Abstract
Head direction (HD) cells fire when an animal faces a particular direction in its environment, and they are thought to represent the neural correlate of the animal's perceived spatial orientation. Previous studies have shown that vestibular information is critical for generating the HD signal but have not delineated whether information from all three semicircular canals or just the horizontal canals, which are primarily sensitive to angular head rotation in the horizontal (yaw) plane, are critical for the signal. Here, we monitored cell activity in the anterodorsal thalamus (ADN), an area known to contain HD cells, in epstatic circler (Ecl) mice, which have a bilateral malformation of the horizontal (lateral) semicircular canals. Ecl mice and their littermates that did not express the mutation (controls) were implanted with recording electrodes in the ADN. Results confirm the important role the horizontal canals play in forming the HD signal. Although normal HD cell activity (Raleigh's r > 0.4) was recorded in control mice, no such activity was found in Ecl mice, although some cells had activity that was mildly modulated by HD (0.4 > r > 0.2). Importantly, we also observed activity in Ecl mice that was best characterized as bursty—a pattern of activity similar to an HD signal but without any preferred firing direction. These results suggest that the neural structure for the HD network remains intact in Ecl mice, but the absence of normal horizontal canals results in an inability to control the network properly and brings about an unstable HD signal.
SIGNIFICANCE STATEMENT Cells in the anterior dorsal thalamic nucleus normally fire in relation to the animal's directional heading with respect to the environment—so-called head direction cells. To understand how these head direction cells generate their activity, we recorded single-unit activity from the anterior dorsal thalamus in transgenic mice that lack functional horizontal semicircular canals. We show that the neural network for the head direction signal remains intact in these mice, but that the absence of normal horizontal canals results in an inability to control the network properly and brings about an unstable head direction signal.
Introduction
Head direction (HD) cells, first discovered in the postsubiculum (Ranck, 1984; Taube et al., 1990a), discharge when an animal is facing a specific direction, independent of its location and behavior in space. Under normal conditions, the HD signal is primarily controlled by visual landmark information (Goodridge and Taube, 1995; Goodridge et al., 1998; Yoder et al., 2011), but several studies have also shown the importance of idiothetic information (vestibular, motor, and proprioceptive cues; Taube et al., 1990b; Taube and Burton, 1995; Goodridge et al., 1998; Stackman et al., 2003; van der Meer et al., 2010; Valerio and Taube, 2012).
The influence of the vestibular system on HD cell firing has been a focus of many studies (Taube et al., 1990b; McNaughton et al., 1991). Indeed, the information coded by the horizontal semicircular canals (the acceleration of rotational head movements) makes it an ideal candidate for providing information about the animal's head movements in the yaw plane. Furthermore, the anatomical projections from the vestibular nuclei to the HD network via the nucleus prepositus hypoglosis, supragenual nucleus, and dorsal tegmental nucleus (DTN) reinforced this hypothesis (Biazoli et al., 2006). Recordings from DTN neurons have shown that they are sensitive to angular head velocity (AHV; Bassett and Taube, 2001; Sharp et al., 2001), and lesion studies have demonstrated that the DTN is necessary for a normal HD signal in the anterodorsal thalamus (Bassett et al., 2007). Lesions or inactivation of the vestibular end organs has also provided evidence pointing to the importance of vestibular information for generating a stable HD signal (Stackman and Taube, 1997). Muir et al. (2009) observed that bilateral occlusions of the semicircular canals disrupted the directionality of HD cells recorded in the ADN of chinchillas. Yoder and Taube (2009) used mutant tilted mice, which lack the development of otolith organs, to show that the elimination of the otolith signal (which encodes linear acceleration and gravity) altered both the number and the stability of HD cells. Nonetheless, some cells in the tilted mice still maintained an HD signal, suggesting that lesion of the otolith organs has less effect on the HD signal than interventions on the semicircular canals.
Currently, no study has investigated the specific role of the horizontal semicircular canals. Thus, the purpose of the present study was to address this issue by using the preexisting mutant epstatic circler (Ecl) mice that have bilateral malformation of the horizontal semicircular canals (Cryns et al., 2004). Importantly, other parts of the vestibular apparatus, including the otolith organs and the two remaining canals (superior and posterior), were shown to have normal morphology. We hypothesized that selective disruption of the horizontal canals would have the same effect as the global lesion of the vestibular organs and consequently disrupt the HD signal. We report here that the HD signal was disrupted in Ecl mice, but the cells that remained displayed bursty firing patterns that were similar to those observed in HD signals in Ecl littermates that did not express the mutation. Our results, based on the analyses of the burst characteristics, have important implications for the role the vestibular system plays in generating and controlling the HD signal.
Materials and Methods
Subjects.
C57L/J mice were mated with SWR/J mice, and an intercross was set up between the F1 progeny. The F2 progeny was scored phenotypically for their circling behavior at the age of 3 weeks (Cryns et al., 2002). This study compared the F2 circler animals of both sexes (Ecl, n = 9; one female, eight males; there was no difference between the one female and the eight males in terms of the results reported) with their noncircling littermates (control, n = 8; five females, three males). Mice were individually housed postoperatively and received food and water ad libitum. Recordings were conducted from Ecl mice that ranged in age from 344 to 512 d (mean, 411.9 d) and from controls that ranged in age from 364 to 489 d (mean, 420.2 d).
Electrodes.
The electrode array consisted of a bundle of eight stereotrodes (two 25-μm-diameter insulated nichrome wires twisted together; California Fine Wire) encased by a 26 gauge stainless steel canula. Each wire of the stereotrode contacted one gold pin of an electrode interface board (EIB 18; Neuralynx). Dental acrylic was used to encase the cannula, wires, and connectors and to attach the heads of three drive screws.
Surgery.
All procedures were performed in compliance with institutional standards as set forth by the National Institutes of Health Guide for the Care and Use of Laboratory Animals and the Society for Neuroscience. Mice were anesthetized with ketamine and xylazine (90 and 10 mg/kg, respectively) and positioned in a stereotaxic apparatus with bregma and lambda in the same horizontal plane. The scalp was retracted, and a hole was drilled above the ADN. Additional holes were drilled, and anchor screws (Lomat Precision) were inserted along the dorsal surface of the skull. These screws were reinforced to the skull with a drop of superglue. The electrode bundle was sterilized and coated (except for the tip) with polyethylene glycol before being positioned dorsal to the ADN using the following coordinates: 0.5 mm posterior to bregma, 0.7 mm lateral to bregma, and 2.0 mm ventral to the cortical surface. The drive was then fastened to the skull screws with grip cement (Dentsply International). The scalp was sutured around the electrode drive, and the animals were allowed to recover 1 week before recording.
