NeuroMorpho.Org - from Conception to Big Data

Since the first release 7 years ago,more than 2.5 million digital reconstructions of neuronal morphology have been downloaded from NeuroMorpho.Org (approximately half a billion branches or 16 km of neuropil length, roughly equivalent to 350 person-centuries of manual tracing) in the course of over 100,000 unique visits from 140 countries. These data have resulted in more than 200 publications on topics ranging from neuronal development to circuit mapping, from synaptic integration to network dynamics, and from comparative analyses to computational models. All along, dendritic and axonal tracings downloaded from the database have also been extensively employed as educational resources in and outside of the classroom to illustrate and quantify the variability and diversity of neuron types within and across brain regions, species, and experimental conditions. As neuroscientists continued to trace neurons and increasingly recognized the potential impact of their reconstructions beyond their originally intended purpose, the content of NeuroMorpho.Org has grown by an order of magnitude, from ~1000 to ~10,000 reconstructions. Such an exemplary success story in neuroscience data sharing unfolded through tremendous community participation. Synergistic interactions of experimentalists, curators, modelers, and tool developers fostered the emergence of a vibrant "ecological niche" of research around digital reconstructions of neuronal morphology. The most recent technological breakthroughs have already started generating effective, high-throughput automated reconstruction algorithms that will soon produce a game-changing explosion of tracing data. Harnessing the full power of this massive amount of data will require progressive automation of the all-important curation process.

Molecular Genetics of Retinal Ganglion Cell Type differentiation

Using conditional knock-in reporters targeted to each of three transcription factors, Brn3a, Brn3b and Brn3c, we are able to describe the dendritic arbor structure and axonal projections of a variety of Retinal Ganglion Cell (RGC) types to the retinorecipient areas in the brain. Using various combinations of genetic tools we then study the roles of these neurons within the neuronal circuit, by ablating with increasing specificity individual cell types, and assessing the effect of these manipulations on the neuronal circuit. For instance, we were able to define the requirements of different Brn3b expressing RGC types in a variety of visual circuits sub serving distinct visual functions, such as Vestibulo-Ocular Response and Pupil Constriction.

In addition, we are investigating at high resolution the function of these transcription factors in the regulation of RGC development, and cell type definition. We find that Brn3s are expressed in overlapping, but distinct RGC populations and in a similar combinatorial fashion in a wide array of projection sensory neurons (Dorsal Root Ganglia, Auditory Spiral Ganglion, etc.). It thus appears that Brn3s participate in a developmental program aimed at differentiating projection sensory neurons, a function which was also documented for the orthologues of Brn3s in Drosophila (Acj 6) and C. Elegans (Unc 86). Ablation of individual Brn3s has very distinct effects on RGC development, implying that these transcription factors may have distinct functions. We have previously documented axonal arbor defects in RGCs upon loss of Brn3b, and dendritic arbor defects in populations or individual RGCs from which Brn3a had been deleted, but a code composed of three transcription factors is unlikely to account for the generation of 20 or more RGC cell types. In order to study how the genetic interactions between Brn3s participate in the generation of RGCs, we have generated pair wise double knock-outs of Brn3a, Brn3b and Brn3c, and report the phenotypes of these mice with respect to RGC development, and the effects on the expression of other transcription factors important for RGCs. In parallel, we are taking advantage of our reporters to label RGCs at early stages of development and follow their fate under wild type and mutant circumstances.

Reduced axonal fields of dentate/hilar border interneurons following chronic NR2B inhibition and its impact on GABAergic input to dentate granule cells
Y. Wang, D. Lapides, K. Quinn, S.B. Bausch

We showed previously that chronic treatment with distinct classes of N-methyl-D-aspartate type glutamate receptor (NMDAR) antagonists exerted differential effects on seizure-like events (SLE). SLE were dramatically reduced in organotypic hippocampal slice cultures treated chronically (17-21 days) with the NR2B-selective antagonist, Ro 256981, but were increased in cultures treated with the non-subtype-selective antagonists, memantine or D-APV. Since GABAergic transmission plays a critical role in the regulation of hippocampal activity, we hypothesized that distinct classes of NMDAR antagonists would have differential effects on GABAergic circuitry and be inversely related to our previous SLE data. To investigate potential changes in interneuron morphology that may contribute to differential changes in SLE, interneurons at the dentate/hilar border were filled with neurobiotin and digitally reconstructed. Treatment with Ro25,698, but not the other NMDAR antagonists dramatically reduced total axonal length and sphere of axonal arborization. Axonal branch points and putative boutons mirrored these results, but branch point and bouton density were unchanged. Dendritic length and sphere of dendritic arborization also were reduced following chronic Ro25,6981, but this reduction was not specific to NR2B inhibition as other NMDAR antagonists showed similar reductions. Double immunofluorescence for GAD 65/67 (GABA synthetic enzymes) and NeuN (for somata) or MAP2 (for dendrites) was then performed to quantify neuron number and numbers of putative GABAergic synaptic boutons apposed to granule cell somata and dendrites, respectively. The data showed the number of neurons and putative GABAergic synapses onto granule cells were lower following treatment all NMDAR antagonists except Ro25,6981 compared to vehicle. Similar decreases have been described in epilepsy models and may contribute to the increased SLE expression in slice cultures following chronic treatment with D-APV and memantine. In contrast, we found equivalent numbers of putative GABAergic synapses onto granule cells in vehicle- and Ro25,6981-treated cultures, despite reduced axonal fields in individual interneurons and no change in the number of GABAergic neurons following chronic Ro25,691. These changes together with electrophysiological results suggest that one possible mechanism for reduced SLE following chronic Ro25,6981 is that individual interneurons may have a reduced ability to synchronize activity across widespread populations of neurons. [Supported by CDMRP award W81XWH-04-1-0065/PR030035 and NINDS grant NS045964]

Amygdala dendrite morphology and spine patterning in a selected mouse line of fear resilience and susceptibility
Vitor de Castro Gomes, Hadley C Bergstrom, Jennifer McGuire, Jennifer Coyner, Anna Sharova, Luke R Johnson

Introduction:The lateral amygdala (LA) is a key site for the establishment of mammalian associative fear memories. Remodeling of dendrites and spines in the LA has been linked with a variety of fear and anxiety phenotypes. Here we investigated how genetically determined differences in fear memory consolidation relate to dendrite complexity and spine density and structure in the LA. Methodology: Lines were derived from an F8 advance intercross of C57BL/6J and DBA/2J mice (originally from Dr A Palmer, University of Chicago, IL) and then bi-directionally selected for 4 generations (S4) based on the level of the conditioned freezing response (fear susceptible (FS) versus fear resilient (FR) mice) in a Pavlovian fear conditioning procedure. LA dendrites and spines of S4 naïve FS (n=5) and FR (n=5) mice were visualized using a Golgi-Cox stain. Reconstruction of dendritic morphology was restricted to principal neurons located in the dorsal, ventrolateral and ventromedial subdivisions of the LA. Dendrite volume was also measured. We segregated the apical and basilar trees based on the length of the arbor. Spine density and morphology (i.e., thin, mushroom, stubby) was recorded from all branch orders of the dendritic tree. Digital reconstructions were conducted in three dimensions using Neurolucida software (MBF Bioscience). Preliminary Results: Sholl analysis was conducted on dendrites to assess the complexity of the tree and on spines to determine their distribution along both basal and apical segments. Overall, no differences in dendritic structure were detected between FS and FR lines. For spines, there was a greater density located on both the basilar and apical dendritic tree (p<0.05) in the FR compared to FS line mice. These preliminary results show that basal alterations in the patterning of spines, but not dendrites, segregated with the fear memory phenotype. Conclusions: These findings suggest that alterations in the intrinsic synaptic connectivity of the LA may underlie and possibly contribute to genetically determined differences in fear memory consolidation.

Dendritic structure of layer V pyramidal neurons projecting to the primary visual cortex of the mouse

Cortical pyramidal neurons are recognizable by their characteristic dendritic morphology. However the basic layout of dendrites exhibits a wide range of variation of dendritic length, branching structure and spine density. More specifically, layer V pyramidal neurons exhibit a range of dendritic morphologies, even within a single cortical area (Tsiola et al. 2003). The dendritic arborization of these neurons appears to be related to their connections and by the cortical area to which they belong. For instance, in monkeys, it was shown that callosal projection neurons possess longer apical and basal dendrites and more dendritic spines than ipsilaterally projecting neurons (Soloway et al. 2002) and that the complexity of the dendritic tree increases with the hierarchical level of cortical areas of a same sensory modality (Elston and Rosa 2000). Cortical pyramidal neurons are often classified by features of their apical dendrites. However, this parameter alone does not account for the structural complexity of the entire dendritic arborization. The developmental pattern of layer V pyramidal neurons strongly suggests that there are likely four distinct functional dendritic compartments namely the basal dendrite, apical trunk, oblique dendrite, and tuft dendrite (Romand et al. 2011). Therefore, the use of several dendritic morphological parameters, pertaining to the apical, oblique and basal dendrites, may provide a more comprehensive classification of layer V pyramidal neurons. Mathematical tools that ordinate objects in multidimensional space, such as the principal component analysis (PCA) and that objectively group similar objects using multiple descriptors such as cluster analysis are appropriate for this approach. These analyses have previously been used to demonstrate the existence of distinct subgroups of layer II/III (Benavides-Piccione et al. 2006), V (Tsiola et al. 2003), and VI (Chen et al. 2009) neurons in the mouse cerebral cortex. To study the morphology of the entire dendritic arbors of individual layer V pyramidal neurons that project onto V1, we injected an adenovirus that expresses enhanced green fluorescence protein (EGFP) under a synapsin promoter (AdSynEGFP) in V1 in C57BL/6 mice. With this technique, a Golgi-like retrograde labeling of complete dendritic arbors was achieved (Tomioka and Rockland 2006; Ichinohe et al. 2008). Complete 3D reconstructions of dendritic arbors of retrogradely labeled layer V pyramidal neurons were performed for neurons of the primary auditory (A1) and somatosensory (S1) cortices and from the lateral (V2L) and medial (V2M) parts of the secondary visual cortices of both hemispheres. The morphological parameters extracted from these reconstructions were subjected to principal component analysis (PCA) and cluster analysis. The PCA showed that neurons are distributed within a continuous range of morphologies and do not form discrete groups. Nevertheless, the cluster analysis defines neuronal groups that share similar features. Each cortical area includes neurons belonging to several clusters. We suggest that layer V feedback connections within a single cortical area comprise several cell types.

