| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
The Neurosciences Institute, San Diego, California
Submitted 14 November 2006; accepted in final form 27 March 2007
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
demonstrated that hippocampal pyramidal neurons exhibit peaks in activity, also known as "place fields," that are highly specific to particular regions of an environment. Subsequent to this landmark finding, a large body of experimental work has detailed the spatially specific activity patterns of excitatory neurons in the CA1, CA3, and dentate gyrus (DG) subregions of the hippocampus (Eichenbaum et al. 1990; Ferbinteanu and Shapiro 2003; Knierim et al. 1995; McNaughton et al. 1983; O'Keefe and Conway 1978; Wilson and McNaughton 1993; Wood and Dudchenko 2003). From such work it can be concluded that the spatial firing fields of excitatory neurons, also known as "principal" neurons, arise from a complex interaction between the spatial arrangement of environment cues and self-motion information. The exact spatiotemporal dynamics of the excitatory and inhibitory synaptic inputs underlying the spatially discrete output of principal cells is unknown, but insight into the contribution of the local network inhibitory synaptic inputs can be garnered by an in-depth analysis of the spatial activity patterns of local interneurons.
The extent to which hippocampal interneurons are predisposed to exhibit fine-scale spatial firing fields is unknown. In behaving rats, the activity of some CA1 GABAergic neurons (hereafter referred to as "interneurons") is reliably spatially modulated (Ego-Stengel and Wilson 2007; Frank et al. 2001; Kubie et al. 1990; Marshall et al. 2002; McNaughton et al. 1983). However, they are often reported to exhibit broad firing fields and are not thought of as "place cells" per se. As a result, interneurons reportedly exhibit much less spatial specificity than that of principal neurons (Frank et al. 2001) and their firing rate is more considered for its relationship to other variables such as network oscillations (Csicsvari et al. 1999; O'Keefe 1976; Ranck 1973) and speed of movement (McNaughton et al. 1983). Taken together, there is a lack of consensus regarding the degree to which interneurons exhibit spatially specific activity and it is unknown to what extent their activity could shape the spatial activity patterns of excitatory (or "place") cells.
The characterization of interneuron spatial firing fields in previous work may have been hampered by the fact that the metrics designed to quantify or identify "place fields" were optimized based on principal neuron firing patterns. These commonly used metrics may not be objective when applied to interneurons that exhibit much higher mean firing rates and are active, to some degree, over nearly the full extent of a given environment. Consequently, the picture put forth by previous experiments may not have provided a uniformly accurate depiction of interneuron spatial activity patterns.
To overcome such issues, we used a recently developed measure referred to as "positional information" (Olypher et al. 2003) to directly compare the spatial firing patterns of principal cells and interneurons from the subiculum, CA1, and dentate gyrus. This measure computes relative entropy, which is the distinction (sometimes termed the "distance") between two probability distributions (Cover and Thomas 2006). In this case, the "distance" computed is that between the probability distribution of firing rates for a given location and the distribution of firing rates across all locations. From the perspective of an ideal observer, this analysis yields a positional information value for each position within the environment that essentially reflects the degree to which knowledge of the firing rate at any given time is increased by knowledge of the animal's position in space. High values for positional information are obtained when the average firing rate for a given location is both unique to that location and associated with minimal variability. As demonstrated by Olypher et al. (2003), the measurement of positional information captures the attributes of an ideal "place cell" that fires reliably at one location and is relatively silent everywhere else.
In the present work, we determined whether interneurons were capable of providing a similar degree of positional information. Because interneuron firing fields were previously described as spatially broad (relative to principal neuron place fields), we also tested whether interneuron firing fields were similar in size to those of principal neurons. In addition, we compared the speed in which (i.e., how abruptly) peak in-field firing rates were reached, relative to baseline firing rates, for interneurons and principal cells. Finally, the directional dependency of interneuron firing fields was examined because principal neuron "place fields" on a track-like environment are often dependent on the animal's direction of travel (McNaughton et al. 1983).
Interneurons in all three regions exhibited robust, spatially specific "ON" and "OFF" firing fields that were comparable in width, positional information, abruptness of rate change, and directional dependency to the classic "place fields" of hippocampal principal cells. Of the interneurons for which the animal's running speed did not contribute to variability in firing rate, >75% had one or more discrete fields. These findings suggest that a large subpopulation of interneurons provide a discrete inhibitory and/or disinhibitory output. Thus it would seem reasonable to hypothesize that inhibition and disinhibition shape the place-specific firing fields of principal neurons in the subiculum, CA1, and DG.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Eight adult male Sprague–Dawley rats were trained to run back and forth between two reward locations on a track-style maze. The distance traveled between the reward locations was 232 cm and the width of the maze at any given spot was 12 cm. Initially, rats received a piece of a Cheerio at both reward locations regardless of the route the rat chose to get to the reward location. Eventually the rats were trained to run to the reward location by a specific path after the center intersection or they would not receive a reward at the reward location. The path segment to use was indicated by an interchangeable textured portion of the wall near the beginning of the traversal. To summarize, the rats began a traversal, encountered the path indicator, freely chose straight or left at the first intersection without contingency of reward, arrived at the center intersection, chose one of the two paths that would get them to the reward location, and then received a reward at the goal location only if they chose the path segment that corresponded with the indicator at the start of the traversal. The indicator was randomly interchanged throughout a session with at least eight changes per day. Training took place over 2–4 wk and by the end of training all animals would traverse to the reward location by the correct path segment >80% of the time. The maze was in a dimly lit room surrounded by a black curtain.
