Developments from Bulk Optogenetics to Single-Cell Strategies to Dissect the Neural Circuits that Underlie Aberrant Motivational States

Abstract

Motivational states are regulated by complex networks across brain regions that are composed of genetically and functionally distinct neuronal populations. Disruption within these neural circuits leads to aberrant motivational states and are thought to be the root cause of psychiatric disorders related to reward processing and addiction. Critical technological advances in the field have revolutionized the study of neural systems by allowing the use of optical strategies to precisely control and visualize neural activity within genetically identified neural populations in the brain. This review will provide a brief introduction into the history of how technological advances in single-cell strategies have been applied to elucidate the neural circuits that underlie aberrant motivational states that often lead to dysfunction in reward processing and addiction.

Determining causal relationships between neural function and behavior is crucial to understand the neuropathology underlying aberrant motivational states.⁵ Studies that discovered the relationship between motivated behaviors and specific brain regions were historically accomplished by tissue lesioning techniques, electrical stimulation, and pharmacological activation or inactivation. While these methods were crucial to uncover the basic neuroanatomical components that mediate motivated behavior, they were unable to determine how a specific neural pathway or neuronal cell type mediates a given behavioral response. Site-directed pharmacological manipulations were initially able to study genetically defined pathways; however, this was only possible if a given population of neurons locally expressed a specific receptor. Pharmacological manipulations also last long periods of time, thus preventing the ability to study how neural activity is required for discrete behavioral events that occur at a much shorter timescale. Optogenetics circumvented this problem by allowing for the manipulation of brain circuitry with millisecond precision. Single-unit electrophysiological recordings have also circumvented the limitation of time resolution by allowing the comparison of behavioral events to single-action potentials within individual neurons. However, advances in single-cell RNA-sequencing techniques have shown us that neurons can be segregated into genetically identified cell types, thus exponentially expanding the anatomical blueprint used in systems neuroscience to study the relationship between motivated behaviors and brain function. In combination with the recent advances that have occurred in the field of single-cell calcium imaging, we now have the capability of tracking the activity dynamics of thousands of genetically identified cells with single-cell resolution across days, weeks, and months. Therefore, these tools have catalyzed discoveries that elucidate the precise neural function underlying aberrant motivational states that often lead to dysfunction in reward processing and addiction.

BULK OPTOGENETIC STRATEGIES TO MANIPULATE NEURAL CIRCUITS OF MOTIVATION

The application of optogenetic strategies for rodent behavioral studies have revolutionized the field of systems neuroscience (Boyden et al. 2005). The most commonly used opsin to activate neural circuits is channelrhodopsin-2 (ChR2). ChR2 is a light-gated cation channel that was originally isolated from blue-green algae (Nagel et al. 2003). ChR2 is maximally activated by blue light (450–490 nm wavelength). When activated, absorbed photons cause a light-induced isomerization of the all-trans retinal protein, which leads to the opening of a channel allowing sodium and other cations to flow through the cell. When expressed in a neuron, this influx of cations causes depolarization of the cell membrane at resting membrane potentials, which leads to the opening of endogenously expressed voltage-gated sodium channels to initiate an action potential. Red-shifted channelrhodopsin proteins have also been developed, which allow for the possibility of exciting two genetically distinct populations of neurons within the same brain site and/or facilitating combination with other optical techniques simultaneously (Zhang et al. 2008; Hegemann and Möglich 2011; Yizhar et al. 2011; Pégard et al. 2017).

Optogenetic inactivation of neural circuits is most commonly accomplished using the light-gated chloride pump, halorhodopsin (NpHR), which was first discovered in archaebacteria (Matsuno-Yagi and Mukohata 1977). Introduction of wild-type NpHR into neurons demonstrated that photoinhibition was possible, but initially, exogenous NpHR protein was not sufficiently expressed at neuronal membranes for consistent results in vivo (Gradinaru et al. 2010). Further modification of NpHR (NpHR3.0), with an added endoplasmic reticulum (ER) export signal and membrane trafficking peptide sequence, results in robust expression at neuronal membranes, which facilitated its use in vivo for neuronal circuit element inhibition (Gradinaru et al. 2010). NpHR3.0 is maximally activated by a yellow/orange light (∼590 nm wavelength) but can respond to a broad wavelength range that extends to green light (∼520–620 nm wavelength). When activated, NpHR3.0 pumps chloride from the extracellular space into the cytoplasm of the cell. When expressed in a neuron, this results in hyperpolarization of the cell membrane, and can decrease neuronal firing rates (Fenno et al. 2011). Optical inhibition can also be achieved by the use of outward proton pumps, such as Arch (Chow et al. 2010; Fenno et al. 2011). Arch is maximally activated by a 560 nm wavelength of light, and activation of Arch has been shown to result in robust currents at relatively low light outputs (Chow et al. 2010). However, Arch has been shown to lead to an equally robust rebound excitation when inhibition is conducted at presynaptic terminals (Mahn et al. 2016), thus limiting its use and feasibility for circuit dissection.

