Mammalian forelimb evolution is driven by uneven proximal-to-distal morphological diversity

  1. Priscila S Rothier
  2. Anne-Claire Fabre
  3. Julien Clavel
  4. Roger BJ Benson
  5. Anthony Herrel

  • Curated by eLife

    eLife logo

    eLife assessment

    This study reports an interesting analysis of evolutionary variation in forelimb/hand bone shapes in relation to functional and developmental variation along the proximo-distal axis. The authors found expected and compelling patterns of evolutionary shape variation along the proximo-distal axis but less expected, yet equally compelling, patterns of shape integration. This paper will be of interest to researchers working on macroevolutionary patterns and sources of morphological diversity.

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

Vertebrate limb morphology often reflects the environment due to variation in locomotor requirements. However, proximal and distal limb segments may evolve differently from one another, reflecting an anatomical gradient of functional specialization that has been suggested to be impacted by the timing of development. Here, we explore whether the temporal sequence of bone condensation predicts variation in the capacity of evolution to generate morphological diversity in proximal and distal forelimb segments across more than 600 species of mammals. Distal elements not only exhibit greater shape diversity, but also show stronger within-element integration and, on average, faster evolutionary responses than intermediate and upper limb segments. Results are consistent with the hypothesis that late developing distal bones display greater morphological variation than more proximal limb elements. However, the higher integration observed within the autopod deviates from such developmental predictions, suggesting that functional specialization plays an important role in driving within-element covariation. Proximal and distal limb segments also show different macroevolutionary patterns, albeit not showing a perfect proximo-distal gradient. The high disparity of the mammalian autopod, reported here, is consistent with the higher potential of development to generate variation in more distal limb structures, as well as functional specialization of the distal elements.

Article activity feed

  1. Version published to 10.7554/elife.81492 on eLife
  2. eLife

    Author Response

    Reviewer #1 (Public Review):

    1. This study performs an interesting analysis of evolutionary variation and integration in forelimb/hand bone shapes in relation to functional and developmental variation along the proximo-distal axis. They found expected patterns of evolutionary shape variation along the proximo-distal axis but less expected patterns of shape integration. This study provides a strong follow-up to previous studies on mammal forelimb variation, adding and testing interesting hypotheses with an impressive dataset. However, this study could better highlight the relevance of this work beyond mammalian forelimbs. The study primarily cites and discusses mammalian limb studies, despite the relevance of the suggested findings beyond mammals and forelimbs. Furthermore, relevant work exists in other tetrapod clades and structures related to later-developing traits and proximo-distal variation. Finally, variations in bone size and shape along the proximo-distal axis could be affecting evolutionary patterns found here and it would be great to make sure they are not influencing the analysis/results.

    We appreciate the reviewer’s comments, and we acknowledge the importance of including examples of non-mammalian lineages in our study. We attended to the recommendation and included more examples of other tetrapod taxa in our text and in our references, providing a more inclusive discussion of limb bone diversity beyond mammals. We also explain below why the results obtained are not inflated by variation of bigger versus smaller sizes of bones.

    Reviewer #2 (Public Review):

    1. Congratulations on producing a very nice study. Your study aims to examine the morphological diversity of different mammalian limb elements, with the ultimate goal seemingly to test expectations based on the different timing of development of the limb bones. There's a lot to like: the sample size is impressive, the methods seem appropriate and sound, the results are interesting, the figures are clear, and the paper is very well written. You find greater diversity and integration in distal limb segments compared to proximal elements, and this may be due to the developmental timing and/or functional specialization of the limb segments. These are interesting results and conclusions that will be of interest to a broad readership. And the large dataset will likely be valuable to future researchers who are interested in mammalian limb morphology and evolution. I have one major concern with how you frame your discussion and conclusions, which I explain below. But I think you can address this issue with some text edits.

    We sincerely thank the reviewer for his constructive recommendations and for his appreciation of our work. We addressed the issue raised as detailed below.

