Divergence of ovipositor length and egg shape in a brood parasitic bitterling fish through the use of different mussel hosts


Teiji Sota, Department of Zoology, Graduate School of Science, Kyoto University, Kitashirakawa-oiwake-cho, Sakyo-ku, Kyoto, 606-8502, Japan.
Tel.: +81 75 753 4078; fax: +81 75 753 4101; e-mail: sota@terra.zool.kyoto-u.ac.jp


Bitterling fishes deposit their eggs on the gills of living mussels using a long ovipositor. We examined whether ovipositor length (OL) and egg shape correlated with differences in host mussel species in the family Unionidae among populations of the tabira bitterling (Acheilognathus tabira) in Japan. Bitterling populations that use mussels in the sub-family Anodontinae possessed longer ovipositors and more elongated eggs than those using mussels of Unioninae, as expected from the difference in host size between the sub-families (anodontine mussels are larger than unionine mussels). Based on a robust phylogeny of A. tabira populations, we demonstrated that the evolution of both OL and egg shape were correlated with host differences, but not with each other, suggesting that these traits have been selected for independently. Our study demonstrates how adaptive traits for brood parasitism may diverge with host shift due to different host availability and/or interspecific competition for hosts.


Brood parasitism is known in birds, fish and insects. This phenomenon is associated with complex adaptations by the brood parasite to use the host, as well as counter adaptations by the brood host to eliminate the deleterious effect caused by the parasites, which can lead to coevolutionary arms races between brood parasites and hosts (Davies et al., 1989). Thus, brood parasite and host systems are intriguing subjects for evolutionary studies. Bitterling fish are obligatory brood parasites of mussels. Specifically, they deposit their eggs in the gill chambers of mussels using a long ovipositor. The ovipositor must be long enough to transfer eggs to the gill cavity through the siphon of a mussel, and thus, should be subject to selection via oviposition success. In addition, egg shape can have a vital role in securing the lodging of eggs in the gill cavity until hatching because mussels respond to bitterling oviposition by rapidly contracting their valves and expelling the eggs. Eggs that are expelled before completing development are quickly eaten or die (Kitamura, 2005). Notably, there is marked variation in egg shape among bitterling species (Nakamura, 1969), which may be adaptive in ensuring eggs are able to remain wedged within the host gill. Mussel gill structure varies among species, and it is hypothesized that bitterling egg shape is under selection to match the locally available mussels (Liu et al., 2006; Reichard et al., 2007). Some species of bitterling are host specialists, while others are generalists that use a range of host species (Smith et al., 2004; Kitamura, 2007; Reichard et al., 2007).

