Cannabinoid Receptors Mediate Methamphetamine Induction of High Frequency Gamma Oscillations in the Nucleus Accumbens

Abstract

Patients suffering from amphetamine---induced psychosis display repetitive behaviors, partially alleviated by antipsychotics, which are reminiscent of rodent stereotypies. Due to recent evidence implicating endocannabinoid involvement in brain disorders, including psychosis, we studied the effects of endocannabinoid signaling on neuronal oscillations of rats exhibiting methamphetamine stereotypy. Neuronal network oscillations were recorded with multiple single electrode arrays aimed at the nucleus accumbens of freely moving rats. During the experiments, animals were dosed intravenously with the CB1 receptor antagonist rimonabant (0.3 mg/kg) or vehicle followed by an ascending dose regimen of methamphetamine (0.01, 0.1, 1, and 3 mg/kg; cumulative dosing). The effects of drug administration on stereotypy and local gamma oscillations were evaluated. Methamphetamine treatment significantly increased high frequency gamma oscillations (~ 80 Hz). Entrainment of a subpopulation of nucleus accumbens neurons to high frequency gamma was associated with stereotypy encoding in putative fast-spiking interneurons, but not in putative medium spiny neurons. The observed ability of methamphetamine to induce both stereotypy and high frequency gamma power was potently disrupted following CB1 receptor blockade. The present data suggest that CB1 receptor-dependent mechanisms are recruited by methamphetamine to modify striatal interneuron oscillations that accompany changes in psychomotor state, further supporting the link between endocannabinoids and schizophrenia spectrum disorders.

Introduction

Psychomotor activation is a classically described behavioral phenomenon observed in rodents administered amphetamines or other psychostimulants. Following low drug doses, psychomotor activation is characterized by increased locomotor behavior at the expense of grooming, feeding and other normal rodent behaviors [1]. Higher challenge doses elicit repetitive head movements, excessive sniffing and other stereotyped behaviors (stereotypies) [2]. Human patients suffering from amphetamine psychosis display repetitive, stereotyped behaviors similar to rodent stereotypy [3]. Moreover, both rodent [4] and human [5] stereotypies can be alleviated by treatment with typical antipsychotics. These observations suggest that amphetamine-induced stereotypy may be an appropriate animal model for the psychomotor expression of psychosis [4].

Recent evidence suggests that cannabinoid CB1 receptor signaling mediates rodent stereotypy. Both wild-type rodents pretreated with CB1 receptor antagonists and CB1 KO mice express diminished amphetamine- [6] and methamphetamine- [7] induced stereotypy. Furthermore, CB1 receptor antagonist microinjections applied locally in the nucleus accumbens (NAc) reduce methamphetamine-induced stereotypy [7]. These findings implicate endocannabinoid(eCB) signaling in the ventral striatum as a crucial neural substrate for psychostimulant stereotypies.

Amphetamine doses capable of stereotypy induction have recently been shown to increase high frequency (~ 80 Hz), local field potential (LFP) gamma oscillations (γ) in the ventral striatum of behaving rodents [8]; however, the relationship between γ and stereotyped behavior is unknown. We recently demonstrated that neuronal firing patterns in the ventral striatum encode methamphetamine-induced stereotypy [7]. Moreover, this phenomenon was preferentially expressed in putatively identified fast spiking interneurons (FSIs).It is therefore possible that γ oscillations in the ventral striatum are a reflection of local, coordinated FSIactivity required for stereotypy expression.

The current study aimed to examine the relationship between γ oscillations in the ventral striatum and the expression of methamphetamine-induced stereotypy. We show that methamphetamine dose-dependently increaseshigh frequency (70-94 Hz, ~ 80 Hz peak)γpower(γ₈₀)in the NAc of freely-moving rats. Furthermore, we demonstrate that CB1 receptor signaling contributes toγ₈₀ induction. Finally, we show that the extent to which FSIs cohere with localγ₈₀predictsthat unit’spropensity to encode stereotypy. Interestingly, this phenomenon was not recapitulated in putatively identified medium spiny neurons (MSNs).These findings further substantiate a role for ventral striatal FSIsin the mediation of eCB-dependent psychostimulant stereotypy.

Results

Methamphetamine Dose-Dependently Increases Gamma Power in the Nucleus Accumbens

Amphetamine has been shown to increase γ in the striatum [8]. To determine if other amphetamines (such as methamphetamine) possess γ-inducing properties, we recorded LFP activity in the ventral striatum following administration of saline, followed by methamphetamine (0.01, 0.1, 1 and 3 mg/kg; i.v.; cumulative dosing). Similar to amphetamine, methamphetamine dose-dependently increased γ LFP power in the NAc(Figure 1).

