Elucidating the diversity and spatial organization of cell types in the brain is an essential goal of neuroscience, with many emerging technologies helping to advance this endeavor. Using a new in situ hybridization method that can measure the expression of hundreds of genes in a given mouse brain section (amplified seqFISH), Shah et al. (2016) describe a spatial organization of hippocampal cell types that differs from previous reports. In seeking to understand this discrepancy, we find that many of the barcoded genes used by seqFISH to characterize this spatial organization, when cross-validated by other sensitive methodologies, exhibit negligible expression in the hippocampus. Additionally, the results of Shah et al. (2016) do not recapitulate canonical cellular hierarchies and improperly classify major neuronal cell types. We suggest that, when describing the spatial organization of brain regions, cross-validation using multiple techniques should be used to yield robust and informative cellular classification. This Matters Arising paper is in response to Shah et al. (2016), published in Neuron. See also the response by Shah et al. (2017), published in this issue.

Cognitive maps confer animals with flexible intelligence by representing spatial, temporal, and abstract relationships that can be used to shape thought, planning, and behavior. Cognitive maps have been observed in the hippocampus, but their algorithmic form and the processes by which they are learned remain obscure. Here, we employed large-scale, longitudinal two-photon calcium imaging to record activity from thousands of neurons in the CA1 region of the hippocampus while mice learned to efficiently collect rewards from two subtly different versions of linear tracks in virtual reality. The results provide a detailed view of the formation of a cognitive map in the hippocampus. Throughout learning, both the animal behavior and hippocampal neural activity progressed through multiple intermediate stages, gradually revealing improved task understanding and behavioral efficiency. The learning process led to progressive decorrelations in initially similar hippocampal neural activity within and across tracks, ultimately resulting in orthogonalized representations resembling a state machine capturing the inherent structure of the task. We show that a Hidden Markov Model (HMM) and a biologically plausible recurrent neural network trained using Hebbian learning can both capture core aspects of the learning dynamics and the orthogonalized representational structure in neural activity. In contrast, we show that gradient-based learning of sequence models such as Long Short-Term Memory networks (LSTMs) and Transformers do not naturally produce such representations. We further demonstrate that mice exhibited adaptive behavior in novel task settings, with neural activity reflecting flexible deployment of the state machine. These findings shed light on the mathematical form of cognitive maps, the learning rules that sculpt them, and the algorithms that promote adaptive behavior in animals. The work thus charts a course toward a deeper understanding of biological intelligence and offers insights toward developing more robust learning algorithms in artificial intelligence.

To study how the brain drives cognition and behavior we need to understand its cellular composition. Advances in single-cell transcriptomics have revolutionized our ability to characterize neuronal diversity. To arrive at meaningful descriptions of cell types, however, gene expression must be linked to structural and functional properties. Axonal projection patterns are an appropriate measure, as they are diverse, change only gradually over time, and they influence and constrain circuit function. Here, we consider how efforts to map transcriptional and morphological diversity in the mouse brain could be linked to generate a modern taxonomy of the mouse brain.

To study how the brain drives cognition and behavior we need to understand its cellular composition. Advances in single-cell transcriptomics have revolutionized our ability to characterize neuronal diversity. To arrive at meaningful descriptions of cell types, however, gene expression must be linked to structural and functional properties. Axonal projection patterns are an appropriate measure, as they are diverse, change only gradually over time, and they influence and constrain circuit function. Here, we consider how efforts to map transcriptional and morphological diversity in the mouse brain could be linked to generate a modern taxonomy of the mouse brain.

Understanding the principles governing neuronal diversity is a fundamental goal for neuroscience. Here we provide an anatomical and transcriptomic database of nearly 200 genetically identified cell populations. By separately analyzing the robustness and pattern of expression differences across these cell populations, we identify two gene classes contributing distinctly to neuronal diversity. Short homeobox transcription factors distinguish neuronal populations combinatorially, and exhibit extremely low transcriptional noise, enabling highly robust expression differences. Long neuronal effector genes, such as channels and cell adhesion molecules, contribute disproportionately to neuronal diversity, based on their patterns rather than robustness of expression differences. By linking transcriptional identity to genetic strains and anatomical atlases we provide an extensive resource for further investigation of mouse neuronal cell types.