Surgery
After behavioral training, the animals were implanted with custom-fabricated microdrives containing 24 stereotrodes. Animals were anesthetized with isoflurane and placed on a stereotaxic table. The skull was cleaned, etched, and dried after which approximately seven to nine anchor screws were inserted into the skull. Bilateral craniotomies were then made and the underlying dura mater was excised. The electrodes were inserted 0.5–1.0 mm deep in the brain over their respective targets and the exposed brain was covered in Neuroseal (Neuronexus Technologies, Ann Arbor, MI). Dental cement was used to fix the microdrives to the skull and support screws. Rats were allowed to recover for 6 days after surgery during which time feeding was ad libitum. The target coordinates relative to bregma were 2.9 mm lateral and 6.10 mm posterior for the subiculum and 4.0 mm lateral and 5.30 mm posterior for CA1 and dentate gyrus. There were four stereotrode bundles, each with five to seven stereotrodes, placed in cortex above each target.
Electrophysiology
Stereotrodes were constructed by twisting pairs of 25-µm, polyimide-insulated tungsten wire (California Fine Wire, Grover Beach, CA) and lightly heating with a heat gun. The wires were gold-plated to a final impedance of approximately 100 k
. The drives were fabricated from a combination of screws, cannulas, and fun plastic. The stereotrodes were lowered between 0.08 and 0.32 mm each day. The majority of recordings were made before moving the electrodes on any given day.
During recording sessions, the microdrive was connected to a custom-made head stage (NBLabs, Denison, TX) that used field-effect transistors for current amplification, and light-emitting diodes (LEDs, red-emitting) for behavioral tracking. Signals were relayed to amplifiers (Neuralynx, Tucson, AZ) and then computers running AD acquisition software (M. Wilson and L. Frank). The signals were filtered between 600 and 6 kHz, digitized at 32 kHz, and amplified by 5,000–30,000. A video camera and tracking system (DragonTracker, Boulder, CO) were used to record the position of the rat, by the LEDs on the head stage, with a sampling rate of 60 Hz and spatial resolution of 5 mm/pixel. Custom-made software (M. Wilson) was used to discriminate individual units based on waveform parameters from the records on each electrode of a stereotrode pair. The discrimination method was previously described in detail (Kargo and Nitz 2003; Nitz and McNaughton 2004).
Histology
Animals were perfused under deep anesthesia with a formalin solution and the brains were removed for dissection. The brains were sliced in 50-µm sections and Nissl stained. The mediolateral dispersion of each track was obvious as was the final depth. The anatomical depth was then compared with the recorded microdrive depth to ensure that the two were in register. In addition, the anatomical evidence and microdrive records were corroborated by EEG signals such as sharp-wave ripple events, theta oscillations, and gamma oscillations.
Delineating putative interneurons and putative principal cells
Cells were categorized as being either putatively excitatory or inhibitory based on spike width and firing rate (Fig. 1). This method was previously used for extracellular recordings from different brain regions including the hippocampus (Fox and Ranck 1981; Frank et al. 2001; Simons 1978). The spike width was taken from the electrode of the stereotrode pair with the largest amplitude spike. The extracellular record is, in theory, the first derivative of the intracellular record of the action potential. In practice, the extracellular spike is a slightly slower approximation of the first derivative, but the width of the action potential can be accurately determined by distinct points on the extracellular record and interneurons and pyramidal cells can be delineated by this measure (Henze et al. 2000). In the extracellular record, there is an obvious spike onset representing initiation of the action potential, a negative peak representing the upstroke of the action potential and a positive peak representing the peak rate of the rapid hyperpolarizing phase of the action potential. The spike width reported here is from the onset of the extracellular spike record to the positive peak. This is thought to represent the full action potential width (Henze et al. 2000). Our values ranged from 0.24 to 0.42 ms, which is a relatively short-duration intracellular action potential, but this indicates only that the extracellular record may better represent the first derivative of the action potential. That being the case, the values reported here may better reflect the half-width of the action potential and the width of interneuron spikes should be <0.32 ms. This is a rather conservative cutoff and limits the potential misclassification of excitatory cells. In addition, putative GABAergic fast-spiking units in the subiculum are clearly delineated from other classes of cells by a very brief spike width (Anderson and O'Mara 2003). Thus it is unlikely that any of the identified subicular putative interneurons were falsely classified. Because interneurons are known to have higher firing rates and excitatory neurons lower firing rates, all interneurons (Fig. 1, gray box at left) had to have an average firing rate >15 Hz and all excitatory cells <15 Hz (Fig. 1, gray box at bottom right). Because interneurons are also known to have a characteristic theta modulation and lack a prominent complex-spike burst, interneurons that lacked evidence of theta modulation or displayed a prominent burst were eliminated (n = 3).
|
There is no strict uniform set of criteria used to determine the existence or size of spatial firing fields. Somewhat recently, researchers have begun to use information theoretic analyses and other related analyses to quantify the spatial activity of hippocampal cells. The most commonly used analysis is given by the formula (Skaggs et al. 1993)
 |
i is the firing rate at that position; and is the mean firing rate over all positions. In this formula there is an obvious inverse relationship between the information value and mean firing rate and thus the formula is optimized to yield high values for classic CA1 "place cells" that exhibit high discharge rate at one position, but are relatively silent everywhere else.