Deciphering the Functional Organization of Cortical Circuits through Cell-Type Identity

Pyramidal neurons represent the majority of neurons in the neocortex. Each class of pyramidal cell sends long range axons to a distinct set of cortical and subcortical brain regions and represents a different information channel of the cortex. To understand the synaptic interactions of pyramidal cells classified by the long-range targets of their axons, we asked whether or not local intracortical connectivity among pyramids reflects their long-range axonal targets. By retrogradely labeling pyramids projecting to different brain regions and recording simultaneously from up to four labeled pyramids in cortical slices, we demonstrated that the probability of synaptic connection depends on the functional identity of both the presynaptic and postsynaptic pyramids. By reconstructing the intrcortical axonal and dendritic arbors of the different classes of pyramids, we further showed that the average axodendritic overlap of the presynaptic and postsynaptic pyramids could not fully account for the differences in connection probability that we measured physiologically. Taken together, these results show that each pyramidal cell type participates in different cortical circuits that generate their unique functional properties.

Old Tricks Lead to New Insights in How Sensory Activity Shapes Neuronal Structure in Mouse Barrel Cortex
Joshua C. Brumberg^1,2 and Chia-Chien Chen^2
1 Department of Psychology, Queens College, CUNY ² Neuropsychology Sub-Program (Psychology), The Graduate Center, CUNY

Over a century after its invention the Golgi technique can still be utilized to gain insights into neuronal structures. The Golgi technique was used to determine the characteristics of neurons in baseline conditions as well as how their structure is impacted by sensory activity in the mouse barrel cortex. Specifically, neurons were reconstructed in control animals and in response to sensory deprivation induced by whisker trimming (every other day from birth). Analyses of adult neurons using unsupervised statistical techniques showed that based on quantitative morphological features it is possible to conclusively classify neurons into distinct groups. To investigate how morphological features react to changes in sensory experience we deprived animals of their normal sensory inputs via trimming their prominent facial whiskers every other day for the first post-natal month. In response to sensory deprivation pyramidal neurons in layer 6 showed retraction of their apical dendrites and elaboration of their basilar dendrites, while layer 6 non-pyramidal cells showed elaboration of their dendrites. Focusing on the dendrites of pyramidal cells revealed that sensory deprivation increased spine density in layer 4 but decreased it in layer 6. Restoring sensory inputs by subsequently allowing the whiskers to regrow for a month only restored the normal spine density in layer 6 not layer 4. Taken together, the results indicate that sensory deprivation has profound impacts on neuronal structures in a laminar dependent pattern and that “classic” techniques can still provide novel insights on neuronal structure and function.

The anterior olfactory nucleus/cortex: a simple model system for understanding forebrain cortices
Rachel B. Kay and Peter Brunjes

Brain function will only be understood by unraveling the roles of the neurons that make up its component circuits. Considerable effort has been made to categorize neocortical neurons. Examinations have revealed that excitatory cells can be divided into subtypes based upon their projection targets as well as their physiology and morphology (e.g., Kumar and Ohana, J. Neurophys.2008). The diverse population of inhibitory cells has been categorized on the basis of morphology, molecular markers, and physiology (e.g., Ascoli et al., Nature Rev Neurosci. 2008).

Due to its relative simplicity and high degree of lamination, the olfactory system provides an attractive region for studies of brain organization. Odor information encoded by sensory neurons in the olfactory epithelium is relayed to the olfactory bulb (OB) where it is processed and then distributed via the lateral olfactory tract to olfactory cortex. The circuitry of the caudal-most olfactory cortex, piriform cortex (PC), has been studied in detail (e.g., Suzuki and Bekkers, J. Neurosci.2011). The region includes pyramidal cells as well as several species of interneurons resembling those in neocortex.

A second region of the olfactory cortex, the anterior olfactory nucleus/cortex, has received much less attention. Located between the OB and PC, the AON is the first region to receive OB input and is well positioned to influence activity in the entire olfactory circuit. It provides feedforward projections to the proximal dendrites of pyramidal cells in the PC and thus is able to control their activity. The AON also feeds back to every stage of processing in the OB and to the AON in the contralateral hemisphere. The AON is composed of two subdivisions. Pars principalis, a large ring of cells encircling the olfactory peduncle, has hallmarks of a simple cortex: it has an outer plexiform layer and an inner cell zone containing pyramidal cells. Pars externa is a small band of neurons found at the junction with the OB. If the AON is a rudimentary cortex, then it should have cell types similar to those seen in more well examined areas. The present study provides the first comprehensive examination of the morphology, electrophysiology, and molecular phenotype of neurons in the AON. Two types of projection neurons, deep and superficial pyramidal cells, were identified on the basis of different physiological signatures. Five subtypes of GABAergic cells were recognized based upon firing patterns, morphology and antigen expression: multipolar fast spiking, multipolar regular firing, neurogliaform superficial, neurogliaform deep, and horizontal cells. Cell types in pars externa shared both similarities and remarkable differences with those in pars principalis. Neocortical correlates were found for all of the cell types in pars principalis. The simplicity of the region may provide an important model for examining general rules of cortical organization.
[Supported by Grant DC-000338 from the National Institutes of Health]

Working with morphologically realistic models in NEURON

The NEURON simulation environment offers many powerful software and GUI tools that simplify creating and working with neuronal models based on detailed orphometric data. This presentation surveys tools, strategies, and workflow for importing morphometric data (Import3D), specifying biophysical properties (CellBuilder), automating spatial discretization (d_lambda), exploring model properties (ModelView, Impedance class), NeuroML interoperability (CellBuilder, ModelView), and prototyping small networks and generating cell class definitions for use in large scale network models (CellBuilder, Network Builder).

The investigation of 3 synaptic inputs on the frog spinal motoneuron by different methods including 3 – D computer reconstruction of the individual connections
N.M.Chmykhova¹, H.-P.Clamann^2
1Sechenov Institute of Evolutionary Physiology and Biochemistry, Russian Academy of Sciences, St-Petersburg, Russia; ²Institute of Physiology, Bern University, Switzerland;
As the final common path of control for the contraction of skeletal muscles, spinal cord motoneurons integrate information from various sensory, descending, supraspinal and intraspinal pathways. In mammals it has been difficult to define the intrinsic organization of the spinal networks with the clarity and precision that has been possible for work performed on primitive vertebrates. In this work we present some results produced by comparing structural and functional parameters of frog lumbar motoneuron inputs.

The connections of three lumbar motoneurons with dorsal root afferent, reticulospinal and propriospinal fibers were recorded by means of pairs of intracellular microelectrodes filled with a marker (HRP or neurobiotin) solution in the isolated brainstem - spinal cord of frogs.
The recording of the monosynaptic dorsal root afferent, reticulospinal and propriospinal EPSPs showed that the mean values of EPSP amplitudes were 154 , 735, 836 μV ; rise-time (RT) – 2.7, 2.7, 2.4 ms and half-width (HW)– 14.3, 6.4, 7.5 ms, respectively. After EPSP recording, the marker was injected into the connected axon-motoneuron pairs to reveal the neuronal trees and their contacts. After histochemical procedures, light microscopic investigation and 3-D computer reconstructions (Eutectic Neuronal System) from the serial sections were performed.

All three motoneurons had rich dendritic trees. The dendrites filled an ellipsoid space whose long axis was directed rostrocaudally. It was found that the number of revealed putative contacts of the dorsal root afferent, reticulospinal and propriospinal fibers with the motoneurons were about 20 ( 19, 19, 21, respectively). All were located at the branches of dendritic trees in the spinal cord gray matter. The dorsal root afferent fiber contacted the dorsal dendrites; the other afferents terminated on the ventromedial dendrites. In each connection 1, 4, and 6 fiber collaterals made contacts with 2, 4, 2 motoneuron dendrite trees respectively. Synaptic bouton density per mm of collateral was 74, 62 and 64. The size of synaptic boutons was 1.7, 1.3 and 2.0 μm.

Quantal analysis of transmitter release at the individual pairs of afferent fiber – motoneuron with binomial model was used to estimate the release parameters which characterize the labeled synapses and to compare these parameters to the morphological and statistical analysis of these three afferent fiber – motoneuron connexions. Binomial probability of release (0.29, 0.59, 0.55) amplitude of unit potential (32, 67, 128 μV) and number of binomial units (16, 19, 12) were calculated.

The comparison of the obtained functional and structural parameters of the three inputs on lumbar motoneurons was carried out and discussed. Unlike previous assumption, it was found that in the frog spinal cord the fiber contacts were located only at the motoneuron dendrites. As in previous studies from our lab, (Babalian and Chmykhova, 1987; Dityatev et al., 2001) a rather close correspondence between numbers of contacts and numbers of binomial units was noted. The detailed analysis of the afferent fiber structure and function should be performed in further studies, as it could be shown that axonal operations are important determinants for information processing.

Complete reconstruction of axonal and dendritic processes of adult-born neurons reveals highly organized integration
Kurt A. Sailor, Gerald J. Sun, Qasim Mahmood, Nikhil Chavali, Kimberly M. Christian, Hongjun Song and Guo-li Ming

Adult neurogenesis occurs primarily in two discrete regions of the mature brain. One of these, the subgranular zone of the dentate gyrus, resides within the hippocampal formation and newborn neurons in this region are believed to contribute to various cognitive and affective behaviors. Adult born hippocampal granule cell neurons undergo a dramatic series of developmental steps to integrate into mature circuitry. Previous studies examined their electrophysiological properties, primarily focusing on their perforant pathway input and dendritic morphological structure. Due to technical hurdles, reconstruction of the complete granule cell axonal trajectories was not feasible. We developed a novel technique which involved 2-photon serial end-block imaging (SEBI) of agarose embedded brains. This technique cycled between vibratome sectioning and imaging to reconstruct the whole hippocampus. Custom post-acquisition stitching and normalization software was developed for seamless reconstruction. As a result, we were able to show, for the first time, the developmental stages of morphological development in adult-born neurons, including complete dendritic and axonal structures of individual neurons.

miRNome analyses of class-specific dendrite development via high-throughput systems neurobiology assays and neuromorphometry

While microRNAs (miRNAs) have recently emerged as critical post-transcriptional modulators of gene expression in neuronal development, very little is known regarding the roles of miRNA-mediated regulation in the specification of cell-type specific dendritic complexity. The dendritic arborization (da) sensory neurons of the Drosophila peripheral nervous system offer an excellent model system for elucidating the molecular mechanisms governing class specific dendrite morphogenesis and for exploring miRNA-mediated control of this process. To facilitate functional analyses of miRNA regulation in da neurons, we have conducted whole-genome miRNA expression profiling as well as mRNA expression profiling of three distinct classes of da neurons, thereby generating a comprehensive molecular gene expression signature within these individual subclasses of da neurons. To further validate the role of the significantly expressed miRNAs in directing dendritic architecture, we conducted a genome-wide UAS-miRNA phenotypic screen using live-image confocal microscopy to directly assess the effect of over/mis-expression of individual and clustered miRNAs on neurons of varying dendritic complexity. Through this approach, we have identified numerous miRNAs with previously unknown functions in dendritic development, including the K box family of miRNAs. Both gain-of-function and loss-of-function analyses, via miRNA sponge transgenes, reveal that K box miRNAs repress the expression of genes required to restrict dendritic branching complexity in da neuron subclasses. Moreover, we have implemented an integrative bioinformatic analysis approach involving inverse correlation between miRNA and mRNA expression profiling data in combination with existing target prediction algorithms to identify putative target of miRNAs in regulating da neuron dendritic development. Finally, we have developed a quantitative neuromorphometry data processing pipeline for high-throughput, semi-automated dendritic digital reconstructions.