Expressing opsin proteins under the control of cell-type-specific promoters is a method to manipulate genetically defined neuronal subtypes without transgenic mouse lines. For example, early studies using this technique manipulated projections from glutamatergic basolateral amygdala (BLA) neurons to the nucleus accumbens (NAc) and found this pathway to be important for reward seeking (Stuber et al. 2011). Calcium-calmodulin-dependent protein kinase IIα (CamKIIα) is preferentially expressed in glutamatergic projection neurons in the BLA (McDonald 1992). ChR2 or NpHR3.0 was introduced into these glutamatergic neurons using an adeno-associated virus (AAV) vector with the opsins under the control of a fragment of the CamKIIα promoter. Stereotaxic injection of viral constructs encoding these proteins into the BLA results in opsin-positive neurons constrained to glutamatergic projection neurons within the BLA. As discussed in detail below, implantation of an optical fiber into the NAc, allows for precise control over excitatory BLA inputs into the NAc. Other studies using the CamKIIα promoter have elucidated multiple circuit components necessary for regulating motivational states (Stuber et al. 2012; Britt and Bonci 2013; Nieh et al. 2013; Juarez et al. 2019).

A transgenic approach was initially a common method to achieve targeted ChR2 or NpHR expression in genetically defined cells (Arenkiel et al. 2007; Zhao et al. 2011). While this method ensures that virtually all neurons of specific genetically defined population will express opsin proteins, it oftentimes does not provide anatomical specificity of expression to a discrete brain region of interest. Thus, to reliably target neuronal populations within specific brain nuclei, Cre recombinase-inducible expression systems have been used in conjunction with transgenic animals expressing cre in specific populations of neurons. Using this method, Cre-inducible opsins are stereotaxically injected into transgenic rodents expressing Cre recombinase in genetically identified neuronal populations (Atasoy et al. 2008; Cardin et al. 2009; Sohal et al. 2009; Tsai et al. 2009; Witten et al. 2011). Cre-inducible AAV vectors contain DNA cassettes with two pairs of incompatible lox sites (LoxP and lox2722) with an opsin inserted between the two lox sites in the reverse orientation. Cre recombinase catalyzes recombination between the two lox sites, resulting in the opsin reversing its orientation, allowing messenger RNA (mRNA) of the opsin to be transcribed. Delivery of these Cre-inducible opsins into a specific brain region results in opsin expression in only the genetically identified cell type in the brain region of interest. Cholinergic interneurons in the NAc have been targeted using this method (Witten et al. 2010). Here, bacterial artificial chromosome (BAC) transgenic choline acetyltransferase (ChAT)::Cre mice were injected with a cre-inducible double-floxed recombinant AAV vector coding for ChR2 or NpHR3.0 into the NAc. Dopaminergic (DAergic) neurons in the ventral segmental area (VTA) have also been targeted using a transgenic approach in which tyrosine hydroxylase (TH)-Cre (Tsai et al. 2009) in mice or rats (Witten et al. 2011) or dopamine transporter (DAT)-cre mice (Stuber et al. 2010; Cohen et al. 2012) are injected with a double-floxed cre-inducible opsin vector. The use of cre-mice paired with double-floxed opsins, or the use of cell-type promoters, allows for precise control over genetically defined populations of neurons.

Although we have focused our discussion on optogenetics, similarly notable advances have been made in the application of chemogenetic strategies with the application of designer receptors exclusively activated by designer drugs (DREADDs) to dissect the neural circuits of reward and addiction, albeit losing the temporal resolution obtained with optogenetics but gaining tractability for being a more physiologically relevant approach (Roth 2016).