    1. Major concern - is developmental timing the best hypothesis?

    You discuss two potential drivers for the relatively greater diversity in distal elements: 1) later development and 2) greater functional specialization. Your data doesn't allow you to fully test these two hypotheses (e.g. you don't have detailed evo-devo data to infer developmental constraints), and I think you realize this - you use phrases like "consistent with the hypothesis that ...". You seem to compromise and conclude that both factors (development + function) are likely driving greater autopod diversity (e.g. Lines 302-306). Being unable to fully test these hypotheses weakens the impact of your conclusions, making them a bit more speculative, but otherwise, it isn't a critical issue.

    But my concern is that you seem to favor developmental factors over functional factors as the primary drivers of your results, and that seems backwards to me. For instance, early in the Abstract (Line 32) and early in the Discussion (Line 201) you mention that your results are consistent with the developmental timing hypothesis, but it's not until later in the Abstract or Discussion that you mention the role of functional diversity/specialization/selection. The problem with favoring the development hypothesis is that your integration results seem to contradict that hypothesis, at least based on your prediction in the Introduction (Line 126; although you spend some of the Discussion trying to make them compatible). Later in the paper, you acknowledge that functional specialization (rather than developmental factors) might be a better explanation for the integration results (Lines 282-284, 345-347), but, again, this is only after discussions about developmental factors.

    When you first start discussing functional diversity, you say, "high integration in the phalanx and metacarpus, possibly favoured the evolution of functionally specialized autopod structures, contributing to the high variation observed in mammalian hand bones." (Line 282). This implies that integration led to functional diversity in the autopod. But I'd flip that: I think the functional specialization of the hand led to greater integration. Integration does not result solely from genetic/developmental factors. It can also result from traits evolving together because they are linked to the same function. From Zelditch & Goswami (2021, Evol. & Dev.): "Within individuals, integration is customarily ascribed to developmental and/or functional interdependencies among traits (Bissell & Diggle, 2010; Cheverud, 1982; Wagner, 1996) and modularity is thus due to their developmental and/or functional independence."

    In sum, I think your results capture evidence of greater functional specialization in hands relative to other segments. You're seeing greater 1) disparity and 2) integration in hands, and both of those are expected outcomes of greater functional specialization. In contrast, I think it's harder to fit your results to the developmental timing hypothesis. Thus, I recommend that throughout the paper (Abstract, Intro, Discussion) you flip your discussion of the two hypotheses and start with a discussion on how functional specialization is likely driving your results, and then you can also note that some results are consistent with the development hypothesis. You could maintain most of your current text, but I'd simply rearrange it, and maybe add more discussion on functional diversity to the Intro.

    Or, if you disagree and think that there's more support for the development hypothesis, then you need to make a better case for it in the paper. Right now, it feels like you're trying to force a conclusion about development without much evidence to back it up.

    We thank the reviewer for his thoughtful and thorough comment. We agree that the results provided, particularly those of integration, support the hypothesis that functional specialization contributes to the uneven diversity of limb bones. We addressed the concerns by substantially changing our discussion, particularly moderating (and removing) sections on the developmental constraints and adding new arguments for other possible drivers for the diversity of limb bones, such as function. However, the goal of the paper was to test whether the data corroborate - or not - the predictions derived from the developmental hypothesis, and they largely do. Therefore, we decided to keep the developmental hypothesis presented first in the introduction and in the discussion section, as we believe this sequence provides more coherence considering the hypothesis tested (we believe that detailing the role of functional specialization particularly in the introduction would mislead the reader to think that we directly tested for these parameters). Following the discussion of the integration results, we then go on to discuss the possible role of functional specialization on the results obtained (lines 262-285, see also lines 216-234). Yet, these are not tested in this paper and remain to be tested in a future analysis focusing specifically on the role of ecology and function in driving variation in the mammalian limb.

    1. Limitations of the dataset

    Using linear measurements is fine, but they mainly just capture simple aspects of the elements (lengths and widths). You should acknowledge in your paper the limitations of that type of data. For example, the deltoid tuberosity of the humerus can vary considerably in size and shape among mammals, but you don’t measure that structure. The autopod elements don’t have a comparable process, meaning that if you were to measure the deltoid tuberosity then you’d likely see a relative increase in humerus disparity (although my guess is that it’d still be well below that of the autopod). And you omit the ulna from your study, and its olecranon process varies considerably among taxa and its length is a very strong correlate of locomotor mode. In other words, your finding of the greatest disparity in the hand might be due in part to your choice of measurements and the omission of measurements of specific processes/elements. I recommend that you add to your paper a brief discussion of the limitations of using linear measurements and how you might expect the results to change if you were to include more detailed measurements and/or more elements.

