Research Article

Optogenetic Dissection of Entorhinal-Hippocampal Functional Connectivity

See allHide authors and affiliations

Science  05 Apr 2013:
Vol. 340, Issue 6128, 1232627
DOI: 10.1126/science.1232627

This article has a correction. Please see:

Structured Abstract

Introduction

The mammalian space circuit contains several functionally specialized cell types, such as place cells in the hippocampus and grid cells, head direction cells, and border cells in the medial entorhinal cortex (MEC). The interaction between entorhinal cell types and hippocampal place cells is poorly understood. Hippocampal place fields are thought to be generated by transformation of spatial signals from the MEC, but which cell types contribute to this process remains elusive.

Embedded Image

Grid cells and border cells fire at similar latencies in response to local photostimulation. (Left) Spike rasters showing discharge in response to successive 3.5-ms light pulses for a grid cell and a border cell. Dots indicate spike times. (Right) Spike rasters showing color-coded normalized firing rate 0 to 50 ms after the light pulse for all grid cells and border cells.

Methods

We used a combined optogenetic-electrophysiological strategy to determine functional identity of entorhinal cells with output to place cells in the hippocampus. Channelrhodopsin-2 (ChR2) was expressed selectively in the hippocampus-targeting subset of entorhinal projection neurons by injecting retrogradely transportable ChR2-coding recombinant adeno-associated virus (rAAV) into the hippocampus. Hippocampus-projecting MEC cells could then be identified, after expression of ChR2 transgenes, as neurons that responded reliably, at a minimal latency, to photostimulation of either cell bodies in the MEC or axons of MEC cells in or near the hippocampus.

Results

Many light-responsive MEC cells were grid cells, but short-latency firing could also be induced in border cells and head direction cells, as well as neurons with irregular firing fields or no fields at all. In each cell group, the majority of neurons discharged at minimal response latencies, suggesting that they had direct projections to the hippocampus. The same cells could also be backfired, at slightly longer latencies, by photostimulation of the cells’ axons in the hippocampus. MEC cells could also be activated by photostimulation in the hippocampal CA1 subfield, but response latencies were more than 150% longer than with somatic or axonal stimulation, suggesting the activations was now synaptic.

Discussion

These findings suggest that place signals are generated by convergence of signals from a variety of entorhinal functional cell types, of which grid cells are the most abundant spatial cell type. A dual spatial input from grid cells and border cells is consistent with the idea that place cells have access to both self-motion and landmark-based information and raises the possibility that the spatial metric of the place-cell population originates from grid cells, whereas boundary- and landmark-induced firing is derived directly from border cells. The dual nature of the spatial input may explain the observation that place cells precede mature grid cells during ontogenesis of the spatial representation system and that place cells can maintain location specificity under conditions that reduce grid-cell periodicity in adult rats. Convergent input from a broad spectrum of entorhinal cell types may also enable individual place cells to respond dynamically, so that different types of input are favored in different behavioral states or behavioral circumstances.

From Grid to Place

Grid cells are considered one of the key sources for place-cell signals in the hippocampus. The entorhinal circuit also contains other functional cell types, but it is unclear which project to the place cells of the hippocampus. Zhang et al. (10.1126/science.1232627, see the Perspective by Poucet and Sargolini) addressed this question using optogenetics and in vivo multi-electrode electrophysiology. Hippocampal cells received input from a broad spectrum of entorhinal neuronal cell types. Grid cells represented the biggest group of spatial inputs, but border cells, head-direction cells, and a large fraction of nonspatial cells also provided inputs. Thus, hippocampal circuits have local mechanisms for processing specific types of functional input from the entorhinal cortex to generate place-specific signals.

Abstract

We used a combined optogenetic-electrophysiological strategy to determine the functional identity of entorhinal cells with output to the place-cell population in the hippocampus. Channelrhodopsin-2 (ChR2) was expressed selectively in the hippocampus-targeting subset of entorhinal projection neurons by infusing retrogradely transportable ChR2-coding recombinant adeno-associated virus in the hippocampus. Virally transduced ChR2-expressing cells were identified in medial entorhinal cortex as cells that fired at fixed minimal latencies in response to local flashes of light. A large number of responsive cells were grid cells, but short-latency firing was also induced in border cells and head-direction cells, as well as cells with irregular or nonspatial firing correlates, which suggests that place fields may be generated by convergence of signals from a broad spectrum of entorhinal functional cell types.

Neuronal systems investigations have matured to the point where computation can be understood at the microcircuit level. In sensory systems, electric signals can be followed through networks of increasing complexity, from receptors to cortex, and experiments have established how neural representations change from one level to the next. Less is known, however, about how neural codes are formed and transformed in higher-order nonsensory cortices. One example of a high-level neural representation is the place code of the hippocampus (1, 2). Place cells are hippocampal cells that fire when the animal is at a particular location in a particular environment, independently of specific sensory inputs (3). Almost all pyramidal cells in the CA areas of the hippocampus are place cells (46). The origin of the place signal has not been established, but the recent identification of functionally specialized cell types in the medial entorhinal cortex (MEC), upstream of the place cells, provides some clues (2). The first entorhinal cell type to be characterized was the grid cell, which has periodic firing fields that span the entire available space in a hexagonal pattern (7, 8). Grid cells intermingle with head-direction cells, which fire whenever animals face a particular direction, irrespective of where they are or what they are doing (911), as well as “border cells,” which signal proximity to specific geometric boundaries of the local environment (1113). The entorhinal cortex also contains cells with irregular spatial firing fields or no spatial firing fields at all (1012, 14). All of these cell types are active in all environments, and their spatial and directional firing relationships are preserved when the animals move from one environment to another (12, 15), which suggests that these cells, unlike the place cells of the hippocampus (16), are part of a path integration–dependent metric applied universally across environments (2, 8, 17).

Although the entorhinal and hippocampal cortices are known to be strongly connected, the functional interactions between the two mapping systems, as well as the interdependence between the different cell types, have not been determined. Hippocampal place fields may be generated by transformation of spatial signals from the entorhinal cortex (2, 1723), but which cell types contribute to this process remains elusive, given that only a subset of the entorhinal projection neurons target the hippocampus (24). The present study set out to identify the functional cell types that project to the hippocampus by combining tetrode recording in the MEC with optogenetic tagging of the hippocampus-projecting cell population in this region.

Results

Retrograde Transduction of Entorhinal Projection Neurons

Entorhinal neurons with projections to the hippocampus can be identified by antidromically stimulating axons in the projection area at the same time as spikes are recorded from the parent cells in the entorhinal cortex. The power of this strategy is limited, however, by the small fraction of entorhinal-hippocampal fibers that can be discharged from a single electrode position in the hippocampal terminal region (25). A substantially larger proportion might be reached if it were possible to tag the incoming axons by hippocampal infusion of a widely diffusible viral vector carrying a light-responsive transgene. We used recombinant adeno-associated virus (rAAV) to target transgenes to entorhinal axon terminals across large parts of the dorsal hippocampus. To maximize transgene expression in the cell bodies of the projection neurons, we improved the retrograde axonal transport properties of classic rAAV-based transduction protocols (26, 27).

Transduction potency, transgene expression, and retrograde transport efficiency were evaluated by stereotaxic injections of rAAV2/1-EYFP-Nuc into the rat dorsal hippocampus (EYFP, enhanced yellow-green fluorescent protein) (28). Two to 3 weeks after viral injection, robust labeling of EYFP fluorescence could be observed in all subfields of the dorsal hippocampus at and around the injection sites (Fig. 1). In addition, EYFP-positive neurons could be identified in layers II and III of MEC, which indicates that the rAAV2/1 vector had infected cell bodies of MEC neurons with projections to the hippocampal injection site. EYFP-positive neurons were nearly absent in the deep layers of MEC (Fig. 1 and fig. S1), which indicates that entorhinal transduction was not caused by passive diffusion of injected virus from the hippocampal infusion site. Labeling was also observed in other brain regions with projections to the hippocampus (29), such as the parasubiculum, lateral entorhinal cortex, nucleus reuniens, and claustrum (Fig. 1). The EYFP expression pattern in superficial layers of the MEC was replicated in an animal injected with rAAV coding for the reporter gene β-galactosidase (β-gal or LacZ) fused with a C-terminal FLAG tag (fig. S2).

