Cortical and hippocampal EEG show differences 1 Running head: CORTICAL AND HIPPOCAMPAL EEG SHOW DIFFERENCES Cortical and Hippocampal EEG Show Different Simultaneous Sleep States after Learning

Several lines of evidence have challenged the assumption that sleep is a whole brain phenomenon. The idea of covert rapid-eye-movement sleep (REMS) processes (Nielsen, 2000) suggests that activity in different areas of the brain participate in different behavioral states at once. Posited covert REMS processes are consistent with a model of brain organization that begins at a local, neuronal group level and eventually leads to sleep as defined on macroscopic levels through collective outputs (Krueger & Obál, 1993). Furthermore, the assumption of brain state homogeneity is methodologically convenient, but may lead to inconsistencies in observations of sleep states underlying learning processes. We found that that a hippocampaldependent learning condition elicits heterogeneity in rapid-eye-movement sleep (REMS) and transition-to-REMS (TREMS) states in the hippocampus and cortex. These results depict that the states of these local structures differ when scored from respective EEG. Consequently, placement of electrodes influences the characterization of sleep states. Measuring sleep from the structures predicted to be affected by experimental manipulation may be a first step in reconciling inconsistencies in the extant literature. Cortical and hippocampal EEG show differences 3 Cortical and Hippocampal EEG Show Different Simultaneous Sleep States after Learning Introduction Traditionally sleep has been considered to be a property of an organism as a whole. Consistent with this model, sleep scoring in humans (Rechtschaffen & Kales, 1968) and rats (Benington, Kodali, & Heller, 1994) assumes cortical state homogeneity and characterizes whole organism sleep state based on predominate EMG and cortical EEG. Recently the assumption that the brain as a whole must participate in the same state simultaneously has been challenged. Covert REMS processes have been suggested as a means to reconcile NREMS mentation through the suggestion that REMS processes “combine” with NREMS sleep processes (Nielsen, 2000). It remains to be determined whether these REMS processes could be isolated to a noncortical site while NREM processes occur in the cortex. Directly challenging the traditional view of sleep, the neuronal group theory of sleep places the origin of sleep squarely at a local level within neuron groups (Krueger & Obál, 1993). REMS processes that occur covertly within NREMS may simply be the product of subcortical local, neuronal group activity. Ample evidence in support of this alternate model of brain state regional heterogeneity exists at various levels of brain organization. Unihemispheric sleep has been shown in dolphins (Goley, 1999), seals (Lyamin, Mukhametov, & Siegel, 2004), and birds (Rattenborg, Amlaner, & Lima, 2001) under certain conditions. Cortical columns, the basic processing unit of the waking brain, have been shown to oscillate between waking and sleep-like states, with a minority of columns existing in a state different than that of the whole-animal (Rector, Topchiy, Carter, & Rojas, 2005). Serotonergic antidepressants have been shown to increase EMG activity in submental leads during tonic REMS (Winkelman & James, 2004) which is characterized by muscle atonia (Rechtschaffen & Kales, 1968; Benington, Kodali, & Heller, 1994) suggesting that Cortical and hippocampal EEG show differences 4 atonia centers in the brain are exhibiting an atypical state. Dissociated states of wakefulness and sleep, termed parasomnias (e.g., REM behavior disorder, sleepwalking, night terrors, and narcolepsy) have been reported in humans and described as mixtures of wakefulness, NREMS, and REMS (Mahowald & Schenck, 1991). Dissociated state in some animals may be utilized when necessary. One idea is that regional difference in state could maintain synapses infrequently used during wakefulness (Krueger & Obál, 1993). Assuming slow wave sleep (SWS) reflects a reset of synaptic connectivity (Moruzzi & Magoun, 1949), heavy use of one area of the cortex during wakefulness results in an increase in slow wave activity in that area relative to others during sleep in mice, rats, chicken, pigeons, cats, and humans (Miyamoto, Katagiri & Hensch, 2003; Iwasaki, Karashima, Tamakawa, & Nakao, 2004; Vyazovskiy, Borbély, & Tobler, 2000; Cottone, Adamo, & Squires, 