|Home | About | Journals | Submit | Contact Us | Français|
Neural phase signaling has gained attention as a putative coding mechanism through which the brain binds the activity of neurons across distributed brain areas to generate thoughts, percepts, and behaviors. Neural phase signaling has been shown to play a role in various cognitive processes, and it has been suggested that altered phase signaling may play a role in mediating the cognitive deficits observed across neuropsychiatric illness. Here we investigated neural phase signaling in two mouse models of cognitive dysfunction: mice with genetically induced hyperdopaminergia (DAT-KO mice) and mice with genetically induced NMDA receptor hypofunction (NR1-KD mice). Cognitive function in these mice was assessed using a radial-arm maze task, and local field potentials were recorded from dorsal hippocampus and prefrontal cortex as DAT-KO mice, NR1-KD mice, and their littermate controls engaged in behavioral exploration. Our results demonstrate that both DAT-KO and NR1-KD mice display deficits in spatial cognitive performance. Moreover, we show that persistent hyperdopaminergia alters inter-structural phase signaling, while NMDA receptor hypofunction alters inter-structural and intra-structural phase signaling. These results demonstrate that dopamine and NMDA receptor dependent glutamate signaling play a critical role in coordinating neural phase signaling, and encourage further studies to investigate the role that deficits in phase signaling play in mediating cognitive dysfunction.
Neural phase signaling has gained attention as a putative coding mechanism through which the brain binds the activity of neurons across brain structures to generate, percepts, and behaviors (Lisman and Buzsaki, 2008). For example, the synchronization of hippocampal (HP) and prefrontal cortex (PF) theta oscillations (4-10Hz) is enhanced during spatial navigation in rodents (Jones and Wilson, 2005). The synchronization of hippocampal and amygdalar oscillations may play a role in fear conditioning (Seidenbecher et al., 2003), and the phase synchronization of rhinal and hippocampal gamma (33-55Hz) oscillations has been suggested to coordinate memory formation (Fell et al., 2001). Moreover, some authors have reported that the long range phase synchronization of cortical gamma oscillations underlies processes such as perception (Rodriguez et al., 1999). Together, these studies suggest that the long range phase synchronization of theta and gamma oscillatory activity may play a role in coordinating inter-structural neural activity.
Studies in rodents have also shown that neural activity within brain structures is coordinated by the phasic interaction of theta and gamma oscillations in which the amplitude of gamma oscillations is modulated by the phase of theta oscillations (cross frequency phase coupling–CFPC) (Bragin et al., 1995; Chrobak and Buzsaki, 1998). More recently, it has been shown that phasic interactions between theta and gamma oscillations also coordinate neural activity across brain structures (Sirota et al., 2008). Importantly, studies have shown that phasic interactions between theta and gamma oscillations occur across the human neocortex as well (Palva et al., 2005; Canolty et al., 2006). These studies also demonstrate that theta-gamma coupling strength increases with cognitive load, suggesting that CFPC plays a role in coordinating neural oscillatory activity during high-end cognitive processes. Multiple studies in rodents demonstrate that phase signaling also coordinates brain activity at the neuronal level (phase locking) (Jones and Wilson, 2005; Siapas et al., 2005). Thus together, long range phase synchronization, CFPC, and phase locking could define a putative code in which phase signaling binds the activity of neurons across distributed brain structures to generate thoughts and behaviors. Several studies in rats have been aimed at investigating local field potential and single unit phase signaling across the hippocampal-prefrontal cortex pathway (Jones and Wilson, 2005; Siapas et al., 2005). These studies have demonstrated that both HP and PF neurons phase lock to HP theta oscillations during exploration, and that HP and PF theta oscillations phase synchronize during accurate spatial task performance, suggesting that phase signaling plays an important role in coordinating HP-PF network function.
Here we investigated neural phase signaling in two mouse models of cognitive dysfunction: mice with genetically induced hyperdopaminergia and mice with genetically induced NMDA receptor hypofunction. In order to assess spatial cognitive processing in the two mouse models, mutant mice and their littermate controls were subjected to a radial arm maze special learning task. Additionally, mice from each group were surgically implanted with recording electrodes in HP and prelimbic cortex, and HP-PF phase signaling was quantified while mice were in their home cage and while they explored a novel environment.