Signal processing.
Neural recordings were preamplified by unit-gain operational amplifiers located on the head stage of the animal (Neuralynx). Signals were then amplified (5000–20,000×) and bandpass filtered (0.3–6 kHz; Neuralynx). When a signal crossed a specified threshold, the two channels of the stereotrode were digitized at 32 kHz and recorded onto an HP XW6200 workstation. Some cells were recorded using a wireless recording system (Triangle BioSystems). The telemetry system consisted of a 15 channel wireless head stage connected to the mouse's head cap, which contained the electrode pins. The head stage sent the electrical signals to a receiver that was connected to the abovementioned computer and recording system. No differences were observed between the wireless and cable recordings, and therefore all data were grouped together.
Recording procedure.
Mice were screened daily while foraging for randomly distributed food pellets (20 mg) in a black cylinder (40 cm diameter) that contained a prominent white cue card occupying ∼90° arc attached to the cylinder wall. The cylinder was surrounded by a floor-to-ceiling black circular curtain, and an overhead speaker, which was controlled by a white noise generator, emitted white noise to discourage the use of other auditory cues. When a cell of interest was identified and its electrical waveform was well isolated from background noise, it was recorded for 8 min in the cylinder. HD cells were recorded across five consecutive 8 min sessions (Fig. 1c): (1) standard 1, in which the cue card was positioned in the standard position; (2) rotation, in which the cue card was rotated 90° clockwise (CW) or counterclockwise (CCW); (3) standard 2, in which the cue card was returned to the initial location; (4) darkness, in which the white cue card was removed and the overhead lights were extinguished; and (5) standard 3, in which the white cue card was replaced at the standard location and the lights were turned on. Before each recording session (except darkness), the animals underwent a short disorientation period in which the mouse was put in an opaque box and then slowly rotated and translated back and forth for ∼1 min. An overhead video camera (XC-711; Sony) was used to monitor the animal's HD at 60 Hz by tracking the positions of a red and a green light-emitting diode (LED). These diodes were attached to the animal's head stage along the mouse's longitudinal body axis and were separated by ∼10 cm.
Characteristics of HD cells. a, Polar plots showing the Rayleigh r value and the PFD of each HD cell recorded from control animals. b, Example of a PFD shift observed when plotting the tuning curve of an HD cell during the first half versus the second half of a session. c, Overhead view of the landmark rotation and dark test sessions. d, Distribution of the amount of PFD shifts observed in HD cells within a recording session (first half vs second half). e, Polar plot showing the angular shift in the PFD between the standard 1 and the cue card rotation sessions. f, Polar plots showing the amount the cell's PFD shifted between the standard 1 and standard 2 sessions.
Spike sorting.
Single units were isolated “offline” manually using graphical cluster cutting software (Spikesort 3D; Neuralynx). Neurons were separated based on several characteristics of the spike waveforms (peak, valley, amplitudes, and energy). Evaluation of the biologically realistic interspike intervals and temporal autocorrelations were used to confirm single-unit isolation.
Data analysis.
HD was determined from the x–y coordinates of the two LEDs. Using 60 6° bins, the average firing rate as a function of HD was calculated by dividing the total number of spikes by the amount of time the animal's HD was within the range of each bin. For each cell, this function was subjected to a Rayleigh's test (Batschelet, 1981) to determine whether firing occurred randomly or clustered in a particular direction. A mean vector length (r) and its corresponding angle were calculated for each cell based on the tuning function of the cell. r values can range between 0 and 1, with values near 1 associated with strong directional activity and values near 0 associated with minimal directional activity. As in previous studies (Muir et al., 2009; Yoder and Taube, 2009), we adopted the criterion of r ≥ 0.4 for a cell to be considered an HD cell. Cells showing a 0.2 ≥ r ≥ 0.4 were considered as HD-modulated cells. Cells with r values <0.2 were considered to have no directional activity.
To examine the effect of the absence of horizontal canals on landmark control, we determined the response of a cell to rotation of the cue card. To measure the shift in the preferred firing direction (PFD) of the cell after rotation of the cue card, we calculated the cross-correlation between the tuning curves for sessions 1 and 2 as described previously (Taube et al., 1990a). Briefly, the firing-rate/head-direction function for session 1 was shifted CW in 6° steps, and the cross-correlation between the curves was recalculated at each step. The angular shift of the PFD of the cell was defined as the angle at which the cross-correlation was maximal.
For each cell, we calculated a burst index to represent the proportion of time a cell fired in high-frequency bursts or was inactive relative to the time during which action potentials occurred at a relatively constant rate (Yoder and Taube, 2009). For this measure, spikes were sorted into 1 s bins from the beginning to the end of a recording session. The burst index was defined as follows:
where F̅R̅ represents the mean firing rate over the entire session, and binsTotal represents the total number of bins during the session. Burst index values can range between 0 and 1, with a value of 0 indicating a firing rate that remains near the mean rate for the entire session and a value of 1.0 indicating the cell is bursty and either remains silent or fires near its maximal rate for the entire session. A firing rate × HD × time analysis was used to illustrate the bursting activity of the cells. This analysis calculates the average HD and firing rate for bins of 10 consecutive
To compare bursting activity across the two groups, bursty cells were submitted to several analyses. We adopted the burst definition used by Muir et al. (2009) in which a burst was defined as a minimum of three consecutive 166.67 ms samples that contained firing. Then, for each cell showing a burst index >0.4 (including HD and HD-modulated cells), we calculated the number of bursts per session, the duration of each burst, the number of spikes per burst, the firing range within a burst (HD at the beginning − HD at the end of the burst), and the angular deviation observed between successive bursts. To assess the extent each burst was directional, we also calculated a Rayleigh's r based on the distribution of directional headings for each spike in the burst (referred to as rburst) and the angular SD of these orientations. Thus, two r values were used in the analyses: (1) Rayleigh's r for the cell based on the entire session samples; and (2) an r value for each individual burst (referred to as rburst), based on the angular distribution of spiking activity within a burst. The second parameter (rburst) was used to compare the directionality of individual bursts between the two groups. For these circular measures and statistics, we used a MATLAB toolbox (CircStat) provided by P. Berens (University of Tubingen, Tubingen, Germany; Berens, 2009).
In addition to the HD, angular velocity, and angular acceleration measures mentioned previously, we also examined several behavioral parameters for each session: (1) the cumulative distance the mouse moved; (2) its linear velocity; and (3) its linear acceleration. To calculate the total cumulative distance (centimeters), the x–y coordinates were smoothed using a sliding window of 60 samples. The average linear velocity (cm/s) and linear acceleration (cm/s2) were then computed for each session.
