Toward a universal measure of robustness across model organs and systems

The development of an individual must be capable of resisting the harmful effects of internal and external perturbations. This capacity, called robustness, can make the difference between normal variation and disease. Some systems and organs are more resilient in their capacity to correct the effects of internal disturbances such as mutations. Similarly, organs and organisms differ in their capacity to be resilient against external disturbances, such as changes in temperature. Furthermore, all developmental systems must be somewhat flexible to permit evolutionary change, and understanding robustness requires a comparative framework. Over the last decades, most research on developmental robustness has been focusing on specific model systems and organs. Hence, we lack tools that would allow cross‐species and cross‐organ comparisons. Here, we emphasize the need for a uniform framework to experimentally test and quantify robustness across study systems and suggest that the analysis of fluctuating asymmetry might be a powerful proxy to do so. Such a comparative framework will ultimately help to resolve why and how organs of the same and different species differ in their sensitivity to internal (e.g., mutations) and external (e.g., temperature) perturbations and at what level of biological organization buffering capacities exist and therefore create robustness of the developmental system.

. When perturbations do happen, the degree to which development can counteract, or be resilient, can make the difference between normal variation and disease (Gibson & Lacek, 2020;Hallgrimsson et al., 2018).Yet, while developmental systems must be robust to safeguard the heritability of traits, it is their flexibility and evolvability that ultimately permits evolutionary change (Masel & Trotter, 2010;Payne & Wagner, 2019).
Not all organs are equally robust.For example, skin characteristics such as fingerprints (Pankanti et al., 2002) or the pigmentation pattern of a zebra (Prinsloo et al., 2020) show more variation than the tooth cusp patterning of a mammalian tooth (Grimes et al., 2019).Different organs (and different characteristics of these organs) have acquired various degrees of robustness during evolution, and even the same organs can differ in different taxonomic groups in the way they respond to perturbations (Salomies et al., 2021;Savriama et al., 2018).Furthermore, internal effects including mutations, and external factors such as temperature and nutrition can presumably challenge the faithfulness of the developmental programs differently.All the differences among organisms, organs, and types of perturbations indicate that calling a system simply "robust" is relatively uninformative on its own.To be useful, robustness should be considered in a comparative context, and ideally, we should be able to measure the degree of an organ's robustness.Here we will discuss what would be needed to achieve such a comparative framework in the study of robustness, and its connections to the concept of agency (Sultan et al., 2022).

| TOWARD A MORE COMPARATIVE APPROACH TO STUDYING ROBUSTNESS
Whereas robustness has been studied in individual taxa by measuring and classifying phenotypic variation, we argue that we lack a framework to compare the development of robustness among different organs and vertebrate species.For example, in evolutionary studies, fluctuating asymmetry (random deviation from symmetry) is an important measure of robustness (Graham & Özener, 2016).However, fluctuating asymmetry is often subtle and difficult to quantify (Palmer & Strobeck, 1986).This means that fluctuating asymmetry, and phenotypic variation in general, may be simply ignored in developmental biology studies.This contrasts with the gross phenotypic effects that are often sought after in developmental experiments (using, e.g., genetic or surgical manipulations).Yet, these kinds of experiments are generally not focused on assessing robustness or quantifying phenotypic variation (but see [Green et al., 2017]).Cumulatively, the experimental research program has provided a nearly exhaustive inventory of the effects of loss of function of different genes on diverse organs.What we lack are ways to integrate all these different experimental studies to systematically measure robustness.More specifically, we lack a framework to examine whether organs that are vulnerable to internal perturbations are also sensitive to external factors affecting development.In the simplest terms, we lack a comparative framework to evaluate how "noisy" different organs and their development really are.The development of new approaches is therefore needed to shed light on the mechanistic underpinnings of robustness by testing the faithfulness of their developmental programs against both internal and external perturbations.

