OPTOGENETICS : Past, Present and Future
Optogenetics refers to a range of techniques that control cellular activity through genetically delivery of light-sensitive proteins. The ability to precisely modulate the activity of a defined population of cells, with little to no invasion or interference of other systems has been a revelation in deconstructing the function of individual neurons and the neural circuits that underpin behaviour alike, and has enabled mammoth leaps in our neuroscientific understanding. Given the central role optogenetic techniques in modern neuroscience, it is often difficult to comprehend that it is a trailblazing field of research is a little more than 10 years old. In this essay, I will review the history, current state and future of optogenetic techniques and applications.
The Development of Optogenetics – Development and Optimisation
The recognition of light as an ideal tool in the investigation of neuronal activity is largely given to Francis Crick. Crick suggested, ‘that the major challenge facing neuroscience was the need to control one type of cell in the brain while leaving others unaltered.’ Introduction of electrodes into neuronal tissue inevitably caused damage, and could not be targeted to precisely defined populations of cells. Pharmacological techniques were, on a neuronal scale, too slow and often not specific enough in their action. If neurons could be appropriately sensitised, Crick postulated that ‘to be able to turn the firing of one or more types of neuron on or off in the alert animal in a rapid manner, the ideal signal would be light.’
Early Light-Sensitising Techniques
As mentioned, controlling neuronal activity via light signals, has long been an appetising thought for neuroscientists. In the early 1980s, Farber and Grinvlad reported stimulation of neurons through insertion of molecular probes that could form transient channels in neuronal membranes in the presence of light. Whilst this did depolarising the membrane, pore generation also severely damaged neurons. Thus, it saw little practical use as neurons could only be stimulated a few times before succumbing to phototoxicity.
The early 1990s saw further refinements in neuronal photosensitivity by the work of Callaway and Katz. Their technique, Scanning Laser Photostimulation (SLP) involved bathing neurons in a solution containing a caged form of glutamate, before liberating the neurotransmitter in a spatially specific fashion by irradiating only restricted areas of the slice (50-100µm at a time). This technique, when paired with patch clamp recordings from a neuron of interest, could be used to identify the locations of neurons making functional synaptic connections to the neuron of interest, with, as Katz himself put it, ’superb’ spatial resolution. Whilst the method showed some promise and was utilised to successfully map some neuronal circuits (see Dantzker and Callaway, 2000) it carries several limitation. Firstly, the experimenter had little ability to control neuronal spiking – action potential generation, and the reversal of the effects of the uncaged transmitter was largely determined passively through diffusion kinetics and neuronal pharmacokinetics. Second, whilst the connectivity of a few, individual neurons, could be mapped, the technique was simply too inefficient to map most, larger neuronal networks, and it could not interrogate neuronal function in vivo. Lastly, photostimulation was not specific to cells, but to illuminated areas. It was not possible therefore to elucidate the behaviour of functionally distinct, but anatomically related, populations of neurons.
Metabotropic Light Sensitive Proteins
Undoubtedly, a breakthrough in optogenetic development occurred in 2002 with the first demonstration of sensitisation of vertebrate neurons to light. Zemelmann and Miesenböck transfected and co-expressed 3 drosophila photoreceptor genes in rat hippocampal neurons to form a powerful metaborhodopsin (i.e. a GPCR connected to retinal) they termed ChARGe. In the presence of light on, the ChARGed neurons would fire. Significantly, this firing was not sensitive to AMPA antagonist AP5 or the NMDA antagonist CV2X, demonstrating that activation of a transfected rhodopsin was sufficient to generate neuronal spiking. Importantly, chARGe did not act indiscriminately thought the illuminated area, rather it localised the responsiveness to light exclusively to the cells that it was expressed in, a character that is particularly valuable for deducing neural circuitry. However, ChARGEd neuronal firing in response to light was not well timed or controlled. Activation of the metabotropic receptor, and it’s secondary messenger systems, may have allowed for powerful depolarisation, but it failed provide much meaningful temporal precision or control of neuronal firing.
