TSPO expression in a Zika virus murine infection model as an imaging target for acute infection-induced neuroinflammation

[³H]PK11195 (NET885250UC, 9.25 MBq, 2912 GBq/mmol) was purchased from Perkin-Elmer, Singapore. [¹⁸F]FEPPA (4.24 GBq, 533 GBq/μmol) was synthesized at > 99% radiochemical purity by National University of Singapore Clinical Imaging Research Center (NUS-CIRC).

All animal experiments were conducted with approval from the Institutional Animal Care and Use Committee (IACUC) of Duke-NUS Medical School and Singhealth (IACUC approval no. 2020/SHS/1607) and conformed to the National Institutes of Health (NIH) guidelines and public law. The animals were housed in individually ventilated cages and provided with food pellets and water ad libitum.

The ZIKV virus strain Paraiba01/2015 (Genbank Accession No. KX280026.1) was isolated from an anonymous patient in the state of Paraiba, Brazil and was a kind gift from Dr. Pedro Vasconcelos (Instituto Evandro Chagas, Brazil). The AG129 mice (adult male) deficient in expression of IFN-α/β and γ receptors were obtained from an in-house breeding colony.

The ZIKV infection model was established as described previously [27, 31]. Briefly, mice were infected with ZIKV by intraperitoneal (i.p.) injection of 10⁵ plaque-forming units (PFU) in 100 μL. Mock-infected controls were similarly inoculated with an equal volume of sterile phosphate buffered saline (PBS, pH 7.4). Both groups of animals were observed daily for signs of disease and underwent general health and neurological examinations. ZIKV-infected mice were euthanized when moribund, based on these criteria: (1) loss of > 20% original body weight, (2) persistent limb paralysis of > 48 h, and (3) immobility or severe lethargy. Animals were euthanized by 10% CO₂ inhalation for 5 min.

Whole brains were harvested as described previously [32]. For immunohistopathological evaluation, brains were fixed in 10% neutral buffered formalin (Sigma-Aldrich, USA) for 72 h and processed for embedding in OCT. Tissue cryosections (10 μm thick) were incubated overnight with either rabbit antibody against translocator protein (TSPO) (Invitrogen MA5-31,966) or rabbit antibody against glial fibrillary acidic protein (GFAP) (Dako Agilent GA524) followed by secondary antibody incubation. Stained tissue sections were imaged using Leica Cell DIVE (Leica Microsystems, Singapore).

Brains used for in vitro TSPO-binding with [³H]PK11195 were freshly embedded in OCT without tissue fixation and rapidly frozen. Cryosections (20 μm thick) were dried in 37 °C for 1 h prior to rehydration in high salt buffer (0.17 M Tris–Cl, pH 7.4) for 30 min. Tissues were incubated with radiotracer (0.01 mM; 0.087 MBq) in room temperature for 1 h and subsequently washed 3 × with high salt buffer and twice with ice-cold Milli-Q water. Slides were air-dried overnight. Digital autoradiography (DAR) images were acquired for 48 h using BeaVacq (Ai4R, France) and subsequently analyzed using the BeaQuant (Ai4R, France).

Frozen brains were homogenized to extract either total protein or total RNA. Total protein was subjected to ELISA for detection of IL-6 and TNF-α expression using the Ready-Set-GO! Kit following the manufacturer’s protocol (eBioscience, USA). Cytokine expression was reported as pg protein per mg (pg/mg) tissue. Viral RNA was isolated using RNEasy RNA extraction kit (Qiagen, Germany) according to manufacturer’s instructions. Real-time RT-PCR was carried out with primers targeting the E gene of ZIKV genome (ZK_F: 5′- CCGCTGCCCAACACAAG– 3′; ZK_R: 5′- CCACTAACGTTCTTTTGCAGACAT- 3′) as previously described [33]. In vitro transcribed RNA product of ZIKV-H/PF/2013 (accession number: KJ776791.2) E gene was used to generate the standard curve for quantification of viral genome copy.

[¹⁸F]FEPPA (2 MBq) was injected into the retroorbital space, and mice were sacrificed following 2 h uptake. Tissues were harvested and brains were micro-dissected as described previously [34]. Tissues were immediately loaded into a gamma counter (2470 Wizard2, PerkinElmer, USA). Tracer distribution was reported as % injected dose per gram tissue (% ID/g).