Figure 1.

Methamphetamine increases LFP γ oscillations in the NAc. (A, top)Power spectrogram from a representative recording site in a subject given saline followed by cumulative methamphetamine administration (0.01, 0.1, 1 and 3 mg/kg; i.v.). All PSD values normalized from 0 to 1 (color scale).(A, bottom) Raw LFP data from the same experiment recorded following saline (left) and methamphetamine (right). Note the prominent increase in fast oscillations following methamphetamine. (B) PSDs for the raw traces displayed in (A), demonstrating the methamphetamine-induced enhancement of γ power.

Rimonabant Attenuates Methamphetamine-Induced Increases in γ₈₀

We previously demonstrated that rimonabant attenuates methamphetamine-induced changes in firing rate in the ventral striatum [7]. To determine if CB1 receptor antagonism also curbs methamphetamine-induced γ₈₀, we recorded LFP activity in the NAc of behaving rats treated with either rimonabant (0.3 mg/g; i.v.; n = 4) or vehicle (n = 5), followed by saline and increasing doses of methamphetamine thereafter (0.01, 0.1, 1 and 3 mg/kg; i.v.; cumulative dosing). The single recording site with the greatest γ response in each animal was chosen for further analysis. Comparison of maximal response sites in the core and shell revealed that Nac sub-region had no effect on methamphetamine-induced changes in LFP frequencies between 0-100 Hz (data not shown; F_(1,127) = 0.06, p > 0.05). Therefore, maximal response sites from the core and shell were pooled for further analysis. Methamphetamine dose-dependently increased LFP power in the 70 to 94 Hz bandwidth (figure 2A; 2 way ANOVA, F_(4,127) = 5.07, p < 0.001;Bonferroni posttest p values ranging from < 0.01 to < 0.001), but not LFP power across other bandwidths (figure 2A; Bonferroni posttest p values > 0.05) in vehicle-treated subjects. Therefore, subsequent analyses were limited to the γ₈₀ bandwidth. Rimonabant attenuated methamphetamine-induced γ₈₀ (Figure 2B, 2 way ANOVA, F_(1,75) = 3.02, p < 0.001), whereas rimonabant alonehad no effect on γ₈₀ power(Figure 2B, 1 way ANOVA, F_(1,75) = 4.43, p > 0.05).

Figure 2.

Methamphetamine-induced increases in γ₈₀require CB1 receptor activation. (A) Mean PSDs, normalized to baseline values, for LFPs at a single recording site in subjects given vehicle followed by saline (sal), followed thereafter by methamphetamine (0.01, 0.1, 1 and 3 mg/kg; meth).Note the dose-dependent increase in power across the high frequency γ(70 – 94 Hz) bandwidth (**-***, 2 way ANOVA, F_(4,127) = 5.07, p < 0.001, Bonferroni posttest p values ranging from < 0.01 to < 0.001).Methamphetamine fails to alter LFP power across other bandwidths (Bonferroni posttest p values > 0.05). Error bars excluded for clarity. (B)γ₈₀across time in both veh-and rimo-treated subjects (1 min bins). Meth significantly augments γ₈₀(1 way ANOVA,F_(12,75) = 7.13, p < 0.001). Rimo pretreatment significantly attenuates meth-induced γ₈₀ (2 way ANOVA, F_(1,75) = 3.02, p < 0.001). Error bars ± SEM.

Phasic Changes in γ₈₀do not Accompany Psychomotor Activation

We previously reported that phasic changes in NAc firing patterns encode the onset of stereotypy [7]. We show in Figure 3A that rimonabant was effective in reducing methamphetamine-induced stereotypy (unpaired t test, t₍₆₎ = 3.36, * p < 0.05), but not locomotion (unpaired t test, t₍₆₎ = 1.16, p > 0.05).To determine if stereotypy initiation is temporally encoded by changes in γ₈₀ we examined brief changes in γ₈₀ time-locked to the onset of each behavioral bout. We observedphasic increases in γ₈₀ both at the onset of stereotypy and throughout the remainder of each behavioral bout in a subset of experiments (Figure 3B). However, we failed to replicate this finding in our main pharmacological experiment, as neither rimonabant nor vehicle-treated rats exhibited γ₈₀ encoding of stereotypy (Figure 3C).To confirm that methamphetamine-induced γ₈₀ representsa direct pharmacological effect on striatal circuitry and not an epiphenomenon secondary to changes in overall psychomotor state, we similarly examined changes in γ₈₀ power time-locked to the onset of each bout of locomotion. As expected, we did not observe evidence of γ₈₀ locomotor encoding (Figure 3D). However, this does not completely rule out that some of the changes in power may arise from a combination of motor and pharmacological effects.

