Viral vectors were originally developed to deliver genes into host cells for therapeutic potential. However, viral vector use in neuroscience research has increased because they enhance interpretation of the anatomy and physiology of brain circuits compared with conventional tract tracing or electrical stimulation techniques. Viral vectors enable neuronal or glial subpopulations to be labeled or stimulated, which can be spatially restricted to a single target nucleus or pathway. Here we review the use of viral vectors to examine the structure and function of motor and limbic basal ganglia (BG) networks in normal and pathological states. We outline the use of viral vectors, particularly lentivirus and adeno-associated virus, in circuit tracing, optogenetic stimulation, and designer drug stimulation experiments. Key studies that have used viral vectors to trace and image pathways and connectivity at gross or ultrastructural levels are reviewed. We explain how optogenetic stimulation and designer drugs used to modulate a distinct pathway and neuronal subpopulation have enhanced our mechanistic understanding of BG function in health and pathophysiology in disease. Finally, we outline how viral vector technology may be applied to neurological and psychiatric conditions to offer new treatments with enhanced outcomes for patients.
- adeno-associated virus
- basal ganglia neuronal phenotypes
- confocal and electron microscopy
viral vectors have been used to increase the knowledge gained in basic neuroscience experiments and in preclinical and clinical studies to test potential treatments for neurological and psychiatric disorders. Vectors are often combined with cutting-edge technologies such as advanced tracing and microscopy techniques and optogenetic and chemogenetic stimulation and can be used for gene therapy strategies to alter the production of therapeutic or defective proteins. This review discusses how viral vectors have been used in basal ganglia (BG) studies and outlines how their use has advanced the anatomical and physiological knowledge gained. To aid readers from different fields, we provide a short summary of BG gross anatomy and the viral vectors used in neuroscience research. Within these sections we provide tables that list the neuronal phenotypes in each BG nucleus and the generic and specific promoters used in BG research as resources for readers. We then explain how viral vector use has enabled BG anatomy to be investigated at the gross level in vivo and also at pathway and ultrastructural levels in vitro. We clarify how specific BG subpathway functions have been elucidated by using optogenetic and chemogenetic stimulation combined with electrophysiology, electrochemistry, or behavior. We describe viral vectors used to create or treat animal models of BG disorders because they advance understanding of disease mechanisms or provide preclinical studies of potential therapies. Finally, we outline how viral vectors and new technologies may be used in the future for preclinical and clinical treatment of BG disorders.
Introduction to Basal Ganglia Structure and Function
BG are a group of interconnected subcortical nuclei in the forebrain and midbrain of birds, reptiles, and mammals (Deschêne et al. 1996; Wilson 1998) involved in cognitive, motor, associative, memory, and learning processes (Bolam et al. 2000), including reward-related learning, behavioral reinforcement, addiction, arousal, selection, and initiation of movement (Albin et al. 1989; Bolam et al. 2000). BG nuclei receive inputs from the cerebral cortex and the thalamus (Wilson 1998) and then process information through multiple parallel loops (Graybiel et al. 1994) before feeding it back to selected regions of the cerebral cortex via the thalamus (Alexander et al. 1986; Alexander and Crutcher 1990).
It is widely recognized that BG nuclei are highly conserved across animal species, indicating their key role in controlling behavior. In mammals, dorsal BG nuclei are important for controlling motor and associative functions and include the caudate nucleus, putamen, globus pallidus [external (GPe) and internal (GPi) segments], substantia nigra [pars compacta (SNc) and pars reticulata (SNr)], and the subthalamic nucleus (STN) (Tepper et al. 2007) (Fig. 1A). The caudate nucleus and putamen are collectively called the dorsal striatum, and in some lower mammals (e.g., rodents) these nuclei are indistinguishable (Wilson 1998). The caudate-putamen and GPi are known as the lentiform or lenticular nucleus in humans, which is the collective target site for treating Parkinson's disease (PD) and other movement disorders (Carpenter and Sutin 1983; Russmann et al. 2003). In cats and rodents, the GPi is known as the entopeduncular nucleus (Gerfen and Wilson 1996; Wilson 1998). Information primarily enters the dorsal BG at the striatum, which receives extensive inputs from the cortex and some information from the intralaminar and ventral motor thalamic nuclei (Bolam et al. 2000; Villalba et al. 2015; Young and Penney 2002), whereas the STN represents a secondary input nucleus (Villalba et al. 2015). The dorsal striatum also receives dopaminergic input from the SNc, which is important in selecting the motor programs to promote or suppress, a process called action selection.
Information is processed through motor-related BG in the hyperdirect, direct, and indirect pathways (Fig. 1A), with the names based on the number of connections and the latency of information transmission between the cortex and BG output nuclei (GPi and SNr) (Albin et al. 1989; Alexander and Crutcher 1990; Nambu et al. 2002). The hyperdirect pathway contains the fewest connections; cortical information enters the BG at the STN (Kita 1994), where it is projected to the GPi and SNr with the shortest latency (Bosch et al. 2011, 2012). For both the direct and indirect pathways, cortical information enters the BG at the dorsal striatum. Medium spiny neurons (MSNs) in the dorsal striatum were once thought to be morphologically and electrophysiologically similar. More recent studies show that MSNs have different membrane properties and that indirect pathway MSNs are more excitable (Galvan et al. 2012). MSNs contain neurochemicals that are distinct to each pathway (Tepper et al. 2007), and they preferentially express different types of dopamine receptors (Gerfen and Wilson 1996; Kravitz et al. 2010). Direct-pathway MSNs express D1-like receptors (Gerfen 2000) and project information to the output nuclei. In contrast, indirect-pathway MSNs express D2-like receptors (Albin et al. 1989; Gerfen 2000; Terman et al. 2001) and send information to the GPe, which then projects to the STN before information reaches the GPi and SNr. The indirect pathway has the longest latency for information to pass from the cortex to the BG output nuclei (Nambu et al. 2000). For all of these pathways, information is sent from the GPi and SNr to the ventral motor and intralaminar thalamic nuclei before it is returned to associative and motor cortical areas or to the brain stem and spinal cord (Bosch-Bouju et al. 2013; Fisher and Reynolds 2014).
The ventral division of the BG controls limbic functions and includes the nucleus accumbens (NAc) (or ventral striatum), ventral pallidum (VP), and ventral tegmental area (VTA) (Alexander et al. 1986; Bolam et al. 2000; Galvan et al. 2015) (Fig. 1B). Here, the NAc is the major input nucleus, receiving glutamatergic input from the prefrontal cortex (PFC), hippocampus, and amygdala and dopaminergic innervation from the VTA (Russo and Nestler 2013). It also receives serotonergic input from the dorsal raphe and noradrenergic input from the locus coeruleus (Lorrain et al. 1999; Unemoto et al. 1985; Young and Penney 2002). In addition, the VTA receives glutamatergic input from the lateral habenula (Balcita-Pedicino et al. 2011; Brinschwitz et al. 2010; Omelchenko et al. 2009) and dorsal raphe (Qi et al. 2014) and GABAergic inputs from the rostromedial tegmental nucleus (Jhou et al. 2009; Kaufling et al. 2010; Russo and Nestler 2013) and the NAc (Russo and Nestler 2013). The ventral BG output nuclei include the VP and rostromedial GPi, which project to thalamic nuclei [mediodorsal, rostral parafascicular nucleus, the magnocellular part of the ventral anterior (VA) nucleus, and the medial part of the ventral lateral (VL) nucleus; Galvan et al. 2015; Swanson et al. 1987] before information reaches the PFC and anterior cingulate cortex (Alexander and Crutcher 1990; Bolam et al. 2000; Galvan et al. 2015; Haber and Knutson 2009; Ligorio et al. 2009). The role of VTA dopamine release in the ventral BG is to provide context and motivational significance to behaviorally important signals, particularly in relation to rewarding, motivational, and cognitive functions, and it also has a role in drug addiction and psychiatric disorders (Bourdy and Barrot 2012; Wang and Tsien 2011).
Initially, knowledge about the functions of BG nuclei was based on observations and symptoms associated with neurodegenerative disorders (Wilson 1998). Many disorders are implicated consistent with its complex, highly interconnected circuitry (Wilson 1998), signifying that the BG have multiple functions (Gerfen and Wilson 1996). Generally, these degenerative disorders are characterized by abnormal movements, ranging from hypokinesia to hyperkinesia (Young and Penney 2002). Hypokinetic disorders are characterized by paucity of movement such as the symptoms observed with PD, where patients have difficulty initiating movement (akinesia), movements are slower (bradykinesia), and patients experience postural rigidity and resting tremor (Albin et al. 1989). The primary pathology of PD is loss of midbrain dopamine neurons. Although PD is considered a motor disorder, many patients experience cognitive decline and depression later in the disease process (Starkstein et al. 1989). The most noted example of a hyperkinetic disorder is Huntington's disease (HD), a progressive genetic neurological disorder caused by an autosomal dominant mutation resulting from expansion of CAG repeats in exon 1 of the huntingtin gene. The symptoms most commonly occur in adults during and after child-bearing age (30–50 yr) and are characterized by excessive uncontrollable movements known as chorea (from its dancelike appearance) (Young and Penney 2002). The exact sequence of HD pathogenesis is not fully understood but involves dysfunction and degeneration within the corticostriatal circuit (Ross and Tabrizi 2011). A decrease in striatal volume occurs before symptom-based diagnosis of HD and is associated with subtle behavioral changes (Ross et al. 2014). Cortical dysfunction occurs early and is associated with cognitive, psychiatric, and metabolic decline. The earliest motor symptom is chorea, which occurs because of striatal dysfunction, while other motor impairments occur later in the disease and are thought to be due to neuronal degeneration (Estrada-Sánchez and Rebec 2013). HD causes selective degeneration of striatal MSNs due to nuclear inclusions and cytosolic aggregation of the huntingtin protein (Ramaswamy et al. 2007; Southwell and Patterson 2011). Nonmovement disorders involve the limbic BG nuclei and include attention deficit hyperactivity disorder (ADHD) and obsessive-compulsive disorder (OCD) (Delong and Wichmann 2007). Psychiatric disorders arising from BG dysfunction include Tourette's syndrome, schizophrenia, depression, OCD, and drug addiction (Ring and Serra-Mestress 2002).
