Surprising Data About Visual Connections in the Brain Revealed by ‘Brainbow’

Neuroscientists know that some connections in the brain are pruned through neural development. Function gives rise to structure, according to the textbooks. But scientists at the Virginia Tech Carilion Research Institute have discovered that the textbooks might be wrong.

The research is in Cell Reports. (full open access)

Research: “Multiple Retinal Axons Converge onto Relay Cells in the Adult Mouse Thalamus” by Sarah Hammer, Aboozar Monavarfeshani, Tyler Lemon, Jianmin Su, and Michael Andrew Fox in Cell Reports doi:10.1016/j.celrep.2015.08.003

Image: Using a technique dubbed “brainbow,” the Virginia Tech Carilion Research Institute scientists tagged synaptic terminals with proteins that fluoresce different colors. The researchers thought one color, representing the single source of the many terminals, would dominate in the clusters. Instead, several different colors appeared together, intertwined but distinct. Credit: Virginia Tech.

“Brainbow” zebra fish.

Neurons are labeled in multiple colors with Brainbow fluorescence microscopy. Three fluorescent proteins (cyan, yellow and red) are randomly taken up by various neurons, offering a palette of dozens of colors to help scientists follow complex neural pathways. Shown here is a five-day-old zebra fish larva viewed from the dorsal side, captured using a 20X objective.

Image credit: Dr. Albert Pan, Harvard University, Cambridge, Mass., U.S.

The human brain is estimated to have over 100 billion neurons. Each neuron connects hundreds to thousands of times, creating a dense, complicated network of over 100 trillion connections. Scientists have the obvious challenge of mapping neuronal networks in a forest where all the trees and roots look identical. To differentiate between individual neurons in a network, researchers have developed the Brainbow technique. Brainbow labels each neuron with a specific color, allowing researchers to follow the pathways from one individual cell to the next. Seen here is the Brainbow technique used in a mouse brain.

Image by Mr. Pierre Mahou, Dr. Emmanuel Beaurepaire, and Dr. Karine Loulier, Ecole Polytechnique.

The “Brainbow” technique is applicable to many cell types, including he muscle fibers of the tongue. Just as in the brain, the fluorescence is generated by the “Brainbow” transgene randomly recombined by the Cre/lox system. Different cell types handle the recombination step differently, leading to a unique mixture of fluorescent proteins and the colorful pattern.

Image: Here the intercrossed muscle fibers in the tongue of a mouse embryo (day 14) are labeled with random combinations of the fluorescent proteins dTomato (red), YFP (green) and Cerulean (blue). This image displays the maximal intensity projection of a confocal image stack (20x 0.8 NA oil objective).

Mapping the entire brain with new and improved Brainbow II technology

Among the many great talks at the recent annual meeting of the Society for Neuroscience were three special lectures given sequentially during the evenings. The first described how we might translate the known circuit diagram of the worm, and the range of neural activities it supports, into it’s play in a 2D world. The second followed with how we might trace the trickle of information from the larger 3D world, through the more complex theater of the fly brain, and back out again. The third, and most gripping story in the trilogy, was Jeff Lichtman’s talk about using his new technology—known as Brainbow II— to turn the wild synaptic jungle into a tame and completely taxonomized arboretum which we can browse at our leisure.

A movie of a millimeter-sized worm learning to recognize and wriggle free from a mini-lariat may not be the critics choice. However, considering that the critical neurons and synapses involved in this particular behavior can now be genetically isolated, and watched in detail, many neurobiologists are fairly excited. We still don’t have whole-brain electrical activity maps for the 302 neurons (and 50 glial cells) in this creature, or even high resolution calcium clips of these cells—but that may not be required. Many neurons do not bother to use discrete spikes when they are only sending signals across short distances, and sometimes they don’t even bother to build axons.

In this case, if we want to understand how the worm acquires the lariat escape trick, perhaps we might instead just watch its mitochondria as their host neurons stir in seeming alarm. Indeed if we were to watch nothing but mitochondria, most of what we might learn about a given neuron through the use of a whole host of other imaging technologies, is already contained within their dynamics. One could probably infer not just the membraneous outlines of a neuron by watching the limits of mitochondrial excursions, but also infer the changes in the shape of the individual neurites. Further in this vein, we also now appreciate that mitochondria don’t just respond to the calcium flows mentioned above, they are in fact calcium-controlling organelles by trade.

