Resolving the cellular specificity of TSPO imaging in a rat model of peripherally-induced neuroinflammation

Highlights • [18F]DPA-714 PET showed an increased brain signal following peripheral LPS challenge.• The cellular origin of TSPO signal appears to depend on the brain region examined.• TSPO signal increase in the hippocampus arises from microglia and astrocytes.• Microglia, macrophages and astrocytes are the main contributors in the substantia nigra.• Macrophages and microglial cells expressing TSPO are distinguished by RNAscope.


Introduction
Neuroinflammation, the activation of inflammatory pathways in the brain and central nervous system (CNS), plays a complicit role in several neurological and psychiatric disorders, and prolonged activation of microglia, the resident immune cells in the brain, is likely associated with pathological progression (Guzman-Martinez et al., 2019). The proinflammatory role of microglia has been widely documented in the past decade due to the availability of imaging biomarkers designed to detect in vivo brain inflammation. Most of these biomarkers belong to a family of positron emission tomography (PET) radiotracers that bind to an 18kD Translocator protein (TSPO) located on the outer mitochondrial membrane of various cells, with a relatively high expression in microglia (Turkheimer et al., 2015). Studies using TSPO PET radiotracers have shown increased TSPO binding, corresponding to increased neuroinflammation, in the brain of patients with Alzheimers disease (AD), Parkinson disease (PD), stroke, traumatic brain injury (TBI), and multiple sclerosis (MS), amongst others (Dimitrova-Shumkovska et al., 2020;Nutma et al., 2019). Altered TSPO binding has also been associated with psychiatric disorders such as depression and schizophrenia (Bauer and Teixeira, 2019;Gilhus and Deuschl, 2019), but the evidence is somewhat ambiguous (Fried et al., 2019;Miller and Goldsmith, 2019;Perry, 2018).
Attesting to the enormous importance of detecting neuroinflammation with molecular imaging is the large array of clinically available TSPO PET radiotracers. Despite this, much is still unknown and controversial about the interpretation of TSPO imaging (Guilarte, 2019), partly due to a lack of sufficient knowledge about its cellular functions and distribution. To date, TSPO has been implicated in a variety of cellular & mitochondrial functions, including biosynthesis of steroids and haeme, porphyrin transport, apoptosis, cell proliferation, and neuromodulation (Papadopoulos et al., 2006;Rupprecht et al., 2010), although some of those have recently been challenged based on the lack of confirmatory findings in studies using TSPO knockout animals (Banati et al., 2014;Middleton et al., 2015;Tu et al., 2014;Wang et al., 2016). In light of all available evidence, there is now an emerging consensus that the primary role of TSPO may be for the regulation of cellular energy turnover processes (Betlazar et al., 2020). As mounting a proinflammatory response by microglia is energetically expensive, this could explain the observed increase in TSPO binding observed in pathological conditions (O'Neill et al., 2016). Additionally, in an inflammatory state, TSPO-rich immune cells from outside the brain (Zhao et al., 2020), such as monocytes and macrophages may infiltrate the brain, and some of the observed central TSPO signal may therefore originate from this peripheral population. The role of TSPO in energy maintenance could also explain its abundance in various other brain cells, all of which have not only immunomodulatory but also energetically demanding roles in the brain, such as ependyma and choroid plexus, which are involved with regulating the cerebrospinal fluid (Hofman and Chen, 2016;Wilson, 1980); the endothelium, which is involved with maintaining the blood brain barrier (Oldendorf et al., 1977); the astrocytes which proliferate and reconfigure during inflammation (Liddelow and Barres, 2017); and even neurons whose increased activation was recently reported to increase TSPO (Notter et al., 2020). All of this greatly complicates the interpretation of 'neuroinflammation' based solely on TSPO image data.
A single systemic administration of low-to-moderate dose of the bacterial endotoxin, lipopolysaccharide (LPS), is known to induce transient microglia proinflammatory activation (Biesmans et al., 2013;Buttini et al., 1996), as well as the characteristic "sickness behaviour", in a variety of experimental animals and humans (Dantzer et al., 2008;DellaGioia and Hannestad, 2010;Ferreira Mello et al., 2013;Park et al., 2011;Schedlowski et al., 2014). This acute LPS challenge approach has been utilized to study neuroinflammation, as measured by an increase in the TSPO PET signal, in humans (Sandiego et al., 2015), primates (Hannestad et al., 2012) and mice (Vignal et al., 2018). As none of those studies were able to explore cellular origins of the signal changes, they mostly attributed them to be of microglial origin. Moreover, many human PET studies continue to view TSPO signal as exclusively labelling changes in microglial activation. However, altered TSPO expression should be understood as the total result of the interactions between various cell types involved in neuroimmune modulation, in addition to microglia (Lee et al., 2020).
To address this point, we aimed to more comprehensively examine the cellular contributors of the TSPO signal after an acute peripheral LPS challenge in rats. We aimed to confirm the upregulation of TSPO utilising in vivo PET imaging, followed by quantitative ex vivo autoradiographic, immunohistochemical and in situ hybridization examination of the CNS tissue, in order to map the contribution of glial, neuronal, endothelial and peripheral monocyte-derived macrophage cells attributing to the overall PET signal.

