A couple of years ago, I ran an optogenetics experiment with bilateral light stimulation in the hypothalamus. Or rather, that was how I planned it. However, when it came to tethering and stimulating the mice, the fibres were too close to each other. I ended up doing unilateral stimulations, and realised that I would need to do angled opto fibres for any future bilateral studies.
Luckily, a stereotaxic frame comes with a pivot to allow exactly this kind of thing. There’s usually a pin holding it vertically; take that out and you can tilt the top half as needed to your desired angle. There’s an “angle scale” to do this accurately, and then you can clamp the whole thing in place.
So now comes the sticky question: how do you work out a new set of coordinates based on an angled implant? Turns out, all you need is a bit of trigonometry (Mr Turner did tell me it would be useful later in life, I just never believed him).
I drew a diagram to calculate the new coordinates based on my chosen angle of 10⁰. I used sine and cosine to derive the unknown lengths:
Now, I can tell you, these calculations were a faff, so I will avoid them in the future if possible. And, I will also say that I am the only person in the lab who has attempted angled fibres. Everyone else just does unilateral, and I thing a large part of their reticence is caused by ignorance of how to calculate the angled opto fibres.
With that in mind, I have developed a handy tool to do these calculations for you. Just input your starting lateral and dorsal coordinates, and the angle to tilt your fibres. I recommend 10⁰. Then press “Calculate” and the script will output your new lateral and dorsal lengths.
I hope people find my tool useful, and encourage them to use angled opto fibres in their studies.
Last blog post I had a revelation about the best numerical aperture to use for in vivo implanted optical fibres. Today, as part of my indepth study planning, I’ll be investigating the best opto flash time. My default has always been 10 ms, because a) it seems to be what most others use and b) it’s always worked well for me.
However, I like to be sure, and it never hurts to optimise your methodology. But where to start? I’ve mentioned in the past about the EC50 of ChR2 being 1 mW/mm2. However, this is actually misleading, as it doesn’t take into account the duration of illumination.
Power, energy and time
The important point here is that mW is a unit of Power (Watts), which is energy (Joules) over time (seconds). And the thing that actually determines activation of the opsin is the energy that it is exposed to. What this means is that, in principle, you could have wildly differing power output activating ChR2 to the same extent, so long as you adjusted the length of time of illumination accordingly.
If we think of a typical in vivo light flash of 10 mW for 10 ms from a 200 µm fibre, we can calculate the energy emitted in this flash with the equation Power (W) = Energy (J)/Time (s):
Energy (J) = 0.01 W x 0.01 s
= 0.0001 J
So we can say that 0.0001 J (or 100 µJ) of 470 nm light is enough energy from a 200 µm diameter fibre to robustly activate ChR2 in the brain in an experimental setting.
Low opto power
Now let’s say we could only produce 100 µW from our fibre (100-fold less than in the previous example). We could theoretically activate ChR2 by adjusting the illumination time accordingly:
Time (s) = 0.0001 J / 0.0001 W
= 1 s
What this means is that if we had a pitifully weak light source, we could still activate ChR2. Although, I’m not sure how useful 1 Hz neuronal stimulation would be biologically. However, there is a way to make this dim level of illumination biologically relevant, as Anpilov et al. did in their recent wireless opto study1. They did this by using a stabilised step-function opsin (SSFO), which acts more like a toggle switch – a single activation turns it on for 30 mins or so.
Fast opto flashing
We can also look in the other direction, power wise. Let’s say you were interested in making neurones fire at 100 Hz. To maintain a 10 % duty cycle (to allow the neurones to recover electrically and to limit tissue warming), we might want a 1 ms light flash, and we could calculate the required optical power like this:
Power (W) = 0.0001 J / 0.001 s
= 0.1 W
So, to drive a fast-frequency neurone like this with an equivalently robust activation of ChR2, we would need to be able to produce 100 mW out the end of a 200 µm fibre, which would be possible with a laser system. A quick note: 100 mW is actually a lot of light power to pump into a mouse’s brain. So, I would not advise aiming that high. I would worry about heating or damaging the tissue, so better to limit yourself to 15 mW or so, and validate your experiment accordingly.
