Functional and anatomical relationships between the medial precentral cortex, dorsal striatum, and head direction cell circuitry. I. Recording studies

Abstract

Head direction (HD) cells fire as a function of the animal’s directional heading and provide the animal with a sense of direction. In rodents, these neurons are located primarily within the limbic system, but small populations of HD cells are found in two extralimbic areas: the medial precentral cortex (PrCM) and dorsal striatum (DS). HD cell activity in these structures could be driven by output from the limbic HD circuit or generated intrinsically. We examined these possibilities by recording the activity of PrCM and DS neurons in control rats and in rats with anterodorsal thalamic nucleus (ADN) lesions, a manipulation that disrupts the limbic HD signal. HD cells in the PrCM and DS of control animals displayed characteristics similar to those of limbic HD cells, and these extralimbic HD signals were eliminated in animals with complete ADN lesions, suggesting that the PrCM and DS HD signals are conveyed from the limbic HD circuit. Angular head velocity cells recorded in the PrCM and DS were unaffected by ADN lesions. Next, we determined if the PrCM and DS convey necessary self-motion signals to the limbic HD circuit. Limbic HD cell activity recorded in the ADN remained intact following combined lesions of the PrCM and DS. Collectively, these experiments reveal a unidirectional functional relationship between the limbic HD circuit and the PrCM and DS; the limbic system generates the HD signal and transmits it to the PrCM and DS, but these extralimbic areas do not provide critical input or feedback to limbic HD cells.

NEW & NOTEWORTHY Head direction (HD) cells have been extensively studied within the limbic system. The lesion and recording experiments reported here examined two relatively understudied populations of HD cells located outside of the canonical limbic HD circuit in the medial precentral cortex and dorsal striatum. We found that HD cell activity in these two extralimbic areas is driven by output from the limbic HD circuit, revealing that HD cell circuitry functionally extends beyond the limbic system.

INTRODUCTION

Animals must maintain a sense of direction to efficiently navigate within their environment. At the neural level, direction is represented by the activity of head direction (HD) cells. These neurons fire only when the animal’s head is pointing in a specific direction in the horizontal plane, referred to as the preferred firing direction (PFD), and remain inactive when the head is oriented elsewhere. Within a population of HD cells, PFDs are uniformly distributed across 360°. These populations operate much like a compass, providing a continuous directional signal used for navigation (Taube et al. 1990a; Valerio and Taube 2012; van der Meer et al. 2010).

HD cells are found in multiple structures within the rodent limbic system; collectively, these interconnected structures form the limbic HD circuit (Taube 2007). This circuit originates in the brain stem where the reciprocally connected dorsal tegmental nucleus and lateral mammillary nucleus are thought to generate the HD signal in a manner dependent on input from the vestibular system (Sharp et al. 2001a; Stackman and Taube 1997; Valerio and Taube 2016; Yoder and Taube 2009). Vestibular input provides critical self-motion information used to update HD cell activity when the animal rotates its head and assumes a new directional heading (Clark and Taube 2012). The HD signal is first conveyed from the lateral mammillary nucleus to the anterodorsal thalamic nucleus (ADN) (Bassett et al. 2007; Blair et al. 1998); subsequently, the ADN transmits the HD signal cortically to the retrosplenial cortex, postsubiculum, and medial entorhinal cortex (Cho and Sharp 2001; Goodridge and Taube 1997; Winter et al. 2015a). The retrosplenial cortex and postsubiculum project back to subcortical areas of the limbic HD circuit, transmitting information about visual landmarks used to anchor directional representations to the external environment (Clark et al. 2010; Goodridge and Taube 1997; van Groen and Wyss 1990, 1992, 2003; Yoder and Taube 2011; for review see Taube 2007).

Interestingly, small populations of HD cells have also been found in two extralimbic areas that are not traditionally associated with the limbic HD circuit: the medial precentral cortex (PrCM) in the frontal lobe and the dorsal striatum (DS), a component of the basal ganglia (Mizumori et al. 2000, 2005; Ragozzino et al. 2001; Wiener 1993). Each structure has diverse functions relating to motor control (Kravitz et al. 2010; Sinnamon and Galer 1984), self-motion encoding (Gardiner and Kitai 1992; Kim et al. 2014; Mizumori et al. 2005; Trytek et al. 1996; Wiener 1993; Yamin et al. 2013; Yeshenko et al. 2004), spatial processing and navigation (Devan and White 1999; Mizumori et al. 2005; Schmitzer-Torbert and Redish 2004; Whishaw et al. 1987; Yamin et al. 2013), goal-directed behavior (Balleine et al. 2007; Ostlund et al. 2009; Yin et al. 2005), and habitual behavior (Devan et al. 2011; Yin and Knowlton 2006). The properties of PrCM and DS HD cells have not been systematically characterized, and the origin of these extralimbic HD signals remains unclear; HD cell activity in the PrCM and DS could be driven by limbic HD circuit output, or it may be generated intrinsically. Regarding the latter possibility, the DS contains neurons modulated by the animal’s head movements (Gardiner and Kitai 1992; Kim et al. 2014; Trytek et al. 1996). Similar cells in the dorsal tegmental nucleus and lateral mammillary nucleus, referred to as angular head velocity (AHV) cells, are considered necessary for the generation of the limbic HD signal (Clark and Taube 2012). Therefore, the DS may be capable of generating HD cell activity independently of the limbic system via locally encoded head movement information.

Conversely, it is possible that HD cells in the limbic system require output from the PrCM and/or DS for their normal functioning. The PrCM and DS are involved in motor control and therefore may provide the limbic HD circuit with critical self-motion information in the form of motor efference signals. Previous studies have shown that the PrCM is involved in generating horizontal head movements associated with orienting responses (Erlich et al. 2011; Sinnamon and Galer 1984). Given that the HD signal must be updated during such head movements, motor efference signals arising from the PrCM may be of particular relevance to the limbic HD circuit. Previous findings suggest that HD cells are indeed influenced by motor information. For example, when motor efference signals are disrupted by passively transporting the animal in a cart, HD cell PFDs can become unstable (Stackman et al. 2003; Yoder et al. 2011; see also Winter et al. 2015b). Moreover, eliminating motor efference signals by head-fixing and passively rotating the animal decreases the signal-to-noise ratio of the HD signal (Shinder and Taube 2014; see also Shinder and Taube 2011). Finally, lesioning the interpeduncular nucleus, a structure thought to convey motor information to the limbic HD circuit, disrupts HD cell activity (Clark et al. 2009). In addition to motor efference signals, DS neurons modulated by movements of the animal’s head may convey a necessary self-motion signal to the limbic HD circuit in a manner similar to vestibular input.

In this study, we explored these possibilities by conducting a pair of lesion and recording experiments in rats. First, we recorded the activity of PrCM and DS neurons in control animals and animals with ADN lesions, a manipulation known to eliminate HD cell activity in downstream nodes of the limbic system (Goodridge and Taube 1997; Winter et al. 2015a). Next, we performed the converse experiment in which we recorded the activity of ADN neurons in animals with combined PrCM and DS lesions. Collectively, our results show that 1) PrCM and DS HD cells are largely identical to those in the ADN in terms of their tuning properties and responses to environmental manipulations; 2) the PrCM and DS HD signals, but not AHV signals, are dependent on output from the limbic HD circuit; and 3) combined lesions of the PrCM and DS do not significantly influence the activity of HD cells in the ADN. Taken together, these results suggest a unidirectional functional relationship between the limbic HD circuit and the PrCM and DS; these two extralimbic areas receive their HD signals from the limbic system but do not provide critical input or feedback to limbic HD cells. The anatomical basis of this functional relationship is revealed in our accompanying report (Mehlman et al. 2019).

MATERIALS AND METHODS

Subjects

Female Long-Evans rats (n = 39) weighing ~300 g were used in the experiments (Harlan Laboratories, Indianapolis, IN). Prior to surgery, animals were pair-housed with food and water available ad libitum. Following surgery, animals were housed individually with water available ad libitum and food access restricted to maintain an 85% ad libitum weight. Colony rooms were kept on a 12:12-h light-dark cycle at all times. All experimental procedures were approved by the Dartmouth College Institutional Animal Care and Use Committee and conformed to the standards outlined in the National Institutes of Health Guide for the Care and Use of Laboratory Animals.

Recording Electrodes

Electrodes were constructed as described previously (Kubie 1984). Briefly, each electrode consisted of an array of ten 25-μm nichrome single-wires or eight 16-μm nichrome stereotrodes (i.e., two wires spun together) threaded through a single 26-gauge stainless steel cannula. Wire tips were gold plated to lower the impedance to ~0.5 MΩ. Dental acrylic was used to encase the electrode in an assembly containing three drive screws threaded into plastic cuffs. These screw cuffs served as the anchor point between the electrode and the skull such that, throughout the course of the experiment, the electrode array could be gradually advanced ventrally (2.5–3 mm) by turning the drive screws.

Surgical Procedures

Animals were anesthetized with a ketamine-xylazine cocktail (90 mg/kg ketamine + 10 mg/kg xylazine injected intramuscularly) or isoflurane (3% vaporized in oxygen) and then placed in a stereotaxic frame (David Kopf Instruments, Tujunga, CA). An incision was made to expose the skull, and small holes were drilled above the structures to be lesioned or implanted with an electrode array. Neurotoxic lesions were performed by injecting 0.1–0.2 μl of a 100 mM solution of N-methyl-d-aspartate (NMDA; dissolved in 0.9% saline; Sigma-Aldrich, St. Louis, MO). Injections were delivered through a blunt-tip 33-gauge stainless steel cannula connected via polyethylene tubing to a 10-μl Hamilton syringe (Hamilton, Reno, NV) that was depressed at a constant rate by a syringe pump (0.1 μl/min; Razel Scientific Instruments, Stamford, CT). The cannula was left in place for 3–6 min following injection delivery to aid diffusion. Electrodes were implanted by positioning the electrode array within or above the structure of interest and then anchoring the screw cuffs to skull screws and the surrounding skull surface with dental cement (Dentsply International, York, PA). For animals that received neurotoxic or sham lesions, the electrode array was implanted immediately following the final injection of NMDA or saline. Animals recovered for 7 days following surgery before commencing recording.

In experiment 1, a total of 18 animals were implanted with an electrode array in the left or right PrCM. Seventeen of these animals were divided into two main groups: unlesioned control animals (n = 9; 8 implanted with single-wire arrays and 1 implanted with a stereotrode array) and animals with ADN lesions (n = 8; all implanted with single-wire arrays). Animals in the lesion group received three bilateral injections of NMDA at the following coordinates: −1.3, −1.7, and −2.1 mm anterior/posterior (A/P), 1.4 mm medial/lateral (M/L), and −5.2 mm dorsal/ventral (D/V). For these coordinates and all that follow, A/P and M/L measurements are relative to bregma and D/V measurements are relative to the cortical surface. In some cases, these injections also damaged a small portion of the hippocampus dorsal to the ADN. To control for this unintended damage, one additional animal was given a small lesion confined to this portion of the hippocampus. This animal received two bilateral injections of NMDA at the following coordinates: −1.8 and −2.2 mm A/P, 1.4 mm M/L, and −4.2 mm D/V.

