Zika virus-induced hyper excitation precedes death of mouse primary neuron

All experimental assays were done using Zika virus (ZIKV) human isolate from Cambodia 2010 (Genbank KU955593 [30]) grown in Vero cells. Dengue virus type 2 ET300 (DENV2) isolated in Australia from a soldier returning from East Timor was passaged 7 times in C6–36 cell line.

Neuronal cultures for confocal microscopy assays, TCID₅₀ and RT-PCR: Tissue culture plate wells and coverslips for confocal microscopy were pre-coated with 100 μl of polyethyleneimine (PEI, Sigma, 0.05% in Borate-buffered solution) at room temperature (RT) for 30 min, rinsed 3 times with Tissue culture treated water and let to dry out in the Biosafety Cabinet class II (BSCII). Cells were plated in a 24-well plate with 25 μl from a stock solution of 3.10⁶ cells/ml. The plate was placed in incubator for half an hour for cells to adhere before adding 1 ml of media. Cells were cultured at 37 °C with 5% CO2. Wells were topped up with Neurobasal media for mouse neurons completed with B27 Supplement, Glutamax and Gentamycin.

Neuronal cultures on Microelectrodes array: The method modified from Hales et al. [31] was used. Microelectrodes array (MEA) were pre-coated before cell seeding. The coating consists of 100 μl PEI at RT for 30 min, followed by TC-water rinses (3 times); laminin (0.02 mg/ml, Sigma L-2020) is finally added for 20 min at 37C and 5% CO2, just before plating. Cells were plated in the center of the MEAs at 7.5.10⁴ cells per device, corresponding to a volume of 25 μl from a 3.10⁶ cells/ml stock cell solution. After half an hour in the respective incubator for cells to adhere to the bottom of the MEA well, 1 ml of cell media was added. Half of the culture medium was changed every 3–4 days excluding the day before recording.

For virus dynamics titration, 1 ml of supernatant from neuron cultures was sampled in triplicates at 24, 48 and 72 h post infection (hpi) and used for TCID₅₀ assays.

Viral titers at the different time points were determined by end point titrations (TCID₅₀) in VERO monolayer cell cultures. In a 96-wells TC plate, seven 10-fold dilutions of sampled supernatants were used onto Vero cells in triplicates at 37 °C with 5% CO2. Each plate was screened under inverted microscope for cytopathogenic effect at 3 and 5 days post infection and to determine TCID₅₀ using Spearmann-Karber calculation method [32].

For gene expression quantification, total RNA was collected from cell culture at 12, 24, 48 and 72 hpi. Media was discarded and replaced with 200 μl fresh media. Cells were removed from the bottom by gentle scraping and vigorous pipetting. Total RNA was extracted using RNeasy Plus Mini Kit (Qiagen Sciences, Maryland, MA). cDNA was prepared using random hexamers and Superscript-III reverse transcriptase (Thermo Fisher Scientific Inc. Australia) as per manufacturer’s protocol. Real-time PCR assay was performed using the SYBR® Premix Ex Taq™ II (Takara-Bio Inc., China) and running on a QuantStudio™ 6 Flex Real Time PCR System (Applied Biosystems). Forward and reverse primers of individual targeted gene are given in Additional file 1: Table S1. Settings are: 95 °C for 30 s, followed by 45 cycles of 95 °C for 5 s, 55 °C for 40 s, followed by melt-curve stage. The 2^ΔΔCt values were calculated at each time point for each gene as the fold-increase over uninfected control at the same time point. Samples are made in triplicates for each value.

