Optogenetic probing of functional brain circuitry


  • James J. Mancuso,

    1. Laboratory of Synaptic Circuitry, Program in Neuroscience and Behavioral Disorders, Duke-NUS Graduate Medical School, Singapore
    Search for more papers by this author
  • Jinsook Kim,

    1. Laboratory of Synaptic Circuitry, Program in Neuroscience and Behavioral Disorders, Duke-NUS Graduate Medical School, Singapore
    Search for more papers by this author
  • Soojung Lee,

    1. Laboratory of Synaptic Circuitry, Program in Neuroscience and Behavioral Disorders, Duke-NUS Graduate Medical School, Singapore
    Search for more papers by this author
  • Sachiko Tsuda,

    1. Laboratory of Synaptic Circuitry, Program in Neuroscience and Behavioral Disorders, Duke-NUS Graduate Medical School, Singapore
    Search for more papers by this author
  • Nicholas B. H. Chow,

    1. Laboratory of Synaptic Circuitry, Program in Neuroscience and Behavioral Disorders, Duke-NUS Graduate Medical School, Singapore
    Search for more papers by this author
  • George J. Augustine

    1. Laboratory of Synaptic Circuitry, Program in Neuroscience and Behavioral Disorders, Duke-NUS Graduate Medical School, Singapore
    2. Center for Functional Connectomics, Korea Institute of Science and Technology, 39-1 Hawolgokdong, Seongbukgu, Seoul, Republic of Korea
    Search for more papers by this author

Corresponding author G. J. Augustine: Laboratory of Synaptic Circuitry, Program in Neuroscience and Behavioral Disorders, Duke-NUS Graduate Medical School, 2 Jalan Bukit Merah, Singapore 169547, Singapore. Email: georgea@neuro.duke.edu


Recently developed optogenetic technologies offer the promise of high-speed mapping of brain circuitry. Genetically targeted light-gated channels and pumps, such as channelrhodopsins and halorhodopsin, allow optical control of neuronal activity with high spatial and temporal resolution. Optogenetic probes of neuronal activity, such as Clomeleon and Mermaid, allow light to be used to monitor the activity of a genetically defined population of neurons. Combining these two complementary sets of optogenetic probes will make it possible to perform all-optical circuit mapping. Owing to the improved efficiency and higher speed of data acquisition, this hybrid approach should enable high-throughput mapping of brain circuitry.

To understand how the brain works, it is essential to understand the function of the synaptic circuits that are responsible for all processing and storage of information within the brain. Anatomical studies have provided us with impressive structural insights into the synaptic connectivity of neurons, but provide minimal information about circuit function. At its most fundamental level, circuit function can be defined by stimulating presynaptic neurons while detecting responses in postsynaptic neurons (Fig. 1A). Observation of such coupling formally defines a functional circuit and can illuminate the physiological properties of the circuit, such as whether it is excitatory or inhibitory, fast or slow and strong or weak.

Figure 1.

Optogenetic mapping of neuronal circuits
A, the minimal information required to define a functional circuit is to activate a presynaptic neuron population (green) and observe a response in a postsynaptic neuron population (pink). B, examples of currently available optogenetic effectors that can be used to control the activity of presynaptic neurons (green) and optogenetic sensors that can be used to detect postsynaptic responses (pink). Abbreviations: D3cpV, D3cpVenus; VSFPs, voltage-sensitive fluorescent proteins. C, combining photostimulation (blue arrow) of presynaptic neurons that express optogenetic effectors such as channelrhodopsin-2 (green) with monitoring of postsynaptic neuronal activity (red) via optogenetic sensors such as Clomeleon enables precise mapping of functional connectivity within neuronal circuits.

In principle, electrophysiological techniques are capable of stimulating presynaptic neurons and detecting responses in their postsynaptic partners. However, in practice these tasks are not easily accomplished. With extracellular electrical stimulation it is difficult to define precisely which neurons are being stimulated and the alternative, stimulation via intracellular electrodes, is tedious and low yield. Likewise, while electrophysiological recordings can provide information about neuronal activity, they offer minimal spatial information and typically examine the activity of only one neuron (or at best a few neurons) at a time among the billions of neurons within the brain. Furthermore, the tremendous diversity of neurons within the brain often makes it challenging to identify the neuron whose activity is being recorded. As a result of these limitations, description of the functional wiring of the brain has proceeded at a very slow pace.

