Neurotransmitter Classification from Electron Microscopy Images at Synaptic Sites in Drosophila Melanogaster

Abstract

High-resolution electron microscopy (EM) of nervous systems enables the reconstruction of neural circuits at the level of individual synaptic connections. However, for invertebrates, such as Drosophila melanogaster, it has so far been unclear whether the phenotype of neurons or synapses alone is sufficient to predict specific functional properties such as neurotransmitter identity. Here, we show that in Drosophila melanogaster artificial convolutional neural networks can confidently predict the type of neurotransmitter released at a synaptic site from EM images alone. The network successfully discriminates between six types of neurotransmitters (GABA, glutamate, acetylcholine, serotonin, dopamine, and octopamine) with an average accuracy of 87% for individual synapses and 94% for entire neurons, assuming each neuron expresses only one neurotransmitter. This result is surprising as there are often no obvious cues in the EM images that human observers can use to predict neurotransmitter identity. We apply the proposed method to quantify whether, similar to the ventral nervous system (VNS), all hemilineages in the Drosophila melanogaster brain express only one fast acting transmitter within their neurons. To test this principle, we predict the neurotransmitter identity of all identified synapses in 89 hemilineages in the Drosophila melanogaster adult brain. While the majority of our predictions show homogeneity of fast-acting neurotransmitter identity within a single hemilineage, we identify a set of hemilineages that express two fast-acting neurotransmitters with high statistical significance. As a result, our predictions are inconsistent with the hypothesis that all neurons within a hemilineage express the same fast-acting neurotransmitter in the brain of Drosophila melanogaster.

The connectome of the adult Drosophila mushroom body provides insights into function

“…MBON names are color-coded based on neurotransmitter. MBONs whose neurotransmitter assignment is based on prediction ( Eckstein et al, 2020 ) are underlined. …”

“…The percentage indicates the fraction of individual synapses scored that agreed with the overall prediction. See Eckstein et al, 2020 for details. Note that predictions for acetylcholine are particularly robust.…”

“…For each of the atypical MBONs we provide a figure supplement ( Figure 8-figure supplements 1-14) that provides information on its top inputs and outputs as well as a video that displays additional morphological and connectivity features. We used newly developed machine vision methods (Eckstein et al 2020;Methods) to predict the neurotransmitters of these MBONs ( Figure 6-figure supplement 2), as the specific GAL4 drivers that would be required to determine transmitters by antibody staining or RNA profiling were not available. As these MBONs are described here for the first time, no experimental data yet exists on their function(s) or physiology.…”

Serotonergic modulation of visual neurons in Drosophila melanogaster

“…Serotonergic release sites appear to lack synaptic partners in the lamina of the blowfly Calliphora [125] and our data using sybGRASP to test whether serotonergic neurons synapse onto L2, T1, and L1 was negative ( S8 Fig). Recent advancements such as the full adult fly brain EM dataset [126] combined with predictive neurotransmitter mapping [127] will be important resources to determine the precise relationship between serotonergic boutons and the nerve terminals within the medulla neuropil. Further studies will also be needed to define the function of putative serotonin autoreceptors which we and others detect in serotonergic neurons (S5 Fig, S6 Fig and Fig 3B) [56,89].…”

Whole-animal connectome and cell-type complement of the three-segmentedPlatynereis dumeriliilarva

“…In Platynereis, for example, we know the neurotransmitter or neuromodulator content only of a small fraction of neurons (Figure 4) as these molecules are invisible to EM. However, artificial convolutional neural networks were recently used to predict synaptic transmitter content in the Drosophila adult brain from EM images alone (Eckstein et al, 2020). This approach still has to be tested in other organisms including Platynereis.…”