Sun, ^* David L. Dilcher Shaoling Zheng, Zhekun Zhou

Angiosperm fruiting axes were discovered from the Upper Jurassic of China. Angiosperms are defined by carpels enclosing ovules, a character demonstrated in this fossil. This feature is lacking in other fossils reported to be earliest angiosperms. The fruits are small follicles formed from conduplicate carpels helically arranged. Adaxial elongate stigmatic crests are conspicuous on each carpel. The basal one-third of the axes bore deciduous organs of uncertain affinities. No scars of subtending floral organs are present to define the individual fertile parts as floral units, but the leaf-like structures subtending each axis define them as flowers. These fruiting axes have primitive characters and characters not considered primitive.

<HW_SUPPLEMENTAL<HW_AFFILIATIONG. Sun, Nanjing Institute of Geology and Palaeontology, Academia Sinica, Nanjing 210008, China. D. L. Dilcher, Florida Museum of Natural History, University of Florida, Gainesville, FL 32611, USA. S. Zheng, Shenyang Institute of Geology and Mineral Resources, Shenyang 110032, China. Z. Zhou, Kunming Institute of Botany, Academia Sinica, Kunming 650204, China.</HW_AFFILIATION
^* To whom correspondence should be addressed. E-mail: gsun@jlonline.comi; dilcher@flmnh.ufl.edu

It has been thought that angiosperms first appeared about 130 million years ago in the Lower Cretaceous (1, 2). There are several recent reports of Triassic, Jurassic, and lowermost Cretaceous-aged fossils identified as angiosperms (3-7), but none of these reports can be accepted as conclusive evidence for the presence of angiosperms. Many reports of early angiosperms are based on pollen, leaves, and wood with vessels, none of which are definitive characters of angiosperms. Some are based on flowers and fruits that are too poorly preserved to demonstrate ovules or seeds enclosed in the carpels. The unique character of angiosperms is that the ovules are completely enclosed in a carpel. Here, we describe such early angiosperm fruits collected from the Upper Jurassic "Jianshangou Bed" in the lower part of the Yixian Formation of Huangbanjiegou village near Shangyuan Town of Beipiao City, western Liaoning Province, northeast China

Fig. 1. Map showing the geographic location of the angiosperm fruiting axes Archaefructus liaoningensis gen. et sp. nov. Vertical lines represent Liaoning Province. Fossil localities are southwest of Beipiao

The Yixian Formation (8-12) consists of layers of volcanic rocks sandwiched between sedimentary rocks. The sedimentary rocks contain abundant freshwater and terrestrial fossils, including plants, bivalves, fish, conchostracans, ostracods, gastropods, insects, turtles, lizards, shrimps, dinosaurs, birds, and mammals that constitute the Jehol biota (13, 14). The Yixian Formation is about 2000 to 2500 m thick and is considered to be latest Jurassic in age (8-15). We classify our discovery as follows:

Division Magnoliophyta

Class Magnoliopsida

Subclass Archaemagnoliidae

G enus Archaefructus Sun, Dilcher, Zheng et Zhou, gen. nov.

Type-species: Archaefructus liaoningensis Sun, Dilcher, Zheng et Zhou, sp. nov.

Generic diagnosis: Reproductive axes branched or unbranched, bearing helically arranged fruits (follicles) on short pedicels. Fruits mature distally, occupying the distal two-thirds of an axis; carpels or stamens deciduous, leaving short peg-like pedicel bases on the proximal one-third of an axis. Fruits derived from conduplicate carpels commonly bearing three (two to four) ovules. Fertile axes are subtended by leaf-like structures (16). Details of the diagnosis of Archaefructus liaoningensis are presented in (17).

The fruits presented here are recognized as angiosperms on the basis of the ability to remove seeds completely enclosed within them. The occurrence of this angiosperm-defining character in Archaefructus is important because it demonstrates that Archaefructus has angiosperm affinities, and it establishes a benchmark in time for when the closed carpel is first found. This character occurs in combination with other reproductive characters, resulting in a new mixture of characters. This unique set of characters should change our understanding of the nature of the early angiosperm flower.