Measured opto flash times
Anyway, back to my planned experiment. The question was: do I need my full 10 ms flash time to produce the firing I want? A recent paper by Herman et al. investigated the silencing of ChR2-expressing neurones at higher light exposures2. It includes a nice overall picture of light pulse duration-dependent spike probabilities in a variety of neurones (Figure 1).
What they find, flashing various neurone types at 20 Hz, is that with light pulses of 5 or 10 ms they have increasing spike probabilities up to 95 – 100 % depending on the neurone type. Then at on-times of 25 ms or longer, the spiking fidelity drops in all neurone types except for fast-spiking neurones in the cortex. Based on this work, I would suggest 5-10 ms appear to be optimal across various neurone types. At any pulse length above or below that, the spiking falls away.
Right, while 5-10 ms looks like a good time duration, that study was performed at a single light intensity, so only provides a partial answer. However, I found an early paper that investigated the threshold light power needed to stimulate an action potential at various distances from the end of the fibre, across a range of pulse widths (Figure 2)3.
A couple of things are clear from Figure 2:
Longer pulse widths drop the power threshold needed to trigger an action potential.
The threshold power needed to trigger an action potential increases with distance from the fibre tip.
It’s difficult to tell from the tiny scale on this graph, but it looks like 5 ms might just be enough to trigger an action potential at 1 mm from fibre tip at the ~9 mW power we get from our system. However, this is dependent on other factors, such as the NA of the implanted fibre.
The best opto flash time
My verdict form this investigation is that 5 ms would likely be fine to trigger a response. However, increasing the flash duration to 10 ms would increase your likelihood of triggering action potentials without any noticeable drawbacks. So after all that, we come back to 10 ms as the best opto flash time (in my opinion).
1. Anplilov et al. Neuron 107(4), 644-655 (2020) Wireless Optogenetic Stimulation of Oxytocin Neurons in a Semi-natural Setup Dynamically Elevates Both Pro-social and Agonistic Behaviors
2. Herman et al. eLife 3, e01481 (2014) Cell type-specific and time-dependent light exposure contribute to silencing in neurons expressing Channelrhodopsin-2
3. Foutz et al. J Neurophysiol 107, 3235-3245 (2012) Theoretical principles underlying optical stimulation of a channelrhodopsin-2positive pyramidal neuron
Planning another optogenetics study, and I needed to cut the optic fibre cannulae ready for implantation. One of the other postdocs in the lab had been super organised and bought in a bunch of implants from Thorlabs at a variety of numerical apertures (thanks Amy). But, which is the best numerical aperture (NA) for my experiment?
I won’t go into details (because I’m not a physicist), but Wikipedia defines the NA of an optical system as “a dimensionless number that characterises the range of angles over which the system can accept or emit light”.
Essentially, as far as we are concerned for fibre optics, the NA is relevant for two things:
The bigger the NA, the more light from the source will travel down the optic fibre – for a laser system, this doesn’t matter much because the coherent light can easily be focused down it, but for an LED, this can make a big difference for how much light is captured by the fibre (rather than scattering away)
It determines how much the light spreads after exiting the fibre (for in vivo opto’s, this will be in the mouse’s brain) – the higher the NA, the greater the cone of light dispersion
So, back to cutting fibres, and I had to decide which ones to use – I normally use the 0.22 NA fibres out of habit, but I have read multiple recommendations to use as high an NA fibre as possible when using an LED system (which is what we have); the idea being to get as much light power as possible into the mouse’s brain, which is important considering LED systems can struggle to be bright enough for in vivo opto’s. Both Prizmatix and Doric suggest using 0.66 NA fibres for LED-connected systems, which is actually higher than the ones we have available from Thorlabs.
