Pond chironomid communities revealed by molecular species delimitation reflect eutrophication

Abstract Farm ponds, a valued habitat for freshwater organisms, are being negatively affected by the recent changes in the environment as well as anthropological activities. In these ponds, biodiversity researchers have tended to focus on species that prefer natural habitats and/or can be identified based on morphological characters. In contrast, this study focused on the insect family Chironomidae, which is widely distributed from clear to polluted waters of ponds, but is hard to identify morphologically as an aquatic larva. We adopted DNA barcoding and molecular species delimitation to identify every single specimen of quantitative collections. From bottom sediments of 17 ponds in summer in the Banshu Plain of Japan, a total of 62 species were delimited based on the DNA sequences of the mitochondrial COI region. Chironomid communities from these ponds were classified into four groups in a two‐dimensional ordination of multivariate analysis (NMDS). One of the dimensions was well correlated with the gradient of eutrophication, while another dimension was not clearly assigned to any general feature of the environmental gradient, but rice cultivation could possibly be involved.


| INTRODUC TI ON
Farm ponds in Japan originally constructed for agricultural irrigation are now at the brink of disappearance not only due to shrinking local agricultural production, but also in order to prevent flooding of the ponds due to abnormally heavy rainfall. They are, however, habitats for endemic aquatic flora and fauna, and thus sources and refuges for aquatic organisms that can also be found in lakes, rivers, and wetlands (Takamura, 2012). They are highly valued in terms of conserving freshwater biodiversity even with the on-going changes.
The structure, size, water inflow and outflow, influx of inorganic and organic substances as well as the surrounding landscapes of the ponds have given rise to a wide variety of pond environments. These respective environments are habitats that actually or potentially nurture a unique set of aquatic flora and fauna (Casas et al., 2012;Cereghino et al., 2008;Chester & Robson, 2013;Fuentes-Rodriguez et al., 2013;Takamura, 2012).
A large number of farm ponds, most of which are/were smallscale, have been created in a seasonally dry temperate climate, such as the coastal area of the Setonaikai Sea (hereafter the Setonaikai Coast) in Japan. Researchers have focused on their value as a habitat of biodiversity in Japan as well as in other areas with such ponds (Casas et al., 2012). They have focused on the study of vascular plants, insects like odonates, hemipterans, and beetles, and vertebrates like fishes, annulans, and frogs (Fukumori et al., 2016;Iwai et al., 2017;Natsumeda et al., 2015;Usio et al., 2017). In such studies, most of the taxa studied are sensitive to environmental deterioration, now in progress among farm ponds, and thus they can be indicators of well-conserved waters. However, these species have been disappearing in highly deteriorated waters and indicate the deterioration by their absence (Ito et al., 2020). There are other taxa containing species that can inhabit those waters and contribute to ecosystem function through their presence. Including them in studies of pond ecosystems would provide a more detailed and vivid picture of pond biodiversity.
Adding to this, those taxa widely focused on in studies of pond biodiversity have often been ones for which their taxonomy is accessible through morphological identification. However, there are quite a few taxa which are not easily identified that abundantly appear in farm ponds. They are also an indispensable part of the pond biodiversity and may play a role in the ecosystem inside and outside the pond. One excellent way of revealing their taxonomic entity is the molecular identification of those organisms, that is, DNA barcoding (Hebert et al., 2003(Hebert et al., , 2016. This method of research will enhance recognition of the value of biodiversity in ponds and that of the ponds themselves in the arena of the freshwater biodiversity landscape. Integrating these two points together, chironomids are one of the representative taxa that should be studied among a wide range of pond environments. A large number of chironomid species are observed in ponds (Figure 1). They are one of the major groups of animals that occur in freshwater habitats like farm ponds, paddy fields, rivers, and lakes (Casas & Langton, 2008;Fuentes-Rodriguez et al., 2013;Medeiros & Quinlan, 2011). The species composition of chironomids varies drastically between different waters, so they F I G U R E 1 Adult males of chironomids. They were collected on the shores of the study ponds and identified to species morphologically, but not on DNA barcodes. have been regarded as good indicator organisms for freshwater environments (Lindegaard, 1995). For example, an index was introduced for a variety of lakes based on the composition of indicator chironomid species collected from the lakes, reflecting the level of eutrophication observed there (Wiederholm, 1980).
Aquatic chironomid larvae are found throughout the seasons in freshwaters that have a wide variety of water quality and substrate composition and can be collected with simple gear like an Ekman-Birge grab sampler in lakes and ponds or a Surber net in rivers and streams.
So, they are good materials for investigating the actual state of the freshwater environment. But one of the difficulties of understanding the chironomid fauna in the water is identifying all chironomid species as larvae correctly from their morphologies. Based on the morphological keys and diagnoses that have been developed worldwide or for the specific local area (e.g., Andersen et al., 2013;Japan Chironomid Workshop, 2010), a number of chironomid species can be identified as larvae, but most species are hard to identify. Morphological features the larva have are often not taxonomically informative or require F I G U R E 2 Locations of the study ponds on the Banshu Plain. The ponds are outlined with thick lines. Pond numbers are the same as in Table 1 painstaking measure/high magnification and expertise to distinguish.
However, recent developments in DNA barcoding (Hebert et al., 2003(Hebert et al., , 2016 have offered a more objective and effective way to perceive the taxonomy of morphological character-poor creatures like larval chironomids, given a wealth of DNA barcodes (species-specific DNA sequences) for identifying biological species on a molecular basis.
As stated before, among farm ponds densely spaced on the Banshu Plain of the Setonaikai Coast, biodiversity research has been intensively undertaken since the 1990s (Takamura, 2012). They are based on species identification of organisms collected or observed in and around ponds with the assistance of taxonomical specialists proficient in morphological identification. However, many taxa that are not easily identified by morphological taxonomy have remained un-

