Axial muscle-fibre orientations in larval zebrafish

Abstract

Most teleost fish propel themselves with lateral body waves powered by their axial muscles. These muscles also power suction feeding through rapid expansion of the mouth cavity. They consist of muscle segments (myomeres), separated by connective tissue sheets (myosepts). In adult teleosts, the fast axial muscle fibres follow pseudo-helical trajectories, which are thought to distribute strain (relative fibre length change) approximately evenly across transverse sections during swimming, thereby optimizing power generation. To achieve strain equalization, a significant angle to the longitudinal axis on the frontal plane (azimuth) is necessary near the medial plane, increasing strain. Additionally, a deviation from longitudinal orientation on the sagittal plane (elevation) is required laterally to decrease strain. Despite several detailed morphological studies, our understanding of muscle-fibre orientations in the entire axial musculature of fish remains incomplete. Furthermore, most research has been done in post-larval stages, leaving a knowledge gap regarding the changing axial muscle architecture during larval development. Larval fish exhibit different body size, body shape and swimming kinematics compared to adults. They experience relatively high viscous forces, requiring higher tail-beat amplitudes to overcome increased drag. Additionally, larval fish swim with higher tail-beat frequencies. Histological studies have shown that in larval fish, muscle fibres in the anal region transition from an almost longitudinal orientation to a pseudo-helical pattern by 3 dpf (days post-fertilization). However, these studies were limited to a few sections of the body and were prone to shrinkage and tissue damage. Here, we introduce a novel methodology for quantifying muscle-fibre orientations along the entire axial muscles. We selected 4 dpf larval zebrafish for our analyses, a stage where larvae are actively swimming but not yet free-feeding. High-resolution confocal 3D scans were obtained from four genetically modified zebrafish expressing green fluorescent protein in fast muscle fibres. Fluorescence variation allowed segmentation of individual muscle fibres, which were then converted to fish-bound coordinates by correcting for the fish's position and orientation in the scan, and normalized to pool results across individuals. We show that at 4 dpf, muscle-fibre trajectories exhibit a helical pattern tapering towards the tail. Average fibre angles decrease from anterior to posterior, with azimuth varying over the dorsoventral axis and elevation varying over the mediolateral axis. Notably, only the anteriormost 20% of the body displayed higher azimuth angles near the medial plane. Angles between neighbouring fibres were substantial, particularly at the rim of the epaxial and hypaxial muscles. The revealed muscle-fibre architecture at this age presumably contributes to the swimming performance of these larvae, but that swimming performance is probably not the only driving factor for the fibre pattern. Our methodology offers a promising avenue for exploring muscle-fibre orientations across ontogenetic series and provides a foundation for in-depth functional studies on the role of muscle architecture in facilitating swimming performance of larval fish.

Keywords: Danio rerio; 3D muscle architecture; confocal laser scanning microscopy; muscle development; swimming muscles.

FIGURE 1

Diagram of a lateral view of an adult zebrafish, showing the expected general direction of a selection of the muscle‐fibre trajectories. Inspired by Alexander (1969).

FIGURE 2

Schematic dorsoventral view of a fish segment on the frontal plane (see left bottom). The thick central lines represent the non‐compressible longitudinal axis of the fish, the thick grey lines indicate muscle fibres. (a) Arrangement in rest. The angle α is the azimuth; the angle between the fibre's projection on the frontal plane and the longitudinal axis. (b) Arrangement for left‐side activation, which leads to shear deformation with shear angle γ. The shear deformation results from an anterior shift of muscle tissue relative to the longitudinal axis on the concave (left) side, and a posterior shift on the convex side. The dashed line signifies the original position of the schematic myosept and is used to measure the shear angle. Based on Van Leeuwen et al. (2008).

FIGURE 3

Image processing steps for fibre segmentation. For each step, the same lateral view (on the left) and transversal cross‐section (on the right) are displayed. (a) Original scan, after LAS X Lightning filtering. (b) The image intensities were re‐scaled to cut off the lowest 5% and highest 0.5% pixel intensities. Intensity normalization was followed by a (c) median filter (kernel size 5 px) and a (d) Gaussian filter (σ = 5 px) to reduce noise. (e) A Canny edge filter was used to find borders between neighbouring fibres (black lines) and subsequently a distance transform (grey values) quantified the distance to the nearest boundary. Local maxima in these transforms were used as markers (coloured circles) for the watershed segmentation. (f) The boundaries for the segmentation process were obtained by applying a Sobel edge detection. (g) Result for the watershed segmentation. Each colour defines a different segmented object, which includes individual fibres but also merged fibres and other objects such as the central chorda. (h) Individual fibres were then found by filtering the 3D objects based on size.