One thing that we learned from Brainbow I, which was further highlighted with the expanded palette of Brainbow II, is that labeling everything can be as bad as labeling nothing at all. Part of Brainbow II’s feature set, is more control for the selective labeling of synapses from different kinds of interneurons, and also the processes of glial cells. In order to reap the benefits of Brainbow II technology and create detailed computer reconstructed images of these cells, Lichtman’s group had to build high speed brain slicing and processing instruments, as well as high power electron microscopes to create the images.

Lichtman reported that together with Zeiss, a new high-throughput 61-beam scanning electron microscope is currently under development. This massive device does not look like something that could just be slid into an elevator and sent to a fourth-floor lab. I asked @zeiss_optics about pricing and availability on this behemoth, along with focused ion beam attachment, and they said that they are offering a nice rebate on orders of two or more. Even still, the result of many months of protected effort has thus far only yielded the structure of just a small piece of brain.

But what a structure it is. The crowning achievement, shown at the convention was distilled into a cylindrical EM reconstruction of a piece of mouse brain smaller than a grain of sand. In the center of this volume was the proximal shaft of a pyramidal cell apical dendrite surrounded by all manner of synaptic elements. If you were ever confounded by the famous 4-color mapping thereom, then Brainbow-style synapse tracing may not be for you. In this volume there are around 680 nerve fibers that can be resolved, together with 774 synapses. A key finding by Lichtman is that mere contact alone, does not a synapse make. By tracking perfectly resolved synaptic vesicles, he was able to show that of every ten plausible synaptic options, perhaps only one or two neighboring profiles turned out to be an actual synapse.

The final point Lichtman made is that now that it is possible to extract the complete membrane topology, including organelles, of an arbitrary region of the brain, formerly unimagined questions might be posed and answered with the click of a mouse. The question he alluded to is the one I raised above, namely, how are the mitochondria distributed, and what are they doing? While this is in large part, a question for live, video microscopy, much can be learned about the state of a given synapse just prior to being fixed by it’s mitochondria. Similarly, much might be also be inferred about the next plausible state of the neural geometry under consideration, provided one knows what to look for.

The one finding here that Lichtman mentioned was that axons have relatively small mitochondria compared to those in the body and dendrites. That may be a seemingly sterile finding when considered alone. But that same afternoon at the conference, there was an exciting talk describing how certain mitochondria are extravasated, or expelled, by axons in the visual system. They are then taken up by astrocytes for processing—a rather surprising finding. It has been known that in some organs mitochondria can be exchanged between cells, much to the benefit of the recipient cell, though for neurons, this is the first report of such phenomena. I did look later at the literature, and this fractionation of mitochondria by size in the polar elements of neurons has actually been known for some time, leading one to guess what other potential findings the Lichtman group might actually possess.

What Lichtman presented is really not a connectome, or a “netlist” of circuit board connections, per say. To date, nobody has even put force a reasonable transform to derive a connectome from a given 3D membrane mesh topology, or even of what use it would be if we had one. Meanwhile, attempts to model the fissions, fusions, and general ramblings of the mitochondria as a function of their genetic makeup, and the positions they take up inside the cell, have already begun. If genetically questionable mitochondria with expired membrane potentials tend to be degraded by fusion with lysosomes near the nucleus, we might ask, can they be blamed for pumping out axons and transporting themselves as far away as possible—even out of the cell entirely?

Clearly, anthropomorphizing mere motile sacks of DNA and enzymes is not the only tool we have to hack the brain. But insofar as the brain is just a complex system of microscopic tubes, it may make sense to take a closer look at the creatures that build and maintain them. In this light, the science of connectomes becomes the science of mitochondria, the mitochondriome perhaps. As much as we can better understand the collective activity of the brain through the remembrance of neurons as once-feral protists now encased in the skull, our understanding of neurons is enhanced by recalling their mitochondria as once-free bacteria now largely trapped in them.