The cross-correlograms (±10 s; bin size, 10 ms) and the power spectral density (PSD) analyses were conducted using data analysis software (NeuroExplorer). PSD raw power values (see Fig. 8) were calculated for frequencies ranging from 0 to 5 Hz (bin size, 5/128 Hz). For each group, the 10 cells that showed the highest peak in power value were included in the analysis.
For all analyses, means are reported along with SEM. Group comparisons were conducted on a personal computer with statistical software (Statview 5.0.1; SAS Institute).
Histology.
After electrophysiological recording, mice received an overdose of pentobarbital, and the electrode tip locations were marked by a small anodal current (15 μA, 20 s) to later conduct a Prussian blue reaction. Mice were then perfused intracardially with saline, followed by a 10% Formalin solution. Each brain was removed from the skull and was postfixed in a 10% Formalin solution containing 2% potassium ferrocyanide for at least 24 h. The brains were then cryoprotected in a 20% sucrose solution for 24 h and then frozen and cut coronally at 30 μm sections with a cryostat. Brain sections containing ADN were mounted on gelatin-coated microscope slides, stained with thionin, and examined under light microscopy to determine the location of recording sites.
Results
Histological analyses revealed that the recording electrodes penetrated the ADN of eight control and nine Ecl animals. A representative electrode track from an Ecl mouse is shown in Figure 2d. The waveforms of 133 cells were isolated from background noise, whereas the electrode bundle was estimated to be located between the dorsal and the ventral bounds of the ADN (73 cells in Ecl mice, 60 cells in control mice). From the analyses, isolated cells were classified into one of four types: (1) HD cells; (2) HD-modulated cells; (3) bursty cells; and (4) undefined cells. A χ2 test performed on cell type (Table 1) recorded from Ecl and control mice indicates that the two groups have significantly different cell type proportions (χ(3)2 = 30.74, n = 133; p < 0.0001).
HD-modulated cells (0.2 > r > 0.4). a, Polar plots showing the Rayleigh r values and the PFD of HD-modulated cells in control (blue dots, n = 9) and Ecl mice (gray dots, n = 7). b, c, Examples of tuning curves of HD-modulated cells recorded from a control (b) and an Ecl (c) mouse. d, Photograph showing electrode track (arrow) passing through the left ADN in an Ecl mouse. Dashed circle area outlines the ADN.
Number of cells showing bursty activity (burst index >0.4) recorded from the ADN of each animal
HD cells and HD-modulated cells
Of the 60 cells recorded in control mice, 21 cells (35%) were classified as HD cells, indicated by a significantly increased firing rate to a specific HD (Rayleigh's r ≥ 0.4). These HD cells from the heterozygous littermates, which did not display any circling behavior, had similar properties compared with HD cells from wild-type mice reported in a previous study (Table 2). Although some of the properties were significantly different between the two groups, if anything, the HD cells from the heterozygous littermates had stronger directional properties than the wild-type controls (based on directional information content and Rayleigh's r values). Possible reasons for this difference include different recording systems (window discriminators vs cluster cutting techniques), differences in ages and gender of the animals at the time of recording [all wild-type mice were males with a mean age of 169 d at the time of recording in the previous study vs a mixed gender group (five females, three males) recorded at a mean age of 420 d in the current study], and different experimenters performing the experiments.
HD cell firing properties of C57BL/6J and Ecl littermates
Using this criterion for defining the presence of direction-specific firing, no HD cells were identified in Ecl mice of 73 recorded cells. In contrast, Figure 1a shows the distribution of Rayleigh's r values from HD cells recorded in control mice. To determine the stability of the PFDs of the cells within a recording session, we computed the PFD of each cell in the first half of the session and compared it with the PFD of the cell recorded in the second half of the session. Figure 1b shows a representative example of intrasession PFD shift. As suggested by this example and visible in the distribution of PFD shifts in Figure 1d, the PFDs of the cells from control mice remain stable within an 8 min session, with a mean PFD shift of 8.1 ± 1.5°.
Additional tests were conducted to verify that the HD cells recorded in control mice were normal (Fig. 1c). We first assessed the influence of the visual landmark on HD cell activity by rotating the white cue card 90° CW or CCW. Thirteen cells were sufficiently stable (i.e., waveforms were well isolated from background noise) to test in this second 8 min session (cue rotation). The polar plots of the amount of angular shift in the PFD of each cell between standard 1 and cue rotation sessions (Fig. 1e) shows that the PFD shifted in the same direction and the same amount as the cue card, with a mean shift of 89.8 ± 4.4° (range, 69.9 to 123.3°), indicating that the cue card had strong visual control over the PFDs of the cells. Figure 1f shows polar plots of the amount the PFDs of the cells shifted between standard 1 and standard 2 sessions. The mean PFD shift between these two sessions was 11.9 ± 2.0°. A V test showed that the mean PFD shift was significantly clustered around 0° (u(13) = 4.95, p < 0.0001), indicating that the PFDs of the cells remain relatively stable across recording sessions.
In addition to HD cells that were defined based on r values >0.4, some cells had r values between 0.2 and 0.4 and were classified as HD-modulated cells because a directional signal was still visually evident in their tuning curves. However, the tuning curves of these HD-modulated cells were more irregular and less smooth compared with tuning curves from cells with r values >0.4. Nine HD-modulated cells (15%) were recorded in control mice, and seven HD-modulated cells (9.6%) were identified in Ecl mice. Figure 2a shows polar plots of these HD-modulated cells, with two representative examples of tuning curves: (1) one from a control mouse (Fig. 2b); and (2) the other from an Ecl mouse (Fig. 2c).
Bursty cells
As reported in previous studies (Muir et al., 2009; Yoder and Taube, 2009), bursty cells exhibit a burst firing pattern, similar to HD cells, but this pattern is not related to a specific HD of the animal. The occurrence of bursts was not associated with any particular type of movement made by the mouse, nor were they linked to any noticeable external event. Figure 3 shows four representative examples of this firing pattern. The firing rate × HD × time plot (Fig. 3a) represents the firing pattern of a normal HD cell, with this characteristic bursting activity. Importantly, in HD cells, these bursts occur when the animal is facing a specific direction: the PFD (Fig. 3a, dotted line). The bursts occur because the mouse is only in the cell's PFD momentarily as it is constantly moving around in the environment. Figure 3b plots the tuning curve of the same cell and confirms the HD-related activity of this cell. The mean burst index across all HD cells in control animals was 0.84 ± 0.02.