In this study, we focus on the tabira bitterling, Acheilognathus tabira, which exhibits varied ovipositor length (OL) and egg shape among local populations (Arai et al., 2007). This species occurs as disconnected populations on the islands of Kyushu and Honshu of the Japanese archipelago. Tabira bitterling use several mussel species in the Unionidae and Margaritiferidae families (order Unionida) (Hirai, 1964; Kondo et al., 1984; Fukuhara et al., 1998; Kitamura, 2007; Oshiumi & Kitamura, 2009; Kitamura & Morosawa, 2010). Most tabira populations use mussels of either the Unioninae or Anodontinae subfamilies of the family Unionidae, and the differences in host taxa likely influence variations in OL and egg shape. The primary importance of the length of the ovipositor is to enable it to reach the appropriate part of the gill cavity for oviposition through the exhalant siphon of the mussel (Kitamura, 2006a,b). Therefore, OL is likely to be selected for in response to the internal structure of particular mussel species. Shell size is generally larger for Anodontinae species (average shell size, 70–90 mm; maximum, ∼200 mm) than for those of Unioninae (average shell size, 50–70 mm; maximum, ∼100 mm; Kondo, 2008; Masuda & Uchiyama, 2010; Kitamura, 2011). In other bitterling species in Japan, populations using anodontine mussels possess longer ovipositors than those using unionine mussels (J. Kitamura, unpublished data; Kitamura, 2006c, 2007; Kitamura et al., 2009a,b; Oshiumi & Kitamura, 2009; Kitamura & Nishio, 2010). Therefore, a longer ovipositor may be selected for in bitterling populations using Anodontinae. In tabira bitterling, deposited eggs are found mainly in the mussel’s suprabranchial cavity, which is connected to the exhalant siphon (Kitamura & Morosawa, 2010). Because the suprabranchial cavity is also larger in Anodontinae than in Unioninae, a more elongated egg shape with a larger surface area may have been selected for in tabira populations using Anodontinae to enable firm attachment of the egg to the cavity wall, thus reducing the probability of being expelled. The larger egg surface area would enhance the contact with cavity wall and the effect of adhesive chemicals secreted from the egg (Nakamura, 1969; Aldridge, 1999). Japanese bitterling species that use small Unionidae usually produce round eggs, whereas those using large Anodontinae exhibit elongated or uniquely-shaped eggs (Nakamura, 1969; Kondo et al., 1984; Kitamura, 2005, 2006c, 2007; Kitamura et al., 2009a,b; Oshiumi & Kitamura, 2009; Kitamura & Morosawa, 2010; Kitamura & Nishio, 2010). The elongated egg shape may bear some cost in terms of its ability to be deposited into small shells as well as a reduction in the total number of eggs deposited per oviposition trial. Thus, an elongated egg shape may only be advantageous when bitterlings use large-sized mussels. Some eggs may eventually settle in the interlamellar space of the gills. The structures of the interlamellar space and septum differ between Anodontinae and Unioninae (Liu et al., 2006) and among species of Unioninae (Kondo, 2008), although it is difficult to predict how these structures affect egg shape evolution. Here, we tested whether the evolutionary changes in these reproductive features coincide with host changes, as predicted above, using a robust phylogenetic tree of tabira bitterling and comparative phylogenetic analysis.

Materials and methods

Field sampling

Reproductive females of A. tabira were collected at 24 localities spanning the entire distributional range of the species (Fig. 1; Table 1). The classification of subspecies followed that of Arai et al. (2007). Habitats were categorized as lotic (rivers and the large lake Biwa) and lentic (small ponds). The bitterling habitat in Lake Biwa was assigned as lotic because the water movement there was rather active and the habitat condition differed largely from that of the small ponds. Host mussels for A. tabira oviposition were determined wherever possible based on field observations and published studies (Hirai, 1964; Kondo et al., 1984; Fukuhara et al., 1998; Kitamura, 2007; Oshiumi & Kitamura, 2009; Kitamura & Morosawa, 2010). Host mussels included Inversidens brandti Obovalis omiensis, Inversiunio jokohamensis, I. reinianus, I. yanagawensis, Unio douglasiae and Pronodularia japanensis in the sub-family Unioninae, Anodonta spp. and Anemina arcaeformis in sub-family Anodontinae of the family Unionidae, and a single species in the Margaritiferidae, Margaritifera laevis. Samples for DNA extraction were collected at all but two localities.

Figure 1.

 Sampling sites of Acheilognathus tabira in Japan. Locality numbers are the same as those given in Table 1. Subspecies names are indicated.

Table 1.   Sample locality, habitat type, host mussel, and dimensions of females and mature eggs.
No.Locality (preferred: site)HabitatHost musselNo. of samples*.Female body length (SL), mmOvipositor length (OL), mmEgg length (L), mmEgg diameter (D), mmEgg L/D ratioEgg volume (mm3)
  1. Dimensions are presented as mean ± SD.

  2. * Number of females for SL and OL; number of females for dimensions of mature eggs in parentheses.

  3. †DNA sequences were not available.