Dorsal and ventral BG nuclei contain GABAergic, glutamatergic, cholinergic, and dopaminergic neurons, with GABAergic neurons being the dominant phenotype in the majority of nuclei (Fig. 1). In addition, there are many subtypes of GABAergic neurons in the BG, each with different electrophysiological characteristics and calcium binding proteins, which will also affect how these neurons respond to inputs (Bolam et al. 2000; Ellens and Leventhal 2013). Furthermore, some of these GABAergic neurons corelease a peptide, for example, somatostatin, which affects responses at postsynaptic neurons (Gittis et al. 2014; Lévesque and Parent 2005; Smith and Bolam 1990; Tepper and Bolam 2004; Yelnik et al. 1984). This review does not focus on these differences; however, for clarity we present the neuronal phenotypes and subtypes in each BG nucleus (motor and limbic) in Table 1.
Much of our understanding of BG circuitry comes from historical tracing technology. The advent of viral vectors has rapidly advanced our knowledge of the anatomical and functional intricacies of these pathways, including the role of individual neuronal phenotypes within these nuclei. Next, we outline the viral vectors used in the BG and explain how their use has added to our knowledge of BG circuitry.
Viral Vector Use in Neuroscience
Since the concept of gene therapy was first described in the 1970s (Friedmann and Roblin 1972), their use and development in the brain for both network discovery and therapy has exploded. Viral vectors utilize the ability of a wild-type virus to infect target cells, hijacking the cellular machinery to express their genome. However, in most cases viral vectors differ in that they are rendered replication incompetent; they cannot replicate their genome or produce a productive infection. Vectors are a delivery vehicle for expression of reporter and/or functional proteins under the control of constitutive, inducible, or cell-type promoters. Here we describe the general characteristics of the major viral vectors used in neuroscience and their advantages and disadvantages. Each virus has characteristic tropism (targeting of cells) and spread from injection sites, in some cases via retrograde or anterograde transport of viral particles, which are important to consider when designing experiments.
Lentiviruses, a group including the immunodeficiency viruses (human, HIV; feline, FIV; equine infectious anemia, EIAV), are part of the Retroviridae family of RNA viruses. Lentiviruses, in contrast to many members of the family, are capable of transducing nondividing cells and hence are more commonly used in the adult brain. Wild-type HIV, the basis of most lentiviral vectors, is modified to improve tropism and safety. The HIV1 genome is deleted of all viral encoded genes (gag, pol, and env), with only the viral long-terminal repeats (LTR) and packaging signal remaining. Third-generation vectors also have a deletion in the 3′-untranslated region (UTR) preventing transcription from the integrated viral promoter and a recombinant 5′-UTR, which is not reliant on the viral tat protein (Dull et al. 1998). The major contributing factor to tropism and transduction of lentiviral vectors is the envelope used. Traditionally, the wild-type envelope from HIV is replaced by the glycoprotein from the vesicular stomatitis virus (VSVg). This envelope mediates wide tropism to the majority of cell types via the low-density lipoprotein receptor (Finkelshtein et al. 2013). The overall size of the vector particles and high affinity for receptors limit spread of lentivirusVSVg vectors to a maximum of <1 mm in most brain regions (Linterman et al. 2011). LentivirusVSVg has no retrograde or anterograde transport properties (Fig. 2). However, using glycoproteins from other viruses or VSVg chimerics with rabies glycoproteins can allow strong retrograde uptake (Kato et al. 2011, 2014; Schoderboeck et al. 2015). Additional envelopes have been used to select specific neural types, e.g., the lymphocytic choriomeningitis virus (LCMV) glycoprotein results in preferential transduction of astrocytes in the rat substantia nigra (Cannon et al. 2011), or rabies to confer retrograde transport (Kato et al. 2011, 2014; Schoderboeck et al. 2015). The packaging capacity of lentiviruses is ∼10 kb, allowing relatively large and complex constructs to be introduced. Unless specifically packaged with an integrase-deficient system (Wanisch and Yanez-Munoz 2009), lentiviral vectors will integrate into the host genome, but there are no reports of integration-related pathologies.
Adeno-associated viruses (AAVs) are members of the Parvoviridae family of single-stranded DNA nonenveloped viruses. AAV is a dependovirus, meaning that alone it cannot generate a productive infection; it requires coinfection with adenovirus or herpes simplex virus (HSV). The AAV genome is packaged in a proteinaceous capsid that differs in amino acid sequence and tropism depending on serotype. AAVs are widespread in mammals; many serotypes have been isolated from humans and monkeys and developed as viral vectors (reviewed in Castle et al. 2016). Historically, the AAV2 serotype has been used in viral vectors and gene therapy trials. In the brain, it shows minimal spread (though further than lentiviruses) and limited anterograde transduction, moving out from local soma down axons to transduce synaptically connected soma, without retrograde transport (Fig. 2) (Murlidharan et al. 2014; Salegio et al. 2013). AAV5 is a retrograde tracer in some situations (Aschauer et al. 2013), although it is mainly restricted to the site of injection within the BG with anterograde spread of the transgene-encoded protein (Salegio et al. 2013). AAV8 and 9 serotypes will trace neurons in both directions (Aschauer et al. 2013; Castle et al. 2014; Ciesielska et al. 2011; Hutson et al. 2012; Salegio et al. 2013). AAV9 and AAVrh10 (rhesus monkey-derived serotype 10) can also be used to deliver genetic material to the brain via systemic administration (Schuster et al. 2014).
Packaging of AAV1, 4, 5, and 6, and more recently 8, 9, and rh10, generally incorporate the AAV2-based genome into capsids from the other serotypes. In many studies these are denoted as AAV2/1, AAV2/rh10, etc. to indicate their chimeric nature. The change to other serotypes has had a significant impact on both tropism and targeting of neurons via anterograde and retrograde transport (reviewed in Castle et al. 2016). Note that transduction patterns can also vary between brain structures targeted and promoters, even when constitutive promoters are used to drive transgene expression. AAV genetic constructs are limited to 4.7 kb by the smaller capacity for effective packaging. This limits applications for some larger genes and more complex multipart regulators. Trans-splicing and minigenes have been used to overcome this (reviewed in Hirsch et al. 2016). Notably, recombinant AAV, unlike their wild-type counterparts, do not routinely integrate into the host genome, instead remaining as an episome (extragenomic circular DNA). In nondividing cells this can still result in long-term transgene expression (at least 6 yr has been demonstrated in monkey BG; Rivera et al. 2005).
Adenovirus was the first to be used as a viral vector in the early 1990s. With a large capacity (at least 10 kb, but up to 36 kb if completely gutted), adenovirus has been widely used in proof-of-principle gene therapy studies (Bemelmans et al. 1999; Ikari et al. 1995) and for tracing (Kuo et al. 1995; Papp et al. 2012; Tomioka and Rockland 2007). The recombinant adenoviral genome can also be packaged into one of several capsids, the most common based on adenovirus 5. The serotype (group A-F) determines transduction efficiency (Chillon et al. 1999). Tropism of adenovirus is mediated by the fiber knob, which binds coxsackie and adenovirus receptor (CAR) and integrin αVβ5 (Arnberg 2009). Adenovirus today is most commonly used as a retrograde neuronal tracer (Kuo et al. 1995; Papp et al. 2012; Tomioka and Rockland 2007). The genome remains episomal in both wild-type virus and recombinant vectors, and expression is limited by a high immunogenicity against the virus (Lowenstein et al. 2007).
Transsynaptic viruses include the alpha-herpes viruses, e.g., pseudorabies and HSV, and the Rhabdoviridae family member, rabies, that have been used extensively as retrograde tracers. Each have double-stranded DNA genomes with large packaging capacity as viral vectors, up to 150 kb. These viruses can enter the CNS from the periphery, for example, via a muscle infection, and “jump” synapses after replication (sometimes referred to as self-replicating tracers). While these vectors have been invaluable for identifying microcircuit structure and connectivity in the brain, their replication eventually kills infected neurons, preventing their use for longer-term functional studies. Newer-generation vectors have been developed to limit transsynaptic spread to a single synapse, eliminating the toxicity. These self-limiting vectors have also been used for traditional gene therapy studies (HSV-1, Goins et al. 2014; Neve 2012) and monosynaptic tracing (rabies and pseudorabies; Arenkiel and Ehlers 2009).
Neuronal specificity of viral vectors.
In neuroscience research, viral vectors were initially used to dissect the anatomy of brain circuits. However, viral vector use has continued to increase substantially in the last decade with the development of tools to examine the function of circuits with light (optogenetic) and designer drug (chemogenetic) stimulation. Modifying the innate tropism of viruses has enabled viral vectors to effectively transduce neurons and glia. A further advancement has been the addition of a promoter sequence to the transgene to target a specific population or subpopulation of neurons or glia in the brain. The promoter transgene encodes a protein that is uniquely expressed in the target cell population [e.g., glutamate decarboxylase (GAD) for GABAergic neurons] and limits transduction of nonnative proteins, such as expression of a reporter fluorophore, to the target population. Here we do not focus on the function of each promoter, but a list of promoters that have been used in BG research is provided in Table 2. Comparison of neuronal phenotypes in BG nuclei in Table 1 and available promoters that have been used in the BG in Table 2 may be useful for designing future viral vector BG experiments.
Tracing and Imaging
Historically, traditional neuronal tracers have been used to investigate brain circuits such as the cortico-BG-thalamocortical pathway (Lanciego et al. 2012; Seeger-Armbruster et al. 2015). These neuronal tracers include biotinylated dextran amine, Phaseolus vulgaris leucoagglutinin, wheat germ agglutinin, horseradish peroxidase, and cholera toxin subunit B. They will trace anterograde, retrograde, or bidirectionally depending on their uptake mechanism and molecular weight (Callaway 2008; Huh et al. 2010; Parr-Brownlie et al. 2015). Like neuronal tracers, some viral vectors can travel retrogradely or anterogradely along the axon (see Fig. 2), and vectors can be pseudotyped with different glycoprotein envelopes (lentiviruses) or chimeras created (lentiviruses and AAVs) to control the direction of transport (Castle et al. 2016; Mazarakis et al. 2001). Viral vector technologies are starting to replace neuronal tracers for mapping neuronal circuits because they have the advantages of selectively labeling specific neuronal phenotypes and pathways (Betley and Sternson 2011; Oztas 2003). Using a specific promoter sequence that controls gene expression allows targeting of proteins for a specific neuronal population, phenotype, or component in the cell membrane, nucleus, or cytoplasm in a specific brain region, pathway, or multisynaptic circuit. Consistent with traditional tracers, the number of neurons transduced and the spread of the virus can also be controlled by the volume injected and by the vector used (Callaway 2008; Huh et al. 2010; Oztas 2003). In addition, pseudorabies virus enables transsynaptic tracing across sequential synapses within a circuit, and the number of synapses crossed is determined by the time the virus is left to transduce neurons (Callaway 2008). It is also possible to combine traditional tracers with viral vectors so that a specific transsynaptic pathway can be labeled. This has been achieved by inserting cDNA for a transsynaptic protein, like wheat germ agglutinin, as a transgene into viruses (Huh et al. 2010). Currently, combining viral vectors with imaging techniques permits anatomical analysis of circuits and functional manipulation; therefore, consequential changes in anatomy that arise from changes in physiology or pathology can be investigated in the same tissue. The next section reviews both gross and ultrastructural imaging techniques that have been combined with viral vector technology to investigate BG circuits.
