Astrocytic Atrophy Following Status Epilepticus Parallels Reduced Ca2+ Activity and Impaired Synaptic Plasticity in the Rat Hippocampus.

Pilocarpine model of epilepsy All procedures were carried out in accordance with University of Nizhny Novgorod regulations. 3–6 weeks old male Sprague-Dawley (Wistar for LTP experiments) rats were injected with lithium chloride (127 mg/kg, Sigma Aldrich) 20–24 h prior to pilocarpine and methylscopolamine (1 mg/kg, Sigma Aldrich) 20 min prior to pilocarpine. Then pilocarpine (Tocris), 10 mg/kg was injected every 30 min (but no more than 60 mg/kg) to induce SE which characterized with generalized seizures lasting for at least 20 min (Supplementary Figure 1 ). To reduce mortality, phenazepam 1 mg/kg was injected every 10 min for 30–40 min after 20 min of generalized seizures.

Nissl staining

Brain tissue was prepared according to routine histologic methods (Singh et al., 2008 ). Briefly, brains were removed immediately after decapitation, immersed in ethanol 96% and embedded in paraffin after dehydration. Paraffin sections (5 μm) were cut in a coronal plane and stained with Nissl's method. Each sixth staining section was chosen for quantitative analysis for each animal. Images of CA1, CA3, hilus, and dentate gyrus were obtained using an x40 magnification. The neurons were counted per 100 μm for cell layer in each area using the plugin “Cell counter” for ImageJ.

Hippocampal slice preparation

The slices were prepared 2–4 weeks after SE. The animals were anesthetized with Isoflurane (1-Chloro-2,2,2-trifluoroethyl difluoromethyl ether) and then decapitated. The rest of the procedure was slightly different for whole cell and field potential recordings.

Preparation for whole-cell recording and imaging

The brains were exposed, and then chilled with ice-cold solution containing (in mM): 50 sucrose; 87 NaCl; 2.5 KCl; 8.48 MgSO4; 1.24 NaH 2 PO 4 ; 26.2 NaHCO 3 ; 0.5 CaCl 2 ; 22 D-Glucose. Hippocampi from both hemispheres were dissected, isolated, and transverse slices (350 μm) were cut using a vibrating microtome (Microm HM650 V; Thermo Fisher Scientific) and left to recover at 34°C for 1 h in a submerged incubation chamber with “storage” solution containing (in mM): 119 NaCl; 2.5 KCl; 1.3 MgSO 4 ; 1 NaH 2 PO 4 ; 26.2 NaHCO 3 ; 1 CaCl 2 ; 1.6 MgCl 2 ; 22 D-Glucose. Then the slices were transferred to the recording chamber and were continuously perfused with a solution containing (in mM): 119 NaCl; 2.5 KCl; 1.3 MgSO 4 ; 1 NaH 2 PO 4 ; 26.2 NaHCO 3 ; 2.5 CaCl 2 ; 11 D-Glucose. All solutions were saturated with carbogen gas mixture containing 95% O 2 and 5% CO 2 . Osmolarity was 295 ± 5 mOsm, pH 7.4. All recordings were done at a temperature of 34°C.

Preparation for field potential recording and LTP induction

The cerebellum and a small section of the frontal cortex were removed. A flat surface for mounting the brain was created by making a cut on the dorsal surface parallel to the horizontal plane. The brain was then mounted onto the stage of the vibratome, and horizontal sections (400 μm thick) were cut in ice-cold artificial cerebrospinal fluid (ACSF). ACSF composed of (in mM): 126 NaCl, 2.5 KCl, 1.25 NaH 2 PO 4 , 1 MgSO 4 , 2 CaCl 2 , 24 NaHCO 3 , and 10 D-glucose was saturated with carbogen. The prepared slices were immersed in a chamber with ACSF, which was placed in a temperature-controlled water bath (35°C) for 1 h. After the incubation, the slices were transferred to the recording chamber, where they were kept for 15–20 min prior to the electrophysiological study. In this chamber, hippocampal slices were perfused with a constant flow of oxygenated ACSF at a rate of 5 ml/min at room temperature. One to five slices from each rat were used in the experiment.