Here we describe an alternative approach to mapping the functional circuitry of the brain. This alternative is based on optogenetic probes that harness the combined power of molecular genetics and optical physiology. These probes take advantage of genetic targeting strategies to express light-sensitive proteins in genetically defined populations of neurons, allowing unambiguous identification of the neurons under investigation and thereby neatly side-stepping the challenge of neuronal diversity. By using light-sensitive probes, it is possible to control the activity of entire populations of potential presynaptic neurons and/or monitor the responses of populations of potential postsynaptic neurons (Fig. 1B). The point we would like to emphasize here is that, in principle, it is possible to combine these two optogenetic approaches (Fig. 1C). The relative technical ease of each will give this all-optogenetic combination the potential to provide high-throughput analysis of brain circuit function and thereby radically transform our understanding of neuronal circuits.

An effective all-optical circuit mapping approach requires the following components: (1) optogenetic actuators of neuronal activity that are precisely targeted, sensitive and capable of reproducing normal patterns of neuronal activity; and (2) compatible optogenetic monitors of neuronal activity that are precisely targeted and are sensitive to normal levels of postsynaptic signalling. This review focuses on recent progress towards achieving these two requirements and lays out the technical hurdles remaining before optogenetic circuit breaking can live up to its vast potential.

Optogenetic actuators of neuronal activity

Several microbial opsins, such as channelrhodopsins, halorhodopsin and archaerhodopsins, have recently been established as high-performance optogenetic actuators. Remarkably, expressing these proteins in neurons allows light to either activate or silence neuronal action potential firing. All of these proteins can be genetically targeted to specific populations of neurons and provide a high degree of spatial and temporal control over neuronal activity via light. Thus, these molecules enable exquisitely specific optogenetic control over potential presynaptic neurons.

Channelrhodopsins. The light-gated ion channel, Channelrhodopsin-2 (ChR2), is almost single-handedly responsible for the tremendous surge in interest in optogenetic technologies. Channelrhodopsin-2 comes from the algae Chlamydomonas reinhardtii and is a seven-transmembrane cation channel that opens when activated by blue-green light (∼400–510 nm; Nagel et al. 2003). Expression of ChR2 in neurons allows light to generate large photocurrents (up to a nanoampere or greater) that are sufficient to trigger action potential firing (Boyden et al. 2005; Zhang et al. 2006). The photostimulation of action potentials via ChR2 is temporally precise, with reported values for jitter in action potential timing typically much less than a millisecond. Performance of ChR2 has been improved even further by molecular engineering strategies. For example, mutations permit more rapid gating of ChR2, allowing action potential firing up to 200 Hz (Gunaydin et al. 2010), or on–off switching of neuronal activity via different wavelengths of light (Berndt et al. 2009). Expressing ChR2 in transgenic mice causes no significant changes in neuronal morphology or intrinsic electrical properties (Wang et al. 2007), indicating that ChR2 can be inert even after chronic expression in mammals.

To date, most ChR2-based photostimulation has relied on infection of brain tissue with virus containing ChR2 DNA. For example, such an approach has been used to evaluate the role of cortical interneurons in generating brain gamma rhythms (Cardin et al. 2009; Sohal et al. 2009). More consistent and predictable ChR2 expression is observed in transgenic mice. The first reported transgenic mouse lines expressed ChR2 under the control of the Thy1 promoter, a neuron-specific promoter that preferentially drives expression in projection neurons. The use of these mice allows photostimulation in vivo (Arenkiel et al. 2007) and in brain slices prepared from these mice (Wang et al. 2007). This line is now in use in more than 100 laboratories around the world. Many other ChR2 mouse lines have now been prepared, including transgenics based on bacterial artificial chromosomes that drive expression under the genomic sequence containing the promoter for a gene of interest (Heintz, 2004). Despite potential concerns about the level of expression achievable with bacterial artificial chromosomes, in fact it has been found that such mice express sufficient amounts of ChR2 to allow efficient photostimulation (Kim et al. 2010).

