Recent origin and diversification accompanied by repeated host shifts of thallus-mining flies (Diptera: Agromyzidae) on liverworts and hornworts

Abstract

Despite the vast diversity of phytophagous insects that feed on vascular plants (tracheophytes), insects that feed on bryophytes remain understudied. Agromyzidae, one of the most species-rich phytophagous clades in Diptera, consists mainly of leaf-mining species that feed on tracheophytes. However, a recent discovery of thallus-mining species on liverworts and hornworts within the Liriomyza group of Phytomyzinae provides an opportunity to study host shifts between tracheophytes and bryophytes. This study aimed to explore the origin and diversification of thallus-miners and estimate the pattern and timing of host shifts. Phylogenetic analysis of Phytomyzinae has revealed that the thallus-mining agromyzids formed a separate clade, which was sister to a fern pinnule-miner. The diversification of bryophyte-associated agromyzids since the Oligocene involved multiple host shifts across various bryophyte taxa. The diversification of the thallus-mining Phytoliriomyza may have occurred at the same time as the leaf-mining agromyzid flies on herbaceous plants, indicating a dynamic history of interactions between bryophytes and herbivores in angiosperms-dominated ecosystems.

1. Introduction

Embryophytes, or land plants, colonized land around 470–551 Mya [1,2], roughly at the same time as the colonization of land by arthropods [3]. The embryophytes comprise tracheophytes (vascular plants) and bryophytes (non-vascular plants), which are likely sister groups [4,5]. The bryophytes, which include mosses, liverworts and hornworts, are the oldest land plants, appearing between Late Ordovician to the latest Silurian. While each of the three bryophyte groups is monophyletic, the relationships between them have long been debated [6] and remain unresolved [4,7].

As soon as terrestrial ecosystems were established in the Silurian, arthropods started to consume land plants [8,9]. Adaptations for plant-feeding allowed insects to access potentially enormous resources at the base of the food web. Most phytophagous insect species are monophagous or oligophagous, and as a result, they can partition niches extensively [10]. Today, phytophagous insects, especially those feeding on angiosperms, are a major component of biodiversity [11]. The diversity of today's plants and insects was formed largely during the Angiosperm Terrestrial Revolution, which occurred approximately 100–50 Mya [12].

The predominance of angiosperm-feeding insects is often attributed to ecological speciation resulting from divergent selection exerted by host plants [13]. The role of ecological speciation in the diversification of phytophagous insects has been highlighted in recent years, and several evolutionary hypotheses have been tested using phylogenetic methods. The classical scenario predicts that the rapid diversification of insects is driven by their ability to circumvent the diverse secondary compounds in plants [13]. In many cases, the insects subsequently settle in new host groups over the course of evolution [14,15]. The trajectory of insect-host associations can be described as host shift, which involves an expansion of diet breadth followed by subsequent host specialization. These ecological changes can be coupled with speciation and further contribute to the diversity of phytophagous insects [14,15].

Bryophytes have a simple morphology and gametophyte-dominant life cycle. The bryophytes lack vascular tissue and the lignin biosynthetic pathway [16]. However, the water-conducting system in bryophytes is homologous to those of tracheophytes [17]. Bryophyte-feeding insects are generally scarce [18]. However, there are several lineages of obligate bryophyte-feeders in some insect orders, Lepidoptera (e.g. Micropterigidae) [19,20] (figure 1a,b), Diptera (e.g. Cylindrotomidae, Rhagionidae, Sciaridae) [21–23] (figure 1c,d), Coleoptera (e.g. Byrrhidae, partly Chrysomelidae) [24,25], Hemiptera (e.g. Coleorrhyncha) [26], and Mecoptera (Boreidae) [27]. The obligate bryophyte-feeding insects tend to specialize in one genus or species of bryophyte. For example, the Japanese micropterigids diversified into 20+ species on Conocephalum liverwort in the Paleogene (approx. 20 Ma) [19].

Figure 1.