Samples preparation and confocal microscopy: At different time points after infection (0, 1, 2, 3 and 7 days post infection or dpi), primary neuron cultures grown on coverslips were fixed by adding 300 μl of 4% paraformaldehyde in 0.05 M Phosphate Buffered Saline (PBSA) for 20 min under gently shaking. The coverslips were washed gently three times for 5 min using 1 ml of PBSA. Cells were permeabilized with 1 ml of 0.1% Triton X-100 (Sigma-Aldrich) in PBSA for 10 min and rinsed three times in PBSA. Non-specific binding was blocked with 0.5% BSA in PBSA for 30 min. Primary antibodies were diluted in 0.5% BSA in PBSA and incubated for 1 h at room temperature. The following primary antibodies were used: DAPI (32,670, Sigma-Aldrich), 1:200 guinea pig anti NeuN (266,004, Synaptic System), 1:500 rabbit anti Synapsin 1/2 (106,002 Synaptic System), 1:1000 chicken anti MAP2 (ab5392, ABCAM), 1:500 rabbit anti GFAP (ab7260, ABCAM), 1:200 human 4G2 anti pan-flavivirus (ab00230–10.0, Focus Bioscience). Coverslips were washed three times with PBSA for 5 min. Coverslips were then incubated in species-specific secondary antibodies diluted in 5% BSA in PBSA for 1 h at room temperature, followed by two washes of 5 min with PBSA. Coverslips were then washed twice with water and counterstain nuclei with DAPI for 10 min, before a final wash in water and mounted carefully on glass slides. The slides were observed with a 63 × objective (with oil) using a Leica SP5 confocal microscope for signal quantification and a ZEISS LSM 800 confocal microscope for virus detection in the cultures. For the number of coverslips and images taken per treatment and condition to quantify MAP2, Synapsin, NeuN and DAPI antibody signals, see Additional file 2: Table S2.

Pictures data analysis: To quantify antibody signals, raw images were extracted with LAS AF Lite software (Leica Microsystems, Germany) in single channel format for analysis. Image analysis is done with ImageJ software [33]. Images were first converted into 16-bits format with default threshold (dark background, B&W parameters). Results were normalized with the number of cells by count of DAPI particles. For DAPI counting, the option “Watershed” was applied before analysis to avoid counting merged particles. The Plugin “Analyze particles” was then used with the parameters “0.00–1.00” for circularity and “100-Inf” for size pixel.

The same type of analysis was done for NeuN counting. For Synapsin, the signal was automatically calculated by the “Measure”, and “Raw integrated density” options. For MAP2 analysis, the same processing of the images was done by converting into 16-bits and applying a default threshold. The signal of the antibody was automatically calculated by the “Measure” and “mean gray value” options.

The presence of the viruses in the primary neuron cultures was assessed by using the ZEN software from Zeiss microscopy, with the Z-stack capture option. An optical section area was captured every 0.34 μm on a total section distance of 6.8 μm. The final image was generated with the maximum intensity projection from the Extended Focus module to obtain a depth of field from the previously acquired Z-stacks.

Recording material: Microelectrodes Array (MEA) were used for electrophysiological recording of neuronal networks (60MEA 200/30iR-Ti-gr, Multi-Channel Systems, Reutlingen, Germany). The glass devices consist of 59 TiN/SiN planar round electrodes (10–30 μm diameter; 100–200 μm center-to-center inter-electrode intervals) arranged in a square grid without corners. A single larger electrode served as reference ground electrode, replacing one recording electrode. Online cellular activity was recorded using the MEA60inv System (Multi Channel Systems, Reutlingen, Germany). Action potential was recorded at each electrode sampled at 10 kHz. MC_Rack software (Multi Channel Systems), installed on the acquisition computer, allowed files acquisition. Data analysis was performed off-line using MC_Rack software by Multi Channel Systems and NeuroSigX software developed by researchers at Institute for Intelligent Systems Research and Innovation, Deakin University, Australia.