Localized ChR2-based photostimulation permits spatial mapping of brain circuitry. Scanning small laser light spots over brain tissue causes activation of only ChR2-positive neurons within the photostimulation volume, generating synaptic activity only in postsynaptic neurons connected to these cells (Fig. 2A–C). The spatial distribution of synaptic circuits between ChR2-positive neurons and their postsynaptic targets can then be mapped by correlating the location of the light spots with the magnitude of the postsynaptic responses, typically measured by electrophysiological recordings (Fig. 2D). To date, the utility of ChR2-mediated photostimulation for brain circuit mapping has been demonstrated in only a few studies performed with acute brain slices (Wang et al. 2007; Petreanu et al. 2007). The example shown in Fig. 2 illustrates how the use of this approach has revealed the spatial distribution of connections between cortical pyramidal neurons and nearby interneurons. Similar techniques have demonstrated interactions between cortical and thalamic neurons (Cruikshank et al. 2010) and cerebellar Purkinje cells (Kim et al. 2010). These early results indicate that photostimulation via genetically targeted ChR2 is a general technology for selective stimulation of specific populations of presynaptic neurons and mapping functional synaptic circuitry in the mammalian brain.

Figure 2.

Mapping of local excitatory circuits innervating cortical interneurons via Channelrhodopsin-2
A, schematic diagram of the experimental arrangement. Postsynaptic responses are recorded, via a patch pipette (right), from an interneuron (red) during photostimulation (blue arrow) of Channelrhodopsin-2-positive presynaptic pyramidal cells (green). Grey cells indicate neurons that are not involved in the local circuit under study. B, dye-filled interneuron from a cortical slice, with numbered squares indicating locations where laser spot (488 nm) was positioned when the responses indicated in C were evoked. C, excitatory postsynaptic currents detected when light spot (during time indicated by bar) was positioned at the locations indicated in B. D, map of spatial distribution of light-evoked postsynaptic currents (PSC); the magnitude of these currents is indicated by the pseudocolour scale on the right (Wang et al. 2007).

Channelrhodopsin-2-mediated mapping of synaptic circuitry has also been accomplished in vivo. The first such study used illumination through surgically implanted crianial windows to photostimulate axons from neurons originating in the somatosensory cortex and demonstrated selective activation of well-known long-range thalamocortical connections (Petreanu et al. 2007). More recently, a non-invasive transcranial optogenetic stimulation method has been developed based on making the skull bone transparent via treatment with a dental resin (Hira et al. 2009). This technique has been applied to the motor cortex of intact ChR2 transgenic mice and has demonstrated that illumination of pyramidal cells expressing ChR2 produces limb movements. A motor map representing the distribution of neurons associated with movement of the limb is then obtained by correlating the location of photostimulation with the magnitude of limb movements (Hira et al. 2009). This technique allows repeated non-invasive mapping of motor circuits over several weeks, opening the way for analyses of long-term circuit plasticity. Therefore, optogenetic photostimulation of brains expressing ChR2 should prove to be an extremely useful approach for exploring functional connectivity of complex neural circuits and behaviours.

Recent studies demonstrate that a second channelrhodopsin from the algae Volvox carteri (VChR1) can also be used to photostimulate neurons in a rapid and precise manner. The attraction of this alternative channelrhodopsin is its activation by yellow light (∼590 nm). This red-shifted light spectrum for activation of VChR1 makes it possible not only to deliver the light to deeper brain areas, but also to combine it with other optogenetic effectors and sensors activated by different wavelengths of light such, as ChR2 or optogenetic activity indicators activated by blue light (Zhang et al. 2008). However, VChR1 has a couple of drawbacks. First, its kinetics of deactivation at the end of a light pulse are approximately 10-fold slower than ChR2, meaning that it is more useful for generating prolonged photostimuli than repetitive brief photostimuli. Making chimeras between ChR2 and VChR1 provides a way to optimize channel gating properties (Lin et al. 2009; Wang et al. 2009). Second, while VChR1 is sensitive to yellow light, it is rather sensitive to blue light as well. This limits its utility as a partner with ChR2 for two-colour photostimulation experiments.