    We followed the recommendation and included a discussion about the dataset limitations, acknowledging for the possible impact of the measurements and the bones chosen in the results obtained (Lines 235-260).

    Reviewer #3 (Public Review):

    1. This paper uses a large (638 species representing 598 genera in 138 families) extant sample of osteologically adult mammals to address the question of proximodistal patterns of cross-taxonomic diversity in forelimb bony elements. The paper concludes, based on a solid phylogenetically controlled multivariate analysis of liner measurements, that proximal forelimb elements are less morphologically diverse and evolutionarily flexible than distal forelimb elements, which the paper concludes is consistent with a developmental constraint axis tied to limb bud growth and development. This paper is of interest to researchers working on macroevolutionary patterns and sources of morphological diversity.

    Methodological review Strengths:

    The taxonomic dataset is very comprehensive for this sort of study and the authors have given consideration to how to identify bony elements present in all mammalian taxa (no small task with this level of taxonomic breadth). Multivariate approaches as used in this study are the gold standard for addressing questions of morphological variations.

    The authors give consideration to two significant confounders of analyses operating at this scale: phylogeny and body size. The methods they use to address these are appropriate, although as I note below body size itself may merit more consideration.

    We sincerely thank the reviewer for his appreciation of our study. We addressed the main concerns pointed out below.

    Weaknesses:

    1. The authors assume a lot of knowledge on the part of the reader regarding their methods. Given that one of their key metrics (stationary variance) is largely a property as I understand it of OU models, more explanation on the authors' biological interpretation of stationary variance would help assess the strength of their conclusions, especially as OU models are not as straightforward as they first appear in their biological interpretation (Cooper et al., 2016).

    We acknowledge that this may not be straightforward and now include a more extensive explanation of the approach and the metrics used. We detailed the explanation about the stationary variances in the methods, contextualizing the biological meaning (lines 456-469).

    1. It is unclear what the authors mean when they say they "simulated the trait evolution under OU processes on 100 datasets". Are the 100 datasets 100 different tree topologies (as seems to be the case later "we replicated the body mass linear regressions with 100 trees from Upham et al (2019)." If that is so, what is the rationale for choosing 100 topologies and what criteria were used to select the 100 topologies?

    We understand the explanation may have been confusing. Globally, we used a parametric bootstrap approach to assess the uncertainty around point estimates for morphological diversity and integration. That is, we first simulated 100 datasets on the maximum clade credibility tree (MCC tree, that summarizes 10,000 trees from Upham et al. 2019) – using the best fit model on our original data (i.e., an OU process) with parameters estimates from this model fit. The model (an OU process) was then fit to these 100 simulated traits, and the distribution of parameters estimates obtained was used to assess the variability around the point estimate (for the determinant, the trace, and the measure of integration) obtained on empirical data. We did not used the simulated dataset to estimate the significance of the stationary variances. We fitted the empirical datasets with 100 trees randomly sampled from the credible set of 10,00 trees of Upham et al (2019) – instead of using the MCC – to further assess the variability due to the tree topology and branching times uncertainties. We included this expanded explanation in the methods in lines 421-428 and 471.

    1. The way the authors approach body mass and allometry, while mathematically correct, ignores the potential contribution of body mass to the questions the authors are interested in. Jenkins (1974) for example argued that small mammals would converge on similar body posture and functional morphology because, at small sizes, all mammals are scansorial if they are not volant. Similarly, Biewener (1989) argued that many traits we view as cursorial adaptations are actually necessary for stability at large body sizes. Thus size may actually be important in determining patterns of variation in limb bone morphology.