| Study ponds
Twenty study ponds were selected for sampling chironomids from among farm ponds in the eastern part of the Banshu Plain  to compare chironomid communities from different environmental conditions in the same year. Of these ponds, 17 ponds were ultimately included in this study (Figure 2), while the other three ponds were omitted due to the poor collection of chironomids.
Their landscape surroundings were regarded as six arable, six forest, and five urban ponds (Table 1), but all of the locations were in the Satoyama (the diverse mosaic of agricultural and nonagricultural lands: Kadoya & Washitani, 2011;Washitani, 2001) area and contained all of the three elements to various degrees. They were small, up to 2 ha in area, and as deep as 6 m in maximum depth. They could be separated into two types. One was the excavated type, which was formed with natural or artificial excavation on a relatively flat landscape, and the other was the embankment type, which was built by embanking a valley. Note: Pond types are excavated (Exc) and embanked (Emb). Dpth: maximum bottom depth of the pond (m), IL: organic content of the bottom substrate as ignition loss, BGAr: phycocyanin concentration of water (RFU: relative fluorescent unit), chla: chlorophyll a concentration of water (µg/l), DO: dissolved oxygen (mg/l), Secci: water transparency (m), spcond: specific conductivity (mS/cm), SS: suspended solid (mg/l), TN: total nitrogen (mg/l), TP: total phosphorus (mg/l), wplantR: sum of the cover scores for water plant, bank: ratio of pond perimeter covered with artificial bank, Trapa: cover score for Trapa japonica, city, water, rice, woods, and field: ratios of area covered with urban area, open water, rice paddy, forest, and agricultural field except for rice paddy.

| Chironomid collection
The The remains were kept in polyethylene bags and brought back to the laboratory in coolers on ice. The chironomids were sorted on ice in order to restrain degradation of DNA and preserved individually in 1.5 ml plastic tubes in a freezer for DNA analysis. Morphological identification was performed during the sorting based on the morphological keys (e.g., Andersen et al., 2013;Japan Chironomid Workshop, 2010), and a small number of larvae with unique morphological characters were identified as species.

| DNA extraction and sequencing
Chironomid DNA was extracted in lysis buffer (1 mM

| Species delimitation based on DNA sequences
The COI DNA sequences of the chironomid specimens were collapsed to haplotypes on the FaBox platform (Villesen, 2007).  (Table S1).

| Ordination of chironomid communities
Chironomid communities of the 17 farm ponds were classified by nonmetric multi-dimensional scaling (NMDS) using the R program package vegan (Oksanen, 2015) with the application of the similarity index of Chao et al. (2006). Indicator species were screened for the classified groups using the package labdsv (Roberts, 2016).   The substrate was heated at 360 ˚C for 2 hr (Salehi et al., 2011).
All of the environmental variables (Table 1)  The PCA analysis was performed with the "prcomp" function in the R. Land use ratios (forest (wood), urban area (city), ponds, rivers, and lakes (water), rice paddy (rice), and agricultural field except for rice paddy (field)), ignition loss as well as T. japonica cover and the ratio of artificial bank were logit-transformed before the NMDS and PCA analyses. If the divided ratios were 1 and 0, they were replaced with 0.99 and 0.01, respectively, in the transformation.

