Calcium activation of cortical neurons by continuous electrical stimulation: Frequency dependence, temporal fidelity, and activation density.

Surgery and electrode implantation

Transgenic mice C57BL/6J-Tg(Thy1 GCaMP6s)GP4.3Dkim/J (n=5, male, 22–28 g; IMSR Cat# JAX:024275, RRID:IMSR_JAX:024275) were used in this experiment. All surgical interventions were performed on adult mice (>4 weeks of age) that were housed in social housing with 12h light / 12h dark cycles and free access to food and water until acute experimentation. All animals were induced with an anesthetic mixture consisting of 75 mg/kg ketamine and 7 mg/kg xylazine administered intraperitoneally and updated with 40 mg/kg as needed. Rectangular craniotomies (~4 mm per side) were made over each somatosensory cortex and electrodes were implanted at a 30° angle as previously described ( Eles et al., 2017 ; T. D. Kozai, Vazquez, Weaver, Kim, & Cui, 2012 ; Takashi D. Y. Kozai, Eles, Vazquez, & Cui, 2016 ; Takashi D Y Kozai, Jaquins-gerstl, Vazquez, Michael, & Cui, 2016 ; Michelson et al., 2018 ; S. M. Wellman & Kozai, 2018 ). Electrical stimulation was conducted through a single-shank 16-channel Michigan style functional silicon probes with 703 μm 2 electrode sites (NeuroNexus Technologies, Ann Arbor, MI). In order to prevent mechanical strain-related Ca activity caused by microelectrode insertion ( Eles, Kozai, Vazquez, & Cui, 2018 ) from influencing stimulation evoked GCaMP activity, the tissue was allowed to rest 20 minutes after insertion prior to stimulation experiments. All experimental protocols were approved by the University of Pittsburgh, Division of Laboratory Animal Resources and Institutional Animal Care and Use Committee in accordance with the standards for humane animal care as set by the Animal Welfare Act and the National Institutes of Health Guide for the Care and Use of Laboratory Animals.

Two-photon Imaging

A two-photon laser scanning microscope (Bruker, Madison, WI) and an OPO laser (Insight DS+; Spectra-Physics, Menlo Park, CA) tuned to a wavelength of 920 nm at 15 mW was used for the entirety of this study. A16X 0.8 numerical aperture water immersion objective lens (Nikon Instruments, Melville, NY) was selected for its 3 mm working distance. Imaging was carried out over a 407 × 407 μm ROI, collected with a 512 × 512 pixel matrix. Before imaging sessions, animals were injected IP with 0.1 ml of 1mg/ml SR101 for visualization of blood vessels and electrode contacts. Imaging plane was positioned 10 μm above the electrode site, 120-150 μm below the surface of the brain. In order to image neural dynamic during electrical stimulation, time series images were recorded at 30 Hz and 2x optical zoom. The imaging plane for time series was set right over the electrode contact, so that the stimulating contact is visible during imaging session.

Electrical stimulation

Concurrent stimulation and imaging typically commenced 20 minutes after electrode implantation to allow time for displaced tissue to settle. An A-M Systems 2100, single channel Isolated Pulse Stimulator (A-M Systems, Sequim, WA) or TDT IZ2 stimulator on an RZ5D system (Tucker-Davis Technologies, Alachua, FL) were used to apply current controlled cathodic leading biphasic symmetric (50 μs each phase) electrical microsimulation at 50μA and frequencies of 10, 30, 50, 75, 90, 130 and 250 Hz. However, in one animal, the impedance of the stimulated site was an order of magnitude lower than that listed by the manufacturer, suggesting minor insulation failure at the electrode/insulation interface, as noted in previously ( Prasad et al., 2014 ). This has the net effect of slightly increasing the electrode interface area, requiring greater current to achieve the same effective charge density seen in other animals. In this animal, 100 μA stimulation current was necessary to drive GCaMP activity. Amplitude of stimulation was selected based on minimum current required to evoke activation during two photon live scanning. Charge/phase, charge density/phase and k values were calculated based on the Shannon’s equation ( Shannon, 1992 ). The k value used during stimulation was between −0.12 and 1.3, well within safety limits to avoid tissue damage. After initial collection of resting state activity for 2min, concurrent imaging and stimulation trials were acquired for 2 min at each frequency, which included 30 s of pre-stimulation baseline, 30s of stimulation and 60s of post-stimulation baseline. An additional 3 minutes was required between each trial to store the data after data collection. The two-photon microscope was synchronized to the stimulator via transistor-transistor logic (TTL) and 30 s stimulation was carried out after a 30 s delay from the start of the time series. Stimulation was carried out in increasing frequency and were not repeated in the same animal. However, in two animals, 10 Hz stimulation was repeated at the end to determine if earlier microstimulation trials change the neuronal response properties for subsequent trials. The second 10 Hz stimulation data showed similar activation patterns to the first 10 Hz stimulation data.

