Who is our most distant animal relative?
There is great interest in a basic question about animal evolution —Who is our most distant animal relative? Until relatively recently it was widely thought that sponges are our most distant animal relatives, i.e. that sponges are sister to all other animals. According to this hypothesis, the first divergence in the animal tree of life was between the lineage that gave rise to sponges and the lineage that gave rise to all other living animals (including ourselves).
Recently, there have been multiple analyses that instead found support for ctenophores (comb jellies) as the sister group to other animals. My colleagues and I first reported this result (Dunn et al. 2008), and since then it has also been recovered in a variety of studies by ourselves and others (Hejnol et al. 2009; Ryan et al. 2013; Moroz et al. 2014; Whelan et al. 2014; Borowiec et al. 2015). Other papers have continued to argue that sponges are the sister group to other animals (Philippe et al. 2009; Pick et al. 2010; Nosenko et al. 2013; Dohrmann and Wörheide 2013).
The authors of a recent paper in PNAS (Pisani et al. 2015) claim, in their manuscript and their public discussion of their results, to have demonstrated that all results of ctenophores as the sister group to other animals were artifacts, and that sponges are the sister to other animals.
Multiple research groups are working on the root of the animal tree, and new data and methods will continue to inform this discussion. In the mean time, colleagues have asked for my thoughts on the Pisani et al. (2015) paper, and I hope it will be productive to share them here.
Sorting out the placements of ctenophores and sponges is what is known as a rooting problem. Does the root of the animal tree divide ctenophores from all other animals, or sponges from all other animals? It is important to note that this isn’t a matter of figuring out whether ctenophores or sponges are older or first, or which of them is the ancestor of other animals — it is a matter of figuring out the historical evolutionary relationships between organisms that are alive today. All living animals have had the exact same amount of time to evolve since their most recent common ancestor, and this distant ancestor was likely very different from any animal alive today. Modern sponges or ctenophores are no older or younger than we are. Just as it wouldn’t make sense to think of humans as the ancestors of ctenophores or sponges, it doesn’t make sense to think of ctenophores or sponges as the ancestors of humans. The most direct way to address a rooting question is to add outgroups to the phylogenetic analysis. Outgroups are organisms that you have prior reason to believe fall outside your focal clade. In this case the animals are the ingroup, and relatives of animals are the outgroup. The point where the ingroup attaches in the phylogeny to the outgroup tree is the ingroup root.
The new study
Pisani et al. (2015) reanalyzes several datasets from previous papers that addressed the phylogenetic positions of sponges and ctenophores. They performed multiple variations on the previously published analyses: they used different models of molecular evolution, they changed outgroup species sampling, and they changed gene sampling.
What are the claims of the new study?
The authors of the Pisani et al. (2015) make two primary claims. First, they claim to demonstrate that the placement of ctenophores as the sister group to other animals in other studies was an artifact. Second, they claim that there is strong support for the placement of sponges as the sister group to other animals. They state these claims strongly and unambiguously:
Genomic data do not support comb jellies as the sister group to all other animals — Title of Pisani et al. 2015
Here, we analyzed the current genomic evidence from comb jellies and found no convincing support for this hypothesis. Instead, when analyzed with appropriate methods, recent genomic data support the traditional hypothesis. — Significance statement of Pisani et al. 2015
We found no support for Ctenophora-sister and conclude it is an artifact resulting from inadequate methodology, especially the use of simplistic evolutionary models and inappropriate choice of species to root the metazoan tree. —Abstract of Pisani et al. 2015
We were able to demonstrate that the placement of Ctenophora at the base of the evolutionary tree of animals is artefactual. When the more powerful models were applied to these datasets, we found that sponges are indeed the earliest diverging animal group. — Gert Wörheide
Our results offer a simple solution to the problem of animal origin: ctenophores are not the first animals, the sponges are, and this is a group of filter feeding organisms much like our more distantly related eukaryotic relatives, the choanoflagellates. — Davide Pisani
What are the results of the new study?
The new paper reanalyzes selected transcriptome and genome datasets from three earlier studies (Ryan et al. 2013; Moroz et al. 2014; Whelan et al. 2014). They evaluate several models of molecular evolution. These include CAT, which accommodates among-site variation in equilibrium frequencies, and CAT-GTR, which shares the same features as CAT and also estimates exchange rates from the data. They find CAT-GTR, the most complex model, to be the best fit and use it in their analyses. This in itself isn’t surprising; the manual of the tool they used explains that the computationally intensive CAT-GTR is almost always selected to be the best fit.
They also vary models and gene sampling to investigate the support of gain and loss of genes for the phylogenetic placement of sponges and ctenophores. This builds on a previous analysis that my colleagues and I published with the Mnemiopsis genome (Ryan et al. 2013). This in many ways is the most interesting part of the paper. They find our original analyses sensitive to model selection and also question the criteria we used for selecting genes. This is productive criticism, and it will be interesting to follow up on.