To test the light output, I hooked up fibres of different NA’s to our LED optogenetics system, and recorded the light power out the end of the fibre using a light meter, both under constant illumination and during 10 Hz flashing with 10 ms on times (Table 1).
True to form, the higher the NA of a fibre, the more light that is passed down it. Great, so at this point I’d pretty much settled on the 0.50 NA fibre, because it emitted approx. 50 % more power than the 0.22 NA fibre. However, for the sake of completeness, I decided to input the values into Karl Deisseroth’s irradiance predictor, to check how deep I would get good ChR2 activation. This is a useful step when planning placement of your optic fibres.
I plotted the values for all three NA fibres (Figure 1), and I’ve included the threshold level of 1 mW/mm2 that I’ve talked about previously (this is the measured EC50 of ChR2 H134R, which I use as a threshold irradiance to assume good activation).
Now I’ll be honest, I was surprised by this outcome – despite having lower light output from the lower NA fibres, the irradiance was higher as soon as you go deeper than about 0.2 mm into the tissue. I can only assume this is because the lower NA results in less light spread coming out of the fibre – the 0.50 NA fibre remains above the critical 1 mW/mm2 down to about 1.0 mm, whereas the 0.22 NA fibre goes to about 1.4 mm.
The answer is simple – I’m going to use the 0.22 NA fibres, because they have the dual benefit of activating ChR2 to a greater depth, and also having lower brightness at the end of the fibre, which means less heating of the tissue and phototoxicity.
My interest in wireless optogenetics has come up a couple of times. In fact, I’ll start with a quick correction: I prefer to call it fibre-free optogenetics, after multiple people mistook my wireless system I was designing as meaning controlled via Bluetooth or WiFi. Which it ain’t. And, for me at least, the whole point of going “wireless” is to do away with the optic fibres, which really embody all the issues and difficulties with in vivo optogenetics:
Impacts to the animal – the need to have the animals in an open cage, with an open lid and a sterile environment to prevent damage to the fibres. Also, they tend to be stiff, having severe behavioural impacts.
Loss of optical power – the optic fibres require additional optical connections, which inevitably leads to light loss, and therefore difficulties obtaining a high enough brightness.
Expensive and fragile – not much more to say, other than we have spent thousands of pounds maintaining the optic fibres for our optogenetics system. This may be more than is typical, but I think that’s because the Plexon fibres we use are very fine and lightweight – I have used more durable ones that were even worse for the mouse behaviour because of the added stiffness.
The most important reason to do away with the optic fibres, as far as I’m concerned, is the impact to the animal. Quite apart from minding the 3R’s with regards to animal welfare, tethering will inevitably cause stress, which is detrimental to the data you can acquire (Figure 1). In fact, it is to the NC3R’s that I am applying for funding to take my fibre-free opto system to the next level.
There is of course the added bonus with wireless optogenetics that you can do optogenetic stimulation in otherwise impossible setups. For example, I am very keen to use my fibre-free opto’s in our calorimetry system to measure energy expenditure in response to opto stim. This is done in an air-tight sealed container, which to my knowledge this has never been done with optogenetic stimulation in the brain.
After a fair bit of research, I have found 4 commercially available wireless in vivo optogenetics systems (Figure 2).
Helios by Plexon and Teleopto by Amuza are both very similar, except that the Helios headstage attaches to “normal” implants, whereas Teleopto make their own custom implants. Both require you to point an IR remote at the headstage constantly (ie. the flashing stops if the signal stops). Fi-Wi from Doric connects over radio signal to drive opto flashing; similar to Teleopto they use custom implants. Neurolux is a very different system to the other three, and uses electromagnetic induction to remotely power the implants. Hence the Neurolux implants are tiny and custom (the LED is actually on the end of the fibre that gets implanted).
