Bloodmeal host identification with inferences to feeding habits of a fish-fed mosquito, Aedes baisasi

The mosquito, Aedes baisasi, which inhabits brackish mangrove swamps, is known to feed on fish. However, its host assemblage has not been investigated at the species level. We amplified and sequenced the cytochrome oxidase subunit I barcoding regions as well as some other regions from blood-fed females to identify host assemblages in the natural populations from four islands in the Ryukyu Archipelago. Hosts were identified from 230 females. We identified 15 host fish species belonging to eight families and four orders. Contrary to expectations from previous observations, mudskippers were detected from only 3% of blood-engorged females. The dominant host was a four-eyed sleeper, Bostrychus sinensis (Butidae, Gobiiformes), in Iriomote-jima Island (61%), while it was a snake eel, Pisodonophis boro (Ophichthidae, Anguilliformes), in Amami-oshima and Okinawa-jima islands (78% and 79%, respectively). Most of the identified hosts were known as air-breathing or amphibious fishes that inhabit mangroves or lagoons. Our results suggest that A. baisasi females locate the bloodmeal hosts within the mangrove forests and sometimes in the adjacent lagoons and land on the surface of available amphibious or other air-breathing fishes exposed in the air to feed on their blood.

www.nature.com/scientificreports www.nature.com/scientificreports/ similarity to a conger eel Uroconger lepturus (Congridae, Anguilliformes) in the GenBank achieved ≥99% similarity to a snake eel Pisodonophis boro (Ophichthidae, Anguilliformes) in our database. We confirmed that BLASTn searches for the 12S rRNA sequences from some of these samples against the GenBank database resulted in same-species identification (Pi. boro) with ≥98% similarity (AMA9 and AMA10, for example, Supplementary  Table S2). Accordingly, we concluded that Pi. boro was the correct host. We found considerable discrepancies among the sequences of both COI and 12S rRNA of a spaghetti eel, Moringua microchir (Moringuidae, Anguilliformes) within the GenBank database. Many of the 12S rRNA sequences from our samples matched one of them (accession no. LC020870) closely; thus, they were identified as M. microchir.
Combined with the GenBank database and ours (Supplementary Table S1), we could identify host fish species with ≥99% similarity from all 230 female mosquitoes analyzed ( Table 2, Supplementary Fig. S1), but for one sample, intermingled DNA sequences were obtained for both COI and cytochrome b regions, which we considered as originated from two different blood sources (Supplementary Table S2), most likely from Pi. boro and B. sinensis. This sample was excluded from further analysis.
Host assemblage. Although we identified 15 host species from four islands, host assemblages were quite different among the islands, except for between Amami-oshima and Okinawa-jima (Fig. 2, Table 2). We identified bloodmeal hosts from 46 female mosquitoes in Amami-oshima, and only three host species were found, with Pi. boro as the most frequent host (78%). About 20% of all females fed on B. sinensis, and the blood from a goby Mugilogobius sp. 'Izumi-haze' was detected in one female. A similar pattern was observed in Okinawa-jima; five host species were found, again with Pi. boro as the most frequent host (79%) and B. sinensis as the second most frequent (14%). Some fed on M. microchir (5%) and the blood from a mudskipper, Periophthalmus argentilineatus, was detected in one female and from a bearded eel goby, Trypauchenopsis intermedia, in another.
On the other hand, we identified 11 host species from 97 female mosquitoes in Iriomote-jima. Bostrychus sinensis was the most frequent host (61%), followed by M. microchir (18%), Pe. argentilineatus (6%) and Pi. boro (5%). Some host species were found only in Iriomote-jima, but they are detected from only one or two female  Although we collected only 6 blood-engorged females on Ishigaki-jima, host composition was quite different from the other islands; M. microchir was the most frequent (50%), followed by two blennies Blenniella bilitonensis and Istiblennius edentulous.

Discussion
Host identification based on DNA sequences. The DNA-sequence-based technique has brought great progress in bloodmeal host identification of mosquitoes and other arthropods (e.g., biting midges 21-23 and ticks 7,24 ). In our study, using a combination of some DNA regions used for phylogenetic studies and DNA barcoding, host identification for fish-fed A. baisasi was much improved. Tamashiro et al. 9 aimed at revealing feeding habits of 35 mosquito species from 11 genera with broad spectra of potential host animals. They also used the mitochondrial 16S rRNA region for this species and this enabled identification at higher taxonomic levels (i.e., at the order level). In our study, we applied the COI region to host identification 4,7 and also benefited from recently developed universal PCR primers, which were originally designed for metabarcoding environmental DNA from fishes 20 . This resulted in host identification at the species level. Therefore, this study was the first to demonstrate host assemblages of mosquito species that are parasitic upon fish in natural populations.
While the COI region is well established for DNA barcoding and the amount of available data has been increasing, many of our samples failed to surpass 99% similarity to any taxa in GenBank. There are several www.nature.com/scientificreports www.nature.com/scientificreports/ possible reasons for this. First, the GenBank database does not include sequence diversity among localities. For example, B. sinensis is widely distributed and regional differentiation may account for the low similarity (92-94%). Second, taxonomy has often not been clearly established and that makes it difficult to complete the database. Regardless, we developed a database including more species from more localities with reliable identification and the genomic data. The sequence database of candidate fishes from the sampling areas greatly improved the similarity, and the short fragment of 12S rRNA (using a "MiFish" primer set) 20 helped to further resolve these problems (Supplementary Table S2).
