Consistent coordination patterns provide near perfect behavior decoding in a comprehensive motor program for insect flight

Precise spike timing encoding is present in many motor systems, but is not frequently utilized to decode behavior or to examine how coordination is achieved across many motor units. Furthermore, testing whether the same coordinated sets of muscles control different movements is difficult without a complete motor representation at the level of the currency of control – action potentials. Here, we demonstrate nearly perfect decoding of six hawk moth flight behavior states elicited in response to wide-field drifting visual stimuli about the flight axes – pitch, roll, and yaw – using a comprehensive, spike-resolved motor program. A simple linear decoding pipeline is sufficient to predict behavior, but only if precise timing information in included. We also demonstrate that as few as half the muscles can be used to retain this near perfect decoding performance, linking coordination to redundancy in encoding across the entire moth flight motor program. We then use this comprehensive motor representation to test if muscle covariation present in one pair of visual stimulus conditions can be used to decode behavior in a different pair of visual stimulus conditions. We find conserved muscle coordination patterns at the level of motor unit spike timings in these functionally distinct behaviors.

Methods
EMGs were taken from the 10 most important muscles for wing control from Manduca sexta while tethered moths flapped in a visual arena. The responded to a visual stimulus while simultaneous recordings of the spike resolved motor program and 6 DoF forces and torques were acquired. The comprehensive motor program of the moth, Manduca sexta, is composed of the 10 muscles (5 bilateral pairs) that accomplish nearly all control of the wings. On each side of the animal there is a primary flight power muscle for the downstroke (Dorsolongitudinal muscle –DLM) and one for the upstroke (Dorsoventral muscle –DVM). There are also three small muscles that produce an order of magnitude less force, but tweak, tension, and pitch the wingstroke to achieve control (“steering” muscles –basalar, subalar, 3rdaxillary). Each of these muscles is composed of one or few motor unit(s) and so we can read off the individual spikes from each recording. This gives us a precisely timed, spike resolved motor program that is nearly complete –any sensory information thataffects wing movements should manifest as changes in the number and timing of these muscle spikes In each case the animal is flapping its wings at roughly 20-22 Hz but there is some variation. Using a Hilbert transform of the z-force recordings, we have cut the data up into comparable wingstrokes which can be used as a unit of analysis. Data has been collected from ~10 moths, with multiple flight bouts in each case. A flight bout typically lasts about 20 secs. We have checked that the animal is flapping ata reasonable frequency and trimmed the bouts if the animal stops or starts flapping before the recording starts or stops. Wingstrokes from all bouts are currently considered equivalently. There are significant differences between individual animals, but each animal had typically 300-1000 wingstrokes recorded. In all cases the neuromuscular recordings have been spike-sorted to identify the spikes. We do not recommend using the continuous voltage traces from the electrophysiological recordings because they can contain movement artifacts which could contain motor information. The identified spikes have been organized into spike trains (for contiguous data) and spike times (for wingstroke cut data).

“Hard Turns” This dataset was collected with projected wide-field motion stimuli in each of the three rotational axes (yaw, pitch, and roll). We rotated the wide-field motion stimulus at a constant rotational velocity in one of the six directions: 3 axes, two signs in each case to produce hard left, hard right, hard pitch up, hard pitch down, hard roll left, hard roll right turns). Each individual has 100s of wingstrokes sampled for each condition which we can consider as repeated observations of the same behavioral condition. There is some variation from flight bout to flight bout that could be considered and there is individual-to-individual variation as in other datasets. This data has been cut up into individual wingstrokes and includes forces and torques sampled for 50 ms for each wingstroke (again 500 samples). This dataset includes 100s of wingstrokes form each of the six conditions with complete (10 muscle) recordings from 4 animals (complete data) and comparable amounts of data from 5 other moths that have only 9 muscle recordings in each case (incomplete data).