I have collated a summary table of the various systems, including a number of parameters (Table 1). Included is the cost to buy a complete setup to stimulate 1 mouse at a time, which usually comes with a few implants. However, I was unable to find out the irradiance available from the Plexon Helios system, despite asking the sales people for those details.
Overall, the Doric system seems the best of the bunch; despite being the heaviest it is very compact and produces by far the highest irradiance. In fact, it provides higher irradiance than the system I’ve been developing, which comes out around 150 mW/mm2. Stay tuned, and I’ll be talking more about my system in the coming months.
1. Won et al., Nat Biomed Eng (2021) Wireless and battery-free technologies for nanoengineering.
I have previously written about the importance of brightness for in vivo optogenetics experiments. It’s just as important for in vitro optogenetics, which is what I’ll be looking at today. This came about because we’re planning a publication, and we need quantification of the light irradiance that we get on the brain slices.
When I started in vitro optogenetics, I tested the brightness of the LED system I had bought for the purpose, but not with a light meter – instead I tested directly on ChR2-expressing neurones, and found that 2 % brightness of the 470 nm LED was sufficient to elicit action potentials.
This was enough for me at the time, and I never bothered doing the metred quantification because the light meter didn’t fit under the objective (even after removing the tissue perfusion bath). However, for publishing I wanted a proper irradiance value, which meant me and our ephys technician spent an afternoon trying to dismantle the condenser under the stage. I say trying to, because microscope has been in pretty heavy use for at least a decade without any kind of service, and we found a lot of salt residue from past aCSF leakages.
Hopefully y’all cringed at the thought of that, because salt build-up inevitably means corrosion of expensive microscope parts. And, surprise surprise, we found the screws and bolts holding the condenser together and onto the microscope are all rusted in place. In the end, we managed to unscrew the top part of the condenser’s lens and wiggle the light meter in place under the objective. Phew! So now we went through a range of LED brightness and measured the brightness coming out the bottom (Figure 1).
As ever, the brightness is not the important parameter here. What matters for activating opsins is the irradiance hitting the slice (irradiance being intensity of light per unit area). So now comes the difficult bit – how do I know the area that the light is hitting on the slice? It is possible to get a microscope ruler, put that under the objective and measure the diameter of your field of view. However, how do I know the camera or eyepiece are visualising the entirety of the illuminated area?
The answer is to go back to physics, and field of view of the objective in use. I found a useful guide from the makers of my LED’s1:
A quick investigation shows me that the Field Number for my objective is 22. Dividing that by the magnification of 40 gives a diameter of 0.55 mm. I know the area of a circle is πr2, which gives me a surface area of 0.238 mm2. So, adjusting the brightness values obtained earlier gives the irradiance output from the LED’s (Figure 2).
I have also plotted a simple linear regression line on the irradiance graph to give an easy formula to give a rough estimate of the irradiance at any given LED brightness. However, I will still make sure to use actual measured values in any publication, rather than the estimates obtained from the regression. Anyway, this does match up nicely with my earlier test data, as the EC50 for ChR2h134r is 1 mW/mm2, and 2 % brightness on my blue LED gives an irradiance of just over 2 mW/mm2.
For any future in vitro optogenetics studies on this rig, I will aim to use 10 % LED intensity, as this will give a solid irradiance of 10-15 mW/mm2, without going into higher saturating irradiance levels.
I have exciting news for my loyal readers: I have finally completed the prototype for my EasyTTL Mega controller! This is a 4-channel optogenetics TTL driver, which I plan to use for in vivo optogenetics experiments.
I’ve mentioned parts of the development of this projectbefore. But, for those who haven’t read the previous blog posts, the idea behind it was to make an optogenetics TTL driver that is simple and easy to use. I get so annoyed by the unnecessary complexity (and associated costs and difficulties) inherent in most neuroscience research equipment. As such, I have produced a massively simplified device that is controlled by the user with knobs (teehee) and switches.