Gross and ultrastructural imaging of circuits.
Combining viral vector technology with imaging techniques enhances knowledge gained from neuroanatomical studies. Traditionally, cell phenotypes were determined by immunohistochemically staining for proteins that are selectively expressed within the cells of interest. While phenotype was established, the pathway (inputs or targets) remained unknown. For example, using GAD immunostaining to label GABAergic neurons in the GPe of rats would show which somata and axons are GABAergic, but we would not know whether the axons were from the striatum or from local neurons. Additional neural circuitry information is gained by examining the target innervations of, or inputs onto, transduced neurons. If GABAergic inputs from the striatum to the GPe were to be investigated, injection of a viral vector under the control of the GAD67 promoter will enable transduced axon terminals from the striatum to be identified (Fig. 2A). To do this, the vector contains the genetic code to express a reporter fluorophore (e.g., green fluorescent protein, yellow fluorescent protein, or mCherry) in virally transduced neurons, which is then visualized with fluorescence or confocal microscopy. Neuron phenotype-specific tracing has been achieved on small (Rah et al. 2015) and large (Beyer et al. 2013; Kasthuri et al. 2015; Lakadamyali et al. 2012) connectome scales by combining brainbow mice and superresolution imaging of neural cell cultures or large blocks of brain, respectively.
When viral vectors are used, transduction specificity is verified by immunohistochemically staining for a neuronal or glial phenotype-specific protein and also showing that markers for other types of neurons and glia are not coexpressed. Nonspecific promoters used in the brain, including the BG, for neuronal tracing studies include the human cytomegalovirus immediate-early enhancer promoter (CMV), CMV coupled with chicken β-actin promoter and first intron and rabbit β-globin splice acceptor (CAG), phosphoglycerate kinase I (PGK), and mammalian elongation factor 1α (EF1α) promoters (see Table 2). Phenotype-specific promoters include GAD67, parvalbumin (PV), Ca2+/calmodulin-dependent kinase IIα (CaMKIIα), choline acetyltransferase (ChAT), tyrosine hydroxylase (TH), and glial fibrillary acidic protein (GFAP) to label GABAergic neurons, GABAergic neurons with the PV calcium binding protein, glutamatergic, cholinergic, and dopaminergic neurons and astrocytes, respectively (Huh et al. 2010; Oh et al. 2009).
Neural circuits can be investigated at gross and ultrastructural levels depending on how the tissue is processed. At the gross level, improved microscope optics and refined tissue clearing techniques (Chung et al. 2013; Deisseroth and Schnitzer 2013; Hama et al. 2011; Ke et al. 2013; Richardson and Lichtman 2015; Tomer et al. 2014) enable visualization of circuits in large blocks of tissue; depth of tissue is only limited by the working distance of the microscope's objective lens. This offers the advantage that one fluorescently labeled population of cells can be imaged along its entire length and the branch points and targets of collaterals can be determined. This can be undertaken in intact brains, with little damage done to the tissue because sectioning is not required (Chung et al. 2013), compared with mapping pathways by examining 5- to 50-μm sections for expression of the fluorophores at injection and target sites (Kuo et al. 1995). To increase the knowledge gained about neural circuits, it is becoming more common to combine viral vector neural tracing with other methods to visualize subcellular components (i.e., synapses) or examine the location of proteins within transduced neurons to either confirm the specificity of the vector or delineate a circuit (Galvan et al. 2012; Konermann et al. 2013; Kuramoto et al. 2009; Lobbestael et al. 2010; Pastrana 2010; Shu et al. 2011).
Fluorophores expressed by virally transduced neurons can be immunolabeled with anti-fluorophore antibodies and in turn labeled with an electron-dense chromogen such as diaminobenzidine (DAB) or immunogold particles, which can be visualized at the ultrastructural level by electron microscopy (Dautan et al. 2014; Galvan et al. 2012; Scotto-Lomassese et al. 2011; Sosinsky et al. 2007). Here, technical advances in electron microscopy and imaging in the last decade enable greater visualization of cellular components. Combining fluorescent viral vector technology with electron microscopy reconstruction techniques such as serial section transmission electron microscopy, serial block face scanning electron microscopy, serial section electron tomography, array tomography, and correlative light and electron microscopy allows ultrastructure visualization in intact blocks of tissues (Arenkiel and Ehlers 2009). However, to date, not all of these electron microscopy techniques have been combined with viral vector technology. Viral vector-induced labeling of mini-singlet oxygen generator (mini-SOG) allows polymerization of DAB after exposure of the transduced neurons to blue light, thus producing cell phenotype- and pathway-specific DAB labeling (Pollock et al. 2014; Shu et al. 2011). Furthermore, by combining light activation of neuronal activity (optogenetic stimulation, see Optogenetics), high-pressure freezing, and electron microscopy, functional changes to circuitry can be explored at the subcellular level (Konermann et al. 2013; Pastrana 2010; Watanabe et al. 2014).
Noninvasive gross anatomical circuitry can be investigated in situ with bioluminescence or functional magnetic resonance imaging (fMRI) by transducing specific cell populations with viral vectors. For bioluminescence imaging, a reporter gene encoding an enzyme that generates light is detected by a biosensor (Massoud et al. 2008; Shah et al. 2008). This low-resolution imaging method shows gross structural circuitry and has been used to investigate differences in spread and transduction efficiency of AAV, adenovirus, and lentiviral vectors (Cho et al. 2005; Contag and Bachmann 2002; Deroose et al. 2006; Massoud et al. 2008). Furthermore, optogenetic stimulation and gross anatomical imaging techniques have been combined to look at functional changes in circuits. For example, fMRI has been used to image blood oxygen level changes when neurons are stimulated (Adriani et al. 2010; Desai et al. 2011; Lee et al. 2010; Schmid et al. 2016) and positron emission tomography (PET) has been combined with viral vector technology to investigate changes in brain glucose metabolism (Thanos et al. 2013).
To date, not all viral vectors have been combined with all in situ imaging techniques, but there are no limitations preventing this. In the future, enhanced in situ imaging resolution by fMRI or bioluminescence would greatly improve circuitry knowledge in health and disease. Viral vectors could be combined with tissue clearing techniques, gene therapy, optogenetic stimulation, tissue processing, and imaging (Galvan et al. 2012; Watanabe et al. 2014; Weber 2012) to address unanswered questions relating to brain circuitry, particularly in the BG.
Investigating BG circuitry.
The use of viral vectors in basic research has revealed that BG circuitry is more complex than previously thought based on traditional neuronal tracers (Akkal et al. 2007; Dautan et al. 2014; Dum and Strick 2013; Fujiyama et al. 2011; Hoover and Strick 1999; Hoshi et al. 2005; Kelly and Strick 2004; Kuramoto et al. 2009; Middleton and Strick 2002; Oztas 2003). Vectors that have been commonly used to investigate afferent and efferent BG connections are HSV, rabies, and sindbis virus, which transsynaptically label pathways. Rabies virus has shown the existence of a closed-loop circuit between the BG and cerebral cortex as well as a bidirectional circuit between the BG and cerebellum (Dum and Strick 2013; Hoshi et al. 2005). Using rabies and HSV, Akkal et al. (2007) revealed that there is a disynaptic connection from GPi to the supplementary motor cortex (SMA), with rostral GPi conveying associative information to pre-SMA whereas sensorimotor information goes from caudal GPi to SMA proper. GPi, SNr, STN, GPe, and striatum bi- or trisynaptically innervate the primary motor cortex and PFC, but information is conveyed by distinct parallel pathways (Hoover and Strick 1999; Kelly and Strick 2004). GPi and SNr innervate specific and different parts of PFC, and these connections are segregated from GPi and SNr neurons that innervate the motor cortex (Middleton and Strick 2002). Sindbis virus has been used in single-cell tracing studies to confirm that thalamocortical neurons from the rostromedial region of VA-VL also innervate the striatum and striatofugal neurons arising in striatal striosomes project to the SNc and send collaterals to GPi, entopeduncular nucleus, and SNr (Fujiyama et al. 2011; Kuramoto et al. 2009).
More recently, AAV and lentiviral vectors have revealed the complexity of inputs to, and microcircuits within, the BG. Acetylcholine is released into the striatum and NAc from local cholinergic interneurons and also from projection neurons with somata in brain stem nuclei. The Cre-lox system is an example of site-specific recombination that uses viral vector technology to deliver either the Cre-recombinase enzyme or a pair of short target sequences, known as the lox sequence, to direct site-specific deletion or insertion of genes or carry out gene translocations or inversions (Sauer and Henderson 1988). Combining ChAT-Cre rats with AAV2 encoding Cre-dependent ChR2 constructs, Dautan et al. (2014) found that cholinergic neurons innervate the striatum and NAc directly but also indirectly through thalamic nuclei, SNc, and VTA. Furthermore, the topography differs: cholinergic neurons from the pedunculopontine nucleus innervate the dorsolateral striatum, whereas neurons in the laterodorsal tegmental nucleus innervate medial striatum and NAc and also send collaterals to the thalamus, and SNc, and VTA in the midbrain (Dautan et al. 2014). Circuit tracing was taken a step further by preparing the tissue for electron microscopy to verify that these cholinergic innervations formed asymmetric (excitatory) synapses (Dautan et al. 2014). The use of ChAT-Cre mice and a modified rabies virus revealed that cholinergic interneurons and MSNs in dorsolateral striatum receive similar extrastriatal inputs; however, cholinergic interneurons receive more extensive intrastriatal cholinergic inputs (Guo et al. 2015).
Future anatomical viral vector studies will likely continue to show greater complexity within the BG and between the BG and other areas of the brain. In addition, the use of viral vectors will enable greater understanding of how changes in structure affect function (and vice versa) because spatially and temporally specific physiological and anatomical techniques can be conducted in the same animals.