Figure 3.

Phasic changes in γ₈₀ are not associated with onset of stereotypy. (A) Stereotypy (top) and locomotion (bottom)expressed following methamphetamine (3 and 1 mg/kg; i.v., respectively) in vehicle- and rimonabant-treated subjects. Systemic rimonabant (0.3 mg/kg, i.v.) significantly attenuates stereotypy compared to vehicle-treated animals (unpaired t test, t₍₆₎ = 3.36, * p < 0.05). Rimonabant fails to curb methamphetamine-induced locomotion (unpaired t test, t₍₆₎ = 1.16, p > 0.05)(B) Mean power spectrogram for peri-event LFPs centered at the onset of individual bouts of stereotypy from a single recording site in an animal given saline followed by methamphetamine (0.01, 0.1, 1 and 3 mg/kg) (top). Mean LFP power is represented by the color scale; peri-event changes in γ power are depicted by the white line plot overlay. Brief fluctuations in γ appear to encode the onset of stereotypy(bottom). (C) Meanperi-event γ₈₀ patterns averaged across all recording sites in vehicle- and rimonabant-treated subjects reveal that phasic changes in γ₈₀are maintained at the onset of neither stereotypy nor locomotion in either treatment group. Error bars +/−SEM.

Spiking Activity in the Nucleus Accumbens is Entrained to Gamma Oscillations

Our prior electrophysiological findings demonstrate that local, ventral striatal CB1 receptor signaling mediates stereotypy [7]. However, it is possible that LFP signals recorded in the NAc are generated by passive volume transmission of extrastriatalγ, rather than local neuronal activity. To determine if NAc γ is related to local neuronal activity, we examined the temporal relationship between γ and neuronal spiking activity simultaneously recorded in the ventral striatum.

We observed γ₈₀-entrained spiking activity in a subpopulation of putative medium spiny neurons (MSNs; Figures4A, B and C; 14 out of 90 recorded neurons, 15.6%), fast-spiking interneurons (FSIs; Figures 4A, B and C; 1 out of 12 recorded neurons, 8.3%) and non-FSI interneurons (5 out of 20 recorded neurons, 25.0%, not shown). To confirm the validity of our averaged peri-spike waveforms, we compared these segments to surrogate segments generated when a random jitter (between −10 and 10 ms) was applied to the raw reference timestamps (Figures 4B and 4C). As expected, this jitter disrupted γ₈₀-spike coherence without altering entrainment to lower frequency waveforms. Subjectively identified entrainment was confirmed by objective statistical analysis using a custom bootstrap algorithm (data not shown; see Methods, Coherence).

Figure 4.

A subpopulation of NAc neurons exhibit spiking activity entrained to γ₈₀ oscillations. (A) Inter-spike interval histograms for a representative MSN (left) and FSI (right). Spike waveform shapes for the entire recording session shown in the insets. (B)Corresponding autocorrelograms for the units shown in (A).(C) Mean LFP segments time-locked to neuronal spiking activity for the units displayed in (A). If spiking activity and LFP activity occur irrespective of one another, averaging peri-spike LFP segments across large numbers of trials theoretically yields a flat line. Note the prominent peri-spike LFP waveforms. To confirm the validity of this analysis, we compared averaged peri-spike waveforms derived from raw data to similar waveforms derived using surrogate spike timestamps artificially jittered between −10 and +10 ms.(D)PSDs for mean peri-spike LFP segments displayed in (B). The robust peak at ~70 to 80 Hz suggests that the above peri-spike waveforms contain a significant γ component. Note that spike jitter interferes with γ entrainment, but not with lower frequency coherence.(E)Spike-LFP coherence for neurons displayed in (A). Note the peak in coherence in the γ bandwidth, confirming that these neurons exhibit γ-entrained spiking activity. Histogram insets depict changes in spike-γ₈₀ coherence following both saline (sal) and methamphetamine (meth) in both vehicle- (veh) and rimonabant- (rimo) treated animals. Pharmacological manipulations had no gross effect on γ₈₀-spike coherence in MSNs (left inset; methamphetamine vs. saline, 1 way ANOVA, F(1,1) = 2.42, p > 0.05; vehicle vs. rimonabant, 1 way ANOVA, F(1,1) = 0.45, p > 0.05) or FSIs (right insert; methamphetamine vs. saline, 1 way ANOVA, F(1,1) = 1.66, p > 0.05; vehicle vs. rimonabant, 1 way ANOVA, F(1,1) = 1.78, p > 0.05).