The EasyTTL Mega (Figure 1) is a 4-channel optogenetics TTL driver built on an Arduino Mega core1. The Arduino Mega is a robust open-source microcontroller board with a massive array of I/O pins. I have given it four TTL outputs, each individually switchable by tactile toggle switches. The stimulation settings are determined by three dials: one for the flash duration, one for the frequency of stimulation and one for the brightness. The flash settings have been taken from a decade of optogenetics literature and experiments by yours truly, and covers 99% of flashing paradigms that I have seen.
Being a good scientist, I needed to test the system, so I connected a couple of the outputs to an oscilloscope and turned on the pulsing. First, I checked the pulse durations (Figure 2), which are consistently accurate across the range.
Next, I tested the range of flashing frequencies (Figure 3), which are also bang tidy.
Finally, I needed to check the brightness control (Figure 4). A quick note: the brightness control is based on pulse width modulation, which means that the laser or LED controller can be set to a constant current. However, the frequency of the PWM is 980 Hz for TTL outputs 1 and 2, but only 490 Hz for TTL outputs 3 and 4. What this means practically is that there will be a threshold flash duration below which the PWM makes the brightness unstable from one flash to the next, and this will be worse for the pins 3 and 4 which run a slower PWM.
Based on my recording of the PWM outputs, the dimming control is unusable for flash durations of 1 ms on all outputs, and 2 ms or less on outputs 3 and 4. Despite my earlier misgivings, I think the 2 ms flash on outputs 1 and 2 looks fine. I’ve put the EasyTTL Mega in the shop, in case anyone wants one for their own research.
My interest today is in optimising optogenetic terminal stimulation. I have previously talked about optimising your opto stim frequency, and I had a little dig at a collaborator who stimulates at 30 Hz.
Optogenetic stimulation fidelity
To summarise my previous post regarding optogenetic stimulation fidelity: if you optogenetically stimulate a neurone too fast (and the definition of “too fast” depends on a number of factors including the next type, opsin, duration and intensity of light flash etc.), they cannot keep up, and instead of firing action potentials they become chronically depolarised without firing.
Essentially, the neurones need time to come recover their membrane potential back below a certain threshold (typically around -50 mV) before they can produce another action potential, thus they effectively become silenced when you want to be activating them.
However, it has since occurred to me that my collaborator only ever stimulates the projection sites of his ChR2-expressing neurones. The neurones he is interested in are found in the hindbrain, and as such he can’t reliably stimulate the soma.
So I had a thought: does stimulating neurone terminals work differently from soma because you don’t need to send the signal down an axon, and as such allow his high frequency stim to work? In particular, my collaborator maintains that high frequency stim results in more release of neuropeptides (eg. AgRP/NPY), rather than fast amino acid transmitters (eg. GABA). This has, in principle, been known for a long time1, but it doesn’t mean a) it’s true for all neurones everywhere b) it’s possible to stimulate neurones that fast in vivo using optogenetics.
In this post, I’ll be exploring the second point in more detail, by looking at the fundamental biology of a neuronal synapse, what causes release of neurotransmitter and how we can successfully control that with optogenetics.
Biology of a synapse
A quick biology lesson: we all know that neurones have action potentials, which is a transient spike in electrical activity across the membrane, and this is how they send information down an axon. However, it isn’t the voltage change that causes neurotransmitter release at the nerve terminal. At least, not directly. Instead, the increase in membrane potential causes an influx of calcium, and it’s the increase in calcium that causes vesicle fusion and release of the neurotransmitter (Figure 1)2.
So, my question regarding optogenetic stimulation of nerve terminals is this: if you overstimulate a nerve terminal into a chronically depolarised (silent) state, do you still drive release of neurotransmitter? This might sound paradoxical, but it could theoretically happen if the calcium release occurs at a slightly depolarised membrane potential (maybe -30 mV, which is easily obtainable by opto-overstimulation). In that case, action potentials would not be necessary, as we are already at the terminal end of the axon and don’t care about sending electrical signals, only about triggering vesicular fusion by increasing calcium.
