How deep is the conflict between molecular and fossil evidence on the age of angiosperms?

I. Introduction

The age of the angiosperms is a long-standing topic of debate. Beginning with Darwin, many botanists took the supposedly sudden appearance of diverse angiosperm leaves in the mid-Cretaceous as evidence that the group originated and radiated extensively before the Cretaceous in some area with no known fossil record. For example, Axelrod (1952, 1970) hypothesized that angiosperms originated in tropical uplands in the Permo-Triassic but only invaded lowland basins in the Cretaceous. In the 1960s, opinion began to shift toward the view that angiosperms originated not long before their Cretaceous appearance. Scott et al. (1960) and Hughes (1961) stressed the failure of palynologists to find angiosperm pollen in pre-Cretaceous rocks, despite nearly world-wide sampling, arguing that some pollen should have been transported into lowland basins even if angiosperms were restricted to the uplands. They also rejected reports of Triassic and Jurassic angiosperms that earlier authors had cited as support for a pre-Cretaceous origin. Others noted that the first Cretaceous angiosperm pollen was much less diverse than expected from earlier leaf identifications, and pollen morphological types appeared in an order corresponding to the course of pollen evolution inferred from comparative studies of extant plants (Doyle, 1969; Muller, 1970). Subsequent workers reinterpreted the record of angiosperm leaves (Hickey & Doyle, 1977) and flowers (Friis et al., 2011) as showing a similar pattern of diversification. Some authors took these observations as evidence that angiosperms originated in the Cretaceous, but others cautioned that they might allow an earlier origin if pre-Cretaceous angiosperms were at a low level of morphological diversification (Doyle, 1969; Muller, 1970).

By contrast, in the past three decades, a great number of molecular dating analyses, based on the divergence of DNA sequences of living plants, have supported a pre-Cretaceous origin of crown-group angiosperms (see Box 1 for definitions) – sometimes Jurassic, but sometimes Triassic or even Permian (see Magallón et al., 2015 and references therein). Late Jurassic ages might be consistent with the existence of low-diversity and ecologically restricted angiosperms (Feild et al., 2004), but older dates are hard to reconcile with the congruence of the fossil record and neobotanical ideas on the evolution of pollen, leaves, and flowers, which have been independently confirmed by molecular phylogenetic analyses (Doyle, 2012). There have been new reports of pre-Cretaceous angiosperm fossils, but these have been questioned on various grounds (Doyle, 2012; Herendeen et al., 2017).

Box 1. Definitions of phylogenetic and palynological terms

Crown group: the most recent common ancestor of a living clade and all its descendants, both living and fossil. All clades recognized in molecular analyses are crown groups.

Stem lineage: the evolutionary line connecting a crown group with its common ancestor with the most closely related living group (sister group).

Stem relative: an extinct group on or attached to the stem lineage.

Monosulcate: with a single elongate furrow-like polar aperture (sulcus) of thinner exine for pollen tube germination; the polar axis is defined by an imaginary line running through the center of the pollen grain and the center of the meiotic tetrad in which it was formed (see accompanying sketch of pollen tetrad and Fig. 1a–g).

Tricolpate: with three elongate furrow-like apertures (colpi, singular colpus) for pollen tube germination running along lines of longitude (relative to the polar axis as already defined; see accompanying sketch of pollen tetrad and Fig. 1k–m).

Tricolporate: with three colpi plus an internally differentiated pore or os (plural ora) in the middle of each colpus. The resulting apertures are described as compound (see Fig. 1n–p).

Triporate: with three round apertures. Pores may be simple or compound (with differentiated inner and outer apertures; see Fig. 1q,r).

Pantoporate, periporate, polyforate: with numerous round apertures scattered over the surface of the grain.

Disulculate, zonasulculate: with two furrows or one ring-like furrow, respectively, on or parallel to the equator.

Inaperturate: with no differentiated aperture (and usually a thin exine, in which the pollen tube can germinate at any point).

Most molecular studies have not addressed these conflicts directly, but recently Barba-Montoya et al. (2018) argued that they reflect deep flaws in interpretation of the fossil record. In this review we summarize paleobotanical evidence on the early history of the angiosperms, organizing our discussion around the particularly extensive pollen record. We show that an informed reading of the fossil record may be consistent with a later Jurassic origin of crown-group angiosperms, but it militates against an older origin, and proposed direct fossil evidence for the existence of angiosperms in the Triassic and Jurassic is either erroneous, highly questionable, or inconclusive. We then discuss potential biases in molecular dating analyses that may have contributed to the conflict between fossil and molecular data and consider briefly the future role of fossil data.

