Guide to Playing Mozak
- Neuron Introduction
- Project Introduction
- Image Stack Introduction
- Spiny vs Aspiny Neurons
- Tracing Faint Branches
- Creating Connected Reconstructions
Welcome to Mozak! In order to help you get better acquainted with this application, we strongly encourage you to read the information below. Some of the tips and tricks should help you improve your skills and the work that you contribute to science!
Getting to know a Neuron
Let us start with the basics about neurons. The main compartment of a neuron is its soma. The soma contains a cell’s nucleus and many other organelles controlling the cell’s basic functions. Here are some images to help you identify the soma:
You will notice that there are many long, skinny processes exiting the soma. We call these processes neurites. However, we more commonly separate them into two groups: dendrites or axons.
Most neurons have multiple dendrites exiting the soma. The function of a dendrite is to receive the chemical signal from the axons of thousands of other neurons. Dendrites tend to travel directly away from the soma and branch a few times before ending. For the cells posted on Mozak, dendrites are usually thick, well-labeled and easy to trace. Here is an example to help you identify the dendrites:
To show you the axon, we are going to have to zoom in. This is because the axon is thin and intermittently labeled. Each neuron has only one axon, and it may exit the soma or it may exit a dendrite very close to the soma. The function of an axon is to transmit the signal of a neuron to all the neurons it synapses with. Unlike most dendrites, an axon can take a very complex path, branching, twisting and turning in all directions. An axon can also travel long distances to synapse with neurons in other areas of the brain or body. Luckily, we are only reconstructing the local axon within a few hundred microns of the soma (for now)! Here is an example image to help you identify axon. In this example the axon exits the soma (can you find where?):
You will notice that the axon is hard to identify in the image above. This is because the example is a 2D image. When you trace in Mozak, the axon will be easier to follow because we provide a 3D image stack of each neuron.
Getting to know the Project
Each neuron is contained in a thin slice of brain tissue. A researcher at the Allen Institute for Brain Science places a small glass pipette in the tissue, and uses a technique called Patch-Clamp to record the electrophysiological profile of the neuron. The electrophysiological profile basically tells us how the neuron communicates to other neurons by measuring the strength and timing of its action potentials (among other things).
Before the pipette is removed from the tissue, it is used to fill the target neuron with a dye called biocytin that makes the neuron visible. The slice of tissue is then imaged at 63X magnification with a special microscope. The microscope takes hundreds of pictures of the tissue at different depths to create a 3D image stack. Image stacks are posted on Mozak so players like you can reconstruct each of the labeled neurons. Your reconstructions capture the “morphology” of a neuron, which is just a fancy way of saying its shape. Our overall goal is to use the morphology and electrophysiological profile of each neuron to discover how many different types of neurons there are in the mouse visual cortex!
Check out the Cell Types Database being generated at the Allen Institute for Brain Science. This is the project your work will be assisting!
Getting to know an Image Stack
When you load a neuron on Mozak, the first thing you will see is the whole image stack. From this initial view, the cell will appear pixelated. This is because the full resolution image stacks are massive; they can be over 100GB!! We have to down-sample them to show you the whole image stack, otherwise the loading times would be unbearable. If you use the zoom function in the upper-left corner of the window, you will notice that the resolution of the cell improves as the area shown gets smaller. Here is an example:
It is most efficient to find a compromise between resolution and visible area. You will want to have enough resolution to be able to trace accurately, but enough visible area that you will not have to pan every few seconds.
The 3D image stack is made up of hundreds of 2D images stacked on top of one another. To see a single plane, you will need to use the connect the dots tool.
Tracing from the whole stack view using the virtual finger tool is appropriate for capturing thick, well-labeled neurites (usually dendrites), while tracing from the single plane view using the connect-the-dots tool is often used for faintly-labeled neurites (usually axon). When you are in the single plane view, use your mouse wheel to scroll between sections!
The last thing to know about 3D image stacks is that they have different zones, and in each zone the neurites will look a little different. Below is a side view of an image stack separated by zone with the left side being closest to the viewer.
- Zone one contains the planes that are closest to the viewer. These are also the deepest planes in the brain slice. Because they are deep, the signal here is usually much harder to see. All neurites (dendrites and axon) will appear faint and more diffuse. Boost the brightness slider to find all the signal here!
- Zone two contains planes that are shallow in the slice, where the signal is easily seen and captured. The soma is almost always in zone two.
- Zone three is very thin and contains the planes at the surface of the slice. The thing to know about this zone is that there is usually a bunch of false signal on the surface of the slice. Try to avoid tracing here unless you see a continuous, easily identifiable neurite.