Bursting activity. a, Firing of a normal HD cell (red histogram) relative to the animal's HD (blue line). The dotted line represents the PFD of the cell at ∼40°. b, Tuning curve plotting the firing rate versus HD function for the cell shown on the left, based on the entire 8 min recording session. c, Bursty cell recorded from an Ecl mouse. Bursty cells also fire in bursts similar to HD cells, but, as visible in d, there is no relationship between the animal's HD and the occurrence of bursts. e, Two bursty cells recorded simultaneously from an Ecl mouse. Importantly, these two cells appear to fire in register. Indeed, the gray cell fires before the blue cell as would occur in normal HD cells. f, Tuning curves plotting the firing rate versus HD function for the two cells shown on the left. g–i, Cross-correlograms of pairs of bursty cells co-recorded. g, Cross-correlogram of the two bursty cells shown in Fig. 3e. The delay between the peak of activity observed in cell 1 (in gray) and cell 2 (in blue) was 620 ms. h, i, Cross-correlograms of two other sessions when more than one bursty cell was recorded (3 cells recorded in h, 2 cells recorded in i). Note that, although the pattern observed in g is interesting, the pairs shown in h and i are more representative of the six pairs of co-recorded bursty cells.
As mentioned previously, no HD cells were recorded from Ecl mice, but 33 cells showed a burst index >0.4. Figure 3, c and e, shows the firing rate × HD × time plot of three bursty cells recorded from Ecl mice (two bursty cells are depicted in Fig. 3e). Before analyzing the bursting activity in detail, it is important to note three points from these plots. First, the burst index of these three cells is comparable with the burst index for HD cells (Fig. 3b), despite the fact that the bursting patterns of these cells are not related to the animal's HD (Rayleigh's r < 0.2; Fig. 3, compare b with d or f). Second, it is noteworthy that for the episode shown in Figure 3e, the two bursty cells appear to fire in register. Indeed, in this case, the burst in the gray-colored cell always precedes the burst in the blue-colored cell, suggesting that these bursty cells exhibit network characteristics very similar to ones observed in HD cells. However, we only co-recorded multiple bursty cells in six instances; among these cases, this example is the only one showing a clear dynamic of burst succession. Figure 3g plots the cross-correlograms of the two bursty cells shown in Figure 3, e and f. Note that the cross-correlograms shown in Figure 3, h and i, are more representative of the behavior of the six pairs of bursty cells co-recorded from Ecl mice. Nevertheless, this one example in Figure 3e remains interesting because the order of firing succession (i.e., which cell fires first) here does not change with the direction of rotation as it did in the study by Muir et al. (2009) after canal occlusions. In other words, whether the animal rotates CW or CCW, the order of cell firing remains gray cell first, followed by firing in the blue cell. Third, in Figure 3, when comparing the plot in a with those in c or e, note that, in the same amount of time (60 s), the Ecl mice (Fig. 3c,e) make more head rotations, or movements in general (shown by the blue lines), than the control mouse (Fig. 3a). The hyperactivity of the Ecl mice, coupled with their circling behavior, were reported by Cryns et al. (2004) as a consequence of the malformation in the canals. However, it is unlikely that the bursty firing pattern was caused by the hyperactivity itself, because in previous studies, burstiness was observed in vestibular-deficient animals showing normal locomotor behavior (Muir et al., 2009).
Table 1 summarizes the number of bursty cells recorded in each animal for both control and Ecl mice. One difference from a previous study in mice (Yoder and Taube, 2009) is that we recorded 10 bursty cells in control animals (mean burst index, 0.65 ± 0.05) that were neither HD cells nor HD-modulated cells, whereas the previous study did not report any nondirectional bursty cells in control mice. In addition to these bursty cells, we recorded 21 HD cells (mean burst index = 0.84 ± 0.02) and 8 HD-modulated cells that had a burst index >0.4 (mean burst index, 0.76 ± 0.06). Nondirectional bursty cells showed a significantly lower mean burst index than HD cells (t(37) = 3.38, p < 0.01), but both cell types displayed a similar number of bursts per minute (nondirectional cells, 6 ± 1.25 bursts/min; HD cells, 6.1 ± 0.84 bursts/min; p > 0.05). Because our goal was to assess the effect of vestibular dysfunction, we grouped all bursty cells (HD cells, HD-modulated cells, and pure bursty cells that had a burst index >0.4) recorded from control animals and compared them with bursty cells found in Ecl mice.
Bursty cells from Ecl mice were composed of seven HD-modulated cells (mean burst index, 0.65 ± 0.11) and 24 non-directional cells. These cells (n = 29) showed a significantly lower mean burst index than bursty cells recorded in control animals (Ecl, 0.62 ± 0.03; control, 0.78 ± 0.02; t(66) = 3.87, p < 0.001). In addition, bursty cells from Ecl mice displayed significantly more bursts per minute than bursty cells from control mice (Ecl, 8.6 ± 1.0 bursts; control, 6.1 ± 0.7 bursts; t(66) = 2.57, p < 0.05).
In summary, these results indicate that, although there were differences in the burst index values between control and Ecl mice, the main difference between the two groups remains the absence of HD cells in Ecl mice.
Burst analyses: hypotheses
The HD network is usually modeled using a ring attractor network, which arises from two processes (Skaggs et al., 1995; Zhang, 1996; McNaughton et al., 2006; Clark and Taube, 2012). First, the connective architecture of the network generates a “bump,” or hill of activity. HD cells are connected with one another in a way that, when the animal is facing a given direction, ψ, a set of HD cells is activated (their PFD ≈ ψ), whereas the other HD cells (PFD ≠ ψ) are inhibited. Second, information about the animal's movements leads to updating the network, which results in a new perceived directional heading. This second aspect of the HD signal results from sensory information coming from various sources, including the vestibular organs, but also from proprioceptive and motor efference cues. Thus, a normal HD signal results from a hardwired network that generates a hill of activity and a landmark-based sensory signal that enables the updating of the HD signal as the animal moves through its environment. Does the vestibular system play more of a generative role or an updating role? As reported above, there are a high proportion of cells that can be characterized as bursty in Ecl mice, suggesting that the network architecture that generates the hill of activity (the individual burst) may remain intact, despite an abnormal vestibular signal. In the context of the present experiments, if the vestibular system is only playing an updating role to the HD signal, then the characteristics of the bursts recorded in the Ecl and control mice should be very similar to one another. In this case, the absence of an HD signal would result from the fact that the occurrence (timing) of the bursts appear disorganized and are not tied to any environmental or internal event, including the types of movements made by the animal. Alternatively, if burst activity is absent or is qualitatively dissimilar, then this result would argue that the vestibular system plays a more generative role in the HD signal. Note that these hypotheses purposely neglect the possible influence of motor efference, proprioceptive, or visual signals, which could partly compensate for the absence of vestibular information. However, because no HD cells were identified in Ecl mice, we considered that these compensatory mechanisms must play only a minor role in our experimental conditions.