II. Patterns in the Cretaceous record

In this survey, we concentrate on the pollen record, which is far better sampled spatially and temporally than the record of other plant parts. Before the Aptian (Fig. 1), practically the only convincing angiosperm megafossils are from the Barremian Las Hoyas flora of Spain (Gomez et al., 2015) and the Yixian flora of northeastern China (Sun et al., 2002, 2008), which straddles the Barremian–Aptian boundary (Chang et al., 2017). Fossil flowers (usually mesofossils, in the millimeter size range, which often have pollen in stamens or on stigmas) are easier to associate with modern clades than pollen, but their record is still limited to relatively few formations, mostly Albian and younger in age. By contrast, there is an extensive pre-Aptian pollen record of angiosperms.

Some early workers took the first appearance of tricolpate pollen (now known back to the late Barremian) as the first definite record of angiosperms (e.g. Scott et al., 1960; Brenner, 1963). Tricolpate pollen (Box 1) certainly represents evidence for crown-group angiosperms, since it is the most securely established morphological synapomorphy of the eudicot clade, which includes c. 72% of living angiosperm species and is strongly supported as monophyletic by molecular data. Tricolpate pollen was modified within eudicots into derived types such as tricolporate (the most common type today) and triporate. However, many angiosperms have monosulcate pollen (Box 1), which has long been considered ancestral based on its association with other ‘primitive’ features and its occurrence in gymnospermous seed plants (Wodehouse, 1936; Takhtajan, 1959). Significantly, palynological work in the 1960s showed that the first fossil tricolpates are preceded by monosulcate angiosperm pollen (Doyle, 1969; Muller, 1970). Subsequently, the ancestral status of monosulcate pollen has been amply confirmed by molecular phylogenetics (Doyle, 2005). Eudicots are one of five clades making up the Mesangiospermae, which include 99.9% of angiosperm species, along with monocots, Magnoliidae, Chloranthaceae, and the aquatic genus Ceratophyllum. The remaining 0.01% constitute the basal ‘ANITA’ lines, namely Amborella, Nymphaeales (water lilies), and Austrobaileyales. All of these groups except eudicots have monosulcate pollen or nontricolpate pollen types thought to be derived from monosulcate (e.g. disulculate, zonasulculate, inaperturate; Box 1).

The concept of tricolpates as the first definite evidence of crown angiosperms was reaffirmed by Barba-Montoya et al. (2018), who took the well-nested position of eudicots as support for a long prior history of angiosperms. They recognized the existence of an earlier monosulcate phase but considered it phylogenetically ambiguous. However, this phase is actually quite extensive and informative. In the following sections we survey the monosulcate and subsequent phases and their phylogenetic implications. For a quantitative assessment of the congruence of the pollen record with molecular dating studies, we complement this survey with an analysis of the diversity of major pollen types through time predicted by a dated molecular tree.

1. Temporal and spatial patterns in the pollen record

The first indication of a monosulcate phase in the angiosperm record was the description by Couper (1958) of Clavatipollenites from the Barremian (upper Wealden) of England. This fossil differs from the monosulcate pollen of gymnosperms in having a columellar exine structure, with radial rods connecting the inner (nexine) and outer (tectum) layers (Doyle et al., 1975; Fig. 1). Later scanning electron microscope studies (Hughes et al., 1979) showed that Couper's (1958) material consisted of several types that differ in microsculpture, but they are all columellar and reticulate, with the tectum consisting of bridges (muri) linking the heads of the columellae to form a network. Many authors have used the name Clavatipollenites for finely reticulate pollen with spinules on the muri and a sculptured sulcus. Transmission electron microscope studies (Doyle et al., 1975; Walker & Walker, 1984) showed that Aptian pollen of this type from the Potomac Group of the eastern USA is also angiosperm-like in lacking the laminated, distinctly staining inner nexine layer (endexine) all around the grain in gymnosperms. Instead, endexine is restricted to the sulcus, and the rest of the nexine consists of ectexine (foot layer), like the columellae and tectum.

Intensive scanning electron microscope studies on the Wealden (Hughes et al., 1979; Hughes & McDougall, 1990; Hughes, 1994) showed low-frequency but diverse angiospermous monosulcates extending back to the Hauterivian, which vary in coarseness of the reticulum and microsculpture of the muri. A divergent element is Tucanopollis (‘Barremian-ring’ of Hughes, 1994), first described from the Barremian and Aptian of Brazil (Regali et al., 1974a,1974b; Regali, 1989), which has a continuous rather than a reticulate tectum but internal exine structure and sulcus sculpture like Clavatipollenites (Doyle & Hotton, 1991).