Fig. 1 rAAV-induced retrograde labeling of hippocampus-projecting MEC neurons.

Sagittal brain sections show viral transduction of hippocampus-projecting cells in the MEC of a rat injected with EYFP-coding rAAV in the dorsal hippocampus (all sections from the same rat; top, lateral; bottom, medial). (Left) Low magnification; (right) high magnification of framed area at left. Note high density of EYFP-expressing neurons in layers II and III but not V and VI in the dorsal MEC (right). Arrowhead indicates retrogradely labeled cells in the claustrum, which has massive unidirectional projections to the hippocampus (63).

Targeting Channelrhodopsin-2 (ChR2) to Entorhinal Projection Neurons

To obtain optogenetic control of the virally transduced MEC neurons, we engineered rAAV2/1 encoding the gain-of-function (H134R) photocurrent-enhanced mutant (30, 31) of the light-activated cation channel ChR2 (3234). EYFP was replaced with a nonfluorescent FLAG tag in this vector. The rAAV was stereotactically injected into the dorsal hippocampus. Immunohistochemical staining with FLAG antibody showed that, within 2 to 3 weeks after intrahippocampal injection, there was reliable widespread expression of the transgene in all subfields of the dorsal hippocampus, particularly the CA fields, as well as in superficial layers of the MEC (28 animals with rAAV-ChR2-FLAG) (Fig. 2 and fig. S3). FLAG was expressed widely across the somato-dendritic surface membrane of the MEC cells (Fig. 2 and fig. S3, right panels). The tetrodes were always located near FLAG-expressing neurons (Fig. 3 and fig. S3). A similar retrograde-transport strategy was developed to transduce entorhinal projection neurons with genes for microbial opsins that hyperpolarize the cell in response to local photostimulation (NpHR and Arch) (figs. S4 to S7) (35). Injections in the overlying neocortex did not result in labeling in MEC (fig. S8).

Fig. 2 Sagittal brain sections showing distribution of ChR2-expressing neurons in the MEC of a rat injected with rAAV-ChR2-FLAG.

(Left) Low resolution; (right), high resolution of the framed area. ChR2-expressing neurons were visualized by immunofluorescent staining using a fluorescent conjugated secondary antibody against the primary antibody to FLAG. Note widespread expression in superficial layers of the dorsal MEC. Strong labeling is also seen in the angular bundle (AB), which contains the perforant-path and temporoammonic fibers of hippocampus-targeting entorhinal projection cells (Fig. 9 and fig. S21). In the hippocampus, all strata are heavily labeled, reflecting expression in both cell bodies and axons.

Fig. 3

Photoexcitable cells in dorsal MEC.(A to C) grid cell; (D to F) border cell; (G to I) head-direction cell; (J to L) nonspatial cell. (A) ChR2-expressing neurons in MEC after rAAV-ChR2-FLAG transduction in the dorsal hippocampus. ChR2-expressing neurons were visualized by immunofluorescent staining using a fluorescent conjugated secondary antibody against the primary antibody to FLAG. (Top) Sagittal brain section showing FLAG expression in hippocampus (left) and dorsal MEC (framed area to the right); (middle) magnification of the framed area; (bottom), adjacent Nissl-stained section corresponding to the high-magnification FLAG stain. End of tetrode trace is indicated (arrowhead, approximate recording position). Rat number and hemisphere (L, left; R, right) are indicated. (B) Firing pattern of a responsive grid cell recorded at the tetrode location in (A). Text label indicates rat number, recording trial, and cell number (T, tetrode; C, cell). (Top left to bottom right): Trajectory with superimposed spike locations in red; color-coded rate map with scale bar indicating minimum and maximum firing rates; polar plot showing firing rate as a function of head direction, with peak rate indicated; and color-coded autocorrelation matrix for the rate map with scale bar showing minimum and maximum correlation values. The scale of the autocorrelogram is twice the scale of the rate map. (C) (Top) Spike rasters showing discharge in response to successive light stimuli for the grid cell in (B) from 50 ms before the stimulus to 100 ms after. The optic fiber was dorsal to the tetrode in MEC. Light was on from 0 to 3.5 ms. Each row corresponds to the beginning of a 1-s period. Dots indicate spike times. Note reliable discharge at a fixed latency (9.0 ms) after the onset of stimulation. (Bottom) Peristimulus spike histograms showing distribution of spikes during the poststimulus interval. (D) Location of tetrodes and virally transduced neurons in MEC in an animal with a light-responsive border cell [symbols as in (A)]. (E) Firing pattern of a photoresponsive border cell recorded at the tetrode location indicated in (D). (Left) Trajectories with spike locations; (right) color-coded rate maps. The first trial (top) was recorded in an empty 1-m-wide square box. In the subsequent trial (below), a wall (white stripe) was placed in the box, in parallel with the peripheral wall that maintained the border field when the box was empty. On the third trial, the rat was tested again in the empty box. The creation of a new field on the distal side of the insert, with an orientation matching the original field along the peripheral wall, is a defining characteristic of border cells in the MEC (12, 37). (F) Spike rasters and peristimulus spike histogram as in (C). (G to I) and (J to L) Location of tetrodes and virally transduced neurons, spatial firing patterns, and rasters and peristimulus spike histograms, as in (A to C).

Photoexcitation of Entorhinal Projection Neurons

To identify the subset of entorhinal neurons with direct projections to the hippocampus, we used a photoexcitation strategy where direct and indirect activation could be distinguished based on response latencies. ChR2-coding rAAV2/1 was first injected into the hippocampus and then, during the same surgical session, an assembly of tetrodes and a laser-coupled optical fiber was implanted into the MEC (fig. S3). Several weeks later, spike activity was recorded from cells in layers II and III of MEC [665 putative principal cells and 114 high-rate narrow-spiking cells likely to be γ-aminobutyric acid–releasing (GABAergic) interneurons] while the animals ran to collect food morsels in a 1-m-wide square enclosure. Grid cells, border cells, and head-direction cells were defined, respectively, as cells with scores for six-fold rotational symmetry, border proximity, and directional modulation that exceeded the 99th percentile for a distribution of corresponding scores from shuffled data (4, 5, 11) (figs. S9A, S10A, and S11A). Spatially modulated cells with irregular nonperiodic firing fields (1012, 14) were defined as cells that did not satisfy criteria for grid cells, border cells, or head-direction cells but had spatial information values that exceeded the 99th percentile of the shuffled data (fig. S12A). Nonspatial cells were defined as cells that satisfied neither of the above criteria. Using these definitions, we found that 158 cells were grid cells (71 clearly in layer II, 60 clearly in layer III); 41 were border cells (12 clearly in layer II, 18 clearly in layer III); 166 were head-direction cells (5 clearly in layer I, 44 clearly in layer II, 75 clearly in layer III); 47 were irregular spatial cells (26 clearly in layer II, 16 clearly in layer III); and 253 were nonspatial (118 clearly in layer II, 100 clearly in layer III). Many head-direction cells had conjunctive properties, satisfying criteria for grid cells or border cells in addition (9.0% and 11.4%, respectively, with a 99th percentile threshold).

