Carrie Mongle

Bayesian tip-dating approaches to inferring evolutionary relationships among hominin taxa

One of Richard’s lasting legacies is undeniably the extraordinary number of hominin fossils he was able to add to our collective understanding of human evolution. Phylogenetic inference is a critical step in figuring out how all of these fossils come together to form the human family tree. This talk will review our recent work in reconstructing the hominin phylogeny, including new analyses to incorporate geochronological information into these inferences.

FULL TRANSCRIPT 

Thank you all for coming back from lunch on time. I'm really excited to be here and I really appreciate the organizers for inviting me and especially to the Leakey family for very recently welcoming me into the TBI fold. So today I'm going to be talking about more broadly than the Turkana Basin, how do we infer how all of these things are related to one another? And I wanted to start just with this very traditional kind of linear view of human evolution because I think this sets the stage for kind of the predominant line of thinking when Richard was first kind of leading these expeditions, especially in the early seventies. So we could think of this as some kind of a chimp like ancestor with a direct line going through maybe something like Lucy discovered in the seventies. And then we had Homo erectus, Neanderthals and then the culmination with modern humans. 

And this was also the source of one of Richard's more notorious disputes scientifically when he was interviewed alongside Donald Johanson by Walter Cronkite and asked to draw a hominin tree. So Don very famously had a tree prepared and drew this very linear relationship with Australopithecus afarensis ultimately leading directly towards humans. And Richard drew an X over that and put a big question mark there instead. In later years with reflection, he more calmly stated that he thought that the story was probably a little bit more complex than it was being presented. And so that was in an NPR interview in 2011. I think that the last 50 or 60 years of field work have probably proven Richard right, that it is a very complex hominin family tree with as many as six or seven different hominin species coexisting at a given time. And we know from talks that were given in the last few days that that's getting pushed earlier and earlier where we have a diversity of hominin species not just at 2 million years, but probably also at 3 million years. And who knows, maybe at four. 

So my big research question is how do we figure out how all of these things are related to one another? Always keeping in mind that this is still an incomplete picture, but what can we do to say something about the evolutionary relationships of all of these hominin species? I would argue that figuring out phylogenetic relationships forms the essential foundation for how we argue about adaptations and when they took place. So did it happen at a single point in time and then all of the hominins after that had that adaptation? Or do we see an adaptation happening multiple times across the family tree, which fossils, if any, that we have today represent a good candidate for our earliest common ancestor? And something that's been the topic of a lot of conversation lately is which of these many, many Australopithecine species, again, if any, represent the candidate for the origin of the genus homo? 

So I would argue again that figuring out the hominin family tree or phylogenetic tree is essential for at least starting to reconstruct how we make those inferences. I'm not the first person to ask this question, obviously. And so in the late 1960s, the methods for this were really formalized in the evolutionary biology community. And very shortly thereafter, many of the people in this room and a great many other put forward some different hypotheses about hominin phylogenetic relationships. And so what I'm doing is building on a lot of this previous work, but then expanding it. So, the starting matrix that I use in my analysis was from the Strait and Grine 1999 and 2004 series of matrices. But as many of you know and have made these discoveries, the search for fossils wasn't complete in 1999. So, I think it was about time that we updated the hominin matrix and included all of these new discoveries, not just new species, but also new fossils to expand our knowledge of something like A. afarensis or variation in that hypodigm.

One example of this has been the Ardipithecus ramidus hypodigm. And so, what you're looking at here just to orient you is a visual representation of our cranio-dental matrix. And when I say cranio-dental matrix, this is your very classic sagittal crest present or absent or orientation of the zygomatic. And each of the white squares I have represented here are missing data for A. ramidus at the time that Strait and Grine published their original matrix and any analysis that directly incorporated their data. So, for example, Dembo and many others have directly included this matrix would've had the same sampling. What we wanted to do as kind of a first step before we think about including new species is include the 2009 publication of ramidus fossils. And this brought our sampling up for ramidus to about 80% of the cranio-dental character matrix. And for anyone who knows anything about parsimony, you know that the characters that you have at the very base of your tree, the very earliest part of your tree, are going to influence how we infer evolution along the rest of the tree. So filling in this really early gap was really critical for the stability of the rest of the analysis. We were also able to directly include some other new fossils that had been discovered for existing species. And so that included the new rudolfensis material as well as several new fossils from Hadar. 

