A multi-muscular, redundant strategy for free-flight roll stability

Reviewer #1 (Public review):

Summary:

The authors developed a free flight perturbation assay that induces roll instability. They set out to understand how steering muscles control and stabilize roll instability. In addition to the previously reported changes in wing amplitude, they find that a pure roll perturbation induces changes in stroke deviation. They silence motor neurons of individual steering muscles to show that silencing any one muscle alone is not sufficient to disrupt the recovery from roll perturbations. Through aerodynamic modelling, they argue that this occurs despite the fact that each muscle alone can induce changes in wing amplitude and/or stroke deviation. They suggest that this robustness to roll perturbation may be due to redundancy in the motor control program. Finally, they confirm redundant projections from the published haltere connectome study and show that indeed muscles which produce similar effects on wing kinematics receive redundant connections from the haltere afferents.

Strengths:

This study's strength lies in integrating findings from several adjacent areas in insect flight control research. The authors combine steering muscles physiology, wing kinematic quantification, aerodynamic modelling, and sensory (haltere) control of wing kinematics. This integrated approach provides a comprehensive discussion of the mechanisms that may underlie recovery from roll perturbations.

Weaknesses:

The biggest weakness is that, although the authors generate a very plausible and interesting hypothesis, much of the supporting evidence already occurs in existing datasets. In fact, a distributed or redundant many-to-one muscle control for flight kinematics is not a new idea. It has been suggested wherever researchers have examined muscle control for wing kinematics, across studies in flies, moths and other insects (Heide and Gotz, 1996; Balint and Dickinson, 2001; Lindsay et al, 2017; Melis et al, 2024, etc; Wood et al., 2024, etc). Therefore, it is not particularly surprising that Drosophila employs a multi-muscle strategy for roll stabilization. Although it is interesting to see that inhibition of even the phasic muscles (b3/ I2) alone did not disrupt the recovery, further experiments are required to address how these muscles contribute to wing kinematics responsible for roll control. Unfortunately, although the new evidence provided in this manuscript strengthens the idea of redundant muscle control, it does not test it directly.

Reviewer #2 (Public review):

Summary:

This manuscript investigates the kinematics, aerodynamics, and neural control of free-flight roll perturbation in fruit flies.

Strengths:

The paper employs a variety of appropriate methods, including magnetically sourced in-flight perturbations, free-flight wing kinematic measurement, and optogenetic silencing of specific motor units (and thus steering muscles). The results are generally consistent with prior work, showing that 5 different bilateral pairs of steering muscles contribute to the roll response, affecting the wing stroke amplitude and wing pitch. Furthermore, the roll response - both the overall animal performance and the details of the wing motion - is not detectably altered by knocking out any one of the five muscle pairs.

The reverse approach, optogenetic activation of specific phasic muscle pairs or silencing of tonic pairs, confirms that the selected muscles produce changes to wing kinematics appropriate for a roll response (confirmed by quasi-steady aerodynamic modeling). This set of results corroborates the main conclusions.

Weaknesses:

The authors refer to this as robust control of roll, though exactly what is meant by this is not clearly defined, and the word "detectably" may be important to understanding the limitations of the findings, since the large amount of variability in many of the experimental measurements would make it challenging to detect differences among treatments. As with the silencing experiments, the wing kinematics after optogenetic activation were highly varied, making it challenging to identify differences between the effects of individual muscles.

