Reproductive biology of Garra regressus and Garra tana (Cypriniformes: Cyprinidae) from Lake Tana, Ethiopia

(Cypriniformes: Cyprinidae) from Lake Tana, Ethiopia Akewake Geremew 1, Abebe Getahun 2 & Eshete Dejen 3 1 Department of Biology, Dilla University, P.O. Box 419, Dilla, Ethiopia 2 Department of Zoological Sciences, Addis Ababa University, P. O. Box 1176, Addis Ababa, Ethiopia 3 Agriculture and Environment Division (AED), Inter-Governmental Authority on Development (IGAD), Djibouti 1 khaliger@yahoo.com (corresponding author), 2 abebe12002@yahoo.com, 3 eshete.Dejen@igad.int A)


INTRODUCTION
Lake Tana, an exceptional hotspot of freshwater fish diversity and endemism in Ethiopia, is an important natural laboratory for evolutionary studies (Sibbing et al. 1998). The lake contains an ichthyofauna represented by seven genera, viz.: Barbus, Clarias, Garra, Labeobarbus, Nemacheilus, Oreochromis and Varicorhinus. Out of the 28 species currently recognized, 20 are endemic to the lake and its catchments (Vijverberg et al. 2009). However, most endemic species are facing serious population declines as a result of habitat degradation and overfishing (de Graaf et al. 2006;Darwall et al. 2011). An important group of fish threatened by the above stressors in Lake Tana is the genus Garra (Getahun 2010a,b;Darwall et al. 2011), represented by four species (Garra tana, G. regressus, G. dembecha and G. dembeensis), of which two (Garra tana and G. regressus) are endemic to the lake (Stiassny & Getahun 2007).
While G. regressus mainly inhabits the sub-littoral zones with rare occurrences in the pelagic and littoral zones, G. tana occurs in the pelagic zone with rare occurrences in the littoral zone of the lake (Geremew 2007). Both species are, however, absent from the tributary rivers of Lake Tana indicating its lacustrine adaptation (Geremew 2007). The two species are heavily impacted by overfishing primarily through bycatch, and have been listed as 'Vulnerable' in the IUCN Red List of Threatened Species (Getahun 2010a,b).
The purpose of this paper is to investigate the various aspects of the reproductive biology of G. regressus and G. tana from Lake Tana, including sex ratio, gonadosomatic index, fecundity, spawning season, and maturity so as to generate baseline information for conservation action.

Study Area
Lake Tana (12 0 N & 37 0 21'E) is Ethiopia's largest lake (3150km 2 that accounts for 50% of the lacustrine area in Ethiopia). It is formed by the volcanic blocking of the Blue Nile and is situated in the northwestern highlands, 560km north of the capital, Addis Ababa, at an altitude of 1830m (Nagelkerke et al. 1994). It is a shallow, oligomesotrophic lake (Dejen et al. 2004) fed by seven big perennial rivers with the Blue Nile being its only outflow (Fig. 1). The Lake Tana basin is isolated from the lower parts of the Nile basin by a 40m high waterfall   (1999). Accordingly, the fish were categorized as immature (I), recovering spent or developing virgin (II), ripening (III), ripe (IV) and spent (VI). Gonadosomatic index, which is the gonad weight as a percentage of total body weight, was calculated for ripe stage females.
The mean size at first sexual maturity, i.e., the standard length at which 50% of the males and females mature (L 50% ), was determined using the software package called "pasgear 2" (Kolding & Skålevik 2010) for both the sexes. Sexual maturity in stage III were considered mature. The percentage of mature fish per length class was calculated and SL 50 was estimated following Gunderson et al. (1980).
1 + e (-aL+b) Where, P m is percent mature (= % Mature) at length L, and bare fitted constants.
The relationship between the percentage of mature fish (P m ) per length class and fish length (Standard Length, SL in mm) was described using a logistic curve: 1n P m / (1 -p m ) = a + bSLi where, a is the intercept and b the slope of the curve. From the sigmoid curve SL 50% = -a/b was determined.
Ripe (stage IV) ovaries preserved in Gilson's fluid were considered for fecundity and egg-size determination. Fecundity estimation was carried out using the gravimetric method (Bagenal & Braum 1978). For ova diameter estimation, eggs having a diameter ≥0.75mm were measured to the nearest 0.01mm using an ocular micrometer.
Chi-square test was used to determine general and seasonal variations in sex ratio.
All statistical analyses were performed using Minitab v14 and the results were reported as mean ± SE (standard error).

