|Home | About | Journals | Submit | Contact Us | Français|
Recent advances in optogenetics have permitted investigations of specific cell types in the nervous system with unprecedented precision and control. This review will discuss the use of optogenetic techniques in the study of mammalian neural circuitry in vivo, as well as practical and theoretical considerations in their application.
For hundreds of years, neuroscientists have gleaned information about the function of specific brain regions by observing what happens when they are injured. More recently, advances in imaging and electrophysiology have allowed scientists to characterize the activity of living brains while animals and humans are exposed to stimuli or perform tasks. These approaches have convincingly determined what different brain regions are doing, but have revealed little about how they function. For example, the striatum has been linked to voluntary movement for over a century , and since the 1960s it has been identified as a major site of dysfunction in Parkinson’s disease [2–4]. However, these facts reveal little about how the striatum controls movement or how striatal dysfunction contributes to Parkinson’s disease. In order to understand how our brains function, new technologies are needed to probe the function of specific cell types and circuits in behaving animals. Historically, the application of pharmaceuticals or electrical stimulation has been used for testing the function of specific cell types and circuits, but these techniques are often not very specific. The recent development of optogenetic methods in neuroscience has allowed researchers to sensitize distinct cell types to light, enabling the non-invasive activation, inhibition, and modulation of specific neuronal populations in living animals with millisecond precision. Techniques like these will lead to a richer understanding of how specific brain regions function, and how their dysfunction might be corrected.
Although a number of methods for optically stimulating specific cell types have been developed over the past two decades [5–10], they have generally relied on expression of multiple proteins or application of exogenous cofactors, limiting their utility in vivo. With the discovery that expression of a single protein—channelrhodopsin-2 (ChR2)—can mediate light-sensitive cation currents  that enable robust and temporally-precise control of neural activity in vitro [12,13] and in vivo [14–18], the use of optogenetic approaches in neuroscience has exploded. Since its initial discovery, many channelrhodopsin variants have been discovered or engineered to confer additional properties on the cells in which they are expressed [19–22]. At the time of this writing, at least 12 channelrhodopsin variants have been reported, each with a unique set of properties .
Whereas activating specific neural subtypes can reveal whether neurons are sufficient for driving a behavior, inhibiting specific neural types can determine whether they are necessary for that behavior. Methods for achieving this include physical or toxigenic methods to kill targeted cell types [24–26], expression of modified potassium or chloride channels that are sensitive to light or exogenous compounds [27–29], and expression of engineered or non-mammalian G-protein-coupled receptors [13,30–33]. Despite the utility of these techniques, newer optogenetic manipulations that are rapidly reversible and do not require a cofactor can be more useful in vivo, especially in applications that benefit from fast kinetics. In 2007, a chloride pump from the archaeon Natronomonas pharaonis (NpHR) was engineered to express in mammalian cells. When illuminated with yellow (~590nm) light, this pump inhibited spiking in neurons [34,35]. More recently, proton pumps have also been used to inhibit neurons . Expressing combinations of these proteins may allow for rapid inhibitory control over multiple neural populations using different frequencies of light. In addition, the activation maxima of certain inhibitory proteins is spectrally separated from ChR2, enabling bidirectional control of individual neurons using yellow and blue light .
In addition to modulating neural activity with light-activated ion channels or pumps, it is possible to control intracellular signaling pathways with light . Adenyl cyclase is a ubiquitous signaling molecule in many tissues of the body. While most forms of adenyl cyclase are not light sensitive, a light-activated adenyl cyclase was discovered in the protist Euglena gracilis . This light-activated cyclase was subsequently expressed in neurons, where it rapidly modulated cAMP levels with illumination [39,40]. Another approach for controlling intracellular signaling involves engineering G-protein coupled receptors to respond to light by integrating their intracellular domains with the light-sensing domain of rhodopsin. This approach has been successfully applied with Gq coupled adrenergic α1a, Gi coupled β2 receptors, and Gi/o coupled serotonin 5-HT1A receptors [41,42]. Finally, expressing naturally occurring opsins such as melanopsin (a Gq coupled receptor) can also confer optical control over G-protein signaling .
Although optogenetics has numerous advantages over other techniques, a number of practical and conceptual issues arise when these techniques are applied to mammals in vivo. These include relatively simple issues such as effectively illuminating large volumes of tissue, as well as more complex issues such as interpreting experimental results. Gross estimations have revealed that tissue can be safely and effectively illuminated up to about 1.5mm from the tip of an optical fiber . Most mouse brain regions are small enough to illuminate with one fiber, but experiments in larger rodents or primates may require a larger area of illumination. One way to achieve this is to implant multiple optical fibers, but this is not ideal as it increases tissue damage. This can be mitigated to some extent by etching fiber tips to reduce their size . An elegant alternative is to use a fiber with a machined tip that either distributes light omnidirectionally or diffuses light along the length of the fiber (Figure 1A). These machined fibers can be coupled to higher laser powers without increasing tissue damage, as they release light over a broader surface area than a flat-cleaved fiber. Alternatively, illumination can be achieved with head mounted light emitting diodes (LEDs) [45,46]. LEDs have several advantages over laser-coupled optical fibers, as they are much cheaper than lasers, more durable than optical fibers, and can be powered by small head mounted batteries that do not require the animal to be tethered throughout the experiment.