COMBINING BULK OPTOGENETIC STRATEGIES WITH SLICE ELECTROPHYSIOLOGY TO PARCEL OUT CIRCUITS OF MOTIVATION

Anatomical tracing studies and electrophysiological techniques using electrical stimulation have been historically used to study the synaptic connectivity within neural circuits. Electrophysiological studies using electrical stimulation can address functionality but cannot determine the connectivity that emerges from cell-type-specific projections, since electrical stimulation will typically activate all afferents from a given neuron. Patch clamp electrophysiology paired with optogenetics circumvents the limitations associated with both of these methods because it allows for cell-type-specific activation and assessment of the strength and functionality of these connections. Using this method, we can photoactivate genetically identified neurons that express ChR2 and record from other genetically identified postsynaptic neurons using mice expressing fluorescent proteins within the target neuron or by performing post hoc immunohistochemistry. These techniques have been successful in parsing out neural circuits involved in driving aberrant motivational states. In early examples of this application, optogenetics was used in NAc brain slices to define the functional connectivity of medium spiny neurons. By conditionally expressing ChR2 in medium spiny neurons, these authors were able to investigate connections within the striatum and projections to the globus pallidus and substantia nigra, as well as examining how striatal cholinergic interneurons can regulate function of other populations of striatal neurons (Chuhma et al. 2011; English et al. 2012).

Optogenetics paired with slice electrophysiology has also been used to examine the possibility of neurotransmitter co-release. DA and glutamate coincident signaling is crucial for a variety of motivated behaviors including responding to motivationally significant stimuli. A subset of TH-positive DA neurons in the VTA also express vesicular glutamate transporter-2 (VGluT2), indicating that these DA neurons are capable of packaging glutamate into synaptic vesicles (Hnasko et al. 2010). Furthermore, pharmacological and electrophysiological studies have suggested that DA neurons co-release glutamate (Sulzer et al. 1998; Bourque and Trudeau 2000; Chuhma et al. 2009); however, these studies only provided indirect evidence because of technical limitations. Selective optogenetic stimulation of ChR2-positive DAergic terminals in the NAc shell results in excitatory postsynaptic currents (Stuber et al. 2010; Tecuapetla et al. 2010), confirming that midbrain DA neurons are capable of co-releasing glutamate in the NAc. Similar studies have now confirmed that other neurons that release neuromodulators, such as acetylcholine, are also capable of glutamate co-release, such as projection neurons in the medial habenula (Ren et al. 2011).

SINGLE-CELL GENOMICS STRATEGIES TO IDENTIFY THE NEURAL CIRCUITS OF MOTIVATION

To achieve precise functional specificity, the adult brain is comprised of a diverse catalog of neuronal cell types that exhibit unique functional and morphological complexity. All neurons share the same genome within an individual, and interregional and intraregional neuronal diversity occurs from distinct patterns of gene expression across distinct cellular populations. Thus, the observation that unique transcriptional features can be used to discriminate distinct functional populations has significant precedent for systems neuroscience research and the study of reward encoding and addiction (Lewis et al. 1988; Lightman 1988; Edlund and Jessell 1999; Fremeau et al. 2001; Shirasaki and Pfaff 2002; Sunkin et al. 2012; Zeisel et al. 2015; Tasic et al. 2016). However, it has become increasingly recognized that neuronal subtypes are often marked by either unknown individual features or a multiplexed combination of features, thereby requiring sufficient numbers of unique cells to be profiled to achieve sufficient tissue-level resolution. Recently, a number of high-throughput genomic strategies have enabled the examination of thousands-to-millions of cells simultaneously at sufficient depth (Klein et al. 2015; Macosko et al. 2015; Cao et al. 2017; Rosenberg et al. 2018). This revolution in high-throughput single-cell sequencing approaches, originally catalyzed by specialist methods, have recently reached a point in which they have become accessible to most laboratories both in terms of technical prerequisites and cost. Although this discussion will be focused on single-cell RNA sequencing, similarly notable advances have been made in other avenues of single-cell genomics, including single-cell DNA sequencing (Lan et al. 2017) and single-cell epigenomics (Cusanovich et al. 2015; Stevens et al. 2017).

High-throughput single-cell sequencing approaches can generally be grouped into droplet-based (Klein et al. 2015; Macosko et al. 2015) or plate-based methods (Campbell et al. 2017; Cao et al. 2017; Moffitt et al. 2018; Rosenberg et al. 2018). Droplet-based methods use microfluidic devices to deliver cells, reagents, and barcoded oligonucleotide-coated microparticles into nanoliter-sized emulsions that serve as a compartment for the capture of polyadenylated transcripts. These barcoded microparticles are then pooled and subjected to solid-phase complementary DNA (cDNA) amplification to ultimately generate sequencing libraries. Rather than relying on microfluidic devices to isolate cells, plate-based strategies instead generate multiplexed sequencing libraries by distributing cells into 96-well plates and performing multiple successive rounds of in situ reverse transcription and barcode ligation with pooling and redistribution of cells between each round. Thus, both droplet-based and plate-based strategies are able to generate next-generation sequencing libraries from thousands-to-millions of cells in which each library molecule is barcoded by transcript-, cell-, and sample-of-origin.