Much of Pisani et al. (2015) focuses on the impact of outgroup selection on the placements of sponges and ctenophores. Several previous studies found that the placements of sponges and ctenophores are sensitive to whether more distant outgroups are included in addition to close outgroups (e.g. Nosenko et al. 2013; Whelan et al. 2014). Pisani et al. (2015) further characterizes this same sensitivity.
The greater the number of outgroups that are sampled, the greater the support for ctenophores as sister. Conversely, as outgroups are removed support for sponges as sister increases. This applies across the datasets they examined.
Why I don’t think the claims follow from the results
While the results are interesting, I don’t think that they support either of the two primary claims of the authors of Pisani et al. (2015), i.e. that the placement of ctenophores as the sister group to other animals is an artifact and that sponges are the sister group to other animals.
Contrary to their statement in the abstract that “We found no support for Ctenophora-sister”, many of their analyses do in fact recover Ctenophora-sister. This can be seen in their Figure 2 (included above). They just dismiss this support as an artifact. This makes their criterion for establishing an artifactual result the pivotal feature of the paper. What is this criterion? Their central argument is that when the rooting of the ingroup is sensitive to how many outgroups are included, the analyses with only a small number of closely related outgroups should be accepted, and analyses with additional outgroups should be rejected. They claim that artifacts can be reduced by reducing outgroup taxon sampling:
In particular, the inclusion of outgroups very distant from the ingroup can cause reconstruction artifacts by attracting fast-evolving (long-branched) in- group species toward the root (25, 27–31). A typical solution is to introduce more closely related outgroups to “break up” the long branch leading to the ingroup, but long-branch attraction artifacts can be further minimized by also removing the distant outgroups. — Pisani et al. 2015
All analyses under consideration include closely related outgroups, so the variation under consideration is whether more distant outgroups are included or removed. As seen above, the authors find that rooting of the animal tree is sensitive to outgroup sampling, with analyses that include only closely related outgroups tending to recover sponges as sister and analyses that also include more distant outgroups tending to recover ctenophores as sister.
But the authors then go beyond describing this sensitivity. They interpret this particular pattern of support as evidence that sponges are sister to other animals. In essence, the authors are arguing that this rooting problem has a Goldilocks Zone — without any outgroups you cannot root the tree, with too many outgroups you get artifacts, but in the middle is a sweet spot where the sampling is just right and you get an unbiased answer that you can trust. In this case, the just-right sampling gives sponges as sister and too much sampling gives ctenophores.
There are several problems with the interpretations of these outgroup experiments as evidence for sponges as sister. First, the topology of the tree is being used to gauge the boundaries of the Goldilocks Zone, not independent explicit criteria. This is circular when the question is about the topology of the tree. Second, a major focus of modern phylogenetic studies is to improve taxon sampling, because it has been shown time and time again to have positive impacts on the ability to resolve tough nodes. I am highly suspect of being asked to reject a hypothesis that is recovered with broader taxon sampling because reduced taxon sampling supports a different hypothesis. Third, the extrapolation of the claim that we need to be careful to limit outgroup sampling has nonsensical implications when applied more broadly. It would suggest, for example, that we can’t simultaneously root clades while placing them in a broader phylogenetic context. For example, this assertion would suggest that we cannot have a well-sampled eukaryote phylogeny that tests the relationship of animals in this broader context and correctly roots animals. But the distinction of an ingroup is contingent only on the definition of the question at hand, and every node on a tree designates a clade that could be an investigator’s ingroup. If I am a mammalogist and consider mammals my ingroup, do I need to remove all but their closest vertebrate relatives to get a correct rooting? That would suggest that it is impossible to resolve any node on a phylogenetic tree.
The lack of explicit criteria other than tree topology for identifying artifacts is also highlighted by comparison to recently published studies:
Thus, while strong support for Ctenophora-sister may be obtained from phylogenomic datasets (2–6, 46, 55), our analysis suggests these results are caused by undetected systematic bias. — Pisani et al. 2015
Citation 55 above is another PNAS paper (Chang et al. 2015) that was published two weeks before Pisani et al. 2015 . These two papers share a co-author (Hervé Philippe), have taxon sampling relevant to the phylogenetic placement of sponges and ctenophores, and address issues that are claimed to introduce artifacts that result in ctenophores being placed as the sister group to animals. In particular, more distantly related fungi are not included (the sampling here corresponds to Holozoa), and a site heterogenous model (CAT) is used. Here is the tree from Chang et al. 2015:
It places ctenophores as the sister group to other animals with strong (100%) support. Chang et al. (2015) was focused on the position of Myxozoa, and didn’t discuss the ctenophore position. Pisani et al. (2015) cites Chang et al. but declares that the position of ctenophores is due to “undetected systematic bias” without any specific support for this claim. The only detected criteria that Pisani et al. (2015) present as evidence that this study is biased is that it gets ctenophores as sister, a circular argument since this is the hypothesis under consideration.
Other recent analyses also find support for ctenophores sister with exclusion of all outgroups except Choanoflagellata (Borowiec et al. 2015). This indicates that the specifics of outgroup sensitivity are even more complicated than previous studies found.
