Search

The fluid movement of an arm is controlled by multiple parameters that can be set independently. Recent studies argue that arm movements are generated by the collective dynamics of neurons in motor cortex. But how these collective dynamics simultaneously encode and control multiple parameters of movement is an open question. Using a task where monkeys made sequential, varied arm movements, we show that the direction and urgency of arm movements are simultaneously encoded in the low-dimensional trajectories of population activity: each movement’s direction by a fixed, looped neural trajectory and its urgency by how quickly that trajectory was traversed. Network models showed this latent coding is potentially advantageous as it allows the direction and urgency of arm movement to be independently controlled. Our results suggest how low-dimensional neural dynamics can define multiple parameters of goal-directed movement simultaneously.

The striatum's complex microcircuit is made by connections within and between its D1- and D2-receptor expressing projection neurons and at least five species of interneuron. Precise knowledge of this circuit is likely essential to understanding striatum's functional roles and its dysfunction in a wide range of movement and cognitive disorders. We introduce here a Bayesian approach to mapping neuron connectivity using intracellular recording data, which lets us simultaneously evaluate the probability of connection between neuron types, the strength of evidence for it, and its dependence on distance. Using it to synthesize a complete map of the mouse striatum, we find strong evidence for two asymmetries: a selective asymmetry of projection neuron connections, with D2 neurons connecting twice as densely to other projection neurons than do D1 neurons, but neither subtype preferentially connecting to another; and a length-scale asymmetry, with interneuron connection probabilities remaining non-negligible at more than twice the distance of projection neuron connections. We further show that our Bayesian approach can evaluate evidence for wiring changes, using data from the developing striatum and a mouse model of Huntington's disease. By quantifying the uncertainty in our knowledge of the microcircuit, our approach reveals a wide range of potential striatal wiring diagrams consistent with current data.SIGNIFICANCE STATEMENTTo properly understand a neuronal circuit's function, it is important to have an accurate picture of the rate of connection between individual neurons and how this rate changes with the distance separating pairs of neurons. We present a Bayesian method for extracting this information from experimental data and apply it to the mouse striatum, a subcortical structure involved in learning and decision-making, which is made up of a variety of different projection neurons and interneurons. Our resulting statistical map reveals not just the most robust estimates of the probability of connection between neuron types, but also the strength of evidence for them, and their dependence on distance.

Investigating the strategies engaged by subjects in decision making and learning requires tracking their choice strategies on a trial-by-trial basis. Here we present a simple but effective probabilistic approach to tracking choice strategies at trial resolution, using Bayesian evidence accumulation. We show this approach identifies both successful learning and the exploratory strategies used in decision tasks performed by humans, non-human primates, rats, and synthetic agents. We find learning occurs earlier and more often than estimated using classical approaches. Also, win-stay and lose-shift strategies, often considered as complementary, are consistently used independently, with the adoption of lose-shift preceding both learning new rules and switching away from old rules. Our approach can be extended to any discrete choice strategy, and its low computational cost is ideally suited for real-time analysis and closed-loop control.

In monogenetic generalized forms of dystonia, in vitro neurophysiological recordings have demonstrated direct evidence for abnormal plasticity at the level of the cortico‐striatal synapse. It is unclear whether similar abnormalities contribute to the pathophysiology of cervical dystonia, the most common type of focal dystonia. We investigated whether abnormal cortico‐striatal synaptic plasticity contributes to abnormal reward‐learning behavior in patients with focal dystonia. Forty patients and 40 controls performed a reward gain and loss avoidance reversal learning task. Participant's behavior was fitted to a computational model of the basal ganglia incorporating detailed cortico‐striatal synaptic learning rules. Model comparisons were performed to assess the ability of four hypothesized receptor specific abnormalities of cortico‐striatal long‐term potentiation (LTP) and long‐term depression (LTD): increased or decreased D1:LTP/LTD and increased or decreased D2: LTP/LTD to explain abnormal behavior in patients. Patients were selectively impaired in the post‐reversal phase of the reward task. Individual learning rates in the reward reversal task correlated with the severity of the patient's motor symptoms. A model of the striatum with decreased D2:LTP/ LTD best explained the patient's behavior, suggesting excessive D2 cortico‐striatal synaptic depotentiation could underpin biased reward‐learning in patients with cervical dystonia. Reversal learning impairment in cervical dystonia may be a behavioral correlate of D2‐specific abnormalities in cortico‐striatal synaptic plasticity. Reinforcement learning tasks with computational modeling could allow the identification of molecular targets for novel treatments based on their ability to restore normal reward‐learning behavior in these patients., Here we demonstrate that reward‐learning in patients with cervical dystonia is abnormal and that this can be directly related to impaired plasticity at the cortico‐striatal synapse. These results support the idea that a common link between focal and generalized forms of dystonia (where animal models have provided direct evidence aberrant of striatal plasticity) is the presence of abnormal cortico‐striatal plasticity. In the future, treatments which target this may be used clinically to alleviate symptoms of this common movement disorder.