PET/CT imaging in mice was performed as described previously [27, 35]. Briefly, mice were sedated with 2% isoflurane, and intravenously (i.v.) injected with [¹⁸F]FEPPA (30 MBq). Following a 2-h tracer uptake phase, mice underwent head-focused PET scan for 10 min (VECTOR⁴CT, MILabs, Netherlands) using the M7 UHR 0.7-mm-pinhole collimator, immediately followed by total-body CT scan for 5 min (tube current, 0.24 mA; tube voltage, 50 kV). The resulting images were reconstructed at 0.8-mm voxel resolution using a SROSEM algorithm with 8 subsets, 14 iterations, and 1.2-μm Gaussian filter. PET and CT images were automatically co-registered, and attenuation correction of the PET image was done using the CT image.

Image analysis was performed using PMOD v3.7 (PMOD Technologies, Switzerland). Volumes of interest (VOIs) were drawn around the brain guided by the skull in CT images. Brain segmentation analysis was performed by drawing VOIs around the cerebral cortex (CTX), diencephalon (Th + Hy), and combined hindbrain (CBX & MY) in the CT images. VOIs were transferred to the co-registered PET images, and tracer uptake was reported as % ID (injected dose).

Following PET imaging, animals were euthanized and brain tissues collected for ex vivo DAR imaging. Cryosections (20 μm) were air-dried for 30 min prior to DAR imaging for 24 h using BeaVacq (Ai4R, France). Images were analyzed using the BeaQuant software (Ai4R, France).

Tissue immune profiling was conducted as previously described [27, 36]. Briefly, brains were harvested, mechanically disaggregated, and incubated in digestion buffer containing DNaseI and collagenase. Following homogenization, single cells were collected by passing the suspension through a nylon mesh strainer (70 μm). Fat was subsequently removed by centrifugation in Percoll. Cells were labelled with fluorescent antibody cocktails, and immune cell subsets were identified by flow cytometry (Fortessa, BD Biosciences, USA) and analyzed using FlowJo V10.8.0 (BD Biosciences, USA). Detected events were normalized to absolute cell number using CountBright Absolute Counting Beads (ThermoFisher, USA) [37]. The following markers were used to identify the immune landscape: total immune cells (CD45⁺); granulocytes (Ly6G⁺Ly6C⁺CD11b⁺); microglia (CD11b⁺F4/80⁺Ly6c⁻Ly6G⁻CD3⁻); monocyte-derived macrophages or mo-MACs (CD11b⁺F4/80⁺Ly6c⁺Ly6G⁻CD3⁻); monocytes (CD11b⁺CD115⁺); dendritic cells (CD11c⁺MHCII⁺); B cells (CD3⁻CD19⁺B220⁺MHCII⁺); total T cells (CD19⁻CD49b⁻B220⁻LY6G⁻CD3⁺); cytotoxic T cells (CD19⁻CD49b⁻B220⁻LY6G⁻CD3⁺CD8⁺), and helper T cells (CD19⁻CD49b⁻B220⁻LY6G⁻CD3⁺CD4⁺). TSPO expression was reported as mean fluorescence intensity (MFI) from samples tagged with anti-mouse TSPO-AlexaFluor 488 (ab199779; Abcam, USA).

Statistical analyses and graphing were performed with Prism v8.0 (GraphPad Software, USA). Means from two groups were compared by Mann–Whitney test, while means from multiple groups were compared by Kruskal–Wallis test with Dunn’s post-hoc correction. Linear regression analyses on scatter plots were performed using Spearman’s coefficient (ρ) to assess the correlation between tissue viral load and cytokine expression. Results were considered statistically significant at p < 0.05.

