Dopamine (DA) plays a central role in the executive functions undertaken by the prefrontal cortex (Björklund and Dunnett, 2007; Puig et al., 2014), with evidence of involvement in working memory (Chudasama and Robbins, 2004), decision-making (Schweimer and Hauber, 2006; Zhang et al., 2007), and inhibitory control (Navailles et al., 2014). However, physiological mechanisms underlying dopamine action in PFC remain elusive since its effect on different cell-types can be different and even opposite. Recent work proposed a two-stage mechanism by which dopamine D4 receptors (D4Rs) activation exerts opposite effects on pyramidal neurons and GABAergic interneurons in the PFC: (1) in the early phase DA release increases interneuron firing rate, leading to the suppression of PFC output signal; and (2) in the late phase DA promotes a global activity suppression over both pyramidal and interneuron, which could balance PFC network to basal levels (Zhong and Yan, 2016). This finding could explain the cortical hyper-excitability exhibited in D4R-deficient mice (Rubinstein et al., 2001) and provide a physiological mechanism of action for DA on PFC networks. Furthermore, it has been proposed that GABAergic inhibitory interneurons activity in the neocortex has a critical involvement in the generation of rhythms in different frequency bands (McBain and Fisahn, 2001). Yet, the effects of dopamine release on PFC neural activity remain hard to measure directly. The interpretation of electrophysiological recordings after intracerebral dopamine injection is complicated by time lags and concentration gradients related to diffusion, tissue damage, and other caveats. Optogenetic experiments offer a much better alternative, with millisecond precision and cell-mediated release (Tye and Deisseroth, 2012; Rosen et al., 2015). Another method of interest is the light-induced uncaging of dopamine, which we have recently developed as a novel technique for the fast delivery of dopamine in living brain slices (Araya et al., 2013). The technique employs RuBiDopa, a two-photon caged dopamine compound based on ruthenium-bipyridine chemistry that allows for very fast dopamine application by visible light irradiation. These caged compounds (RuBi compounds) release a neurotransmitter with an uncaging time in the nano-second scale and can be used with two-photon (2P) excitation (Filevich et al., 2010; Salierno et al., 2010), thus providing administration of molecules of interest with high temporal and spatial resolution. Until now, this new tool has only been used in vitro. Here we set out to assess the functionality of RuBiDopa on LFP recorded from freely-behaving rats. Other important improvement that optical methods present over the exogenous injection of neurotransmitter is that the latter, due to diffusion limitations, does not correctly represent physiological conditions, while the rapid activation of receptors via optical methods yields activation timing that mimics the tonic (slow) or phasic (fast) release of neurotransmitters (Rosen et al., 2015). Tonic and phasic DA release are underlied by physiologically distinct mechanisms. For instance, while dopaminergic neurons releasing tonic levels of DA (yielding concentrations in the nM scale) are under the influence of GABAergic inhibition, phasic DA release depends on modulation by glutamatergic inputs and yields a much higher, transient DA concentration (hundreds of μM–mM range; Goto et al., 2007). Although dopamine uncaging from RuBiDopa does not mimic synaptic conditions because of the need of diffusion from uncaging site to receptors and subsequent slower clearance of neurotransmitter (which also depends on diffusion), the kind of DA action obtained through this method is physiologically comparable to dopamine receptors activation with tonic dynamics.

LFP oscillations are typically classified into different frequency ranges and have been linked to a myriad of cognitive functions (Buzsáki and Draguhn, 2004; Buzsaki, 2006). The delta band (1–4 Hz) is a staple of slow-wave sleep (SWS) (Hobson and Pace-Schott, 2002), as its amplitude increases to the highest levels in this stage of the sleep-wake cycle. The theta band (5–12 Hz) has been extensively studied in the rodent hippocampus in relation to rapid-eye-movement (REM) sleep (Whishaw and Vanderwolf, 1973; Winson, 1974), spatial navigation (O'Keefe, 1976; Pavlides and Winson, 1989), and memory (Louie and Wilson, 2001; Hasselmo et al., 2002). Theta oscillations are also present in other cortical regions, including the PFC. Importantly, the amplitude of theta oscillations in the PFC and its coupling with hippocampal theta has been related with spatial memory in rodents (Jones and Wilson, 2005; O'Neill et al., 2013) and humans (Canolty et al., 2006; Ferreira et al., 2014; Kaplan et al., 2014). Gamma oscillations (>20 Hz) have been implicated in episodic memory in humans (Nyhus and Curran, 2010), as well as memory encoding and retrieval in the hippocampus of rodents (Colgin and Moser, 2010).

