We’re on the cusp of solving how all the neurons in the fruit fly brain are connected. This represents a huge milestone for the field of neuroscience and a giant leap forward in the understanding of human brains, but how?
As many people spent the summer trying to keep the flies away from their fruit-bowls, an international group of scientists published one of the biggest and most important datasets in the field of connectomics to date, the complete 3D electron micrograph volume of the fruit-fly (Drosophila melanogaster) brain.
The study represents a significant milestone in the field of connectomics, the study of how all the individual cells in the brain are connected in their circuits. It is the largest complete brain volume to be imaged at the synaptic level, and the culmination of a big international team effort.
The result is a three-dimensional architectural blueprint upon which Drosophila neuroscientists can map their circuits of interest and understand how they interact with neighboring cells and their synaptic targets. Meaning they can understand exactly which neurons are involved in a behavior or a memory, and how they connect in their circuits.
An ambitious project
I spoke to Davi Bock (@dddavi), Group Leader at Howard Hughes Medical Institute’s (HHMI) Janelia Research Campus in Ashburn Virginia, US, via webcam. Standing at a desk in his office off the lab, Davi has the pragmatic air of a man that gets things done. His office behind him is typical of the functional space and utility provided to group leaders at Janelia.
The campus is built into a hillside outside of Ashburn overlooking a man-made lake and, beyond the tree-line, the Potomac river. The modern glass-fronted building, which curves along the hillside in a shallow ‘S’ shape akin to that of a C. elegans worm fuses form, function, and purpose. Behind the façade is a brightly-lit wooden wall-lined lobby area behind which lies the auditorium. To the right the building opens into a space with scattered sofas for informal discussions among colleagues, before a small library, and a canteen behind. To the left is the heart of Janelia, Bob’s Pub, which crucially houses the free coffee and encourages scientific debate and discussion into the night. Up one of the large staircases sit the wet lab areas. Floor-to-ceiling glass-fronted spaces full of benches that fill the length of the building. One wing houses the Drosophila research labs, and the other the mammalian labs. The lack of dividing walls between the groups fosters interaction and collaboration.
The building of the campus was almost as ambitious and inspired as the scientific projects that are carried out within its walls.
As Davi reflected on the project: “The project was so ambitious, it’s an only-at-Janelia kind of project.”
A true team effort
Davi explained it was the collaborative nature of Janelia and the HHMI’s support of visionary projects that enabled the ambitious project to succeed:
“It was a true team effort, requiring input from over 20 people.”
He explained how each stage of the five-step process to obtain the three-dimensional volume from sample preparation and screening to neural reconstruction and analysis required custom-building of equipment or the overcoming of technical boundaries.
For example, a team led by Richard Fetter, then at Janelia Research Campus, now at UCSF, fine-tuned the sample preparation and then undertook the serial sectioning of over 7,000 40-nanometre-thick brain sections.
Davi’s group developed a next generation transmission electron microscope camera array, which took two years of development. The system incorporated modern cameras that can image at 100 frames per second, and a custom piezo-driven stage that could move samples under the cameras with nanometer precision.
A robotic sample handler was also developed that facilitated the automated collection of the last 20% of the imaging data. However, prior to this 80% of the data was meticulously acquired manually by co-first author Zhihao Zheng.
The image processing was another feat. As Davi explained:
“If you have an image on your computer it’s straightforward to adjust the contrast. However, the dataset contained 21 million images. To manipulate these required the input of computational experts.”
Michael Kazhdan at Johns Hopkins University, and Bill Karsh at Janelia Research Campus wrote the core image processing algorithms. Whilst Stephan Saalfeld, Khaled Khairy, and Eric Trautman at Janelia Research Campus worked to design a custom database to enable automation of the image processing pipeline.
Adding: “And then you need to do neuronal tracing on the assembled image volumes, and you need custom software to do that. But, for this we were able to benefit from CATMAID, a software that enables collaborative annotation of neurons.”
In total the project took six years from conception and initiation to publication. From beginning to end, the imaging alone took 14-16 months.
Davi estimates that each of the 23 authors on the paper spent at least one year or more of solid effort on the project, meaning even a conservative effort puts the cumulative duration of this project at 23 years in total.
Greg Jefferis (@gsxej), Group leader at the Medical Research Council’s Laboratory of Molecular Biology and The University of Cambridge’s Department of Zoology, winner of this year’s Francis Crick Medal and Lecture by the Royal Society, and one of the authors on the paper, explained his involvement in the project and the implications the findings have for the field.
Speaking to me from an office at the Champalimaud Centre for the Unknown in Lisbon, Portugal, he is visibly excited by the publication of the 3D brain volume. Framed by a floor-to-ceiling window, Greg’s mood reflects the sunny scene behind him as the Tagus River sparkles beneath a cloudless sky. Greg is bright, beaming and animated.