In the 1970s, it became clear that some aspects of the pollen record in other geographic areas differed from what is seen in England and the eastern USA. These differences indicate that there are migrational as well as evolutionary patterns in the record, but they do not contradict the general evolutionary scheme. The British and Potomac floras represent Brenner's (1976) Southern Laurasia province, which extends east to Kazakhstan and China. The data support a modified version of the poleward migration theory of Axelrod (1959), originally based on leaf floras. The picture is summarized in Fig. 2, with four paleogeographic maps showing important pollen sequences (selected because they are particularly well studied and well dated by marine fossils or by palynological correlations with marine sediments in the same province), and Fig. 3, with first occurrences in these sequences of monosulcate and tricolpate angiosperm pollen, plotted against paleolatitude and time. These sequences are only a fraction of those that are known, but most others are less intensively studied, less confidently dated, and/or cover only short stratigraphic intervals. Many of the publications involved are several decades old, but the picture is corroborated by more recent studies in both previously and newly investigated areas.

An important advance was recognition that tricolpate pollen (initially with reticulate sculpture) appears consistently earlier in Brenner's (1976) Northern Gondwana province, including the late Barremian and Aptian of Brazil (Brenner, 1976; Regali & Viana, 1989), Gabon (pollen Zone C-VII; Doyle et al., 1977), and Israel (Brenner, 1976, 1996). Some occurrences originally considered Aptian are now thought to be late Barremian, based in part on association with the distinctive pollen genus Afropollis, which appears in the dated late Barremian of several areas (Doyle et al., 1982; Hughes & McDougall, 1990; Doyle, 1992). Nonangiospermous dominants indicate arid tropical climates in some parts of this province (Classopollis, ephedroid pollen) and wetter conditions (more fern spores) in others (Brenner, 1976, 1996; Doyle et al., 1982; Mejia-Velasquez et al., 2018).

Significantly, however, Zone C-VII and older sediments in Gabon and Congo (Zones C-V and C-VI; Doyle et al., 1977; Doyle, 1992) and Brazil (Regali & Viana, 1989) contain many of the same monosulcate angiosperm types seen in the Wealden, including reticulate monosulcates and especially abundant Tucanopollis. Later, in Zones C-VIII and C-IX (Aptian), angiosperms are more diverse, including tricolpates with striate as well as reticulate sculpture. Consistent pollen sequences have been described from Egypt (Schrank, 1983, 1992; Penny, 1991; Ibrahim, 2002; Schrank & Mahmoud, 2002) and Israel (Brenner, 1996). Reticulate monosulcate pollen is known from the Valanginian of Israel (Brenner, 1996) and Italy (Trevisan, 1988), which was located at the northern edge of Gondwana.

In Asian paleoequatorial areas, pre-Albian floras are like those of Northern Gondwana in their nonangiospermous dominants (Smiley, 1970a,1970b; Li & Liu, 1994; Racey & Goodall, 2009). In South China, Clavatipollenites and tricolpates have been reported from presumed pre-Albian beds, plus tricolpates and tricolporates from the Albian–early Cenomanian (Li & Liu, 1994), but age control is poor.

The dynamics of relations between Southern Laurasia and Northern Gondwana are becoming increasingly clear. In England, Kemp (1968) reported the first tricolpates (with reticulate sculpture) in the marine early Albian, together with the distinctive Clavatipollenites rotundus group. Similarly, in well-dated Barremian through middle Albian marine sequences in Portugal, Heimhofer et al. (2007) found the first reticulate tricolpates and C. rotundus in the earliest Albian, joined later in the early Albian by striate tricolpates. Reticulate and striate tricolpates also occur in the late early Albian of Texas (Tanrikulu et al., 2018). In the Potomac Group, reticulate tricolpates and C. rotundus appear consistently in upper Zone I (Doyle & Robbins, 1977; Hickey & Doyle, 1977). However, Hughes & McDougall (1990) illustrated exceedingly rare earlier tricolpates (one grain in the late Barremian, one in the early Aptian), and Doyle (1992) reported two tricolpate grains in lower Zone I (presumably Aptian).

Although these data imply that very rare eudicots existed in Southern Laurasia since the late Barremian, the consistent early Albian appearance of reticulate and striate tricolpates, which occurred in Northern Gondwana since the late Barremian and early Aptian, respectively, appears to represent a major migrational influx of eudicots, possibly due to global warming (Heimhofer et al., 2005; Coiffard & Gomez, 2012; Zhang et al., 2018). The Potomac leaf record shows a consistent pattern. Zone I contains rare simple angiosperm leaves comparable to members of the ANITA grade, Chloranthaceae, magnoliids, and monocots, plus ternately lobed leaves. The latter have been compared with the basal eudicot order Ranunculales (see Jud, 2015), but the low number of vein orders, presence of mesophyll secretory cells, and absence of tricolpate pollen at some localities have led to suspicions that some of these leaves may not be crown-group eudicots (Doyle, 2012; Doyle & Upchurch, 2014). By contrast, Zone II (middle and late Albian) shows a variety of new eudicot leaf types, as discussed later.

