Popul Ecol (2005) 47:41–51 DOI 10.1007/s10144-004-0201-0

ORIGINAL ARTICLE

Jyun-ichi Kitamura Factors affecting seasonal mortality of rosy bitterling (Rhodeus ocellatus kurumeus) embryos on the gills of their host mussel

Received: 1 June 2004 / Accepted: 4 October 2004 / Published online: 20 November 2004 Ó The Society of Population Ecology and Springer-Verlag Tokyo 2004

Abstract I investigated the seasonal change in factors asphyxiation. Increased embryo mortalities may arise affecting embryonic mortality in the rosy bitterling, through competition among embryos, between embryos Rhodeus ocellatus kurumeus, a freshwater fish that and mussel, and ambient dissolved oxygen levels. The spawns on the gills of living unionid mussels. Research optimal period for bitterling to spawn may represent a was conducted in a small pond during 1999 and 2001 in balance between two opposing factors; with positive which bitterling were provided with Anodonta sp. and negative effects of a seasonal rise in temperature mussels for spawning. Bitterling spawned between April directly affecting embryonic growth rate and oxygen and July, peaking mid–late May. Seasonal survival rate availability. of bitterling embryos in their mussel hosts was uni- modal, with a peak between late April and mid May Keywords Freshwater mussel Æ Oxygen availability Æ (about 70% of total spawnings). In mid April, survival Adaptation Æ Oviposition Æ Density dependence Æ was about 50%. The lowest survival was from late May Egg size to July (0%). Losses of bitterling embryos from mussels were identified by ejections from the mussel host. Ejections were categorized as either ejections of live embryos, or ejections of embryos that died in the Introduction mussel and were subsequently expelled from the mussel. Ejection rates of live embryos were higher in the earlier In oviparous species, embryo traits and parental ovi- part of the spawning period (early–mid April) and dead position decisions can directly affect the survival of embryo ejections in the later period (after June). The offspring and thereby affect parental fitness. However, ejection rate of live embryos was higher among younger the effects of these adaptive traits and oviposition deci- embryos earlier in the season, probably because of the sions may vary seasonally with changes in biotic and incomplete development of morphological and abiotic factors. The fitness consequence of seasonal behavioural traits associated with maintaining the em- timing of reproduction in relation to habitat quality has bryo inside the mussel gill chambers, and as a conse- been studied for insect herbivores (e.g. Hunter and quence of a more protracted developmental period at McNeil 1997) and plants (e.g. Suzuki 2002, 2003). low temperatures making them more susceptible to However, the consequences of such seasonal changes for ejection. The ejection rate of dead embryos was higher embryo mortality are not well described in fish. Under- in older embryos later in the season, and in larger standing seasonal variation in offspring mortality among mussels and at high embryo densities. The survival of oviposition sites allows an understanding of seasonal embryos in mussels was probably related to oxygen habitat choice and the consequences of seasonality for availability, with mortalities probably caused by individual fitness. Here I investigated the consequences of seasonality for offspring mortality in the bitterling, Rhodeus ocellatus kurumeus. Bitterling (family Cyprinidae, J. Kitamura subfamily Acheilognathinae) are freshwater fish that Department of Zoology, Faculty of Science, use only the interlamellar spaces of the paired inner Kyoto University, Kitashirakawa-Oiwakecho, and outer gills of living unionid freshwater mussels as Sakyo-ku, Kyoto 606-8502, Japan E-mail: [email protected] a reproductive substrate (Kanoh 1996, 2000; Aldridge Tel.: +81-75-7534077 1997, 1999; for review see Smith et al. 2004). Female Fax: +81-75-7534100 bitterling use a long ovipositor to place their eggs 42 onto the gills of a mussel through the mussel’s exhalant siphon. Male bitterling fertilize the eggs by Methods releasing sperm into the inhalant siphon of the mussel. Fertilized eggs reside in the mussel until development Study site is completed, finally emerging from the mussel as well- developed larvae. The time to complete development The field survey and experiment, and sampling of mus- sels were conducted from April–July in 1999 and 2001, inside host mussels (embryo development time) is 2 3–6 weeks. Parental bitterling do not care for their in two small artifical ponds (approximately 80 m ) cre- eggs and offspring; however, bitterling embryos in the ated more than 70 years ago in Higashiosaka and Yao gill cavity of a mussel are protected from predators City, Osaka, Japan. The bitterling population in the and typically have relatively high survival rates study ponds was identified as R. ocellatus kurumeus by (Nagata 1985). Kawamura et al. (2001) using mitochondrial gene-se- Because bitterling use a discrete spawning site that quence data. This subspecies is endemic to western Ja- can be readily manipulated, and because they display pan and listed as a critically endangered species in the remarkable morphological, physiological and behavio- Red Data Book of brackish and freshwater fishes of the ural adaptations by using mussels as spawning sites, Environment Agency of Japan (Kawamura 2003). The they represent a valuable model in behavioural, popu- study ponds were designated as sampling stations 18 and lation and evolutionary ecology (Smith et al. 2004). 23 by Kawamura et al. (2001). They are situated in the Previous studies of European bitterling (R. sericeus) foothills of Mount Ikoma-Shigi, an area listed as have provided evidence that the mortality of bitterling ‘‘invaluable wetlands to be conserved’’ by the Environ- embryos on mussel gills is positively density dependent ment Agency of Japan (Ministry of the Environment (Smith et al. 2000a, 2001). However, these studies 2002). Only this bitterling species and one freshwater estimated mortality rates over relatively short periods mussel species, Anodonta sp. B (sensu Kondo 2002), and did not consider the possibility of seasonal varia- were present in these ponds. tion in mortality rates. In addition, the results were calculated with an assumption that mortality rates of all embryo developmental stages were constant. How- Seasonal mortality of bitterling embryos in mussels ever, this assumption has never been explicitly tested, though it has important implications for estimates of The aim of this experiment was to explore the effects of density-dependent mortality of bitterling embryos and mussel characteristics (sex and size), embryo develop- the adaptive value of oviposition decisions. Density- ment time, embryo density, season and larvae size on the dependent mortality of bitterling embryos in mussels is mortality of bitterling during embryonic development in believed to arise through competition for oxygen in the their mussel host. A total of 441 and 180 mussels of mussel gill chamber (Smith et al. 2000a, b, 2001). Anodonta sp. B were collected in mid-March 1999 and Competition for oxygen may occur among bitterling 2001 respectively from the sampling pond before the embryos, between the host mussel and bitterling em- onset of the bitterling reproductive season. The mussels bryos, and between bitterling embryos and the devel- were transferred to a holding basket in the experimental oping mussel embryos, called glochidia, which are pond. The basket was half-filled with a substrate of sand brooded in the outer gill chambers of female mussels and covered with netting that prevented the bitterling (Smith et al. 2004). from obtaining access to the mussels, but allowed the Bitterling embryo mortality rates may also be related mussels to filter normally. The mean length of mussels to maternal yolk resources allocated to eggs. Yolk sup- was 67.7±5.5 (SD) mm in 1999 and 64.2±5.6 mm in ply may vary seasonally with parental changes in phys- 2001. iological condition, and is often related to diet, and/or For each replication, a group of five (until 22 April in social status, particularly if spawning periods are pro- 1999 and throughout 2001) or three (from 23 April in tracted (Kamler 1992). 1999) mussels of each sex was randomly selected from In this study I conducted a field survey and experi- the holding basket and each placed in a flowerpot (9 cm ment to estimate the seasonal change in mortality rates deep, 11 cm diameter) with a sand substrate. The flow- of bitterling embryos in mussel gill chambers. I also erpots were placed in the experimental pond, approxi- tested whether the mortality rates were affected by four mately 15–30 cm from the bank, spaced approximately variables: mussel sex and size, and bitterling embryo 10 cm apart around the pond margin and in a water development time and density in the host mussel. In depth of approximately 15–30 cm. The mussels were addition, I tested whether the effect of these four factors placed in the pond at sunset and exposed to bitterling varied among developmental stages of embryos and spawning until they were retrieved at 17:00 the follow- periods. Finally, I related embryo mortality rates to ing day. Bitterling do not spawn during the night three extrinsic variables (water temperature, dissolved- (J. Kitamura, unpublished data), so mussels were ex- oxygen levels and bitterling oviposition choice) and posed to spawning for a single day only. This procedure three intrinsic variables (condition of females and size of was conducted daily in 1999 and weekly in 2001 eggs and larvae). throughout the bitterling spawning season. 43