The 18 animals described above were implanted with the electrode array starting 1–1.5 mm above the DS within the intermediate or deep layers of the PrCM. During subsequent screening (see below), the electrode array was gradually advanced ventrally over weeks into and through the DS. These electrode arrays were implanted within the following range of coordinates: 0.5 to −0.5 mm A/P, 1.4 to 2.0 mm M/L, and −1.0 to −2.0 mm D/V. To further examine the distribution of HD cells across different portions of the DS, an additional group of four animals, referred to as “extended-DS” animals, were implanted with a single-wire electrode array starting deeper in the brain within the most dorsal portion of the DS; these electrode arrays were implanted within the following range of coordinates: 2.4 to 0.5 mm A/P, 1.5 to 2.5 mm M/L, and −2.5 to −3.4 mm D/V. HD cells and AHV cells recorded in the extended-DS group were included in analyses examining HD cell and AHV cell characteristics in control animals; however, because this range of implant coordinates is more anterior, lateral, and ventral compared with that of the lesioned animals, these four animals were excluded from all analyses comparing the control and lesion groups in an effort to make the two groups more equivalent.

In experiment 2, a total of 17 animals were implanted with a single-wire electrode array directly above the right ADN at the following coordinates: −1.9 mm A/P, 1.3 mm M/L, and −3.7 mm D/V. These animals were split into two groups: control animals (n = 7) and animals with combined PrCM and DS lesions (n = 10). Animals in the lesion group received four bilateral injections of NMDA at the following coordinates: 0.8 mm A/P, 1.7 mm M/L, and −4.0 and −4.5 mm D/V (sites 1 and 2); −0.3 mm A/P, 2.0 mm M/L, and −3.4 and −3.9 mm D/V (sites 3 and 4). These lesions targeted the portions of the PrCM and DS in which HD cells and AHV cells were recorded in experiment 1. A subset of animals in the control group (n = 4) received a sham lesion in which the lesion procedure was performed using 0.9% saline instead of NMDA; the remaining animals (n = 3) were unlesioned. HD cell characteristics did not differ significantly between sham-lesioned and unlesioned animals (data not shown); therefore, these animals were combined into a single control group for all analyses.

Recording Procedures

A head stage containing a preamplifier was connected to the electrode array implanted in each animal. From the head stage, the electrophysiological signals detected by each wire were passed through a flexible cable, supported by a counterweight, to an overhead commutator (Biela Engineering, Irvine, CA). The commutator, in turn, passed the signals to equipment located in an adjacent room, where they were amplified by a factor of 10,000–50,000 (Grass Instruments, West Warwick, RI) and bandpass filtered (300–10,000 Hz; Peavey Electronics, Meridian, MS). For each single wire in the electrode array, a dual-window discriminator (BAK Electronics, Sanford, FL) was used to manually isolate spike waveforms from background noise online. An overhead color video camera (Sony, Tokyo, Japan) connected to an automated video tracking system (Ebtronics, Elmont, NY) measured the x and y locations of two light-emitting diodes (LEDs) attached to the head stage, at a rate of 60 Hz. One LED (red) was centered above the animal’s head, and the other LED (green) was centered above the animal’s back. A computer (Apple, Cupertino, CA) running custom software (LabVIEW; National Instruments, Austin, TX) acquired and stored spike time stamps from isolated neurons and the video tracking data. The animal implanted with a stereotrode array was connected to a different recording system (Neuralynx, Bozeman, MT), and spike waveforms were manually isolated offline using a spike sorting procedure in which waveform characteristics were compared across the two wires in each stereotrode (see Winter et al. 2015a).

Once or twice a day over the course of 4–8 wk, each electrode wire was screened for electrophysiological activity while the animal freely moved in a gray cylindrical arena (76-cm diameter, 51-cm height) foraging for food pellets (20 mg; Bio-Serv, Frenchtown, NJ) dropped at random intervals from an overhead dispenser. The cylinder was surrounded by and centered within a black curtain (~2.5-m diameter) hanging from the ceiling to the floor to mask visual cues outside of the arena. A single polarizing visual landmark was provided by affixing a white cue card to the inside wall of the cylinder (subtending ~100° of arc). Eight symmetrically arranged overhead lights provided illumination, and an overhead speaker generated white noise to mask auditory cues. The floor of the cylinder was covered with a new sheet of gray paper at the beginning of each screening and recording session to minimize residual olfactory cues. In the adjacent room, the experimenter monitored electrophysiological activity via an oscilloscope (Tektronix, Beaverton, OR) and an audio speaker; the animal’s behavior was monitored via a video feed from the overhead video camera. When the activity of single neurons was detected and their waveforms isolated, the cells were recorded for an 8-min session (referred to as a “preliminary” session). In some cases, isolated neurons displaying activity clearly unrelated to the animal’s HD or movement were noted but not recorded. If no cells were recorded, or when recording sessions concluded, the electrode array was advanced ventrally ~50–100 μm and the animal was returned to the colony room before the electrode wires were screened again 3 h later or the next day. If a neuron satisfied the criteria for a HD cell (see below), the experimenter performed the following series of environmental manipulations (Fig. 1).

Fig. 1.

Diagrams of head direction (HD) cell recording arenas and procedures. A: during HD cell recording, animals foraged for food pellets in a cylinder containing a single polarizing visual landmark (i.e., the cue card). Each HD cell was recorded across a sequence of 5 sessions: standard (STD1) → rotation (ROT) → standard (STD2) → dark (DRK) → standard (STD3). B: one HD cell per animal was recorded in the dual-chamber apparatus (DCA). The HD cell was first recorded in a familiar cylinder (CYL1), and then the animal walked through a novel passageway into a novel rectangle and the cell was recorded again (REC). Finally, the animal walked back through the passageway into the cylinder and the cell was recorded a third time (CYL2). In A and B, the curved or straight lines along the outside edge of the arenas represent cue card locations.

Landmark Rotation

The landmark rotation was performed to evaluate the capacity of a visual landmark to control a HD cell’s PFD. This test consisted of three separate 8-min sessions (Fig. 1A). First, a standard session (STD1) was recorded in the cylinder with the cue card occupying the same position as during screening. This session was followed by a rotation session (ROT) in which the cue card was rotated 90° in the clockwise (CW) or counterclockwise (CCW) direction. Subsequently, a second standard session (STD2) was recorded in which the cue card was returned to the same position it occupied during STD1.

Before each session, the animal was disoriented by placing it in an opaque box that the experimenter slowly rotated in alternating directions while simultaneously walking around the cylinder in alternating directions. Following this disorientation procedure, the animal was placed into the cylinder from a random direction and recording commenced. All manipulations of the cue card were performed while the animal was outside of the cylinder in the opaque box. The animal foraged for food pellets during each session.

Dark Session

The dark session (DRK) immediately followed STD2 (Fig. 1A) and was performed to measure changes in HD cell tuning properties induced by the removal of visual information and to examine the extent to which HD cells maintain a stable PFD in the absence of a salient visual landmark. DRK consisted of an 8- to 16-min session, which was identical to the standard sessions except that the cue card was removed and the overhead lights were turned off. Following DRK, the cue card was returned to the same position it occupied during STD1 and STD2, the overhead lights were turned on, and a final 8-min standard session (STD3) was recorded.

Before each session, the animal was disoriented and placed into the cylinder from a random direction. All manipulations of the cue card and overhead lights were performed while the animal was outside of the cylinder in the opaque box. The animal foraged for food pellets during each session.

Dual-Chamber Apparatus

The dual-chamber apparatus (DCA) was used to examine how extralimbic HD cells respond when the animal walks from a familiar environment into a novel environment and then back into the familiar environment. For this test, the cylinder described above was replaced with a new arena consisting of two gray compartments, a cylinder (75-cm diameter, 41-cm height) and a rectangle (51-cm width, 69-cm length, 41-cm height), connected by a narrow U-shaped passageway (15-cm width, 157-cm length, 41-cm height; Fig. 1B). The animal’s access to this passageway was controlled by the experimenter via manually operated doors located where the passageway connected to each compartment. A white cue card was affixed to the inside wall of each compartment. In the cylinder, the cue card occupied the same position as the cue card in the cylinder used previously during screening and the three standard sessions. The rectangle contained a cue card rotated 90°CCW with respect to the cue card in the cylinder. The floors of the cylinder and rectangle were covered with a new sheet of gray paper at the beginning of testing. Because the cylinder used in the DCA was similar to the cylinder used previously, it was considered a familiar environment. The rectangle and passageway, unlike anything encountered previously by the animals, constitute a novel environment. In the limbic system, the PFDs of HD cells remain relatively unchanged as the animal exits the familiar cylinder and enters the novel passageway and rectangle; these HD cells are presumably using self-motion information to maintain a stable directional representation across the environments (Stackman et al. 2003; Taube and Burton 1995; Yoder et al. 2011).

The procedure for the DCA consisted of three separate sessions. The animal was first disoriented and placed into the cylinder from a random direction, and then an 8-min session was recorded while the animal was in the cylinder with the door closed (CYL1). Next, the cylinder door was opened, allowing the animal to walk through the passageway and into the rectangle. Once in the rectangle, the rectangle door was closed and an 8- to 10-min session was recorded (REC). Finally, the rectangle door was opened to allow the animal to walk back through the passageway and into the cylinder. The cylinder door was then closed, and a second 8-min session was recorded (CYL2). During each session the animal foraged for food pellets. This test was performed only once per animal because the rectangle and passageway lose their novelty after the initial test.

Some HD cells could not be recorded across the full series of environmental manipulations due to the loss of waveform isolation.

Recording and Behavioral Analyses

Data analyses were performed offline using custom software (LabVIEW). For each recorded cell, the preliminary session or STD1 was used to generate a directional tuning curve by binning the animal’s HD, determined by the location of the two LEDs, into 60 6° bins and computing the average firing rate within each bin. Tuning curves were then used to compute the firing characteristics of each cell. A Rayleigh’s test (Batschelet 1981) was performed to determine the magnitude of directional tuning. This test measured the extent to which spiking activity was clustered around a single HD, yielding a Rayleigh r value ranging from 0 to 1, with 0 representing a neuron that fires uniformly across 360° and 1 indicating that spiking occurs exclusively within a single bin. We adopted criteria used previously in our laboratory for classifying cells as HD cells (e.g., Valerio and Taube 2016; Yoder et al. 2015): cells with Rayleigh r ≥ 0.4 were defined as HD cells. To further quantify the directional tuning of HD cells, we calculated their directional information content (Skaggs et al. 1993), a measure of the extent to which spikes predict the animal’s HD. Values approaching or ≥1 indicate strong directional tuning (see Stackman and Taube 1998). The peak firing rate of each HD cell was defined by the value of the bin with the highest firing rate, and the direction associated with that bin defined the cell’s PFD. The directional firing range was determined using a triangular model (see Taube et al. 1990a). We computed the anticipatory time interval (ATI) of each HD cell using a time-shift analysis (Blair and Sharp 1995). This measure describes the temporal offset at which a HD cell’s firing is maximally correlated with the animal’s HD, with positive values indicating firing predictive of the animal’s future HD. The ATI was defined by the x-intercept of a line fit to points in a scatterplot illustrating the “separation angle” between CW and CCW directional tuning curves after the recorded spike train was incrementally shifted forward and backward in time relative to the video tracking data (see Clark et al. 2010). The correlation coefficient (Pearson’s r) of this best-fit line had to be ≥0.85 for a HD cell’s ATI value to be included in group analyses.