The adult AG129 murine model of acute Zika virus (ZIKV) infection exhibited median survival of 9 days with mortality from day 8 post-infection onwards (Fig. S1a). Disease was accompanied by overt decline of health status (Fig. S1b), drastic weight loss (Fig. S1c), and neurological deficits—such as ataxia, unsteady gait, and weakness or paralysis of the limbs (Fig. S1d)—that manifested from day 4 post-infection. The study schedule was optimized to capture disease- and host response-related biomarker dynamics at increasing stages of disease severity (Fig. 1a). Increasing ZIKV burden was detected in mouse brains at mid disease (day 4) and late disease (day 8) (Fig. 1b). Disease was accompanied by hallmarks of neuroinflammation: elevated expression of pro-inflammatory cytokine interleukin-6 (IL-6) in homogenized brain tissue (Fig. 1b); infiltration of immune cells in representative tissue sections from different brain regions (Fig. 1c, d); and high expression of glial fibrillary acidic protein (GFAP) detected by immunofluorescence imaging of brain tissue sections (Fig. 1e). Moreover, increased brain expression of translocator protein (TSPO) was observed in tissue sections (Fig. 1f). Thus, ZIKV neurological infection in our model is characterized by global elevated expression of both GFAP and TSPO throughout the brain cortex and parenchyma at late ZIKV disease, which was not overt at mid disease (data not shown). These establish that ZIKV-associated neuroinflammation is accompanied by brain TSPO overexpression at late disease.

Tissue sections from two selected regions (Fig. 1c) incubated with [³H]PK11195 and imaged by DAR (Fig. 1g) revealed twofold higher radioligand binding in late ZIKV brains compared to similarly treated pre-infection mouse brains (region 1, p = 0.008; region 2, p = 0.049) (Fig. 1h). Tracer binding was mostly localized to the cerebral cortex (CTX), hippocampal formation (HPF), diencephalon (thalamus and hypothalamus; Th + Hy), and cerebellum (CBX) (Fig. 1g). No increase in tracer binding to brain tissue was observed at mid ZIKV disease relative to pre-infection (Fig. 1h), consistent with TSPO immunostaining findings (Fig. 1f). Hence, subsequent studies focused on comparing late versus pre-disease stages.

Tissue biodistribution of [¹⁸F]FEPPA in ZIKV-infected brains was determined by ex vivo gamma counting at various stages of disease (Fig. 2a). Tracer uptake in whole brains was 2.4-fold higher at late disease versus pre-infection (2.26 vs. 0.95 %ID/g; p = 0.03) (Fig. 2b; Table S1). All micro-dissected brain regions assessed also exhibited ≥ 2.2-fold-higher [¹⁸F]FEPPA uptake relative to pre-infection (Fig. 2b; Table S1). However, [¹⁸F]FEPPA blood pool activity was notably 2.3-fold greater than pre-infection levels (1.48 vs. 0.65 %ID/g; p = 0.03); and brain tracer uptake at late disease was only 1.53-fold higher than in the blood (Fig. 2b; Table S1). These indicate a significant contribution of blood pool activity to whole brain or regional [¹⁸F]FEPPA brain uptake.

[¹⁸F]FEPPA-PET images of ZIKV-infected brains revealed a visible increase in signal at late disease relative to pre-infection (Fig. 2c), which upon VOI analysis was 2.6-fold higher in late disease than pre-infection (3.72 vs. 1.42 %ID; p = 0.002) (Fig. 2d). This trend was recapitulated in the different brain regions evaluated with brain segmentation analysis: 2.7-fold increased uptake in cerebral cortex (CTX) (3.15 vs. 1.16 %ID; p = 0.002); 2.7-fold increase in the diencephalon (Th + Hy) (3.89 vs. 1.46 %ID; p = 0.001); and a two-fold increase in the hindbrain (CBX & MY) (3.74 vs. 1.86 %ID; p = 0.02) (Fig. 2d). The VOI volumes in PET tracer uptake analysis were comparable between groups (Fig. S2). DAR imaging after PET/CT also revealed global increased brain [¹⁸F]FEPPA uptake at late disease (Fig. 2e). In addition, radioligand uptake in whole brains exhibited high positive correlation with ZIKV viral load (ρ = 1.0; p < 0.0001) and expression of pro-inflammatory cytokines IL-6 (ρ = 1.0; p < 0.0001) and TNF-α (ρ = 0.82; p = 0.006) in whole brains (Fig. 2f). These strongly suggest that TSPO expression can potentially be used as a surrogate biomarker for interrogating brain tissue viral burden and neuroinflammation status in our ZIKV model by in vivo [¹⁸F]FEPPA-PET imaging. However, radioligand activity in the brain may also be confounded by substantial blood pool activity.