| TARGET ORGANS FOR ASSAYING ROBUSTNESS
Most organ systems are thought to employ comparable signaling networks, core genes, and logic of patterning (Cardoso-Moreira et al., 2019;Li et al., 2019;Villar et al., 2014).This raises the question of whether different organs have also comparable robustness, and what regulatory features might make these systems behave differently, including, for example, differences between organs and species.Here, we claim that why and how organ features differ in their robustness is a largely understudied yet highly relevant research question.To make this question operational, one should select target organs allowing effective comparisons.A key approach to assay and compare robustness would be to include target organs that (a) are relatively well-conserved across organisms, (b) ideally differ in their robustness (either between species or between organs if several organs are studied), (c) can be analyzed both in vivo and in organ culture conditions, and that (d) can be analyzed at different levels of biological organization (from gene regulation to phenotypes).
For example, epithelial organs such as teeth, scales, feathers, hair follicles, and glands are partially highly conserved and share many aspects of their genetic architecture, regulatory network topology, and logic of patterning (Jernvall & Thesleff, 2012;Pispa & Thesleff, 2003).Yet, these organs also greatly differ in their robustness.Mammalian teeth for example show a highly canalized development and even faithfully develop in organ culture conditions (Jernvall & Jung, 2000).In contrast, other mammalian epithelial organs such as mammary glands (Veltmaat et al., 2013), or teeth in nonmammalian lineages (Lafuma et al., 2021;Salomies et al., 2021) can be highly variable.They have also specific advantages.For example, teeth and to some extent also scales can be investigated and measured in live specimens, as well as in fossils.Experimentally, teeth, hair follicles, and glands can be analyzed in organ cultures using live imaging.Thus, rather than being limited to the final, phenotypic end product, we can compare the process of development and when and how organs differ in their robustness.

| A UNIVERSAL RULER FOR ROBUSTNESS
A major challenge when assessing the robustness of a system is the difficulty to disentangle variation caused by changes in robustness and the inevitable random variation of biological systems, as well as those caused by for example variation between experiments (Masel & Trotter, 2010).Therefore, to measure the variation directly caused by decreased developmental robustness we need conceptual approaches to assess the faithfulness of development.We propose that bilateral symmetry as a distinct feature of many vertebrate organs could be taken advantage of to study robustness across species.Specifically, random deviations from symmetry, which is often called fluctuating asymmetry, have been already used as a measure of robustness (Graham & Özener, 2016;Palmer & Strobeck, 1986).Whereas genetic and environmental factors (stressors) are typically inferred to contribute to fluctuating asymmetry, identifying the cause is often challenging (Graham & Özener, 2016;Savriama & Klingenberg, 2011).This is in part because fluctuating asymmetry is typically a small fraction of the total phenotypic variation.Yet, pushing organ systems experimentally might greatly increase their variation, including fluctuating asymmetry.In the next chapter, we will explain the power of using fluctuating asymmetry as a ruler for measuring system robustness.

| MEASURING ROBUSTNESS AGAINST DIFFERENT PERTURBATIONS ACROSS LEVELS OF BIOLOGICAL ORGANIZATION
Besides the analysis of robustness in a comparative framework (comparing organs and species), at least two additional research gaps can be noticed.The first gap concerns the comparative analysis of different types of perturbations, that is, internal versus external perturbations.The second research gap is the analysis of robustness in an integrative manner.Overall, we lack an understanding of the mechanisms and levels of biological complexity underlying buffering capacities.Moreover, it is unclear how these buffering capacities are distributed across developmental stages.Key questions include: is the capacity of buffering systems evenly distributed throughout development?Are there critical periods and limitations of certain buffering systems at specific developmental stages?Integrating all the different steps should be included in a framework for the analyses of robustness (Figure 1).

