by Yijie Yin
Another milestone in Fly Connectomics has been achieved!
We are happy and *deeply* honoured to have contributed significantly to the mapping and annotation of the first synapse-resolution, full brain connectome of the fruit fly Drosophila melanogaster (‘Full Adult Fly Brain’, FAFB dataset1). The adult fruit fly is just small enough that we can get nanometre-resolution images of the entire brain, but complex enough to display interesting behaviours (e.g. courtship, learning). This connectome is made up of ~130,000 neurons, and ~50 million synapses2–7.
The main paper from our group, Schlegel et al. 20235, reports the following accomplishments:
We have annotated all neurons for broad information flow: afferent neurons go into the brain, efferent neurons go out of the brain, and intrinsic neurons are restricted to the brain. We have also assigned them to anatomical superclasses. For example, optic lobe superclasses based on projection patterns are optic (intrinsic to the ocelli or optic lobe), visual projection neurons (transmit information from the optic lobe to central brain), and visual-centrifugal neurons (transmit information from central brain to the optic lobe).
Figure 1. Hierarchy of annotations in the fruit fly connectome. Figure adapted from Schlegel et al. 2023. A. Illustration of the hierarchy of annotations in the fruit fly connectome. B. Snapshots of all neurons in each superclass.
We have assigned neurons to inferred hemilineages and morphological groups within hemilineages. ‘hemilineages’ are the discrete developmental units based on stem cell (‘neuroblast’) of origin and Notch signalling.
Figure 2. Hemilineage identification in the fly connectome. Figure adapted from Schlegel et al. 2023. A. Illustration of the developmental process, where a neuroblast typically gives rise to two hemilineages. B. The cell body fibre tracts of the identified hemilineages. Some hemilineages are named on the right.
We have annotated specific cell types, based on previous electron-microscopy datasets (such as the hemibrain8, an EM volume containing ~half an adult brain) and the light level literature. Comparing the data across three hemispheres (FAFB and hemibrain) allowed us to identify more robust and reproducible cell types than had been possible in the hemibrain.
Figure 3. Cell type annotation in the fly connectome. Figure adapted from Schlegel et al. 2023. A. Both classes (from an information flow perspective) and morphology groups (from a developmental perspective) can be further divided into cell types. B. An example cell type found consistently across individuals and hemispheres.
Finally, we analysed different aspects of variation and stereotypy. To what extent are fly brains similar to each other? To what extent can discoveries made in FAFB be used for other flies?
In addition to global quantifications, we present an interesting case, where a two-fold difference (across individuals) in the number of neurons for a cell type made little difference on the cell-type-level connectivity. Experimental manipulations had previously demonstrated that (unlike for most neural stem cells) mushroom body neuroblast cycling can continue under starvation conditions, resulting in an increased number of Kenyon cells9. We therefore suspect that starvation of the FAFB fly might underlie the two-fold difference in neuron number.
Figure 4. A surprising two-fold difference in neuron number for cell type Kenyon Cell (KC) gamma-m. Figure adapted from Schlegel et al. 2023. A. KCs are involved in the learning and memory circuit in the fruit fly: since KCs are much more numerous than ALPNs which relay sensory information from the periphery, KCs are thought to provide a rich sensory representation for learning to take place. B. Different sub-types of KCs in the mushroom body. C. Count of KC types in the currently-available hemispheres. KCg-m are twice as numerous in the FlyWire dataset compared to hemibrain. Analysis on connectivity (see Schlegel et al. 2023) revealed that despite the difference in neuron number, cell-type level connectivity remains conserved across individual flies, and individual neurons adjust the synaptic budget they occupy for their pre/post-synaptic partners accordingly.
The annotations we have produced are available in this github repository: https://github.com/flyconnectome/flywire_annotations. You can also query some of the annotations through this neuroglancer link: https://neuroglancer-demo.appspot.com/#!gs://flyem-user-links/short/2023-10-18.104728.json.
Complementing this paper, Dorkenwald et al. 20234 – on which we are also authors – describes the process of reconstructing all the neurons in the brain to produce the raw connectome; our group in Cambridge contributed about 40% of the central brain and 25% of the whole brain. That paper provides global descriptions of neuron connectivity and innervation and global information flow, as well as an intriguing interpretation of the ocelli circuit. They also highlight the ‘package’ of papers generated alongside the connectome, e.g. the integrate and file model by Shiu et al. 202310.
Try fly connectomics today!
References (main papers from our group highlighted)