For each recorded MEC neuron, we then asked if the cell discharged in response to local photostimulation at a 473-nm wavelength, as might be expected if it expressed ChR2. Photostimulation was given after the foraging trial while the rat rested in a deep flower pot on a pedestal. After a 2-min no-stimulation baseline, the rat received 2 min of shuttered 3.5-ms 473-nm pulses at a light power density of 10 mW/mm2 and a frequency of 1 Hz, followed, in 16 of the rats, by 2-min blocks of 5-pulse trains at 5 Hz and 10-pulse trains at 20 Hz (train rates 0.2 Hz). Photostimulation did not induce detectable changes in the spike waveforms of the recorded cells under the described conditions (fig. S13). A large fraction of the cells responded to the photostimulation (Figs. 3 and 4). Light-responsive cells were identified and counted by comparing, for each cell, the proportion of spikes emitted after the light stimulus at 1 Hz stimulation with the proportion expected to fire at similar latencies by chance. The expected proportion was determined for each cell by a shuffling procedure in which the timing of each spike was displaced randomly forward or backward in time within an interval (–100 ms to 100 ms] around the light stimulus. The displacement was repeated 10,000 times. Spikes were then counted for bins of 1 ms in 100-ms time windows after the stimulation, both for the real data and for the aggregated shuffled data, and the 3-ms block with the maximum number of spikes was identified for each of the two distributions. The cell was defined as light responsive if the proportion of spikes in the most-active 3-ms block in the real data exceeded the 99.9th-percentile value for the proportion of spikes in the most-active 3-ms block in the shuffled distribution. On the basis of this criterion, a total of 183 MEC cells were defined as photoresponsive (47.5% of all cells recorded in MEC layers II and 23.5% of all cells in layer III). A total of 135 of these neurons had waveform and firing-rate properties thought to be typical of principal cells (figs. S13 and S14A). The number of spikes emitted at the peak latency in the responsive cells was many times higher than the background firing at any other time (Figs. 3 and 4 and figs. S9 to S12 and S14).

Fig. 4 Firing latencies for the entire sample of photoresponsive MEC neurons.

(A) Spike rasters showing color-coded firing rate as a function of time after the start of photostimulation for all categories of principal cells that responded to the light pulse. Firing rate is shown for successive time bins of 0.5 ms during the 50 ms after the light pulse. Each row shows data for one cell. Dark blue is silent; red is high rate; and rates are normalized to peak rate (color scale at right). Among the head-direction cells, number 11 is a conjunctive grid × head-direction cell, and numbers 3, 8, and 13 are conjunctive border × head-direction cells. The absence of spike latencies shorter than ~9 ms suggests that this is the time it takes to discharge a ChR2-expressing principal cell by direct optical stimulation under the conditions described herein (36). The almost identical minimal discharge latency of the four cell groups suggests that all groups have direct projections to the hippocampus. The few cells with longer latencies (e.g., grid cells numbered 35 to 37) were probably activated indirectly (synaptically). (B) Distribution of modal peak latencies across bins of 0.5 ms during the first 20 ms after the onset of photostimulation (all cells of each category). Green vertical line indicates modal latency for all principal cells.

The response latency of the cells that passed the criterion was short and varied minimally within and between animals (Fig. 4 and fig. S15). The modal latency of all responsive principal cells at the standard power density (10 mW/mm2) was 9.50 ± 0.14 ms, measured from the beginning of the light pulse (mean ± SEM across cells). This latency is comparable to the 9- to 10-ms response time obtained with similar light flashes in ChR2-activated principal neurons of other cortical regions (36). Response latencies varied minimally with changes in power density above ~2.5 mW/mm2 (Fig. 5 and fig. S16) (interquartile ranges of 9.9 to 10.6 ms at 2.5 mW/mm2, 9.5 to 10.5 ms at 5.0 mW/mm2, and 9.1 to 10.2 ms at 10 mW/mm2). The relative constancy of response latencies across trials and cells with the chosen stimulation parameters points to photoillumination-induced discharge as a powerful tool for identifying cells with monosynaptic connections to a ChR2-infected target region.

Fig. 5 Effect of power density on response latency for a representative light-responsive MEC cell.

The example cell is a grid cell (5-digit rat number, trial number, and cell number are indicated). (Left column, from top to bottom) Trajectory with spike locations, color-coded rate map, directional rate map, and autocorrelation map, as in Fig. 3B. (Right four columns): (Top row) Raster diagram (as in Fig. 3C) showing firing in response to light pulses at increasing power densities, from 1 mW/mm2 to 10 mW/mm2. Light was on from 0 to 3.5 ms in each case. (Bottom row) Peristimulus time histogram. Note reliable discharge at an almost fixed latency (~9 to 10 ms) at 5 and 10 mW/mm2. Few cells discharged at less than 5 mW/mm2, but when they discharged, the latencies were most often similar to those of the higher power densities.

A large fraction of the photoresponsive principal cells were grid cells (37 cells or 27.4%), but the sample of excitable principal cells also included 10 border cells (7.4%) and 16 head-direction cells (11.9%), as well as three irregular spatial cells (2.2%) and 69 nonspatial cells (51.1%) (Figs. 3 and 4 and figs. S9 to S12). Among the excitable head-direction cells, three had conjunctive properties (one grid × head direction; three border × head direction). Responsive grid cells were recorded in both layer II and layer III (grid cells: 22 and 9 cells, respectively; border cells: 5 and 1 cell; head-direction cells: 1 and 5 cells; irregular spatial cells: 1 in each layer; nonspatial cells: 42 and 14 cells, respectively). Because the counts of nonresponsive cells are likely to include hippocampus-projecting cells outside the illuminated region, we followed up with analyses where the sample was limited to responsive and nonresponsive cells from tetrode bundles with at least two responsive cells. Within this subset, likely to be nearer the light source, the five cell types responded at approximately similar frequencies (grid cells: 24 of 66 cells, or 36.4%; border cells: 7 of 13 cells, or 53.8%; head-direction cells: 6 of 36 cells, or 16.7%; irregular spatial cells: 0 of 23 cells; nonspatial cells: 41 of 80 cells, or 51.3%). The exact proportion of inputs from each cell class cannot be inferred, considering that uptake and retrograde transport of virus may be influenced by possible differences in terminal distribution, axonal diameter, and myelination.

The functional identity of putative border cells was verified on separate trials where a 50-cm-long barrier was inserted centrally in the recording box in parallel with the wall that maintained the original border field (Fig. 3E and fig. S10B). In cells with high border scores, this generally leads to the emergence of a new field along the insert on the side that faces away from the original field (the distal side) (12, 37, 38). In the present border cells, the firing rate increased from 0.64 ± 0.11 to 7.1 ± 1.3 Hz within a 20-cm band on the distal side of the wall insert [t(6) = 5.18, P < 0.005; paired-samples t test], whereas no increase was detected on the proximal side [from 3.6 ± 1.3 to 3.1 ± 0.8 Hz; t(6) = 0.83, P > 0.40]. Firing rates returned to baseline after removal of the wall (1.2 ± 0.5 Hz on the distal side).

Direct Activation of ChR2-Expressing Cells

A photoresponsive cell might fire because ChR2 channels are activated, not in the recorded cell itself, but in cells with synaptic connections to the recorded cell. Such connected cells might include local entorhinal cells as well as hippocampal cells with axons branching into the photostimulated region of the MEC (39). We first evaluated the presence of indirectly activated neurons by measuring the variation in latency from stimulation to discharge in the entire population of responsive cells. The minimum modal latency in the cell sample should correspond to the time it takes to elicit an action potential by direct illumination of the recorded cell, whereas longer latencies could reflect indirect activation. In the majority of MEC cells, local photostimulation caused firing only at short latencies, with little variation from cell to cell or from trial to trial (Fig. 4 and figs. S9 to S12). Short latencies predominated in all functional cell types. Delayed firing and secondary firing peaks were observed only in exceptional cases (figs. S9C, S10C, S14, and S17). The number of long-latency outliers varied slightly between the groups, but the values for the lower 10th, 20th, and 50th percentiles were highly similar (Fig. 3) (grid cells: 8.69, 9.2, and 9.69 ms, respectively; border cells: 8.63, 8.86, and 9.31 ms; head-direction cells: 8.30, 8.35, and 9.18 ms; irregular and nonspatial cells: 7.87, 8.46, and 9.41 ms), which suggests that cells with direct hippocampal projections exist in all functional cell populations. The latency distributions of most cells were symmetric, with skewness values not differing from zero in any of the groups (grid cells: −0.28 ± 0.25; border cells: −0.58 ± 0.40; head-direction cells: −0.47 ± 0.47; t values < 1.5, P > 0.18), except for the nonspatial cells [−0.49 ± 0.18; t(67) = 2.7, P < 0.01]. The general lack of positive skew in the latency distributions of individual cells supports the idea that the vast majority of light-induced responses were directly activated. The low number of long-latency cells is consistent with the lack of direct excitatory connections in layer II of MEC (40, 41).