The term optogenetics describes technology in which genes encoding light-sensitive proteins are expressed in a target cell population to modulate excitability using light stimulation of specific wavelengths (see, e.g., Yizhar et al. 2011). Optogenetics provides scientists with tools to control neuronal activity and glial function in a temporal, spatial, and phenotype-specific manner relevant for physiologically intact tissue and behavioral tests (Fenno et al. 2011), which could not be achieved previously with other neuroscience methods such as deep brain stimulation (DBS) or drug administration. Since the first report of optogenetics (Boyden et al. 2005), application of this technique has significantly increased the understanding of physiological networks and pathways in mammalian brains.
Optogenetics is frequently used in combination with neuronal cell recordings or electrochemistry ex vivo (e.g., Cepeda et al. 2013; Tsai et al. 2009; Zhang et al. 2015) or in vivo (e.g., Bass et al. 2010; Galvan et al. 2012; Gradinaru et al. 2009). One of the aims of neuroscience is to understand the mechanisms and circuits involved in generating and controlling behavior, and optogenetics increases knowledge about distinct behaviorally relevant circuits in healthy and pathological states (Aquili et al. 2014; Chaudhury et al. 2013; Kravitz et al. 2010; Seeger-Armbruster et al. 2015). A key advantage of optogenetic stimulation is that a specific neuronal subpopulation and/or pathway can be selectively targeted by using viral vectors with neuron phenotype-specific promoters (see Table 2) and/or transgenic animals and appropriate placement of light probes. Below we discuss the impact optogenetics has had on the understanding of BG function.
Viral vectors represent the most popular means of introducing microbial opsin (light-sensitive protein) genes into neurons. The first article about optogenetics in neuroscience (Boyden et al. 2005) and some recent studies used lentiviruses as the viral vector of choice (see Parr-Brownlie et al. 2015). However, the majority of optogenetic studies use AAVs and occasionally HSVs (see Yizhar et al. 2011; Zhang et al. 2010). Most optogenetic studies in the BG have introduced genes for transmembrane channel or carrier proteins. The two most widely used opsins are channelrhodopsin-2 (ChR2) and halorhodopsin (NpHR), or mutated variants of these, which were cloned from Chlamydomonas reinhardtii (green algae) and Natronomonas pharaonis (halobacteria), respectively. Light of an appropriate wavelength causes a conformational change in the opsin, which opens ChR2 or NpHR. For ChR2, blue light (∼470 nm) opens a cation channel allowing Na+ to enter the cell, thus depolarizing it (Nagel et al. 2003). For NpHR, yellow light (∼590 nm) activates a pump that transports Cl− into cells and hyperpolarizes transduced neurons (Zhang et al. 2007). Today, many other variants and new opsins exist, such as C1V1 [a channelrhodopsin-1 (ChR1)-red-shifted channelrhodopsin-1 (VChR1) chimera] and archaerhodopsins (Arch and ArchT), with different activation wavelengths and times (Fig. 3A). For comparisons of activation wavelengths and kinetic properties of different opsins see Yizhar et al. (2011), Mattis et al. (2012), Tye and Deisseroth (2012), and Klapoetke et al. (2014). Therefore, using opsins with distinguishable peak activation profiles enables neuronal populations in one area of the brain to be selectively excited and/or inhibited (Han and Boyden 2007). The continuous development of new opsin variants with enhanced or altered properties will increase the use of optogenetics in the future.
Optogenetic stimulation has moved beyond activating or inhibiting neuronal activity so that it can be used to control intracellular signaling, DNA binding, and epigenetics (see Chow and Boyden 2013). More recently, with rhodopsin-G protein-coupled receptor chimeras, Opto-XRs, green light (500 nm) stimulation modulates cellular signaling by affecting intracellular G protein-coupled pathways (Fenno et al. 2011; Yizhar et al. 2011; Zhang et al. 2010), and this was first described by Airan et al. (2009) in the BG of mice. The advantage of Opto-XRs is that they permit cell signaling pathway-derived effects to be separated from changes in receptor activation and the neuron's membrane potential, and genetic adaptations would increase their use in more brain regions. Other opsins [light oxygen voltage (LOV) domains, cryptochromes, and phytochromes] are fused to effector proteins (e.g., Rac) and located intracellularly (Pastrana 2013; Zhang and Cui 2015). Here, light of appropriate wavelength activates the effector protein, altering its function and those of downstream mediators, resulting in polymerization of actin, protein dimerization, altering DNA binding proteins, or other changes in cellular function, which can be detected by including a reporter fluorophore in the genetic construct or measured as changes in cell activity, accumulation of a downstream protein, or changes in behavior. Optogenetics can also be used to alter DNA binding and mRNA expression by using light-inducible transcription effectors (LITEs) that interact with transcription activator-like effector (TALE) DNA binding domains (Kennedy et al. 2010; Konermann et al. 2013; Schindler et al. 2015). To date, such in vivo experiments have examined effects on gene expression but have not been combined with behavior and other physiological measures (cell recordings, electrochemistry, or microdialysis), but this is critical to understand the consequences of altering gene expression in normal and diseased brains. None of these constructs has been used within BG nuclei.
BG function in health.
The BG can be divided into parallel networks (see Fig. 1), and we focus on motor and limbic pathways. Optogenetic stimulation is frequently applied in rodents to modulate neuronal activity in brain structures, including the BG, and alter behavior (e.g., Gradinaru et al. 2009; Seeger-Armbruster et al. 2015; Tsai et al. 2009), but most studies in nonhuman primates have been limited to the cerebral cortex (e.g., Dai et al. 2015; Han et al. 2011), with only one targeting subcortical nuclei (Galvan et al. 2012). In this section we summarize how viral vectors and optogenetic experiments have significantly increased the understanding of the BG networks by dissecting the function of subpathways and subpopulations within or between BG nuclei or between BG nuclei and other structures that project to or receive information from the BG in health.
Like electrical stimulation, optogenetic stimulation-induced neuronal responses are often difficult to interpret and understand because of their complexity. However, compared with electrical stimulation, selective activation of one cell phenotype and verification of the maximum brain area stimulated (determined by examining the location of transduced cells postmortem) aid interpretation of neuronal responses. Striatal, GPe, and VL thalamus responses following light stimulation of dendrites, soma, axons, or terminals of virally transduced GPe or dorsal striatal neurons varied depending on postmortem ChR2 expression (Galvan et al. 2012). More neurons responded in the striatum and VL (∼32% of neurons responded) than GPe (3% responses). Responses included direct and indirect ChR2-induced effects, with indirect effects most likely mediated through local inhibitory network activation (Galvan et al. 2012). Together, these data highlighted a difference in the type and magnitude of responses depending on the BG nucleus. This study shows that while optogenetic stimulation is a powerful tool that enables anatomy and function to be dissected, responses are complex and the viral construct, injection parameters, and light stimulation conditions need to be optimized for each structure.
Optogenetic experiments have shown that the medial prefronto-striosomal circuit selectively affects decision-making in a cost-benefit task under approach-avoidance conflict conditions known to evoke anxiety in humans. The function of inputs from the PFC has been thought to be important in addiction behavior and associated plasticity, mostly through NAc afferents (see Yager et al. 2015). In contrast to this, a recent study showed the involvement of prefrontal afferents on striosomes in the dorsal striatum in decision-making (Friedman et al. 2015). Striosomes are too small to be identified in human fMRI, but abnormalities in postmortem tissue are reported in patients with mood changes and motor deficits (Crittenden and Graybiel 2011). Modulation of prefrontal-prelimbic cortex (PFC-PL) afferents innervating dorsal striatum striosomes with enhanced NpHR (eNpHR) or C1V1 (excited by 532-nm light) (Friedman et al. 2015) underlies balanced decision-making disturbances. In particular, eNpHR-mediated intrastriatal inactivation of PFC-PL afferents increased the choice for the high-cost option, while C1V1-mediated excitation changed the behavior toward the low-cost and low-benefit option.
The existence of rapid cerebello-striatal communication, independent of slower cortical loops, has been proven with optogenetics. Dorsolateral striatum activity is modulated by cerebellar projection neurons via a disynaptic pathway through the centrolateral nucleus (CL) of the thalamus (Chen et al. 2014). Cerebellum-induced responses in the striatum after electrical or optogenetic stimulation (ChR2) of neurons in the cerebellar dentate nucleus were prevented by inactivation of CL by using local injection of tetrodotoxin to block voltage-gated sodium channels or by optogenetically silencing CL neurons expressing ArchT, a light-sensitive (566 nm) extrusion proton transporter (Chen et al. 2014). This cerebello-striatal pathway may play an important role coordinating motor outputs in real time.
The functions of SNc- or VTA-derived dopaminergic afferents in the striatum have been separated and characterized with optogenetics. By combining stimulation of ChR2 expressed in spatially restricted SNc neurons with fast-scan cyclic voltammetry (FSCV) in the striatum, optical stimulation restricted dopamine release to the dorsal region (Bass et al. 2010). This technical advancement is of special interest for understanding the neural microcircuitry underlying dopamine transmission since it is nearly impossible to electrically stimulate SNc alone without also recruiting VTA neurons in adjacent brain regions or fibers of passage (Bass et al. 2010). Furthermore, Cre-transgenic mouse lines (DAT-Cre and VGAT-Cre) combined with viral vector-based optogenetics (ChR2 activation) enabled cell type-specific identification of VTA neurons in a reward paradigm and verified that dopamine neurons and their GABAergic neighbors in the VTA have distinct physiological responses to reward (Cohen et al. 2012), which was previously hypothesized but could not be definitively shown without these techniques.
Direct optogenetic manipulation of MSN subpopulations has been used to confirm their roles underlying behavior. Albin et al. (1989) hypothesized that subpopulations of dorsal striatal MSNs express D1- or D2-like dopamine receptors and are part of the promovement, direct or antimovement, indirect pathway (Fig. 1), respectively, but this could not be explicitly tested without using viral vectors and optogenetics. Selective optogenetic stimulation of either D1- or D2-like-expressing MSN populations confirmed that the pathways differentially influence BG output nuclei and movements (Freeze et al. 2013; Kravitz et al. 2010, 2013). Specifically, the use of D1- and D2-Cre-transgenic mice in combination with stimulation of virally introduced ChR2 revealed that the D2-like-expressing neurons excited SNr neurons and suppressed movements to the point of inducing parkinsonism, whereas stimulation of D1-like neurons inhibited SNr activity and promoted movements in behaving mice (Freeze et al. 2013; Kravitz et al. 2010). In the limbic BG pathway, behavior flexibility can be increased by inhibiting NAc shell MSNs during reward or error feedback intervals (Aquili et al. 2014). The precise timing of the bilateral optogenetic inhibition (via eNpHR activation) revealed critical time periods when NAc shell neurons integrate reward or error feedback history and use this integrated history to make subsequent decisions (Aquili et al. 2014).