Animals
Male Sprague-Dawley (SD) adult rats weighing 250-300 g (Charles River, UK) were used in all experiments. Rats were housed under controlled environmental conditions (22 ± 1 • C and a 12-h light/12-h dark cycle) with free access to food and water. This study was ethically reviewed and conducted in accordance with Animals (Scientific Procedures) Act 1986 and the GSK Policy on the Care, Welfare and Treatment of Animals.

Intracerebral LPS (ic LPS)
We initially used a rat model of intracerebral administration of LPS (ic LPS model) since it has been extensively used for the evaluation of brain nuclear medicine imaging tracers (Dickens et al., 2014;Ory et al., 2016;Sridharan et al., 2017;Venneti et al., 2007) and is known to induce a robust focal inflammatory lesion (Sridharan et al., 2017). Rats (n = 11) were treated with unilateral stereotaxic injection of 1 µg (in 0.5 µg/µL final concentration) of LPS from Escherichia coli 0111: B4 (Sigma) mm AP to the Bregma) using a 2 μL Neuros™ microsyringe (Hamilton, USA) and injection rate of approx. 0.5 μL/min (Espinosa-Oliva et al., 2013). PET image acquisition was carried out 4 days after intracerebral injection of LPS using [ 11 C]PK11195 (n = 3) and [ 18 F]DPA-714 (n = 4) after which the brains were removed for autoradiography analysis. For corroborative immunofluorescence a different cohort of rats (n = 4) were utilised. See also experimental design in Table 1.

Systemic intraperitoneal LPS (ip LPS)
In order to induce neuroinflammation via a peripheral administration of LPS, rats (n = 40) were injected intraperitoneally (ip) with either LPS from Escherichia coli 0111: B4 (Sigma) (0.5 mg/kg, ip), or vehicle injection (saline 1 mL/kg) as a control. PET imaging was carried out 24 hours following challenge using [ 18 F]DPA-714 (n = 8/group) and the brains were removed for immunohistochemistry and autoradiography analyses. For immunofluorescence and RNAscope in situ hybridization, two different cohort of rats (n = 6/group and n = 5-6/group respectively) were utilised. See experimental design in Table 1. The timing of the PET imaging (24 hours after peripheral LPS administration), was chosen as this represents a period where the occurrence of depressivelike behavior has previously been observed (Dantzer et al., 2008;Park et al., 2011).

[ 11 C]PK11195 synthesis
[ 11 C](R)-PK11195 was produced at the PET Centre, St Thomas's Hospital, London, UK. Carbon-11 was used to radiolabel PK11195 with an overall synthesis time of 35 min from the end of bombardment. (R)-N-Desmethyl PK11195 was the precursor used for the synthesis. The process involved (a) synthesis of [ 11 C]CH 3 I from [ 11 C]CO 2 using Trac-erLab FX MeI, (b) transfer of [ 11 C]CH 3 I to Trasis AiO module, (c) reaction of [ 11 C]CH 3 I with KOH (4 mg) and precursor (1 mg) in DMSO at room temperature for 2 min, followed by (d) semi-preparative HPLC purification on Gemini C18 250x10mm 10µ, Phenomenex and finally (e) product formulation on Sep-Pak C18 light and sterilisation on Sterifix Paed filter. On average, 0.5-1 GBq [ 11 C]PK11195 was obtained from 20 GBq of [ 11 C]CO 2 (greater than 95% radiochemical purity and greater than 50 GBq/µmol) with activity yields of 4 ± 2%.

PET image acquisition
Image acquisition was conducted on both rat models of neuroinflammation (i.e. ic LPS and ip LPS). Anaesthesia was induced with 4% isoflurane and maintained at 2-2.5% of isoflurane in 100% O 2 .
Ipsilateral/contralateral ratios were calculated by dividing the area under the curve (AUC) for basal ganglia in the ipsilateral vs. in the contralateral over the entire time frame.

Ip LPS rat image acquisition
On the day of image acquisition, twenty-four hours (±2 h) following systemic administration of LPS or vehicle, rats were surgically (under 2.5% isoflurane anaesthesia) prepared by implanting a femoral venous and arterial catheter for radiotracer administration and blood sampling, respectively. They were then placed in a supine position and scanned for 60 mins following intravenous administration of [ 18 F]DPA-714 (12.79 ± 1.75 MBq). Image acquisition and reconstruction was carried out as described in the previous section for [ 18 F]DPA-714 PET. Arterial blood samples (20 μL per sample) were collected and plasma isolated by centrifugation at 3,000 rpm, room temperature for 5 min. Radioactivity in whole blood and plasma samples were measured by gamma counting, decay corrected, and plotted using Graph-Pad Prism 8 (San Diego, CA, USA) to calculate AUC. The plasma and whole blood radioactive counts obtained were converted to MBq using a calibrated conversion scale, and % injected dose in MBq/g (%ID/g) calculated for each rat.

