Oscillatory activity in the monkey hippocampus during visual ...

Published in final edited form in: Proceedings of the National Academy of Sciences of the United States of America 110(32), 13144-13149. (2013) doi: 10.1073/pnas.1302351110

Oscillatory activity in the monkey hippocampus during visual exploration and memory formation Michael J. Jutrasa,1, Pascal Friesb, and Elizabeth A. Buffaloa,c,1,2 a

Yerkes National Primate Research Center, 954 Gatewood Road, Atlanta, GA 30329, USA. Ernst Strüngmann Institute (ESI) for Neuroscience in Cooperation with Max Planck Society, 60528 Frankfurt, Germany. c Department of Neurology, Emory University School of Medicine, 1440 Clifton Road, Atlanta, GA, 30322, USA. b

1

Present address: Department of Physiology and Biophysics, University of Washington, Seattle, WA 98195

2

Email address for correspondence: [email protected] Elizabeth Buffalo Yerkes National Primate Research Center 954 Gatewood Road NE Atlanta, GA 30329 Email: [email protected] Voice: 404-712-9431 Fax: 404-727-9294 Abstract Primates explore the visual world through the use of saccadic eye movements. Neuronal activity in the hippocampus, a structure known to be essential for memory, is modulated by this saccadic activity, but the relationship between visual exploration through saccades and memory formation is not well understood. Here, we identify a link between theta-band (3-12 Hz) oscillatory activity in the hippocampus and saccadic activity in monkeys performing a recognition memory task. As monkeys freely explored novel images, saccades produced a theta-band phase reset, and the reliability of this phase reset was predictive of subsequent recognition. In addition, enhanced theta-band power prior to stimulus onset predicted stronger stimulus encoding. Together, these data suggest that hippocampal theta-band oscillations act in concert with active exploration in the primate and possibly serve to establish the optimal conditions for stimulus encoding.

Jutras, M. J., Fries, P., Buffalo, E. A. (2013) Oscillatory activity in the monkey hippocampus during visual exploration and memory formation. Proceedings of the National Academy of Sciences of the United States of America 110(32), 13144-13149. (2013) doi: 10.1073/pnas.1302351110

Introduction The use of saccadic eye movements to acquire information about the surrounding environment is perhaps the most conspicuous example of exploratory behavior in the primate. This behavior provides a mechanism for parsing incoming information into discrete, stable segments, i.e., snapshots of individual elements comprising a complex visual scene, allowing time for sufficient processing to occur before moving to the next fixation target. This mechanism of actively sampling sensory information from the environment may be similar to the behaviors engaged in by rodents exploring their environment through such activities as sniffing and whisking. Specifically, the fixation period following each saccade may be homologous to the period of incoming sensory information accompanying each sniff cycle in the rodent (1). Recently, it has been suggested that motor behaviors associated with information gathering are integral to the “active sensing” process in natural behavior (2). It is plausible that there may exist certain common neuronal elements across species that are associated with active sensing processes, such that the neuronal mechanisms underlying the encoding of information are intimately connected with the motor activities involved in acquiring that information. In many mammalian species, voluntary, exploratory behaviors are often associated with theta-band activity, a prominent 3-12 Hz oscillatory activity in the hippocampus and other limbic structures. This activity has been studied extensively in the rodent hippocampus (3-5), but has also been described in bats (6), cats (7), and, more recently, humans (8-11). In rodents, theta appears to show close temporal relationships with running (3, 12) and sniffing (13), suggesting an association between theta and the rate of sensory input. While hippocampal theta has been identified in anesthetized monkeys (14), the lack of a clear demonstration of hippocampal theta in awake monkeys has been attributed to the fact that the recording methods typically require immobile, headaffixed monkeys, in contrast to rodent studies using freely-moving animals (15). However, it is possible that exploration through saccadic eye movements may approximate the sensory acquisition that rodents engage in as they explore their physical environment. Accordingly, we examined hippocampal activity related to visual exploration and memory formation in monkeys performing a free-viewing visual recognition memory task. Results Behavioral Results We recorded hippocampal activity in two rhesus monkeys performing the Visual Preferential Looking Task (VPLT) (16-18) (Fig. 1A). In each recording session, monkeys were presented with two-hundred novel complex stimuli (11° x 11° in size) on a computer screen. Each stimulus was presented twice during a given session, with up to 8 intervening stimuli between successive presentations. The monkeys’ eye movements were measured and each stimulus remained on the screen until the monkey’s gaze moved off the stimulus or for a maximum of 5 seconds. Fig. 1B depicts an example of a monkey’s eye movements during the first (“novel”, yellow trace) and second (“repeat”, blue trace) presentation of a stimulus. Monkeys demonstrated recognition memory by spending less time exploring the stimulus when it was repeated compared to when it was novel. Across 45 sessions, the monkeys demonstrated robust recognition memory performance. There was a significant decrease in looking time for the repeated presentation (average looking times for novel and repeat trials were 2.7 s and 0.8 s, respectively; paired t test, p < 0.001; Fig. 1C). To control for varying interest in individual stimuli, recognition memory performance was calculated as the 2


difference in looking time between presentations expressed as a percentage of the amount of time the monkey spent looking at the first presentation of each stimulus. The median reduction in looking time was 70.7% (67.3% in Monkey A and 72.8% in Monkey B; Fig. 1D). Similar tasks which examine recognition memory through preference for novelty have been shown to depend on the integrity of the hippocampus in rodents, monkeys, and humans (19-22). Because the task involves minimal training and the monkeys are not explicitly rewarded for viewing the pictures, the behavior thus measured may approximate natural exploratory activity.

Figure 1 Task, behavioral results, and theta-band oscillations in the hippocampus. (A) VPLT design. (B) Representative example of one monkey’s scan path showing that the monkey spent more time looking at the image when it was novel (yellow) compared with when it was repeated (blue). Circles represent points of fixation between saccades, with the size of each circle proportional to the duration of the fixation period. (C) Percentage of change in looking time for all stimuli (gray, monkey A; white, monkey B). (D) Perievent histograms for two example units showing the cumulative firing activity across fixation periods during stimulus exploration. Error bars represent SEM. (E) Power spectra for two example LFP channels across all VPLT blocks showing peaks around 8–11 Hz. (F) Examples of theta bouts during VPLT performance. For each example, the theta bout in the raw LFP is marked by a gray square and the power spectrogram of the LFP signal is presented below. (G) Autocorrelograms (Autocorr.) of each example LFP shown in F during the theta bout.

Saccadic modulation of hippocampal firing rates We recorded single unit activity and local field potentials (LFPs) from a total of 114 locations in the hippocampal formation of two rhesus monkeys (58 from Monkey A and 56 from Monkey B,

3


primarily in the anterior hippocampal formation; recording locations are provided in ref. 18). Many units exhibited clear modulation of firing activity surrounding each saccade (Fig. 1E). Because this response was variable in time across units and was often biphasic in nature, we tested this modulation statistically using a template matching procedure (23), which allowed us to examine the consistency of the firing response pattern following each saccade. Based on this analysis, the spike patterns of 58 out of 131 single units (44%) were predictable across individual saccades to a degree that was significantly greater than chance (test of proportions, p < 0.01). This proportion is in agreement with previously reported saccadic modulation in the monkey medial temporal lobe (MTL) (23, 24). In addition, we compared the incidence of saccade-modulated neurons with neurons that were significantly responsive to the onset of visual stimuli (16), and found no correlation between the two distributions (Spearman’s rank correlation coefficient, ρ = 0.09, p > 0.1), indicating that the tendency of neurons to show saccade modulation was not simply a reflection of visual sensitivity. Theta-band activity in hippocampal LFPs Power spectral analysis of hippocampal LFPs during VPLT blocks revealed the presence of peaks throughout the theta frequency range (3-12 Hz), with the most prominent of these peaks occurring between 8 and 11 Hz (Figs. 1F and S1). The median saccade rate (5.1 Hz; Fig. S2) was outside the range of the peak hippocampal theta power observed in these data, yet was within the 3-12 Hz theta band. Visual inspection of LFP signals revealed that unlike the long, continuous trains of oscillations seen in the rodent hippocampus during “active states”, oscillatory activity occurred in intermittent bouts separated by desynchronized, non-theta activity, similar to the activity previously described in the human hippocampus (8). Previous studies have evaluated the incidence of oscillatory components in continuous LFP data using the Pepisode measure (8, 25-27), which detects episodes of oscillatory activity exceeding amplitude and duration thresholds at each frequency of interest while at the same time ignoring the transient voltage fluctuations that may accompany artifacts or evoked potentials. We used this method to detect bouts of oscillatory activity in the theta frequency range that exceeded 3 cycles in duration and determined the incidence of these bouts across blocks of the VPLT. Four example bouts are depicted in Fig. 1F (Upper). The regular rhythmicity of these example oscillatory episodes was confirmed by calculating autocorrelograms of the LFP contained within each epoch, which revealed prominent oscillations in the 8-11 Hz range (Fig. 1G). Bouts of theta oscillations occurred with a median bout duration of 508 ms and a median inter-bout interval of 1194 ms (Fig. S3). The average theta Pepisode across LFPs, taken across all components of the VPLT, was 0.24 ± 0.01. Similar measures, albeit with a lower incidence of bouts, were obtained from data recorded during a calibration task that monkeys performed during alternate blocks with the VPLT (see SI Methods). As an additional measure to assess the prevalence of detected oscillations in particular frequency bands, the mean Pepisode of each LFP at each frequency was evaluated against the background value of 0.05, or 5%, using a two-tailed t-test [p < 0.0014, after applying the Bonferroni correction; (25)]. The resulting distribution reveals a majority of significantly high Pepisode values across LFPs falling within the range of 8-11 Hz (Fig. S4), indicating that bouts of oscillatory activity are most prevalent within this frequency band across hippocampal LFPs. Saccade-triggered phase reset of hippocampal theta Previous work in rodents and humans has shown that ongoing theta oscillations undergo a phase reset during salient events (e.g. stimulus onset) in certain memory tasks (28-30), suggesting that 4