The study of corelease of neurotransmitters has been revolutionized by combining cell type-specific viral constructs, Cre-transgenic animals, and optogenetics. The role of corelease of glutamate from some midbrain dopamine neurons has been resolved by combining optogenetic stimulation of ChR2-expressing midbrain dopamine neurons with FSCV or electrophysiological recordings (Stuber et al. 2010; Zhang et al. 2015). In particular, blue light stimulation of transduced dopamine neuron terminals in the NAc shell or dorsal striatum resulted in dopamine release in both structures but only caused glutamate-mediated excitatory postsynaptic currents (EPSCs) in NAc MSNs (Stuber et al. 2010; Zhang et al. 2015). The additional use of vesicular glutamate transporter 2 (vGlut2) conditional knockout or Cre-transgenic mice showed that this glutamate release from dopamine neurons is vGlut2 mediated (Stuber et al. 2010) and that dopamine and glutamate can be released from the same mesoaccumbens axon, but not at the same site or from the same synaptic vesicle (Zhang et al. 2015). Further to glutamate, midbrain dopamine neurons also corelease GABA from their axons (Tritsch et al. 2012, 2014). Activation of ChR2-expressing nigrostriatal or mesolimbic dopaminergic terminals in the dorsal or ventral striatum showed that GABA corelease rapidly inhibited MSN activity, which relied on GABA uptake across the plasma membrane rather than de novo synthesis and the activity of vesicular monoamine transporter 2 (not vesicular GABA transporter) (Tritsch et al. 2012, 2014).
Optogenetic stimulation experiments will continue to increase our knowledge of BG connections with other brain regions as well as interactions of neuronal subpopulations within healthy BG because specific cell bodies or their projections can be targeted. Given the extensive BG interconnections, optogenetics allows the function of one projection over another to be differentiated. Extended knowledge of BG during health is necessary to understand changes in pathophysiological conditions.
Changes in BG activity underlying neurological and psychiatric conditions have been investigated by optogenetic stimulation in animal models of disease. These studies have provided new insights into the mechanism of action of electrical DBS currently used in patients or have identified potential new treatment strategies.
Animal models for PD have investigated the underlying mechanisms or new options for DBS. Kravitz et al. (2010) provided the first evidence that direct-pathway MSN stimulation can ameliorate motor deficits in parkinsonian mice. In particular, bilateral activation of ChR2-expressing D1-like MSNs (direct pathway) completely rescued deficits in freezing, bradykinesia, and deficits initiating locomotion in bilaterally 6-OHDA-lesioned mice, while stimulation of D2-like MSNs (indirect pathway) induced a parkinsonian state in normal mice (Kravitz et al. 2010). Optogenetic experiments showed that inactivation of STN neurons, the hypothesized target effect of electrical STN DBS in PD patients, may not be the mechanism responsible for improving motor control (Gradinaru et al. 2009). Neither direct inhibition (via eNpHR stimulation) nor activation (via ChR2 stimulation) of STN neurons improved motor symptoms in unilateral 6-OHDA-lesioned rats. On the other hand, selective stimulation of afferent axons originating in the motor cortex that project to the STN showed a therapeutic behavioral effect (Gradinaru et al. 2009). These studies encourage review of current DBS mechanisms and strategies and provide additional target sites for stimulation therapies. For an overview on optogenetic insights for PD see Vazey and Aston-Jones (2013) and Rossi et al. (2015).
The short-latency cerebellar input to dorsolateral striatum (Chen et al. 2014) rapidly modulates symptom onset in dystonia-Parkinsonism. Bilateral optogenetic silencing of CL thalamic neurons, functionally severing the link between the cerebellum and the striatum, abated cerebellum-induced dystonia shortly after light activation was started and dystonia returned soon after stimulation was terminated. Thus, under pathological conditions, the cerebello-striatal pathway may transfer abnormal cerebellar activity to the BG, contributing to movement disorders such as dystonia (Chen et al. 2014).
Optogenetic experiments have also increased knowledge of changes in striatal MSNs (Cepeda et al. 2013) and cholinergic function (Holley et al. 2015) in the R6/2 transgenic mouse model of HD with a severe, rapidly progressing phenotype. The cholinergic deficit in HD was previously unresolved, since large cholinergic interneurons in the striatum are spared in the disease. In slice preparations, activation of ChR2-expressing PV interneurons strongly inhibited MSNs in HD model mice but not in control mice (Cepeda et al. 2013), indicating that PV interneurons are a possible source of increased GABA synaptic activity on MSNs. Furthermore, optogenetic stimulation of somatostatin-expressing interneurons inhibited striatal cholinergic interneurons in the R6/2 mouse model (Holley et al. 2015); thus the cholinergic deficit arises from a change in activity and not degeneration of these neurons.
The BG, particularly limbic structures, are involved in the pathology of OCD, and ablation or stimulation can decrease symptoms in patients (Ahmari and Dougherty 2015). Two recent studies have investigated the impact of optogenetically manipulating orbitofrontal cortico (OFC)-striatal projections on OCD pathology and treatment. Burguiere et al. (2013) selectively stimulated ChR2-expressing lateral OFC-striatal neurons and their striatal terminals in SAPAP3 (corticostriatal postsynaptic density protein)-knockout mice, which display anxiety and the OCD-like symptom of perseverative grooming. Stimulation reinstated normal responses, i.e., inhibited conditioned grooming (Burguiere et al. 2013). Another study (Ahmari et al. 2013), published in the same journal issue, activated OFC-ventromedial striatal neurons in normal mice to mimic hyperactivity in the corticostriatal-thalamocortical circuit, previously observed in OCD patients (Ahmari and Dougherty 2015). While acute stimulation of ChR2-expressing neurons did not generate repetitive behaviors, brief repeated activation over several days resulted in a progressive increase in grooming behavior that correlated with an increase in stimulation-evoked firing rate of neurons in the ventromedial striatum (Ahmari et al. 2013).
The specific timing of dopamine release and NAc activity could partly underlie conditioned place preference (CPP) behavior. Changes in limbic BG reward system and synaptic modifications in substance abuse disorders have been investigated with viral vector-mediated optogenetic stimulation in animal models (for review see Britt and Bonci 2013; Stuber et al. 2012). Stimulation of ChR2-expressing mesoaccumbens dopamine neurons resulted in CPP in drug-naive behaving mice. Phasic, but not tonic, optogenetic stimulation of dopamine somata in VTA induced drug addiction behavior as well as transient dopamine release measured by FSCV (Tsai et al. 2009). Additional neuronal targets for optogenetic induction of CPP in mice, without additional drug application, are MSNs in the NAc (Airan et al. 2009). Activation of an opsin-adrenoceptor chimera, Opto-α1, in NAc MSNs increased NAc neuronal activity. More interestingly, this stimulation induced CPP behavior in drug-free mice (Airan et al. 2009). Thus optogenetic stimulation studies have refined the timing of dopamine release in reward pathways.
Optogenetic stimulation and CPP tasks have also been used to dissect the structures and pathways involved in drug-induced CPP. One of the drugs often used to induce CPP in rodents is the dopamine reuptake blocker cocaine. Similar to the dorsal striatum, MSNs in the NAc selectively express D1- or D2-like receptors, which have been selectively targeted in Cre-transgenic mice (Lobo et al. 2010). Combined cocaine administration and stimulation of D1-like MSNs increased CPP, while cocaine combined with D2-like receptor stimulation attenuated it. In a control experiment, a HSV-ChR2 vector was injected into the NAc of wild-type animals to globally activate NAc neuron activity during cocaine + blue light stimulation, which enhanced CPP (Lobo et al. 2010). In light of their results, the authors concluded that there might be an imbalance in these two NAc MSN subtypes in the addicted brain, highlighting that correcting this imbalance could be a treatment strategy. Besides MSNs, the NAc also contains interneurons. Witten et al. (2010) expressed eNpHR or ChR2 in NAc cholinergic interneurons in ChAT-Cre mice. In vivo NAc light stimulation revealed that activation of ChAT interneurons inhibited the majority of MSNs, while inhibition of ChAT interneurons excited most MSNs. Furthermore, photoinhibition (via eNpHR stimulation) of ChAT interneurons resulted in reduced cocaine CPP, illustrating that cholinergic neuron activity has a role facilitating cocaine conditioning in freely moving animals (Witten et al. 2010). Recently, selective optogenetic VTA dopamine neuron self-stimulation in DAT-Cre mice induced and progressed cocaine addiction and was associated with increased NAc MSN synaptic plasticity (Pascoli et al. 2015).
In addition to generating or modulating animal models for addiction behavior, virus-mediated optogenetics has been used to investigate potential reward system-based treatment options. Optogenetic inhibition of prelimbic cortico-accumbens neurons can prevent reinstatement of cocaine seeking behavior (Stefanik et al. 2013). In particular, photoinhibition (via eNpHR or ArchT) of prelimbic cortical neurons or their afferents in the NAc core prevented reinstatement of cocaine seeking behavior. This study strengthens knowledge obtained from pharmacological blockers about the role of prelimbic cortex and NAc core in cocaine reinstatement behavior (Kalivas et al. 2005). Optogenetic activation of medial PFC afferents in NAc shell induced long-term potentiation in D1-expressing MSNs, showing that synaptic plasticity is required for locomotor sensitization to cocaine (Pascoli et al. 2012). Specifically, bilateral low-frequency stimulation of ChR2-transduced medial PFC afferents in the NAc shell depotentiated EPSC in D1-expressing MSNs, abolished locomotor sensitization to cocaine, and reset behavioral sensitization induced by chronic cocaine injections. More recently, optogenetic experiments have revealed that synaptic plasticity of PFC and hippocampal inputs onto NAc D1 MSNs are associated with cocaine seeking behavior, with hippocampus conveying information on intensity of cocaine seeking and PFC differentiating outcomes related to cocaine and noncocaine actions (Pascoli et al. 2014). Creed et al. (2015) confirmed that low-frequency optogenetic stimulation prevents cocaine addiction and can be mimicked by electrical low-frequency DBS when combined with injection of a D1-like antagonist. These studies show that knowledge gained by optogenetic experiments can direct electrical DBS regimens that could be rapidly translated into the clinic to treat addictive behavior.