The pollen-bearing organs of Archaefructus are unknown. They were not present with carpels in the fossil material examined. The proximal one-third of each fertile axis has what appears to be pedicel bases that may have borne deciduous fruits or other organs such as stamens. Archaefructus may have been either unisexual (monoecious or dioecious) or bisexual. No pollen was found attached to any surface of the fruits or axes, and no angiosperm pollen has been isolated from the matrix. The only sterile organs associated with Archaefructus are two poorly preserved leaf-like structures (Fig. 2A). The lateral axis is borne in the axil of a leaf-like structure occurring on the main axis. Examination by epi-illumination and fluorescence epi-illumination of the surfaces of this fertile complex revealed some cellular detail of the epidermal cells covering the fruits and axes, but no pollen or evidence of scars of any deciduous organs were found except for the peg-like pedicels basal in each fertile axis. Therefore, there may have been deciduous floral organs of an unknown nature associated with these fruiting axes when they were young. These fossils are fruiting axes bearing individual conduplicate carpels (Fig. 2, A and B), and each axis should be regarded as originating from a floral unit (Fig. 2A). The elongate nature of the axes may have been more extended in the fruiting stage than at pollination. The crowded carpels at the apices suggest this (Fig. 2B). Also, in the young carpels the stigmatic tissue occupies proportionately more area, and the apical prominence appears to continue to enlarge as the carpel matures.

Fig. 2. Archaefructus liaoningensis Sun, Dilcher, Zheng et Zhou gen. et sp. nov. (A) Holotype, SZ0916; fruiting axes and remains of two subtending leaves. Scale bar, 5 mm. (B) Enlarged view of the carpels showing remains of the adaxial crest, abaxial venation, seeds in each carpel, and finger-like prominences. Scale bar, 5 mm. (C) Portion of a seed removed from a carpel, as viewed by scanning electron microscopy. Scale bar, 25 µm

The carpels of Archaefructus are closed in a conduplicate fashion, contain more than one ovule, and are clustered together. Subtending each "flower" is a leaf-like structure consisting of a petiole that terminates in a branched pattern of possibly three major veins extending into a crumpled leaf lamina (Fig. 2A). A few of the basal reproductive organs were deciduous at maturity while the subtending leaves and distal carpels remained attached. It is possible that the crumpled leaf-like organs subtending each fertile axis were colored or patterned in some way to attract the attention of insect pollinators. The stigmatic surface may have produced an exudate on which the dipterians, known from the same sediments, may have fed (12). It is also possible that the extended tips on the stigmatic crests of each carpel functioned similarly to those of Ascarina of the Chloranthaceae, which is wind pollinated (18). Thus, there is no single pattern of pollination biology present, as is found for specific fossil angiosperm taxa occurring during the latter Cretaceous (19), but both insects and wind may have been involved. Insect pollination offers a biological environment that would have contributed to an early and rapid diversification of the angiosperms.

Archaefructus has helically arranged carpels, and the placement of the two leaf-like organs suggests this pattern continued in the foliage. In contrast, many members of Gnetales (20), found in the Mesozoic, are characterized by oppositely placed leaves, branches, and reproductive organs. Probable fossils of Gnetales that co-occur with Archaefructus in the Yixian Formation include Chaoyangia liangii (21) and Eragrosites changii (22). These were both described recently as the earliest record of angiosperms. Chaoyangia liangii has ribbed stems with conspicuous nodes, each bearing two oppositely arranged leaves. The stems branch oppositely to produce a cyme-like pattern on which winged fruits or seeds are borne. These winged fruits or seeds are similar to those previously described as Gurvanella (23, 24) and have a distinct resemblance to the winged seeds of Welwitchia mirabilis. Chaoyangia liangii is an interesting fossil plant, but the ribbed stems, opposite branching, and winged fruits or seeds suggest that it has affinities with Gnetales rather than the angiosperms. It is unlike any living Gnetales, and careful analysis of the described specimen and additional material needs to be carried out. Before it can be accepted unequivocally as an angiosperm, the nature of the winged fruits or seeds must be clearly understood, and we conclude at this time that it most probably is an extinct genus of Gnetales.

Eragrosites changii is a name given to fossils interpreted as grass-like remains (22). These fossils also have reproductive organs borne on ribbed axes with distinct nodes that are oppositely branched, characters typical of Gnetales. In addition, the tightly crowded grass-like seed heads have oppositely arranged bracts that are reminiscent of the seed-bearing organs of Ephedra and Welwitchia. We consider this fossil grass to represent fossil remains of an extinct Gnetales, and it definitely is not an angiosperm.

