Final Report Summary - COMP-OPTOGEN (Computational optogenetics for the characterization and control of cortical activity)
Since the recent introduction of optogenetics into neuroscience, the ability to optically control populations of neurons through the insertion of light-gated ion channels (‘opsins’) allows the perturbation of neural activity with a high spatial and temporal precision that is currently unparalleled by other available methods. These traits, together with the specificity as to which type of neuron and the spatial location of distinct populations can be targeted, have made optogenetics an attractive tool for neuroscientists.
As new opsins are created, the possibilities for artificially controlling neural activity are extended. For example, an important innovation within optogenetics was the creation of both excitatory and inhibitory neurons, allowing the activation and silencing respectively of neurons. Even more fundamentally, new opsins can have activation kinetics that differ greatly to previously studied opsins. As different classes of neuron themselves have varying kinetics, the interaction between opsin and neuron dynamics can be unpredictable. The combinatorics of potential interactions further increases as multiple opsins, such as excitatory and inhibitory variants, may be utilized within the same neuronal population. The range of opsins that are available increases yearly, yet tools to effectively evaluate which opsins should be matched to specific neuron populations and, just as importantly, how they should be optically driven, are currently lacking.
This project aimed to move optogenetics from a gross tool for manipulating neuronal activity, by improving the functional application of opsins to their target population by examining the interaction of opsins, both with each other within a neuron and also within the network, in order to propose more effective stimulation protocols. This was done by identifying three main objectives:
(i) Identification of matching opsin characteristics with neuron dynamics, to identify the degree to how opsins can be chosen such that their kinetics were best matched to the intrinsic characteristics of specific neuronal classes;
(ii) Characterisation of co-expression and co-activation of opsins in neuronal populations, to investigate the inclusion of both excitatory and inhibitory opsins within a single neuron, and how their activation could be optimally controlled;
and (iii) Defining stimulation protocols for 'loose' control of neural activity, to determine methods to provide naturalistic control, rather than overwriting neuronal output as is typical in optogenetic protocols.
This project combined both experimental and computational neuroscience, with significant feedback and cooperation between these two streams in order to tackle the three objectives of this project. By including both models and electrophysiology within my project, I could tune my computational model to reflect experimental reality to refine my predictions. Similarly, the close relationship between experiment and model allowed me to evaluate various experimental scenarios to identify worthwhile experimental protocols and expected results.
In the first approach, I developed computational models of neurons and opsins, based on existing, available models in addition to preliminary experimental results. Opsin models were created as ion channels, allowing theme to be combined with multi-compartmental models of single neurons. This allowed combinations of opsin and neuron class to be evaluated in an efficient manner. I provided the capability to drive neurons via a range of methods, devising protocols those that are identical to those used experimentally as well as naturalistic inputs that appear under in situ conditions. These model options provided me with the tools needed to probe my model to identify the likely experimental interactions between opsin and neuron as per each of the three objectives, and matched to each of the experimental conditions required. Critically, my simulations allowed me to identify which stimulation protocols and neuron-opsin combinations were expected to result in the largest differences, thereby allowing me to refine my experimental design to focus on interactions of interest between opsins and neurons. A key feature of these models were the inclusion of temporal kinectics which are opsin-specific and whose importance are promoted when considering the interaction of excitation to inhibition over small timescales. My model also allowed me to consider scenarios that are not easily manipulated experimentally, such as the variance of expression patterns and light scatter.
The results from my computational model fed directly into the second stream, in which I developed and performed in vitro experiments. Based on the predictions generated from my models, I chose to investigate the interaction of a dual opsin construct within excitatory and inhibitory neurons in cortical areas. I aimed to verify these results for dual opsins by using a construct that contains both excitatory and inhibitory opsins (ChR2-2A-Halo) that allows for stoichiometric expression by providing a fixed ratio for the two opsins, channelrhodopsin-2 (ChR2) and halorhodopsin (Halo). This required the viral packaging of the opsin construct, injecting viral loads to target sites, slicing and performing whole-cell patch clamping to obtain intracellular recordings. Lastly, this required devising methods to match the optimal activation protocol that were identified to be most effective from the predictions of my model to obtain ‘loose’ control over a neuronal population.
From my computational models and experimental results, I identified three novel findings: firstly, that there was a clear relationship between how the opsin interacted with types of neurons in different ways, based on the internal dynamics of each neuron (Objective I).