Reviewer #3 (Public review):

Summary:

Ludlow et al. investigate the control strategy that flies use to stabilize flight during small roll perturbations. Using 3D kinematic analysis of freely flying Drosophila, Ludlow et al. ask how manipulating steering motor neurons alters the fast stabilization reflex, which spans only a few wingbeats. Bilaterally activating i1 or i2 wing steering motor neurons during free flight pitches the fly down via decreases in the wing stroke amplitude, whereas inhibiting the b3 wing steering motor neurons pitches the fly up via increases in the wing stroke amplitude. These results suggest that asymmetric recruitment of these steering muscles may be used to rotate the fly around the roll axis. Then, using quasi-steady aerodynamic modeling, they linearly interpolate changes in six kinematic features of wing strokes to describe which wing parameters produce the greatest corrective roll torque. They repeated this analysis with data from optogenetic activation of wing steering motor neurons. They argue that modeling of these optogenetic perturbations supports the hypothesis that i1, i2, and b3 muscles contribute to rotation around the roll axis by calculating roll torque changes from changing kinematics of a single wing, despite bilateral optogenetic activation. Finally, they support their claims about the redundancy of steering motor neurons by presenting connectomic analyses of the direct pathways from haltere sensory neurons to wing steering motor neurons. This analysis reveals two independent pathways from halteres to two wing muscle groups (b1 and b2 vs. i1, i2, and b3). Taken together, these data provide some evidence for redundant roll stabilization control strategies in Drosophila.

Strengths:

The central strength of this work is the high-quality free-flight kinematic dataset, which provides a detailed picture of how wing-stroke parameters change during natural roll perturbations and correction. The integration of quasi-steady aerodynamic modeling with these kinematic data offers a principled framework for linking muscle activity to torque generation. The use of split-Gal4 lines to target individual wing steering motor neurons with cell-type specificity provides a potentially precise approach to probe the contribution of specific muscles to roll torque. Together, these tools position this study to make a meaningful contribution to understanding the sensorimotor control strategies underlying flight stabilization in Drosophila.

Weaknesses:

The GtACR1 silencing experiments lack validation that the optogenetic manipulation actually suppresses motor neuron activity. Without a positive control to calibrate light intensity and duration, the absence of kinematic effects cannot be interpreted with confidence. The confocal images of the driver lines provided are insufficient to uniquely identify the targeted motor neurons and do not clearly show expression in the brain and nerve cord. The kinematic modeling relies on flight profiles derived from a single fly and a single trial, raising questions about whether they capture the full range of natural variation. Finally, the connectomics analysis largely recapitulates prior work and provides little new insight.

There is no evidence that optogenetic silencing of wing motor neurons with GtACR1 is actually suppressing their activity. There are many factors that could impact the efficacy of this manipulation: transgene expression, light intensity and duration, etc. While electrophysiology experiments would be ideal, this would be challenging. Another option would be to use a positive control with an obvious phenotype to calibrate the light intensity and duration. For example, silencing all motor neurons (e.g., using OK371-Gal4 or another driver labeling glutamatergic neurons) should essentially paralyze the fly. Without additional evidence, the lack of a kinematic effect in the GtACR experiments is not convincing.

The confocal images in Figure 1C are of poor quality and do not help identify the motor neurons. They only point to cell body locations, and the morphologies of the neurons are not clear. The glial sheath also appears to be labeled. The images as they currently exist are not sufficient to uniquely identify the wing motor neurons and should be updated, including brains, since the experiments manipulate activity across the nervous system. It should also be clarified that these split Gal4 lines were not generated in this work.

The connectomics analysis does not add much on top of what was already known. It doesn't necessarily need to be removed, and the authors acknowledge that it is basically repeating prior analyses in prior publications from another connectome dataset. But it could be reduced to a schematic summarizing prior work. The one thing that should be clarified is what the "haltere afferents" actually are. Campaniform sensilla only or other sensory neurons (e.g., haltere chordotonal neurons)?

In Figures 3 and 4, 50 kinematic profiles are created using data from a single fly and a single trial. Are these kinematic profiles representative of the range that flies use? Would the results generalize across different bouts? Figure 1i-m shows a much narrower distribution of kinematic features compared to Figure 3, which might reduce the resulting torque calculations. Why not sample from the distribution of data in 1i-m in Figure 3? Could there be some sort of bootstrapping for the modeling in Figures 3-4?