He explains that solving the complete functional wiring diagram of the adult fly brain is an achievable target on the near horizon as he walks me through the importance of the study and how neuroscientists the world over can harness the data for their experiments:
“The Drosophila brain is incredibly complex for its size.”
Adding: “Its poppy seed-sized brain is just 8*107 µm3 in volume and thought to contain around 100 thousand neurons, which are connected by 40 to 50 million synapses.”
Explaining: “It is also remarkably stereotyped. A neuron that exists in one position in one fly’s brain will be there in the same position in another fly’s brain. This means neurons of interest imaged using light microscopy can be mapped onto this new electron microscopy volume data. Meaning we can see with nanometer resolution where chemical synapses are formed between neurons in their circuits.”
Greg is right to be excited. His group have been investigating olfactory behaviors in Drosophila since he started his lab in 2004. They have explored the physiology and connections of neurons underlying stereotypical behaviors like aversiveness or attractiveness to smells and pheromones in incredible detail. His group can now directly relate their findings to this three-dimensional map to understand how these stereotypical behaviors are elicited.
As Greg explains: “We’re understanding even more about our data now. One of the surprises that came out of the brain volume data set is just how many chemical synapses there are between neurons. There are many more than we previously thought. However, what they mean for brain computations remains to be answered, but they must be important.”
Opening the floodgates to solving the connectome
This volume adds a large missing piece to the puzzle of fly connectome research and there have been multiple papers published since its release. The scientists having benefited from the volume to map their circuits of interest.
Their progress was also aided by the fact that the volume is free to access for anyone. You can browse the imaging data here at the Virtual Fly Brain, a website developed as a hub for sharing of Drosophila neural anatomy and imaging data. It is hoped this community-based approach will lead to faster understanding of the connectome.
Fundamentals of brain organization exist throughout nature
A lot of genes are conserved between fruit flies and humans. When the fruit fly genome was mapped in 2000 the study revealed it shares 60% of its genome with humans.
Architectural themes have also been conserved throughout biology and circuit motifs identified in the small nervous system of the tiny nematode worm C. elegans are also seen in flies and mammalian brains. The more we understand about the intricacies of the fly brain and how it is organized, the more we can understand about our own.
The problem is one of scale. As we heard from Davi, it is difficult and time consuming to produce these efforts of reconstructing brain tissue from electron micrographs of serial sections. And this explains why, in 2018, we have only just managed to obtain a volume of the whole adult fly brain, which contains about 100 thousand neurons.
In comparison the mouse brain contains around 100 million neurons. To date only a volume of mouse retina has been resolved at the synaptic level. However, this volume only contains 950 neurons, many of which are incomplete as they project beyond the edges of the volume.
The 'Machine Intelligence from Cortical Networks' (MICRoNS) project funded by IARPA seeks to revolutionize machine learning by reverse-engineering the algorithms of the brain.
To do this, a collaborative group of scientists from the Allen Institute, Baylor College of Medicine, Harvard University and Princeton University will reconstruct a 1 mm3 volume of visual cortex from the mouse brain to connect with imaging and physiological data they have already amassed.
I contacted Allen Institute project scientist Dan Bumbarger (@DanBumbarger) to find out more about this project and how it compares in scale to the Drosophila effort, as he explains:
“We are mapping out vertebrate cortical circuits at synapse resolution with serial section transmission electron microscopy, in the same way that was done in Davi Bock’s group.”
Adding: “The major thrust of the work is trying to understand computation in the part of the cortex responsible for visual information processing.”
This volume represents only a small portion of the mouse brain but is several times larger than the entire Drosophila brain. Yet the group are using lessons learned from the Drosophila effort to inform their progress and interpretation of the data, as Dan explains further:
“On the technical side, we are adapting the render image processing software developed for the fly brain work to help us with image registration on the petabyte scale. On the theoretical side, many questions generalize well between different animals. Comparisons between different species will help us understand and generalize ideas about what networks of neurons compute and how they go about doing so.”
It’s not the size that counts
Despite the advances in high-throughput electron microscopy and automated imaging processes, it will be some years before we have a full mouse brain volume reconstructed from serial electron micrographs. And we may have to wait longer still for a reconstruction of the human brain which dwarfs the complexity of the mouse brain by another order of magnitude. The human brain is estimated to be between 1200-1450 cm3 in volume, containing 86 billion neurons and a trillion synapses.
But it’s not the size that counts. The structural circuit motifs that are seen in nematode worms and fly brains are also repeated in mice and human brains. Understanding how information flows in these arrangements of neurons and non-neuronal cells will inform how our brains process information.
The fruit fly may be an annoying pest around our fruit bowls, but it could hold the key to unlocking the secrets of our brains.