The key to any successful connectomics effort is data quality. Nowhere is this truer than electron microscopy-based efforts, whether they are TEM, FIBSEM or Serial Block Face SEM. The ideal sample should be of high quality and faithfully represent the biology, maximize the information content, and facilitate downstream automated segmentation and reconstruction. Based on previous large-scale TEM imaging efforts at Janelia (1st instar brain and the single optic lobe medulla column imaged by the FlyEM project), and images of intact adult fly brains using conventional methods, it is clear that future efforts to generate an EM-based fly brain connectome will require significant improvements in membrane staining and detection of synapses, ways of reducing the number of processes that cannot be traced (‘orphans’), and unambiguous visualization of synaptic junctions in the sample. The focus of the lab is to develop tools and methods to resolve these issues.
Improving Membrane and Synapse Staining
Membrane staining intensity in EM is directly related to heavy metal content. Since the majority of membrane staining is due to the reaction of osmium tetroxide with unsaturated fatty acids in membrane lipids during fixation, an obvious approach to differentially increase membrane staining is to optimize the osmium tetroxide fixation step. The ‘reduced osmium’ method widely used on mammalian nervous tissue is known to increase the relative staining of membranes and reduce that of the cytoplasm. Unfortunately, in flies it reproducibly generates membrane structural artifacts. Our lab is working to develop and optimize a ‘fly’-specific reduced osmium method that mimics the effect of the mammalian method on membranes, with the goal of facilitating automated segmentation of processes.
Synapse detection can be substantially improved in automated detection schemes by enhanced staining of the presynaptic density characteristic of most fly synapses, the ‘T-bar’. A side benefit of the effort to increase membrane staining, has been differentially enhanced staining of T-bars relative to the cytoplasm. Another project in our lab is to continue this optimization of T-bar staining.
Transmission EM-based methods require additional heavy metal staining of the sections on support films prior to imaging. This adds 2 elements of risk, the potential for breaking the support film by accidental poking during handling and the deposition of staining artifacts on the sections. Another project in our lab is to develop new methods to further increase heavy metal staining of samples in order to produce samples that DO NOT require additional post-sectioning staining. For a serial section series of the adult fly brain expected to require ~7500 sections on ~2000 grids, eliminating the additional handling and potential staining artifacts of this many grids would be a huge reduction in risk for the project.
Facilitating Segmentation and Reducing Orphan Processes
Typically in EM reconstructions, there are processes that cannot be connected to their cell of origin. In conventionally prepared fly brains, ‘orphan’ processes arise when smaller caliber portions run obliquely in a plane of the section. Their continuity in adjacent sections is obscured and lost by more strongly stained processes surrounding them. An obvious way to address the orphan problem is to add some separation between processes. In mammalian nervous tissue, 15-20% of the volume is known to consist of extracellular space. This natural separation of neuronal processes can be preserved with specialized sample preparation methods. Our lab has developed methods for preserving extracellular space in fly CNS preparations while maintaining excellent ultrastructure. The additional separation between processes facilitates tracing of thin processes, helps distinguish true synaptic partners from processes next to the synapse merely by chance, and highlights areas of neuronal processes known to be in close proximity by function, such as gap and septate junctions.
Tools to Increase the Information Content of an EM Connectome
In collaboration with the Simpson lab at Janelia, our lab is working to develop tools to increase the information content of an EM-based fly brain connectome through direct visualization. For example, knowing the neurotransmitter used by each cell, and thus to sign of its synapses, would be a huge. Unfortunately, the simple morphology of synaptic vesicles does not provide this information in Drosophila. Using the Simpson lab’s HRP reporter constructs targeted to different intracellular compartments, in conjunction with the GAL4-UAS controlled expression, our lab is developing methods to increase the information content of samples while preserving excellent ultrastructure. One such construct, neuronal synaptobrevin-HRP, is targeted to the lumen of synaptic vesicles. The final electron dense product remains confined to the lumen of the vesicle and is easily identified without obscuring other cellular structures. Expressing this UAS-nSyb-HRP reporter with a neurotransmitter synaptic vesicle transporter GAL4 driver, could be used to directly identify all the cells that express the neurotransmitter in a whole brain reconstruction at nanometer resolution. Being able to label all cells of a specific neurotransmitter type, or all excitatory or inhibitory neurons, would provide valuable information for a connectome.
Using cell specific drivers, it is also possible to determine, at the EM level, whether pairs of cells suspected of being part of a circuit, actually make synaptic contact. Our lab is working to develop such pre-embedding double labeling strategies with excellent morphology using the GAL4-UAS and LexA-LexOp systems.