Athanasia Papoutsi

Induction and modulation of persistent activity in a layer V PFC microcircuit model

Working memory refers to the temporary storage of information and is strongly associated with the prefrontal cortex (PFC). Persistent activity of cortical neurons, namely the activity that persists beyond the stimulus presentation, is considered the cellular correlate of working memory. Although past studies suggested that this type of activity is characteristic of large scale networks, recent experimental evidence imply that small, tightly interconnected clusters of neurons in the cortex may support similar functionalities. However, very little is known about the biophysical mechanisms giving rise to persistent activity in small-sized microcircuits in the PFC. Here, we present a detailed biophysically—yet morphologically simplified—microcircuit model of layer V PFC neurons that incorporates connectivity constraints and is validated against a multitude of experimental data. We show that (a) a small-sized network can exhibit persistent activity under realistic stimulus conditions. (b) Its emergence depends strongly on the interplay of dADP, NMDA, and GABAB currents. (c) Although increases in stimulus duration increase the probability of persistent activity induction, variability in the stimulus firing frequency does not consistently influence it. (d) Modulation of ionic conductances (Ih, ID, IsAHP, IcaL, IcaN, IcaR) differentially controls persistent activity properties in a location dependent manner. These findings suggest that modulation of the microcircuits firing characteristics is achieved primarily through changes in its intrinsic mechanism makeup, supporting the hypothesis of multiple bi-stable units in the PFC. Overall, the model generates a number of experimentally testable predictions that may lead to a better understanding of the biophysical mechanisms of persistent activity induction and modulation in the PFC.

Coding and decoding with dendrites

Since the discovery of complex, voltage dependent mechanisms in the dendrites of multiple neuron types, great effort has been devoted in search of a direct link between dendritic properties and specific neuronal functions. Over the last few years, new experimental techniques have allowed the visualization and probing of dendritic anatomy, plasticity and integrative schemes with unprecedented detail. This vast amount of information has caused a paradigm shift in the study of memory, one of the most important pursuits in Neuroscience, and calls for the development of novel theories and models that will unify the available data according to some basic principles. Traditional models of memory considered neural cells as the fundamental processing units in the brain. Recent studies however are proposing new theories in which memory is not only formed by modifying the synaptic connections between neurons, but also by modifications of intrinsic and anatomical dendritic properties as well as fine tuning of the wiring diagram. In this review paper we present previous studies along with recent findings from our group that support a key role of dendrites in information processing, including the encoding and decoding of new memories, both at the single cell and the network level.

Modulatory effects of inhibition on persistent activity in a cortical microcircuit model.

Neocortical network activity is generated through a dynamic balance between excitation, provided by pyramidal neurons, and inhibition, provided by interneurons. Imbalance of the excitation/inhibition ratio has been identified in several neuropsychiatric diseases, such as schizophrenia, autism and epilepsy, which also present with other cognitive deficits and symptoms associated with prefrontal cortical (PFC) dysfunction. We undertook a computational approach to study how changes in the excitation/inhibition balance in a PFC microcircuit model affect the properties of persistent activity, considered the cellular correlate of working memory function in PFC. To this end, we constructed a PFC microcircuit, consisting of pyramidal neuron models and all three different interneuron types: fast-spiking (FS), regular-spiking (RS), and irregular-spiking (IS) interneurons. Persistent activity was induced in the microcircuit model with a stimulus to the proximal apical dendrites of the pyramidal neuron models, and its properties were analyzed, such as the induction profile, the interspike intervals (ISIs) and neuronal synchronicity. Our simulations showed that (a) the induction but not the firing frequency or neuronal synchronicity is modulated by changes in the NMDA-to-AMPA ratio on FS interneuron model, (b) removing or decreasing the FS model input to the pyramidal neuron models greatly limited the biophysical modulation of persistent activity induction, decreased the ISIs and neuronal synchronicity during persistent activity, (c) the induction and firing properties could not be altered by the addition of other inhibitory inputs to the soma (from RS or IS models), and (d) the synchronicity change could be reversed by the addition of other inhibitory inputs to the soma, but beyond the levels of the control network. Thus, generic somatic inhibition acts as a pacemaker of persistent activity and FS specific inhibition modulates the output of the pacemaker.

Dendritic nonlinearities reduce network size requirements and mediate ON and OFF states of persistent activity in a PFC microcircuit model.