The consistently short activation latencies of the space- and direction-modulated principal cells were in contrast to the variable responses obtained in light-responsive MEC cells with discharge characteristics similar to those of hippocampal interneurons. These cells had short waveforms (peak-to-trough width: 177 ± 7 μs), high firing rates (mean rate: 19.7 ± 1.1 Hz), and no clear spatial or directional preferences (Fig. 6, A and B, and fig. S14), properties which all are thought to be typical for fast-spiking GABAergic cells in the hippocampus as well as the entorhinal cortex (42, 43). Forty-eight of the 114 high-rate narrow-waveform units in the cell sample (42.1%) satisfied the criterion of having a larger number of spikes in the most-active 3-ms block after illumination than in the corresponding shuffled distribution. Almost half of these cells discharged at latencies similar to those of the principal cell groups (less than ~10 to 11 ms) (Fig. 6, C to E), which is consistent with previous studies showing that a subset of entorhinal GABAergic neurons has direct projections to the hippocampus (44, 45). However, as a whole, the latency distribution was skewed toward longer latencies, with more than one-quarter of the population firing between 11 and 16 ms and the rest within a broad band ~20 to 40 ms after the light stimulus (Fig. 6, D to E). The mean firing latency of the putative GABAergic projection neurons (12.8 ± 1.0 ms) was significantly longer than for the space and direction-modulated principal cells [9.6 ± 0.1 ms; t(112) = 3.7, P < 0.001]. Medians were also different (10.4 versus 9.4 ms; Mann-Whitney U test: Z = 3.9, P < 0.001). The fact that so many late-activated cells were found in this cell group is consistent with the observation that the majority of local connections from projection neurons in MEC layers II and III are onto inhibitory cells (41).

Fig. 6 Photoinduced spiking in putative GABAergic neurons.

(A) Superimposed waveforms from a high-rate narrow-waveform MEC neuron. Representative waveforms are shown for each of the four electrodes of the tetrode. (B) (Left to right) Trajectory with spike locations, color-coded rate map, directional rate map, and autocorrelation map for a representative high-rate narrow-waveform cell, as in Fig. 3B. (C) Photostimulation-induced discharge in three representative cells in this group, as in Fig. 3C (spike rasters and peristimulus spike histograms). Modal peak latencies are indicated for each cell. The cell in the top panel of (C) is the same as the one in (B). The response latency of this cell is similar to that of the principal neurons in Fig. 4. The middle cell fired at a longer latency, which suggests that it was activated synaptically. The bottom cell fired at a short latency but showed additional peaks during the subsequent 10 ms, which suggests that the cell was discharged both directly and indirectly. (D) Color-coded spike rasters for the entire sample of responsive high-rate narrow-waveform units, as in Fig. 4A. (E) Distribution of modal peak latencies across bins of 0.5 ms during the first 20 ms after the onset of local photostimulation (all high-rate narrow-waveform units). Note the skewed nature of the distributions in (D) and (E). The extended latencies suggest that many of these cells were activated synaptically via photostimulation of other cells.

The symmetric distribution of response latencies in the principal neurons also speaks against the possibility that cells were activated by stimulation of ChR2-expressing axons from projection neurons in the CA1 of the hippocampus. Output from CA1 cells would be expected to skew the response latencies of the MEC cells to the right. To test the contribution of hippocampal axons more directly, we implanted optic fibers in the CA1 pyramidal cell layer or the stratum oriens or alveus of five animals with tetrodes in the MEC. We identified 14 MEC cells that responded to CA1 illumination in these animals (seven principal cells and seven high-rate narrow-waveform cells) (Fig. 7 and fig. S18). These neurons had peak latencies ranging from 15.8 to 30.2 ms. In five of the cells, stimulation was evoked both from the hippocampus and the MEC, always with longer latencies after hippocampal illumination (9.7 to 10.7 ms versus 18.5 to 29.4 ms).

Fig. 7 Synaptic discharge of MEC neurons after stimulation of hippocampal pyramidal cells.

(A) (Left to right) Trajectory with spike locations, color-coded rate map, spatial autocorrelation map, and directional rate map for a nonspatial cell, as in Fig. 3B. (B) Discharge in the same cell after photostimulation in MEC (left) or stratum pyramidale or stratum oriens of CA1 (right; HPC, hippocampus). Spike rasters and peristimulus spike histograms as in Fig. 3C. The red reference line highlights 10 ms. Note extended spike latency with hippocampal stimulation, which suggests that the cell is activated synaptically from the CA1. (C) Color-coded spike rasters, as in Fig. 4A, showing long latencies when superficial MEC cells were discharged by photostimulation in the CA1 pyramidal cell layer (all cells that responded to superficial CA1 stimulation). (D) Distribution of modal peak latencies across bins of 0.5 ms during the 30 ms after the onset of CA1 stimulation [same sample as in (C)]. Green vertical line indicates modal latency for all principal cells in Fig. 4.

In a further test of whether photostimulated MEC cells were directly activated, we determined if the responses of such cells are maintained at high-pulse frequencies where synaptically triggered discharges become unreliable (46). The predominance of 9- to 10-ms latencies was maintained at 5 and 20 Hz in all spatial and directional cells in 16 animals (Fig. 8A and fig. S19). The percentage of pulses that evoked spikes at these frequencies, from the second pulse and onward, was not lower at 5 and 20 Hz than at 1 Hz in any of the cell types (Fig. 8B). The modal spike latencies of grid cells, border cells, and head-direction cells remained unchanged across stimulation frequencies (Frequency: F2,60 = 0.32; Frequency × Cell Type: F4,60 = 0.14; repeated-measures analysis of variance).

Fig. 8 Light-induced discharge at higher stimulation frequencies.

(A) Color-coded spike rasters showing response to stimulation at 1, 5, and 20 Hz for the entire sample of cells recorded at these frequencies that were responsive at 1 Hz stimulation. Each row shows data for one cell. Time bins are 0.5 ms. Color scale to the right; color is normalized to the firing rate at 1 Hz. For 1 Hz, all light pulses are included. For 5 and 20 Hz, all light pulses except the first in each pulse train are included. (Top) Grid cells; (middle) border cells and head-direction cells; (bottom) high-rate narrow-waveform cells (putative GABAergic neurons). In all three principal cell types, light responses were maintained at the higher frequencies, with similar spike latencies, which suggests that the cells were activated directly rather than via other cells that expressed ChR2. (B) Histogram showing percentage of pulses that evoked spikes in grid cells, border cells, and head-direction cells at 1, 5, and 20 Hz photostimulation frequencies (means ± SEM). The first spike of the pulse train is not counted. Note that all functional cell types continued to respond at the highest stimulation frequency.

We also tested, in a small sample, whether firing could still be induced optically after blockade of excitatory neurotransmission in the recording area. Three rats were implanted with cannulae above the fiber-tetrode bundle in the MEC. Once light-responsive neurons were identified in MEC, a cocktail of the competitive AMPA/kainate receptor antagonist 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) and the competitive N-methyl-d-aspartate (NMDA) receptor antagonist d(–)-2-amino-5-phosphonovaleric acid (D-APV) was infused to interrupt glutamatergic neurotransmission at the recording location. In six cells, we were able to follow neural activity until firing recovered after washout of the drug (fig. S20). The infusion substantially reduced all spontaneous firing, but short-latency firing could still be evoked. Evoked activity decreased in some of the cells, as expected if glutamatergic blockade changes excitability by removing excitatory drive within the network, but the relative persistence of short-latency discharges is consistent with the interpretation that the overwhelming majority of light-responsive entorhinal principal cells were activated directly.

