The central complex consists of three main neuropiles — the protocerebral bridge (PB), the ellipsoid body (EB, Central Body Lower in other insects) and the fan-shaped body (FB, Central Body Upper in other insects) — and at least three accessory neuropiles — the noduli (NO), gall (GA) and lateral accessory lobe (LAL) (Figure \ref{288889}A and \citealt{Wolff2015,Lin2013,Hanesch1989}). One of the most striking neural elements of the central complex are the columnar neurons, which innervate one of the eighteen (in Drosophila) glomeruli of the PB, one vertical section of either the FB or EB, and one accessory neuropile — a column being constituted by the PB glomeruli and FB/EB section. A total of 12 different columnar cell types have been described, with stereotypical correspondences between the PB glomerulus and the EB/FB section. In addition to these "principal cells", there are a number of neurons innervating multiple columns of one neuropile. These neurons often innervate subdivisions orthogonal to the columns. Moreover they sometimes also project to neuropiles outside the central complex . This set of neurons includes the ring neurons, which innervate a ring within the EB and an accessory neuropile, and a collection of inputs and interneurons with processes in the FB and PB. From this light level anatomy and putative synaptic polarity, one can derive a hypothetical general information flow diagram of the central complex (Figure \ref{288889}) :
- Ring neurons provide input to the EB columnar neurons.
- Recurrent connections between EB columnar neurons form and sustain a ring attractor for heading direction
- Information is transferred from the EB columnar system to the FB columnar system via the PB (interestingly, only one columnar neuron type displays presynaptic terminals in the PB, the E-PGs)
- FB columnar neurons also receive inputs in the FB
- All columnar neurons but the E-PGs also receive inputs in the PB
- Interneurons in the PB and FB further interconnect the columns
- All accessory structures are potential outputs
We show that this overall flow of information is generally supported functionally for the parts we have tested so far, but with a few potentially important differences: the observed connectivity in the PB is sparse, rendering the function of PB interneurons possibly critical; accessory structures are usually input rather than output areas; and, consequently, output channels of the CX are scarce.
Results
Cell types considered
\citep[see][]{Jenett_2012}, we picked driver lines for central complex cell types based on the strength of expression and the sparseness of the pattern in the central complex. At this stage, the 37 driver lines (for 24 cell types) cover the main columnar neuron types and PB interneurons \cite{wolff_neuroarchitecture_2015}, three types of ring neurons, a LAL-FB neuron and neurons innervating accessory structures, namely a LAL interneuron and three types of neurons connecting the LAL to the noduli. Neuron types are schematized in Figure \ref{831433}A and Supplementary Figure \ref{147797}. The drivers used and the nomenclature are listed in Table \ref{786393}. Throughout this paper we follow the naming convention set out in \cite{wolff_neuroarchitecture_2015} for full names, and abbreviated following the scheme described in \cite{Kakaria2017} and used in \cite{Green2017} and \cite{Turner_Evans_2017}. At the time of writing this manuscript, the dataset explores inputs to the EB system, connections between EB columnar neurons, connections in the PB as well as potential inputs and outputs in the LAL, Gall and noduli. Connectivity of different cell types within the FB have not been explored.
A functional connectivity screen
For any given pair of neurons, the overlap between pre- and post-synaptic looking regions was assessed based on publicly available expression patterns (\citealp{Tirian_2017,Jenett2012}, see Figure \ref{831433}B) digitally aligned on a common reference brain (as described in \citealt{Aso_2014}). For each combination selected, one or several genotypes were prepared to express CsChrimson and GCaMP6m in the potential pre- and post-synaptic partners, respectively (Figure \ref{831433}C, see Materials and Methods). Brains and VNCs were dissected out and tested using a standardized protocol (see Figure \ref{831433}D, and Materials and Methods). The light stimulation intensities were kept at low levels (50uW/mm2, delivered in trains of 2 ms pulses) to reduce the chances of observing indirect connections, while at the same time allowing us to see near-saturating responses in strongly connected pairs (see Supplementary Figure \ref{723169}). Whenever large responses were observed pharmacology was used to check that observed transients were synaptically mediated, and to narrow down the neurotransmitters involved (Supplementary Figures \ref{488013} and \ref{841071}).
We picked driver lines for functional connectivity mapping by visually inspecting the Janelia Gal4-driver collection