phase-locking of hippocampal theta to incoming sensory information is an important element of stimulus processing during encoding. Only neural events for which there is an increase in phase concentration without a concurrent change in power at the dominant frequency meet the criteria for a “true” phase reset (31-33). The former requirement may be assessed through multiple means, including the observation of oscillatory activity in the post-event signal after averaging across multiple events (i.e. ref. 28), or through the calculation of phase concentration in the form of the Rayleigh statistic, after deriving the phase at the frequency of interest using the Hilbert transform (i.e. ref. 34). To assess the relationship between theta oscillations and saccades, we first examined hippocampal LFPs around the saccade to the center fixation cross that was presented at the start of each VPLT trial. Only trials in which no saccades occurred within 600 ms before, or 1000 ms after, the saccade to the center cross were included (see SI Methods). These trials were divided into “presaccade” and “postsaccade” periods, with the former consisting of the 600 ms period leading up to saccade onset, and the latter consisting of 600 ms following a 400 ms “buffer” period immediately following saccade onset. The buffer period was included to account for any visually-evoked voltage response driven by the appearance of the center cross. To assess the degree of pre- and postsaccade phase consistency across trials, we examined LFP signals in two ways: at the single trial level (Fig. 2A, Upper) and after calculating the trial-averaged LFP (Fig. 2A, Lower). We observed an increased tendency of trial-averaged LFPs to display oscillatory activity in the postsaccade period compared to the presaccade period, similar to the increased post-stimulus oscillatory activity seen in trialaveraged LFPs in the rodent hippocampus (28). Power spectra were calculated for the pre- and poststimulus periods, for single-trial LFP signals and for the trial-averaged LFPs. Theta power in individual trials was not significantly different postsaccade compared to presaccade (Fig. 2B, Upper), indicating that each individual saccade did not evoke a change in theta power in the LFP. However, theta power was significantly higher in the postsaccade period than the presaccade period in the trial-averaged signal (Fig. 2B, Lower; p < 0.01). This postsaccade increase in theta-band activity in the averaged LFP, but not the LFP taken from individual trials, suggests that ongoing theta oscillations in the primate hippocampus undergo a reset to a consistent phase following saccadic eye movements, producing event-locked oscillatory patterns in the average signal postsaccade. The prolonged time-course of this effect (which outlasts a typical evoked response that may be present in the trial-averaged LFP signal) further argues in favor of a phase reset of ongoing oscillatory activity, rather than an additive evoked response. Further evidence supporting a phase reset after saccade onset comes from the calculation of the Rayleigh statistic (35), which compares the distribution of phases across fixation periods against a uniform (random) distribution at each time-point relative to saccade onset. Average Rayleigh statistic values for theta-filtered (6-12 Hz) LFPs were higher during the postsaccade period than the presaccade period (Fig. S5; p < 0.01), indicating a higher degree of phase concentration during the fixation period following saccade onset. A similar analysis performed on LFP data taken from the calibration task did not show similar saccade-related changes in power and phase (Fig. S6 and SI Methods), indicating that this effect was specific to saccades made in anticipation of an exploratory state.

5


Figure 2 Saccade-triggered phase reset of hippocampal theta. (A) Example LFP traces (Upper) and trial-averaged LFP (Lower) aligned to saccade onset during the prestimulus fixation period. Theta oscillations show phase alignment across trials during the last 600 ms of the fixation period (represented by the green scale bar), which translates to visible theta oscillations in the trial-averaged LFP during the same time period. (Scale bar = 600 ms.) (B) Theta (6.7–11.6 Hz) power calculated across all single trials (Upper) and for each trial-averaged signal (Lower) for presaccade (red) and postsaccade (green) periods. Theta power was significantly higher for the postsaccade period than for the presaccade period for the trial-averaged signals (*p < 0.05) but not for the individual trial signals (p > 0.1).

Modulation of the phase of theta may have important implications for cognition, as demonstrated in some studies of hippocampal long-term potentiation (36-38) and sensory processing (34, 39). For example, neural activity in the hippocampus may be influenced by theta phase in such a way that resetting to an “ideal” phase upon saccade initiation may set up the optimal conditions for plasticity, and thus, memory formation. Accordingly, we next considered whether there was any relationship between theta phase reset after saccades and memory formation as measured in the VPLT. Due to the relatively short duration of fixation periods during naturalistic viewing compared to cued fixation, we focused on the 200 ms period immediately following each saccade the monkeys made as they explored the visual stimuli during novel trials. We limited our analysis to novel trials in which the monkey made at least 3 saccades during exploration of the image, with all resultant postsaccade fixation periods lasting at least 200 ms. LFPs were filtered in the 3-12 Hz theta band and the time-resolved phase of the filtered signal was calculated using the Hilbert transform. This phase distribution was averaged within each LFP channel for each of two memory conditions: the 30 stimuli for which the monkey showed the best subsequent recognition memory (high recognition, i.e., largest percent reduction in looking time for repeated stimuli) and the 30 stimuli for which the monkey showed the worst subsequent recognition memory (low recognition). These trials represented, on average, the top and bottom 30.8% of all trials within a recording session that met the inclusion criteria. While both conditions showed similar latencies to the first saccade and 6


intersaccade intervals, high recognition trials were significantly longer, and thus contained more saccades, than low recognition trials (Table S1); therefore, only the first three postsaccade fixation periods were analyzed for each trial. The phase distribution for both conditions showed a high degree of organization during the postsaccade period; in other words, at any given time point, the phase was highly predictable across fixation periods (Fig. 3B). This change in oscillatory phase was not accompanied by a concomitant increase in power (Fig. S7), supporting the hypothesis that this shift in phase at saccade onset represents a phase reset (31-33), similar to that seen with pre-stimulus saccades. However, the reliability of theta phase following saccade that occurred during image viewing was variable across memory conditions, with high recognition trials showing greater postsaccade phase specificity than low recognition trials (Figs. 3A & C). Rayleigh statistic values for each LFP tended to be higher during the postsaccade fixation period in the high recognition condition than in the low recognition condition (Fig. 3D). For each LFP, we calculated the difference between memory conditions of the average Rayleigh statistic value in the 40-200 ms time window after saccade onset; the distribution of differences across LFPs was significantly greater than zero (sign test, p < 0.01; Fig. 3E). This time window of significant modulation by memory condition represents times at which the LFP theta phase was significantly more consistent, or predictable, during the encoding of subsequently well-remembered stimuli, than for poorly remembered stimuli. These data suggest that hippocampal theta phase resetting by saccades may contribute to memory formation.