The choice of droplet-based versus plate-based strategy depends on factors including the amount of starting material, the required depth of transcriptome coverage, the number of cells to be achieved, cost, and convenience. In general, plate-based strategies require considerably more starting material and sacrifice depth of coverage but can be used to produce data from a greater number of individual cells at a significantly lower cost. Although home-built droplet microfluidic-based strategies often demand a significant amount of technical expertise and investment, they have recently been largely replaced by commercial options that dramatically simplify the technical process, provide consistent performance across datasets, and generate high-quality data at greater depth (Kim et al. 2019; Sharma et al. 2020). However, such commercial options require specialized proprietary equipment and significant reagent costs, dramatically increasing the cost per cell. Ultimately, however, the choice of droplet-based or plate-based strategy depends mostly on experimental requirements. For small brain nuclei or samples that are cost- or time-prohibitive to generate, commercial droplet-based strategies may be most appropriate, whereas plate-based strategies may be best suited to studies of large brain regions requiring transcriptomes to be generated from a very large number of individual cells.

In terms of data robustness, the primary factor determining the statistical resolution of feature identification and cell-type classification is the balance between depth of transcriptome coverage and total number of cells (Svensson et al. 2017; Zhang et al. 2020). Very roughly, shallow depth can be compensated for by generating transcriptomes from a large number of cells, while a fewer number of cells in a dataset can be overcome by sampling each cell's transcriptome more deeply. Ultimately, however, high-throughput, single-cell RNA-sequencing methods are considered shallow as compared to manual low-throughput methods (Zeisel et al. 2015; Tasic et al. 2016; Hodge et al. 2019), and this lack of depth is overcome by observing and statistically grouping a large number of cells based on transcriptional similarity. The size and complex nature of data generated from single-cell RNA-sequencing experiments has mandated the development of novel computational and analytical approaches to interpret the data (Macosko et al. 2015; Qiu et al. 2017; Ziegenhain et al. 2017; Butler et al. 2018; Vieth et al. 2019), and standardized analytical pipelines are available and validated for most routine applications.

Initial studies of single-cell RNA sequencing have been used to identify transcriptionally discrete subtypes of neurons within anatomically defined brain regions important for motivated behaviors, such as the cortex, striatum, and hypothalamus (Gokce et al. 2016; Tasic et al. 2016; Campbell et al. 2017; Chen et al. 2017; Moffitt et al. 2018; Kim et al. 2019; Mickelsen et al. 2019; Rossi et al. 2019; Sharma et al. 2020). In these studies, data is initially filtered to remove low-quality cells and multiplets normalized across cells, corrected for technical artifacts, and subsequently clustered based on reduced gene expression components. Canonical features are then used to identify clusters corresponding to resident cell types in the tissue, and differential gene expression tests between clusters are used to identify specific transcriptional features for each population or subpopulation of cells. In this way, neuronal populations can be discriminated by identifying transcriptional features that label discrete subtypes of neurons.

Similar differential gene expression tests can be used to identify cell-type-specific transcriptional representations generated by an experimental manipulation, such as diet-induced obesity and chronic exposure to morphine (Avey et al. 2018; Rossi et al. 2019). Furthermore, neuronal subpopulations activated by an acute stimulus such as an acute exposure to morphine, stress, or a social stimulus can be discovered by examining the distribution of immediate early gene expression across neuronal subtypes (Wu et al. 2017; Hrvatin et al. 2018; Kim et al. 2019). Neuronal projection populations can also be identified by virally labeling projection populations and examining the distribution of virally encoded sequences within transcriptionally defined subtypes of neurons at the projection source (Kebschull et al. 2016). More complex analyses can also be performed to examine cell-state changes either during neuronal lineage specification or in response to a chronic perturbation (Qiu et al. 2017; La Manno et al. 2018; Kanton et al. 2019; Rossi et al. 2019; Sharma et al. 2020). Roughly, these methods seek to identify transcriptional trajectories across populations of cells either by ordering cells according to transcriptional similarity (Qiu et al. 2017) or by determining the direction of gene expression changes by comparing gene-wise abundances of nascent and mature transcripts to assign weighted transcriptional vectors across cells (La Manno et al. 2018). As a note of caution, however, chronic manipulations can result in promiscuous transcriptional alterations across neuronal populations within a sampled tissue, and unlike acute challenges, prioritizing perturbed neuronal subtypes in response to chronic treatments can prove difficult. Consequently, datasets produced from such chronic experiments may demand to be interpreted as exploratory evidence rather than conclusive results, unless the experiment has been deliberately designed to test a specific molecular hypothesis.

