Experimental Design
We performed in vivo whole–cell patch–clamp recordings of neocortical neurons of the primary somatosensory cortex to examine tactile stimulus–evoked sensory processing in anesthetized mice and to probe the causal role of endogenous noise sources and parameters for atypical sensory information processing in autism. Throughout the text, the terms autism and autistic people/individuals are used, in line with recent evidence suggesting that these terms are preferred in the autistic community and are less stigmatizing 87.
Ethical statement
All experimental procedures were performed in accordance with the EU directive 2010/63/EU and French law following procedures approved by the Bordeaux Ethics Committee (CE2A50) and Ministry for Higher Education and Research. Mice were maintained under controlled conditions (temperature 22-24°C, humidity 40-60%, 12h/12h light/dark cycle, light on at 07:00) in a conventional animal facility with ad libitum access to food and water. All experiments were performed during the light cycle.
Mice
Second generation Fmr1 knockout (Fmr1−/y) 46 and wild-type littermate mice at P26-42 were used in our study. Mice were maintained in a mixed 129/Sv/C57Bl/6J/FVB background (backcrossed 6 generations into C57Bl/6J) as described in 46. Male wildtype and Fmr1−/y littermates were generated by crossing Fmr1+/− females with Fmr1+/y male mice from the same production, and the resulting progeny used for our experiments was either Fmr1+/y (wild-type) or Fmr1−/y (KO). Mice were maintained in collective cages following weaning (3–5 litter males per cage). Cages were balanced for genotype and supplemented with minimal enrichment (cotton nestlets). Number of mice are given in the figure captions. The genotype of experimental animals was re-confirmed post-hoc by tail-PCR.
Surgery
Mice (P26–42) were anaesthetized with a mixture of ketamine (100 mg.kg− 1) and xylazine (10 mg.kg− 1) injected intraperitoneally and supplemented as necessary throughout the procedure. Proper depth of anesthesia was monitored by testing the absence of a foot-pinch reflex and whisker movement. Mice were head-fixed using non-puncture ear-bars and a nose-clamp (SR-6M, Narishige). Body temperature was maintained at 37°C. Prior to making an incision on the skin to expose the skull, 0.1 ml of a 1:4 Lidocaine to saline solution was administered subcutaneously and waited for 2 to 5 minutes to induce local analgesia. Following a careful removal of the scalp, and the remaining tissue on the skull, a small craniotomy was made above the S1 hindpaw region (1 mm posterior and 1.5 mm lateral from Bregma, confirmed with intrinsic imaging coupled with hind paw stimulation) using a dental drill (World Precision Instruments).
In vivo whole-cell patch-clamp recordings
Blind, in vivo whole-cell recordings were performed from layer 2/3 pyramidal neurons of the hindpaw region of S1 in anaesthetized mice, as described previously 44, 46. Neurons were identified by their electrophysiological properties, and in some cases by their post-hoc morphology. Depth of neurons was on average 263 µm from pia, ranging from 175 µm to 374 µm. There was no genotype difference in the depth of recording (WT = 261.69 ± 34.91 µm; Fmr1−/y = 259.72 ± 49.12 µm; p > 0.05, unpaired student t-test). Data were acquired at 20 kHz sampling rate and low-pass filtered at 3 kHz using Dagan BVC-700A amplifier (Dagan, Minneapolis, USA), Digidata 1320A and Clampex 10.4 software (Axon Instruments). Recording pipettes with an open-tip resistance of 4–6 MΩ were pulled from borosilicate glass using a PC-10 puller (Narishige) and filled with intracellular solution containing (in mM): 130 K-methanesulphonate, 10 Hepes, 7 KCl, 0.05 EGTA, 2 Na2ATP, 2 MgATP, 0.5 Na2GTP (all products from Sigma Aldrich); pH 7.28 (adjusted with KOH); osmolarity was 280–295 osm. In a subset of experiments, biocytin (1.5–2.5 mg/ml) was added to the recording solution for post-hoc neuronal identification and anatomical comparison. The intracellular solution was filtered using a 0.22-µm pore-size centrifuge filter (Costar Spin-X). Cells were excluded from the analysis if the pipette access resistance exceeded 50 MΩ or the neuron was depolarized more than − 50 mV.