Archaefructus presents a new set of characters not previously known in angiosperms. Typically the division Magnoliophyta (25) is used for angiosperms or flowering plants, and the class Magnoliopsida is used for the dicotyledons and Liliopsida for the monocotyledons. We suggest that a new subclass, Archaemagnoliidae, be constructed for angiosperms that do not conform to the character sets of any of the existing subclasses of the Magnoliophyta. This new subclass is characterized by flowers subtended by only a single leaf or leaf-like organ. Flowers consist of elongate receptacles bearing conduplicate carpels helically. The nature of the male floral organs is unknown at this time. Flowers appear to terminate axes and predate the evolution of any floral patterns. The subclass does not fit the concepts of "paleoherb" or of "eoangiosperm," as both represent collections of angiosperm taxa already more specialized and modified (26) than Archaefructus of the subclass Archaemagnoliidae.

Although Archaefructus fits the general plan of the "fundamental axis" for the primitive angiosperm (26), there are no subtending bracts present; the carpels, leaves, and branching are helical; and the development of carpels is conduplicate (plicate) (27) rather than ascidiate (27, 28). Some cladograms (28, 29) suggest that ascidiate carpels with one or two ovules are most primitive, on the basis of the occurrence of these characters in extant angiosperms such as Chloranthaceae. This family is schematically derived through a gnetalian ancestry based on these characters (28). Archaefructus does not support this proposed evolutionary scheme.

In extant angiosperms, ovules are formed on the inner surface of the carpel, which histologically is different from the outer surface. After fusion of the carpel, the ovules are enclosed and isolated from external environmental factors. Endress (27) has maintained that the so-called "open carpels" of some angiosperms are a myth because secretions produced by the inner lining of the carpels fill any gap. The pollen grain and the pollen tube are required to interact with the biochemical barrier as well as, in most carpels, the physical barrier to the male gametophyte presented by the closed carpel (30, 31). This important step in angiosperm reproduction is clearly well developed in the Upper Jurassic in Archaefructus. It allowed for incompatibility to develop between the male gametophyte and the carpel very early in angiosperm evolution.

Overall, Archaefructus looks more like a seed fern-type plant than like bennettitalian or gnetalian plants, which have received support as ancestral groups (26, 32, 33). The leaf-like nature of the fertile shoots, the helical disposition of the carpels, the conduplicate nature of the carpels with multiple ovules, and the subtending leafy structures are characters that would support the possible seed fern ancestry of Archaefructus. Gnetales are considered a sister group of the angiosperms, just as they might be thought of as a sister group of some of the Mesozoic seed ferns. The Mesozoic seed ferns are poorly understood and probably do not represent a natural group of plants. Many seed ferns became extinct during the Triassic or the Jurassic, and all became extinct by the mid-Cretaceous (34). Perhaps some lineages of Mesozoic seed ferns are the ancestors of the Mesozoic radiation of the angiosperms, explaining why Gnetales and angiosperms are often found to be sister clades.

Archaefructus is more than 85 mm long and consists of two fertile axes, which give rise to nearly 60 carpels and two leaves (Fig. 2, A and B). This compression-impression plant material was recovered by cleaving apart sedimentary layers of rock. The fossil is unlike the charcoalified remains, recovered by sieving, that have added much to our knowledge of early angiosperm reproduction (35, 36). Those flowers and fruits are minute relative to the material of Archaefructus. Thus, Archaefructus is a clear indicator that large reproductive axes of angiosperms existed early in angiosperm evolution, even if only a few have been recovered. This may suggest that the small angiosperm flowers and fruits of early angiosperms are derived and reduced to small sizes from an ancestor with large flowers.

By the mid-Cretaceous and into the lower Upper Cretaceous, a tremendous increase in angiosperm diversity appears in the fossil record (37-41). Nearly all of these fossils represent lines of evolution progressing toward extant taxonomic clades of angiosperms at the family or generic level (24). The evolution of modern angiosperm taxonomic groups thus seems to have transpired relatively quickly during the Lower Cretaceous.