Connectomics
As a kid, my favorite game was Connect the Dots, particularly the ones of SpongeBob characters. I remember the feeling of satisfaction I got after connecting all 20 dots of Larry the Lobster! Buttttt as I turned the pages, they got more and more complex… 50 dots of Patrick Star, 100 dots of Sandy the Squirrel, 1,000 dots of Plankton! Damnit, Plankton! You’ve got to be kidding me!
But as it turns out, adults are playing this game now, too…. The New Edition: Connect 100 Billion Dots.
Sounds crazy, I know. But this is basically the goal of a massive project called the Human Connectome Project.
The Human Connectome Project, or HCP for short, aims to create a comprehensive map of all 100 billion neurons in the human brain, called a connectome. Depending on the neuroscientist you ask, the connectome may also include the synaptome (synapses) and epigenome (epigenetic factors) as well. Allow me to take you back to the basics of Biology 101:
A neuron (nerve cell) is the smallest functional unit of the nervous system. The dendrites take in inputs from the axon terminals of other neurons, and if the inputs are excitatory and threshold is reached (-55 mV), an action potential will propagate down the length of the axon. This will trigger the release of neurotransmitters into synapses, the 20 nm extracellular space in between two neurons.
The cell shown in the diagram above is just one type of neuron. Just to put things into perspective, here are a few different types of neurons found in the cerebellum, alone:
Cerebellum Golgi cell
Cerebellum Lugaro cell
Cerebellum Purkinje cell
Cerebellum basket cell
Cerebellum candelabrum cell
Cerebellum granule cell
Cerebellum nucleus reciprocal projections neuron
Cerebellum stellate cell
Cerebellum unipolar brush cell
Notice how I said, “a few.”
In short, the Human Connectome Project is looking at the brain in a completely different way than ever before. This is necessary, as the methods we use to study the brain today fall quite short, to say the least.
Neural connectivity can be mapped at 3 different levels: the macroscale, mesoscale, and microscale. For the sake of this article, we’ll be focusing on the microscale- that is, mapping every neuron individually.
The Brainbow Technique is kind of like the different color wires connecting to a computer; it takes advantage of the fact that if all neurons were a different color, all of the neural connections could be seen. To do this, researchers used red, blue, yellow, and green fluorescent proteins at randomized ratios inside of neurons. They do this by utilizing Cre recombinase-mediated DNA excision and DNA inversion (yes, CRISPR/Cas9 is not the only gene editing technology!!).
Cre-recombinase is an enzyme that can perform either excision or inversion of DNA flanked by Lox sites (the triangles), depending on the triangles’ orientations. Different Lox sites (Lox1 & Lox2) are mutually incompatible, meaning only the RFP/CFP pair or GFP/YFP pair is selected. If the RFP/CFP is selected by DNA excision, DNA inversion will either ‘keep it’ or ‘invert it.’ If kept, red will be expressed, or if inverted, blue will be expressed. This process is the same if the GFP/YFP pair is selected, either expressing green or yellow based on DNA inversion. The process is stochastic — completely randomized!
Multiple vectors can be used in each cell in the process described above. Using various models, neurons have been made ~100 different colors! It provides each cell with a specific barcode, and is especially helpful in cell tracing and lineage analysis, where color is used to differentiate cell populations derived from different progenitors.
And by doing this, we get some really pretty pictures!
The photograph to the left, taken by a light microscope, depicts a sulcus in the cerebral cortex of a transgenic mouse. The dots are the soma (cell bodies, called the grey matter), and the wires are the white matter.
Although the photograph is a masterpiece and is undoubtedly a major feat in the field, there’s nooooo way we’re going to be able to trace all of the axons and dendrites in the cerebral cortex using this method. The network is just too dense!
So, a giant robot machine thingy decided it would come along and save the day! (Jk… unless?)
The Automatic Tape-Collection Mechanism is this crazy machine that automatically collects sections of the brain on thick plastic film. The ultramicrotome moves brain against a diamond knife, cutting each section 30 nm so that every wire is on its own section. After floating on water, the sections are picked up by the conveyor belt. The machine is able to cut 10,300 sections of brain a day (over 100 TB of data)! …which is the size of a grain of salt…
In a sense, it creates a “tape” of the brain. This tape is then cut and pasted onto silicon wafers:
Each piece of brain is like a scene from a movie that can be put together when imaged under an electron microscope — that is, 33,333 scenes to make a cubic millimeter of a movie…
One 3D cubic millimeter of the brain made from 33,333 2D images of cut up brain… appetizing
Researchers used this data to track part of a neuron which is colored in red. By removing everything around it, this was the result.
This is an axon connecting to a dendrite it communicates with.