Abstract We recently described a new form of neural integration and firing in a subset of interneurons, in which evoking hundreds of action potentials over tens of seconds to minutes produces a sudden barrage of action potentials lasting about a minute beyond the inciting stimulation. During this persistent firing, action potentials are generated in the distal axon and propagate retrogradely to the soma. To distinguish this from other forms of persistent firing, we refer to it here as ’retroaxonal barrage firing’, or ’barrage firing’ for short. Its induction is blocked by chemical inhibitors of gap junctions and curiously, stimulation of one interneuron in some cases triggers barrage firing in a nearby, unstimulated interneuron. Beyond these clues, the mechanisms of barrage firing are unknown. Here we report new results related to these mechanisms. Induction of barrage firing was blocked by lowering extracellular calcium, as long as normal action potential threshold was maintained, and it was inhibited by blocking L-type voltage-gated calcium channels. Despite its calcium dependence, barrage firing was not prevented by inhibiting chemical synaptic transmission. Furthermore, loading the stimulated/recorded interneuron with BAPTA did not block barrage firing, suggesting that the required calcium entry occurs in other cells. Finally, barrage firing was normal in mice with deletion of the primary gene for neuronal gap junctions (connexin36), suggesting that non-neuronal gap junctions may be involved. Together, these findings suggest that barrage firing is probably triggered by a multicellular mechanism involving calcium signalling and gap junctions, but operating independently of chemical synaptic transmission.

As animals navigate, they must identify features within context. In the mammalian brain, the hippocampus has the ability to separately encode different environmental contexts, even when they share some prominent features. To do so, neurons respond to sensory features in a context-dependent manner; however, it is not known how this encoding emerges. To examine this, we performed electrical recordings in the hippocampus as mice navigated in two distinct virtual environments. In CA1, both synaptic input to single neurons and population activity strongly tracked visual cues in one environment, whereas responses were almost completely absent when the same cue was presented in a second environment. A very similar, highly context-dependent pattern of cue-driven spiking was also observed in CA3. These results indicate that CA1 inherits a complex spatial code from upstream regions, including CA3, that have already computed a context-dependent representation of environmental features.

Animals can learn general task structures and use them to solve new problems with novel sensory specifics. This capacity of ‘learning to learn’, or meta-learning, is difficult to achieve in artificial systems, and the mechanisms by which it is achieved in animals are unknown. As a step toward enabling mechanistic studies, we developed a behavioral paradigm that demonstrates meta-learning in head-fixed mice. We trained mice to perform a two-alternative forced-choice task in virtual reality (VR), and successively changed the visual cues that signaled reward location. Mice showed increased learning speed in both cue generalization and serial reversal tasks. During reversal learning, behavior exhibited sharp transitions, with the transition occurring earlier in each successive reversal. Analysis of motor patterns revealed that animals utilized similar motor programs to execute the same actions in response to different cues but modified the motor programs during reversal learning. Our study demonstrates that mice can perform meta-learning tasks in VR, thus opening up opportunities for future mechanistic studies.

Electrophysiology is the most used approach for the collection of functional data in basic and translational neuroscience, but it is typically limited to either intracellular or extracellular recordings. The integration of multiple physiological modalities for the routine acquisition of multimodal data with microelectrodes could be useful for biomedical applications, yet this has been challenging owing to incompatibilities of fabrication methods. Here, we present a suite of glass pipettes with integrated microelectrodes for the simultaneous acquisition of multimodal intracellular and extracellular information in vivo, electrochemistry assessments, and optogenetic perturbations of neural activity. We used the integrated devices to acquire multimodal signals from the CA1 region of the hippocampus in mice and rats, and show that these data can serve as ground-truth validation for the performance of spike-sorting algorithms. The microdevices are applicable for basic and translational neurobiology, and for the development of next-generation brain-machine interfaces.

Our ability to remember the past is essential for guiding our future behavior. Psychological and neurobiological features of declarative memories are known to transform over time in a process known as systems consolidation. While many theories have sought to explain the time-varying role of hippocampal and neocortical brain areas, the computational principles that govern these transformations remain unclear. Here we propose a theory of systems consolidation in which hippocampal-cortical interactions serve to optimize generalizations that guide future adaptive behavior. We use mathematical analysis of neural network models to characterize fundamental performance tradeoffs in systems consolidation, revealing that memory components should be organized according to their predictability. The theory shows that multiple interacting memory systems can outperform just one, normatively unifying diverse experimental observations and making novel experimental predictions. Our results suggest that the psychological taxonomy and neurobiological organization of declarative memories reflect a system optimized for behaving well in an uncertain future.