Chronic stability of single-channel neurophysiological correlates of gross and fine reaching movements in the rat

Abstract

While substantial task-related neural activity has been observed during motor tasks in rodent primary motor cortex and premotor cortex, the long-term stability of these responses in healthy rats is uncertain, limiting the interpretability of longitudinal changes in the specific patterns of neural activity associated with learning or motor recovery following injury. This study examined the stability of task-related neural activity associated with execution of two distinct reaching tasks in healthy rodents. A novel automated rodent behavioral apparatus was constructed and rats were trained to perform a reaching task combining a 'gross' lever press and a 'fine' pellet retrieval. In each animal, two chronic microelectrode arrays were implanted in motor cortex spanning the caudal forelimb area (rodent primary motor cortex) and the rostral forelimb area (rodent premotor cortex). We recorded multiunit spiking and local field potential activity from 10 days to 7-10 weeks post-implantation to characterize the patterns of neural activity observed during each task component and analyzed the consistency of channel-specific task-related neural activity. Task-related changes in neural activity were observed on the majority of channels. While the task-related changes in multi-unit spiking and local field potential spectral power were consistent over several weeks, spectral power changes were more stable, despite the trade-off of decreased spatial and temporal resolution. These results show that neural activity in rodent primary and premotor cortex is associated with specific phases of reaching movements with stable patterns of task-related activity across time, establishing the relevance of the rodent for future studies designed to examine changes in task-related neural activity during recovery from focal cortical lesions.

Fig 1. Complex reaching task.

To evaluate the neurophysiological correlates of both gross and fine reaching movements within individual animals, we designed a novel, automated behavior box (A). B. At the beginning of each trial rats were required to use their preferred forearm to depress a lever placed outside an opening at the back corner of the box, causing a door at the front of the box in the corner diagonally opposite from the lever to open, providing access to a food pellet placed on a ledge outside the box. After transitioning to the door, rats reached through an opening to grasp and retrieve the food pellet reward. After detecting an attempt to retrieve the pellet via an infrared beam across the front of the box, the door was automatically closed, preventing repeated retrieval attempts. C. To examine the neurophysiological correlates of task performance, video recordings were used to identify the time points for the reach-to-lever onset, lever press, reach-to-pellet onset, and pellet grasp onset within each trial.

Fig 2. Chronic microelectrode implants.

A. The rodent motor system consists of the caudal forelimb area (CFA), a homologue of M1, as well as a secondary rostral forelimb area (RFA), a homologue of premotor cortex. B. Sixteen-channel chronic microwire electrode arrays were implanted into each rat with one array placed in RFA and a second array implanted into CFA. Intracortical microstimulation mapping was used to confirm the locations of RFA and CFA in three rodents with stereotaxic coordinates used to determine implant locations in the remaining rats. C. Approximate implant locations relative to skull landmarks were determined from intraoperative photographs. With respect to stereotaxic coordinates, the location of the electrode arrays implanted into RFA was very consistent across rats with more variability in the locations of the CFA arrays due to the variable locations and orientations of blood vessels relative to the craniectomy opening.

Fig 3. Electrophysiological analyses.

Multi-unit firing and local-field potential (LFP) activity were examined in each recording session. A. To examine multi-unit firing, signals were re-referenced to the common average and band-pass filtered between 300Hz and 3000Hz. Each trace shows data filtered for multi-unit activity and aligned to the same time point from randomly selected channels in a single rat. A scale of 100 μV is indicated by the vertical red line. The multi-unit detection threshold of -4.5 x the RMS voltage of each channel is shown by the dashed red line below each trace. B. LFP activity was visualized by filtering between 5Hz and 400Hz after re-referencing to the common average. Each trace shows data filtered for LFP activity and aligned to the same time point from randomly selected channels in a single rat. A scale of 250 μV is indicated by the vertical red line.

Fig 4. Exemplar reach-related channel.

A. Task-related neural activity is shown for an exemplar channel from the RFA array of Rat 1 as shown by the red electrode in the topographic plot. B. Raster plots show multi-unit spike times (black ticks) aligned to the onset of the reaching movement towards the lever (left) and food pellet (right), showing an increase in firing rate correlated to the reach onset. C. Multi-unit firing rates were estimated by counting spike times within 100 ms windows shifted by 50 ms between windows and then averaging firing rates across trials, showing a statistically significant (p<0.05) increase in firing rate aligned to the onset of the reach in both task components. The statistically significant peak firing rates are indicated by the yellow highlighting. D. Spectral power changes are plotted aligned to the onset of reaching movements to the lever (left) and pellet (right). Spectral power changes were log transformed and z-scored relative to randomly selected time windows between trials, therefore, positive values indicate increases in spectral power, and negative values indicate decreases in spectral power. A broadband increase in spectral power in frequencies above 60Hz was observed for both the reach to the lever and the reach to the pellet. E. High-gamma band (70Hz-110Hz) power was significantly increased around both the reach to the lever (left) and the reach to the food pellet (right). The statistically significant peak modulations of high-gamma band spectral power are indicated by the yellow highlighting.

Fig 5. Exemplar grasp-related channel.