Secondly, that using dual activation it was possible to achieve gain modulation, whereby the mechanism by which an individual neuron can alter the relationship between the input it receives and it's output (Objective 2). The ability to modulate gain is central for healthy operation and its dysfunction has been proposed to be a basis of multiple neural disorders (such as epilepsy, schizophrenia, autism). However, this was highly dependent on the neuronal morphology. This finding thus suggests that some neuronal populations within the brain are more likely to be responsible for regulating overall activity due to their capacity for gain modulation. The impact for this result extends beyond this project, as if it is indeed able to be verified experimentally, it indicates which neuronal populations are the optimal targets for neurotherapeutics.
Lastly, I established protocols that were able to alter the firing rates of populations of neurons, without disturbing the firing characteristics of neurons. This property allows for ‘loose control’, whereby an external agent can increase or decrease neural activity of populations by choosing the insertion or deletion of spike times in a naturalistic manner (Objective III).
In addition, my computational models also have allowed me to evaluate the experimental limitation of an effect known as 'partial illumination', which occurs when the activating light does not illuminate the entire neuron due to light scattering effects. This has been partly quantified previously for the simple case of having distinct areas illuminated, but has not included the illumination gradient that occurs experimentally. I was able to establish not only how far light had to penetrate but also how quickly it had to taper for the effect to not alter optogenetic activation, and found that even a moderate amount of partial illumination will interfere with opsin activation.
The impact of this project extends threefold. Firstly, it directly benefits optogenetics through the refinement of illumination protocols and the matching of opsin to neuron class. In this respect, it demonstrates that optogenetics can be used as a finely tempered tool with which to manipulate neural activity.
Furthermore, I utilised opsins to explore a fundamental neuroscientific question: namely, how do competing excitatory and inhibitory inputs act in concert to alter the relation between a neuron’s input and output at the level of an individual neuron? This was done using excitatory and inhibitory opsins to mimic the corresponding input that a neuron would receive from populations of neurons in situ. Using this approximation, I was able to establish that the relation between input and output within a neuron is not only dependent on the interaction of excitation and inhibition input but also on the dendritic morphology of a neuron.
Finally, as this project combined a computational tool for guiding optogenetic design, together with experimental validation, it demonstrates a viable possible alternative method to reduce the ethical footprint of biological models. It also provided a useful tool with which to evaluate whether experimental constraints adversely impacted experiments, and how best to compensate for these effects.
As new opsins are created, the possibilities for artificially controlling neural activity are extended. For example, an important innovation within optogenetics was the creation of both excitatory and inhibitory neurons, allowing the activation and silencing respectively of neurons. Even more fundamentally, new opsins can have activation kinetics that differ greatly to previously studied opsins. As different classes of neuron themselves have varying kinetics, the interaction between opsin and neuron dynamics can be unpredictable. The combinatorics of potential interactions further increases as multiple opsins, such as excitatory and inhibitory variants, may be utilized within the same neuronal population. The range of opsins that are available increases yearly, yet tools to effectively evaluate which opsins should be matched to specific neuron populations and, just as importantly, how they should be optically driven, are currently lacking.
This project aimed to move optogenetics from a gross tool for manipulating neuronal activity, by improving the functional application of opsins to their target population by examining the interaction of opsins, both with each other within a neuron and also within the network, in order to propose more effective stimulation protocols. This was done by identifying three main objectives:
(i) Identification of matching opsin characteristics with neuron dynamics, to identify the degree to how opsins can be chosen such that their kinetics were best matched to the intrinsic characteristics of specific neuronal classes;
(ii) Characterisation of co-expression and co-activation of opsins in neuronal populations, to investigate the inclusion of both excitatory and inhibitory opsins within a single neuron, and how their activation could be optimally controlled;
and (iii) Defining stimulation protocols for 'loose' control of neural activity, to determine methods to provide naturalistic control, rather than overwriting neuronal output as is typical in optogenetic protocols.
This project combined both experimental and computational neuroscience, with significant feedback and cooperation between these two streams in order to tackle the three objectives of this project. By including both models and electrophysiology within my project, I could tune my computational model to reflect experimental reality to refine my predictions. Similarly, the close relationship between experiment and model allowed me to evaluate various experimental scenarios to identify worthwhile experimental protocols and expected results.