To answer my question, we need to look at the channels that cause the increase in calcium upon membrane depolarisation, and in particular at what membrane potential the calcium release is triggered. If calcium release occurs at a low (-30 mV or below) membrane potential, then we could happily see neurotransmitter release from chronic depolarisation. However, if it occurs at a more depolarised level, it is extremely unlikely that the neurone terminal would reach that membrane potential from chronic overstimulation.
The channels we’re interested in are called, surprise surprise, voltage-activated calcium channels. This actually comprises a large family of channels, with multiple groups. I won’t go in to depth here, because there is so much literature concerning these channels. However, of the ten mammalian variants, there are three that are important in neurones for synaptic release (Table 1)3.
To quote the review by Dolphin3:
“For most synapses, CaV2.1 (P/Q)- and CaV2.2 (N)-type channels are involved in varying proportions in synaptic transmission, depending on the synapse in question and the developmental stage… At some synapses, CaV2.3 channels, activated by smaller depolarizations, play an important role, rarely as the main channel involved in vesicular release”
So there you have it, straight from the … Dolphin’s mouth. The two channels that are most important for vesicular fusion and the release of neurotransmitters are activated at very high membrane potentials (-5.7 mV and -13 mV), which are far too high to be activated by non-firing chronic depolarisation of the membrane.
I now feel confident in saying that high frequency optogenetic stimulation (eg. 30 Hz) of a nerve terminal, which is likely to induce chronic depolarisation rather than action potentials, is not likely to cause the release of neurotransmitter from the presynaptic terminal. I would therefore urge my fellow researchers to refrain from such high frequency optogenetic stimulation.
1. Dutton and Dyball J Physiol (Lond) 290, 433-440 (1979) Phasic firing enhances vasopressin release from the rat neurohypophysis
2. Südhof Cold Spring Harb Perspect Biol 4(1), a011353 (2012) Calcium control of neurotransmitter release
3. Dolphin Function 2(1), zqaa027 (2021) Functions of presynaptic voltage-gated calcium channels
Why do you eat? We’ve come to realise that it’s far more complex than simply saying you eat because you are hungry. Anyone who’s been reading my blog will understand the complexity of the neural networks involved in the control feeding behaviour.
A paper came out a couple of weeks ago that splits out a number of the drivers for food intake by differentiating a subpopulation within the lateral hypothalamus (LH)1. This stems from an earlier paper that found that GABAergic (VGAT-expressing) neurones of the LH show heterogeneous feeding behaviours2:
Consummatory behaviour – the classic homeostatic drive to consume calories
Appetitive behaviour – the more complex drive for reward and rewarding foods
Anyway, back to the recent Siemian paper, where they investigate leptin receptor (LEPR)-expressing neurones of the LH as a subpopulation (~20 %) of the larger VGAT population. They start with my favourite method for inhibiting neurones, which is to ablate them using cre-dependent caspase (AAV-FlEX-tsCasp3-TEVp; Figure 1A/B). Ablating LH VGAT neurones with caspase causes a lean phenotype with reduced food intake (“consummatory behaviour”; Figure 1C/D), with delayed learning to cue-stimulated sucrose intake (“appetitive behaviour”; Figure 1E). However, LH Lepr-ablated mice had no change to gross food intake or body weight (Figure 1F/G), but did learn a cue-stimulated appetitive response (Figure 1H).
In order to directly control feeding behaviour, they next use my favourite method for driving neuronal activity, which is optogenetic stimulation of ChR2. Injecting cre-dependent ChR2 or the inhibitory NpHR into the LH of VGAT-cre or Lepr-cre mice (Figure 2A/B), they run two behaviour tests to mimic the data seen with caspase. In this case, because optogenetic stimulation is instantaneous (rather than the long-term chronic caspase), they use acute tests.
