Organotypic culture of human brain explants as a preclinical model for AI-driven antiviral studies

Abstract

Viral neuroinfections represent a major health burden for which the development of antivirals is needed. Antiviral compounds that target the consequences of a brain infection (symptomatic treatment) rather than the cause (direct-acting antivirals) constitute a promising mitigation strategy that requires to be investigated in relevant models. However, physiological surrogates mimicking an adult human cortex are lacking, limiting our understanding of the mechanisms associated with viro-induced neurological disorders. Here, we optimized the Organotypic culture of Post-mortem Adult human cortical Brain explants (OPAB) as a preclinical platform for Artificial Intelligence (AI)-driven antiviral studies. OPAB shows robust viability over weeks, well-preserved 3D cytoarchitecture, viral permissiveness, and spontaneous local field potential (LFP). Using LFP as a surrogate for neurohealth, we developed a machine learning framework to predict with high confidence the infection status of OPAB. As a proof-of-concept, we showed that antiviral-treated OPAB could partially restore LFP-based electrical activity of infected OPAB in a donor-dependent manner. Together, we propose OPAB as a physiologically relevant and versatile model to study neuroinfections and beyond, providing a platform for preclinical drug discovery.

Keywords: Artificial Intelligence; Bunyavirus; Neurotropic Virus; Small Molecule; Tahyna Virus.

Figure 1. Schematic procedure and histochemical characterization of organotypic human cortical slice culture (OPAB).

(A) The scheme represents the procedure to obtain and culture OPAB. The first step of the procedure (left panel) is to harvest brain tissue by cutting small cubes (about 0.5 cm³) from the cortical area of interest (frontal or parietal). Then, 300 µm-thick tissue slices are cut using a vibratome (middle panel). The brain slices are transferred and cultured at air–liquid interface (right panel). (B, C) Immunohistochemistry of Hemalin-Eosin (HE; B) or NeuN (neuronal marker; C) of OPAB when harvested from patients (0 DIV), or after 12 days in culture (12 DIV). Scale bar: 100 µm. (D, E) Immunohistochemistry of OPAB at 0 DIV for leukocytes (D, CD45) and T lymphocyte (E, CD3). The right panels correspond to magnifications from the squares, and another zoom-in highlighting blood vessels with CD3-positive cells. Scale bar: 100 µm. Source data are available online for this figure.

Figure 2. Schematic procedure and histochemical characterization of organotypic human cortical slice culture (OPAB).

(A) Three-dimensional confocal snapshot of a parietal OPAB cultured for 5 days ex vivo showing the organization of mature neurons (MAP2, cyan), astrocytes (GFAP, orange), and cell nuclei (Dapi, blue). Scale bar: 100 µm. (B–D) Snapshots from 3D confocal imaging (top view) of frontal OPAB, showing astrocytes (GFAP, B), blood vessel endothelial cells (CD31, C), and microglial cells (Iba1, D). Scale bars: 10 µm. (E) Frontal OPAB cultured for 16 days ex vivo stained with a Live/Dead marker, showing healthy cells (green), dead cells (red), and dying cells in yellow. The micrograph shows a gallery of confocal Z plans acquired at 1 μm interval, from the bottom of the slice near the coverslip (upper left) toward 30 µm within the slice (lower right). (F–H) The bar graphs correspond to the measurement of LDH release, accounting for cytotoxicity in untreated young (F) or older (G) OPAB, or young OPAB treated with 10 µg/ml Gambogic Acid (H), a neurotoxic agent, after indicated days in culture. Each graph corresponds to one donor and at least three individual slices were assessed per condition. Source data are available online for this figure.

Figure 3. Assessment of OPAB electrical activity.

(A, B) The micrographs represent high-resolution images of frontal OPAB stained with the presynaptic marker Bassoon (cyan) and post-synaptic marker Homer-1 (orange). Scale bar: 10 µm. (B) Magnification of Bassoon-Homer-1 colocalizing events from the white squares in (A). (C) Three-dimensional confocal imaging of neurons (MAP2), presynaptic proteins (Bassoon, FLRT3) and post-synaptic proteins (Homer-1 and LPHN3) in frontal OPAB. White arrows highlight trans-synaptic structures, positives for pre- and post-synaptic markers. Scale bar: 2 µm. (D, E) Frontal OPAB were seeded onto 3D-MEA (D) and electrical activity was recorded. (E) Example of traces of extracellular electrical signal as a function of time, recorded from one electrode (left), and overlay cut-outs showing the shape, amplitude, signal to noise ratio, and dynamics of isolated spike signals. (F) Single-donor assessment of OPAB spontaneous electrical activity. Data were processed as per LFP data processing indications (see Methods for details). Data is displayed across slices for each donor and the mean +/− SD from 3 recordings per slice is shown. Conditions for Donor B and B′ correspond to a single donor for which electrical activity recording was performed twice, at 1-week interval. (G) Frequency spectrum of the spontaneous LFP of OPAB displayed across slices across donors. Data shows mean +/− SD from 3 slice per condition. (H) Confusion matrix of the Random Forest Classifier (RFC)-based machine learning algorithm trained across donors, and tested on the same conditions (“unseen” dataset). In every case, the first number describes the number of inputs from the X axis classified as the condition in the Y axis. CUP_min = 0.25. (I) Principal component of OPAB LFP across donors. The confidence ellipses represent the covariance of the given variables principal component 1 (PC1) and principal component 2 (PC2). The percentages of the X and Y axis represent the variance ratio of the components. Source data are available online for this figure.

