Diversication of leech proboscis structure according to prey ingestion behavior

Background Adaptive radiation is a phenomenon in which various organs are diversied morphologically or functionally as animals adapt to environmental inputs such as diet and circumstance. Although previous studies have addressed changes caused by various external pressures, the evidence for variation in invertebrates is not well known. Leeches comprise a carnivorous or ectoparasitic group of animals that feed on a wide range of prey. They exhibit a corresponding variety of ingestion behaviors and morphological diversity of mouthparts and gut specializations. However, research on the diversity of ingestion behaviors and the internal structure of feeding organs in leeches is little known. In this study, we use histological analyses, uorescent labeling and immunohistochemistry to reveal the detailed proboscis structure in the family Glossiphoniidae, while also suggesting the diversication of proboscises. Results We identied the feeding behavior of rhynchobdellid leeches, which have the proboscises. Alboglossiphonia sp. swallows prey whole using its proboscis, whereas other leeches exhibit typical uid-sucking behavior. Glossiphoniid leeches exhibit uid ingestion behavior along with clear arrangement of longitudinal muscles, circular muscles surrounding the lumen, and radial muscles, while Alboglossiphonia sp., which displays macrophagous ingestion like salid Barbronia sp., has a partial circular muscle distribution and spacious lumen that extends to longitudinal muscle layer. To address whether the different feeding behaviors are intrinsic, we investigated the behavioral patterns and muscle arrangements in the earlier developmental stage of glossiphoniid leeches. Juvenile Glossiphoniidae including the Alboglossiphonia sp. exhibit the uid ingestion behavior and have the proboscis with the compartmentalized muscle layers. Conclusions Genetic, morphological and behavioral differences between juvenile and adult stages of Alboglossiphonia sp. suggest their adult feeding biology has diverged from ancestral glossiphoniid leeches, while retaining developmental vestiges of the typical juvenile feeding morphology currently observed across Glossiphoniidae. This study provides the characteristics of leeches with specic ingestion behaviors, and a comparison of structural differences that serves as the rst evidence of the proboscis diversication.

whether the different feeding behaviors are intrinsic, we investigated the behavioral patterns and muscle arrangements in the earlier developmental stage of glossiphoniid leeches. Juvenile Glossiphoniidae including the Alboglossiphonia sp. exhibit the uid ingestion behavior and have the proboscis with the compartmentalized muscle layers.
Conclusions Genetic, morphological and behavioral differences between juvenile and adult stages of Alboglossiphonia sp. suggest their adult feeding biology has diverged from ancestral glossiphoniid leeches, while retaining developmental vestiges of the typical juvenile feeding morphology currently observed across Glossiphoniidae. This study provides the characteristics of leeches with speci c ingestion behaviors, and a comparison of structural differences that serves as the rst evidence of the proboscis diversi cation.

Background
Diverse animals have evolved a great variety of ways to obtain the energy needed for survival. As ingestion methods have diversi ed and evolved across species, they have the functional morphology of the ingestion tube. A few invertebrates and vertebrates use potent jaws to swallow the entire prey (macrophagous) [1][2][3][4][5], while others use various other organs such as proboscises and stylets to penetrate the body wall of the prey and suck out uid ( uid ingestion) [6][7][8][9][10][11]. Leeches (Phylum Annelida), are included in the class Clitellata, superorder Euhirndinea. The group of Clitellata consists of Euhirudinea, Acanthobdellida, Branchiobdellida, and Oligochaeta, all of which are hermaphroditic and deposit cocoons through a special organ called clitellum [12]. Leeches comprise a carnivorous or ectoparasitic group of animals that feed on a wide range of prey. Accordingly, they exhibit a corresponding wide variety of ingestion behaviors and morphological diversity of mouthparts and gut specializations. Various studies have partly classi ed leeches based on the differences in their feeding structures (e.g., Rhynchobdellida, and Arhynchobdellida) [13,14], and these groupings have been supported by modern molecular phylogenies. Representative classi cation of Hirudinea is as follows: Hirudiniformes comprise jawed leeches, which use jaws with teeth to injure the host's body wall and ingest body uids; Rhynchobdellida consists of jawless leeches, which use the proboscis to penetrate the host's body wall and suck soft body uids; and Erpobdelliformes includes jawless leeches, which swallow organic material [13,[15][16][17]. The food ingestion behaviors of leeches have previously been identi ed; however, studies investigating differences in internal organ structure according to behavioral patterns are limited to selected species.