Obligate bryophyte-feeding insects. Larvae and adults of three bryophyte-associated insect lineages with Triassic, Jurassic and Paleogene origins. (a,b) Neomicropteryx nipponensis (Micropterigidae: Lepidoptera) on Conocephalum sp.; (c,d), Spania sp. (Rhagionidae: Diptera) on Pellia endiviifolia; (e–g), Phytoliriomyza ricciae (Agromyzidae: Diptera) on Riccia nipponica.

Thalli of bryophytes lack physical defenses such as trichomes. Although the extent of antiherbivore defenses in bryophytes is not well understood, the low level of direct consumption of bryophytes is often attributed to their deterrent chemical compounds [18]. Notably, the oil bodies in liverworts act as a deterrent to herbivores, as they are a unique feature that helps to protect the thalli from being consumed [28]. The diverse chemical compounds in liverworts are different from those found in tracheophytes and also vary among different taxonomic groups of liverworts. This difference in chemical composition suggests that bryophytes and tracheophytes occupy distinct areas in the resource space [29]. The thalloid bryophytes are phylogenetically diverse and their secondary chemical compounds vary greatly both within and among species. This diversity of thalloid liverworts and hornworts thus may provide a wealth of ecological opportunities for insects, leading to diversification. The macroevolutionary process of bryophyte-feeding insects, however, is poorly understood.

Agromyzidae (leaf mining flies) is a dipteran family belonging to Schizophora. The evolutionary history of Agromyzidae is closely linked with angiosperms [30]. The earliest evidence of Agromyzidae can be found in the form of fossilized leaf mines from the Paleogene period, represented by Phytomyzites biliapchaensis [31]. Some groups of agromyzid flies (e.g. Phytomyza) underwent rapid radiations on angiosperms during the Oligocene [32,33]. This radiation was concurrent with the recent rapid radiation of Schizophora, which occurred ca. 65 Mya [34].

While most agromyzid flies are known for their leaf-mining behaviour on tracheophytes, a recent study has revealed a significant diversity of species that mine the thalli of liverworts or hornworts [35]. In Japan, 39 species of Phytoliriomyza species are known to only mine the thalli of a single genus, demonstrating their high level of host specificity [36]. The high diversity of bryophyte-mining species suggests that multiple agromyzid species can coexist on a single species/genus of bryophytes. For example, 15 species can be found on Conocephalum (Conocephalaceae), six on Reboulia (Aytoniaceae) and five on Marchantia (Marchantiaceae). The Phytoliriomyza species in Japan provides a unique opportunity to study how phytophagous insects can colonize and diversify on both tracheophytes and bryophytes, despite their phylogenetic differences.

This study aims to provide a phylogenetic framework for the bryophyte-associated agromyzid flies and to infer the evolutionary origin of bryophyte feeding. We explore the pattern and timing of host shifts, and tested whether: (i) the bryophyte-feeders and tracheophyte-feeders are phylogenetically clustered, (ii) the divergence events reflect the phylogenetic proximity of bryophytes, and (iii) the origin of thallus-mining. Based on the results, we evaluate whether the evolutionary pattern of host plant use in the Phytoliriomyza follows any of the macroevolutionary scenarios of host–plant driven speciation. Furthermore, we use molecular clock analysis to estimate the divergence times of bryophagy and gain insight into the chronological origin of phytophagy in Phytoliriomyza. Our study sheds light on the history and evolutionary dynamics of the bryophyte–herbivore interactions.

2. Material and methods

(a) . Ecological data

Phytoliriomyza is a diverse genus, containing 155 species described today. The biology of this genus has poorly been known, and host plants are known for 16 species that feed on bryophytes (3 spp.), ferns (7 spp.), and angiosperms (6 spp.) [30]. Recently, we discovered 39 species of bryophyte-associated Phytoliriomyza through extensive sampling effort in Japan [35].

There are 111 recorded species of thalloid liverworts (comprising 2 classes, 5 orders, 19 families and 31 genera) and 17 species of hornworts (1 class, 3 orders, 3 families and 6 genera) in Japan. The liverworts can either be leafy or thalloid in the forms of gametophytes. Thalloid liverworts are further divided into morphologically distinct groups, complex and simple thalloid liverworts. Among these, the larvae of Phytoliriomyza mine thalli of either complex thalloid liverworts or hornworts [35].