Figure 4. TAHV infection and antivirals’ evaluation in Vero E6 and neuronal cells.

(A, B) Vero E6 cells were infected for 1 (A) or 2 (B) days with TAHV at MOI 0.1 either non-treated (NT) or treated with 100 nM Bafilomycin A1 (Baf A1), 10 µM RG10b or 20 µM Rottlerin. Quantification of TAHV RNA, normalized by GAPDH RNA (control), was measured by RT-qPCR and data are represented as fold change to the TAHV infected non-treated (NT) condition. The data correspond to the mean +/− SD from three individual experiments performed in duplicates. Unpaired t-test p value < 0.0001 (****). (C, D) RG10b IC₅₀ and CC₅₀ measurement in Vero E6 cells by RT-qPCR (C) and plaque assay (D). Data normalized to highest Ct values (RT-qPCR) and titer (Plaque assay) = 100% infection. Cytotoxicity measured by LDH assay and normalized to positive control = 100%. The data correspond to the mean +/− SD from two individual experiments performed in duplicates. (E) Neuronal LUHMES cells were infected with TAHV for 48 h at MOI 0.1 and stained with anti-MAP2 antibody and smFISH targeting TAHV RNA. Confocal snapshots show bona fide neuronal differentiation (MAP2, green), and TAHV infection (TAHV smFISH, magenta). Scale bar: 50 µm. (F, G) Neuronal LUHMES cells were infected with TAHV for indicated time at MOI 0.1 and TAHV RNA expression levels were measured by RT-qPCR (F), and cytotoxicity was measured in the supernatant using LDH assay (G). The data are means +/− SD from biological duplicates or triplicates from three independent experiments. Unpaired t-test p value < 0.05 (*), < 0.01 (**); nd: not determined. Source data are available online for this figure.

Figure 5. TAHV infection and antivirals’ evaluation in OPAB.

(A) Parietal OPAB from 5 donors infected with 10⁶ pfu TAHV for indicated days were either DMSO-treated (DMSO) or treated with 10 µM RG10b or 20 µM Rottlerin. Quantification of TAHV RNA, normalized to GAPDH mRNA levels, was measured by RT-qPCR and data are represented as fold change normalized to conditions at 0 dpi to equal 1. The data correspond to the mean +/− SEM from 5 donors (n = 10 slices for 0, 2, and 4 dpi and n = 8 slices for 7 dpi) performed in duplicates/triplicates. Unpaired t-test p value = 0.031 (*). Differences between all other conditions are non-significant. (B) OPAB were either non-infected (Non-inf), infected with 10⁶ pfu TAHV or with equivalent volume of UV-inactivated TAHV. After 48 hpi, quantification of TAHV RNA, normalized by GAPDH RNA, was measured by RT-qPCR. The data correspond to the mean +/− SD from 5 or 6 slices from 2 donors. Unpaired t-test p value < 0.0001 (****). (C) Cytotoxicity in non-infected (NI) and TAHV-infected OPAB with 10 µM RG10b or 20 µM Rottlerin treatment at 0, 2, 4, and 7 dpi. Cytotoxicity was determined by measuring levels of released LDH in the culture supernatant and normalized to a positive control (lysed OPAB) = 100%. Data are mean +/− SD and with n = 4 or 5 slices from at least 2 donors. No significant differences were measured. (D) Example of plaque assays from TAHV production in Vero E6 cells at 2 dpi and from harvested supernatant from TAHV-infected OPAB at 7 dpi. Source data are available online for this figure.

Figure 6. Machine learning algorithm to evaluate OPAB neurohealth based on LFP.