If spikes are the medium, what is the message? Answering that question is driving the development of large-scale, single neuron resolution recordings from behaving animals, on the scale of thousands of neurons. But these data are inherently high-dimensional, with as many dimensions as neurons - so how do we make sense of them? For many the answer is to reduce the number of dimensions. Here I argue we can distinguish weak and strong principles of neural dimension reduction. The weak principle is that dimension reduction is a convenient tool for making sense of complex neural data. The strong principle is that dimension reduction shows us how neural circuits actually operate and compute. Elucidating these principles is crucial, for which we subscribe to provides radically different interpretations of the same neural activity data. I show how we could make either the weak or strong principles appear to be true based on innocuous looking decisions about how we use dimension reduction on our data. To counteract these confounds, I outline the experimental evidence for the strong principle that do not come from dimension reduction; but also show there are a number of neural phenomena that the strong principle fails to address. To reconcile these conflicting data, I suggest that the brain has both principles at play., 27 pages, 5 figures

Calcium imaging is a powerful tool for capturing the simultaneous activity of large populations of neurons. Here we determine the extent to which our inferences of neural population activity, correlations, and coding depend on our choice of whether and how we deconvolve the calcium time-series into spike-driven events. To this end, we use a range of deconvolution algorithms to create nine versions of the same calcium imaging data obtained from barrel cortex during a pole-detection task. Seeking suitable values for the deconvolution algorithms’ parameters, we optimise them against ground-truth data, and find those parameters both vary by up to two orders of magnitude between neurons and are sensitive to small changes in their values. Applied to the barrel cortex data, we show that a substantial fraction of the processing methods fail to recover simple features of population activity in barrel cortex already established by electrophysiological recordings. Raw calcium time-series contain an order of magnitude more neurons tuned to features of the pole task; yet there is also qualitative disagreement between deconvolution methods on which neurons are tuned to the task. Finally, we show that raw and processed calcium time-series qualitatively disagree on the structure of correlations within the population and the dimensionality of its joint activity. Collectively, our results show that properties of neural activity, correlations, and coding inferred from calcium imaging are sensitive to the choice of if and how spike-evoked events are recovered. We suggest that quantitative results obtained from population calcium-imaging be verified across multiple processed forms of the calcium time-series.

Medial prefrontal cortex (mPfC) activity represents information about the state of the world, including present behaviour, such as decisions, and the immediate past, such as short-term memory. Unknown is whether information about different states of the world are represented in the same mPfC neural population and, if so, how they are kept distinct. To address this, we analyse here mPfC population activity of rats learning rules in a Y-maze, with self-initiated choice trials to an arm-end followed by a self-paced return during the inter-trial interval (ITI). We find that trial and ITI population activity from the same population fall into different low-dimensional subspaces. These subspaces encode different states of the world: multiple features of the task can be decoded from both trial and ITI activity, but the decoding axes for the same feature are roughly orthogonal between the two task phases, and the decodings are predominantly of features of the present during the trial but features of the preceding trial during the ITI. These subspace distinctions are carried forward into sleep, where population activity is preferentially reactivated in post-training sleep, but differently for activity from the trial and ITI subspaces. Our results suggest that the problem of interference when representing different states of the world is solved in mPfC by population activity occupying different subspaces for the world states, which can be independently decoded by downstream targets and independently addressed by upstream inputs.Significance statementActivity in the medial prefrontal cortex plays a roles in representing the current and past states of the world. We show that during a maze task the activity of a single population in medial prefrontal cortex represents at least two different states of the world. These representations were sequential and sufficiently distinct that a downstream population could separately read out either state from that activity. Moreover, the activity representing different states is differently reactivated in sleep. Different world states can thus be represented in the same medial prefrontal cortex population, but in such a way that prevents potentially catastrophic interference between them.

