Category Archives: Connectomics

The AII Amacrine Cell Connectome: A Dense Network Hub

AII-connectome

We have a new publication in Frontiers in Neuroscience, The AII Amacrine Cell Connectome: A Dense Network Hub.  Authors are Robert E. MarcJames R. Anderson, Bryan W. Jones, Crystal Sigulinsky and J. Scott Lauritzen.

Abstract:  The mammalian AII retinal amacrine cell is a narrow-field, multistratified glycinergic neuron best known for its role in collecting scotopic signals from rod bipolar cells and distributing them to ON and OFF cone pathways in a crossover network via a combination of inhibitory synapses and heterocellular AII::ON cone bipolar cell gap junctions. Long considered a simple cell, a full connectomics analysis shows that AII cells possess the most complex interaction repertoire of any known vertebrate neuron, contacting at least 28 different cell classes, including every class of retinal bipolar cell. Beyond its basic role in distributing rod signals to cone pathways, the AII cell may also mediate narrow-field feedback and feedforward inhibition for the photopic OFF channel, photopic ON-OFF inhibitory crossover signaling, and serves as a nexus for a collection of inhibitory networks arising from cone pathways that likely negotiate fast switching between cone and rod vision. Further analysis of the complete synaptic counts for five AII cells shows that (1) synaptic sampling is normalized for anatomic target encounter rates; (2) qualitative targeting is specific and apparently errorless; and (3) that AII cells strongly differentiate partner cohorts by synaptic and/or coupling weights. The AII network is a dense hub connecting all primary retinal excitatory channels via precisely weighted drive and specific polarities. Homologs of AII amacrine cells have yet to be identified in non-mammalians, but we propose that such homologs should be narrow-field glycinergic amacrine cells driving photopic ON-OFF crossover via heterocellular coupling with ON cone bipolar cells and glycinergic synapses on OFF cone bipolar cells. The specific evolutionary event creating the mammalian AII scotopic-photopic hub would then simply be the emergence of large numbers of pure rod bipolar cells.

 

A Synaptic Basis for Small World Network Design in the ON Inner Plexiform Layer of the Rabbit Retina

Bipolar cells_

This abstract was presented today at the 2014 Association for Research in Vision and Opthalmology (ARVO) meetings in Orlando, Florida by J Scott Lauritzen, Noah T. Nelson, Crystal L. Sigulinsky, Nathan Sherbotie, John Hoang, Rebecca L. PfeifferJames R. Anderson, Carl B. Watt, Bryan W. Jones and Robert E. Marc.

Purpose: Converging evidence suggests that large- and intermediate-scale neural networks throughout the nervous system exhibit small world’ design characterized by high local clustering of connections yet short path length between neuronal modules (Watts & Strogatz 1998 Nature; Sporns et al.2004 Trends in Cog Sci). It is suspected that this organizing principle scales to local networks (Ganmor et al. 2011 J Neurosci; Sporns 2006 BioSystems) but direct observation of synapses and local network topologies mediating small world design has not been achieved in any neuronal tissue. We sought direct evidence for synaptic and topological substrates that instantiate small world network architectures in the ON inner plexiform layer (IPL) of the rabbit retina. To test this we mined ≈ 200 ON cone bipolar cells (BCs) and ≈ 500 inhibitory amacrine cell (AC) processes in the ultrastructural rabbit retinal connectome (RC1).

Methods: BC networks in RC1 were annotated with the Viking viewer and explored via graph visualization of connectivity and 3D rendering (Anderson et al. 2011 J Microscopy). Small molecule signals embedded in RC1 e.g. GABA glycine and L-glutamate combined with morphological reconstruction and connectivity analysis allow for robust cell classification. MacNeil et al. (2004 J Comp Neurol) BC classification scheme used for clarity.

Results: Homocellular BC coupling (CBb3::CBb3 CBb4::CBb4 CBb5::CBb5) and within-class BC inhibitory networks (CBb3 → AC –| CBb3 CBb4 → AC –| CBb4 CBb5 → AC –| CBb5) in each ON IPL strata form laminar-specific functional sheets with high clustering coefficients. Heterocellular BC coupling (CBb3::CBb4 CBb4::CBb5 CBb3::CBb5) and cross-class BC inhibitory networks (CBb3 → AC –| CBb4 CBb4 → AC –| CBb3 CBb4 → AC –| CBb5 CBb5 → AC –| CBb4 CBb3 → AC –| CBb5 CBb5 → AC –| CBb3) establish short synaptic path lengths across all ON IPL laminae.

Conclusions: The retina contains a greater than expected number of synaptic hubs that multiplex parallel channels presynaptic to ganglion cells. The results validate a synaptic basis (ie. direct synaptic connectivity) and local network topology for the small world architecture indicated at larger scales providing neuroanatomical plausibility of this organization for local networks and are consistent with small world design as a fundamental organizing principle of neural networks on multiple spatial scales.

Support:  NIH EY02576 (RM), NIH EY015128 (RM), NSF 0941717 (RM), NIH EY014800 Vision Core (RM), RPB CDA (BWJ), Thome AMD Grant (BWJ).

Synapse Classification And Localization In Electron Micrographs

Synapse-classification_

We have a new publication, Synapse Classification And Localization In Electron Micrographs in Pattern Recognition Letters.  Authors are: Vignesh JagadeeshJames Anderson, Bryan W. JonesRobert MarcSteven Fisher and B.S. Manjunath.

Abstract:  Classification and detection of biological structures in Electron Micrographs (EM) is a relatively new large scale image analysis problem. The primary challenges are in modeling diverse visual characteristics and development of scalable techniques. In this paper we propose novel methods for synapse detection and localization, an important problem in connectomics. We first propose an attribute based descriptor for characterizing synaptic junctions. These descriptors are task specific, low dimensional and can be scaled across large image sizes. Subsequently, techniques for fast localization of these junctions are proposed. Experimental results on images acquired from a mammalian retinal tissue compare favorably with state of the art descriptors used for object detection.