TSPO-expressing cells in whole brain tissues were profiled using flow cytometry and compared with ex vivo [¹⁸F]FEPPA brain biodistribution analysis (Fig. 3a). Cells that do not express the general immune cell marker CD45—i.e., CD45⁻ cells—exhibited low TSPO expression at < 500 mean fluorescence intensity (MFI) (Fig. S3a). These non-immune cells include neurons, neuronal support cells, and endothelia. TSPO expression on these cells did not change with increasing disease severity (Fig. S3a), and this trend was also observed in the different micro-dissected brain regions (Fig. S3b). These data signify minimal contribution of CD45⁻ cells to brain TSPO expression.

In contrast, TSPO expression in the total population of immune cells (i.e., CD45⁺ cells) in the brain increased during ZIKV disease progression. Compared to pre-infection, CD45⁺ cell TSPO expression at mid and late ZIKV disease were 1.7-fold (p = 0.03) and 1.9-fold (p = 0.02) higher, respectively (Fig. 3b; Table S2a). This trend is reproduced in various micro-dissected brain regions (Fig. S4), where TSPO expression by immune cells at late disease increased by 1.3- to 3.0-fold compared to pre-infection, and the highest increase was recorded in the hippocampus and cerebral cortex (Table S3). TSPO expression in total CD45⁺ immune cells at pre and late ZIKV disease also exhibited strong correlation with ex vivo whole brain uptake of [¹⁸F]FEPPA (ρ = 0.89; p = 0.0006) (Fig. 3c).

The total CD45⁺ immune cells were further divided into subsets of (1) myeloid lineage and (2) lymphoid lineage cells. Myeloid lineage cells include granulocytes, monocytes, dendritic cells, brain-resident macrophages (i.e., microglia), and monocyte-derived macrophages (Mo-MAC). Lymphoid lineage cells include total T cells, helper (CD4⁺) T cells, and cytotoxic (CD8⁺) T cells. Only the myeloid subset of immune cells—specifically granulocytes, monocytes, and microglia, exhibited higher TSPO expression in ZIKV disease relative to pre-infection (Fig. 3b; Table S2). TSPO expression in microglia from whole brains was 3.9-fold higher than pre-infection (p = 0.006) (Fig. 3b; Table S2). In various brain regions, primarily in the hippocampus and cerebellum, microglia TSPO expression increased between 2.8- and 5.7-fold (Fig. S4; Table S3). Microglia TSPO expression in whole brains exhibited strong positive correlation with [¹⁸F]FEPPA uptake (ρ = 0.89; p = 0.0006) (Fig. 3d). Aside from microglia, monocytes also exhibited elevated TSPO expression at late disease, which was 2.9-fold higher than pre-infection (p = 0.02) (Fig. 3b; Table S2). In micro-dissected brains, primarily hippocampus and cerebral cortex, the monocyte TSPO expression at late disease increased by 2.1- to 4.5-fold (Fig. S4; Table S3), with monocyte TSPO expression in whole brains strongly correlated with [¹⁸F]FEPPA uptake (ρ = 0.92; p = 0.0002) (Fig. 3e).

In contrast to microglia and monocytes, TSPO expression in granulocytes transiently increased at mid disease relative to pre-infection (2.1-fold higher, p = 0.004) (Fig. 3b, Table S2), and the strength of correlation between granulocyte TSPO expression during disease and brain tracer uptake (ρ = 0.77; p = 0.009) was weaker than either microglia or monocytes (ρ = 0.89–0.92) (Fig. 3f). [¹⁸F]FEPPA uptake in micro-dissected brains also correlated positively with TSPO expression on both total immune (CD45⁺) cells (Fig. S5a) and myeloid cell subsets—microglia, monocytes, and granulocytes (Fig. S5b) isolated from the different brain regions, replicating our observations in whole brains. These findings confirmed that TSPO expression in ZIKV brains, especially during late disease, is contributed by CD45⁺ myeloid cells—primarily monocytes and microglia.