Channels responsive to Photolytic Neurotransmitter
The next few years fostered promising new developments as several groups worked to try and improve the temporal precision of light-sensitising proteins. Zemelmann and Miesenböck returned just a year later in 2003 and published results on a refined technique to improve on the precision of light-dependent firing.The team expressed puringergic, P2X2, ligand-gated ion channels in hippocampal neurons, and activated them via photolysis of a caged form of the agonist, ATP. When exposed to cognate agonist, the purinergic receptors opened, the membrane depolarised, and the neurons fired, peaking at a firing frequency of ~40 Hz. Spikes initiated and terminated rapidly, and did not attenuate. Current amplitudes and firing frequencies could be tuned by varying the concentration of ligand, allowing for a level of control over firing rate, but dependence on transmitter diffusion and metabolism limited the effectiveness of this approach. The combination of powerful peak currents, and absence of P2X receptors in drosophila has resulted in versions of this technique having some utility in drosophila models.
Artificial Light Sensitive Channels
Clearly, expression of ion channels possessed the favourable kinetics for depolarisation, but lacked light sensitivity. In an attempt to unite the favourable kinetics of ion channels with light sensitivity, Isacoff and Kramer engineered artificial light sensitive ion channels. The team engineered their first channel, a Shaker Potassium channel (termed SPARK) to contain a light sensitive, pore blocking moiety. Long wavelength light would drive the moiety to it’s extended trans configuration, allowing the pore to be fully blocked, while short wavelength light caused the moiety to adopt it’s shorter cis configuration, retracting the blocker and allowing conduction. Thus different wavelengths of light could modulate action potential firing in a rapid, precise and reversible manner. In 2006, another channel had been engineered, engineering an inotropic Glutamate channel to be covalently to a glutamate molecule via a photosensitive tether. (LiGluR) Light activation would cause a conformational shift in the tether, causing Glutamate to bind to it’s receptor, opening the channel. These artificial channels have had some success in ex vivo and in brain slice experiments, but such finely engineered structures proved difficult to design and optimise, as well as to express in vivo. Perhaps more pressing to their relative downfall however, was the development of light sensitive, microbial single component ion channels.
Single Component, Microbial Ion Channels
By far the technique that has gained the most traction in the literature is that of using Single Component, Microbial Ion Channels. In 2002, Sineshchekov et al. identified that two rhodopsins – ChR1 and ChR2 – in C. Reinhardtii acted as low and high light intensity phototaxis receptors and could generate electrical currents. A year later, Nagel et al. performed electrophysiological analysis of Xenopus Oocytes transfected with ChR2. Their patch clamp recordings confirmed that ChR2 was a directly light sensitive, cation-selective ion channel.
One of the most seminal papers in the field of optogenetics was published in 2005, when Karl Deisseroth’s lab demonstrated for the first time that microbial light sensitive ion channels can be easily expressed in mammals, and use them to manipulate brain function. A lentiviral gene delivery system transfected ChR2 into cultured mouse hippocampal neurons, and subsequent electrophysiological analysis indicated that the neurons could reliably spike (with millisecond timescale control) in response to light stimuli. Finally, the prayers of decades of neuroscientists had been answered – a technique that allowed for temporally precise, and non-invasive control of activity of genetically defined neuronal populations! Well, not exactly.
Expanding the Toolkit: Optimisation of Light-sensitive Channels
The discovery of ChR2 as a light sensitive cation channel was undoubtedly a major step forward for neuroscience research. That being said, the protein evolved to be optimised for bacterial motility, not for experimental ease. It’s natural biophysical properties are not particularly well adapted for neuroscientific investigation – it generates small photocurrents, and the kinetics of ChR2 slow down considerably at depolarised membrane potentials. In addition, when expressed at high levels, extra spikes (e.g. doublets) were shown to occur in response to a single light pulse, infringing the coding fidelity of some optogenetic signals.
Even with limited structural models, ChR variants were engineered with more favourable biophysical properties. So called bi-stable ChRs that could be gated into the active state by a single pulse of light were developed by introducing targeted substitutions at C128T or C128S (Berndt et al 2009). These channels were also shown to undergo a precise step termination in the presence of red light. In short, these modified ChR2 channels act as light sensitive, binary switches. Turning a neuron ‘on’ after a brief pulse of light, and ceasing activity after a pulse of red light. An E123T substitution was shown to produce faster gating kinetics, increased photocurrent magnitude and reduced sensitisation (Gunaydin LA, et al. 2010). Combining this mutation with a threonine-to-cysteine substitution at position 159 resulted in drastically increased precision of single action potential induction over a broad range of frequencies. (Berndt et al 2011).