More recently, however, it was shown that other analyses may be better suited to investigate the location-specific informational content provided by individual cells. Here we use one such measure, that of "positional information," described by Olypher et al. (2003). The determination of positional information is a specific application of the formula for "relative entropy" as described by Cover and Thomas (2006). As applied in this work, relative entropy (i.e., positional information) reflects the inefficiency in assuming that the distribution of firing rates across all 100-ms time periods (thus all locations) is the same as that for all 100-ms time periods associated with a particular location. Specifically, the measure computes the distinction between two probability distributions: 1) the probability of observing a given spike count across all 100-ms time windows (thus all locations) and 2) the probability of observing a given spike count across all 100-ms time windows associated with a particular location. The distinction between the two probability distributions is determined separately for each location in the environment. The resulting "positional information" value [or, Ipos(xi)] essentially reflects the degree to which knowledge of firing rate is increased by knowledge of the animal's location (Olypher et al. 2003). The mean value Ipos(xi) across all locations is equal to the mutual information, I(X; K) between the position and spike count variables (Olypher et al. 2003).
For all cells, the amount of positional information for each specific location, Ipos(xi), and the mutual information between position and spike count, I(X; K), was calculated (Olypher et al. 2003)
 |
For both probability distributions, responses (k) constitute the number of spikes in a 100-ms time window. For a given direction, responses and the position (xi) of the animal at the time of each response were recorded for the duration of all trials. Positions (xi) had a spatial resolution of 5 cm in the direction of travel of the rat by the width of the track, which was 12 cm. Only trials in which the animal traversed from the start to goal without stopping for more than 4 s at any given spot or reversing direction were used. For a position to be included in the analysis the rat had to encounter that position on at least eight trials.
The maximum positional information value possible depends on both the number of possible positions that the animal can take and the number of different spike counts possible. In the present experiments, the highest possible Ipos(xi) value is 6.29 bits. This would be the value obtained for a single position if a cell discharged spikes at just that position and nowhere else. Note, however, that in this case, I(X; K) would then be only 0.10 because Ipos(xi) would be 0.02 bit at all other locations. Across all positions the most informative cell would reliably fire at a different rate at each position, thereby producing a high degree of Ipos(xi) at each position and a high value of I(X; K).
Directionality was calculated using the formula
 |
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
From eight animals, recordings of 475 neurons were obtained that were verified anatomically to reside within the hippocampal formation (subiculum: n = 264, CA1: n = 92, dentate gyrus: n = 119). Using the duration of the extracellular spike waveform and the average firing rate (Fox and Ranck 1981; Frank et al. 2001; Simons 1978), 93% of the neurons were categorized as being either a putative principal cell (Fig. 1; n = 383) or a putative interneuron (Fig. 1; n = 59). See METHODS for more information.
In this navigational task, rats were allowed to freely traverse from a start location to a goal location by four different partially overlapping routes. Reward at the goal location was dependent on the particular path segments chosen (see METHODS). A textured indicator near the start of the traversal indicated which path segment to take and the indicator was randomly switched throughout a recording session.
CA1 spatial firing fields
To form the basis for quantitative comparison, the fields of classically identified CA1 "place cells" were defined using a recently developed analysis (Olypher et al. 2003). As opposed to a commonly used information theoretic analysis used in the field (Gothard et al. 1996; Hargreaves et al. 2005; Nitz and McNaughton 1999; Skaggs et al. 1993), this metric quantifies the informational content of the spike discharge of a cell at each specific location within an environment. Firing rate data from two representative CA1 cells are shown (Fig. 2).
|
The informational content provided by the spike discharge of the cell was quantified. Two measures were calculated: the amount of positional information provided for each specific position [Ipos(xi)] and the mean positional information across all positions, which is equal to the mutual information between position and spike count [I(X, K)]. The power of this analytical technique is that 1) informational content is calculated for each position; 2) the reliability of spike discharge across trials for each position is incorporated; and 3) there is no inherent bias for baseline firing rate. The spatial boundary of each position (xi) was 5 cm in the direction of travel of the rat by 12 cm, which is the width of the track. There were 73 total positions and 45 were encountered on any given traversal from the start to goal. Maps of the Ipos(xi) for each position are shown for this cell (Fig. 2A, bottom). The field traveling IN was associated with a peak in Ipos(xi) of 1.02 bits and the field traveling OUT was associated with a peak in Ipos(xi) of 1.61 bits. Both fields were obviously dependent on the direction of travel because the cell did not fire in the opposite direction.