Sholl analysis

Sholl analysis was performed on adaptively thresholded maximal projections of Z-stacks, where each XY-plane has been filtered with anisotropic diffusion filtering. All processing steps were performed using image-funcut library [image-funcut, https://github.com/abrazhe/image-funcut ] and other custom-written Python scripts, using Scikit-Image [scikit, http://scikit-image.org/ ] and Sci-Py [scipy, http://www.scipy.org/ ] libraries (Van Der Walt et al., 2014 ). The step-by-step procedure is summarized in Supplementary Figure 2 . Sholl metric was calculated automatically as a number of intersections of circles with centers at the soma and increasing radii with the thresholded mask obtained as described above. Shearlet-based estimate of spatial complexity and entropy for 2D patterns A spatial pattern can be characterized by a pair of statistical properties, namely entropy and statistical complexity (López-Ruiz et al., 1995 ). An ordered (e.g., periodic) structure with a single spatial scale and preferred feature orientation will have both low entropy and small statistical complexity, as the structure in any part of the system can be reconstructed from a small area. At the other end of the complexity-entropy spectrum, where the state is disordered with no spatial correlations, the entropy of the system will be maximal, while the complexity will again be low (the spatial pattern has the same local statistics). Intermediate cases with high statistical complexity are of more interest, as they represent systems with non-trivial regularities and underlying structure embedded in randomness. We developed an algorithm to map local entropy and complexity values for biologically relevant structures using shearlet transform to induce local probability densities of scale and orientation and Jensen-Shannon divergence to define statistical complexity. Below we describe the two points in more detail. Entropy and statistical complexity Both entropy and complexity (entropic non-triviality) measures for a 2D pattern were defined statistically for a distribution of spatial features, such as orientation, or scale. Here we denoted such a distribution as (1) P : = { P i } for a set of features i = 1 … N . Then entropy was defined simply as Shannon information entropy (2) S [ P ] = - ∑ i P i l o g 2 P i . Entropy will have its maximum for the equiprobable distribution of all features P e , (3) S [ P e ] = S max = 2 N , where N is a number of possible states or features. This allows to introduce normalized entropy: (4) H s [ P ] : = S [ P ] / S [ P e ] , H s [ P ] ∈ [ 0 , … 1 ] Following (López-Ruiz et al., 1995 ) we used the disequilibrium-based complexity measure (5) C [ P ] : = Q [ P , P e ] H s [ P ] i.e., the one based on the statistical distance between the observed ( P ) and equiprobable ( P e ) distributions. Here, following (Lamberti et al., 2004 ; Rosso et al., 2007 ), we employed normalized Jensen-Shannon divergence (6) Q J S = J [ P , P e ] / J max as a measure of distance between two distributions, where Jensen-Shannon divergence is defined as (7) J [ P , P e ] = S [ P + P e 2 ] - 1 2 ( S [ P ] + S [ P e ] ) . Clearly, J [ P, P e ] = 0 if P = P e and reaches its maximum when only one feature, say m , is present, while all others are absent: P i = 1| i = m , and Pi = 0| i ≠ m .

Shearlet transform

Shearlet transform provides a convenient probability density function for spatial entropy and complexity estimates, describing local prevalence of structures with some specific scale and orientation. We used fast finite discrete shearlet transform (FFST) described in detail by (Häuser and Steidl, 2013 ). Here we provide a minimally sufficient description of the FFST and its use in calculation of spatial entropy and complexity. Discrete shearlet transform was based on convolving the digital 2D image (8) I ( x , y ) ∈ R ( N , N ) with scaled, sheared, and shifted copies of a “mother” shearlet function ψ, thus accounting for different scales and orientations of features contained in the image; one uses the dilation matrix A and shear matrix S to create the sheared, scaled and shifted copies of the mother wavelet ψ x : (9) ψ a , s , t = a - 3 / 4 ( A a - 1 S s - 1 ( x - t ) ) . Thus, the scaled and shared copies of ψ pick up dominant anisotropic features at different spatial scales and orientations. In the discrete transform, one uses a fixed number of decomposition scales and shifts as well as scale-dependent number of orientations (more orientations at higher spatial frequencies). Finally, shearlet decomposition of image was given by shearlet coefficients (10) T ( I ) ( j , k , m ) = 〈 I , ψ j , k , m 〉 where discrete shearlet ψ j,k,m = ψ ajsj,ktm ( x ) is the shearlet at discrete scale α j , shear s j,k and shift t m . Thus, T(I) is a set of K images of the same size as I(x,y) , where the value at a specific (x,y) location in the k -th image represents the shearlet coefficient at some specific scale j and shear s . Following ideas from wavelet entropy (Rosso et al., 2001 ) and earlier of spectral entropy of (Powell and Percival, 1979 ), in each location of the studied 2D image I(x,y) , we defined P ( x, y ) = Pk ( x, y ) as normalized power of the shearlet coefficients at this point: (11) P k ( x , y ) = E k ( x , y ) / ∑ j E j ( x , y ) thus, interpreting a spectrum of local feature scales and orientations as a density function. Here ( K σ * · ) denotes convolution with a Gaussian kernel with scale-dependent standard deviation σ j . Volume fraction (VF) of astrocytic leaflets To calculate the VF of the fine process of the astrocyte, we followed a similar method described by Heller and Rusakov ( 2015 ) A line of 45 μm length were drawn from the soma on a single Z plane of the stack. Spatial attention was paid to ensure that fluorescence of soma was not saturated. The estimated VF was calculated with the following: (12) G V ( i , j ) = ( F ( i , j ) - F 0 ) / ( F m a x - F 0 ) where F(I,j) —the fluorescent in particular pixel of the line, F max —the fluorescence of soma, F 0 –the background fluorescence. F 0 was obtained in image area which had no stained astrocytes.

