Diffuse Coevolution between Two Epicephala Species (Gracillariidae) and Two Breynia Species (Phyllanthaceae)

The diffuse coevolution between two moth species (Epicephala lativalvaris and E. mirivalvata) and two plant species (Breynia fruticosa and B. rostrata) is reported based on field observations and indoor experiments conducted in Hainan and Fujian, China. Study results showed that the two Epicephala species jointly pollinated the two Breynia species, which led to a unique obligate pollination mutualism of two−to−two species specificity. A single Epicephala larva exclusively fed on seeds of host plants and developed to maturity by consuming all six seeds of each fruit, whereas a fraction of intact fruits were left to ensure the reproduction of plants within the whole population. Larvae of the two Epicephala species are competitive for resources; the population of E. mirivalvata is much smaller than that of E. lativalvaris, which has resulted from the differences in the female ovipositor structures and oviposition mode. The life history of Epicephala species highly coincides with the phenology of Breynia plants, and different phenology of B. fruticosa resulted in the different life history of the two Epicephala species in Hainan and Fujian. The natural hybridization of two host plants, possibly induced by the alternate pollination of two Epicephala species, is briefly discussed.


Introduction
The obligate pollination relationship between plants and their seed-parasitic pollinators is perhaps one of the most specialized cases in mutualism between insects and plants [1][2][3][4]. Ehrlich and Raven defined coevolution as the interaction between two major groups of organisms with a close and evident ecological relationship, such as plants and herbivores [5]. In the known classical obligate pollination mutualism between figs-fig wasps and yuccasyucca moths, figs and yuccas depend exclusively on pollination of fig wasps and yucca moths respectively, while sacrificing some seeds for larvae of fig wasps and yucca moths to feed [6,7]. The obligate pollination mutualism between Epicephala moths of Gracillariidae and Glochidion plants of Euphorbiaceae was discovered in 2003 [3]. Three pollination relationships were proposed [8]: (1) one Epicephala species obligately pollinates one Glochidion species, (2) two Epicephala species jointly pollinate one Glochidion species, and (3) two Epicephala species jointly pollinate two closely related parapatric Glochidion species. The most distinctive floral features associated with the pollination mode are the structures of the pistils and stamens: styles are usually reduced to entire tips and medially fused, or filaments and anthers are variously fused [9][10].
An obligate pollination relationship between Breynia vitis-idaea and an unnamed Epicephala species was discovered in 2004 [11]. This Epicephala species actively pollinated B. vitis-idaea and was observed loading numerous pollen grains on proboscis. Another Epicephala species on leaves of B. fruticosa was also observed carrying numerous pollen grains on proboscis, which led to a deduction that this species was an active pollinator of B. fruticosa [11]. Okamoto et al. [12] suggested the floral scent of Glochidion is one of the important key signals that mediate the encounters of the species-specific partners in the Glochidion-Epicephala mutualism. In tightly coevolved mutualism between Breynia vitis-idaea and Epicephala species, system-specific chemistry is not necessary for efficient host location to Epicephala moths [13].
By means of field observations and indoor experiments, we studied the phenology, the phylogenetic relationship and the morphological characters of the two Breynia species, as well as the morphological characters, the behavior, the life history, the phylogenetic relationship and the population size of the two Epicephala species in Hainan and Fujian. Based on related literature and results of our study, we discussed herein: (1) the impact of the different phenology of the allopatric B. fruticosa populations on the life history of the Epicephala moths in coevolution; (2) the influence of moth pollinators on the hybridization of Breynia species; and (3) whether the relationship between the pollinators is stable in this diffuse coevolution.

