Publications & Preprints

Natural behaviors are a coordinated symphony of motor acts that drive reafferent (self-induced) sensory activation. Individual sensors cannot disambiguate exafferent (externally induced) from reafferent sources. Nevertheless, animals readily differentiate between these sources of sensory signals to carry out adaptive behaviors through corollary discharge circuits (CDCs), which provide predictive motor signals from motor pathways to sensory processing and other motor pathways. Yet, how CDCs comprehensively integrate into the nervous system remains unexplored. Here, we use connectomics, neuroanatomical, physiological, and behavioral approaches to resolve the network architecture of two pairs of ascending histaminergic neurons (AHNs) in Drosophila, which function as a predictive CDC in other insects. Both AHN pairs receive input primarily from a partially overlapping population of descending neurons, especially from DNg02, which controls wing motor output. Using Ca2+ imaging and behavioral recordings, we show that AHN activation is correlated to flight behavior and precedes wing motion. Optogenetic activation of DNg02 is sufficient to activate AHNs, indicating that AHNs are activated by descending commands in advance of behavior and not as a consequence of sensory input. Downstream, each AHN pair targets predominantly non-overlapping networks, including those that process visual, auditory, and mechanosensory information, as well as networks controlling wing, haltere, and leg sensorimotor control. These results support the conclusion that the AHNs provide a predictive motor signal about wing motor state to mostly non-overlapping sensory and motor networks. Future work will determine how AHN signaling is driven by other descending neurons and interpreted by AHN downstream targets to maintain adaptive sensorimotor performance.

Dopaminergic projections regulate various brain functions and are implicated in many neuropsychiatric disorders. There are two anatomically and functionally distinct dopaminergic projections connecting the midbrain to striatum: nigrostriatal, which controls movement, and mesolimbic, which regulates motivation. However, how these discrete dopaminergic synaptic connections are established is unknown. Through an unbiased search, we identify that two groups of antagonistic TGF-β family members, bone morphogenetic protein (BMP)6/BMP2 and transforming growth factor (TGF)-β2, regulate dopaminergic synapse development of nigrostriatal and mesolimbic neurons, respectively. Projection-preferential expression of their receptors contributes to specific synapse development. Downstream, Smad1 and Smad2 are specifically activated and required for dopaminergic synapse development and function in nigrostriatal vs. mesolimbic projections. Remarkably, Smad1 mutant mice show motor defects, whereas Smad2 mutant mice show lack of motivation. These results uncover the molecular logic underlying the proper establishment of functionally segregated dopaminergic synapses and may provide strategies to treat relevant, projection-specific disease symptoms by targeting specific BMPs/TGF-β and/or Smads.

Across mammalian skin, structurally complex and diverse mechanosensory end organs respond to mechanical stimuli and enable our perception of dynamic, light touch. How forces act on morphologically dissimilar mechanosensory end organs of the skin to gate the requisite mechanotransduction channel Piezo2 and excite mechanosensory neurons is not understood. Here, we report high-resolution reconstructions of the hair follicle lanceolate complex, Meissner corpuscle, and Pacinian corpuscle and the subcellular distribution of Piezo2 within them. Across all three end organs, Piezo2 is restricted to the sensory axon membrane, including axon protrusions that extend from the axon body. These protrusions, which are numerous and elaborate extensively within the end organs, tether the axon to resident non-neuronal cells via adherens junctions. These findings support a unified model for dynamic touch in which mechanical stimuli stretch hundreds to thousands of axon protrusions across an end organ, opening proximal, axonal Piezo2 channels and exciting the neuron.

Our ability to sense and move our bodies relies on proprioceptors, sensory neurons that detect mechanical forces within the body. Different subtypes of proprioceptors detect different kinematic features, such as joint position, movement, and vibration, but the mechanisms that underlie proprioceptor feature selectivity remain poorly understood. Using single-nucleus RNA sequencing (RNA-seq), we found that proprioceptor subtypes in the Drosophila leg lack differential expression of mechanosensitive ion channels. However, anatomical reconstruction of the proprioceptors and connected tendons revealed major biomechanical differences between subtypes. We built a model of the proprioceptors and tendons that identified a biomechanical mechanism for joint angle selectivity and predicted the existence of a topographic map of joint angle, which we confirmed using calcium imaging. Our findings suggest that biomechanical specialization is a key determinant of proprioceptor feature selectivity in Drosophila. More broadly, the discovery of proprioceptive maps reveals common organizational principles between proprioception and other topographically organized sensory systems.

We present an auxiliary learning task for the problem of neuron segmentation in electron microscopy volumes. The auxiliary task consists of the prediction of local shape descriptors (LSDs), which we combine with conventional voxel-wise direct neighbor affinities for neuron boundary detection. The shape descriptors capture local statistics about the neuron to be segmented, such as diameter, elongation, and direction. On a study comparing several existing methods across various specimen, imaging techniques, and resolutions, auxiliary learning of LSDs consistently increases segmentation accuracy of affinity-based methods over a range of metrics. Furthermore, the addition of LSDs promotes affinity-based segmentation methods to be on par with the current state of the art for neuron segmentation (flood-filling networks), while being two orders of magnitudes more efficient-a critical requirement for the processing of future petabyte-sized datasets.