The degree to which "place fields" reflected discrete changes in firing rate was assessed by quantifying field width as well as the abruptness of change from the baseline firing rate to the peak of in-field firing rate. Trials in which the rat traveled to the goal location by the same sequence of path segments were grouped (Fig. 2A, black line). The position data for these trials were linearized and the firing rate for each position is shown (Fig. 2B). A Gaussian function, constrained by the peak firing rate position (xi) and average firing rate (yi), was fit to the linearized firing rate data and the full width at half-height (FWHH) of the curve fit was used to quantify the size of the field. The slope at half-height ("SHH") of the same Gaussian function was taken as a measure reflecting the abruptness with which peak in-field firing rate was achieved. FWHHs for the firing fields on the IN and OUT paths were 28.0 and 34.1 cm, respectively. SHHs for the same IN and OUT firing fields were 0.74 and 0.55 Hz/cm, respectively. Ipos(xi) for each position is also shown for comparison.
Using only the firing rate criterion, cell 2 had one field traveling IN and two fields traveling OUT (Fig. 2C). It is not uncommon for a CA1 "place cell" to have multiple fields in the same environment (Gothard et al. 1996) and 7 of 42 CA1 cells with fields had multiple fields on the same path. When analyzed for informational content, the field on the IN path was associated with a peak in Ipos(xi) of 1.41 bits. Field A on the OUT path was associated with a peak in Ipos(xi) of 0.69 bit and field B on the OUT path was associated with a peak in Ipos(xi) of 0.91 bit (Fig. 2C). The somewhat lower information values associated with fields A and B may be attributable to the fact that the response was approximately the same (
8 Hz) at each location. From the perspective of an ideal observer, a firing rate of 8 Hz on the IN path provides a high degree of information that the animal is located at one specific location; however, a firing rate of 8 Hz on the OUT path provides information only that the animal is at one of two locations.
For the population of CA1 principal cells (n = 74), 61 total fields from 42 cells were identified according to the firing rate definition. According to the information metric (Skaggs et al. 1993) commonly used in the field (see METHODS), these neurons provided, on average, 1.86 bits of information per spike. Note that the information values derived from the two methods (Skaggs et al. 1993 vs. Olypher et al. 2003) are not directly comparable. The mean amount of information per spike (1.86 bits) is consistent with what was previously reported for neurons having a highly place specific spike discharge (Frank et al. 2000, 2001; Hargreaves et al. 2005). With the newer method applied to all fields, the average peak in Ipos(xi) was 1.16 bits and for 95% of the fields the peak Ipos(xi) was >0.65 bit. The average size (FWHH) of the fields was 30.9 cm and 95% of the fields were <57 cm in size. These location-specific positional information values obtained from classically identified CA1 "place cells" were used to define and quantify spatial firing fields. Therefore in this environment and using the positional information measure, a discrete field was deemed to exist when Ipos(xi) exceeded 0.65 bit and the spatial extent of the place field was <57 cm. Notably, there were no false-positive identifications of a field. Of the CA1 cells that lacked a field according to the firing rate criterion, none provided more than 0.65 bit of positional information for any position.
Interneurons with discrete ON and OFF fields
Having established the location-specific information content of CA1 principal cells, the informational content of interneurons was quantified for comparison. In contrast to principal cells, interneurons fire continuously while an animal is traversing a path, as shown for a representative interneuron (Fig. 3A). The average firing rate of this cell across all positions was 20.1 Hz. For this interneuron, there were two obvious places (field A and field B) on the OUT path where the firing rate was noticeably different from that of all other positions. At field A, the rate decreased to 2.2 Hz and at field B the rate increased to 40 Hz. The discrete decrease in activity at field A provided 1.23 bits of positional information and the discrete increase in rate at field B provided 0.68 bit of positional information. Hereafter these fields will be referred to as OFF fields and ON fields, respectively. The OFF field is enlarged (inset, top right) for visualization purposes. Both fields were dependent on the direction of travel of the rat. Directionality was quantified using the formula given in the methods (field A = 1.32; field B = 1.01). A value of 0 indicates that the firing rate was the same traveling IN or OUT and a value of 0.5 indicates the firing rate was twice as large in one direction versus the other direction.
|
For the same neuron, the firing rate at each position on each trial is shown (Fig. 3C). For the black path (top plot), the firing rate at each position for each trial is shown from 0 Hz to the mean rate (20.1 Hz). As can clearly be seen in this plot 1) the cell was reliably silent across most trials at field A and 2) the cell was silent predominantly only at that position. For the blue path (bottom plot) the firing rate at each position and each trial is shown color-coded from the mean to 40 Hz. In this plot it is obvious that 1) the firing rate was 
40 Hz on most trials at field B and 2) firing rates 40 Hz occur predominantly only at field B. In summary, this interneuron had two fields, one ON and one OFF, that yielded positional information values and field widths comparable to the two example CA1 principal cells shown in Fig. 2.