After gathering data on host plants, we assessed the host plant breadth by measuring the range of hosts used by each species. As host range is a quantitative trait, determining whether a species is monophagous or polyphagous depends on how we measure its niche breadth by the phylogenetic distances of its hosts. In line with previous studies [36,37], we treated herbivorous species feeding on a single plant family as specialists (monophagous), and those feeding on multiple plant families as generalists (polyphagous). This definition is typically applied to insects that feed on angiosperms, but we believe it can also be applicable to bryophytes. This is because thalloid bryophyte families are phylogenetically distantly related to one another and often comprise only a few species, indicating that they can be viewed as distinct resource spaces.

All the species analysed in our study were specific to either a single plant species or a group at the genus or family level. This predominance of specialists was not due to any sampling biases, as shown in electronic supplementary material, figure S2, where we collected flies from multiple localities throughout Japan.

(b) . Sampling and rearing

We targeted thallus-miners of agromyzid flies on all species of thalloid bryophytes in Japan. We collected thallus-mining agromyzids on 47 species of thalloid liverworts and 7 species of hornworts. Additionally, we also sampled leaf-mining agromyzids from the leaves of tracheophytes at various locations throughout Japan. In the laboratory, we reared the agromyzid larvae mining the thalli/leaves and collected the emerged adults and/or thallus/leaf-mining larvae. These specimens were stored in 99% ethanol and used for DNA sequencing. We ultimately obtained 201 agromyzid specimens, including 186 individuals emerged from bryophytes and 15 individuals emerged from tracheophytes.

(c) . DNA extraction and sequencing

Total genomic DNA was extracted from isolated legs of flies using the DNeasy Blood and Tissue Kit (Qiagen, Germantown, MD, USA). We sequenced fragments of three nuclear and two mitochondrial genes, including 16S rRNA (16S; approx. 550 bp), 18S rRNA (18S; approx. 1100 bp), 28S rRNA (28S; approx. 1150 bp), caudal (CAD; approx. 1600 bp) and cytochrome c oxidase subunit I (COI; approx. 650 bp). These fragments were amplified using polymerase chain reaction (PCR) with the primers listed in the electronic supplementary material, table S1. The purified PCR products were then subjected to direct sequencing at Eurofins, Tokyo, Japan.

The 28S and COI sequences of all 186 samples were used for species discrimination and examination of host specificity in the barcoding analysis (Dryad: Sequence data for barcoding). The sequences of these genes obtained in this study and additional sequences obtained from GenBank (listed in electronic supplementary material, table S2) were aligned using the program Muscle [38] implemented in Seaview [39,40] with the default settings. A maximum-likelihood (ML) analysis with the concatenated data of 28S and COI gene sequences was performed using the program RAxML v. 7.4.2 [41] implemented in raxmlGUI v.1.31 [42]. We selected GTRGAMMA for the substitution rate model and assessed the node credibility using 1000 bootstrap replications. The sequences of the remaining three genes (i.e. 18S rRNA, 16S rRNA, CAD) were obtained for 47 samples, including 37 Phytoliriomyza species that feed on bryophytes and 10 agromyzid species that feed on tracheophytes. These sequences were used for the subsequent phylogenetic analysis. The sequence data obtained in this study were deposited at the DNA Data Bank of Japan, and the DDBJ/GenBank accession numbers are listed in the electronic supplementary material, table S2.

(d) . Phylogenetic analysis and divergence time estimation

An aligned sequence matrix of 4387 base pairs was created, which included five gene sequences: 16S (524 bp), 18S (1063 bp), 28S (1306 bp), COI (658 bp) and CAD (836 bp). This matrix consisted of the aforementioned five gene sequences for 47 agromyzid samples collected in our study, and three or two gene sequences (primarily 28S, CAD, and COI) for 49 samples (taxa) of Agromyzidae and other Schizophora families obtained from GenBank (as shown in the electronic supplementary material, table S2). COI, 28S and CAD sequences were obtained for all but a few samples. These gene sequences are frequently used in phylogenetic studies of Schizophora, and the 28S and CAD gene sequences are particularly informative for our divergence time estimation due to their relative conservativeness. In the phylogenetic analysis, each gene was treated as a separate partition, and COI and CAD were further divided into three codon positions.