Figure 3 Postsaccade phase reliability during novel image exploration predicts subsequent recognition. (A) Phase distribution of an example theta-filtered LFP signal at 120 ms after saccade onset for the first three saccades in all high recognition (red) and Low recognition (blue) conditions, in which 0° was defined as the trough of a theta cycle. (B) Raw (red) and theta (3–12 Hz)-filtered (blue) segments from an example LFP

7


showing reset to a consistent phase following saccade onset. (C) Phase concentration for the 400-ms period centered on saccade onset for high recognition and low recognition conditions for the example LFP in A. White dashed lines mark the time point of the phase distribution shown in A. (D) Rayleigh statistic as a function of time relative to saccade onset for high recognition and low recognition conditions for the example LFP in A. The black solid line represents the threshold for significant (p < 0.01) deviation from a uniform phase distribution. (E) Rayleigh statistic values for the high recognition (x axis) and low recognition (y axis) trials for all LFPs (n = 74). Each point represents the average Rayleigh statistic value in the 40- to 200-ms time window after saccade onset across all fixation periods analyzed (n = 90 in each condition). Rayleigh statistic values were significantly higher for high recognition than low recognition trials across LFPs (paired t test, p < 0.01).

Pre-stimulus theta-band power and memory We next considered whether there was any relationship between modulations in theta-band power and recognition memory performance on the VPLT. In view of previous studies showing positive correlations between pre-stimulus theta in the human MTL and memory (40-42), we examined the possibility that the strength of theta-band activity during the baseline period leading up to stimulus onset might predict the strength of memory encoding. We calculated the power in the 3-20 Hz range for each LFP during the 800 ms prior to novel presentations of the 30 stimuli for which the monkey showed the best subsequent recognition memory (high recognition) and prior to novel presentations of the 30 stimuli for which the monkey showed the worst subsequent recognition memory (low recognition). These trials represented the top and bottom 22.2% of trials in terms of memory performance. In the representative example LFP power spectra shown in Fig. 4A and across the population (Fig. 4B), there was an enhancement in pre-stimulus theta power for trials with the best subsequent memory performance compared to trials with the worst subsequent memory performance. A non-parametric permutation test (43) revealed a cluster of significant memory modulation starting approximately 500 ms before stimulus onset, initially restricted within a 8-10 Hz band but gradually encompassing a larger frequency range (6-12 Hz) approaching stimulus onset. Due to the temporal smoothing inherent in this method, a post-stimulus onset influence cannot be ruled out in the later part of this effect; therefore, the time-averaged power at 9 Hz was calculated for 444-ms windows immediately preceding and immediately following stimulus onset. In both windows, power was significantly higher for the high recognition condition than the low recognition condition (p < 0.01; Fig. 4C). This result indicates that not only is there a true correlation between theta power immediately preceding stimulus onset and memory performance, but that this pre-stimulus effect persists into at least the initial period of the stimulus presentation.

8


Figure 4 Theta-band power preceding stimulus onset for novel stimuli is predictive of subsequent recognition. (A) Power spectrograms from one example LFP during the 800ms time period immediately preceding stimulus presentation. (B) Modulation of theta power for high recognition and low recognition trials across 114 LFPs. The area of significant power modulation across conditions is outlined in black. (C) Power in the 9-Hz band for the 444-ms window preceding (Left) and following (Right) stimulus onset, for high recognition (red) and low recognition (blue) trials. Power was significantly higher for high recognition than for low recognition in both time windows (paired t test, *p < 0.01).

Discussion In a naturalistic task of recognition memory through free-viewing, we found that single neurons and LFPs in the primate hippocampus were modulated by saccades during visual exploration. Further, saccades produced a phase-resetting of theta activity in the hippocampal LFP. Importantly, we observed significant effects on memory of pre-stimulus theta power and postsaccade theta phase variability during visual exploration. Together, these data suggest that modulations in hippocampal theta during exploratory behaviors may be important for memory storage in the primate hippocampus. Because the timing of neuronal spikes in the hippocampus is influenced by theta oscillations (9, 44-46), theta phase-resetting could represent a mechanism by which stimulus-related activity in the hippocampus, in the form of neuronal firing, may be coordinated with ongoing behavior to optimize memory formation. Previous studies of saccadic modulation in the primate hippocampus have drawn parallels between this modulation and theta activity in rodents (24, 47). Saccadic eye movements partition sensory input into discrete elements in much the same way that whisking and sniffing, behaviors often associated with hippocampal theta in rodents, break up somatosensory and olfactory information, respectively (1). Furthermore, the modulation of hippocampal activity by saccadic eye movements is reminiscent of previous findings showing that theta in rodents becomes phasesynchronized with whisking and sniffing during exploration (13, 48). While this relationship may

9


exist only during certain behavioral states or only during the presence of sensory input (49), it suggests that one function of hippocampal theta across species may be as a mechanism to promote neural processing of information gathered during the performance of “sensing” behaviors. Our finding that hippocampal theta undergoes a phase reset with saccade onset is similar to the phase-resetting observed in both rodent (28, 29) and human (30, 50, 51) hippocampus with stimulus events during performance of memory tasks. This phase-resetting has several important implications for memory formation. For example, the induction of long-term potentiation in the hippocampus is preferential for particular phases of theta (36-38, 52, 53). Thus, phase-resetting of the hippocampal theta rhythm may represent a mechanism to ensure that the hippocampus is in an “ideal phase” to integrate incoming sensory information into memory via synaptic plasticity. Additionally, hippocampal theta may be important for organizing interactions between disparate brain regions. Neurons in multiple brain regions, including prefrontal cortex, are phase-locked to the hippocampal theta rhythm (54-56), and the amplitude of locally-generated gamma-band (30-140 Hz) oscillations is also modulated by hippocampal theta phase (55, 57, 58). This latter phenomenon may be particularly important for memory, as gamma synchrony within and between MTL regions is enhanced during successful memory task performance (18, 59, 60). The degree to which interactions between brain regions are modulated by hippocampal theta phase may thus represent a mechanism by which theta organizes the routing of information within and between functional networks during memory formation and retrieval (57, 61, 62). It has also been suggested that theta serves as a “framesetting” mechanism, where the theta cycle provides a reference in which multiple items are represented at different phases of the cycle (63, 64). In view of these functional models of hippocampal theta, phase-resetting may represent a mechanism to temporally align mnemonic processing with potentially significant behavioral events (e.g., saccade/stimulus onset). We also observed an effect of pre-stimulus theta-band power on memory, such that higher theta power preceding novel stimulus presentation, and persisting through at least the initial period of stimulus presentation, predicted better subsequent recognition. This result is consistent with studies in human epilepsy patients showing positive correlations between pre-stimulus theta in the MTL and memory (40-42) and suggests that theta oscillations are involved in “preparatory” mechanisms that bring the hippocampus into an “online” state (5). Because the monkeys in the current study were not explicitly trained to form a memory of the stimuli, it is unlikely that this preparatory state reflects an overt intention, or motivation, to remember the stimulus. An alternative explanation is that this activity is a component of an attentional process related to the anticipation of a novel stimulus, and that this anticipation is conducive to processes involved in memory storage (65). It was recently demonstrated that target detection performance fluctuates rhythmically within the theta-band frequency range (66), providing evidence that covert attentional processes, like some exploratory behaviors, are rhythmic in nature and possibly synchronize with oscillatory neural patterns. In summary, these results provide a glimpse into a possible mechanism for memory encoding in the hippocampus during active exploration. This mechanism may involve linking saccadic behavior and hippocampal activity in such a way that the former establishes the optimal conditions for synaptic plasticity. The present results suggest that encoding of sensory information in the hippocampus is closely linked to the sampling routines utilized in active sensing, and that the precise timing of hippocampal activity relative to behavioral events affects the strength of memory formation.