The record in Northern Laurasia (Siberia, Alaska, Canada) and Southern Gondwana (southern South America, southern Africa, India, Australia, Antarctica) completes this picture by showing a still more delayed entry of angiosperms. In western Canada, tricolpates are first reported in the middle or late Albian (Norris, 1967; Playford, 1971; Singh, 1975). Angiospermous monosulcates appear at the same horizons or slightly earlier (early middle Albian: Playford, 1971). In Arctic Canada and Alaska, Brenner (1976) found no tricolpates until the Cenomanian, but they have been extended down into the late Albian in the Canadian Arctic Archipelago; angiospermous monosulcates are usually absent (Galloway et al., 2012). In Australia, Clavatipollenites appears in the (Barremian?) Aptian, and reticulate tricolpates in the middle or late Albian (Dettmann, 1973, 1986; Burger, 1993; Korasidis et al., 2016). In southern Argentina, Clavatipollenites and other angiosperm monosulcates occur in the Aptian and are joined by tricolpates in the early Albian (Archangelsky et al., 2009; Llorens & Perez Loinaze, 2016; Perez Loinaze et al., 2016). In the Antarctic peninsula, Clavatipollenites appears in the early Albian and tricolpates in the middle Albian–Cenomanian (Dettmann & Thomson, 1987).

These variations do not contradict the congruence of the sequence of fossil pollen types and ideas on evolution based on extant plants; they simply mean that certain lines immigrated later into some areas. The earlier absences of angiosperms are not a function of absence of suitable rocks, since there are older sediments in many of these areas with rich pollen and spore floras but no reported angiosperms (e.g. white circles in Fig. 2).

4. General evolutionary implications

Perhaps more than the record of angiospermous monosulcates, the stratigraphic record of tricolpate and derived pollen types argues against the long period of unrecorded diversification implied by molecular dating analyses. This is illustrated graphically by Fig. 4, based on a dated molecular tree derived from the data set of Magallón et al. (2015), which is representative of trees with a Triassic age of angiosperms. In this analysis we estimated the number of lineages with sulcate, colpate (mostly tricolpate), colporate, porate, and inaperturate pollen from states in living taxa using stochastic character mapping (see Supporting Information Methods S1; Table S1), juxtaposed with observed curves of pollen types in three Cretaceous sections. Estimated ages center on 220 Ma for angiosperms (Late Triassic), 150 Ma for eudicots (Late Jurassic), and 132 Ma for Pentapetalae (Hauterivian), comparable to dates of Barba-Montoya et al. (2018). The tree implies that colpates were already nearly as diverse as sulcates when they are first observed as fossils in the late Barremian, and more remarkably that colporates were as diverse as colpates, c. 20 Myr before the first tricolporate pollen is seen in the late Albian. It is difficult to attribute this mismatch to failure of palynologists to recognize earlier tricolpate and tricolporate pollen. It is possible to overlook monosulcate angiosperm pollen, which may require close examination to distinguish from gymnospermous pollen, but tricolpates and tricolporates are easily recognized, and palynologists have been aware of their relation to angiosperms since early in the history of the field.

These considerations show paradoxical congruence relations among molecular phylogenetic analyses, molecular dating, and the fossil record. The stratigraphic succession of pollen types is congruent with scenarios for character evolution inferred from molecular phylogenetics, but not with molecular dating analyses. Congruence between inferred pollen evolution and stratigraphy is expected only if there was a relatively short lag between evolution of successively more derived types and their appearance in the fossil record. If molecular dates of the sort in Fig. 4 are correct, one must ask why taxa with new pollen types waited patiently for tens of millions of years before entering the fossil record in the order they had evolved.

A related question is why angiosperms did not radiate in the Jurassic if they originated then, given that they expanded so dramatically in the Cretaceous. The answer cannot be that the whole suite of angiosperm innovations had not yet accumulated, because in that case Jurassic representatives of the angiosperm line would be on the stem lineage rather than in the crown group, and the conflict with molecular dates would remain. It is difficult to imagine extrinsic environmental factors that might have suppressed radiation of angiosperms as a whole, considering how quickly they occupied a wide range of climatic belts and local habitats in the Cretaceous (Doyle & Donoghue, 1993). However, as argued by Feild et al. (2004), if Late Jurassic angiosperms were like extant terrestrial members of the ANITA grade (Amborella, Austrobaileyales), which grow mainly in wet tropical to subtropical forest understory habitats, they might have been restricted geographically and inhibited from diversifying because most of the tropical belt was arid at that time. The observed Cretaceous rise of angiosperms might then represent the radiation of mesangiosperms. There are some Early Cretaceous fossils from the ANITA grade, such as Anacostia/Similipollis in Austrobaileyales (Friis et al., 1997; Doyle & Endress, 2014) and a growing number of Nymphaeales. However, Similipollis is a minor element in palynofloras, and it is possible that the nymphaealean line was still terrestrial in the Jurassic and did not become widespread until it later invaded aquatic habitats (Doyle & Endress, 2014).

