Michel Kerszberg

Stabilization of complex input-output functions in neural clusters formed by synapse selection

Abstract We numerically analyze the self-organization of formal neurons disposed in a two-dimensional layer, and receiving inputs from two sets of afferent axons A and B. The probability for a given afferent to innervate some neuron depends initially on both afferent and target neuron types, which may be excitatory or inhibitory. This early wiring diagram leads to relatively ill-defined functional groups within the neuronal assembly. There follows a period during which the system differentiates, under the presence of external inputs, into groups of neurons with stable input-output relationships. The mechanism proposed for this maturation is based on the management of a limited provision of retrograde trophic factor distributed from postsynaptic neurons to presynaptic terminals whenever a Hebb-like condition is satisfied. Those boutons which do not receive sufficient trophic support ultimately degenerate. The remaining circuitry is characterized by emergent “mexican-hat” type interactions, i.e., short-range excitation vs. longer-range inhibition, and exhibits well-defined functional properties. These final properties are found to depend both on the initial wiring diagram, and on the correlations between the afferent inputs. Thus, increasing the frequency of, e.g., simultaneous activation of A and B, leads to an increased size of those patches which display activity when A and B are active together. Patches can be observed which realize in a stable and reliable manner any of the 16 Boolean functions of the variables A and B. Usually, patches endowed with different functions may coexist in a given system.

Computational diversity in a formal model of the insect olfactory macroglomerulus

We present a model of the specialist olfactory system of selected moth species and the cockroach. The model is built in a semirandom fashion, constrained by biological (physiological and anatomical) data. We propose a classification of the response patterns of individual neurons, based on the temporal aspects of the observed responses. Among the observations made in our simulations a number relate to data about olfactory information processing reported in the literature; others may serve as predictions and as guidelines for further investigations. We discuss the effect of the stochastic parameters of the model on the observed model behavior and on the ability of the model to extract features of the input stimulation. We conclude that a formal network, built with random connectivity, can suffice to reproduce and to explain many aspects of olfactory information processing at the first level of the specialist olfactory system of insects.

How neurons may compute: The case of insect sexual pheromone discrimination

Recognition of pheromone scent by male insects probably depends on analyzing the blends composition in terms of relative concentrations of major and minor molecular components. Based on anatomical, physiological and behavioral data concerning certain moth species and the cockroach, we propose a simple, biologically plausible neural circuit which is able to perform this task reliably. The model employs oscillations as a detecting device. This principle is easily generalized to other systems. As a computational device, ratio detection may find applications in a variety of biological situations, e.g. in the olfactory system of all animals.

Synchronization in Neural Assemblies

We analyze numerically and analytically the interplay of time scales involved in the processing of information by neural networks. Action potential transmission times, delayed synaptic action, time integration of signals by the postsynaptic membrane are taken into account. The effect of dynamic noise at synapses (triggering failures) and at the axon hillock (neuron misfirings) are also considered. We study the stationary retrieval states, their basins of attraction, and the time scale for retrieval. We find that neural assemblies are, to a remarkable extent, impervious to sloppy synchronization. Random delays in a network are also shown to allow for learning (upon presentation) and playback of pattern sequences. The match between the internal delay distribution and the time course of pattern teaching provides a selection mechanism favoring those sequences which best fit the systems built-in delays.

Signal-induced selection among spontaneous oscillatory patterns in a model of honeybee olfactory glomeruli

Do the oscillations observed in many neural assemblies have a cognitive significance? We investigate this question by mathematical modeling of the honeybees olfactory glomeruli, which are a subsystem of the antennal lobe nervous network, involved in food odor recognition during foraging behavior. Our computations reveal spontaneous oscillations. In those units where they manifest themselves, however, application of input signals modulate only slightly the autonomous activity: thus, an intense, synchronized oscillatory background tends to hinder odor discrimination. In contrast, where and when spontaneous oscillations are repressed, due to low excitability, different input signals will re-excite selectively distinct subsets of spontaneous oscillatory modes. These observations agree well with experimental findings and suggest new, quantitative experiments. They further indicate a possible role for the modulation and differential activation of endogenous oscillations in odor identification and possibly in other cognitive activities subserved e.g. by the mammalian cortex.

