Two-photon mouse air hockey

Blogging on Peer-Reviewed Research

Every so often, I encounter a technical advance that is simply so crazy-cool that I have to talk about it. Dombeck et al. publishing in Neuron offer such an advance.

They found a way to image the activity of whole fields of neurons using two-photon fluorescent microscopy -- a technique that I will define in a second. They can do this with in mice that are actually behaving by mounting the mouse in an apparatus that lets the mouse run on a track ball floating on air -- just like air hockey. (I want to meet the person who came up with that. There had to be high-fives all-around.)

Background: Methods for Recording from Many Neurons including Two-photon Fluorescent Microscopy

If you are a behavioral neuroscientist -- like me -- one of your big questions is: what sort of information does neuronal firing encode? We would like to know how the brain encodes information so that we can understand what particular regions of the brain do. Now the more traditional way to figure this out is to put an electrode into the part of the brain you want to study. Then you have the animal -- usually a rat, but sometimes a mouse -- cruise around doing some behavior task of your devising. Then you relate the activity recorded in the electrode to some aspect of the behavior or the task. This is how you determine what the action potentials mean.

There are some drawbacks to this technique, however.

First, there is a limit to the number of neurons that you can differentiate in this manner. At some point, if the two neurons are too close together you cannot tell if the action potentials recorded are coming from one or another. Researchers have solved this problem by lowering bundles of electrodes -- called tetrodes because oddly enough they are bundles of four electrodes. The four electrodes function sort of like radio towers for triangulating position, only they are trying to identify distinct neurons and assign each recorded action potential to them.

The second and perhaps larger issue is that electrical recording provides no spatial or identity information about the recorded cells. Since you cannot say precisely where the electrodes are you cannot triangulate the exact location of the recorded cells or where they are in relation to one another. Because you cannot see the recorded cells, you can't tell whether they are neurons or some other cell type.

There are some other techniques under development. One such technique is called two-photon fluorescent microscopy. To understand this, first you have to understand fluorescent microscopy.

i-92f63c44a71180de022f9b7ac2b1f3f4-fluorescence.gifFluorescent microscopy works something like this:

You shine a high energy lamp on the sample in question -- often the lamp is in the UV spectrum. Photons from the lamp excite electrons in specific fluorophores in the sample -- these are added ahead of time to specifically label proteins or structure...I don't have time to explain how you do that -- to a higher energy state. When these excited electrons decay back to ground state they will emit slightly lower energy photons. You can use a filter to observe only the photons that are emitted, screening out the photons from your lamp and from the background. Thus, fluorescent microscopy allows for specific imaging of whatever you labelled with the fluorophore. Only the things you want to glow will glow.

There are limits to the resolution of fluorescent microscopy. Also, it is very hard to look into thick tissues. The excitation spectra of the lamps typically used for fluorescence are absorbed by the tissue.

So, researchers perfected a different technique of doing this called "two-photon" fluorescent microscopy. Two-photon fluorescent microscopy is the same as the regular, only this time you use two different excitation photons rather than one to bring the electron on the fluorophore up to the excited state. The benefit of this is two fold. 1) Because you are using two photons, each photon now can have lower energy. It turns out that the energy involved now puts the photons in the area of near-infrared which is absorbed very little by thick tissues. The photons go right through. 2) Using two photons and two lamps allows you to focus the beam on a very, very, very thin section in the tissue. You can literally take optical sections through a tissue that are less than a cell thick.

Two-photon microscopy allows you to image a wide field of many cells within a tissue. This can be applied to the problem of measuring activity by using a variety of activity-specific dyes as fluorophores. The most common is probably the Ca influx dye used in this study. Active neurons typically have Ca influx. This Ca activates the fluorophore which can then be observed using two-photon microscopy.

Ouala! You can take a picture of the activity of a large number of neurons' activity in a tissue. Also, you know where they are in relation to one another. You can also determine cell type by morphology or other staining techniques.

Two-Photon in a Behaving Mouse

Sufficeth to say, two-photon microscopy is very complicated. It also requires equipment that is both heavy and expensive. While it is feasible to mount an electrode on a mouse's head while it is running around -- we do this all the time -- it is not feasible to mount a microscope. Further, even the slightest motion -- we are talking microns of motion -- can screw up the picture for two-photon. This is why the technique has mostly been applied either in culture or in anesthetized mice.

Not any more! This is where Dombeck et al. are particularly tricky. Basically they attach a mouse's head to a head stage that is attached to the microscope. The head is immobilized so that it can't move even a little bit, and they use a computer program to subtract out even motion on the level of microns.

Under normal circumstances, the mouse attached to the microscope would not be able to do much beside be pissed. It certainly would not be able to perform meaningful behaviors during which we could measure the activity. So, the authors of the paper place the mouse on a trackball made of styrofoam floating on an air cushion -- sort of like air hockey. The mouse can run in response to light or smell stimuli without large changes in head position.

The set-up for these experiment is shown here (Figure 1 on the paper):


The data for this is pretty astonishing. Here is a video of some of the data:

The video shows what the two-photon microscopy looks like. The grayish blobs are neurons. Neurons are not naturally fluorescent; therefore, to image them the researchers use a particular genetically engineered mouse that expresses a fluorescent protein in the cells (YFP for the initiated). The redish bursts are Ca influxes as imaged by a fluorescent Ca indicator dye. The bar on the left indicates how fast the mouse is running on the track ball.