An ML analysis was performed using the IQ-TREE v2.1.3 software [43]. Sequence data were divided into a total of 9 partitions by gene (ribosomal DNA genes: 16S, 18S and 28S) and by gene and codon position (protein coding genes: COI and CAD). An optimal partitioning scheme and substitution models were determined by ModelFinder [44] implemented in IQ-TREE. The selected substitution models for optimized partitions were: TIM2 + F + R5 for COI pos1; TN + F + I + G4 for COI pos2; TPM3 + F + I + G4 for COI pos3; GTR + F + I + G4 for each of 16S, 28S, CAD pos1 + pos2, and CAD pos3; and TVM + F + I + G4 for 18S. To determine the node credibility, the SH-like approximate likelihood ratio test (SH-aLRT) [45] and the ultrafast bootstrap (UFBoot) approximation [46] were used, each with 1000 iterations. A node was considered credible if the results showed an SH-aLRT ≥ 80% and a UFBoot ≥ 95% [47]. For the estimation of divergence time, the best tree from the IQ-TREE analysis was converted into a clocked tree using the MCMCTree function in the PAML v4.8 software package [48]. The independent rates model (uncorrelated relaxed clock model) assuming a lognormal distribution of the rates was applied in the MCMCTree analysis to allow for heterogeneity of evolutionary rates among the branches. This model allows the evolutionary rate of each branch to be different and accounts for the variation in rates between different branches.

For calibration, the crown age of Agromyzidae was constrained to greater than 64.4 Ma, based on a leaf mine fossil of Phytomyzites [31]; the root age of the tree, which corresponded to the crown age of Schizophora, was constrained to <70 Ma in accordance with the divergence time estimation by Wiegmann et al. [34]. These constraints were set as a lower limit, L(0.644, 0.01, 0.01), and an upper limit, U(0.7), respectively, in the MCMCTree analysis. To obtain reliable results, two independent Markov chain Monte Carlo (MCMC) runs of 50 000 burn-in generations and 500 000 generations were performed, with sampling every 50 generations; the convergence of the estimated node ages within the two runs were confirmed.

3. Results

(a) . Diversity and host specificity of thallus-mining agromyzids

The thallus mines of thalloid liverworts and hornworts were discovered in various habitats in the Japanese Archipelago (figure 1e–g). From the mined thalli, 3096 agromyzid flies emerged. The adult flies were then used for morphological and molecular examinations (electronic supplementary material, figure S1), and 39 Phytoliriomyza species were identified. Of these, 36 species were associated with thalloid liverworts, and 3 species were associated with hornworts. All thallus-mining Phytoliriomyza were specific to a single bryophyte family. Most species were host-specific at the genus level of the bryophytes: five species on Marchantia, two species on Dumortiera, three species on Plagiochasma (Aytoniaceae), two species on Asterella (Aytoniaceae), six species on Reboulia (Aytoniaceae), one species on Wiesnerella (Wiesnerellaceae), fifteen species on Conocephalum (Conocephalaceae), three species on Riccia (Ricciaceae), one species on Folioceros (Anthocerotaceae), and one species on Megaceros (Dendrocerotaceae). Of the 39 species, 23 were specific to a single bryophyte species (e.g. three species were specific to Marchantia polymorpha). Only two species were not host-specific at genus level: P. phaerocerotis, which mines thalli of three hornworts' genera in Anthocerotaceae, Phaeoceros, Notothylas and Anthoceros, and P. plagiochasmatos, which mines thalli of two liverworts’ genera in Aytoniaceae, Plagiochasma and Asterella.