To address this issue concerning whether the vestibular signal influences the generative or updating processes, we analyzed individual bursts characteristics (e.g., burst duration, number of spikes per burst, rburst) to assess whether a vestibular deficit altered the quality of the bursts. Although the time interval of each burst is generally short, from ∼100 ms to a few seconds, each spike can be associated with an HD. Thus, a mean vector can be calculated for each burst based on these samples. In general, this mean vector should be close to 1 when a short angular distance is covered during a burst, and a smaller mean vector length would occur if the burst lasted a longer period of time while the animal's HD covered a larger angular distance.
Individual bursts characteristics
Figure 4, a and b, shows two examples of the distribution of rburst values and mean angles for the bursts recorded from an HD cell in a control animal and from a bursty cell in an Ecl mouse, respectively. A comparison of the polar plots suggests that the angles encompassing the HD cell bursts (Fig. 4a) are much more clustered than those recorded from the bursty cell (Fig. 4b). This finding is consistent with the above analyses based on the Rayleigh's r of the session, which used samples from the entire session. Furthermore, this result shows that the angular instability of bursty cells in Ecl mice is also observed across individual bursts. In the example shown in Figure 4b (Ecl mouse), it is noteworthy that, despite the above mentioned angular instability across the burst, many individual bursts had very high r values (values close to 1 and near the circle periphery). However, Figure 4c shows that the bursts recorded from control animals have significantly higher rburst values than bursty cells recorded from Ecl mice (control rburst, 0.91 ± 0.01; Ecl rburst, 0.76 ± 0.02; t(66) = 7.49, p < 0.0001).
Comparison of burst characteristics in control and Ecl mice. a, b, Polar plots showing the Rayleigh's r value (rburst) and the angle of each burst for an HD cell from a control mouse (a) and for a bursty cell recorded from an Ecl mouse (b). c, Mean Rayleigh's rburst. d, Mean burst index. e, Mean number of spikes per burst. f, Mean burst duration. *p < 0.05, ****p < 0.0001.
We reported above that control animals had a higher mean burst index compared with Ecl mice (Fig. 4d). Additional analyses showed that bursts recorded in control animals contained more spikes/burst (t(66) = 2.27, p < 0.05) than bursts recorded from Ecl mice (Fig. 4e; control, 36.15 ± 6.69 spikes; Ecl, 17.61 ± 2.94 spikes). The analysis of burst duration (Fig. 4f) also indicated that bursts lasted longer in controls than Ecl mice (control, 1279.59 ± 106.78 ms; Ecl, 939.77 ± 56.96 ms; t(66) = 2.55 p < 0.05).
In summary, the burst characteristics in Ecl mice are significantly different on several measures (shorter duration, contain less spikes/burst, less directional) than bursts in control mice. These differences suggest that the yaw/horizontal rotational signal is not simply updating the network about the animals' movement but instead is involved in generating the burst activity. However, before leaving this issue, it is important to consider whether the burst characteristics were different between the two groups because of different behavior—specifically, it is possible that differences in locomotor activity (e.g., see Figs. 3 and 5) might account for the differences in burst characteristics. Thus, in the next section, we compare the locomotor behavior of the two groups.
Behavioral analyses
Figure 5a–d plots, respectively, the total linear distance, the angular distance, the mean linear speed, and the mean angular velocity observed during recording sessions in the two groups (control, n = 23; Ecl, n = 36). The analyses confirm the hyperactivity of Ecl mice (Cryns et al., 2004), revealing significant differences between the two groups in the distance traveled (control, 2143.9 ± 115.7 cm; Ecl, 4194.2 ± 247.8 cm; t(57) = 6.32, p < 0.0001), the angular distance covered by the animal's head (control, 65,318 ± 3236°; Ecl, 164,063 ± 7912°; t(57) = 9.63 p < 0.0001), the average AHV (control, 135.5 ± 6.9°/s; Ecl, 331.8 ± 15.0°/s; t(57) = 10.02, p < 0.0001), and the average linear speed of the animal (control, 4.43 ± 0.23 cm/s; Ecl, 8.49 ± 0.48 cm/s; t(57) = 6.42, p < 0.0001). The analyses performed on two other behavioral parameters gave comparable outputs, with a significant group effect on the average linear acceleration (control, 28.24 ± 0.83 cm/s2; Ecl, 46.83 ± 1.96 cm/s2; t(57) = 7.29, p < 0.0001) and on the average angular acceleration (control, 4592 ± 287°/s2; Ecl, 8931 ± 346°/s2; t(57) = 8.85, p < 0.0001). As shown in Figure 5e, these results indicate that Ecl mice show hyperactive behavior with frequent and high-velocity head rotations (compare with Fig. 5f showing a representative example of head turns in a control mouse). Two questions then arise from this observation. First, does the hyperactivity account for the absence of the HD signal in Ecl mice? Indeed, the absence of an HD signal could result from a disruption of the vestibular signal itself, but it is also possible that bursty cells are HD cells exposed to different behavioral conditions (i.e., very high AHV) that lead to the observed effects. Second, does the hyperactivity explain the differences in the burst characteristics observed between Ecl and control mice? The next section addresses the first question, whereas the section after addresses the second question.
Behavioral analyses. a, Distribution of the mean distance traveled by control (in blue) and Ecl (in gray) mice during an 8 min recording session. b, Distribution of the mean angular distance. c, Distribution of the mean linear speed. d, Distribution of the mean AHV. e, f, Representative examples of movements observed in Ecl (e) and control (f) mice. Note the episodes of fast CW and CCW turns observed in the Ecl animal.
HD signal and hyperactive behavior
Figure 6, a and b, shows the percentage of time spent at different AHVs (bins, 30°/s). These distributions confirm our preceding observations (Fig. 5d). Specifically, although 74.8% of the samples were recorded while the control animals' AHV was <90°/s, only 45.3% of the samples were in the same AHV range in Ecl mice. A previous study in intact rats showed that the HD signal remained stable despite relatively high-velocity rotations (150°/s; Zugaro et al., 2003), although in the current study we observed even higher AHV values in Ecl mice (Fig. 5d).
AHV and directionality of bursty cells. a, b, For each distribution the samples (
s) were grouped in 30°/s bins according to probability across the recording session. Averaged values for all bursty cells recorded from Ecl mice and controls are shown, respectively, in a and b. c–f, Rayleigh's r values calculated from all the samples are compared with r values obtained after excluding samples with high instantaneous AHV (>180°/s). For each bursty cell recorded from Ecl mice (c) or controls (e), the two r values are aligned. The average values for each group are shown in the left panels: Ecl in d and controls in f.