| Robustness against internal perturbations
Internal perturbations that challenge systems include mutations (or more general changes in the genetic makeup of an individual).While most developmental biology, evolutionary genetics, and evo-devo research has focused on the effects of such mutations on the average means of phenotypic traits, much less attention has been paid to the effects on variation itself.
First, looking at the effects of single mutations or of a mutation in a certain key regulator gene can lead to a significant reduction or increase in the size (or any other characteristic) of an organ.For example, tooth size and shape have been shown to be simultaneously controlled by ectodysplasin-A in different vertebrates (Kangas et al., 2004;Harris et al., 2008;Salomies et al., 2021).But at the same time, an increased phenotypic variation would indicate changes in the buffering capacity of the system in response to this perturbation.We, therefore, postulate the existence of mutations that would only affect variation but not the total mean.With the predominant focus on significance and effect sizes, these developmentally and evolutionary relevant phenotypes might have been systematically overlooked.
Second, features like genomic heterozygosity or the introgression of new alleles (both could happen e.g., F I G U R E 1 A framework to study developmental robustness against internal and external perturbations across organs, species, and levels of biological organization. through hybridization of different populations or species) are known to greatly affect trait variation.For example, it has been shown that hybridization increases morphological variation in Darwin finches (Grant & Grant, 2019).Moreover, in cichlid fishes (Gerwin et al., 2021) and salamanders (Dittrich-Reed & Fitzpatrick, 2013), hybridization leads to variation in color patterns that even go beyond the variation observed in the parental species (i.e., transgressive phenotypes).Yet, in the context of evolutionary-developmental biology, the focus of research has been mainly on differences in means and not on differences in the variation itself.
Two approaches might help us to better investigate these changes in variation (i.e., robustness).First, an increased number of replicates will provide better estimates for variation.The small number of replicates that are often performed to access phenotypic or transcriptomic variation (usually between four and eight in vertebrates) might not suffice to reliably quantify variation itself.Second, we can use fluctuating asymmetry as a measure to estimate variation in robustness.A very robust system would be assumed to have a strong correlation between left and right-side phenotypes, while a less robust system would have a weaker correlation.A practical challenge in measuring fluctuating asymmetry is that it is not an easy task empirically.One needs many replicates and negligible measurement error, especially if the system shows large amounts of experiment-caused variation or a tendency for directional asymmetry.Yet, the mathematical and statistical methods are well developed to analyze asymmetry (e.g., Palmer & Strobeck, 1986;Savriama & Klingenberg, 2011), suggesting that the task is feasible.

| Robustness against external perturbations
One of the most fascinating properties of development is the degree/extent of external perturbations it can withstand without affecting the phenotypic outcome.These external perturbations can include factors such as mechanical forces, nutrition, or temperature.They can also include the combination of external factors that may vary when organs are cultured ex vivo.Although the effects of environmental factors such as temperature (especially in connection to heat shock proteins) on gene expression and signaling pathways have been extensively studied in plants and invertebrates (e.g., [van Bergen et al., 2017]), relatively little is known about these robustness mechanisms in vertebrates (Irvine, 2020).Effects on developmental rates have been for example described in fish (Kavanagh et al., 2010;Kratochwil et al., 2015) and reptiles (Mitchell et al., 2018).However, we lack a comprehensive understanding of how environmental perturbations affect developmental outcomes and how environmental variation causes buffering or cascading effects at the transcriptional and cellular levels.Ecological variables such as temperature are less of a concern for developing organs in "externally buffered" homeothermic animals.This has led to a lack of research on the topic.In fact, only a very limited number of studies have so far tested the effects of temperature on tissues (Wöltgens et al., 1993), and it is for example unclear how patterning would react to variation in temperature.One would generally expect that mammalian systems are less robust in the face of external perturbations that are normally buffered by homeostasis -a testable hypothesis.
Similarly, as for internal perturbations a combination of an increased number of replicates and taking advantage of fluctuating asymmetry will provide new insights into how external factors affect the robustness of development.Lastly, it will be of course interesting to assay if the buffering mechanisms for internal and external perturbations differ and if combinations of both lead to more variation and a greater challenge for the system (suggesting that the buffering pathways are not independent).

| Buffering capacities at different levels of biological complexity
Knowing the intrinsic differences in the robustness of different organs in various species will be the starting point for a better investigation of the mechanisms that underlie the buffering capacities of the respective organ systems.Two approaches might be of interest here.The first one is to compare organs with internal or external perturbations that do not alter the "normal" developmental outcome.In those instances, signatures of this buffering might be detectable at the transcriptomic or cellular level, as the system is actively working to maintain the status quo.The second case would be to identify tipping points after which the buffering capacity of the developmental system is no longer sufficient to maintain the "normal" phenotype.This might be particularly applicable for external perturbations such as temperature.Here, contrasting the development of individuals at a "just tolerable temperature" with those at a "just too high temperature" could be informative.
The most adequate current tool to analyze robustness at transcriptional and cellular levels would likely be single-cell RNA sequencing (scRNA-seq) as it enables the identification of normal (and abnormal) endpoints at a cellular level.The methodology also allows assaying changes in cell composition.However, bulk sequencing, especially when the acquisition of rather homogeneous cell populations is feasible, would be a powerful tool as well.Together these tools can be used to quantify the robustness of developmental systems (Figure 2), which can then be combined with phenotypic measures of developing and final morphology.

