The cerebral cortex continuously adapts through learning and changes in sensory experience. A major goal of neuroscience is to understand the cellular and synaptic plasticity mechanisms that allow behavior to adapt to a changing environment. Remarkably, neural circuits maintain stable activity despite ongoing adaptive changes in synaptic strength and connectivity. Recent work has revealed multiple homeostatic plasticity mechanisms that maintain neural firing rates within an optimum range. However, it is not well understood how brain circuits maintain homeostasis in different brain regions and across different time scales.

In primary sensory cortex, changes in sensory input drive adaptive changes in cortical representations. Broadly, depriving a set of sensory inputs rapidly weakens neural responses to those inputs and gradually strengthens spared inputs. Recent studies show that deprivation also triggers rapid plasticity in inhibitory circuits that could stabilize neural activity despite ongoing weakening of deprived inputs. Chapter 2 investigates the cellular and circuit mechanisms that underlie this rapid plasticity of inhibition in the superficial layers of rodent primary somatosensory cortex (S1). This study shows that 1-day whisker deprivation weakens inhibition in layer (L) 2/3 pyramidal neurons by decreasing the intrinsic excitability of L2/3 parvalbumin-positive (PV) interneurons near spike threshold. Deprivation reduces PV spike threshold through an increase in voltage-gated potassium conductances. These findings demonstrate that activity in sensory cortex is rapidly stabilized through plasticity of PV intrinsic excitability.

Often, a single sensory manipulation drives multiple cellular and circuit changes with distinct temporal components. In cortical inhibitory circuits, the time course of plasticity is not well understood. Building on the studies in Chapter 2, we hypothesized that additional plasticity mechanisms may be recruited in L2/3 PV neurons in response to different time scales of whisker deprivation. In Chapter 3, we test this idea by extending the the duration of whisker deprivation to 3 days and asking whether changes in L2/3 PV circuits are consistent with those observed after 1 day of deprivation. We find that 3-day deprivation also decreases PV intrinsic excitability, but through a different mechanism than 1-day deprivation. Deprivation strengthens the medium afterhyperpolarization (mAHP) without affecting spike threshold. Thus, experience-dependent plasticity of cortical PV circuits involves a succession of distinct components.

Sensory information is encoded with sparse spiking in rodent sensory cortex, but the organization and functional basis of this sparse code is not well understood. I conducted two studies to characterize function and learning in the cortex. In the first study, I used population calcium imaging to study the circuit-level factors underlying sparse coding in layer (L) 2 of mouse somatosensory cortex. Whisker deflection elicited low-probability spikes in small, shifting neural ensembles spanning multiple cortical columns. Neurons within a column-sized imaging field were tuned heterogeneously to many different whiskers, contrary to standard models of somatotopy. A spectrum of whisker-evoked response probability existed across neurons that correlated strongly with spontaneous firing rate. This correlation indicates that a major component of responsiveness is independent of experimental stimulus choice. The distribution of responsivity was skewed, indicating the existence of a small population of highly-responsive neurons. Highly-responsive neurons included pyramidal cells and interneurons, and individual whisker deflections were primarily encoded by a small, stable population of highly responsive cells. L2 neurons projecting to motor (M1) and secondary somatosensory (S2) cortex differed in whisker tuning and sparseness, suggesting these intermingled populations send disparate information to their targets. Thus, sparse coding in L2 reflects heterogeneous sensory tuning, low average response probability across neurons, a skewed distribution of inherent responsiveness that includes a small number of more-active neurons, and functional specialization of S1 output streams.

In order to test whether the large pool of unresponsive neurons observed might be important in learning, I developed a novel type of brain-machine interface (BMI) based on calcium imaging in the intact cortex. In this BMI task, the mouse learned to use voluntary modulations of neural activity to control a device. The BMI design allowed for direct control over the relationship between neuronal activity and behavioral output. We trained mice to operantly control an auditory cursor using spike-related calcium signals recorded with 2-photon imaging in motor and somatosensory cortex. Mice rapidly learned to modulate activity in layer 2/3 neurons, evident both across- and within-sessions. Learning was accompanied by striking modifications of firing correlations within spatially localized networks at fine scales (10-100 microns). We found that less-active neurons, and even silent neurons, could dramatically up-modulate their firing to successfully learn the task. Neurons in a `cloud' around the BMI-controlling neurons initially exhibited task-related activity, which dampened out as the animal honed in on the specific cells controlling the device. This suggests an economization of activity, which may be reflective of the sparse firing strategies in the cortex.

These studies both point to the existence of a gradient of activity in cortical neurons in L2/3, which can nevertheless be volitionally manipulated by learning. Neurons in L2/3 of rodent somatosensory cortex had unexpectedly divergent tuning, organized in a salt-and-pepper fashion. The reliability of their responses depended mainly on intrinsic, stimulus-independent responsivity, and more minimally on tuning and downstream targets. However, even extremely inactive neurons could be induced to modify their firing dramatically by coupling their activity to reward. The network rapidly learned to minimize the number of neurons necessary to perform the task, suggesting that an economizing impetus might be at work in superficial cortex.