WT and Dopamine transporter (DAT) knockout (KO) littermates were generated from heterozygotes that had been backcrossed over ten generations onto the C57BL/6J background. DAT-KO mice lack the gene encoding the plasma membrane transporter that regulates spatial and temporal extracellular DA signaling (Giros et al., 1996). Due to loss of the DAT, these mutants exhibit a persistent 5-fold increase in extracellular striatal DA levels (Giros et al., 1996; Gainetdinov and Caron, 2003), and show locomotor hyperactivity, deficits in cognitive functions, deficits in sensorimotor gating, and altered hippocampal activity during awake behavioral periods (Giros et al., 1996; Gainetdinov et al., 1999; Dzirasa et al., 2006).
Functional NMDA receptors are composed of an essential NMDA receptor subunit, NR1, and one of four NR2 NMDA receptor subunits combined in an undetermined ratio to make a heteromeric receptor complex (Kutsuwada et al., 1992; Monyer et al., 1992). Though the global deletion of the NR1 subunit leads to perinatal lethality (Forrest et al., 1994; Li et al., 1994), mice genetically engineered to express only 10% of normal NR1 levels via hypomorphic mutation of Grin1 survive until adulthood (Mohn et al., 1999). Mice were derived by first generating C57BL6 and 129SvJ congenic lines (backcrossed > 12 generations). C57BL6 and 129SvJ mice that were heterozygous for the altered Grin1 allele were then crossed to generate the F1 mice used for this study. Importantly, NR1-knockdown (NR1-KD) mice exhibit a 90% reduction in functional NMDA receptors compared to wild-type (WT) mice, and also display locomotor hyperactivity, deficits in sensorimotor gating, and aberrant social interactions (Mohn et al., 1999; Duncan et al., 2006).
Mice were housed three-five/cage and maintained in a humidity- and temperature-controlled room with a 12-hour light dark cycle. Food and water was available ad libitum. All studies were conducted with approved protocols from the Duke University Institutional Animal Care and Use Committee and were in accordance with the NIH guidelines for the Care and Use of Laboratory Animals.
Radial arm maze spatial learning experiments were conducted in twenty-nine mice: 8 NR1-KD mice/7 WT littermate controls and 7 DAT-KO mice/7 WT litter mate controls. Mice were food deprived for about 5 hours each day before task performance. Following 1-2 shaping sessions, mice were subjected to five four-day block trials (20 consecutive days) of the 8-Arm Radial Maze task. Each trial consisted of confining the animal to the center of the maze for 1 minute. After the confinement period, mice were allowed to explore the maze until 300 seconds elapsed or they entered all eight arms. Each arm was baited with a small piece of sweet breakfast cereal. Cognitive performance was then assessed by the mean number of entries to repeat, and the number of perseverative errors (the number of times an animal re-entered an arm that it had just left) observed in each animal during a trial block.
Five DAT-KO mice and seven WT littermate controls were separated into individual cages, and surgically implanted with microwire recording electrodes. Electrophysiological experiments were initiated following a two week recovery. Four WT and four DAT-KO mice were first subjected to electrophysiological recordings during periods of quiet waking in their home cage. Periods of quiet waking (behaviorally inactive) were identified using video recordings and LFP power spectral analysis (Dzirasa et al., 2006). Subsequently, five WT mice and five DAT-KO mice were subjected to electrophysiological recordings as they explored a novel environment. A radial arm maze was chosen as the ‘novel environment’ used for electrophysiological recordings in DAT-KO mice because radial arm maze exposure does not induce hyperactivity in these mice (Gainetdinov et al., 1999).
Five NR1-KD mice and four WT littermate controls were also separated into individual cages, and surgically implanted with recording electrodes. Experiments were initiated following a two week recovery. All of the WT mice were recorded during quiet waking in their home cage, and as they explored a ‘novel environment.’ All of the NR1-KD mice were recorded as they explored a novel environment, and three of the mice were also recorded during quiet waking in their home cage. The novel environment utilized in NR1-KD and their littermate controls consisted of an open field (two NR1-KD and 2 WT controls) or radial arm maze (three NR1-KD and 2 WT controls). Both open field and radial arm maze exposure induced hyperactivity in NR1-KD mice. NR1-KD mice displayed mild seizure-like neural activity across hippocampus and prelimbic cortex (see Figure S4). A subset of these mice also showed full generalized tonic-clonic seizures (Ramsey, unpublished findings). This latter group of mice was excluded from the study.