10


Methods Behavioral testing and data acquisition. Procedures used for behavioral testing and electrophysiological recording are described in detail in ref. 18 and SI Methods. Neuronal recordings were carried out in two adult male rhesus monkeys (Macaca mulatta). Monkeys were tested on the Visual Preferential Looking Task (VPLT) while their eye movements were recorded using a non-invasive infrared eye-tracking system (ISCAN, Burlington, Massachusetts). The monkey initiated each trial by fixating a white cross (1°) at the center of a computer screen for 1 s. Following this, the cross disappeared and a picture stimulus (11°) was presented and remained on the screen as long as the monkey continued to look at it, up to a maximum looking time of 5 seconds. The VPLT was given in 51 daily blocks of 6, 8, or 10 trials each, chosen pseudorandomly. LFP and single unit recordings were obtained from tungsten microelectrodes placed in the anterior part of the left hippocampus, in the CA3 field, dentate gyrus, and subiculum (see ref. 18 for recording locations). Data amplification, filtering, and acquisition were performed with a Multichannel Acquisition Processor (MAP) system from Plexon Inc. (Dallas, TX). The neural signal was split to separately extract the spike and the LFP components. For spike recordings, the signals were filtered from 250 Hz – 8 kHz, further amplified and digitized at 40 kHz. For LFP recordings, the signals were filtered with a passband of 0.7-170 Hz, further amplified and digitized at 1 kHz; any additional filtering was performed in Matlab. Eye movement data were digitized and stored with a 240 Hz resolution. Data Analysis. All analyses were performed using custom programming in Matlab and using FieldTrip (fieldtrip.fcdonders.nl). LFP spectral analysis, bout detection, and phase concentration. We calculated power spectra from LFP data after first limiting data segments to those recorded during blocks of VPLT performance, including intertrial intervals, pre-stimulus fixation periods, and stimulus presentation/exploration periods. Initial spectral analysis was performed using the multitaper method (67, 68) on non-overlapping LFP segments of 5 seconds each. Bouts of theta activity in LFP data were quantified using an oscillatory episode detection algorithm that estimates the background power spectrum of the LFP in order to determine power and duration criteria; this method is described in detail in refs. 8, 25-27 and SI Methods. Pepisode measures were calculated from LFP segments taken from blocks of VPLT performance, with each segment consisting of a block in its entirety. To calculate theta power for pre- and postsaccade periods: for individual trial power measures, we multiplied the raw LFP traces for each pre- and postsaccade period with 1 orthogonal taper function before Fourier transformation, providing spectral smoothing of ±1.67 Hz (for 600-ms segments taken from the pre-stimulus period or the calibration task) or ±4 Hz (for 200-ms segments taken from the stimulus viewing period). For trial-averaged power measures, the same method was used, with the exception that before taper multiplication and Fourier transformation, we calculated the average LFP signal (i.e., evoked signal) across trials. Taper multiplication and Fourier transformation were then applied to each trial-averaged LFP segment. We then calculated the average power in the theta frequency band (6.7-11.6 Hz for pre-stimulus and calibration task data and 4-12 Hz for data taken from the stimulus viewing period) to obtain a theta power value for each LFP for presaccade and postsaccade periods. 11


Acknowledgments We thank Megan Jutras for technical assistance. Supported by the Emory Alzheimer's Disease Research Center grant AG025688 (E.A.B.), the National Institute of General Medical Sciences and National Institute of Neurological Disorders and Stroke (M.J.J.), grants from the National Institute of Mental Health, MH080007 (E.A.B.), MH093807 (E.A.B.), and MH082559 (M.J.J.), and the National Center for Research Resources P51RR165 (currently the Office of Research Infrastructure Programs/OD P51OD11132). References 1. Kepecs A, Uchida N, & Mainen ZF (2006) The sniff as a unit of olfactory processing. Chem Senses 31(2):167-179. 2. Schroeder CE, Wilson DA, Radman T, Scharfman H, & Lakatos P (2010) Dynamics of Active Sensing and perceptual selection. Curr. Opin. Neurobiol. 20(2):172-176. 3. Vanderwolf CH (1969) Hippocampal electrical activity and voluntary movement in the rat. Electroencephalogr Clin Neurophysiol 26(4):407-418. 4. Winson J (1974) Patterns of hippocampal theta rhythm in the freely moving rat. Electroencephalogr Clin Neurophysiol 36:291-301. 5. Buzsáki G (2002) Theta Oscillations in the Hippocampus. Neuron 33(3):325-340. 6. Ulanovsky N & Moss CF (2007) Hippocampal cellular and network activity in freely moving echolocating bats. Nat Neurosci 10(2):224-233. 7. Grastyán E, Lissák K, Madarász I, & Donhoffer H (1959) Hippocampal electrical activity during the development of conditioned reflexes. Electroencephalogr Clin Neurophysiol 11(3):409-430. 8. Ekstrom AD, et al. (2005) Human hippocampal theta activity during virtual navigation. Hippocampus 15(7):881-889. 9. Rutishauser U, Ross IB, Mamelak AN, & Schuman EM (2010) Human memory strength is predicted by theta-frequency phase-locking of single neurons. Nature 464(7290):903-907. 10. Lega BC, Jacobs J, & Kahana M (2011) Human hippocampal theta oscillations and the formation of episodic memories. Hippocampus 22(4):748-761. 11. Cantero JL, et al. (2003) Sleep-dependent theta oscillations in the human hippocampus and neocortex. J Neurosci 23(34):10897-10903. 12. McFarland WL, Teitelbaum H, & Hedges EK (1975) Relationship between hippocampal theta activity and running speed in the rat. J. Comp. Physiol. Psychol. 88(1):324-328. 13. Macrides F, Eichenbaum HB, & Forbes WB (1982) Temporal relationship between sniffing and the limbic theta rhythm during odor discrimination reversal learning. J Neurosci 2(12):1705-1717. 14. Stewart M & Fox SE (1991) Hippocampal theta activity in monkeys. Brain Res 538(1):59-63. 15. Skaggs WE, et al. (2007) EEG Sharp Waves and Sparse Ensemble Unit Activity in the Macaque Hippocampus. J Neurophysiol 98(2):898-910. 16. Jutras MJ & Buffalo EA (2010) Recognition memory signals in the macaque hippocampus. Proc Natl Acad Sci U S A 107(1):401-406. 17. Killian NJ, Jutras MJ, & Buffalo EA (2012) A map of visual space in the primate entorhinal cortex. Nature 491(7426):761-764. 18. Jutras MJ, Fries P, & Buffalo EA (2009) Gamma-band synchronization in the macaque hippocampus and memory formation. J Neurosci 29(40):12521-12531. 19. Zola SM, et al. (2000) Impaired recognition memory in monkeys after damage limited to the hippocampal region. J Neurosci 20(1):451-463.

12


20.

21. 22. 23.

24.

25.

26.

27. 28. 29. 30.

31. 32.

33. 34. 35. 36. 37. 38.

39.

Nemanic S, Alvarado MC, & Bachevalier J (2004) The hippocampal/parahippocampal regions and recognition memory: insights from visual paired comparison versus objectdelayed nonmatching in monkeys. J Neurosci 24(8):2013-2026. McKee RD & Squire LR (1993) On the development of declarative memory. J Exp Psychol Learn Mem Cogn 19(2):397-404. Clark RE, Zola SM, & Squire LR (2000) Impaired recognition memory in rats after damage to the hippocampus. J Neurosci 20(23):8853-8860. Sobotka S, Nowicka A, & Ringo JL (1997) Activity linked to externally cued saccades in single units recorded from hippocampal, parahippocampal, and inferotemporal areas of macaques. J Neurophysiol 78(4):2156-2163. Ringo JL, Sobotka S, Diltz MD, & Bunce CM (1994) Eye movements modulate activity in hippocampal, parahippocampal, and inferotemporal neurons. J Neurophysiol 71(3):12851288. Caplan JB, Madsen JR, Raghavachari S, & Kahana MJ (2001) Distinct Patterns of Brain Oscillations Underlie Two Basic Parameters of Human Maze Learning. J Neurophysiol 86(1):368-380. Hughes AM, Whitten TA, Caplan JB, & Dickson CT (2011) BOSC: A better oscillation detection method, extracts both sustained and transient rhythms from rat hippocampal recordings. Hippocampus 22(6): 1417-1428. van Vugt MK, Sederberg PB, & Kahana MJ (2007) Comparison of spectral analysis methods for characterizing brain oscillations. J Neurosci Methods 162(1-2):49-63. Williams JM & Givens B (2003) Stimulation-induced reset of hippocampal theta in the freely performing rat. Hippocampus 13(1):109-116. Givens B (1996) Stimulus-evoked resetting of the dentate theta rhythm: relation to working memory. Neuroreport 8(1):159-163. Mormann F, et al. (2005) Phase/amplitude reset and theta-gamma interaction in the human medial temporal lobe during a continuous word recognition memory task. Hippocampus 15(7):890-900. Shah AS, et al. (2004) Neural Dynamics and the Fundamental Mechanisms of Event-related Brain Potentials. Cereb Cortex 14(5):476-483. Klimesch W, Sauseng P, Hanslmayr S, Gruber W, & Freunberger R (2007) Event-related phase reorganization may explain evoked neural dynamics. Neurosci Biobehav Rev 31(7):1003-1016. Sauseng P, et al. (2007) Are event-related potential components generated by phase resetting of brain oscillations? A critical discussion. Neuroscience 146(4):1435-1444. Rajkai C, et al. (2008) Transient Cortical Excitation at the Onset of Visual Fixation. Cereb Cortex 18(1):200-209. Fisher NI (1993) Statistical analysis of circular data (New York, NY, USA) pp xviii, 277 p. Huerta PT & Lisman JE (1995) Bidirectional synaptic plasticity induced by a single burst during cholinergic theta oscillation in CA1 in vitro. Neuron 15(5):1053-1063. Huerta PT & Lisman JE (1993) Heightened synaptic plasticity of hippocampal CA1 neurons during a Cholinergically induced rhythmic state. Nature 364(6439):723-725. Holscher C, Anwyl R, & Rowan MJ (1997) Stimulation on the Positive Phase of Hippocampal Theta Rhythm Induces Long-Term Potentiation That Can Be Depotentiated by Stimulation on the Negative Phase in Area CA1 In Vivo. J Neurosci 17(16):6470-6477. Lakatos P, Chen C-M, O'Connell MN, Mills A, & Schroeder CE (2007) Neuronal Oscillations and Multisensory Interaction in Primary Auditory Cortex. Neuron 53(2):279292.