To further illustrate the spatial specificity and informational content of interneuron fields, six more interneurons (two subicular, one dentate gyrus, and three CA1) recorded from five different animals are shown (Fig. 4). All peak Ipos(xi) values, FWHH, and SHH values for these interneurons fell well within the ranges of the same values calculated for CA1 principal neurons. The peak Ipos(xi) values of all fields are shown for each cell (Fig. 4, A–F). The first subicular cell had an OFF field (A) and an ON field (B) traveling OUT (Fig. 4A). The second subicular cell had an OFF field traveling OUT (Fig. 4B). The dentate gyrus interneuron had an OFF field (A) and an ON field (B) traveling OUT (Fig. 4C). The first CA1 cell had three OFF fields (A, B, and C) traveling IN (Fig. 4D). The second CA1 cell had two ON fields (A and B) and two OFF fields (C and D) traveling OUT (Fig. 4E). The last CA1 cell had one ON field (F). Widths of the fields ranged from 23.1 cm (SUB-1: B field) to 35.3 cm (SUB-1: A-field). SHH values ranged between 0.58 and 0.96 Hz/cm.
|
The firing rate of principal neurons and interneurons can also be affected by changes in movement velocity (McNaughton et al. 1983); however, movement velocity did not contribute to the spatial firing fields of the interneurons shown thus far. The contribution of movement velocity to the variability in firing rate was quantified to ensure that interneuron fields were indeed related to the spatial location of the animal. For the interneuron shown in Fig. 3 (black route going OUT), firing rate versus movement velocity was plotted and a linear least-squares analysis was applied (Fig. 5A). The coefficient of determination (r2) was 0.01, indicating that only 1% of the variation in firing rate could be explained by movement velocity. In addition, the low firing rate associated with the OFF field (filled symbols) was not related to maximal or minimal movement velocities.
|
Interneuron fields versus principal cell fields: similarities and differences
The width, directionality, and location-specific positional information of interneuron and principal neuron fields were very comparable. Histograms of the peak Ipos(xi) for all interneuron (23 cells, 44 fields) and CA1 principal cell (42 cells, 61 fields) fields are shown (Fig. 6A). There was no statistical difference between the average peak in Ipos(xi) of interneuron (1.02 bits) and principal cell (1.17 bits) fields (Fig. 6D). Histograms of field widths are shown (Fig. 6B) and the average widths of the fields for the two populations were nearly identical (Fig. 6E; interneurons 30.7 cm; principal cells: 29.8 cm). In-field firing rates were similarly dependent on the direction of travel of the rat. Histograms of directionality values for all fields are shown (Fig. 6C) and there was no significant difference between interneuron and principal cell fields.
|
Although the individual fields of interneurons and principal cells were similar, the populations differed in the mean positional information across all locations. Mutual information between the position and spike count variables, I(X; K), was greater for interneurons than for principal cells [Fig. 7A: 0.26 vs. 0.13 bit, P < 0.01; note that I(X; K) is equal to the mean of all location-specific positional information values, Ipos(xi)]. It was initially speculated that this may be explained by the fact that interneurons simply fire more spikes. However, when position was randomly shuffled, I(X; K) was negligible for both populations (Fig. 7B). Interneurons also differed from principal cells in that they had, on average, more fields per cell (Fig. 7C: 2.2 vs. 1.4 fields per cell, P < 0.01). There was a weak, statistically insignificant correlation between total positional information content and the number of fields. Therefore the increased number of fields per interneuron may have contributed, in small part, to the increased total informational content of interneurons, but cannot be the sole cause. In summary, interneurons and principal neurons provide a comparable amount of location-specific information at a given field, but interneurons actually provide more location-specific information on average across all locations.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
) to the firing patterns of hippocampal and subicular excitatory and inhibitory neurons clearly demonstrated that a subset of inhibitory interneurons exhibit discrete spatial firing fields having the same fundamental features as the fields of CA1 principal cells. This metric also revealed that interneurons have OFF fields that are as discrete and informative as classic "place fields" of principal neurons. Taken together, these findings suggest that some interneurons register allocentric space more discretely than previously thought and that fine-scale inhibitory and disinhibitory local network actions contribute to the construction of principal neuron "place fields." The location-specific firing fields of interneurons were shown to be like the fields of principal neurons in four distinct ways. First, peak in-field positional information values were well within the range of the values obtained for CA1, DG, and SUB principal neurons. This indicates that the in-field firing rates for interneurons were as unique and reliable as the robust "all-versus-none" in-field firing rates of principal cells. Second, the abruptness with which firing rates changed from field entry to the point of peak in-field discharge was comparable for the two cell types. Third, the place-specific firing fields of both principal neurons and interneurons were strongly dependent on the animal's direction of motion. Thus interneurons and principal neurons are alike in generating distinct spatial activity patterns for different routes taken through track-style environments. Last, and perhaps most important, interneuron and principal cell fields were of the same size, indicating that the registration of allocentric position by interneurons bears the same fine-scale specificity as that of principal neurons.
