activeFly aimed to identify neurons supplying motor information to cells processing visual feedback and elucidate the role of these visuomotor interactions. We first mapped the site of motor signals to HS and H2 cells, two classes of optic flow sensitive cells, specifically to their axon terminals. Using an electron microscopy (EM) dataset and computational analysis, we identified approximately 100 neurons receiving direct convergent information from premotor centers, including the VNC (the insect analogue of the spinal cord). Notably, we identified multiple pathways capable of providing non-visual information to HS and H2 cells, rendering a silencing screen for identifying dedicated cells obsolete.
To explore the function of the identified network, we conducted four distinct studies. Initially, we examined the role of visual feedback in walking control. This study demonstrated that flies swiftly stabilize gaze by reducing the gain of posture control reflexes. The notion that visual feedback plays a critical role in the interplay between mechanical stability and behavioral goals established a framework for exploring the underlying visuomotor circuits governing gaze stabilization. We hypothesized that HS or H2 cells would contribute to gaze stability by regulating leg placements on a moment-to-moment basis to counter postural reflexes.
Using whole-cell patch recordings, a second study described that HS cells received ascending information from the VNC. This modulation operates on a moment-by-moment bases, facilitating the cells’ rapid control of the ipsilateral foreleg. In addition, HS cells either gradually depolarized or hyperpolarized as a function of several high- vs. low-speed steps, respectively. This study unveiled the source of speed modulation in HS cells and its function: to facilitate rapid and context-dependent recruitment of optic flow-sensitive neurons for steering control.
In a third study, using whole-cell patch recordings and 2-photon optical imaging from different elements of the network confirmed their sensitivity to body rotations, thus validating predictions derived from the connectivity analysis. We used visual stimuli to precisely modulate activity at the network's input (the optic flow-sensitive cells) and assess signal propagation throughout the network. Our findings showed that the network architecture fosters competitive disinhibition between the right and left sides, enabling robust extraction of rotational information during translation.
A four study examined the interplay between visual and motor signals within the network by manipulating visual information. This study showed that optic flow-sensitive cells optimally integrate multimodal information, enhancing the weighting for non-visual signals as visual uncertainty escalates. Altogether, these findings strongly support the idea that non-visual information robustly encode self-rotations. Moreover, our work illustrates that inhibitory neurons within the network regulate optimal multimodal integration via normalization.
In summary, activeFly has shown that early visuomotor interactions optimally and robustly encode self-rotations. This information is flexible used based on the fly’s forward speed to control steering. These findings have been published in:
1. Fujiwara, et al, and Chiappe (2017) Nat Neurosci. PMID: 27798632
2. Erginkaya, et al., and Chiappe (2023) BiorXiv 552150
3. Cruz, et al., and Chiappe (2021) Curr Biol PMID: 34499851
4. Fujiwara, et al., and Chiappe (2022) Neuron PMID: 35525243
5. Cruz, Marques and Chiappe, in prep
6. Cruz and Chiappe (2023) Curr Opin in Neurobiol PMID: 37651855
7. Chiappe (2023) Curr Opin in Neurobiol PMID: 37453230