13


40.

41.

42. 43. 44. 45.

46. 47.

48. 49. 50. 51. 52. 53.

54. 55. 56.

57. 58. 59. 60.

Addante RJ, Watrous AJ, Yonelinas AP, Ekstrom AD, & Ranganath C (2011) Prestimulus theta activity predicts correct source memory retrieval. Proc Natl Acad Sci U S A 108(26):10702-10707. Guderian S, Schott BH, Richardson-Klavehn A, & Duzel E (2009) Medial temporal theta state before an event predicts episodic encoding success in humans. Proc Natl Acad Sci U S A 106(13):5365-5370. Fell J, et al. (2011) Medial temporal theta/alpha power enhancement precedes successful memory encoding: evidence based on intracranial EEG. J Neurosci 31(14):5392-5397. Maris E & Oostenveld R (2007) Nonparametric statistical testing of EEG- and MEG-data. J Neurosci Methods 164(1):177-190. Jacobs J, Kahana MJ, Ekstrom AD, & Fried I (2007) Brain oscillations control timing of single-neuron activity in humans. J Neurosci 27(14):3839-3844. Skaggs WE, McNaughton BL, Wilson MA, & Barnes CA (1996) Theta phase precession in hippocampal neuronal populations and the compression of temporal sequences. Hippocampus 6(2):149-172. Fox SE, Wolfson S, & Ranck JB, Jr. (1986) Hippocampal theta rhythm and the firing of neurons in walking and urethane anesthetized rats. Exp Brain Res 62(3):495-508. Sobotka S & Ringo JL (1997) Saccadic eye movements, even in darkness, generate eventrelated potentials recorded in medial sputum and medial temporal cortex. Brain Res 756(12):168-173. Komisaruk BR (1970) Synchrony between limbic system theta activity and rhythmical behavior in rats. J Comp Physiol Psychol 70(3):482-492. Berg RW, Whitmer D, & Kleinfeld D (2006) Exploratory Whisking by Rat Is Not Phase Locked to the Hippocampal Theta Rhythm. J Neurosci 26(24):6518-6522. Rizzuto DS, et al. (2003) Reset of human neocortical oscillations during a working memory task. Proc Natl Acad Sci U S A 100(13):7931-7936. Tesche CD & Karhu J (2000) Theta oscillations index human hippocampal activation during a working memory task. Proc Natl Acad Sci U S A 97(2):919-924. McCartney H, Johnson AD, Weil ZM, & Givens B (2004) Theta reset produces optimal conditions for long-term potentiation. Hippocampus 14(6):684-687. Hyman JM, Wyble BP, Goyal V, Rossi CA, & Hasselmo ME (2003) Stimulation in hippocampal region CA1 in behaving rats yields long-term potentiation when delivered to the peak of theta and long-term depression when delivered to the trough. J Neurosci 23(37):11725-11731. Siapas AG, Lubenov EV, & Wilson MA (2005) Prefrontal Phase Locking to Hippocampal Theta Oscillations. Neuron 46(1):141-151. Sirota A, et al. (2008) Entrainment of neocortical neurons and gamma oscillations by the hippocampal theta rhythm. Neuron 60(4):683-697. Hyman JM, Zilli EA, Paley AM, & Hasselmo ME (2005) Medial prefrontal cortex cells show dynamic modulation with the hippocampal theta rhythm dependent on behavior. Hippocampus 15(6):739-749. Colgin LL, et al. (2009) Frequency of gamma oscillations routes flow of information in the hippocampus. Nature 462(7271):353-357. Bragin A, et al. (1995) Gamma (40-100 Hz) oscillation in the hippocampus of the behaving rat. J Neurosci 15(1):47-60. Fell J, et al. (2001) Human memory formation is accompanied by rhinal-hippocampal coupling and decoupling. Nat Neurosci 4(12):1259-1264. Montgomery SM & Buzsaki G (2007) Gamma oscillations dynamically couple hippocampal CA3 and CA1 regions during memory task performance. Proc Natl Acad Sci U S A 104(36):14495-14500.

14


61. 62.

63. 64. 65. 66. 67. 68.

Fries P (2009) The model- and the data-gamma. Neuron 64(5):601-602. Hasselmo ME, Bodelon C, & Wyble BP (2002) A proposed function for hippocampal theta rhythm: separate phases of encoding and retrieval enhance reversal of prior learning. Neural Comput 14(4):793-817. Lisman J (2005) The theta/gamma discrete phase code occuring during the hippocampal phase precession may be a more general brain coding scheme. Hippocampus 15(7):913-922. Lisman JE & Idiart MA (1995) Storage of 7 +/- 2 short-term memories in oscillatory subcycles. Science 267(5203):1512-1515. Wittmann BC, Bunzeck N, Dolan RJ, & Düzel E (2007) Anticipation of novelty recruits reward system and hippocampus while promoting recollection. Neuroimage 38(1):194-202. Landau Ayelet N & Fries P (2012) Attention Samples Stimuli Rhythmically. Curr Biol 22(11):1000-1004. Jarvis MR & Mitra PP (2001) Sampling properties of the spectrum and coherency of sequences of action potentials. Neural Comput 13(4):717-749. Mitra PP & Pesaran B (1999) Analysis of dynamic brain imaging data. Biophys J 76(2):691708.

15


Supporting Information for “Oscillatory activity in the monkey hippocampus during visual exploration and memory formation” Michael J. Jutras, Pascal Fries, and Elizabeth A. Buffalo* *To whom correspondence should be addressed. E-mail: [email protected] SI Methods Procedures were carried out in accordance with National Institutes of Health guidelines and were approved by the Emory University Institutional Animal Care and Use Committee. Neuronal recordings were carried out in two adult male rhesus monkeys (Macaca mulatta), which were obtained from the breeding colony at the Yerkes National Primate Research Center. Their mean weight at the start of the experiment was 6.8  1.1 kg, and their mean age was 4 years and 5 months. Prior to implantation of recording hardware, monkeys were scanned with magnetic resonance imaging (MRI) to localize the hippocampus and to guide placement of the recording chamber. Using this information, a cilux plastic chamber (Crist Instrument Co., Hagerstown, MD) for recording neural activity, and a titanium post for holding the head were surgically implanted. We performed post-surgical MRI to fine-tune electrode placement and to determine recording locations. Behavioral testing procedures. During testing, each monkey sat in a dimly illuminated room, 60 cm from a 19” cathode ray tube monitor, running at 120 Hz, non-interlaced refresh rate. Eye movements were recorded using a non-invasive infrared eye-tracking system (ISCAN, Burlington, Massachusetts). Stimuli were presented using experimental control software (CORTEX, www.cortex.salk.edu). At the beginning of each recording session, the monkey performed a calibration task, which involved holding a touch-sensitive bar while fixating a small (0.3°) gray square, presented on a dark background at various locations on the monitor. The monkey had to maintain fixation within a 3° window until the fixation point changed to an equiluminant yellow at a randomly chosen time between 500 ms and 1100 ms after fixation onset. The monkey was required to release the touch-sensitive bar within 500 ms of the color change for delivery of a drop of applesauce. During this task, the gain and offset of the oculomotor signals were adjusted so that the computed eye position matched targets that were a known distance from the central fixation point. Visual Preferential Looking Task. Following the calibration task, the monkey was tested on the Visual Preferential Looking Task (VPLT). The monkey initiated each trial by fixating a white cross (1°) at the center of the computer screen. After maintaining fixation on this target for 1 s, the target disappeared and a picture stimulus was presented (11°). The stimulus disappeared when the monkey’s direction of gaze moved off the stimulus, or after a maximum looking time of 5 seconds. Each stimulus was presented twice during a given session, with up to 8 intervening stimuli between successive presentations. The VPLT was given in 51 daily blocks of 6, 8, or 10 trials each, chosen pseudorandomly, for a total of 400 trials each day. The median delay between successive presentations was 8.1 seconds. Stimuli were obtained from Flickr (http://www.flickr.com/). A total of 9000 stimuli were used in this study.