DAT-KO mice and WT littermate controls were anesthetized with ketamine (100mg/kg) and xylazine (10mg/kg), placed in a stereotaxic device, and metal ground screws were secured to the cranium. Two 16-Tungsten microwire array electrodes (30 μm diameter single wires) were implanted bilaterally through a small cranial window into the dorsal hippocampus (HP) (stereotaxic coordinates: -2.3mm AP, +/-1.6mm ML, and 1.5mm DV from bregma), and anchored to ground screws located above the above the cortex using dental acrylic. Two 16-Tungsten microwire array electrodes (30μm diameter single wires) were also implanted bilaterally through a small cranial window into prelimbic cortex (PrL-an anatomical subdivision of medial prefrontal cortex in rats) (stereotaxic coordinates: 1.7mm AP, +/-0.4mm ML, and 1.9mm DV from bregma, and 2.7mm AP, +/-0.4mm ML, and 0.8mm DV from bregma), and anchored to the secured ground screws using dental acrylic. The surgical procedure was identical for the NR1-KD mice and their littermate controls with the exception that the anesthetic dose was reduced by 50% for NR1-KD mice, and only one 16-Tungsten microwire array electrode was implanted into the dorsal hippocampus and one 12-Tungsten microwire array electrode was implanted into prelimbic cortex (PrL).
Neural activity was recorded using the Multi-Neuron Acquisition Processor system (Plexon Inc, TX). Local field potentials (LFPs) were preamplified (500×), filtered (0.5–400 Hz), and digitized at 500 Hz using a Digital Acquisition card (National Instruments, Austin, TX) and a Multi-Neuron Acquisition Processor (Plexon, Dallas, TX). All electrophysiological recordings were referenced to two ground screws, and recording segments demonstrating LFP saturation resulting from movement artifacts were excluded from analysis. It is important to note that LFP recordings were associated with significant phase offsets that varied across frequencies (Nelson et al., 2008). These phase offsets were corrected using the LFPAlign function (Plexon, Dallas,TX).
Behaviors were recorded with a video cassette recorder. Video images were loaded onto a cassette recorder and synchronized with the neural recordings using a millisecond-precision timer.
Ten to sixteen LFPs were recorded per session (up to eight from hippocampus and up to eight from prelimbic cortex). LFPs were filtered using butterworth bandpass filters designed to isolate either theta (4-10Hz) or gamma (33-55Hz) oscillations. The instantaneous phase of the filtered LFP was then determined using the Hilbert transform (Siapas et al., 2005), yielding two phase time series for each frequency, HP(t) and PrL(t) (500 values per second). The Rayleigh test of circular uniformity (Siapas et al., 2005; Jacobs et al., 2007) was then used to quantify phase synchrony as a function of the difference between the two phase time series [HP(t) - PrL(t)]. We also used the Rayleigh test to quantify phase synchrony when time offsets ranging from -3s to 3s in 20ms increments were introduced between the phase time series. This yielded a total of 301 Zstat values, where Zstat = -ln(p). The Zstat value corresponding to phase synchrony was then given by the difference between the Zstat at zero offset and the maximum Zstat observed in the [-3s to -1s] and [1s to 3s] offset windows. This process was repeated for each of the up to 64 HP × PrL LFP combinations (8 HP channels × 8 PrL channels) recorded per mouse. The significance threshold for phase synchrony was then conservatively set at α=0.001/(301*64*5mice), corresponding to Zstat = 18.38. All comparison combinations were used to make statistical comparisons of theta and gamma band phase synchrony across genotypes. This conservative approach was taken given the large coefficient of variation (mean C.V. = 102%) across the Zstat values observed across gamma frequency HP × PrL combinations within mice. The inter-structural phase synchrony data presented in Figure 1 was collected from the WT littermate controls of DAT-KO mice.
LFPs were filtered using butterworth bandpass filters designed to isolate either theta (4-10Hz) or gamma (33-55Hz) oscillations. The instantaneous amplitude and phase of the filtered LFPs were then determined using the Hilbert transform, and the Modulation index was calculated using the MATLAB code provided by Canolty et. al. (Canolty et al., 2006) Briefly, a continuous variable z(t) is defined as a function of the instantaneous theta phase and instantaneous gamma amplitude such that z(t) = AG(t)*eiTH(t), where AG is the instantaneous gamma oscillatory amplitude, eiTH is a function of the instantaneous theta oscillatory phase. A time lag is then introduced between the instantaneous amplitude and theta phase values such that zsurr is parameterized by both time and the offset between the two variables, zsurr = AHG(t+τ)*eiTH(t). The moldus of the first moment of z(t), compared to the distribution of surrogate lengths, provides a measure of coupling strength. The normalized z-scored length, or Modulation index, is then defined as MNORM = (MRAW-μ)/ σ, where MRAW is the modulus of the first moment of z(t), μ is the mean of the surrogate lengths, and σ is their standard deviation. Importantly, MNORM is a Z-score and can be used to determine the probability that a result would be due to chance. The inter- and intra-structural cross frequency phase coupling data presented in Figure 2 was collected from the WT littermate controls of NR1-KD mice.
