Intra-individual Physiomic Landscape of Pyramidal Neurons in the Human Neocortex

Computation within cortical microcircuits is determined by functional properties of the neurons and their synaptic interactions. While heterogeneity of inhibitory interneurons is well established, the anatomical, physiological, and molecular differentiation of excitatory pyramidal neurons is not fully resolved. To identify functional subtypes within the pyramidal neuron population, we focused on human layer 2-3 cortex which greatly expanded during evolution. We performed multi-neuron patch-clamp recordings in brain slices from the temporal cortex of 22 epilepsy patients. We characterized the electrophysiological properties of up to 80 pyramidal neurons per patient, enabling us to assess inter- and intra-individual functional variability. Hierarchical clustering of the high-dimensional parameter space yielded functionally distinct clusters of pyramidal neurons which were present across individuals. This may represent a generic organizational principle converging with previously described transcriptomic heterogeneity. We further observed substantial heterogeneity in physiological parameters with intra-individual variability being severalfold larger than inter-individual variability. The phenotypic variability within and across pyramidal neuron subtypes has important implications for the computational capacity of the cortical microcircuit.

Synaptic Interactions in Scorpion Peg Sensilla Appear to Maintain Chemosensory Neurons within Dynamic Firing Range

Scorpions have elaborate chemo-tactile organs called pectines on their ventral mesosoma. The teeth of the comb-like pectines support thousands of minute projections called peg sensilla (a.k.a. “pegs”), each containing approximately 10 chemosensory neurons. Males use pectines to detect pheromones released by females, and both sexes apparently use pectines to find prey and navigate to home retreats. Electrophysiological recordings from pegs of Paruroctonus utahensis reveal three spontaneously active cells (A1, A2, and B), which appear to interact synaptically. We made long-term extracellular recordings from the bases of peg sensilla and used a combination of conditional cross-interval and conditional interspike-interval analyses to assess the temporal dynamics of the A and B spike trains. Like previous studies, we found that A cells are inhibited by B cells for tens of milliseconds. However, after normalizing our records, we also found clear evidence that the A cells excite the B cells. This simple local circuit appears to maintain the A cells in a dynamic firing range and may have important implications for tracking pheromonal trails and sensing substrate chemistry for navigation.

Spatiotemporal co-dependency between macrophages and exhausted CD8+ T cells in cancer

T cell exhaustion is a major impediment to anti-tumor immunity. However, it remains elusive how other immune cells in the tumor microenvironment (TME) contribute to this dysfunctional state. Here we show that the biology of tumor-associated macrophages (TAM) and exhausted T cells (Tex) in the TME is extensively linked. We demonstrate that in vivo depletion of TAM reduces exhaustion programs in tumor-infiltrating CD8+ T cells and reinvigorates their effector potential. Reciprocally, transcriptional and epigenetic profiling reveals that Tex express factors that actively recruit monocytes to the TME and shape their differentiation. Using lattice light sheet microscopy, we show that TAM and CD8+ T cells engage in unique long-lasting antigen-specific synaptic interactions that fail to activate T cells, but prime them for exhaustion, which is then accelerated in hypoxic conditions. Spatially resolved sequencing supports a spatiotemporal self-enforcing positive feedback circuit that is aligned to protect rather than destroy a tumor.

GABA and IA Independently Regulate rNST Responses to Afferent Input

Taste responses in the rostral nucleus of the solitary tract (rNST) influence motivated ingestive behavior via ascending pathways, and consummatory reflex behavior via local, brainstem connections. Modifications to the afferent signal within the rNST include changes in gain (the overall rate of neuron activity) and changes in gustatory tuning (the degree to which individual neurons respond to divergent gustatory qualities). These alterations of the sensory signal derive from both synaptic interactions within the nucleus and the constitutive cellular membrane properties of rNST neurons. GABA neurons are well represented within the rNST, as is expression of KV4.3, a channel for a rapidly inactivating outward K+ current (IA). GABAergic synapses suppress rNST responses to afferent input and previous studies showed that this suppression is greater in cells expressing IA, suggesting a possible interaction. Here, we examine the potential interaction between GABAergic inhibition and IA channels in a series of patch clamp experiments. Optogenetic release of GABA suppressed rNST responses to afferent (electrical) stimulation and this effect was greater in cells with IA, confirming an earlier report. We further observed that the composite inhibitory postsynaptic potential was larger in IA positive cells, suggesting one mechanism for the greater afferent suppression. Blocking IA with the channel blocker AmmTX3, enhanced the response to afferent stimulation, suggesting a suppressive role for this channel in regulating afferent input at rest. However, pharmacologic blockade of IA did not suppress GABAergic inhibition, indicating that IA and GABA independently regulate excitatory afferent input.