Neocortical application of the specific BK Ca channel agonist, BMS191011
To pharmacologically target BKCa channels, we used the specific channel agonist, BMS191011 (3-[(5-Chloro-2-hydroxyphenyl)methyl]-5-[4-(trifluoromethyl)phenyl]-1,3,4-oxadiazol-2(3H)-one, 100 µM; Tocris). A stock solution with a concentration of 50 mM BMS191011 was prepared in DMSO and stored at − 20°C. For direct neocortical application the drug was diluted to a final concentration of 100 µM in PBS (final concentration of DMSO in PBS: 0.2%). Cortical application of BMS191011 (~ 1 ml) was performed at least 30 minutes prior to the whole-cell patch-clamp experiments. Drug allocation was semi-randomized and balanced for cage composition.
Data analysis
Neuronal morphology
Following biocytin (1.5–2.5 mg/ml Biocytin, Sigma) filling of the neurons during recording, mice were perfused for post-hoc staining 46. Briefly, mice received a lethal dose of pentobarbital (300 mg/kg, i.p.) delivered in the presence of lurocaine (30 mg/kg; i.p.). Following respiratory arrest (and after verifying the absence of reflexes to toe/tail pinch and eye-blink) tissue was fixed by trans-cardial perfusion with 1 X PBS (pH 7.4), followed by 4% paraformaldehyde in 1 X PBS (pH 7.4). Brains were then post-fixed for 2h in 4% PFA (or stored in 1 X PBS prior to slicing). Subsequently, 80-µm-thick slices were cut using a vibratome (Leica), and the slices were stored in 1 X PBS prior to staining. Biocytin was revealed using streptavidin-Alexa Fluor 555 (Invitrogen). Slices were mounted in Mowiol and neuronal morphology was reconstructed using a Neurolucida system (MBF Biosciences) equipped with a 100x oil immersion objective lens.
Spontaneous AP firing
Neurons that spontaneously fired at least one action potential (AP) during a 120-second-time window were considered spontaneously active, otherwise silent. Spontaneous AP rate was calculated as the number of APs elicited during this 120-second-time window. The analysis included data from both active and silent neurons. We acknowledge the limitation of the term ‘silent’, since these neurons would likely become ‘active’ if we would analyze spontaneous AP firing over a longer time window. As a result, many WT neurons had spontaneous AP firing values of zero and we could therefore not include this feature in our correlation matrix and the accompanying node plot for WT neurons.
Up– and down–states
For up– and down–states, both ‘active and silent cells’ were included in this analysis. Custom-made python scripts were used to detect all up– and down–states during a 180-second recording period, and to quantify their duration, frequency, and membrane potential at the respective states. A pre-processing step was performed when necessary to correct for linear drifts in membrane potential. Our algorithm annotated each point of the signal as either an up– or down–state with no intermediate state. A gliding threshold was calculated every second as the median of all points during both a four-seconds-period before and after that point. For each point of the signal the median of the surrounding points (during 50 ms before and after) was computed and compared to the corresponding gliding threshold. If this median was greater than this threshold, the point was considered part of an up-state and vice versa. Our analysis revealed “micro”-up-states lasting between 100 and 150 ms. Events lasting less than 100 ms were considered too short and removed from the analysis.
Power of membrane potential oscillations
The two periodograms (WT and KO) were obtained utilizing the Welch function of the Python open-source library, SciPy. Parameters such as a 4-second Hann sliding window, a 50% overlap and the mean periodogram as the averaging method were used to calculate the Power spectral density (PSD). PSD values for each delta (0.5–4 Hz), theta (4–7 Hz), alpha (8–12 Hz), beta (13–30 Hz), and gamma (30–100 Hz) bands were computed by calculating the area under the curve of the periodogram to the respective frequency band by applying the composite Simpson rule.