Astrocyte coupling analysis

The astrocytes were loaded with 50 μM Alexa Fluor 594 through the patch pipette for 30 min. then Z-stack two-photon images was obtained (emission band-pass filter 565–610 nm, 512 x 512 pixels). The images were then denoised with block matching 4D (BM4D) free scrip for MATLAB (Maggioni et al., 2013 ; Danielyan et al., 2014 ). The distance to neighboring astrocytes coupled to the target astrocyte through gap-junctions was calculated in 3D-space using Pythagorean theorem with custom-written MATLAB script. Fluorescence Intensities of all coupled cells were normalized to fluorescence of soma of the patched astrocyte. The relationship between distance fluorescence of coupled astrocyte and distance to this astrocyte was fitted by monoexponential function to obtain coupling length constant ( C λ ) (Anders et al., 2014 ): (13) I ( d ) = I 0 e x p ( - d / C λ ) , were, d—distance, I o —the normalized fluorescence intensity of the coupled cell.

Electrophysiological recordings Whole-cell recording

Whole-cell voltage-clamp and current-clamp were performed with Multiclamp 700B amplifier (Molecular Devices). The CA1 str.radiatum astrocytes were visualized with BX51WI (Olympus) or Axio Examiner Z1 (Zeiss) microscope equipped with infrared differential interference contrast. Borosilicate patch pipettes (Resistance 3−5 MΩ) were filled with internal solution containing (in mM): 130 KCH 3 SO 3 , 10 HEPES, 10 Na 2 -phosphocreatine, 8 NaCl, 3 l-ascorbic acid, 2 Mg-GTP (pH adjusted to 7.2, osmolarity of 295 ± 3 mOsm). For simultaneous two-photon imaging, 50 μM Alexa Fluor 594 was added to the internal solution. Bipolar extracellular tungsten electrode (FHC) was placed in str. radiatum between CA1 and CA3 areas. Once whole-cell configuration was obtained, the cell was dialyzed for 5 to 10 min before the start of recording. In voltage clamp recordings the astrocytes were held at−80 mV. Voltage steps were applied to obtain current-voltage (I-V) relationship. In current clamp, current steps were applied to corroborate the absence of membrane excitability. Cycles of 1, 4, and 5 electrical stimuli (100 ms, 50 Hz) were applied to Schaffer collaterals. The intensity of stimulation was set to induce synaptic currents in astrocytes of 20 to 50 pA to a single stimulus. Series and input resistances were continuously monitored by a voltage step of−5 mV after each cycle. Signals were sampled at 5 kHz and filtered at 2.5 kHz. Passive astrocytes were taken at 100−200 μm from the stimulating electrode. They were identified by small soma (5–10 μm), low resting membrane potential (~−80 mV), low input resistance (50 MΩ) were considered NG2 or complex cells and were excluded from the study. Membrane currents were analyzed using custom-written MATLAB scrips (MathWorks R2016a). Synaptic currents of 1, 4, and 5 stimuli were baseline subtracted and then averaged. I K (K + current) was measured 200 ms after the stimulus. At this time point I K was not contaminated by the current mediated by field potential and transporter current. From this point the decay was fitted with mono-exponential function and τ decay calculated. To obtain I K in response to 5th stimulus, the response to 4 stimuli was subtracted from the response to 4 stimuli.