Recording method: All manipulations and recordings were made in a biosafety cabinet. Each recording session consisted of 30 min of spontaneous activity recording at 0, 1, 2, 3, and 7 dpi or 15 min of each post stimulus activity recording at 8 dpi for gabazine stimulation assays. Every record started 5–10 min after placing the MEA on the amplifier to avoid neuron response to mechanical stress due to movement and to allow adaptation to the biosafety cabinet conditions and temperature set by the adaptor instead of incubator. MEAs temperature was maintained at 37 degrees Celsius through the recording. Half of the media was changed after each recording session. The raw continuous voltage traces were filtered to remove the traces of field potential below 200 Hz generated by the collectively charged network. The high pass filter was comprised of a 2nd order Butterworth filter with cutoff frequency of 200 Hz. The spike detection threshold was set at 8 times the high-pass filtered signal’s standard deviation [34] (as measured by MC_Rack software, 22 μV) within a 500 ms window.

Recording time points: Once the dissociated neurons from mouse embryos were plated in the MEA, the cultures were let to grow and maturate for 7 days. At 7 div, these primary neuron networks were infected with ZIKV or DENV2 (used as a positive control) at a MOI of 0.2.

For each experimental assay, 6 electrophysiological recordings of spontaneous activity were made. The first was done at 7 div before infection and set as reference for further analysis. After infection, records were sampled at 1, 2, 3 and 7 days post infection (dpi) (Additional file 3: Figure S1A). An additional recording was made at 2 dpi by introducing the D(−)-2-amino-5-phosphonopentanoic acid (APV) (Sigma Aldrich) at 300 μM, a NMDA-R antagonist [29]. The last recording took place at 8 dpi for stimulation with 20 μM of gabazine [35], a GABA_A antagonist (Tocris SR 95531 hydrobromide). Before the introduction of the APV or gabazine, a prerecording was done by adding the same volume of solvent (water) in the well as a reference for the analysis of the stimulus.

Microelectrode array data analysis with MC_Rack: MEA data analysis is performed offline with MC_Rack software (Multi Channel Systems, Reutlingen, Germany). Active electrodes (AE) are selected if the spike rate for the electrode is equal or higher than 0.01 spike per second. Burst electrodes are detected with the following parameters: maximum interval to start burst = 100 ms, maximum interval to end burst = 100 ms, minimum interval between bursts = 100 ms, minimum duration of burst = 10 ms, and minimum number of spikes in burst = 3 (adapted from [36], see (Additional file 3: Figure S1B). The relative total spike (TS) number changes per electrode is calculated by: Ratio = ln (TS at time point dpi)/(TS at 0 dpi) for spontaneous spike activity. For post gabazine analysis, the TS reference is changed with the activity post stimulation by solvent (water).

Microelectrode array data analysis with NeuroSigX: MC_Rack files are cut into 6 filtered files of 5 min for spontaneous activity at the different times post infection and 6 files of 2 min post gabazine stimulation at 8 dpi. Mcd files are then converted into txt files (MC_Data tool, MultiChannel System, Reutlingen, Germany) and analyzed by NeuroSigX. The NeuroSigX software uses novel spike sorting and data analysis algorithms to explore the neural spike activity and spatio-temporal behavior of the neuronal network [37, 38]. A threshold of 22 μV is employed to maintain the analytical consistency between the preliminary analysis by MC_Rack software and analysis by NeuroSigX. Raster plots, electrode activity maps and mean spike activity data and figures are extracted for temporal and spatial analysis and illustration.

Graphs and statistical analysis are done using GraphPad Prism 5 software. All statistical tests are done using a two-tailed analysis. All statistical results are expressed with the p-value using the following annotations: ns for P > 0.05, * for P ≤ 0.05, ** for P ≤ 0.01, and *** for P ≤ 0.001.