After exposure to bitterling spawning, the number of mussel size, embryo development time, and bitterling bitterling eggs deposited on gills of each mussel was embryo density in the host mussel. Because multiple counted by opening a 1-cm gap between the valves of the comparisons were made among variables, all significance mussel using a mussel opening device. A record was levels underwent Bonferroni correction. Embryo devel- made of mussel shell size, mussel sex, and in which of the opment time (days) or rate (speed) at each stage could four gills (inner or outer, left or right) eggs had been not be measured for every embryo, because they re- deposited. After counting the eggs, each mussel was mained inside the host mussel. Consequently, the aver- isolated in a 0.5-l plastic bottle with a 1-mm-mesh net- age duration (days) from oviposition to the emergence ting top and a Styrofoam float and placed back in the of all the swimming larvae from each host mussel experimental pond. The netting allowed the mussel to (incubation period) was used for the duration of incu- filter water but retained any bitterling embryos and bation under the assumption that embryo development larvae that were ejected or that emerged from the mus- time at each stage correlated positively with incubation sel. Each bottle was examined daily and a record kept of period. For the incubation period of host mussels from the numbers of eggs and embryos ejected by the mussel, which no swimming larvae emerged, the average incu- and of fully developed free-swimming larvae that bation period of all mussels exposed to bitterling emerged. Daily checks of mussels continued until all the spawning on the same day was used as a substitute. In bitterling eggs and embryos had been ejected or had addition, the relationship between oviposition date and emerged. All eggs, embryos and larvae were collected embryo development time for each bitterling embryo and staged under a binocular microscope and classified was investigated. according to the scheme in Table 1. The total length It was impossible to make an estimate of the incu- (TL) (length from the tip of the snout to the tip of the bation period after 27 May in 1999 because all bitterling tail fin) of each larva was also measured to the nearest embryos in all mussels died. In 2001, both the number of 0.1 mm at staging. Mussels were used in the experiment mussels used and the number of bitterling in mussels was only once to avoid pseudoreplicating results. After the low. Consequently, the extended Fisher’s exact test was field survey and experiment, all mussels were released used to test among developmental stages for a ‘period’ into the sampling pond. (see below) effect on mortality (ejection rate), and Water temperature was measured daily at the bottom among periods for a developmental-stage effect. The of the pond at a water depth of approximately 80 cm in ejection rate of bitterling embryos for each develop- 1999 and 2001. Dissolved-oxygen levels were measured mental stage in each period represents the ratio of the hourly throughout the day in the same location from 14 number of embryos ejected to the number of eggs June–6 September in 2001 using a DO 9709 dissolved– deposited. The survival rate represents the ratio of the oxygen data logger (Delta OHM). number of free-swimming larvae to the number of eggs Each bitterling embryo ejected from mussels was deposited in each period. In the extended Fisher’s exact identified as either alive or dead. Live bitterling embryos test, all eggs deposited in host mussels in each period prematurely ejected from mussels are vulnerable to were pooled. predation, usually being eaten by adult bitterling as soon To test for an effect of season on embryo mortality, as they are ejected (Smith et al. 2004; J. Kitamura, each day in which spawnings occurred was classified into personal observation), and it was assumed that ejected one of five or six ‘periods’, each of approximately 15- live embryos represented mortalities. On the other hand, days duration: 6–26 April, 27 April–11 May, 12–27 ejected dead embryos represented mortality in the mus- May, 28 May–13 June, 14–27 June and 28 June–10 July sel for some other reason. These two mortality processes in 1999; and 7–21 April, 28 April–12 May, 19 May–2 were analysed separately. June, 9–23 June and 29 June–13 July in 2001. Embryo Multiple logistic regression analysis was used to development time and larva size for each period in 1999 estimate the impact of factors that could affect embryo and on each experimental day in 2001 were tested by mortality. Four factors were investigated: mussel sex, ANOVA and the Tukey-Kramer test.