PET image analyses
Images were analysed using VivoQuant 2.0 (Invicro LLC) software. The 3D rat brain atlas template was used to extract the radiotracer timeactivity curves (TACs) in 13 regions of interest (ROI; olfactory, cortex, basal ganglia, corpus callosum, hippocampus, thalamus, amygdala, hypothalamus, midbrain, septal area, ventricles, white matter and cerebellum) and also as a whole brain ROI. Standardized uptake values (SUV) were calculated by dividing the image derived concentration (in kBq/cc) with the ratio of the injected dose in MBq to the body weight (in g).

Tissue collection
Immediately following scanning, ip LPS rats were euthanized by decapitation, whole brains excised and divided into two hemispheres. One hemisphere was flash frozen in chilled isopentane (− 35 • C) and stored at − 80 • C until required for autoradiography. The other hemisphere was placed in 4% w/v paraformaldehyde (PFA) for 24 h, followed by immersion in 30% w/v sucrose solution and stored at 4 • C until required for assessment by immunohistochemistry. For the ic LPS rats, immediately following scanning rats, whole brains were flash frozen in chilled isopentane (− 35 • C) and stored at − 80 • C until required for autoradiography (Table 1).
Briefly, slides were incubated with 100 mM Tris-HCl, containing 1 nM [ 3 H]PK11195 (specific activity 82.7 Ci/mmol, Perkin Elmer, #NET885001MC) for 30 min, or 1 nM [ 3 H]DPA-714 (specific activity 40.8 Ci/mmol, custom tritiated by RC TRITEC, Switzerland) for 60 min, washed twice for 6 min in 100 mM Tris-HCl, rinse-dipped in dH 2 O and air dried. Non-specific binding (NSB) for both radiotracers were determined on adjacent sections in the presence of unlabelled PK11195 (20 µM; Sigma-Aldrich, #85532-75-8). Slides were exposed to tritiumsensitive film (Amersham Hyperfilm MP, GE Healthcare) in autoradiography cassettes together with a set of tritium standards ([ 3 H] Microscale, American Radiolabelled Chemicals, #art-0123A) for 6 weeks ([ 3 H]PK11195) or 4 weeks ([ 3 H]DPA-714). Quantitative analysis was performed using MCID image analyzer (Image Research, Canada), and the brain structures were identified using the rat brain atlas of Franklin and Paxinos (1997). All regions of interest (ROI) were analyzed by freehand drawing tools in two consecutive sections per brain (i.e. adjacent total and NSB sections). Certain animals were excluded from the analysis given the difficulty to sample the ROI.
Double immunofluorescence of TSPO (NP155) and Iba1 (a marker for both microglia and macrophages) was conducted on rat brain sections acquired from a non-imaged rat cohort from ip LPS and vehicle (n = 6/group). The animals were transcardially perfused with PBS following 4% w/v PFA, 4 days after intracerebral injection of LPS or 24 h after systemic administration of LPS/vehicle. Brains were excised placed in 4% w/v PFA for 24 h, followed by immersion in 30% w/v sucrose solution and stored at 4 • C.
Brains were cut at 35 µm thickness in HM 430 Sliding Microtome (Thermo Fisher Scientific). Free floating sections were incubated in citrate buffer pH = 8 at 80 • C for 30 min and 0.3% v/v H 2 O 2 to block for endogenous peroxidase activity (only for bright field immunohistochemistry), and with 10% w/v milk and 0.3% v/v Triton X-100 in TBS for 40 min. After rinses with tris buffered saline (TBS), sections were incubated overnight at 4 • C with rabbit anti-TSPO antibody (1:2000 in blocking solution, NP155) for bright field immunohistochemistry and with a mix of primary antibodies for immunofluorescence (see Supplemental Table 1). After washes with TBS, sections were incubated with the appropriate biotinylated (Vector Labs) or Alexa-conjugated secondary antibodies (see Supplemental Table 1).

Immunohistochemistry
For bright field immunohistochemistry, following rinsing, the sections were incubated with Vectastain Elite ABC peroxidase kit (Vectors Labs) and visualized with diaminobenzidine (DAB) precipitation. Sections were mounted with DPX and images were captured at 40x magnification with photomicrographs captured using a Virtual Slide Microscope VS120 (Olympus Life Science). ROIs from the olfactory area, hippocampus, substantia nigra, periaqueductal grey midbrain and the cerebellum (3-4 sections/ROI) were analysed by measuring the percentage area covered by TSPO in 1700 × 1700 pixels images using ImageJ (NIH, USA). Certain animals were excluded from the analysis given the difficulty to sample the ROI.
To measure the TSPO expression of other cell types (Iba1-cells), such as astrocytes, neurons or endothelial cells, to the overall LPS induced TSPO expression, double immunofluorescence of TSPO (NP155) and GFAP (astrocytes) and triple immunofluorescence of TSPO (NP155), NeuN (neurons) and CD31 (vascular endothelial cells) were performed, following the protocol previously described (see Supplemental Table 1 for antibodies used).