1.Iwate: HanamakiLoticUnioninae16 (6)57.66 ± 6.1522.38 ± 6.192.24 ± 0.391.40 ± 0.291.60 ± 0.062.27 ± 0.08
2.Iwate: OsyuLenticAnodontinae8 (8)51.39 ± 5.9433.94 ± 4.272.39 ± 1.901.30 ± 0.811.85 ± 0.202.11 ± 0.31
3.Miyagi: NatoriLoticUnioninae9 (9)56.12 ± 6.6523.49 ± 8.692.29 ± 1.141.33 ± 0.891.73 ± 0.132.13 ± 0.32
4.Tochigi: Nasu†LoticMargaritiferdae18 (8)60.73 ± 9.4919.70 ± 3.702.74 ± 2.721.49 ± 0.501.83 ± 0.143.18 ± 0.51
5.Ibaraki: ItakoLoticUnioninae1 (1)64.2019.202.251.371.642.21
6.Ibaraki: Omitama†LoticUnioninae8 (8)48.73 ± 5.9922.70 ± 1.062.23 ± 1.281.38 ± 0.801.61 ± 0.072.25 ± 0.36
7.Mie: Harai R.LoticUnioninae77 (77)44.49 ± 5.8627.61 ± 4.862.01 ± 1.221.24 ± 0.951.63 ± 0.131.63 ± 0.31
8.Shiga: Biwa L.LoticUnioninae10 (10)53.47 ± 5.7324.09 ± 2.782.37 ± 1.151.50 ± 0.571.58 ± 0.072.80 ± 0.32
9.Kyoto: Kizu R.LoticUnknown9 (1)47.96 ± 2.3835.23 ± 4.142.391.461.642.65
10.Okayama: SetouchiLoticUnioninae1 (1)51.8032.901.871.371.361.84
11.Akita: KitaakitaLenticAnodontinae31 (10)53.72 ± 5.7050.87 ± 6.703.13 ± 0.581.17 ± 0.482.69 ± 0.142.21 ± 0.16
12.Akita: GojomeLenticAnodontinae62 (22)49.50 ± 10.9837.49 ± 6.623.08 ± 1.811.25 ± 1.062.47 ± 0.212.56 ± 0.53
13.Akita: YurihonjyoLoticUnknown4 (4)56.78 ± 8.9227.88 ± 3.293.04 ± 2.191.32 ± 1.302.31 ± 0.072.81 ± 0.72
14.Akita: NikahoLenticAnodontinae18 (11)69.62 ± 4.5253.79 ± 11.903.20 ± 1.671.30 ± 1.182.48 ± 0.302.84 ± 0.49
15.Akita: YokoteLenticAnodontinae21 (20)41.55 ± 3.8433.46 ± 5.302.84 ± 1.901.03 ± 0.592.77 ± 0.131.59 ± 0.27
16.Yamagata: YuzaLoticAnodontinae13 (13)62.99 ± 5.4136.12 ± 3.322.98 ± 1.501.28 ± 0.472.34 ± 0.142.54 ± 0.24
17.Yamagata: TsuruokaLenticAnodontinae22 (10)53.39 ± 6.5536.62 ± 5.132.65 ± 1.181.14 ± 0.602.33 ± 0.071.80 ± 0.27
18.Niigata: ShibataLenticAnodontinae2 (2)41.25 ± 4.7441.00 ± 9.762.83 ± 1.061.22 ± 0.572.32 ± 0.192.21 ± 0.13
19.Toyama: HimiLoticUnknown1 (1)61.6042.902.361.132.091.56
20.Ishikawa: ShibayamagataLoticUnknown3 (3)52.43 ± 3.4933.93 ± 10.002.41 ± 1.861.18 ± 0.352.04 ± 0.211.75 ± 0.07
21.Fukui: EchizenLoticUnknown1 (1)61.6032.002.491.072.331.49
22.Shimane: OdaLoticAnodontinae27 (27)54.32 ± 5.9026.53 ± 5.533.06 ± 2.771.27 ± 1.612.41 ± 0.142.70 ± 0.99
23.Fukuoka: KasuyaLoticUnknown1 (1)69.2029.202.811.242.272.24
24Kumamoto: TamanaLoticUnioninae15 (6)56.53 ± 5.1829.05 ± 4.522.60 ± 2.281.29 ± 0.772.02 ± 0.262.25 ± 0.23