    We agree with the observation. We believe that categorizing the groups according to size would provide a meaningful overview on the effect of size on the diversity and evolution of limb bones. Although insightful and worthy of investigation, we were particularly interested in understanding whether developmental timing corresponds to bone diversification more broadly across Mammalia and thus considered only the size residual values. This issue will be addressed in our future works. We discussed in the lines 329-341 the potential contribution of body size to limb segment diversification and the importance of considering this aspect in future studies.

    1. Review of interpretation.

    The authors conclude that their result, in showing a proximo-distal gradient of increasing disparity and stationary variance in forelimb bone morphology, supports the idea that proximo-distal patterning of limb bone development constrains the range of morphological diversity of the proximal limb elements. However, this correlation ignores two important considerations. The first is that the stylopod connects to the pectoral girdle and the axial skeleton, and so is feasibly more constrained functionally, not developmentally in its morphological evolution. The second, related, issue arises from the authors' study itself, which shows that the lowest morphological integration is found in the stylopod and zeugopod, whereas the autopod elements are highly integrated. This suggests a greater tendency towards modularity in the stylopod and zeugopod, which is itself a measure of evolutionary lability (Klingenberg, 2008). And indeed the mammalian stylopod is developmentally comprised of multiple elements (the epiphyses and diaphysis) that are responding to very different developmental and biomechanical signals. Thus, for example, the functional signal in stylopod (Gould, 2016) and zeugopod (MacLeod and Rose, 1993) articular surface specifically is very high. What is missing to fully resolve the question posed by the authors is developmental data indicating whether or not the degree of morphological disparity in the hard tissues of the forelimb change over the course of ontogeny throughout the mammalian tree, and whether changing functional constraints over ontogeny (as is the case in marsupials) affect these patterns.

    We thank the reviewer for sharing such an interesting reinterpretation of the results. Combined to the recommendations from the other two reviewers, we substantially changed our discussion, specially modifying the interpretation of results concerning trait integration. We discussed the possible role of the functional variation at the articulations on element integration in lines 263-285.

  3. eLife

    eLife assessment

    This study reports an interesting analysis of evolutionary variation in forelimb/hand bone shapes in relation to functional and developmental variation along the proximo-distal axis. The authors found expected and compelling patterns of evolutionary shape variation along the proximo-distal axis but less expected, yet equally compelling, patterns of shape integration. This paper will be of interest to researchers working on macroevolutionary patterns and sources of morphological diversity.

  4. eLife

    Reviewer #1 (Public Review):

    This study performs an interesting analysis of evolutionary variation and integration in forelimb/hand bone shapes in relation to functional and developmental variation along the proximo-distal axis. They found expected patterns of evolutionary shape variation along the proximo-distal axis but less expected patterns of shape integration. This study provides a strong follow-up to previous studies on mammal forelimb variation, adding and testing interesting hypotheses with an impressive dataset. However, this study could better highlight the relevance of this work beyond mammalian forelimbs. The study primarily cites and discusses mammalian limb studies, despite the relevance of the suggested findings beyond mammals and forelimbs. Furthermore, relevant work exists in other tetrapod clades and structures related to later-developing traits and proximo-distal variation. Finally, variations in bone size and shape along the proximo-distal axis could be affecting evolutionary patterns found here and it would be great to make sure they are not influencing the analysis/results.

  5. eLife

    Reviewer #2 (Public Review):

    Congratulations on producing a very nice study. Your study aims to examine the morphological diversity of different mammalian limb elements, with the ultimate goal seemingly to test expectations based on the different timing of development of the limb bones. There's a lot to like: the sample size is impressive, the methods seem appropriate and sound, the results are interesting, the figures are clear, and the paper is very well written. You find greater diversity and integration in distal limb segments compared to proximal elements, and this may be due to the developmental timing and/or functional specialization of the limb segments. These are interesting results and conclusions that will be of interest to a broad readership. And the large dataset will likely be valuable to future researchers who are interested in mammalian limb morphology and evolution. I have one major concern with how you frame your discussion and conclusions, which I explain below. But I think you can address this issue with some text edits.

    Major concern - is developmental timing the best hypothesis?