To compute the angular shift of each HD cell’s PFD from one session to another, a cross-correlation was performed in which the tuning curves from the two sessions were offset relative to each other in 6° increments and then correlated; the angular offset producing the maximum correlation was defined as the PFD shift. The circular mean of all shift values within a group was calculated, and a Rayleigh’s test was performed to determine if the shift values displayed significant clustering. Previous studies have shown that when multiple HD cells are recorded simultaneously, their PFDs shift in register (Taube et al. 1990b); therefore, the shift values of HD cells recorded simultaneously were averaged together to avoid overweighting sessions in which multiple HD cells were recorded. To compute the angular shift of PFDs from STD1 to ROT, the direction of shift was defined relative to the rotation of the cue card (CW or CCW); a positive value represents a shift in the same direction as the cue card, whereas a negative value indicates a shift in the opposite direction. For all other shift values reported, a positive value represents a CCW shift, whereas a negative value indicates a CW shift.

To determine the angular stability of each HD cell’s PFD within a session, we first computed the peak firing rate as described above and then computed the firing rate in successive 1-s bins. For each bin in which the firing rate was ≥75% of the peak firing rate (i.e., bins in which the animal was presumably oriented toward the PFD), the average HD of the animal during that bin was computed (see Yoder and Taube 2009). The angular deviation of these HDs was used as a measure of PFD stability; this measure produces a low value when the PFD is stable and a higher value when the PFD is unstable. To compare PFD stability under dark vs. light conditions, the stability of each HD cell’s PFD was measured over identical time periods (i.e., if DRK was recorded for 16 min, the cell’s stability was computed using the first 8 min of DRK and then compared with that cell’s stability during the 8-min STD1 recorded under light conditions).

In experiment 1, we also analyzed the extent to which cell activity was modulated by the animal’s AHV in the horizontal plane. For each recorded cell, the preliminary session or STD1 was used to plot the average firing rate as a function of the animal’s AHV in the CW and CCW directions using 6°/s bins. We then calculated the slope and correlation coefficient (Pearson’s r) of the best-fit lines for the CW and CCW segments of the tuning curve. Next, for each cell, we performed 500 iterations of a shuffling procedure in which the recorded spike train was randomly shifted in time relative to the video tracking data, and then the slope and correlation values for the CW and CCW segments were recalculated. Cells displaying CW and/or CCW tuning with 1) an unsigned slope value ≥0.03 spikes/° and an unsigned correlation value ≥0.6 and 2) unsigned slope and correlation values greater than or equal to the 95th percentile values obtained from the distributions of that cell’s shuffled data were classified as AHV cells. The slope and correlation values used as criteria for classifying cells as AHV cells are based on the properties of AHV cells recorded previously in the dorsal tegmental nucleus and lateral mammillary nucleus (Bassett and Taube 2001; Stackman and Taube 1998). For some cells, both the CW and CCW segments of the tuning curve satisfied the AHV cell criteria. These cells were classified as symmetric AHV cells if both segments were similarly modulated by the animal’s AHV (e.g., the cell’s firing rate increased during CW and CCW head turns). However, if the CW and CCW segments displayed opposite responses (e.g., the cell’s firing rate increased during CW head turns but decreased during CCW head turns), the cell was classified as an asymmetric AHV cell. If only one segment of the tuning curve satisfied the AHV cell criteria, the cell was classified as an asymmetric AHV cell. To examine AHV cells independently of HD cells, we excluded all AHV cells that also met the criteria for HD cells (i.e., we examined “pure” AHV cells). Neurons classified as both HD cells and AHV cells (i.e., conjunctive HD × AHV cells) are discussed in results.

For experiment 1, the depth of each isolated cell was estimated using the drive screw turn records and the reconstructed electrode array paths (see below). All neurons encountered dorsal to −2.3 mm D/V were classified as PrCM cells, and all cells encountered ventral to −2.3 mm D/V were classified as DS cells; this depth represents the dorsal edge of the corpus callosum, which separates the PrCM and DS (Paxinos and Watson 1997, 2009). For experiment 2, all neurons encountered between the first and last recorded HD cell (±250 μm) were classified as ADN cells. The percentage of HD cells within each structure was calculated by first pooling all animals within a group (i.e., control or lesion) and then dividing the total number of HD cells recorded in a brain area by the total number of unique neurons recorded or observed in that area. For experiment 1, this same procedure was used to calculate AHV cell percentages.

Finally, we examined the animals’ movements to determine if the brain lesions impaired locomotor behavior. For each animal in experiment 2, we computed the total distance traveled, mean linear speed, and mean angular head speed during each STD1 session. In experiment 1, STD1 sessions were rarely recorded in the lesion group due to a disruption of HD cell activity (see results); therefore, for each animal in experiment 1, we selected five preliminary recording sessions (i.e., sessions in which the recorded neurons did not necessarily satisfy the HD cell criteria) and computed the total distanced traveled, mean linear speed, and mean angular head speed during each of these sessions.

Between-group and within-group comparisons were performed using a Pearson’s chi-squared test, independent-samples t-test, paired samples t-test, Watson’s two-sample test, one-way ANOVA, repeated-measures ANOVA, or two-way ANOVA. Correlations were examined using a Pearson’s r. For between-group comparisons, Levene’s test was used to examine equality of variances and the appropriate corrections were performed when variances were unequal. For Pearson’s chi-squared tests, the Yate’s correction for continuity was used when one or more cells in the contingency table contained a count ≤5. When multiple post hoc pairwise comparisons were performed, the Bonferroni correction was used. All tests were performed using statistical software (SPSS, IBM, Armonk, NY; and R, R Foundation for Statistical Computing, Vienna, Austria). All means are reported with ±SE.

Histological Processing and Analyses

When the electrode array was fully advanced, recording concluded and the animal received a lethal dose of pentobarbital sodium (150 mg/kg injected intraperitoneally). The terminal location of the electrode array was marked by passing constant anodal current through one of the electrode wires (15 μA for 20 s), and then the animal was transcardially perfused with 0.9% saline followed by 10% formalin. The brain was removed and postfixed in 10% formalin containing 2% potassium ferrocyanide to produce Prussian blue staining around the marking lesion made at the tip of the electrode array. Finally, the brain was cryoprotected in 20% sucrose for 2 days and then sectioned on a cryostat. Coronal sections (30 μm thick) containing the PrCM, DS, and ADN were mounted on glass slides and Nissl-stained with thionine. These sections were viewed under a light microscope (Leitz, Portugal) for histological analyses.

For experiment 1, sections containing the PrCM and DS were used to examine the electrode track and determine the terminal location of the electrode array in each animal, indicated by the presence of the marking lesion and associated Prussian blue staining. The path of the electrode array in each animal was then reconstructed onto a series of nine atlas plates illustrating coronal sections throughout the A/P axis of the PrCM and DS (2.2 to −0.8 mm). The locations of HD cells and AHV cells recorded along the reconstructed paths were estimated using the drive screw turn records. Next, sections containing the ADN were used to examine the neurotoxic lesions; lesioned tissue was readily differentiated from undamaged tissue based on the presence of gliosis and the absence of large, darkly stained neurons. For each animal, the extent of the ADN lesion was reconstructed by tracing the area of lesioned tissue onto a series of four atlas plates illustrating coronal sections throughout the A/P axis of the ADN (−1.3 to −2.1 mm). The percentage of the ADN lesioned in each animal was then computed by measuring these tracings using image processing software (ImageJ; NIH, Bethesda, MD) and then dividing the area of lesioned ADN by the total area of ADN. To illustrate the extent of the average ADN lesion, ImageJ was used to create Z stacks of each atlas plate composed of a tracing from each of the lesioned animals. Next, the “Z Project” function in ImageJ was used to collapse each Z stack into a single image representing the extent of the average lesion at that A/P level, with the darkness of an area corresponding to the number of animals in which that area was lesioned. Finally, representative images of the ADN of control and lesioned animals were acquired using a digital camera (Leica, Buffalo Grove, IL) connected to the microscope.

For experiment 2, sections containing the ADN were used to verify that the electrode array passed through the ADN in each animal, based on the location of the electrode track, marking lesion, and Prussian blue staining. Next, sections containing the PrCM and DS were used to examine the neurotoxic lesions. Lesioned tissue and undamaged tissue were differentiated using the criteria described above. For each animal, the lesions were reconstructed using the procedure described above; the area of lesioned tissue was traced onto a series of eight atlas plates illustrating coronal sections throughout the A/P axis of the PrCM and DS (2.2 to −1.3 mm). Finally, we generated an illustration of the extent of the average combined PrCM and DS lesion using the procedure described above. The atlas plates used in all histological analyses illustrate anatomical borders based on the atlas of Paxinos and Watson (1997, 2009).

RESULTS

HD Cell Recording Locations in the PrCM and DS of Control Animals

In control animals from experiment 1, HD cells were first recorded in the PrCM and subsequently in the DS as the electrode array was advanced ventrally. The depths of HD cells revealed an apparent bimodal distribution; one population of HD cells was confined to the most dorsal recording locations corresponding to the PrCM, and a second more widely distributed population was recorded more ventrally at depths corresponding to the DS (Fig. 2A). The border between the PrCM and DS was defined as −2.3 mm D/V, and it is possible that cells recorded on or immediately adjacent to this border were misclassified. Importantly, the recording locations shown in Fig. 2B illustrate that, for most animals, there was an absence of HD cells (and AHV cells) recorded immediately ventral to −2.3 mm D/V as the electrode array passed through the corpus callosum. This “gap” in recordings between the PrCM and DS can also be observed in the histogram in Fig. 2A, which shows a dramatic decline in the number of HD cells recorded at depths corresponding to the corpus callosum (i.e., −2.3 to −2.6 mm D/V) and a subsequent increase in the number of HD cells recorded at depths ventral to the corpus callosum corresponding to the DS.

Fig. 2.

Head direction (HD) cell and angular head velocity (AHV) cell recording locations in the medial precentral cortex (PrCM) and dorsal striatum (DS). A: histogram illustrating the distribution of the depths of all HD cells recorded in the control, lesion, and extended-DS groups. The border between the PrCM and DS was defined as −2.3 mm dorsal/ventral; all cells dorsal were classified as PrCM cells and all cells ventral were classified as DS cells. B: atlas plates representing coronal sections throughout the anterior/posterior (A/P) axis of the PrCM and DS. For each animal, the path of the electrode array was reconstructed and is represented by a vertical line. Blue lines represent control animals, red lines represent animals with anterodorsal thalamic nucleus (ADN) lesions, the orange line represents the animal with a small hippocampal (HPC) lesion, and green lines represent the extended-DS animals with electrode arrays implanted at more anterior, lateral, and/or ventral coordinates within the DS. Overlaid symbols represent the locations at which HD cells (○) and AHV cells (×) were recorded. Adjacent symbols represent cells that were recorded simultaneously at the same location along the reconstructed path. The number at top left of each plate indicates its A/P location in mm relative to bregma. Scale bar, 1 mm. M1, primary motor cortex; cc, corpus callosum.