We measured the standard length (SL) and OL of live reproductive females following the procedure of Kitamura (2007). Mature eggs were collected from each female by pressing the abdomen, and were preserved in 10% neutralized formalin. Assuming a spherical egg shape, the length (L) and diameter (D) of mature eggs were measured under a binocular microscope. These dimensions were averaged for each female. The ratio of egg length to diameter (L/D) was used as an index of egg shape. The volume of an egg was estimated as 4π LD2/3.

DNA sequencing and phylogenetic analysis

For DNA sequencing, we chose one A. tabira individual from each of the 22 localities because the differentiation among populations was detected in a previous detailed phylogeographic analysis (N. Nagata, unpublished data). Eight cyprinid species were used as outgroup taxa, including seven bitterling species (Acheilognathus cyanostigma from Mie, A. rhombeus from Mie, A. melanogaster from Iwate, A. typus from Akita, Rhodeus ocellatus ocellatus from Yamagata, Tanakia limbata from Mie, and T. lanceolata from Mie), and Zacco platypus from Mie.

Total genomic DNA was extracted from dorsal fin tissue using a Wizard Genomic DNA Purification Kit (Promega). In total, 4986 bp representing one mitochondrial (Cytb, 1126 bp) gene and four nuclear genes (Glyt, 802 bp; Myh6, 732 bp; RAG1, 1504 bp; Ryr3, 822 bp) were used for the phylogenetic analysis. Glyt, Myh6, and Ryr3 were amplified and sequenced following Li et al. (2007). RAG1 was amplified and sequenced using the primers RAG1F1, RAG1R1 and RAG1R3 (López et al., 2004). Cytb was amplified and sequenced using the primers GLUDG-L (Palumbi, 1996) and CB6-2 (5′-CCT CGA TCT TCG GAT TAC AAG-3′; this article). PCR products were sequenced using an ABI 3130XL Genetic Analyzer (Applied Biosystems). All sequences were deposited in DDBJ/GenBank (accession numbers: Cytb: AB620130AB620159; Glyt: AB621420AB621449; Myh6: AB621450AB621479; RAG1: AB621480AB621509; Ryr3: AB621510AB621539).

For the phylogenetic analysis, we treated each gene and each codon position as a different partition. Substitution models for different partitions were selected using KAKUSAN ver. 4 (Tanabe, 2011) based on Akaike’s information criterion (AIC) for maximum likelihood methods (ML) and the BIC2 criterion for Bayesian inference (BI). Reconstruction of a ML tree was carried out using Treefinder (Jobb et al., 2004). The confidence level of each node was estimated with 1000 bootstrap replications. The BI tree was reconstructed using MrBayes 3.1.2 (Ronquist & Huelsenbeck, 2003). We performed 20 million generations of MCMC with a sampling frequency of 100; the initial 20,000 trees were discarded as burn-in.

Because the ML and BI analyses resulted in almost identical topologies, the ML tree was converted to an ultrametric tree using the penalized likelihood method (Sanderson, 2002) implemented in r8s version 1.71 (Sanderson, 2003). An optimal smoothing value of 105 was obtained using the cross validation procedure.

Statistics and phylogenetic comparative analysis

In statistical and phylogenetic comparative analyses, all variables but egg length/diameter ratio were log-transformed. Regression analysis and analysis of variance (anova) were performed using JMP ver 5.0.1J (SAS Institute). Reconstruction of ancestral states and the analysis of independent contrasts between traits based on the phylogenetic tree were performed using Mesquite ver 2.74 (Maddison & Maddison, 2010). For nodes with zero branch length, 1% (= 0.1) of the maximum tree height was given. Parsimonious reconstruction of ancestral states was made for OL and egg shape. To examine the evolutionary relationship between OL and egg shape (length/diameter ratio), PDAP module (Midford et al., 2003) in Mesquite was used. Because an initial assessment with PDAP detected significant phylogenetic dependences of both the OL and egg shape contrasts, branch lengths were transformed by Pagel’s (1992) method to standardize phylogenetically independent contrasts adequately. In the least square regression analysis for the relationship between two independent contrasts, the degree of freedom was reduced according to the number of polytomies on the phylogenetic tree (Purvis & Garland, 1993).