Life History of Epicephala Moths and Phenology of Breynia Plants
In Hainan, B. fruticosa has six fruit periods per year, lasting from the first period in early January to the sixth period in late November. Epicephala lativalvaris and E. mirivalvata have six generations correspondingly, lasting from the first generation in middle January to the sixth generation in late December (Table  S1).
In Fujian, both B. fruticosa and B. rostrata have 4-5 fruit periods per year, which lasted from the first period in early June to the fifth period in late April of the following year. A large number of flowers withered in winter due to absence of pollinators. A small amount of flowers that were pollinated during middle to late November were dormant and no seeds were formed until middle or late March of the following year, and the fifth fruit period lasted from middle to late April. Both E. lativalvaris and E. mirivalvata have 4-5 generations correspondingly, which lasted from the first generation in middle June to the fifth generation in middle May of the following year. Larvae that pupated during early to middle November could emerge into adults, pollinate mature female flowers and lay eggs of the fifth generation during middle to late November. Larvae that pupated in late November overwintered in pupas, and they began to emerge in late April and disappeared in middle May. Pollinated flowers and Epicephala eggs in flowers were dormant until middle or late March of the following year, and the adults of the fifth generation emerged in middle May (Table S2). We smelled flower fragrance of both B. fruticosa and B. rostrata at night only, which suggests that they were blooming at night.
A single larva of both E. lativalvaris and E. mirivalvata needs to consume all six seeds within a fruit to develop into maturity. Mature larvae (the last instar) gnawed a hole in the fruit wall and exited, then produced cocoons and turned into pupae on leaves of hosts or the ambient plants. Pupal stage of E. lativalvaris lasted 9-15 days, and pupal stage of E. mirivalvata lasted 9-12 days. Adults of both Epicephala species could survive 3-5 days by sucking nectar for nutrition ( Figure S1). The result of the hand-pollination hybridization experiment showed that there are no reproductive barriers between B. fruticosa and B. rostrata. 90% B. fruticosa female flowers (n = 50) with pollen grains from B. rostrata developed, so did 94% B. rostrata female flowers (n = 50) with pollen grains from B. fruticosa (Table S7).

Behavior of Epicephala Adults
Epicephala moths became active after dark, which was in accordance with the blossom of B. fruticosa and B. rostrata flowers. Female adults of the two Epicephala species actively collected pollen from male flowers and pollinated female flowers of both Breynia species at night. Female Epicephala adults pushed proboscis tip into the dehiscent pit of calyx lobes, rubbed against stamen column to collect pollen grains, and sometimes they pushed the whole proboscis into the pit (Figure 2: A). During pollen collection, female adults intermittently spread proboscis several times to get more pollen grains. They only collected pollen grains in one male flower each time effectively, which could pollinate more than ten female flowers in the same or different plants.
Epicephala adults repeatedly rubbed proboscis against stigma to deposit pollen grains on all stigmas ( Figure 2: B, C). They would pace back and forth to pollinate all flowers on one branch, and female flowers were pollinated in this way as to pollinate most efficiently.
The two Epicephala species exhibited similar behavior of pollen collection and pollination, but their oviposition behavior was visibly different. Epicephala lativalvaris females bent abdomen to basal calyx on ventral surface and inserted ovipositor through calyx lobes and ovary to lay eggs into ovary tissue ( Figure 2: D-F). Epicephala mirivalvata females bent abdomen to basal calyx on dorsal surface to deposit eggs between ovary and calyx lobes ( Figure 2: G-I). After oviposition, Epicephala moths often gave complementary pollination to make sure there are enough pollen grains for female flowers. On rare occasion, they would not lay eggs if disturbed.

