How can morphology show that organisms are related




















For example, insects use wings to fly like bats and birds, but the wing structure and embryonic origin is completely different. These are called analogous structures. Analogous structures : The c wing of a honeybee is similar in shape to a b bird wing and a bat wing, and it serves the same function. However, the honeybee wing is not composed of bones and has a distinctly-different structure and embryonic origin. These wing types insect versus bat and bird illustrate an analogy: similar structures that do not share an evolutionary history.

Similar traits can be either homologous or analogous. Homologous structures share a similar embryonic origin; analogous organs have a similar function. For example, the bones in the front flipper of a whale are homologous to the bones in the human arm. These structures are not analogous.

The wings of a butterfly and the wings of a bird are analogous, but not homologous. Some structures are both analogous and homologous: the wings of a bird and the wings of a bat are both homologous and analogous.

Scientists must determine which type of similarity a feature exhibits to decipher the phylogeny of the organisms being studied. With the advancement of DNA technology, the area of molecular systematics, which describes the use of information on the molecular level including DNA analysis, has blossomed. New computer programs not only confirm many earlier classified organisms, but also uncover previously-made errors.

As with physical characteristics, even the DNA sequence can be tricky to read in some cases. For some situations, two very closely-related organisms can appear unrelated if a mutation occurred that caused a shift in the genetic code.

An insertion or deletion mutation would move each nucleotide base over one place, causing two similar codes to appear unrelated. Sometimes two segments of DNA code in distantly-related organisms randomly share a high percentage of bases in the same locations, causing these organisms to appear closely related when they are not.

For both of these situations, computer technologies have been developed to help identify the actual relationships. Ultimately, the coupled use of both morphologic and molecular information is more effective in determining phylogeny.

A phylogenetic tree sorts organisms into clades or groups of organisms that descended from a single ancestor using maximum parsimony. After the homologous and analogous traits are sorted, scientists often organize the homologous traits using a system called cladistics.

This system sorts organisms into clades: groups of organisms that descended from a single ancestor. For example, all of the organisms in the orange region evolved from a single ancestor that had amniotic eggs.

Consequently, all of these organisms also have amniotic eggs and make a single clade, also called a monophyletic group. Clades must include all of the descendants from a branch point.

Common ancestors : Lizards, rabbits, and humans all descend from a common ancestor that had an amniotic egg. Thus, lizards, rabbits, and humans all belong to the clade Amniota. Vertebrata is a larger clade that also includes fish and lamprey. Although phylogenetic inference is a rich and complex field, it is based on the simple idea that as long as traits or characters change their state reasonably rarely compared to the rate of lineage branching , then the distribution of traits among species provides evidence of how recently these species last shared a common ancestor.

One profound implication of this way of thinking about trait evolution is the idea that living species are the summation of their evolutionary history. In other words, if you list all the traits of a living species, you can assume that each trait arose on an ancestral branch somewhere in that species' history.

While some traits, such as the presence of a distinct cellular nucleus , evolved in the ancient past and are shared by many organisms, others, like spoken language, arose much more recently in this case, very recently, as speech is a uniquely human trait.

Thus, understanding the evolutionary history of living species amounts to understanding where on the tree of life these species' distinctive traits arose. Moreover, phylogenetic trees serve as extremely powerful tools for organizing this knowledge of biological diversity. In biology, the concept of relatedness is defined in terms of recency to a common ancestor.

As a result, the question "Is species A more closely related to species B or to species C? To help clarify this logic, think about the relationships within human families. The most recent common ancestors of both you and your siblings are your parents; the most recent common ancestors of you and your first cousins are your grandparents; and the most recent common ancestors of you and your second cousins are your great-grandparents.

Note that your parents are situated one generation ago, your grandparents are situated two generations ago, and your great-grandparents are situated three generations ago. This arrangement of ancestors explains why you are more closely related to your siblings than your cousins, and why you are more closely related to your first cousins than your second cousins.