A complementary problem to light delivery is achieving abundant and specific expression of optogenetic proteins. Cell-type specific expression has been accomplished in mammalian tissue with transgenic mice, viral infection strategies, and in utero electroporation. In one group of transgenic lines, ChR2 was expressed in multiple brain regions under control of the Thy1 promoter [17,47]. In another line, ChR2 was expressed in the spinal cord and hind brain under control of the VGlut2 promoter, and used to drive locomotion . New transgenic mice are under development, including mice that contain optogenetic proteins such as ChR2 in floxed configurations that can be crossed into cre driver lines to drive cell type specific expression , an advantage that had previously been limited to viral expression strategies [50–55]. Transgenic mice are advantageous for expressing optogenetic proteins throughout a large brain structure with highly-reproducible expression patterns, although achieving sufficient expression levels for stimulation in vivo remains a challenge.
Although transgenic mice offer a convenient and reliable expression system, viral-based methods are more versatile than mouse lines for many applications. Viruses are easier to generate than transgenic mice, can be engineered to control expression level and cell-type specificity, and can drive expression in species that are not as genetically tractable as mice, such as rats or primates. Furthermore, viral expression is restricted to injected sites, making it possible to inject a virus in one structure and illuminate in another to investigate projections between two structures [16,55–57]. Multiple viruses can also be injected into the same brain structure to achieve multimodal control of distinct neuronal populations [19,34]. An alternative to viral-based methods is in utero electroporation, which can be used to introduce optogenetic proteins into specific cell types or layers, typically in the neocortex [45,58–60].
Many laboratories have used ChR2 to probe the function of specific cell types and circuits in mammalian brains. Among others, these include exploring the minimal number of neurons required to represent a perception , demonstrating that phasic firing of dopamine neurons can mediate conditioned learning , that parvalbumin-expressing cells in cortex contribute to gamma rhythms [51,52], that hypocretin expressing cells in the lateral hypothalamus contribute to sleep to wake transitions , and that striatal output pathways control motor output . Despite strong behavioral results, none of these studies conclusively showed that the targeted cell types were, in fact, efficiently excited by illumination in the awake animal. Nearly all in vivo validation of ChR2 stimulation has been performed in anesthetized animals, which does not accurately reflect the awake state. For example, neurons in awake animals are typically more depolarized and receive more synaptic drive. Under these conditions, neurons can enter a depolarization block when activated too strongly, which could reduce spiking of ChR2-expressing cells. Additionally, network effects such as lateral inhibition could result in a net inhibition of a targeted cell type, despite activation of ChR2. To investigate these issues, it is necessary to combine optogenetics with awake electrophysiology and record from target neurons while they are illuminated. Briefly, this approach involves implanting single or multiple ‘optrodes’ (combined electrode and optical fiber) into the brain of an animal that expresses an excitable optogenetic protein such as ChR2. By examining recordings from this optrode it is possible to conclusively determine what happens to the target cell type during illumination.
The first optrodes were made by simply gluing an optical fiber to a single tungsten microelectrode for use in anesthetized recordings . We, and others, have found that a slightly evolved design in which an optical fiber was glued on a 16-site silicon probe is ideal for anesthetized recording, as it allows 16 channels of recording in a spatially defined manner with minimal tissue damage (Figure 1B) . A similar silicon-probe based approach was recently used in chronic awake recordings . For awake recordings, we have had success with a simple design utilizing a zirconia ferrule and a short piece of optical fiber (Figure 1C). This ferrule assembly can be attached to a traditional microwire array, resulting in a compact optrode array for chronic awake recording and stimulation (Figure 1D). It is also possible to place the electrodes in tetrode configurations or mount optrodes in microdrives to record from different depths in the same animal. While optrodes based on silicon probes allow for the highest density recordings with the smallest damage to the brain, optrodes based on microwires are cheaper to build and can record neurons for months after implantation.
Optrode recordings can be used for more than validating ChR2 activation. Awake electrophysiology has historically been limited by a lack of reliable methods for identifying specific cell types in extracellular recordings. By combining optogenetics with awake recordings, specific cell types can be identified by testing their responsiveness to light. Once identified, the activity of these subtypes can be tracked to determine their behavioral correlates [44,63,64].