Given the established neurotropism of ZIKV during mammalian infection, functional impact of Zika infection in Mus musculus embryonic cerebral cortical cells cultured on MEA was tested. Dengue virus serotype-2 (DENV2) was used as a positive viral control. At 2 dpi, virus infections triggered hyperactivity in primary neuron cultures (Fig. 1 and Additional file 4: Figure S2A and B). At 3 dpi, the activity of infected primary neuron cultures decreased and reached an average individual electrode activity comparable to uninfected cultures (Additional file 4: Figure S2B). However, at 7 dpi, the spiking activity was significantly and dramatically decreased in ZIKV infected neuron cultures, while DENV2 infected neuron cultures remained hyperactive (Fig. 1). GABA_A or ɤ-Aminobutyric acid, is an amino acid inhibitory neurotransmitter widely present in the nervous system [39, 40]. In order to exclude the possibility of a GABA_A-induced inhibitory effect, cultures from primary neuron model were treated with gabazine (SR 95531 [2-(3′-carboxy-2′-propyl)-3-amino-6-p-methoxyphenylpyridazinium bromide]), a GABA_A antagonist, known to remove neurotransmitter inhibition hence stimulating the firing activity of neuronal network [35]. The electrophysiological response [41, 42] was recorded in primary neurons and we found a lack of relative response to gabazine compared to solvent in cultures infected by ZIKV (Fig. 2a and b), as shown on a 3D electrodes map (Fig. 2c). Gabazine or water (solvent) was unable to provoke any response in ZIKV infected primary neurons at 7 dpi, confirming the extreme functional silencing of neuronal activity.

To determine ZIKV replication dynamics in mouse neurons, primary neurons from embryonic mice were infected with ZIKV and viral titers in the media supernatant were measured over time. The results showed that ZIKV replicated in primary neuron cultures until 2 dpi (Fig. 3a). However, at 3 dpi, ZIKV titer reduced dramatically by 4 logs. As control, primary neuron cultures were also infected with DENV2 (Fig. 3a). Cultures infected with DENV2 show a growth dynamic with an initial decrease followed by increase in virus titers until 3 dpi. Virus infections in the primary neuron cultures was confirmed by immunofluorescence. Uninfected cultures showed the presence of neurons with MAP2 staining and glial cells with GFAP stain (Fig. 3b). Zika virus and DENV2 staining were detected from 1 to 3 dpi in mouse cultures (Fig. 3c). DENV2 was visible mostly in neurons and in glial cells, yet at lower frequency, from 1 to 3 dpi. Zika virus was present in the cultures at 1 dpi, with signal detected in both in neurons and glial cells. The amount and frequency of ZIKV signal became lower at 2 and 3 dpi, and mostly co-localized with dense DAPI nuclei.

Zika virus infected neurons showed a significant decrease in the neuronal marker Microtubule Associated Protein 2 (MAP2) staining compared to the uninfected and DENV2 infected cultures (Fig. 4a and b). At the same time, ZIKV infected neuron cultures had a significant decrease in NeuN positive cells compared to uninfected cultures (Mann Whitney test, P = 0.008, U = 7.0) and DENV2 infected cultures (Mann Whitney test, P = 0.001, U = 0.0) (Additional file 4: Figure S2C). Moreover, analysis of 3 dpi ZIKV infected neuron nuclei with DAPI staining showed a significantly decreased percentage of large and heterochromatic nuclei (associated with NeuN staining of mature neurons) and an increase in small sized nuclei or fragments compared to uninfected (Fig. 4c) (Mann Whitney tests, for “bin 0” P = 0.0079, for “bin 300” P = 0.0079 and for “bin 400” P = 0.0112). Later during the time-course, results also showed that at 7 dpi, consistent with the functional silencing in spike detection recorded, the significant decrease in the number of matured mouse neurons in ZIKV infected cultures was verified with the NeuN, neuronal nuclear marker (Fig. 5a and b). In contrast, uninfected as well as DENV2-infected cultures showed a well-matured network with a significantly higher number of mature neurons at 7 dpi. These results confirm that, even at moderate level Multiplicity of Infection (MOI = 0.2), ZIKV infection dramatically reduced the number of mature neurons, as shown before [25].