Feeding habits of Aedes baisasi. Host identification at the species level has some interesting implications.
First, contrary to our expectations, based on the laboratory observation by Okudo et al. 14 , mudskippers were not the main source of bloodmeals in A. baisasi; DNA sequences that originated with a mudskipper Periophthalmus argentilineatus were detected in only 7 out of 229 (3%) host-identified females. Second, bloodmeal hosts fed frequently on fish species which inhabit mangrove swamps, such as B. sinensis, Pi. boro and Pe. argentilineatus 25,26 , confirming that feeding activities of A. baisasi are restricted to the mangroves 16 . Some other species documented less frequently included Mugilogobius sp., Myersina macrostoma, Trypauchenopsis intermedia, and Uropterygius concolor, which also inhabit mangroves [26][27][28] . Some species, such as Entomacrodus striatus, Salarias fasciatus, Istiblennius edentulus, and Gymnothorax pictus inhabit tidepools along rocky shore and/or coral reefs 29 , implying that A. baisasi may sometimes search for hosts out of mangrove forests and may visit nearby lagoons. Third, many bloodmeal hosts can remain out of water for prolonged periods. Bostrychus sinensis is known as a facultative air-breathing fish 30 , and can survive out of water for more than a day 31 . Pisodonophis boro also has air-breathing habits 32 . Some blennies are amphibious, e.g., E. striatus [33][34][35] , I. edentulus 35,36 (but see Platt et al. 33 ) and also B. bilitonensis, sometimes clings to rocks out of water 37 . The moray eel G. pictus also leaves water and wriggles across dry places 38 . Given the behavioral attributes of these fishes, it is likely that A. baisasi searches for and lands on fishes either when they leave the water or when parts of their bodies are exposed to the air.
Many of these air-breathing or amphibious fish are reported to be more active out of the water at night than during the day. Most amphibious blennies emerge from water mainly at night to avoid the risk of desiccation 39,40 . Bostrychus sinensis usually hides in hollows on mud, beneath rocks, or gaps in mangrove roots, and in caves during the daytime and emerges at night 41 . We have seen many individuals feeding in shallow waters at night, which may often expose themselves to the air. Pisodonophis boro is also active at night at shallows 42 , foraging mainly for sesarmid crabs 43 , and is also an accessible host for mosquitoes. These habits match the feeding activity of A. baisasi.
Our results, with support from other observations 14,16 , suggest that females rest in lobster holes during the daytime and leave at night to search for bloodmeal hosts within the mangroves and sometimes in adjacent lagoons. They locate and land on the surfaces of exposed amphibious or other air-breathing fishes and feed there, but are not attracted to humans or warm-blooded animals 14 . However, some ecological factors remain unclear regarding seasonal variation in host selection, the stage of fishes on which A. baisasi feed (e.g., juveniles, immatures or adults) and the cues A. baisasi uses to locate bloodmeal hosts. Seasonal variation is expected because fish communities vary seasonally in mangroves 25,42 . Tamashiro et al. 9 found three A. baisasi females with blood from frogs. These were sampled on Iriomote-jima in May (Ichiro Miyagi and Takako Toma, Laboratory of Mosquito Systematics of Southeast Asia and South Pacific, personal communication). Olfaction is the likely cue for the mosquitoes to locate a host, which is implied by the fact that air-breathing fishes have special excretion systems in their skin [44][45][46] , but laboratory and/or field assays are needed to confirm this hypothesis.
Host preference of Aedes baisasi. Host assemblages were quite different among the islands. This may reflect differences in abundance of available fishes among the islands, considering that B. sinensis is abundant on Iriomote-jima, but less on Okinawa-jima and Amami-oshima 41,47 .
On the other hand, it is puzzling that A. baisasi do not feed frequently on the mudskipper, Pe. argentilineatus, which is apparently the most accessible fish host in this region. Periophthalmus argentilineatus seems less active at night 48 , and we often saw it still staying out of water. The mudskipper may have some kind of mechanism that keeps off A. baisasi. It is also the puzzling that significant blood feeding from spaghetti eels, Moringua microchir, occurred on three of the four islands. Little information is available on the life history and ecology of moringuid eels, including M. microchir 29,49 . Moringua microchir is recognized as a rather rare species in Japan, and it has not reported from mangroves. Keith et al. 50 reported that the juveniles inhabit estuaries and lower reaches of rivers, while adult females stay on shallow marine bottoms. However, frequent predation by mosquitoes implies that spaghetti eels are actually common around mangroves. Our data also suggest that A. baisasi may preferentially feed on it. Although A. baisasi chooses bloodmeal hosts according to their availability (abundance and nocturnal activity out of water), host preferences may exist.