Membrane potential fluctuations/wavelet analysis
Spontaneous resting signals were transformed using a complex Morlet wavelet with 4Hz as mother wavelet frequency 38. The widths used to scale the wavelets were computed using the following equation:
$$\frac{\left(w*Fs\right)}{\left(2*yScale\right)}$$
where w is the mother wavelet frequency, Fs the sampling rate (20kHz), and yScale is scale of the frequencies we are interested in. Absolute values were plotted in color-code with the scale ranging from 0 to 3. This maximum is a tradeoff between being able to detect the differences between genotypes but not saturating the signal.
Intrinsic properties
To study the intrinsic properties of the recorded neurons, we measured the membrane potential responses to 500-ms long step current injections ranging from − 450 pA to 550 pA (step size: 50 pA). To determine the action potential (AP) threshold, we measured the membrane potential where the slope of its rising phase exceeded 10 mV/ms. AP half-width was determined by measuring the duration of the first AP at half maximal amplitude (half-distance from threshold to peak) following the rheobase injection. Maximum AP frequency was calculated from the voltage trace with the largest number of APs. Calculation of AP accommodation was performed using a voltage trace encompassing 5 APs. Briefly, the spike interval (SI, in ms) between the 1st and 2nd AP (1st spike interval, SI), and the 4th and 5th AP (4th SI) were calculated, and AP accommodation was then calculated as 4th SI/1st SI. For analysis of the AP after-depolarization (ADP), trains of three APs at various frequencies were generated by brief somatic current injections (1 nA, 1.08 ms). Only AP trains occurring during down–states were selected for the analysis. Three to six trials were averaged, and the ADP amplitude (from baseline) was measured 5 ms after the peak of the last AP. AP half-width ratio was measured as the ratio of the third and first AP. To measure input resistance, we injected 500-ms-long hyperpolarizing (-100 pA) current pulses and measured the steady-state membrane potential deflection at 300 ms relative to baseline.
Sensory stimulus evoked responses - Hindpaw (HP) and forepaw (FP) stimulation
Sensory responses to tactile paw stimulus were evoked by applying squared current pulses (2 ms duration, 100 V, 30 mA) to the paws via conductive adhesive strips (~ 1 cm2) placed on top of, and underneath the HP or FP, as described previously 44, 46. These conductive strips covered the entire paw. Following the establishment of a somatic whole-cell recording configuration, the contralateral HP or FP was stimulated 40 times at an interval of < 0.3 Hz.
EPSPs and signal-to-noise (SNR) ratios
Parameters of HP stimulus evoked excitatory postsynaptic potentials (EPSPs) from 40 successive trials were calculated for EPSP-only neurons (neurons responding to HP stimulus exclusively in a sub-threshold manner, i.e. an EPSP or a failure) using Clampfit software (version 11.1, Molecular devices, LLC). Briefly, the maximum EPSP amplitude was determined for each trial during a 200-ms time window following the HP stimulation. Trials with a response amplitude of less than two times the standard deviation of the baseline were considered as failures. EPSP duration was calculated by measuring the width of the response at half-maximal amplitude. Response slope was estimated as the rise slope between the 20th and 80th percentile of the EPSP amplitude relative to the baseline. Baseline membrane potential (Vm) fluctuation was calculated as the standard deviation (SD) of the Vm fluctuation during a 200-ms-time window just before the stimulus onset. Signal-to-noise ratio (SNR) was calculated similar as described in Dinstein et al. 6, by dividing the EPSP amplitude of each trial by the EPSP amplitude variance across trials for each cell. EPSP latencies were measured for the averaged response for each cell. EPSP onset latency was measured as the delay following HP stimulation where the Gaussian fit of the response’s rising phase crosses the Vm baseline (averaged Vm potential during 200 ms before stimulus onset). Peak latency was calculated as the delay of the EPSP maximum amplitude with respect to the onset of the response.
Evoked APs
Neurons were included in the evoked AP analysis if HP stimuli elicited at least one AP during the 40 trials. Accordingly, these neurons were classified as AP-EPSP neurons. The quantification of evoked AP responses was adapted from 44, 46. Briefly, spontaneous AP firing (pre-stimulus APs) was calculated as the number of APs elicited within a 200-ms-time window prior to HP stimulus. The evoked AP firing was quantified as the difference between the number of APs fired within a 200-ms-long time window following the HP stimulation (post-stimulus APs) and the pre-stimulus AP number (evoked APs = post-stimulus APs – pre-stimulus APs). Coefficient of variation (c.v.) was calculated by dividing the standard deviation of AP firing by the mean evoked AP firing for individual trials. Mean AP number per successful trial was determined by dividing the number of APs evoked during a 40-trial session by the number of trials eliciting at least one AP. To determine AP dispersion, we measured the onset of the first AP in each trial within a 70-ms-time-window following HP stimulus.
