Figures and data in Disentangling cephalopod chromatophores motor units with computer vision

Figures
Tables
Additional files

8 figures, 1 table and 1 additional file

Figures

Figure 1

Download asset Open asset

Identification and characterization of putative chromatophore motor units in *E. berryi*.

(a) Example HD-video frame extracted from a *E. berryi* recording. (b) Visual representation of presumed motor units. Colors represent motion clusters of chromatophore slices grouped together by covariation using the HDBSCAN algorithm. Segmented chromatophores shown in white, with epicenters marked as magenta dots. Colored clusters indicate presumed motor units based on slice motion correlation. Outlier slices are omitted for clarity. (c) Number of chromatophores per cluster size, and frequency of chromatophores belonging to multiple clusters. (d) Zoom on the frame shown in (a), with detected slices overlaid in magenta. (e) Zoom on two overlapping motor units from (b). The central chromatophore (red circle) contributes slices to both the blue and the purple clusters. Cluster centers of mass, calculated from the epicenters of the chromatophores belonging to each cluster, are shown as colored crosses. This illustrates two ‘virtual chromatophores’, where coordinated activity spans slices from several chromatophores. (f) Surface areas of slices belonging to each cluster over time. In bold, the surface areas of the slices highlighted in white in (e). (g) Distribution of slice orientations relative to the motor unit’s center of mass.

Figure 2 with 1 supplement

Download asset Open asset

Multiple innervation of individual chromatophores.

(a) Activity traces (surface area as functions of time) of 36 polar slices from a single chromatophore, offset vertically pairwise by 5 pixels for clarity. Each trace represents the positive, detrended change in surface area over time, highlighting expansion dynamics while excluding negative (contraction) phases (see ‘Materials and methods’), assumed to be passive. (b) Cumulative explained variance plot for principal component analysis. Relationship between number of principal components and explained variance, with PCs ranked by decreasing contribution. Red dashed line corresponding to greatest slope change (here PC5) (see ‘Materials and methods’). (c) Histogram showing the number of principal components identified per chromatophore across the dataset. (d) Independent sources acting on the 36 slices of a single chromatophore, extracted with independent component analysis. In this case, the chromatophore is controlled by five independent sources that likely correspond to five distinct motor neurons. (e) Polar plot representing the influence of each source (motor neuron activity) on each slice. Black line indicates the first radial slice; influence of a source on a slice expressed in % (radial scale). (f) Examples of PCA-ICA analyses run on six other chromatophores.

Figure 2—figure supplement 1

Download asset Open asset

*Sepia officinalis* (mantle length 20 cm; photograph by Stephan Junek, Max Planck Institute for Brain Research).

The inset shows the analyzed region of interest at the center of the mantle (0.5×0.5 cm).

Figure 3

Download asset Open asset

Identification of putative motor units.

(a) Schematic representation of derived chromatophore innervation by motor neurons. Each chromatophore is depicted as a circle, with arcs around them representing independent components identified by ICA, each interpreted as corresponding to the zone of influence of a single motor neuron on that particular chromatophore. One goal of this clustering was to reveal shared chromatophore innervation and, thus, motor units. Colors and lines connecting IC arcs indicate that they belong to the same motor unit, highlighting the shared innervation pattern across multiple chromatophores. (b) A small number of clusters of size 3–5 are mapped back on the original image for visualization. (c) Distribution of cluster size (size in number of IC). For visualization, the graph was truncated to show over 90% of the data. (d) Distribution of cluster convex hull areas, showing a right-skewed profile with most clusters occupying smaller areas. All values are reported in pixel² (px²); note that the main text reports corresponding measurements in µm² after conversion (10 pixels = 1  µm, thus 1 px²=0.01 µm²). (e) Relationship between chromatophore count and convex hull area for each cluster. (f) Chromatophore density (chromatophores per pixel²) as a function of convex hull area, plotted on a log-log scale. The inverse relationship reflects decreasing density in larger clusters. (g) Histogram showing the distribution of second-order co-occurrence counts—defined as the number of chromatophore pairs co-occurring in at least two clusters—across 1000 randomized datasets (blue). Randomizations preserved both the number of clusters and the number of memberships per chromatophore. The red dashed line indicates the observed experimental value, significantly higher than expected by chance.

Figure 4

Download asset Open asset

Spatial organization of putative motor units.

