Scale dependence decouple life history traits from transposable element evolution in sauropsids

Reviewer #1 (Public review):

Summary:

In this study, Liang et al. characterize the genomic composition of transposable elements (TEs) in 167 sauropsid species, with the aim of testing whether increased TE accumulation is modulated by reduced selection efficacy (Mutational Hazard Hypothesis, MHH), and whether this relationship is recursively observed at multiple scales within the group. By employing life history traits (body mass and generation time) and genetic parameters (dN/dS) as proxies of effective population size, they observe a positive relationship between drift levels and TE content across all species. This relationship, however, fades out in most of the clades within sauropsids.

On one hand, this study highlights that diverging patterns of TE accumulation are modestly but still appreciably driven by differences in selection efficacy across sauropsids, providing support for the MHH within this long-diverged vertebrate group. On the other hand, this same general principle fails to explain TE content variation when closely related species within the same clade are considered. The data presented suggest that TE accumulation is dictated by other lineage-specific mechanisms, and/or that differences in effective population size that emerged later in these clades are not relevant enough to drive divergent TE contents.

Overall, the study provides a further empirical test of the MHH, joining an increasing collection of studies that reject its global validity at macroevolutionary scales. The analysis leverages the plethora of genomic data now available for a previously poorly sampled vertebrate group, and adds insightful points of view on the nearly-neutral evolution of the genome in sauropsids.

Strengths:

The paper is well written, and the dataset is appropriate to test the predictions of the MHH. Sauropsids are characterized by highly diversified life history traits and strategies, likely very diverse effective population sizes, and very wide genome size and TE content ranges. This makes them an ideal group to investigate whether the divergent TE contents and landscapes of different lineages emerged due to differences in purging selection.

A few TE-focused comparative studies already covered squamates and birds separately (Pasquesi et al. 2018; Ji et al. 2022), or included a limited number of sauropsid species in analyses of much wider phylogenetic breadth (Weinstein & Roy, 2026; Lewin & Eyre-Walker, 2026). This work is a first explicit attempt to incorporate them together in one analysis to investigate the determinants of TE dynamics and genome size variation. Given the large comparative dataset, it also represents a considerable effort in terms of data compilation, annotation and analysis, integrating life-history and genomic traits together, and can serve as a resource for future studies.

While most of the methods employed are appropriate, the robustness of some aspects of the analysis could be improved, and interpretation of some results would benefit from a reformulation more aligned with what the data suggest (see just below).

Weaknesses:

Conclusions made on the pooled dataset are partially, sometimes solely, based on Pearson's correlations. This is not a solid practice, as results are affected by the phylogenetic structure of the data. Mentioning that a correlation is initially strong but loses signal when accounting for phylogeny is fine if used to highlight the need for accounting for phylogenetic non-independence. However, phylogenetic correlations should be those on which the conclusions are drawn, and the authors should be clear in their formulations that there is no association when PGLS or SEM suggests so, even though Pearson's correlations would suggest otherwise. For instance, the abstract reads that a correlation is found between generation time and TE content across all species, although this is not supported by PGLS. Or the claim that there is a strong positive association between dN/dS and TE proportion is, although still present, significantly downsized under phylogenetic correlation. These are just a couple of examples, but several conclusions throughout the text should be reformulated or rescaled to fit phylogenetically corrected results.

This is particularly pertinent given the problem of scale dependence that the authors highlight: the significance of a given predictor (selection efficacy) of TE accumulation can not only change according to the phylogenetic scale under consideration, but is also sensitive to compositional differences between clades, which are very pronounced here given the strong phylogenetic inertia displayed by both life-history traits and TE amounts. The purpose of phylogeny-aware methods is precisely to "discriminate conditions under which pooled correlations may arise from compositional differences rather than from repeated within-lineage evolutionary processes", so should not be regarded as a secondary result.

Since the study is based on the quantification of multiple Ne proxies, the authors could expand a bit more on that and check their expected reciprocal relationships. For example, in the SEM, body mass and generation time are untied from dN/dS variation, but an explicit test outside of it would be required. If these different indicators actually go in the same direction, it would add a support layer to them, capturing differences in selection effectiveness. In non-avian sauropsids, for example, body mass was not a predictor of dN/dS and longevity (Figuet et al. 2016): the wider dataset gathered here notably increases the sampling power to assess whether/which LHTs are concordant with dN/dS.