III. Pre-Cretaceous angiosperm reports

There is a long history of reports of pre-Cretaceous angiosperms and their rejection for stratigraphic or morphological reasons (Axelrod, 1952, 1970; Scott et al., 1960; Hughes, 1961; Doyle, 2012; Herendeen et al., 2017; Wang, 2018). Recently, molecular dating studies have cited putative pre-Cretaceous angiosperms as support for early molecular ages for crown-group angiosperms (e.g. Barba-Montoya et al., 2018), and paleobotanists have taken molecular dates as enhancing the plausibility of pre-Cretaceous claims (e.g. Liu & Wang, 2017; Wang, 2018). It should be recalled that fossils relate to molecular ages of the angiosperms only if they belong to the crown group; stem relatives can be either older or younger than the crown node.

Barba-Montoya et al. (2018) suggested that paleobotanical arguments against pre-Cretaceous angiosperms are flawed because of reliance on absence of evidence for ‘key characters’, rather than evidence of their absence; formulation of key characters in ‘the increasingly out-moded parsimony-based phylogenetic framework’; and methodological biases in distinguishing stem- and crown-angiosperms. However, these issues are irrelevant in most cases. Bayesian inference performs better than parsimony in cases of long-branch attraction, but it would be unwarranted to dismiss results based on parsimony out of hand: both methods usually give congruent results with empirical data sets (Rindal & Brower, 2011; Coiro et al., 2018). No putative pre-Cretaceous angiosperms have been examined in an explicit phylogenetic framework of any sort (although parsimony was used to evaluate the putative Cretaceous stem fossil Archaefructus; Sun et al., 2002; Doyle, 2008), and there are few cases where the issue is whether fossils are stem relatives or crown-group members. Usually, disagreement concerns whether morphological features were misinterpreted or whether the fossils are related to other seed plant groups. Some have defined angiosperms typologically (plants with enclosed seeds) while neglecting evidence for the homology of the structures involved or characters of other organs. There are similar problems with some Cretaceous fossils (see Herendeen et al., 2017), but here we focus on pre-Cretaceous records (for more details of our interpretations of particular taxa, see Notes S1).

Several fossils cited by Axelrod (1952) and others as pre-Cretaceous angiosperms were later shown to belong to other plant groups that were unrecognized at the time. A classic example is Eucommiidites, a Jurassic and Cretaceous pollen type with three furrows that Erdtman (1948) compared with the tricolpate pollen of eudicots. Confirming earlier less conclusive indications (e.g. Couper, 1958; Doyle et al., 1975), Pedersen et al. (1989) and Friis et al. (2009a) associated Eucommiidites with male and female structures that they assigned to the new order Erdtmanithecales, which is apparently related to Gnetales.

A recent example of this sort is Schmeissneria, described from the Early Jurassic of Germany by Kirchner & van Konijnenburg-van Cittert (1994) as a ginkgophyte. It has short shoots bearing strap-shaped leaves with an even number of veins, attached female axes bearing units consisting of one or two seeds surrounded by a longitudinally ridged ‘cupule’, and male axes bearing sporophylls with a cluster of pollen sacs, which are not known attached but occur consistently in the same beds. Wang et al. (2007) and Wang (2010, 2018) reinterpreted the female units in the German fossils and Middle Jurassic material from China as flowers with a perianth and two fused carpels with enclosed ovules. Wang et al. (2007) rejected the ginkgophyte interpretation because Kirchner & van Konijnenburg-van Cittert (1994) had excluded Schmeissneria from all known ginkgophyte genera; however, this only meant it was a new genus and in no way excluded it from ginkgophytes as a whole, which it resembles in all the vegetative and male features listed. To this list, van Konijnenburg-van Cittert (2010) added smooth monosulcate pollen, which is typical of (though not limited to) ginkgophytes. Furthermore, the female units can be reinterpreted in ginkgophytic terms, with the carpels as seeds (with apical hairs, as argued by Wang, 2010), and the supposed enclosed seeds, which show no regular relation to the ridges of the cupule and no cellular detail, as resin bodies in the cupule wall. The cupule may correspond to the bivalved capsule of the ginkgophytic order Czekanowskiales, which differs in containing several seeds rather than one or two. Wang et al. (2007) and Wang (2010) described the female structures as having two locules separated by a septum, but identification of such internal features often requires three-dimensional preservation, whereas these fossils are nearly two-dimensional compressions. The reconstruction of Schmeissneria presented by Wang (2018) is most unangiosperm like, but it closely conforms to the morphology of a ginkgophyte group such as Czekanowskiales.