Theta and gamma oscillatory power was determined for each LFP by taking the mean of the amplitude envelope calculated for each of the theta and gamma filtered LFP traces.
The HP-PrL LFP combinations which demonstrated the strongest phase synchrony were used to estimate the coherence (Bendat and Piersol, 1986) and partial directed coherence (Sameshima and Baccala, 1999; Baccala and Sameshima, 2001) for each animal. Squared coherence is a frequency-wise measure of the total coupling strength between two waves. Squared coherence was calculated using previously described methods based on vector autoregressive (VAR) modeling (Bendat and Piersol, 1986; Baccalá et al., 2006). Briefly, the bivariate VAR model of order p is described by the following equation:
where, , m = 1,2, represents the m-th time series at time t and , m = 1,2, is the m-th innovation with covariance matrix
Important quantities are the bivariate AR filter given by:
and the bivariate spectral density matrix given by:
where the superscript H stands for conjugate transpose operation, for any frequency λ [−π, π).
The squared coherence at frequency λ between time series 1 and 2 is then given by:
The modulus square of PDC from signal 2 to 1 which will be discussed below can be written as
The PDC from signal 1 to 2 can be written analogously.
For the analysis a VAR model was fitted sequentially to each non-overlapping 0.5s time segments to assure a stationary and adequate fit. The coherence for each animal was then calculated as the average of the squared coherences calculated from each VAR model for each respective LFP recording. The averages for the WT and DAT-KO group coherences were calculated as the averages of the mean coherences for the mice in respective groups, and the total theta coherence was defined as the average of the squared coherence values in the 4-10Hz range. Similarly, the total gamma coherence was defined as the average of squared coherence values in the 33-55Hz range. The statistical distribution for the null hypothesis of zero total theta and gamma coherences were calculated by the parametric bootstrap method previously described (Baccalá et al., 2006).
Partial directed coherence (PDC) is an established method for the identification of directional interaction between brain areas (Sameshima and Baccala, 1999; Baccala and Sameshima, 2001; Baccalá et al., 2006; Takahashi et al., 2007). Briefly, consider two wave signals, u1 and u2. PDC from u1 to u2 measures the proportion of the power spectra of u2 due to an effect from u1. Similarly, the PDC from u2 to u1 measures the proportion of the power spectra of u1 due to an effect from u2. The key property that allows for calculation of PDC is asymmetry between u1 and u2, such that the PDC from u1 to u2 is generally not equal to PDC from u2 to u1. Nullity of PDC from u1 to u2 indicates the absence of an influence of u1 on u2, and thus no directional interaction. Conversely, a PDC value of one determined from u1 to u2 indicates that u2 is linearly predictable from u1. A useful form of PDC called generalized PDC was utilized for all PDC computations presented in the text (Baccalá et al., 2006). PDC was computed with the same VAR models used to calculate coherence. The same procedures used to estimate the total theta and gamma coherences were utilized to estimate the WT and DAT-KO group total theta and gamma PDC. Parametric bootstrap analysis was used to calculate the threshold values under the null hypothesis of zero total theta and gamma PDC (Baccalá et al., 2006).
Statistical significance was determined for inter-structural phase synchrony and inter-structural CFPC comparisons across genotype using a Bonferroni corrected two-tailed t-test. The significance threshold was conservatively estimated at α = 0.001/(64-combinations*10-mice) or P* < 0.000001. Statistical significance was determined for intra-structure cross frequency phase coupling using ANOVA followed by student's t-test. In this case, we conservatively corrected for six comparisons (the mean number of LFP channels recorded per mouse). A corrected P*-value less than 0.05 was considered significant. All comparisons across genotype were made against WT littermate controls.