Neither methamphetamine (for MSNs Figure 4E left inset; 1 way ANOVA, F_(1,1) = 2.42, p > 0.05; for FSIs Figure 4E right inset; 1 way ANOVA, F_(1,1) = 1.66, p > 0.05) nor rimonabant (for MSNs Figure 4E left inset; 1 way ANOVA, F_(1,1) = 0.45, p > 0.05; for FSIs Figure 4E right inset; 1 way ANOVA, F_(1,1) = 1.78, p > 0.05) significantly altered neuronal coherence to theγ₈₀ LFP bandwidth. To ensure that these negative findings were not due to the inclusion of neurons with relatively low γ₈₀ coherence, a similar analysis was restricted to neurons exhibiting statistical criteria for γ₈₀ coherence (see methods; data not shown; n = 18after exclusion of the pilot experiment; no main effect for saline vs. methamphetamine, 1 way ANOVA, F_(1,1) = 3.00, p > 0.05; no main effect for vehicle vs. rimonabant, 1 way ANOVA F_(1,1) = 0.31, p > 0.5).

However, when analysis was restricted to stereotypy-encoding units, a relationship was observed between γ₈₀-spike coherence and phasic neuronal encoding of stereotypy. The degree to which anFSIexhibited γ₈₀ coherence waspredictive of its propensity to effectively encode the onset of stereotypic behavior(Figure 5B right panel;linear regression, F_(1,6) = 6.88, R² = 0.53, p < 0.05).By contrast, this phenomenon was not recapitulated in MSNs (Figure 5B left panel; linear regression, F_(1,10) = 0.23, R² = 0.02, p > 0.05).

Figure 5.

FSI, but not MSN, gamma entrainment is associated with neuronal encoding of stereotypy. (A) Peri-event histograms of 2 representative neurons (MSN and FSI, left and right, respectively) encoding the onset of stereotypy. Line plot overlays depicting normalized peri-event firing rates allow for comparison of behavioral encoding strength (z score peak heights) between neurons. (B) Linear regression of stereotypy encoding strength as a function of γ₈₀ coherence in stereotypy-encoding MSNs (left) and FSIs (right).γ₈₀and encoding strength are directly correlated across all treatment groups in FSIs (linear regression, F_(1,6) = 6.99, R² = 0.54, p < 0.05), but not MSNs (linear regression, F_(1,10) = 0.23, R² = 0.22, p > 0.05).

Striatal Gamma is Locally Generated

The existence of γ-entrained striatal neurons strongly suggests that striatal γ is generated locally. To confirm that these phenomena are local and not due to volume conductionwithin the striatum orfrom the nearby piriform cortex, we further examined γ oscillations at each recording site. We observed bandpassed (50-100 Hz), averaged LFP segments time-locked to γ waveforms peaks recorded at each electrode. This analysis failed to demonstrate any evidence of γ volume conductance, such as peak amplitude degradation as a function of inter-electrode distance or anatomical distance from the piriform cortex. Figure 6 shows a representative example of this analysis in an 8 electrode row implanted in the NAc of an animal given methamphetamine.

Figure 6.

Ventral striatal γ₈₀is locally generated. (A) A representative segment of LFP data before (raw) and after (γ₈₀bandpass) undergoing frequency filtering to isolate γ₈₀ oscillations for further manipulation. Associated PSDs are shown. (B) LFPs recorded from a row of 8 electrodes in the NAc coreduring a representative experiment are shown (right). Data was bandpassed(70-94 Hz) and averaged about waveform peaks identified at the electrode most proximal to the nearby piriform cortex. The γ₈₀ phase and amplitude relationship at each recording site as a function of anatomical proximity (left). Note that γ₈₀ phase does not shift, nor does amplitude degrade as a function of inter-electrode distance, suggesting that high γ₈₀ coherence between recording sites is attributed to common neurotransmissive influence, rather than passive volume conductance. Moreover, no evidence of volume conduction as a function of anatomic distance to the piriform cortex is observed.

Discussion

Reconciling Differences in Psychostimulant Modulation of Gamma

γ oscillations have been proposed to serve as a mechanism by which appropriate neural ensemble spike timing is achieved in the mammalian brain [9]. γ is widely expressed across brain regions, and has been observed in numerous cortical [8,10,11] and subcortical [12]structures, including the striatum [13,14]. In humans, deficits in cortical γ coherence have been linked to several neuropsychiatric disorders, including schizophrenia [15,16]. It is possible that aberrant schizophrenic γ profiles are related to psychosis-associated hyperdopaminergia; however, few studies have explored this possibility.