Table 1 Classification of stages in the development of rosy bitterling (Rhodeus ocellatus kurumeus) embryos and larvae

Stage Corresponding Time from Description stages of Nagata fertilization and Miyabe (1978) at 22°C (h)

1 1–14 27 Unhatched embryo 2 15–17 34 Free embryo without chorion and incomplete yolk-sac projection 3 18–20 67 Free embryo with complete yolk-sac projection and without eye 4 21–24 139 Free embryo with incomplete eye but with lens visible 5 25–26 260 Free embryo without yolk-sac projection with complete eye containing retinal pigment 6 27–29 400 Free embryo with dorsal fin rays visible 7 30 432 Swimming larva with all fin rays visible and without yolk sac 44

Seasonal changes in maternal characteristics

To explore the effects of maternal characteristics (egg size and female condition) on the mortality of bitterling in mussel hosts, a weekly sampling of female bitterling was conducted using a Mondori trap in the test pond from April–August in 2001 and 2002. All collected fe- male bitterling were identified as possessing ripe ova by gently squeezing their abdomen; eggs were extruded from females with mature ovulated eggs. Fish with mature eggs were measured for standard length (SL) and wet body weight (BW) to the nearest 0.1 mm and 0.1 g respectively. After measurement, all the female’s mature eggs were stripped by squeezing her abdomen and the eggs fixed in a 10% formaldehyde solution. After strip- ping, females were returned to the survey pond. All mature eggs were measured to the nearest 0.1 mm under a binocular microscope. Egg volumes were cal- culated using the following formula:  4 1 1 2 Egg volume ðmm3Þ¼ p 3 2a 2b where a and b are the lengths of the major and minor axes respectively (Coleman 1991). In the analysis of seasonal change of egg size, female bitterling that had more than four eggs were used, and the average egg volume was used as an estimate of egg size for each female. Female somatic condition was expressed using con- dition factor (K) (Tesch 1971), calculated as follows: BWðgÞ K ¼ 105: SL3ðmmÞ To test for an effect of season on the egg size, SL and K for each experimental day were tested by ANOVA and Tukey-Kramer test. Fig. 1 a Seasonal changes in water temperature in 1999 and c seasonal changes in water temperature and average dissolved oxygen (DO) in 2001. The average number of eggs deposited in Results each mussel in b 1999 and d 2001. Dots and vertical bars indicate mean and SD respectively Seasonal and daily changes in environmental variables