BEAN: Interpretable and Efficient Learning With Biologically-Enhanced Artificial Neuronal Assembly Regularization

Deep neural networks (DNNs) are known for extracting useful information from large amounts of data. However, the representations learned in DNNs are typically hard to interpret, especially in dense layers. One crucial issue of the classical DNN model such as multilayer perceptron (MLP) is that neurons in the same layer of DNNs are conditionally independent of each other, which makes co-training and emergence of higher modularity difficult. In contrast to DNNs, biological neurons in mammalian brains display substantial dependency patterns. Specifically, biological neural networks encode representations by so-called neuronal assemblies: groups of neurons interconnected by strong synaptic interactions and sharing joint semantic content. The resulting population coding is essential for human cognitive and mnemonic processes. Here, we propose a novel Biologically Enhanced Artificial Neuronal assembly (BEAN) regularization1 to model neuronal correlations and dependencies, inspired by cell assembly theory from neuroscience. Experimental results show that BEAN enables the formation of interpretable neuronal functional clusters and consequently promotes a sparse, memory/computation-efficient network without loss of model performance. Moreover, our few-shot learning experiments demonstrate that BEAN could also enhance the generalizability of the model when training samples are extremely limited.

Spontaneous neural synchrony links intrinsic spinal sensory and motor networks during unconsciousness

Non-random functional connectivity during unconsciousness is a defining feature of supraspinal networks. However, its generalizability to intrinsic spinal networks remains incompletely understood. Previously, Barry et al. (2014) used fMRI to reveal bilateral resting state functional connectivity within sensory-dominant and, separately, motor-dominant regions of the spinal cord. Here, we record spike trains from large populations of spinal interneurons in vivo in rats and demonstrate that spontaneous functional connectivity also links sensory- and motor-dominant regions during unconsciousness. The spatiotemporal patterns of connectivity could not be explained by latent afferent activity or by populations of interconnected neurons spiking randomly. We also document connection latencies compatible with mono- and di-synaptic interactions and putative excitatory and inhibitory connections. The observed activity is consistent with the hypothesis that salient, experience-dependent patterns of neural transmission introduced during behavior or by injury/disease are reactivated during unconsciousness. Such a spinal replay mechanism could shape circuit-level connectivity and ultimately behavior.

Somatostatin-expressing parafacial neurons are CO2/H+ sensitive and regulate baseline breathing

Glutamatergic neurons in the retrotrapezoid nucleus (RTN) function as respiratory chemoreceptors by regulating breathing in response to tissue CO2/H+. The RTN and greater parafacial region may also function as a chemosensing network composed of CO2/H+-sensitive excitatory and inhibitory synaptic interactions. In the context of disease, we showed that loss of inhibitory neural activity in a mouse model of Dravet syndrome disinhibited RTN chemoreceptors and destabilized breathing (Kuo et. al., 2019; 25). Despite this, contributions of parafacial inhibitory neurons to control of breathing are unknown, and synaptic properties of RTN neurons have not been characterized. Here, we show the parafacial region contains a limited diversity of inhibitory neurons including somatostatin (Sst)-, parvalbumin (Pvalb)- and cholecystokinin (Cck)-expressing neurons. Of these, Sst-expressing interneurons appear uniquely inhibited by CO2/H+. We also show RTN chemoreceptors receive inhibitory input that is withdrawn in a CO2/H+-dependent manner, and chemogenetic suppression of Sst+ parafacial neurons, but not Pvalb+ or Cck+ neurons, increases baseline breathing. These results suggest Sst-expressing parafacial neurons contribute to RTN chemoreception and respiratory activity.