Finally, to confirm the functional diversity of the entorhinal direct inputs to the hippocampus, we adapted an optogenetic backfiring approach (47) for use in in vivo studies. MEC neurons with axons in the hippocampus were photostimulated either in the perforant path–projection zone within the hippocampus or in the angular bundle posterior to the hippocampus, upstream of the point where axons of MEC cells fan into the hippocampus (Fig. 2). We were able to backfire five grid cells, two border cells, four nonspatial cells, and three high-rate narrow-waveform cells in four animals (Fig. 9 and fig. S21). Latencies were 1 to 2 ms longer than with local entorhinal stimulation, as expected because of the added conduction latencies. With stimulation in the perforant-path zone, the mean firing latency (± SEM) was 12.7 ± 0.1 ms; with stimulation in the angular bundle the latency was 11.6 ± 0.1 ms. The latencies were significantly longer with stimulation in the hippocampus than in the angular bundle [t(12) = 6.5, P < 0.001].

Fig. 9 Antidromic discharge of MEC neurons after photostimulation of entorhinal projections to the hippocampus.

(A and B) Two example cells that responded to antidromic stimulation. (A) is a grid cell (left to right: trajectory with spike locations, color-coded rate map, directional rate map, and autocorrelation map); (B) is a border cell (top row, trajectory with spike locations; bottom row, rate maps). Insertion of a wall induced a firing field on one side of the wall (second trial from the left), mirroring the field at the peripheral wall on the baseline trial. Symbols as in Fig. 3, B and E. (C and D) Photoinduced antidromic discharge of the cells in (A) and (B), respectively, after photostimulation in the angular bundle (C) or the perforant path (D). Spike rasters and peristimulus spike histogram as in Fig. 3C. Note short firing latencies, only 1 to 2 ms slower than with local entorhinal stimulation and more than 4 ms faster than with synaptic stimulation from the CA1. The slightly longer delay with stimulation in the perforant path than in the angular bundle likely reflects the longer conduction distance. (E) Color-coded spike rasters showing distribution of spike latencies in all MEC cells that could be discharged by photostimulation in the perforant path–termination zone (cells 1 to 10) or the angular bundle (cells 11 to 14). Symbols as in Fig. 4A. (F) Distribution of modal peak latencies across bins of 0.5 ms during the 20 ms after the onset of antidromic stimulation [same sample as in (E)]. Green vertical line indicates modal latency for all principal cells in Fig. 4.

Discussion

We have shown that optogenetics can be combined with electrophysiological recording to identify the projection pattern of functionally defined cell types in intermingled neural populations of the mammalian spatial representation circuit. Building on previous work showing that virus can be transported retrogradely to cell bodies of projection neurons (26, 35, 46), we engineered rAAV to effectively express microbial opsins not only in the infusion area but also in the soma of cells that project to that area, such as the principal cells of the superficial layers of MEC. Hippocampus-projecting MEC cells could be identified, after expression of ChR2 transgenes, as neurons that responded reliably, at a constant minimal latency, to photostimulation near the cell bodies in the MEC or near axons of these cells in the hippocampus or the angular bundle. The response latencies of the principal cells were substantially shorter than the latencies of many putative GABAergic MEC neurons, which might have been activated synaptically via photostimulation of ChR2-expressing projection neurons. Response latencies were also shorter than after stimulation of CA1 outputs from the hippocampus. These differential time courses imply that, in experiments with minimal latencies, the discharge was caused by activation of ChR2 channels in the recorded neurons themselves.

Responsive cells were present in all functional classes of MEC cells, which suggests that the hippocampus receives direct input not only from spatial cells, like grid cells and border cells, but also from head-direction cells, nonperiodic spatial cells, and nonspatial cells. We did not observe any cell type in MEC that did not project to the hippocampus, despite the fact that many principal neurons in layer II have been reported to send axons exclusively to extrahippocampal targets such as the contralateral MEC (24). The presence of a dual spatial input, from grid cells and border cells, is consistent with the idea that place cells have access to both self-motion and landmark-based information (48) and raises the possibility that the spatial metric of the place-cell population (49, 50) originates from grid cells (17), whereas boundary and landmark-induced firing (37, 38, 51) is derived directly from border cells (18, 19, 52). The dual nature of the spatial input may account for the observation that place cells precede mature grid cells during ontogenesis of the spatial representation system (4, 5) and that place cells can maintain location specificity under conditions that reduce grid-cell periodicity in adult rats (53).

A surprising aspect of the results is that a large component of the MEC input originates from nonspatial cells. If these cells target principal cells, and not only interneurons, they likely provide a major direct-input to the place cell population, considering that nearly all active hippocampal principal cells are place cells (1, 5). The findings thus raise the possibility that individual place cells receive input from a variety of functional cell classes in the MEC. How is then the selective output of the place cells generated? Place cells may have mechanisms for gating particular external inputs (54). These may involve differences in synaptic strength, active dendritic properties, or local circuit mechanisms (22, 23, 55, 56). A similar gating mechanism may be present in layer II/III of the visual cortex, where orientation-selective neurons convert widely tuned synaptic inputs to highly specific output signals (57). Convergent input from a broad spectrum of entorhinal cell types would enable individual place cells to respond more dynamically, such that different types of input could be favored in different states or under different behavioral circumstances. This may greatly increase the representational and computational capacity of the hippocampal network.

Materials and Methods

AAV Vector Constructs

All viral constructs were generated by a polymerase chain reaction (PCR)–based amplification and cloning method. The coding region in all viral vector constructs was verified by double-stranded DNA sequencing to make sure that no shifted open reading frame and no undesired point mutation was introduced by PCR. All proviral plasmids used for packaging rAAV were flanked by AAV serotype-2 inverted terminal repeats (ITRs). rAAV2/1 vectors contained a woodchuck hepatitis virus posttranscriptional regulatory element (WPRE) and a bovine growth hormone (BGH) polyadenylation signal for enhancing transgene transcription and expression. In all viral expression cassettes, transcriptional regulation was regulated by either a hybrid cytomegalovirus early enhancer–chicken β-actin (CAG) promoter (in experiments with halorhodopsin or Arch) or a calcium–calmodulin-dependent protein kinase II α (CaMKIIα) promoter (in experiments with ChR2), which each efficiently drive transduction of rAAV2/1 in both principal cells and interneurons (58). pEYFP-Nuc (Clontech, Mountain View, California, USA) and pcNDA3-LacZ (Invitrogen, Grand Island, New York, USA) were used as PCR templates to generate rAAV-EYFP-Nuc and rAAV-LacZ, respectively. The opsin-expressing rAAV contained either the light-driven cation channel ChR2 or the inward chloride pump halorhodopsin (eNpHR 3.0) or the outward proton pump archaerhodopsin-3 (Arch). For making rAAV encoding mammalian codon-optimized opsins, pAAV-CaMKIIα-hChR2(H134R)-EYFP (gift from Karl Deisseroth, Stanford University, Stanford, California, USA), pAAV-CaMKIIα-eNpHR3.0-EYFP (gift from Karl Deisseroth, Stanford University, Stanford, California, USA) and AAV-FLEX-Arch-GFP (GFP, green fluorescent protein) (gift from Edward Boyden, MIT, Cambridge, Massachusetts, USA) were used as PCR templates to generate rAAV-CaMKIIα-ChR2, rAAV-CaMKIIα-NpHR, and rAAV- CaMKIIα-Arch, respectively. For generating trafficking-enhanced opsins, a FLAG tag was placed at the C terminus of all opsin genes between a 20–amino acid trafficking signal DYKDHDGDYKDHDIDYKDDDDK and an endoplasmic reticulum (ER)–exporting motif FCYENEV, both derived from the inward-rectifier potassium ion channel Kir2.1 and introduced to improve membrane trafficking (34, 35). The 17–amino acid N-terminal signal peptide from the β subunit of nicotinic acetylcholine receptor (nAChR) originally used for membrane insertion in eNpHR2.0 (35) was also removed. This resulted in the creation of trafficking-enhanced microbial opsins to be used for photostimulation.