Based on the reconstructed paths of the electrode arrays (Fig. 2B), HD cells appeared confined to the most medial portion of the DS located immediately adjacent to the lateral ventricle. The electrode array passed through this medial portion in the nine control animals; DS HD cells were recorded in 77% (7/9) of these animals at A/P levels ranging from 0.5 to −0.8 mm (Fig. 2B). When the electrode was implanted at more anterior, lateral, and/or ventral coordinates in the DS (i.e., the extended-DS animals), HD cells were recorded in only 25% (1/4) of the animals (Fig. 2B).

In the nine control animals, the electrode arrays were implanted in the lateral portion of the PrCM overlying the DS. If the electrode arrays were implanted more medially, they would have entered the lateral ventricle, not the DS, when advanced ventrally. Because of this constraint, some cortical recording locations correspond more closely to the primary motor cortex (M1) compared with the PrCM. The PrCM is a relatively ambiguous cortical area with poorly defined borders that vary across studies. This area has been alternatively defined as the medial agranular cortex (Hoover and Vertes 2007; Ostlund et al. 2009; Reep et al. 1990), anteromedial cortex (Sinnamon and Galer 1984), motor cortex (Condé et al. 1995), secondary motor cortex (Paxinos and Watson 1997, 2009; Swanson 1998; Yin 2009), vibrissa motor cortex (Brecht et al. 2004), and frontal orienting field (Erlich et al. 2011). For simplicity and to be consistent with the existing HD cell literature, we refer to all cells recorded in this cortical area as PrCM cells. PrCM HD cells were recorded in 67% (6/9) of the control animals at A/P levels ranging from 0.7 to −0.8 mm (Fig. 2B). Based on the depth at which the electrode arrays were implanted and the reconstructed paths (Fig. 2B), HD cells were recorded primarily from the deep layers of the PrCM (i.e., layers V–VI).

Measurements of a neuron’s firing characteristics (e.g., mean firing rate) and spike waveform properties (e.g., waveform duration) have been used to classify striatal cells as medium spiny neurons, fast spiking interneurons, or tonically active neurons (Berke et al. 2004; Yamin et al. 2013). For most of the HD cells we recorded, spike waveforms were isolated online (see materials and methods), and unfortunately this approach did not provide the quantitative measurements of waveform properties that are typically required for cell type classification. When we examined the waveforms qualitatively, HD cell waveforms appeared variable, with no distinguishing characteristics that were consistent across HD cells and no characteristics that differentiated HD cells from AHV cells or other cells that were not spatially tuned. Schmitzer-Torbert and Redish (2004) classified DS cells as either phasic-firing neurons or tonic-firing neurons based solely on the neurons’ activity levels during the recording session (i.e., no waveform measures were used). We did not classify HD cells using this approach because HD cells can display both phasic and tonic firing patterns depending on the movement of the animal, its orientation within the environment, and the PFD of the HD cell: when the animal is actively moving in and out of the PFD, the cell’s firing will appear phasic, but if the animal remains relatively still while oriented toward the PFD, firing will appear tonic. Therefore, classifying a HD cell as phasic vs. tonic is highly dependent on the incidental movement and orientation of the animal during the recording session and the neuron’s PFD.

In our accompanying report (Mehlman et al. 2019), we describe direct anatomical projections from the medial entorhinal cortex to the PrCM and DS. Given that the medial entorhinal cortex contains conjunctive HD × grid cells (Sargolini et al. 2006), it is possible that the PrCM and/or DS also contain HD × grid cells. We generated firing rate heat maps and spatial autocorrelation plots for each extralimbic HD cell we recorded and did not observe any grid cell-like tuning patterns, indicating that the PrCM and DS do not contain conjunctive HD × grid cells (data not shown).

Comparison of HD Cell Characteristics in the PrCM, DS, and ADN of Control Animals

Tuning properties.

Extralimbic HD cells recorded in the PrCM and DS of control animals from experiment 1 displayed tuning properties largely identical to those of limbic HD cells recorded in the ADN of control animals from experiment 2 (Fig. 3A). In all three structures, HD cell PFDs were uniformly distributed across 360° (data not shown). One-way ANOVAs revealed no significant differences between PrCM, DS, and ADN HD cells in terms of their Rayleigh r [F(2,87) = 0.608, P = 0.547], information content [F(2,87) = 0.858, P = 0.428], directional firing range [F(2,87) = 1.144, P = 0.323], or ATI [F(2,53) = 1.340, P = 0.271]. However, the peak firing rates of HD cells in the PrCM, DS, and ADN differed significantly [one-way ANOVA, F(2,87) = 13.214, P < 0.001]; post hoc pairwise comparisons revealed that the peak firing rates of PrCM HD cells were significantly higher than those of HD cells in the DS (P < 0.01) and ADN (P < 0.001) (Table 1; Fig. 3B).

Fig. 3.

Tuning properties of head direction (HD) cells recorded in the medial precentral cortex (PrCM), dorsal striatum (DS), and anterodorsal thalamic nucleus (ADN) of control animals. A: tuning curves illustrating the activity of a representative PrCM, DS, and ADN HD cell. B: grouped scatterplots illustrating the Rayleigh r (left), directional firing range (middle), and peak firing rate (right) of all HD cells recorded in the control groups. Group means are indicated by the overlaid horizontal lines. **P < 0.01, one-way ANOVA, post hoc pairwise comparison (PrCM and DS). ***P < 0.001, one-way ANOVA, post hoc pairwise comparison (PrCM and ADN).

Table 1.

Tuning properties of HD cells recorded in the PrCM, DS, and ADN of control animals

	PrCM HD Cells		DS HD Cells		ADN HD Cells
Property	Mean ± SE	Range	Mean ± SE	Range	Mean ± SE	Range
Rayleigh r	0.715 ± 0.035	0.470–0.898	0.705 ± 0.036	0.403–0.936	0.675 ± 0.019	0.403–0.923
Information content	0.940 ± 0.117	0.433–1.937	1.174 ± 0.123	0.302–2.873	1.107 ± 0.081	0.306–2.599
Directional firing range, °	116.8 ± 4.8	91.7–160.5	112.2 ± 5.5	57.5–142.6	123.4 ± 5.1	73.1–232.2
Peak firing rate, Hz	65.1 ± 7.1*†	20.6–118.8	37.3 ± 4.3*	8.7–92.1	32.1 ± 3.0†	6.9–98.1
Anticipatory time interval, ms	61.9 ± 11.4	23.8–94.0	51.5 ± 6.6	10.3–88.8	40.0 ± 6.5	−37.4–144.2

Values are means ± SE and ranges of head direction (HD) cell tuning properties recorded in the medial precentral cortex (PrCM; n = 16 cells), dorsal striatum (DS; n = 25 cells), and anterodorsal thalamic nucleus (ADN; n = 49 cells). A subset of HD cells (PrCM, n = 7; DS, n = 12; ADN, n = 37) was included in the anticipatory time interval analysis (see materials and methods).

^*P < 0.01, one-way ANOVA, post hoc pairwise comparison (PrCM and DS).

^†P < 0.001, one-way ANOVA, post hoc pairwise comparison (PrCM and ADN).

Landmark rotation.

During landmark rotation manipulations, PrCM and DS HD cells responded similarly to ADN HD cells (Table 2). The PFDs of HD cells in all three structures shifted in register with rotations of the cue card, as illustrated by the example DS HD cell in Fig. 4A. Across the three brain areas, there were no significant differences in PFD shift values from STD1 to ROT, with HD cells in all areas tending to under rotate (i.e., the PFDs did not shift a full 90° following the 90° cue card rotation) [PrCM, mean shift = 71.9 ± 11.3°, Rayleigh r = 0.832, P < 0.001; DS, mean shift = 71.7 ± 7.3°, Rayleigh r = 0.849, P < 0.001; ADN, mean shift = 64.2 ± 7.1°, Rayleigh r = 0.786, P < 0.001; one-way ANOVA, F(2,62) = 0.283, P = 0.755] (Fig. 4B, left). This tendency to under rotate has been reported previously in limbic HD cells and may be due to an uncontrolled polarizing cue (e.g., an odor source) that remained stable across the standard and rotation sessions (Taube et al. 1990b; Taube 1995). Additionally, the PFDs of HD cells in the PrCM, DS, and ADN remained similarly stable across the standard sessions; PFD shift values from STD1 to STD2 were near 0° and did not differ significantly between brain areas [PrCM, mean shift = 2.5 ± 4.3°CCW, Rayleigh r = 0.972, P < 0.001; DS, mean shift = 9.5 ± 7.6°CW, Rayleigh r = 0.845, P < 0.001; ADN, mean shift = 0.5 ± 5.8°CCW, Rayleigh r = 0.880, P < 0.001; one-way ANOVA, F(2,62) = 1.548, P = 0.221] (Fig. 4B, right). When multiple HD cells in the PrCM, DS, or ADN were recorded simultaneously, their PFDs shifted in register (data not shown).

Table 2.

Responses to environmental manipulations displayed by HD cells recorded in the PrCM, DS, and ADN of control animals

	PrCM HD Cells		DS HD Cells		ADN HD Cells
Manipulation	Mean ± SE	Range	Mean ± SE	Range	Mean ± SE	Range
Landmark rotation
STD1 – ROT	71.9 ± 11.3	−20–102	71.7 ± 7.3	12–168	64.2 ± 7.1	0–162
STD1 – STD2	2.5 ± 4.3	−24–18	−9.5 ± 7.6	−126–36	0.5 ± 5.8	−132–78
Dark session
Directional firing range, °
Light	114.6 ± 5.2	91.7–160.5	106.1 ± 4.0	57.5–142.6	124.7 ± 5.4	73.1–231.1
Dark	162.7 ± 7.5	111.3–215.3	144.5 ± 9.4	99.2–292.3	146.4 ± 6.1	78.9–252.2
PFD stability, °
Light	28.3 ± 4.8	8.0–65.9	33.3 ± 5.4	11.9–100.2	27.5 ± 2.5	7.4–61.4
Dark	41.0 ± 5.5	17.6–90.5	53.2 ± 5.4	17.8–99.6	44.4 ± 3.2	11.5–87.8
Rayleigh r
Light	0.714 ± 0.038	0.470–0.898	0.728 ± 0.037	0.403–0.936	0.686 ± 0.020	0.433–0.923
Dark	0.594 ± 0.060	0.160–0.835	0.581 ± 0.054	0.222–0.923	0.606 ± 0.023	0.229–0.847
Dual-chamber apparatus
CYL1 – REC	25.0 ± 17.0	−18–63	24.7 ± 20.6	−36–60
CYL1 – CYL2	4.5 ± 2.9	0–12	−1.2 ± 2.2	−6–6

Values are means ± SE and ranges of head direction (HD) cell responses to manipulations recorded in the medial precentral cortex (PrCM), dorsal striatum (DS), and anterodorsal thalamic nucleus (ADN) during landmark rotation (PrCM, n = 11 cells; DS, n = 22 cells; ADN, n = 32 cells), dark session (PrCM, n = 14 cells; DS, n = 21 cells; ADN, n = 37 cells), and dual-chamber apparatus (PrCM, n = 4 cells; DS, n = 5 cells; ADN, n = 0 cells) manipulations. Mean values for the landmark rotation and dual-chamber apparatus manipulations indicate the mean preferred firing direction shift values (in degrees) for each brain area. STD1, first standard session; ROT, rotation session; STD2, second standard session; CYL1, first cylinder session; REC, novel rectangle session; CYL2, second cylinder session.