To determine how much influence behavior could have on the HD signal, we tested whether the strength of the directional signal (Rayleigh's r) was correlated to any of the behavioral measures in control animals. No significant correlation was found between these behavioral parameters (distance traveled, angular distance, linear speed, and angular velocity) and directionality (r values). The highest correlation observed was a trend for a negative correlation (r = −0.40, p = 0.07) between the Rayleigh's r and the mean linear speed, suggesting a small reduction in the strength of the directional signal for HD cells in sessions when animals showed high linear speed.
A second approach determined whether bursty cells in Ecl mice could be classified as directional when considering only episodes when the behavior of Ecl mice was comparable with the controls. Figure 6c–f shows the Rayleigh's r calculated using either all the recorded samples or only samples when the animal's AHV was <180°/s (mean, 56.4% of the samples; range, 37–85%). The two Rayleigh's r values for each bursty cell in Ecl mice are shown in Figure 6c. In general, Figure 6d shows that we observed a small but significant increase in the r values when restricting the range of AHV (all samples, r = 0.11 ± 0.01; AHV <180°/s, r = 0.15 ± 0.02; t(28) = 4.48, p < 0.0001). However, importantly, none of the cells started to display direction-specific firing as defined by reaching an r value ≥0.4 when considering only slower AHVs. Figure 6, e and f, shows the same analysis in controls in which there is no evidence for a significant effect of AHV. Similar analyses were conducted for linear speed, in which we calculated the Rayleigh's r after excluding samples when the instantaneous linear speed was >15 cm/s. When using only samples with low linear speed in Ecl mice (83.5% of the samples; range, 44.6–99.3%), we observed a small but significant increase in the mean r values (all samples, r = 0.11 ± 0.01; linear speed <15 cm/s, r = 0.14 ± 0.02; t(28) = 3.47, p < 0.01). However, once again, none of the bursty cells showed an r value >0.4 when excluding samples with high instantaneous linear speed.
Together, these results suggest that the alteration of the HD signal in Ecl mice is not attributable to the behavioral differences observed between the two groups in terms of movement parameters but rather to the absence of vestibular information resulting from the malformation of the horizontal canals.
Burst characteristics and hyperactive behavior
In a previous study (Muir et al., 2009), the bursty cells observed after semicircular canals occlusion displayed bursts that were very similar to the ones observed in normal HD cells. In contrast, in the Ecl mice, we observed several differences in the nature of the bursts (Fig. 4a–d), suggesting a possible involvement of the vestibular signal in the generation of the HD signal. To determine whether these differences were attributable to hyperactivity or the vestibular malformation, we analyzed the bursts characteristics at different AHVs and compared their characteristics across the two groups. Because of limited sampling above 180°/s in control mice (Fig. 6b), we restricted our comparisons to AHVs between 0 and 180°/s and used intervals of 30°/s. In addition, to get a fair comparison of the bursts recorded from the two groups, we randomly selected an equal number of bursts for each group for each AHV category (0–30°/s, n = 565; 30–60°/s, n = 361; 60–90°/s, n = 236; 90–120°/s, n = 113; 120–150°/s, n = 55; 150–180°/s, n = 34). Using this procedure was important because the differences between the two groups in the bursts sampling at a given AHV (Fig. 6a,b) would generate an artificial group effect. We also verified that the mean angular velocity during bursts were comparable between the two groups for each AHV category (p values >0.05). We then analyzed four parameters and paid close attention to group effects, which would indicate differences in the generation of individual bursts when movement behavior (AHV) was held constant.
rburst values
The effect of AHV on rburst values is shown in Figure 7a. An ANOVA reveals an effect of AHV interval, with higher r values for smaller AHVs (F(5,2716) = 37.56, p < 0.0001), as well as a group effect with higher r values for controls compared with Ecl mice (F(1,2716) = 35.02, p < 0.0001) but no interaction between the two factors (p > 0.05). These results indicate that individual bursts are more directional in control mice than in Ecl mice. Post hoc tests show that this difference is significant for AHVs ranging from 0 to 90°/s (0–30°/s, t(1128) = 8.46, p < 0.0001; 30–60°/s, t(720) = 6.42, p < 0.0001; 60–90°/s, t(470) = 60.01, p < 0.0001) but not for AHV intervals > 90° (p > 0.05). Interestingly, this comparison also shows that, for both groups, the bursts are affected the same way by an increase in AHV and that the slight loss in directionality observed with AHV (decreased r) is independent of the vestibular alteration.
Influence of behavior on burst characteristics. a, Rayleigh rburst values in control and Ecl mice for bursts recorded while the animal was rotating within a given AHV range. b, Angular distance traveled by controls and Ecl mice between the start and end of each burst at different AHV ranges. c, Mean firing rate of the burst at different AHV ranges. d, Drift observed between successive bursts at different AHV ranges.
Angular distance covered during burst
Compared with the rburst results, the only significant effect revealed by the analyses was an AHV interval effect (F(5,2716) = 505.22, p < 0.0001). There were no group or group × AHV interval interaction effects (p values >0.05), indicating that, for this parameter, both groups show similar bursts. Thus, the only effect here is that as the animal's AHV increased, as one would predict there was an accompanying increase in the angular distance the animal turned its head during a burst, and this result was true for both groups (Fig. 7b). Therefore, although spike activity within a burst is less tuned to a single HD in Ecl mice than in controls (Fig. 7a), the angular distance covered by the animal's head turn in a burst is not different between groups. Thus, unlike the rburst values reported above, this finding indicates that the burst appears to encode the amount the animal will turn its head, and that this activity is independent of the vestibular signal.
Burst firing rates
Given the differences in burst duration (Fig. 4f) and number of spikes per burst (Fig. 4e), we compared the firing rates of the burst between the two groups (Fig. 7c). This analysis revealed no AHV interval effect (p > 0.05), but rather a global group effect (F(1,2716) = 23.70, p < 0.0001), and only a trend for an interaction (F(5,2716) = 2.11, p = 0.06). Therefore, when comparing similar behavior based on AHV, some differences remain between the two groups, suggesting that the vestibular signal influences the firing within each burst of the cell.