- Zone four contains the planes furthest from the viewer. These planes are images taken above the surface of the slice, which means they will not contain any signal or tissue at all. Essentially, the microscope focused on the air above the slice. Note that this zone may be much thinner than in the example above.
Spiny vs Aspiny Neurons
The goal of this project is to learn all the different cell types in the mouse brain, but we already know a lot. For example, most neurons fall into two main categories: spiny and aspiny. Learning the difference between a spiny and aspiny neuron will help you trace more accurately!
A characteristic of all spiny neurons is that their dendrites have small projections called dendritic spines. Spines usually contain a synapse with the axon of another neuron, and having lots of spines increases the surface area available for a neuron to make synapses with lots of other neurons. You do not need to trace spines! Here is an example image to help you identify dendritic spines:
Spiny neurons generally have a “pyramidal” shape, which means they have one longer dendrite that travels towards the surface of the brain. The longer dendrite is called an apical dendrite (all the others are called basal dendrites). It is very common for the axon to exit the soma opposite the apical dendrite in spiny neurons. Here are some images to help you identify the pyramidal shape of spiny neurons and an apical dendrite:
The apical dendrite and its branches are in pink. The axon is in red, notice its location relative to the apical dendrite. Finally, can you see the pyramid shape the neuron has?
Aspiny neurons, by contrast, tend to have smooth dendrites. They still have synapses with the axons of other neurons, but their synapses are on the main shaft of their dendrites. Here is an example image of a smooth dendrite from an aspiny neuron:
Aspiny neurons also have lots of variety in their shape. Because of their smooth dendrites and the fact that their axon can exit the soma from any direction (or even from a distal dendrite) it can be much more difficult to figure out which neurite is the axon for aspiny neurons.
Tracing faint branches
Accurately tracing faint branch is one of the most challenging parts of creating a complete reconstruction. Here are some tips that can help!
- Remember to use the brightness slider to find the faintest signal. Inverting the images can also be useful (your eyes may prefer black signal on white background). The images below show the same location with the brightness adjusted. Can you find more branches in the image on the right?
- Rotating the image can be a useful way to find and follow a neurite. This is especially true if the neurite is moving directly towards or away from you. Turning the image stack 90 degrees, so the narrow side is facing you, can allow you to see when neurites are truncated at the surface of the slice.
- Capturing faint axon ends up being like a game of connect-the-dots. Axon labeling can be discontinuous. It is helpful to go into connect the dots mode and scroll through planes with your mouse wheel, looking for dots that travel in a consistent direction. Here is an example to show you what intermittent labeling looks like, and how to trace it:
Creating a fully connected reconstruction
On the left is a view of a neuron that other people have contributed to (a consensus reconstruction). At first glance it might look okay, but switching to 'Disconnected Mode' from the menu in the upper-right corner reveals the many segments that are not connected to the main structure. Having lots of disconnected segments makes the reconstruction less valuable to the scientists who want to use it in their project.
In the picture on the right (Disconnected Mode), the reddish segments are not connected to any segment that can be connected back to the soma, while the bluish segments do connect to the soma. In this viewing mode, your goal will be to find a way to connect the reddish segments to a bluish segment, turning as much of the structure blue as possible. HOWEVER, you should also avoid connecting any axon to dendrite and vice versa. The dendrites should all initiate at the soma, travel for a short distance, branching a few times and then end. The axon should initiate at the soma (or sometimes from a dendrite near the soma) and travel a great distance, with lots of branches, twists and turns. Generally, you should only connect segments that have the same quality of signal. For example, if a branch is spotty and faint, it should probably be connected to another spotty and faint branch. If a branch is thick and bright with spines, it should probably only be connected to another segment that is thick and bright, with spines.
To connect discontinuous segments, switch to Disconnected Mode, and use the connect-the-dots tool to select two existing nodes and draw a straight line between them (it is very helpful to have ‘Snap To Node’ turned ON). Below is a before and after view of a segment being extended using the connect-the-dots tool.
You will not want to connect every node you see back to a blue branch. Particularly nodes that are distant from any other nodes can be ignored, as they are probably places where people traced noise rather than actual signal. Below is an example of a node that is probably not marking real signal:
Other tips you should keep in mind when creating a connected structure:
- Make sure the 'Snap to Node' setting in the menu is set to 'On'.
- In Disconnected Mode, look for the reddish segments that are close to blue segments and try to extend the blue segment to reach them.
- Go to the highest resolution level when joining nodes. Finer details help make the trace look smoother and more accurate since you can capture all the small bends and turns a neurite might make.
- Right click and drag to view the reconstruction from more than one angle. This will help you double check that the traces you are trying to connect are as close as you think they are.