The human brain is the most complex object in the known universe. But what about a fruit fly’s? While only the size of a pinhead, it’s still pretty motherf'ing complicated. Which is why a team of scientists at the Howard Hughes Medical Institute used an ingenious technique called “Brainbow” to make their job easier, revealing fine neural structures with unprecedented clarity.

Before Brainbow, scientists interested in tracing the structural connections between neurons could only color-code them one or two at a time using crude dyes. That meant slicing, staining, and combining many separate samples to build up a map of even tiny portions of a brain. Brainbow uses genetically engineered fluorescent proteins to make the neurons color-code themselves, right in the brain, in up to 100 different glowing colors. This makes it easier for scientists to clearly map the overall structure and tangled mass of connections, and also zero in on tiny individual areas of interest. Call it neuroscience by way of Massimo Vignelli.

Brainbow had already been demonstrated on lab mice, so why use it on lowly drosophila melanogaster instead of a bigger, “cooler” animal? As Technology Review explains, “these organisms have a very sophisticated set of existing genetic tools, [so] researchers can exert even greater control over when and where the fluorescent proteins are expressed.” Which means better, more informative pictures, and better, more informative science. (The researchers’ modified color-coding techniques for the fruit fly are called dBrainbow and Flybow.)

More photos over at Co. Design.

Oculomotor nerve

The oculomotor nerve controls the eye muscles that allow our pupils to focus light, eyelids to blink, and eyes to track moving objects. The bundle of oculomotor nerves in this image is from a Brainbow mouse, which was genetically engineered to randomly color neurons different hues.

Some brain injuries like concussions can be difficult to diagnose quickly in an emergency room or on the sidelines of sports. Recently, researchers at NYU Langone Medical Center developed a new technology that uses eye tracking as a readout for brain injury. They tracked the eye movements of healthy and brain-injured patients as they watched music videos (including “Under the Sea” from The Little Mermaid and Shakira’s “Waka Waka”) and could accurately diagnose brain injury and severity. While the study was small, they hope larger studies will make the technology more sensitive and will help validate other brain injury tests.

Image by Dr. Katie Matho, Dr. Jean Livet, and Raphaëlle Barry/Nikon Small World.

Brainbow: Neurons are works of art. 

Introducing a modified version of green fluorescent protein into the genomes of mice resulted in the synthesis of proteins that fluoresced in different colors. Because each neuron is labeled with a distinct color, the pathways of neural axons can be traced to their destinations.

So this is an amazing thing we talked about in bio the other day. It’s a method called Brainbow. It’s really hard to track neurons when they can get so long and tangly, as you can see in this light microscopy of a mouse hippocampus, so scientists can mark each neuron with a different color of fluorescence via transgenes that produce fluorescent proteins. What a brilliant solution! 

Neurons in a zebrafish embryo

Zebrafish have proven invaluable for understanding what we know about nerves and the brain. Observing brain development and interrogating how growing neurons find their correct targets are possible thanks to the transparent, genetically malleable nature of zebrafish embryos. Recently, scientists have developed a technique called “Brainbow” that individually colors each neuron, allowing researchers to map the start and end points of neural circuits. Applying Brainbow to zebrafish will allow researchers to visualize how neurons connect with one another during development and how different diseases disrupt this process.

Image by Dr. Albert Pan, Harvard University.


BrainBow is a technique where cells are made to express several fluorescent proteins, in essentially random amounts. The randomness derives from feedback loops in gene expression. Mixing of fluorescence wavelengths yields a remarkable colour contrast on the single-neuron level.

The method was originally developed by Jeff W. Lichtman and Joshua R. Sanes at the Department of Neurobiology, Harvard Medical School.

Read more about BrainBow on Wikipedia or an introduction at the Harvard Gazette.


Purkinje neurons reach their arbor-like dendrites into the molecular layer of the developing cerebellum of a mouse.

Credit: Tamily Weissman, Harvard University. The original Brainbow mouse was produced by Livet J, Weissman TA, Kang H, Draft RW, Lu J, Bennis RA, Sanes JR, Lichtman, JW. Nature (2007) 450:56–62.