Moreover, single-cell genomic strategies such as single-cell RNA sequencing have become readily accessible to the nonspecialist both in terms of cost and technical expertise in recent years. In the field of reward processing and addiction, single-cell RNA sequencing can be leveraged to identify precise transcriptionally defined neuronal subtypes within brain regions of interest, describe which subtypes serve as functional substrates during experimental conditions, and to distinguish specific projection populations within heterogeneous brain regions. These methods are poised to become an essential tool in systems neuroscience that will dramatically aid in dissecting the circuit that regulates aberrant motivated behaviors.

SINGLE-CELL OPTICAL IMAGING STRATEGIES TO STUDY THE NEURAL CIRCUITS OF MOTIVATION

Recent advances in single-cell calcium-imaging approaches have yielded unprecedented precision to study the neural circuits of motivated behaviors within genetically identified neuronal populations. The use of genetically encoded calcium indicators, such a GCaMP (Chen et al. 2013), in combination with two-photon microscopy (Denk et al. 1990) and head-mountable miniature microscopes (Ghosh et al. 2011) has again revolutionized the field of behavioral neuroscience. The combination of these tools allows for single-cell calcium imaging and provide novel means to measure neuronal activity. Two-photon microscopy allows for calcium imaging in awake head-fixed animals with very high spatial resolution. Head-mountable miniature microscopes have less spatial resolution; however, they allow for calcium imaging to be performed in freely moving unrestrained animals. These approaches lack the exquisite temporal precision of in vivo electrophysiology as they do not directly measure the voltage change due to action potential firing. Instead, they rely on measuring calcium concentration changes in the cell, and rely on this signal as a proxy for neuronal activity (Wei et al. 2019). Despite the loss of temporal precision, imaging approaches come with a unique set of advantages. Single-cell imaging approaches allow the recording of a large number of genetically or projection-defined neurons simultaneously (McHenry et al. 2017; Otis et al. 2017; Namboodiri et al. 2019; Rossi et al. 2019; Resendez et al. 2020), whereas phototagging approaches using in vivo electrophysiology result in much lower yields (Hangya et al. 2015). Single-cell imaging approaches, especially two-photon calcium imaging, allows the longitudinal tracking of activity of the same set of neurons across many days of testing, such as learning to associate an auditory cue with a reward delivery (Otis et al. 2017; Namboodiri et al. 2019). The high spatial resolution from a two-photon microscope also allows the recording of subcellular neural structures, such as distal axonal targets (Otis et al. 2019). Overall, these advantages have driven the rapid adoption of such approaches in recent years in systems neuroscience research.

As mentioned above, single-cell calcium imaging can be performed using two largely different modalities. In one, the imaging is performed using a tabletop two-photon microscope, which requires head fixation of animals. Within this approach, different laboratories have either used chronic windows for surface cortical imaging (Komiyama et al. 2010; Huber et al. 2012), implantation of optical cannulae for brain regions at an intermediate depth (Kaifosh et al. 2013; Otis et al. 2017), or implantation of gradient refractive index lenses for deep brain regions (McHenry et al. 2017; Namboodiri et al. 2019; Rossi et al. 2019; Resendez et al. 2020). In the other approach, the imaging is performed using a miniaturized microscope (often called a “miniscope”) in combination with single-photon fluorescence excitation (Ghosh et al. 2011). The primary benefit of such an approach is that the animals are able to freely move during behavior (albeit being typically tethered). However, this benefit comes at the cost of lower resolution and, hence, higher potential for cross talk between nearby neurons and subcellular structures (Ghosh et al. 2011; Zhou et al. 2018).

