Degradation of Head Direction Cell Activity during Inverted Locomotion

Introduction

Head direction (HD) cells, found primarily in the rat limbic system, carry information about the HD of the animal in the horizontal plane (for review, see Sharp et al., 2001a; Taube and Bassett, 2003). Each HD cell is tuned to a specific “preferred direction,” and different HD cells have different preferred directions, presumably allowing the combined activity of the population to reflect the perceived directional heading of the animal. HD cells use a number of different cues to maintain a consistent allocentric directional reference including both landmark cues and internal self-movement (idiothetic) cues (Taube et al., 1990; Taube and Burton, 1995). One feature that has received little experimental attention is the influence of body orientation on the activity of HD cells. Stackman et al. (2000) recorded HD cells while rats used a vertically oriented mesh ladder to move between the floor of the apparatus and an elevated ledge. They found that if the animal approached the ladder with its head oriented in the preferred direction of the recorded cell, it continued to fire as the animal ascended the ladder, but activity diminished to baseline on the return trip as the animal descended in a head-first orientation. Conversely, when the animal approached the ladder in the direction 180° opposite the preferred direction of the recorded cell, the cell was quiet while the animal ascended the ladder but was active during the descent. This pattern of activity suggests that the HD cells treated the ladder as a vertical extension of the floor (i.e., the cell responded as would be expected if the wall containing the ladder was folded down to enlarge the floor of the recording enclosure). This finding is not surprising if changes in the HD signal are primarily driven by angular velocity information carried by the horizontal semicircular canals (Taube and Bassett, 2003), because the transition between a horizontal and vertical surface requires a transient change in head pitch, but no head turn to the left or right. The change of surfaces leads to a different overall otolith signal but does not affect the signal conveyed by the semicircular canals. Once the animal is in the vertical plane, activation of the horizontal canals occurs as before whenever the animal rotates its head from side to side. However, if this principle is taken further and the animal performs a second right-angle turn into an inverted orientation (hanging upside down from the ceiling), then there will be a conflict between the actual HD and the HD as determined purely by horizontal angular velocity.

In an attempt to characterize how HD cell activity might change during 0 g, Taube et al. (2004) recorded HD activity during brief periods of weightlessness produced by parabolic flight. During weightlessness, the animal was positioned manually so that it was hanging from the ceiling in an inverted orientation. In this condition, HD cells showed a loss of directional specificity. However, it was not clear whether this loss of tuning was caused by the inverted body position, the 0 g condition, or because the animal was placed manually in the inverted orientation rather than allowing it to self-locomote into this position. In the same experiment, HD cells maintained directional specificity during the 0 g condition while the animal locomoted upright on the floor, providing evidence against 0 g as the sole mechanism of the effect.

In this investigation, we recorded HD cells from the anterodorsal thalamic nucleus (ADN) when the animal was inverted on the ceiling as described by Taube et al. (2004), but recordings were made in normal gravity, and the animal selflocomoted onto and across the ceiling rather than being passively placed in position. We report that HD cells showed dramatically distorted and often a complete loss of directional tuning during inverted locomotion.

Materials and Methods

Subjects and behavioral apparatus. Five female Long-Evans rats served as subjects. The weights ranged between 250 and 300 g at the start of the study. A food-restricted diet was used to maintain the rats at ∼85% of free feeding weight. Water was available ad libitum in the home cage. The animals were housed singly in clear plastic cages in a room maintained on a 12 h light/dark cycle.

During cell recording, animals traversed the inside of a track formed in the shape of a square ring for a food reward (Fig. 1). Each side was 1.22 m in length and 0.30 m wide. The floor, walls, and ceiling of the track were composed of galvanized wire mesh that provided traction for climbing. The bottom surface of the track was enclosed by clear Plexiglas walls and was divided into two equal halves by a clear Plexiglas barrier. Four video cameras were each oriented toward a different surface of the track, allowing the determination of animal location and heading regardless of where the animal was located in the apparatus. The apparatus was mounted on a central pivot point attached to the floor, allowing 360° rotation of the track relative to the room. Several stable visual landmarks were visible from the track, including a doorway and a black curtain along one wall of the room. The room was dimly illuminated by light from the adjacent doorway.