Halorhodopsin. Light-triggered inhibition of neuronal activity can be obtained by expression of the Natronomonas pharaonis halorhodopsin (NpHR; Han & Boyden, 2007; Zhang et al. 2007). A screen of several related proteins revealed that NpHR provides prolonged inhibitory photocurrents (Zhang et al. 2007). The NpHR inhibits electrical activity by pumping chloride ions into neurons; this anion influx yields an outward current that inhibits neuronal activity. The excitation maximum of NpHR is around 580 nm, which is red-shifted in comparison to the action spectrum of ChR2, allowing simultaneous expression of ChR2 and NpHR to confer bidirectional optogenetic modulation of membrane potential.

Thus far, three generations of NpHR have been developed. While the first-generation NpHR can photoinhibit neuronal firing (Zhang et al. 2007), this NpHR, however, tends to be retained in the endoplasmic reticulum when expressed at high levels in mammalian neurons (Gradinaru et al. 2008; Zhao et al. 2008). Fusing an endoplasmic reticulum export motif onto the C-terminus of NpHR, creating an ‘enhanced’ NpHR, improves membrane targeting of NpHR by reducing endoplasmic reticulum retention. This second-generation tool (eNpHR, eNpHR2.0) has been successfully applied for studies in vivo and in brain slice preparations (Gradinaru et al. 2009; Sohal et al. 2009; Tønnesen et al. 2009). A transgenic mouse expressing this version of NpHR has been developed and allows good photoinhibition, both in brain slices and in vivo (unpublished observations). Furthermore, a third-generation NpHR (eNpHR3.0) offers increased potency of optical inhibition without requiring increased light power (Gradinaru et al. 2010). Addition of an endoplasmic reticulum export motif improves the membrane targeting of this version and increases the amplitude of photocurrents more than 20-fold compared with the first-generation NpHR. Importantly, this enhanced eNpHR3.0 enables optogenetic control with red/far-red light (−680 nm), which allows deeper penetration of light into biological tissues and therefore increases the volume of brain tissue that can be inhibited. Similar properties have been observed for archaerhodopins, which are light-activated proton pumps (Chow et al. 2010).

In summary, channelrhodopsins and halorhodopsins are at an advanced stage of development. These optogenetic actuators provide rapid and spatially defined control of neuronal activity, fulfilling the first requirement for high-throughput optogenetic circuit mapping (Fig. 1B). These two groups of optogenetic actuators provide complementary information for mapping neuronal circuits. Optogenetic activators such as the channelrhodopsins allow us to define neurons whose activation is sufficient to elicit a specific postsynaptic response or behavioural response. Optogenetic inhibitors, such as halorhodopsin, define which neurons are necessary for a particular postsynaptic response or behaviour.

Optogenetic sensors of neuronal activity

All-optogenetic circuit mapping also demands genetically encoded sensors whose optical properties change in response to neuronal activity. Such optogenetic sensors must be capable of reliably detecting the subthreshold changes in membrane potential that typically occur in postsynaptic neurons in response to activation of presynaptic inputs, as well as the larger swings in membrane potential associated with postsynaptic action potentials. To date, no optogenetic sensor completely fulfills these requirements. However, there have been advances in the development and refinement of several different types of optogenetic sensors that promise the ability to image postsynaptic activity. We will highlight these here, acknowledging the omission of other promising approaches, such as optogenetic imaging of synaptic vesicle exocytosis (Miesenbock et al. 1998; Granseth et al. 2006) or released neurotransmitters (Hires et al. 2008).

Ion indicators. Although not direct indicators of neuronal membrane potential, optogenetic sensors of intracellular ion activity, specifically calcium (Ca2+) and chloride (Cl) ions, have proven useful for detection of neuronal activity.

Genetically encoded calcium indicators. As action potential firing is usually associated with opening of voltage-gated Ca2+ channels, organic Ca2+ indicator dyes have long been used to image neuronal activity (Yuste et al. 1992; Ohki et al. 2005; Homma et al. 2009). Inspired by the invention of the calmodulin-based genetically encoded Ca2+ indicator (GECI) Cameleon (Miyawaki et al. 1997), much recent effort has gone into optimizing GECIs for detection of neural activity. The introduction of GCaMP (a single emission wavelength GECI), based on circularly permuted green fluorescent protein and calmodulin, presented a leap forward in affinity and signal-to-noise ratio (SNR; Nakai et al. 2001). The Förster resonance energy transfer (FRET)-based indicator D3cpVenus (D3cpV) has been reported to detect single action potentials in pyramidal neurons from brain slices and in vivo (Wallace et al. 2008). The most recent GCaMP version, GCaMP3, has been shown to outperform other available GECIs in absolute response and SNR in acute cortical slices and in vivo in a variety of invertebrate systems (Tian et al. 2009).