(b) . Origin and diversification of Phytoliriomyza through host shifts

The evolutionary pattern of host–plant use in bryophyte-mining Phytoliriomyza species was examined using an ML phylogeny based on a combined dataset of five gene sequences (figure 2). The Phytomyzinae clade was recovered with robust node support values (SH-aLRT = 100%, UFBoot = 100%), but the monophyly of Agromyzinae was not well supported (SH-aLRT = 17%, UFBoot = 71%). Phytomyzinae was divided into four clades: Cerodontha, Phytobia, Phytomyza and Liriomyza groups. Within the Liriomyza group, Phytoliriomyza was polyphyletic, while the 37 Phytoliriomyza species associated with bryophytes were monophyletic and had high support values (SH-aLRT = 94%, UFBoot = 94%). However, the species using the same host plant species did not form a clade. For example, 15 species of Conocephalum liverwort-feeders were polyphyletic, many nodes were not well resolved with low support values. The clade of the bryophyte-miners was sister to Phytoliriomyza felti, which is a fern-pinnule miner on Asplenium platyneuron; this clade was moderately well supported (SH-aLRT = 88%, UFBoot = 89%). The sister clade of these Phytoliriomyza species was a lineage comprising Liriomyza, Galiomyza, Metopomyza and Phytoliriomyza, all of which are associated with angiosperms.

Figure 2.

Clocked maximum-likelihood tree of Agromyzidae based on the combined dataset of five gene sequences (16S, 18S, 28S, COI, CAD) for 49 taxa of Agromyzidae, including 39 taxa of thallus-mining Phytoliriomyza of bryophytes and 10 taxa of leaf-mining agromyzids of tracheophytes, newly sampled in Japan. Numerals on the branches are node credibility values (%) obtained by the SH-like approximate likelihood ratio test (SH-aLRT) and the ultrafast bootstrap (UF Boot) both with 1000 replications. Red bars indicate 95% confidence interval of the estimated divergence time. Tips indicate sample numbers (corresponding to those in electronic supplementary material, electronic supplementary material, table S2) and species names, followed by the host–plant genus. Symbols show the host plant for the taxa feeding on bryophytes or ferns; the host plant is not indicated for angiosperm-associated species. Photos in the right corner show examples of 22 thallus- (a–p) and leaf- (q–v) mining agromyzid species (a–v).

The chronogram indicated that the split between tracheophyte-miners and bryophyte-miners took place during the Eocene (ca. 40.4–43.8 Mya). The diversification of bryophyte-associated Phytoliriomyza species accompanied by host shifts across bryophyte genera took place from the Oligocene to the Miocene. This diversification was observed in two clades: one associated with complex thalloid liverworts in Marchantiales, such as Marchantia, Dumortiera, Plagiochasma, Reboulia, Wiesnerella and Conocephalum (Clade I); another associated with hornworts and two genera of complex thalloid liverworts (Clade II). As we used only one calibration point, the divergence times may be older than what is indicated in our study. Nevertheless, the estimated appearance and subsequent diversification of Phytomyza in the Eocene is consistent with the age estimated by the molecular study [33].

4. Discussion

(a) . Diversification and host-plant shifts in Phytoliriomyza flies

The results of a phylogenetic analysis indicate that the shift from angiosperms to thalloid bryophytes only occurred once, indicating that these two groups of plants are distantly related in terms of their resource space. The monophyly of thallus-mining insects of thalloid liverworts and hornworts is not supported, suggesting that the division between liverworts and hornworts in resource space is not clear-cut, despite their chemical differences [49]. This outcome underscores the complicated history of Phytoliriomyza, as the observed pattern of host–plant utilization cannot be explained solely based on chemical differences.

We examine the roles of host-shifts in the macroevolutionary process of Phytoliriomyza in the context of scenarios of host–plant driven speciation, including speciation driven by oscillation [50], escape-and-radiate [11], host–plant specialization [51] and the musical chairs [37]. Our results indicate that Phytoliriomyza repeatedly colonizes new host plants and subsequently returns to previous hosts. The results are thus incompatible with the oscillation hypothesis which predicts that the generalist ancestors give rise to specialist daughters [50]. Although the order of host shifts is not fully resolved, they do not show clear host–plant conservatism (i.e. the use of related plant species by related insects) or phylogenetic tracking (i.e. the congruence of the phylogenies of herbivorous insects and their host-plants). These trends are not consistent with the escape-and-radiate hypothesis [11].

