During embryonic and larval stages the zebrafish epidermis consists of an outer layer, the periderm, and an inner basal layer attached to the basement membrane; it becomes stratified only at metamorphosis (Le Guellec et al., 2004). The two layers differ in origin and in cytoskeletal organisation. Periderm cells carry on their free surface a labyrinth of actin-based protrusions called microridges, which have been imaged in living animals and shown to form from smaller precursors under the control of cortical contraction (Lam et al., 2015; van Loon et al., 2020). Basal cells lack microridges and face the basement membrane, to which they are anchored. A protocol must therefore specify which layer is being imaged, since the same reporter expressed in both would give superimposed signals a few micrometres apart.

Three kinds of genetically encoded probe are in common use for filamentous actin. Lifeact is a short peptide of 17 amino acids derived from a yeast actin-binding protein and fused to a fluorescent protein (Riedl et al., 2008). A second probe uses the calponin homology domain of utrophin (Burkel et al., 2007). The third is fluorescently tagged actin itself. They are not interchangeable: a systematic comparison found that each is enriched on some filament networks and under-represented on others, and that high expression can change network architecture (Belin et al., 2014). For microtubules, a tagged tubulin shows the whole array, whereas a tagged end-binding protein such as EB3 marks only growing plus ends and allows polymerisation to be followed as moving comets (Stepanova et al., 2003).

The reporter should be chosen for the question. For the overall shape of cortical actin and microridges, Lifeact or the utrophin domain is suitable; for turnover within a structure, tagged actin is more appropriate. Whatever the choice, the protocol recommends that a key observation be repeated with a second, unrelated probe before it is interpreted. Expression should be restricted to one layer by a keratin promoter. The promoter characterised by Gong et al. (2002) drives expression in the superficial layer, and promoter fragments specific for basal cells have also been described. Constructs are conveniently assembled with the Tol2kit, which combines a promoter, a coding sequence and a polyadenylation signal in a transposon backbone (Kwan et al., 2007).

Two modes of expression are useful. Injection of the construct with transposase messenger RNA at the one-cell stage gives mosaic animals in which isolated labelled cells sit among unlabelled neighbours, so that cell outlines and protrusions can be resolved. A stable transgenic line gives uniform labelling for measurements across the tissue. For either, the dimmest animals in which the structure can still be seen should be selected, because the brightest are the most likely to be perturbed.

Embryos are raised at the standard temperature of 28.5 °C and staged by morphology and not by hours alone (Kimmel et al., 1995). If imaging will extend beyond the onset of pigmentation, melanin can be suppressed with phenylthiourea (Karlsson et al., 2001), although epidermal imaging is often possible between melanophores and the drug is best avoided when it is not needed. Unhatched embryos are dechorionated with fine forceps shortly before mounting. Screening for fluorescence should be done quickly on a stereomicroscope at low power, since exposure during screening counts against the light budget described in Step 4.

Animals are anaesthetised with buffered tricaine at the concentration given in standard husbandry references (Westerfield, 2000) and embedded in low-melting-point agarose on a glass-bottomed dish, with the region to be imaged against the glass. The agarose should be cooled to just above its gelling point before the animal is added. The lowest concentration that prevents drift should be used, because embryos held in stiff agarose grow abnormally over long recordings, a problem documented for light sheet microscopy and equally relevant here (Kaufmann et al., 2012). Once the gel has set it is covered with embryo medium containing tricaine. The specimen should not be pressed against the glass. The periderm is the outermost cell layer of the animal, and compression flattens microridges and can wound it.

Phototoxicity is the main threat to a recording of cytoskeletal dynamics, since its early signs are subtle changes in exactly the rates being measured. The protocol therefore asks the user to decide the total light dose before the experiment and to treat it as a fixed budget to be divided among time points and focal planes. The recommendations of Icha et al. (2017) and Laissue et al. (2017) are followed: use the longest excitation wavelength for which a suitable fluorescent protein is available, prefer a camera-based spinning disk or a light sheet system to a point scanner where there is a choice, lower laser power and lengthen exposure instead of the reverse, and stop illumination between frames. The power at the objective should be measured with a meter and recorded, since a percentage setting in the software means nothing to another laboratory.

The interval between frames should be set by the process. Plus-end comets and actin flow require intervals of seconds over a few minutes; the remodelling of microridges or the rearrangement of cells requires intervals of minutes over hours. Sampling faster than the process demands spends the budget for no gain. Because the epidermis is thin, a stack of a few planes spanning one layer is sufficient, and a single plane is often enough for a flat region over the yolk or the trunk. Focus drift is better corrected by a hardware autofocus or by acquiring a slightly larger stack than by re-imaging. The stage should be held at the rearing temperature throughout.

A recording should be accepted only if the animal was demonstrably healthy at its end. The protocol specifies three checks. Heartbeat and circulation are inspected by transmitted light before and after the session. The imaged field is examined for signs of damage, such as cell rounding, blebbing or extrusion of cells from the periderm. Where the recording is long enough, the occurrence of normal cell divisions in the field is noted, since mitotic delay is a sensitive indicator of light damage (Laissue et al., 2017). Siblings mounted in the same dish and not illuminated serve as controls for the effects of mounting and anaesthesia alone. If any check fails, the light budget is reduced and the experiment repeated, which is the return loop in Figure 1.

Raw files are kept unaltered, and all processing is carried out by script on copies so that it can be repeated. Fiji provides the necessary tools for registration, kymograph construction and segmentation (Schindelin et al., 2012), and TrackMate can follow plus-end comets or other discrete objects (Tinevez et al., 2017). Measurements should be summarised per animal before any statistical comparison, because cells within one animal are not independent. The methods section of any resulting paper should give the transgene and its mode of expression, stage, mounting medium and anaesthetic, objective, measured power, exposure, interval and duration, together with the animal details required by the ARRIVE guidelines (Percie du Sert et al., 2020).