rAAV Preparation and Titering

Chimeric rAAV2/1 was serotyped with AAV1 capsid proteins. The pseudotyped rAAV2/1 was prepared by cotransfection of human embryonic kidney cell line HEK293 by standard calcium phosphate transfection along with an adenoviral helper plasmid pHelper from Stratagene’s AAV Helper-Free System (La Jolla, California, USA). Twelve hours after transient transfection, the DNA/CaCl2 mixture was replaced with normal growth medium. After an additional 60 hours in culture, the transfected cells were collected and subjected to three freeze/thaw cycles. The clear supernatant was then purified using heparin affinity columns (HiTrap Heparin HP, GE Healthcare, Uppsala, Sweden). The purified rAAV2/1 was concentrated in an Amicon Ultra-4 centrifugal filter 100K device (Millipore, Billerica, Massachusetts, USA). Viral titers of all prepared rAAV2/1 were determined by real-time quantitative PCR using StepOnePlus Real-Time PCR Systems (Applied Biosystems, Foster City, California, USA) and TaqMan Universal Master Mix as previously described (59). All titered viruses were diluted and matched to 1.0 × 1012 viral genomic particles/ml by 1× phosphate-buffered saline (PBS).

Subjects

Neuronal activity was recorded from MEC in 34 male Long-Evans rats (3 to 5 months old, 350 to 450 g at surgery). The animals were housed individually in transparent Plexiglas cages (45 × 30 × 35 cm) in a temperature- and humidity-controlled vivarium 5 to 10 m from the recording rooms. All rats were maintained on a 12-hour light/12-hour dark schedule. Testing occurred in the dark phase. The rats were kept at 85 to 90% of free-feeding body weight and deprived of feed 18 to 24 hours before each training and recording trial. Body weight at the time of surgery was 350 to 400 g. Water was available ad libitum.

Surgery

On the day of surgery, the rats were anesthetized with isoflurane (induction chamber level of 5.0% with an air flow at 1400 ml/min, gradually reduced after the rats were secured in the stereotaxic apparatus to 1% isoflurane with an air flow at 1000 to 1200 ml/min). Levels of anesthesia were monitored regularly by testing toe and tail pinch reflexes. High titer–matched solution of rAAV was injected into the hippocampus over a period of 5 to 10 min at three locations within the dorsal hippocampus (AP: 4.1 mm; ML: ±2.6 mm; DV: 3.5, 2.8, and 2.1 mm), using a 10-μl NanoFil syringe (World Precision Instruments, Sarasota, Florida, USA) and a 33-gauge beveled metal needle. Injection volume (0.5 to 1 μl at each location) and flow rate (0.1 μl/min) were controlled with a Micro4 Microsyringe Pump Controller (World Precision Instruments). After injection, the needle was left in place for 10 additional minutes before it was withdrawn slowly. During the same surgical session, after hippocampal rAAV injection, the animals were implanted with tetrodes and optic fibers in the MEC and the hippocampus.

All animals were implanted with microdrives in MEC, which each were connected to an assembly of four tetrodes, cut flat at the same level, as well as a 125-μm wide optic fiber with the tip 500 μm above the tetrode tips. The tetrode-fiber assembly was implanted 0.1 to 0.5 mm in front of the transverse sinus, 4.5 to 4.7 mm from the midline, and 1.6 to 1.8 mm below the dura. Implants were oriented at an 8 to 20 degree angle in the anterior direction in the sagittal plane. Five of the animals had an additional optic fiber implanted in the ipsilateral pyramidal cell layer (or stratum oriens or alveus) of dorsal CA1 (3.8 mm behind bregma, 3.0 mm lateral to the midline, 1.8 mm below dura). Five other animals had an optic fiber in the perforant-path termination zone of the dorsal hippocampus (perforant path: 4.3 mm behind bregma, 2.4 mm lateral to the midline, ~2.6 mm below dura; angular bundle: 7.8 mm behind bregma, 4.2 mm lateral to the midline, ~2.8 mm below dura), in addition to tetrodes and optic fiber in the MEC. In three of the animals, cells were also recorded with tetrodes in CA3 in response to local stimulation in this region. Three of the animals had a 26-ga cannula (C315G; Plastics One, Roanoke, Virginia, USA) in the MEC, near the tetrode-fiber assembly. Finally, two animals with fibers and tetrodes in MEC received rAAV infusions in the cortex overlying the hippocampus.

The tetrodes were made of 17-μm polyimide-coated platinum-iridium (90 to 10%) wire. The electrode tips were platinum-plated to reduce electrode impedances to ~200 kΩ at 1 kHz. A jeweler’s screw fixed to the skull served as a ground electrode. The microdrives were secured to the skull using additional jewelers’ screws and dental cement. After surgery, the rats were allowed 1 week of recovery before handling and/or habituation to the test environments was resumed.

Training and Neuronal Recording Procedures

Tetrode turning started 2 to 3 days after viral infection and implantation, but data collection from MEC did not start until 2 to 4 weeks later. Before each recording trial, the rat rested on a towel in a large flower pot on a pedestal. The rat was connected to the recording equipment via AC-coupled unity-gain operational amplifiers close to the rat’s head, using a counterbalanced cable that allowed the animal to move freely in the pot and the recording boxes. Over the course of 10 to 20 days, the tetrodes were lowered in steps of 50 μm or less until single neurons could be isolated at appropriate depths. When the signal amplitudes exceeded ~four times the noise level (root mean square 20 to 30 μV) and the units were stable for more than 1 hour, data were recorded. After each finished set of experiments, the tetrodes were moved further until new well-separated cells were encountered.

Recorded signals were amplified 8000 to 25,000 times and band-pass filtered between 0.8 and 6.7 kHz. Triggered spikes were stored to disk at 48 kHz (50 samples per waveform, 8 bits/sample) with a 32-bit time stamp (clock rate at 96 kHz). An electroencephalogram (EEG) was recorded single-ended from one of the electrodes. The EEG was amplified 3000 to 10,000 times, low pass–filtered at 500 Hz, sampled at 4800 Hz, and stored with the unit data. By means of an overhead video camera, the recording system tracked the position of two light-emitting diodes (LEDs), one large and one small, on the head stage (sampling rate 50 Hz). The LEDs were separated by 5 to 10 cm and aligned with the body axis of the rat.

In parallel with the turning of the tetrodes, over the course of 3 to 6 weeks, the animals were trained to run around in a square black aluminum enclosure polarized by a white cue card. Running was motivated by randomly scattering crumbs of chocolate at 10- to 15-s intervals in the recording enclosure. Each trial lasted 10 or 15 min.

In rats with putative border cells, the recording trial in the square box was succeeded by a test in the same box in which a separate wall (50 cm long × 50 cm high) was inserted between the center of one of the external walls and the center of the box. This test was followed by another trial without the wall. These trials were 10 min each.

Recording of Light-Responsive Neurons

Identification of ChR2-expressing functional cell types was performed by first recording neural activity while the rat ran for 10 or 15 min in the square black aluminum enclosure. The functional identity of cells recorded during this trial was determined by estimating the spatial periodicity, the spatial information values, the directional modulation, and the proximity to borders of each cell’s neural firing pattern and then comparing these values to shuffled data from the entire population of recorded neurons, as described below.

After the recording in the square enclosure, the rat was moved back to the pot and, after 2 min of baseline recording during rest, the photostimulation started. In the main experiment, photostimulation was delivered into MEC, above the tetrode tips. The stimulation consisted of 3.5-ms 473-nm light pulses delivered repeatedly at a power density of 10 mW/mm2 for one or three periods of 2 min. All animals received stimulation at a frequency of 1 Hz for 2 min. Sixteen of the animals received additional blocks of stimulation with light pulses delivered at 5 and 20 Hz, respectively, in blocks of 5 pulses for 5 Hz and 10 pulses for 20 Hz, at intervals of 5 s for a duration of 2 min each. In a subset of experiments, photostimulation was delivered in the CA1 region of the hippocampus, in the perforant path–termination zone at a deeper location, or in the angular bundle posterior to the hippocampus. Virally transduced cells were in all cases identified as cells that fired reliably at fixed latencies in response to the photostimulation.

