Morphological and phylogeographic evidence for budding speciation: an example in hominins

Abstract

Parametric phylogenetic approaches that attempt to delineate between distinct ‘modes’ of speciation (splitting cladogenesis, budding cladogenesis and anagenesis) between fossil taxa have become increasingly popular among comparative biologists. But it is not yet well understood how clearly morphological data from fossil taxa speak to detailed questions of speciation mode when compared with the lineage diversification models that serve as their basis. In addition, the congruence of inferences made using these approaches with geographical patterns has not been explored. Here, I extend a previously introduced maximum-likelihood approach for the examination of ancestor–descendant relationships to accommodate budding speciation and apply it to a dataset of fossil hominins. I place these results in a phylogeographic context to better understand spatial dynamics underlying the hypothesized speciation patterns. The spatial patterns implied by the phylogeny hint at the complex demographic processes underlying the spread and diversification of hominins throughout the Pleistocene. I also find that inferences of budding are driven primarily by stratigraphic, versus morphological, data and discuss the ramifications for interpretations of speciation process in hominins specifically and from phylogenetic data in general.

1. Introduction

Parametric phylogenetic approaches that aim to distinguish between competing ‘modes’ of speciation between fossil taxa have rapidly proliferated over the past several years. These approaches typically employ extended models of lineage diversification that allow taxa to be related by one of several possible speciation patterns that have long been discussed by palaeontologists [1]. These include: (i) splitting cladogenesis, where a lineage splits evenly into two daughter lineages, (ii) budding cladogenesis, where a smaller daughter lineage splits off an older ancestral lineage, and (iii) anagenesis, where a single lineage continuously evolves without splitting. These approaches aim to reconstruct ancestor–descendant (AD) sequences of fossil taxa through the stratigraphic record using a unified model of lineage diversification and stratigraphic preservation [2,3]. These approaches have been increasingly leveraged to address empirical patterns in the fossil record [4], yielding results that authors have interpreted to highlight the ubiquity of budding speciation as a dominant process in the diversification of new taxa [5]. Nevertheless, few investigations have directly examined morphological support for mode. The use of geographical data has also remained under-explored [6–8].

Hominins are a strong candidate taxon in which to examine alternative speciation modes in the fossil record, with a history of disagreements over evolutionary mode spanning back decades. Early approaches employed cladograms and stratophenetic diagrams to examine relationships [9–11]. Repeated attempts to reconstruct hominin phylogeny [12–16] left major controversies concerning specific hypotheses of direct ancestry unresolved [17–19]. Parametric approaches have since been applied to hominins, showing promise in quantifying statistical support for competing hypotheses [20]. Further explorations have incorporated AD relationships [21]. Despite this improved resolution, many important issues remain in hominin phylogeny. Nevertheless, the role of more complex speciation scenarios such as budding has not been statistically evaluated.

In this study, I examine the morphological and phylogeographic evidence for budding speciation in shaping hominin phylogeny. I extend a previous parametric approach for ‘stratophylogenetic’ inference, including AD relationships, from morphological and stratigraphic data to accommodate budding cladogenesis, which is considered here as a directly ancestral arrangement where a putative ancestor overlaps with a descendant in temporal range. I then use the phylogeny to explore the phylogeographic patterns underlying the spread and morphological divergence of Pleistocene Homo populations. Lastly, I evaluate whether morphological data alone support the AD relationships that underpin inferences of budding when the stratigraphic record is not considered.

2. Material and methods

(a). Morphometric data

I used a dataset of cranial landmark coordinates from the literature gathered from 18 fossil hominin specimens and two extant outgroup taxa [22]. I aligned the 46 three-dimensional landmarks using Procrustes superimposition in MorphoJ [23]. I then transformed the data using a principal component analysis (PCA) in R [24]. I retained the 13 principal component (PC) axes needed to approach 100% of the variance in the dataset. Phylogenetic inference from morphometric landmarks is viewed as controversial by some [25,26] but appears to perform fairly well in young clades [27–30], such as is the case here.

