Adaptive behavior is coordinated by neuronal networks that are distributed across multiple brain regions. How cross-regional interactions reorganize during learning remains elusive. We applied multi-fiber photometry to chronically record simultaneous activity of 12-48 mouse brain regions while mice learned a tactile discrimination task. We found that with learning most regions shifted their peak activity from reward-related action to the reward-predicting stimulus. We corroborated this finding by functional connectivity estimation using transfer entropy, which revealed growth and stabilization of mesoscale networks encompassing basal ganglia, thalamus, cortex, and hippocampus, especially during stimulus presentation. The internal globus pallidus, ventromedial thalamus, and several regions in frontal cortex emerged as hub regions. Our results highlight the cooperative action of distributed brain regions to establish goal-oriented mesoscale network dynamics during learning.

The medial prefrontal cortex (mPFC) and the dorsomedial striatum (dmStr) are linked to working memory (WM) but how striatum-projecting mPFC neurons contribute to WM encoding, maintenance, or retrieval remains unclear. Here, we probed mPFC→dmStr pathway function in freely-moving mice during a T-maze alternation test of spatial WM. Fiber photometry of GCaMP6m-labeled mPFC→dmStr projection neurons revealed strongest activity during the delay period that requires WM maintenance. Demonstrating causality, optogenetic inhibition of mPFC→dmStr neurons only during the delay period impaired performance. Conversely, enhancing mPFC→dmStr pathway activity—via pharmacological suppression of HCN1 or by optogenetic activation during the delay— alleviated WM impairment induced by NMDA receptor blockade. Consistently, cellular-resolution miniscope imaging resolved preferred activation of >50% mPFC→dmStr neurons during WM maintenance. This subpopulation was distinct from neurons showing preference for encoding and retrieval. In all periods, including the delay, neuronal sequences were evident. Striatum-projecting mPFC neurons thus critically contribute to spatial WM maintenance.

No abstract

We introduce the dune-curvilineargrid module. The module provides the self-contained, parallel grid manager dune-curvilineargrid, as well as the underlying elementary curvilinear geometry module dune-curvilineargeometry. Both modules are developed as extension of the DUNE [3] project, and conform to the generic dune-grid and dune-geometry interfaces respectively. We expect the reader to be at least briefly familiar with the DUNE interface to fully benefit from this paper. dune-curvilineargrid is a part of the computational framework developed within the doctoral thesis of Aleksejs Fomins. The work is fully funded by and carried out at the technology company LSPR AG. It is motivated by the need for reliable and scalable electromagnetic design of nanooptical devices, achieved by HADES3D family of electromagnetic codes. It is of primary scientific and industrial interest to model full 3D geometric designs as close to the real fabricated structures as possible. Curvilinear geometries improve both the accuracy of modeling smooth material boundaries, and the convergence rate of PDE solutions with increasing basis function order [9], reducing the necessary computational effort. Additionally, higher order methods decrease the memory footprint of PDE solvers at the expense of higher operational intensity, which helps in extracting optimal performance from processing power dominated high performance architectures [30]. dune-curvilineargeometry is capable of modeling simplex entities (edges, triangles and tetrahedra) up to polynomial order 5 via hard-coded Lagrange polynomials, and arbitrary order via analytical procedures. Its most notable features are local-to-global and global-to-local coordinate mappings, symbolic and recursive integration, symbolic polynomial scalars, vectors and matrices (e.g. Jacobians and Integration Elements). dune-curvilineargrid uses the dune-curvilineargeometry module to provide the following functionality: fully parallel input of curvilinear meshes in the gmsh [10] mesh format, processing only the corresponding part of the mesh on each available core; mesh partitioning at the reading stage (using ParMETIS [16,21]); unique global indices for all mesh entities over all processes; Ghost elements associated with the interprocessor boundaries; interprocessor communication of data for shared entities of all codimensions via the standard DUNE data handle interface. There is also significant support for Boundary Integral (BI) codes, allowing for arbitrary number of interior boundary surfaces, as well as all-to-all dense parallel communication procedures. The dune-curvilineargrid grid manager is continuously developed and improved, and so is this documentation. For the most recent version of the documentation, as well as the source code, please refer to the following repositories

An important goal in systems neuroscience is to understand the structure of neuronal interactions, frequently approached by studying functional relations between recorded neuronal signals. Commonly used pairwise measures (e.g. correlation coefficient) offer limited insight, neither addressing the specificity of estimated neuronal relations nor potential synergistic coupling between neuronal signals. Tripartite measures, such as partial correlation, variance partitioning, and partial information decomposition, address these questions by disentangling functional relations into interpretable information atoms (unique, redundant and synergistic). Here, we apply these tripartite measures to simulated neuronal recordings to investigate their sensitivity to noise. We find that the considered measures are mostly accurate and specific for signals with noiseless sources but experience significant bias for noisy sources. We show that permutation-testing of such measures results in high false positive rates even for small noise fractions and large data sizes. We present a conservative null hypothesis for significance testing of tripartite measures, which significantly decreases false positive rate at a tolerable expense of increasing false negative rate. We hope our study raises awareness about the potential pitfalls of significance testing and of interpretation of functional relations, offering both conceptual and practical advice.

We report on the 3-dimensional full-wave analysis of optical array antennas that employ a dipole element as the fundamental building block. We use a finite element time domain (FETD) method discretized on unstructured tetrahedral grids in order to efficiently resolve the geometry which has a wide range of characteristic length scales, from the nanometer to the micrometer range. Such devices are useful in a number of applications in order to convert propagating electromagnetic energy into localized energy which is concentrated within a spot whose dimension is significantly smaller than the wavelength. This capability is especially useful for field emitter arrays (FEA) used in novel, ultra-low emittance photocathodes. The antenna elements are modeled with gold metallic properties in the optical region of the electromagnetic spectrum. There, gold is a dispersive dielectric material and desribed with a Drude dielectric material model. To support the validity of our analysis we numerically analyze electromagnetic problems that can be solved analytically, thus benchmarking the algorithm. We then computationally analyze a single dipole element in free space and a logarithmically periodic array of dipoles, similar to the concept of the Yagi-Uda array antenna in the microwave region. We demonstrate the existence of resonant modes on the dipole rod elements. Eventually, we comment on further development work.

An important goal in systems neuroscience is to understand the structure of neuronal interactions, frequently approached by studying functional relations between recorded neuronal signals. Commonly used pairwise metrics (e.g. correlation coefficient) offer limited insight, neither addressing the specificity of estimated neuronal interactions nor potential synergistic coupling between neuronal signals. Tripartite metrics, such as partial correlation, variance partitioning, and partial information decomposition, address these questions by disentangling functional relations into interpretable information atoms (unique, redundant and synergistic). Here, we apply these tripartite metrics to simulated neuronal recordings to investigate their sensitivity to impurities (like noise or other unexplained variance) in the data. We find that all considered metrics are accurate and specific for pure signals but experience significant bias for impure signals. We show that permutation-testing of such metrics results in high false positive rates even for small impurities and large data sizes. We present a conservative null hypothesis for significance testing of tripartite metrics, which significantly decreases false positive rate at a tolerable expense of increasing false negative rate. We hope our study raises awareness about the potential pitfalls of significance testing and of interpretation of functional relations, offering both conceptual and practical advice.