Field potential recording

Field excitatory postsynaptic potentials (fEPSPs) were recorded from CA1 str. radiatum using glass microelectrodes (0.2–1.0 MΩ) filled with ACSF. Synaptic responses were evoked with extracellular stimulation of the Schaffer collaterals using a bipolar twisted stimulating electrode made of insulated nichrome wire placed in the str. radiatum at the CA1–CA2 border. The stimulation was performed with rectangular paired pulses (duration, 0.1 ms; interstimulus interval, 50 ms) every 20 s via an A365 stimulus isolator (WPI). Responses were amplified using a microelectrode AC amplifier model 1800 (A-M Systems) and were digitized and recorded on a personal computer using ADC/DAC NI USB-6211 (National Instruments) and WinWCP v5.2.2 software by John Dempster (University of Strathclyde). Electrophysiological data were analyzed with the Clampfit 10.2 program (Axon Instruments). The dependence of field response amplitude on stimulation strength was determined by increasing the current intensity from 25 to 300 μA. For each fEPSP, the amplitude and the slope of the rising phase at a level of 20–80% of the peak amplitude were measured. The presynaptic fiber volley (PrV) was quantified by the peak amplitude. The maximum rise slope of the input-output (I/O) relationships (fEPSP amplitude vs. PrV amplitude) was calculated for every slice by fitting with a sigmoidal Gompertz function (Equation 14) using OriginPro 8 (OriginLab Corporation). (14) y = a e - e ( - k ( x - x c ) ) , where a is an asymptote of the maximum fEPSP amplitude; e is Euler's Number ( e = 2.71828 …); k and x c are positive numbers describing the shape of the curve; x c is the PrV amplitude at which the maximum slope of the curve is observed; ak / e is a maximum slope of the curve. The stimulus intensity used in the experiment was chosen so that the amplitude of fEPSP would be 40–50% of the amplitude where the population spike appeared for the first time. The strength of stimulation was unvaried during the experiments, usually being 50–150 μA. The paired-pulse ratio (PPR) was measured as a ratio of the second and first fEPSP amplitude. The LTP induction was started only if a stable amplitude of the baseline fEPSP had been recorded for 20 min. Three trains of high-frequency stimulation (HFS, 100 pulses at 100 Hz, with an inter-train interval of 20 s protocol) was applied to induce LTP. The fEPSPs were recorded after induction protocol during 60 min. The baseline fEPSP and the potentiated fEPSP (recorded 47–60 min after HFS) were averaged separately to measure LTP in a slice. Plasticity value was calculated as a ratio of the slope of the rising phase in the averaged potentiated and baseline fEPSP. Ca 2+ imaging Ca 2+ activity was recorded with a confocal microscope, Zeiss LSM DuoScan 510, in CA1 str.radiatum of acute hippocampal slices pre-incubated with Ca 2+ dye, Oregon Green 488 BAPTA-1 AM (Invitrogen) and an astrocyte specific marker, sulforhodamine 101 (100 nM, Invitrogen). After the preparation, the slices were transferred to a 3 ml incubation chamber with constantly gassed ACSF containing both dyes. Oregon Green 488 BAPTA-1 AM was initially dissolved to 0.795 mM in 0.8% Pluronic F-127 in DMSO. Then 3 μl of the dye was added to the chamber. After incubation for 40–45 min at 37°C in the dark, the slices were transferred to the recording/imaging chamber for time-lapse imaging (one frame/s). Oregon Green 488 BAPTA1 was excited with a 488 nm argon laser and imaged with an emission band-pass filter 500–530 nm; sulforhodamine 101 was excited with a 543 nm HeNe laser and imaged with an emission band-pass filter 650–710 nm. The imaging was performed for 10 min at 34°C in normal ASCF, then 30 dark noise images were recorded. The raw imaging data were exported to MATLAB. The median of the dark noise was calculated for each pixel and subtracted from the corresponding pixel intensity value of the fluorescence images. Then denoising was done with the BM3D algorithm (Danielyan et al., 2014 ). The movement artifacts were corrected with the single-step DFT algorithm (Guizar-Sicairos et al., 2008 ). The whole Ca 2+ events (x-y-time 3D Ca 2+ signals) were detected with the adapted algorithm which we described previously (Wu et al., 2014 ). Briefly, each pixel of the image series was analyzed independently. Firstly, we roughly estimated a baseline fluorescence F 0temp applying 60-s 3rd order Savitzky-Golay polynomial filter which smoothed all Ca 2+ signals on the fluorescent signal F. Then, we estimated a temporary (ΔF/F) temp = (F − F 0temp ) / F 0temp to find Ca 2+ transients exceeding a statistical threshold. Then these transients were excluded from the baseline which was further smoothed with 100-s filter. This filter interpolated the intervals left by the excluded transients, and, thus, we obtained the uninterrupted final baseline F 0 which was used to obtain ΔF/F = (F − F 0 ) / F 0 . Ca 2+ transients exceeding a statistical threshold were detected and binarized. The active neighboring pixels were grouped into x-y 2D Ca 2+ events, which were reconstructed into x-y-time 3D Ca 2+ events. For each Ca 2+ event the maximal projection (S max ), the integral and the duration were calculated. To avoid noise detection the events excluded from further analysis if the integral was less than 4 μm 2 s, or the S max was less than 10 μm 2 , or the duration less than 2 s. The probability density of the events sizes and the durations appeared linear in log-log scale, suggesting that the obtained distributions can be described by a power law. Therefore, the power law fit was applied, and the corresponding exponent was calculated for each slice.

Statistical analysis

All data are presented as mean ± standard error of mean (SEM). Statistical significance was assessed using non-parametric Mann-Whitney test, parametric Student's t -test and repeated measures two-way ANOVA as stated in the text. P < 0.05 was considered statistically significant.

Supplementary material The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnmol.2018.00215/full#supplementary-material Click here for additional data file. Click here for additional data file.