Behavioral task. Rats were trained to move between the two floor compartments of the track for a food reward (Fig. 1). To successfully complete the task, the animal had to climb the outer wall of the start compartment, locomote across the ceiling in an inverted orientation, and climb down the adjoining wall into the reward compartment where it could consume several small sucrose pellets from a food bowl. The current compartment then became the new start compartment, and the animal was rewarded for repeating the task in the opposite direction. Training the animals in this task required several weeks of shaping. Initially, a wooden platform touching both walls was hung a few inches from the ceiling. Animals were placed on the platform and reinforced for descending the wall to the reward compartment. After mastering this phase, the animals were placed in the start compartment and coaxed to climb up to the platform, cross, and descend to receive a reward. At this point, the wooden platform was removed, and the animal was coaxed across the ceiling while inverted with physical support provided from an experimenter. After much practice, most animals were able to perform the task without support. Within a few more weeks, the animals were able to perform the task in the opposite direction, and then they quickly learned to alternate directions between trials.

Surgical procedures and electrophysiological recording. After completion of behavioral training, each animal received an implantation of a recording electrode array to monitor neural activity during the behavioral task. Electrodes were constructed using methods described by Kubie (1984) and consisted of a bundle of 10 25-μm-diameter wires that were insulated, except at the tip. The wires were threaded through a stainless steel cannula that was moveable in the dorsal/ventral direction after being fixed to the skull using dental acrylic. Standard surgical procedures were followed for implantation of the recording electrodes (Taube, 1995). Briefly, animals were anesthetized with a ketamine-acepromazine-xylazine mixture. After exposure of the skull, a hole was drilled, and the electrode was implanted just dorsal to the right ADN based on coordinates (1.5 mm posterior to bregma, 1.3 mm lateral to bregma, and 3.7 mm below the cortical surface) provided by Paxinos and Watson (1998). After surgery, animals were allowed at least 1 week of recovery before commencement of experimental procedures. All procedures were conducted under an institutionally approved animal care and use protocol.

Electrical signals recorded in the brain were passed through a field-effect transistor in a source-follower configuration. Signals were amplified by a factor of 10,000-50,000, bandpass filtered (300-10,000 Hz, 3 dB/octave), and sent through a dual window discriminator (BAK Electronics, Mount Airy, MD) for spike discrimination. Screening for HD cells involved examining the electrical signal on each of the 10 implanted wires while the animal foraged in a wooden cylinder for food pellets dropped from above by an automatic food pellet dispenser. If no HD cells were detected, the electrode was advanced 25-50 μm, and the animal was returned to its home cage. If a HD cell was isolated, the animal was moved to the recording track for data collection.

Recording sessions began by placing the animal in one of the two floor compartments, at which time the animal would typically begin performing the task in alternating directions. During sessions, the position and directional orientation of the rat was monitored using an automated tracking system (Ebtronics, Elmont, NY). This video tracking hardware provided X and Y coordinates (256 × 256) of red and green light-emitting diodes (LEDs) that were secured 6 cm apart above the head and back of the animal, respectively. As the animal moved between the different surfaces of the track, an experimenter manually switched the input of the video tracker to the camera oriented toward that particular surface. The firing rate of recorded cells and LED positions were saved onto a personal computer at a rate of 60 Hz. Data analysis was accomplished off-line using custom software (LabView; National Instruments, Austin, TX). The HD of the animal was determined by the relative position of the red and green LEDs. For samples in which the animal was positioned on the floor or hanging from the ceiling, the calculated HD of the animal corresponded to the azimuth that the animal's head was facing using Cartesian coordinates, with 0° and 90° represented by the directions of east and north, respectively. With respect to the walls, the directional frame of reference was considered to be an extension of the floor [i.e., if an animal approached the eastern wall from the floor while facing 0° (east) and ascended the wall vertically (without rotating its head to one side or the other), the HD of the animal was still considered to be 0°]. If the animal then moved from the east wall onto the ceiling (without rotating its head to the left or right), the animal was facing 180°. Finally, during the descent of the west wall from the ceiling (without rotating its head), the animal was considered to be once again facing 0°. Because the linear movements of the animals were typically in the direction of the track, the ranges of HDs sampled while actively performing were usually aligned with the long axis of the track. This occurrence was especially problematic on the ceiling and walls, where there were sometimes two narrow ranges of unsampled HDs running at right angles to the long axis of the track. In cases in which the preferred direction of the recorded HD cell was >45° from the orientation of the track, the track was rotated to align it with the preferred firing direction of the cell to increase the number of samples in which the animal was facing the preferred direction of the cell. This procedure was not entirely successful, because most cells would rotate somewhat with the long axis of the track (i.e., the track itself exerted partial landmark control of HD activity).