| Agential dynamics of developmentally robust systems
A key question in our understanding of robustness is how it scales from individual organs to the whole organism.When agency is considered a property of an integrated whole (Sultan et al., 2022), we can ask whether measuring robustness of individual organs can be used as a proxy for the whole individual.In other words, is it possible to "atomize" agency by studying many individual organ systems separately and only then infer the system as a whole?Even if agency is principally considered a system or individual level property, measuring robustness at the organ level may provide links to the biological mechanisms how the whole system modulates when facing, for example, a change in the environment.
Considering individual organ systems, it can also be debated whether organ development is generally more machine-like, or whether organ development has complex emergent properties that make it impossible to describe system behavior, including robustness, from components alone.In the latter case developing organs can be considered to possess agency in generating their own robustness.Currently we lack a complete understanding of the repertoire of outcomes that a developing organ with emergent properties can have.Although we start to gain increasingly detailed insights into the causative processes that shape system robustness (e.g., enhancer redundancy [Osterwalder et al., 2018]), the Hypothetical examples of readouts testing robustness at the levels of morphology and/or transcriptomes.(a) Mutations may not only increase the mean but also the variation of a phenotype or the expression of a gene.An increased variation would suggest a limited ability of the system to buffer against the perturbation.(b) A further experimental approach to investigate robustness might be the analysis of fluctuating asymmetry.If the perturbation induces variation in a random fashion, we assume an increase in fluctuating asymmetry (in addition to the increased variation).(c) Analysis at the cellular level could reveal changes in cellular trajectories (here visualized by single-cell RNA sequencing) that affect either the endpoints of cellular trajectories and/or the cellular composition of a tissue.(d) Something similar could also be seen using bulk RNA-sequencing, where the transcriptional trajectories of developing tissues might diverge.Such changes could happen even if the phenotype remains unchanged.
generative aspect of robustness in producing phenotypic variation remains to be explored.More integrative approaches that span levels of biological organization from single-cell and spatial transcriptomes to protein activity and cell behavior will be needed.These approaches have to be linked to experimental assays that control (and manipulate) environmental variation to test how the whole system reacts to perturbations.
Moving from development to macroevolutionary diversity, we can ask whether developmentally robust organs are evolutionarily more diverse or versatile.Robustness and plasticity of organ development may also bias or direct evolution.These kinds of examples are known for different systems, such as beetle horns (Rohner et al., 2022) and mammalian teeth (Kavanagh et al., 2007).An interesting avenue of research linked to agency would be to explore whether robustness, by affecting the probability of different evolutionary outcomes, promotes or inhibits the emergence of evolutionary innovations.Operationally, geometric morphometrics and topographic measures provide excellent means for a fast-throughput tabulation of large sets of morphological features (Di-Poï & Milinkovitch, 2016;Morita et al., 2020;Rohner et al., 2022).Morphological analysis with a larger phylogenetic sampling will certainly allow a direct comparison of evolutionary patterns and variational properties that could be ultimately linked to the robustness of the respective organ systems.

| CONCLUSION
Here we state that a comparative framework to study robustness across organ systems and species is a research gap within evo-devo.It remains unclear whether there is a conserved link between robustness against internal and external perturbations.Comparative analysis will be able to test whether the robustness of a developmental system against external perturbations translates also to robustness against internal perturbations and vice versa.We anticipate that highly canalized or less noisy developmental systems are also robust against internal disturbances.A comparative framework will provide benchmarks to measure the effects of external perturbations (environmental robustness).Here, quantifying disturbances from epigenomes to transcriptomes to phenomes might provide an empirical "robustness metric" across species, organs, and levels of biological organization.Conceptually, bridging experimentally produced variation with phenotype-based inferences of robustness, such as fluctuating asymmetry, allows interdisciplinary advances in linking processes to patterns.By building from individual studies on robustness interorgan and interspecies comparisons, and all the way to evolutionary patterns, it will be possible to establish variational properties, or "variational envelopes" that scale from individual experiments to species, and from molecules to organisms.Potential impacts would also include regenerative medicine, as a better mechanistic understanding of the underpinnings of robustness is expected to advance/improve protocols used for organoid differentiation and tissue engineering.Finally, a comparative framework for robustness research should allow empirical appraisals of different conceptual frameworks, such as agency, in explaining biological systems.