A similar case is Solaranthus, from the Middle Jurassic of China, which Zheng & Wang (2010) interpreted as a flower with numerous carpels on a peltate receptacle and reflexed tepals and stamens. Deng et al. (2014) synonymized Solaranthus with Aegianthus (a resemblance already noted by Zheng & Wang, 2010), which has peltate microsporophylls with pollen sacs below the polygonal cap and resin bodies in the cap. They considered Aegianthus a cycad, but because its microsporophylls are nearly radial, whereas those of cycads are bilateral, a more likely relationship may be with the ‘seed fern’ order Peltaspermales. The smooth monosulcate pollen would be consistent with either affinity. The supposed carpels appear to be resin bodies in the cap, whereas the stamens and tepals are clearly pollen sacs viewed at various angles. Their reflexed orientation would be typical for cycad or peltasperm pollen sacs but bizarre for parts of a flower. Wang (2018) discounted the interpretation of Deng et al. (2014) but offered no additional evidence for Solaranthus being anything other than a gymnosperm pollen organ.

Liu & Wang (2016) described Euanthus from the Middle Jurassic of China as a flower with sepals, petals, anthers, and a gynoecium with an apical style and pentamerous ovary. However, the texture of the supposed sepals and petals is distinctly woody, more like the cone scales of a conifer than perianth parts of an angiosperm, as illustrated graphically by Herendeen et al. (2017). The style may represent the cone axis where scales have fallen off, but it is not clear whether it is actually part of the specimen or just underlies it. The pentamerous receptacle appears to be a basal view of the broken cone axis and surrounding smaller scales, whereas the so-called anthers (Liu & Wang, 2016; Wang, 2018) have no clear morphological features. Overall, Euanthus is more readily interpreted as a fragmentary conifer cone than an angiosperm flower.

A similar case may be Nanjinganthus from the Early Jurassic of China (Fu et al., 2018), known as numerous compressions with a ‘perianth’ of appendages similar to those of Euanthus, with ridges suggesting robust vascular bundles. Ten specimens have a branched prolongation of the axis described as a ‘dendroid style’. This resembles an axis with spirally arranged appendages more than the style of any angiosperm. It recalls the male cones of some conifers, which have basal bracts resembling the ‘perianth’. Fu et al. (2018) rejected this interpretation because they found no pollen on the dendroid structure, but this could be due to loss at many points prior to burial. Alternatively, the ‘style branches’ may be degraded ovuliferous scales of a female conifer cone. Fu et al. (2018) interpreted the basal part of the fossils as an inferior ovary with one to three enclosed ovules, but the morphology of this area is obscure because of poor preservation and irregularity in the number and appearance of the supposed ovules.

Other proposed pre-Cretaceous angiosperms are less clearly related to any particular nonangiospermous group but show little evidence of homologies with angiosperm structures, or have anomalous features that cast doubt on such homologies. One is Xingxueanthus, from the Middle Jurassic of China (Wang & Wang, 2010; Wang, 2018), interpreted as an axis with bracts and axillary gynoecia with free-central placentation. However, the supposed axis of the flower (placenta) is perpendicular to the bract and parallel to the axis, not at an angle arising from the axil. The presence of a bract subtending the ‘gynoecium’ and the woody appearance of the ‘placenta’ could indicate that this fossil represents the ovuliferous shoot of a coniferophyte, but again it is highly compressed, and the morphological nature of its parts is obscure.

Two other fossils from the Middle Jurassic of China have herbaceous leafy stems. Han et al. (2016) described Juraherba as consisting of a corm-like stem bearing roots, linear one-veined leaves, and axes terminating in longitudinally ridged fructifications that contain seeds. One-veined leaves occur in a few angiosperms, such as Hydatellaceae in the Nymphaeales (as noted by Han et al., 2016), but they are more typical of lycophytes and conifers. The ridges appear to be narrow pointed appendages of a strobilus; the supposed seed lacks any visible structure. Yuhania (Liu & Wang, 2017) consists of a stem bearing linear leaves with five or six parallel veins and structures described as aggregate fruits. However, the putative fruits vary in size by an order of magnitude, and the one with the most visible structure (Fig. 2h) is not clearly attached. Liu & Wang (2017) and Wang (2018) interpreted the units making up the fruits as carpels subtended by bracts (contrary to the definition of an aggregate fruit as derived from one flower with free carpels), but the supposed bracts and carpels have a highly anomalous reflexed orientation, and the structures identified as seeds have no visible morphology. Better preserved specimens of these enigmatic plants are required to obtain enough detail for robust systematic conclusions.