16


Because the monkey controlled the duration of stimulus presentation, the duration of gaze on each stimulus provides a measure of the monkey’s preference for the stimulus. We compared the amount of time the monkey spent looking at each stimulus during its first (“Novel”) and second (“Repeat”) presentation. Adult monkeys show a strong preference for novelty; therefore, a significant reduction in looking time between the two presentations of a stimulus indicated that the monkey had formed a memory of the stimulus and spent less time looking at the now familiar stimulus during its second presentation. To control for varying interest in individual stimuli, recognition memory performance was calculated as the difference in looking time between presentations as a percentage of the amount of time the monkey spent looking at the first presentation of each stimulus: (novel – repeat) ÷ novel. Reward was not delivered during blocks of the VPLT; however, 5 trials of the calibration task were presented between each block to give the monkey a chance to earn some reward and to verify calibration. The number of trials in each VPLT block was varied to prevent the monkey from knowing when to expect the rewarded calibration trials. Electrophysiological recording methods. The recording apparatus consisted of a multi-channel microdrive (FHC Inc., Bowdoinham, Maine) holding a manifold consisting of a 23-gauge guide tube containing 4 independently moveable tungsten microelectrodes (FHC Inc., Bowdoin, Maine), with each electrode inside an individual polyamide tube. Electrode impedance was in the range of 1-2 MΩ, and electrode tips were separated horizontally by 190 µm. For each recording, the guide tube was slowly lowered through the intact dura mater and advanced to ~3.5 mm dorsal to the hippocampus with the use of coordinates derived from the MRI scans. The electrodes were then slowly advanced out of the guide tube to the hippocampus. No attempt was made to select neurons based on firing pattern. Instead, we collected data from the first neurons we encountered in the hippocampus. At the end of each recording session, the microelectrodes and guide tube were retracted. All recordings took place in the anterior part of the left hippocampus (1). Recording sites were located in the CA3 field, dentate gyrus, and subiculum. Data amplification, filtering, and acquisition were performed with a Multichannel Acquisition Processor (MAP) system from Plexon Inc. (Dallas, TX). The neural signal was split to separately extract the spike and the LFP components. For spike recordings, the signals were filtered from 250 Hz – 8 kHz, further amplified and digitized at 40 kHz. A threshold was set interactively, in order to separate spikes from noise, and spike waveforms were stored in a time window from 150 µs before to 650 µs after threshold crossing. Each recording typically yielded 2 to 6 units; single units were sorted offline using Offline Sorter (Plexon, Inc.). For LFP recordings, the signals were filtered with a passband of 0.7-170 Hz, further amplified and digitized at 1 kHz; any additional filtering was performed in Matlab. Only LFPs obtained from electrodes on which single unit activity was also recorded were analyzed in order to ensure that all LFP data were taken from cell layers of the hippocampus. Eye movement data were digitized and stored with a 240 Hz resolution. Data Analysis. All analyses were performed using custom programming in Matlab (The Mathworks, Inc., Natick, MA) and using FieldTrip (fieldtrip. fcdonders.nl), an open source toolbox for the analysis of neurophysiological data.

17


Saccade detection. Eye movement data were analysed in order to isolate fixation periods occurring between saccades. Saccades were detected by first applying a low-pass filter with a high-cut frequency limit of 40 Hz to the horizontal and vertical eye position data to remove high-frequency noise, differentiating and combining these signals to obtain the eye velocity, and setting a threshold of 40 degrees/second in order to define saccades. The start and end of each saccade was considered to occur when the first order derivative of the eye velocity (i.e., acceleration) reached zero before the upward crossing and after the downward crossing of this threshold, respectively. Powerline artifact removal. We estimated the amplitude of the powerline fluctuations with a Discrete Fourier Transformation (DFT) of long data segments which contained the data epochs of interest. We then computed the DFT at 60 and 120 Hz. Because the powerline artifact is of a perfectly constant frequency and amplitude, and because the long data segments contained integer cycles of the artifact frequencies, essentially all the artifact energy is contained in those DFTs. We constructed sine waves with the amplitudes and phases as estimated by the respective DFTs, and subtracted those sine waves from the original long data segments. The epoch of interest was then cut out of the cleaned epoch. Power spectra of the cleaned epochs demonstrated that all artifact energy was eliminated, leaving a notch of a bin width of 0.1 Hz in the monkey recordings. Single unit saccade modulation. We recorded from 131 hippocampal single units in two monkeys. For each neuron, the firing activity aligned to saccades was examined by calculating peri-saccade time histograms for the 1 second period centered at the onset of each saccade. We used a template matching procedure to test the modulation of the firing response following saccades for each neuron (2). Spike counts from the 300-ms period following each saccade were partitioned into 15 bins of 20 ms each, and the average baseline firing rate from the 300-ms period before saccade onset was subtracted from each bin. In this way, a 15-element vector was calculated for each saccade. The consistency of response patterns across saccades was determined by calculating the dot product of the vector for each saccade and the 15-element vector created by averaging all other response vectors, i.e. the average of vectors for all saccades excluding the saccade under consideration. The products from all bins were then added together to produce a dot product measure for each saccade: 15 n   M i    yij   y kj  j 1  k 1,  i 

where n is the number of saccades in the data set, Mi is the dot product measure for saccade i, yij is the amplitude of the response for saccade i and bin j (measured with respect to the presaccade baseline), and ykj is the same as yij, with index k substituted for index i. A test of proportions (3) was used to determine if the percentage of saccades for which the dot product measure, Mi, had positive values was significantly greater than 50%, according to the formula

z

pˆ  p p(1  p ) n

18


where pˆ is the observed proportion of saccades for which Mi is positive, p is the hypothesized proportion, and n is the sample size. Fifty percent is the value expected by chance if there is no consistent response across saccades; thus, a value significantly greater than 50% indicates predictive power from calculating the average response to all other saccades. A one-tailed test with a = 0.01 was used, corresponding to a critical z-value of 2.33. LFP spectral analysis and bout detection. We calculated power spectra from LFP data after first limiting data segments to those recorded during blocks of VPLT performance, including intertrial intervals, pre-stimulus fixation periods, and stimulus presentation/exploration periods. Initial spectral analysis was performed using the multitaper method (4, 5) on non-overlapping LFP segments of 5 seconds each, multiplying each with 3 orthogonal taper functions before Fourier transformation. This provided spectral smoothing of ±0.4 Hz. Local maxima were detected using the localmax function in Matlab, designating a minimum peak distance of 1.6 Hz. Bouts of theta activity in LFP data were quantified using an oscillatory episode detection algorithm that estimates the background power spectrum of the LFP in order to determine power and duration criteria (6-9). LFP segments were taken from blocks of VPLT performance, with each segment consisting of a block in its entirety (median block duration across all recordings was 32.5 sec.). After isolating LFP segments recorded during blocks of VPLT performance, we wavelet transformed the raw LFP traces using Morlet wavelets with a window of 7 cycles (10). As a result, for each frequency f, 7/(2f) seconds at the beginning and the end of each block were excluded from the calculation of wavelet power. Frequencies were sampled at 35 logarithmically-spaced steps between 2 and 38 Hz. The background power spectrum for each LFP was determined by fitting a linear regression to this wavelet power spectrum in log-log coordinates. At each frequency, a power threshold, PT, was set to the 95th percentile of the c2(2) distribution of wavelet power values at that frequency, and a duration threshold, DT, was set to 3 cycles. For each frequency, all time intervals during which LFP power exceeded PT for a time period exceeding DT cycles was designated an oscillatory episode. Theta bouts were defined as any contiguous time points of oscillatory episodes within the 3-12 Hz frequency band (3.08-12.34 after log2-spacing). Finally, we defined Pepisode as the proportion of time periods across all blocks of VPLT performance that exceeded both thresholds. Data used to plot example theta bouts in Fig. 1 underwent a prewhitening step prior to wavelet multiplication/Fourier transform for presentation. The example time-frequency spectrograms shown in Figure 1F, and the corresponding autocorrelograms in Figure 1G, were calculated from the following (from left to right): an LFP segment starting during the intertrial interval and persisting through the pre-stimulus fixation period; a segment starting during the pre-stimulus fixation period and ending shortly after stimulus onset; and two LFP segments occurring entirely within the stimulus presentation (free-viewing) period. Theta phase resetting with saccade onset. To examine phase resetting of theta oscillations during the pre-stimulus fixation period, we selected trials in which the monkey did not make any saccades in the 600-ms period before the appearance of the center cross or in the 1000-ms period between the saccade to the center cross and the onset of the stimulus. Because the monkey was only required to fixate within a 4° window around the center cross, it was possible in some excluded trials for a small