Chromatophores belonging to the same correlated-motion cluster, or putative MU, are shown as colored dots connected by lines. (**a–d**) Compact clusters. (**e, f**) Overlapping clusters with shared chromatophores. (g) Elongated cluster with a linear structure. (h) Mixed profile featuring a core group of adjacent chromatophores connected to a single, distant chromatophore.

Figure 5 with 1 supplement

Download asset Open asset

Composite summary of chromatophore slices expansion–contraction dynamics.

Events are first filtered to remove very small peaks (expansion amplitude ≤ 15% of the observed amplitude range). Remaining events are split by expansion amplitude percentiles: SMALL = lower tail, MEDIUM = central band, LARGE = upper tail. Within each tail band, outliers in either amplitude or speed are removed using Tukey’s IQR rule, and statistics are computed on the remaining events (see ‘Materials and methods’). (a) Peak-aligned event segments (10 randomly selected events per column) for each group. Traces are shifted to align at the peak (t=0). Same y calibration in the four panels. (b) Mean event profile for each size group, where each segment is color-coded by instantaneous size change (|Δradius| for pairs of consecutive frames); LUT at right. Same y calibration in the four panels (c) Means ± SDs of chromatophore size change (px/frame) for expansion (green) and contraction (red) phases. Statistical significance evaluated using paired t-tests between expansion and contraction phases (n=2157 event pairs). Asterisks indicate significant paired differences *p<0.05, **p<0.01, ***p<0.001, and ****p<0.0001. (d) Same as in (c) but for event duration (in frames).

Figure 5—figure supplement 1

Download asset Open asset

Relative temporal composition of chromatophore events across amplitude groups.

For each amplitude category (SMALL, MEDIUM, LARGE), the proportion of frames spent in the expansion (green) versus contraction (red) phase was computed across all valid events (n=2157). Percentages were calculated relative to the total duration of each event (expansion + contraction). This analysis corresponds to the same dataset shown in Figure 5.

Figure 6 with 1 supplement

Download asset Open asset

Patch-clamp stimulation of chromatophore-lobe motor neuron.

(a) Patch-clamp recordings from a chromatophore-lobe motor neuron. Ten consecutive sweeps are shown, vertically offset for clarity and downsampled ×4 (from 100 to 25 kHz). (b) Sequence of video stills showing chromatophore expansion elicited by stimulation of the same motor neuron as in a. In orange, chromatophore epicenters, defined as the center of mass at the frame where the chromatophore is in its most contracted state. Each chromatophore expands anisotropically toward a common center of innervation (approximated as a red cross). (c) Composite images showing chromatophore outlines before (blue) and during stimulation (green). The first three examples display anisotropic expansion across several chromatophores within a single motor unit—the first example is the same motor unit shown in (a, b). The last example shows isotropic expansion of a motor unit composed of a single chromatophore.

Figure 6—video 1

Download asset

posterframe for video — Video example of patch-clamp stimulation.

On the left, a video showing the expansion of five chromatophores composing a motor unit—the same motor unit shown in Figure 6b. On the right, membrane-potential trace from the corresponding motor neuron, synchronized with the video on the left.

Figure 7 with 3 supplements

Download asset Open asset

Designs of experimental arenas.

(a) Arena for *Euprymna berryi* hatchlings. (b) Rig for *Sepia officinalis*. Four adjustable LED lamps in yellow. Transparent tank in cyan. Structure in gray.

Figure 7—figure supplement 1

Download asset Open asset

Sepia filming setup.

(a) Remotely controlled linear rails. (b) Z-axis manual translation stage. (c) Camera and lens. (d) LED lights.

Figure 7—figure supplement 2

Download asset Open asset

Visible implant elastomer tagging.

(a) *Sepia officinalis* marked at the center of the mantle with tags of various colors in a row. (b) Marked animal under fluorescent light. (c) Fluorescent lamp mounted next to the camera on the filming rig. (d) Footage captured by the high-resolution camera. Close-up on the blue fluorescent tag indicated by an arrow.

Figure 7—figure supplement 3

Download asset Open asset

Schematics of experimental setup with head fixation.

The camera (green) is positioned above the tank, which is filled with artificial seawater (blue). A constant flow of water (indicated by arrows) ensures proper oxygenation. LEDs (yellow) surrounding the tank provide uniform illumination. A bracket (black) supports a thick glass panel (gray) to minimize water distortion during filming. The same bracket also holds a head-fixation mount (white) to stabilize the animal’s head.

Figure 8

Download asset Open asset

Confocal image of *E. berryi*’s stained skin captured with ZEISS LSM 980.