Physiology and morphology of Dbx1-derived respiratory rhythm-generating neurons of the preBötzinger Complex of neonatal mice

Breathing in mammals depends on an inspiratory-related rhythm that is generated by glutamatergic neurons in the preBötzinger complex (preBötC) of the lower brainstem. A substantial subset of putative rhythm-generating preBötC neurons derives from a single genetic line that expresses the transcription factor Dbx1, but the physiological and morphological properties that contribute to rhythmogenesis remain incompletely understood. To elucidate these mechanisms we comparatively analyzed Dbx1-derived (Dbx1+) neurons and unrelated Dbx1- neurons in the preBötC. We performed whole-cell patch-clamp recordings in newborn mouse slice that retain the preBötC and spontaneously generate respiratory related motor rhythms in vitro. Biocytin was deposited intracellularly to image the recorded neurons offline and produce digital reconstructions. We show that Dbx1+ neurons activate earlier in the respiratory cycle and discharge greater magnitude inspiratory bursts compared to Dbx1- neurons. Furthermore, Dbx1+ neurons require less input current to discharge spikes (rheobase) in the context of network activity. The expression of intrinsic membrane properties indicative of A-current (IA) and hyperpolarization-activated current (Ih) tends to be mutually exclusive in Dbx1+ neurons. In contrast, there is no such relationship in the expression of currents IA and Ih in Dbx1- neurons. Confocal imaging and digital morphological reconstruction of recorded neurons revealed dendritic spines on Dbx1- neurons, but Dbx1+ neurons were spineless. Dbx1+ neuron morphology was largely confined to the transverse plane whereas Dbx1- neurons projected dendrites to a greater extent in the parasagittal plane. The putative rhythmogenic nature of Dbx1+ neurons may be attributable, in part, to a higher level of intrinsic excitability in the context of network synaptic activity. Furthermore, Dbx1+ neurons exhibit a functional morphology that may facilitate temporal summation and integration of local synaptic inputs from other Dbx1+ neurons, taking place largely in the dendrites, which could be important for initiating and maintaining bursts and synchronizing activity during the inspiratory phase of the respiratory (breathing) cycle. This work may be significant, in part, because the morphological properties of a specific neuron class can be directly analyzed in the context of a behaviorally relevant activity pattern in vitro.

Regional differences in intrinsic excitability and dendritic morphology of medium spiny neurons during stages of habit learning
Evans RC, Hawes S, Benkert EA, Gillani F, Unruh BA, Arshadi C, Cadima GJ, Santa-Maria L, Blackwell, KT

The learning of a habit requires many repetitions of a task. This learning process is thought to be encoded in the neurons and pathways of the basal ganglia, within which the dorsomedial striatum is primarily engaged early in learning, and the dorsolateral striatum is primarily engaged later in learning (Yin et al., 2009). Here we investigate how the transition between strategies in procedural learning is reflected in neurons in the dorsomedial and dorsolateral striatum. Specifically, we test whether changes in neurons’ intrinsic excitability and dendritic morphology correspond to maze performance strategies typical of early/attentive and late/habit learning stages.
We find that dorsomedial neurons are more intrinsically excitable than dorsolateral neurons, and that this difference is most prominent in the early trained animals which use an attentive strategy to find a food cup. This medial-lateral excitability difference is much less pronounced in untrained mice or mice which have transitioned to using a habit strategy. The same neurons from which these excitability measures were taken were filled with biocytin and stained with a DAB reaction. These cells were then digitally reconstructed using Neurolucida. We found that the dorsomedial neurons had a lower spine density than the dorsolateral neurons, while there was no difference in dendritic length between regions. However, this medial-lateral spine density difference does not appear to be dependent on length of training or strategy used. The lower spine density (indicative of fewer synaptic inputs) in combination with the higher intrinsic excitability of medial neurons could indicate a homeostatic compensatory mechanism is at work.
Our results suggest that changes in synaptic activity during task-learning differentially alter the intrinsic excitability of medial and lateral striatal neurons, but do not alter spine density.

Neuronal reconstructions as a tool to study cell-specific synaptic transmission and neuromodulation in the neocortex

Morphological classifications of both excitatory and inhibitory neurons are a prerequisite to understanding their functional role in the synaptic microcircuitry of the neocortex. In order to obtain a quantitative rather than qualitative classification of neurons digital reconstructions of their dendritic and axonal domains are necessary.

The somatosensory barrel cortex and its columnar architecture provides an ideal framework for a correlated structural and functional description of both excitatory and inhibitory synaptic connections. Based on the outlines of the barrel column it is possible to identify neurons by their axonal morphology, i.e. whether they possess on average a predominantly intralaminar, translaminar, intracolumnar or more lateral projection pattern. Together with electrophysiological and/or immunocytochemical characteristics this allows a classification neuronal cell types. Such a classification is often required because functional and structural properties of synaptic connections, e.g. the release probability, the synaptic efficacy, the neuronal connectivity ratio and the sub-cellular target regions of synaptic contacts dependent strongly on the pre- and/or postsynaptic neuronal cell types.

In addition, the sensitivity to neuromodulators such as acetylcholine (ACh) appears to depend on the neuronal cell type. For example, ACh is a neuromodulator released in the neocortex during states of wakefulness and attention. Pyramidal cells in layer 2/3 and 5 of the neocortex are depolarised by ACh acting on muscarinic receptors while excitatory layer 4 spiny neurons show a persistent hyperpolarising response. This suggests a layer- and neuron-specific responsiveness to ACh. We also observed a neuron-specific ACh response with a single cortical layer: In the barrel cortex, thalomocortically-projecting layer 6 pyramidal cells are depolarised by ACh while those projecting corticocortically show a hyperpolarising response. The thalamocortical and corticocortical layer 6 pyramidal neurons have very distinct axonal domains. The axon of the former is largely confined to a barrel column while that of the latter exhibits long-range horizontal projections. Thus, the different muscarinic ACh response types are highly correlated to the morphological layer 6 neuron subtype. Since the hyper- and depolarising responses are mediated by different muscarinic receptor types this suggesting a neuron-specific expression of receptors.

I my talk I will present several examples of synaptic connectivity, functional properties and neuromodulatory effects that dependent on the neuronal cell type. They underscore the importance of digital neuronal reconstructions since this enables us to use quantitative parameters for the identification and classification of cortical neuron types.

Biophysical mechanisms influencing the firing of grid cells and place cells
Michele Ferrante and Michael E. Hasselmo

We are implementing realistic biophysical models of neurons in the entorhinal cortex and hippocampus, supported by experimental data from our laboratory, to investigate how the brain encodes spatial information. The medial entorhinal cortex (mEC) (Hafting et al., 2005; Brandon et al., 2011) and hippocampus (HC) (Morris et al., 1982; Eichenbaum, 2004) exhibit distinct spatially-modulated cellular activity when an animal is randomly foraging in an open field. HC place cells fire in precise locations (place fields), while mEC grid cells fire in a triangular array of locations throughout the environment. Understanding how grid cells and place cells compute the coordinate system for spatial representations may provide new etiological insights for disorders impairing spatial coding for episodic memories (e.g. Alzheimer’s disease and Topographical Disorientation).

Intrinsic properties (i.e. neuronal morphology and ion channels) may contribute to dorsal-ventral differences in mEC and HC firing properties that metrically encode space. Data from our lab and others shows that the distance between grid fields in mEC increases for grid cells at different dorsal-to-ventral anatomical positions correlating with stellate cells resonance properties (Giocomo et al. 2007), and morphological properties (Garden et al., 2008). Additionally, Ih knockout in mice produces bigger place fields and grid fields (Hussaini et al., 2011; Giocomo et al., 2011). Using these and other experimental constraints we are implementing realistic biophysical models of grid cells and place cells.

Our models indicate how differences in neuronal morphology, ions channels and sensory cues for spatial location may affect basic firing field properties (size, location, and firing frequency). A model based on interactions of oscillations predicted that the difference in spacing could arise from differences in the intrinsic oscillation frequency of entorhinal neurons. One variant of the model (Hasselmo, 2008) uses velocity modulation of neurons showing rhythmic persistent spiking in the entorhinal cortex (Fransen et al., 2006; Tahvildari et al., 2007; Yoshida et al., 2008) and in the postsubiculum (Yoshida&Hasselmo, 2009). Newer versions of the model have overcome the problem of variance in single cell oscillations or persistent spiking (Zilli et al., 2009) by enhancing reliability via network interactions between spiking neurons (Zilli&Hasselmo, 2010). Another recent model has combined oscillations with network attractor dynamics to generate the spatial periodicity of grid cells (Hasselmo&Brandon, 2012). Models of the generation of grid cell firing from head direction input have been incorporated in network models of episodic memory function that associate events in an episode with their position along a spatio-temporal trajectory that links the events in the episode (Hasselmo et al., 2009; 2012). Our novel results show how resonance amplitude is affected by neuronal morphology, and how ion channels affect resonance properties. Furthermore, our simulations suggest how realistic neuronal morphology allows dendrites to respond as independent computational units (i.e. to rhythmic synaptic input) while somatic interactions of oscillations can generate location-dependent firing of grid cells. Finally, we suggest possible mechanisms for shifting attractor dynamics based on a voltage phase shift (relative to sinusoidal inputs) that depends on the membrane holding potential, and beat patterns arising from the interaction of rhythmic glutamatergic and GABAergic synapses.