Technological advances have unraveled the existence of small clusters of co-active neurons in the neocortex. The functional implications of these microcircuits are in large part unexplored. Using a heavily constrained biophysical model of a L5 PFC microcircuit, we recently showed that these structures act as tunable modules of persistent activity, the cellular correlate of working memory. Here, we investigate the mechanisms that underlie persistent activity emergence (ON) and termination (OFF) and search for the minimum network size required for expressing these states within physiological regimes. We show that (a) NMDA-mediated dendritic spikes gate the induction of persistent firing in the microcircuit. (b) The minimum network size required for persistent activity induction is inversely proportional to the synaptic drive of each excitatory neuron. (c) Relaxation of connectivity and synaptic delay constraints eliminates the gating effect of NMDA spikes, albeit at a cost of much larger networks. (d) Persistent activity termination by increased inhibition depends on the strength of the synaptic input and is negatively modulated by dADP. (e) Slow synaptic mechanisms and network activity contain predictive information regarding the ability of a given stimulus to turn ON and/or OFF persistent firing in the microcircuit model. Overall, this study zooms out from dendrites to cell assemblies and suggests a tight interaction between dendritic non-linearities and network properties (size/connectivity) that may facilitate the short-memory function of the PFC.

An Architecture for the Acceleration of a Hybrid Leaky Integrate and Fire SNN on the Convey HC-2ex FPGA-Based Processor

Neuromorphic computing is expanding by leaps and bounds through custom integrated circuits (digital and analog), and large scale platforms developed by industry or government-funded projects (e.g. TrueNorth and BrainScaleS, respectively). Whereas the trend is for massive parallelism and neuromorphic computation in order to solve problems, such as those that may appear in machine learning and deep learning algorithms, there is substantial work on brain-like highly accurate neuromorphic computing in order to model the human brain. In such a form of computing, spiking neural networks (SNN) such as the Hodgkin and Huxley model are mapped to various technologies, including FPGAs. In this work, we present a highly efficient FPGA-based architecture for the detailed hybrid Leaky Integrate and Fire SNN that can simulate generic characteristics of neurons of the cerebral cortex. This architecture supports arbitrary, sparse O(n2) interconnection of neurons without need to re-compile the design, and plasticity rules, yielding on a four-FPGA Convey 2ex hybrid computer a speedup of 923x for a non-trivial data set on 240 neurons vs. the same model in the software simulator BRAIN on a Intel(R) Xeon(R) CPU E5-2620 v2 @ 2.10GHz, i.e. the reference state-of-the-art software. Although the reference, official software is single core, the speedup demonstrates that the application scales well among multiple FPGAs, whereas this would not be the case in general-purpose computers due to the arbitrary interconnect requirements. The FPGA-based approach leads to highly detailed models of parts of the human brain up to a few hundred neurons vs. a dozen or fewer neurons on the reference system.

