The largest early-diverging angiosperm family is mostly pollinated by ovipositing insects and so are most surviving lineages of early angiosperms

1. Introduction

Pollination in ancient flowering plants is usually assumed to have involved insects with mouth parts for foraging on liquids or pollen. This is based on structural adaptations in fossilized flowers and insects from the Early-mid Cretaceous [1–3]. Most surviving early-diverging lineages are insect-pollinated [4,5], although pollination mechanisms are known in few of the 283 species of Amborellaceae, Nymphaeaceae (including Cambombaceae), Hydatellaceae, Schisandraceae (including Illiciaceae), Trimeniaceae and Austrobaileyaceae that make up the ANITA grade [6,7]. Pollen transport from anthers to stigmas has been studied in the field in only about 24 ANITA species, in several cases without successful documentation of both pollen removal from the anthers and stigma pollen receipt (table 1). Of the 17 species with confirmed pollen vectors, three or four are exclusively or mostly fly-pollinated, six beetle pollinated and seven can have their pollen transported from anthers to stigmas by either flies, bees or beetles (table 1). Such scarce field data may handicap our understanding of early-angiosperm pollination.

The most widely accepted hypothesis still is that early angiosperms are pollinated by ‘generalist' insects and tend to have ‘generalized' pollination systems [2,5,19–22]. ‘Generalized' pollinators are animals that forage on several phylogenetically unrelated plant species and often are mandibulate insects that feed on solid food, including spores or pollen [1]. Generalized pollination systems are documented for four species of Nuphar in which pollen is transferred by ephedrid flies feeding on exudates, but also pollen-collecting bees and tissue-chewing beetles (table 1).

The scarcity of pollination data for ANITA species is partly due to most of their species being tropical shrubs or aquatics, usually with nocturnal anthesis. Another problem is the scarcity of taxonomists capable of identifying tropical flies, microlepidoptera and beetles, which prevents assessments of possible species-specific pollinators. The most diverse ANITA family, the Schisandraceae, has at least 91 species and is thought to be mostly fly-pollinated, with some studies invoking pollen feeding as the reward ([21,23–26], but see [27]). The family comprises three genera, Illicium (49 species), Kadsura (15–20 species [28]) and Schisandra (27–29 species [29]), with the last two consisting of woody vines in tropical and subtropical Asia, except for two species that occur in temperate China (S. chinensis) and North America (S. glabra). Illicium species are shrubs to small trees and occur in Southeast Asia, except for seven species in southeastern North America, Mexico and the Caribbean ([30]; all 91 species and their ranges are listed in electronic supplementary material, table S1).

We studied pollination in Schisandraceae species from all three genera during nine field seasons (2008–2016), using close-up filming of gall midge behaviour, examination of midges with scanning electron microscopy (SEM), molecular clock-dated phylogenies for Schisandraceae and their pollinators, and barcoding of multiple samples of the pollinators associated with each species. Because male and female gall midges use chemoreception to detect sex pheromone and plant-produced chemicals [31], we earlier analysed the chemical composition of their floral scent and floral resin exudates in several species [27]. Further chemical analyses of floral scents and exudates were, therefore, not part of this study.

Our main questions were (i) what are the pollination mechanisms in Schisandraceae and what is the extent of flower and pollinator specialization, i.e. do pollinator species pollinate only one or several species of Schisandraceae and vice versa; (ii) what is the evolutionary time frame of any coevolutionary interactions; and (iii) is there evidence of ‘generalized' fly or beetle pollinations in Asian Schisandraceae [7,22,23]? We then place the Schisandraceae results in the broader context of other ANITA species to infer which pollination mechanisms predominate in the surviving ancient flowering plant lineages.

2. Material and methods

(a) Study species, study sites, and observation periods and methods

Between 2008 and 2016, we carried out nocturnal and diurnal observations on 21 species of Schisandraceae at the locations mapped in figure 1. The locations are widely distributed in tropical and subtropical South China and temperate east, central and west China, and Japan (one species). All of them are in well-protected nature reserves; their geographical coordinates are provided in electronic supplementary material, table S2, which also lists plant vouchers deposited in the herbarium of the South China Botanical Garden (IBSC). Observation times (in days) for all sites are listed in electronic supplementary material, table S2, and the methods used to study the behaviour, morphology and life cycle of the gall midges or beetles that visited the flowers are in the electronic supplementary material. To test if gall midges might be species-specific pollinators, we sequenced multiple midges per study species and site as well as midges from two or three populations of a few widespread species studied at multiple sites, and we also collected midges from species co-occurring at the same location (as shown in electronic supplementary material, table S2).