Fig. 4.

Responses to environmental manipulations displayed by head direction (HD) cells recorded in the medial precentral cortex (PrCM), dorsal striatum (DS), and anterodorsal thalamic nucleus (ADN) of control animals. A: example HD cell recorded in the DS during the first standard (STD1), rotation (ROT), and second standard (STD2) sessions. B: circular histograms illustrating the angular shift of each HD cell’s preferred firing direction (PFD) from STD1 to ROT (left) and from STD1 to STD2 (right). Each circle represents the PFD shift of a single HD cell recorded in the PrCM, DS, or ADN. For each brain area, the circular mean of all shift values is indicated and represented by the arrow in the center of the plot. The length of each arrow indicates the Rayleigh r of the shift values (ranging from 0 to 1, with the circle’s radius representing 1). C: example HD cell recorded in the DS during the STD2, dark (DRK), and third standard (STD3) sessions. D: same recordings as in C, illustrating HD cell activity over time. Each circle represents the HD of the animal when the cell’s firing rate was ≥75% of its peak firing rate. E: mean directional firing range (left), PFD stability (middle), and Rayleigh r (right) of all HD cells recorded in the PrCM, DS, and ADN under light and dark conditions for 8 min. Error bars are SE. ***P < 0.001, repeated-measures ANOVA, main effect of condition (light and dark). F: example HD cell recorded in the PrCM during the first session in the familiar cylinder (CYL1), the session in the novel rectangle (REC), and the second session in the cylinder (CYL2). G: circular histograms (as in B) illustrating the response of each PrCM and DS HD cell recorded in the dual-chamber apparatus. PFD shifts from CYL1 to REC (left) and from CYL1 to CYL2 (right) are shown.

Dark session.

Extralimbic HD cells responded similar to limbic HD cells when the overhead lights were turned off. Compared with light conditions (i.e., STD1), HD cells in the PrCM, DS, and ADN recorded during DRK displayed a significant 1) increase in directional firing range, 2) decrease in PFD stability, and 3) decrease in Rayleigh r (Table 2; Fig. 4E). For directional firing range, a repeated-measures ANOVA revealed a significant main effect of condition [F(1,69) = 61.418, P < 0.001] and no significant main effect of brain area [F(2,69) = 1.296, P = 0.280], but there was a significant interaction [F(2,69) = 3.211, P = 0.046]. Exploring this interaction, post hoc univariate comparisons revealed a significant directional firing range difference between areas during STD1 [F(2,69) = 3.218, P = 0.046], but not during DRK [F(2,69) = 1.172, P = 0.316]. Importantly, this difference in directional firing range under light conditions was only significant when the subset of HD cells recorded under both light and dark conditions was examined; some HD cells were recorded during STD1 but not DRK (due to the loss of waveform isolation before DRK), and when these cells were included in an analysis of HD cell directional firing ranges under light conditions, there was no significant difference between brain areas (Table 1; Fig. 3B). In terms of PFD stability, a repeated-measures ANOVA revealed a significant main effect of condition [F(1,69) = 41.991, P < 0.001] and no significant main effect of brain area [F(2,69) = 1.444, P = 0.243] or interaction [F(2,69) = 0.552, P = 0.579]. Similarly, in terms of Rayleigh r, a repeated-measures ANOVA revealed a significant main effect of condition [F(1,69) = 39.310, P < 0.001] and no significant main effect of brain area [F(2,69) = 0.028, P = 0.972] or interaction [F(2,69) = 1.499, P = 0.231]. The example DS HD cell in Fig. 4, C and D, illustrates the increased directional firing range and decreased PFD stability, respectively, exhibited during DRK. Despite this loss of precision and stability, HD cells in the PrCM, DS, and ADN maintained much of their directional tuning in the dark, presumably via the availability of self-motion cues. During DRK, 86% (12/14), 71% (15/21), and 95% (35/37) of HD cells in the PrCM, DS, and ADN, respectively, continued to satisfy the HD cell criteria (i.e., Rayleigh r ≥ 0.4). As described above, the animals may have had access to an uncontrolled polarizing cue, which could account for the tendency of HD cells to under rotate during ROT. If this cue was an odor source, it would have been available to the animals during DRK and may have contributed to the maintenance of directional tuning. Nevertheless, the tuning properties of HD cells in all three brain areas were significantly affected under dark conditions.

Dual-chamber apparatus.

When the animals were placed in the cylinder of the DCA, HD cells displayed PFDs similar to those observed in the cylinder used for the standard sessions (data not shown), suggesting that the DCA cylinder was indeed treated as a familiar environment. When the animals walked between the familiar cylinder and the novel rectangle, PrCM and DS HD cells responded similarly to limbic HD cells recorded in previous studies (Clark et al. 2010; Taube and Burton 1995; Yoder et al. 2015). As illustrated by the example PrCM HD cell in Fig. 4F, PFDs remained relatively stable in the DCA. The magnitude and direction of the PFD shifts from CYL1 to REC did not differ significantly between brain areas (PrCM, mean shift = 25.0 ± 17.0°CCW, Rayleigh r = 0.873, P < 0.05; DS, mean shift = 24.7 ± 20.6°CCW, Rayleigh r = 0.755, P < 0.05; Watson’s two-sample test, U² = 0.070, P > 0.1) (Table 2; Fig. 4G, left). In the absence of familiar landmarks, HD cells are thought to utilize self-motion information to continuously estimate the animal’s current HD relative to the animal’s HD when it initially entered the novel environment (Taube and Burton 1995). This estimation process, termed path integration, is imperfect, and the errors that accrue over time likely resulted in the PFD shifts we observed in the rectangle. When the animals subsequently returned to the cylinder (i.e., CYL2), PFDs returned to their initial direction. PFD shift values from CYL1 to CYL2 were near 0° and did not differ significantly between brain areas (PrCM, mean shift = 4.5 ± 2.9°CCW, Rayleigh r = 0.996, P < 0.01; DS, mean shift = 1.2 ± 2.2°CW, Rayleigh r = 0.997, P < 0.01; Watson’s two-sample test, U² = 0.048, P > 0.1) (Table 2; Fig. 4G, right). PFDs likely “reset” to their initial direction during CYL2 due to the presence of the familiar cue card. This visual landmark presumably allowed HD cells to correct for the errors that accumulated while the animal was path integrating within the novel environment.

In sum, PrCM and DS HD cell tuning properties and responses to environmental manipulations largely reflected those of ADN HD cells recorded in the current study and previously (Taube 1995; Taube and Burton 1995). These extralimbic HD cells 1) are controlled by a visual landmark, 2) maintain directional tuning, albeit slightly degraded, in the absence of visual information, and 3) largely preserve their representation of direction as the animal walks from a familiar to a novel environment. The striking similarities between the extralimbic and limbic HD signals suggest that HD cell activity in the PrCM and DS is driven by limbic HD circuit output; this hypothesis is directly tested in experiment 1.

AHV Cells in the PrCM and DS of Control Animals

We recorded AHV cells intermixed with HD cells in the PrCM and DS of control animals (Fig. 2B). These neurons fired as a function of the animal’s AHV in the horizontal plane and displayed a range of different responses to CW vs. CCW rotations. Some AHV cells [PrCM, 47.1% (8/17); DS, 5.9% (2/34)] were insensitive to the direction of rotation – these symmetric AHV cells responded similarly to CW and CCW head turns, either increasing (Fig. 5A) or decreasing (Fig. 5D) their firing rate during rotation in either direction. In contrast, other AHV cells [PrCM, 52.9% (9/17); DS, 94.1% (32/34)] exhibited asymmetric responses in which their firing rate either increased or decreased in response to head turns in one direction, but were not significantly responsive to rotation in the other direction. The preferred turn direction of asymmetric AHV cells was independent of the hemisphere in which the neurons were recorded, indicating that these AHV cells do not encode exclusively ipsiversive or contraversive head turns [PrCM, Pearson’s chi-squared test, χ²(1) = 0, P = 1; DS, Pearson’s chi-squared test, χ²(1) = 0.086, P = 0.770]. The strength of AHV cell tuning in the PrCM and DS (i.e., the mean unsigned slope and correlation values for the tuning curves) is discussed below in Effects of ADN Lesions and is illustrated in Fig. 5H. In both brain areas, a subset of symmetric and asymmetric AHV cell tuning curves displayed greater modulation (i.e., steeper slopes) across low AHVs (~0–60°/s) compared with high AHVs (Fig. 5A); this “scallop” property has been observed previously in the tuning curves of AHV cells recorded in the dorsal tegmental nucleus and lateral mammillary nucleus (Bassett and Taube 2001). Although the origin of the scallop is unclear, it is unlikely an artifact of the recording or analysis procedures because some AHV cell tuning curves contained a scallop, whereas others did not contain one (Fig. 5B). In both brain areas, we observed that AHV cell responses were maintained during behavioral epochs in which the animal’s head rotated in the horizontal plane while the trunk and limbs remained stationary, suggesting that these neurons encode head movements independently of full body movements.

Fig. 5.

Tuning properties of angular head velocity (AHV) cells recorded in the medial precentral cortex (PrCM) and dorsal striatum (DS) of control and lesioned animals. A: example symmetric AHV cell recorded in the PrCM of a control animal. B: example symmetric AHV cell recorded in the PrCM of a lesioned animal. C: example symmetric AHV cell recorded in the DS of a lesioned animal. D: example symmetric AHV cell recorded in the DS of an extended-DS animal. In contrast to the AHV cells in A–C, which displayed increased firing rates during head turns, this AHV cell displayed a firing rate that decreased during head turns. E: example asymmetric AHV cell recorded in the PrCM of a lesioned animal. F: example asymmetric AHV cell recorded in the DS of a lesioned animal. The AHV cell in E responded oppositely to clockwise (CW) vs. counterclockwise (CCW) rotation, displaying a firing rate that decreased during CW head turns and increased during CCW head turns. In contrast, the AHV cell in F was responsive to head turns in one direction (CW), but unresponsive when the head turned in the other direction. In A–F, positive velocities indicate rotation in the CCW direction and negative velocities indicate rotation in the CW direction. G: scatterplot illustrating the CW and CCW slope of all AHV cells recorded in the PrCM and DS of control, extended-DS, and lesioned animals. Symmetric AHV cells appear near the diagonal in quadrants II and IV (indicated by the shaded areas), whereas asymmetric AHV cells appear in quadrants I and III, and away from the diagonal in quadrants II and IV (i.e., outside of the shaded areas). The labeled arrows point to the AHV cells illustrated in A–F. H: mean unsigned slope values (left) and correlation values (right) for the tuning curves of all PrCM and DS AHV cells recorded in control and lesioned animals. Error bars are SE. **P < 0.01, two-way ANOVA, main effect of brain area (PrCM and DS).