A. Task-related neural activity is shown for an exemplar channel from the RFA array of Rat 2 as shown by the red electrode in the topographic plot. B. Raster plots show multi-unit spike times (black ticks) aligned to the lever press (left) and onset of the pellet grasp (right), showing an increase in firing rate correlated to the grasp onset. C. Multi-unit firing rates were estimated by counting spike times within 100 ms windows shifted by 50 ms between windows and then averaging across trials. Mean firing rates synchronized to the lever press and grasp were both significantly (p<0.05) higher than expected by chance with a greater depth of modulation during the grasp. The statistically significant peak firing rates are indicated by the yellow highlighting. D. Spectral power changes are plotted aligned to the lever press (left) and pellet grasp (right). Spectral power changes were log transformed and z-scored relative to randomly selected time windows between trials, therefore, positive values indicated increases in spectral power, and negative values indicate decreases in spectral power. A broadband increase in spectral power in frequencies above 60Hz was observed for the pellet grasp. E. High-gamma band (70Hz-110Hz) power was significantly increased around both the lever press (left) and the pellet grasp (right) with a greater increase in high-gamma band power associated with the grasp. The statistically significant peak modulations of high-gamma band spectral power are indicated by the yellow highlighting.

Fig 6. Exemplar press-related channel.

A. Task-related neural activity is shown for an exemplar channel from the RFA array of Rat 5 as shown by the red electrode in the topographic plot. B. Raster plots show multi-unit spike times (black ticks) aligned to the lever press (left) and onset of the reach to the pellet (right), showing a decrease in firing rate correlated to the lever press followed by an increase in firing rate prior to the reach to the pellet. C. Multi-unit firing rates were estimated by counting spike times within 100 ms windows shifted by 50 ms between windows and then averaging across trials. Mean firing rates show a complex task response with firing rates significantly (p<0.05) decreased around the lever press and 650 ms prior to the reach to the pellet and then significantly increased immediately prior to the reach to the pellet. The statistically significant peak firing rates are indicated by the yellow highlighting. D. Spectral power changes are plotted aligned to the reach to the lever (left) and pellet grasp (right). Spectral power changes were log transformed and z-scored relative to randomly selected time windows between trials, therefore, positive values indicated increases in spectral power, and negative values indicate decreases in spectral power. A broadband increase in spectral power in frequencies above 60Hz was observed for the grasp of the food pellet (right) with a lower amplitude increase in broadband spectral power observed before the reach to the lever. E. High-gamma band (70Hz-110Hz) power did not show a decrease around the lever press but instead showed a small, but statistically significant (p<0.05), increase in high-gamma band power prior to the reach to the lever. The high-gamma band power maintained the significant increase in activity near the time of the pellet retrieval observed in the multi-unit firing rate. The statistically significant peak modulations of high-gamma band spectral power are indicated by the yellow highlighting.

Fig 7. Microelectrode classification.

Each microelectrode was classified based upon whether it was significantly modulated by the reach to the lever, the lever press, or not modulated during the lever press. Similarly, each electrode was also classified as modulated when aligned to the reach to the food pellet, when aligned to the grasp of the pellet, or not modulated during the retrieval. Each classification was made using both the multi-unit firing rate (A) and high-gamma band (70-110Hz) spectral power (B). When examining changes in multi-unit firing rate, more channels were modulated by the reaching movements than either the lever press or grasping movements. When examining changes in high-gamma band power, while more channels were modulated by the reach than the lever press, slightly more channels were modulated by the grasp than the reach to the pellet.

Fig 8. Exemplar task-related activity across recording days.

A. The stability of task-related neural activity is shown for an exemplar channel from the RFA array of Rat 5 as shown by the red electrode in the topographic plot. B. Raster plots show multi-unit spike times (black ticks) aligned to the onset of the grasping of the food pellet on several recording days, showing an increase in firing rate correlated to the pellet grasp that is conserved across days. C. Multi-unit firing rates were estimated by counting spike times within a 100 ms window shifted by 50 ms between windows and then averaging firing rates across trials, showing a statistically significant (p<0.05) increase in firing rate aligned to the grasp. While the timing of this increase in firing rate is conserved across days, there is some variance in the depth-of-modulation observed. D. Spectral power changes are plotted aligned to the onset of the grasping of the pellet on the same days. Spectral power changes were log transformed and z-scored relative to randomly selected time windows between trials, therefore, positive values indicate increases in spectral power, and negative values indicate decreases in spectral power. A broadband increase in spectral power in frequencies above 60Hz is observed aligned to the grasp in each day. E. High-gamma band (70Hz-110Hz) power was significantly increased around the grasp with similar depth-of-modulations observed on each recording day. All plots were generated from the same exemplar channel within RFA of Rat 5.

Fig 9. Cross-day stability.

The stability of task-related changes in neural activity during the pellet retrieval was examined by determining the correlation between the daily mean firing rate and the overall mean firing rate as well as the correlation between the daily high-gamma band power change and the overall mean high-gamma band power change for each channel. Plots show the mean daily correlation across channels for the pellet retrieval and the error bars show the standard error. While a decrease in signal quality was observed at 9 weeks and 5 weeks in rats 1 and 2 respectively, overall, task-related changes in multi-unit activity and high-gamma band power were stable across days for each rat. While changes in multi-unit firing rate and high-gamma band power were both stable across recording days, the task-related the high-gamma band power change had higher correlations between the daily and overall mean than was observed for task-related multi-unit firing rate changes. Additionally, high-gamma band power changes show a slower decrease in stability from the degradation in signal quality observed in rats 1 and 2.