Branching Patterns of Neuronal Morphologies Revealed by Sequence Analysis Techniques
Todd A Gillette, Parsa Hosseini, Sridevi Polavaram, Giorgio A. Ascoli

The diversity in neuronal tree morphology lends itself to quantification by a wide variety of morphological analyses, from whole tree to branch level, geometric to topological (i.e. branching structure), and univariate to Sholl-like analyses. Many of these analyses have aided in the understanding of the relationship between morphology and electrophysiology, signal integration, growth mechanisms, and connectivity. However, characterizing neuronal branching in terms of averages over bifurcations (in metrics such as partition asymmetry) or as a function of distance from the soma provide only rough descriptions of the complexity of these tree structures. Here we present a method for analyzing the branching patterns of entire neuronal arbors and looking for stereotypical signatures of particular arborization and cell types. To this end, neuronal morphologies from the NeuroMorpho.Org database were converted to character sequences based on the local branching properties of bifurcations. The resulting strings were then analyzed using sequence analysis techniques to find the patterns that distinguish cell classes. This method, combined with cluster analysis, showed clear pattern distinctions between axons, pyramidal apical dendrites, and basal and non-pyramidal dendrites, as expected. Moreover, substantial differences were found between pyramidal apical dendrites of different brain regions and between pyramidal axons of different species. Significant but smaller differences were also found in other class comparisons. The substantial pattern differences between classes are distinct from, and in some cases larger than, differences in standard morphometrics. In many cases combining pattern differences with such metrics increases the distinctiveness of neuronal classes as determined via cluster analysis. Thus, branching patterns are stereotyped in certain neuronal classes, and those features are not captured in any other previously used morphometric.,br> [Supported by NIH R01 NS39600]

Advances in Neuron Tracing and Digital Reconstruction

The state-of-the-art in 3D neuron tracing and reconstruction is reaching an exciting juncture. Advancements in computer methods and technology are meeting with the technological innovations in microscopic imaging methods to produce more robust and anatomically accurate digital reconstructions of neurons than ever before. New software, capable of automatically tracing dendrites, axons, dendritic spines, and synapses through image datasets hundreds of gigabytes in size is now a commercial reality. Neurolucida, the mainstay of neuroscience laboratories worldwide for neuron tracing, has been steadily evolving, year-by-year, to increase the throughput of neuronal analysis via improved automation and imaging methods. Our AutoNeuron technology now achieves unprecedented accuracy in both fully-automated and user-guided automated methods, dramatically increasing the efficiency of digital reconstruction of neuronal morphology.
Neurons are among the most complex biological structures in existence, making automated reconstruction a challenge. In addition, the different labeling and staining techniques in use, along with the variation in microscopic imaging methods, makes the development of robust software to work well with these many diverse image sets even more challenging. MBF, with more than twenty years of recognized achievements, is focused on creating methods that will accommodate the diverse and evolving needs of researchers engaged in the analysis and quantification of neuronal morphology. Furthermore, the neuroscientist’s need for accuracy means that combining these digital neuron reconstruction techniques with consistent and accurate microscopic imaging methods to yield optimal results is critical.
This presentation will highlight some of the latest capabilities in neuron morphological analysis, combining microscopic data collection and automated tracing algorithms with confocal and brightfield specimens. It will also provide a glimpse of some of the upcoming developments, soon to be available.

The Open Source Brain repository: enabling collaborative development, sharing and critical evaluation of complex neuronal and network models

Computational modelling is important for understanding how brain function and dysfunction emerge from lower level neurophysiological mechanisms. However, computational neuroscience has been hampered by poor accessibility, transparency, validation and reuse of models. The Open Source Brain (OSB) repository will address these issues. By providing models in a standardised format, OSB will allow their detailed properties and metadata to be exposed in a transparent and accessible form. OSB will also provide the software infrastructure required to collaboratively develop and critically evaluate models, ranging from single cells to detailed 3D microcircuits and brain regions.

We are combining advanced open source technologies for tracking, annotating and combining models developed across research teams, with software for building, validating, visualising, simulating and analysing models. This builds on our previous work developing software applications for model construction and a widely used standardised model description language . OSB will benefit from close interaction with other important neuroinformatics resources like NeuroMorpho, ModelDB and NeuroElectro.

By increasing the scientific rigour of model construction, improving their robustness and transparency and lowering technological barriers, OSB will increase the power of computational approaches and make them accessible to a wider range of neuroscientists

Automatic Morphological Reconstruction of Neurons from Multiphoton and Confocal Microscopy Images
I.A. Kakadiaris, A. Santamaria-Pang, D. Jimenez, P. Hernandez, D. Labate, M. Papadakis

The challenges faced in analyzing optical imaging data from neurons include a low signal-to-noise ratio of the acquired images and the multi-scale nature of the tubular structures that range in size from hundreds of microns to hundreds of nanometers. In this talk, I will present a computational framework (ORION) for an automatic, three-dimensional (3D) morphological reconstruction of live nerve cells. These reconstructions can serve as virtual platforms for performing computationally guided experiments simulating neuronal function with direct validation from imaging. To accomplish this long term goal we propose a multi-step approach for producing a 3D geometrical representation of a nerve cell. The key aspects of this approach are: (i) detection of neuronal dendrites through learning and predicting 3D tubular models, and (ii) extraction of the dendritic/axonal centerline.

To carry out these tasks our group has developed a number of algorithms for segmenting a neuron from its background and for extracting the dendritic/axonal centerline. We employ a probabilistic segmentation algorithm employing multiscale image analysis for binarization. Centerline extraction is performed with a variety of algorithms such as a Minimum Shape-Cost (MSC) Tree or with a variation of Dykstra’s algorithm. Most recent developments employ the use of shearlets to characterize prevailing directions in image patches in order to identify somas from dendrites and axons when multiple neurons are present in the same image. Our methods have been extensively validated on DIADEM datasets as well as on other data sets provided by our neuroscience partners. Our algorithms require minimal training and operator intervention and perform with average sub-voxel/pixel accuracy. We will present extensive quantitative and qualitative results demonstrating the accuracy and robustness of our method.

Matching of serial sections for single cell reconstruction and subsequent alignment with functional brain maps

In the cerebral cortex each neuron establishes synaptic contacts with hundreds of other neurons up to 3-4 mm laterally. The complex structure of intracortical connections represents a major challenge of neuroscience, i.e. to find out how the axons of different cell types are distributed spatially, how individual axons relate to functional representations at local and global scales. Here we use intracellular labeling cortical neurons and combine with functional brain imaging (intrinsic signal optical imaging) in the primary visual cortex of the cat. In order to unravel the spatial distribution of connections established by single pyramidal and non-pyramidal cells, labeled dendrites, axons and axon terminals were reconstructed in 3-dimensions (3d). To this end, serial sections (80 um thick) were cut parallel to the cortical surface and processed for histochemical visualization of biocytin labeled processes. Reconstruction of dendritic and axonal processes as well as axon terminals of single cells was made from 9-16 consecutive sections using an Olympus BX50 light microscope (x100 objective) fitted with a computer controlled moving stage and a z-motor (Neurolucida, MBF). Then, the 3d location of axon terminals was aligned with corresponding functional maps using custom made software and their distribution evaluated quantitatively.

There are two important points to be considered. (i) A major difficulty of serial reconstruction from large sections (<10 mm long) in which a single axon can span 5-6 mm laterally (i.e. parallel to cortex surface) is the occurrence of tissue distortion due largely to uneven tissue shrinkage and mechanical damage. These distortions may cause several tens or even hundreds of um mismatch between labeled processes that continue from one section to another. On the other hand, alignment of cell reconstructions with in vivo functional images requires to setting tissue sectioning plane as closely matching the camera image plane as possible.

I will present data how to control and validate the above mentioned technical pitfalls and offer solutions to overcome their effects.

[Supported by FP7-ICT-2009-4 (Brainscales-enlargEU)]

Genetic labeling to quantify synaptic development during adult neurogenesis

Our laboratory is interested in the assembly of neuronal circuits. We focus on the process of neuron addition into the brain of vertebrates, and seek to understand how new neurons integrate into the circuits of the adult brain, and their role in information processing and storage.

To quantify synaptic formation during the maturation of new neurons while they integrate into the circuit we have generated viral vectors encoding fluorescently-tagged synaptic markers. To visualize glutamatergic input synapses, we expressed a PSD-95:GFP fusion protein, a scaffolding protein that localizes to the postsynaptic density of glutamatergic synapses and can be used as a postsynaptic marker of glutamatergic synapses. To label presynaptic synapses (output synapses), we used a synaptophysin:GFP fusion protein that can be used to study the distribution and density of presynaptic sites in neurons both in vitro and in vivo.

We have used retroviral vectors encoding PSDG and sypG to label input and output synapses, and to monitor their development in new neurons added during adulthood in the olfactory bulb (OB) and dentate gyrus (DG) of mice and rats.

In addition, we have generated viral vectors encoding both synaptic-tagging constructs and ion channels that increase or decrease electrical activity in single neurons.

We will discuss how increasing electrical activity in single neurons affect their synaptic development and the involvement of the NPAS4 transcription factor in the regulation of synapse formation.

Structure-function relationships in primate neocortical pyramidal neurons
Jennifer I. Luebke^1,2 and Christina M. Weaver^2,3
1Department of Anatomy and Neurobiology, Boston University School of Medicine, Boston, MA 02118, USA; ²Computational Neurobiology and Imaging Center, Mount Sinai School of Medicine, New York, NY 10029, USA; ³Department of Mathematics, Franklin and Marshall College, Lancaster, PA 17604, USA.

Neocortical pyramidal neuron dendrites and dendritic spines vary significantly across different brain areas and undergo significant changes in both normal and pathological aging. Because of the determinant role of these highly dynamic structures in signaling by individual neurons and ultimately in the functionality of neuronal networks that mediate cognitive functions, a detailed understanding of dendritic parameters and changes to them is of paramount importance. The multidimensional approach taken by our collaborative group includes analyses of empirical high-resolution 3D morphological and electrophysiological data acquired from frontal cortical pyramidal neurons using confocal laser scanning microscopy and whole-cell patch-clamp recording techniques, combined with computational modeling methodologies.

Using this approach we have demonstrated that layer 3 pyramidal neurons in in vitro slices of monkey primary visual cortex (V1) and dorsolateral prefrontal cortex (dlPFC) possess highly distinctive structural and functional properties. Area V1 pyramidal neurons are much smaller than dlPFC neurons, with significantly less extensive dendritic arbors and far fewer dendritic spines. Relative to dlPFC neurons, V1 neurons have a significantly higher input resistance, depolarized resting membrane potential and higher action potential (AP) firing rates. Spontaneous postsynaptic currents are lower in amplitude and have faster kinetics in V1 than in dlPFC neurons, but are no different in frequency. Three-dimensional reconstructions of V1 and dlPFC neurons were incorporated into computational models containing Hodgkin-Huxley and AMPA- and GABAA-receptor gated channels. Morphology alone largely accounted for observed passive physiological properties, but led to AP firing rates that differed more than observed empirically, and to synaptic responses that opposed empirical results. Accordingly, modeling predicts that active channel conductances differ between V1 and dlPFC neurons. The unique features of V1 and dlPFC neurons are likely fundamental determinants of area-specific network behavior. The compact electrotonic arbor and increased excitability of V1 neurons support the rapid signal integration required for early processing of visual information. The greater connectivity and dendritic complexity of dlPFC neurons likely support higher level cognitive functions including working memory and planning.