Introduction to the Computational Neuroscience Special Section

Many of us played computer games during our childhood and youth, and some of us still do, while others decided to do something seemingly more useful – like trying to better understand the mammalian brain. Many computer games have a linear progression; individual levels are subdivisions of a larger, more complex world. The practical advantage of having levels is that they divide a game into manageable sections. Upon completion of the easy-to-master entry levels, difficulty and complexity increase and often, prior knowledge and acquired skills are necessary to advance further. A computer game analogy can be applied quite easily to the task of understanding the mammalian brain. Obviously, in this ‘game’, we enter a multiplayer environment; prior knowledge has been and is acquired by many ‘players’ in all areas of neuroscience. While physically coexisting within the same structure, each level presents its own themes, rewards and challenges. In the areas of molecular and cellular, cognitive, developmental, behavioural and clinical neuroscience, ‘players’ progress towards prevailing milestones and research goals. In many cases, it is while we are dealing with the difficulties and specifics that come with our individual entry level that we realize for the first time the complexities and begin to grasp the dimension and complexity of the games’ world map. We find out how little we know as individuals and realize how important it is to join forces with other ‘players’ to crosslink different areas and to link different levels of understanding. This is where theory becomes important. It can provide the quantitative and intellectual framework for linking different areas and different levels of understanding of the nervous system. Models that recapitulate biology on one level can help to generate testable predictions from empirical data and motivate experiments on other levels and in other areas. It can be used, for example, to predict how individual proteins like gap-junction channels may affect the synchrony of networks of many neurons within disease-relevant brain regions such as the basal ganglia. Schwab et al. (2016) use such an approach in this issue and predict that dopamine-regulated gap-junction conductance could be involved in the development of synchrony in the basal ganglia in Parkinson’s disease. The spatially localized collective activity of neuronal populations is described by the current-source density. Gratiy et al. (2017) challenge the basic assumptions of the current-source density analysis and show that the extracellular potential is determined not only by the transmembrane currents, but also by extracellular diffusive currents. The authors further estimate the effect of extracellular diffusion in in vivo LFP recordings from the visual cortex and show an effect only at low frequencies. At the single neuron level, empirical data on how ion channels determine the excitability of neurons are robust for some neurons such as the CA1 pyramidal neuron of the hippocampus and detailed biophysical models are already available. The study by Migliore et al. (2016) uses such a full-morphology biophysical model of single neurons to predict the effect of electrical fields, for example those generated by power lines, on neuronal excitability. This theoretical framework could now be extended to the network and cognitive levels to predict potential influences of external electric fields on higher brain functions. One plasticity mechanism that has been extensively integrated into circuit models and models of learning and memory is synaptic modification. While bidirectional changes in synaptic strength have been convincingly demonstrated to occur at many different types of synapses involving different signalling pathways and induction mechanisms in vitro, the relevance of particular synaptic learning rules for learning and memory is still quite unclear. In this issue, Wilmes et al. (2016) simulate how synaptic plasticity of inhibitory synapses may impact the directed flow of information within single neurons and elegantly link the single neuron functional level with the circuit level. Jezdrzejewska-Szmek et al. (2016) develop a calcium-based plasticity rule in a model of spiny projection neurons of the striatum. This model integrates calcium dynamics including diffusion, buffering and extrusion to predict the direction of synaptic plasticity. It is a perfect example of how a computational model could link the molecular level (calcium concentration) to network dynamics and memory processes. These levels are also linked computationally by the work of Sweeney & Clopath (2016). They provide a computational approach to a phenomenon, for which the empirical data are clearly not complete, because it is hard to test experimentally in reduced preparations and even harder in vivo. However, the consequences on network organization caused by neurotransmitter diffusion through cellular membranes and the distancedependent effects on the induction of synaptic plasticity could be of more general relevance for many other networks. Overall, theory is an essential tool for interpreting experiments. It is also allowing us to generate cross-level experimentally testable predictions and thus leads us to greater understanding of brain function. In neuroscience and in computer games, access to new levels often requires learning and progression gained from other levels. Only in rare cases does a single area provide direct access point to many other areas. Understanding the mammalian brain is a ‘game’ too complex to be completed and solved by individual ‘players’. Correspondence: Panayiota Poirazi, as above. E-mail: poirazi@imbb.forth.gr

The role of microcircuits in the pre-frontal cortex in detecting and encoding temporally patterned information

Working memory is the capacity of the brain to hold information temporarily for immediate use. The neural correlate of this type of short-term memory is persistent spiking activity (PA) of both excitatory and inhibitory neurons, mainly in the pre-frontal cortex (PFC). Neurons embedded in PFC microcircuits integrate widespread information from various brain regions. PFC microcircuits of ~ten neurons are characterized by highly reciprocal connections and non-linear integration in their dendrites. Understanding how dendrites integrate information from multiple sources is crucial to explain their functional role. Spatial and temporal integration of signals occurs at dendrites before propagating to the soma, and plays a role in coding of information [1,2]. The main question is whether and how pyramidal neurons in such circuits can detect temporally structured information in order to timely adjust behavior. Dendrites of cortical pyramidal neurons have been shown to exhibit temporal sensitivity [1]. To address this question, we constructed a detailed PFC microcircuit, implemented in the NEURON simulation environment. We used reconstructed morphologies of layer 5 PFC pyramidal neurons, validated against experimental findings. The microcircuit consisted of nine pyramidal neurons and two interneurons, all interconnected [3]. To investigate temporal coding, we delivered temporally structured input to four out of the nine pyramidal neurons and assessed, given persistent activity emergence, a) the time-to-first-spike (ttfs) of each pyramidal cell and b) the Inter-Spike Intervals (ISIs) during PA. We find that temporally patterned inputs (simulated as different temporal orders of activated neurons) induce different responses, in both the ttfs and ISI distributions. To investigate the mechanisms that mediate this type of coding, we varied both the stimulus frequency and the pyramidal neuron morphologies in the microcircuit. We found that lower stimulus frequencies resulted in increased differences between various temporal orders of activation. The same result was observed when using neurons with morphologically complex dendritic trees, indicating that dendritic morphology may play a key role in the ability of PFC microcircuits to detect and encode temporally patterned inputs. Further, we investigated whether this type of temporal coding exhibits specificity, i.e. whether the PFC can consistently interpret similar temporally patterned inputs in a spectrum of background activity environments. Overall, this study seeks to understand at what level and how neuronal circuits implement the timely firing of PFC pyramidal neurons. Improving our understanding of temporal coding in complex areas like the PFC is essential for disentangling how the brain detects, encodes and transmits information. While dendrites of cortical neurons have been shown to detect differences in the order of incoming signals, our study breaks new ground in finding whether such a temporal code is preserved at the microcircuit level. These findings can be tested experimentally to further investigate the role of temporal coding in higher-level areas of the brain.