First we investigated if WT mice displayed neural phase synchrony across the HP-PF pathway during behavioral exploration. We simultaneously recorded local field potentials (LFPs) from hippocampus and prelimbic cortex as mice explored a novel environment. To measure low and high-frequency phase synchrony across the HP-PF pathway, we band-pass filtered LFPs in the theta (4-10Hz) or gamma (33-55Hz) range, and decomposed filtered oscillations into instantaneous phase and amplitude components (Figure 1a). We then quantified theta and gamma band phase synchrony by calculating the circular uniformity of the phase time series difference between the HP and PrL electrode pairs (Figure 1b and 1c). Using this process, we identified significant theta band and gamma band phase synchrony across the HP-PF pathway (Figure 1d-g). Incidentally, we also investigated theta and gamma band phase synchrony during periods of quiet waking (behaviorally inactive) while WT mice were in their home cage. Our results indicated that novelty exposure significantly increased theta band synchrony (Theta Z-stat: 12910±490 in the novel environment, 2960±170 in the home cage; tstat1,510=15.2;P*<0.001), and decreased gamma band synchrony in WT mice (Gamma Z-stat: 1350±100 in the novel environment, 3180±210 in the home cage; tstat1,510=9.02;P*<0.001).
Next, we quantified HP and PrL intra-structural synchrony by calculating theta-gamma cross frequency phase coupling (CFPC) using the Modulation index (Figure 2a-b) (Canolty et al., 2006). Our results indicated that WT mice displayed significant HP and PrL intra-structural synchrony during periods of novel exploration. Moreover, our results also indicated that novelty exposure significantly increased hippocampal and cortical CFPC in WT mice when compared to periods of behavioral inactivity while mice were in their home cage (tstat1,58=2.46; P<0.05 and tstat1,61=4.87; P<0.001, respectively; see Figure 2c-d). Finally, in order to quantify the influence of hippocampal theta oscillatory activity over cortical gamma oscillatory activity, we calculated inter-structural CFPC using the Modulation Index. Consistent with previous reports (Sirota et al., 2008), our results indicated that WT mice displayed significant phase coupling between cortical gamma oscillations and hippocampal theta oscillations during periods of novel exploration. Our results also indicated that novelty exposure significantly increased inter-structural CFPC when compared to periods of behavioral inactivity while mice were in their home cage (M: 0.9±0.0 in the home cage, 1.7±0.1 in the novel environment; tstat1,470=8.81, P*<0.000001; see Figure 2e).
After demonstrating that neural phase signaling was indeed present in WT mice, we investigated if specific aspects of neural phase signaling were altered by persistent hyperdopaminergia or NMDA receptor hypofunction.
In order to assess spatial cognitive function in dopamine transporter-knockout (DAT-KO) mice, we subjected mice to a classic eight-arm radial maze learning task. Our results indicated that C57BL/6J DAT-KO mice displayed deficits in spatial cognitive performance. As shown in Figure 3, two way ANOVA of the entries to repeat found the main effects of genotype [F1,69=11.09, P<0.01] and trial [F4,69=3.53, P=0.013], as well as the genotype by condition interaction [F4,69=3.46, P=0.014] to be significant. Moreover, two way ANOVA of the perseverative errors committed found the main effects of genotype [F1,69=11.60, P<0.01] to be significant. These observations, which are consistent with observations previously gained in DAT-KO mice bred on a different background (129×C57Bl6 mixed background) (Gainetdinov et al., 1999), indicate that altered dopamine signaling leads to cognitive dysfunction.
After confirming that DAT-KO mice displayed spatial cognitive deficits compared to their WT littermates, we set out to quantify HP-PF phase signaling during spatial cognitive processing in the mutant mice. DAT-KO mice were surgically implanted with microwire electrodes, and LFP recordings were conducted as they engaged in novelty induced exploration. We found that DAT-KO mice displayed significantly higher HP-PrL gamma band phase synchrony compared to WT mice during exploration of a novel environment (DAT-KO gamma Z-stat: 3640±310; n=296 channel combinations, tstat1,614=7.26, P*<0.0000001, compared to novelty exposed WT mice; compare Figure 1 and Figure 4c). Conversely, theta band phase synchrony was statistically indistinguishable in DAT-KO and WT mice (DAT-KO theta Z-stat: 12740±700; n =296 channel combinations, tstat1,614=0.20, p>0.05 compared to novelty exposed WT mice; compare Figure 1 and Figure 4a). When we compared these data to LFP recordings conducted while DAT-KO mice were in their home cage, we found that novelty exposure potentiated theta band phase synchrony (tstat1,534=10.88, P*<0.001), but failed to attenuate HP-PF gamma band phase synchrony as it had in WT mice (tstat1,534=0.98; P*>0.05). Interestingly, our results also showed that hippocampal and cortical theta-gamma CFPC in DAT-KO mice and inter-structural CFPC (hippocampal theta – cortical gamma coupling) was statistically indistinguishable from that observed in WT mice (P>0.05 for all comparisons).