To investigate the effect of habitat, host, and OL on the evolution of egg shape, taking phylogenetic history into account, a comparative analysis using the generalized estimating equation (GEE; Paradis & Claude, 2002) was implemented using the R-package APE (Paradis et al., 2004). In this analysis, because six A. tabira populations without host data were eliminated, 16 populations were used in the analysis.


Ovipositor length was significantly longer in lentic than in lotic populations (F1,22 = 14.2, = 0.0011), as well as in populations using anodontine rather than unionine or margaritiferid mussels (F2,15 = 12.6, = 0.0006; Fig. 2a). OL did not depend on body length (t22 = −0.33, = 0.7461). Anodontinae hosts occurred mainly in lentic habitats, while Unioninae (and Margaritiferidae) hosts were only identified in lotic habitats (Fisher’s exact test, = 0.0015). With this tight association between habitat type and host taxon, the multiple regression model with all these variables contained no significant variable (habitat type: F1,17 = 3.7, = 0.0757; host taxon: F2,17 = 2.1, = 0.1593; body length: F1,17 = 0.66, = 0.4305).

Figure 2.

 (a) Frequency distribution of mean ovipositor length among 24 populations. (b) Frequency distribution of mean egg length/diameter ratio (egg shape) among 24 populations. In (a) and (b), different colours indicate differences in host mussels: black, Anodontinae; white, Unioninae; pale grey, Margaritiferidae; dark grey, unknown. (c) Relationship between ovipositor length and egg shape (length/diameter ratio). Different symbols indicate differences in host mussels. Closed circle, Anodontinae; open circle, Unioninae; triangle, Margariferidae; grey circle, unknown. (d) Relationship between phylogenetically independent contrasts for ovipositor length and egg shape. The contrast for ovipositor length is made positive.

Egg volume did not depend on body size (t22 = 1.24, = 0.2280) or OL (t22 = −0.95, = 0.3539), and did not differ between lotic and lentic habitats (F1,22 = 0.06, = 0.8056) or among mussels (F2,15 = 0.03, = 0.1510). Egg shape (length/diameter ratio) was more elongate in lentic than in lotic populations (F1,22 = 11.9, = 0.0023), and in populations using anodontine mussels than in unionine or margaritiferid mussels (F2,15 = 23.7, < 0.0001; Fig. 2b). Egg elongation was positively correlated with OL (t22 = 3.53, = 0.0019; Fig. 2c), but not body length (t22 = 0.11, = 0.9130) or egg volume (t22 = −0.26, = 0.7974). In 18 A. tabira populations for which at least six female specimens were collected, egg shape was not significantly regressed on body length (t-test; all > 0.05). In the multiple regression model with five variables, only host taxon had a significant effect (host taxon: F2,17 = 4.4, = 0.0393; habitat type: F1,17 = 0.0025, = 0.9608; body size, F1,17 = 0.00018, = 0.9608;OL: F1,17 = 0.19, = 0.6727; egg volume: F1,17 = 0.0018, = 0.9667).

The simultaneous analysis of combined sequence data resulted in a robust topology, and different subspecies were recovered as different lineages except that A. t. jordani was paraphyletic with two lineages from different regions (Fig. 3). Ancestral state reconstruction of OL and egg shape on the phylogeny of A. tabira (Fig. 4) showed that evolution of elongated ovipositors and elongated eggs occurred repeatedly in different lineages in separate geographic regions (see Fig. 1). Despite the correlation of egg shape and OL among populations (Fig. 2c), the evolutionary pattern of OL (Fig. 4 left) was not consistent with that of egg shape (Fig. 4 right); i.e. there was no correlation between the independent contrasts of these traits (Fig. 2d; least squares regression, d.f. = 18, = 1.7, two-tailed = 0.1160).