RNAscope in situ hybridization
Finally to further investigate Tspo gene expression in different immune-cell types, such as microglia and infiltrated monocytes-derived macrophages, and to overcome the limitations in the use of reliable histological markers to discriminate these two immune cells populations, RNAscope Multiplex Fluorescence V2 technology for in situ hybridization was used to simultaneously detect three RNA targets (Tspo, Tmem119 and Ccr2) on fresh-frozen brain sections from vehicle and LPS-treated rats. RNAscope in situ hybridization was performed according to the manufacturers protocol for RNAscope Multiplex Fluorescent Reagent Kit v2 (Advanced Cell Diagnostics, #320293), using 3 different probes against the genes of interest (see Supplemental Table 2). For quantification, images were acquired as previously described from one dorsal CA1 of the hippocampus section (Bregma − 3.24 mm) and one substantia nigra section (Bregma − 3.24 mm) per animal. Analysis was performed using HALO Image Analysis Software (Indica Lab) to quantify the number of Tspo mRNA copies following a cell-type discrimination based on the expression of a recently discovered specific microglial marker in the brain; Tmem119 that is reported to differentiate activated microglia from monocyte-derived macrophages (Bennett et al., 2016); and Ccr2, a marker that although expressed in both microglia and monocyte-derived macrophages, is primarily expressed in monocytederived macrophages (Haage et al., 2019;Li et al., 2019). Tmem119 mRNA expression is highly enriched in the CNS and is abundandly expressed in microglial cell population (Bennett et al., 2016;Van Hove et al., 2019). However, Tmem119 is co-expressed in a very low proportion in monocyte-derived macrophages (Li et al., 2019;Van Hove et al., 2019). In order to ensure specific selection of microglia and monocytederived macrophages and to minimise nonspecific hybridization events, we defined a specific threshold to distinguish these two cell populations: microglial cells were defined as Tmem119 positive (more than 5 Tmem119 mRNA copies) and Ccr2 negative (less than 5 Ccr2 mRNA copies); and monocyte-derived macrophages as Tmem119 negative (less than 5 Tmem119 mRNA copies) and Ccr2 positive (more than 5 Ccr2 mRNA copies) (Fig. 5d).

Statistics
Time activity curve -standardised uptake value (TAC SUV) data from both the ic LPS and ip LPS rat models, and AUC from the TACs of [ 18 F] DPA-714 in the ip LPS model, were analysed using a mixed-effects model ANOVA with ROI or time as within-subject factor and treatment as between-subject factor followed by post hoc Sidak tests.
For the ic LPS model, AUC measured in contralateral and ipsilateral striatum were confirmed to follow a normal distribution with Shapiro-Wilk normality tests and were subsequently analysed using paired ttests and ipsilateral/contralateral ratios with an unpaired t-test.
The remaining data were also analysed by mixed-effects ANOVA with cells or ROI as within-subject factor and treatment as betweensubject factor followed by post hoc Sidak tests. Number of Iba1+ cells and TSPO+ cells were analysed using unpaired t-test. Quantitative data are expressed as mean ± SEM. All statistical analyses were carried out using GraphPad Prism (version 8), and p value < 0.05 was considered to be statistically significant.