The scatterplot in Fig. 5G displays the CW and CCW slope of each AHV cell and illustrates the variety of tuning properties observed. In this plot, symmetric AHV cells that increased their firing rate during head turns in either direction, thus displaying a negative CW slope and a positive CCW slope, appear near the diagonal in quadrant IV (indicated by the shaded area). Conversely, symmetric AHV cells with firing rates that decreased during head turns in either direction, thus displaying a positive CW slope and negative CCW slope, appear near the diagonal in quadrant II (indicated by the shaded area). Asymmetric AHV cells appear in quadrants I and III, and away from the diagonal in quadrants II and IV (i.e., outside of the shaded areas). Neurons modulated by the animal’s head movements have been observed previously in the DS (Gardiner and Kitai 1992; Kim et al. 2014; Trytek et al. 1996), but to our knowledge, the current study is the first to report AHV cells in the PrCM. Compared with HD cells, AHV cells appeared more widely distributed across different portions of the DS; AHV cells were recorded in all (4/4) of the extended-DS animals, whereas HD cells were recorded in only 25% (1/4) of these animals.

Interestingly, a subset of the HD cells recorded in control animals also met the criteria for AHV cells (i.e., conjunctive HD × AHV cells). In the PrCM, 37.5% (6/16) of HD cells were classified as HD × AHV cells. In the DS, 24% (6/25) of HD cells were classified as HD × AHV cells.

Effects of ADN Lesions

ADN lesions.

To determine the origin of the extralimbic HD signals in the PrCM and DS, we recorded PrCM and DS neurons in animals that had received bilateral ADN lesions, a manipulation known to eliminate the HD signal in downstream nodes of the limbic system (Goodridge and Taube 1997; Winter et al. 2015a). If the PrCM and DS receive their HD signals from limbic HD circuit output, HD cell activity in these extralimbic areas should be eliminated in animals with ADN lesions.

Injections of NMDA produced focal lesions surrounding the injection sites, evidenced by the presence of gliosis and the absence of large, darkly stained neurons (Fig. 6A). The extent of the average ADN lesion is illustrated in Fig. 6B. When the total area of the ADN was considered (i.e., both hemispheres combined), 69.4% was lesioned on average, ranging from 11.6% to 97.4%. In most animals, lesions were confined to the ADN, with minor spreading into the adjacent thalamic nuclei (i.e., the anteroventral thalamic nucleus, anteromedial thalamic nucleus, laterodorsal thalamic nucleus, and mediodorsal thalamic nucleus). Additionally, some animals displayed minor damage to portions of the hippocampus overlying the ADN, presumably due to NMDA spreading up the cannula track. The recording locations in the lesioned animals were comparable to the recording locations in the control animals (Fig. 2B).

Fig. 6.

Anterodorsal thalamic nucleus (ADN) lesions. A: representative images illustrating the right ADN of a control animal (left; the ADN is outlined by the dashed line) and a lesioned animal (right; the lesioned area is outlined by the dashed line). The number at top right of each image indicates its anterior/posterior (A/P) location in mm relative to bregma. Scale bars, 500 μm. B: atlas plates representing coronal sections throughout the A/P axis of the ADN. Red shading illustrates the extent of the average ADN lesion; the darkness of an area corresponds to the number of animals in which that area was lesioned (i.e., darker areas are lesioned across more animals compared with lighter areas). The number at top left of each plate indicates its A/P location in mm relative to bregma. Scale bar, 1 mm. AVN, anteroventral thalamic nucleus; AMN, anteromedial thalamic nucleus; LDN, laterodorsal thalamic nucleus; MDN, mediodorsal thalamic nucleus.

Effects of lesions on HD cell and AHV cell activity in the PrCM and DS.

We first examined the percentage of HD cells and AHV cells recorded in each group. In all analyses comparing the control and lesion groups, we excluded the control animal implanted with a stereotrode array to make the two groups more equivalent, since no lesioned animals were implanted with a stereotrode array (including this animal in the analyses, however, did not change the overall results; data not shown).

In the PrCM, the HD cell percentage decreased significantly from 16.3% (13/80) in the control group to 0% (0/40) in the lesion group [Pearson’s chi-squared test, χ²(1) = 5.705, P = 0.017]; PrCM HD cells were recorded in 62.5% (5/8) of the control animals and in 0% (0/8) of the lesioned animals. In the DS, the percentage of HD cells decreased from 8.2% (16/196) in the control group to 2.9% (5/173) in the lesion group; this effect closely approached significance [Pearson’s chi-squared test, χ²(1) = 3.829, P = 0.050] (Fig. 7A, left). DS HD cells were recorded in 75% (6/8) of the control animals and in 37.5% (3/8) of the lesioned animals. Importantly, PrCM and DS HD cells were recorded in the animal that received a small hippocampal lesion, and the quantity and quality of these HD cells were comparable to that of control animals (data not shown). Therefore, the effects of the ADN lesions are unlikely to be attributed to the minor hippocampal damage caused by the surgical procedure (Fig. 6B).

Fig. 7.

Effects of anterodorsal thalamic nucleus (ADN) lesions. A: percentage of head direction (HD) cells (left) and angular head velocity (AHV) cells (right) recorded in the medial precentral cortex (PrCM) and dorsal striatum (DS) of control and lesioned animals. *P < 0.05; +P = 0.05, Pearson’s chi-squared test. B: scatterplots illustrating the relationship between lesion size and the number of HD cells recorded in the DS of lesioned animals. The number of HD cells is plotted relative to the size of the ADN lesion in the hemisphere ipsilateral (top) or contralateral (bottom) to the recording electrode. The number at bottom left of each plot indicates the correlation between lesion size and number of HD cells recorded. ***P < 0.001, Pearson’s r. C: grouped scatterplots illustrating the directional firing range (top) and preferred firing direction (PFD) stability (bottom) of all HD cells recorded in the DS of control and lesioned animals. Group means are indicated by the overlaid horizontal lines. ***P < 0.001, independent-samples t-test. D: circular histograms (as in Fig. 4) illustrating the angular shift of each DS HD cell’s PFD from the first standard session (STD1) to the rotation session (ROT; left) and from STD1 to the second standard session (STD2; right). Arrows represent the circular mean and Rayleigh r of all shift values within each group.

Interestingly, AHV cell activity in both structures was relatively unaffected by the ADN lesions (Fig. 5, B, C, E, and F). In the PrCM, equivalent percentages of AHV cells were recorded in the control and lesion groups [control, 18.8% (15/80); lesion, 22.5% (9/40); Pearson’s chi-squared test, χ²(1) = 0.059, P = 0.809]. The same was true for the DS [control, 6.6% (13/196); lesion, 10.4% (18/173); Pearson’s chi-squared test, χ²(1) = 1.244, P = 0.265] (Fig. 7A, right). PrCM AHV cells were recorded in 75% (6/8) of the control animals and in 75% (6/8) of the lesioned animals. DS AHV cells were recorded in 62.5% (5/8) of the control animals and in 62.5% (5/8) of the lesioned animals. Furthermore, the strength of AHV cell tuning in lesioned animals was similar to that of control animals. Tuning strength was examined by computing the mean unsigned slope and correlation values for the AHV cell tuning curves; only segments of the tuning curves that satisfied the AHV cell criteria contributed to these calculations (i.e., for symmetric cells, the unsigned slope and correlation values associated with the CW and CCW segments were averaged together; for asymmetric cells, the unsigned slope and correlation values associated with the preferred turn direction contributed to these calculations, whereas values associated with the unresponsive turn direction were excluded). Mean unsigned slope values did not differ significantly between the control and lesion groups. However, these slope values displayed a significant difference in terms of PrCM vs. DS AHV cells: PrCM AHV cells were more strongly tuned compared with DS AHV cells [two-way ANOVA, main effect of group, F(1,51) = 0.662, P = 0.419; main effect of brain area, F(1,51) = 7.715, P = 0.008; interaction, F(1,51) = 2.716, P = 0.105] (Fig. 5H, left). Mean unsigned correlation values did not display significant differences between groups or brain areas [two-way ANOVA, main effect of group, F(1,51) = 2.253, P = 0.140; main effect of brain area, F(1,51) = 0.237, P = 0.628; interaction, F(1,51) = 1.785, P = 0.187] (Fig. 5H, right). This overall pattern of results reveals a dissociation between extralimbic HD and AHV signals; although ADN lesions eliminated PrCM HD cell activity and disrupted DS HD cell activity, AHV cell activity in both structures remained largely intact.

Given that HD cells were recorded in the DS of a subset of lesioned animals, we next examined the relationship between lesion size and the number of DS HD cells recorded. Figure 7B illustrates the number of DS HD cells recorded in each animal as a function of ADN lesion size in the hemisphere ipsilateral or contralateral to the recording electrode. The number of DS HD cells was significantly and negatively correlated with the size of the ADN lesion in the ipsilateral hemisphere [Pearson’s r = −0.970, t(6) = −9.698, P < 0.001], and no HD cells were recorded in animals with complete ipsilateral lesions (i.e., greater than ~80% lesioned) (Fig. 7B, top). In contrast, there was no significant correlation between the number of DS HD cells and the size of the ADN lesion in the contralateral hemisphere [Pearson’s r = −0.339, t(6) = −0.883, P = 0.206], and HD cells were recorded in two animals with complete contralateral lesions (i.e., greater than ~80% lesioned) (Fig. 7B, bottom). It is important to note that these correlation values are driven by the relatively small number of lesioned animals in which HD cells were recorded (n = 3), and there is a floor effect in that the remaining animals pool at the lower limit of the y-axis. Therefore, the relationship between lesion size and the number of DS HD cells recorded may not be strictly linear. Nevertheless, these analyses suggest that the HD signal is likely conveyed from the limbic system to the DS within, not across, hemispheres.

Next, we examined the tuning properties of the five DS HD cells recorded in animals with incomplete ADN lesions (i.e., ipsilateral lesion less than ~80%). Compared with control animals, DS HD cells in lesioned animals displayed a significantly larger directional firing range [independent-samples t-test, t(19) = −5.234, P < 0.001] (Fig. 7C, top). Importantly, DS HD cells in control and lesioned animals did not differ significantly in terms of PFD stability under light conditions [independent-samples t-test, t(19) = 0.458, P = 0.652] (Fig. 7C, bottom), indicating that PFD instability over time is unlikely to account for the increased directional firing range displayed by DS HD cells in lesioned animals. Rather, these HD cells were simply tuned to a wider range of directions. Independent-samples t-tests revealed no significant differences between control and lesioned animals in terms of Rayleigh r [t(9.783) = −0.186, P = 0.856], information content [t(19) = −0.091, P = 0.929], peak firing rate [t(4.507) = −0.459, P = 0.667], or ATI [t(13) = −1.771, P = 0.100] (Table 3). Thus, although DS HD cells in lesioned animals displayed a significantly larger directional firing range, these neurons did not display significantly degraded directional tuning in terms of Rayleigh r and information content. Of the five HD cells recorded in the lesioned animals, two (40%) were classified as HD × AHV cells.