While advances in GECIs have yielded ever-improving means of monitoring action potential firing, caveats will always exist in the use of Ca2+ as a surrogate for neuronal action potentials. Specifically, intracellular Ca2+ concentration can rise in response to cellular signalling processes independent of postsynaptic depolarization, such as release of Ca2+ from intracellular stores, potentially leading to an overestimation of action potential activity. Furthermore, because very little Ca2+ influx occurs in response to subthreshold changes in membrane potential, GECIs do not offer sufficient sensitivity to detect inhibitory or subthreshold excitatory synaptic responses. In summary, GECIs are useful for detecting action potential activity but probably are not the optogenetic sensor of choice for all-optogenetic circuit mapping.

Clomeleon. Synaptic inhibition is an essential part of information processing by neural networks and usually relies on transmembrane fluxes of Cl through GABA and glycine receptors. As a result, the FRET-based Cl indicator Clomeleon (Kuner & Augustine, 2000) offers the possibility of imaging synaptic inhibition (Fig. 3). Clomeleon consists of a yellow fluorescent protein acceptor, whose fluorescence is quenched by Cl and other halides (Wachter & Remington, 1999), linked to a Cl-insensitive cyan fluorescent protein donor. Increases in Cl concentration reduce FRET and thereby change the colour of Clomeleon from yellow to cyan. In comparison to earlier organic dyes used as Cl sensors, Clomeleon offers excitation by visible light (440 nm), safer (genetic) loading procedures and negligible leakage from cells. Clomeleon has been expressed in neurons in transgenic mice without behavioural aberrations, indicating that it is inert.

Figure 3.

Optogenetic imaging of inhibitory circuit activity via Clomeleon
A, schematic diagram of the experimental arrangement. Stimulation of interneurons (grey), via the electrode on the left, causes changes in intracellular Cl concentration in portsynaptic CA1 pyramidal cells expressing Clomeleon (red). B, image of changes in intracellular Cl concentration produced by electrical stimulation (pseudocolour scale) superimposed on yellow fluorescent protein fluorescence image (grey scale) of hippocampal slice prepared from a Clomeleon transgenic mouse (Berglund et al. 2006).

These transgenic mice have made it possible for Clomeleon imaging to reveal the spatiotemporal dynamics of the activity of inhibitory circuits in several brain areas, including hippocampus, cerebellum and deep cerebellar nuclei (Berglund et al. 2006, 2008). Clomeleon was also used successfully to demonstrate local disynaptic inhibitory responses of principal neurons within the basolateral nucleus of amygdala and premotor neurons in the superior colliculus (Berglund et al. 2008). Clomeleon has also been used to image non-synaptic tonic inhibition of cerebellar granule cells (Lee et al. 2010).

However, Clomeleon suffers from a relatively low SNR and narrow dynamic range, largely due to the twin problems of low affinity for Cl and relatively high background levels of intracellular Cl. A modified version offers improved binding affinity (Markova et al. 2008), though its SNR for FRET-based measurements of synaptic inhibition is no better than that of Clomeleon (Grimley et al. unpublished observations). Thus, while Clomeleon is capable of imaging the activity of inhibitory synaptic circuits, its SNR still is not optimal for this purpose. Use of in vitro protein engineering technology has led to the production of many new Clomeleon variants with improved binding affinity and dynamic range; one of these (SuperClomeleon) exhibits a more than fourfold improvement in SNR (Grimley et al. unpublished observations). Another new sensor, ClopHensor, allows simultaneous ratiometric measurement of Cl and pH (Arosio et al. 2010).