What is my take on whether sponges or ctenophores are the sister group to animals?
My take is that we don’t have a final answer yet on the root of the animal tree. All recent well-sampled analyses suggest that the sister group is ctenophores or sponges, but we can’t completely exclude either yet.
This is a very hard problem, involving nodes that are more than 500 million years old and very short internal phylogenetic branch lengths. But I don’t think that demonstration of sensitivity to outgroup sampling (such as including fungi or not) or closely related models (such as CAT vs. CAT-GTR) alone can be used as evidence to reject some hypotheses outright as artifacts. So while the authors of Pisani et al. (2015) claim to exclude ctenophores as the sister group and find sponges as the sister group, I don’t think we can exclude either sponges or ctenophores as sister outright at this point. Much of the disagreement regarding the placements of sponges and ctenophores now comes down to how to interpret sensitivity to outgroup sampling, and how this may interact with model selection.
Where does this uncertainty about the placement of sponges and ctenophores leave our understanding early animal evolution? While there is still intense interest over some open questions, such as the root of the animal tree, there is broad consensus regarding many nodes that are critical to understanding broad patterns in animal evolution. Colleagues and I recently reviewed this consensus and open questions (Dunn et al. 2014). We can still say a lot about animal evolution even without a fully resolved tree.
Consider, for example, the evolution of traits such as the nervous system that are often categorized as “complex”. Pisani et al. (2015) end their paper by stating:
Our results do not support the currently emerging point of view according to which the origin of complex characters, such as nervous systems, was far more complicated than previously thought (e.g., 7, 8). — Pisani et al. 2015
The two citations are for reviews I co-authored about animal phylogeny and trait evolution (Dunn et al. 2014; Dunn et al. 2015). In these, we describe that many characters exhibit much more homoplasy than previously thought and explain that many clade-specific traits have been overlooked when making generalizations about trait evolution. In stark contrast to the first animal phylogenies that were constructed by minimizing the number of changes in complex traits, it is now clear that many traits exhibit an evolutionary history that is far more complicated than previously thought, with many independent gains and/or losses:
This “emerging point of view” that complex characters have a complicated evolutionary history does not hinge on the phylogenetic placements of sponges and ctenophores, but builds on the work of many investigators working on many groups of organisms. With regard to the nervous system evolution, the trait that Pisani et al. (2015) raise, we explicitly pointed out that there is homoplasy regardless of whether sponges (Porifera) or ctenophores are the sister group to all other animals:
Nerve cells are present in Ctenophora, Cnidaria, and Bilateria but are absent in Placozoa and Porifera… Nerves may have arisen independently in Ctenophora and the clade composed of Cnidaria and Bilateria. Alternatively, nerves may have had a single origin followed by loss in Placozoa alone (if Porifera is the sister group to all other animals) or both Placozoa and Porifera (if Ctenophora is the sister group to all other animals). — Dunn et al. 2014
The trees of Pisani et al. (2015) also place Placozoa in a position that indicates multiple gain and/or loss of the nervous system, but the authors do not consider this taxon in their character evolution discussion. Of course the focus of their paper is phylogenetic relationships, but their statement about nervous system evolution is insufficient without considering Placozoa as well. Their placement of Placozoa also calls into question their conclusion that a recent paper on ion channel evolution (Liebeskind et al. 2015) should be reassessed.
It’s OK to have open questions about deep animal phylogeny. I do hope that methods are developed and data are collected that will lead to broad consensus on a well supported placement of the animal root, and given how quickly the field is advancing this could happen soon. There has been considerable progress on understanding the animal phylogeny in recent years (including many important contributions by the authors of Pisani et al.). There will continue to be important gains in our understanding of character evolution even as questions persist about key relationships in the animal tree of life. It isn’t time to get out the coffins, let alone the nails, when we consider whether sponges or ctenophores are the sister group to other animals.
Addendum
I’ve done some further analyses on the data from Chang et al. (2015), which was focused on the placement of Myxozoa. As described above, this paper found strong support for the placement of ctenophores as the sister group to other animals. Without presenting additional analyses, Pisani et al. (2015) claimed that Chang et al.’s position of ctenophores was due to “undetected systematic bias”. It is circular to assert that the placement of ctenophores as the sister group to other animals is alone evidence of bias if the question at hand is the placement of ctenophores, so additional analyses were needed to address this.
The first step was to remove Myxozoa, which was well sampled in Chang et al. but not Pisani et al. This clade has a highly accelarated rate of molecular evolution that may have impacted the rooting of animals. I then performed four analyses that looked at two specific factors that Pisani et al. focused on as potential sources of bias, model selection and outgroup selection. Even when a more complex model (CATGTR) is used and outgroups were reduced to include only choanoflagellates, support for the placement of ctenophores as the sister group to other animals was still strong. There were a couple relationships within Cnidaria and Bilateria that did not converge between runs, but the placements of sponges and ctenophores did. These preliminary analyses therefore do not find any evidence of “undetected systematic bias” impacting the placement of ctenophores in Chang et al. Please see the git respository for all data, code, and results.