Substantial improvements were made to the functioning of ChR2 with only minimal knowledge of it’s structure. By 2012, X-ray crystallographic techniques had deduced the ChR2 crystal structure to a near atomic resolution of 2.3Å (Kato et al 2012). Such a high resolution 3D model of an ion channel’s structure is scarce, and has proved to be of enormous value in guiding the more complex design of novel ChR variants. For example in 2015, Wietek et al introduced no fewer than 5 targeted mutations to the structure of ChR2, typically substituting key negatively charged resides with more positively charged alternatives at the gate and in the pore of the channel. The resulting channel, termed iChloC (improved Chloride Conducting ChR) no longer shows depolarizing activity in patch-clamped neurons and reliably inhibits synaptically evoked spikes in undisturbed CA1 pyramidal cells. The ability to develop channels with unique properties is a valuable tool for researches, and has significantly broadened the horizons of experimental design.
Discovery and Development of Other Light-sensitive Ion Channels
ChR is not alone in being the only bacterial light sensitive ion channel that has been implanted in optogenetic techniques. Indeed, in 1971, Stoeckenius and Oesterhelt discovered that a protein in the purple membrane of Halobacteria, bacteriorhodopsin, acted as an ion pump that can be rapidly activated by visible-light photons.
Perhaps a more experimentally relevant protein discovery would be that of Halorhodopsins. Halorhodopsins were discovered in 1977 by Matsuno-Yagi and Mukohat. In the mid 2000s, their electrophysical properties were investigated and they were shown to possess several distinct biophysical properties from ChR. Whereas ChR was activated mainly by blue light, Halorhodopsins where activated by more yellow light. In addition, whereas ChR was cation permeable, halorhodpsin is primarily permeable to Chloride ions, meaning it’s activation will hyper-polarise the cell and result in neuronal silencing. More families of microbial ion channels continue to be discovered, and their unique and useful properties exploited. In 2015, Govorunova et al described the discovery of two channels, GtACR1 and GtACR2 (Guillardia theta anion channel rhodopsins 1 and 2), which generated stronger hyper polarising currents, was more selectively permeable to chloride and possessed a more negative reversal potential than currently available optogenetic proteins. The discovery of a diverse range of ion channels, each of which can have their biophysical properties modified further, has been central in designing and generating the fantastic application of optogenetic techniques.
Genetic Delivery Strategies
The expression of light sensitive channels in defined populations of neurons is central to the efficacy of optogenetic techniques. Thankfully, researchers are presented with several routes to introduce their desired protein into their desired population of cells.
In-Utero Electroporation
In 2001, Santo and Nakatsuji developed a surgical method of gene delivery – in-utero electroporation (IUE). In-utero surgery was performed in the developing mouse foetus, and the gene of interest injected into the developing ventricle of the embryo. Application of a small current across the developing skull electroplates the charged DNA into the cells lining the ventricle. As inner cortical layers develop before outer layers, and thus this method allows for expression of proteins early in development and in relation to developmental stage.
IUE is useful has been particularly useful in developmental circuit mapping experiments. For example, Petreanu et al 2007 used IUE to deliver ChR2-Venus together with mCherr to layer 2/3 pyramidal neurons in mouse somatosensory cortex. Brains were then surgically extracted and sliced. As ChR2 expressing axons can be stimulated even if they are severed from their parent somata, even in brain slices where circuits may have been severed, the group were able to investigate the local and long range cortical projections of L2/3 pyramidal neurons. The group confirmed L2/3 neurons connect to pyramidal cells in L2, L3, L5 and a subset of L6 cells in their home column of their ipsilateral hemisphere, but also notably found that L2/3 axons contact the same subset of excitatory neurons in the contralateral hemisphere. IUE allows for reliable spatial and temporal discrimination of neural gene expression in development, and has shown to be particularly useful if a DNA promoter region is not known for a developmental population.