In both years of the study, water temperature increased ited daily was high from mid-April to late June but de- from April to August, whereas dissolved-oxygen levels creased by early August. In both years, there was a decreased from June to August (Fig. 1). During a single significant difference in the number of eggs between diurnal cycle, dissolved-oxygen levels increased from periods (P<0.001; two-way ANOVA), but there were approximately 08:00 to 17:00, then decreased overnight no significant differences between mussel sexes, or in the from 17:00 to 08:00 the following day. Thus, dissolved- interaction between period and sex of mussels (P>0.05). oxygen levels were lowest at approximately 08:00 and The number of bitterling eggs deposited in each host highest at approximately 17:00 on all dates. mussel was not significantly correlated with the size of the host mussel in any period (linear regression, all P values>0.05) except for male host mussels on 12–27 Patterns of oviposition in mussels May in 1999 (F=11.49, df=47, P=0.001) and those on 29 June–13 July in 2001 (F=14.25, df=14, P=0.0023). The number of bitterling eggs deposited daily in the gill In both years the number of bitterling eggs deposited cavities of mussels was high from late April to mid-June differed significantly among the inner and outer and left but greatly decreased by early July in 1999 (Fig. 1). and right sides of mussel gills (P<0.001; two-way AN- Similarly, in 2001 the number of bitterling eggs depos- OVA), though the number did not differ with sex of 45 mussels nor was there an interaction between location in belonging to stage 7 in Table 1. Embryo development the gill and mussel sex (P>0.05). Eggs were deposited time showed a decline over time in both years, but with more frequently in the inner gill chamber than outer an increase in late June in 1999 (Fig. 3). The TL of (Scheffe test, all P values<0.05), but there was no dif- larvae at emergence from mussels increased from early ference between the right and left sides of the inner and April to early May in 2001, but decreased from early outer gill chambers (Scheffe test, all P values>0.05). May to early July in both years (Fig. 4).

Seasonal survival and development of bitterling embryos Factors affecting embryo mortality

The survival of bitterling embryos during incubation in Multiple logistic regression analysis was applied to bit- mussels was high, approximately 70% from early April terling embryo mortality in mussels from 6 April to 27 to late May (Fig. 2). However, survival decreased to 10– May in 1999 (Table 2). For the ejection of live embryos, 30% from early June to early July in both years. The mussel sex and incubation period had a significant effect ejection rate of dead embryos per 15 days was low (10– on the ejection rate during the egg stage only, i.e. the 30%) for embryos deposited from early April to late ejection rates of eggs were higher in female mussels than May, but increased to 40–90% for those deposited from male and increased with the length of the incubation early June to early July. The ejection rate of live embryos period. At other developmental stages, none of the was relatively high at 20–30% in early to mid-April but measured variables were significantly correlated with low thereafter. From 9 June to 13 July in 2001, the ejection rates of live embryos. ejection rate of live embryos again increased to about 20%. The number of larval bitterling that completed development and emerged successfully from mussels totaled 1,389 in 1999 and 283 in 2001 over the course of the experiment. All these larvae were classed as

Fig. 3 Seasonal changes in embryo development time (in days) Fig. 2 Seasonal changes in rate of survival and live and dead from oviposition date to emergence from mussel hosts in 1999 and ejections of bitterling embryos in mussels in 1999 and 2001. 2001. Numbers above bars refer to number of larvae and bars with Numbers above and in parentheses refer to numbers of host mussels different letters were significantly different in a Tukey-Kramer test and bitterling eggs respectively (P<0.05) 46

The ejection rates of dead embryos differed signifi- cantly among developmental stages in each period in both years (extended Fisher’s exact test, all P val- ues<0.05; Fig. 7). Older embryos tended to have a higher ejection ratio. The ejection rate of dead embryos also differed significantly among periods at each devel- opmental stage (extended Fisher’s exact test, all P val- ues<0.05). The ejection rate of dead embryos tended to increase over time.

Seasonal changes in egg size and female condition at spawning

Females with less than three mature eggs in their ovaries were removed from the analysis of the effects of season on egg size and female condition, as were all data for dates when there were fewer than three females with more than four mature eggs (21 and 28 April and 2 and 29 June 2001; 26 June 2002). In both years, egg volumes were lowest early in the spawning season and highest from mid- to late April, but decreased from early May to late July (one-way ANOVA, both P values<0.001; Fig. 8). Female SL did not significantly differ among oviposition dates in either year (one-way ANOVA, both P values>0.05) (mean±SD=36.9±2.5, range 30.2– 41.2 in 2001; mean±SD=35.7±2.6, range 30.0–40.4 in 2002) except for 24 July 2002 (mean±SD=31.6±4.5, range 25.2–35.1). The condition factors (K) differed Fig. 4 Seasonal changes in total length (TL) of larvae at emergence significantly among oviposition dates in 2001 (one-way from mussel host in 1999 and 2001. Numbers above bars refer to ANOVA, F=3.53, df=9, 67, P<0.01), but not in 2002 number of larvae and bars with different letters were significantly different in a Tukey-Kramer test (P<0.05) (one-way ANOVA, F=1.17, df=14, 124, P>0.05). In 2001, condition factors were higher on 6 and 13 April than 26 May and 2 June (Tukey-Kramer test, all P Ejection rates of live embryos differed significantly values<0.05), but except for these dates there were no among developmental stages in each period in both other significant differences among oviposition dates years (extended Fisher’s exact test, all P values<0.05; (Tukey-Kramer test, all P values>0.05). Fig. 5); younger embryos had a higher probability of To test for an effect of female SL and K on egg vol- ejection than older embryos. When ejection rates where ume, data were categorised in the first and later periods standardized per 24 h within periods for variation in of the spawning season with each period being of embryo development time, stage 2 embryos were shown approximately 40-days duration: 13 April–19 May (first to have the highest rates of ejection among develop- period) and 26 May–13 July (later period) in 2001, and 10 mental stages in all periods (Fig. 6). The ejection rates of April–15 May (first period) and 22 May–2 July (later live embryos differed significantly among periods during period) in 2002. Egg volume was significantly positively developmental stages 1 and 2 (extended Fisher’s exact correlated with the SL in the first period in 2001 (simple test, all P values<0.05), but not significantly during the r=0.070, R2=0.159, F=4.74) and both first (simple other stages (extended Fisher’s exact test, all P val- r=0.061, R2=0.092, F=4.48) and later (simple r=0.062, ues>0.05; Fig. 5). The ejection rates of live embryos R2=0.183, F=13.18) periods in 2002. K was correlated tended to decrease with season for the eggs and stage 2 with egg volume only in the later period in 2002 (simple embryos. r=0.586, R2=0.254, F=5.53) (stepwise multiple Mussel size and embryo density in mussels had sig- regression analysis, all P values<0.05). nificant effects on the ejection rates of dead embryos during stages 5 and 6 (Table 2), and incubation period also had a significant effect at stage 5 only, i.e. the Discussion ejection rates of embryos at stages 5 and 6 increased with mussel size and embryo abundance in mussels (density Bitterling mortality in host mussels changed seasonally dependent), but decreased with incubation period at and was affected by four factors, mussel sex and size, stage 5. At other developmental stages there was no incubation period and bitterling embryo density in association with ejections. mussels. The effects of these factors varied for live and 47