Number of Epicephala Eggs and Consumption of Breynia Seeds
Epicephala lativalvaris laid eggs by piercing through calyx lobes and ovary wall, leaving oviposition scars on fruits and calyx lobes. We could acquire their egg amount by counting scars on each fruit. In all sampled fruits (n = 1269 in B. fruticosa, n = 867 in B. rostrata), 72.0% in B. fruticosa (n = 914) and 88.6% in B. rostrata (n = 768) with oviposition scars loaded E. lativalvaris eggs (Table  S3). Epicephala mirivalvata laid eggs between calyx lobes and ovary wall. Therefore their eggs could only be observed by dissecting female flowers. Only 2.6% female flowers in B. fruticosa (n = 2) and 4.0% in B. rostrata (n = 5) loaded E. mirivalvata eggs in all dissected flowers (n = 76 in B. fruticosa, n = 126 in B. rostrata) (Table S4). We collected 28 E. lativalvaris adults and two E. mirivalvata adults in the field, and reared 109 E. lativalvaris larvae and four E. mirivalvata larvae from fruits. We observed E. lativalvaris visiting female flowers 35 times in the field, but E. mirivalvata twice only. The above statistical figures indicated that the population size of E. mirivalvata was much smaller than that of E. lativalvaris.

Phylogenetic Analysis of Epicephala Moths and Breynia Plants
We analyzed the COI genes of E. lativalvaris and Epicephala sp. ex B. fruticosa [11], and found that the pairwise distance between them is 0. Moreover, Epicephala sp. ex B. fruticosa would leave scars on host flowers after ovipositing, which is similar to E. lativalvaris. Based on the above information, we could conclude that the Epicephala species pollinating B. fruticosa in Kawakita and Kato [11] is E. lativalvaris.
The topological structure of the maximum-parsimonious (MP) tree and the maximum-likelihood (ML) tree of plants are completely consistent (Figure 3), which indicates that B. fruticosa and B. rostrata are highly surpported to be sister taxa. Both MP and ML trees of Epicephala moths ( Figure 4)

Diffuse Coevolution between Two Epicephala Species and Two Breynia Species
Diffuse coevolution firstly advanced and defined by Janzen (1980) [14]. He defined coevolution as an evolutionary change in a trait of the individuals in one population in response to a trait of the individuals of a second population, followed by an evolutionary response by the second population to the change in the first. Diffuse coevolution occurs when either or both populations in coevolution are represented by an array of populations that generate a selective pressure as a group [14].
Our study revealed that B. fruticosa and B. rostrata distributed in Hainan and Fujian were dependent on the obligate pollination of E. lativalvaris and E. mirivalvata, forming a mutualism of two2to2two species specificity. Two Epicephala species deposited eggs on female flowers before or after pollination, and their larvae exclusively fed on seeds of two Breynia species. The life history of two Epicephala species correlated with the phenology of two Breynia species; Breynia plants bloomed after dark, which coincided with the activity of Epicephala moths. According to the definition of diffuse coevolution and the classic examples of figs-fig wasps and yuccas-yucca moths, we confirm that diffuse coevolution exists between the two Epicephala species and the two Breynia species, which is firstly revealed in this study. Though a single Epicephala larva needs to consume all six seeds of each fruit to develop to maturity, enough intact fruits developed by Epicephala pollinated flowers are left to ensure the reproduction of the hosts within the whole population. The impact of the phenology of the allopatric B. fruticosa on the life history of Epicephala moths was obvious. Different climates in Hainan and Fujian led to different phenology of B. fruticosa, which resulted in different life history of Epicephala species. In Hainan B. fruticosa has six fruit periods without dormant stage, so do two Epicephala species (Table S1). While in Fujian, pollinated female flowers of B. fruticosa were dormant in winter, so a minority of Epicephala individuals overwintered in eggs in female flowers and a majority overwintered in pupas until hosts bloom in the following year (Table S2).
Breynia fruticosa and B. rostrata are sister taxa, which suggests that the pollinators they are sharing might have pollinated their ancestor, and that this pollination system maintained after the speciation of B. fruticosa and B. rostrata.

Hybridization of Two Breynia Plants Induced by Epicephala Moths
The study of the genetic structure of closely related figs and their associated pollinators in the Neotropical Region showed that  (Figure 1). Breynia rostrata was recorded to have distribution in Hainan [16], but we did not find it in this study. We only found three individuals with characters of both B. rostrata and B. fruticosa in Yingge Mountain, which indicated that the population of B. rostrata was either very small or has disappeared. The plants with double characters in Hainan may imply that B. rostrata once existed there and might have hybridized with B. fruticosa in nature.