Rodents studies reveal 2, functionally independent γ bandwidths, a low frequency γ bandwidth (30-70 Hz, peak at ~60 Hz) (γ₆₀) and the higher frequency γ₈₀[8, 17,18].A recent study demonstrated that the psychotomimetic agents ketamine and MK-801 robustly enhance γ₆₀in the neocortex of behaving rats. In contrast, amphetamine or the direct dopamine (DA) agonist apomorphine only mildly increased γ₆₀ [19]. However, a subsequent study examined γ over a wider bandwidth (40-100 Hz) in rats treated with either amphetamine or apomorphine [8]. Both DA agonists were shown to cause a dramatic power shift from γ₆₀toγ₈₀in the cortex and striatum of behaving rats [20].

The current study aimed to expand upon burgeoning work implicating the importance of γ in the NAc [18, 21]. We demonstrate that, unlike its congener amphetamine, methamphetamine profoundly increases γ₈₀without altering its low frequency counterpart. It has been suggested that γ₆₀is primarily associated with olfactory processing in the piriform cortex, whereas γ₈₀is likely related to frontal cortex activity [8]. Moreover, both brain regions send non-reciprocal efferents to the NAc[22,23] and exhibit high levels of striatal γ coherence [20]. Therefore, although the modulation of olfactory processing may differ between amphetamine and methamphetamine, both psychostimulants perturb coordinated frontostriatal activity during motivated behavior[13], which may underlie their similar psychotomimetic properties.

Cannabinoid Signaling Drives Neurophysiological Responses to Methamphetamine

We previously reported that eCB signaling drives methamphetamine-induced changes in NAc firing rates, and that this phenomenon is observed preferentially in putative FSIs[7]. Consistent with these findings, we now show that methamphetamine-induced NAcγ₈₀ also requires CB1 transduction. This may indicate that γ₈₀ is functionally related to local interneuron activity. Indeed, a subpopulation of FSIs has previously been reported to exhibit spiking activity phase locked to γ₈₀[8,13,24]. Likewise, we observed that a subpopulation of FSIs is entrained toγ₈₀. We additionally note that a subpopulation of MSNs and non-FSI interneurons also exhibit robust γ₈₀ entrainment. Taken together, these findings suggest that eCB transmission in striatum may drive coordinated, γ₈₀-tuned neuronal activity. Likewise, our findings add to a growing literature suggesting that CB1 receptor activation is crucially important for network dynamics in the hippocampus and prefrontal cortex [25,26].Our data confirm and extend these findings byshowing that oscillatory brain activity is mediated by eCBs, whereas other studies in freely-moving animals have either failed to see an effect of CB1 receptor antagonists or have only used agonists to delineate an effect of CB1 receptor activation[25,26].

Congruent with aforementioned theoretical frameworks attributing γ₈₀ to cortical efferent activity, methamphetamine increases corticostriatal glutamate release in the NAc[27]. Augmented efferent drive could induce coordinated high frequency oscillations in target striatal neurons, whichis ultimately reflected by extracellular γ₈₀. eCBs could play a permissive role in this sequence. Briefly, CB1 receptor stimulation curbs both glutamatergic and GABAergic neurotransmission [28,29,30,31]; however, in the striatum, GABAergic synapses are more sensitive to cannabinoid inhibition than their excitatory counterparts [32]. It is possible that endogenous CB1 tone acts to dampen inhibitory drive onto corticostriatal neuronal targets, thereby allowing high-frequency excitatory input to dominate their oscillatory patterns. Alternatively, precisely timedCB1 receptor signaling on corticostriatal terminals themselves maytune presynaptic glutamate release such that high frequency, post-synaptic reverberations occur. Further investigation is clearly required if these speculations are to be delineated.

Surprisingly, neither methamphetamine nor rimonabant alteredgrossneuronal γ₈₀ entrainment. By definition, γ₈₀ spike coherence is more readily identified as the analysis incorporates increasing numbers of spikes. This consideration is particularly relevant to striatal units, which typically have low firing rates and γ coherence. It is therefore foreseeable that rimonabant alters γ₈₀ power without altering spike coherence to this frequency band. Since rimonabant is presumably working at presynaptic receptors, its effect is most easily accounted for by presuming that rimonabant interferes with γ₈₀–entrained, presynaptic neurotransmitter release, rather than by directly affecting post-synaptic fluctuations in membrane potential that are already γ-entrained. Because rimonabant has the potential to alter presynaptic signal transduction, without directly affecting post-synaptic properties, post-synaptic entrainment to a residual γ₈₀ signal should not be modulated by rimonabant. Therefore, the degree of γ₈₀ entrainment may be a good predictor for how much stereotypy-relevant information a given neuron is receiving as compared to other neurons, but a poor predictor for how much stereotypy is being expressed behaviorally (which is indeed what we observed).