• Download figure
• Open in new tab
• Download powerPoint

(b) Co-phylogenetic analyses and molecular clock dating

Our molecular-phylogenetic analyses of the Schisandraceae and their flower-visiting midges as well as our molecular clock models and their calibrations are described in the electronic supplementary material.

3. Results

(a) Ovipositing gall midges (Cecidomyiidi: Resseliella) as pollinators of Schisandraceae

In all species of Schisandraceae studied here, flowers open at night (figures 2a,d and 3a–e), emit a strong odour when fresh and wilt after 2–3 days, sometimes 6 days. Illicium has bisexual flowers that are functionally female during the first night, male during the second; Kadsura has unisexual flowers with both sexes produced on each plant individual (monoecious); Schisandra also has unisexual flowers but plants dioecious. Stamens are free (Illicium) or massively fused (Kadsura, Schisandra). Electronic supplementary material, table S1 summarizes floral traits and pollination observations for the 91 species of Schisandraceae, including the 21 that we studied at the locations shown in figure 1.

• Download figure
• Open in new tab
• Download powerPoint

• Download figure
• Open in new tab
• Download powerPoint

All flowers were visited and pollinated immediately upon opening by female gall midges in the genus Resseliella (Cecidomyiidae, Diptera, assigned by DNA and morphology) that oviposit in flowers of both sexual stages and morphs. Electronic supplementary material, videos S1–S4 show midges pollinating flowers of Kadsura and Schisandra; after landing, midges walked around, probing the flower surface with their cerci to identify suitable oviposition sites. The cerci bear tactile setae and four conspicuous chemosensory setae that are from 11.7 to 20.4 µm long, depending on species (figures 2c,g,i; 4a). Pollen grains often become lodged between the cerci as seen in the SEM of S. henryi (figure 4a,g; electronic supplementary material, video S2).

• Download figure
• Open in new tab
• Download powerPoint

In the bisexual flowers of Illicium, tepals do no fully expand during the first night of anthesis (figure 2a,d), instead forming a chamber in which midges lay their eggs (figure 2e) and where their larvae feed (figure 2b,f). Egg numbers per flower ranged from 6.4 ± 1.5 (I. micranthum) over 12.9 ± 2.7 (I. lanceolatum) and 14.1 ± 4.9 (I. majus) to 40.5 ± 2.3 eggs and 39.5 ± 3.9 larvae (I. verum), thus being highest in species with a floral chamber, such as I. verum (electronic supplementary material, table S5, which also lists the sites of oviposition and resin secretion for each species). The midges' larvae feed on the resinous exudates, and in a parallel study, we provide a chemical analysis of the resin of K. coccinea (Discussion).

In the unisexual flowers of Kadsura and Schisandra, the first midges arrived at the virgin flowers as soon as tepals had spread sufficiently to create a small orifice, and sometimes, midges arrived even on buds that were about to open; they also sheltered in the flowers during the daytime (figure 3h of S. sphenanthera). In male flowers, midges deposited their eggs between the thecae of adjacent stamens (figure 3a,c,d,f,g,i); in female flowers, they deposited their eggs in the clefts between the clustered carpels (figure 3b,e; electronic supplementary material, table S5). Pollen release and stigma receptivity ended on the second day, and by the third day, tepals of both male and female flowers dropped to the ground as a unit, along with the developing midge larvae (figure 3i). Egg numbers in male and female flower resembled those in Kadsura (e.g. 17.6 ± 5.3 and 17 ± 4.9 in male and female flowers of S. sphenanthera; 14.5 ± 3.3 and 8.8 ± 2.0 in male and female flowers of Schisandra spec. Wang ZW 193; cf. electronic supplementary material, video S3), except for K. coccinea, which has enlarged inner tepals in which midge larvae develop.