And I just want to very briefly go through some of the highlights. I know not everyone likes looking at trees, and these are very small to see on the screen, but a few things that I wanted to point out here, one interesting and in the context of the latest round of post cranial papers that have come out for Sahelanthropus is that we find pretty weak support for Sahelanthropus as a basal hominin, we find very strong support for ramidus, but I think Sahelanthropus maybe knew fossils or access to the original fossils would change our interpretation of that. But as it stands currently, fairly weak support on that front. And we also in our 2019 paper found that the best candidate for the origin of the genus Homo was Kenyanthropus. And so if you're familiar with bootstrap supports, you might be wondering why that's an eight there. 

I'm happy to talk during coffee. These are not absolute values, but GC values. So, it is still weak support, but if we were to put Kenyanthropus somewhere in this analysis, we came out with it kind of as the basal member of the genus Homo. If you are able to see the names on this tree, and I apologize if you're not, you might be asking yourself if I'm talking about the origin of the genus Homo, where is Australopithecussediba on this tree? So we wanted to make sure our fossils were completely sampled and our matrix was updated before we started adding new species, but that was our next step. So what about A.sediba? As we know, a lot of people have proposed that it might share a unique link with the genus Homo on a combination of primitive and more derived characteristics, and particularly in the cranium. 

However, others have argued that a lot of these Homo-like characteristics are simply because we're looking at a juvenile. So MH1, which is shown on the bottom here, is a young individual. And so it has reduced post orbital constriction. It has a fairly flat face, and Kimbel and others have argued that if we were looking at it as a grownup, it might look like A. africanus. So we wanted to test this explicitly in a phylogenetic framework. So this was our most recent phylogenetic analysis that came out in 2023. We were very grateful to Lee and others for letting us look at the original fossils. So we examined MH1 and MH2 two ourselves in person, and we really wanted to evaluate the impact of those ontogenetically driven characters. So, what happens if you just treat those all as missing data for A. sediba? We assume we don't know what it would've looked like as an adult at all. 

We also wanted to explicitly test the hypothesis that it might share a sister relationship with Australopithecus africanus. And so we use Bayesian methods to do that. Happy to talk about those later and not to miss an opportunity to continuously expand the matrix. We also took this opportunity to incorporate new and really beautiful discoveries that had been found since 2019. And so, this includes the Woranso-Mille skull, which we're very grateful that Johannes actually coded for our matrix in their supplemental information. So that made that very easy for us. And then the new Drimolen skull that I got to be part of in 2021. 

So there were a number of trees that came out of this. I'm not going to make you look at all of them. We are going to just focus on two. And these are two of the most parsimonious trees that came out of the analysis. When we have multiple most parsimonious trees, basically we can say we have equivocal results, they're equally probable. And what are those? In one scenario, we find sediba as the sister to Habilis, rudolfensis and later Homo, but Kenyanthropus actually is still the basal member of that grouping. And in the other scenario, Kenyanthropus hops down and forms a much more basal position where it's actually kind of earlier or more basal to the origin of both Paranthropus and Homo. Generally, at this point in talks especially about fossils, we say we need more data, so we need to go out and we need to find more fossils. 