Correlation matrix and node plot
The correlation graphs were created with python custom-made script using NetworkX and Netgraph libraries. Seven categories of parameters (in WT neurons six, since spontaneous AP firing could not be included, see above) were defined: Trial-by-trial variability parameters, up–/down–state parameters, spontaneous AP firing, AP parameters, membrane potential (Vm) variance parameters (PSD + SD baseline Vm variance), SNR parameters, and EPSP parameters. Parameters were ordered depending on these categories, and each category is displayed in a different color in the graph. The nodes were arranged on a circular layout and the size of the nodes is proportional to their degrees – in this case the number of statistically significant correlations. Only correlations with a p-value < 0.05 using the Pearson test are shown. Edge size and color depend on the correlation coefficient, larger coefficients (absolute value) have edges with greater width and darker color (blue for negative and red for positive correlations).
Trial correlation parameters
The time window chosen to compute Vm baseline parameters (baseline Vm, baseline Vm SD, PSD) on a trial-by-trial basis was a range of 200 ms before the onset of the HP stimulus. To estimate the influence of baseline Vm fluctuation and PSD on the strength, duration, and reliability of HP stimulus evoked EPSPs, these parameters were normalized by the baseline Vm. For correlating these parameters for each trial, we used Pearson correlation tests.
Data normalization in Fig. 5
To assess whether BMS191011 application corrected the altered physiological features of Fmr1−/y neurons, we statistically compared WT and Fmr1−/y-BMS191011 values after they were both normalized by Fmr1−/y values (Fig. 5, panels C, F, G, H, I, M, O, P, Q, S, T). For spontaneous AP firing (panel D) it was not possible to normalize this data because of the high proportion of zero values for WT and Fmr1−/y-BMS191011 neurons. Values were either normalized by the mean of the Fmr1−/y values if these values were normally distributed, or by the median in case of non-normally distributed data. This is stated for each panel in the legend of Fig. 5.
Overall experimental design and analysis
Sample sizes were determined based on our published work 44, 46. In addition, we performed posthoc statistical tests of power. Mice of both genotypes were littermates and randomly assigned. Recordings and analysis were performed blind to the genotype.
Statistical analysis
Values were first tested for outliers (Grubb’s outlier test with alpha = 0.05). These outliers were removed from the statistical analysis and the resulting plots. Values were also tested for normality using the Shapiro-Wilk normality test. If the values were normally distributed an unpaired t-test was used to compare two groups. For non-normally distributed parameters we used Mann-Whitney’s U-test. A mixed ANOVA model was used for repeated measurements. As we combined silent neurons (no firing in 2 min time window) and active neurons for the calculation of spontaneous properties, we have performed a two-sided non-parametric permutation test to calculate the p-Value. Boxplots indicate the median value (middle line), the mean (green line), as well as the 25th and 75th percentiles (box). The lower whisker will extend to the first datum greater than Q1–1.5*IQR where IQR is the interquartile range (Q3-Q1). Similarly, the upper whisker will extend to the last datum less than Q3 + 1.5*IQR (matplotlib boxplot function default parameters). Correlation matrices were made with R Pearson tests, resulting in a coefficient of correlation and an associated p-value. Trial-by-trial variability was calculated as standard deviation of the parameter values across all trials for each cell. The F-test of equality of variances or Bartlett test were used to explore the difference in variance between genotypes at the cell-population level (trial-wise average) for normally distributed data. For non-normally distributed data the Levene test was used with the mean as center parameter. Density plots (Rugg plots) were made with a gaussian kernal density estimation using the function scipy.kde.gaussian_kde from the python library scipy. P values < 0.05 were considered significant (* P < 0.05, ** P < 0.01, *** P < 0.001).