Connected to the point above, retrieval of expected correlations between dN/dS and life-history traits in birds has been more challenging, but this has been attributed to the presence of a substitution bias (GC-biased gene conversion). Because of this, dN/dS can become an imprecise Ne proxy if some methodological precautions are not integrated (see Bolivar et al. 2016; https://doi.org/10.1186/s13059-018-1613-z). At the same time, the avian dataset used in Marino et al. does display some expected correlations between dN/dS and several LHTs, while still calculating the metric without correction for substitution biases, but across a much higher number of species (~200). So I wonder whether the combination of substitution noise in dN/dS and less abundant sampling (and maybe less contrast in Ne?) in birds might contribute to confounding the relationships between Ne proxies and a potential correlation of those with TE content.

Reviewer #2 (Public review):

Summary:

This study aims to assess the drivers of the interspecific distribution of transposable elements (TEs), using genomes from the Sauropsids clade. In particular, the authors test the mutational hazard hypothesis (MHH), which stipulates that slightly deleterious TEs should preferentially accumulate in populations with smaller population sizes, where purifying selection is weaker.

Strengths:

The manuscript is clearly written, and the underlying concepts (MHH, Simpson's paradox) are well introduced and discussed. In that respect, the manuscript is highly didactic and an enjoyable read. The authors employ state-of-the-art phylogeny-aware statistical procedures, including structural equation modeling to infer causal relationships. The results are of high interest: the authors report that correlations between TE frequencies and proxies of population size (life-history traits - such as body mass, and molecular traits - such as the dN/dS ratio), which appear at large evolutionary scales, disappear when within-clade relationships are examined. The authors discuss these results and propose that at shorter time scales, clade- and TE family-specific bursts dominate the dynamics of TE dynamics.

Weaknesses:

While I generally agree with the authors' interpretation and find these results of great interest, I have several concerns about the underlying statistical analyses and think that some possible alternative explanations of a methodological nature should be checked to confirm the results.

(1) Three types of statistical analyses are employed: Pearson correlation coefficient, phylogenetic generalised least-squares (PGLS), and structural equation modelling (SEM). Both the Pearson correlation test and the PGLS assume normality (binormality in the case of Pearson's correlation) and linearity (and this is very likely the case for SEM as well, although it depends on the exact method employed). Deviation from these hypotheses can create erroneous P-value computations (in both directions). I could not find any assessment of how well the data fit these hypotheses in the manuscript; differences in significance could reflect differences in how the data fulfil these assumptions across time scales. Figure 2a suggests at least a deviation from normality, be it because of the Aves clade alone. A non-parametric correlation (such as Kendall's tau) should be used instead of Pearson's correlation, as it assumes neither normality nor linearity. The PGLS analysis is trickier in that respect: the authors should check the normality and independence of the model residuals (and likely homoscedasticity too, unless the underlying model allows for lineage-specific variance). An alternative would be to transform the variables using phylogenetic independent contrasts (PIC, which still assumes a Brownian model - as does the PGLS) and use a non-parametric correlation on the resulting PICs. I don't have expertise in SEM, but a quick search suggests that non-parametric (or at least non-linear) SEM implementations are available.

(2) The positive correlation (l208, l299) with generation time should be further commented on, as it has the opposite direction as expected. Following the proposed rationale, species with shorter generation times should have more TEs, shouldn't they? Unless "generation time" here actually refers to the number of generations per year, instead of the duration of a generation, as the text suggests?

Author response:

Thank you very much for your constructive assessment and the thoughtful comments provided by both reviewers. We greatly appreciate their valuable and insightful suggestions, which will help us strengthen the analyses and interpretation of our study. We are pleased to provide the following provisional responses outlining our planned revisions. A fully revised manuscript, together with detailed point-by-point responses to all reviewer comments, will be submitted in due course.

Responses to Reviewer #1:

Phylogenetically informed statistical analyses. We fully agree that phylogenetically informed analyses should take precedence when interpreting cross-species associations. In the revised manuscript, we will reorganize the Results to present PGLS and SEM findings as the primary evidence. We will also carefully revise the Abstract, Results, and Discussion to ensure that our conclusions are consistent with the phylogenetically corrected results, particularly regarding the relationship between generation time and TE content across all species and the strength of the association between dN/dS and TE content.