Viral Vector Techniques
Living in a post-genomic era, many of the mysteries of our DNA sequence have became unravelled, and more and more DNA promoter regions have been discovered. This is a fact that has been exploited by the neuroscience community through viral vector development.To target specific neuronal populations in vivo strategies are required that can elegantly and decisively discriminate between subtlety different neuronal populations. An elegant solution to this problem is to package the opsin in a viral vector. The gene can be tagged with a fluorescent marker, and packaged the downstream of a promoter specific to a neural population. The virus is then introduced to the subject, and infects essentially all cells. However, only those cells expressing transcription factors that up regulate the promoter region will express the opsin. As viruses tend to introduce multiple gene copies into each target cell, they are capable of mediating high levels of opsin gene expression, overcoming the low transcriptional activity of some cell-specific promoters and allowing experimentally meaningful expression of the opsin.
Cre-Log Approaches
Some promoters are, unfortunately, too large to insert into a viral vector. Typically a binary method of gene expression will be used instead, typically a Cre-Log approach. A viral vector containing an antisense version of the gene, will be injected into a transgenic mouse line. The mice will have been developed to express the necessary Cre-recombinase to extract and reorientate the gene in the correct order so expression can occur. The use of an specific Cre-recombinase dependent protein transcription ensures that expression leak in non-targeted cells does not occur, and thus protein expression is specifically defined to a subset of cells.
A fairly early example of utilising a binary method of expression in optogenetics comes from Tsai et al in 2009. Using a Cre-inducible adeno-associated virus (AAV) vector with a double-floxed inverted open reading frame (ORF), containing an antisense ChR2-EYFP sequence nearly all neuronal cells were infected. Only a specific subset of Cre-expressing TH cells were able to irreversibly invert the ChR2-EYFP ORF, and thus activate sustained ChR2-EYFP expression under the strong, constitutively active elongation factor 1α (EF-1α) promoter. This resulted in expression of ChR2 exclusively in Dopaminergic neurons of the Ventral Tegmental Area of the midbrain, allowing the group to start investigating the roles of these neurons specifically in behavioural conditioning.
Present Optogenetics – Applications
The temporal precision, cell type specificity, and minimal invasiveness of optogenetic techniques mean they are well suited as an investigative tool in real time behavioural studies. A potent example of this is a 2013 study from Chaudhury et al. The group utilised the temporal precision and cell-type and projection-pathway specificity of optogenetics to show, in mouse, phasic activation of VTA neurons projecting to the nucleus accumbens (NAc), but not to the medial prefrontal cortex (mPFC), induced susceptibility to social-defeat stress. Conversely, optogenetic inhibition of the VTA–NAc projection induced resilience, whereas inhibition of the VTA–mPFC projection promoted susceptibility to social defeat stress. The neural-circuit-specific activation of the neurons with optogenetics, has thus helped untangle the neural mechanisms that underpin depression, as well as many other neural phenomena.
In addition to writing neural activity, optogenetic techniques have been utilised to indirectly report emotional states in rodents. Animals cannot verbally report ratings of reward, and as such, researchers have often grounded their ‘liking’ assays on the subjective monitoring of orofacial expressions. Understandably, this reliance on subjective states limits the usefulness of such assays, and weakens the conclusions drawn from behavioural that rely on them. Domingos et al 2011, developed an optogenetic assay for quantification of the reward value of nutrients in the context of different metabolic states and leptin levels. Domingos measured the mouse’s preference for a nutrient versus a reference stimulus in which ingestion of water (or solutions of sweeteners) induces optogenetic stimulation of DA neurons. The reliable precision of optogenetic induced firing allowed for quantitative analysis in changes in preference relative to a reference stimulus reflect changes in the reward value of the ingested agent, giving a more objective rating scale for ‘liking’ as opposed to the that predominated in the field prior. Domingo has used quantitive analysis of mice behaviour in this assay to show that food restriction increased the ‘value’ of sucrose relative to sweeteners plus optogenetic stimulation, and that leptin decreased the ‘value’ of sucrose. This suggest that leptin suppresses the ability of sucrose to drive taste-independent DA neuronal activation, and could provide insights into increasing the effectiveness of dieting in human patients. Optogenetic techniques do not just harbour the potential to conduct more exotic behavioural experiments in rodents; through dissection and control of precise neural circuitry, they allow for a new level of control and objectivity in vivo and with it, the potential to draw firmer conclusions and a deeper understanding of neural circuitry and behaviour.