Table 3.

Tuning properties of HD cells recorded in the DS of control animals and animals with ADN lesions

	DS HD Cells
	Control		ADN Lesion
Property	Mean ± SE	Range	Mean ± SE	Range
Rayleigh r	0.668 ± 0.042	0.403–0.905	0.681 ± 0.052	0.548–0.817
Information content	1.028 ± 0.116	0.457–1.877	1.052 ± 0.281	0.568–2.053
Directional firing range, °	109.2 ± 4.9*	57.5–142.6	166.4 ± 12.0*	121.6–189.9
Peak firing rate, Hz	44.1 ± 5.5	13.6–92.1	54.6 ± 22.2	12.0–135.3
Anticipatory time interval, ms	51.5 ± 6.6	10.3–88.8	76.7 ± 10.2	57.6–92.5

Values are means ± SE and ranges of head direction (HD) cell tuning properties recorded in the dorsal striatum (DS) of control animals (n = 16 cells) and animals with anterodorsal thalamic nucleus (ADN) lesions (n = 5 cells). A subset of HD cells from control (n = 12) and lesioned animals (n = 3) was included in the anticipatory time interval analysis (see materials and methods).

^*P < 0.001, independent-samples t-test.

Finally, we examined how the DS HD cells recorded in lesioned animals responded to environmental manipulations. The PFDs of HD cells in lesioned animals shifted in register with the cue card from STD1 to ROT, but tended to under rotate, similar to HD cells in control animals (control, mean shift = 63.5 ± 10.4°, Rayleigh r = 0.806, P < 0.001; lesion, mean shift = 64.8 ± 20.2°, Rayleigh r = 0.783, P < 0.05; Watson’s two-sample test, U² = 0.110, P > 0.1) (Fig. 7D, left). Additionally, the PFDs of DS HD cells in control and lesioned animals remained similarly stable across standard sessions; in both groups, PFD shift values from STD1 to STD2 were near 0° (control, mean shift = 15.1 ± 10.9°CW, Rayleigh r = 0.788, P < 0.001; lesion, mean shift = 2.0 ± 8.4°CW, Rayleigh r = 0.958, P < 0.01; Watson’s two-sample test, U² = 0.177, P > 0.05) (Fig. 7D, right). Two DS HD cells in lesioned animals were recorded in the DCA; these HD cells responded similarly to HD cells in control animals, displaying only small PFD shifts from CYL1 to REC and a relatively stable PFD from CYL1 to CYL2 (data not shown). Finally, DS HD cells recorded in lesioned animals displayed significantly less PFD stability under dark conditions compared with light conditions, similar to HD cells in control animals [dark = 66.4 ± 13.1°; light = 31.5 ± 9.2°; paired samples t-test, t(4) = 5.692, P = 0.005], and the amount of PFD stability measured during 16 min in the dark was nearly identical for DS HD cells in control and lesioned animals [control = 66.1 ± 6.8°; lesion = 65.7 ± 10.3°; independent-samples t-test, t(18) = 0.030, P = 0.977]. Thus, although incomplete ADN lesions altered the tuning curves of DS HD cells, these lesions did not affect their responses to environmental manipulations.

In sum, no PrCM HD cell activity was observed in animals with ADN lesions. In the DS, HD cells were recorded only in animals with incomplete ADN lesions; these DS HD cells displayed a notable disruption, firing over a wider range of directions compared with control animals. This pattern of results suggests that the extralimbic HD signals in the PrCM and DS are conveyed from the limbic system in a manner dependent on the ADN. Extralimbic AHV cells, however, were largely unaffected by the ADN lesions; this dissociation suggests that the HD and AHV signals are transmitted to the PrCM and DS via separate pathways that are differentially dependent on the ADN. Alternatively, AHV cell activity could be generated locally within the PrCM and DS independently of the ADN and limbic system. These hypotheses are further addressed in discussion.

Effects of lesions on locomotor behavior.

The ADN lesions did not significantly affect locomotor behavior. During preliminary recording sessions, control and lesioned animals traveled comparable distances [control, 40.0 ± 2.4 m; lesion, 44.7 ± 2.1 m; independent-samples t-test, t(78) = −1.507, P = 0.136] while moving at similar mean linear speeds [control, 8.3 ± 0.5 cm/sec; lesion, 9.3 ± 0.4 cm/s; independent-samples t-test, t(78) = −1.507, P = 0.136]. Additionally, both groups displayed similar mean angular head speeds [control, 30.0 ± 1.5 °/s; lesion, 30.0 ± 1.0 °/s; independent-samples t-test, t(66.797) = −0.017, P = 0.986]. Therefore, the differences in HD cell activity observed between the control and lesion groups cannot be attributed to differences in locomotor behavior.

Effects of Combined PrCM and DS Lesions

Combined PrCM and DS lesions.

Given their involvement in motor control, the PrCM and DS may convey critical self-motion information to the limbic HD circuit in the form of motor efference signals. Additionally, the AHV signals in these extralimbic areas may provide self-motion information that influences limbic HD cells. To test these hypotheses, we examined the effects of combined PrCM and DS lesions on HD cell activity in the ADN. Figure 8 illustrates the extent of the average PrCM and DS lesion. The injections of NMDA produced relatively large areas of lesioned tissue displaying gliosis and lacking large, darkly stained neurons. Importantly, these lesions were largely confined to the portions of the PrCM and DS in which HD cells and AHV cells were recorded in experiment 1. Across animals, lesions were consistently confined to the lateral portion of the PrCM and the adjacent M1, and the most medial portion of the DS located immediately adjacent to the lateral ventricle. These locations are indicated by the shaded areas in Fig. 8. Examination of electrode tracks and marking lesions confirmed that the electrode array in each animal passed through the ADN.

Fig. 8.

Combined medial precentral cortex (PrCM) and dorsal striatum (DS) lesions. Atlas plates representing coronal sections throughout the anterior/posterior (A/P) axis of the PrCM and DS. As in Fig. 6, red shading illustrates the extent of the average PrCM and DS lesion; darker areas are lesioned more consistently across animals compared with lighter areas. The number at top left of each plate indicates its A/P location in mm relative to bregma. Scale bar, 1 mm. Note that the lesions were largely confined to the portions of the PrCM and DS in which HD cells and AHV cells were recorded in experiment 1. M1, primary motor cortex; S1, primary somatosensory cortex.

Effects of lesions on HD cell activity in the ADN.

The percentage of HD cells recorded in the ADN did not differ significantly between the control group and the group of animals with combined PrCM and DS lesions [control, 54.4% (49/90); lesion, 48.7% (37/76); Pearson’s chi-squared test, χ²(1) = 0.548, P = 0.459] (Fig. 9A). Moreover, HD cells were recorded in all (7/7) of the control animals and in all (10/10) of the lesioned animals. Compared with those in control animals, HD cells in lesioned animals displayed a significantly higher peak firing rate [independent-samples t-test, t(84) = 2.485, P = 0.015]. Independent-samples t-tests revealed no significant differences between control and lesioned animals in terms of all other HD cell tuning properties [Rayleigh r, t(84)= 0.391, P = 0.696; information content, t(84) = 0.187, P = 0.852; directional firing range, t(84) = −0.514, P = 0.608; ATI, t(58) = −0.004, P = 0.996] (Table 4). It is unclear why the peak firing rate of ADN HD cells was increased in animals with combined PrCM and DS lesions. Given that the majority of projection neurons in the DS are GABAergic (Yin and Knowlton 2006), damage to this structure may produce disinhibition that in turn increases activity levels in the thalamus.

Fig. 9.

Effects of combined medial precentral cortex (PrCM) and dorsal striatum (DS) lesions. A: percentage of head direction (HD) cells recorded in the anterodorsal thalamic nucleus (ADN) of control and lesioned animals. B: circular histograms (as in Fig. 4) illustrating the angular shift of each ADN HD cell’s preferred firing direction (PFD) from the first standard session (STD1) to the rotation session (ROT; left) and from STD1 to the second standard session (STD2; right). Arrows represent the circular mean and Rayleigh r of all shift values within each group. C: mean PFD stability of all ADN HD cells recorded in control and lesioned animals under dark conditions for 8 min. D: average distance traveled (left), mean linear speed (middle), and mean angular head speed (right) during STD1 for the control and lesion groups. Error bars are SE.

Table 4.

Tuning properties of HD cells recorded in the ADN of control animals and animals with combined PrCM and DS lesions

	ADN HD Cells
	Control		PrCM and DS lesion
Property	Mean ± SE	Range	Mean ± SE	Range
Rayleigh r	0.675 ± 0.019	0.403–0.923	0.686 ± 0.020	0.485–0.937
Information content	1.107 ± 0.081	0.306–2.599	1.129 ± 0.081	0.420–2.447
Directional firing range, °	123.4 ± 5.1	73.1–232.2	119.2 ± 6.6	63.7–251.7
Peak firing rate, Hz	32.1 ± 3.0*	6.9–98.1	44.0 ± 3.9*	8.0–91.1
Anticipatory time interval, ms	40.0 ± 6.5	−37.4–144.2	39.913 ± 8.5	−78.7–113.1

Values are means ± SE and ranges of head direction (HD) cell tuning properties recorded in the anterodorsal thalamic nucleus (ADN) of control animals (n = 49 cells) and animals with combined medial precentral cortex (PrCM) and dorsal striatum (DS) lesions (n = 37 cells). A subset of HD cells from control (n = 37) and lesioned animals (n = 23) was included in the anticipatory time interval analysis (see materials and methods).

^*P < 0.05, independent-samples t-test.

Combined lesions of the PrCM and DS did not significantly affect the responses of ADN HD cells to environmental manipulations. HD cells in control and lesioned animals displayed equivalent PFD shift values from STD1 to ROT, shifting in register with the cue card but tending to under rotate (control, mean shift = 64.2 ± 7.1°, Rayleigh r = 0.786, P < 0.001; lesion, mean shift = 63.1 ± 8.4°, Rayleigh r = 0.760, P < 0.001; Watson’s two-sample test, U² = 0.072, P > 0.1) (Fig. 9B, left). Additionally, HD cell PFDs in both groups remained stable across standard sessions; PFD shift values from STD1 to STD2 were near 0° and did not differ significantly between control and lesioned animals (control, mean shift = 0.5 ± 5.8°CCW, Rayleigh r = 0.880, P < 0.001; lesion, mean shift = 12.1 ± 5.9°CCW, Rayleigh r = 0.874, P < 0.001; Watson’s two-sample test, U² = 0.124, P > 0.1) (Fig. 9B, right).