Distinct genetic forms of autism are hypothesized to share a common increase in excitation inhibition (E-I) ratio in cerebral cortex, causing hyperexcitability and excess spiking. We provide a systematic test of this hypothesis across 4 mouse models (Fmr1-/y, Cntnap2-/-, 16p11.2del/+, Tsc2+/-), focusing on somatosensory cortex. All mouse models with autism risk mutations showed reduced feedforward inhibition in layer 2/3 (L2/3) coupled with more modest, variable reductions in feedforward excitation, driving a common increase in E-I conductance ratio. Contrary to the classic, naïve prediction that increased E-I ratio drives spiking hyperexcitability, synaptic conductance modeling suggested that sensory inputs to this circuit would actually evoke stable or reduced firing in L2/3 in vivo. Indeed, spiking evoked by single whisker deflections was normal in the two strains whose modeling predicted stable synaptic depolarization (Cntnap2-/-, 16p11.2del/+) and was reduced in the Fmr1-/y mice, also matching that strain's modeling prediction. In addition, spiking of putative inhibitory neurons was reduced in each strain. Thus, increased E-I ratio does not necessarily yield increased spiking responses to simple sensory stimuli.

Stable or reduced spiking was surprising given the dramatically reduced inhibition observed in each strain. Thus, we tested whether the Fmr1-/y mouse exhibits excess spiking in response to multi-whisker sequences, which drive more spiking in S1 compared to single-whisker deflections and may thus engage a different synaptic E-I ratio. Total spiking to whisker sequences, and cross-whisker suppression, were reduced in Fmr1-/y mice. These effects were explained by reduced firing to the anatomically corresponding columnar whisker and increased tuning heterogeneity within each cortical column. This suggests that weak or heterogeneous sensory maps may underlie circuit abnormalities in disorders expressing autism spectrum behaviors.

Understanding how cortical circuits process sensory information and support perception is a fundamental problem in neuroscience. Rodents, being tactile experts, actively use their whiskers to sense complex tactile features like surface texture, object shape and location. In this dissertation I address how cortical neurons integrate sensory information from individual whiskers to support accurate and precise representations of complex tactile features.

Natural whisking during tactile exploration generates complex spatiotemporal sequences of whisker stimulation. Objects with different textures and shapes result in different patterns of whisker stimulation, sequentially stimulating different combinations of whiskers across time. I thus hypothesized that individual neurons in primary whisker somatosensory cortex (S1) are sensitive to specific features of tactile sequences. Chapter 2 describes the timescales at which S1 neurons integrated sensory input while rats discriminated between whisker impulse sequences that varied in single-impulse kinematics. While discrimination performance was consistent with integration at a relatively slow timescale (approximately 150ms), most S1 neurons integrated whisker input at a fast timescale (<20ms), generating a precise code for vibrotactile sequences in S1. Neurons with slower integration windows (>60ms) did not accurately represent the stimulus but were instead related to the rat’s behavioral choice. These findings show that S1 neurons encode whisker input at a fast timescale and suggest that areas downstream of S1 temporally integrate this information to guide perceptual discrimination.

Given the precise representations of tactile sequences in S1, Chapter 3 explores the elementary computations underlying tactile stimulus integration by S1 neurons. Tactile sequences vary in spatial identity of stimulated whiskers and inter-whisker-deflection-intervals (Δt). Dense stimulation of local whisker pairs over a physiological range of Δt revealed a somatopically organized rate code for whisker combinations that was precise in space and coarser in time. Sublinear suppression for suboptimal combinations sharpened tuning relative to that expected from linear integration alone; analogous to the computation of motion direction selectivity in many visual circuits, thus suggesting a common computation for spatiotemporal feature extraction. Taken together, this dissertation shows that S1 neurons integrate sensory input in space and time to generate robust tuning for spatiotemporal features of tactile scenes.

CHAPTER 1: Somatosensory maps in rodents & primates—a review

This chapter reviews basic principles and recent findings in primate, human, and rodent somatosensory maps. Topographic maps of the body surface are a major feature of somatosensory cortex. In primates, parietal cortex contains four somatosensory areas, each with its own map, with the primary cutaneous map in area 3b. Rodents have at least three somatosensory areas, and the whisker map in rodent primary somatosensory cortex is a canonical system for studying cortical microcircuits, sensory coding, and map plasticity.

Maps are not isomorphic to the body surface, but magnify behaviorally important skin regions, which include the hands and face in primates, and the whiskers in rodents. Within each map, intracortical circuits process tactile information, mediate spatial integration, and support active sensation. Functional representations are more overlapping than suggested by textbook depictions of map topography. Maps may also contain fine-scale representations of touch sub-modalities, or direction of tactile motion. In addition, somatosensory maps are plastic throughout life in response to altered use or injury.