Among the uid ingestion leeches, the glossiphoniid leeches display diverse trophic levels in the ecosystem [6, 13,[17][18][19][20][21]. Their food consumption behavior shows a consistent and stereotyped pattern involving a structure called proboscis used to penetrate the body wall and ingest body uids of hosts. In the well-studied leech model Helobdella, a clitellate annelid, proteoblast DM'' (labeled as the 4d cell in Spirallian nomenclature) contributes to development of an unsegmented prostomium which develops into a proboscis during organogenesis [22,23]. The proboscis is comprised of longitudinal, radial and circular muscles. It extends forward and backward to penetrate the body wall of the host and ingest body uids [13]. In the Lophotrochozoans, the striated muscle speci c gene st-mhc and Troponin complex are known to be involved in the development of foregut muscles as well as somatic muscles [24,25]. These previous studies suggest the possibility of the presence of foregut speci c factors in the leech model, and imply that the corresponding factors can be visualized in the foregut muscle structure. Also, it suggests that these foregut-speci c muscle complexes may be regulated by innervation [24,26]. Thus, we con rmed the presence of uorescence-labeled muscle and nerve bers by F-actin marker phalloidin and nerve ber marker anti-acetylated tubulin with aforementioned markers to characterize the muscle arrangement in the foregut region.
Glossiphoniid leeches are known for their typical dietary feeding behaviors that rely on their whip-shaped proboscises [12,13]. Theromyzon tessulatum, parasitic in the nasal passages of aquatic birds, and Placobdella costata, ectoparasitic in freshwater turtles, have proboscises that consist of outer longitudinal muscle, circular muscle which circumscribe the lumen, and radial muscle bers [13].
However, the relationship between internal structure and ingestion behavior is not well understood. Here, we investigated the internal structure of proboscis in the family Glossiphoniidae. Interestingly, despite Alboglossiphonia sp. representing the distinct morphological and molecular phylogenetic features of genus Alboglossiphonia, which belongs to the family Glossiphoniidae, it shows the macrophagous behavior by surrounding and swallowing a prey like a sali d leech Barbronia sp.
In this study, we demonstrate the characteristics of leeches with different types of ingestion behavior, and compare structural differences that provide the rst evidence of proboscis diversi cation.

Results And Discussion
External morphological features and phylogenetic status of leeches The present molecular phylogenetic analysis shows a clear separation of the four main clades of leeches, Erpobdelliformes, Hirudiniformes, Glossiphoniidae, and Piscicolidae with strong branch-support values (Fig. 1B). This result is generally congruent with the conventional classi cation based on their morphological characteristics ( Fig. 1 and Additional le1: Fig. S1) [27]. All species of Alboglossiphonia with a proboscis are placed within Glossiphoniidae. Though placement within a jawless ( uid-ingestion leech group), Alboglossiphonia sp. has an extraordinary behavior on feeding, i.e., swallowing preys using the proboscis (Additional le2: Video S1). Macrophagy is one of the characteristics of the members of Erpobdelliformes [15]. The phylogenetic relation between the erpobdelliformes group and Alboglossiphonia sp. is, however, unsupportable with the current molecular result due to their separation at an earlier node. The genus Alboglossiphonia forms a monophyletic group within Glossiphoniidae with relatively high branch supporting values (BS=100%, and PP=1.00) but the origin of the ingestion behavior, which differs from that of the close congener Alboglossiphonia lata reamains unclear. In other words, it is di cult to explain the behavioral features based on external and molecular phylogenetic characteristics. From another perspective, these issues question the unique changes that may occur within the same family. Also, the ndings suggest that the ingestion characteristics of Alboglossiphonia sp. arise from changes in the internal structure of esophagus.