Because evolutionary trees depict common ancestry, they also contain information on the degree of relatedness of the terminal nodes. For example, Figure 4 can be used to determine whether a lizard is more closely related to a salamander or to a human. The first step in answering this question is to trace downward from the appropriate tips of the tree to find the most recent common ancestor of lizards and salamanders marked x in the diagram.

The next step is to use a similar process to locate the most recent common ancestor of lizards and humans marked y in the diagram. Once you have done this, you can clearly see that node y is a descendant of node x , as indicated by the fact that you must pass through node x to get from the root of the tree to node y. This tells you that lizards are more closely related to humans than they are to salamanders. Many people are surprised by this conclusion regarding lizards, humans, and salamanders.

This conclusion makes sense, however, because there is a lineage specifically, the internal branch between x and y that is ancestral to both lizards and humans but not to salamanders. Any trait that evolved on this lineage will therefore tend to occur in lizards and humans but not salamanders.

One such trait is the amnion, a protective layer that surrounds the embryo and originally served to allow eggs to mature away from water. When it comes to this conclusion, one frequent source of confusion is the fact that salamanders and lizards look somewhat alike. However, while it is true that closely related organisms often have similar appearances, this is not always the case. The primary reason for this observation is that morphological evolution can occur at different rates on different branches of a phylogeny.

In the example from Figure 4, the similarities of salamanders and lizards e. Among the land vertebrates, these are ancestral traits, or plesiomorphies.

While the lizard and salamander lineages both retained these ancestral traits, the mammalian lineage underwent dramatic evolutionary divergence , evolving an erect gait, fur, mammary glands, and many other features. These evolutionarily derived features, or apomorphies , are shared by all mammals but are not found in other living vertebrates. Nonetheless, the presence of these uniquely mammalian features does not change the fact that lizards are more closely related to mammals than to salamanders.

Remember, relatedness is about descent, not similarity. This should not be surprising. After all, you would still be more closely related to your first cousin than to your second cousin even if you happened to look more similar to the latter!

Since the time of the ancient Greeks, the prevailing guide to thinking about nature was the so-called "ladder of life," also known as scala naturae. This idea imagines that living species represent various degrees of perfection, with humans as the "most perfect" species and therefore at the top of the ladder see, for example, Figure 5.

Although the ladder of life idea was central to the evolutionary theories advanced by many of his predecessors, Charles Darwin largely rejected the ladder concept in favor of a tree; indeed, the only figure in Darwin's Origin of Species was a tree, reproduced here as Figure 6.

Darwin's view of the evolutionary tree is beautifully expressed in the following quote:. The affinities of all the beings of the same class have sometimes been represented by a great tree. I believe this simile largely speaks the truth The green and budding twigs may represent existing species; and those produced during former years may represent the long succession of extinct species.

Darwin — In the years following Darwin's work, biologists formally rejected the ladder of life in favor of the tree concept. Nonetheless, many current discussions of evolution, especially in the general media, retain vestiges of the ladder view. This is often betrayed by the perception that some organisms are "advanced," whereas others are "primitive. Why is this the case?

For one, "ladder thinking" leads to statements that incorrectly imply that one living species or group is ancestral to another; examples of such statements include "tetrapods land vertebrates evolved from fish " or "humans evolved from monkeys.

And is the howler monkey really an ancestor of you or any other human? Clearly, the answer to both of these questions is no. Why, then, does a statement like "tetrapods evolved from fish" seem reasonable at first glance? However, this is not strictly true, because while the last common ancestor of both clades may have had more obvious physical similarities to living fish than to living tetrapods, it was not identical to any living organism fishlike or otherwise.

Both lineages—the one leading to living fish e. Over this period, all aspects of fish physiology and the fish genome have changed, though perhaps in ways that are not obvious to the human eye. Thus, it is not accurate to say that the common ancestor of both fish and tetrapods was a fish. The best you could do would be to say that the common ancestor had a body form and ecology that were more similar to that of living fish than to that of living tetrapods.

