Animal surgery and fluorescence labeling
The study and all experiments conducted were fully approved by the Institutional Animal Care Unit Committee (IACUC) in Administration Office of Laboratory Animals Beijing China (ID# B10831). FVB mice in postnatal days (PND) 20~35 were anesthetized by the intraperitoneal injections of urethane (1.5g/kg). Anesthetic depth was judged based on lack of reflexes in pinch withdrawal and blink eyelid, and was maintained by giving the supplemental dosage of urethane (0.25g/kg) throughout the experiments. Body temperature was maintained by using a computer-controlled heating blanket at 37°C. The barrel cortex was located based on the distributional map of superficial vessels and confirmed by histological reconstructions (Figure 1A). A craniotomy (1~2mm in diameter) was made on the skull above the center of the barrel cortex (Figure 1B), which was located at 1mm posterior to the bregma and 3.5mm lateral to the midline. It is noteworthy that the dura was intact throughout all experiments, and the care was taken to avoid any damage to superficial vessels and cortices.
In our studies, Oregon Green BAPTA-1-AM (OBG-1, Invitrogen, CA USA), a Ca2+ dye, was used to measure the activities from neurons and astrocytes. OGB-1 was dissolved in DMSO and 20% Pluronic F-127 (2g Pluronic F-127 in 10ml DMSO, Invitrogen, USA) to have its stock solution at 10mM. The stock solution was diluted in ACSF to yield its final concentration at 1mM. This OBG-1 solution was injected into layer I~II of barrel cortex by the pressure (1bar, 5min) through a patch pipette (200μm below the pia) to label the population of nerve cells, termed as multicell bolus loading [55–57]. In the meantime, 100μM sulforhodanmine-101 (SR101, Invitrogen, USA) was co-injected to label the astrocytes specifically . The volumes of these dyes were controlled at ~0.5μl. After micro-injections, the craniotomy well was filled by low-melted agarose (1~2%) in saline and then was sealed with glass coverslip. The exposed skull was adhered to a custom-made recording chamber with dental acrylic cement, which was surperfused with saline (mM): 125 NaCl, 2.5 KCl, 26 NaHCO3, 1.25 NaH2PO4, 2 CaCl2, 1 MgCl2, 20 glucose (pH 7.4). The saline was warmed up to 37°C and bubbled with 95%O2/5% CO2.
Whisker stimuli and barrel cells’ responses
All major whiskers in the contralateral sides of the imaged barrel cortices were deflected in a caudal-to-rostral direction by air-puffing during the experiments, which is more similar to the natural movement of whiskers. Whisker stimuli were done by giving the sequential brief pulses of air-puffing (50psi, 50ms) through a tiny steel tube that was mounted on a micromanipulator and controlled with costume-made LabVIEW program. The stimulus patterns were the paired burst-pulses, in which each burst had a given frequency. The frequency patterns in these paired bursts (10 seconds for each) were 8 and 12Hz (8-to-12Hz), closely to natural frequency in exploratory whisking [25, 49, 58]. Burst-pulse intervals were 10 seconds. To avoid the stimulations to the skin and furs, a tip of the stimulator was positioned in a way that it did not blow on the snout.
The amplitudes of Ca2+ signals are correlated positively with spike frequency, and Ca2+ levels in a neuron indicate its functional activities, so that spiking activity at the neurons can be estimated from their somatic Ca2+ signals in vivo[19, 31, 32, 59]. Different from the spikes as a functional index of neuronal activity, the activities of astrocytes are associated with the changes of Ca2+ signals [30, 60]. In addition, the synchrony of Ca2+ signals provide a measurement for the timing of cellular activities [4, 7, 54, 61]. Therefore, we studied the activity patterns of barrel neurons and astrocytes in response to whisker stimuli by using in vivo two-photon cellular imaging in order to reveal the processes of neuronal encoding in brain networks, in which the peak amplitudes and temporal synchrony of Ca2+ signals were analyzed.