Comparative ingestion behavior of leeches with different food sources Systematically different leech species exhibit overlapping or unique trophic niche in the fauna and display diverse feeding behaviors for target food sources [6,13,20,28]. In order to observe the exact feeding behavior patterns of different species, behavioral experiments involving various feeding conditions are required. Nevertheless, most reported analyses have focused on quantitative evaluation or positive reaction based on serological tests [13,[28][29][30]. Therefore, we rst investigated the food preferences of leeches and their behavioral differences during ingestion using various preys have been reported [3,6,13,19]. We tested the following prey species in the present study to determine the speci c ingestion behavior of leeches: Limnodrilus hoffmeisteri, a vermiform freshwater oligochaete, which can be swallowed whole; Biomphalaria sp. and Physella sp., freshwater snails that cannot be swallowed by leeches; and Chironomus sp., an insect larva that is unswallowable due to its large size and/or presence of cuticle (Fig. 2). H. austinensis ingests body uids of bloodworms and snails by inserting its proboscis into the host's body wall thereby sucking out the body uids as reported previously [6]. A second glossiphoniid species, A. lata, exhibits a feeding behavior pattern very similar to H. austinensis. It attacks only snails, and does so by inserting its anterior end into the snail's shell and sucking the body uids through the inserted proboscis (Fig. 2C, Additional les: Video S2-3). Unlike other glossiphoniid leeches, Alboglossiphonia sp. intakes only freshwater earthworms and exhibits macrophagy by wrapping the prey and eating it as a whole, similar to Barbronia sp. (Fig. 2C, Additional les: Video S1 and S4). Through food preferences, we identi ed the unique trophic niche under non-competitive conditions of the sympatric leeches Alboglossiphonia sp. and A. lata. They can be assumed that, despite the same habitat environment, due to the trophic niche partitioning, they show coexistence without competition [31][32][33].
Also, unique ingestion behaviors support the contention that the structure of feeding organ in Alboglossiphonia sp. is different from that of the other glossiphoniid leeches and resemble that of macrophagous leeches.
Comparative structural morphology of leech feeding organs Species in different ecosystems have evolved in habitats by undergoing changes in their external and internal structure, and manifested behavioral variations to survive against various external pressures [34][35][36]. These changes resulted in behavioral convergence despite differences across the species (e.g., Rodent turbinate; Arachnid web architectures) [37,38]. Among various environmental factors, the speci c behavioral convergence about food that is essential for survival manifests in various aspects (e.g., Ultrasonic predator whale and bat) [39,40]. These behavioral convergences beyond species cannot be explained phylogenetically, suggesting that behavioral convergence is the result of evolutionary convergence depending on the choice of similar food sources via speciation. Thus, similar patterns of ingestion behavior among Alboglossiphonia sp. and Barbronia sp., suggest the presence of possible differences in the proboscis of Alboglossiphonia sp. despite belonging to Glossiphoniidae. In the present study, the structure of the ingestion tube was elucidated via histological analysis and molecular methods in order to identify the structural similarities and differences between these leeches.
The proboscis of glossiphoniid leeches such as H. austinensis and A. lata is a muscularized tubelike organ specialized for penetration into the prey and ingestion of the prey's blood or other body uids and soft tissues. Developmentally, the proboscis arises primarily from mesodermal precursor cells known as M teloblasts. Structurally, it is characterized by a sharply de ned complement of longitudinal, radial, and circumferential muscles [22]. Longitudinal muscles form the outer edge of the proboscis and radial muscles span the thick wall of the proboscis from just within the longitudinal muscles to its three-fold symmetric lumen in cross-section (Figs. 3A-B). The lumen assumes a narrow three-pronged stellate shape when the radial muscles are relaxed and expands to an approximately triangular form when the radial muscles contract [41,42]. Finally, prominent circular muscles lie roughly halfway between the center and the edge of the proboscis, thereby forming a circular band de ned by the three tips of the lumen [13,22]. This compartmentalized structure may facilitate independent movement of proboscis from the anterior to the posterior direction and suggests that it is associated with uid ingestion ( Fig. 3 and Additional le 1: Fig. S2A) [43,44].
Compared with the well-de ned proboscis of the uid-feeding species, the feeding organ of the macrophagous leech Barbronia sp. shows differences. Barbronia sp. does not have a proboscis, instead, it has an integrated esophageal structure connected to buccal cavity with a band of circular muscles that circumscribes a tri-radiate and spacious lumen (Fig. S2B) [45]. Intriguingly, the circular muscle of Alboglossiphonia sp. in proboscis is partially distributed, and the tri-radiate tips of its lumen also extend further radially towards the outer band of longitudinal muscles, causing spacious (Figs. 3C-D). We speculate that this morphological difference in macrophagous species compared with uid-ingesting species is an evidence of speciation and evolutionary adaptation [46,47], in that facilitates further expansion of the lumen for ingestion of intact prey ( Fig. 3 and additional le 1: Fig. S2).