Optogenetic stimulation experiments have revealed new firing patterns and neural circuit-specific mechanisms of depression that could improve DBS treatment in patients. The limbic-reward BG pathway is also involved in depression and can be targeted with electrical DBS to treat patients (see Williams and Okun 2013) and optogenetic stimulation in animal models (see Lobo et al. 2012). Phasic, but not tonic, stimulation of ChR2-expressing dopaminergic VTA-NAc neurons mediates susceptibility to subthreshold social-defeat stress in freely behaving mice, measured by social avoidance and decreased sucrose preference, while VTA-medial PFC stimulation was not effective (Chaudhury et al. 2013). Photoinhibition (via NpHR stimulation) of VTA mesoaccumbens neurons reversed “depression-like” symptoms in previously susceptible mice, whereas inhibition of VTA-medial PFC neurons promoted susceptibility (Chaudhury et al. 2013).
The use of optogenetics in animal models of BG pathologies has immensely increased our understanding of neuronal subpopulations (e.g., D1- and D2-expressing MSNs in PD or addictive behavior) or projection pathways (e.g., VTA-NAc vs. VTA-mPFC in depression-related stress). Furthermore, this technology raises questions about previously assumed mechanisms in pathology and treatment that may need to be reassessed. Overall, this knowledge has the potential to improve treatment options for BG-related disorders, either promptly by applying new findings to existing treatment options or by developing new treatment strategies in the long term.
Chemogenetic stimulation, a technique that uses small-molecule-mediated activation of engineered proteins, is another major technology that has increased our understanding of neuronal circuits in health and disease (see Sternson and Roth 2014). The chemogenetic concept uses a viral vector to introduce a nonnative protein, such as an engineered G protein-coupled receptor that is only activated by a specific nonnative chemical, into specific brain regions. Similar to optogenetic stimulation, nonnative protein expression can be restricted to specific neuronal phenotypes. Following this, systemic administration of a drug that crosses the blood-brain barrier activates this nonnative chemoreceptor and associated changes in neuronal activity, chemistry, and behavior can be quantified. In contrast to optogenetic stimulation, the latency (minutes) for activating or inhibiting nonnative proteins is slow, and once the chemical has been administered the effects last for hours. Therefore, chemogenetic stimulation is ideal for investigating long-term changes within a circuit, chronic rather than acute effects, or for mimicking physiological parameters that occur over longer time frames (hours-days), e.g., circadian rhythms. Finally, chemogenetics has the advantage of selectively activating cellular signaling systems without the need to introduce a fiber-optic probe into the brain.
The idea to activate receptors by nonnative ligands goes back to the 1990s (e.g., Coward et al. 1998; Strader et al. 1991) and led to the generation of several “receptors activated solely by synthetic ligands” (RASSLs) (for review see Conklin et al. 2008). However, early RASSLs were not widely used or adapted in neuroscience because of restricted selectivity of either the engineered receptors or the nonnative ligands (for review see Sternson and Roth 2014). A solution for these problems arose with the development of a platform termed “designer receptors exclusively activated by designer drugs” (DREADDs), where intracellular signaling is modulated via the selective activation of genetically engineered G protein-coupled receptors by pharmacologically inert, small druglike molecules (Armbruster et al. 2007). In the last decade, DREADDs have tremendously increased our understanding of neuronal circuits and the number of publications using DREADDs has increased exponentially (English and Roth 2015). Chemogenetics can also be used for cell type-specific control of ion conductance by targeting genetically engineered ligand-gated ion channels (LGICs), which enables mimicking of ionotropic, rather than slower G protein-coupled, modulation of neuronal activity (see Sternson and Roth 2014).
Similar to optogenetics, virus-mediated gene transduction is the most common approach for introducing chemogenetic constructs into neurons. AAVs, HSVs, and lentiviruses are viral vectors routinely used to introduce DREADDs, with canine adenovirus providing a recent additional option (for overview see Urban and Roth 2015; Zhu and Roth 2015). Transgenic animals have also been produced (Alexander et al. 2009; Farrell et al. 2013; Guettier et al. 2009). The original DREADDs (Armbruster et al. 2007) were engineered from human muscarinic acetylcholine receptors and are activated by clozapine N-oxide (CNO), an inactive clozapine metabolite. Three engineered muscarinic DREADDs (see Fig. 3B) that have been frequently used in neuroscience share two point mutations that make them responsive to CNO and unresponsive to acetylcholine (Farrell and Roth 2013; Urban and Roth 2015; Wess et al. 2013; Zhu and Roth 2015): hM4Di (human M4 muscarinic DREADD receptor) is coupled to Gαi signaling and silences neuronal activity (Armbruster et al. 2007), hM3Dq is coupled to Gαq signaling and increases neuronal activity (Alexander et al. 2009), and rM3Ds (chimeric rat M3 muscarinic DREADD receptor with intracellular loops from the turkey β1-adenoceptor) is coupled to Gαs signaling and modulates neuronal activity (Guettier et al. 2009). The chimeric rM3Ds (Guettier et al. 2009) expanded the DREADD toolbox to include Gs-coupled receptors, which no native muscarinic receptors couple to.
Most DREADDs are activated by CNO, an inert pharmacological agent with known high bioavailability and blood-brain barrier permeability in humans and mice (Rogan and Roth 2011). CNO undergoes extensive back-metabolism to the pharmacologically active clozapine in humans (Chang et al. 1998), which might limit rapid translation of this approach to humans and may also cause confounding effects in animal experiments. Use of CNO as the common inert ligand limits the effectiveness of these DREADDs for multiplexed and bidirectional chemogenetic control within the brain. To overcome this, Vardy et al. (2015) recently developed a new kappa-opioid receptor DREADD (KORD) coupled to Gi (see Fig. 3B), which is activated by salvinorin B (SalB), an inactive, druglike metabolite of the KOR-selective agonist salvinorin A. Vardy et al. (2015) demonstrated proof of concept that combining different DREADDs in the same animals can successfully facilitate multiplexed chemogenetic control of behavior.
Parallel to the ongoing development of DREADDs, different classes of chemogenetic LGIC tools for neuron phenotype-specific perturbation in mammalian brains have been developed (for overview see Sternson and Roth 2014). In the single BG study, administration of the antiparasitic drug ivermectin selectively silenced virally transduced BG neurons expressing a mutated glutamate-gated chloride (GluCl) channel from Caenorhabditis elegans (roundworm) (Lerchner et al. 2007; Slimko et al. 2002); therefore, LGICs are not discussed below.
Motor and limbic BG pathways (Fig. 1) have been targeted with chemogenetic experiments, with the majority of studies performed in mice or rats. Below we discuss the impact DREADDs have had on increasing knowledge of BG function.
Applications in health.
Distinct functional properties of direct and indirect striatal MSNs have been explored with DREADDs. Chemogenetic perturbations of MSNs or cortical inputs to MSNs during striatal excitatory synaptogenesis (mouse postnatal days 8–14) disrupted neuronal plasticity and MSN morphology (Kozorovitskiy et al. 2012). Extensive bilateral inhibition by CNO activation of hM4Di in direct or indirect MSNs in the dorsolateral striatum resulted in opposing changes: inhibition of direct (D1-like) MSNs decreased miniature EPSC frequency and spine density, whereas these measures increased for indirect (D2-like) MSNs. In contrast, CNO-mediated inhibition of hM4Di-expressing corticostriatal neurons during excitatory synaptogenesis decreased miniature EPSC frequency and spine density for both direct- and indirect-pathway MSNs. For indirect-pathway MSNs, hM4Di activation effects persisted into early adulthood (Kozorovitskiy et al. 2012); however, the effect on direct-pathway MSNs was not reported. Activity in direct- and indirect-pathway MSNs and cortical inputs is important for determining formation and maintenance of corticostriatal synapses and thus has important implications for overall BG function.
Chemogenetics have been used to explore the role of specific striatal MSNs and VTA-SN neuron subpopulations in the control of movement. Direct-pathway MSNs were either inhibited (hM4Di) or activated (rM3Ds) by CNO administration during discrete periods of training. Inhibition impaired performance, whereas activation significantly enhanced it and altered behavior by causing a high-reward preference (Ferguson et al. 2013). Recently, Gi-DREADD KORD was combined with hM3Dq in GABAergic VTA-SN neurons to allow bidirectional control of behavior in the same animals. While SalB silenced these neurons and enhanced locomotion, CNO activated them and decreased locomotion (Vardy et al. 2015).
In addition to DREADD-induced changes in behavior or neuronal recordings, it can be combined with in vivo imaging. DREAMM (DREADD-assisted metabolic mapping) generates whole-brain metabolic maps of cell-specific functional circuits (Anderson et al. 2013; Michaelides et al. 2013). hM4Di-mediated inhibition of MSNs in the NAc shell combined with [18F]fluorodeoxyglucose (FDG) μPET imaging revealed concurrent engagement of distinct corticolimbic networks: inhibition of direct MSNs significantly increased FDG uptake in posterior VP and VTA-SN, and inhibition of indirect MSNs decreased FDG uptake in rostral VP (Michaelides et al. 2013). The same DREAMM technique was employed by Anderson et al. (2013) in rat models of opiate addiction and depression, showing that hM4Di-mediated inhibition of prodynorphin-expressing neurons in periamygdaloid cortex increased metabolic activity in the extended amygdala, a key structure of the extrahypothalamic brain stress system. DREADDs have also been used to image the location and intensity of lentivirus-mediated hM4Di expression in the putamen of nonhuman primates with in vivo PET (Nagai et al. 2014) and matched with postmortem immunohistochemical analysis of hM4Di expression. These researchers have shown in other monkeys that CNO administration to AAV-mediated hM4Di expression in the ventral striatum altered performance in a reward-size task, similar to effects seen with muscimol inactivation (Nagai et al. 2014). Thus DREADDs can produce behavioral changes in higher-order species. These multidisciplinary studies show that DREADDs and PET imaging can enhance our understanding of the neuronal mechanisms underlying changes in behavior.
Applications for neurological and psychiatric disorders.
One strategy to treat neurological disorders is to directly target neuronal populations affected in the condition. DREADD activation of cholinergic pedunculopontine tegmental nucleus (PPTg) neurons using hM3Dq increased PPTg spiking activity and improved motor deficits in PD model rats (Pienaar et al. 2015); thus selective PPTg activation might be a potential target to treat PD. Another strategy has been to express a DREADD in transplanted neurons, enabling subsequent control by systemic CNO injections (Dell'Anno et al. 2014). In particular, unilateral 6-OHDA-lesioned rats were transplanted with hM3Dq-expressing induced dopaminergic (iDA) neurons, generated from skin fibroblasts, in the lesioned striatum. The grafted neurons were functionally integrated, and CNO-activated iDA neurons improved motor deficits, similar to primary dopamine neurons grafted in lesioned rats (Dell'Anno et al. 2014). Two recent reviews (Sharma and Pienaar 2014; Vazey and Aston-Jones 2013) discuss the current state and possibilities for future applications of DREADDs in the field of PD research.