Increased ipsilateral striatal uptake of both [ 11 C]PK11195 and [ 18 F] DPA-714 with ic LPS treatment
A robust focal inflammatory response following ic LPS has been detected in the present study (Supplemental Figures 1 and 2) including increased activation of microglia and astrocytes, as well as upregulation of TSPO, accompanied by induction of mRNA expression for various proinflammatory cytokines such as Tnf-α, Il-6, Il-1β, intercellular adhesion molecules (Icam-1) and peripheral monocyte recruitment (Ly6c) (see Supplemental Figure 2). This response was expected and aligns with previous reports studying different TSPO radiotracers (Ory et al., 2016;Sridharan et al., 2017). We aimed to replicate these findings with two TSPO radiotracers, [ 11 C]PK11195 and [ 18 F]DPA-714, in order to select the more sensitive radiotracer for the peripherally-induced neuroinflammation study (ip LPS), which was anticipated to present a more physiological neuroinflammatory phenotype.
We compared ipsilateral (lesioned) side of the brain with the contralateral (non-lesioned) side as a control, since contralateral side shows no inflammatory reaction (activation of microglia or astrocytes)  Figure 1). A significant increase in uptake of radiotracer was observed in the LPS-lesioned hemisphere compared to the non-lesioned hemisphere for both [ 11 C]PK11195 and [ 18 F]DPA-714 (F (1, 4) = 55.27 , p = 0.0017 and F (10, 60) = 122.3, p < 0.0001, respectively). Additionally, a significant effect of interaction time × hemisphere (F (7, 28) = 2.855, p = 0.0222; F (10, 60) = 4.148, p = 0.000224) was observed in the TACs for both [ 11 C]PK11195 and [ 18 F] DPA-714 in the basal ganglia ( Fig. 1a-d). As expected, ic LPS induced a significant increase in both [ 11 C]PK11195 and [ 18 F]DPA-714 uptake in the basal ganglia region of interest, where the AUC from the TACs confirmed a significantly higher uptake of both [ 11 C]PK11195 and [ 18 F] DPA-714 in the LPS-lesioned striatum vs. the non-lesioned striatum, 4 days after ic LPS (t (2) = 9.558, p = 0.012; t (3) = 7.456, p = 0.005, Fig. 1e,f). The ipsilateral/contralateral ratio was significantly higher in case of [ 18 F]DPA-714, demonstrating better signal, compared to [ 11 C] PK11195 (t (5) = 3.533, p = 0.0167, Fig. 1g). The superiority of DPA-714 was confirmed by ex vivo autoradiography using [ 3 H]PK11195 and [ 3 H]DPA-714, where both showed increased binding in the LPSlesioned side, but [ 3 H]DPA-714 signal is demonstrably clearer (Fig. 1h). Subsequently, [ 18 F]DPA-714 was selected to assess neuroinflammation in the systemic LPS rat model. No significant differences in [ 18 F]DPA-714 SUV AUC values were observed in any other ROI assessed in the LPS-lesioned hemisphere compared to the non-lesioned hemisphere, although there were a few regions that demonstrated elevated uptake in the lesioned-hemisphere, such as the cortex, corpus callosum, thalamus, amygdala, ventricles and white matter (Supplemental Table 3). The effect of the injection procedure at this particular time-point was assessed by immunofluorescence analysis of Iba1 and TSPO expression in ic vehicle-injected rats. No significant differences were detected in the ipsilateral vs. contralateral side of the brain in Iba1 and TSPO expression 4 days after intrastriatal vehicle injection (see Supplemental Figure 1c).
Importantly, there was no significant difference between the LPSand vehicle-treated rats in either the peripheral distribution of the  Figure 3a); or in the injected dose/weight (ID/g) of the radiotracer (Supplemental Figure 3b).

Increased global binding of [ 3 H]DPA-714 to CNS structures following ip LPS treatment
To corroborate the results obtained by [ 18 F]DPA-714 PET in ip LPS rats, we performed [ 3 H]DPA-714 ex vivo autoradiography on the brain tissue from the imaged animals. Densitometry analysis confirmed a significant 32% overall increase in [ 3 H]DPA-714 binding in the whole brain (all ROIs, 7.01 ± 0.73 nCi/mg LPS, 5.31 ± 0.47 nCi/mg VEH). A significant effect of treatment (F (1, 11) = 11.84, p = 0.0055) and ROI was observed (F (4, 40) = 31.69, p < 0.0001), but no significant treatment × ROI interaction was found (F (4, 40) = 0.56, p = 0.6885). assessed in the LPS treated-rats compared to vehicle-treated controls, with a qualitative observation suggesting this to be the greatest in the substantia nigra (65.5% increase compared to vehicle-treated rats) and lowest in the olfactory area (15.6%, Fig. 3a-b).

Microglia/macrophages together with astrocytes are the major contributors to increased TSPO expression following ip LPS treatment in rats
To investigate the extent of expression of TSPO within the different types of cells we performed co-localization immunofluorescence. We examined two ROIs, hippocampus and substantia nigra. We selected these ROIs for the following reasons: (1) both were found to have a large increase in TSPO following ip LPS treatment, as measured by PET (Fig. 2) and autoradiography (Fig. 3); (2) previous clinical PET studies have shown significant TSPO changes in these two brain areas of patients with neurodegenerative disease and psychiatric disorders (Haarman et al., 2014;Kreisl et al., 2016;Ouchi et al., 2005) and (3) both brain regions are known to have a particularly high microglia density (De Biase et al., 2017;Tan et al., 2020).
We next examined the co-localization, defining two types of cells, based on TSPO and Iba1 expression: TSPO positive that are also positive for Iba1 (TSPO+Iba1+), and TSPO positive that are not positive for Iba1 (TSPO+Iba1-). We looked at the effect of LPS treatment, the type of cells, and their interaction. In the hippocampus, we observed significant effect of LPS treatment (F (1, 10) = 9.409, p = 0.0119) and of cell type (F (1, 10 = 5.097, p = 0.0476) but no interaction (F (1, 10) = 2.858, p = 0.1218), indicating that after LPS both TSPO+Iba1+ and TSPO+Iba1cells increased evenly (Fig. 4d). Similarly, in the substantia nigra, there was a significant effect of LPS (F (1, 10 = 10.42, p = 0.0091) and of cell type (F (1, 10 = 19.49, p = 0.0013) with no significant interaction (F (1, 10) = 4.032, p = 0.0724), although there appeared to be a trend toward a greater increase in Iba1-cells after LPS (Fig. 4e). This suggests that LPS treatment induced an overall increase in various TSPO expressing cells that are not only of macrophage/microglia lineage. It is also interesting to note that even in the vehicle treated rats there was a similar number of TSPO+Iba1-and TSPO+Iba1+ cells, suggesting that in basal conditions there is as much TSPO in microglia/macrophage cells as in other cell types.
In order to investigate the identity of TSPO+Iba1-cells, we performed a further immunofluorescence study: double immunofluorescence of TSPO and GFAP (astrocyte marker), and triple immunofluorescence of TSPO, NeuN (neuronal marker) and CD31 (vascular endothelium marker). These multi-immunofluorescence investigations confirmed the expression of TSPO in astrocytes, neurons and endothelial cells in both the hippocampus and substantia nigra from both LPS and vehicle-treated rats (Fig. 4g,h). Statistical analysis demonstrated a significant effect of cells (F (2, 20 = 29.54, p < 0.0001) and significant cells × treatment interaction (F (2, 20 = 6.431, p = 0.007) in the hippocampus. Post hoc analysis confirmed that only the TSPO+GFAP+ cells increased significantly after LPS treatment (p = 0.0081). This trend of increased TSPO+GFAP+ cells also occurred in the substantia nigra, although there was no significant treatment × cells interaction and hence no post hoc comparisons were performed. For the substantia nigra there appeared to be only an overall significant effect of cells (F (2, 19 = 28.22, p < 0.0001). Overall these data suggest that although TSPO is present in all the cell types examined, astrocytes appear to be the main non-microglial contributor to the increased expression of TSPO following ip LPS treatment.

