaging.
The target preferred in clinical investigations remains, currently, the well-known mitochondrial TSPO1. But during the last decade also proteins of plasma membrane became potential binding sites for PET tracer.
Overexpression of TSPO in brain tissue subjected to inflammatory diseases has been confirmed by postmortem studies [93] as well as in animal experiments.
Three generations of small molecule TSPO ligands have been proposed as PET tracers: 1) azepines and quinolones; 2) predominantly, compounds with three separate non-azepine and non-isoquinoline ring systems 3) tri-and tetracyclic compounds. The third generation of TSPO ligands was expected to reduce the inter-subject variations caused by the presence of three different binding types of study participants due to the rs6971 polymorphism. erties with exception of a low uptake into the brain. In first clinical tests with healthy volunteers, no differences between the binding types were observed.
However, an influence of the various genotypes in patients is yet not excluded.
The broader clinical experience with second generation compounds suggests that the tracers among these with the best properties, like [ 18 F]FEPPA or [ 11 C]PBR28 and its fluorinated analogue remain on the schedule of clinical trials [138] [139] [140].
New insights on protein-ligand interactions and protein-protein relationships in the mitochondrial membrane can be expected now also from molecular dynamics studies using the crystal structure of TSPO [38] [39] [40]. This should offer not only new information on the physiological and pathophysiological role of TSPO in inflammatory diseases but also on potential protective mechanisms in infectious diseases like malaria [66].
For the PET targets on the plasma membrane, CX3CR1 and FR2, polymorphisms of relevance for the imaging results, similar to that of second generation ligands of TSPO, are not known, currently. CX3CR1 ligands proposed recently by Karlström et al. [109] showed appropriate binding affinities and selectivity. Their high lipophilicity was a tribute to their original purpose as an orally applicable therapeutic in multiple sclerosis. With regard to the application as a PET tracer it is rather a drawback. AZD8797 as well as related derivatives open a door for further modifications of potential lead structures.
Hitherto, small molecule ligands or folate conjugates binding to FR1 and FR2 specifically are not available. A PET labelling of folate receptors in inflammatory processes becomes specific for the FR2 subtype only by the distribution of the FR2-expressing M2 subtype of activated microglia.
On the other hand, distribution of FR1 and FR2 as well as their contributions to the cerebral uptake of folate through the epithelial cells of the choroid plexus and from the ventricular or spinal cerebrospinal fluid raise new questions also for potential imaging applications apart from tumor diseases. Additionally, the time course of such processes and their role in comparison to a direct transfer of folate, folate conjugates and analogues across the blood vessels would be an important item for the estimation of a potential diagnostic value of FR ligands in clinical PET imaging methods. Moreover, the recognition that the permeability of the blood brain barrier increases with rising age earns some attention [141].