    You discuss two potential drivers for the relatively greater diversity in distal elements: 1) later development and 2) greater functional specialization. Your data doesn't allow you to fully test these two hypotheses (e.g. you don't have detailed evo-devo data to infer developmental constraints), and I think you realize this - you use phrases like "consistent with the hypothesis that ...". You seem to compromise and conclude that both factors (development + function) are likely driving greater autopod diversity (e.g. Lines 302-306). Being unable to fully test these hypotheses weakens the impact of your conclusions, making them a bit more speculative, but otherwise, it isn't a critical issue.

    But my concern is that you seem to favor developmental factors over functional factors as the primary drivers of your results, and that seems backwards to me. For instance, early in the Abstract (Line 32) and early in the Discussion (Line 201) you mention that your results are consistent with the developmental timing hypothesis, but it's not until later in the Abstract or Discussion that you mention the role of functional diversity/specialization/selection. The problem with favoring the development hypothesis is that your integration results seem to contradict that hypothesis, at least based on your prediction in the Introduction (Line 126; although you spend some of the Discussion trying to make them compatible). Later in the paper, you acknowledge that functional specialization (rather than developmental factors) might be a better explanation for the integration results (Lines 282-284, 345-347), but, again, this is only after discussions about developmental factors.

    When you first start discussing functional diversity, you say, "high integration in the phalanx and metacarpus, possibly favoured the evolution of functionally specialized autopod structures, contributing to the high variation observed in mammalian hand bones." (Line 282). This implies that integration led to functional diversity in the autopod. But I'd flip that: I think the functional specialization of the hand led to greater integration. Integration does not result solely from genetic/developmental factors. It can also result from traits evolving together because they are linked to the same function. From Zelditch & Goswami (2021, Evol. & Dev.): "Within individuals, integration is customarily ascribed to developmental and/or functional interdependencies among traits (Bissell & Diggle, 2010; Cheverud, 1982; Wagner, 1996) and modularity is thus due to their developmental and/or functional independence."

    In sum, I think your results capture evidence of greater functional specialization in hands relative to other segments. You're seeing greater 1) disparity and 2) integration in hands, and both of those are expected outcomes of greater functional specialization. In contrast, I think it's harder to fit your results to the developmental timing hypothesis. Thus, I recommend that throughout the paper (Abstract, Intro, Discussion) you flip your discussion of the two hypotheses and start with a discussion on how functional specialization is likely driving your results, and then you can also note that some results are consistent with the development hypothesis. You could maintain most of your current text, but I'd simply rearrange it, and maybe add more discussion on functional diversity to the Intro.

    Or, if you disagree and think that there's more support for the development hypothesis, then you need to make a better case for it in the paper. Right now, it feels like you're trying to force a conclusion about development without much evidence to back it up.

    Limitations of the dataset

    Using linear measurements is fine, but they mainly just capture simple aspects of the elements (lengths and widths). You should acknowledge in your paper the limitations of that type of data. For example, the deltoid tuberosity of the humerus can vary considerably in size and shape among mammals, but you don't measure that structure. The autopod elements don't have a comparable process, meaning that if you were to measure the deltoid tuberosity then you'd likely see a relative increase in humerus disparity (although my guess is that it'd still be well below that of the autopod). And you omit the ulna from your study, and its olecranon process varies considerably among taxa and its length is a very strong correlate of locomotor mode. In other words, your finding of the greatest disparity in the hand might be due in part to your choice of measurements and the omission of measurements of specific processes/elements. I recommend that you add to your paper a brief discussion of the limitations of using linear measurements and how you might expect the results to change if you were to include more detailed measurements and/or more elements.

  6. eLife

    Reviewer #3 (Public Review):

    This paper uses a large (638 species representing 598 genera in 138 families) extant sample of osteologically adult mammals to address the question of proximodistal patterns of cross-taxonomic diversity in forelimb bony elements. The paper concludes, based on a solid phylogenetically controlled multivariate analysis of liner measurements, that proximal forelimb elements are less morphologically diverse and evolutionarily flexible than distal forelimb elements, which the paper concludes is consistent with a developmental constraint axis tied to limb bud growth and development. This paper is of interest to researchers working on macroevolutionary patterns and sources of morphological diversity.