A somewhat similar case is Sanmiguelia, originally based on pleated leaves from the Late Triassic of Colorado (Brown, 1956). Cornet (1986, 1989b) associated Late Triassic leaves from Texas with woody stems and male and female structures, the latter interpreted as flowers with a perianth and several carpels. He argued that the leaves are monocot-like in having two orders of parallel venation and cross-veins, but the low frequency and irregularity of the cross-veins suggest they may be artifacts of degradation and shredding of leaf tissue. The male organs are strobili that appear more ginkgophytic than angiospermous; the pollen is smooth and monosulcate, consistent with many possible affinities. The female organs are so highly compressed that features such as enclosed seeds are not demonstrable (Doyle & Donoghue, 1993). The leaf morphology might be most plausibly derived from a coniferophyte (ginkgophyte, conifer, gnetophyte) type with an even number of parallel veins.

A general problem is that the proposed Jurassic angiosperms known as isolated reproductive structures (Solaranthus, Euanthus, Nanjinganthus, Xingxueanthus) occur in compression floras that are rich in typical Jurassic ferns, ginkgophytes, cycadophytes, and conifers (e.g. Na et al., 2017; Pott & Jiang, 2017; Fu et al., 2018) but contain no leaves with distinctive angiosperm apomorphies (several orders of reticulate venation, etc.), whereas those known as nearly whole plants (Schmeissneria, Juraherba, Yuhania) have leaves with no angiosperm features. By contrast, in Cretaceous and Cenozoic impression and compression floras, flowers are far less common than leaves of the same taxa. If the Jurassic fossils were angiosperm stem relatives then they might not have typical angiosperm leaves, but they would also have no bearing on molecular ages. Furthermore, one would expect angiosperms to be more like each other in the Jurassic than in the Early Cretaceous; but, if anything, proposed Jurassic angiosperms are more disparate morphologically and lack common features that might provide a coherent picture of the first angiosperms.

There are pre-Cretaceous fossils that share clearer potential synapomorphies with angiosperms. These include the Late Triassic Crinopolles pollen group, first described by Cornet (1989a) from the Newark sequence of Virginia, which includes monosulcate pollen with columellae and a reticulate tectum. However, in well-preserved specimens, transmission electron microscopy shows that the nexine consists of laminated endexine of uniform thickness (Cornet, 1989a; Doyle & Hotton, 1991; Fig. 5a–d), as in gymnospermous pollen, whereas in Early Cretaceous and extant monosulcate angiosperms endexine is either lacking or nonlaminated, except under the sulcus (Doyle et al., 1975). This could mean that Crinopolles represent either a convergent gymnospermous group or angiosperm stem relatives. As already discussed, it is not resolved whether the ancestral tectum in angiosperms was continuous or reticulate. Reticulate–columellar monosulcates are also known from the Middle Triassic of the Barents Sea and Switzerland (Hochuli & Feist-Burkhardt, 2004, 2013), but their nexine structure is unknown.

Equally tantalizing is Phyllites sp. of Seward (1904) from the Middle Jurassic Stonesfield Slate of England (see also Cleal & Rees, 2003; Fig. 5e,f). This is a single ovate leaf with acrodromous venation (palmate venation with a midvein and arcuate lateral primary veins). Unfortunately, no cuticle or finer venation is preserved, so it is not known if there were several orders of reticulate veins. The major venation suggests a position in the crown group, since, in terms of parsimony, ovate leaf shape is ancestral but palmate venation is derived, arising in Nymphaeales, Piperales, some Laurales, and eudicots (Doyle, 2012).

IV. Potential biases in molecular dating

This survey indicates that the angiosperm fossil record is indeed difficult to reconcile with molecular analyses that date the angiosperm crown group as much older than the Cretaceous, and the conflict cannot be readily explained as a result of misinterpretation of the fossil record. Molecular dating analyses have sometimes been perceived as more trustworthy, because they supposedly overcome biases and weaknesses of the fossil record (e.g. Kenrick, 2011). However, molecular dating methods are not free from their own potential sources of bias, and not necessarily as accurate as desired, or as robust against violation of their many assumptions (Bromham et al., 2018).