Another strategy to treat disorders is to find neuroprotective treatments. Chemogenetics have elucidated neuroprotective effects of the cannabinoid 1 (CB1) receptor in the dorsolateral striatum in health (Chiarlone et al. 2014). CNO activation of hM3Dq-expressing corticostriatal neurons enhanced glutamatergic transmission and excitotoxicity, which was associated with reduced DARPP-32 immunoreactivity and impaired RotaRod performance. CNO effects were abrogated by the selective NMDA receptor antagonist MK-801 or tetrahydrocannabinol (THC), a CB receptor agonist, indicating that striatal CB1 cannabinoid receptors are neuroprotective. This physiological study has important implications for HD and other neurological disorders (Chiarlone et al. 2014).
Chemogenetic approaches are increasing knowledge of the limbic BG pathways involved in substance abuse disorders (for review see Ferguson and Neumaier 2015; Yager et al. 2015). HSV vectors with promoters for enkephalin and dynorphin that express the DREADD hM4Di have been used to examine the role of striatopalldial or striatonigral pathways in amphetamine-induced locomotor sensitization (Ferguson et al. 2011). Stimulation of hM4Di by CNO hyperpolarized and silenced transduced striatopalldial or striatonigral neurons, respectively, and was associated with increased and decreased amphetamine-induced sensitization. Furthermore, Ferguson et al. (2011) demonstrated that acute drug effects can be determined by using behavioral adaptations associated with repeated drug exposure. Inhibition (via hM4Di + CNO injection) of dorsal striatum direct- or indirect-pathway MSNs had no effect on acute amphetamine-induced hyperlocomotion in rats but caused drug addiction with locomotor sensitization after repeated amphetamine exposure. Transient indirect-pathway inhibition enhanced the development and persistence of sensitization, while inhibition of direct-pathway MSNs impaired sensitization persistence (Ferguson et al. 2011). These studies indicate that different neuronal populations in the BG are involved in specific aspects of amphetamine-induced locomotor activity. Chemogenetics can target subpathways to treat drug addiction, and findings may help direct future drug development (Ena et al. 2011).
Another animal model of drug addiction is based on the fact that cocaine administration increases the amount of ΔFosB in NAc neurons of mice (Kelz et al. 1999). In transgenic mice with increased ΔFosB expression, recombinant HSV-GluR2 injection into the NAc caused GluR2 overexpression and enhanced rewarding effects of cocaine. In contrast, overexpression of GluR2(Q) caused drug aversion, highlighting a cocaine drug addiction mechanism that could be manipulated to treat this disease (Kelz et al. 1999). VTA and SN dopamine neuron activity appears to underlie this cocaine-induced behavior. Inhibition of VTA and SN midbrain neurons expressing KORD or hM4Di (via SalB or CNO application, respectively) decreased spontaneous and acute cocaine-induced locomotor activity in rats and also inhibited midbrain putative dopamine neuron activity in brain slices (Marchant et al. 2016).
DREADDs have been used to investigate limbic BG nuclei cell populations involved in reinstatement of drug seeking, a model of addiction relapse. Distinct neuronal projections from VP to VTA underlie specific aspects of reinstatement of cocaine seeking (Mahler et al. 2014). Transient inhibition of hM4Di-expressing rostral VP neurons blocked reinstatement of drug-associated, cue-induced cocaine seeking behavior, while caudal VP neuron inhibition blocked cocaine-primed reinstatement. Chemogenetic manipulation of NAc, inspired by the finding that NAc lesions reduce relapse rates in alcohol-dependent patients (Wu et al. 2010), altered ethanol consumption and self-administration, while sucrose consumption was unaffected (Bull et al. 2014; Cassataro et al. 2014). Transient inhibition (via hM4Di) of NAc core and shell neuronal populations reduced ethanol consumption, whereas hM3Dq-mediated transient activation had no effect, consistent with electrolytic lesions of NAc core (Cassataro et al. 2014). NAc core astrocyte-specific DREADDs (via hM3Dq) can modulate ethanol-related drug-seeking behavior (Bull et al. 2014); activation facilitated intracranial self-stimulation and reduced the motivation for ethanol self-administration after 3 wk of abstinence. Scofield et al. (2015) recently demonstrated that transient CNO activation of hM3Dq-expressing NAc core astrocytes increased NAc core extracellular glutamate levels in vivo and inhibited cue-induced reinstatement of cocaine seeking, probably by inhibiting cue-induced synaptic glutamate spillover. Neuronal subpopulations in VP as well as neuronal and astrocyte populations in NAc are involved in drug-seeking behavior.
Virus-mediated chemogenetic experiments have increased our understanding of physiological interactions within the BG and BG connections with other brain regions. Furthermore, chemogenetics has revealed that distinct cell populations (e.g., NAc neurons and astrocytes) are involved in addiction-related behavior and has provided new perspectives for potential PD treatments (e.g., long activation periods of PPTg neurons or transplanted neurons). Further development of chemogenetic technology has the potential to improve pharmacological therapies of BG-related disorders.
Creating and Treating Animal Models of BG Disorders
Viral vector technology has been routinely used to alter cellular and genetic functions by increasing or decreasing protein expression or delivering genes for therapeutic benefit. However, more recently viral vectors have been used to create genetically modified animals, create disease models, or alter transcription to study epigenetic changes.
Cre-lox transgenic rodents have been used to develop “brainbow” labeling (Cai et al. 2013), which has been used to elucidate the roles of striatopallidal and striatonigral pathways in health and disorders like PD (Ena et al. 2011; Kravitz et al. 2010). The technology has been combined with bacterial artificial chromosome (BAC) Cre-recombinase driver lines and AAVs to target expression in specific neuron or glial populations (Gerfen et al. 2013). Cerebral cortex and BG BAC Cre-driver lines have been characterized (Gerfen et al. 2013) and will be useful for exploring neuronal circuits.
Viral vectors have been used to insert or delete genes for gene therapy options. Zinc finger nucleases, transcription activator-like effector nucleases (TALENs) (Joung and Sander 2013), and CRISPR-Cas can be used to edit the genome; CRISPR/Cas technology is superseding zinc finger nucleases (Ledford 2015). CRISPR is an acronym for clustered, regularly interspaced, short palindromic repeats; it identifies the gene to be edited. Cas stands for CRISPR-associated protein and is a RNA-guided DNA endonuclease enzyme that cleaves the host DNA to add or delete the target gene. CRISPR-Cas technology is still in its infancy and has not been utilized in the BG but could be used in the future to create animal models of disease and compile genetic libraries that will aid genetic screening and eventually therapeutic applications (Sander and Joung 2014).
Animal models of BG dysfunction have been created with viral vectors. In some, primarily early-onset cases, PD has a genetic component. Based on this, viral vectors (lentivirus, AAV, HSV, and adenovirus) have been used to create animal models of these genetic forms (Löw and Aebischer 2012). These viral vector models have been used in rats, mice, and marmosets and are based on dominant mutations of alpha-synuclein (in Lewy body aggregates) and leucine-rich repeat kinase-2 (LRRK-2) genes linked to PD, which cause overexpression and aggregation of these proteins in the BG and degeneration of dopaminergic neurons. Recessive mutations of genes associated with early-onset PD have also been created by overexpressing parkin substrates, which causes toxic accumulation of parkin and eventually dopaminergic neuron loss in the substantia nigra (see Löw and Aebischer 2012). Parkin substrates that have been used are Pael-R (Parkin-associated endothelin receptor-like receptor), CDCrel (protein at presynaptic terminals), p38/JTV (a substrate aminoacyl-tRNA synthetase cofactor), and synphilin-1 (a neuronal cytoplasmic protein). AAVs have delivered small interfering RNA to the BG to decrease TH gene expression, a dopamine biosynthesis enzyme, which resulted in PD-like behavioral deficits, illustrating that small interfering RNA could be used to create other animal models or gene therapy treatments (Hommel et al. 2003). Optogenetic activation in D1- or D2-Cre mice has also resulted in a behavioral animal model of PD. Activation of ChR2-expressing D2-like MSNs in D2-Cre transgenic mice induced parkinsonism, but activation of the direct pathway in D1-Cre mice increased movements and decreased freezing and bradykinesia (Kravitz et al. 2010). This model could be used to study the relationships in changes in BG circuits and motor function (Kravitz et al. 2010).
To date, the only viral vector-derived animal model for HD involves lentivirus-mediated expression of the mutant human huntingtin gene that has 82 CAG repeats, causing formation of inclusions in striatal neurons and loss of striatal MSNs in rodents (Ramaswamy et al. 2007), which is similar to the pathology that occurs in the BG of humans. Currently, these researchers are developing a nonhuman primate model of HD (Ramaswamy et al. 2007). Finally, circuitry has been selectively lesioned, using viral vectors to examine how a pathway normally functions and to consider how it might contribute to BG disorders. Takada et al. (2013) developed a new vector, NeuRet, based on HIV with a hybrid of rabies and VSVg glycoproteins, to deliver immunotoxin to ablate the hyperdirect pathway neurons in primates, showing that the motor cortex excites GPi neurons through this pathway, which was later confirmed with pharmacophysiological and electrical stimulation methods.
Treating BG disorders.
To best answer any research question or develop gene therapy strategies using viral vectors, the optimal vector needs to be chosen. The advantages of one vector over others is described in Viral Vector Use in Neuroscience; however, most studies that use vectors rarely outline the rationale for their choice. Choosing the best vector is a critical step for progressing gene therapy technologies into clinical trials.
The longevity and specificity of vectors in BG nuclei have been examined in rodents, pigs, and primates. Lentiviral vectors have been compared to a Moloney murine virus in rodent striatum and were shown to provide sustained expression for at least 6 mo and transduced >88% of striatal neurons. A major advantage of lentiviral vectors is that they can transduce nondividing cells (Blömer et al. 1997). However, AAVs are more commonly used for gene therapy in the BG. Recombinant AAV5 has been preferred over AAV1 because AAV5 also transduced nonneuronal cells in rat and pig striatum (Kornum et al. 2010). Studies comparing viral vectors in monkey brains, because they more closely resemble the human brain, found that AAV2 injected into the putamen, thalamus, and corona radiata was easier to visualize by fMRI than AAV1 (Fiandaca et al. 2009; Fiandaca and Federoff 2014) and thus may translate effectively to patient applications in the clinic. To this end, AAV2 serotypes have been compared in primates to determine transduction efficiency. Injection of AAV2 serotypes 1, 2, 5, and 8 into the caudate putamen of rhesus monkeys revealed that more cells were transduced with AAV2/8 than the others; however, this was not specific to neurons (Sanchez et al. 2011).