Table 2 Summary of multiple logistic regressions on bitterling embryo mortality in their mussel hosts for four independent variables (mussel sexa and size, incubation period and embryo density) at each developmental embryo stage from 6 April–27 May in 1999

Embryonic stage Living embryo Dead embryo

Independent variable Coefficient P Odds ratio Coefficient P Odds ratio

Egg Mussel sex 1.028 <0.0001** 0.357 0.776 0.0300 0.460 Mussel size 0.010 0.6223 0.774 0.046 0.1407 3.169 Incubation period 0.097 0.0002* 9.290 0.083 0.0312 6.715 Embryo density 0.023 0.0551 0.282 0.079 0.0056 0.014 Free embryo, stage 2 Mussel sex 0.364 0.1322 0.694 1.205 0.0887 0.299 Mussel size 0.058 0.0129 0.232 0.038 0.5238 0.381 Incubation period 0.041 0.1889 2.580 0.139 0.0694 24.604 Embryo density 0.003 0.7191 1.202 0.003 0.9316 0.878 Free embryo, stage 3 Mussel sex 0.337 0.4708 0.714 0.684 0.4256 1.982 Mussel size 0.039 0.3483 2.650 0.048 0.4921 3.394 Incubation period 0.016 0.7784 1.455 0.044 0.6642 0.366 Embryo density 0.009 0.6513 0.643 0.024 0.5029 0.317 Free embryo, stage 4 Mussel sex 0.033 0.9569 0.968 0.097 0.8771 0.907 Mussel size 0.008 0.8863 1.217 0.085 0.1304 8.446 Incubation period 0.067 0.3566 4.699 0.046 0.5165 2.898 Embryo density 0.003 0.9042 0.860 0.029 0.3475 0.254 Free embryo, stage 5 Mussel sex 0.451 0.4664 0.637 1.158 0.008 3.184 Mussel size 0.098 0.1249 0.085 0.139 <0.0001** 33.200 Incubation period 0.048 0.5542 3.035 0.508 <0.0001** <0.001 Embryo density 0.012 0.631 1.773 0.048 <0.0007* 9.617 Free embryo, stage 6 Mussel sex 0.456 0.5591 0.634 0.181 0.15 0.835 Mussel size 0.018 0.8022 1.576 0.050 <0.0001** 3.532 Incubation period 0.077 0.3733 5.918 0.042 0.0156 0.383 Embryo density 0.166 0.1354 0.002 0.019 <0.0001** 2.492

*P<0.05; **P<0.01 after Bonferroni correction aMale=1, female=0 dead bitterling embryos and among embryo develop- 1963; Nakamura 1969; Nagata and Miyabe 1978; Suzuki mental stages. and Hibiya 1984; Suzuki 1988). Both the projections and the tubercles may also serve to keep the embryo wedged between the gill lamellae. The early development of Mortality of live embryos due to ejection by mussels motility along with positive rheotaxis may enable the embryo to maintain its position in the gill. However, no Bitterling embryos are known to have several unique evidence has been shown to demonstrate that these adaptations to enable them to maintain their position in characteristics serve these functions. The present study the gill cavities of their host mussel, including early demonstrates that early embryos do suffer potentially hatching, an anaerobic ethanol pathway for glycolysis, high rates of ejection throughout the period they are in a and an unusually well-developed embryonic respiratory mussel gill chamber (Fig. 5). In particular, eggs (stage 1) system (see Smith et al. 2004 for review). The appear- and stage 2 embryos have higher ejection rates than the ance of embryos changes dramatically during develop- other stages. The transition period, stage 2, was most ment. Bitterling eggs are relatively larger than other vulnerable to ejection from the mussel gills (Fig. 6). cyprinids (Aldridge 1999), shaped like electric light bulbs Differences in the degree of completion of the adaptive, (Nagata and Miyabe 1978). Smith et al. (2004) consid- morphological and behavioural characteristics may ered that this large size and shape might be adaptive in underlie the differences in vulnerability to the exhalant providing a tight fit for the egg between the mussel gill flow from the mussel gill among the stages. Thus, lamellae. After hatching, the embryo develops a pair of ontogenetic changes associated with hatching appear to dorsal yolk-sac projections and small scaly tubercles on be associated with high potential mortality rates due to the yolk sac, and is also motile (Nagata and Miyabe ejection by the mussel. The present study supports the 1978; Suzuki and Hibiya 1984, 1985; Suzuki 1988). The contention that the unusual embryonic characteristics of projections and tubercles are not fully developed at the bitterling may have adaptive value. embryonic stage 2, but are complete at embryonic stages The morphological and behavioural characteristics of 3 and 4, and reduced after embryonic stage 5 (Nikolsky more-developed bitterling embryos may be effective in 48