To validate the interpretation that the short latency–activated neurons were directly stimulated, we tested whether firing could be induced optically after blockade of excitatory neurotransmission in the recording area. Three rats were implanted with cannulae above the fiber-tetrode bundle in the MEC. Once light-responsive neurons were identified in MEC, a cocktail of the competitive AMPA/kainate receptor antagonist CNQX and the competitive NMDA receptor antagonist D-APV was infused to interrupt glutamatergic neurotransmission at the recording location. CNQX-APV cocktail (60), dissolved in sterile 0.9% saline (3 mM for CNQX and 30 mM for APV) and with final pH adjusted to 7.2, was infused via a 33-ga internal cannula (C315I; Plastics One Roanoke, Virginia, USA) connected by polyethylene tubing to a 25-μl syringe mounted in a CMA/100 infusion pump. The tip of the inner cannula protruded 0.9 mm beyond the implanted guide cannula. The infusion rate (0.1 μl/min) was controlled by the syringe pump. The total volume of each infusion was 0.5 to 2.0 μl. The inner cannula was retracted 10 min after the infusion. After the infusion, recording in the flower pot was resumed.

Spike Sorting, Cell Classification, and Rate Maps

Spike sorting was performed offline using graphical cluster-cutting software (fig. S13). Position estimates were based on tracking of one of the LEDs on the head stage. Only epochs with instantaneous running speeds of 2.5 cm/s or more were included. To characterize firing fields, the position data were sorted into 2.5-cm × 2.5-cm bins, and the path was smoothed with a 21-sample boxcar window filter (400 ms; 10 samples on each side) (11). Maps for number of spikes and time were smoothed individually using a quasi-Gaussian kernel over the surrounding 5 × 5 bins (11). Firing rates were determined by dividing spike number and time for each bin of the two smoothed maps. The peak rate was defined as the rate in the bin with the highest rate in the firing rate map.

Identification of Photoresponsive Cells

Photoexcitable cells were formally identified by comparing firing rates as a function of stimulation latency during the 100 ms after each light pulse with the firing rates obtained for similar time blocks after shuffling the spike times of each cell within an interval [–100, 100 ms] around the light stimulus. After each shuffling, spikes were counted for bins of 1 ms after the stimulation, and the three successive bins with the maximum number of spikes during the 100 ms period after the stimulation were identified. A similar block of three successive bins was identified for the real data. For each cell, the spike times were shuffled 10,000 times. Cells were classified as photoresponsive if the number of spikes in the block with maximal number of spikes in the real data exceeded the 99.9th percentile value of the distribution of number of spikes in the most active triple for the shuffled data. The latency of the response was taken as the mean latency of all spikes contributing to this 3-bin block.

Analysis of Grid Cells

The structure of the rate maps was evaluated for all cells with more than 100 spikes by calculating the spatial autocorrelation for each smoothed rate map (10). Autocorrelograms were based on Pearson’s product moment correlation coefficient with corrections for edge effects and unvisited locations. With λ(x, y) denoting the average rate of a cell at location (x, y), the autocorrelation between the fields with spatial lags of τx and τy was estimated as:r(τx,τy)=nλ(x,y)λ(xτx,yτy)λ(x,y)λ(xτx,yτy)nλ(x,y)2[λ(x,y)]2nλ(xτx,yτy)2[λ(xτx,yτy)]2where the summation is over all n pixels in λ(x, y) for which rate was estimated for both λ(x, y) and λ(x – τx, y – τy). Autocorrelations were not estimated for lags of τx, τy where n < 20.

The degree of spatial periodicity (“gridness” or “grid scores”) was determined for each recorded cell by taking a circular sample of the autocorrelogram, centered on the central peak but with the central peak excluded, and comparing rotated versions of this sample (5, 10, 11). Gridness (the cell’s grid score) was defined as the minimum difference between any of the elements in the first group and any of the elements in the second. Grid cells were defined as cells in which rotational symmetry–based grid scores exceeded the 99th percentile of a distribution of grid scores for shuffled recordings from the entire population of MEC cells. Shuffling was performed by time-shifting, for each permutation trial, the entire sequence of spikes fired by the cell along the animal’s path by a random interval between 20 s and 20 s less than the length of the trial (usually 600 – 20 s = 580 s), with the end of the trial wrapped to the beginning. The shuffling procedure was repeated 100 times for each of the 779 recordings in the photoexcitation study and each of 270 cells in the photoinhibition study, yielding a total of 77,900 and 27,000 permutations, respectively. For each permutation, a firing rate map and an autocorrelation map were constructed, and a grid score was calculated. The 99th percentile was then read out from the overall distribution of grid scores. Grid cells were defined as cells with grid scores higher than the 99th percentile of the grid scores of the distribution for the shuffled data (fig. S9A).

Analysis of Head-Direction Cells

The rat’s head direction was calculated for each tracker sample from the projection of the relative position of the two LEDs onto the horizontal plane. The directional tuning function for each cell was obtained by plotting the firing rate as a function of the rat’s directional heading, divided into bins of 3 degrees and smoothed with a 15 degrees mean window filter (two bins on each side) (4, 5, 11). In order to minimize the contribution of inhomogeneous sampling on directional tuning estimates, data were accepted only if all directional bins were covered by the animal.

The strength of directional tuning was estimated by computing the length of the mean vector for the circular distribution of firing rate (5, 11). Head direction–modulated cells were defined as cells with mean vector lengths significantly exceeding the degree of directional tuning that would be expected by chance for the MEC population. Threshold values were determined by a shuffling procedure performed in the same way as for grid cells. For each permutation trial, the entire sequence of spikes fired by the cell was time-shifted along the animal’s path by a random interval between on one side 20 s and on the other side 20 s less than the length of the trial, with the end of the trial wrapped to the beginning, a head-direction tuning curve was then constructed, and the mean vector length was calculated. The distribution of mean vector lengths was computed for the entire set of permutations from all cells in the sample (77,900 permutations for the photoexcitation study, 27,000 permutations for the photoinhibition study, and 100 permutations per cell). Cells were defined as directionally modulated if the mean vector from the recorded data was longer than the 99th percentile of mean vector lengths in the distribution generated from the shuffled data (fig. S11A).

Analysis of Border Cells

Border cells were identified by computing, for each cell, the difference between the maximal length of a wall touching upon any single firing field of the cell and the average distance of the field from the nearest wall, divided by the sum of those values (11, 12). Firing fields were defined as collections of neighboring pixels with firing rates higher than 0.3 times the cell’s peak firing rate that covered a total area of at least 200 cm2. Border scores ranged from –1 for cells with central firing fields to +1 for cells with fields that perfectly line up along at least one entire wall. Border cells were defined as cells with border scores significantly exceeding the degree of wall-related firing that would be expected for MEC cells by chance. The significance level was determined by a shuffling procedure performed for experiments in the square boxes in the same way as for grid cells and head-direction cells. For each permutation trial, the entire sequence of spikes fired by the cell was time-shifted along the animal’s path by a random interval between 20 s and 20 s less than the length of the trial, with the end of the trial wrapped to the beginning, a rate map was then constructed, and a border score was calculated. The distribution of border scores was computed for the entire set of permutations from all cells in the sample (77,900 permutations for the photoexcitation study; 27,000 permutations for the photoinhibition study; 100 permutations per cell), and the 99th percentile was determined. Cells were defined as border cells if the border score from the recorded data was higher than the 99th percentile for border scores in the distribution generated from the shuffled data (fig. S10A).