19


saccade to occur during the pre-stimulus fixation period without aborting the trial. An average of 20.3 ± 2.2 trials per session met these criteria, representing, on average, 10.4% of trials in each session. Power values were calculated for the presaccade period (the 600-ms period preceding saccade onset) and the postsaccade period (the 600-ms period following a 400-ms postsaccade “buffer” period) using separate methods to obtain individual trial power measures and trial-averaged power measures. For individual trial power measures, we multiplied the raw LFP traces for each 600-ms period with 1 orthogonal taper function before Fourier transformation, providing spectral smoothing of ±1.67 Hz. We then calculated the average power in the 6.7-11.6 Hz frequency band, across trials, to obtain a theta power value for each LFP for presaccade and postsaccade periods. For trial-averaged power measures, the same method was used, with the exception that before taper multiplication and Fourier transformation, we calculated the average LFP signal (i.e., evoked signal) across trials. This resulted in a single trial-averaged segment for the 600-ms presaccade period and another single trial-averaged segment for the 600-ms postsaccade period, for each LFP. We then applied taper multiplication and Fourier transformation to each LFP segment and calculated the average power in the 6.7-11.6 Hz frequency band, producing a theta power value for presaccade and postsaccade trial-averaged signals. A paired t-test was used to determine whether theta power was significantly different for pre- and postsaccade periods, separately for individual trial theta power values and trial-averaged theta power values. To calculate phase concentration for pre- and postsaccade LFPs, each LFP signal (encompassing the entire recording, i.e. before separating into pre- and postsaccade segments) was first filtered with a band-pass filter of 6-12 Hz, using a zero-phase-shift fourth-order Butterworth filter. LFP segments were designated which included 1000 ms of presaccade LFP data and 1300 ms of postsaccade LFP data. The local phase estimate at each time point in the LFP segment was calculated using the Hilbert transform, and phase distributions were calculated across all pre- and postsaccade fixation periods for each LFP channel. A 500-ms window was designated for the presaccade period, encompassing the period of -600 to -100 ms preceding saccade onset, allowing for a 100-ms buffer immediately preceding saccade onset to account for any distortion due to filtering. Similarly, a 500-ms window encompassing the period of 400 to 900 ms following saccade onset was designated for the postsaccade period. The degree of phase concentration across pre- and postsaccade fixation periods was evaluated by calculating the Rayleigh statistic (11) from the distribution of phases at each time-point in each 500-ms window. Briefly, given n phases fi, the first trigonometric moment ( )∑ was defined. The preferred phase m is given by the orientation of m and the mean resultant value R is given by the modulus of m. The Rayleigh statistic is given by Z = nR2. For each LFP, we calculated the average Rayleigh statistic value for the entire 500-ms window in each condition (pre- and postsaccade). A paired t-test was used to determine the statistical significance of the difference in pre- and postsaccade Rayleigh statistic values. Phase distributions following saccade onset during visual exploration. To examine the relationship between the LFP theta phase following saccade onset and recognition memory, the 20


stimuli from each session were ranked in order of increasing recognition performance, quantified as the percent change in looking time between Novel and Repeat presentations for each stimulus. The 30 Novel trials with the highest subsequent percent reduction were designated “High Recognition” and the 30 Novel trials with the lowest subsequent percent reduction were designated “Low Recognition”. The first three postsaccade fixation periods in each trial were included in the analysis; thus, only those trials in which the monkey made at least three saccades during Novel stimulus presentation, with each postsaccade fixation period lasting at least 200 ms, were used. This was done to control for the number of fixation periods in each condition and the amount of time included following each saccade. Finally, recording sessions in which fewer than 60 trials met these criteria were excluded from analysis in order to maintain an adequate separation between High and Low Recognition conditions. Seventy-four LFP recordings met these inclusion criteria. Average looking times, latencies to the first saccade following stimulus onset, and inter-saccade intervals for the first 3 200-ms postsaccade fixation periods in each trial are presented in Table S1. LFPs recorded during Novel trials were filtered with a band-pass filter of 3-12 Hz, using a zero-phase-shift fourth-order Butterworth filter. LFP segments were designated which included 500 ms of presaccade LFP data and 500 ms of postsaccade LFP data. The local phase estimate at each time point in this LFP segment was calculated using the Hilbert transform to extract the phase of the LFP at that time point. Phase distributions were calculated across all 90 postsaccade fixation periods in each condition, for each LFP (the first 3 fixation periods taken from each of 30 trials in each condition). This ensured that the final distributions used for the statistical analysis contained an equal number of measurements. The degree of phase consistency across saccades was evaluated by calculating the Rayleigh statistic from the distribution of phases at each time-point in the 400-ms window centered on saccade onset, for all the fixation periods in each condition, using methods described above. The probability that the null hypothesis of sample uniformity holds (i.e., the threshold for determining whether the phase distribution follows a random vs. non-uniform distribution) was calculated as P = e-Z (an adequate approximation for n > 50) (11, 12). For each LFP, we averaged the Rayleigh statistic values obtained in each condition for all time-points falling between 40 and 200 ms after saccade onset. A paired t-test was used to compare Rayleigh values between conditions across LFPs. To evaluate the degree to which changes in postsaccade theta phase represent a reset in the phase of ongoing oscillatory activity, rather than an additive, evoked response, we calculated power values for a 250-ms presaccade period and a 250-ms postsaccade period, after first selecting saccades occurring during stimulus viewing that were not preceded by another saccade within a 400 ms window, nor followed by another saccade within a 250 ms window. This 400-ms presaccade window was chosen in order to prevent the inclusion of a potential evoked response during the preceding fixation period in the calculation of presaccade power, while still including a substantial number of trials for analysis. Within each recording session, an average of 40.2 ± 2.7 peri-saccade periods met these criteria. Power values were calculated for the 250-ms periods preceding and following saccade onset, for individual trials (saccades) as well as again for trial-averaged LFP segments. Each LFP segment was multiplied with 1 orthogonal taper function before Fourier transformation, providing spectral smoothing of ±4 Hz. After averaging power values across the range of theta-band frequencies (4-12 Hz), a paired t-test was performed to compare theta power across LFPs for pre-

21


and postsaccade periods, separately for individual trial power values and trial-averaged power values. Theta power preceding stimulus onset and relationship with memory. To compare prestimulus theta power for subsequently well-recognized and subsequently poorly recognized stimuli, trials were ranked by increasing memory performance and the 30 trials with the largest/smallest difference in looking times between Novel and Repeat presentations were designated High/Low Recognition (as above). Only those trials in which the monkey looked at the Novel stimulus presentation for at least 750 ms were used. Each category represented a median of 22.2% of the total number of trials considered for each session. For each trial, the period of 800 ms immediately preceding stimulus onset was selected, buffered, and wavelet transformed, using Morlet wavelets with a window of 7 cycles. Time-frequency representations of spectral power were calculated from wavelet-transformed data segments, producing a mean time-resolved measure of spectral power for each LFP and each condition. To test for statistical significance of differences in pre-stimulus theta power for High and Low Recognition conditions, we performed a nonparametric multi-sample permutation test, with the median difference in power between conditions as our test statistic. The test involves a comparison of the observed difference against a reference distribution of differences under the null hypothesis of no significant modulation of power between conditions. The reference distribution was obtained by performing the following procedure 10,000 times. For each LFP, a random decision was made to which condition the data from either condition was assigned. We then calculated the test statistic at each time and frequency for these randomly assigned conditions and stored only the minimal and maximal difference across frequencies. From the resulting distribution of 10,000 minimal and maximal differences, we determined the 2.5th and the 97.5th percentile. The empirically observed, nonrandomized difference at a particular frequency was considered statistically significant (p < 0.05), when it was larger than the 97.5th or smaller than the 2.5th percentile of the reference distribution. This procedure corresponds to a two-sided test with a global false positive rate of 5% and correction for the multiple comparisons across frequencies (13, 14). We used this non-parametric permutation approach, because 1) it is free of assumptions about the underlying distributions, 2) it is not affected by partial dependence among the time-frequency tiles, and 3) it allows for correction for multiple comparisons without additional assumptions. To investigate the extent to which LFP data immediately following stimulus onset might have some influence on the enhancement in pre-stimulus theta-band power associated with better recognition memory, we designated 444-ms windows immediately preceding and immediately following stimulus onset for the same High and Low Recognition trials. Each LFP segment was multiplied with a Hanning taper function before Fourier transformation. Because the band of memory-related pre-stimulus power enhancement was centered at 9 Hz, a paired t-test was performed to compare power values at 9 Hz in each window across LFPs for the High and Low Recognition conditions. Theta oscillations and investigation of phase resetting during calibration task. To quantify the presence of theta-band activity during the calibration task, in which saccadic eye movements tend to