Finally, the control and lesion groups did not differ significantly in the amount of PFD stability displayed by ADN HD cells under dark conditions [control, 44.4 ± 3.2°; lesion, 51.1 ± 4.4°; independent-samples t-test, t(60) = 1.249, P = 0.217] (Fig. 9C). Presumably, HD cell tuning is maintained under dark conditions via self-motion cues. If the PrCM and DS provide critical self-motion information to limbic HD cells, we would expect combined PrCM and DS lesions to significantly disrupt ADN HD cell stability in the dark; however, no such disruption was observed. In sum, the results of experiment 2 suggest that the PrCM and DS do not transmit self-motion signals or other critical information to the limbic HD circuit.

Effects of lesions on locomotor behavior.

Combined PrCM and DS lesions had no significant effects on locomotor behavior. During STD1, control and lesioned animals traveled comparable distances [control, 50.8 ± 3.1 m; lesion, 57.3 ± 5.0 m; independent-samples t-test, t(71) = 1.124, P = 0.265] (Fig. 9D, left) while moving at similar mean linear speeds [control, 10.7 ± 0.6 cm/s; lesion, 12.2 ± 1.0 cm/s; independent-samples t-test, t(71) = 1.274, P = 0.207] (Fig. 9D, middle). Additionally, both groups displayed similar mean angular head speeds [control, 32.0 ± 1.6°/s; lesion, 31.5 ± 1.4°/s; independent-samples t-test, t(71) = −0.228, P = 0.821] (Fig. 9D, right).

DISCUSSION

Our results provide the first quantitative descriptions of HD cell populations in the PrCM and DS. These extralimbic HD cells are largely identical to ADN HD cells in terms of their tuning properties and responses to environmental manipulations, suggesting that the PrCM and DS HD signals are conveyed from the limbic system. Indeed, the results of experiment 1 confirmed this hypothesis, demonstrating that disruption of the limbic HD circuit eliminates HD cell activity in the PrCM and DS. Conversely, experiment 2 revealed that limbic HD cell activity recorded in the ADN is unaffected by combined lesions of the PrCM and DS, despite these structures being involved in self-motion processing. This pattern of results reveals a unidirectional functional relationship in which the HD signal is first generated in the limbic HD circuit and subsequently conveyed to the PrCM and DS. These extralimbic areas, however, do not provide critical input or feedback to HD cells in the limbic system.

Extralimbic HD Cell Characteristics Reflect Those of Limbic HD Cells

Previous descriptions of PrCM and DS HD cells by Wiener (1993) and Mizumori and colleagues (Mizumori et al. 2000, 2005; Ragozzino et al. 2001) provided the first evidence of these extralimbic HD signals. In the present study, we systematically characterized the properties of extralimbic HD cells using the same methodologies that have been used previously to characterize limbic HD cells located in the postsubiculum (Taube et al. 1990a), ADN (Taube 1995), lateral mammillary nucleus (Stackman and Taube 1998), dorsal tegmental nucleus (Sharp et al. 2001b), retrosplenial cortex (Cho and Sharp 2001), and medial entorhinal cortex (Sargolini et al. 2006). This standardized approach across studies enables a direct comparison of extralimbic vs. limbic HD cell characteristics.

Our findings revealed that the tuning properties of PrCM and DS HD cells are similar to those of the ADN and other limbic areas in terms of Rayleigh r, information content, directional firing range, peak firing rate, and ATI (although PrCM HD cells displayed a significantly higher peak firing rate compared with HD cells in the DS and ADN) (Stackman and Taube 1998; Taube 1995). Additionally, we showed that PrCM and DS HD cells are strongly influenced by familiar visual landmarks (i.e., during ROT), yet largely maintain their directional tuning when visual cues are removed entirely (i.e., during DRK) or when familiar visual information is unavailable (i.e., in the DCA). These findings demonstrate the multimodal nature of PrCM and DS HD cells—their directional representations are anchored to the external environment via visual landmarks, but in the absence of these external cues, internally generated self-motion information suffices to maintain their directional tuning. Limbic HD cells recorded in the current study and previously share these characteristics (Clark et al. 2010; Taube et al. 1990b; Taube 1995; Taube and Burton 1995; Yoder et al. 2015).

Extralimbic HD Signals Are Conveyed from the Limbic System

Given that the PrCM and DS HD signals are essentially replicas of the limbic HD signal, the most parsimonious hypothesis is that these extralimbic HD signals are conveyed from the limbic system and not generated intrinsically. This hypothesis is supported by the results of experiment 1: when we disrupted output from the limbic HD circuit by lesioning a central node, the ADN, HD cell activity in the PrCM and DS was eliminated. Given this functional relationship, there must be underlying anatomical connections capable of conveying information from the limbic HD circuit to the PrCM and DS; these pathways are explored in our accompanying report (Mehlman et al. 2019). Importantly, the HD signal does not appear to be significantly modified at the level of the PrCM and DS; if this were the case, we would expect more pronounced differences between PrCM, DS, and limbic HD cells in terms of their tuning properties and/or responses to environmental manipulations.

AHV Cells Intermix with HD Cells in the PrCM and DS

We identified populations of neurons in the PrCM and DS that were modulated by the animal’s AHV. These AHV cells were located in portions of the PrCM and DS that also contained HD cells. Although complete ADN lesions eliminated HD cell activity in the PrCM and DS, AHV cells in both structures were largely unaffected by the lesions. This dissociation suggests that the PrCM and DS receive the HD signal via pathways sensitive to ADN lesions, whereas the AHV signal is conveyed to the PrCM and DS via pathways insensitive to ADN lesions. In our accompanying report (Mehlman et al. 2019), we describe anatomical connections supporting this hypothesis. Alternatively, AHV signals in the PrCM and DS could be generated intrinsically. Given their involvement in motor control, it is possible that the PrCM and DS generate AHV cell activity from local motor efference signals. These signals are thought to originate in the motor cortex and propagate to the striatum via corticostriatal projections arising from pyramidal tract neuron collaterals (Reiner et al. 2010).

Given what is known about HD signal generation in the limbic system, it is somewhat surprising that the PrCM and DS contain AHV cells yet are incapable of generating HD cell activity intrinsically. The subcortical structures thought to generate the limbic HD signal (the dorsal tegmental nucleus and lateral mammillary nucleus) contain both HD cells and AHV cells; these limbic AHV cells display tuning curves that appear strikingly similar to those of the extralimbic AHV cells we recorded in the PrCM and DS (Bassett and Taube 2001; Sharp et al. 2001b; Stackman and Taube 1998). The AHV signals in the dorsal tegmental nucleus and lateral mammillary nucleus are thought to be dependent on input from the vestibular system and are considered critical for HD signal generation (Clark and Taube 2012). Indeed, manipulations thought to disrupt AHV cell activity in these subcortical areas also disrupt limbic HD cell activity (Butler and Taube 2015; Clark et al. 2012), and computational models of HD signal generation incorporate inputs conveying AHV information (Redish et al. 1996; Skaggs et al. 1995; Song and Wang 2005). Given the availability of a local AHV signal unaffected by ADN lesions, one might expect the PrCM and DS to be capable of generating HD cell activity intrinsically, even in animals with ADN lesions; however, this is not the case. Although local AHV cell activity may be necessary for HD signal generation in the limbic system, it is not sufficient for HD signal generation in the PrCM and DS.

The proportion of DS neurons classified as AHV cells was relatively small compared with that reported previously in studies examining head movement responses in the basal ganglia of rats (Gardiner and Kitai 1992; Trytek et al. 1996) and mice (Kim et al. 2014). There are three critical differences between the current study and these prior reports. First, we recorded in the most medial portion of the DS, whereas previous studies targeted more central and lateral portions of the DS, as well as the globus pallidus. Second, our animals were untrained and simply foraged for food pellets during recording sessions. In contrast, the previous experiments were performed in trained animals executing learned and stereotyped movements in response to cue presentations. This difference is noteworthy given that movement-related activity in the basal ganglia can be influenced by task demands (Gardiner and Kitai 1992; Kimura 1990). Finally, we classified AHV cells using relatively stringent criteria based on the properties of AHV cells observed in the dorsal tegmental nucleus and lateral mammillary nucleus (Bassett and Taube 2001; Stackman and Taube 1998), identifying neurons strongly modulated by the animal’s AHV in the horizontal plane. The previous studies employed more permissive criteria, examining responses to horizontal and vertical head movements, and either did not quantify head velocity (i.e., spiking was simply correlated to movement onset) or measured linear head velocity instead of AHV.

The Limbic HD Circuit Does Not Receive Self-Motion Information from the PrCM and DS

Experiment 2 revealed that the functional relationship between the limbic HD circuit and the PrCM and DS is unidirectional. We observed normal ADN HD cell activity in animals with combined PrCM and DS lesions, indicating that these extralimbic areas do not provide critical input or feedback to the limbic HD circuit. This finding is somewhat unexpected given that the PrCM and DS are functionally and anatomically situated to provide self-motion information to the limbic HD circuit. These structures are involved in motor control and thus could provide critical motor efference signals to the limbic system. Indeed, limbic HD cells can become unstable when motor efference signals are disrupted by passively transporting the animal (Stackman et al. 2003; Yoder et al. 2011; see also Winter et al. 2015b). Additionally, AHV signals arising from the PrCM and DS could contribute to the generation of limbic HD cell activity in a manner similar to AHV signals in the brain stem. The PrCM and DS, however, do not appear to convey critical motor efference or AHV signals to the limbic HD circuit because ADN HD cell activity in lesioned animals remained intact, even under dark conditions in which self-motion information is most critical. Additionally, the PrCM and DS do not appear to convey critical visual information to the limbic HD circuit as ADN HD cells in lesioned animals continued to shift in register with cue card rotations. This finding is somewhat surprising because 1) projection neurons in the medial portion of the DS receive direct excitatory input from the primary visual cortex (Khibnik et al. 2014), and 2) it has been proposed that striatal systems process information about local landmarks used for navigation (Burgess 2008). Importantly, these null effects are not due to a lack of the necessary anatomical connections; in our accompanying report (Mehlman et al. 2019), we describe pathways capable of conveying information from the PrCM and DS to the limbic HD circuit.

Our lesion and recording experiments revealed a functional relationship between the limbic HD circuit and the PrCM and DS. Next, in our accompanying report (Mehlman et al. 2019), we examine the underlying anatomical relationships in a series of retrograde and anterograde tracing studies that uncover a detailed pattern of connectivity between spatial processing circuitry and the PrCM and DS. We conclude with a general discussion connecting the findings from both reports to provide a comprehensive description of how the limbic HD circuit functionally and anatomically interacts with the PrCM and DS. Additionally, we discuss the potential functional roles of extralimbic HD cells in the PrCM and DS.

GRANTS

This work was supported by National Institutes of Health Grants DC009318 (to J. S. Taube), NS053907 (to J. S. Taube), and NS096888 (to M. L. Mehlman).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

M.L.M., S.S.W., S.V., and J.S.T. conceived and designed research; M.L.M., S.S.W., and S.V. performed experiments; M.L.M., S.S.W., and S.V. analyzed data; M.L.M., S.S.W., and J.S.T. interpreted results of experiments; M.L.M. prepared figures; M.L.M. drafted manuscript; M.L.M. and J.S.T. edited and revised manuscript; M.L.M. and J.S.T. approved final version of manuscript.