DAPI in blue (nuclei), phalloidin in magenta (muscle fibers), and tubulin in green (nerve fibers). A single chromatophore is highlighted (white box) to illustrate its cellular organization. Radial muscle fibers (magenta) extend outward from the central pigment sacculus, with nuclei (blue) arranged in a rosette at the base of each fiber. Tubulin-labeled nerve fibers (green) approach the muscles, suggesting potential neuromuscular connections, though the precise contact points cannot be confirmed in the absence of a neuromuscular junction marker.

Tables

Key resources table

Reagent type (species) or resource	Designation	Source or reference	Identifiers	Additional information
Software, algorithm	CHROMAS (a1fe9de3)	Ukrow et al., 2025 Ukrow and Renard, 2025	—	Pipeline for chromatophore segmentation and analysis
Software, algorithm	Python (3.9+)	Python Software Foundation https://www.python.org/	RRID:SCR_008394
Software, algorithm	Fully Convolutional Network—ResNet-50	Long et al., 2015	—	Chromatophore segmentation
Software, algorithm	scikit-learn (1.6.1)	Pedregosa et al., 2011	RRID:SCR_002577	Clustering and dimensionality reduction
Software, algorithm	Matplotlib (3.10.0)	Hunter, 2007	RRID:SCR_008624	Visualization
Software, algorithm	HDBSCAN	Campello et al., 2013	—	Clustering
Software, algorithm	Numpy	Harris et al., 2020	RRID:SCR_008633	Numpy package
Software, algorithm	Pandas	McKinney, 2010	RRID:SCR_018214	Pandas package
Software, algorithm	kneed (KneeLocator)	Satopaa et al., 2011	—	Determined PCA elbow for selecting number of ICs
Other	Euprymna berryi (bobtail squid)	Max Planck Institute for Brain Research (Frankfurt); State Key Laboratory of Membrane Biology (Peking)	NCBITAXON:153281	(Organism) Lab-reared; used for computational analysis and electrophysiology
Other	Sepia officinalis (European cuttlefish)	Max Planck Institute	NCBITAXON:6610	(Organism) Lab-reared; used for computational analysis
Other	Aqua Medic Tri Complex	Aqua Medic https://www.aqua-medic.de/	—	Supply of macro- and trace elements
Other	Instant Ocean	Instant Ocean https://www.instantocean.com/	—	Artificial seawater
Other	Visible Implant Elastomer (VIE) Tags	Northwest Marine Technology, Inc https://www.nmt.us/visible-implant-elastomer/	—	Marking the same cuttlefish skin region across sessions
Other	Basler Ace 2 a2A2590-60ucPRO camera	Basler https://www.baslerweb.com/	—	E. berryi hatchling imaging
Other	Kowa LM25JC10M macro lens	Kowa https://www.kowa-lenses.com/	—	High-resolution imaging of E. berryi
Other	Custom-built LED ring	Max Planck Institute	—	Adjustable illumination (E. berryi)
Other	Custom sound- and light-proof enclosure	Max Planck Institute	—	Eliminating external vibrations and light (E. berryi)
Other	Basler CoaXPress 2.0 boA9344-70cc (65 MP) camera and accessories	Basler	—	High-resolution recordings in S. officinalis
Other	Qioptiq Apo-Rodagon-D 2×4/75 lens	Qioptiq	—	Macro magnification
Other	Custom-built E. berryi imaging arena	Max Planck Institute	—	Arena for filming
Other	Custom-built Sepia officinalis imaging arena	Max Planck Institute	—	Arena for filming
Other	BX51W1, Olympus, fluorescence microscope	Olympus http://olympusconfocal.com	—	Patch-clamp microscope
Other	UMPlanFI 20, Olympus, immersive lens	Olympus	—	Patch-clamp lens
Other	a2A4096-30ucBAS, Basler camera	Basler	—	Chromatophore recording during patch clamp
Other	MLM-3X-MP, Computar, lens	Computar https://www.computar.com/	—	Chromatophore recording during patch clamp
Other	Borosilicate glass, BF150-86-10, WPI	WPI https://www.wpi-europe.com/	—	To make glass electrodes
Other	P-1000, Sutter Instrument	Sutter Instrument https://www.sutter.com/micropipette/p-1000	RRID:SCR_021042	Pipette puller
Other	HEKA amplifier (ESC100-USB)	HEKA	—	E-phys recording