    Methodological review

    Strengths:

    The taxonomic dataset is very comprehensive for this sort of study and the authors have given consideration to how to identify bony elements present in all mammalian taxa (no small task with this level of taxonomic breadth). Multivariate approaches as used in this study are the gold standard for addressing questions of morphological variations.
    The authors give consideration to two significant confounders of analyses operating at this scale: phylogeny and body size. The methods they use to address these are appropriate, although as I note below body size itself may merit more consideration.

    Weaknesses:

    The authors assume a lot of knowledge on the part of the reader regarding their methods. Given that one of their key metrics (stationary variance) is largely a property as I understand it of OU models, more explanation on the authors' biological interpretation of stationary variance would help assess the strength of their conclusions, especially as OU models are not as straightforward as they first appear in their biological interpretation (Cooper et al., 2016).
    It is unclear what the authors mean when they say they "simulated the trait evolution under OU processes on 100 datasets". Are the 100 datasets 100 different tree topologies (as seems to be the case later "we replicated the body mass linear regressions with 100 trees from Upham et al (2019)." If that is so, what is the rationale for choosing 100 topologies and what criteria were used to select the 100 topologies?
    The way the authors approach body mass and allometry, while mathematically correct, ignores the potential contribution of body mass to the questions the authors are interested in. Jenkins (1974) for example argued that small mammals would converge on similar body posture and functional morphology because, at small sizes, all mammals are scansorial if they are not volant. Similarly, Biewener (1989) argued that many traits we view as cursorial adaptations are actually necessary for stability at large body sizes. Thus size may actually be important in determining patterns of variation in limb bone morphology.

    Review of interpretation.

    The authors conclude that their result, in showing a proximo-distal gradient of increasing disparity and stationary variance in forelimb bone morphology, supports the idea that proximo-distal patterning of limb bone development constrains the range of morphological diversity of the proximal limb elements. However, this correlation ignores two important considerations. The first is that the stylopod connects to the pectoral girdle and the axial skeleton, and so is feasibly more constrained functionally, not developmentally in its morphological evolution. The second, related, issue arises from the authors' study itself, which shows that the lowest morphological integration is found in the stylopod and zeugopod, whereas the autopod elements are highly integrated. This suggests a greater tendency towards modularity in the stylopod and zeugopod, which is itself a measure of evolutionary lability (Klingenberg, 2008). And indeed the mammalian stylopod is developmentally comprised of multiple elements (the epiphyses and diaphysis) that are responding to very different developmental and biomechanical signals. Thus, for example, the functional signal in stylopod (Gould, 2016) and zeugopod (MacLeod and Rose, 1993) articular surface specifically is very high. What is missing to fully resolve the question posed by the authors is developmental data indicating whether or not the degree of morphological disparity in the hard tissues of the forelimb change over the course of ontogeny throughout the mammalian tree, and whether changing functional constraints over ontogeny (as is the case in marsupials) affect these patterns.

    References

    Biewener, A. A. (1989). Scaling body support in mammals: limb posture and muscle mechanics. Science, 245(4913), 45-48.
    Cooper, N., Thomas, G.H. and FitzJohn, R.G. (2016), Shedding light on the 'dark side' of phylogenetic comparative methods. Methods Ecol Evol, 7: 693-699. https://doi.org/10.1111/2041-210X.12533
    Gould, F.D.H. (2107), Testing the Role of Cursorial Specializations as Adaptive Key Innovations in Paleocene-Eocene Ungulates of North America. J Mammal Evol 24, 453-463. https://doi.org/10.1007/s10914-016-9359-4
    Jenkins, F. A. (1974). Tree shrew locomotion and the origins of primate arborealism. In F. A. Jenkins (Ed.), Primate locomotion. New York: Academic Press.
    Klingenberg, C. P. (2008). Morphological Integration and Developmental Modularity. Annual Review of Ecology, Evolution, and Systematics, 39, 115-132. http://www.jstor.org/stable/30245156
    MacLeod, N., & Rose, K. D. (1993). Inferring locomotor behavior in Paleogene mammals via eigenshape analysis. Am J Sci, 293(A), 300-355.

  7. Version published to 10.1101/2022.07.26.501575v1 on bioRxiv