In order to investigate the increased HP-PF gamma phase synchrony in DAT-KO mice, we measured HP-PrL partial directed coherence (PDC). PDC is a method that quantifies the preferential direction of wave activity propagation between two fields based on power spectral changes (Sameshima and Baccala, 1999; Baccala and Sameshima, 2001; Kus et al., 2004; Baccalá et al., 2006; Takahashi et al., 2007). DAT-KO mice displayed increased PrL to HP directional wave propagation in the high frequency range compared to WT mice (tstat1,29=5.78, respectively, P*< 0.01).
Conversely, there was no difference in high frequency HP to PrL wave propagation (Figure 4e). Together, this suggests that HP gamma oscillatory activity becomes increasingly synchronized to the gamma source responsible for modulating high frequency PrL oscillatory activity in DAT-KO mice.
Overall, our data demonstrated that persistent hyperdopaminergia alters high frequency inter-structural HP-PF synchrony, while leaving low frequency inter-structural HP-PF synchrony and HP-PF intra-structural synchrony intact.
Since the spatial cognitive processing of NR1-KD mutants had not been assessed, we subjected WT and NR1-KD mice to our radial-arm maze spatial learning task (Figure 5). Two-way ANOVA of entries to repeat found the main effects of genotype [F1,74=14.01, P<0.001] but not trial [F4,74=1.36, P>0.05], or the genotype by condition interaction [F4,74=0.93, P>0.05] to be significant. This was due to the rapid learning rates observed in 5/7 WT mice observed across the first block. When entries to repeat analysis was limited to data collected on the first day of each trial block, two-way ANOVA found the main effects of genotype [F1,74=16.95, P<0.01] and trial [F4,74=4.30, P<0.01], as well as the genotype by condition interaction [F4,74=3.36, P=0.016] to be significant. Moreover, two way ANOVA of perseverative errors found the main effects of genotype [F1,69=37.91, P<0.001] to be significant, demonstrating that NR1-KD mice displayed cognitive inflexibility (Clarke et al., 2004). Together, these results indicated that global NMDA receptor hypofunction was indeed associated with cognitive dysfunction.
After confirming that NR1-KD mice displayed spatial cognitive deficits compared to their WT littermates, we set out to quantify if HP-PF phase signaling was altered during periods of spatial cognitive processing in the mutant mice. NR1-KD mice were surgically implanted with microwire electrodes, and LFP recordings were conducted as they engaged in novelty induced exploration. NR1-KD mice displayed HP-PF theta and gamma band phase synchrony that was statistically indistinguishable from that observed in WT mice (p>0.05 for both comparisons). Conversely, NR1-KD mice displayed profound deficits in hippocampal and prelimbic cortex theta-gamma CFPC compared to WT mice during periods of novel exploration (tstat1,66=13.83 and tstat1,70=8.18, respectively, p*<0.001 for both comparisons, compare Figure 2 and Figure 6a-b). Our data recorded from NR1-KD mice during periods of quiet waking in their home cage also demonstrated that novelty exposure failed to potentiate HP and PrL CFPC as it had in WT mice (tstat1,60=0.63 and tstat1,60=0.77, respectively; P>0.05; see Figure 6a-b). Importantly, these deficits were not due to changes in hippocampal/prelimbic cortex theta or gamma oscillatory power (see Figure S8 and S9). When data recorded from WT mice was pooled across all sessions, we found that hippocampal gamma oscillatory power was preferentially modulated by a precise phase band of the theta oscillation trough, and a broad phase band following the theta oscillation peak. All of the WT mice displayed preferential gamma modulation at both the precise and broad phase band; conversely, NR1-KD mice only showed preferential modulation of gamma oscillatory activity by a broad phase band following the theta oscillation peak (Figure 6c). These results suggest that NMDA receptor function is critically involved in coordinating processes central to the precise coupling of gamma oscillatory power to the theta oscillation trough. Our results also showed that cortical gamma oscillatory power in WT mice was preferentially modulated by a broad phase band during the falling phase of theta oscillations. Similarly, NR1-KD mice showed preferential modulation of gamma oscillatory power by a broad phase band during the falling phase of theta oscillations, though cortical CFPC was significantly diminished overall (Figure 6d). Next, we investigated inter-structural phase coupling (hippocampal theta-cortical gamma) in NR1-KD mice. Our results indicated that novelty exposed NR1-KD mice displayed significantly enhanced inter-structural CFPC compared to their WT littermate controls. Moreover, enhanced inter-structural CFPC was also observed during quiet waking while NR1-KD mice were in their home cage (NR1-Novel M: 2.7±0.1, NR1-Home M: 3.6±0.2, tstat1,478=4.23, P*<0.0001 for both comparisons; compare Figure 2e with Figure 6e).