Data analysis Identification of activated neurons

The acquired time series images were converted to a single tiff image file using ImageJ (NIH). The tiff image was then analyzed using a custom MATLAB script (MATLAB, RRID:SCR_001622). Non-overlapping elliptical regions of interest (ROIs) were manually drawn around neurons to fully encompass their somas, which were identified by their morphology as bright rings of fluorescence ( Fig 1b,d ). ROIs were obtained using the CROIEditor MATLAB script, written by Jonas Reber ( https://www.mathworks.com/matlabcentral/fileexchange/31388-multi-roi-mask-editor-class ). Fluorescence intensity time-courses for each neuron were generated by averaging all pixels within a cell’s ROI. Baseline fluorescence intensity (F 0 ) was calculated as the average signal during the 30-second pre-stimulation interval, and each time-course was then converted to changes in fluorescence (ΔF/F 0 ). To identify activated neurons, each neuron’s ΔF/F 0 signal during stimulation was compared with their pre-stimulation ΔF/F 0 signal. A fluorescence threshold of the mean pre-stimulation signal + three standard deviations (SD) was then established, such that neurons which showed ΔF/F 0 changes that exceeded the threshold during stimulation were considered to be activated by the electrical stimulation. While it is possible that the threshold selection might not have identified all active cells, GCaMP6s is very sensitive to neuronal activity even for single action potentials (>30% ΔF/F reported) ( Chen et al., 2013 ). Our criteria was robust for detection of low levels of activity even in the presence of small, unwanted signal changes due to motion or photomultiplier tube (PMT) dark current. The spatial locations and the activation properties of these neurons were then quantified for further analysis. Neuropil Activation Neuropil intensity was quantified by averaging the intensity level of a 10 × 25 μm region that did not contain neuronal somas. For each frequency, the neuropil was measured in the following distances from the edge of the electrode; 10-20, 30-40, 50-60, 100-110, 150-160, and 245-255 μm. Similarly, representative neuronal somas were selected at 15 and 115 μm from the electrode. F/F 0 of GCaMP intensities were computed where F 0 was the average GCaMP intensity of the 30 s pre-stimulus period. Temporal activation pattern Somas were classified as onset-responsive if they were activated during the first 0-2 seconds of the 30 second pulse train and ceased activation before the final 2 seconds of the pulse train (28-30 s). Somas were classified as steady-state-responsive neurons if they were active during 28-30 s post-stimulation onset. Duration of activity was calculated for each neuron by considering all time-points during stimulation which exceeded the intensity threshold. To examine the temporal characteristics of activated neurons in local vs distant or dense vs sparse groups, a weighted response time was estimated for each neuron. The ‘Activation Time’ was calculated as the time for which the cumulative ΔF/F sum reached 50% of the total ΔF/F sum over the 30s stimulation period. This expression describes the time it takes to cover half of the response energy. For example, a soma that reaches peak intensity at 0 seconds into the pulse train and maintains the exact same ΔF/F for all 30 s would have an ‘Activation Time’ of 15 s. Similarly, a soma that maintains a peak ΔF/F from 0 s to 5 s, but then falls to 20% of peak ΔF/F from 5 s to 30 s would have an ‘Activation Time’ of 5 s. Spatial activation pattern The spatial relationship between stimulation frequency and activity was determined by measuring the distance from the center of each active neuron to the nearest edge of the stimulating electrode site. The location of the center of the stimulating electrode site was visually identified from the same image series that was used to identify activated neurons and was confirmed with a high-resolution image taken immediately before the trial. The distances from the nearest edge of the stimulating electrode site to each activated neuron during the stimulation period were estimated for all animals and stimulation frequencies. The mean distance + 1SD was then calculated for each stimulation frequency and animal. Because the distance of onset and steady state activated neurons were considered for these calculations, distance thresholds remained fairly consistent across stimulation frequencies within animals. Therefore, the stimulation frequency with the minimum mean + 1SD for each animal was chosen to represent a distance threshold (139.7±18.8 μm for n=5), whereby cells in this region were considered local neurons, while those beyond the distance threshold were considered distant neurons. Activation density Density was examined as an independent measure by creating concentric circular bins, with an increasing radius of 20 μm drawn around the center of the electrode site. The number of fully encompassed, activated neurons within the area defined by each bin were counted from 0 to 30 s post stimulation onset, and then divided by the total area encompassed by that bin. Activated neurons were separated into dense or sparse populations, by establishing a threshold of mean density + 1 SD. To compare high and low-density activation patterns across stimulation frequencies, density thresholds were established for all stimulation frequencies, and the lowest mean density threshold was applied to all trials.

Statistical analysis

All data in boxplots show the sample median (eyes), 25 th and 75 th percentiles (top and bottom edges of box), 1.5 times the interquartile range (whiskers), and outliers (individual dots). To assess significance between stimulation frequencies, a repeated measures two-way ANOVA was performed (SPSS, RRID:SCR_002865), followed by post-hoc two-sample unequal variance t-test with a Bonferroni correction.