Since TSPO is expressed by all immune cells assayed in the brain (Fig. 3b), albeit at varying levels, we hypothesized that an increase in immune cell numbers in the brain—either by infiltration from the periphery or in situ proliferation—may also contribute to the overall increase in TSPO expression in the brain. We therefore evaluated absolute counts of immune cells at different disease stages. The total CD45⁺ immune cell population in late ZIKV whole brains was 6.2-fold higher than at pre-infection (p < 0.001) (Fig. 4a; Table S2b). Similarly, total CD45⁺ immune cell counts at late disease increased by 3.0- to 7.2-fold in various brain regions, most prominently in the hippocampus (Fig. S6; Table S4). The trend of increased cell numbers with disease progression was also observed in different myeloid and lymphoid cell subsets. In whole brains at late disease, granulocytes increased by 21.5-fold (p < 0.001); dendritic cells increased by 13.5-fold (p < 0.001); and monocyte-derived macrophages (Mo-MAC) increased by 6.4-fold (p < 0.001) (Fig. 4a; Table S2b). Of the various brain regions, the hippocampus recorded the greatest fold-change increase in infiltrating myeloid cells at late disease (53-fold increase in granulocytes; 47-fold increase in dendritic cells; and 20-fold increase in Mo-MAC) (Fig. S6; Table S4). Similarly, in whole brains, total T cells at late ZIKV increased by 14.6-fold (p < 0.001), and helper T cells increased by 4.1-fold (p = 0.02). Cytotoxic T cells, which were the least abundant CD45⁺ cell subtype detected pre-infection, increased by a substantial 141.8-fold in late ZIKV disease (p < 0.001) (Fig. 4a; Table S2b). Of the various brain regions, the hippocampus again exhibited the highest increase in lymphoid cell counts (20-fold increase in total T cells; tenfold increase in helper T cells; and 53-fold increase in cytotoxic T cells) (Fig. S6; Table S4). Importantly, absolute counts of total CD45⁺ cells in the brain exhibited strong positive correlation with brain [¹⁸F]FEPPA uptake (ρ = 0.76; p < 0.0001) (Fig. 4b). Strong positive correlation was also observed between brain tracer uptake and absolute cell counts of all myeloid and lymphoid cell subsets evaluated, the most notable of which were the granulocytes, dendritic cells, and monocyte-derived macrophages (Mo-MAC) (Fig. 4c–d). These results established that increased TSPO expression in the brain is also contributed by increased numbers of immune cells—mainly the granulocytes and dendritic cells. Altogether, the TSPO expression profile and immune cell landscape in ZIKV brains provide strong support of TSPO as a biologically relevant imaging target for ZIKV neuroinflammation in our infection model.

Despite the demonstrated utility of both [³H]PK11195 and [¹⁸F]FEPPA to identify ZIKV-driven neuroinflammation with in vitro, ex vivo, and in vivo assays (Figs. 1–2), we noted a 2.3-fold increase in [¹⁸F]FEPPA blood pool activity at late disease (Fig. 2b). We therefore assessed TSPO expression in blood cells over the course of ZIKV disease. TSPO expression on total CD45⁺ immune cells in the blood decreased during ZIKV disease, and similar trends were observed with circulating granulocytes and monocytes in the blood (Fig. 5a; Table S5a). Moreover, TSPO expression in various immune cell subsets (i.e., total CD45⁺ cells, myeloid, and lymphoid cells) did not correlate with blood [¹⁸F]FEPPA activity (Fig. S7). These contrasted with TSPO expression of these immune cells in the brain (Fig. 3b, c, f). On the other hand, total immune cell counts in the blood exhibited 3.7-fold increase in late ZIKV relative to pre-infection (p = 0.04), and the increased numbers were contributed primarily by granulocytes (6.6-fold increase, p = 0.02) and cytotoxic T cells (4.5-fold increase; p = 0.04) (Fig. 5b; Table S5b). There was a strong positive correlation between total CD45⁺ cell counts in the blood and [¹⁸F]FEPPA blood uptake (ρ = 0.93; p = 0.0003) (Fig. 5c), and this trend was reflected in granulocytes (ρ = 0.95; p < 0.0001), monocytes (ρ = 0.78; p = 0.01) (Fig. 5d), as well as lymphoid cells evaluated (Fig. 5e). These findings demonstrate that the increased blood pool [¹⁸F]FEPPA activity in late ZIKV disease mice is due to the increased presence of true TSPO targets.