To delineate the muscular arrangement of the proboscis, we carried out molecular analyses to assess differences in the expression of a common muscle-patterning gene, and molecular markers of muscle anatomy, among four leech species. Previous studies reported on muscular differentiation in lophotrochozoan animals. In Platynereis dumerillii, visceral muscles in the foregut consist of striated and smooth muscles. Within this region, troponin T proteins and myosin heavy chain (MHC) genes are involved in muscle cells and are known as annelid foregut striated muscle markers [24]. Instead of visualizing the speci c st-mhc ortholog in each species, we tried to express the striated muscle in the esophagus in each species based on the similarity between the orthologs of Hirudo nipponia (Hirudinidae) and Helobdella austinensis (Glossiphoniidae), which are systematically distant. The two sequences showed a high level of similarity (81% identity at the nucleotide level and 93% identity at the amino acid level), suggesting that the esophagus muscle layer can be indirectly visualized using st-mhc orthologs of two species (Additional le 1: Fig. S3). In addition, nerves are distributed in the foregut region, and it can be assumed that muscle movement, which is controlled by the innervation pattern and the detailed arrangement of muscles in the foregut, can be con rmed by co-visualizing the nerve and muscle bers [26]. Various muscle markers, st-mhc transcript, and the innervation marker acetylated tubulin show detailed intra-structural and nerve distribution according to muscle bers in each species. Within 4 species, we con rmed that the nerve bers are distributed along the arrangement of muscles in the esophagus (Fig. 3). Innervation of the muscle suggests that the muscles in the esophagus are regulated by neuronal stimulation, and the spatial expressions of st-mhc orthologs reveal the possibility of existence of those st-mhc orthologs and conservation of foregut muscle components in different leech species ( Fig. 3 and Additional le1: Fig. S4). Fluid-ingesting leeches have a distinct longitudinal, circular, and radial muscle arrangement, and the lumen extends to the circular muscle layer (Figs. 3A-B, and Additional le1: Fig. S4). The con guration of muscles in the leech proboscis exhibits structural similarity to that of vertebrate iris muscles consisting of circular sphincter and radial muscle [13,48]. In the case of Alboglossiphonia sp., the circular muscle layer is partial, with more spacious lumen that extended to the longitudinal muscle layer than in H. austinensis and A. lata, which present clear muscle layers (Figs. 3A-C). Due to these structural differences, the tip of the proboscis of uid-sucking leeches shows condensed apical structure, while Alboglossiphonia sp. exhibits an uncondensable cylindrical tip with expanded proboscis pore (Fig. 4A). These features are likely related to the limited ability to condense the proboscis tip given the partial distribution of circular muscle, suggesting possible macrophagy via loose internal space construction (Figs. 3C-D, 4B). Also, numerous cilia bundles are clearly visible at the tip of the proboscis of Alboglossiphoia sp. (Fig. 4A). It is assumed that the cilia bundles are related to the recognition of prey as sensory cilia (sensilla) and the use of the proboscis for macrophagy [49,50].
These results provide the rst evidence suggesting that muscular organization, including differences in muscle-type composition within the proboscis, may facilitate macrophagous feeding behavior in fresh water leeches. Barbronia sp. is a macrophagous leech with radial musculature of the esophagus extending outwardly beyond longitudinal musculature, as seen in macrophagous oligochaetes [51], without forming a proboscis organ (Fig. 3D, Additional le 1: Figs. S2B and S4). This feeding organ is not an isolated proboscis, and cannot elongate the structure or penetrate the prey, resulting in altered feeding behavior. Esophageal intramuscular complexes generate a strong suction force, resulting in a unique pattern of ingestion behavior such as swallowing of organic matter or popping of soft parts (Additional le 5: Video S4). The peristaltic movement of the esophagus in vertebrates suggests that the action of extending food in the aboral direction via complex interaction between circular and longitudinal muscles results in radial contraction [52]. In Barbronia sp., the muscular arrangement of the feeding organ is similar to that of vertebrates, thereby suggesting macrophagy via peristalsis of longitudinal and circular muscles in the esophagus (Fig 3D and Additional le 1: Fig. S2C). In addition, it is speculated that the dense longitudinal muscles located outside the esophagus may be related to the locomotion of Barbronia sp. (Figs. 3D and 4B). In summary, Alboglossiphonia sp. exhibits and alternative distribution of circular musculature, along with an expanded luminal space, designed in combination for feeding behavior that is intermediate between uid-sucking and macrophagous leeches. Furthermore, this structural organization is hypothesized to facilitate a pattern of ingestion behavior similar to that of Barbronia sp.

