Vectorial efference copy and visuomotor transformation through gap junctions

Background

Movement isn’t solely about getting from point A to B. It affects the sensory input we receive in a continuous, moment-by-moment manner. Consider the movements of your eyes as you read this text: you flick through the words across the screen in a series of rapid ‘saccades’, and yet you still perceive the screen as a stable, coherent image. Rather than viewing the world as a series of incredibly jerky (and nauseating!) images, your brain recognizes and ignores these ‘self-generated’ signals. How?

Signals generated from self-motion can cancel sensory input produced during movement, termed as an ‘efference copy’. They occur across diverse organisms and sensory modalities, including mammalian eye movements, but identifying where they originate, and how the underlying neural circuits implement them, has proven a challenging feat.

To address this question, Falt and colleagues leveraged recent findings on retinal movements and whole-brain connectomes in the genetic fruit fly model, Drosophila melanogaster. These fruit flies rely strongly on the visual properties of the world, exhibiting spontaneous retinal movements, and the ‘optokinetic reflex’ (OKR). Here, external motion on the retina causes the eyes to follow a moving scene before rapidly resetting gaze with a saccade (see Figure 1 below). Because these saccades also shift the retinal image, visual input must be suppressed to prevent the reflex from responding to self-generated motion, which would cause a reflexive feedback loop.

 

Experimental Design

Using head-fixed flies, the authors combined high-speed retinal tracking, whole-cell patch-clamp recordings, anatomical staining and genetic manipulations. Experiments were performed in darkness, as well as while flies viewed moving black-and-white gratings, during which the ‘deep pseudopupil’ was tracked to quantify retinal gaze orientation.

They recorded from neurons involved in vision, primarily the ‘horizontal system’ (HS) and ‘vertical system’ (VS) cells, alongside the two retinal motor neurons (RMNs), each innervating a single retinal muscle (musculus orbito-scapalis, MOS; musculus orbito-tentoralis, MOT) (muscles highlighted in Figure 1). Genetic mutants and optogenetic manipulations then provided the causal evidence for their involvement in the visuomotor pathway.

From these data, the authors established two linear models: 1) an oculomotor model predicting retinal shift based on RMN activity. This was later combined with the RMNs’ visual tuning to predict the optokinetic response with great precision, and 2) an electrical wiring model which predicts the electrical information transmission between retinal motor neurons and the upstream visual cells via gap junctions.

Highlighted Results

There are various exciting results that emerged:

• Each retinal muscle is innervated by a single motor neuron, and gaze shift direction can be well predicted by the sum of the activity of the two neurons innervating the muscles.
• Rather than being primarily chemically connected, the authors found that the visual and retinal motor neurons are electrically coupled through gap junctions, providing an incredibly rapid and direct transmission of information between cells.
• A subset of visual neurons (VS1-4 and HS, specifically) are distinctly coupled to each type of retinal motor neuron. Their activity can be linearly summated to predict RMN activity, closely matching the experimental data.
• Conversely, hyperpolarizing the RMNs produced corresponding hyperpolarization in HS cells, demonstrating direct transmission from motor neurons back to upstream sensory neurons.
• Activity levels in HS are modulated by the amplitude and direction of retinal shifts. This precedes retinal movement, indicating that the neurons are being ‘predictively’ suppressed before a gaze shift (inhibiting the optokinetic reflex). This convincingly demonstrates a visual efference copy being transmitted along the motor –> sensory pathway (in ED Figure 10: information transmitted from point 6 back to point 5).

 

 

Two features appear critical to this rapid response. First, retinal muscles rest in a tonic state, meaning they have a constant, slight contraction at baseline, and via de-/hyperpolarization, can rapidly contract/relax further to adjust the retinal position. Second, direct communication between motor and visual neurons allows motor neurons to rapidly respond to visual input. This also provides what the authors term an “off-switch” for the gaze stability reflex during gaze shifting. Beyond describing this mechanism in remarkable detail, this work provides a base for asking how other animals may uniquely (or not!) implement similar computations.

Most excitingly, the authors seem to provide the final piece of what was a ‘nearly complete visuomotor transformation pathway’: with single-cell resolution, it is possible to map the pathway from light reaching the photoreceptor, to the corresponding behavioural output (albeit with some additional contributing components possibly unknown). ED Figure 10 visualizes this pathway, starting at point 1 with the photoreceptor receiving photons, and ending at point 7, where the retina is actively shifted in response.

What I like about this preprint:

Harnessing the breadth of experimental tools available for Drosophila, the authors elegantly dissect the process of efference copy generation, demonstrating an admirable amount of work to do so. Each experimental step has a clear motivation and rationale, and I particularly liked that the authors started from rather simple predictions and models, before elaborating with additional components (e.g. starting from the oculomotor model derived from individual RMN responses, and then adding visual tuning to make predictions). I also appreciated their use of genetic mutants and optogenetics – rather than simply matching physiology to model predictions, they provide compelling causal evidence for the mechanisms driving the results.

Finally, I think it’s thrilling to see open-source databases of the fruit fly brain (e.g. the connectome) used as a reference for new findings. This preprinted work highlights that even comprehensive, seemingly complete descriptions of the brain can still lack critical detail – here, gap junctions coupling visuo-motor neurons – and that there are still endless exciting discoveries to be made about how animals sense and act in a dynamic world.

Questions for the Authors:

1. From a simplicity perspective, I wasn’t expecting that HS cells receive efference copies from two retinal motor neurons, whereas VS cells receive a copy from only one. Is there a functional or computational advantage to this circuitry? Could similar results be expected if the pathways were simply HS –> MOT and VS –> MOS? Could your models be perturbed or rearranged to test how different circuitry or connections influence both the resulting efference copy and retinal movement?
2. You mention that efference copies from the retina are likely to influence neck motor neurons, and vice versa. Could the neck and retina ever attempt to stabilize in opposite directions, potentially producing conflicting efference copies, or do gap junctions inherently prevent this?
3. How might this translate to a freely-moving fly, whether on a substrate or in the air? Would you expect the same computations and circuitry under much faster and more dynamic sensory input (and more complex motor decisions)?
4. You describe a key missing link in a ‘nearly complete sensory-motor transformation’. With this level of resolution now available, and confident predictions about the computations each component performs, what’s next? What might this mean for future Drosophila research on efference copies and active sensing? How might it extend to other animals?

Tags: drosophila, efference copy, fruit fly, vision

Read preprint 1 Star (2 votes)