LPS-induced increased Tspo mRNA expression is limited to microglia specifically in the hippocampus, and to both microglia and monocytederived macrophages in the substantia nigra
In order to further characterize the identity of TSPO+Iba+ cells in both ROIs and to determine whether they were monocyte-derived macrophages or microglia, RNAscope Multiplex Fluorescence V2 technology for in situ hybridization was used to investigate Tspo gene expression in these different immune-cell types. Ip LPS treatment caused a significant increase in Tspo mRNA positive cells in both the hippocampus and substantia nigra compared to vehicle treated rats (t (9) = 3.091, p = 0.0129; t (9) = 3.164, p = 0.0115, respectively, Fig. 5b,c). For the purposes of distinguishing microglial cells from monocyte-derived macrophages, microglial cells were defined as Tmem119 positive and Ccr2 negative; and monocyte-derived macrophages as Tmem119 negative and Ccr2 positive (Fig. 5d). In the hippocampus, we observed significant effects of LPS treatment (F (1, 10) = 5.442, p = 0.0418) and of cell type (F (1,9 = 10.08, p = 0.0113) but not of interaction (F (1, 9) = 5.931, p = 0.0376). Similarly, in the substantia nigra, there was a significant effect of LPS treatment (F (1, 9 = 17.66, p = 0.0023) and of cell type (F (1, 9 = 25.79, p = 0.0007), but also of interaction (F (1, 9) = 10.16, p = 0.0110). Post hoc analysis of the number of Tspo mRNA copies in both cell populations revealed that in the hippocampus, Tspo mRNA increased expression was confined to microglia specifically (p = 0.0077, Fig. 5e), but in contrast, the substantia nigra had an elevated number of Tspo mRNA copies which were significantly higher in both microglia and monocyte-derived macrophages in LPS-treated rats (p = 0.0003, p = 0.0004, Fig. 5f).