| RE SULTS
In all 17 farm ponds surveyed, 48 species were delimited from 177 haplotypes (Table S1). This species delimitation was based on 421 DNA haplotype sequences out of 786 chironomids from total 20 ponds ; Tables S3 and S4, Figures 3 and 4). The chironomid community compositions of 17 farm ponds (Table S2) F I G U R E 4 Species delimitation results by PTP on a nonclock tree reconstructed by MrBayes software. A cluster of branches continuously colored in red represents a delimited species (molecular operational taxonomic unit: MOTU). A singleton also represents a single species. Each haplotype sequence is specified with the specimen ID of each sample individual. Specimen IDs and species names are in Table S4 were ordinated in a two-dimensional space and classified into four groups: A, B, C, D (Table 1, Figure 5). The stress value was 0.1064, indicating the ordination result was reasonable (Zurr et al., 2007).
The four-group classification gave a slight, but visible, hump on the graph line in Figure 6 of the Calinski criterion and so was judged as the most likely (Calinski & Harabasz, 1974), although it was not much more so than the five-group classification ( Figure 6).
Group A was comprised of four ponds (Kashitanirokugo, Kuriyanigo, Hiroharamuko, and Tosaka). Eleven chironomid species were associated with this group, but no species was a definite indicator species. Pond numbers and abbreviations of species names are the same as in Table 1 and Table S1, respectively