As new analytical tools are developed, it becomes increasingly clear that LFP oscillations can interact in multiple, non-excludent ways (Onslow et al., 2011; Lisman and Jensen, 2013). An interaction of particular interest is phase-amplitude comodulation, which occurs when the amplitude of a higher frequency oscillation is modulated by the phase of a slower rhythm (Tort et al., 2008, 2010; Canolty and Knight, 2010). This phenomenon has been studied in the hippocampus of behaving animals undergoing learning and memory retrieval (Tort et al., 2009), decision-making and working memory (Tort et al., 2008; Axmacher et al., 2010). In this brain region, theta-gamma coupling positively correlates with rodent's performance while learning a new task (Tort et al., 2009) and increases during working memory or retrieval demands, as in the central arm of the T-maze during alternation (Schomburg et al., 2014) and cued tasks (Tort et al., 2008). In human hippocampus, it has been proposed that comodulation plays a role in working memory by representing different items in gamma sub-cycles temporally ordered inside theta oscillations (Axmacher et al., 2010). And Kaplan et al. (2014) also found that theta-gamma coupling between theta phase from mPFC and gamma amplitude of the medial parietal cortex increases during retrieval of a spatial memory. These results suggest that cortical networks in the mammalian brain actively use comodulations to perform neural computation. The role of DA in neural phase signaling has been studied, showing that this neuromodulator plays a key role in triggering and maintenance of high-coherence states of LFP bands across brain regions. Infusion of DA in PFC increased coherence in the theta range between PFC and hippocampus CA1 region (Benchenane et al., 2010). Moreover, coherence between these two areas is impaired in a genetic model of schizophrenia, a pathology highly related to DA dysfunction (Sigurdsson et al., 2010). Mice with hyperdopaminergia showed increased hippocampus-PFC gamma phase synchrony (Dzirasa et al., 2009), while a DA receptor antagonist attenuated theta phase coupling and theta-gamma phase-amplitude comodulation between these regions (Xu et al., 2016).

Given the well-known implication of DA in PFC functioning in health conditions and disease and increasing evidence over the last years about frequency coupling playing a role in different key processes of neural functions, we aimed to study the presence and mechanisms underlying coupling in this brain region. In this work we specifically aimed to probe theta-gamma comodulation in the PFC and whether it is modulated by dopamine. To this end, we performed electrophysiological recordings in the medial PFC of freely behaving animals chronically implanted with multielectrode arrays. Dopamine was uncaged using green light delivered by a fiber optic to the mPFC, and its effect was measured as changes in the theta-gamma and delta-gamma comodulations.

We characterized phase—amplitude coupling patterns of the LFP recorded in the mPFC during spontaneous behaviors and under the influence of dopamine uncaging.

Our results suggest that theta phase modulates the amplitude of gamma oscillations in the mPFC. Also, this modulation is state-dependent, only occurring during active waking; during REM sleep we observed the absence of theta-gamma coupling and reduced power in gamma band (Figure 2D). And finally, it does not depend on the theta amplitude (Figure 5B). We also found that dopamine uncaging by light activation produces a sudden change in the coupling of gamma oscillations, which become modulated by the phase of delta oscillations rather than by the theta rhythm phase. We did not find strong delta-gamma coupling during other states also associated to high delta power, as SWS (Supplementary Figure 3) and quiet waking (Supplementary Figure 1). It should be noticed that the comodulation seen during SWS (Supplementary Figure 3), despite present, is not structured into a clear comodulation island and has a small effect size. We also showed that the observed increase in delta-gamma MI is not the result of an increase in the power of the frequency bands involved. For instance, PSDs of all the channels analyzed show no changes before and after laser onset, and subtle changes found when extracting the bands of interest for the different animals shows decreasing rather than increasing power. Also, it has to be noted that delta-gamma coupling is stronger for small delta amplitude values (Figure 4B), suggesting that the increase in MI does not involve a signal gain in the frequency band being modulated.