Another problem with ladder thinking is that even with such clarifications, it is still easy to make errors of reasoning. For example, suppose you are told that goldfish have body outgrowths in this case, fins with cartilaginous structures called rays. This is an open-access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors have declared that no competing interests exist. Whether such a tendency exists is a long standing open question [2] — [6]. While it seems evident that more complex organisms exist today than at the advent of life, simple single-celled organisms continue to persist in large numbers, so it is clear that evolution does not guarantee complexity must increase.

Moreover, loss of complexity has been observed in many species [7] — [9]. This begs the question: under what circumstances will complexity increase or decrease over evolutionary time? It is likely that particular environmental conditions are more likely to select for increased complexity than others, especially if this complexity comes at a cost.

As argued by proponents of embodied cognition, intelligent behavior emerges from the interplay between an organism's nervous system, morphology, and environment [10] — [14]. Therefore, if the ecological niche of a species remains constant and its body plan is evolutionarily constrained, then the neural system must adapt in order to succeed under this particular set of circumstances.

This may be investigated experimentally through the use of evolving robots [15] , [16] which stand in for biological organisms. For instance, it has been demonstrated [11] , [17] that the complexity of an evolved neural system depends on the particular morphology it is controlling: in a given task environment certain morphologies can readily succeed with simple neural systems, while other morphologies require the discovery of more complex neural systems, or may prevent success altogether.

This can be studied by observing how organisms evolve in different environments. For instance, Passy [18] demonstrated that the morphological complexity of benthic colonial diatoms measured as their fractal dimension is significantly correlated with the variability of the environmental niches in which they are found. However, the biological evidence for a correlation between environmental and morphological complexity is sparse.

This is in part because it is difficult to isolate systems where this may be studied effectively and to develop metrics that quantify morphological and environmental complexity. Ideally, it would be desirable to perform controlled investigations in which environmental complexity is under experimental control.

Given enough time and resources it may be possible to carry out these investigations directly on living organisms. However, by performing experiments in silico , it is possible to do so with much greater speed and more precise control over experimental conditions. Specifically, by evolving virtual organisms [19] in physically realistic simulations, it is possible to faithfully model the relevant interactions between organisms and their environments.

Previously, the evolution of complexity has been investigated in silico using an alternative computational model [20]. In that work, populations of computer programs competed among themselves for the energy required to execute their instructions and gained energy by executing specific logic functions.

With their system, Lenski et al. However, in that system the programs did not have bodies with which to physically interact with their environment. On the contrary, the evolutionary model employed here evolves embodied virtual organisms with evolutionarily determined body plans in physically realistic simulation environments. This provides a testbed for investigating how environment may influence the complexity of evolving physical morphologies. Using in silico evolution to act on both the morphologies and nervous systems of simulated organisms or robots was first demonstrated by Sims [19] , and has since been followed by a number of other studies e.

These studies employed a variety of experimental techniques, including different genetic encodings, morphological systems such as branching structures or cellular aggregations , and evolutionary models. However, by constructing morphologies out of a relatively small number of geometric primitives, all of these studies were severely limited in the complexity of the morphologies which they could evolve, and therefore do not offer good test beds for investigating morphological complexity.

Recently, we introduced a new method for evolving virtual organisms that is capable of producing a greater diversity of morphologies than previous systems [33]. By using it to evolve organisms with restricted nervous systems in a variety of environments it was possible to demonstrate how such a system could be used for investigating the relationship between environmental and morphological complexity.

Here, the results of [33] are refined and extended to demonstrate that selection for locomotion tends to induce selection pressures favoring more complex morphologies than would be expected solely due to random chance, and is therefore a driven rather than passive trend [3] , [6] , [34].

In subsequent experiments we employ a multi-objective selection mechanism to select for simplicity in addition to behavioral competency. This selection mechanism filters out morphological complexity that arises due to biases in the underlying evolutionary model or because of genetic drift, and only allows for complexity that confers a selective advantage on the simulated organism. Moreover, this selection mechanism acts to impose a cost on complexity as is thought to occur in biological organisms [35] , [36].

Under this regime complex environments tend to induce selection for greater morphological complexity when compared to a simpler environment. This result supports the hypothesis that the environment plays an active role in determining morphological complexity.