That methamphetamine fails to increase γ₈₀spikecoherence, despite its robust effect on overallγ₈₀ power, is puzzling. Perhaps methamphetamine augments striatal γnot through local dopaminergic effects, but bypromotingγ₈₀ oscillations in non-striatal brain regions from which the γ₈₀ signal originates (i.e. the prefrontal cortex). Such a mechanism would allow methamphetamine to have robust effects on grossγ₈₀ generation without altering striatal coherence to synaptic input across the γ₈₀ bandwidth. Future investigations of this matter are justified.

Striatal Gamma-Spike Coherenceis Associated with Stereotypy Encoding

Phasic changes in ventral striatal firing patterns have been shown to encode the onset of stereotypy in methamphetamine-treated animals [7]. In contrast, we show that fluctuations in γ₈₀ on an identical timescale fail to encode stereotypic initiation. These negative findings are not entirely unexpected given previous reports demonstrating that phasic increases in γ₆₀, but not γ₈₀, are associated with movement initiation [17, 33].Nevertheless, the finding that γ₈₀ entrainment predicts neuronalstereotypy encoding strength suggests that information required to select behaviorally relevant interneuron ensembles is transmitted across the γ₈₀ bandwidth. That the correlation coefficient for this relationship is low is to be expected;γ₈₀oscillations are in large part an indication of the frequency at which the cortex relays information to the striatum. Therefore vastly disparate data are no doubt embedded within this bandwidth, withstereotypy-related information comprising only a fraction of the total corticostriatal information mass. A similar line of thought may explain why fluctuations in overall γ₈₀ power fail to encode the onset of stereotypy. It is possible that total corticostriatal information flow, and its putative index γ₈₀, remain unchanged during stereotypy, while the portion of γ₈₀ signal pertinent to stereotypic behavior increases relative to its irrelevant counterparts during such behavioral bouts.

Building a Case for the Preferential Involvement of Fast-Spiking Interneurons in Stereotypy

We previously published that FSIs, as opposed to MSNs, preferentially encode the onset of methamphetamine stereotypy [7]. Moreover, multiple research groups have reported γ₈₀ entrainment of striatal FSI spiking activity [8, 9, 12-14]. When our findings are considered within this context, it seems reasonable to posit that γ₈₀-entrained FSI activitycontributes to stereotypy. Indeed, we observed γ₈₀ entrainment in both MSNs and FSIs; however, only FSI entrainment was shown to correlate with a neuron’s propensity to relay stereotypy-pertinent information. It should be noted that this observation likely relied upon small subpopulations of heavily entrained neurons, suggesting that any interpretation rendered here is best entertained with caution. That said, our findings may suggestthat FSIs, rather than MSNs, are preferentially tuned to the stereotypy.

This can perhaps be explained by observed differences in afferent innervation to these two striatal populations. Specifically, it is known that glutamatergic drive excites discrete MSN ensembles through monosynaptic transmission and can silence other MSNs through distal activation of FSIs[34, 35, 36, 37].Our results support the notion that an organization of feedforward inhibition exists, in which corticostriatal input to one site of the FSI network may relay excitation to other FSIs within the ensemble to produce a selective inhibitory influence on MSNs situated at other loci within the same microcircuit. A release of FSI-mediated inhibition can then function as a gating mechanism for MSN action and perhaps this occurs through FSI γ₈₀-entrained input, leading ultimately to the production of stereotypy.

Psychosis May Recruit Endogenous Cannabinoid Signaling

Human studies implicate a role for eCB signaling in psychosis. Recent reports have linked a 9-time AAT triplet repeat in the 3′ region of the CNR1 gene, the gene encoding the human CB1 receptor, to the development of hebephrenic schizophrenia [40,41,42]. Moreover, serum [41,43] and cerebrospinal fluid [41,44,45] levels of the eCBanandamide are elevated in unmedicated, first episode paranoid schizophrenic patients, while levels are normalized in schizophrenic patients receiving typical antipsychotics [41,45].