The Future of Optogenetics – Potential Developments and Applications
Optogenetics has proven to be a powerful tool for investigating neuronal activity. That being said, it has it’s shortcomings and limitations and as such, a number of groups have their goals set on developing techniques to expand and improve the ability of neuroscientists to understand neuronal activity.
Reading and Writing Neuronal Activity in Real Time
The ability to read and write neural activity in real time in a behaving animal, and measuring it’s effects has long been very desirable for neuroscientists. Such experiments would permit a greater understanding of neural circuitry, neural coding and a more physiological, causal diagnosis of mental illnesses that are currently functionally defined. More complex applications of optogenetics techniques are being developed to allow for such an experiment.
In 2018, Mardinly et al engineered a set of soma-targeted optogenetic channels, ST-ChroME and IRES-ST-eGtACR1 and engineered/optimised them for multiphoton stimulation using computer-generated holography (CGH). The team were able to, in a 3D pattern, simultaneously photostimulate up to 50 neurons at once to express complex neural activity patterns. It is likely only a matter of time until this technique is utilised to design more diverse behavioural experiments, and gain insight into the principles of neuronal coding.
DREADDs
Whilst light as a stimulus is generally very favourable, it’s delivery to the CNS can be problematic. It can often be difficult to illuminate widely-distributed neuronal populations, and to do so is often invasive. In addition, optical fibres (particularly when activated over longer periods of time) can generate a lot of heat, further contributing to potential tissue damage. An attractive alternative technique to selectively activate a precise population of neurons particularly is comes from use of chemogenetic DREADDs. (Designer Receptors Exclusively Activated by Designer Drugs). Bryan Roth’s lab were the first to generate transgenic mice expressing a mutated GPCR. The receptor has lost affinity to it’s natural cognate ligand, but has high affinity and efficacy for pharmacologically inert, orally bioavailable drug. Due to it’s ease of use and suitability for both acute and chronic regulation, DREADD techniques have been well implanted in the study of feeding behaviours. (e.g. Krashes et al 2011)
Magnetogenetics
Again, another method that aims to improve upon the invasiveness of optogenetics is magneto genetics – a process of sensitising neurons to, and subsequently controlling neurons with magnetic stimuli. Such an approach requires no surgery or tethering to an energy source, as magnetic fields can pass freely through organic tissue.
The existence of naturally occurring proteins that have evolved to be sensitive to magnetic fields is unclear, and their successful implementation into neurons is yet to be successfully reported. Engineering proteins under the thinking of magneto-thermo-genetics (MTG) has so far shown the most promise. MTG relies on the principle of thermal relaxation, whereby small magnetic nanoparticles are heated up by an alternating magnetic field, such as a radiofrequency field. Stanley and colleagues have successfully controlled some neuronal activity through ectopic expression chimeric ferritin tethered to TRPV1 via a GFP nanobody. In 2016, the group delivered this protein complex to glucose sensing neurons in mouse ventromedial hypothalamus with an adenovirus vector. They reported that, in a manor similar to optogenetic activation, the application of radio-frequency fields increased blood glucose levels dramatically, and that the degree of this physiological response correlated fluorescence studies of TRPV1-GFP nano body expression.
Magnetogenetics does suffer from some key deficiencies compared to it’s light-dependent cousin. Firstly, the mechanism of many magnetogenetic techniques remains cloudy, making meaningful optimisation of these proteins difficult. Secondly, whereas the infrastructure required for optogenetics experiments is fairly minimal, magnetogenetic techniques require powerful magnetic fields, which are not easy to produce and often risk introducing artefacts. Lastly, the current kinetics of action potential generation from magnetically sensitised cells is very slow – whilst optogenetic techniques produce reliably produce firing after milliseconds of light exposure, MTG requires seconds to induce action potentials. Work continues to improve magnetogenetic improvements, and to dethrone optogenetics.
Conclusions
It is difficult to understate the significance of optogenetics as a tool in modern neuroscience. Controlling the activity of a defined population of neurons, has allowed for the dissection of neural circuitry at an unprecedented rate. From humble microbial beginnings, single ion channels have been engineered and modified to create an experimental toolbox of impressive breadth. Optogenetic techniques will continue to be applied in novel fashions to elucidate more complex neural function, and have inspired new techniques (e.g. Magnetogenetics, DREADDs) that aim to improve on optogenetics’ invasive limitations.