First is free-access feeding to assess consummatory behaviour; second is real-time place preference to assess appetitive behaviour. I use both these tests frequently, and find them to be both robust and informative. The data show, as expected, that optogenetic stimulation of VGAT neurones drives consummatory behaviour (Figure 2D/E) and a strong preference (Figure 2H/I), whereas stimulation of Lepr neurones induced place preference (Figure 2J/K) without any impact on consummatory behaviour (Figure 2F/G).
Next comes the data that I think is the most interesting part of this paper, and also the most complex, where the authors use a fluorescent miniscope to investigate GCaMP activity in the two populations of interest (Figure 3A/B). What’s really interesting here is how they split out the neuronal populations based on the timing of their response to whether they were responsive during the cue or after it (they show “pre-responsive” neurones, but in a world without midi-chlorians, I don’t think we should put too much stock in neurones that predict when a cue is going to happen).
So anyway, what we’re interested in is the subpopulations of neurones that respond differently between the CS+ and CS– stimuli, which for LHVGAT is the post-lick reward-responsive neurones (Figure 3G), but for LHLEPR is both the cue-responsive and reward-responsive neurones (Figure 3J/K). This is quantified in Figure 3L-O, where they show that LHLEPR neurones are strongly and significantly predictive of reward from cues. This means that in addition to the general LH drive for reward, the LHLEPR neurones are the ones that can distinguish cues and therefore may be important for the behavioural discrimination of food cues.
There are a couple more figures that go into further detail to show the importance of the LHLEPR projections to the VTA for mediating appetitive learning, and then show that the LHLEPR neurones are not relevant for cocaine preference. But, I’ve shown here the results that I found most interesting, and how they inform us on the control of feeding behaviour. In particular, I want to highlight the use of miniscopes to pick out subpopulations of neurones based on behavioural responses.
1. Siemian et al. Cell Reports 36, 109615 (2021) Lateral hypothalamic LEPR neurons drive appetitive but no consummatory behaviors
2. Jennings et al. Cell 160, 516-527 (2015) Visualizing hypothalamic network dynamics for appetitive and consummatory behaviors.
A paper came out recently that looks at optogenetics in a way I would have never thought of1. It’s funny how much effort we put into lasers and optics and everything in order to deliver light into a mouse’s brain, because visible light just doesn’t pass through tissue well enough for us to try activating from outside the brain (Figure 1A). But it never occurred to me to use non-visible wavelengths of light for the purpose; in this case it’s x-ray optogenetics.
Obviously, this comes with its own challenges – if a wavelength of light passes straight through tissue, it can’t interact with the proteins, which includes any opsins. So, how do you do this? Well, today we’ll find out from Matsubara et al.1
I actually remember an undergraduate practical that gave the answer to this, although my hazy memories suggest it was gamma waves from a radioactive isotope, rather than X-rays. Either way, the principle remains, which is the use of “scintillant”, which as far as I know is a fancy word for a chemical that is fluorescent under high energy light waves.
Anyway, Matsubara et al. show us the effects of UV and X-rays on their scintillant, which is Ce:GAGG, which emits yellow light upon stimulation (Figure 1B). They then test this scintillant for activating opsins in cultured cells, and find robust activation of the red-shifted opsins (Figure 1C/D), particularly of our favourite super-sensitive red-shifted opsin chRmine2.
After some optimisation in brain slices, Matsubara et al. then take their system in vivo, and inject AAV-DIO-chRmine and their scintillant into the VTA of a DAT-cre mouse (Figure 2A). They get nice c-fos induction from X-ray stimulation of their model system (Figure 2B/C).
After next showing that their scintillant particles are not cytotoxic when injected into mouse brains, they test their opto system in a conditioned place preference experiment with stimulation of the VTA (Figure 3A). Due to the nature of X-rays, they set up the in vivo experiment in lead-lined 2-compartment preference cages (Figure 3B/C).