Perceptual decision making is an active process where animals move their sense organs to extract task-relevant information. To investigate how the brain translates sensory input into decisions during active sensation, we developed a mouse active touch task where the mechanosensory input can be precisely measured and that challenges animals to use multiple mechanosensory cues. Mice were trained to localise a pole using a single whisker and to report their decision by selecting one of three choices. Using high-speed imaging and machine vision we estimated whisker-object mechanical forces at millisecond resolution. Mice solved the task by a sensory-motor strategy where both the strength and direction of whisker bending were informative cues to pole location. We found competing influences of immediate sensory input and choice memory on mouse choice. On correct trials, choice could be predicted from the direction and strength of whisker bending, but not from previous choice. In contrast, on error trials, choice could be predicted from previous choice but not from whisker bending. This study shows that animal choices during active tactile decision making can be predicted from mechanosenory and choice-memory signals; and provides a new task, well-suited for future study of the neural basis of active perceptual decisions.SIGNIFICANCE STATEMENTDue to the difficulty of measuring the sensory input to moving sense organs, active perceptual decision making remains poorly understood. The whisker system provides a way forward since it is now possible to measure the mechanical forces due to whisker-object contact during behaviour. Here we train mice in a novel behavioural task that challenges them to use rich mechanosensory cues, but can be performed using one whisker and enables task-relevant mechanical forces to be precisely estimated. This approach enables rigorous study of how sensory cues translate into action during active, perceptual decision making. Our findings provide new insight into active touch and how sensory/internal signals interact to determine behavioural choices.

Movement disorders arise from the complex interplay of multiple changes to neural circuits. Successful treatments for these disorders could interact with these complex changes in myriad ways, and as a consequence their mechanisms of action and their amelioration of symptoms are incompletely understood. Using Parkinson’s disease as a case-study, we review here how computational models are a crucial tool for taming this complexity, across causative mechanisms, consequent neural dynamics, and treatments. For mechanisms, we review models that capture the effects of losing dopamine on basal ganglia function; for dynamics, we discuss models that have transformed our understanding of how beta-band (15-30 Hz) oscillations arise in the parkinsonian basal ganglia. For treatments, we touch on the breadth of computational modelling work trying to understand the therapeutic actions of deep brain stimulation. Collectively, models from across all levels of description are providing a compelling account of the causes, symptoms, and treatments for Parkinson’s disease.

The joint activity of neural populations is high dimensional and complex. One strategy for reaching a tractable understanding of circuit function is to seek the simplest dynamical system that can account for the population activity. By imaging Aplysia’s pedal ganglion during fictive locomotion, here we show that its population-wide activity arises from a low-dimensional spiral attractor. Evoking locomotion moved the population into a low-dimensional, periodic, decaying orbit - a spiral - in which it behaved as a true attractor, converging to the same orbit when evoked, and returning to that orbit after transient perturbation. We found the same attractor in every preparation, and could predict motor output directly from its orbit, yet individual neurons’ participation changed across consecutive locomotion bouts. From these results, we propose that only the low-dimensional dynamics for movement control, and not the high-dimensional population activity, are consistent within and between nervous systems. DOI: http://dx.doi.org/10.7554/eLife.27342.001, eLife digest In all animals, neurons in the brain work together to generate movement. From a slug’s ability to crawl, to your ability to move your hand, movement is dependent on hundreds or thousands of neurons being active at the same time. Rhythmic movements such as crawling or swimming show this clearly: groups of neurons fire together and remain silent together in a repeating sequence, producing waves of muscle contraction. But do we need to understand the activity of each of the hundreds or thousands of individual neurons to know how they generate these movements? Bruno et al. argue that we do not, and propose instead that brain circuits that generate movement show a few set patterns of activity. By recording the activity of a population of neurons, we can identify the pattern of activity that generates a particular movement. To illustrate this point, Bruno et al. examined the network of neurons that drives the rhythmic crawling movement of the sea slug Aplysia. The results show that the network of neurons seems to contain many different patterns of activity during crawling. Yet collectively these different patterns are reflections of a simpler hidden system, a spiral of ever-decreasing, oscillating activity. This pattern is referred to as a spiral attractor because whenever the network is activated, the overall pattern of activity is always pulled into this spiral regardless of its starting point. The same applies whenever the network is disturbed. The key thing to note, however, is that individual neurons within the network do not show the same activity each time the network is active. This means that only the spiral attractor itself, and not the contribution of the individual neurons, is constant every time the sea slug crawls. What do we need to know to understand the brain? The results presented by Bruno et al. suggests that identifying the hidden systems that underlie seemingly complex and varying neural activity is the key to understanding how brains generate movement. This may also be true for how brains form memories, make decisions, and give rise to sight, hearing and touch. DOI: http://dx.doi.org/10.7554/eLife.27342.002