In the first approach, I developed computational models of neurons and opsins, based on existing, available models in addition to preliminary experimental results. Opsin models were created as ion channels, allowing theme to be combined with multi-compartmental models of single neurons. This allowed combinations of opsin and neuron class to be evaluated in an efficient manner. I provided the capability to drive neurons via a range of methods, devising protocols those that are identical to those used experimentally as well as naturalistic inputs that appear under in situ conditions. These model options provided me with the tools needed to probe my model to identify the likely experimental interactions between opsin and neuron as per each of the three objectives, and matched to each of the experimental conditions required. Critically, my simulations allowed me to identify which stimulation protocols and neuron-opsin combinations were expected to result in the largest differences, thereby allowing me to refine my experimental design to focus on interactions of interest between opsins and neurons. A key feature of these models were the inclusion of temporal kinectics which are opsin-specific and whose importance are promoted when considering the interaction of excitation to inhibition over small timescales. My model also allowed me to consider scenarios that are not easily manipulated experimentally, such as the variance of expression patterns and light scatter.
The results from my computational model fed directly into the second stream, in which I developed and performed in vitro experiments. Based on the predictions generated from my models, I chose to investigate the interaction of a dual opsin construct within excitatory and inhibitory neurons in cortical areas. I aimed to verify these results for dual opsins by using a construct that contains both excitatory and inhibitory opsins (ChR2-2A-Halo) that allows for stoichiometric expression by providing a fixed ratio for the two opsins, channelrhodopsin-2 (ChR2) and halorhodopsin (Halo). This required the viral packaging of the opsin construct, injecting viral loads to target sites, slicing and performing whole-cell patch clamping to obtain intracellular recordings. Lastly, this required devising methods to match the optimal activation protocol that were identified to be most effective from the predictions of my model to obtain ‘loose’ control over a neuronal population.
From my computational models and experimental results, I identified three novel findings: firstly, that there was a clear relationship between how the opsin interacted with types of neurons in different ways, based on the internal dynamics of each neuron (Objective I).
Secondly, that using dual activation it was possible to achieve gain modulation, whereby the mechanism by which an individual neuron can alter the relationship between the input it receives and it's output (Objective 2). The ability to modulate gain is central for healthy operation and its dysfunction has been proposed to be a basis of multiple neural disorders (such as epilepsy, schizophrenia, autism). However, this was highly dependent on the neuronal morphology. This finding thus suggests that some neuronal populations within the brain are more likely to be responsible for regulating overall activity due to their capacity for gain modulation. The impact for this result extends beyond this project, as if it is indeed able to be verified experimentally, it indicates which neuronal populations are the optimal targets for neurotherapeutics.
Lastly, I established protocols that were able to alter the firing rates of populations of neurons, without disturbing the firing characteristics of neurons. This property allows for ‘loose control’, whereby an external agent can increase or decrease neural activity of populations by choosing the insertion or deletion of spike times in a naturalistic manner (Objective III).
In addition, my computational models also have allowed me to evaluate the experimental limitation of an effect known as 'partial illumination', which occurs when the activating light does not illuminate the entire neuron due to light scattering effects. This has been partly quantified previously for the simple case of having distinct areas illuminated, but has not included the illumination gradient that occurs experimentally. I was able to establish not only how far light had to penetrate but also how quickly it had to taper for the effect to not alter optogenetic activation, and found that even a moderate amount of partial illumination will interfere with opsin activation.
The impact of this project extends threefold. Firstly, it directly benefits optogenetics through the refinement of illumination protocols and the matching of opsin to neuron class. In this respect, it demonstrates that optogenetics can be used as a finely tempered tool with which to manipulate neural activity.
Furthermore, I utilised opsins to explore a fundamental neuroscientific question: namely, how do competing excitatory and inhibitory inputs act in concert to alter the relation between a neuron’s input and output at the level of an individual neuron? This was done using excitatory and inhibitory opsins to mimic the corresponding input that a neuron would receive from populations of neurons in situ. Using this approximation, I was able to establish that the relation between input and output within a neuron is not only dependent on the interaction of excitation and inhibition input but also on the dendritic morphology of a neuron.
Finally, as this project combined a computational tool for guiding optogenetic design, together with experimental validation, it demonstrates a viable possible alternative method to reduce the ethical footprint of biological models. It also provided a useful tool with which to evaluate whether experimental constraints adversely impacted experiments, and how best to compensate for these effects.