CNS*2018 workshop, Seattle, July 17, Allen Institute for Brain Science

4th workshop on
Neuronal morphology and structure

Alexander Bird (Ernst Strüngmann Institute and FIAS, Frankfurt)
André Castro (Ernst Strüngmann Institute and FIAS, Frankfurt)
Hermann Cuntz (Ernst Strüngmann Institute and FIAS, Frankfurt)



Hollis Cline (Scripps Research Institute) keynote

Ruth Benavides-Piccione (Instituto Cajal, Spain)
Staci Sorensen, (Allen Institute, USA)
Erik De Schutter (OIST, Japan)
Uygar Sümbül (Allen Institute, USA)
Lida Kanari (EPFL, Switzerland)
Casey Schneider-Mizell (Janelia Research Campus, USA)
Kurt Haas (University of British Columbia, Canada)
Sophie Laturnus (Universität Tübingen, Germany)
Hongkui Zeng (Allen Institute, USA)

Hollis Cline: In vivo time-lapse imaging analysis of neuronal structure and functional plasticity

Ruth Benavides-Piccione: The microanatomy of pyramidal cells
Pyramidal neurons are the most abundant and characteristic neuronal type in the neocortex and their dendritic spines are the main postsynaptic elements of cortical excitatory synapses. In turn, pyramidal cell axons constitute the main source of these synapses. Thus, our understanding of the synaptic organization of the neocortex largely depends on the knowledge available regarding synaptic inputs to pyramidal cells. Most of the present knowledge has been obtained from experimental animals resulting in considerable voids in the information available for the microanatomy of the human brain. However, systematic light microscopy methods can now be applied to human tissue (biopsy or autopsy) that offer images of extraordinary quality. This is the case for intracellular injection in fixed material using markers like Lucifer Yellow. The morphology can be visualized with fine morphological details (e.g., dendritic spines) and 3D reconstructions of the dendritic arbors can be performed in the human cortex. Although the process of data acquisition and reconstruction is time-consuming, recently available semi-automatic software and implementation of specific tools, allow for the human pyramidal cell dendritic architecture, including the position and detailed morphology of their dendritic spines along the apical and basal arbors, to be reconstructed in 3D. This large volume of data available on the microanatomy of pyramidal cells will help to model and recognize these neurons allowing their virtual creation in different cortical areas and layers. In particular, since dendritic spines are key elements in learning, memory and cognition we believe it is of great importance to build models which include these structures.

Casey Schneider-Mizell: The neuroanatomy of connectivity in the Drosophila larva
Neuronal morphology is an integral part of the development and function of neuronal circuits, but simultane- ously measuring detailed neuroanatomy and synaptic connectivity remains hard. Recent advances in electron microscopy methods have enabled nanometer-scale mapping of the structure of large populations of whole neurons and their connectivity in small organisms such as the Drosophila larva. Through mapping and analysis of the larval sensorimotor and adult fly visual systems, we demonstrated that, in Drosophila, dendrites make numerous small, microtubule-free 'twigs' that are the dominant sites of synaptic input. Across developmental stages of the larva, we found that even though no new neurons are born, neurons increase in both arbor size and synaptic count. However, in central neurons this change is dominated by the growth of new postsynaptic twigs relative to backbone growth. In first-order nociceptive processing neurons, five-fold increases in size, number of terminal dendritic branches, and total number of synaptic inputs were accompanied by cell type-specific connectivity changes that preserved the fraction of total synaptic input associated with each pre-synaptic partner. Taken together, our results describe the cellular neuroanatomy underlying synaptic connectivity in Drosophila and suggest that precise patterns of structural growth act to conserve the computational function of a circuit, for example determining the location of a dangerous stimulus.

Erik De Schutter: The Purkinje cell dendrite causes its unique firing rate-dependent phase response curve
The electrophysiological properties of Purkinje cells (PCs) have been well explored, but less is known about how PC dendrites regulate somatic firing properties. To unmask the interaction between the soma and dendrites of PCs, we have built a new PC model, which is validated by a plethora of experimental data. We use this model to explore the phase-response curves (PRCs) of PCs, which get larger and broader at higher firing rates. We find that PC dendrites determine the firing rate-dependent changes of the PRC. We further test the effect of firing rate-dependent PRCs on oscillations in the cerebellar PC layer and find that larger and broader PRCs at higher firing rates increase the oscillation power density in a sparsely connected PC network model.

Kurt Haas: Dynamic morphometrics: Rapid time-lapse imaging and quantification of experience-driven dendrite growth
The computations a neuron performs are determined by the synaptic inputs it receives, their spatial arrangement along its dendrites, and the structure of the dendritic arbor. However, it remains poorly understood how functional arrangements of synapses and dendritic morphology arise during development. We have developed platforms for in vivo rapid time-lapse imaging of dendritic arbor growth and activity in the awake developing brain, and paradigms inducing experience-driven structural and functional plasticity. Our software, Dynamo, has been particularly useful for tracking dendritic growth behaviors across short intervals over long periods to generate rich data sets which describe how dynamic growth culminates into long term patterning.