Fig. 5 Seasonal changes in ejection rate of live bitterling embryos from mussels at each developmental stage in 1999 and 2001 withstanding the exhalant flow in the gill cavities of their host mussel. In general, embryo development time of the endogenous feeding period of fish decreases with increasing temperature (Kamler 1992), and the same was Fig. 6 Ejection rate of live bitterling embryos from mussels in each shown in the present study (Fig. 3). During spring, wa- oviposition period in 1999 and 2001 ter temperature increased from the onset of bitterling spawning to the end of the spawning season. The present hypothesis). However, the experimental design for the study shows that embryo development was relatively study was flawed and failed to control for the direct slow for eggs deposited early in the spawning season effects of phosphate on bitterling embryos (Smith et al. (Fig. 3), when the average water temperature experi- 2004). On the other hand, Smith et al. (2004) showed enced by embryos was low. Thus, the length of incuba- that when mussels were disturbed or handled, the mus- tion tended to decrease over time as water temperature cular contraction of the mussel as it closes its valves increased in the study pond. The higher rate of ejection could inadvertently dislodge bitterling embryos from the of eggs and stage 2 embryos earlier in the season (Ta- gills at high embryo densities. The same phenomenon ble 2, Fig. 5) may have been because of a lengthening of was observed for embryos of R. ocellatus kurumeus at these susceptible periods of development due to rela- higher densities (>100 embryos per mussel) (J. Kitam- tively low temperatures. For later developmental stages, ura, unpublished data). Smith et al. (2004) suggested which may be less susceptible to ejection, the rate of that mussels may not detect the presence of embryos, ejection of bitterling embryos does not appear to be and ejections are accidental (accidental hypothesis). related to the length of incubation (Fig. 5). From the current study it is impossible to verify either Previous studies have discussed whether embryo hypothesis. ejection is an active or passive process by the host mussel (for review see Smith et al. 2004). However, critical evidence regarding this question has not been found. In Mortality of bitterling embryos in mussels R. sericeus, Reynolds and Guillaume (1998) found that the ejection of bitterling embryos from mussels increased In accordance with Aldridge (1997, 1999) and Smith with high levels of phosphate, and they speculated that et al. (2000, 2001, 2004), survival of bitterling embryos in mussels had control over the embryos in their gills and mussels is expected to be low when the oxygen avail- could actively eject them (mussel-ejection-behaviour ability per embryo in each host mussel is below the 49

Fig. 7 Seasonal changes in ejection rate of dead bitterling embryos from mussels at each developmental stage in 1999 and 2001

Fig. 8 Seasonal changes in egg size of female bitterling in 2001 and threshold of an embryo’s minimum requirement. A 2002. Numbers above bars refer to number of larvae, and bars with reduction in oxygen availability may result from in- different letters were significantly different in a Tukey-Kramer test creased embryo activity, competition with other em- (P<0.05). Bars without letters represent dates with too few data bryos in the same host, consumption by the host mussel itself, or a decline in the ambient dissolved-oxygen concentration (Kamler 1992). These processes may all deformities and reduced embryo viability (Kamler increase with elevated temperatures. In fact, bitterling 1992). In the present study, the development time of embryo activity increases with embryo age (Smith et al. embryos after 15 June was relatively high, irrespective of 2001), and mussel activity increases with elevated tem- temperature (Fig. 3). This might have resulted from re- perature and mussel size (Higashi 1965a, b; Huebner duced embryo development rates arising from insuffi- 1982; Bayne and Newell 1983; Fujikura et al. 1988; cient oxygen availability at high temperatures. Pusch et al. 2001). In addition, the concentration of In fish, offspring viability is related to egg quality, i.e. dissolved oxygen in temperate ponds tends to decrease egg size and yolk composition (Kamler 1992). In gen- with depth, and to decrease from sunset until dawn as a eral, the size of eggs deposited by females correlates consequence of plant and animal respiration (Saijyo and positively with the size and condition of females, and Mitamura 1995). The present study showed that the negatively with water temperature (Imai and Tanaka mortality of bitterling in mussels, which may be due to 1987; Kamler 1992; Chambers 1997). Larva size at asphyxiation, increased with age of embryos, later birth hatching increases with egg size and yolk chemical date, larger host mussels and embryo density (density- properties (e.g. protein and lipid); egg quality is lowest in dependent mortality) (Table 2; Figs. 2, 7), findings young and old females (Kamler 1992). In the present which are consistent with mortality factors related to study, egg size increased with size and condition of fe- oxygen requirement. males in some periods of the spawning season, and the The death of bitterling embryos in mussels may also size of eggs and larvae at emergence from mussels was have been a result of direct disruption of embryo smallest earlier in the spawning season (early April), development at high temperatures. High temperatures largest in mid- and late April and then subsequently are known to directly affect embryo development in fish, decreased (Figs. 4, 8). Female size was unchanged arresting development and resulting in an increase in over time though their somatic condition declined. 50

