How does experience shape the brain?
Research in the Barth Lab is focused on understanding how experience assembles and alters the properties of neural circuits in the cerebral cortex, in both normal and disease states. The lab has a specific focus on somatosensation in the mouse model system, where specific types of sensory input from the skin are used to drive neural activity to change the strength of synaptic connections and the firing output of cortical neurons. This neural plasticity can result in enhanced perceptual capabilities and influence subsequent learning. A detailed examination of how synapses are changed by experience is revealing fundamental principles about both perception and learning across many neural systems. In addition, researchers in the lab are using electrophysiological recordings, electron microscopy, and computational modeling to understand how functional networks are constructed and optimized in the neocortex. Experiments take advantage of transgenic mice to manipulate gene expression and label defined neural subsets and whole-cell recording and imaging to quantitate the electrical properties of cortical neurons. Ongoing projects include:
1. Algorithms for learning:
How is information from the external world transformed by cortical circuits during both normal sensory processing as well as adaptively, during learning? Decades of research have focused on neural decoding – a stimulus focused-approach to estimate the probability or properties of a sensory stimulus from the pattern of spikes collected from increasingly large populations of neurons. In contrast, we are interested in neural algorithms: the small-scale, laminar and translaminar computations deployed by defined subtypes of neocortical neurons in the receipt and transfer of sensory input. This bottom-up approach seeks to identify biologically-grounded procedures by which neurons transform inputs to outputs, constrained by deep knowledge about the identity of molecularly-distinct subsets of neurons connected through highly-specified networks. Our experimental data drive predictions about how these processes can be altered during different brain states, such as attention, reward, or fatigue, as well as how they enable experience-dependent plasticity. Using state-of-the-art genetic tools for pathway-specific activation and cell-type specific electrophysiological recordings, we are developing insight into the low- and high-level computations that are carried out by increasingly large assemblies of neurons. We have identified critical cellular and synaptic control points, embedded in the 6-layered architecture of the circuit, that enable the cortical algorithm to adapt and change during behaviorally-relevant training. The sequence of these changes – which synapses, on which cells, in which layers – shed light onto how this neocortical circuit has become specialized for plasticity. Insights from biological systems can thus provide inspiration for engineered circuits for artificial intelligence.
2. High-throughput, fluorescence-based methods for synapse detection and connectomics:
Anatomical methods for determining cell-type specific connectivity are essential to inspire and constrain our understanding of neural circuit function across development, during learning, and in disease states. We have developed new genetically-encoded reagents for fluorescence-synapse labeling and connectivity analysis in brain tissue that are designed for high-throughput, compartment-specific localization of synapses across diverse neuron types in the mammalian brain. High-resolution confocal image stacks of virally-transduced neurons can be used for 3D reconstructions of postsynaptic cells, automated detection of synaptic puncta, and multichannel fluorescence alignment of dendrites, synapses, and presynaptic neurites to assess cell-type specific connectivity. We are using these fluorescence-based reagents to quantitatively evaluate changes in synaptic connectivity during learning and in mouse models of neurological disorders such as Alzheimer’s and Parkinson’s disease. The vast number of fluorescently-labeled, input- and target-specified synapses we are collecting offers new and exciting opportunities for data analysis and machine learning.
3. Network assembly and optimization using principles of neural design:
Neurons within the neocortex are connected in stereotyped ways to generate complex but reproducible patterns of activity. We are interested in how these principles enable effective information transfer and network plasticity. We are working with computer scientists to identify and adapt biological principles into engineered networks to inspire new architectures for information storage, working memory, and learning.