Our results show that the decreasing number of mouse mature neurons at 3 dpi is preceded by neuronal spiking hyperactivity (Additional file 4: Figure S2C). In mammals, glutamate is the main excitatory neurotransmitter [43]. A recent report demonstrates the efficacy of memantine, a blocker of NMDA-R, in decreasing Zika-induced neurodegeneration in the brain of infected mice, implying a possible toxic role of the excitatory neurotransmitter glutamate [25]. Real time RT-PCR results showed an early significant increase in the mRNA expression of EAAT (Excitatory Amino Acid Transporter, a glutamate transporter at post-synaptic level) at 24 h post infection (hpi), but returned to baseline level at 2 dpi (Fig. 6a). On the other hand, mRNAs of the Glutamate Dehydrogenase (GD1), a mitochondrial enzyme responsible for glutamate oxidation and recycling towards the tricarboxylic cycle by mouse astrocytes [44] (Fig. 6b), as well as the VGlut (Vesicular Glutamate Transporter, a cytoplasmic pre-synaptic glutamate transporter) showed no substantial change over time after infection in primary neurons (Additional file 5: Figure S3). Regarding the GABA_A pathway, pre-synaptic GABA_A transporter 1 (GAT1) mRNA was significantly over-expressed only transiently in ZIKV infected neurons at 24 hpi (Fig. 6c) while the expression profile of the post-synaptic GABA receptor remained unchanged over time (Additional file 5: Figure S3). Finally, the Voltage-gated sodium channel (VGNaC), responsible for the propagation of action potential, mRNA expression increased in mouse neurons with a peak at 48 hpi (Fig. 6d), corresponding to the hyper-excitatory phase recorded with MEA (Fig. 1). These results indicate that ZIKV infection leads to differential gene expression of glutamate and GABA pathway genes in mouse primary neurons at early stage of infection with ZIKV. The early and transient upregulation of EAAT could decrease the synaptic concentration of the excitatory neurotransmitter glutamate by increase of its transporter. As well, the brief upregulation of GAT1 may improve the presynaptic availability of the inhibitory neurotransmitter GABA. However, these gene regulation are not enough to compensate the long-lasting excitation post infection and first detrimental effects are observed at 3 dpi (Figs. 3 and 5).

Immunofluorescence result showed that at 7 dpi, there was a decrease of the phosphoprotein Synapsin I/II signal in ZIKV infected neuron networks compared with uninfected control (Fig. 5a and c). Along with mRNA data, the decrease in Synapsin protein in mouse primary neurons after ZIKV infection confirms the importance of this protein in neuronal networks and its potential role in neuronal damage mechanisms post infection, as it has been shown to accelerate synaptic vesicle traffic during repetitive stimulation [45]. Except the EAAT and GAT1 transient and potentially protective regulations, the absence of durable regulatory mechanisms in mouse neural tissue to offset the excitatory effect of glutamate, notably the absence of upregulation of GD1 for recycling glutamate by astrocytes [46], could explain the extensive neuronal death [25], over and above the viral cytopathogenic effect.

To test the hypothesis that the hyper excitation observed in ZIKV infected mouse neurons (Fig. 1) was due to an increase of glutamate neurotransmitter, NMDA-R antagonist, APV, was introduced to the mouse neuron cultures on MEA at 2 dpi. The analysis of the electric activity post APV introduction was compared to the neuron activity after introduction of the solvent (water). The antagonist caused a decrease of the network activity with 20% less AE for uninfected and ZIKV infected cultures (Fig. 7a). However, uninfected neuron cultures had a significantly lesser decrease of individual spike activity compared to ZIKV infected neuron cultures with ln(TS_APV/TS_Water) = − 0.034, ± 0.08 SEM for uninfected cultures and ln(TS_APV/TS_Water) = − 0.66, ± 0.08 SEM for ZIKV infected cultures (Kruskal-Wallis test, P = 0.019) (Fig. 7b). These results indicate that the electrode hyper activity observed at 2 dpi in ZIKV infected cultures was mostly due to the excitation of NMDA-R and glutamate neurotransmitter.