Applying these same techniques to electrophysiologically characterized neurons from young and aged monkeys, computational modeling predicts that age-related morphologic changes significantly reduce inward and outward dendritic voltage attenuation (presented in a poster here by Christina Weaver). However, morphology alone cannot account for an age-related increase of firing rates in response to somatic current injections. These multidisciplinary studies provide unique information on the cellular phenotypes of pyramidal neurons in distinct cortical areas of the primate brain, essential for understanding how neurons operate within local circuits. They also provide testable hypotheses regarding the functional repercussions of dystrophic changes to these neurons.

[Supported by NIH/NIA grants P01 AG00001, R01 AG025062, R01 AG035071 and NIH/NCRR RR00165]

Putting it All Together: Linking and Federation of Neuroscience Data through the Neuroscience Information Framework

Understanding the brain strains the limits of current human ingenuity. Perhaps more than any other organ system, the problem of understanding the brain is fundamentally multiscale, with relevant data derived from spatial and temporal scales spanning many orders of magnitude. Because of the complexity and breadth of these networks, unraveling functional circuits underlying complex behaviors or pinpointing the locus of disease processes, even where the genetic defect is known, has confounded scientists, who by the limitations of experimental methods glimpse only a pinhole view of a vast interconnected landscape.

Neuroscientists rely heavily on technological advances to expand our capacity to deal with this enormous complexity. Such advances in data acquisition and computing power are producing views of the brain of increasing size, breadth and detail, as we acquire data spanning multiple scales across increasing expanses of brain tissue. The potential power of these integrated approaches is exemplified in large-scale projects such as the Blue Brain and the Allen Brain project . These projects realize huge monetary and manpower investments into the generation of large amounts of data. Because data within these projects are mainly acquired within a single framework, they are able to build powerful informatics infrastructure to serve and analyze these data. Mining these richly integrated data sets is starting to yield new insights into how the brain is organized.

The vast majority of neuroscience, however, is still conducted by individual researchers, who contribute their data and unique insights through less well structured venues such as the literature and websites or the creation of smaller custom databases. Although the amount of data is increasing daily, neuroscience as a whole, with its exceptionally large scope and diverse research community, until recently lacked a coherent community framework for bringing these data together. I will present the Neuroscience Information Framework a 5 year project from the Neurosciences NIH Blueprint Consortium to provide a platform and semantic framework for cataloging, integrating and linking the large number of digital resources that are now available across scales within neuroscience. As neurons and neuroanatomy are foundational objects within neuroscience, a significant focus has been on linking databases of digital morphologies (NeuroMorpho.org) to other cellular and anatomical databases. We have also worked with an international group of experts via the International Neuroinformatics Coordinating Facility to define an ontological structure for describing neurons and their properties.

A test case: the structural analysis of the retina

Because of its small dimensions and well-studied function, the retina has sometimes provided an advance view of developments in understanding other structures of the central nervous system. Current concepts of retinal structure will be summarized, comparing the contributions of light microscopic reconstructions, genetic techniques, and high-throughput electron microscopy. The fundamental cellular architecture of the retina is now clear, as are the basic pathways of information flow. The contributions of the three methodologies, and the prospects for CNS structures of a larger scale, will be discussed.

Dendritic Spine Pathology in Cortical Pyramidal Cells of GABA Neuron-Specific NMDA Receptor Hypofunction Mice

Schizophrenia is a devastating and complex psychiatric disorders characterized by impairment in multiple domains, including cognitive symptoms, such as deficits in perception, attention and learning and memory. These features suggest that the pathophysiology of schizophrenia reflects altered connectivity within and between multiple brain regions. These connections involve dendritic spines, small protrusions from the dendritic shafts that are the synaptic targets of axon terminals from nearby cells. Postmortem studies have revealed dendritic abnormalities of pyramidal neurons in the dorsal lateral prefrontal cortex (DLPFC) of subjects with schizophrenia. These abnormalities include smaller somal volumes, decreased dendritic arbor size and complexity, and reduced dendritic spine density on subpopulations of pyramidal neurons. For example, Glantz and Lewis (2000) reported that spine density on the basilar dendrites of pyramidal neurons in deep layer 3 of DLPFC is significantly lower in subjects with schizophrenia relative to normal comparison subjects. However, the molecular mechanisms that may contribute to these alterations are unknown.

We recently generated a new transgenic mouse line, in which GluN1 (Grin1) subunit of NMDA receptor channels is eliminated selectively from cortical and hippocampal GABAergic interneurons (mainly parvalbumin (PV)-positive) during early postnatal period. We found a constellation of cellular, physiological and behavioral phenotypes that resembled schizophrenia pathophysiology (Belforte et al., 2010). For example, cognitive-like symptoms included impairments in spatial working memory and short-term social memory and attenuated sensorimotor gating (decreased PPI of the startle reflex). Negative-like symptoms were also observed including reduced preference for sweet solution (i.e., anhedonia) as well as deficits in social interaction and nesting/mating behaviors. On a neurophysiological level we observed disinhibition of cortical excitatory neurons and reduced neuronal synchrony, both of which are consistent with the hyperactivity in the dorsolateral prefrontal cortex observed in individuals with schizophrenia during working memory tasks. Furthermore, we found positive-like symptoms including an exacerbation of amphetamine-induced hyperlocomotion, which was accompanied by an increased dopamine release in the ventral striatum. We also observed impairments in tone-evoked gamma frequency local-field potential (LFP) oscillations in the auditory cortex, which is consistently reported in schizophrenia subjects. Taken together, these findings suggest that a selective dysfunction of corticolimbic PV GABAergic neurons induced through NMDAR hypofunction recapitulated a remarkable number of pathophysiological characteristics of schizophrenia. This model unifies the dopamine hypothesis, neurodevelopmental hypothesis, and stress hypothesis with the NMDAR hypofunction-induced GABAergic dysfunction hypothesis (Nakazawa et al, Neuropharmacology, 2012).

In order to explore whether selective reduction of NMDARs in interneurons results in the morphology change in the pyramidal neurons, we have analyzed a development-dependent changes in the dendritic morphology and spine density of apical dendrites of layer V pyramidal neurons in mPFC with modified Golgi-Cox staining technique. We found that spine density of the primary apical dendrites of layer V pyramidal neurons is significantly reduced in the mutants at 8 weeks of age (N=4). Preliminary results that this spine density reduction was not observed at 4 weeks of age, suggesting that synaptic pruning during the adolescent period is exacerbated in the mutants. These findings are consistent with a hypothesis of abnormal synaptic pruning as a potential cause for schizophrenia (Feinberg, 1982). We are studying possible mechanisms underlying the abnormal synaptic pruning which should be originating from NMDA receptor hypofunction in the cortical GABAergic interneurons.

In the Pursuit of Connectomes - Combining in vivo Recordings with automated Reconstructions of complete 3D Neuron Morphology

Soma location, dendrite morphology and synaptic innervation are key determinants of neuronal function. Unfortunately, conventional functional measurements of sensory-evoked activity in vivo yield limited structural information. In particular, when trying to infer mechanistic principles that underlie perception and behavior, interpretations from functional recordings of individual or small groups of neurons often remain ambiguous without detailed knowledge of the underlying network structures.

I will present a novel reverse engineering approach that allows investigating sensory-evoked signal flow through individual and ensembles of neurons within the context of their surrounding neural networks. To do so, spontaneous and sensory-evoked activity patterns are recorded from individual neurons in vivo. In addition, the complete 3D dendrite and axon projection patterns of such in vivo characterized neurons are reconstructed using a custom- designed semiautomatic tracing pipeline. Our semiautomatic tracing system rivals manual tracings from human experts in both accuracy and completeness. Specifically, the combination of these in vivo filling and reconstruction approaches recovered total axonal lengths far greater than previously observed. For example, intracortical axonal lengths of most pyramidal neurons in rodent sensory cortex exceed 10 centimeters per neuron. The axon of an individual thalamocortical neuron can even reach a length of up to 25 centimeters.

Further, our latest developments allow recording and reconstructing of connected pairs in vivo. This approach has been largely used for preparations in vitro, for example resulting in various detailed computer models that relate the structure of an individual neuron to its measured function after somatic current injections. Unfortunately, the typical thickness of an in vitro brain slice is 300µm. Consequently, in vitro tracings usually suffer from cut off dendrites and axons, restricting connectivity measurements to close-by neurons. Our approach overcomes these limits and allows reconstructing the connectivity of neurons separated even by millimeters.

Further, our latest developments allow recording and reconstructing of connected pairs in vivo. This approach has been largely used for preparations in vitro, for example resulting in various detailed computer models that relate the structure of an individual neuron to its measured function after somatic current injections. Unfortunately, the typical thickness of an in vitro brain slice is 300µm. Consequently, in vitro tracings usually suffer from cut off dendrites and axons, restricting connectivity measurements to close-by neurons. Our approach overcomes these limits and allows reconstructing the connectivity of neurons separated even by millimeters.

Navigating NeuroMorpho.Org

NeuroMorpho.Org is a central repository of 3D digital reconstructions of neuronal morphology. The repository contains over 10,000 reconstructions from 21 species representing contributions from 120 laboratories. Each reconstruction in the database is associated with relevant metadata details such as brain region, cell type, experimental protocol, reconstruction software, and date of deposition, to name a few. Visitors to the website are able to browse and search this vast collection of digital morphologies by selecting various metadata categories such as the ones listed above. This presentation will be a live demonstration of navigating NeuroMorpho.Org. Various functionalities of the website that allow users to browse, search and download desired data will be highlighted. Furthermore, I will present the vast literature database built during the historical and ongoing literature mining process to identify publications containing digital reconstructions.

Vaa3D-Neuron2 for automated 3D neuron reconstruction and efficient editing

We present the latest Vaa3D-Neuron2 tool, which has been developed as a software suite for automated neuron reconstruction from light microscopy images. This software has been tested on thousands of neuronal images for automated tracing. In addition, the software facilitates easy and quick neuron editing of the morphology. Therefore, as a suite, Vaa3D-Neuron2 is suitable for large-scale reconstruction effort in the computational neuroscience field.

L-Measure: A cross-platform, open-source tool for quantitative morphometric analysis, statistical comparison, and feature-based search of digital reconstructions of neuronal morphology
Sridevi Polavaram and Giorgio A. Ascoli

L-Measure is a user-friendly, multi-platform software for morphometric quantification of digitally reconstructed neurons. Over 100 unique scalar features, such as the mean, range, variance, and total sum (as applicable) of the number of branches, angles, length, fractal measures etc., can be measured. Moreover, the distributions of feature dependencies (e.g. diameter vs. path distance from the soma) can also be extracted. The analyses can be applied to groups of arbors or selected sub-parts (dendrites only, axon only, etc.). The output is saved to a tab-separated text file, which can be easily loaded into external statistical and plotting software, such as Excel, MATLAB, or R. The built-in statistical tests (with optional corrections for multiple-testing) provide for comparisons between two groups of data, such as different cell types, lab archives, brain regions, developmental stages, control vs. experimental conditions etc. The search functionality allows identification and selection of subset of reconstructions based on their morphometric features.

