Comment on a PREreview of "On theindependent irritabilityof goldfish eggs and embryos – a living communication on the rhythmic yolk contractions in goldfish"

This Zenodo record is a permanently preserved version of a comment on a PREreview. You can view the complete PREreview and comments at https://prereview.org/reviews/11444773. This work presents the first modern descriptions of the rhythmic contractility of the yolk of the domesticated goldfish Carassius auratus. The authors use image analysis tools to quantify the contractions using the mean pixel intensity of the yolk as it oscillates over time. They demonstrate that these oscillations are not present in either zebrafish or carp embryos. They then go on to show that contractions are an autonomous property of the goldfish yolk using carp/goldfish hybrids and unfertilized eggs. Finally, they show that manipulations of early embryonic patterning correlate with changes to speed and/or the magnitude of yolk contractions. Overall this study of early goldfish development is a valuable contribution to the existing body of work on early teleost development and highlights the variation in the evolution of embryo and egg behaviour that we can observe when we sample widely across taxa. However, the data currently in this manuscript is not sufficient to support the more detailed conclusions of the authors. Major points: The various metrics used to quantify yolk contractility in Figure 1 are a great starting point however, the focus on the mean pixel intensity of the yolk/embryo as the central metric for quantifying yolk contractions throughout the paper is a cause for concern. My concern is that the mean pixel intensity metric is too tangentially related to the biological phenomena of interest to be a consistent readout. It is not clear what high or low values of this metric would mean biologically or whether they would even be the same between imaging sessions. This is in contrast with a metric like circularity which has a well-defined meaning for shape analysis. The reasoning in the first section of the results on page 2 for using this metric, that frames at similar phases of oscillation will have similar intensities, does not seem to account for any irregularity in the contractions (particularly as they get stronger) which might cause differences in where the light hits the yolk. Similarly, there is clear variation in the illumination of the yolk between different embryos (see supplementary movies M1 and M5). While the detrending performed on the oscillating trace adjusts for the absolute intensity differences between embryos it will not account for differences in the range of pixel intensities, an example of which would be in movie M1 where some embryos have completely saturated pixels and others do not. Instead, I would suggest that the authors use circularity or a similar metric as their main readout of yolk contraction. While the circularity trace shown in Figure 1 is less smooth than the mean pixel intensity trace it is likely to be a much more consistent and biologically meaningful quantification. Alternatively, if the authors wanted to quantify the contractions in even greater detail, I would suggest generating a set number of radii from the centre of mass of the embryo/yolk and measure their change in length as the yolk contracts. This would pick up more of the subtleties of contraction and would give a nice way to investigate its symmetry. Response: Thank you for this comment. We understand that the reason for the choice of quantification method was not elaborated in the first version of the manuscript. We expanded our discussion on this in the Extracting the mean pixel value of an egg / embryo over time allows quantification of the period of yolk contractions from simple timelapse stereomicroscopy section of the second version of the manuscript, which we quote here: "To quantify these contractions, we segmented the yolk and determined its circularity, its perimeter, and its projected area over time. However, we immediately realized that quantifying yolk contractions from measurements of the shape of the yolk requires precise image segmentation (with the yolk as foreground) over long periods of time (spanning > 12 hours post-fertilization, from cleavage to epiboly stages). This is challenging in our simple experimental and imaging conditions as the yolk has low contrast compared to the background and to the ooplasm / embryo proper, the yolk changes its shape and position in 3D, and the entire egg / embryo is not fixed in one axis. Accordingly, we also extracted the timeseries from the mean pixel value of the entire egg / embryo, circumscribed by its chorion, over time (Figure 1, Supplementary Movie M2, reasoning that still-frames of an egg / embryo after a full contraction of the yolk would look most similar, and hence would have most similar mean pixel values (Figure 1 1 and 3, 2 and 4, 5 and 8). For the analysis, we detrended the extracted timeseries via sinc-filter detrending after specifying a cut-off period. We subjected