2004; Ferrara, De Gennaro, Curcio, Cristiani, & Bertini, 2002; Huber, Ghilardi, Massimini, & Tononi, 2004; Yasuda, Yasuda, Brown, & Krueger, 2005; Kattler et al., 1994), pointing to regional independence of intensity and, possibly, state. Songbirds in a migratory state have been reported to sleep nearly two-thirds less than when in a non-migratory state (Rattenborg, Mandt, Obermeyer, Winsauer, Huber, Wikelski, & Benca, 2004). Though this has been interpreted as evidence against sleep during flight, brain regions other than cortical sites have yet to be investigated and could prove to exist in a sleep-like state. In sleeping flocks, mallard ducks located at an edge of the group exhibit a 150% increase in unihemispheric slowwave sleep (USWS) keeping one open eye as a means of predator detection (Rattenborg, Lima, & Amlaner, 1999). Fur seals display two fundamentally different patterns of sleep: bilaterally symmetrical slow-wave sleep (BSWS), the predominate pattern when sleeping on land; and SWS with a striking interhemispheric EEG asymmetry (ASWS), the predominate pattern when Cortical and hippocampal EEG show differences 5 sleeping in the water (Lyamin, Mukhametov, & Siegel, 2004). Furthermore, fur seals have shown an increase in BSWS when sleep deprived while on land (Lyamin, Kosenko, Lapierre, Mukhametov, & Siegel, 2008). Similarly, domestic chicks have been shown to spend more time in bihemispheric sleep in the recovery period after sleep deprivation (Bobbo, Nelini, & Mascetti, 2008). The assumption of brain state homogeneity is methodologically convenient as characterization would require one pair of cortical leads, but may explain inconsistencies in the research that remain to be resolved. For example the involvement of REMS in learning processes remains controversial. Individuals taking antidepressant pharmaceuticals; monoamine oxidase inhibitors (MAOIs), selective serotonin reuptake inhibitors (SSRIs), and tricyclic antidepressants(TCAs); have shown significant reduction of REMS, but these classes of drug typically do not disrupt normal daily functioning (Vertes & Eastman, 2000). Yet in learning animals, sleep deprivation during the REMS window, a period after task training in which REMS has been shown to increase, causes performance deficits. The latency to onset and window duration have been shown to vary depending on the conditions of a task (e.g., 2-way shuttle avoidance task in rodents: 1-4 hr, 100 trials, single session [Smith, J. Young, & W. Young, 1980); 9-12 hr or 53-56 hr, 50 trials/day, 2 consecutive days (Smith & Lapp, 1986; Smith & MacNeill, 1993); 9-12 hr or 17-24 hr, 20 trials/day, 5 continuous days (Smith & Butler, 1982; Smith et al., 1980); 9-12 hr, 50 trials/day, 2 consecutive days (Smith, Tenn, & Annett, 1991)]. Additionally when animals learn there is a rise in sleep in only certain stages, and such higher percentage of that state is not consistently reported (e.g., pursuit rotor learning task in humans has shown statistically significant (Fogel, C. Smith, & Cote, 2007) and non-significant (Peters, V. Smith, & C. Smith, 2007) increases in Stage 2 sleep on acquisition night). Cortical and hippocampal EEG show differences 6 We suggest that brain state heterogeneity exists at the site level, contributing to the idea that sleep is not a whole brain phenomenon, and that we can observe such heterogeneities in the hippocampus and cortex for REMS and TREMS states in animals involved in a hippocampaldependent learning condition. Additionally this work may serve as a first attempt to explain inconsistencies similar to those aforementioned in the sleep literature. In order to further test the model of regional heterogeneity against the current assumption of brain state homogeneity in all conditions, learning or otherwise, we examined hippocampal and cortical EEG of rats involved in a hippocampal-dependent spatial memory task for simultaneous differences in REMS and TREMS states between recording sites. Methods Five male Fisher 344 rats of weight and age, on average, 360.5 ± 44.54 g and 5.8 ± 1.89 months, performed a visual platform variation of the Morris water maze (Morris, 1984) (5 trials/d, 2 d) to test for visual acuity. Subsequently these animals were trained on a hippocampaldependent spatial learning task, the Poe 8-box maze (Poe et al., 2002), for food reward. During training rats were food restricted, but maintained a minimum of 80% of their pre-training weight. Animals