In this work organisms are evolved in a variety of simulated environments in order to better understand the role of the environment in shaping morphological complexity. While inspired by the above mentioned studies in which the morphologies and controllers of virtual organisms were also evolved [19] , [21] — [32] , the system presented here has several advantages which make it better suited for studying the evolution of morphological complexity.

The first advantage relates to the task environments within which organisms evolve. The majority of the studies mentioned above were restricted to evolving for locomotion over flat terrain. While investigating this task has yielded interesting results, it suffers from its simplicity: simple morphologies composed of just a few cuboids or spheres are all that are needed to be successful. Even when more challenging task environments have been explored e. In the current work, a variety of task environments with interesting properties are investigated, and morphologies with greater geometric detail are used, so it is possible to study the evolution of morphological complexity.

Another advantage of the current system is the way in which the genetic material that the evolutionary model acts on is encoded. As has been demonstrated in the past [25] , [26] , genetic encodings that simulate development to some extent offer demonstrable benefits over those that do not.

This is because such encodings tend to produce regularities and symmetries in the phenotype; such patterns in nature are the inevitable result of biological development, which biases the kinds of phenotypes that biological evolution may act on [38]. For this reason, here we employ a particular form of genetic encoding that produces three-dimensional shapes with regular patterns see Methods for more details [39].

Each genome generated from this encoding generates a triangular mesh trimesh that forms the body plan of the virtual organism. Trimeshes allow evolution to craft morphologies with greater geometric detail compared to other systems in which evolution composes a small number of simple three-dimensional shapes together [19] , [21] — [32] see Figs.

Finally, populations of these genetic encodings are evolved with a commonly-used evolutionary model which has been demonstrated to be more evolvable than other evolutionary models [40].

The control environment top left and three icy environments are shown with organisms that evolved to successfully move in each. The five morphologies with smallest top , values from left to right: 0. The morphologies with high values are visually more complex than those with small values. The behavior of each virtual organism is simulated in a three-dimensional, physically-realistic virtual environment in order to assess its fitness.

Because of the organisms' triangular mesh body plans and the complex environments in which they are evolved, evaluating the fitness of each organism requires considerable time. Moreover, many evolutionary trials were conducted in each of several environments to allow for meaningful statistical analysis.

For these reasons all of the experiments were carried out on a 7. In order to study the relationship between the morphological complexity of the virtual organisms and the task environments within which they evolve, evolutionary trials are conducted in each of 50 different environments. The first environment in which organisms are evolved is composed only of a uniform, flat, high friction ground surface refer to Fig. The organisms evolved in this simple environment are considered control cases to compare against organisms evolved in other environments.

Subsequent environments are more complex: they all consist of an infinite series of low friction rectangular solids over which an organism must locomote see below for a characterization of this complexity. This requires the evolution of morphologies with appropriate physical forms. The icy environments vary according to two parameters: the height of the blocks and the spacing between them.

Each of these parameters varies from 0. These two parameters and the their exponential scaling are employed in order to produce a variety of qualitatively different environments that roughly approximate natural surfaces, but yet are also amenable to analysis and efficient simulation.

There are certainly many ways in which the environments could be created to more closely approximate natural terrain, and there are many other factors which could influence the complexity of an environment, however the parameterization employed here provides a set of environments within which it is largely possible to evolve organisms capable of successful locomotion with the bare minimum of neural complexity.

This allows for isolating the influence of environment on morphological complexity, which is the property of interest in this study see Conclusions for further discussion. For each icy environment, evolutionary trials are conducted in that environment and a corresponding evolutionary trials are conducted in the control environment for a total of evolutionary trials; see Methods for details.

This figure demonstrates that there is a clear relationship between the environmental parameters and the difficulty of the task. Specifically, moving to the lower right in Fig. Keeping the spacing constant and decreasing the block height moving left in Fig.

Once the height has been reduced to 0. This plot shows the mean distance achieved by the final generation champion taken across the independent trials of CPPN-NEAT in each of the 49 icy environments. For comparison, the mean distance achieved across all independent trials in the control environment is 7. As the spacing between the blocks is reduced moving upward in Fig. The height of the blocks loses importance in this part of the parameter space but still has an effect though opposite to when the spacing is large.