Several issues need to be addressed when identifying cell types based on ChR2 expression. The first concerns stimulation parameters for identifying ChR2-expressing neurons. It is possible to use brief (<10msec), high powered (>10mW) laser pulses to drive single action potentials in neurons [63,64]. However, illuminations like this can cause large populations of ChR2-expressing neurons to fire at once, which can increase synchronous multi-unit activity on a recording electrode, making spike sorting difficult . This multi-unit activity can also infiltrate the recording of a well isolated unit and spuriously lead a researcher to conclude that the isolated unit expresses ChR2. Finally, high powered laser pulses can cause photoelectric artifacts on recording electrodes [44,63,65]. In the striatum, we have found that longer (1s), low power (~0.1 to 3.0mW) laser pulses are sufficient to drive spiking in ChR2 expressing neurons, while avoiding both large increases in multi-unit activity and photoelectric effects. Useful stimulation parameters are dependent on the recorded brain structure, and different stimulation parameters have been used to mitigate these issues in the hippocampus .
After determining useful stimulation parameters, identifying ChR2-expressing neurons can still be difficult due to the high interconnectivity among neurons in most brain regions. For example, a non-ChR2 expressing neuron could appear to be light-responsive if it receives excitatory input from a neighboring ChR2-expressing neuron. In brain regions that are interconnected by excitatory synapses exhibiting short-term depression of glutamate release, this issue can be addressed with trains of moderate frequency laser pulses. Due to the rapid kinetics of ChR2, neurons that express ChR2 reliably respond to each light pulse in a train (at rates <40Hz). In contrast, synaptically activated neurons respond unreliably after the first few pulses, presumably as the readily releasable pool of glutamate is depleted .
This specific confound is less likely in inhibitory structures. For example, the striatum does not contain any excitatory neurons, so increases in spiking that follow the laser pulse cannot be due to local excitatory drive. However, inhibitory structures like the striatum often exhibit lateral inhibition which can overpower light-activated currents in ChR2-expressing neurons. In these cases, a lower laser power can reveal ChR2-mediated responses. For example, some striatal neurons respond most strongly to low power laser illumination, because higher laser powers either depolarize neurons too much (inactivating sodium channels and driving neurons into a depolarization block) or drive more lateral inhibition from other ChR2-expressing cells (Figure 2A). Other striatal neurons respond better to higher laser powers (Figure 2A), likely because they express lower levels of ChR2 or do not receive adequate illumination at lower laser power. In such cases, it can be useful to illuminate with a series of pulses of incrementing laser powers to determine which neurons express ChR2. As a final note, while it is possible to identify ChR2 expressing neurons with these techniques, it is difficult to conclude anything about the cell type of neurons that do not respond to the light. Non-responsive neurons may not express high enough levels of ChR2 to drive spiking, may not be effectively illuminated, or could be synaptically inhibited.
Theoretically, it should also be possible to identify and track more than one cell type simultaneously. This would involve targeting multiple optogenetic proteins to different cell types, and using the unique characteristics of these proteins to identify each cell type. For example, spectrally-distinct channelrhodopsin variants could be targeted to two cell populations, which can be selectively stimulated with different wavelengths of light . The unique kinetics of different optogenetic proteins can also be used to identify different cell types. We have expressed ChR2 in one neuron type in the striatum and a step-function variant of ChR2 (ChR2-SFO) in another. Both of these proteins are rapidly activated by blue light, but the ChR2 expressing neurons stop firing when the light turns off, while ChR2-SFO expressing neurons continue firing after the light is shut off (Figure 2B). Using combinations of spectrally and kinetically distinct optogenetic proteins under control of cell-type specific promoters, it may be possible to identify more than two cell types in the same recording.
Optogenetics has evolved extremely rapidly in the years since light was first used to modulate the activity of neurons. Future years are expected to bring additional optogenetic tools, as well as new applications for these tools. Some of these tools may include proteins with new functionalities such as more restricted activation spectra, activation by additional frequencies such as infrared that better penetrate through tissue, and better targeting to subcellular domains or organelles. It is also likely that newer optrode designs will allow for higher density recordings and efficient illumination of all recording sites with less damage to the brain. Finally, in addition to tagging neurons based on genetic expression patterns, it may eventually be possible to control expression of optogenetic proteins using activity-sensitive promoters for immediate early genes such as c-fos or arc. Using this strategy, neurons that are activated during a certain behavior would express ChR2 and could be re-activated at a later time, potentially re-activating that behavior. These types of experiments may allow researchers to tease apart different functional ensembles of neurons (cell assemblies) that drive a particular behavior. For example, it may be possible to identify a functional ensemble early in a behavioral learning paradigm, and track the activity of that ensemble as the animal learns and improves the behavior. The growing application of optogenetic techniques promises to increase our understanding of how different brain regions function, and how to intervene when they malfunction.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.