The subgenus Geoskusea, to which A. baisasi belongs, includes 10 species, all of which also inhabit brackish water 51 . In addition to A. baisasi and A. longiforceps 13 , other species also probably use air-breathing fishes as bloodmeal hosts. Further studies of host identification in these species will help our understanding of the evolution in host preference and exploitation in niche adaptation.

Methods
Sampling. Adult mosquitoes were collected during the daytime (between 9:00 a.m. and 6:00 p.m.) in mangrove forests on four islands in the Ryukyu Archipelago, Japan (Fig. 2 www.nature.com/scientificreports www.nature.com/scientificreports/ Mosquitoes resting in crab or mangrove lobster burrows during the daytime 16 were forced out by disturbing them with a twig and trapped with a sweep net. We found from preliminary surveys that it was more effective than using a handheld vacuum 16 . Collected mosquitoes were preserved in 99.5% ethanol and brought to the laboratory. Bloodmeal host identification. Each mosquito was individually examined to determine species, sex, and blood-feeding status under a microscope. For blood-engorged females, their abdomens were isolated using forceps. Genomic DNA was extracted from the abdomen either using a DNeasy Blood & Tissue Kit (Qiagen) or by the HotSHOT method 52 .
All sequenced specimens were identified using BLASTn searches against the GenBank nucleotic acid sequence database (NCBI website, http://www.ncbi.nlm.nih.gov/BLAST/) and/or a similar search using BLAST+ ver. 2.3.0+ against the additional reference sequences stated below. The most similar fish species (≥99% sequence identity) based upon blood from engorged female mosquitoes was considered to be the parasitized host.
When we could not find any data that matched our query sequences at ≥99%, we conducted PCR using the same DNA polymerase kit for other regions with primers designed to amplify exclusively from vertebrate or fish DNA: the mitochondrial cytochrome B region with CytB(f)(5′-GAG GMC AAA TAT CMT TCT GAG G-3′) and CytB(r)(5′-TAG GGC VAG KAC TCC TCC TAG T-3′) 4 , or the mitochondrial 16S rDNA region with 16Sa-L (5′-CGC CTG TTT ACC AAA AAC ATC GCC T-3′) and 16Sb-H (5′-CCG GTC TGA ACT CAG ATC ACG T-3′) 53 . PCR cycling conditions were as follows: initial denaturation at 94 °C for 2 min, then 35 cycles of 98 °C for 10 s, 55 °C for 40 s, and 68 °C for 1 min for cytochrome B, and 35 cycles of 98 °C for 10 s, 50 °C for 40 s, and 68 °C for 1 min for 16S rDNA. Positive amplicons were sequenced with one of the primers used in PCR reactions. Identities of all sequenced specimens were determined as described above.
Reference sequences of fishes. Total genomic DNA of 15 specimens belonging to six anguilliform species and six gobiiform species collected from Okinawa-jima and Iriomote-jima was extracted from the right pectoral fins (Gobiiformes) or muscle pieces (Anguilliformes) preserved in 99.5% ethanol, using a DNeasy Blood & Tissue Kit (Quiagen, Hilden, Germany) or a Maxwell RSC Blood DNA Kit (Promega, Fitchburg, Wisconsin, USA).
Whole genome shotgun sequencing libraries were prepared using a KAPA HyperPlus Kit, PCR-free (KAPA Biosystems, Wilmington, Massachusetts, USA). Extracted genomic DNA was enzymatically fragmented into pieces of 200-1000 bp. After repairing the protruding ends and A-tailing, sequencing adaptors were ligated onto both ends of the DNA fragments. Shotgun libraries were then sequenced on either an Illumina MiSeq sequencer (Illumina, San Diego, California, USA) with MiSeq V3 600 cycle kit (Illumina) or an Illumina HiSeq 2500 sequencer in Rapid Run mode version 2 using a HiSeq Rapid Cluster Kit v2-Paired-End (Illumina) and HiSeq Rapid SBS Kit v2 (Illumina) or an Illumina HiSeq 4000 sequencer with HiSeq 3000/4000 PE Cluster Kits and HiSeq 3000/4000 SBS kit (300 cycles, Illumina) following manufacturer instructions.
Sequencing data from each library were assembled with the IDBA_UD assembler version 1.1.1 54 with different kmer lengths (60, 80, 100). Identification of complete mitochondrial genomes from assembled contigs was performed by (1) comparing them with the complete Stiphodon alcedo mitochondrial genome (accession: AB613000.1) (BLASTN e-value B 1e-100), and by (2) confirming that 100 bp of both head and tail DNA sequences of a contig were identical, indicating that the sequence was circular. Complete mitochondrial genomes were aligned using MAFFT v7.244 55 and all positions with gaps were removed using trimAl 56 . All sequenced raw data are available in the DDBJ Sequence Read Archive under BioProject accession number PRJDB5763. Assembled mitochondrial genome sequences with gene annotations are available in the DDBJ database under accession numbers: AP019348-AP019362. Accession numbers for each individual are shown in Supplementary  Table S1.
Procedures used to handle fish specimens in this study were approved by the Animal Care and Use Committees of both Okinawa Institute of Science and technology Graduate University and Gifu University. All experiments and samplings were performed in accordance with relevant guidelines and regulations of the committees.