But here’s the issue: this was wayyyyy too much labor for such a small amount of data. Scientists had to manually color in the neuron on all 33,333 images! This is what coloring all components of a single image looks like, in a process called hand segmentation, or as I like to call, “Color in the Dots”:
1,000 cubic microns of everything around a single dendrite of a pyramidal neuron
But luckily, computers are helping us with this tedious process. The segmentation is now fully automated with the use of artificial intelligence!
The image above has a whopping 675 synapses, 530 axons, and 90 dendrites. It displays highly non-random connectivity (P<.00002), meaning some axons prefer to innervate certain dendrites. Microscopes are currently being made to increase the efficiency of this meticulous process, such as a 61 beam scanning electron microscope.
This is all great… if you want me to cut your brain into a bajillion pieces. To study the brain of a living organism on the microscale, we use two-photon excitation microscopy. Get excited! (see what I did there, hehe)
Two-photon excitation microscopy is a fluorescence imaging technique used to take pictures of living tissue up to one millimeter in depth. In short, it uses near-infrared excitation light which can excite fluorescent dyes in neurons. For each excitation, two photons of infared light are absorbed. Using infrared light minimizes scattering in the tissue, and due to absorption of more than one photon, the background signal is filtered out. This makes it possible for us to look reeeealllyyy deep into the tissue.
This is all really cool, but what we’re actually doing with all the data is another story…
So we have all this data. Great. But what exactly do we do with it?
We make these things called brain networks! Brain networks are collections of nodes (neuronal elements) and edges (the interconnections), put simply. They are made from measurements of structural or functional relationships between pairs of neurons or brain regions, depending on what the neuroscientists defines a “node” as.
The diagram to the left depicts the connection matrix, describing the network’s topology. The circles are the nodes, and the lines are the edges (A). A node is said to be “low degree” if it doesn’t have a lot of connections. A node with many connections is “high degree” (B).
A network can be further divided into two modules (C). A high degree node can either be a connector hub, maintaining connections that link different modules, or a provincial hub, one within many connections within a single module.
With the matrix we made, we can use special statistical tools from graph theory to help us! Briefly, measures of brain connectivity fall into three different categories: segregation, integration, and influence. We’ll go a little deeper on segregation and integration.
Measures of segregation analyze the extent to which nodes aggregate into separate clusters, which can be expressed by computing the network’s clustering coefficient, or by its tendency to form distinct modules.
Measures of integration quantify the ease with which communication occurs along network paths, an important factor in how nodes can exchange information key measures of integration relate to communication efficiency and path length. For more on that, click here, or read about graph theory! (Message me if you do)
Basically, the combination of high clustering (segregation) and short path length (integration) indicates the presence of “small world” network, as seen here.
We can analyze these small world networks in patients with different disorders to understand them on a neurological level
By understanding the causes of these complex disorders, we can make more effective treatments (who doesn’t want that?!)
Neural connectivity of depression, ADHD, and schizophrenia
An organism that we’ve done a ton of connectivity analyses on is, you guessed it, C. elegans!
C. elegans is a 1 millimeter nematode that we have the whole entire connectome of! Sounds impressive (because it is), but these little guys only have 302 neurons… we have 100 billion.
My favorite model organism!
Connectivity matrix of C. elegans… Damn.
Thank you, Barack
Still, we have learned a lot from C. elegans. Researchers used theoretical network control principles to predict neuron function in their connectome. If you read this, which I highly suggest you do, let’s chat about it.
None of this would have been possible without huge companies, like Google, helping with the computational methods using AI. Even Barack Obama is helping with this; he launched the BRAIN Initiative in 2013 to support technologies that will help create the human connectome.
Basically, this whole “connectomics” thing is pretty damn important. With a better understanding of the brain, we could make more effective treatments for neurodegenerative diseases like Alzheimer’s and mental illness. We could make diseases, or even memories, actually look like something!
So when we actually map the whole human connectome… could we extend it to a robot?! That, my friend, is a topic for another article.
Key Takeaways
A connectome is a map of the brain on the level of individual neurons. The Human Connectome Project is trying to map the connectome of the brain
The Brainbow technique, using Cre-recombinase, makes really pretty pictures! Buttt when it comes to the cerebral cortex, it just ain’t gonna cut it
The Automatic Tape Collection Mechanism and electron microscopy has made it possible to more efficiently create images of the brain along with fully automated segmentation
Two-photon excitation microscopy can look at neurons in living tissue
The brain is organized in “small world” networks, which we analyze using statistical and modeling techniques and graph theory
We have the full connectome of C. elegans, and it was extended to a FRACKING ROBOT!
HCP is going to revolutionize our understanding of the brain and treatments for neurological diseases and mental illnesses… a memory could look like something on the level of the neuron!