| D ISCUSS I ON
The present 17 pond chironomid communities were classified into four groups according to the compositions of delimited species.
These results clearly show that there were different types of chironomid communities among the ponds which were different in the composition of surrounding land use. As this study is unique in adopting DNA barcoding and offering a well-objective taxonomy of molecular species delimitation, it is worth mentioning what species is unique to the pond environments.
The second axis of the ordination was correlated with the eutrophication gradient. A similar gradient has been reported for the chironomid communities in lakes and ponds of the eastern Canadian Arctic (Medeiros & Quinlan, 2011). There was no definite indicator chironomid species of the clear water ponds (group A), but Tanytarsus oyamai was the indicator of the turbid water ponds (group C). This species is commonly found in Japanese rice paddy fields (Takamura, 1993(Takamura, , 1996Takamura & Yasuno, 1986) and prefers water of high specific conductivity (>90 µS/cm: Kawai et al., 1998).
Propsilocerus akamusi, the indicator species of the group B ponds, is common to Japanese eutrophic lakes (Iwakuma, 1987(Iwakuma, , 1992Takamura & Iwakuma, 1990;Yamagishi & Fukuhara, 1971), but this species did not characterize group C, the typical eutrophic-water chironomid community. P. akamusi is notorious for breaking out in eutrophic lakes. Actually, this species might also have another aspect of habitat preference that is not simply on the eutrophic-oligotrophic gradient. The group B ponds were negatively characterized with rice paddy cover in the 50-m buffer zone around the ponds. Rice paddies are habitats for many chironomid species (Darby, 1962;Takamura, 1993), but P. akamusi is not included among them. This species might be fairly sensitive to any chemicals like pesticides applied in and around rice paddies.
As shown in Figure 6, the classification of the pond chironomid communities into four groups was not fully conclusive, although it seemed to fit well into this two-dimensional ordination. Among the classifications of the three, four ( Figure 5), and five groups, no identical groups were recognized, but P. akamusi was considered as an indicator species in any classification. On the contrary, T. oyamai, the indicator species of group C, was regarded as such only in the four-group classification (Takamura, unpublished). So, P. akamusi is strongly representative of a certain type of chironomid community, which may be negatively related to the rice paddy landscape.
One of the important points made in this study is to identify every single specimen of chironomids from quantitative sampling on a molecular basis, even if it did not fully succeed. This type of work is laborious, still possesses methodological limitations in sampling (Shelton et al., 2016), and involved low percentages of successful sequencing (DNA barcoding) at some ponds in this study. In doing so, the specimens, the DNA of which had been sequenced, were delimited to species even though their scientific names were not fully determined due to the shortage of DNA barcode references, difficulties in molecular species delimitation (Tang et al., 2014), or more or less putative delimited species (Zhang et al., 2013). The DNA sequence information collected from these specimens somehow describes the actual states of populations and communities and will be the basis for future species identification or taxonomical reviews.  Table 1 That could provide a deeper understanding of the biodiversity under investigation together with studies on morphological taxonomy and ecology.
The present species delimitation was originally made for the chironomid samples collected from bottom substrates and aquatic plants in the summer and fall of 2012 with the procedures of GMYC: Generalized Mixed Yule Coalescence (Fujisawa & Barraclough, 2013) and PTP: Poisson Tree Process (Zhang et al., 2013)  . Although their effectiveness has been proven for a variety of data sets, it is clear that they produce slightly different delimitation results for the same data (e.g., Tang et al., 2014). The present results showed the same thing. There were four cases where one GMYC species contained two PTP species and two cases where one contained three ; Tables S3 and   S4, Figures 3 and 4). As chironomid species are hard to identify in their larval stages based on morphological characters (Andersen et al., 2013;Sasa & Kikuch, 1995), it is difficult to determine which delimitation result matches the actual species composition better, based on the comparison between the molecular and morphological identifications. In this study, we selected the result which produced the lowest number of taxa. This decision could be tested by multilocus delimitation, thorough morphological identification of larvae backcasted from the present delimitation, and/or identifying adult chironomids reared from future larval collection.
It is worth considering recent advances in species delimitation based on DNA sequences. They have relied more on multiplelocus data rather than single-locus data (e.g., Brix et al., 2018;Lin et al., 2017;Luo et al., 2018;Vitecek et al., 2017). Single-locus data, especially that of the most frequently used gene COI, have a nonnegligible deficiency in delimiting species with gene flow. So, species delimitation based on a larger number of loci is ideal, especially for closely related species (Dupuis et al., 2012;Luo et al., 2018). There were some delimited species that might contain two or more closely related species in this study. For example, Microchironomus tener was delimited as a single species by GMYC, but as three species by PTP (Tables S3 and S4, Figures 3 and 4).
The present collection of chironomid larvae came from 17 ponds situated within an area of about 30 km by 30 km. Chironomids are generally not regarded as actively dispersing insects, but they are likely to be easily dispersed on wind and water currents (Armitage, 1995). In this regard, the collection of each species could be supposed to have come from a population or a meta-population.
So, the samples and their phylogenetic relationship may well reflect the coalescence processes supposed to be proceeding in the area.
The clump of short branches at the tip of the phylogenetic tree ( Figures 3 and 4) can be seen as a species population entity in which the coalescence process has proceeded. This branching structure seems to clearly demonstrate the threshold level where the processes of speciation and coalescence are discriminated from each other. If this were not the case, the discrimination process central to the species delimitation models would not have worked efficiently.
This implies that the present collection of chironomids is ideal for molecular species delimitation.
The present results on diversity of chironomid species composition between the farm ponds were observed almost ten years ago.
Now the environmental conditions in and around the farm ponds have changed along with the trend stated in Introduction. Although it is the object of future research how chironomid communities have responded to the change, it is noteworthy that the toxic effects of neonicotinoids have been increasingly noticed and investigated not only for terrestrial arthropods like bees, but also for aquatic insects (e.g., Hladik et al., 2018;Morrissey et al., 2015;Takeshita et al., 2020).
The association of chironomid community ordination with rice paddy land use may suggest the effect, while no pesticides were monitored in this study. As discussed by Imai et al. (2016) and Usio et al. (2013), Usio et al. (2017), the style of pond management and people's (farmers and nonfarmers) concerns on the activity is also worth considering for the conservation of pond biodiversity. In both points, monitoring the status of biodiversity in ponds is a key in assessing and making a feedback to the conservation. DNA barcoding and molecular species delimitation would make a great contribution.

ACK N OWLED G M ENTS
We thank Toshikazu Kizuka and Noriko Takamura for helping to design this study, and Yoko Oikawa and Shiori Okuda for sample sorting and genetic analysis. We also thank the pond managers for granting permission to study the farm ponds and the water-

CO N FLI C T O F I NTE R E S T
The authors declare no competing interests.

DATA AVA I L A B I L I T Y S TAT E M E N T
DNA sequences of COI mitochondrial gene of delimited specimens were deposited in DNA Data Bank of Japan (LC494722-LC495141).