Analysis of HD data. Cellular activity was examined to determine whether a given cell maintained directional tuning while the animal locomoted on each surface. Two methods were used for this analysis. First, firing rate × HD tuning curves was plotted across each surface for each cell (for examples, see Fig. 3), and the resulting plots were used to subjectively determine whether a cell showed directional modulation of firing activity. The second approach involved performing a Rayleigh analysis (Batschelet, 1981) to determine whether the HDs associated with cellular activity were distributed significantly nonrandomly, as would be expected if a cell was tuned directionally. The critical statistic of the Rayleigh test is the mean vector length, r, which varies between 0 and 1, with higher values indicating that the distribution of directions is clustered. The critical significance level of the mean vector length is determined by the number of observations (e.g., r_crit = 0.173 for n = 100 using our α criterion of 0.05), and if the r value meets this significance level, the distribution can be said to be nonrandom (Batschelet, 1981). To transform our firing rate × HD data into a format appropriate for this test, we divided the 360° directional range into 15 24° bins and calculated the number of spikes that occurred in each bin. Ideally, this distribution could then be subjected to a grouped Rayleigh test. However, the test assumes that each direction has an equal chance of being sampled (i.e., a spike has an equal chance of occurring in each directional bin), and the raw data do not meet this requirement because some HDs were invariably sampled more than others. To overcome this limitation, the cellular data on each surface were normalized by adjusting the number of spikes occurring in a given directional bin based on what would be expected if it had only been sampled as often as the fewest sampled bin. As an example, assume the fewest sampled bin was sampled 50 times. If another bin was sampled 100 times and during these samples a total of 30 spikes were detected, the number of spikes in this second bin was normalized to 15 spikes. Another assumption of this test is that each bin is sampled at least once. To improve the reliability of the data, we required that each bin be sampled at least 10 times. For cells in which this minimum sampling criterion was not met, the subjective analysis was used exclusively to determine whether the cell maintained directional specificity on a given surface.

To further assess the directional characteristics of recorded cells, the 360° directional range was divided into 6° bins, and the following measures were calculated for each cell on each surface: (1) directional information content; (2) preferred direction; (3) peak firing rate; (4) background firing rate; and (5) signal/noise ratio. Directional information content is a measure of how many bits of HD information is conveyed by each spike (Skaggs et al., 1993) and was calculated by the following formula: directional information content =∑ p_i (λ_i/λ) log₂ (λ_i/λ), where p_i is the probability that the head pointed in the ith directional bin, λ_i is the mean firing rate for bin i, and λ is the mean firing rate across all directional bins. The preferred direction was the directional bin with the highest average firing rate. The peak firing rate was the firing rate corresponding to the preferred direction. The background firing rate was the average firing rate of the three directional bins located 180° opposite the preferred direction. The signal/noise ratio was the ratio of the peak firing rate divided by the background firing rate. For the firing rate of a directional bin to be considered valid, it had to have been sampled a minimum of 10 times. This minimum sampling criterion was always met for all directional bins on the floor but was not always met for all bins on the walls and ceiling. However, we do not believe this caused a bias in these measures, because cells were selected for recording and analyses based on the criterion that the preferred directions on the floor tended to line up with the long axis of the track, directions that tended to be sampled heavily on the walls and ceiling. Statistical comparisons of these measures were performed using a Wilcoxon signed rank test.

Histology. At the completion of the experiments, animals were anesthetized deeply, and a small anodal current (20 μA, 10 s) was passed through one electrode wire to later conduct a Prussian blue reaction. The animals were then perfused transcardially with saline, followed by 10% formalin in saline. The brains were removed and placed in 10% formalin for at least 48 h. The brains were then placed in a 10% formalin solution containing 2% potassium ferrocyanide for 24 h and then reimmersed in 10% formalin (24 h) before being placed in 20% sucrose for 24 h. They were then frozen, sectioned (40 μm) in the coronal plane, stained with cresyl violet, and examined microscopically for localization of the recording sites. All recording electrodes were localized to the ADN.