There are at least three possible reasons why previous reports did not conclude that interneurons are capable of exhibiting location-specific activity that is on a par with that of principal neurons. First, reports directly comparing interneuron and principal cell spatial activity patterns were based primarily on CA1 interneurons (Frank et al. 2001; Kubie et al. 1990; Marshall et al. 2002). The number of CA1 cells in our data set was somewhat limited and there are reportedly anywhere from 10 to 16 different types of interneurons in CA1 (Somogyi and Klausberger 2005). Therefore although there are CA1 interneurons that exhibit discrete "place fields" (see Fig. 3, also see Ego-Stengel and Wilson 2007), it cannot presently be determined exactly how prevalent this capacity is among CA1 interneuron subtypes. The majority of velocity-insensitive subicular and DG interneurons exhibited discrete fields and it is possible that the fine-scale mapping of space by interneurons is more prevalent in these regions. Thus many interneurons of the hippocampal formation are capable of discretely encoding space, but a much more exhaustive survey is needed to characterize this property across hippocampal formation interneuron subtypes.
A second reason for the apparent contradiction is that the most commonly used spatial information theoretic metric (Skaggs et al. 1993) is only a global measure and does not quantify the informational content provided at each location. It also does not consider the reliability of a response across separate visits to a particular position and it is best suited for detecting the place fields of neurons having very low mean firing rates (Frank et al. 2001). In contrast, the analysis used here characterizes the spatial information provided by neuronal activity at each location within the environment, it incorporates reliability, and it considers any firing rate deviation from any baseline rate as potentially informative (Olypher et al. 2003). The analysis yields high information values for a given location when the discharge rate is both unique to that location and is reliably observed across multiple visits. These are the exact properties that characterize the spatially specific activity patterns of classic "place cells" that have reliably high in-field firing rates that stand out against a background of near-zero firing rates.
Finally, the third possible explanation may be the pooling of velocity-correlated and velocity-uncorrelated neurons in previous experiments. In any given task, reliable changes in movement velocity tend to occur over specific regions of an environment. If a transition in movement velocity occurs over a long distance, it follows that the firing rate may exhibit a corresponding broad modulation in rate.
Because we have shown that interneuron spatial firing fields are as spatially discrete as those of principal neurons, it follows that interneurons may directly shape the place-specific firing fields of principal neurons. This may equally occur by inhibitory or disinhibitory processes. The intrinsic excitability of interneurons contributes to an elevated baseline firing rate that enables them to discretely register spatial information by either activity maxima or minima. In the case of an OFF field, the discrete decrease in inhibitory input to a principal cell may in theory directly produce a field by enabling even a nonspecific excitatory input to evoke suprathreshold events at that position. From an engineering standpoint, the discrete decrease in inhibitory input could also act as an "AND" gate with like spatially specific excitatory inputs (Klyachko and Stevens 2006). The discharge rate of a principal cell at any given position could therefore be dependent on both a specific increase in excitatory input and a specific decrease in inhibitory input to that cell.
Recent findings suggest that ON fields of interneurons reflect excitatory input from one or more principal cells (Klausberger et al. 2005; Marshall et al. 2002; Maurer et al. 2006). Such data are consistent with our finding that interneuron ON fields are as spatially discrete as those of principal cells. Together, these findings suggest that interneuron ON fields are largely driven by inputs from single principal cells or by groups of principal cells sharing identical spatial registers. Irrespective of the mechanism by which ON fields are produced, they should effectively decrease the probability of spiking in their principal neuron targets. In this way, interneuron ON fields could serve to maximize the segregation of principal neurons into active and inactive groups and thereby maximize the specificity with which principal neurons register position in the environment.
Alternatively, if an interneuron contacts other inhibitory inputs, the ON field response of an interneuron could actually facilitate the ON field response of a principal cell by indirectly providing disinhibition (Tsodyks et al. 1997). This may be especially true if the synaptic target of an interneuron is the axon of a nonspecific inhibitory input innervating the proximal dendrite of a principal cell. Proximal inhibitory inputs may effectively compartmentalize a principal cell by shunting distal excitatory inputs. It follows that spatially discrete inhibition of those proximal inhibitory inputs can effectively decompartmentalize a principal neuron and facilitate the conduction of distal excitatory inputs known to be important for place cell activity (Ang et al. 2005; Brun et al. 2002; Kocsis et al. 1999). Thus interneuron ON field responses could, like OFF fields, effectively act as an "AND" gate for distal excitatory inputs that share the same spatial specificity.