Both of these approaches have been used to study neuronal mechanisms underlying processes related to our understanding of aberrant motivational states as they relate to reward processing and addiction. For instance, two-photon imaging approaches have been used to study the role of distinct projection pathways in the medial prefrontal cortex (mPFC) (Otis et al. 2017), ventral/medial orbitofrontal cortex (vmOFC) (Namboodiri et al. 2019), or medial thalamus (Otis et al. 2019) in cue–reward associative learning and memory. These studies have collectively highlighted projection-specific encoding of information related to cues predictive of rewards. Indeed, one study even found that there is an explicit correlate of a long-term cue–reward memory in the activity patterns of vmOFC neurons projecting to the VTA, despite complete behavioral extinction of the cue–reward association (Namboodiri et al. 2019). In other words, despite a previously reward-associated cue no longer being predictive of reward, these neurons (vmOFC→VTA) respond to the cue the same way they responded when the cue was predictive of reward. While such a long-term memory of the original association has long been known to exist, the ability to longitudinally track neuronal activity across days made it possible to identify an explicit correlate of this memory in vmOFC neurons. Of course, such approaches can be easily extended to study drug-related cognitive processes to study neuronal network mechanisms underlying extinction, reinstatement, renewal, or memory consolidation and reconsolidation.

While longitudinal tracking is more challenging under freely moving settings, freely moving imaging approaches have been used to study the role of genetically defined neuronal populations or specific projections within the circuits that regulate reward processing and drug-related behaviors. One of the first studies to investigate reward-related behaviors using miniscopes in combination with a GRIN lens to image deep in the brain found that GABAergic neurons in the lateral hypothalamus (LH) encode both appetitive and consummatory behaviors (Jennings et al. 2015). However, these processes were encoded by distinct populations of LH GABAergic neurons, as the neurons that encoded appetitive behaviors did not spatially overlap with those that encoded consummatory behaviors. Another more recent study used miniscopes and found that a projection from mPFC to the periaqueductal gray is related to the development of compulsive alcohol-drinking behavior (Siciliano et al. 2019). Using a similar approach, another study found that degradation in spatiotemporal representations within single neurons in the rat infralimbic cortex that project to the NAc are correlated with the incubation of cocaine craving (Cameron et al. 2019). Furthermore, miniscope recordings in freely moving mice have also been used to identify that hippocampal CA1 neuronal ensembles encode an association between a spatial location and the availability of nicotine (Xia et al. 2017). While there are many such published studies, the above examples highlight the exciting range of possibilities for uncovering neuronal network mechanisms underlying reward encoding and drug-related processes.

CONCLUSIONS AND FUTURE DIRECTIONS IN THE APPLICATION OF ALL-OPTICAL STRATEGIES TO STUDY THE NEURAL CIRCUITS OF MOTIVATION

Single-cell sequencing methods are expanding the blueprint for precise circuit dissection of the neural circuits involved in reward processing and addiction. The field of miniscope calcium imaging is evolving rapidly with new adaptations that allow us to study the neural circuit components that regulate motivated behaviors in very novel ways. For example, we can simultaneously perform calcium imaging at two distal cites (de Groot et al. 2020) to compare how activity dynamics of genetically identified neurons in two regions are related to each other. We can combine calcium imaging with optogenetics to study circuit interactions in freely moving animals (Stamatakis et al. 2018). The use of wireless/untethered calcium-imaging approaches will allow us to improve the study of behaviors that are impaired by having an animal tethered (Shuman et al. 2020). Further advances in the field also allow us to perform real-time analysis of single-cell calcium imaging in freely moving animals (Friedrich et al. 2020), opening up an entirely new realm of possibilities for closed-loop experiments to further understand how a distinct ensemble of neurons interact to regulate motivational states related to reward processing and addiction.

Perhaps the biggest advantage yet of using imaging approaches to monitor neuronal activity is that over the last couple of years, it has become possible to selectively stimulate or inhibit neurons with identified activity patterns in a task using two-photon microscopy (Pégard et al. 2017; Carrillo-Reid et al. 2019; Chettih and Harvey 2019; Jennings et al. 2019; Marshel et al. 2019). For instance, one study demonstrated that OFC neurons related to feeding interact with other neurons related to social behaviors (Jennings et al. 2019). Intriguingly, activation of just ten identified feeding-related neurons was sufficient to cause changes in licking behavior. This raises the possibility that many behaviors may be under the control of a select subset of neurons that can be identified using task-related activity patterns. As of now, such selective single-cell photostimulation is only possible under head-fixed two-photon microscopes. Yet, it is conceivable that such approaches may be extended to freely moving settings in the near future.

Advances in single-cell tools used for the identification, manipulation, and visualization of neural circuits involved in reward and addiction have aided in supporting and refuting many hypotheses that were previously untestable due to technical limitations associated with traditional techniques. Therefore, the new wave of single-cell tools will undoubtedly allow us to ask questions that were previously unreachable and permit us to enter a new era of systems neuroscience research.