Here the general pattern is for taller blocks to make the task easier, because taller blocks provide more voluminous gaps which more easily support a variety of ways to gain purchase. Finally in the top row of Fig. For a better understanding of how the evolved organisms behave in each of these environments it is helpful to observe their behavior. It is clear that different environments in this parameterization present the evolutionary system with varying degrees of difficulty, but the question now becomes: how does environment influence the evolution of morphological complexity?

There are many approaches to quantify the complexity of an evolved morphology. Commonly, the variability of part types such as the number of cell types [41] has been used to measure the morphological complexity of biological organisms.

But, the parts under consideration may vary in scale from organelles [42] to limbs [43] , and it is unclear what should be considered a part in the current work. More geometric measures describing how space-filling a morphology is could also be employed see Text S1 and Figure S2.

Alternatively, a morphology's surface area to volume ratio could be measured, or its concavity could be computed e. However each of these measures may be deceived by relatively simple body shapes, such as those that are very flat or contain large, simple concavities e. Instead, it is useful to think about the complexity of a body plan in information theoretic terms.

One commonly used measure of complexity is Shannon's Entropy [44] , which measures the uncertainty of a random variable. Recent work [45] , [46] has demonstrated how Shannon Entropy can be applied to measure the complexity of a 3D object by considering the curvature of the object as a random variable.

This means that in order to have higher complexity it is necessary to have more angles regions of non-zero curvature that can not simply be a repeating pattern, exactly what humans would think of as more complex shapes. And in fact, quantifying the complexity of 3D objects in this way has been shown to strongly correlate with human observers' notions of complexity [46]. In this work, the complexity of an organism's morphology is computed as the quantity which is the morphology's entropy of curvature or, in terminology which may be more familiar to biologists, it is the Shannon diversity [47] of the curvature on the organism's exterior see Methods for details.

Does capture the complexity of evolved morphologies? To answer this question, is calculated for all best-of-trial virtual organisms from all environments icy and control.

Out of those , the five morphologies with the smallest value and the five morphologies with the largest value are selected. Images of these morphologies are shown in Fig.

Looking at these two sets of morphologies, those with high values appear more complex than those with low values. In light of this observation and the previous work in this area it is concluded that successfully captures morphological complexity. Similarly, the concept of entropy may also be applied to characterizing the complexity of an environment. In the current formulation, environments are differentiated by variability in surface friction and terrain elevation.

In the flat ground environment both the height of the terrain and the surface friction are uniform throughout, thus conveying zero entropy. On the other hand, in all of the icy environments there is variability in both of these properties. The surface friction is low on the ice blocks, but high on the ground between them. Likewise, the terrain is one height on the blocks and another in the intervening space. Therefore each of the icy environments has non-zero entropies of friction and elevation and so is considered to be more complex than flat ground.

However, since each icy environment consists of a uniform series of ice blocks, the relative complexity between these environments is not considered.

Armed with these measures, it is now possible to characterize how different environments influence the morphological complexity of evolving organisms.

In order to understand the evolutionary pressures which lead to virtual organisms that are more or less morphologically complex, it is interesting to consider how morphological complexity varies over evolutionary time in different environments, and how these changes correspond to variations in fitness.

Towards that goal, Fig. Here it can be seen that morphological complexity tends to increase over time along with fitness. This means that in these environments selection for locomotion corresponds to an increase in complexity. This plot depicts morphological complexity and fitness displacement over evolutionary time for three sample icy environments left : spacing 0.

For the icy environments morphological complexity is plotted in blue and displacement is plotted in red. For the corresponding trials in the control environment morphological complexity is plotted in black and displacement is plotted in green. Solid lines denote means taken across all best-of-generation individuals from all trials in the set and dotted lines denote one unit of standard error.