Ca2+ fluorescence signals for cellular responses to whisker stimuli were acquired by using Fluoviewer-10 software (Olympus Inc. Japan) and analyzed in the regions of interest (ROI) from cell bodies by using NIH ImageJ and MATLAB (MathWorks). To reduce the photon and PMT noise, a median filter (radius, 1 pixel) was applied to all images. Ca2+ signals in cellular responses were normalized and presented as relative fluorescence changes (ΔF/F)[29, 62, 63]. Baseline fluorescence (F) was an averaged value in the ROI before stimuli, and ΔF values were differences between Ca2+ signals from the evoked responses in the ROI and baseline fluorescence. It is noteworthy that all of the fluorescence signals were subtracted from the noise signals of unstained blood vessels, as well as Ca2+ signals in the astrocytes were normalized to SR101 signal to eliminate motion artefacts. The normalized Ca2+ signals were smoothed by low-pass Butterworth filter to remove low-amplitude fluctuations and to minimize the distortions from fast Ca2+ transients [31, 32]. Effective signals from each of active cells were judged according to a criterion that their relative fluorescence changes were greater than 2.5 SD of baseline values and lasted for 500ms. In our experiments, whisker stimuli induced the robust changes of Ca2+ signals in barrel cells, and the criteria above was found effective for sorting evoked signals from noise.
In our study, the paired burst-stimuli were given to induce two sequential responses. As the fluctuation of fluorescence signals influenced a precise measurement of response amplitudes, we thought the magnitude differences of two responses if their differences were above 2 SD of baseline values; however, we defined no difference in the magnitude of two responses if their net changes were less than 2 SD. If cellular response one (R1) was larger than response two (R2) above 2 SD of baseline, the pattern was defined as the decrement (R1>R2). On the other hand, R2>R1 above 2 SD was the increment. No differences for R1 and R2 were called as parallel. This classification is similar to synaptic transmission patterns . It is noteworthy that R2 values were the absolute changes of responses induced by stimulus two. That is, if the calcium signals in response to stimulus one were not back to their baseline levels, R2 was measured as a difference between the magnitude of response two and the residual level of response one.
The pairswise cross-correlations of normalized and smoothed Ca2+
signals (ΔF/F) in the neurons and/or astrocytes between each of cell pairs were analyzed as Pearson correlation [7
]. Although the cross-correlations between neurons from raw fluorescence traces were higher than the deconvolved traces over 2 fold [61
], we computed raw traces without temporal deconvolution in the neurons consistently with those in astrocytes which had no spikes firing [65
]. Consider two signals x (t) and y (t) of a real variable t; the cross-correlation r
at delay d
is defined as:
mx and my are the means of the corresponding series. Correlation coefficients normalized to the autocorrelation at zero lag were calculated. Based on these calculations, the correlation matrices were plotted using MATLAB 7.0.
In the study of functional connectivity [38
] among network cells, we converted correlation coefficient matrix (r
) into binary adjacency matrix (A
) by setting a threshold (thresh
]. It was the averaged correlation coefficients plus two-time standard deviations corresponding to spontaneous cellular activities without whisker stimuli. If r
during whisker stimuli is lareger than thresh
, i.e., A
equals to one, the functional connection is present between cell i
and cell j
. On the other hand, if r
during whisker stimuli is less than thresh
, i.e., A
is equal to zero, the functional connection is not present between cell i
and cell j
. The formula is
It is noteworthy that the definition of functional connection, whose threshold is set at mean+2SD of correlation coefficients during the spontaneous activities of network cells, is based on an assumption that their activities are random in nature (no coordination). In other words, there are no interactions, or functional connections, among these network cells without input signals .
Based on these criteria of binary adjacency matrices and the spatial positions of network cells, we plotted the graphs which consisted of a set of nodes (the cells activated by stimulus bursts) and their functional connections (lines) under the conditions of response one to 8Hz stimuli and response two to 12Hz. In these graphs of neural networks, two parameters for each cell were merited to indicate how each cell is connected with others. The cell that connects with one at least is called as a function-connected cell in neural network. The percentages of function-connceted cells present how many cells are functionally connected with others. If a cell connects with others, the percentages of functional connections for its actually connected cells in the total cells are calculated to present the connection strength for each of network cells. The folmula are given below.
For a neural network consisting of activated neurons (N) in complete graphs, the number of connctions for each neuron with others is N-1. The number of function-connected neurons is n, and the averaged number of connections for each neuron with others is k. Thus, P
=n/N stands for the percentages of function-connected neurons. P
=k/(N-1) presents the percentages of functional connections of each neuron. The astrocytes are connected tightly and widely via gap junctions, the number of connected neuron-astrocyte pairwise shows no variation during different stimuli in our studies. In a network comprising of N neurons and M astrocytes, we calculate the percentages of neuron-astrocytic functional connections (l) in total potential links, which is P
All data are presented as mean±SEM. Student’st tests (two-tailed, paired, or unpaired assuming unequal variances) were done in R software package, version 2.10.1 (http://www.r-project.org/) to evaluate statistical significance. A p≤0.05 is defined as statistical significance.