Conserved uid ingestion behavior and compartmentalized foregut musculature in glossiphoniid juveniles
Glossiphoniid leeches bearing a cocoon with a thin membrane and without the hardened shell similar to erpobdelliformes have the embryos attached to the abdomen until they grow to a su cient size. These growing embryos can parasitize the host and ingest body uids. After receiving parental care, the individuals exhibit a parasitic life on the host and use their developed proboscis to ingest body uids [3,17,[53][54][55]. However, the uid-ingestion behavior of Alboglossiphonia sp. larvae is unknown apart from leeches belonging to glossiphoniid known as uid-ingesting leeches. To investigate their ingestion behavior, we rst analyzed the feeding behavioral patterns of different leech species in the juvenile stages (Fig. 5A). Alboglossiphonia sp. adults do exhibit macrophagous feeding behavior. However, the juveniles of this species exhibit uid-sucking behavior similar to A. lata and H. austinensis (Fig. 5A and Additional le 6: Video S5). The differences in behavioral patterns are thought to be due to variation in the proboscis structure within the foregut. Thus, we conducted immuno uorescence staining to analyze the proboscis structures in juvenile stages of three species that exhibit food-ingestion activity. Our analyses of uorescent muscle markers revealed the presence of a well-developed independent proboscis in the foregut, even at the juvenile stage of food intake (Fig. 5B). Cross-sectional analyses of the juvenile proboscis showed a well-partitioned musculature in all three species, although Alboglossiphonia sp. showed differences compared with its adult form. The arrangement of circular muscle in the adult proboscis was observed as comparatively less structured than the other leech species, while the circular muscle layer in the juvenile stage exhibited a well-de ned partition, as in H. austinensis and A. lata (Figs. 3A-C, and 5B). These results indicate that Alboglossiphonia sp. manifests uid-sucking behavior using well-developed muscles in the juvenile stage. Subsequently, Alboglossiphonia sp. undergoes gradual changes in the structural arrangement of muscles in the proboscis along with the ingestion pattern changing to macrophagy. These ndings explain the presence of an intermediate proboscis structure in Alboglossiphonia sp. compared with uid-sucking and macrophagous structure seen in other leeches. Within glossiphoniid leeches, speci c food preferences vary widely across species. For example, the Amazon leech Haementeria ghilianii is a large rhynchobdellid species adapted to feeding on mammalian blood [13,21], and Helobdella stagnalis consumes diverse foods, whereas Glossiphonia complanata known as a specialist leech prefers gastropods [13,20]. Similarly, A. lata and Alboglossiphonia sp., which belong to the same genus, have different food niches in the same habitat (Figs. 2C and 5A). These diverse food preferences suggest that ancestral glossiphoniid leeches may have ingested different prey. Subsequent divergence may have arisen from a combination of differences in morphological development that were associated with preferences for speci c prey items. Furthermore, ingestion of selective prey may alter the structure of the feeding organ, and accordingly, results in differences in feeding behavior [56]. As representative examples, H. austinensis and A. lata show similar feeding behavior in the larval and adult stages, and the proboscis exhibits similar muscle structure. However, in the case of Alboglossiphonia sp., the larval stages ingest uids with their proboscis, while the adults show macrophagous behavior, attributable to the differences in the arrangement of muscle layers. Therefore, within Glossiphoniidae, it appears that juvenile structural morphology facilitates ingestion of body uids by two of the leech species we investigated, while the particular proboscis structure and feeding behavior in juvenile of Alboglossiphonia sp. may persist as vestiges of the ancestral, or most common, pattern observed among glossiphoniid leeches [57,58] (Fig. 5C).