Consequently, the higher mortalities of bitterling em- position rate. Thus, the seasonal timing of spawning by bryos in mussels later in the spawning season might also female bitterling may be adaptive and related to embryo be related to a decrease in egg size due to an increase in mortality in mussels. water temperature and a decline in the mother’s condi- tion. However, because no dead embryos were found to Acknowledgements I am grateful to E. Tanaka, S. Kitagishi and have depleted yolk sacs, nutrient deficiency of small eggs F. Wakano (Section of Biology Higashiosaka City), and Dr. Y. Kanoh (Section of Biology, Seifu High School) for help with was unlikely to be the major mortality factor. the field survey; Professors Y. Nagata and I. Maki, H. Yamane, K. Tani, Y. Iokura and other members of the Laboratory of Animal Ecology, Osaka Kyoiku University, for their support and useful Mortality of bitterling embryos in relation to mussel sex advice; Professor M. Hori, Associate Professors T. Sota and K. Watanabe and other members of the Laboratory of Animal The higher ejection rate of live bitterling eggs from fe- Ecology, Kyoto University, for useful suggestions and discussion; Dr. K. Kawamura (Division of Genetics, National Research Insti- male than male mussels (Table 2) might be a conse- tute of Aquaculture) for comments. I also thank Dr. C. Smith quence of the higher filtration rate from the inner gills of (Department of Biology, University of Leicester) for useful sugges- females. Female unionid mussels incubate their own tions and help with English corrections. This study was partly sup- embryos, termed glochidia, in a pair of outer gills during ported by the Sasakawa Scientific Research Grant from the Japan their spawning season (Kondo 2002). Anodonta sp. B is Science Society. gravid for most of the year (Kondo 2002) and overlaps in its spawning season with R. ocellatus kurumeus. The flow rate from the gills of female mussels may be higher References from the inner gills than the outer because the presence of glochidia might inhibit normal ventilation in the outer Aldridge DC (1997) Reproductive ecology of bitterling (Rhodeus gills. Consequently, water flow rates may be higher in sericeus Pallas) and unionid mussels. PhD Thesis, Cambridge the inner gills of female mussels compared with males, University, Cambridge Aldridge DC (1999) Development of European bitterling in the because the females must adjust their ventilation rates to gills of freshwater mussels. J Fish Biol 54:138–151 account for reduced ventilation in their outer gills. In the Bayne BL, Newell RC (1983) Physiological energetics of marine present study, bitterling spawned most frequently in the mollusks. In: Saleuddin ASM, Wilbur KM (eds) The Mollusca. two inner gills in all seasons. Academic, London, pp 407–515 Chambers RC (1997) Environmental influences on egg and prop- A difference between the mussel sexes in the mortality agule sizes in marine fishes. In: Chambers RC, Trippel EA (eds) of dead ejected embryos may not be detected mainly Early life history and recruitment in fish populations. Chapman because most eggs tend to be deposited in inner rather and Hall, London, pp 63–102 than the outer gills (Smith et al. 2004). The glochidia Coleman RM (1991) Measuring parental investment in nonspher- ical eggs. Copeia 1991:1092–1098 consume oxygen and fill the outer gill cavity (Tankersley Fujikura K, Segawa S, Okutani T (1988) Respiration rate and and Dimock 1993; Kondo 2002). Therefore, the presence ammonia excretion rate of freshwater mussel, Unio douglasiae of glochidia may adversely affect normal gill function (Griffith et Pidgeon) (in Japanese). Venus 47:207–211 and prevent the gills from receiving bitterling eggs Higashi S (1965a) The respiration of principal mollusks in Lake (Aldridge 1997; Smith et al. 2004). In the present study, Biwa-ko (in Japanese). Venus 23:229–237 Higashi S (1965b) Basal metabolism of principal mollusks in Lake bitterling embryos were rarely found in the outer gills, Biwa-ko (in Japanese). Venus 24:152–155 and some bitterling embryos in outer gills loaded with Huebner JD (1982) Seasonal variation in two species of unionid glochidia were killed by the attachment of the glochidia clams from Manitoba, Canada: respiration. Can J Zool 60:560– (J. Kitamura, personal observation). Despite this, the 564 Hunter MD, McNeil JH (1997) Host-plant quality influences dia- mortality of dead ejected embryos was not significantly pause and voltinism in a polyphagous insect herbivore. Ecology different between the mussel sexes (Table 2). 78:977–987 Imai C, Tanaka S (1987) Effect of sea water temperature on egg size of Japanese anchovy. Nippon Suisan Gakkaishi 53:2169–2178 Seasonal timing of bitterling spawning Kamler E (1992) Early life history of fish. Chapman and Hall, London The optimal period for bitterling to spawn may be bal- Kanoh Y (1996) Pre-oviposition ejaculation in externally fertilizing anced between the positive effects, i.e. higher embryonic fish: how sneaker male rose bitterlings contrive to mate. Ethology 102:883–899 growth rates, and the negative effects, i.e. lower oxygen Kanoh Y (2000) Reproductive success associated with territoriality, concentration, of a seasonal rise in temperature. The sneaking, and grouping in male rose bitterlings, Rhodeus ejection rate of younger live embryos was highest earlier ocellatus (Pisces: Cyprinidae). Env Biol Fish 57:143–154 in the spawning season (early to mid-April), whereas the Kawamura K (2003) Rhodeus ocellatus kurumeus In: Ministry of the environment in Japan. Threatened wildlife of Japan, Red mortality of older embryos in mussels increased as the data book, 2nd edn, vol 4. Pisces, brackish and freshwater fishes season progressed (after early June). The survival of (in Japanese). Japan Wildlife Research Center, Tokyo, pp 44– bitterling embryos in mussels was the highest from late 45 April to mid-May (Fig. 2). Oviposition by bitterling Kawamura K, Nagata Y, Ohtaka H, Kanoh Y, Kitamura J (2001) Genetic diversity in the Japanese rosy bitterling, Rhodeus occurred from early April to early July, with a peak in ocellatus kurumeus (Cyprinidae). Ichthyol Res 48:369–378 mid- to late May (Fig. 1). The period of highest embryo Kondo T (2002) Unionid mussels of Japan (in Japanese). Osaka survival rate overlapped with the period of highest ovi- Kyoiku University, Kashiwara City 51