Assembly quality as a potential confounding factor. We will examine whether BUSCO completeness scores and contig N50 are associated with genome size and TE content. Where appropriate, we will incorporate assembly quality metrics as covariates in the PGLS and SEM models or perform additional sensitivity analyses. We will also consider conducting separate analyses for avian and non-avian sauropsids to evaluate the robustness of our findings and further explore their phylogenetic scale dependence.

SEM methodology. We will expand the Methods section to provide a clearer explanation of how the piecewise SEM was constructed, the biological rationale underlying the proposed pathways, and the criteria used for model evaluation and selection, including comparisons of alternative model configurations where appropriate. We will also explore the feasibility of conducting separate SEM analyses for avian and non-avian sauropsids.

Residual support for the MHH and contrast dependence. We agree that the weak but persistent associations between TE content and proxies of effective population size (Ne) in the pooled dataset deserve greater emphasis. We will discuss these relationships more explicitly, including the association between dN/dS and genome size. We will also expand our discussion of contrast dependence, particularly the possibility that reduced variation in life-history traits and TE content, together with differences in sample size and lineage-specific evolutionary dynamics, may contribute to the weakening of associations at smaller phylogenetic scales.

dN/dS estimation. We will clarify the procedures used for gene-level filtering and dN/dS estimation, particularly the treatment of genes with very low dS values, and evaluate the potential influence of alternative estimation approaches. We will also further examine the relationships among body mass, generation time, and dN/dS to assess the consistency of these variables as indirect proxies of Ne and selection efficacy. In addition, we will discuss the potential influence of GC-biased gene conversion (gBGC) on dN/dS estimates in birds, with reference to the relevant literature.

Recent TE elements. We agree that examining recently accumulated TEs would provide a valuable additional perspective. We plan to investigate whether the abundance of recent TEs (Kimura divergence <5%) is associated with life-history traits or dN/dS and present the results in the supplementary materials.

We will also address all minor comments, including moving interpretive statements from the Results to the Discussion, incorporating the suggested references (including doi:10.1111/mec.70265 and RepeatMasker), applying log transformations to genome size where appropriate, correcting the order of figure citations, defining the abbreviation "RE", clarifying or removing the unexplained close-up in Figure 1, and refining the relevant wording throughout the manuscript.

Responses to Reviewer #2:

Statistical assumptions (normality, linearity, and homoscedasticity). We will systematically evaluate the statistical assumptions underlying our analyses, including the distributional properties of the data and the residuals of the fitted models. Specifically, we will (i) complement Pearson correlations with non-parametric alternatives, such as Kendall's tau, to assess the robustness of the observed associations; (ii) examine PGLS residuals for normality, homoscedasticity, and remaining phylogenetic dependence, and consider non-parametric correlations based on phylogenetic independent contrasts (PIC) as an additional sensitivity analysis; and (iii) assess the assumptions and goodness of fit of the SEM component models and explore alternative modeling approaches where appropriate.

Direction of the generation time correlation. We will clarify the definition and units of generation time used in the dataset (duration of a generation, in years) and discuss this counterintuitive positive association explicitly, relating it to the expectation under the MHH and to previous findings.

Reproducibility. We will improve the reproducibility of our analyses by providing the input datasets alongside the corresponding R scripts, detailed README files, and relevant software and package information. We will also make the analysis code available under an appropriate open-source license to facilitate the reproduction of our statistical models and figures.

Minor comments. We will provide additional justification for selecting Sauropsida as the phylogenetic scope of our study, clarify the definition of the "matched" dataset, revise the use of "superficially", and further explain the interpretation of body mass as a potential physiological factor or an indirect proxy for population density across different taxonomic groups.

We have no factual concerns regarding the eLife Assessment or the public reviews.

Thank you again for the opportunity to strengthen our work. We believe that these revisions will improve the statistical robustness of our analyses and provide a more nuanced interpretation of the associations between TE accumulation, life-history traits, and molecular evolutionary indicators across phylogenetic scales.