Conclusions
The results of this investigation suggest that there is an observable correlation between internal morphological structure and ingestion behavior in the proboscis of leech annelids. The organization of tissue and musculature in the proboscis of macrophagous leeches enables ingestion of whole organisms, unlike uid-ingestion mechanisms. Alboglossiphonia sp. exhibits an esophageal structure intermediate between macrophagous and uid-feeding leeches and manifests similar uid intake behavior during the juvenile stage as other proboscis leeches. This behavioral pattern suggests that the feeding behavior of leech is not intrinsic and may change to other patterns of ingestion depending on the development of feeding organ structure. Also, similar food preferences reveal structural and behavioral convergence among other species, despite the species diversity. Genetic, morphological and behavioral differences between juvenile and adult stages of Alboglossiphonia sp. suggest their adult feeding biology has diverged from ancestral glossiphoniid leeches, while retaining developmental vestiges of the typical juvenile feeding morphology currently observed across Glossiphoniidae. The authors declare that they have no competing interests.

Methods And Materials
Animals Adult Alboglossiphonia lata, Alboglossiphonia sp., and Barbronia sp. specimens were collected by examining submerged plants, leaves, and plastic bags in selected localities of Bangjook reservoir in Cheongju, Chungcheongbuk-do (South Korea). Adult Glossiphonia sp. was collected in selected localities of Dal stream in Goesan-gun, Chungcheongbuk-do (South Korea). Adult Hemiclepsis sp. was collected in selected localities of Miho stream in Cheongju, Chungcheongbuk-do (South Korea). Helobdella austinensis was bred in the laboratory. All adult specimens except Glossiphonia sp. and Hemiclepsis sp., which cannot be incubated in arti cial conditions, were incubated in a bowl containing arti cial pond water. Glossiphonia sp. and Hemiclepsis sp. were xed with 100% EtOH until used for the histological analysis. The specimens were cared for once daily by changing solution and the bowl was scrubbed manually to get rid of any residual waste. They were stored in a BOD incubator at 22 ℃.

Phylogenetic analysis
Three partial nucleotide sequences of the mitochondrial cytochrome c oxidase subunit 1 (CO1) (Alboglossiphonia sp., A. lata, and Barbronia sp., about 700 bp), and four partial sequences of 18S ribosomal RNA (18S rRNA, about 1.8 kb) from the same three species plus Helobdella austinensis were obtained in this study by PCR ampli cation. Additional sequences of both genes were obtained from the GenBank, and two alignments of 62 COI and 62 18S genes from the same group of species (see Additional le1: Table S1 for GenBank accession number) were prepared using ClustalW implemented in MEGA7 software (ver. 7.0.26) [59], and then concatenated. Phylogenetic tree hypotheses were prepared from the concatenated matrix using Maximum Likelihood (ML) and Bayesian Inference (BI). The best-t model was searched based on the corrected Akaike Information Criterion (AICc) using IQ-TREE [60] webserver (http://www.iqtree.org). The ML and BI analyses were conducted using RAxML-NG software (v 0.9.0) [61] and MrBayes software (ver. 3.2.7a) [62] under the General Time Reversible model (GTR) with a proportion of invariable sites (I) and a gamma-shaped distribution rates (G4). The ML tree reconstruction was initially attempted by generating 3,000 bootstrap replicates with "autoMRE" command. The bootstrapping support values for branches were estimated under the transfer bootstrap expectation (TBE) [63]. Markov Chain Monte Carlo (MCMC) for the BI tree was run with 5,000,000 generations and the BI tree was constructed by discarding the rst 25% generations. The trees were visualized with FigTree software (ver. 1.4.4).

Prey selection test and tracking analysis
In order to compare feeding behaviors of leeches, we conducted a survey in the laboratory environment using various food types: Limnodrilus hoffmeisteri (Clitellata, Annelida), swallowable and worm shape; Biomphalaria sp. and Physella sp. (Gastropoda, Mollusca), unswallowable and carrying a shell; and Chironomus sp. (Insecta, Arthropoda), unswallowable and exhibiting a worm shape. First, several individuals of each leech species were placed in the 55 mm petri-dish, and ingestion patterns were observed under mixed prey species (single leech species vs. multiple prey species) and each prey species (single leech species vs. single prey species) to con rm exact preferences and ingestion behavior. After observation, one or two prey organisms were provided to each leech. Each experimental dish was videorecorded using a DCR-SR200 camcorder (SONY, Minato, TYO, Japan) over 8 hr, or until the leeches completed feeding under room temperature. Ingestion behavior tests were performed on three biological replicates in the same condition as described above. Among the recorded videos, location analysis on ingestion behavior was conducted using one representative video for each species. To analyze the behavior of both leeches and the prey, the location of all individuals present in the petri dish was tracked every 3 min using EthoVision software (Noldus Information Technology, Wageningen, GE, Netherlands). When the predators were supplied with two species of prey, only the behavior of prey that was ingested was tracked. However, when Barbronia sp. was provided with L. hoffmeisteri or Chironomus sp., the individual location was tracked every 30 s due to their relatively rapid ingestion. Distances between leeches or preys and a reference point established on the 12 o'clock edge of the petri-dish were recorded.