Results

A total of 24 HD cells from five animals were monitored while the animal performed the task. Five cells were discarded from analysis because the preferred directions of the cell were sufficiently oriented away from the long axis of the track that the animal did not adequately sample the preferred direction while on surfaces other than the floor, and rotating the track failed to sufficiently improve this sampling. This left a total of 19 cells for additional analysis.

The number of total iterations performed while recording each cell depended on the motivation of the animal to perform the task. In a typical session, the animal would perform 5-10 complete iterations of the task in alternating directions while pausing 0.5-2 min between iterations. As expected from this pattern of behavior, animals typically spent much more time on the floor than on the other three surfaces of the track. Across all recording sessions, animals spent, on average, 17.7 min on the floor, 2.6 min on the east wall, 2.6 min on the west wall, and 3.9 min on the ceiling. On surfaces other than the floor, animals frequently did not sample all directional bins a minimum of 10 times, the number of samples considered necessary for reliable firing rate calculation. Using the more conservative 6° bin criterion, 11 cells did not meet this requirement for all directional bins on the east wall, for an average inadequately sampled directional range of 45.3 ± 12.9°. Thirteen cells fit this description on the west wall (average inadequately sampled range, 41.5 ± 1.0°), and nine cells did not meet the minimum sampling requirement for all directional bins while on the ceiling (average inadequately sampled range, 59.3 ± 7.0°). In each of these instances, the inadequately sampled bins were away from the direction of the long axis of the track and were usually divided between two directional ranges falling to either side of the long axis. To provide a sense of the distribution of the less sampled directional bins, the dotted line superimposed on the tuning curves shown in Figure 3 plots the number of samples used to calculate the firing rate plot at each directional bin. Note the pair of lower sampled directional ranges that often occurred on ceiling and wall plots around the 90° and 270° directional bins (directions that were orthogonal to the long axis of the track).

Discussion

The discovery of cells carrying information about the HD of the animal (HD cells) and cells carrying information about the location of the animal (place cells) has greatly influenced theorists interested in the neural mechanisms of navigation. Although these two cell types share many important features such as control by both external landmarks and internal idiothetic cues, an important difference is the degree to which spatially relevant activity is maintained across a variety of situations. Place cells exhibiting location-specific activity in a given environment or behavioral context will commonly become silent if the situation changes (O'Keefe and Conway, 1978; Kubie and Ranck, 1983). HD cells, on the other hand, typically maintain directionally specific responding across different environments or behavioral tasks (Taube et al., 1990; Golob et al., 2001). Although the preferred direction of HD cells will often shift if the animal is placed in an environment in the absence of polarizing cues, no previous study has shown an instance in intact animals under normal gravity in which HD cells lose directional specificity during freely moving locomotion. Here we show that the directional specificity of HD cells was lost or dramatically distorted when the animal locomoted in an upside-down body orientation.

A number of explanations might account for this finding, including the possibility that the effect is an artifact of some feature of the inverted recording procedure. For instance, the number of samples collected while the animal was traversing the ceiling and walls was much less than collected in a typical HD cell experiment, and so the apparent lack of directional tuning on the ceiling might perhaps be attributable to insensitivity because of the fewer number of samples in each directional bin. We reject this possibility because the number of samples collected on the ceiling was greater than the number of samples on the walls (an average of 3.9 min on the ceiling vs 2.6 min on the walls) and directional tuning was preserved in virtually every instance on the walls. Similarly, the animals often sampled an incomplete range of HDs while inverted, particularly for bins corresponding to directions orthogonal to the longitudinal axis of the track, bringing up the possibility that some cells remained tuned, but their preferred directions fell in unsampled bins. The fact that CD cells often showed random shifts potentially away from the long axis of the track lends credibility to this explanation. However, because undersampled regions typically occurred in two very narrow ranges lying orthogonal to the track (dotted lines of Fig. 3), we do not believe there was much opportunity to miss directional cells even if they had experienced a dramatically shifted preferred direction while inverted, because these narrow ranges were considerably smaller than the typical 90° firing range of HD cells, and the directional bins surrounding this undersampled range gave little indication of direction-specific firing.