They then show that the mice show increased place preference with chRmine stimulation (Figure 3D), and decreased place preference with stGtACR1 (Figure 3E), as expected.
I’m also happy to report that the authors checked for long-term damage to the mice caused by exposure to X-rays. They found no change to locomotor activity or blood-brain barrier function.
However, after prolonged stimulation with the high-dose X-rays, mice did have reduced numbers of immature neurones in the detate gyrus. The low dose flashing of X-rays had no impact though, so I think this method would be fine to use so long as you were careful with your experimental planning to limit the X-ray exposure of the animals.
Having said that, x-ray optogenetics is not a technique that I ever envisage myself actually wanting to use. There is a high level of difficulty and complexity, which I don’t think outweigh the improvements to animal behaviour. I think other wireless opto methods have a much better balance of complexity to impact on the animal.
1. Matsubara et al. Nat Commun 22(1), 4478 (2021) Remote control of neural function by X-ray-induced scintillation.
A colleague sent me a paper about a novel opsin the other day, because he knows about my interest in optogenetics, and particularly in new tools that we can use to improve our experiments1. And then a few days later I received an email alert of a second paper2 that fulfils the same purpose as the first, namely producing new inhibitory opsins.
So, in this post I will investigate and compare these papers and what their results might mean for doing opto experiments. To begin, both papers aim to solve the same problem that has plagued optogenetics since its inception: the inability to optogenetically inhibit neurone terminals.
If this sounds untrue, let me quickly explain that while we have a number of inhibitory opsins available, none of them can produce reliable inhibition at the terminal. For example, ArchT is a proton pump, which causes hyperpolarisation, but in the tiny volume of the terminal also has a dramatic impact on the pH, which causes spontaneous neurotransmitter release.3
I’ll start with the common aspects of these new opsins: both are light responsive Gi/o-coupled GPCR’s, which means that they inhibit synaptic fusion by blocking production of cAMP and by suppression of Ca2+ release. However, the lamprey parapinopsin (PPO) is bistable, activated by UV and turned off by amber light (Figure 1A/B), whereas the mosquito panopsin homolog (OPN3; Mahn’s variant is called eOPN3) is activated by green light (Figure 1D/E).
Next, each paper goes on to demonstrate potent inhibition of neurone terminals in vitro. Both papers show extensive in vitro analysis, but for today I’m interested in the action at terminals, where they both show decreased amplitude of evoked post-synaptic currents (Figure 2A for Copits; Figure 2C for Mahn). They also both show they can decrease spontaneous post-synaptic current frequency without changing amplitude (Figure 2B for Copits; Figure 2D for Mahn).
Lastly, they both show they can impact animal behaviour in vivo by stimulating neurone terminals with their new opsins. For example, Copits et al. were able to block cocaine-induced conditioning in a VTA -> NAc projections (Figure 3A), whereas Mahn et al. managed to influence which direction mice were turning in an open field (Figure 3B).
All in all, I was very impressed by these new inhibitory opsins. If they ever become available, for example through Addgene, I would definitely look into them. It is important to be able to inhibit neurone projections like this.
However, from a purely practical point of view, I think I would lean towards the mosquito eOPN3 from Mahn et al, due to the stimulation wavelength of 500-550 nm as opposed to the UV stimulation of lamprey PPO from Copits et al.
1. Mahn et al. Neuron 109, 1621-1635 (2021) Efficient optogenetic silencing of neurotransmitter release with a mosquito rhodopsin.
2. Copits et al. Neuron 109, 1791-1809 (2021) A photoswitchable GPCR-based opsin for presynaptic inhibition.
3. Mahn et al. Nat Neurosci 19(4), 554-556 (2016) Biophysical constraints of optogenetic inhibition at presynaptic terminals.