Figure 3.

 Maximum-likelihood tree of Acheilognathus tabira based on five gene sequences. The most distantly related outgroup species Zacco platypus, has been removed. Node supports are ML bootstrap percents followed by Bayesian posterior probabilities (shown when > 50% or > 0.5).

Figure 4.

 Reconstructed ancestral states for ovipositor length (OL) and egg shape (L/D; length/diameter ratio) on the phylogenetic tree of Acheilognathus tabira. Ancestral states are indicated by the different grey-scale colours of the branches. For each locality, host mussel sub-family (A, Anodontinae; U, Unioninae; ?, unknown) and the average egg shape are shown. See Table 1 for locality numbers.

When taking phylogeny into account in the analysis of variation in egg shape by the GEE approach, the factors affecting its evolution included habitat type (dfP[phylogenetic df] = 3.96, = −0.172 ± 0.0390 [SE], t = −4.4, = 0.0499) and host taxon (dfP = 3.96, b = −0.202  ± 0.0395 [SE], t = −5.1, = 0.0379). A model including these factors returned no significant effect (dfP = 3.96; habitat type: t = −0.3, = 0.8012; host taxon: t = −1.8, = 0.3306). Similarly, habitat (dfP = 3.96, b = −0.170  ± 0.0082 [SE], t = −20.7, = 0.0026) and host (dfP = 3.96, b = −0.194 ± 0.0093 [SE], t = −20.9, = 0.0025) influenced OL individually, whereas the model comprising both factors again returned no significant effect (dfP = 3.96; habitat type: t = −2.8, = 0.2261; host taxon: t = −6.7, = 0.1014).


As predicted, ovipositors and eggs were more elongated in populations from lentic habitats than those from lotic habitats, and in populations using anodontine hosts than those from lotic habitats using unionine and margaritiferid hosts. Because unionine and margaritiferid mussels occur mostly in lotic habitats, while anodontine mussels are mostly found in lentic habitats (Table 1; see also Negishi et al., 2008a,b; Kitamura, 2008), it is difficult to predict whether habitat type or mussel taxon is the most important factor. Indeed, the multiple regression analysis for OL failed to determine which was the responsible variable, while that for egg shape identified host taxon as the sole variable with significant effect. It would be biologically reasonable that host taxon, rather than habitat type, affects OL as well as egg shape, but further study of adaptive aspects in these traits is required. Indeed, many studies have indicated that the adaptive egg size of fish varies among habitat types (e.g. Saitoh, 1990; Iguchi & Yamaguchi, 1994; Johnston & Leggett, 2002; Einum & Flemming, 2002). However, the egg volume of the tabira bitterling did not differ significantly between habitat types, and the variation in egg shape was not related to egg size variation.

The longer ovipositors possessed by the tabira bitterling using Anodontinae as hosts could be attributed to the larger size of Anodontinae compared to Unioninae. It appears that the longer ovipositor is advantageous in using both anodontine and unionine mussels, but that a trade-off may exist between OL and quickness of motion. Egg deposition requires instantaneous insertion of the ovipositor into the gill cavity of the mussel though a siphon before the mussel quickly retracts its siphon and closes its shell in response to insertion of the ovipositor. How the elongated egg is adapted to anodontine mussels remains unclear. Our hypothesis was that elongated eggs, rather than round eggs, may be more resistant to expulsion from the suprabranchial cavity. To confirm this, it is necessary to conduct transplant experiments to assess whether round eggs are more likely to be expelled by an anodontine host than elongated eggs. In reptiles, production of elongated eggs may be an adaptive response by small-sized females to maintain optimal egg volume (Congdon & Gibbons, 1987; Elgar & Heaphy, 1989). In A. tabira, however, egg shape was not correlated with female size within or among populations. Additionally, egg volume did not vary with egg shape. Therefore, the possibility that elongated eggs enable the deposition of larger eggs can be rejected.