Animal studies confirm the importance of eCB signaling in psychosis-related behavioral paradigms. Deficits in prepulse inhibition induced by ketamine or MK-801 are ameliorated by CB1 receptor antagonists [46,47,48]. Likewise, CB1 KO mice [6] and wild-type rodents treated with CB1 antagonists [6,7] exhibit reduced psychostimulant stereotypy (although see Martin et al. 2003 [4]). Moreover, local CB1 receptor blockade in the NAc also attenuates stereotypy, implicating the ventral striatum as a fundamental neuroanatomical substrate of cannabinoid-mediated psychosis [7]. This is consistent with our current work, which shows that stereotypy encodingaccompaniescannabinoid receptor-mediated γ. Future experiments should examine whether antipsychotics prevent methamphetamine-induced LFP gamma oscillations and whether CB1 receptor agonists (direct or indirect) increase accumbal γ power. Together, the novel findings presented here support a growing body of literature suggesting that CB1 receptor antagonists may possess clinically relevant antipsychotic properties.

Materials and Methods

All experiments were performed according to United States Public Health Service Guide for the Care and Use of Laboratory Animals and were approved by the Albany Medical College Institutional Animal Care and Use Committee and the University of Maryland School of Medicine Institutional Animal Care and Use Committee.

Animals and Surgery

Adult, male Sprague Dawley rats (n = 10; 300-400 g; Charles River) were individually housed in a 12 hr light/dark environment. Animals were given ad libitum access to food and water. Each animal was surgically implanted with a unilateral, 8 × 2 stainless steel microelectrode array (700 kΩ impedance as measured in saline with a 100 Hz sine wave input; Microprobe Inc.) as previously described [7].Ground electrodes were implanted in the contralateral cortex. Subjects were permitted to recover from surgery for 1 week prior to experiments.

Electrophysiological Recordings

LFP and multiple single-unit activity were simultaneously recorded in awake, behaving rats. Animals were connected to a flexible cable with a printed circuit board headstage containing a micro-operation amplifier for individual unity gain amplification from each electrode. Signals were routed to a differential preamplifier (fixed 50 × gain; Plexon Inc.) and a relayed Multineuron Acquisition Processor (Plexon Inc.) which allows for computer-controlled, channel-specific signal amplification (gain steps 1-30, total gain 1000 × to 32000 x), filtering (second-order 500 Hz low cut, 5 kHz high cut) and analog to digital conversion (32 simultaneous sampling 12-bit converters, 75 kHz). Multiunit signals were referenced to ground, sorted from noise online and spike sorted offline as previously described [7]. LFP signals were referenced to ground on a skull screw in the contralateral occipital bone and collected at a 1 kHz sampling frequency. All raw LFP voltage traces and whole-session power spectral densities (PSDs) were visually inspected to confirm recording quality prior to further analysis.

A subset of recording sites exhibited brief, sporadic bouts of high amplitude 60 Hz noise artifact. To exclude such channels from further analyses could introduce selection bias into our data by which channels exhibiting prominent low frequency γ would be discounted. We instead employed a signal processing strategy through an offline, custom-written (Matlab) artifact filter, which reduced oscillatory signals > 6 to 8 standard deviations from the mean signal to the mean LFP value. Thisoutlier cut-off was guided by signal gain settings used for a given channel, and determined to address high amplitude anomalies, without disrupting lower amplitude 60 Hz signals. Examination of pre- and post-filtered PSDs revealed that this manipulation removed sharp peaks in the PSD at 60Hz without altering non-60 Hz bandwidths. Comparison of filteredPSDs to naturally artifact-free PSDs revealed comparable peaks at 60 Hz, confirming that biologically relevant 60 Hz oscillations remained unfiltered.

Dosing Regimen

Rats were individually habituated daily for 3 days to a custom-modified, sound-attenuated activity chamber (Med Associates) equipped with a rotating commutator to allow for electrophysiological recording and intravenous drug administration during ongoing behavior. On the fourth day, animals were subjected to 20 min of baseline recording. Rats were then administered either rimonabant (0.3 mg/kg; i.v.; n = 4), vehicle (n = 5) or no treatment (n = 1) followed 5 min later by saline (i.v.) and increasing doses of methamphetamine thereafter (0.01, 0.1, 1, 3 mg/kg; i.v.; cumulative dosing; 5 min inter-dose interval). Subjects were then permitted to behave freely until their motor behavior returned to pre-drug values (~ 3-6 hrs).

Behavior

Stereotypy was quantified by a experimenter blinded to treatment as previously described [7]. Briefly, stereotypy was scored as either “present” or “not present” in 1 s bins for the first 75 min of each experiment. A score of “present” was assigned when each of 3 criteria was satisfied: (1) all 4 of the rats paws were stationary, (2) the animal was excessively sniffing and (3) the subject performed at least 3 rhythmic (~ 4-8 Hz) head bobs. Each stereotypy bout began at the onset of the first rhythmic head bob.