Relationship between E. Lativalvaris and E. Mirivalvata
In Breynia-Epicephala diffuse coevolution mutualism, Epicephala lativalvaris and E. mirivalvata overlap in the ecological niche, and are competitive for resources. Epicephala mirivalvata has a much smaller population size than E. lativalvaris does and plays a secondary role in this mutualism. This is closely associated with female ovipositor structures and oviposition mode. Epicephala lativalvaris has an apically spine-shaped ovipositor ( Figure 5: A), which could pierce through calyx lobes and ovary wall to lay eggs into the ovary tissue ( Figure 2: D-F). So the external influence imposed on egg hatching could be greatly reduced, and the newly hatched larvae could instantly feed on seeds to ensure a higher survival rate. On the contrary, Epicephala mirivalvata has a clavate ovipositor with blunt apex (Figure 5: B), which could deposit eggs only between calyx lobes and ovary (Figure 2: G-I). So both eggs and newly hatched larvae were exposed to the influence of environmental factors. Moreover, the newly hatched larvae had to bore into the ovary, which could increase the mortal hazard and eventually decrease the population size. The phylogenetic analysis of Epicephala trees reveals that E. mirivalvata diverged firstly within all known pollinators of Breynia plants, that E. mirivalvata firstly established relationship with B. fruticosa and B. rostrata. However, its ovipositior structure and oviposition mode have resulted its inferiority in competition with E. lativalvaris. So it can be deduced that E. mirivalvata is likely to disappear or shift host to maintain population size, and E. lativalvaris may become the sole pollinator of B. fruticosa and B. rostrata in the future evolutionary course.

Conclusion
In Breynia-Epicephala diffuse coevolution, two morphologically and behaviorally distinct Epicephala species jointly pollinated two Breynia species, forming a new obligate pollination mutualism of two-to-two species speciality. We firstly describe this model of mutualism system in this study. Different phenology of B. fruticosa resulted in different life history of Epicephala species in Hainan and Fujian, and two hosts sharing pollination of the two Epicephala species has led to hybridization of Breynia plants. In this mutualism, two competitive Epicephala species overlap in the ecological niche. Epicephala mirivalvata is much smaller than E. lativalvaris in population size due to different oviposition mode, and therefore faces more selection pressure.

Studied Insects and Plants
Epicephala lativalvaris Li, Wang & Zhang, 2012 and E. mirivalvata Li, Wang & Zhang, 2012 are two small nocturnal moths that are distinctly different in morphology [17]. Females have a different oviposition mode due to having a differently shaped ovipositor ( Figure 5). However, they obligately and jointly pollinated B. fruticosa in Hainan and Fujian, and pollinated B. rostrata in Fujian.
Breynia fruticosa and B. rostrata are similar in morphology: male flowers ( Figure 6: A, B) have a funnel-shaped calyx with six dentations at apex; the three stamens fused into a column and concealed in calyx, which makes it hard to be touched by flower visitors except Epicephala moths; female flowers bear free calyx and styles, which is likely to be accessed by other flower visitors. The two Breynia species can be distinguished by having different calyx and stigmas: in B. fruticosa ( Figure