Several technologies introduced above provide important potential mechanisms for gene therapy. For example, CRIPSR/Cas technology could eventually be used for gene therapy treatments in the BG. First, effective delivery needs to be achieved, aberrant mutations in the genome avoided, and ethical issues around being able to change the human genome addressed (Ledford 2015). Currently, gene therapy is the most advanced viral vector technology for treatment applications in humans (Denyer and Douglas 2012; Lowenstein et al. 2014).
Viral vectors have also been used to treat animal models of BG dysfunction as a preclinical step toward gene therapy clinical trials. To date, most of the studies focus on PD or HD. To determine which gene therapy tool may be most efficacious for PD, viral vector delivery is sometimes combined with imaging techniques. Glial cell-derived neurotrophic factor (GDNF) has been delivered in vivo to the striatum with AAV2 vectors in 6-OHDA-lesioned PD model rats and monkeys. In rodents, the striatum was the most effective site of GDNF delivery because it reached all affected BG areas compared with delivery in the substantia nigra (Ciesielska et al. 2011). In monkeys, low levels of GDNF delivery protected neurons, did not interrupt dopamine synthesis, and provided behavioral attenuation of symptoms (Eslamboli et al. 2005). Other approaches include increasing dopamine-synthesis enzymes like aromatic-amino acid decarboxylase (AADC), GTP cyclohydrolase 1 (GCH), and TH in 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) parkinsonian monkeys. Delivery of these dopaminergic enzymes increased the number of dopamine-positive neurons in the putamen, increased dopamine levels in the brain, and restored motor function (Muramatsu et al. 2002).
Gene therapy has also been investigated in animal models of HD. Neurotropic factors are being used to investigate neuronal survival in quinolinic acid (QA) or transgenic animal models of HD. Viral delivery of brain-derived neurotrophic factor (BDNF) reduced striatal cell loss in QA HD model rats (Henry et al. 2007; Southwell and Patterson 2011). A member of the GDNF family, neuretin, similarly rescued striatal MSNs in models of HD when delivered via viral vector (Southwell and Patterson 2011). Another use of viral vectors in the treatment of HD could be the delivery of growth factors like BDNF to neural stem cells in the subventricular zone, which in turn proliferate into MSNs and migrate to the striatum (Henry et al. 2007; Southwell and Patterson 2011). Similarly, AAV1/2 delivery of BDNF or GDNF enhanced striatal and dopamine neuron survival in vitro (Kells et al. 2007). Finally, calmodulin delivery via AAV has improved motor function, reduced the number of inclusions and cell loss, and increased body weight, possibly by improving HD-associated deficits in calcium signaling (Southwell and Patterson 2011).
Although less studied than PD and HD, gene therapy targets have also been identified in models of psychiatric disorders. Anatomical and molecular BG targets involving serotonin and BDNF have been identified (Gelfand and Kaplitt 2013) in depression, schizophrenia, and OCD, possibly because these disorders have overlapping symptoms. BDNF levels are decreased in postmortem tissue from depression patients (Deisseroth 2011), and serum response factor, which is an activator of early genes, is also decreased in NAc in postmortem tissue (Ramanan et al. 2005). The serotonin receptor 5-HT1b is thought to be involved in depression. It interacts with a cytoplasmic protein, pII, which causes depression-like symptoms in knockout mice. AAV-mediated delivery of small interference RNA has been used to block pII production in the NAc in normal mice, which caused depression-like behaviors (Gelfand and Kaplitt 2013). Optogenetic stimulation in the pII knockout produced antidepressant effects. Similar targets have been identified for treating drug addiction. pII-knockout mice have increased CPP after cocaine administration, which can be reversed after AAV pII gene therapy (Kaplitt et al. 2010). Mice with knockdown of the BDNF receptor TrkB in D1-expressing MSNs also had increased CPP in response to cocaine administration, while TrKB in D2 MSNs or optogenetic stimulation of D2 MSNs in the NAc (Lobo et al. 2010) had the opposite effect. Schizophrenia can have a genetic basis, but further studies are needed to understand the complexity of schizophrenia symptoms. Genetic deletion of Dgcr8, which is a microRNA processor required for proper dendrite development, affects working memory (Stark et al. 2008), whereas knocking out the DISC 1 gene (disrupted in schizophrenia 1) affects progenitor proliferation in the dentate gyrus (Mao et al. 2009). A deficiency in the postsynaptic scaffolding protein SAPAP3 (SAP90/PSD95-associated protein 3) may underlie OCD symptoms because SAPAP3-knockout mice produce OCD behaviors, and OCD behaviors are reversed with lentiviral vector delivery of SAPAP3 to the dorsolateral striatum or administration of serotonin uptake inhibitors (Welch et al. 2007). Another therapeutic target for OCD is the transmembrane protein Slitrk5 (SLIT and NTRK-like protein 5), because knockout mice express OCD-like behaviors (Gelfand and Kaplitt 2013; Shmelkov et al. 2010). Both SAPAP3 and Slitrk5 could be used as gene therapy targets in OCD. Thus these molecular BG targets in preclinical studies of psychiatric disorders could eventually lead to treatment options in patients.
Some preclinical gene therapy experiments have been moved into patient clinical trials. These human trials were monitored with PET scanning or ELISA assays and clinical tests. For PD, clinical trials have reached open label, phase I–II. Gene therapy delivery of GAD, neurturin, or ProSavin (AADC, TH, and cyclohydrolase 1) via an AAV or lentiviral vector to the STN or striatum saw some improvements in movement deficits in PD patients (Feigin et al. 2007; Kaplitt et al. 2007; Olanow et al. 2015; Palfi et al. 2014). Although gene therapy did not always significantly improve clinical measures (Kaplitt et al. 2007; Olanow et al. 2015), patients did not suffer any adverse effects from the treatment, indicating that gene therapy may become an effective treatment for BG disorders. In the future, gene therapy may become a conventional treatment for many neurological disorders.
Viral vector use in BG research is expected to continue to increase in the next decade as more groups employ vectors for optogenetic and chemogenetic stimulation, to trace pathways and create or treat animal models of BG disorders. The specificity of viral vectors to target a neuronal or glial population could be used to further delineate BG circuits, connectomes, and microcircuits. Imaging can be done in vitro or in vivo from the ultrastructural level through to large blocks of brain, and continued developments in microscopy and bioimaging (fMRI, PET) will improve the resolution of images at all levels. Although injection of viral vectors combined with bioimaging may not become a routine clinical scan, it is possible that these technologies could be combined in rare conditions in patients. Nevertheless, the anatomy and physiology of inputs to and from the BG will continue to be explored in preclinical studies to better understand the role of BG in controlling behavior and the deficits that occur with BG dysfunction. Such basic science knowledge will be critical for guiding new treatment strategies to treat BG disorders such as PD, HD, OCD, depression, and drug addiction.
While viral vector technologies have been used in the BG, additional information would be gained by using multidisciplinary approaches. For example, viral vectors could be combined with tissue clearing techniques, optogenetic or chemogenetic stimulation, tissue processing, and imaging to examine anatomical and physiological changes at gross and ultrastructural levels in the same tissue (Pastrana 2010; Watanabe et al. 2014). Such approaches would address unanswered questions such as, “Does altered BG activity in PD change synaptic structure (number of synapses or dendritic spines) or function (plasticity mechanisms) downstream from the striatum?” These new technologies will also permit previously unimagined experiments such as investigating the relationship between changes in gene expression and behavior underlying drug addiction by using TALEs and LITEs (Konermann et al. 2013). An initial challenge for BG researchers is to extend their understanding and use of these multidisciplinary viral vector approaches.
Although gene therapy is in its infancy, it has been trialed in PD patients. Importantly, phase I–II trials have reported that gene therapy applied to the brain is safe, with few adverse effects and no reports of tumor development (Feigin et al. 2007; Kaplitt et al. 2007; Olanow et al. 2015; Palfi et al. 2014). Currently, gene therapy is not more efficient than other treatments, but these studies may be biased because they are usually offered to patients at later stages of the disease for whom conventional treatments have side effects. Now that gene therapy has been shown to be safe, it will be important to trial future strategies on patients earlier in the disease process. Controversial future gene therapy options include the use of CRISPR/Cas technology to edit the human genome to prevent a BG disorder or reduce its severity, perhaps by removing many of the CAG repeats in the huntingtin gene of HD patients. At this stage, CRISPR/Cas gene editing would need to be performed in preclinical studies aimed at reducing BG dysfunction and shown to be highly efficacious. Then, the ethics of CRISPR/Cas use in patients (Ledford 2015) would need to be debated before the first clinical study could be conducted.
Preclinical studies continue to explore potential new treatment options for BG conditions. Optogenetic stimulation in the BG has improved understanding of function (Kravitz et al. 2010) and has continued to tease out the mechanisms of DBS (Gradinaru et al. 2009). More recently, recordings of VA motor thalamus showed impaired movement-related modulations in firing rate and low-threshold calcium spike bursts in PD model rats (Bosch-Bouju et al. 2014). On the basis of this finding, we optogenetically stimulated VA motor thalamus with complex patterns (neuronal activity previously recorded from the VA of controls) and found that reaching performance improved in PD model rats (Seeger-Armbruster et al. 2015). Thus application of findings from basic science can successfully identify novel treatment approaches for BG disorders, and future chemogenetic studies will provide additional opportunities for creating new treatment strategies. An obvious step in the future will be the use of optogenetic and chemogenetic stimulation in patients. Development of other viral vector-based treatment options for human BG disorders is initially limited by imagining new ways to combine these tools and then applying them in preclinical studies.
This work was supported by grants from the Neurological Foundation of New Zealand (to L. C. Parr-Brownlie and S. M. Hughes), Health Research Council of New Zealand (to L. C. Parr-Brownlie and S. M. Hughes), Royal Society of New Zealand Marsden Fund (to S. M. Hughes), and Brain Research New Zealand, Centre of Research Excellence (to L. C. Parr-Brownlie and S. M. Hughes).
No conflicts of interest, financial or otherwise, are declared by the author(s).
Author contributions: R.J.S., S.S.-A., S.M.H., and L.C.P.-B. prepared figures; R.J.S., S.S.-A., S.M.H., and L.C.P.-B. drafted manuscript; R.J.S., S.S.-A., S.M.H., and L.C.P.-B. edited and revised manuscript; R.J.S., S.S.-A., S.M.H., and L.C.P.-B. approved final version of manuscript.
- Copyright © 2016 the American Physiological Society