PREreview of Deficiency of the lipid flippase ATP10A causes diet-induced dyslipidemia in female mice

This review reflects comments and contributions by Veronica Busa, Hector Fugihara Kroes, Pablo Ranea-Robles, Cláudia Gil and Pearlyn Toh. Review synthesized by Pablo Ranea-Robles.

The manuscript reports that Atp10A deficiency in mice results in female-specific dyslipidemia when mice are fed a high-fat diet, characterized by elevated plasma triglycerides, free fatty acids and cholesterol levels. These changes were accompanied by changes in circulating metabolites, which may help understanding ATP10A functions in the membrane for lipid translocation. Additionally, ATP10A deficiency led to hepatic insulin resistance without perturbations in glucose homeostasis. The authors also report a high basal insulin signaling in endothelial cells in vitro when ATP10A was knocked down, which led to desensitization of the insulin response after stimulation. With these data, the authors propose that ATP10A, a flippase, modulates the cellular lipid environment, causing changes in localization of bioactive lipids which have an impact on insulin signaling, facilitating the attenuation of the insulin receptor. We found this preprint interesting and relevant, revealing new roles for ATP10A, which had been associated before with metabolic homeostasis, but whose role was not well defined. We found that, in general, the conclusions of the paper are well supported by the data, data visualization was excellent, most of the data analysis is available in repositories, and the limitations of the study are well addressed in the discussion section. We summarize here the comments of the crowd in order to improve the understanding and readability of this study.

Major comments:

• Even though we agree in general on the role of ATP10A after the data generated in this manuscript, we think that the main conclusion and title of this preprint may be misleading. Authors reached this conclusion based on ATP10A KO mice or KD cells. To fully support the claim in the title, a reversal experiment should be performed with re-introduction of ATP10A or overexpression of ATP10A in diet-induced obesity animals (after phenotypes were established) to conclude that “ATP10A promotes endothelial cell insulin sensitivity and protects against diet-induced dyslipidemia”. We understand this can be a whole other project itself, so instead of asking for those experiments, we suggest changing the title accordingly, to state that lack of ATP10A leads to insulin resistance and dyslipidemia in female mice.

• The female-specific phenotype is very intriguing. A great addition to this paper would be to evaluate ATP10A expression in male and female tissues, at least in the liver and adipose tissue. If the ATP10A staining showed in liver and visceral fat was performed in female mice, do authors observe similar staining in male mice?

• The claim in the abstract “Here, we generated Atp10A knockout mice and show that Atp10A deficiency results in female-specific dyslipidemia, independent of diet-induced obesity” does not seem correct as the phenotype was only present in mice with HFD. One key experiment that we think is missing is the measurement of the parameters that led to the conclusion of female-specific dyslipidemia and changes to lipoprotein metabolism in chow-fed mice. That would prove that the dyslipidemia is diet-induced and not only caused by ATP10A deletion.

• In Figure 1, the authors claim that “the area under the curve (AUC) from the OGTT shows a trend toward the Atp10A-/-female mice having greater glucose excursion compared to Atp10A+/+controls”. We are not convinced by this claim by looking at the data, there is no comparison between males and females, and the genotype difference is very similar in both sexes.

• Following on the changes in these parameters, it would be really interesting if authors had data from the lipidomics analysis related to these molecules, as it would confirm the result obtained on Figure 2. For example, levels of unconjugated fatty acids.

• The difference observed in plasma FFA and cholesterol between males and females, regardless of genotype, is quite striking (4-5 fold-change) and not expected. Could the authors provide an explanation for that? Were the conditions different? There should be a mention to this difference in the results or the discussion

• Authors found a decrease in the activating phosphorylation of cPLA2 together with changes in associated signaling pathways. This is linked to the observed changes in circulating eicosanoids. However, it appears as the decrease in cPLA2 phosphorylation is parallel to a decrease in the amount of total cPLA2 protein. The graph seems to indicate that the quantification of p-cPLA2 was done relative to beta-actin, but perhaps it will be better to quantify relative to total levels of cPLA2.

Minor comments:

• We found it was not entirely clear the diet condition on some panels across the manuscript. It would improve the readability of the manuscript if it was clear under which conditions every measurement was done, i.e, when the experiment was performed in chow mice and when in HFD mice. 

• Considering the importance of ATP10A and the existence of previous studies on its role in metabolic homeostasis we think it would be great to add a bit more information of the results of those previous studies, as it is done in the discussion.

• The validation of the KO is well supported with data in Figure 5, but authors could add a picture of a gel with bands after amplification in WT and KO allele, so the result of the PCR is available to everyone.