Analysis of Irregular Spatial Cells and Nonspatial Cells

Spatially modulated cells with irregular firing fields were defined as cells that did not satisfy criteria for grid cells, border cells, or head-direction cells but had spatial information values that exceeded the 99th percentile of the shuffled data (fig. S12A). Nonspatial cells were defined as cells that satisfied neither of the above criteria.

For each cell, the spatial information content in bits per spike was calculated as information content=ipiλiλlog2λiλwhere λi is the mean firing rate of a unit in the ith bin, λ is the overall mean firing rate, and pi is the probability of the animal being in the ith bin [(occupancy in the ith bin)/(total recording time)] (61).

An adaptive smoothing method (62) was used before the calculation of information scores in order to optimize the trade-off between blurring error and sampling error. The raw data were first divided into bins of 2.5 cm × 2.5 cm. Then, the firing rate at each point in the environment was estimated by expanding a circle around the point until rαinswhere r is the radius of the circle in bins, n is the number of occupancy samples within the circle, s is the total number of spikes in those occupancy samples, and the constant α is set to 10,000. With a position sampling rate of 50 Hz, the firing rate at that point was then set to 50 · s/n.

The chance level for spatial information was determined by the same random permutation procedure as used for the other cell types. Spatial cells were defined as cells in which spatial information values exceeded the 99th percentile of a distribution of such values for shuffled recordings from the entire population of MEC cells. Shuffling was performed by time-shifting, for each permutation trial, the entire sequence of spikes fired by the cell along the animal’s path by a random interval between 20 s and 20 s less than the length of the trial (usually 600 − 20 = 580 s), with the end of the trial wrapped to the beginning. The shuffling procedure was repeated 100 times for each of the 726 recordings in the photoexcitation study, yielding a total of 72,600 permutations. For each permutation, the spatial information value of the firing rate distribution was determined. The 99th percentile was then read out from the overall distribution and used as a threshold to define spatial (but nonperiodic) MEC cells.

Histology, Immunohistochemistry, and Reconstruction of Recording Positions

Electrodes were not moved after the final recording session. The rats received an overdose of sodium pentobarbital and were transcardially perfused with 0.9% saline followed by 4% formaldehyde. The perfused brain with skull and microdrive was soaked into 4% formaldehyde for 2 to 4 hours, the electrodes were turned all the way up, and the brain was extracted and stored in 4% formaldehyde. At least 24 hours later, the brains were quickly frozen and cut by a Micron Cryo-Star HM560 Cryostat (Micron International, Germany) at 30 μm in the sagittal plane. All sections around the area of the tetrode trace were collected and mounted on glass. For every pair of consecutive sections, the first was stained with cresyl violet (Nissl) and the second was assigned for staining by antibodies. The latter group of sections was divided into six interleaved sets by placing the sections sequentially into a row of 6 wells in a 24-well plate containing 1× PBS, such that each section set could be used against different antibodies (FLAG, NeuN, or GFP).

For the immunostaining, sections were rinsed three times for 10 min in 1× PBS at room temperature and preincubated for 2 hours in 10% normal goat serum in PBST (1× PBS with 0.5% Triton X-100). All rinses between incubation steps were with PBST. After rinsing, processed sections were incubated with different primary antibodies against FLAG (Sigma, 1:2000), NeuN (Millipore, 1:500), GFP (Clontech, 1:2000), c-Fos (Calbiochem, 1:2000), and glial fibrillary acidic protein (GFAP) (Sigma, 1:1000) for 72 hours in antibody-blocking buffer at 4°C. After three times of 15-min washing in 1× PBST at room temperature, sections were incubated either in a mouse antibody directed against a donkey secondary antibody or a donkey antibody directed against a rabbit secondary antibody conjugated with either fluorescein isothiocyanate or Cy3, respectively (Jackson ImmunoResearch, West Grove, Pennsylvania,  USA, 1:2000), for 2 hours at room temperature. After intensive rinse with 1× PBST, sections were mounted onto glass slides with 4′,6′-diamidino-2-phenylindole (DAPI)–containing Vectashield mounting medium (Vector Laboratories, Burlingame, California, USA), and a cover slip was applied. Expression of microbial opsins was estimated by anti-FLAG antibody, and the C-terminal FLAG tag was fused with the opsin in the viral construct. NeuN was used for staining neuronal cell types only. GFP was used for rAAV-EYFP-Nuc infected neurons.

The positions of the tips of the recording electrodes were determined from digital pictures of the Nissl-stained sections. Scanning and measurements were made using MIRAX MIDI software (Carl Zeiss, Germany). A shrinkage coefficient was calculated by measuring the distance (on the digital image) from the surface of the brain to the tips of the recording electrodes and then dividing this by the final depth of the electrodes, read out from the turning protocol. Only recordings obtained in the superficial layers of MEC or at the MEC-parasubiculum border were analyzed.

For rats infused with rAAV-LacZ, β-galactosidase activity was also carried out in parallel with anti-FLAG immunostaining. Briefly, sections were first fixed for 2 hours with 4% paraformaldehyde in 1× PBS and then washed with 100 mM phosphate buffer (pH 7.4), 2 mM MgCl2, and 5 mM EGTA (pH 8.0) for four times of 15 min each at room temperature. To develop blue precipitates, sections were then immersed into X-gal staining solution with 100 mM phosphate buffer (pH 7.4), 2 mM MgCl2, 0.01% sodium desoxycholate, 0.02% Nonidet P-40, 5 mM potassium-ferricyanide, 5 mM potassium-ferrocyanide, and 0.5 mg/ml of X-gal in the dark at 37°C with a humidified chamber under gentle shaking. Once the blue color was developed to a maximum, X-gal staining was stopped by rinsing sections with 1× PBS.

Supplementary Materials

http://www.sciencemag.org/content/340/6128/1232627/suppl/DC1

Figs. S1 to S21

References (6567)

References and Notes

  1. AAV2 is a commonly used serotype for transgene delivery to restricted brain regions but its transport efficiency is limited (26, 64). We resolved this drawback by cross-packaging rAAV2 with viral capsids from other AAV serotypes to generate hybrid species such as rAAV2/1 (59, 64).
  2. EYFP was expressed under the control of a hybrid CAG promoter. Three tandem repeats of the nuclear localization signal (NLS) from the simian virus 40 large T antigen were added to the C terminus of EYFP to facilitate translocation of EYFP into the nucleus of infected neurons.
  3. To increase rAAV packaging capacity and to avoid bleaching during in vivo photostimulation, EYFP was replaced with a nonfluorescent FLAG tag. FLAG was placed between a 20–amino acid trafficking signal and an ER-exporting motif, both derived from the inward-rectifier potassium ion channel Kir2.1 and introduced to improve membrane trafficking (35).
  4. Acknowledgments: The work was supported by an Advanced Investigator Grant from the European Research Council (CIRCUIT grant agreement no. 232608), the Louis-Jeantet Prize for Medicine, the Kavli Foundation, and the Centre of Excellence scheme of the Research Council of Norway. We thank M. P. Witter for discussion; A. M. Amundsgård, A. Burøy, K. Haugen, K. Jenssen, E. Kråkvik, R. Skjerpeng, H. Waade, J. Wu, and Q. Zhang for technical assistance; K. Deisseroth for providing ChR2-H134R and eNpHR3.0 plasmids; and E. Boyden for providing Arch plasmid. S.-J.Z., J.Y., M.-B.M., and E.I.M. planned and designed the study; S.-J.Z. designed and performed all molecular biology, modified and made all viruses, and performed all surgeries; J.Y., C.M., I.C., S.-J.Z., and D.L. performed experimental recordings; J.Y. performed analyses and made figures; A.T., C.M., and J.Y. designed the laser system; A.T. made stimulation and shuffle programs; M.-B.M. and E.I.M. supervised experiments and analyses; and E.I.M. wrote the paper with input from all authors. The molecular biology sections were written by S.-J.Z, with some support from E.I.M. All authors discussed analyses and results.
View Abstract

Stay Connected to Science

Navigate This Article