Constraining uncertainty in the timescale of angiosperm evolution and the veracity of a Cretaceous Terrestrial Revolution

Abstract

Through the lens of the fossil record, angiosperm diversification precipitated a Cretaceous Terrestrial Revolution (KTR) in which pollinators, herbivores and predators underwent explosive co-diversification. Molecular dating studies imply that early angiosperm evolution is not documented in the fossil record. This mismatch remains controversial. We used a Bayesian molecular dating method to analyse a dataset of 83 genes from 644 taxa and 52 fossil calibrations to explore the effect of different interpretations of the fossil record, molecular clock models, data partitioning, among other factors, on angiosperm divergence time estimation. Controlling for different sources of uncertainty indicates that the timescale of angiosperm diversification is much less certain than previous molecular dating studies have suggested. Discord between molecular clock and purely fossil-based interpretations of angiosperm diversification may be a consequence of false precision on both sides. We reject a post-Jurassic origin of angiosperms, supporting the notion of a cryptic early history of angiosperms, but this history may be as much as 121 Myr, or as little as 23 Myr. These conclusions remain compatible with palaeobotanical evidence and a more general KTR in which major groups of angiosperms diverged later within the Cretaceous, alongside the diversification of pollinators, herbivores and their predators.

Keywords: Bayesian analysis; Cretaceous Terrestrial Revolution; angiosperms; divergence time; fossil record.

Figure 1

RAxML tree estimated from the 83 genes and 644 taxa of tracheophytes. The major angiosperm lineages and grades are highlighted: ANA grade (red), magnoliids (green), monocots (yellow), Ceratophyllales (pale blue), basal eudicots grade (pink), Dilleniales (orange), superasterids (purple) and superrosids (blue). Species names and bootstrap support values are indicated in Supporting Information Fig. S3.

Figure 2

Summary tree of tracheophytes showing fossil calibrations. Calibrations are represented for 52 nodes, consisting of (>) soft minimum (closed red dots) or both ([min, max]) soft minimum and soft maximum (open red dots). Calibrated nodes are numbered as in Supporting Information Fig. S2. Justifications for these minima and maxima are provided in Notes S1 and an overview in Table S3. The dagger symbol shows a species that is extinct. The tree has been scaled to time on the basis of the minimum constraints.

Figure 3

The effect of calibrations on posterior divergence time estimates of major groups of tracheophytes and angiosperms. (a) Summary chronogram for tracheophytes (including two lycophytes, two ferns, eight gymnosperms and 64 orders of angiosperms) with terminals collapsed to represent angiosperm orders showing divergence time estimates. Nodes are drawn at the posterior means obtained and horizontal bars represent 95% high posterior density (HPD) credibility intervals (CIs). Estimates were obtained using the HKY85 + Γ5 substitution model, independent rates model (IR), with the 83 genes subdivided into three partitions: 1^st and 2^nd codon positions for plastid genes; 1^st and 2^nd codon positions for mitochondrial genes; and nuclear RNA genes. Five nodes are connected (purple open dots) across the analyses to facilitate comparison: tracheophytes (n = 645), seed plants (n = 647), angiosperms (n = 648), eudicots (n = 655) and monocots (n = 1193). (b–d) Calibration, prior and posterior densities for three angiosperm nodes in the tracheophyte phylogeny. Colouring relates to the calibration strategy as in (a). The phylogeny with clade names is provided in Fig. 6. Nodes in parentheses are numbered as in Supporting Information Fig. S2.

Figure 4

Sensitivity of time estimates to the number of partitions, rate model, birth–death process, exclusion of lycophytes + ferns and fossil calibrations. The posterior mean times (black dots) and 95% credibility intervals (CIs) (red lines) of 643 nodes under calibration strategy A (SA), independent rates (IR) model, and gene alignments and three partitions are plotted against (a) estimates using one partition, (b) mixed partitions, (c) autocorrelated rates (AR) model, (d) birth–death parameters adjusted to generate a tree with long internal branches and short tip branches (BD1), and (e) large node ages with nodes close to the root (BD2), (f) excluding ferns and lycophytes (EP), (g) calibration strategy B (SB), (h) calibration strategy C (SC), (i) calibration strategy D (SD) and (j) calibration strategy E (SE).

Figure 5

Branch rate differences inferred from competing calibration strategies. All rate differences are plotted as positive regardless of whether they represent rate accelerations or decelerations. (a) Absolute and (b) proportional rate differences based on calibration strategy A (SA) that does not force an Early Cretaceous diversification of crown‐angiosperms (but remains compatible with this scenario). (c) Absolute and (d) proportional rate differences based on calibration strategy E (SE), which forces an Early Cretaceous diversification of crown‐angiosperms. Key early angiosperm nodes are labelled. Note that the eudicot crown is an outlier in all, but early angiosperm clade rates fall within the bounds exhibited by other, younger nodes in the tree, indicating that the independent rates (IR) model can accommodate heterogeneous rates required by an Early Cretaceous diversification of crown‐angiosperms.

Figure 6

The time tree of tracheophytes encompassing uncertainty of calibration strategies. Holistic timescale for tracheophytes with terminals collapsed to represent angiosperm orders. Node ages are plotted at the posterior mean for calibration strategy A (SA), three partitions (3P), independent rates model (IR) and HKY85 + Γ5 substitution model. The node bars are composites extending from the minimum 2.5% high posterior density (HPD) limit to the maximum 97.5% limit across all calibration strategy analyses (excluding results from calibration strategy B). This timescale should be read in terms of the span of clade age uncertainty, not from the absolute position of the nodes, which are placed at an arbitrary midpoint. The interval of residual uncertainty associated with the angiosperm crown is highlighted.