The electrophysiological properties and connectivity patterns of several interneuron subtypes in the hippocampus, entorhinal cortex, and subiculum strongly indicate that they are positioned to play a prominent role in generating the discrete spatial firing fields of principal neurons (Freund and Buzsáki 1996). For example, one subtype selectively innervates specific dendritic segments of principal cells, suggesting that they serve to shunt specific afferent inputs (Han et al. 1993), and basket and chandelier interneurons may potently inhibit action potential generation in principal neurons by powerful synapses at the soma and axon initial segments (Somogyi et al. 1983a,b). In vitro work demonstrates that interneurons strongly regulate the bursting behavior of principal neurons of the subiculum (Menendez de la Prida 2003). The activity of single interneurons can potently suppress activity in excitatory neurons (Cobb et al. 1995) and activity bursts of CA1 principal cells may be amplified by selective disinhibition (Klyachko and Stevens 2006). Basket and chandelier interneurons were previously proposed to be critical in the timing of pyramidal neuron action potentials (Buzsáki and Chrobak 1995) and a recent work examining the timing of cholecystokinin-expressing neurons relative to the population theta rhythm suggests a role for this interneuron subtype in differentiating subgroups of pyramidal neurons forming neuronal assemblies (Klausberger et al. 2005). Thus the hypothesis that interneurons direct the fine-scale spatial specificity of principal cell field is consistent with anatomical and in vitro data.
Currently, there are no in vivo data detailing the dynamics of the inhibitory and excitatory synaptic inputs underlying the "place field" responses of principal cells. In sensory systems and neocortex, where it is much easier to combine in vivo intracellular recordings with the presentation of relevant sensory stimuli, a variety of inhibitory input dynamics were found to underlie the response tuning of individual neurons (Monier et al. 2003; Wehr and Zador 2003; Wilent and Contreras 2005). The inhibitory input to a cell can be exquisitely tuned for the preferred stimulus (Ferster 1986; Wehr and Zador 2003), but a diversity of tuning combinations is also encountered (Monier et al. 2003) and preferred responses may also be associated with decreased inhibition (Fried et al. 2005; Rao et al. 2000). Thus the possible roles of interneurons in shaping the spatial firing fields of principal cells discussed here are consistent with the known functional role of interneurons in neocortex.
Although it remains to be determined whether the shape of principal cell place fields is indeed critically dependent on the fine-scale positional firing patterns of interneurons, it is notable that alterations in the spatial firing properties of interneurons and principal cells tend to occur in parallel. For instance, physical restraint depresses interneuron discharge and abolishes spatially selective discharge of CA1 principal cells (Foster et al. 1989). The time courses for novelty-induced depressions in CA1 interneuron firing rates and changes in the accuracy of principal cell–based reconstruction of position are similar (Wilson and McNaughton 1993). Regions of an environment undergoing the greatest novelty-induced change in interneuron firing rates overlap regions where the greatest differences in principal cell spatial firing patterns are observed (Nitz and McNaughton 2004). Finally, the spatial coherence of interneuron and principal cell activities is similarly affected by pharmacological manipulation of cholinergic receptors (Brazhnik et al. 2003).
The distinct afferent input patterns and architectures of hippocampal and subicular networks suggest that a thorough understanding of how spatial firing fields in the brain are generated is potentially attainable. Much is known concerning how environmental stimuli, self-motion information, and network oscillations affect the place-specific discharge properties of hippocampal and subicular excitatory neurons (Barnes et al. 1990; Eichenbaum et al. 1990; Ferbinteanu and Shapiro 2003; Knierim et al. 1995; McNaughton et al. 1983; O'Keefe and Conway 1978; Wilson and McNaughton 1993; Wood and Dudchenko 2003). In addition, recent work identified differences in the form in which spatial information is registered across different subregions of the hippocampus, the subiculum, and the entorhinal cortex (Frank et al. 2000; Hafting et al. 2005; Knierim et al. 1995; Sharp and Green 1994). It is likely that subregion-specific transformations in the way that space is registered are more than simply a product of excitatory afferent input dynamics. The role of local networks is undoubtedly important and it has been shown here that location-specific inhibition and disinhibition are prominent within the hippocampal formation.
|
GRANTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
FOOTNOTES |
|---|
Address for reprint requests and other correspondence: D. Nitz, The Neurosciences Institute, 10640 John J. Hopkins Drive, San Diego, CA 92121-1120 (E-mail: nitz{at}nsi.edu)
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Ang CW, Carlson GC, Coulter DA. Hippocampal CA1 circuitry dynamically gates direct cortical inputs preferentially at theta frequencies. J Neurosci 25: 9567–9580, 2005.
Barnes CA, McNaughton BL, Mizumori SJ, Leonard BW, Lin LH. Comparison of spatial and temporal characteristics of neuronal activity in sequential stages of hippocampal processing. Prog Brain Res 83: 287–300, 1990.[ISI][Medline]
Brazhnik ES, Muller RU, Fox SE. Muscarinic blockade slows and degrades the location-specific firing of hippocampal pyramidal cells. J Neurosci 23: 611–621, 2003.
Brun VH, Otnass MK, Molden S, Steffenach HA, Witter MP, Moser MB, Moser EI. Place cells and place recognition maintained by direct entorhinal-hippocampal circuitry. Science 296: 2243–2246, 2002.
Buzsáki G, Chrobak JJ. Temporal structure in spatially organized neuronal ensembles: a role for interneuronal networks. Curr Opin Neurobiol 5: 504–510, 1995.[CrossRef][ISI][Medline]
Cobb SR, Buhl EH, Halasy K, Paulsen O, Somogyi P. Synchronization of neuronal activity in hippocampus by individual GABAergic interneurons. Nature 378: 75–78, 1995.[CrossRef][Medline]
Cover JA, Thomas TM. Elements of Information Theory (2nd ed.). Hoboken, NJ: Wiley, 2006.