There was a tendency for elongated eggs to be associated with females with longer ovipositors. However, the evolutionary correlation between OL and egg shape was not significant. Indeed, there were inconsistencies between host use and physical characters. For example, at locality 22, where tabira bitterling used Anodontinae hosts in a lotic habitat, eggs were elongated but ovipositors remained shorter. This may indicate that egg shape responds to host mussel more quickly than OL, or that there are other factors affecting the evolution of OL. In contrast, at localities 23 and 24, lotic habitats where tabira use Unioninae hosts, ovipositors were short, but eggs were elongated. More detailed information regarding the host mussel used by each local bitterling population is necessary to resolve these confounding cases.

In the regions where A. tabira with short ovipositors and globular (short-length) eggs use unionine mussels, congeners with long ovipositors and elongated eggs that use anodontine mussels are also found (Kitamura, 2006c; J. Kitamura, unpublished data). For example A. melanogaster overlaps in distribution with A. tabira erythorpterus, and A. cyanostigma with A. tabira tabira (Nakamura, 1969; Watanabe, 1998). Although these congeners are often not sympatric today, probably due to the overall decline of bitterlings in Japan (Kitamura, 2008), these species might have competed for hosts previously and thus may have specialized in different host mussels. Additionally, a human-mediated colonization of tabira bitterling in a pond segregated from other bitterling species may have promoted evolution of characters for host use. Among A. tabira erythropterus populations, which usually inhabit lotic habitats and have unionine hosts, short ovipositors and round eggs, a population in an irrigation pond (locality 2) with only anodontine mussels exhibited the longest ovipositor observed and produced the most elongated eggs among the con-subspecific populations (Table 1).

The present study suggests that OL and egg shape can undergo divergent selection, and may evolve rapidly at a local scale. This finding of intraspecific variation in reproductive traits in response to different hosts contradicts findings for the European bitterling, Rhodeus amarus (Reichard et al., 2010). Specifically, in the European bitterling, generalized responses to mussels in different regions are observed, largely in relation to the historical age of the association between bitterling populations and mussels. In East Asia, two or more species of bitterling often co-occur, causing competition for mussels, whereas only a single species is present in Europe. Inter-specific competition among East Asian bitterling species may have promoted the divergence of bitterling reproductive traits, not only among species (e.g. Kitamura, 2007; Reichard et al., 2007), but also among local populations. To elucidate the intraspecific evolutionary changes in reproductive characters in tabira bitterling, further experimental studies are required to gain an understanding of the genetic basis of the traits. Experimental hybridizations between local populations combined with measures of reproductive success would also be useful.


We thank H. Sugiyama, T. Kusanagi, M. Kumagaya, K. Konno, S. Kimura, A. Matsui, H. Terui, Y. Watanabe, Y. Suzuki, Y. Sugawara, N. Okabe, T. Nishimura, Y. Fujimoto, K. Shindo, T. Morosawa, M. Kumagai, T. Hagiwara, T. Abe, I. Kobayashi, N. Onikura, C. Oshiumi, M. Nishio, O. Inamura, T. Ueda, N. Kusamitsu, K. Yoshitani, K. Tominaga, Y. Kano, Y. Nagata, G. Maeda and J. Kawahara for helping with sampling, M. Hori, K. Watanabe, S. Mori, R. Arai and T. Tanabe for advice and comments, and C. Smith for reviewing the English text. We also thank two anonymous reviewers for helpful comments. The study was supported by grants-in-aid from the Sumitomo Foundation, (#070543) and from the Japan Society for the Promotion of Science (#20770014), and the Global COE Program (06) from the Ministry of Education, Culture, Sports, Science and Technology of Japan.

Data deposited at Dryad: doi: 10.5061/dryad.qh1n088f