Histological analyses
To visualize differentiation of proboscis muscle structure, adult leeches were treated with relaxation buffer (4.8 mM NaCl2, 1.2 mM KCl, 10 mM MgCl2, 8% EtOH) and xed in 4% PFA (Electron Microscopy Sciences, Hat eld, PA, USA) in 1X phosphate buffered saline (PBS) overnight at 4°C. For H&E staining, leeches were dehydrated in EtOH series and cleared in Xylene (Central Drug House, New Delhi, DL, India) for 2 hr. The leeches were embedded in para n (Leica, Wetzlar, HE, Germany) and stored at -20°C. Para nized samples (10 µm thickness) were cut with a RM2235 microtome (Leica, Wetzlar, HE, Germany) and stained with Mayer's Hematoxylin (Cancer Diagnostics, Durham, NC, USA) and Eosin (Cancer Diagnostics, Durham, NC, USA). Samples were mounted on glass slides with an Organo Mount (ImmunoBioScience, Mukilteo, WA, USA) and dried overnight at room temperature. Sections were imaged with a LEICA DM6 B compound light microscope (Leica, Wetzlar, HE, Germany) and a LEICA DFC450 C camera (Leica, Wetzlar, HE, Germany). The obtained images were edited using Las X software (Leica, Wetzlar, HE, Germany) and Adobe Photoshop CS5 (Adobe, San Jose, CA, USA). The edited images were prepared as gure plates using Adobe Illustrator CS6 (Adobe, San Jose, CA, USA). To obtain cryo-sections, leeches were embedded in O.C.T. compound (VWR, Radnor, PA, USA) and rapidly frozen in liqui ed nitrogen. Cryo-sectioned samples (15 µm in thickness) were cut with a CM1520 cryostat (Leica, Wetzlar, HE, Germany) and stored at -70°C until use.

Fluorescent labeling and immunohistochemistry
Whole-mount immunostaining was performed according to previously published protocols [54], with the following details: The cross-sections were dried and washed in PBT (0.1% Tween-20 with 1X PBS) ve times. The nerve and muscle bers were visualized after double immunostaining as follows. After washing with PBT, the sections were incubated in diluted blocking solution (1:9 = 10X Roche Western Blocking Reagent : PBT) for 2 h. Samples were incubated with primary antibodies (anti-acetylated-α-Tubulin produced in mouse, Sigma Aldrich, T-7451; or anti-cardiac TroponinT produced in rabbit, Abcam, ab115134) in diluted blocking Solution (1:500) at 4 ℃ for 48 h. After ve consecutive washes with PBT, the sections were incubated with a secondary antibody (goat anti-mouse IgG H&L Alexa Fluor 488, Abcam, ab150113; goat anti-rabbit IgG (H+L) cross-adsorbed secondary antibody Alexa Fluor 568, Invitrogen, A11011) in diluted blocking Solution (1:1000) at 4 ℃ for 24 h. After checking the labeled signal, the samples were washed ve times with PBT, and then stained with Texas Red™-X Phalloidin (ThermoFisher, T7471) for 1 h to visualize F-actin. After checking the labeled signal, the samples were washed ve times with PBT and labeled with DAPI in PBT (1:100) at room temperature in the dark overnight. After washing with PBT ve times, the samples were mounted with Fluoromount-G (SouthernBiotech, Birmingham, AL, USA). Fluorescence-stained embryos and slide samples were imaged using a LEICA DM6 B with a LEICA DFC450 C camera (Leica, Wetzlar, HE, Germany). The obtained images were edited using Las X software (Leica, Wetzlar, HE, Germany). To con rm the detailed muscle structure and innervation in the proboscis, slides co-labeled with F-actin and acetylated tubulin were imaged with a LSM 710 confocal microscope (Carl Zeiss, Oberkochen, BW, Germany). The obtained images were edited using ZEN software (L Carl Zeiss, Oberkochen, BW, Germany). The edited images were prepared as gure plates using Adobe Illustrator CS6 (Adobe, San Jose, CA, USA).