L-Measure is compatible with all main reconstruction file formats, computer platforms, and browsers. The tool has been used in at least 20 publications (outside of our lab) in the past 3 years. The online version of L-Measure has been utilized more than 3000 times in the last 2 years to process nearly 20,000 files. The standalone version has been downloaded over 1000 times in the past year. The tool caters to a wide range of applications with a Java-based point-and-click graphical user interface as well as the programmable command-line call to a C++ compiled executable, which enables batch processing of thousands of files. L-Measure is actively developed and maintained per bug reports or user requests, such as processing new file formats, computing new functions, enabling sub-tree level analysis, etc. The code, freely shared, has been successfully incorporated in other popular neuroinformatics tools. L-Measure is also a critical component of the NeuroMorpho.Org data processing pipeline. This demo will introduce new and established functionalities with hands-on use-case examples. Feel free to bring your reconstructions for live testing!
[Supported by NIH R01 NS39600]

Employing NeuGen 2.0 to automatically generate realistic morphologies of hippocampal neurons and neural networks in 3D

Detailed cell and network morphologies are becoming increasingly important in Computational Neuroscience. Great efforts have been undertaken to systematically record and store the anatomical data of cells. This effort is visible in databases, such as NeuroMorpho.org. In order to make use of these fast growing data within computational models of networks, it is vital to include detailed data of morphologies when generating those cell and network geometries. For this purpose we developed the Neuron Network Generator NeuGen 2.0, that is designed to include known and published anatomical data of cells and to automatically generate large networks of neurons. It offers export functionality to classic simulators, such as the NEURON Simulator by Hines and Carnevale. NeuGen 2.0 is designed in a modular way, so any new and available data can be included into NeuGen 2.0. Also, new brain areas and cell types can be defined with the possibility of constructing user-defined cell types and networks. Therefore, NeuGen 2.0 is a software package that grows with each new piece of anatomical data, which subsequently will continue to increase the morphological detail of automatically generated networks. In this paper we introduce NeuGen 2.0 and apply its functionalities to the CA1 hippocampus. Runtime and memory benchmarks show that NeuGen 2.0 is applicable to generating very large networks, with high morphological detail.

Long-term gene therapy causes transgene-specific changes in the morphology of normal adult retinal ganglion cells
Jennifer Rodger¹, Marissa Penrose¹, Mats Hellstrom², Alan R Harvey^1,2,*
1Experimental and Regenerative Neuroscience, School of Animal Biology, The University of Western Australia, Perth, Australia, ²School of Anatomy, Physiology and Human Biology, The University of Western Australia, Perth, Australia,

Recombinant adeno-associated viral (rAAV) vectors can be used to introduce neurotrophic genes into injured CNS neurons, promoting survival and axonal regeneration. Gene therapy holds much promise for the treatment of neurotrauma and neurodegenerative diseases. We previously demonstrated that intraocular injection of rAAV encoding neurotrophic factors significantly changed the dendritic morphology of retinal ganglion cells (RGCs) regenerating an axon into a peripheral nerve graft. Here we examined whether vector-delivered neurotrophins also altered the morphology of normal adult rat RGCs. We used rAAV2 encoding: (i) green fluorescent protein (GFP), or (ii) bi-cistronic vectors encoding GFP and ciliary neurotrophic factor (CNTF), brain-derived neurotrophic factor (BDNF) or growth-associated protein-43 (GAP43). After 5-8 months, RGCs were retrogradely labeled with fluorogold (FG) by placing FG on the cut optic nerve two days prior to sacrifice. Live retinal whole mounts were prepared and GFP positive (transduced) or GFP negative (non-transduced) RGCs injected iontophoretically with 2% Lucifer yellow. Dendritic morphology was analyzed using Neurolucida software. Significant changes in dendritic architecture were found, in both transduced and non-transduced populations. Multivariate analysis revealed that dendritic fields of RGCs in all injected retinae had lost complexity. Retinae injected with AAV-BDNF-GFP were the most severely affected: dendrites of transduced cells were significantly more tortuous compared to GFP transduced controls, and non-transduced cells had reduced branching. The proportion of RGCs with aberrant morphology was significantly increased in all transgene groups compared to uninjected controls (no abnormal cells), representing fewer than 10% of cells in the AAV-GFP, AAV-CNTF-GFP and AAV-GAP43-GFP groups, but 46% of cells in the AAV-BDNF-GFP group. Thus vector-mediated expression of neurotrophic factors has measurable, gene-specific effects on the morphology of intact adult neurons. Such changes will likely alter the functional properties of neurons and may need to be considered when designing vector-based protocols for the treatment of neurotrauma and neurodegeneration.

Principles governing the input-output properties of motoneurons: lessons learned due to digital reconstructions of motoneuron dendritic trees
Ken Rose, Monica Neuber-Hess, Keith Fenrich, John Grande, Rob Maratta, Stephen Montague, Ethan Zhao, Farin Bourojeni, and Heather Nichol

In the words of Sir Charles Sherrington, motoneurons represent the ‘final common pathway’ of all networks involved in motor control. The simplicity of this description stands in sharp contrast to the myriad of intrinsic factors that regulate the input-output properties of motoneurons. These properties are determined by the collective interactions of 30,000-50,000 excitatory and inhibitory synapses, over 10 types of voltage-dependent and independent channels, and at least 4 sets of neuromodulatory systems. Most of these interactions take place on the dendritic tree. This suggests that the structure of the dendritic tree and the arrangement of synapses and channels on the dendritic tree may determine the spatial limits of these interactions and their effect on the input-output properties of motoneurons. To test this hypothesis, we have quantitatively mapped the distribution of several classes of synapses and channels on the dendrites of motoneurons.

Serotonin (5-HT) and noradrenalin (NA) amplify the responses of motoneurons to activation of excitatory or inhibitory synapses. Our studies suggest that this regulation is highly compartmentalized. The density of NA and 5-HT synapses is 6 to 10 times higher on small (<1.0 um), compared to large (>8.0 um), diameter dendrites. Since the actions of 5-HT and NA are mediated by second messengers that have limited mobility within dendrites, their local concentrations will be 50 to 80 times higher in small, compared to large, diameter dendrites. Moreover, measurements of the distance between neighbouring NA/NA synapses or 5-HT/5-HT synapses indicate that these synapses are 2 to 3 times closer to together than predicted by computer-generated distributions in which the position of each synapse is unrelated to the position of its neighbours. Thus, the global and local distribution of 5-HT and NA synapses are both designed to exploit the intrinsic amplification provided by second messenger systems.

The highly ordered distribution of 5-HT and NA synapses underscores the value of an equally comprehensive description of the distribution of voltage-dependent and independent channels that are regulated by 5-HT and NA. As a first step to meeting this goal, we have examined the distribution of HCN1 subunits of the channels mediating Ih. The density of HCN1 was higher on distal dendrites of motoneurons with small dendritic trees. On motoneurons with large dendritic trees, the proximal to distal gradient was either modest or absent.

To determine the functional consequences of these distribution patterns, we used anatomically realistic compartmental models to determine the intrinsic resonance properties of motoneurons. In the absence of modulation by 5-HT, the resonance frequency increased from 6 Hz at the soma to 8-9 Hz on distal dendrites. If the conductances for the channels responsible for Ih were proportional to the relative density of 5-HT synapses of dendrites with different diameters, the slope of the gradient doubled in small motoneurons, but was not affected in large motoneurons. This suggests that different regions of the dendritic tree are tuned to different frequencies of rhythmic synaptic activity. In keeping with our hypothesis, these frequencies depend on the distribution of Ih and 5-HT synapses.
[Supported by CIHR and the Barbara Turnbull Foundation]

Microglia Arbor Analytics

Brain tissue contains large ensembles of cells (neurons and glia) with complex and dynamic three-dimensional arbor morphologies. The sizes and shapes of these arbors vary across the brain especially as a function of brain region, vascular proximity, and cell layers. These arbors are also dynamic – changing in response to perturbations such as disease and injury caused by implanted neuroprosthetic devices. Recent advances in automated arbor tracing algorithms have made it practical to reconstruct these arbors on a large scale (tens of thousands to hundreds of thousand cells) from tiled series of 3-D multi-parameter optical microscope generated images. The logical next step is to quantify these arbors into an informative feature vector, and then to profile the collection of cells to extract spatio-temporal patterns in an informative and scalable manner – we call this “quantitative arbor analytics.”

In this presentation, I will describe recent advances in the direct integration of 3-D automated image analysis tools and novel bio-informatics tools as part of the FARSIGHT project. This integration enables direct and scalable approaches to arbor analytics. As a driving example, I will focus on changes to the gliovascular brain tissue inflicted by implanted neuroprosthetic devices. FARSIGHT is a free and open source computational toolkit for quantitative three-dimensional (3D) delineation (segmentation) of cellular anatomy, the topology of the vascular network, association of molecular analytes with specific cell types and multi-cellular units (e.g., stem cell niches, and cell layers), and statistical analysis of spatial cell distributions relative to various structures including the vasculature and implanted neural recording devices. It has an active multi-scale linked software architecture that has enabled us to embed novel active and transfer machine learning algorithms, multi-variate co-clustering algorithms, and progression analysis algorithms, alongside traditional pattern analysis tools. Using these tools, it is now possible to compute detailed measurements of cell arbors for developing a quantitative understanding of tissue structure and function. For this, we have extended the L-measure developed by Scorcioni et al., These analytics tools are versatile and serve a variety of practical purposes. For a start, we can statistically validate the automatically traced arbors. Next, we can build a quantitative profile describing the distribution of arbor morphologies in the normal, unperturbed brain. Finally, it is now practical to detect and quantify anomalies and changes in the perturbed brain.