Drift between successive bursts
Figure 7d plots the drift observed in the PFD of the cell between successive bursts: the amount of angular distance separating the animal's mean HD during one burst from its mean HD of the next one. The analysis revealed an AHV interval effect (F(5,2716) = 4.29, p < 0.001), a group effect (F(1,2716) = 64.74, p < 0.0001), but no interaction between the two factors (F < 1). This result indicates that the drift observed in Ecl mice is significantly larger than the drift observed in control animals at all AHV intervals and that, as suggested previously (Fig. 6c–f), this instability in the angle of individual bursts (Fig. 4b) is not attributable to AHV differences between the two groups. This last result confirms, for individual bursts, our previous analyses showing that the absence of directionality in bursty cells recorded from Ecl mice is not caused by hyperactivity.
Burst analysis summary
In this section, we addressed whether the differences in locomotor activity could account for the differences we observed in the burst characteristics in Ecl mice (Fig. 4). We found that the firing properties of bursty cells (r value, burst firing rate) were significantly different between the two groups, even when AHV was held constant. However, Ecl mice and controls show similar angular distances covered during their bursts, suggesting that the network that generates the hill of activity is able to activate and inhibit bursty cells in very similar ways, despite the absence of a vestibular signal from the horizontal canals. Furthermore, the mechanisms that control these properties must be independent of the vestibular signal.
Periodicity in bursty cell activity
When looking at the activity of the two bursty cells in Figure 3e, one can notice that, in addition to the linkage in their firing, the two cells appear to display a periodic firing pattern. Whether this periodic rhythm (4.34 s) is attributable to (1) intrinsic properties of the cell, (2) network properties, or (3) behavior of the animal is unclear. Two possible hypotheses could explain this pattern. First, bursty cells, when deprived of the vestibular signal, could exhibit some network dynamics that contain a specific rhythmicity. Thus, when the cells are deprived of vestibular information about how the head is turning, the (attractor) network defaults to a mode in which the activity hill rotates rhythmically. An alternative hypothesis is that the periodicity does not reflect a pure network dynamic but rather a dysfunctional HD network that is subjected to sensory/motor stimulation when the animal circles repeatedly. However, before addressing this issue, we first analyzed all bursty cells to determine how often this phenomenon occurred compared with HD cells. Because the activity of HD cells is associated with a specific HD, one would expect it to be less periodic than the one observed from bursty cells. We computed PSD analyses for 10 other bursty cells, which were selected based on the highest power values from their autocorrelograms (Fig. 8b). We compared these results from the bursty cells with 10 HD cells that contained the highest power values from their autocorrelograms (Fig. 8a). Figure 8c shows that the mean of the maximum power values is significantly higher in HD cells (mean peak power, 0.61 ± 0.06) compared with bursty cells (mean peak power, 0.27 ± 0.10; t(18) = 2.89, p < 0.01). These analyses indicate that, except for a few cases (three bursty cells from the same animal), the bursty cells appear to be less periodic than HD cells. This analysis confirms that, even if the bursty rhythmicity is an interesting phenomenon, the periodicity observed for the bursty cells depicted in Figure 3e is more of an exception than a general finding.
Analyses on the periodicity in bursty cells. a, b, PSD computed between 0 and 5 Hz for the 10 HD cells (a) and bursty cells (b) with the highest power values. c, Mean of the peak power values for HD cells (n = 10) and bursty cells (n = 10). **p < 0.01. d, Example of bursting firing pattern recorded from animal sv 26. The black dots plot the animal's HD. The red vertical bars plot the bursting activity of the cell, which appears periodic when the animal is rotating (bottom left), but the cell stops firing when the animal is motionless (bottom left, HD ≈ 280°) from 115 to 135 s. The rhythmic activity is then visible again when the animal restarts its stereotypical rotation movement.
Nonetheless, for the few cells that did display periodic bursting, one can ask whether this firing pattern is an intrinsic network dynamic or whether it is modulated by the animal's behavior. Figure 8e shows the bursting patterns of the “periodic” bursty cells. The bottom left inset of Figure 8d depicts a period of time when the animal makes fast head turns; note that the cell activity appears periodic. However, when the animal is immobile (bottom right inset), the bursting activity stops, suggesting that it does not reflect an intrinsic property of the network, but rather is more likely a consequence of motor behavior (including proprioceptive feedback) acting on the network.
Discussion
The present study assessed the activity of cells recorded from the anterior thalamus in Ecl mice, a preexisting mutant with a characterized vestibular pathology consisting of a malformation of the lateral (horizontal) semicircular canals (Cryns et al., 2004). Although a few HD-modulated cells (Rayleigh 0.2 < r < 0.4) were recorded from these mice (n = 7), we were unable to identify any HD cells, whereas normal HD cells were recorded from their littermate controls that did not express the vestibular deficient phenotype. Thus, despite the presence of stable visual landmarks, cells could no longer generate and maintain normal direction-specific firing. We focused our analyses on two issues regarding the importance of the horizontal canal for the HD signal. We first compared the directionality of ADN cells in the two groups, and then we analyzed the quality of the bursts to assess whether the structure of the signal was altered in Ecl mice.
Critical involvement of the horizontal canals in HD signal stability
Previous studies have reported that a lesion of the vestibular organs disrupts the activity of HD cells in rats (Stackman and Taube, 1997), chinchillas (Muir et al., 2009), and mice (Yoder and Taube, 2009). In this last study, the authors tested the effect of an alteration in the otolith organs, which sense linear acceleration, and observed that, although an HD signal remained present, the cells preferred firing directions were unstable, even in a stable environment. Because no HD cells were found in studies with complete lesions of the vestibular organ, this finding suggested that the semicircular canals play a prominent role in generating a normal HD signal. More specifically, one would predict that the horizontal canals, which sense angular acceleration primarily in the horizontal plane, is the key structure conveying angular head movement information to the HD network. The present study confirms this prediction by showing that a specific alteration of the horizontal canals is sufficient to produce the same effect as a global lesion of the vestibular end organs (Stackman and Taube, 1997; Stackman et al., 2002) or an occlusion of the vestibular canals (Muir et al., 2009). Interestingly, bursty cells were recorded from Ecl mice: these cells, which were observed in previous studies after lesions of the vestibular system, display a bursting firing pattern similar to the one observed with HD cells, except that the bursty cells show no PFD in their bursting activity. Because of the similarities of their firing pattern with HD cells (Fig. 3), these bursty cells were interpreted as HD cells disconnected from the updating system that conveys information about the animal's head movements.
Bursty cell: the influence of behavior
The fact that Ecl mice were selected based on a behavioral abnormality (circling behavior) raised the question of the effect of these high-velocity rotations on the HD signal. More specifically, we examined whether the bursty cells recorded from Ecl mice were simply HD cells exposed to unusual behavioral conditions of high AHVs. Our findings suggest that the alteration of the HD signal is independent of the rotational velocity and that, although both circling behavior and HD signal alteration may have the same cause, this behavior does not explain the absence of HD cells in Ecl mice.