However, it is unclear whether this increase of complexity is the result of a passive or a driven trend [3] , [6] , [34]. Passive trends may result from envelope expansion without any directional bias. On the other hand, driven trends exhibit a consistent, directional bias. This corresponds to active selection for greater complexity. In this case not only will there be an increase in mean and maximum complexity, but the minimum level of complexity will increase over evolutionary time as well.

In order to separate the influence of these factors it is useful to compare the evolving populations to a neutral shadow model [49] , [50]. For a generational evolutionary model, such as that employed here, a neutral shadow of a given experiment is equivalent to re-running the evolutionary model with the same parameters but with random selection.

It is known that the evolutionary system employed here [40] has an inherent bias to increase genotypic complexity over evolutionary time. The increasing purple curve in Fig. In fact, random selection alone produces morphologies that are more complex than those selected in any of the environments investigated.

However, this comparison is not entirely fair. At any given generation, individuals in the random selection experiments will be the end product of many more reproduction mutation and crossover events than the corresponding individuals evolved for displacement, because under random selection it is unlikely that any individual will persist in the population for very long.

Therefore, individuals in the random selection experiments will have had many more opportunities to increase the complexity of their genomes and hence the complexity of their morphologies. This plot compares morphological complexity over evolutionary time for all single-objective experiments in the control environment black and all icy environments blue along with several neutral shadow models.

Solid lines denote means taken across all best-of-generation individuals from all trials in that environment and dotted lines denote one unit of standard error. The remaining lines depict the alternate shadow models with reproduction depths matched to the two real evolutionary experiments see Text S1 for details.

In order to correct for this discrepancy in the number of reproduction events, alternative shadow models are employed. Specifically, neutral shadow models of both the flat ground experiments and a representative icy environment spacing 0. In each of the independent trials evolving for locomotion in both of these environments, a record of every reproduction event is kept, and alternative shadow models are created for each trial such that they maintain the same rate of reproduction.

These shadow models are detailed in Text S1. All model alternatives have similar complexity curves see yellow, green, red and gray lines in Fig. This implies that greater morphological complexity is being actively selected for in these environments: there is a driven trend towards increased morphological complexity. While the results reported so far support the hypothesis that there is a driven trend for increased morphological complexity in all environments, they do not differentiate between the complexities of organisms evolved based on which environment they are evolved in.

Specifically, Fig. In fact, when the morphological complexities of organisms evolved in each of the 49 icy environments are compared with independent sets of trials conducted in the control environment see Figs. It is hypothesized that without a cost to becoming more complex the driven trend towards increased morphological complexity will dominate in all of the investigated environments.

On the other hand, it is hypothesized that when complexity does come at a cost—as is thought to occur in biological organisms [35] , [36] —there will be greater pressure towards increased morphological complexity in more complex environments. In an an attempt to test this hypothesis, a second set of experiments is conducted which uses Pareto based multi-objective selection [51] , [52] to evolve organisms that can locomote in their given task environment and are as simple as possible, therefore imposing a cost on complexity.

As was done for the single-objective experiments, independent trials of a multi-objective model are run in each of the 49 icy environments along with a corresponding 49 independent sets of trials apiece in the high friction, flat ground control environment.

By selecting for both maximal displacement and minimal morphological complexity these experiments should evolve organisms that are no more complex than necessary to succeed in their task environment. If indeed more complex environments induce greater selection pressure favoring morphological complexity than simple environments when morphological complexity comes at a cost, then these differences should be observable under this regime.

Comparing the results of these multi-objective experiments, we indeed see that more complex environments tend to select for organisms with greater morphological complexity when compared with organisms evolved in the simple, control environment. Since selecting a single representative individual from each trial is not as straightforward as in the single-objective case see Methods , several different techniques are employed to compare the results of these experiments.

This plot compares the ways in which the complexity of morphologies from icy environments differ from the complexity of morphologies evolved in the control environment under multi-objective selection. This plot is created from the multi-objective results by comparing the mean values across each trial's final Pareto front in each icy environment to the mean values across each trial's final Pareto front in a corresponding set of trials in the control environment.

All p -values calculated using the Mann-Whitney U test.



0コメント

  • 1000 / 1000