Cases in which molecular analyses estimate a much older age than fossil evidence in other groups, such as mammals (dos Reis et al., 2012; O'Leary et al., 2013; Phillips & Fruciano, 2018), have led others to reinvestigate the assumptions of molecular dating. Indeed, even though new methods to deal with issues such as rate heterogeneity and topological error are constantly being developed, it could be that some biological problems are still outside the scope of our current implementations.

Many Bayesian studies dealing with the age of the angiosperms have tested the sensitivity of molecular dates to different prior assumptions or partitioning schemes (Foster et al., 2016; Barba-Montoya et al., 2018). However, other authors have directly investigated systematic biases that could lead to overestimation of the age of the angiosperms. Beaulieu et al. (2015) used a simulation approach to detect potential biases created by molecular rate heterogeneity linked to habit (herbaceous vs woody), as well as by the diversified sampling strategy necessarily employed in such a large group. Their results demonstrated that even for sequences generated on a simulated tree of angiosperms with a young crown-group age (140 Ma), relaxed-clock methods tended to overestimate the age of the crown group by as much as 70 Ma when high-rate, herbaceous clades are present near the base of the group. More puzzlingly, the age of the crown angiosperms was overestimated by 50 Ma even when simulations were conducted under favorable conditions for relaxed-clock methods.

Another explanation for this perplexing phenomenon was offered by Brown & Smith (2018), who identified a substantial issue with the implementation of node priors: the interaction between the many different user-defined priors and the topology-generated effective priors that pushed the ages towards a much older origin of crown-group angiosperms than expected from the user-specified priors.

These methodological and biological complications have been identified in placental mammals, where the fossil record suggests an explosive radiation in the Paleocene and only stem relatives are known from the Cretaceous, but molecular analyses place the radiation well back in the Late Cretaceous. In this case, Phillips & Fruciano (2018) identified a connection between the use of calibrations at the base of clades that show parallel shifts towards slow molecular rates (primates, proboscidians) and older inferred ages for crown placentals. Two clades used for calibration in angiosperms, Proteales and Fagales, may represent similar cases, since they both have unusually short branches on molecular phylograms.

These considerations could mean that the tempo and mode of molecular evolution during the origin and early radiation of the angiosperms are extremely hard to model. A deeper investigation of the implications of fossil-informed ages for the mode of molecular evolution in angiosperms could provide novel insights not only on the time of their origin, but also on mechanisms acting at the molecular level during their radiation.

V. Conclusions

This survey reaffirms the view that stratigraphic and morphological patterns in the Cretaceous angiosperm record are difficult to reconcile with molecular dating analyses that place the origin of the group long before the Cretaceous. Reports of pre-Cretaceous angiosperms do not offer serious support for older molecular dates, and most are clearly erroneous. These results, like those involving other taxa such as placental mammals (Phillips & Fruciano, 2018), underline the need to investigate whether current molecular methods have inherent biases that are responsible for conflicts of this type with the fossil record.

The Cretaceous record permits and may even favor some Jurassic diversification of the angiosperms, and the nearly universal molecular support for a pre-Cretaceous origin should give pause to a literal reading of the known record. It would be unwarranted to dismiss scenarios in which low-diversity ANITA-grade angiosperms were ecologically and geographically restricted in the Jurassic; for example, if they grew in wet tropical understory habitats like woody members of the ANITA grade (Feild et al., 2004). In this case, the observed rise of reticulate monosulcate pollen beginning in the Valanginian might mark the radiation of mesangiosperms. However, the amount of pre-Cretaceous diversification implied by earlier Jurassic, Triassic, or Permian dates would conflict with the Cretaceous record, in which successive new pollen types appear in the order of evolution inferred from molecular phylogenetic analyses, but much later than predicted by molecular dating.

Our review demonstrates that the fossil record provides what appears to be fundamental evidence on the timescale and pattern of the origin and early evolution of angiosperms. This underlines the continued importance of expertise in paleobotany and related fields. Many of the erroneous interpretations of pre-Cretaceous fossils discussed here are due to misunderstandings of basic plant morphology. From the viewpoint of molecular dating alone, paleobotanical expertise is pivotal for successful calibration, especially now that the incorporation of fossil information using well-curated morphological matrices is becoming common practice in the broader field (Gavryushkina et al., 2017). Furthermore, discovery of pre-Cretaceous crown-group angiosperm fossils would confirm molecular evidence that angiosperms are older than their currently accepted record, but recognition of angiosperm stem relatives could be vastly more significant for understanding the origin of angiosperms and their distinctive features. Thus, better evidence on venation or cuticle of leaves such as Phyllites from the Stonesfield Slate or association of Crinopolles pollen with other plant organs could result in major breakthroughs in angiosperm paleobotany. However, full use of this information would require a broad and deep perspective on plant morphology, anatomy, and taphonomy.