While the small sample size is an important limitation of the study, the data altogether suggests that the shift in comodulation from theta to delta observed after RuBiDopa uncaging can be attributed to the neurotransmitter dopamine acting on dopamine receptors in the mPFC. Another caveat worth mentioning is that theta-gamma comodulation was only present during the limited periods of time when the animals were actively exploring the environment. It is therefore difficult to reach a definite conclusion regarding dopamine-related changes in theta-gamma comodulation. On the other hand, the increase in delta-gamma comodulation caused by dopamine uncaging was quite robust. As it was stated in the Methods section, the experimental conditions reached in this work imply that the dopamine uncaging only can be done in the waking stage of the sleep-wake cycle. Therefore, it remains an open question if DA uncaging in the mPFC would also underlie delta-gamma coupling during sleep states.

Regarding the dopamine concentrations reached in these experiments, we did not expect differences in the effect observed in electrodes at different distances from the fiber, since we applied light of such power so as to saturate the uncaging capacity of dopamine from RuBiDopa at the concentration indicated in the experiment. Hence, possible differences in light intensity reaching the electrodes did not interfere with dopamine concentrations reached around them. Regarding kinetics and concentration of DA in physiological conditions, this is an issue that have been controversial over the years, as there has not been consensus about the factors that govern and determine DA concentrations in DA synapses. As most DARs are located extrasinaptically, DA needs to “spillover” the synaptic release site to be able to act on receptors. This spillover is known to be mediated by different mechanisms such as uptake by DA transporters (DATs) and diffusion. The kinetics of these two processes compete and diffusion seems to have more influence in determining the net extracellular concentration of DA and thus the amount of DA that reaches DARs (Cragg and Rice, 2004). Factors such as diffusion rate, distance between release site and receptors, and also DATs-mediated uptake, will determine if the amount of DA present at the DAR is sufficient to activate it and gate a response. In vitro studies determining affinity constants and EC₅₀ values for DARs have shown that receptor activation occurs in a range between 10 nM (high-affinity states) and 1 μM (low-affinity states; Richfield et al., 1989; Neve and Neve, 1997). The final dopamine concentration in the present experiments is difficult to calculate properly because of lack of control of diffusion through the brain tissue. Nevertheless, taking into account the diffusion factor from injection site to electrodes (which we roughly estimate in a factor of 10), and knowing that with a light intensity of 1 mW (see Methods) all molecules of RuBiDopa will release DA, we estimate the DA concentration reached in this work to be in a value between 25 and 30 μM. Thus, this concentration should be highly efficient in activating DARs in low as in high-affinity states. It is well-established that DA release from dopaminergic neurons can be slow, low-amplitude and constant (tonic) or fast, high-amplitude and transient (phasic). These two mechanisms are physiologically different and importantly, can trigger opposite responses in postsynaptic neurons (Rosen et al., 2015). The kind of DA action obtained through RuBiDopa uncaging in this work is physiologically comparable to dopamine receptors activation with tonic dynamics. Physiologically, there are significant differences in DA concentration intrasynaptically (where 25–30 μM could be in the range or even lower) and extrasynaptically, where this represents a rather substantial increase beyond the normal 10–20 nM of this compartment. Therefore, this could be a constraint in the interpretation of the data, since concentrations reached in this work, despite being sufficient to activate DARs in extrasynaptic space, are also significantly higher than in physiological conditions.

Previous studies with RuBi compounds did not show any toxic effect at the concentration used in this work (300 μM) (Rial Verde et al., 2008; Lopes-dos-Santos et al., 2011; Araya et al., 2013). It should also be noted that the animals used in the present study showed normal LFP and spike recordings for up to a month (data not shown), which reflects intact brain functionality in the region where RuBiDopa was administered for experiments. This suggests that the caged compound did not affect physiology through a toxic effect. As stated before (see Methods), it is not possible to use an “empty” caging molecule as a negative control for our experiments. Nevertheless, future experiments administering DA antagonists remain necessary as a negative control.