ST-MHC gene identi cation, probe synthesis and in situ hybridization
Total RNA was isolated from H. austinensis mixed-stage embryos and Hirudo nipponia head tissue using TRIzol reagent (Invitrogen, Carlsbad, CA, USA). We selected mRNA from total RNA using Oligo (dT) primer (Promega, Madison, WI, USA) and synthesized cDNA (SuperScript II First Synthesis System for RT-PCR, Invitrogen, Carlsbad, CA, USA). To isolate the H. austinesis striated myosin heavy chain (ST-MHC) gene, a previously published sequence [64] was used and screened using a BLAST implemented in the whole draft-genome reference (http://genome.jgi.doe.gov/Helro1/Helro1.home.html). Two candidate genes (protein id 64397 and 129847) were screened, and the foregut speci c st-mhc gene was isolated by con rming the foregut speci c expression pattern (protein id: 129847) (Additional le 1: Fig S3B). In the H. nipponia transcriptome data, only a single striated myosin heavy chain transcript was found, which showed a high degree of similarity to the H. austinensis foregut speci c st-mhc gene (nucleotide similarity: 81%, translated sequence similarity: 93%) (For nucleotide similarity, see Additional le 1: Fig   S3C). The st-mhc speci c primers were designed to amplify the consensus region of the two sequences producing similar length (product sizes about 850 nucleotides -protein id 129847, Hau st-mhc forward: 5'-GCCACCAAAGGTGAAGAG-3'; Hau st-mhc reverse: 5'-GTCCTCAACGAGCTGCAT-3'). H. nippoinia st-mhc transcript (Hni st-mhc forward: 5'-GCCACCAAGGGCGAAGAA-3'; Hni st-mhc reverse: 5'-TCCTCGACCAATTGCATTTCC-3'). These ampli ed fragments were cloned into pGEM T vector (Promega, Madison, WI, USA). RNAprobes labeled with digoxigenin were made using the MEGAscript kit (Ambion, Austin, TX, USA) and DIG RNA Labeling Mix (Roche, Basel, Switzerland), according to the manufa cturer's instructions. The synthesized RNA probes were applied to each sample at a nal concentration of 2 ng/μl, and the probe labeled samples were incubated with an Anti-Digoxigenin-POD Fab fragments produced in sheep (Roche, Basel, Switzerland) in diluted blocking solution (1:1000). The detail procedure of in situ hybridization was followed using previously published methods [54,65,66]. After cryosection, stored samples were dried to remove residual moisture. Dried samples were treated with 0.2N HCl buffer to inhibit endogenous enzymes and rinsed three times with PBT. After this process, the following experiments were carried out using the same protocol as described above. and are presented as "BS/PP". Dashes (-) after BS indicate PP that has not been applicable for the ML tree mainly due to the topological discrepancies between ML and Bayesian Inference (BI) trees. For BI topology, see Additional le 1: Fig. S1B.

Figure 2
Comparative ingestion behavior of leeches with different food sources. (A) Representative leech species and prey types. See also Additional les: Video S1-4. Scale bars 2 mm. (B) Schematic procedure of different ingestion tests depending on prey types. The recorded locomotion of a leech and its prey was analyzed using EthoVision, a target tracking program. The relative distance from a leech (shown in blue) or a prey (shown in red) to the reference point (asterisk) was measured. Ingestion period is indicated by green box. (C) Representative behaviors of leeches in the presence of speci c prey. Each graph represents the distance between the leech (blue arrowhead) and the food (red arrowhead) from the reference point (white asterisk). When a uid sucking leech adhered to food, its position was consistent (green box) for a period of time. After ingestion of food, the remaining prey that cannot be swallowed persisted. In the case of macrophagous leeches, only locations of the leech remained detectable (green box with green arrowhead) after ingestion of whole prey targets by the leech. Only Limnodrilus hoffmeisteri was fully ingestible by macrophagy.  leeches have compartmentalized muscle layers, with a distinct ring of circular muscles surrounding the proboscis cavity and the radial muscles extending from inner to the outer region of the proboscis. In contrast, Alboglossiphonia sp. shows three sets of separate circular muscles, radial muscles, and an expanded lumen within the proboscis. The esophagus of Barbronia sp. is surrounded by circular muscles, radial muscles extend throughout the body, inner radial muscles are well developed, and the lumen expands to the circular muscle layer.