Burst quality: minor effect of vestibular alteration
A second issue we assessed relates to the quality of the burst activity observed in the two groups. Are the bursts recorded from Ecl mice similar to those observed in controls? In other words, except for the directional instability of the bursty cells in Ecl mice, are the bursts characteristics identical for the two groups? The first analysis on individual bursts (Fig. 4) suggested substantial differences, with bursts recorded from Ecl mice being significantly less directional, containing a lower number of action potentials and showing shorter durations. These results suggested that, in addition to playing a critical role in maintaining a stable PFD, the vestibular signal was also critical for the generation of the burst activity. However, an analysis of burst properties based on the animals' movement patterns found that Ecl mice and controls displayed very different behavior, with Ecl mice showing a markedly increased proportion of high AHVs compared with control mice (Fig. 5). This behavioral disparity could have contributed to the differences in the burst characteristics. Therefore, we compared the bursts characteristics at different AHV intervals using a comparable number of bursts (Fig. 7). The burst analyses revealed that individual bursts were better tuned to a specific HD in control animals at low AHVs (0–90°/s) and that controls showed higher firing rates in bursts. Importantly, however, the analyses also showed that the angular distance covered by the head during the burst was not different between the two groups, suggesting that the mechanism that generates the burst is not affected by the absence of a vestibular signal. Thus, taken together, the network that generates the HD signal appears to remain relatively intact in Ecl mice, thereby leading to the generation of burst events that are generally similar to one another and which are equally affected by increases in AHV. Furthermore, the malformed horizontal canals in Ecl mice leads to an altered vestibular signal, which no longer acts on the attractor network sufficiently to bring about a normal HD signal. Thus, although the vestibular system appears to play both a modulatory and a generative role in forming the HD signal, the fact that (1) the network can still generate burst events, (2) these bursts events cover a similar angular distance, despite the absence of a vestibular yaw signal, and 3) bursts recorded in Ecl mice are as resistant to high AHV as cells from controls, indicates that burst generation remains quite stable in the absence of a signal coming from the horizontal semicircular canals. In summary, in contrast to our previous views (e.g., Taube, 2007; Yoder and Taube, 2014), these results suggest that the vestibular signal plays a more important role in providing information to the HD network than in generating the hill of activity observed in bursty cells.
Is the burst generator a “default mode” for HD cells that are disconnected from their vestibular inputs?
Does the network that generates the hill of activity, here disconnected from AHV information in Ecl mice, show any stable behavior that would be informative about its properties? In other words, does the network behave in a particular way in the absence of a vestibular signal that conveys primarily horizontal head movements? When we normalized the AHV, we observed that the activity within the burst covers similar angular distances for bursty cells and HD cells. Because Ecl mice cannot compute their AHV based on vestibular inputs, this finding suggests that the sequence of activity that initiates the generation of the bursts is hardwired and is a characteristic default mode for the network. We also observed that a few bursty cells (3 of 24) showed a periodic pattern of activity, possibly revealing additional information about the default mode that may be attributable to network properties (Figs. 3e–g, 8d). However, closer examination of this phenomenon indicates that the periodicity was not a pure network behavior, but rather more likely attributed to sensorimotor influences (Fig. 8d). Indeed, a pure network activity would be independent of the animal's behavior; in our case, we observed that the periodicity of the cells is clearly, at minimum, modulated, if not caused by the animal rotational behavior (Fig. 8d).
HD signal generation
The HD cell network (Fig. 9) has been shown to have an overall hierarchical organization (Taube, 1998, 2007). Indeed, lesions of the postsubiculum, or medial entorhinal cortex, do not alter the HD signal recorded in the ADN (Goodridge and Taube, 1997; Clark and Taube, 2011), but lesions of any area afferent to it leads to the loss of direction-specific firing in the ADN. However, how the HD signal is disrupted depends on which brain area that is lesioned. Disrupting subcortical structures before the DTN, such as the supragenual, prepositus, or vestibular apparatus (Fig. 8, red dashed line) generates a bursty signal in the ADN comparable with the one we observed in the present study (Stackman and Taube, 1997; Muir et al., 2009; Yoder and Taube, 2009; Clark et al., 2012; Butler and Taube, 2015). In contrast, lesions to more rostral brain areas (DTN, lateral mammillary nucleus) above the red line results in the loss of direction-specific firing but no burst activity in ADN cells (Blair et al., 1998; Bassett et al., 2007). Consistent with attractor network models, these results suggest that this division between the different groups of subcortical nuclei play different roles in signal processing. Indeed, the bursting activity induced by either vestibular disruption or damage to brainstem nuclei (supragenual and prepositus nuclei) supports the hypothesis that the network organization remains intact rostrally at the level of the dorsal tegmental/lateral mammillary nuclei but lacks accurate updating that is normally provided via AHV information. Our results reinforce this view by demonstrating that burst qualities are weakly affected—even at high AHVs—by an alteration of the horizontal semicircular canal signal.
Circuit diagram showing principal nuclei generating and processing the HD signal. Subcortical areas that, when lesioned, induce a bursty firing pattern in ADN are shown in red. Areas that when lesioned induce an abolition of the HD signal in the ADN without burstiness are represented in green. Confirming previous results our data suggest that the generation of the HD signal results from the reciprocal interactions between the lateral mammillary nucleus and the dorsal tegmental nucleus.
Summary
From a historical perspective, we have long considered the vestibular system important for generating the HD signal, in which without an intact vestibular system there was an absence of direction-specific activity throughout the HD cell network (Stackman and Taube, 1997). Our results support this view and extend it by demonstrating that the horizontal semicircular canals, which are particularly sensitive for processing angular rotation in the horizontal plane, are critical for producing a normal HD signal. Importantly, however, our findings suggest that the vestibular system is not responsible for generating the attractor network, but rather only controlling it. The similarity of burst firing in Ecl mice to that of controls suggests that the attractor network remains intact after vestibular disruption, but the network is no longer controlled properly. This dysfunction gives rise to a rapidly drifting network that manifests itself as bursty cells in rostral brain areas that contain HD cells.
Footnotes
This work was supported by National Institutes of Health Grants DC009318 and NS053907. We thank Jennifer Marcroft for technical assistance. Parts of this work have been published previously in abstract form at the Society for Neuroscience meeting, New Orleans, Louisiana, in 2012.
The authors declare no competing financial interests.
- Correspondence should be addressed to Jeffrey Taube, Dartmouth College, Department of Psychological and Brain Sciences, 6207 Moore Hall, Hanover, NH 03755. jeffrey.taube{at}dartmouth.edu