(A) Accuracy of RFC-based machine learning algorithm trained on the Mock-TAHV classification of LFP recording at 48 hpi, tested on the same OPAB at 0, 0.5, 48, and 96 hpi. The model shows high efficiency when tested on 48 and 96 hpi LFP recordings, and presents quasi-random predictions when tested on 0 and 0.5 hpi LFP recordings. Data are mean +/− SD from 3 donors, 3 samples per donor, 3 recordings per samples. (B) The dot plot represents the first three Principal components after a principle component analysis (PCA) of the amplitude-frequency spectrum for the Mock and TAHV conditions recorded at 48 hpi. (C) Frequency spectrum for Mock and TAHV infected conditions recorded at 48 hpi, across all donors. The X-axis are the 300 values used to train the RFC model, and the unit is reported to the equivalent frequencies. The standard deviation is obtained across all slices of all donors. The outlier OPAB slice 7 shown in Fig. 3F donor B′ has not been included in the analysis. (D) The graphs show Features importance for classification between Mock and TAHV-infected conditions recorded at 48 hpi, across donors. The X axis are the 300 features used to train the model, and the importance is scaled and quantified by arbitrary units. For each donor, the feature importance profiles are strongly different, showing that the prominent frequencies affected were different in each donor. (E) Confusion matrix of RFC model trained on condition Mock and TAHV-infected recordings at 48 hpi, and tested on those conditions in the presence of 10 µM RG10b (RG10b-treated TAHV) and recorded at 48 hpi. The first number describes the number of inputs from the X axis classified as the condition in the Y axis. For the donors A and B, RG10b-treated TAHV is partially classified as Mock and as non-treated TAHV. CUP_min = 0.5. Source data are available online for this figure.

Figure EV1. Immunohistochemistry of OPAB.

(A, B) Immunochemistry of Hemalin-Eosin (H&E), neurons (NeuN), oligodendrocytes (Oligo2) and astrocytes (GFAP) from parietal (A) or frontal (B) areas of OPAB at indicated days in vitro (DIV). The condition 0 DIV correspond to immediate sample collection at autopsy, before vibratome slicing and OPAB. The black squares correspond to the zoomed area on the right panels. Images were acquired on Leica Thunder using 20x magnification. Scale bar: 100 µm.

Figure EV2. Characterization of astrocytic and immunoreactivity of OPAB.

(A, B) OPAB were cultured for indicated days in vitro, RNA was extracted and RT-qPCR was performed to measure the expression levels of GFAP- (A) and SOX9-coding (B) mRNA over time. Data are mean +/− SEM and each dot corresponds to the measurement of the mRNA levels per slice, coming from at least two donors. (C, D) OPAB were fixed at 4 DIV (C) or indicated times. (D) and stained with anti-SOX9 antibody, labeling the nucleus of astrocytes. (C) Representative 3D confocal image of SOX9 staining. Scale bar: 50 µm. (D) Quantification of the number of SOX9-positive nuclei per field of view at indicated time post in vitro culture. Unpaired t-test p value < 0.05 (*); ns: non-significant. Data are mean +/− SD from 8 fields of view taken from two slices from two donors. (E) Samples processed as in A-B were assessed for expression of IL1β-coding mRNA. Of note, one slice at day 4 post culture exhibit oddly high values preventing reliable statistical analysis. Data are mean +/− SEM and each dot corresponds to the measurement of the mRNA levels per slice, coming from at least two donors. (F) Immunohistochemistry of OPAB at 0 DIV showing absence of monocyte/macrophage/microglia (CD163, brown) staining in the cortex. The white arrow highlights a blood vessel at which CD163 staining is elevated, indicating that monocyte/macrophages did not extensively infiltrated the cortex.

Figure EV3. Characterization of OPAB by immunofluorescence.

(A) Snapshots from 3D confocal imaging (top view) of frontal OPAB, showing parallelly aligned post-mitotic neurons (MAP2 in cyan) and nuclei (Dapi in white). Scale bar: 50 µm. (B, C) Three-dimensional imaging of neurons (Tuj1 and MAP2), astrocytes (GFAP) and microglial cells (Iba1) from a frontal OPAB at 5 DIV. (D) Neuronal LUHMES cells treated for 24 h with 10 µg/ml Gambogic acid (Gambo), and cytotoxicity was measured in the supernatant using LDH assay. Unpaired t-test p value < 0.05 (*). Data are mean +/− SD from 2 individual experiments performed in triplicates. Controls were performed once per experiment. (E) OPAB were treated for 24 h with 10 µg/ml Gambogic acid and anti-Cleaved-Caspase-3 antibody labeling was performed. Confocal snapshots highlight brighter Cleaved-Caspase-3 staining upon Gambogic acid treatment. Scale bar: 50 µm.

Figure EV4. Characterization of TAHV smFISH in Vero E6 cells.

(A) Vero E6 cells were infected with TAHV for 24 h at indicated MOI and stained with phalloidin (actin, red), TAHV smFISH (viral RNA, orange), and Dapi (nuclei, blue). Scale bar: 25 µm. Lower right panels correspond to two magnified crops from indicated conditions, highlighting dotted structures likely corresponding to viral factories. (B) Quantification of the number of large dotted structures from the TAHV smFISH staining in (A). Data are mean +/− SD from 10 fields of view per condition from an experiment. Unpaired t-test p value < 0.01 (**) or <0.0001 (****).