Morphologically distinctive dentate projection neurons show unique developmental profile and post-traumatic plasticity
Elgammal FS, Gupta A, Chika-Nwosuh O, Swietek B, Kella K and Santhakumar V

Moderate forms of concussive brain trauma increase the risk of multiple organic pathologies, including spontaneously occurring seizures and memory dysfunction. Our investigation of post-traumatic plasticity among molecular layer neurons, revealed a class of neurons with somata in the inner molecular layer and axonal projections to CA3, which an earlier study had identified as glutamatergic semilunar granule cells (SGCs), distinguished from dentate granule cells by morphological and physiological features (Williams et al.,2007). The most striking morphological difference between SGCs and granule cells is the wide dendritic span of the SGCs compared to granule cells. Our earlier studies in one month old rats have used this morphological classification to demonstrate that SGCs receive significantly greater synaptic and tonic GABAergic inhibition compared to granule cells (Gupta et al., 2012). Likewise, SGCs also receive significantly greater excitatory synaptic inputs than granule cells. Within a week after concussive brain injury, SGCs and granule cells demonstrate contrasting changes in inhibitory tone, with a post-injury increase in tonic and synaptic GABA currents in dentate granule cells as opposed to a decrease in both parameters in SGCs after brain injury. Moreover, intrinsic excitability of SGCs is enhanced after brain injury (Gupta et al., 2012) which was unique among dentate excitatory neurons. Interestingly, while the frequency inhibitory synaptic currents in granule cells remained unchanged between one and four month old uninjured rats, there was a significant decrease in synaptic inhibition in SGCs during the same time period indicating that SGCs show developmental plasticity into early adulthood. We propose that location and morphological characteristics contribute, in part, to the differences in synaptic inputs and post-traumatic plasticity between SGCs and granule cells. Using the open source, L-Measure software we are analysing morphometric data from 3D Neurolucida reconstructions of a set of SGCs and granule cells recorded during slice physiology experiments. In addition to the expected differences in somatic size, location, and maximum dendritic spread we compare several morphometric characteristics of dendrites and axons between SGCs and granule cells in uninjured brain slices. We find that compared to granule cells, SGCs have a greater total axonal length with significantly higher branching in the dentate hilus. The implications of SGC morphology to its distinctive synaptic physiology and post-traumatic and developmental plasticity and network function will be discussed.

Adolescent nicotine induces persisting dendritic alterations in structures of the extended amygdala

Digital 3-D reconstruction of cells from Golgi-stained tissue from a number of brain areas after adolescent nicotine exposure has revealed that there is a selective stimulation of new dendritic branches after nicotine. In n. accumbens, this growth is seen at a dose of nicotine not effective in adults. In medial prefrontal cortex and basolateral amygdala, there are adult-adolescent differences in the particular neurons undergoing dendritic branching after nicotine dosing. Adolescents also undergo additional nicotine-induced branching in bed n. of stria terminalis and dentate gyrus, although adult comparison cells are not yet available. These data suggest that adolescent nicotine induces growth in a somewhat different circuitry than adult nicotine. Inferences from observed dendritic elaboration, and reference to the literature on activity-dependent synaptogenesis, suggest that adolescent nicotine may alter functional connectivity of that circuitry, which then persists into adulthood and affects CNS functioning/behavior.

The logic of inferring synaptic connectivity changes from individual dendritic tree reconstructions, and the relative value of Sholl analysis and branch order analysis of reconstructions, will be discussed.

Neural computation in dendritic and axonal arbors in the inner retina

The retina performs many specific signal processing functions in two dozen parallel pathways that carry the visual scene to the brain. The anatomy and synaptic connections within these microcircuits have been studied using 3D reconstructions from electron microscopy and more recently confocal microscopy, so much of the circuit structure is known. Morphological differences and specific patterns of synaptic connectivity define anatomical cell types, which correlate well with cell type defined by physiological properties such as their synaptic currents and their spatio-temporal responses to light stimuli. However, the details of how many of these microcircuits process the visual input are unknown.

Using realistic computational models defined by 3D reconstructions of retinal neurons along with some of their physiological properties, we hypothesized that the dendritic arbor of the On-Off direction-selective ganglion cell comprises independent computational subunits. The subunits are critically dependent on thin dendrites to give partial electrotonic isolation. Each subunit sums local excitatory and inhibitory inputs that are modulated by presynaptic circuits. Signals in each dendrite reflect the direction of moving light stimuli, so that the summed postsynaptic signal within each subunit is larger for motion in a specific direction. This directional preference is further amplified by nonlinear regenerative processing by sodium channels. Each subunit independently generates spikes that reflect directional preferences of presynaptic and postsynaptic processing. The spikes propagate throughout the cell to its axon and its dendritic arbor, quenching spike initiation in other subunits, so that the subunit with the most signal controls the cell's spike rate.

Recently several types of retinal neuron that were thought to have exclusively graded responses have been shown to spike in response to light stimuli. Bipolar cells contain separate dendritic and axonal arbors that each comprises a proximal stalk and 3-5 thinner branches emanating 10-30 µm. Several types of bipolar cell also contain sodium channels located in their axon that can generate full-blown spikes or smaller spikelets. Using realistic computational models calibrated with responses to stimuli, we showed that spikelets and graded signals propagate along the axonal branches but can be modulated by inhibitory feedback. Moreover, signals in different axon terminal branches can generate independent calcium spikes, implying that the axon terminal comprises separate subunits that modulate synaptic outputs. This independent processing depends on the pattern of thin branches in the axon terminal. Because each bipolar cell may contact several types of postsynaptic cell, this arrangement allows the synaptic output signals from a bipolar cell to be tailored in a different way for each postsynaptic cell.

In summary, computational modeling of retinal microcircuits defined by 3D reconstruction, calibrated with biophysical details of the real cells such as the location of ion channels along with physiological responses, suggests that local processing within subunits of dendritic and axonal arbors contributes in a substantial way to the retina's highly specific signal processing.

Geometric feature learning is essential for automated tracing of neurites
Rohan Gala, Julio Chapeton, and Armen Stepanyants

Automating tracing of axonal and dendritic arbors from light microscopy stacks of images is vital for quantitative analyses of large-scale neural circuits. In this work we focus on branch merging, which is an underlying component of many automated tracing algorithms. Branch merging is required for connecting neurites which appear broken due to imperfect labeling or for resolving neurites which appear fused due to the limited resolution of light microscopy. Such problems are routinely solved by trained users performing manual tracing tasks. We noticed that the users learn to discriminate between correct and erroneous branch merging scenarios by combining information from various sources: distances between branches, branch orientations, and average intensities, variations in intensities, branch thicknesses, curvatures, tortuosities, and presence of spines or boutons. Therefore, to evaluate different branch merging scenarios automatically, we designed a cost function which combines many of the above mentioned features. Modified perception and SVM classifier algorithms were trained during user-assisted branch merging procedure to determine the best weights for combining these features. The weights were next used to automatically trace new image stacks obtained under similar experimental conditions.

To evaluate the performance of the automated tracing algorithm, a variety of image stacks were traced automatically, as well as manually by several trained users, which was done to establish the baseline inter-user variability (gold standard). Several geometrical and topological features of the traces were selected for the comparison of automated and manual traces. These features include the overall characteristics of the traces, such as the total length, number of trees, number of branch- and terminal-points, as well as measures that reflect the affinity of the automated traces to the gold standards, such as distance between corresponding branches, distance between corresponding branch- and terminal-points, number of stolen (false positive) and lost (false negative) branches. Our results show that when the density of labeled neurites is sufficiently low, automated traces do not differ significantly from the gold standards.

The online training and branch merging algorithms are implemented as part of the Neural Circuit Tracer software for automated and manual tracing of neurites from light microscopy stacks of images. Information on the Neural Circuit Tracer can be found at www.neurogeometry.net. [Supported by the NIH grant NS063494]

The regulation of the actin cytoskeleton in dendrite morphogenesis

The fundamental processes underlying the development of dendritic branches with very heterogeneous morphologies and the key molecules involved in controlling of dendrite outgrowth and branching are fundamental to our understanding of neurodevelopment, neuronal remodeling and brain wiring.

We performed a systematic RNAi-based screen to identify genes critical for dendrite formation in multi-dendritic arborization (da) neurons of the peripheral nervous system (PNS) in Drosophila. Thus, we analyzed the effect of cell-autonomous knock-down of genes of interest on the development of 4 morphologically distinct neuronal subclasses. Genetic and time-lapse analysis indicated that the Arp2/3 complex, involved in the nucleation of branched actin filaments, is required for the de novo formation of dendrite branches in all da neuron classes. On the contrary, over-expression of Arp2/3 activators was sufficient to induce excessive branching in all da neuron classes. Among the potential activators of the Arp2/3 complex, only WAVE knock down elicited the same effect as loss of Arp2/3 components. Our results suggest that Arp2/3 complex-induced actin nucleation is a general mechanism at the core of new branch point formation. Nonetheless, dendrite branches of different neuronal types display remarkably different morphologies. We sought to identify molecular components underlying distinct branch morphologies. Among the da neurons, class III are decorated by actin rich, short and straight branches. We found that the actin-bundling protein fascin is a specific marker of this type of branch and is specifically required for its formation and dynamics. Thus, cytoskeletal regulation is distinct in branches of different morphologies and dynamics.

Morphological and computational properties of thalamocortical neurons from VPL nucleus of cats

Thalamus is the main gateway of ascending sensory information arriving to cerebral cortex. Thalamocortical (TC) neurons also receive direct excitatory corticothalamic feedback drive and inhibitory drive from reticular thalamic nucleus. The aim of the present study was to investigate the morphological properties of TC neurons responsible for integration of excitatory influences coming from periphery and corticothalamic fibers. TC neurons from VPL nucleus were stained with retrograde tracer Fluoro Ruby injected in white matter underneath the somatosensory cortex within forelimb area. A part of the brain containing VPL nucleus was cut on 80-m-thick consecutive sections. After revealing tracer, in order to minimize shrinkage, the sections were incubated with OsO4 for 10-20 min. The sections were reconstructed on Neurolucida, using oil immersive objective x100. Six-eight consecutive sections were needed to reconstruct each neuron. Z-axis was corrected for dry shrinkage and for virtual shrinkage caused by optical density of the imbedding medium and oil. Major morphological properties of TC neurons (soma size, number of dendrites, number of dendritic branches etc.) in our study were similar to the ones described in literature. However, since we took particular care of shrinkage factors, we were able to provide additional information. The largest total dendritic membrane area was fond at distances 120-160 m from the soma. The mean diameter of distal dendrites was in the order of 1 m. The mean geometrical ratio at branching points for back propagating signals was 1.43, however for orthodromically travelling signals the geometrical ratio was above 3 for middle and distal dendrites and it was above 4 for proximal dendrites suggesting dramatic attenuation of forward propagating synaptic inputs. Indeed, modeling effects of single vesicle release in distal dendrites induced local EPSPs of tens mV in amplitude which were attenuated during propagation to the soma to become 0.4-0.5 mV in amplitude. Stimulation of proximal dendrites elicited local EPSPs of much smaller amplitude, which was only slightly, but significantly larger at the level of soma for AMPA mediated EPSPs, but smaller for NMDA mediated EPSPs as compare to distally evoked EPSPs. Modeling of simultaneous activation of several release sites on proximal dendrites led to nearly linear summation of local responses independently of release site distribution. However, simultaneous activation of closely located several release sites on distal dendrites in a model resulted in remarkable sub-linear summation due to approaching of reversal potential for EPSPs, however the summation was nearly linear when different dendritic branches were activated. Altogether, our results demonstrate that morphological features of VPL TC neurons are well situated to integrate biologically relevant excitatory inputs. Multi-release site lemniscal synapses are located on proximal dendrites. Multiple single-release site corticothalamic synapses are formed on separated branches of distal dendrites. Both arrangements are efficient for integration of incoming onto TC neurons excitatory influences.