Dendritic nonlinearities enable PFC microcircuits to serve as predictive modules of persistent activity

The ability to monitor and probe the activity of large neuronal networks both in vivo and in vitro has recently established that neurons of various brain regions are organized into spatially restricted clusters (or small assemblies) that are bi-directionally connected, share common inputs and are co-activated during behavioral tasks [1,2]. Investigations regarding the functional implications of such neuronal clustering have proposed that this modularity may underlie the spiking irregularities seen in cortical activity in vivo [3] or code for the execution of a voluntary movement [4]. In the prefrontal cortex (PFC), such microcircuits are proposed to support the spontaneous emergence of Up and Down states [5], a phenomenon linked to persistent activity, which is the cellular correlate of working memory. In this work we investigate the functional role of PFC microcircuits in the expression of persistent activity, focusing on the contribution of nonlinear dendritic properties to the induction, termination, and coding of upcoming state transitions. Towards this goal we developed a layer V PFC microcircuit consisting of 7 pyramidal neurons and 2 interneurons implemented in the NEURON simulation environment. Modelling equations for the biophysical mechanisms used have been reported in [6,7]. All neuron models were biophysically detailed but morphologically simplified and were validated regarding their intrinsic, synaptic and connectivity properties (e.g. number of synapses, latencies etc). Our results show that the non-linear integration of synaptic inputs at the basal dendrites of pyramidal neurons, mediated by the induction of NMDA-spikes, is imperative for the emergence of the persistent state in the microcircuit, but this necessity disappears when increasing the network size. Moreover dendritic versus somatic specific alterations of ionic currents (such as the R type VGCCs) differentially modulate persistent activity induction, substantiating the critical role of location specific effects of various neuromodulators. Finally, we find that several features of the network activity prior to the induction and/or termination of persistent firing contain predictive information of the upcoming state-transition(s), which is readily available to downstream neurons. These findings suggest that PFC microcircuits may serve as tunable and predictive modules of persistent activity and subsequently working memory.

Mechanisms underlying the emergence of Up and Down states in a model PFC microcircuit

Up and Down states are oscillations between periods of prolonged activity (Up state) and quiescence (Down state) and are recorded both in vivo and in vitro in layer V prefrontal cortex (PFC) pyramidal neurons. Biophysical mechanisms that have been proposed to underlie this phenomenon include the balance of excitation and inhibition within local PFC networks [1] along with certain intrinsic membrane mechanisms such as the afterdepolarization [2]. Using a biophysical compartmental network model of PFC layer V pyramidal neurons that incorporates anatomical data (as described in [3]), we investigated the role of synaptic input, intrinsic currents and local interconnectivity in the following features of Up and Down states: (a) the emergence of Up and Down states, (b) the duration of Up states, (c) the frequency of Up states and (d) the firing frequency during the Up state. We found that Up and Down states could emerge in our model microcircuit (see Figure ​Figure1),1), provided the existence of background synaptic activity. Among the various conditions we examined, statistically significant results were obtained when: Figure 1 Representative trace (black) of Up and Down states. Blue trace is the signal after it has been filtered with the Butterworth filter. Red boxes are indicative of Down States and an Up state that meets the criteria (500 ms duration and above -60 mV depolarization ... - Increasing the firing frequency of the background synaptic input or the number of activated background synapses (Up frequency increased by ~150% and 60%, respectively, firing frequency increased by ~30% and 50%, respectively). - Blocking the NMDA current, while compensating for the reduced excitability by enhancing the AMPA current (no emergence of Up and Down states). - Increasing the iNMDA-to-iAMPA ratio from 1 to 1.5 (Up frequency increased ~190%, firing frequency increased by 25%, Up duration doubled). - Activating the dADP mechanism at a physiological value (4mV) (Up frequency increased by ~ 200%, firing frequency increased by 60%, Up duration doubled).