22


be cued, rather than exploratory, we applied the Pepisode analysis to data recorded during blocks of the calibration task. The methods for this analysis were identical to those used to quantify theta-band activity during blocks of the VPLT. The average theta Pepisode across LFPs, taken across all blocks of the calibration task, was 0.19 ± 0.01. Bouts of theta-band activity occurred with a median bout duration of 502 ms and a median inter-bout interval of 1,551 ms. The extent to which oscillatory phase was reset with the onset of cued saccades during the calibration task was investigated using identical analyses to those performed for saccades during the pre-stimulus period. Briefly: Power values were calculated for the presaccade period (the 600-ms period preceding saccade onset) and the postsaccade period (the 600-ms period following a 400-ms postsaccade buffer period) using separate methods to obtain individual trial power measures and trial-averaged power measures (as described in Methods). Taper multiplication and Fourier transformation were used to calculate the average power in the 6.7-11.6 Hz frequency band for each LFP segment, producing a theta power value for presaccade and postsaccade trial-averaged signals. A paired t-test was used to determine whether theta power was significantly different for pre- and postsaccade periods, separately for individual trial theta power values and trial-averaged theta power values. We also calculated the Rayleigh statistic as follows: each LFP signal (encompassing the entire recording) was filtered with a band-pass filter of 6-12 Hz. Saccades occurring following the appearance of the gray square in the calibration task were selected, and trials were limited to saccades which were followed by a fixation period lasting at least 1 second. LFP segments were designated which included 1,000 ms of presaccade LFP data and 1,300 ms of postsaccade LFP data. The local phase estimate at each time point in the LFP segment was calculated using the Hilbert transform, phase distributions were calculated across all pre- and postsaccade fixation periods for each LFP channel, and the degree of phase concentration across pre- and postsaccade fixation periods was evaluated by calculating the Rayleigh statistic from the distribution of phases at each time-point in the 500-ms windows designated for each period (-600 to -100 ms preceding saccade onset and 400 - 900 ms following saccade onset).

1. 2.

3. 4. 5. 6.

Jutras MJ, Fries P, & Buffalo EA (2009) Gamma-band synchronization in the macaque hippocampus and memory formation. J. Neurosci. 29(40):12521-12531. Sobotka S, Nowicka A, & Ringo JL (1997) Activity linked to externally cued saccades in single units recorded from hippocampal, parahippocampal, and inferotemporal areas of macaques. J. Neurophysiol. 78(4):2156-2163. Sokal RR & Rohlf FJ (1995) Biometry (W. H. Freeman and Company, New York, NY). Jarvis MR & Mitra PP (2001) Sampling properties of the spectrum and coherency of sequences of action potentials. Neural Computation 13(4):717-749. Mitra PP & Pesaran B (1999) Analysis of dynamic brain imaging data. Biophys. J. 76(2):691708. Hughes AM, Whitten TA, Caplan JB, & Dickson CT (2011) BOSC: A better oscillation detection method, extracts both sustained and transient rhythms from rat hippocampal recordings. Hippocampus 22(6): 1417-1428.

23


7.

8. 9. 10. 11. 12. 13. 14.

Caplan JB, Madsen JR, Raghavachari S, & Kahana MJ (2001) Distinct Patterns of Brain Oscillations Underlie Two Basic Parameters of Human Maze Learning. J Neurophysiol 86(1):368-380. Ekstrom AD, et al. (2005) Human hippocampal theta activity during virtual navigation. Hippocampus 15(7):881-889. van Vugt MK, Sederberg PB, & Kahana MJ (2007) Comparison of spectral analysis methods for characterizing brain oscillations. J Neurosci Methods 162(1-2):49-63. Grossman A & Morlet J (1985) Decomposition of functions into wavelets of constant shape, and related transforms. Mathematics + Physics, ed Streit L (World Scientific, Singapore). Fisher NI (1993) Statistical analysis of circular data (Cambridge University Press, Cambridge, UK) pp xviii, 277 p. Siapas AG, Lubenov EV, & Wilson MA (2005) Prefrontal Phase Locking to Hippocampal Theta Oscillations. Neuron 46(1):141-151. Nichols TE & Holmes AP (2002) Nonparametric permutation tests for functional neuroimaging: A primer with examples. Human Brain Mapping 15(1):1-25. Maris E & Oostenveld R (2007) Nonparametric statistical testing of EEG- and MEG-data. J. Neurosci. Methods 164(1):177-190.

Supporting figure legends

Fig. S1. Distribution of spectral peaks across hippocampal LFPs. Distribution of local maxima in the 2-12 Hz range in the power spectra of pre-whitened LFP data recorded on all channels during VPLT performance.

24


Fig. S2. Distribution of saccade rates across recordings. Distribution of saccade rates for all saccades measured across 45 test sessions of the VPLT. Red dashed line: median (5.1 Hz).

Fig. S3. Theta bout durations and inter-bout intervals. (A) Distribution of theta bout durations across 114 LFPs. Red dashed line: median (508 ms). (B) Distribution of theta inter-bout intervals across 114 LFPs. Red dashed line: median (1194 ms).

25


Fig. S4. LFPs displaying significant levels of oscillatory activity. The histogram shows the distribution of LFP channels with significant Pepisode values at each frequency tested (two-tailed t-test, p < 0.014).

Fig. S5. Pre- and postsaccade theta-band phase reliability during pre-stimulus period. Average Rayleigh statistic values, representing phase reliability, for theta-filtered (6.7-11.6) LFP segments taken from the 500-ms pre- and postsaccade periods during the pre-stimulus period. Phase reliability was significantly higher for the postsaccade period than the presaccade period (paired-sample t-test, p < 0.01).

26


Fig. S6. Pre- and postsaccade theta-band phase reliability during calibration task. Theta (6.7-11.6 Hz) power calculated across all single trials (A) and for each trial-averaged signal (B), for presaccade (red) and postsaccade (green) periods. Theta power was significantly higher for the postsaccade period than the presaccade period for the individual trial signals (p < 0.05) but not for the trial-averaged signals (p > 0.1). (C) Average Rayleigh statistic values, representing phase reliability, for theta-filtered (6.7-11.6) LFP segments taken from the 500-ms pre- and postsaccade periods.

Fig. S7. Pre- and postsaccade theta-band LFP power for individual saccades and for average saccade-locked LFPs. (A) Theta (4-12 Hz) power in the 250 ms window immediately preceding saccades (red) and immediately following saccades (green), calculated for individual LFP segments. (B) Theta power in 250 ms pre- and postsaccade windows, calculated for LFP segments after averaging across all saccades in a session (n = 87 LFP channels). Theta power was significantly higher for the postsaccade period than the presaccade period for the trial-averaged signals (p < 0.01) but not for the individual trial signals (p > 0.1).

27


Table S1. Novel trial looking times, latency to first saccade following stimulus onset, and intersaccade intervals for the high recognition and low recognition conditions Total number of sessions: High Recognition 27 Novel trial looking time (ms) Latency to first saccade after stimulus onset (ms) Inter-saccade interval (ms)

Low Recognition

p value

4448.6 ± 43.0

2818.4 ± 124.3

p < 0.01

229.2 ± 2.4

226.6 ± 2.9

p > 0.1

241.3 ± 3.0

243.3 ± 2.8

p > 0.1

Measures are presented as average ± SEM.

28