Lida Kanari: Randomness and structure in artificially generated neuronal networks
The rodent cerebral cortex consists of a few millions neurons that are highly connected by a complex network of synapses. Many factors contribute to the generation of structural connections of neurons, among which are the geometrical constraints of the neuronal morphologies (such as the shape and their relative positions in space). However, the precise effect of the anatomical constraints on the neuronal connectivity are not well understood. To what extent is the structure of the neuronal network encoded in the genetic information of an organism and to what extent do the connectivity patterns emerge from a combination of stochastic events and interactions between growing structures? To address this question, we have designed a simple generative model based on the theory of random walks that enables the study of fundamental mathematical and physical properties of interacting morphologies. We thereby assess which anatomical properties of neurons are crucial in the formation of brain networks and illustrate that stochastic interactions play a significant role in the generation of the complex connectivity patterns within these networks.

Sophie Laturnus: A systematic comparison of neural morphology representations for cell type discrimination
The morphology of neurons is typically considered a defining feature of neural cell types. For example, 14 types of bipolar cells can be discriminated in the mouse retina based on their morphology (Helmstaedter et al. 2013, Kim et al. 2014, Greene et al. 2016), leading to a classification in good agreement with genetic and physiological data (Shekhar et al. 2016, Franke et al. 2017). Similarly, many types of retinal ganglion cells or cortical interneurons can be discriminated based on morphological properties (Sumbul et al., 2014, Jiang et al 2015). Given recent advances in automatic reconstruction and crowd-based tracing techniques, the amount of available data is rapidly increasing (see e.g. However, a systematic comparison of feature representations to automatically classify neurons using machine learning techniques is missing. A number of different feature spaces have been used in the past, ranging from expert defined summary statistics (Scorcioni et al. 2008) over neurite density maps (Helmstaedter et al. 2013, Jiang et al. 2015) to graph theoretical or topological concepts such as tree-edit distance (Heumann et al. 2009, Gilette et al. 2009) and persistence Li et al. 2017). This diversity of concepts, each validated on its own data set, make it difficult to judge which representation is the most useful to distinguish cell types within a class of neurons. Here, we systematically compare the performance of machine learning classifiers building on different feature representations using data sets from the retina and visual cortex of mice. We investigate which morphological representations allow reliable discrimination of neural types within defined cell classes. We find that expert defined features and 2-dimensional persistence diagrams allow to predict class labels best across all data sets.

Staci Sorensen: Morphological, electrophysiological and transcriptional descriptions of cortical cell types
Understanding the diversity of cell types in the brain has been a major challenge in neuroscience. One approach to solving this problem has been to characterize salient biological properties of individual neurons, including their electrophysiology, morphology and gene expression, and to use them to define cell types. We established a single cell characterization process using standardized patch clamp recordings in brain slices and image- based neuron reconstructions to measure intrinsic physiological and morphological properties of hundreds of cortical neurons in the adult mouse and human brain. For this talk, I will focus mainly on the morphological descriptions of excitatory and inhibitory neurons across all layers of the cortex. I will relate these findings to electrophysiological types and gene expression patterns aiming for a more comprehensive description of cell types.

Uygar Sümbül: Quantifying neuroanatomy
Brain cells display fascinating shapes and positioning patterns. Despite the importance of this diversity for brain function, its organizing principles are not well understood. To obtain a set of cell descriptors, researchers traditionally identify the features that explicitly cover multiple aspects of the anatomical variability. This approach typically produces a heterogeneous set: different features may have different units, and it is not clear how to assign relative importances to those features. Here, I will talk about our attempts at quantifying neuroanatomy and objectivizing comparative studies without explicitly calculating classical features.

Hongkui Zeng: Morphology as a key feature for neuronal cell type classification
To understand the function of the brain and how its dysfunction leads to brain diseases, it is essential to have a deep understanding of the cell type composition of the brain, how the cell types are connected with each other and what their roles are in circuit function. Neuronal cell types need to be defined by a combination of molecular, morphological and physiological properties. Furthermore, the full extent of single neuron morphologies provide critical information about how neural signals are organized and transmitted across the brain to their target regions. At the Allen Institute, we have built multiple platforms, including single-cell transcriptomics, single and multi-patching electrophysiology, 3D reconstruction of neuronal morphology, and brain-wide connectivity mapping, to characterize the transcriptomic, physiological, morphological, and connectional properties of different types of neurons in a standardized way. In particular, we obtain neuronal morphologies from both biocytin-filled neurons after patch-clamp recordings in brain slices and fluorescently labeled neurons through fMOST whole mouse brain imaging. We perform quantitative analysis of morphological features for cell type classification. We correlate morphological features with electrophysiological and/or gene expression features at single cell level using Patch-seq or multiplexed FISH techniques. These efforts are leading to the creation of a cell type taxonomy for the mouse cortex which lays the foundation towards the understanding of circuit function.