Ministry of the Environment in Japan (2002) Important wetland Smith C, Reynolds JD, Sutherland WJ (2000b) The population 500 selection in Japan (in Japanese). Ministry of the Environ- consequences of reproductive decisions. Proc R Soc Lond B ment in Japan, pp186–229 267:1327–1334 Nagata Y (1985) Estimation of population fecundity of the bit- Smith C, Rippon K, Douglas A, Jurajda P (2001) A proximate cue terling, Rhodeus ocellatus, and ecological significance of its for oviposition site choice in the bitterling (Rhodeus sericeus). spawning habit into bivalves (in Japanese). Jpn J Ichthyol Freshw Biol 46:903–911 32:324–334 Smith C, Reichard M, Jurajda P, Przybylski M (2004) The repro- Nagata Y, Miyabe H (1978) Developmental stages of the bitterling, ductive ecology of the European bitterling (Rhodeus sericeus). Rhodeus ocellatus ocellatus (Cyprinidae). Mem Osaka Kyoiku J Zool Lond 262:107–124 Univ Ser III 26:171–181 Suzuki N (1988) Development and incubation method of Rhodeus Nakamura M (1969) Cyprinid fishes of Japan. Studies on the life ocellatus smith in the laboratory. In: Nagata Y (eds) Research history of cyprinid fishes of Japan (in Japanese). Research club of Japanese rosy bitterling, Rhodeusocellatus kurumeus (in Institute for Natural Resources, Tokyo Japanese). Shinseisha, Tokyo, pp 42–51 Nikolsky G V (1963) The ecology of fishes (in Russian). Japanese Suzuki M (2002) Individual flowering schedule, fruit set, and flower translation by K. Kamei, published 1970. Takara-shobo, and seed predation in Vaccinium hirtum Thumb (Ericaceae). Yonago Can J Bot 80:82–92 Pusch M, Siefert J, Walz N (2001) Filtration and respiration rates Suzuki M (2003) Effects of flower and seed predators and pollin- of two unionid species and their impact on the water quality of ators on fruit production in two sequentially flowering cong- a lowland river. In: Bauer G, Wachtler K (eds) Ecology and eners. Plant Ecol 166:37–48 evolution of the freshwater mussels unionoide. Springer, Berlin Suzuki N, Hibiya T (1984) Minute tubercles on the skin surface of Heidelberg New York, pp 317–326 larvae of Rhodeus (Cyprinidae) (in Japanese). Jpn J Ichthyol Reynolds JD, Guillaume H (1998) Effects of phosphate on 31:198–202 the reproductive symbiosis between bitterlings and freshwa- Suzuki N, Hibiya T (1985) Minute tubercles on the skin surface of ter mussels: implications for conservation. J Appl Ecol 35:575– larvae of Acheilognathus and Pseudoperilampus (Cyprinidae) (in 581 Japanese). Jpn J Ichthyol 32:335–344 Saijyo Y, Mitamura O (1995) New book, study method of lake and Tankersley RA, Dimock RV (1993) The effect of larval brooding pond (in Japanese). Koudansya, Tokyo on the respiratory physiology of the freshwater unionid mussel Smith C, Reynolds JD, Sutherland WJ, Jurajda P (2000a) Adaptive Pyganodon cataracta. Am Mid Nat 130:146–163 host choice and avoidance of superparasitism in the spawning Tesch, FW (1971) Age and growth. In: Ricker WE (ed) Methods decisions of bitterling (Rhodeus sericeus). Behav Ecol Sociobiol for assessment of fish production in fresh waters. IBP Hand- 48:29–35 book 3. Blackwell, Oxford, pp 98–130