Taken together, our data suggested that NMDA receptor hypofunction disrupts HP and PF intra-structural coupling between theta and gamma oscillations while enhancing inter-structural coupling between hippocampal theta and cortical gamma oscillatory activity (Figure 7).
Neural phase signaling has gained attention as a putative coding mechanism through which the brain binds the activity of neurons across distributed brain structures to generate thoughts, percepts, and behaviors. Several studies in rodents have been aimed at investigating hippocampal-prefrontal cortex pathway phase signaling during spatial cognitive processes. These studies have shown that both hippocampal (HP) and prefrontal cortex (PF) neurons (Siapas et al., 2005), and HP-PF gamma oscillatory activity (Sirota et al., 2008), phase lock to HP theta oscillations during periods of exploration in a novel environment. Moreover, studies have demonstrated that HP and PF theta oscillations phase synchronize during accurate spatial task performance (Jones and Wilson, 2005), suggesting that phase signaling plays an important role in coordinating HP-PF pathway function during periods of enhanced spatial cognitive processing.
Here we identify altered phase signaling across the hippocampal-prefrontal cortex pathway in two mouse models of cognitive dysfunction: mice with genetically induced hyperdopaminergia (DAT-KO mice) and mice with genetically induced NMDA receptor hypofunction (NR1-KD mice). Though our findings only serve to correlate deficits in HP-PF phase signaling with gross dysfunction in spatial cognitive task performance, they provide a novel frame work to investigate the molecular mechanisms underlying phase signaling, and ultimately the role that phase signaling plays in coordinating brain activity during spatial cognitive processing.
Our findings indicated that WT mice display theta and gamma band phase synchrony at zero lag. Zero lag phase synchrony can result from defined neurological mechanisms, or from artifacts such as generic noise or electrical diffusion. While our data demonstrating the presence of HP and PrL theta-gamma coupling in WT mice suggested that theta band phase synchrony was not simply due to mechanical noise, our findings did not exclude the possibility that the theta band zero-lag phase synchrony measured using our recording technique may have resulted from electrical diffusion (Sirota et al., 2008). Thus, DAT-KO and NR1-KD mice may have indeed displayed alterations in theta-band synchrony that were not identified in this study. On the other hand, our findings did suggest that gamma band phase synchrony resulted from neurological mechanisms. Firstly, gamma oscillations are local (Sirota et al., 2008), Thus, it is unlikely that HP gamma oscillations could be recorded from PrL cortex given the distance between the two structures. Secondly, our analysis demonstrated that there was limited coherence at frequencies between theta and gamma coherence bands (see Supplemental Figure S4). Given that WT and DAT-KO (Dzirasa et al., 2006), and NR1-KD mice (see Supplemental Figure S5) both display higher hippocampal power at frequencies ranging from 10-33Hz than the gamma frequency range, a continuous coherence band from 4-55Hz would be observed if the gamma coherence observed throughout the study was due to electrical diffusion. Finally, we found a high coefficient of variance of gamma phase synchrony across HPxPrL channel combinations even within the same mice. In fact, we even found channel combinations that displayed high phase synchrony and channel combinations that displayed no phase synchrony at all in the same mice. If gamma phase synchrony was due to electrical diffusion, common noise, or technical cross talk, gamma phase synchrony would be detected in all of the channel combinations.