Optogenetic voltage sensors. The most direct way to image the activity of synaptic circuits is to use fluorescent probes that detect the transient changes in postsynaptic membrane potential associated with synaptic activity. Towards this goal, several generations of voltage-sensitive dyes have been developed (Zecevic et al. 2003; Kee et al. 2008). While even the best of these suffer from a number of shortcomings, most notably low SNR, they have nonetheless proved useful for imaging the activity of neural circuits and for other applications.

More recent efforts have been directed towards development of genetically encoded fluorescent voltage sensors. These optogenetic probes of membrane potential consist of a voltage-sensing moiety, often part of an ion channel, fused to a fluorescent protein. Early generations of optogenetic voltage sensors have been plagued by low SNR, relatively slow response kinetics and/or poor membrane targeting (Baker et al. 2008). However, very recent developments offer considerable promise for the future of this imaging modality. One promising sensor is the FRET-based Mermaid, consisting of a voltage-sensing domain from a voltage-sensitive phosphatase and a pair of coral fluorescent proteins, mUKG and mKO (Tsutsui et al. 2008). This sensor is capable of resolving action potentials at a frequency of 50 Hz in mouse neuroblastoma cells, but its SNR still is insufficient to detect individual postsynaptic potentials in brain slices or in vivo. Still other promising variants based on the voltage-sensitive phosphatase, the voltage-sensitive fluorescent proteins (VSFPs), are described in the article by Mutoh et al. (2011) in this issue of Experimental Physiology. Finally, because all voltage sensors necessarily add to the transmembrane capacitive load, high levels of expression can perturb neuronal function (Chanda et al. 2005; Sjulson & Miesenbock, 2007; Akemann et al. 2009). This places an upper limit on the amount of improvement in SNR that can come from increasing VSFP concentration in the membrane.


The optogenetics revolution has provided us with tools that can either control or monitor neuronal activity on a large scale. Combining these two capabilities promises a powerful new means of studying neuronal circuits with high speed and precision. However, hurdles remain for full implementation of this approach.

Development and subsequent improvement of a multitude of optogenetic actuators, such as Channelrhodopsin-2 and halorhodopsin, have allowed for precise and bidirectional optical control of single-cell activity with high temporal resolution (Boyden et al. 2005). These optogenetic actuators are effective for implementation in all-optical circuit mapping in a number of neuronal subtypes. The continued search for novel light-activated actuators (Zhang et al. 2008) and genetic alterations to existing actuators (Gradinaru et al. 2010) may help to create channels with sufficient spectral separation to allow effective simultaneous use with either other actuators or optogenetic monitors.

While the development of optogenetic monitors, such as Clomeleon, Mermaid, VSFPs and the GCaMPs (Kuner & Augustine, 2000; Tsutsui et al. 2008; Tian et al. 2009; Mutoh et al. 2011), has also been impressive, it clearly remains the rate-limiting step in the effective implementation of all-optogenetic circuit mapping. For the most well-developed ion indicators, the most straightforward method for continued improvement in affinity, kinetics and SNR remains incremental changes in the structure of their ion sensing regions through site-directed mutagenesis. Such improvements will continue to provide ever-better tools for optogenetic circuit mapping and other brain imaging applications.

Difficulties in the use of VSFPs for circuit mapping in slices and in vivo are more problematic, as described in Mutoh et al. (2011), and therefore require more aggressive approaches to improvement. The in vitro protein engineering strategy employed by Grimley and co-workers (unpublished observations) offers a very promising means of quickly modifying protein structure and screening the resultant changes in sensor performance.

All-optogenetic circuit mapping will require the combination of optogenetic actuators and sensors with compatible spectral properties. At present, the most useful actuator (ChR2) and sensors (GCaMpS) have overlapping excitation spectra, which makes their simultaneous use problemeatic. Further refinement of optogenetic probes will need to take into account such issues of spectral compatibility. Finally, it is important to acknowledge that the use of light for control and monitoring of neuronal activity in the intact brain is limited to a depth of a few hundred micrometres because of light scattering. Further development of non-invasive methods that accurately excite actuators and monitors with better depth penetration, such as multiphoton excitation with adaptive optics, will aid application of all-optogenetic circuit mapping to the intact brain. Such improvements will greatly increase our ability to investigate functional neural circuits underlying complex behaviour and disease in ways that were unimaginable only a short time ago.