Methods
Breynia fruticosa flowered for the 10 warmer months of the year, fruited sparsely and irregularly in Hong Kong [18]. In order to acquire detailed phenological data of B. fruticosa and B. rostrata as to compare with the life history of E. mirivalvata and E. lativalvaris, we made a tracking observation of the flower and fruit developmental stages of B. fruticosa in Hainan, as well as of B. fruticosa and B. rostrata in Fujian. We randomly selected ten individuals of B. fruticosa in Yingge Mountain Nature Reserves, and five individuals of both B. fruticosa and B. rostrata in Tianzhu Mountain National Forest Park and Wanshi Botanical Garden, respectively. We recorded the development of flowers and fruits, and counted the number of flowers, the fruits and developing fruits of the selected plant individuals once a week.
To make sure whether the hybridization could happen between B. fruticosa and B. rostrata, we performed a hand-pollination hybridization experiment in Wanshi Botanical Garden. We respectively selected 50 non-pollinating female flowers of B. fruticosa and B. rostrata. We gathered pollen grains from male flowers of B. fruticosa and hand-pollinated female flowers of B. rostrata, and from male flowers of B. rostrata and hand-pollinated female flowers of B. fruticosa. We bagged these pollinated flowers with fine netting (0.5 mm mesh), and seven days later, we counted and calculated their developmental rate ( Figure S2).
We observed Epicephala moths during full anthesis and recorded their visiting behavior in detail, with particular attention paid to their nocturnal activities. We recorded the time that the moths spent on pollination, oviposition and pollen collection, and observed how they used their proboscis to collect pollen and pollinate flowers, and where they oviposited. We collected developing and mature fruits and put them in a cylindrical plastic box (8.5 cm612.0 cm) to rear Epicephala larvae. We then recorded how and when the larvae left fruits to cocoon and emerge.
The presence of oviposition scars and eggs on fruits was considered to have a one-to-one correspondence [11]. We randomly examined 23 flowers in Hainan (Yingge Mountain) and 179 flowers in Fujian (Tianzhu Mountain and Wanshi Botanical Garden) to search for eggs (Table S3) (Table S4).
Adult specimens examined in this study were collected from their hosts or reared from fruits in captivity. Dissection of female flowers of Breynia plants and female genitalia of Epicephala moths was conducted under an Olympus SZ11 stereo-microscope. Photographs were taken with a Canon G10 digital camera in the field and the illustration of genitalia were prepared with an Olympus C-7070 Wide Zoom digital camera.
We extracted genomic DNA of Epicephala moths from the fruits we reared, and extracted genomic DNA of B. rostrata from silicagel dried leaves collected in Tianzhu Mountain, Fujian. The method of DNA extraction followed the CTAB procedure [19]. We PCR-amplified the COI gene fragments using primers LepF1 and LepR1 [20], and PCR-amplified the matK gene fragments using primers 570F and 1710R [21].
We analyzed the phylogenetic relationship between E. lativalvaris and E. mirivalvata based on the mitochondrial cytochrome oxidase subunit I (COI) gene of 20 Epicephala sequences, which is widely used in the phylogenetic analysis of insects [9,22,23]. The sequences of E. lativalvaris, E. mirivalvata and E. relictella were newly obtained during this study and 17 sequences were adopted from the previous studies [8][9] (Table S5). Alignment of COI gene and matK gene were aligned using Mega v. 5.05. The phylogeny was rooted with Epicephala sp. ex Flueggea suffruticosa according to the phylogenetic analysis of Epicephala and Phyllantheae [10].
We analyzed the phylogenetic relationship between B. fruticosa and B. rostrata based on the maturase K (matK) gene, which has previously been shown to be particularly useful for phylogenetic analysis of the family and tribe [8,9,21,24]. Breynia rostrata sequence was newly obtained during this study and the remaining sequences were adopted from the previous studies [8,9,21] (Table S6). The phylogeny was rooted with F. suffruticosa according to the phylogenetic analysis of Epicephala and Phyllantheae [10].
Maximum-parsimony and maximum-likelihood analyses of COI and matK were performed using PAUP v. 4.0 b10 [25]. Both maximum-parsimony and maximum-likelihood heuristic searches were conducted with equal weight, 1000 and 100 replicates of random addition analyses respectively, and tree bisection-reconnection branch swapping. The robustness of maximum-parsimony and maximum-likelihood trees were assessed by non-parametric bootstrapping with 1000 and 100 replicates respectively.