the detrended timeseries to continuous wavelet transform using the Morlet wavelet as mother wavelet (Mönke et al., 2020), allowing us to recover the instantaneous period along a ridge tracing wavelet with maximum power for every timepoint. Comparing the timeseries and the evolution of period over time, we noted that the mean pixel value of the embryo is a proxy of the projected area of the segmented yolk. However, as it does not rely on image segmentation, extracting the timeseries of yolk contractions from the mean pixel value of the entire egg / embryo is additionally more robust, at least in quantifying the period of contractions (which does not rely on sample orientation, unlike quantifying phase and amplitude), compared to measurements of the perimeter, circularity, or projected area of the yolk when some segmentation is possible. As exemplified in Figure 1, the timeseries extracted from the mean pixel value provides a suitable approximation of the period of yolk contractions, matching the actual period of a full contraction (e.g. Figure 1, timepoints 1 and 3, or 2 and 4, or 5 and 8) determined from manual inspection of timelapse imaging (i.e. the ground truth). Thus, the method from mean pixel value provides a simple and flexible measure of the yolk contractions, and we henceforth used this method in our quantification." While I am not too familiar with the details of the wavelet transformation method, I can see the use for extracting the period and amplitude etc of the oscillating trace in question. However, I do not understand why it would be applied to a non-oscillating trace such as those acquired from the zebrafish or carp embryos. In this case I would assume that the resulting properties of the waveform do not actually correspond to true changes in their behaviour but rather minor variation in quantification/imaging over time that has been erroneously amplified. This undermines the conclusion, which would appear to be otherwise strongly supported from the live movies presented, that yolk contractions are found in goldfish but not carp or zebrafish. Instead of the wavelet analysis I would suggest presenting the traces currently found in supplementary figure 1 which clearly show a lack of oscillation in these species. Response: We now included comparisons of the timeseries from the embryos to those of background regions in close proximity to the fish embryos (Supplementary Figure F2B) in the second version of the manuscript. We showed that the rhythms we detect are not an artefact of image acquisition or due to background effects, as we reported in the Goldfish, unlike closely-related common carp and zebrafish, exhibit persistent rhythmic yolk contractions during embryonic development section of the second version: "The rhythms we had detected for the three fish species had high wavelet powers (Supplementary Figure F1A), correlating well with considered wavelets and not with white noise (Mönke et al., 2020) at least at specific timepoints during development (see differences in temporal profile of wavelet power between the three fish species as represented by black bars in Supplementary Figure F1A). Comparison with timeseries from background regions (in close proximity to analyzed fish embryos) verified that the detected embryo rhythms were not an artefact of the image acquisition nor were due to background effects (Figure 2B versus Supplementary Figure F2B)." In addition, we actually also detect a similar long (slow) period rhythmic trend in goldfish, as we also report: "Interestingly, we also observed a trend in the timeseries of goldfish embryos (see for e.g. raw timeseries of representative goldfish sample in Figure 2B) seemingly matching the slow (long-period) rhythms detected in common carp and zebrafish. For some goldfish samples (e.g. 15/61 goldfish samples plotted in Figure 2C), this slow rhythm could even register higher wavelet power than the faster yolk contractions when the cut-off period was set at 850 seconds. We are still not certain what these slow rhythms are." The data presented in Figure 2 and 3 very nicely demonstrate, in two different ways, that yolk contractions are an autonomous property of the yolk cell itself. However, my major point about these figures is the need for more samples in order to be able to conclude that (a) unfertilized eggs begin contracting at the same time as fertilised ones (one embryo does, while the other begins contracting later), and (b) the period of unfertilised yolk contraction is meaningfully faster than in fertilised eggs (there is considerable variation in period between the samples presented). If the latter remains true with further repeats then this result could relate to the increased period of contraction seen when treating with nocodazole or the oranda mutant, it could be possible that changes to the embryonic development generally may alter the dynamics of contraction. Additionally, the authors should consider whether these manipulations don't just change the period or amplitude but are actually affecting the