(b). Evaluating ancestor–descendant relationships

I extended a previous approach for the evaluation of AD relationships from discrete traits [21] to accommodate both (i) continuous characters [31] and (ii) ‘budding’ patterns of ancestry where an ancestral taxon persists after a descendant splits off (electronic supplementary material, figure S1). Trees were scored using Brownian [32] and Poisson [33] models of morphological evolution and stratigraphic preservation, respectively. The implementation developed for this study accepted a fully bifurcating starting tree as input and identified statistical support for collapsing candidates into direct ancestors. I performed the test using several bifurcating starting trees: the first consistent with the results of Parins-Fukuchi et al. [21] and successive trials within a nearest-neighbour interchange (NNI) move from the first. Taxa with a lower stratigraphic boundary than their sister taxon or clade were treated as candidate ancestors. I implemented a procedure that iterates through the candidates, collapsing them into ancestors by constraining them to zero branch lengths and assessing statistical support from both the morphological and stratigraphic data. I retained the topology, including AD relationships, from the tree with the best Akaike information criterion (AIC) score. AIC was used rather than raw log-likelihoods because the addition of an AD relationship reduces the number of free parameters by one. Model uncertainty for AD assignments was estimated using AIC weights [34]. All AD analyses were performed using the mandos executable available in the cophymaru code repository (https://github.com/carolinetomo/cophymaru).

(c). Biogeographic analyses

I inferred biogeographic ranges using Lagrange (https://github.com/blackrim/lagrange). Internal nodes associated with ancestral taxa were constrained to geographical ranges corresponding to the associated taxon. Ranges were treated as discrete units corresponding to four areas occupied by hominin populations throughout the Plio-Pleistocene: Africa, Asia, the Middle East (encompassing the Levant, the Arabian Peninsula and the Steppes) and Europe.

3. Results and discussion

(a). Budding speciation in hominin evolution

When considered in a stratophylogenetic framework, morphometric and temporal data suggest several possible episodes of budding speciation (figure 1; electronic supplementary material, figure S2). Neither of the Australopithecus samples included in the analysis was identified as an ancestral taxon. Like several previous studies, Australopithecus afarensis is recovered as an outgroup to a clade comprising all of the later-occurring hominin morphotaxa. Many authors have interpreted those cladistic results to imply that A. afarensis is ancestral to Homo. However, these results suggest that A. afarensis is not directly ancestral but is instead a terminal outgroup. Since Australopithecus anamensis appears to be a chronospecies that is ancestral to A. afarensis [21,35], it is possible that the specimens assigned to A. anamensis represent a population ancestral to both A. afarensis and the later-occurring hominin lineages. Paranthropus boisei is identified as the direct ancestor of Paranthropus robustus. This finding is consistent with previous qualitative interpretations [36]. Paranthropus aethiopicus, on the other hand, is reconstructed with high confidence to be a distinct, branching taxon (table 1).

Figure 1.

Best stratophylogenetic tree with geographical range reconstructions mapped to internal nodes. Asterisks denote nodes fixed in the biogeographic analysis owing to their reconstruction as ancestral taxa. Geographical ranges representing tips and sampled ancestral taxa correspond only to the particular specimens contributing to the morphological data within each taxon in this study (see electronic supplementary material, table S1 for full list of specimens).

Table 1.

Support for each candidate taxon's position as a direct ancestor on the best stratophylogenetic tree. Model support is given by the AIC weight of the ancestral arrangement calculated relative to the model support for the bifurcating arrangement. The resulting support indices fall between 0 and 1, with values above 0.5 indicating greater support for a budding or anagenetic arrangement and below 0.5 indicating preference for a bifurcating relationship. When calculating support for each taxon, all other relationships and AD assignments in the tree were held at their ML placements. The first column reflects AIC support yielded by the combined stratigraphic and morphological dataset, while the second column reflects support displayed by morphology alone. OTU, operational taxonomic unit.

OTU	combined AIC weight	morphological AIC weight
P. aethiopicus	0.03	0.02
P. boisei	0.75	0.51
A. afarensis	0.02	0.02
A. africanus	0.14	0.02
H. habilis	0.41	0.21
H. erectus	0.77	0.26
H. heidelbergensis	0.75	0.49
H. rhodesiensis	0.02	0.01