Discussion
The aim of the current study was to ascertain if in vivo PET imaging with the second generation TSPO PET radiotracer, [ 18 F]DPA-714, can be used to detect neuroinflammation induced via peripheral admnistration of the endotoxin, LPS, in a rodent model, and also provide insight into the origin of the signal changes in terms of their cellular sources and regional distribution.
We performed an initial study to compare two TSPO PET radiotracers, [ 11 C]PK11195 (first generation) and [ 18 F]DPA-714 (second generation), and confirmed an increased uptake of both in a focallylesioned area of the brain. We used a well-characterised model of unilateral intracranial LPS administration to model the focal inflammation, and we performed our measurements 4 days after LPS by when a potentially confounding disruption of the blood brain barrier largely subsides (Sridharan et al., 2017), while microglia are robustly activated around the lesioned area (Supplemental Figure 1) (Choi et al., 2009). These findings corroborated previous studies which demonstrated increased TSPO expression in the same type of model with various TSPO PET radiotracers including [ 11 C]PK11195 and [ 18 F]DPA-714 (Ory et al., 2016(Ory et al., , 2015Sridharan et al., 2017) and are consistent with published data indicating that [ 18 F]DPA-714 appears to be more sensitive than [ 11 C]PK11195 to detect TSPO in acute neuroinflammation (Boutin et al., 2013;Chauveau et al., 2009). The increased sensitivity of [ 18 F]DPA-714 and other "second generation" TSPO radiotracers over the first generation tracer, [ 11 C]PK11195, is becoming increasingly recognised (Alam et al., 2017). Nevertheless the reason some clinical studies continue to use [ 11 C]PK11195 may be due to it not suffering from the problematic binding profile, unlike second generation radiotracers wherein approximately third of the population must be excluded due to low binding caused by a particular genetic polymorphism in their human Tspo gene (Owen et al., 2012) that does not appear to feature in rodents. Due to the increased sensitivity of the [ 18 F]DPA-714 PET tracer demonstrated in this initial study, we selected this PET radiotracer to assess neuroinflammation in the ip LPS rat model. In the rat model of peripherally induced neuroinflammation, we measured an increased [ 18 F]DPA-714 uptake in the whole brain 24 hours after systemic LPS administration. We chose this time-point because of in-house data confirming robust microglial activation and Tspo mRNA upregulation (Supplemental Figure 4). In addition, RT-qPCR analysis showed that the cytokine Tnf-α and the intercellular adhesion molecule-1 (Icam-1) were also elevated in the brain at this time (Supplemental Figure 4), both of which are highly relevant to brain inflammation in context of psychiatric disorders (Capuron and Miller, 2011;Müller, 2019). The increased expression of Icam-1 particularly points towards activation of endothelial cells as a result of alterations of the blood brain barrier and increased peripheral immune cell recruitment, adhesion and infiltration (Müller, 2019). Tnf-α is additionally capable of stimulating microglia to produce monocyte chemoattractant protein-1 (Mcp-1) which was also increased in our samples (Supplemental Figure 4) and which is likely to mediate the recruitment of monocytes into the brain (D'Mello et al., 2009). Collectively these data confirm the immune system activation and suggest peripheral immune cell infiltration in the brain parenchyma at this time-point. Interestingly, two earlier PET studies, one in humans (Sandiego et al., 2015) and one in non-human primates (Hannestad et al., 2012) demonstrated increased TSPO binding at the much earlier time-points of 1 and 4 hours following iv LPS administration, respectively. We did not investigate such early time-points because our pilot rat experiments showed no evidence of either microglia hypertrophy or hyperplasia 6 hours after systemic LPS. This is also borne out by the literature where the majority of studies using similar doses of LPS in rodents report activated microglia at 24 hours (Biesmans et al., 2013;Buttini et al., 1996;Hoogland et al., 2015). It is nevertheless possible there are species differences in the timing of microglia and TSPO changes, given that humans are known to be much more sensitive to LPS than rodents (Copeland et al., 2005).
The increased ip LPS-induced binding of DPA-714 to TSPO in the rat brains appeared to be global, in that all ROIs showed a similar increase, of which notably the olfactory, hippocampus, mid brain, ventricles, white matter and cerebellum increased the most. This finding matches the only other published rodent study that reported a similarly widespread global TSPO increase in a mouse after a much higher dose (5 mg/ kg ip) of LPS (Vignal et al., 2018). Given that none of the aforementioned studies in any species explored the cellular origins of the TSPO signal in more detail, we followed up the rat PET study by post mortem autoradiographic and immunohistochemical examinations.
High resolution autoradiography with [ 3 H]DPA-714 corroborated the increased overall TSPO binding in all ROIs following ip LPS treatment (Fig. 3a, b). Immunohistochemical analysis also confirmed increased TSPO expression (Fig. 3c, d), which was particularly evident in the olfactory area, hippocampus and substantia nigra (Fig. 3d). Interestingly, neither TSPO immunohistochemistry nor [ 3 H]DPA-714 autoradiography detected any significant TSPO increase in the cerebellum (Fig. 3b, d), whereas there appeared to be a clear cerebellar increase in [ 18 F]DPA-714 uptake with in vivo PET (p = 0.002, Fig. 2d). This could be due to the complex TSPO distribution in the cerebellum and suggests a possibility that LPS-induced [ 18 F]DPA-714 signal is comprised of multiple sources including peripheral TSPO bound to plasma proteins and blood cells (Turkheimer et al., 2015;Veronese et al., 2018), in addition to the brain parenchyma. Despite these potentials confounds, our data demonstrate that peripherally induced neuroinflammation in rats can be detected in vivo by [ 18 F]DPA-714 PET and corroborated by [ 3 H]DPA-714 ex vivo autoradiography and TSPO immunohistochemistry.
In order to more precisely explore the cellular specificity of TSPO we performed double immunofluorescence studies combining TSPO with markers of different CNS cells subtypes. We analyzed the hippocampal and the substantia nigra regions since these are the most densely microglia populated areas of the brain (De Biase et al., 2017;Tan et al., 2020) and because several clinical studies have demonstrated changes in TSPO PET signal in these areas of patients with neurodegenerative and psychiatric disorders (Haarman et al., 2014;Kreisl et al., 2016;Ouchi et al., 2005). We first looked at TSPO in microglia/macrophage cells (Iba1 positive cells) versus all other cells (Iba1 negative cells) and found that even in the vehicle treated rats, TSPO expression was approximately evenly distributed between these two classes of cells in both ROIs. Following ip LPS treatment, as expected, there was an overall increase in TSPO+Iba1+ cells, but also a surprisingly large fractional increase (≃1 and 1.5 fold increase in hippocampus and substantia nigra respectively) in the TSPO+Iba1-cells compared to vehicle, confirming the presence of TSPO in the non-microglial cells at rest, and as well as in the inflamed state.
As previous studies have shown a low, albeit positive TSPO expression in non-microglial cells including neurons (Veenman and Gavish, 2000), endothelial cells (Betlazar et al., 2018) and astrocytes (Pannell et al., 2020), in both rodents and humans (Gui et al., 2020;Wimberley et al., 2018), we further examined the contribution of those cells to the TSPO+Iba1-signal. We demonstrated the presence of TSPO in all three cell types, in both LPS-and vehicle treated rats (Fig. 4g, h) corroborating that there appears to be an additional non-microglial expression of TSPO. Of the three cell types (neurons, endothelial cells and astrocytes), TSPO after LPS was only significantly increased in hippocampal astrocytes (p = 0.0081), with a similar trend observed in astrocytes in the substantia nigra (Fig. 4g-j) (p = 0.0497). Neither neurons nor endothelial cells demonstrated an upregulation of TSPO in response to LPS, at least at this time-point, in these two brain regions. These results confirm the findings of several studies describing TSPO upregulation in astrocytes in different neurological disorders (Cosenza-Nashat et al., 2009;Lavisse et al., 2012;Nutma et al., 2019). Our results also confirm our original hypothesis that a significant proportion of TSPO imaging signal arises from cells that are not of microglia/macrophage lineage, even under basal conditions.
One of the limitations of using a cellular marker such as Iba1 to study Fig. 4. TSPO co-localization with different cell types in ip LPS-or vehicle-injected rats (a) Confocal photomicrographs from Iba1 (green) and TSPO (red)-immunostained hippocampus and substantia nigra sections of ip LPS-and vehicle-treated rats (top left panel). Scale bar whole brain section = 1 mm, scale bar magnification = 50 μm. Nuclear counterstaining was performed using DAPI (blue  neuroinflammation is that it does not distinguish between CNS resident microglia and peripherally infiltrating macrophages, due to their common lineage. To further establish the cellular specificity of the TSPO+Iba1+ signal, as microglial cells versus monocyte-derived macrophages, we utilised RNAscope in situ hybridization. We found that LPSinduced expression of Tspo mRNA in the hippocampus was limited to microglia specifically (1-fold increase), whereas in the substantia nigra Tspo mRNA was significantly increased in both microglia (2-fold) and monocyte-derived macrophages (1.3-fold) vs. vehicle (Fig. 5). It is possible that this increased Tspo gene expression in monocyte-derived macrophages induced by LPS in the substantia nigra could be responsible for the more pronounced increase in colocalized TSPO+Iba1+ protein signal from cells in the substantia nigra compared to the hippocampus (Fig. 4). Additional studies to more accurately characterise the regional specificity of Tspo expression in monocyte-derived macrophages are warranted especially in conjunction with blood brain barrier permeability. Another limitation of this study is that we restricted our measurements to a single time-point after LPS. Given the reported differences in timing of neuroinflammatory changes after LPS and similar peripheral inductors, it would be of interest to conduct a study combining longitudinal PET imaging with cellular/molecular characterisation of signal changes to capture the timecourse of the immune response and regional brain effects.