Deep brain stimulation (DBS) is a remarkably successful treatment for the motor symptoms of Parkinson's disease. High-frequency stimulation of the subthalamic nucleus (STN) within the basal ganglia is a main clinical target, but the physiological mechanisms of therapeutic STN DBS at the cellular and network level are unclear. We set out to begin to address the hypothesis that a mixture of responses in the basal ganglia output nuclei, combining regularized firing and inhibition, is a key contributor to the effectiveness of STN DBS. We used our computational model of the complete basal ganglia circuit to show how such a mixture of responses in basal ganglia output naturally arises from the network effects of STN DBS. We replicated the diversification of responses recorded in a primate STN DBS study to show that the model's predicted mixture of responses is consistent with therapeutic STN DBS. We then showed how this 'mixture of response' perspective suggests new ideas for DBS mechanisms: first, that the therapeutic frequency of STN DBS is above 100�€ƒHz because the diversification of responses exhibits a step change above this frequency; and second, that optogenetic models of direct STN stimulation during DBS have proven therapeutically ineffective because they do not replicate the mixture of basal ganglia output responses evoked by electrical DBS.

Izhikevich (2003) proposed a new canonical neuron model of spike generation. The model was surprisingly simple yet able to accurately replicate the firing patterns of different types of cortical cell. Here, we derive a solution method that allows efficient simulation of the model.

The prefrontal cortex is thought to learn the relationships between actions and their outcomes. But little is known about what changes to population activity in prefrontal cortex are specific to learning these relationships. Here we characterise the plasticity of population activity in the medial prefrontal cortex of male rats learning rules on a Y-maze. First, we show that the population always changes its patterns of joint activity between the periods of sleep either side of a training session on the maze, irrespective of successful rule learning during training. Next, by comparing the structure of population activity in sleep and training, we show that this population plasticity differs between learning and non-learning sessions. In learning sessions, the changes in population activity in posttraining sleep incorporate the changes to the population activity during training on the maze. In non-learning sessions, the changes in sleep and training are unrelated. Finally, we show evidence that the non-learning and learning forms of population plasticity are driven by different neuron-level changes, with the non-learning form entirely accounted for by independent changes to the excitability of individual neurons, and the learning form also including changes to firing rate couplings between neurons. Collectively, our results suggest two different forms of population plasticity in prefrontal cortex during the learning of action-outcome relationships, one a persistent change in population activity structure decoupled from overt rule-learning, the other a directional change driven by feedback during behaviour.Significance statementThe prefrontal cortex is thought to represent our knowledge about what action is worth doing in which context. But we do not know how the activity of neurons in prefrontal cortex collectively changes when learning which actions are relevant. Here we show in a trial-and-error task that population activity in prefrontal cortex is persistently changing, irrespective of learning. Only during episodes of clear learning of relevant actions are the accompanying changes to population activity carried forward into sleep, suggesting a long-lasting form of neural plasticity. Our results suggest that representations of relevant actions in prefrontal cortex are acquired by reward imposing a direction onto ongoing population plasticity.