(b) Population- and community-level midge/Schisandraceae interactions and their evolutionary duration

Sequencing of the COI arthropod barcoding marker in multiple larvae and adult midges per Schisandraceae species showed that most species have their own midge pollinator except for Illicium majus, which at the single location where it was studied, was pollinated by two midge species (electronic supplementary material, figure S3D–G, table S4 and figure S4 shows the complete midge sampling, including outgroups and identical sequences excluded from the molecular clock tree in figure 4c). Six species were studied at more than one site, and five of them were pollinated by the same midge species at each location, while one was pollinated by a different species at each of three locations (electronic supplementary material, tables S2 and S4).

Figure 4a shows the locations at which species were studied (numbered 1–19 as on the map in figure 1; electronic supplementary material, table S2 provides further details on the locations and study periods). Four locations had more than one Schisandraceae species occurring within metres of each other and with overlapping flowering times (electronic supplementary material, table S2), namely Emeishan (S. henryi, S. micranthum, S. rubriflora and S. sphenanthera); Hengshan (Schisandra spec. Wang 193, S. lanceolatum and S. sphenanthera); Tianjinshan (Illicium spec. Cui 184, llicium jiadifengpi and Kadsura coccinea); Daweishan (Illicium majus and Schisandra spec. Luo 761). In each such community, coexisting flower types differed from each other in structure and colour and were pollinated by midge species that differed in abdomen morphology (figure 4a). For example, the chemosensory setae of the midge that pollinates the widely distributed S. sphenanthera are 19.9 µm long (figure 4a), while those of the midge that pollinates the sympatric species Schisandra spec. Wang 193 are 17.3 µm long (figure 4a; electronic supplementary material, video S3). Schisandra spec. Luo 761 and S. henryi have similar flowers (both lack a flower chamber), but the chemosensory setae of the midge that pollinates the former are 17.1 µm long, while those of the midge that pollinates the latter are 19.1 µm long (figure 4a).

The geological times over which Resseliella midges may have coevolved with Schisandraceae can be inferred from the midge and plant chronograms (electronic supplementary material, figures S1 and S2, with error bars; the inferred highest posterior density ages for each node are also shown in figure 4b,c). Under a relaxed molecular clock, constrained with three internal fossils (electronic supplementary material), we inferred a crown age of the Schisandraceae of 135.8 Myr (79.7–268.2 Myr 95% HPD), with most species having evolved from 20 to 5 Ma. The Resseliella midges that pollinate the studied Schisandraceae stem from a most recent common ancestor that lived about 11.2 Ma, an estimate that needs to be regarded with caution because it is based on only 460 bp of the COI gene and because we only sequenced midges captured in Schisandraceae flowers, while related Resseliella might exist on other plants.

(c) First records of beetle pollination in Illicium species

The female- and male-stage flowers of Illicium jiadifengpi, I. macranthum and I. simonsi were never visited by egg-laying midges but instead by species of Eusphalerum (Staphylinidae: Omaliinae). The beetles were mating in the flowers, and during the flowers' male stage they appeared to feed on pollen, as is typical of Eusphalerum [32]. Eusphalerum beetles in the female-stage flower of I. jiadifengpi were covered with pollen and contacted the stigmas (electronic supplementary material, figure S3I,J).

In the Schisandraceae phylogeny, the sequenced two beetle-pollinated species (figure 4b, marked in turquoise) are not sister species, suggesting independent shifts from midge to beetle pollination. All three beetle-pollinated Schisandraceae have white tepals (electronic supplementary material, figures S3H and I; table S1 gives the flower colours of all Schisandraceae).

(d) Preponderance of pollination by egg-laying insects in the ANITA grade

Table 1 shows the pollen vectors of all ANITA grade species that have been studied. Of the approximately 40 such species, including the ones studied here, most rely on ovipositing flies. Beetle pollination is now documented for nine ANITA grade species (including our three Schisandraceae), but probably occurs in more species of Nymphaea, a genus of aquatics.