CHAPTER 2: Somatotopic precision of whisker tuning in layer 2/3 of rat barrel cortex

Although cortical maps of the sensory periphery are topographically organized at a large scale, the fine-scale precision of the map at the level of neighboring cells varies between species, sensory modalities, and cortical layers. In rodent somatosensory cortex, where each whisker is mapped to a dedicated cortical column, cells in layer 4 of each column are predominantly tuned to the appropriate whisker (columnar whisker; CW). However, in mice, cells in the upper cortical layers display locally heterogeneous "salt-and-pepper" tuning. It remains unclear whether the same heterogeneity is found in the upper layers of the rat whisker map. To address this question, we examined whisker tuning in regular-spiking (RS) and fast-spiking (FS) units recorded from layer 2/3 whisker columns in anesthetized rats. Among RS units, we observe a narrow majority (59%) best tuned to the CW; a clear majority (88%) with the CW among a statistically comparable group of "equal best" whiskers; and a small number (12%) of units best driven by one or more surround whiskers and poorly responsive to the CW. CW-tuned units display sharper tuning and faster responses to their best whisker than non-CW-tuned units, but are distributed evenly along vertical and horizontal dimensions of the column. These results demonstrate tuning in rat L2/3 that is more homogeneous than in mouse layer 2, but less so than in rat or mouse layer 4. FS units are tuned to the CW more frequently (78%) than RS units, consistent with broadly sampled input from local excitatory cells most responsive, as a group, to the columnar whisker.

The cerebral cortex is thought to generate sensory experience and facilitate broadly integrative processing used in attention, decision making, and learning. A major goal in neuroscience is to understand how neurons in cortex interact to represent sensory change in the world. Rodents use whiskers to actively palpate their surroundings and gather tactile information. While it is known that rodents can discriminate different textures using whisker input, how the whisker system encodes local spatial features, like shape, is not well understood.

For many perceptual tasks rodents must pay attention to whisker input in order to extract relevant sensory information. Neural oscillations (gamma) may indicate attentive brain states, and Chapter 2 explores how sensory experience may shape the brain’s ability to generate these rhythms. Superficial layers of primary somatosensory cortex (S1) showed a profound reduction in recurrent inhibition after sensory deprivation, as well as a sharp reduction in spontaneous and evoked gamma oscillations. These findings suggest cortical rhythms change with sensory experience in a recurrent inhibition-dependent manner.

How the whisker system encodes spatial features, like shape is unknown. In Chapter 3 we describe an experiment designed to study basic elements of shape encoding, using a set of tactile gratings. We trained mice to discriminate smooth surfaces from rough surfaces composed of raised gratings or sandpaper. High-speed whisker imaging revealed that stick-slip micromotions clustered on discrete grating ridges, evoking spatially and temporally precise spikes in S1. At the same time, slip-responsive neurons also responded to whisker contact and texture, forming an overlapping population of tactile feature responses. Mean firing rate was higher on rough surfaces due to stick-slip-evoked spikes, and by weighting individual units we could accurately predict rough vs. smooth trials from neural activity alone. Surprisingly, 20% of neurons showed selectivity among rough textures, including tuning for specific gratings. Thus, S1 neurons simultaneously coded microscopic and macroscopic features of touch during active palpation of tactile gratings.

Presynaptic NMDA receptors ( NMDARs) modulate release and plasticity at many glutamatergic synapses, but the specificity of their expression across synapse classes has not been examined. We found that non-postsynaptic, likely presynaptic NR2B-containing NMDARs enhanced AMPA receptor-mediated synaptic transmission at layer 4 ( L4) to L2/3 ( L4-L2/3) synapses in juvenile rat barrel cortex. This modulation was apparent at room temperature when presynaptic NMDARs were activated by elevation of extracellular glutamate or application of exogenous NMDAR agonists. At near physiological temperatures, modulation of transmission by presynaptic NMDARs occurred naturally, without the need for external activation. Blockade of presynaptic NMDARs depressed unitary and extracellularly evoked EPSCs at L4 -L2/3 synapses, accompanied by increases in paired-pulse ratio and coefficient of variation, indicative of a decrease in presynaptic release probability. NMDAR agonists increased the frequency of miniature EPSCs in L2/3 neurons, without altering their amplitude or kinetics. Focal application of NMDAR antagonist revealed that the NMDARs that modulate L4-L2/3 transmission are located in L2/3, not L4, consistent with localization on terminals or axons of L4-L2/3 synapses, rather than on the somato-dendritic compartment of presynaptic L4 neurons. In contrast, presynaptic NMDARs did not modulate L4 -L4 synapses, which originate from the same presynaptic neurons as L4-L2/3 synapses, or cross-columnar L2/3-L2/3 horizontal projections, which synapse onto the same postsynaptic target neurons. Thus, presynaptic NMDARs selectively modulate L4-L2/3 synapses, relative to other synapses made by the same neurons. Existence of these receptors may support specialized processing or plasticity by L4-L2/3 synapses.