Pleiotropic effects of trisomy and pharmacologic modulation on structural, functional, molecular, and genetic systems in a Down syndrome mouse model

Joint Public Review:

Using Ts65Dn - the most commonly used mouse model of Down syndrome (DS) - the goal of this study is two-pronged: 1) to conduct a thorough assessment of DS-related genotypic, physiological, behavioral, and phenotypic measures in a longitudinal manner; and 2) to measure the effects of chronic GTE-EGCG on these measures in the Ts65Dn mouse model. Corroborating results from several previous studies on Ts65Dn mice, findings of this study show confirm the Ts65Dn mouse model exhibits the suite of traits associated with DS. The findings also suggest that the mouse model might have experienced drift, given the milder phenotypes than those reported by earlier studies. Results of the GTE-EGCG treatment do not support its therapeutic use and instead show that the treatment exacerbated certain DS-related phenotypes.

Strengths:
The authors performed a rigorous assessment of treatment and examined treatment and genotypic alterations at multiple time points during growth and aging. Detailed analysis shows differences in genotype during aging as well as genotype with treatment. This study is solid in the overarching methodological approach (with the exception of RNAseq, described below). The biggest strength of the study is its approach and dataset, which corroborate results from a multitude of past studies on Ts65Dn mice, albeit on adult specimens. It would be beneficial for the dataset to be made available to other researchers using a public data repository.

Weaknesses:
There are several primary weaknesses, described below:

Sex was not considered in the analyses
The number of experimental animals of each sex are not clearly represented in the paper, but are buried in supplemental tables, and the Ns for the RNAseq are unclear. No analyses were done to examine sex differences in male/female DS or WT animals with or without treatment. Body measurements will greatly vary by sex, but this was not taken into consideration during assessments. As such, there is a high amount of variability within each cohort measured for body assessments (tibia, body weight, skeletal development etc.). Supplemental table 14 had the list of each animal, but not collated by sex, genotype or treatment, making it difficult to assess the strength of each measurement.

Key results are not clearly depicted in the main figures
Rigorous assessment of each figure and the clarity of the figure to convey the results of the analysis needs to be performed. Many of the figures do not clearly represent the findings, with authors heavily relying on supplemental figures to present details to explain results. Figure legends do not adequately describe figures; rather, they are limited to describing how the analysis is performed. For example, LDA plots in Figure 4 do not clearly convey the results of metabolite analysis.
Overall, the amount of data presented here is overwhelming, making it difficult to interpret the findings. Some assessments that do not add to the overall paper need to be removed. Clarifying the text, figures and trimming the supplement to represent the data in a manner that is easily understood will improve the readability of the paper. For example, perhaps measures which are not strongly impacted by genotype could be moved to the supplement, because they are not directly relevant to the question of whether GTE-EGCG reverses the impact of trisomy on the measures.

Lack of clarity in the behavioral analyses
Behavioral assessments are not clearly written in the methods. For example, for the novel object recognition task, it isn't clear how preference was calculated. Is this simply the percent of time spent with the novel object, or is this a relative measure (novel:familiar ratio)? This matters because if it is simply the percent of time, the relevant measure is to compare each group to 50% (the absence of a preference). The key measures for each test need to be readily distinguished from the control measures.
There are also many dependent behavioral measures. For example, speed and distance are directly related to each other, but these are typically reported as control measures to help interpret the key measure, which is the anxiety-like behavior. Similarly, some behavioral tests were used to represent multiple behavioral dimensions, such as anxiety and arousal. In general, the measurements of arousal seem atypical (speed and distance are typically reported as control measures, not measures of arousal). Similarly, measures of latency during training would not typically be used as a measure of long-term memory but instead reported as a control measure to show learning occurred. LDA analysis requires independence of the measures, as well as normality. It does not appear that all of the measures fed into this analysis would have met these assumptions, but the methods also do not clearly describe which measures were actually used in the LDA.

Unclear value of RNAseq
RNAseq was performed in cerebellum, a relatively spared region in DS pathology at an early time point in disease. Further, the expression of 125 genes triplicated in DS was shown in a PCA plot to highly overlap with WT, indicating that there are minimal differences in gene expression in these genes. If these genes are not critical for cerebellar function, perhaps this could account for the lack of differences between WT and Ts65Dn mice. If the authors are interested in performing RNAseq, it would have made more sense to perform this in hippocampus (to compare with metabolites) and to perform more stringent bioinformatic analysis than assessment by PCA of a limited subset of genes. Supplementary Table S14, which shows the differentially expressed genes, appears to be missing from the manuscript and cannot be evaluated. Additionally, the methods of the RNAseq are not sufficiently described and lack critical details. For example, what was the normalization performed, and which groups were compared to identify differentially expressed genes? It would also be worthwhile to describe how animals were identified for RNAseq-were those animals representative of their groups across other measures?