Our work also suggests that frequency band power in the delta and gamma ranges does not significantly change as a result of DA uncaging. For instance, during REM sleep, theta power is high but theta-gamma coupling is absent (Figures 2C,D), and during SWS the same is valid for delta band (high delta power and low delta-gamma coupling; Figure 3A, Supplementary Figure 3). Figures 4B, 5B also show that delta and theta amplitude are not positively correlated with the comodulation strength in the mPFC respectively. Previous articles point to the importance of coupling strength alone in cognitive functions such as learning and memory retrieval, decision-making and working memory (Tort et al., 2008, 2009, 2010; Axmacher et al., 2010; Kaplan et al., 2014). Our results suggest that dopamine release in the mPFC could act not by increasing amplitude gain of slower oscillations, but by fine-tuning inhibitory networks (Zhong and Yan, 2016), which interact with principal excitatory cells (Tseng and O'Donnell, 2004) and generate gamma oscillations (McBain and Fisahn, 2001; Bartos et al., 2007).

It is known that abnormal dopamine activity plays a major role in the pathophysiology of schizophrenia (hyperdopaminergia in the mesostriatal pathway, frontal hypodopaminergia in the mesocortical pathway), and impaired neuronal synchronization within cortical networks in the gamma frequency band has been proposed to contribute to several psychiatric disorders also characterized by cognitive deficits. In the PFC, it has been proposed that dopamine modulates working memory capacity (Seamans and Yang, 2004). Dopamine has also been shown to modulate the gamma band, which is involved in cognitive function (Furth et al., 2013). Moreover, a shift in the phase-frequency of the comodulation to slower rhythms is involved in the increase in working memory capacity in the hippocampus (Axmacher et al., 2010). Shifting the modulating frequency of comodulation from theta to delta rhythms as a result from dopamine action can reflect synchronization features between mPFC and ventral tegmental area VTA. Peters et al. (2004) showed that VTA slow oscillations drive PFC pyramidal neurons up-states, and both structures are coherent in the delta range. Also, VTA inactivation by lidocaine injection lead to suppression of slow oscillations in the PFC. Therefore, dopamine uncaging in our work could mimic the dopaminergic component of the VTA-PFC projection. Our results are in accordance with previous results showing that DA has a key role in triggering and maintaining synchrony and coupling between different LFP bands and across different brain regions (Dzirasa et al., 2009; Benchenane et al., 2010; Sigurdsson et al., 2010; Xu et al., 2016).

The present data also constitute the first evidence of theta-gamma comodulation in the mPFC of freely behaving rats during active exploration, in agreement with recent results in mice (Zhang et al., 2016). Also, the lack of theta-gamma comodulation in the mPFC during REM sleep, despite the presence of theta rhythm generated in the hippocampus during this state, has recently been described (Zhang et al., 2016) and remains to be elucidated. Systemic features of REM sleep, such as the interruption of sensory inputs or the shut-down of the noradrenergic/serotonergic drive, are unlikely to explain this absence, because theta-gamma comodulation occurs during REM sleep in the hippocampus (Scheffzük et al., 2011; Scheffer-Teixeira et al., 2012). While hippocampal ripples and mPFC spindles episodes are synchronized during SWS (Siapas and Wilson, 1998), the present results along with the observations that spike-timing correlations between HPC and PFC decrease (Wierzynski et al., 2009) and theta—gamma coupling increases in the hippocampus during REM sleep (Scheffzük et al., 2011) suggest a decoupling event between both structures.

Plastic changes related to DA input may also underlie delta-gamma comodulation observed after laser activation of RubiDopa or DA injection. DA fibers terminals arising from VTA innerve the same area as hippocampal inputs to PFC (Carr and Sesack, 1996). Also, it was shown that DA injection and VTA stimulation modulate long term potentiation and depression in the hippocampal—PFA synapses (Jay et al., 1996, 2004; Gurden et al., 1999). Our current investigation does not allow us to completely elucidate which events are at play; further research, therefore, must be performed to address the exciting possibility that theta-to-delta coupling change may reflect a PFC gating mechanism for controlling external influences.

The low number of animals presented in this study is a highly significant limitation in the conclusions we can obtain from our results. Nevertheless, the effect observed after RuBiDopa uncaging is robust and we propose these findings as the preliminary stage from which further investigation can be founded. Future directions point to the replication of these results and additional experiments aiming to strengthen our conclusions and broaden our understanding about the role of DA in the modulation of cognitive processes in the prefrontal cortex.

The study was designed by VA, RE, and SR. The chemical syntheses were performed by VA and RE. The experiments were performed by VA and AS. The analyses were performed by VA and RS. The figures were prepared by VA, RS, and SR. The manuscript was written by VA, RS, AT, and SR.