The sheer number of sophisticated techniques that went into making this paper possible is staggering. Just to summarize in case you weren't paying attention, they include:

  • Fluorescent microscopy
  • Two-photon fluorscent microscopy
  • Ca-sensitive fluorophores
  • Genetically-engineered mice with fluorescent proteins
  • Mouse air hockey

Whoa. And, sometimes I think my job may be becoming a little too complicated...

Anyway, as if this weren't enough, the authors discovered some interesting stuff -- besides of course that they could pull this all off.

First, one of the limits of recording with electrodes is that if a cell doesn't fire an action potential, you don't know it is there. You cannot effectively calculate the number of silent cells. Using this technique you definitely can. They authors calculate about a quarter of the neurons are silent in this region during the task.

Second, the authors observed calcium influxes into astrocytes. These had hitherto been only observed in cultured tissues or anesthetized animals. It would appear that the electrical activity in non-neurons does occur physiologically and might be relevant.

I doubt that this technique will fully supplant electrical recording for several reasons. (Equating the imaged activation with the electrical activity must be difficult. Also, it is just too involved a technique to become standard. This whole set up must have cost a bundle.) Those concerns aside, however, this experiment hasn't even scratched the surface of what this technique is capable of. Investigating the electrical interaction between neurons and between neurons and astrocytes in a spatial framework has been an ongoing issue in neuroscience. This technique could solve it that problem.

Congratulations to Dombeck et al. for a truly excellent paper.

The full citation:

Daniel A. Dombeck, Anton N. Khabbaz, Forrest Collman, Thomas L. Adelman and David W. Tank. "Imaging Large-Scale Neural Activity with Cellular Resolution in Awake, Mobile Mice." Neuron Volume 56, Issue 1, , 4 October 2007, Pages 43-57.

Hat-tIp: Faculty of 1000


More like this

Blood flow in the brain is linked to neuronal activity. Therefore, various 'brain scanning' techniques can be used to observe neuronal activity in the brain. This has led to an astonishing revolution in knowledge of how the brain works. Of course, you knew that already. Also astonishing is that…
A paper by researchers from Princeton University, just published in the open access journal PLoS One, describes a new virus-based technique for probing the connections between neurons while simultaneously monitoring their activity in live animals. Various methods are available for studying the…
In his 1941 book Man on His Nature, the Nobel Prize-winning physiologist Sir Charles Sherrington described the brain as "an enchanted loom where millions of flashing shuttles weave a dissolving pattern." Little could he have known that within 50 years neuroscientists would have at their disposal…
USING an inventive new method in which mice run through a virtual reality environment based on the video game Quake, researchers from Princeton University have made the first direct measurements of the cellular activity associated with spatial navigation. The method will allow for…

"Using two photons and two lamps allows you to focus the beam on a very, very, very thin section in the tissue."

They actually only used one Ti:sapphire laser. The high resolution from two-photon fluorescence microscopy with a single laser cames from the fact that this type of fluorescence goes as the intensity squared, so the process is only efficient at the very center of the laser's focal spot, and there's basically no fluorescence from outside this very small area. Normal (one-photon) fluorescence is efficient over the whole focal spot because it is directly proportional to the source intensity.

Using two lasers with overlapping spots is possible, but then you have to get two laser spots of a few microns each to overlap and move perfectly together. If they don't move perfectly together, you'll get intensity fluctuations in your image caused by this imperfect alignment rather than by any real features in what you're imaging.


Speaking as one who was involved in some very early electrode implantation experiments in rats and cats (1961-63) with recordings using an EEG machine, this is simply amazing.

One area where part of this technique might be quite useful (without needing the air hockey) would be tracking the firing of visual, auditory, and olfactory neurons.

Another variant, with an apparatus that could essentially hold a conscious animal motionless with the exception of moving one limb at a time or moving the entire animal in various positions, would be to study the neuronal circuits involved in proprioception -- an area of study that has had too little attention.

Thanks for that Jake, that is truly amazing.

What's the blue signal?

The blue signal is the excitation spectrum from the excitation lamp. The green signal is the emission spectrum from the excited fluorophore.

Of course, the two are not always blue and green. Sometimes they are different colors or not in the visible light spectrum. Particularly with two-photon, the "blue" is actually near infrared. (Although, I think the emission for this experiment was still visible red. I will have to look it up...)

This is certainly a great paper, and demonstrates that many experiments in sleeping animals aren't going to cut it anymore.

The excitation wavelength for the calcium imaging is 890nm. Infrared.

Also, the following conclusion is misleding. "First, one of the limits of recording with electrodes is that if a cell doesn't fire an action potential, you don't know it is there. You cannot effectively calculate the number of silent cells. Using this technique you definitely can."

This is true only for calcium events that are above the detection threshold. This paper has not demonstrated that single AP events can be reliably discriminated from noise. As I recall from discussions with the authors at a poster session, while the occasional single AP event may be detected, a short burst is required for a reasonable detection accuracy. This is a challenge because, depending on the brain region, sparse spiking rates are very prevelant and encode real information.

Other issues :

The bulk loading of dye makes this a terminal experiment that is limited to a few hours at most. Ideally one would want to go back to the same region after a training paradigm.

Bulk loading also stains promiscuously, preventing the identification of the circuit architecture. We want to see how these cells are wired up, otherwise the spatial information does not add more value than dropping a tetrode in.

A better solution would be using genetically encoded calcium indicators in this experimental system. This solution is dependent on the quality of the indicators, which are still significantly worse dF/F than organic dyes. However, these are getting very close, with G-CaMP2 from Junichi Nakai, d3cpv Cameleons from Amy Palmer and TN-XXL from Oliver Griesbeck all looking good for single AP detection in vivo in sleeping mouse.

If you want to read more on this stuff, just check out my blog Brain Windows.