A Neuronal Model of a Global Workspace in Effortful Cognitive Tasks

Abstract: A minimal hypothesis is proposed concerning the brain processes underlying effortful tasks. It distinguishes two main computational spaces: a unique global workspace composed of distributed and heavily interconnected neurons with long‐range axons, and a set of specialized and modular perceptual, motor, memory, evaluative, and attentional processors. Workspace neurons are mobilized in effortful tasks for which the specialized processors do not suffice. They selectively mobilize or suppress, through descending connections, the contribution of specific processor neurons. In the course of task performance, workspace neurons become spontaneously coactivated, forming discrete though variable spatio‐temporal patterns subject to modulation by vigilance signals and to selection by reward signals. A computer simulation of the Stroop task shows workspace activation to increase during acquisition of a novel task, effortful execution, and after errors. We outline predictions for spatio‐temporal activation patterns during brain imaging, particularly about the contribution of dorsolateral prefrontal cortex and anterior cingulate to the workspace.

Genetics and epigenetics of neural function: A model

A neural model for one- or few-trial irreversible behavior learning such as occurs in imprinting is introduced. It is assumed that synaptic connections in the relevant parts of the central nervous system are initially set up in a largely, but not totally random fashion, as a result, for instance, of differential cellcell adhesion. The behavior to be learned is then sometimes exhibited, but not in a reproducible, mature way. During early neural activity, active postsynaptic neurons may, however, deliver a putative retrograde trophic factor to some of their afferent synaptic boutons. This is taken to occur according to a Hebb-type rule. At a later stage, only those synapses that have accumulated enough trophic factor are stabilized selectively. We show explicitly how this process may lead to a perfectly wired circuit. The calculations indicate that if the connections were relatively well defined from the beginning, then random pulses at the inputs suffice for this refinement process to take place. This is analogous to the maturation of neural circuits under spontaneous electrical activity (unsupervised learning). If the initial connections are fuzzy, however, well-defined patterns of activation are needed at the inputs so that selective stabilization leads to a correct functional system (the model now behaves in an instructionist mode). Experiments suggested by the model are discussed, and involve the manipulation of afferent inputs, of the initial synapse distribution, or of the stabilization phase.

A model for motor endplate morphogenesis: Diffusible morphogens, transmembrane signaling, and compartmentalized gene expression

A mathematical model for the formation and maintenance of synaptic contacts at the motor endplate is proposed. It is based on diffusion between sarcoplasmic nuclei of limiting amounts of a morphogen substance. The morphogen is postulated to act on genetic switch-like intranuclear units and to regulate positively both the transcription of its own gene and that of acetylcholine receptor (AChR) subunit genes. The efficacy of autoregulation is assumed to be depressed by electrical activity; while AChR genes transcription is enhanced by anterograde neural factors. Thus the model involves Turings classical ingredients: autocatalysis and short range activation by the morphogen, and long range inhibition by electrical activity. Our predictions include: the stabilization of a single, transcriptionally active nucleus located in the central region of the developing muscle fiber (or myotube); the frequent occurrence of transcriptional activity in nuclei at the tendinous ends; and the onset, upon denervation of adult muscle, of transcription waves, starting from both the central site and the tendinous nuclei. In noninnervated fibers, the calculations show that spontaneous, irregular electrical activity leads to a variety of near-periodic spatial patterns of transcription; these are also predicted in innervated fibers when the depressing effect of electrical activity is weak, giving rise to the stabilization of multiple endplates as occurs in muscles with distributed innervation.

Developmental canalization can enhance species survival

We investigate the behavior of haploid, asexual populations undergoing an evolutionary process. Each individual is endowed with a genotype, and one of several possible developmental mechanisms mapping this genotype onto a phenotype. We show that various properties of the mapping itself have important consequences for the survival of the groups. The populations which are most successful, both alone (but in a changing environment) as well as in competition against other groups (for which the mapping is different) consist of organisms where gene expression is characterized by pleiotropism, polygenic inheritance, and some amount of canalization (i.e. error damping). These same features lead to the appearance of patterns of punctuated equilibrium during evolution. Punctuated evolution was sometimes observed even in the absence of stabilizing selection; it then arose solely from the internal developmental constraints.

Pheromone detection, ratio discrimination and oscillations: a new approach to olfactory coding.

Several models of olfactory processing in the insect and mammalian olfactory systems have demonstrated the presence of oscillations. Whether the oscillatory behavior is a carrier of information is however not clear, nor do we know to what extent these oscillations contribute to odor decoding and discrimination. We have proposed a model, based on anatomical, physiological and behavioral data pertaining to insects, which shows how odor quality can actually be coded by modes of oscillation. A functional role for the major and minor pheromone components in conspecific mate recognition was proposed. Here we show how the components can, by modulating oscillations intrinsic to the antennal lobe glomeruli (macroglomerular complex, MGC), achieve robust pheromone detection at low computational cost.