were observed during daily training sessions and scored for errors on each lap. Rats walked or ran around an elevated track in a clockwise direction for 30 minutes, feeding from three of eight baited boxes each containing 1 mL food (powdered LabDiet® 5001 Rodent Diet® pellets mixed with water) delivered from syringes through tubing. To avoid aiding animals with visual or scent cues, all boxes were attached to similar tubing and syringes and contained inaccessible food. After every 5 laps, animals were removed from the maze and placed on a platform to rest for 2 min. Animals were returned after resting to a random position on the maze Cortical and hippocampal EEG show differences 7 to begin the next lap. The maze was rotated 180o after every 10 laps to remove any egocentric cues from the animal. The location of baited boxes was adjusted in order to hold the 3 food-box configuration unchanged relative to allocentric cues. Possible error types were: commission, an investigation of an empty (non-baited) box with sniffing or a nose poke into the box; hesitation, a pause and 45o head turn toward an empty box; and omission, a failure to eat from or investigate a baited box. Animals participated in training until committing at maximum an average of one error per lap or fewer at a rate of 45 laps per hour. Following training, animals were anesthetized with an intraperitoneal injection of 60 mg/kg sodium pentobarbital then implanted using stereotaxic surgery with a hyperdrive, a set of microdrives each capable of recording single cell activities from the hippocampus (Venkatachalam, Fee, & Kleinfeld, 1999). This device housed a dozen 12 μm wire tetrodes placed deep in the dorsal hippocampus and a reference tetrode placed 0.5 mm dorsal to the hippocampal cell layer in the deep white matter of the neocortex. The hippocampal EEG was 0gain current amplified and obtained by referencing one of the twelve deep hippocampal tetrodes to the neocortical reference tetrode. The cortical EEG was obtained separately from a jeweler’s screw electrode placed in the skull over the left parietal lobe (n=4) or the left frontal lobe (n=1), differentially referenced to a similar screw electrode placed in the skull over the frontal cortex. The EMG was differentially recorded from a pair of wires threaded through the nuchal muscles and the two EMGs were referenced together for one channel of EMG recording. After surgery animals were given an intramuscular injection of 1 mL Pro-Pen-G® (Penicillin G Procaine Injectable Suspension), orally administered liquid Children’s Tylenol®, placed on a heating blanket and monitored until they had regained consciousness. Cortical and hippocampal EEG show differences 8 Each rat was habituated to recording conditions and resumed pre-surgical 8-box maze performance (45 laps/hr, <1 error/lap) during a minimum 10-day recovery period. After recovery the animals performed the 8-box maze daily at the beginning of the light period for food reward for 5 consecutive days. This task reused the 3 food-box configuration that rats had learned during previous training (familiar configuration) and included an initially novel configuration as well. The novel configuration was located on the opposite side of the room, previously hidden from the animals’ visual field by a patterned divider. Learning trials consisted of 45 total laps: 15 on the familiar configuration, 15 on the novel configuration, and 15 again on the familiar configuration. Within a 15 lap sub-session, as during training, animals rested for 2 minutes after every 5 laps, and the maze was rotated 180o after the 10th lap to remove the predictive power of egocentric cues. Rats were scored for errors as previously described. The largest task performance gain across the learning sessions was calculated within each animal’s error record. The largest performance gain, which typically occurred between day 2 and 3 or later, was a reduction in the errors on laps 11-15 after maze rotation on the novel maze. Errors on these post-rotation laps in days prior to the gain were committed at boxes that had contained food before the 180o rotation, suggesting that animals were not performing the task using allocentric cues that persist independently of maze rotation. The EEG and EMG were recorded after task performance for 4 hr roughly 60-90 minutes into the light period. The EEG and EMG recordings were read into the Sleepscorer program (Mathworks) where states were