Among the above-mentioned four scenarios, our results are consistent with the hypotheses of ecological speciation driven by host–plant specialization. Further testing is needed to determine the correlation between the number of host shift events and diversification rates. Nevertheless, host shifts may be a significant driving force in the diversification of Phytoliriomyza. Although the species ranges of Phytoliriomyza did not exhibit clear allopatry (see Supplementary Material and electronic supplementary material, figure S2), it should be noted that non-ecological factors, such as geographical isolation, can be more important than host shifts in speciation events [52–54]. The complex geotectonic events of the Japanese archipelago during the Miocene [55], and the paleoclimatic shift during the Eocene–Oligocene transition [56] may have also played a role in their evolution and diversification.

(b) . Bryophyte–arthropod interactions in the angiosperm-dominated world

It has been believed that the bryophyte-feeding insects in the modern maintain the antique diet and habit choices. Their relationship with host plants thus is referred to as ‘ancient association’ [57]. Some of the bryophyte-feeding insect lineages, indeed, exhibit morphological stability and low species diversity. The origin of the mandibulate moths, Micropterigidae, the sister group of all other lepidopterans, dates back to the Triassic [58]. The fossil species of Litoleptis (Rhagionidae), thallus-miners of complex thalloid liverworts [22], is known from the Jurassic [59]. The moss bugs in the hemipteran group Coleorrhyncha (Peloridiidae) were present in the Permian [60]. The estimated origin of the moss beetles (Byrrhidae) is between the latest Triassic and Early Jurassic, with the oldest fossils dating back to the Late Cretaceous [61].

While fossils can provide profound insight into the plant–herbivore interactions in geologic time [62], bryophyte–arthropod interactions are poorly documented in the fossil record. The oldest evidence of external feeding on liverworts was found in Middle Devonian body fossils, making it one of the few records of herbivory in bryophyte fossils [63]. Most body fossils of bryophytes have been found as amber inclusions deposited during the Cenozoic era, and they are scarce in pre-Cenozoic rock formations [64]. The paleobryological studies are hindered by their small and cryptic nature [65], but possibly also due to the preservation potential [66,67].

The notion of evolutionary stasis in bryophyte–arthropod associations is undermined by increasing evidence of the dynamic history of bryophytes. The macroevolutionary history of the three groups of bryophytes, liverworts, hornworts and mosses, differs greatly [68]. The hornworts are a probable sister to a moss and liverwort clade, and maintained a low level of diversity [69]. The majority of the genera of complex thalloid liverworts appeared during the Cretaceous and have since steadily increased in taxonomic diversity [70]. Later, the turnover from gymnosperm-dominated to angiosperm-dominated floras, resulting in increased humidity and new ecological niches [12], led to the rise of liverworts and mosses in epiphytic realms [71,72].

Our study has provided a timeline for the evolution of the thallus-mining Phytoliriomyza. The bryophyte-mining Phytoliriomyza flies originated in the Eocene, and diversified in the Oligocene to Miocene accompanied by repeated host–plant shifts. This timing of origin and diversification is similar to that of Phytomyza, the agromyzid flies that mine leaves of herbaceous plants in temperate, open habitats [33]. During this time, there was a sharp increase in the fossil records of leaf mines made by agromyzids on angiosperm leaves [31]. These results suggest that the agromyzid flies experienced periods of rapid diversification during the Oligocene to Miocene on both bryophytes and angiosperms in temperate regions. With these cases, the herbivore insect community on bryophytes no longer conforms to the stereotyped scenario of the ancient association through evolutionary stasis. The bryophyte–arthropod interactions have been dynamically affected by the spurred diversification of angiosperms.

Data accessibility

Sequence data and electronic materials are available from the Dryad Digital Repository [73].

Authors' contributions

M.K.: conceptualization, data curation, formal analysis, investigation, writing—original draft, writing—review and editing; L.Y.: investigation, methodology, writing—original draft; Y.I.: writing—review and editing; T.S.: investigation, methodology, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This work was supported by the Japan Society for the Promotion of Science (KAKENHI #15370012, #18207002, #22247003, #22405009, #15H02420, #20H03321).