Optical recording with fast voltage sensitive dyes makes it possible, in suitable preparations, to simultaneously monitor the action potentials of large numbers of individual neurons. Here we describe methods for doing this, including considerations of different dyes and imaging systems, methods for correlating the optical signals with their source neurons, procedures for getting good signals, and the use of Independent Component Analysis for spike-sorting raw optical data into single neuron traces. These combined tools represent a powerful approach for large-scale recording of neural networks with high temporal and spatial resolution.

The existence of multiple parallel loops connecting sensorimotor systems to the basal ganglia has given rise to proposals that these nuclei serve as a selection mechanism resolving competitions between the alternative actions available in a given context. A strong test of this hypothesis is to require a computational model of the basal ganglia to generate integrated selection sequences in an autonomous agent, we therefore describe a robot architecture into which such a model is embedded, and require it to control action selection in a robotic task inspired by animal observations. Our results demonstrate effective action selection by the embedded model under a wide range of sensory and motivational conditions. When confronted with multiple, high salience alternatives, the robot also exhibits forms of behavioral disintegration that show similarities to animal behavior in conflict situations. The model is shown to cast light on recent neurobiological findings concerning behavioral switching and sequencing.

Neurobehavioral data from intact, decerebrate, and neonatal rats suggests that the reticular formation provides a brainstem substrate for action selection in the vertebrate central nervous system. In this article, Kilmer, McCulloch and Blum’s (1969) landmark reticular formation model (see also Kilmer, 1997) is described and re-evaluated, both in simulation and, for the first time, as a mobile robot con troller. Particular model configurations are found to provide effective action selection mechanisms in a robot survival task using either simulated or physical robots. The model’s competence is dependent on the organization of afferents from model sensory systems, and a genetic algorithm search identi fied a class of afferent configurations which have long survival times. The results support our proposal that the reticular formation evolved to provide effective arbitration between innate behaviors and, with the forebrain basal ganglia, may constitute the integrative, “centrencephalic” core of vertebrate brain architecture. Additionally, the results demonstrate that the Kilmer et al. model provides an alternative form of robot controller to those usually considered in the adaptive behavior literature.

Background The recent discoveries of genes implicated in striatal dopamine transmission in patients with cervical dystonia have emphasised the central role of the basal ganglia in the pathogenesis of the condition. Dopamine’s principle role within the striatum is to bias the action selection function of the basal ganglia towards the best outcome. Our hypothesis was that abnormalities of dopamine neurotransmission would result in a measurable bias in reward based learning in patients with cervical dystonia. Methods We used a reversal learning task to assess dopamine based learning in a group of 40 patients with cervical dystonia and 40 age matched controls. Exclusion criteria included previous psychiatric diagnosis or current use of psychotropic medication. Results Patients demonstrated a consistent impairment in reversal learning performance compared to controls (15% less rewards p Conclusions Patients with cervical dystonia have impaired reward-reversal learning. Treatments which aim to enhance reinforcement learning could be explored as future options for both motor and non-motor symptoms in these patients.

We previously proposed that the basal ganglia (BG) play a crucial role in action selection. Quantitative analysis and simulation of a computational model of the intrinsic BG demonstrated that its output was consistent with this proposition. Here we build on that model by embedding it into a wider circuit containing the motor thalamocortical loop and thalamic reticular nucleus (TRN). Simulation of this extended model showed that the additions gave five main results which are desirable in a selection/switching mechanism. First, low salience actions (i.e. those with low urgency) could be selected. Second, the range of salience values over which actions could be switched between was increased. Third, the contrast between the selected and non-selected actions was enhanced via improved differentiation of outputs from the BG. Fourth, transient increases in the salience of a non-selected action were prevented from interrupting the ongoing action, unless the transient was of sufficient magnitude. Finally, the selection of the ongoing action persisted when a new closely matched salience action became active. The first result was facilitated by the thalamocortical loop; the rest were dependent on the presence of the TRN. Thus, we conclude that the results are consistent with these structures having clearly defined functions in action selection.