Our data demonstrated that novelty exposure increased HP-PF theta phase synchrony and theta-gamma intra-structural phase coupling in normal mice. Thus, theta phase synchrony and intra-structural phase coupling likely plays a prominent role in coordinating hippocampal and cortical gamma oscillatory activity during periods of behavioral exploration. Conversely, novelty exposure decreased gamma band inter-network phase synchrony. Our results in DAT-KO mice demonstrate that persistent hyperdopaminergia potentiates HP-PF gamma band inter-structural phase synchrony. We also show that the enhanced HP-PF gamma band phase synchrony observed in DAT-KO mice is associated with a unidirectional increase in gamma band LFP wave propagation from PrL to HP. Thus, our findings suggest that persistent hyperdopaminergia enhances the synchronization of HP oscillations to the gamma source responsible for generating PF gamma oscillatory activity. Incidentally, hippocampal gamma oscillatory activity becomes phase synchronized to cortical gamma oscillations as learning progresses (Fell et al., 2001), and evidence suggests that this process may be mediated by efferent projections from prefrontal cortex to lateral entorhinal cortex that control the transfer of information from hippocampus to cortex (Paz et al., 2007). Moreover, several studies indicate that modest increases in dopaminergic tone enhance neuronal activity in entorhinal cortex (Caruana et al., 2006), and DAT-KO mice have been shown to display increased gamma oscillatory power across hippocampus (Dzirasa et al., 2006). Thus, the increased HP-PF gamma band synchrony observed in DAT-KO mice may be due to enhanced dopaminergic activity at entorhinal cortex. Importantly, our results also demonstrate that hyperdopaminergia does not alter HP-PrL theta band phase synchrony or HP and PrL intra-structural synchrony. Together, these results suggest that HP-PrL gamma-band phase synchrony represents an independent mechanism for processing high frequency information across the HP-PF pathway.
Our results observed in the NR1-KD mice also showed that diminished NMDA receptor function disrupts hippocampal and cortical intra-structural synchrony, while enhancing inter-structural CFPC and leaving low- and high-frequency inter-structural phase synchrony intact. This indicates that the inter-structural synchronization of HP-PF gamma oscillatory activity does not simply result from coordinated theta-gamma interactions occurring locally within each of these two structures. Rather, our data collected in NR1-KD mice also suggests that gamma-band phase synchrony results from a complex mechanism responsible for coordinating high precision information processing across the HP-PF pathway.
Altogether, our findings suggest that gamma-band phase synchrony and theta-gamma phase coupling likely represent distinct phase signaling mechanisms for binding the activity of inter-structural and intra-structural neuronal populations. Moreover, the coordination of gamma oscillatory activity using these phase signaling mechanisms is differentially affected by novelty, persistent hyperdopaminergia, and NMDA receptor hypofunction.
Dopamine (DA) signaling has been suspected as an important mediator of the cognitive function (Kellendonk et al., 2006), and glutamatergic signaling has been shown to play an important role in learning and memory via activation of NMDA receptors (Larson and Lynch, 1988; O'Dell et al., 1991; Tsien et al., 1996; Tang et al., 1999; Shimizu et al., 2000). As such, it is believed that alterations in dopamine and/or NMDA receptor dependant glutamate neurotransmission contribute, at least in part, to the cognitive endophenotypes observed across neuropsychiatric disorders such as schizophrenia, attention-deficit-hyperactivity disorder, and bipolar disorder (Coyle, 1996; Tamminga, 1998; Laruelle and Abi-Dargham, 1999; Coyle and Tsai, 2004; Horschitz et al., 2005; Jucaite et al., 2005; Kellendonk et al., 2006). While the electrophysiological mechanisms underlying the manifestation of these brain disorders have yet to be identified, multiple studies have demonstrated that the cognitive impairments observed in patients with schizophrenia are associated with altered coordination of high frequency neural oscillatory activity (Spencer et al., 2004; Cho et al., 2006; Light et al., 2006; Yeragani et al., 2006). Moreover, several electroencephalographic (EEG) studies have demonstrated that individuals with schizophrenia display abnormal activation of prefrontal cortical networks during cognitive tasks (Winterer et al., 2000; Winterer et al., 2004). Taken together with our findings, this suggests that altered high frequency phase signaling may indeed play a role in mediating cognitive dysfunction; however, further studies will be necessary in order to definitively test this critical assumption.
We would like to thank our colleagues Shih-Chieh Lin and Ranier Gutierrez for comments on this manuscript; G. Lehew for technical assistance; L. Oliveira, T. Jones, and G. Wood for miscellaneous support; and S. Halkiotis for proofreading of this manuscript. This work was supported by funding from the Wakeman Foundation, Ruth K. Broad Foundation, and UNCF/Merck to KD; and by Award Number RO1-NS19576 and MH73853 from NARSAD and NIH to MGC; and by Award Number P50MH060451 and R33NSO49534 from NIH to MALN. A special thanks to Freeman Hrabowski, Robert and Jane Meyerhoff, and the Meyerhoff Scholarship Program.