4. Discussion

Our study shows that most Schisandraceae are pollinated by female gall midges whose larvae feed on resinous exudates produced in wounds resulting from oviposition and larval feeding. As is typical of plant resins, Schisandraceae resins are a lipophilic mixture of volatile and non-volatile terpenes, with caryophyllene a main compound [27]. The gall midge genus Resseliella has about 50 described species that lay their eggs in the bark, cones or fruits of (mostly) Pinaceae or Rosaceae, with the larvae then feeding on resin [33,34]. The adults of cecidomyiid midges typically emerge synchronized with their host plants' phenology [35].

One species of Kadsura and two of Schisandra have been reported as pollinated by Resseliella adults feeding on pollen [21,24,26,36], but when we restudied these species (albeit at different sites than in the original papers; electronic supplementary material, video S4), we could not observe any pollen feeding. Pollen feeding is otherwise unknown in cecidomyiids [37]. Ovipositing, not pollen feeding, pollinators also match the finding that flowers of at least four Schisandraceae are thermogenic ([22]: I. floridanum; [38]: S. glabra; [39]: I. dunnianum, I. tsangii), with experiments showing that warming persists after the flowers' sexual function is over and is beneficial to larval development. Further work on this obligate midge pollination system is required to determine the chemical signalling that is likely to govern how resin gall midges finding their host plants.

Beetles have long been suspected to co-pollinate—along with cecidomyiid midges and other insects—the American I. floridanum and S. glabra [23,38], but ours are the first documented cases of beetle pollination in the family.

(a) Population- and community-level interactions

We only obtained DNA sequences from midges captured on Schisandraceae flowers and therefore cannot know where any non-Schisandraceae-ovipositing Resseliella would fall in the phylogeny. The much younger midge divergence times than Schisandraceae divergence times suggest that now extinct midges may have pollinated the earliest Schisandraceae, while ‘modern’ Resseliella may have only radiated onto Schisandraceae since the Pliocene.

Of the six Schisandraceae studied at more than one location, five were pollinated by the same midge species at each location; one was pollinated by a different species at each of three locations. Only one species was pollinated by two midge species at the same location, suggesting that most Schisandraceae have a single pollinator species, but that a few may ‘sample' the local Resseliella pool, rather than being strictly dependent on a one species. The gall midge/Schisandraceae system thus is an instance of diffuse coevolution sensu Janzen [40], where interacting guilds of plants and arthropods coevolve with each other and where strictly coinciding divergence times for all interacting partners are not expected [41].

(b) Preponderance of pollination by egg-laying insects in the ANITA grade

Of the approximately 40 ANITA grade species with documented pollen transfer vectors, most rely on ovipositing flies whose larvae feed on flower exudates; beetle pollination is documented for nine species (table 1 and our data). As Amborella trichopoda is sister to the remaining angiosperms, its mode of pollen transfer would be especially interesting to know; the sole study of its pollination so far found flies, beetles and microlepidoptera visiting the flowers, but was unable to detect a pollen vector during the 3 h of observation just after sunset [8].

Insects that oviposit in flowers typically distribute their eggs over multiple sites to avoid larval competition, and because larvae have specific feeding niches, oviposition needs to be tissue-specific. Pollination by egg-laying insects is, therefore, economic in terms of pollen grains/ovule, matching the minute flowers and tiny anthers of many ANITA grade species (also the absence of nectar spurs). In addition, many ANITA species occur in dense tropical vegetation, and reliance on nocturnal scent-oriented egg-laying female insects that search for suitable sites for their larvae's development may be favoured in such habitats. While more fieldwork on ANITA grade species is needed, it seems unlikely that the picture will change drastically because most of the 283 living ANITA species are Schisandraceae.

Data accessibility

All supporting data are included as additional files.

Authors' contributions

S.-X.L. designed the research, participated in fieldwork, and analysed the data, S.S.R. did all molecular clock work; S.S.R. and S.-X.L. wrote the manuscript; L.-J.Z., S.Y., Z.-H.M., and S.-X.L. performed experiments, conducted fieldwork and participated in the interpretation of the results. D.-X.Z. financially supported S.Y. and Z.-H.M.

Competing interests

The authors declare that they have no competing interests.

Funding

Financial support came from the National Natural Science Foundation of China (grants nos. 31170217 and 31370268 to S.-X.L.).

Footnotes

Electronic supplementary material is available online at https://dx.doi.org/10.6084/m9.figshare.c.3954091.