manually scored in 10-second epochs for sleep/waking states. One of five sleep/waking states were assigned to each epoch (Bjorness, 2008): Active waking (AW) = theta activity and high EMG activity Cortical and hippocampal EEG show differences 9 Quiet waking (QW) = low amplitude, desynchronized EEG and relatively little EMG activity Quiet sleep (QS) [NREMS] = high amplitude synchronized EEG and low EMG activity Transition-to-REMS (TR) [TREMS] = high amplitude spindle activity and low EMG activity rapid-eye-movement sleep (RE) [REMS] = clear, sustained theta activity and phasic muscle twitches on a background of low EMG activity Recordings of sleep/waking activity prior to the most significant intersession task performance gain across the 5-day learning session were selected for further analysis since the largest increase of REMS intensity has been shown to occur at the day prior to the largest task performance gain (Smith, Nixon, & Nader, 2004). Custom Matlab programs (Mathworks; Gross, Walsh, Booth, & Poe, 2008) were used in the state analysis. Power spectral density values in the delta (0.4-4 Hz), theta (5-9 Hz), sigma (10-14 Hz), and beta (15-20 Hz) frequency bands for each epoch were calculated. Mean power values were found for each band in each 4 hr recording. Additionally power spectral density values ± SDM in dB were calculated from total RE, TR, and QS states for frequencies 0 20.04 Hz at 0.244 Hz intervals in each 4 hr recording. Normalized power spectral density values obtained from the frontal-frontal and hippocampal EEG represent one animal and were compared separately. Data from this animal were isolated in this power spectral density value analysis because relative wave powers for each state differ in the frontal and parietal cortices and cannot be directly compared. This data is not excluded from subsequent category analyses because recording site should not affect state if scored using accepted parameters. RE and TR epochs were identified in hippocampal and, separately, cortical recordings. For any epoch scored as RE or TR the state at the alternative site was also noted. The denotation Cortical and hippocampal EEG show differences 10 of states in the two sites was assigned as hippocampal state/cortical state, e.g., TR/RE = (hippocampal state was TR and cortical state was RE, simultaneously). In this investigation we sought to explore the pressures for TR and RE states in particular as learning conditions have been shown to influence these states. Eight distinct categories including RE, TR, or ‘n’ (non-RE, non-TR) and describing both hippocampal and cortical state simultaneously were produced: TR/TR, RE/RE, TR/’n’, ‘n’/TR, RE/’n’, ‘n’/RE, TR/RE, and RE/TR. TR/TR and RE/RE categories were termed similar epochs as state was uniform between sites. The remaining six categories (TR/’n’, ‘n’/TR, RE/’n’, ‘n’/RE, TR/RE, and RE/TR) were termed dissimilar epochs as state was found to differ between sites. A mock example is provided below depicting the state of the two separate sites at simultaneous epochs, the resulting category denotation at each epoch, and the nature (dissimilar or similar) of the category: Hippocampal state QS QS QS TR RE RE RE RE QS Cortical state QS TR QS TR TR RE RE QS QS Epoch Number 100 101 102 103 104 105 106 107 108 Category Denotation none n/TR none TR/TR RE/TR RE/RE RE/RE RE/n none Similar categories Dissimilar categories In a single site’s recording: e density (power) value. That single band power value respective band’s power for all epochs in th normalized band power for a category ± SEM power within each category and band was compared representing this process for delta power Mean normalized DP for all TR/TR epochs in cortical site Mean ± SEM DP for all TR/TR epochs from cortical site across all animals Single animal Addition of other animals’ data Cortical and hippocampal EEG show differences ach band (4 in total) of a category epoch had a was normalized to the mean of at single site’s recording. The mean of the was then calculated within a site across all animals between sites. A diagram is provided on the following page: Mean ± SEM DP for all TR/TR epochs from hippocampal site across all animals Single site (single recording): Cortical site Single epoch: TR/TR epoch Mean delta power within all epochs = mean DP Delta power within this epoch = DP = normalized DP for 1 TR/TR epoch in the cortical site Comparison Addition of other TR/TR epochs’ data DP Mean DP 11