Diverse properties and synaptic targets of long-range projecting inhibitory interneurons in the CA1 hippocampus
Dept. Biochemistry, Microbiology and Bio-informatics, Université Laval; Axis of Cellular and Molecular Neuroscience, IUSMQ, Québec, PQ, Canada

In the hippocampus, GABAergic interneurons serve the role of local inhibitory elements regulating the activity of principal cells and interneurons. In addition, some types of hippocampal interneurons can project to distant brain areas, providing a means for interareal coordination of neuronal activity. However, the properties and postsynaptic targets of long-range projecting interneurons remain largely unknown. Using a combination of whole-cell patch-clamp recordings, immunohistochemistry, cellular imaging and anatomical reconstruction, we identified two distinct types of long-range projecting interneurons in the mouse hippocampal CA1 area: (1) VIP- and muscarinic M2R-expressing interneurons in stratum oriens/alveus project to the subiculum, and (2) CCK-positive interneurons in stratum radiatum project to the subiculum and the entorhinal cortex. The two cell types differed by their anatomical and physiological properties, molecular profiles and synaptic targets, consistent with a functional dichotomy in interregional synchronization of neuronal activity in the temporal lobe.

Can Quantitative Dendritic Reconstructions Assess Dendritic Pathology?

By 1975 there were multiple Golgi analysis of normal and abnormal neuronal structure, particularly dendritic aberrations, and a key question from these studies was how much the abnormal dendrites contributed to dendritic function and signal processing. These Golgi studies included assessment of normal aging and senescence, Alzheimer’s disease, epilepsy, and genetic conditions, where the dendrites are distorted (such as intracellular storage disorders. At that time analysis of electrotonic function of dendrites was being developed by Rall and others, but only on model cells under ideal conditions. The principal method for analyzing dendrites was Golgi staining and there were quantitative estimates for dendritic extent and Scholl analysis, particularly across pathological dimensions.

To assess the hypothesis that either normal or abnormal dendritic structure might influence dendritic signal processing required the development of permanent intracellular staining combined with neuronal recordings (with HRP or biocytin), which reveals much more dendritic structure than Golgi staining, detailed neuronal reconstructions (initially with camera lucida then Neurolucida in 1990), and a mathematical method for electrical reconstruction from small, cylindrical dendritic segments (initially analog cable models but then digital semiconductor models). Based on these new approaches of digital reconstruction and electrotonic assessment of specific, realistic dendrites, mapping signal processing differences between neurons, and with aging, epilepsy and other diseases has now become possible. However, the assessment of dendritic pathology also requires companion physiological data on membrane and channel distribution and micropathology, which remains at the forefront of knowledge for estimation of dendritic signal processing.

The role of gap junctions between inhibitory interneuron dendrites: Beyond synchrony

We know far less about the dendrites of inhibitory interneurons compared to pyramidal cells. One interesting ubiquitous feature of interneuron dendrites is that they are electrically coupled by gap junctions. So how does this affect the behavior of such a network? I will present data where the accurate reconstruction of cerebellar Golgi cells was crucial to show that, 1) gap junction coupling between these cells allows them to share synaptic charge, making them act like a syncytium, 2) despite extensive electrical coupling, the heterogeneity of coupling puts strong constraints on spike time synchrony in the network.

NeuronStudio and beyond: using digital reconstructions to explore neuronal structure/ function

Christina M. Weaver^1,2 Patrick R. Hof^2,3 and Jennifer I. Luebke^2,4
1Department of Mathematics, Franklin and Marshall College, Lancaster, PA 17604, USA; ²Computational Neurobiology and Imaging Center, Mount Sinai School of Medicine, New York, NY 10029, USA; ³Department of Neuroscience, Mount Sinai School of Medicine, New York, NY 10029, USA; ⁴Department of Anatomy and Neurobiology, Boston University School of Medicine, Boston, MA 02118, USA

In collaborative work between our laboratories we have used automated software packages to reconstruct dendrites and spines in several different studies, including complete 3D reconstruction of more than 200 neurons. In particular, NeuronStudio is a user-friendly, robust, and freely available tool developed at the Mount Sinai School of Medicine for automated reconstruction of neuronal dendrite and spine morphologies. Here we review our approach for obtaining digital reconstructions of neurons from high resolution 3D confocal microscopic images, including integration of adjacent tiled images with the Volume Integration and Alignment System (VIAS) and automated tracing of dendrites and spines with NeuronStudio. We then present two recent examples of using our digital reconstructions to test hypotheses about the contribution of morphology to the function of layer 3 pyramidal neurons of the dorsolateral prefrontal cortex (dlPFC) of the rhesus monkey.

First, we show that morphology alone cannot account fully for recent findings that in vitro firing rates of these neurons are increased with aging. Using the NEURON simulation environment to construct compartment models of 6 young and 6 aged dlPFC neurons with equal passive parameters, we found that somatofugal and somatopetal dendritic voltage attenuation was significantly lower in aged model neurons than in young ones. We added voltage- and calcium-gated ion channels in order to fit the input resistance and firing rates of one young model neuron to its empirical data, and then applied those same parameters to all young and aged model neurons. Opposing the empirical results, there was no significant difference in either input resistance or firing rates of the young versus aged model neurons. Thus, these studies predict that age-related morphologic differences do affect dendritic signal integration, but do not account for changes in neuronal excitability observed in vitro.

Second, we show that our 3D reconstructions provide morphologic evidence for the spatial clustering of dendritic spines on individual apical branches of dlPFC neurons. We applied clustering algorithms and Monte Carlo simulations to quantify the probability that the level of spine clustering observed on dendrite branches occurs randomly. Upon analyzing ~40,000 spines on 280 apical branches, we found that spine clusters occur significantly more frequently on apical terminal branches than expected by pure chance. These findings indicate that spine clustering on these branches is likely supported by systematic biological processes. We also found that mushroom- and stubby-shaped spines are predominant in clusters on dendritic segments that display prolific clustering, independently supporting a causal link between spine morphology and spatial clustering. Future application and extension of these approaches will continue our quantitative, comparative studies between neurons from different brain regions (such as presented in the talk by Jennifer Luebke), age groups, and in various states of neurodegeneration and dysfunction.
[Supported by NIH/NIA grants P01 AG00001, R01 AG025062, R01 AG035071, R01 MH071818, R01 DC05669 and NIH/NCRR RR00165]

Hippocampome.Org: Open access to a knowledge base for the rodent hippocampus
D. W. WHEELER, C. WHITE, C. L. REES, D. J. HAMILTON, A. O. KOMENDANTOV, M. BERGAMINO, G. A. ASCOLI.

Hippocampome.Org hosts a knowledge base of neuron types for the rodent hippocampus. A neuron type is primarily defined by the presence or absence of axons and dendrites in 26 neuroanatomical parcels delineated by the partition of the hippocampal complex into subregions (dentate gyrus, CA3, CA2, CA1, subiculum, and entorhinal cortex) and their respective histochemical layers (e.g. CA1 oriens, radiatum, etc.). With such an approach, sufficient morphological information is available in the literature to characterize over 100 neuronal types. Each neuronal type is assigned a human- and machine-readable unique identifier to unambiguously link it to the relevant references and to bridge the sometimes confusing nomenclature. This initial characterization is then augmented by including information on the electrophysiological properties and the molecular marker expression profile for each neuron type. The collation of evidence from multiple publications into a centralized repository enables data mining for new discoveries, such as the potential connectivity of the network of hippocampal classes. Hippocampome.Org, which is currently undergoing alpha testing, is designed to facilitate analytics with user-friendly browsing and searching and to allow for the universally accessible machine-readable download of information.
[Supported by NIH R01 NS39600]

Modelling Signal Processing in Neurons

The crucial feature of neuronal ensembles is their high complexity and variability. This makes modelling and computation very difficult, in particular for detailed models based on first principles. The problem starts with modelling geometry, which has to extract the essential features from those highly complex and variable phenotypes and at the same time has to take in to account the stochastic variability. Moreover, models of the highly complex processes which are living on these geometries are far from being well established, since those are highly complex too and couple on a hierarchy of scales in space and time. Simulating such systems always puts the whole approach to test, including modeling, numerical methods and software implementations. In combination with validation based on experimental data, all components have to be enhanced to reach a reliable solving strategy.

To handle problems of this complexity, new mathematical methods and software tools are required. In recent years, new approaches such as parallel adaptive multigrid methods and corresponding software tools have been developed allowing to treat problems of huge complexity.

In the lecture we present a three dimensional model of signaling in neurons. First we show a method for the reconstruction of the geomety of cells and subcellular structures as three dimensional objects. With this tool, NeuRA, complex geometries of neuron nuclei were reconstructed. We present the results and discuss reasons for the complicated shapes. We further present a tool for the automatic generation of realistic networks of neurons (NeuGen) and a tool to classify neuron morphologies.

We further show some new model for the electric signal on realistic 3d morphologies. The model is derived directly from the Maxwell equations and allows the computation of the electric field in the exteerior and interior space in adtion to the membarne potential.

Morphologic diversity of cutaneous sensory afferents revealed by genetically directed sparse labeling
Hao Wu¹, John Williams^1,4 and Jeremy Nathans^1,2,3,4
1Department of Molecular Biology and Genetics, ²Department of Neuroscience, ³Department of Ophthalmology, and ⁴Howard Hughes Medical Institute, Johns Hopkins University School of Medicine, Baltimore, Maryland 21205.

The diversity of cutaneous sensory afferents has been studied by many investigators using behavioral, physiologic, molecular, and genetic approaches. Largely missing, thus far, is an analysis of the complete morphologies of individual afferent arbors. Here we present a survey of cutaneous sensory arbor morphologies in hairy skin of the mouse using genetically-directed sparse labeling with a sensory neuron-specific alkaline phosphatase reporter. Quantitative analyses of 719 arbors, among which 77 were fully reconstructed, reveal ten morphologically distinct types. Among the two types with the largest arbors, one contacts ~200 hair follicles with circumferential endings and a second is characterized by a densely ramifying arbor with one to several thousand branches and a total axon length between one-half and one meter. These observations constrain models of receptive field size and structure among cutaneous sensory neurons, and they raise intriguing questions regarding the cellular and developmental mechanisms responsible for this morphological diversity.

Towards a quantitative classification of neuronal cell types, a hundred years after Cajal and Lorente

I will discuss the need and importance of arriving at a quantitative description of neuronal phenotypes in order to classify the complement of neuronal cell types with objective metrics. I will illustrate some of the progress (and pitfalls) of this research program using our data and analysis from our work in the cerebral cortex.