Csicsvari J, Hirase H, Czurko A, Mamiya A, Buzsáki G. Fast network oscillations in the hippocampal CA1 region of the behaving rat. J Neurosci 19: RC20, 1999.
Ego-Stengel V, Wilson MA. Spatial selectivity and theta phase precession in CA1 interneurons. Hippocampus 17: 161–174, 2007.[CrossRef][ISI][Medline]
Eichenbaum H, Stewart C, Morris RG. Hippocampal representation in place learning. J Neurosci 10: 3531–3542, 1990.[Abstract]
Ferbinteanu J, Shapiro ML. Prospective and retrospective memory coding in the hippocampus. Neuron 40: 1227–1239, 2003.[CrossRef][ISI][Medline]
Ferster D. Orientation selectivity of synaptic potentials in neurons of cat primary visual cortex. J Neurosci 6: 1284–1301, 1986.[Abstract]
Foster TC, Castro CA, McNaughton BL. Spatial selectivity of rat hippocampal neurons: dependence on preparedness for movement. Science 244: 1580–1582, 1989.
Fox SE, Ranck JB Jr. Electrophysiological characteristics of hippocampal complex-spike cells and theta cells. Exp Brain Res 41: 399–410, 1981.[ISI][Medline]
Frank LM, Brown EN, Wilson M. Trajectory encoding in the hippocampus and entorhinal cortex. Neuron 27: 169–178, 2000.[CrossRef][ISI][Medline]
Frank LM, Brown EN, Wilson MA. A comparison of the firing properties of putative excitatory and inhibitory neurons from CA1 and the entorhinal cortex. J Neurophysiol 86: 2029–2040, 2001.
Freund TF, Buzsáki G. Interneurons of the hippocampus. Hippocampus 6: 347–470, 1996.[CrossRef][ISI][Medline]
Fried SI, Munch TA, Werblin FS. Directional selectivity is formed at multiple levels by laterally offset inhibition in the rabbit retina. Neuron 46: 117–127, 2005.[CrossRef][ISI][Medline]
Gothard KM, Skaggs WE, Moore KM, McNaughton BL. Binding of hippocampal CA1 neural activity to multiple reference frames in a landmark-based navigation task. J Neurosci 16: 823–835, 1996.
Hafting T, Fyhn M, Molden S, Moser MB, Moser EI. Microstructure of a spatial map in the entorhinal cortex. Nature 436: 801–806, 2005.[CrossRef][Medline]
Han ZS, Buhl EH, Lorinczi Z, Somogyi P. A high degree of spatial selectivity in the axonal and dendritic domains of physiologically identified local-circuit neurons in the dentate gyrus of the rat hippocampus. Eur J Neurosci 5: 395–410, 1993.[CrossRef][ISI][Medline]
Hargreaves EL, Rao G, Lee I, Knierim JJ. Major dissociation between medial and lateral entorhinal input to dorsal hippocampus. Science 308: 1792–1794, 2005.
Henze DA, Borhegyi Z, Csicsvari J, Mamiya A, Harris KD, Buzsáki G. Intracellular features predicted by extracellular recordings in the hippocampus in vivo. J Neurophysiol 84: 390–400, 2000.
Kargo WJ, Nitz DA. Early skill learning is expressed through selection and tuning of cortically represented muscle synergies. J Neurosci 23: 11255–11269, 2003.
Klausberger T, Marton LF, O'Neill J, Huck JH, Dalezios Y, Fuentealba P, Suen WY, Papp E, Kaneko T, Watanabe M, Csicsvari J, Somogyi P. Complementary roles of cholecystokinin- and parvalbumin-expressing GABAergic neurons in hippocampal network oscillations. J Neurosci 25: 9782–9793, 2005.
Klyachko VA, Stevens CF. Excitatory and feed-forward inhibitory hippocampal synapses work synergistically as an adaptive filter of natural spike trains. PLoS Biol 4: e207, 2006.[CrossRef][Medline]
Knierim JJ, Kudrimoti HS, McNaughton BL. Place cells, head direction cells, and the learning of landmark stability. J Neurosci 15: 1648–1659, 1995.[Abstract]
Kocsis B, Bragin A, Buzsáki G. Interdependence of multiple theta generators in the hippocampus: a partial coherence analysis. J Neurosci 19: 6200–6212, 1999.
Kubie JL, Muller RU, Bostock E. Spatial firing properties of hippocampal theta cells. J Neurosci 10: 1110–1123, 1990.[Abstract]
Marshall L, Henze DA, Hirase H, Leinekugel X, Dragoi G, Buzsáki G. Hippocampal pyramidal cell-interneuron spike transmission is frequency dependent and responsible for place modulation of interneuron discharge. J Neurosci 22: RC197, 2002.
Maurer AP, Cowen SL, Burke SN, Barnes CA, McNaughton BL. Phase precession in hippocampal interneurons showing strong functional cou