FIGURE 4

Transformation of world‐bound to fish‐bound coordinates. Due to slight differences in embedding, the fish can be slanted, bent and/or torsioned. To normalize the fibre coordinates across fish and discount differences in embedding, we expressed locations and orientations relative to the centre of the notochord. Hereto, we determined the location and orientation of the notochord along the length of the fish (see Section 2) and corrected the data based on the closest point on the notochord. The yellow and blue vector components represent the directions of the x and z‐axis, respectively.

FIGURE 5

Definitions of axes, planes and angles, with the x‐axis in yellow and the y and z‐axes in blue. (a) Schematic overview of axes origins and plane definitions. The coordinate system is defined relative to a straight notochord and may deviate from a central axis of the fish, as the notochord is naturally slightly curved, especially in the anteriormost part. The origin corresponds to the anteriormost point on the centre of the notochord where the axial musculature starts. (b) Axes definitions for fibre orientations. α: azimuth, β: elevation, θ: angle with longitudinal direction of straightened notochord.

FIGURE 6

Fibre coordinates in cross‐sections. Different colours correspond to different fish. For cross‐sections, the fish is divided into 5 equal length bins, as indicated in the insets. All axes are normalized to the length of the axial muscles, on average 0.01 $\approx$ 27.6 μm. Larger empty circles indicate the centre of the notochord at y,z = 0,0.

FIGURE 7

Lateral view of 3D scan. Data represent a color‐coded maximum projection of the optimized, raw 3D data of a 4 dpf larval fish, before muscle segmentation and before corrections for positioning and curvature/torsion of the fish. Colours correspond to raw z coordinate, in a gradient from yellow to dark‐blue from medial to lateral, as indicated by the legend. (a) Full overview of a 4 dpf larval fish. (b–d) Magnification of the fibres at square ‘b’, ‘c’, ‘d’ in image ‘a’. 1: Cranial levator, 2: Ventral muscle going around the yolk sac. (e) legend for the colour encoding of z values. The scale lines in each bottom right corner of the panels represent 0.1 mm.

FIGURE 8

The interfibre angle δ for fibres between 10 and 30 μm apart (roughly represents fibre diameter) calculated per fish. Colours correspond to different bins along the longitudinal axis, numbered 1, 2, 3, 4, 5 from anterior to posterior.

FIGURE 9

Fibre orientations for five cross‐sections represented as quiver plots. Arrows quantify the fibre orientations. The x‐axis is oriented perpendicular to the shown transverse sections, so the arrow length is zero for fibres oriented parallel to the longitudinal axis. Red arrows indicate the arrow length for an angle of 90° relative to the x‐axis. Empty circles indicate the centre of the notochord at y,z = 0,0 .

FIGURE 10

Fibre angles relative to the longitudinal axis (θ). (a) Fibre coordinates on the sagittal plane (cross‐sections) in 5 bins along the anteroposterior axis, colour coded for θ according to colour legend. Larger empty circles indicate the notochord centre. (b) Violin plot of theta in five bins along the anteroposterior axis, with the hypaxial muscles in light grey and the epaxial muscles in dark grey for each bin. These data represent the variation in angles due to the observed helical fibre arrangement, rather than the deviation from an expected mean value.

FIGURE 11

Fibre angles relative to the frontal plane (β). (a) Fibre coordinates on the sagittal plane in 5 bins along the anteroposterior axis, colour coded for β according to colour legend. Larger empty circles indicate the notochord centre. (b, c) Comparison of fibre elevation between bins, relative to the mediolateral axis. The lighter the colour, the more posterior (see bottom for legend). The data is split between hypaxial muscles in panel ‘b’ and epaxial muscles in ‘c’. The smaller the slope, the less variation in elevation over the width of the body.

FIGURE 12

Fibre angles between medial plane and the fibre's projection on the frontal plane (α). (a) Fibre coordinates on the sagittal plane in 5 bins along the anteroposterior axis, colour coded for α according to colour legend. Larger empty circles indicate the notochord centre. (b–e) Comparison of fibre azimuth between bins. The lighter the colour, the more posterior (see bottom for legend). The data is plotted against the mediolateral axis in panels ‘b’ and ‘c’, and against the dorsoventral axis in ‘d’ and ‘e’. The data is split between hypaxial muscles in panels ‘b’ and ‘d’, and epaxial muscles in ‘c’ and ‘e’. The smaller the slope, the less variation in α over that body axis.