Conclusions
Overall, our data have shown an increased [ 18 F]DPA-714 signal following a peripheral challenge with LPS, with analysis revealing this increased signal to arise from both microglia and non-microglia cellular sources in the specific brain regions that were examined. It is important to note that the LPS paradigm utilised in these studies is a model of peripherally-induced inflammation that subsequently causes a neuroinflammatory state but that cannot be directly related to human neurodegenerative disease or psychiatric disorders. It will therefore be imperative to investigate TSPO expression on human tissue acquired from diseased patients vs. matched controls, and to examine the changes at different time-points, as there may be more pronounced/different neuroinflammatory responses exhibited in chronic human neurodegenerative disease or psychiatric disorders. Subsequently, our data may only reflect the "tip of the iceberg" for comprehending how the nonmicroglial component of the TSPO PET signal can be understood and therefore interpreted in neuroinflammation.
Consequently, as a result of these findings we suggest that an important contribution of microglia, astrocytes and peripheral infiltrated cells needs to be considered when assessing neuroinflammation with TSPO PET radiotracers. Moreover, the cellular origin of TSPO signal appears to depend on the brain region examined, and we further propose that this will also be dependant on the type and the time-course of the inflammatory response.

Ethics approval
This study was ethically reviewed and conducted in accordance with Animals (Scientific Procedures) Act 1986 and the GSK Policy on the Care, Welfare and Treatment of Animals.

Consent for publication
All the authors agree to the publication of this work.

Availability of data and material
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Funding
This study was part funded by GSK and a grant from the Wellcome Trust (Grant number: 104025/Z/14/Z). The Centre for Radiopharmaceutical Chemistry is funded in part by the Department of Health's NIHR Biomedical Research Centres funding scheme; K. Sander is additionally funded by Mallinckrodt.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.