The analysis revealed support for a budding event between Homo erectus and the clade encompassing Homo heidelbergensis, Homo rhodesiensis, Neanderthals and modern humans. The mid-Pleistocene specimens corresponding to H. heidelbergensis and H. rhodesiensis grouped polyphyletically, forming clades with Neanderthals and Homo sapiens, respectively. This result is notable, given the historical tendency to treat specimens corresponding to each taxon (as treated here) as geographical variants of H. heidelbergensis. Homo heidelbergensis, represented by only European specimens in this study, was reconstructed as a budding ancestor to Neanderthals, while H. rhodesiensis was reconstructed as a sister taxon to modern humans. The placement of European H. heidelbergensis specimens as directly ancestral to Neanderthals is consistent with both the genomic results of Meyer et al. [37] and the morphological results of Mounier & Caparrós [38], the former of which inferred the Atapuerca specimens as most closely related to Neanderthals and the latter of which reconstructed both Atapuerca 5 and Steinheim (the two specimens representing H. heidelbergensis in this study) as outgroups to Neanderthals. It does, however, disagree with a morphospecies-level phylogenetic analysis, which inferred H. heidelbergensis, defined as containing both the European and African samples included here, as ancestral to both Neanderthals and modern humans [21]. The discrepancy between studies can, therefore, be explained most simply by their differing treatments of H. heidelbergensis. The results here further demonstrate that careful examination of geography and morphology below the morphospecies level is needed to develop a thorough understanding of the spatial and demographic processes that shaped the divergence and reticulation of hominin populations throughout the Pleistocene.

Support of AD relationships differs when temporal and morphological data are considered together and when morphology is considered separately (table 1; electronic supplementary material, figures S3 and S4). In contrast with the combined dataset, the morphological data alone are either equivocal (H. heidelbergensis, P. boisei) or prefer a bifurcating arrangement (H. erectus). This may reflect the limits of morphospecies-level data. The temporal and spatial heterogeneity expected in a long-lived and widespread lineage such as H. erectus [39] may simply yield too much demographic complexity to map neatly to such a coarse representation of process. Adopting a finer scale of analysis will facilitate testing of the phylogenetic cohesiveness of named Pleistocene taxa such as H. erectus and H. heidelbergensis and identify whether breaking down spatial heterogeneity can yield a more detailed understanding of the processes underlying the divergence of Pleistocene populations.

(b). Pleistocene biogeography

The biogeographic range reconstructions suggest the presence of a geographically widespread, mid-Pleistocene ancestor to humans, Neanderthals, H. heidelbergensis and H. rhodesiensis. Pleistocene Homo can thus be characterized by patterns in widespread dispersal, followed by gradual fragmentation into geographically distinct subpopulations. Homo heidelbergensis and H. rhodesiensis are separated into geographically separate clades, disagreeing with treatments lumping them together into a single geographically widespread H. heidelbergensis. These results further underscore the need for examination at lower taxonomic scales than is typically undertaken. Overall, the picture of hominin evolution presented here represents an alignment of phylogenetic results with the heterogeneous geographical patterns observed in H. erectus and early African H. sapiens [40,41] and the complicated network of genetic interactions [42–45] towards a deeper statistical understanding of the complexity in evolutionary and demographic patterns between hominin populations throughout the Pleistocene.

(c). Phylogenetics and speciation processes in the fossil record

This study highlights the difficulty in identifying ancestral fossil populations from taxonomically coarse data. The comparatively weak morphological support for the AD relationships identified here suggests that researchers should be cautious when interpreting palaeontological phylogenies. In general, I suggest that reconstructions of AD relationships that are generated from phylogenetic data should generally be treated as contingent and coarse reconstructions as opposed to detailed inferences of the processes underlying the divergence of species. Population-level phenotypic (or molecular) and spatial data are typically needed to generate a detailed understanding of speciation processes in living and fossil taxa. If satisfied with the particular scheme of OTU (operational taxonomic unit) resolution and a coarse scale of inference, the framework here may be broadly useful to discern general patterns in phylogenetic ancestry. However, researchers using approaches that seek to delineate between competing modes of speciation should be mindful of scale. Further development of a clear understanding of the capabilities and limitations of data at multiple timescales will enable a continued expansion of the boundaries of fossil data when speaking to fundamental questions in evolutionary biology.

Data accessibility

All data and tree files used in the analyses are available on Figshare (https://figshare.com/articles/dataset/Morphological_and_phylogeographic_evidence_for_budding_speciation_an_example_in_hominins/13283507). The stratophylogenetic approach was implemented in the Go language as part of the cophymaru software package and is available as development code on Github (https://github.com/carolinetomo/cophymaru). More detailed information concerning the statistical procedure, including intermediate steps and treatment of the data, is contained within the electronic supplementary material. The code implemented to examine budding ancestor–descendant relationships is available on Github: https://github.com/carolinetomo/cophymaru/tree/dev. Input/output files, scripts, raw data, etc. are available on Figshare: https://figshare.com/articles/dataset/Morphological_and_phylogeographic_evidence_for_budding_speciation_an_example_in_hominins/13283507.

Competing interests

I declare I have no competing interests.

Funding

I received no funding for this study.