Arguably, for this type of analysis, we actually do already have some more data that could come to bear on this. So, two sources of data that we could think about as far as reconstructing or resolving this problem would be geochronology. So how old are some of these fossils and does that influence the probability that they would be more or less basal? And then we can also, I'll talk about this later in the talk, think about the role of postcrania. So our analyses of hominin taxa are entirely head up. So, I'm going to start with geochronology, and I'm going to approach this through a method called tip dating. And so, this is a Bayesian method that we can use to calibrate the hominin tree. And basically, what we're doing is we're taking that morphological data, the kind of traditional craniodental character matrix, and we're combining that with molecular data for the living taxa as well as some ancient DNA for more recent. 

And that gives us a calibration on our kind of scale of how quickly things are changing. And then the geochronology of the actual fossils themselves, so time bins of how old these things actually are. The data that went into this are all available. So the morphological data we posted on Morpho bank, so that's available for download from our 23 paper, the mitochondrial DNA that goes into this was all from GenBank, so that's for both the apes as well as the modern humans, and Denisovans, Neanderthals, all of those, the first and last appearance dates I think are really important. And so we spent a lot of time arguing about what would be the most appropriate first appearance date for a given fossil, and how do we incorporate that into the analysis? We've posted that online as well as a supplementary that people can download and read our full explanation for why we pick certain dates. 

But if there are geologists in the room that would like to lay in on that, I'm happy to have input. And then all of this was run on the CIPRES science gateway in Mr. Bays. It takes a really, really long time. So what did we find first, just to orient you, because there's a lot of data being shown on this tree. The distributions in blue at each of the nodes are kind of the range of date estimates in our analysis. And the value that's shown right in the middle of that distribution is the mean estimation for that divergence date. And again, this takes into account both the age of the fossils, the relationships of the fossils as well as the molecular clock. What was reassuring, I think first of all, is that the topology of this tree, so the relationships among the different hominin taxa does not differ from our pure cranial dental analysis. 

So we get the same evolutionary relationships between taxa as we do just thinking about skulls and teeth. So that's reassuring. What is interesting is that we recover a very deep divergence between Paranthropus and Homo. So, we're estimating a divergence time between Paranthropus and Homo at around 4.5 million years. This makes sense when you think about the basic mechanics of this analysis character change over units of time, and there's a lot of character change between these two groups, so we need a lot of time to accumulate that. But I think with some new papers that have been published recently, so that tooth that was associated with the stone tools at 3.3 years, maybe some other fossils that we might hear about one day soon, I wouldn't be surprised to see much earlier Paranthropus. And I think this analysis is in line with that. Even with a very long ghost lineage, we are still recovering sediba as the kind of sister to the genus Homo. So a lot has been said about it's far too young to be in this position in our analysis. It's still being placed with a long lineage still as the sister to that group. 

But a lot of people have also said, well, maybe we should think about bones that are not in the head. So, we should think about post crania as far as reconstructing these phylogenetic relationships, especially related to sediba. That makes sense. When you think about hominins kind of evolution more broadly, our defining characteristic is bipedalism, so maybe that should be a character in our matrix. But for a really long time that has been very difficult to do. So aside from rare cases like the Turkana Boy and Lucy, we have very few species where we can confidently say that post cranial elements are attributable to a species associated with teeth or with craniodental remains. That is changing rapidly as we reviewed in a recent paper. And this is because of discoveries like sediba, like Little Foot, like the ramidus skeleton. I think we actually are now in a position where we can incorporate post cranial data. 

So this started as part of my postdoctoral work with Ashley Hammond at the a AMNH and continues now because my postdoc was during Covid. If you think about the entirety of the data that we use to estimate these evolutionary relationships, we have around 110 characters, so about 110 data points. By my estimate, by the time we finish this, we should be closer to around 600 data points across the skeleton that we're using to infer evolutionary relationships. And so, I think that puts us at a really exciting time to rethink how all of these hominin taxa might be related to one another, and how we might go about sorting out some of that diversity that we see throughout the Australopithecines and now that we're seeing within the genus. Homo, thank you very much. 

View all Human Evolution videos