Neuron Classification

Striatal neurons were sorted based on their spike waveform shape and firing rates as previously described [7]. In short, neurons with relatively slow spike waveforms (peak width at half maximum > 120 μs; valley width at half minimum > 165 μs) and low firing rates (< 5 Hz) were classified as putative medium spiny neurons (MSNs). Neurons with relatively fast spike waveforms (peak width at half maximum < 120 μs; valley width at half minimum < 165 μs) were classified as putative fast-spiking interneurons (FSIs). Statistical analyses excluded neurons which failed to meet the criteria for either MSNs or FSIs.

Neuronal Encoding of Behavior

Neuronal encoding strength was determined as previously described [7]. Briefly, peri-event raster plots about the onset of either locomotion or stereotypy (+/− 2 s, 100 ms bins) were constructed for each recorded neuron. Peri-event histograms, with firing rates normalized to baseline (−2 to −1 s), were averaged across 15 to 60 trials, depending on the number of high quality behavioral trials that could be isolated in a given animal (see Figure 5A). Only trials in which the criteria for locomotion and stereotypy were not met for a full 2 seconds prior to behavioral initiation were included. Locomotion trials were restricted to behavior occurring after the administration of 0.1 mg/kg methamphetamine, whereas stereotypy trials were restricted to onset times following 3 mg/kg. Neurons whose mean peri-event firing rates exceeded an absolute z score 2 between −1 and +1 s were said to encode the behavior of interest. Mean z scores at time 0 were designated as a neuron’s encoding “strength”.

Power Spectral Densities

LFP rate histograms and standard 2-dimension PSDs were calculated and graphed using NeuroExplorer (Nex Technologies). Power expressed in dB was derived as the log₁₀ of the raw PSD. Power spectrograms were calculated and graphed using custom-written MATLAB scripts (The Mathworks). Whole- session power spectrograms were restricted to frequencies from 0 to 100 Hz and were calculated using a 1 s Hamming window with a 1 s step size. Peri-event power spectrograms were restricted to frequencies from 0 to 100 Hz and were calculated using a 100 ms Hamming window with a 50 ms step size. Changes in highγ power over time were calculated by averaging the PSD from 50 to 100 Hz using a 50ms Hamming window with a 50 ms step size.

Coherence

LFP-spike coherence was calculated and graphed using NeuroExplorer, which employs the following equation for coherence:

$${\mathrm{C}}_{\mathrm{xy}}=\left({{\mathrm{P}}_{\mathrm{xy}}}^{\ast }{\mathrm{P}}_{\mathrm{xy}}\right)∕\left({{\mathrm{P}}_{{\mathrm{xx}}^{.}}}^{\ast }{\mathrm{P}}_{\mathrm{yy}}\right)$$ (1)

where C_xy equals coherence between variables x and y, P_xx equals the average of the squared spectra of x, P_yy equals the average of the squared spectra of y and P_xy equals the product of the spectra for x and y.

For each recorded unit, γ entrainment was statistically determined using a custom bootstrap Matlab function, similar to functions employed in previously published reports [17]. Briefly, for each recorded spike train, inter-spike intervals were randomly shuffled 100 times to create 100 surrogate spike trains with identical inter-spike interval histograms and mean firing rates, but randomly varying spike distributions with respect to local γ. Spike-LFP coherence at 80 Hz for each actual spike train was then converted to a z score referenced to its surrogates’ coherence distribution. A z score greater than 3 was said to indicate significant high γ coherence. Spike-field coherence peak values at 80 Hz wereused for any further analyses of high γ spike coherence.

Histology

Electrode tip placement was confirmed as previously described [7]. Signals recorded from electrodes positioned outside the NAc were excluded from further analysis.

Data Analysis and Statistics

Bar graphs, line graphs, scatter plots were created and all statistical analyses were performed using Prism 5 (GraphPad Software). Statistical tests included standard 1 and 2 way ANOVAs followed by Tukey’s multiple-comparisons test or Bonferroni posttests. The criterion for significance was set to p < 0.05. All error bars represent +/− SEM. Linear regression and peri-event analyses used to assess the relationship between γ and stereotypy were restricted to behavioral timestamps and time bins following 3 mg/kg methamphetamine, as this dose was previously shown to produce robust stereotypy [7].

Drugs

Methamphetamine (Sigma-Aldrich) was delivered in saline. Rimonabant (SR141716A; Research Triangle Institute – National Institute on Drug Abuse) was delivered in a 1:1:18 ratio of ethanol, Emulphor and saline.

Highlights.

• Methamphetamine increases high frequency gamma oscillations in the accumbens

• These effects are attenuated by CB1 receptor blockade

• The endocannabinoid system is required for accumbal ensemble response to methamphetamine