• This study found that ATP10A deficiency has a stronger impact on female mice. However, when placed on HFD, at the end of the 12-week period of HFD feeding, the trajectory of BW between WT and KO male mice seem to diverge. We consider this finding is at least worth to be discussed in the discussion

• In Fig 1 legend it is said that the percentage of lean and fat mass was obtained by dividing the fat or lean by the combined sum of lean+fat. Did the authors check that the number was equal if BW was used as the denominator?

• The text in the results section related to Supp Fig 2 does not fit with the data in the figure, where a difference can be seen based on genotype during the light period. We also thought that the table is kind of difficult to grasp, one possibility would be to show the plots with hourly changes in EE and other parameters obtained with the Sable system so the differences are better seen.

• The last sentence of the first paragraph in the results section was a bit confusing for us. We think that it could be better understood if "cause" is changed to "modulate" or "alter the development", because the diet indeed causes obesity, the genotype can only alter this.

• Probably not many people are familiar with P4-ATPases, but many more are with flippases, so that information that is present in the Introduction we think would be useful for the reader

• The authors mention use of the Gtex database, but there is no reference regarding that database. We suggest including one.

• Since the Infinity kit was used to measure TG, it should be mentioned that the measurement includes free glycerol

• Since there is no significant difference in the analysis of data from Fig 2G, the conclusion should be toned down a bit and reflect the fact that it is a trend. We agree that the difference is there and is not negligible but adding this nuance will better reflect the data.

• We suggest adding that mice were fasted in the results section when reporting the changes in the lipidome.

• In figures such as supplemental 1A in which there are many data points with a low spread, the y-axis could be adjusted to show a smaller range. This would allow a better utilization of plot space and a clearer visualization of the data.

Comments on reporting:

• There is a mention to Supplementary Table 1, but we could not find it across the preprint

• This preprint includes multiple western blots, but molecular weight indicators are missing. Addition of these data and inclusion of full uncropped blots in Supplementary Data would be important for reproducibility.

• In supplemental figure 8 the significance levels are incorrectly captioned: “*p≤0.05; **p≤0.005; *p≤0.0005; ***p≤0.0001” (when it should probably be “*p≤0.05; **p≤0.005; ***p≤0.0005; ****p≤0.0001”).

• In supplemental figure 8D, there is an indication of statistical significance even when the captions deny that statistical analysis were made due to low n. If no statistical analysis was made, there should be no annotation. In that same figure, there are 4 data points for 10A+/+ when n should be 2 according to captions.

• Avoid using different significance level notations across figures and if you do so, make sure to be transparent. In supplemental figure 1A and supplemental figure 4E p-values >0.005 are annotated as in the significance level “**” while following the reference established in other figures, it should be “*”.

• Even though the number of mice is well defined in the figure legends, we found it unclear in Figure 3, when authors report 5 samples from 8 mice, so please specify what constitutes n there.

• The authors should be commended for providing links to -omics raw data deposited in public repositories. That certainly improves the reproducibility of the study. For complete reproducibility, we encourage the authors to share energy expenditure data and their code for data analysis or at least provide more information about tools and packages used to analyze all sequencing and -omics data in the methods section. Some of this information is in the figure legends, where it is ill-suited, but much of it is missing in case someone wanted to reproduce the analysis.

• We found statistical analysis in figure legends a bit repetitive sometimes. For example in Figure 2, the same information could be given once at the end of the text of the legend and there will be no information missing. That would help readability.

• Authors did a very good job representing individual data points in the figures. However, in Figure 2H and 2I they are missing (the individual data points).

• As a positive, Figure 1A is very clear. It is easy to understand what was done, even though the authors do not show the KO confirmation.

• If one checks the online version of the preprint, not the pdf, Supplemental figure 4 and 5 are identical, authors may want to fix this.

Suggestions for future studies

• Use an endothelial-specific ATP10A KO model to interrogate whether this expression difference alone is sufficient for the observed phenotypes (it is mentioned that ATP10A is highest expressed in endothelial tissue, but is that the only relevant tissue? E.g. is female-specific liver insulin resistance rescued with only the very low hepatocyte ATP10A expression?). This is well-addressed in the discussion by the authors

• The authors hypothesize that the sex-specific phenotypes are attributable to differences in hormones and their receptors. Although outside the scope of this manuscript, future studies should use the four-core mouse genotypes to deconvolute genetic versus hormonal causes.

Competing interests

The author declares that they have no competing interests.