Breeding season, sexual maturity, fecundity, egg size determination and sex ratio
Although there was insignificant variation in the mean monthly GSI of female G. regressus (ANOVA, F 6, 36 = 0.9, P = 0.507), the mean values ranged from 7.22 to 14.98 from April to October. Initiation of breeding activity is closely tied to the beginning of the rainy season, followed by the breeding during the rains (see Fig. 3a and Fig. 4a).
Seasonal variation in the GSI was evident for female specimens of G. tana. The mean monthly GSI values ranged from 7.61-17.88. The GSI varied significantly between sampling periods (ANOVA, F 11, 172 = 3.77, P< 0.001) with higher values recorded between March and June (Fig. 3b). After June, the GSI values progressively dropped, and lower values were recorded during January, October and November. The highest individual GSI (40.45%) as well as the maximum mean monthly GSI (17.94%) in G. tana was observed during June. The cycle in GSI was also reflected in the monthly variation in the frequency of females with ripe ovaries (Fig. 4b) with low frequencies encountered during periods of low GSI values. Although, female G. tana in breeding conditions were caught throughout the sampling period, considerable proportions were found in the catches from March to July.
There was no noticeable sexual dimorphism in both species. Though it was difficult to observe the spawning behavior, it is possible to indicate the preferred spawning location from the distribution patterns of stage IV females. A higher proportion (78.8%, n=41) of stage IV females of G. regressus occupied the sub-littoral zone than other regions.
For G. tana, the preferred spawning location as seen through the distribution patterns of stage IV females was the pelagic area (87%, n=269), and were rarely seen in the littoral habitats.
The mean size at first maturity depicted in Fig. 5 was significantly different between the species only for females, since the confidence interval of males overlapped at SL 50 (Table 1). Only individuals in the third or higher stages of gonadal maturity have been considered to be mature.
Garra regressus females start maturing earlier (82.6mm) than males (85.3mm). All G. regressus males and females over the lengths 124.4mm and 125.0mm respectively are mature.

7229
frequency distribution of eggs showed bimodality and ovaries contained two kinds of oocytes (Fig. 7a): eggs with yolk (vitellogenic oocytes) and eggs without yolk.
Only the vitellogenic eggs were counted to determine fecundity because only these oocytes would be spawned in subsequent peak. The size frequency distribution of eggs showed unimodality in G. tana (Fig. 7b). From a total of 149 G. regressus examined, the sex ratio did not significantly differ from 1:1 during any period, except that in the total catch the ratio was female-biased (chi-square= 4.2, df= 1, P= 0.0404) ( Table  3). From a total of 745 individuals of G. tana examined in 12 sampling occasions, a significant deviation from a 1:1 (female: male) ratio occurred only during March and April, when females were more numerous, and during August when males were numerous (Table 3).

DISCUSSION
The results of GSI and frequency of females with ripe gonads, suggest that the two Garra species in Lake Tana has an intensive breeding activity coinciding with the major rainy season of the area. Therefore, biotic and abiotic factors that are associated with rainfall seem to be important cues to trigger breeding in the two species. Synchronization of sexual maturation and reproduction with rainy season is well documented in tropical

Reproductive biology of Garra regressus and G. tana
Geremew et al.

Figure 7. Ova diameter frequency histograms of (a) Garra regressus and (b) G. tana from Lake Tana
regime (i.e., the decrease in maximum and increase in the minimum air temperature (Fig. 2)) during the rainy season could possibly trigger reproductive activity, as observed in Labeo horie (Dadebo et al. 2003) and G. rufa (Bardakci et al. 2000). The relationship between intensive breeding activity and the rainy season has also been reported for other species of the genus Garra? (Jha et al. 2005).
The study also revealed considerable differences in reproductive periodicity of the two Garra species with a moderately long breeding season in G. regressus to an essentially continuous breeding in G. tana. Similar patterns of reproductive activity are known for other species of Garra in Iran (Abedi et al. 2011) and in characid fishes inhabiting a tropical stream in Panama (Kramer 1978). In this study no spatial differences in distribution patterns of Garra were observed between peak breeding periods and other periods.
The mean size at sexual maturity was similar for both species (and sexes) of Garra. Mean length at maturity (SLm) is determined to predict size-assortative mating as well as management options by restricting mesh sizes to avoid over-exploitation, though exploitation is non-existent in this case. Size at maturity could vary among closely related species, among populations within species and among individuals within populations, suggesting that they can respond rapidly to natural selection (Stearns 1992cited in Tsikliras & Antonoupoulou 2006. Length of maturity in many fish species depends on demographic condition, and is determined by both genes and the environment (Fryer   1972;Lowe-McConnell 1987;Wootton 1998); but can also be influenced by other factors such as longterm fishing pressure, which cannot be the case here. The fecundity of the two Garra species increased in proportion to length, gonad weight and body weight. The positive correlation of absolute fecundity and body length has been reported for other Garra species (e.g., G. ceylonenesis in Sri Lanka) (de Silva 1991). The parameter 'b' of the relationship of length with fecundity for G. tana (b = 2.2) is higher than that of G. rufa (b =1.8) in Iranian waters (Abedi et al. 2011) and G. ceylonensis (b = 1.22) (de Silva 1991). But for G. regressus, this value (b = 1.5) is very close to the Asian species (G. ceylonensis and G. rufa). However, the 'b' value is generally comparable to previously reported values for other fish species (Bagenal 1978). In many fish species 'b' value is usually around three when fecundity is related to length, and about one when related to weight (Bagenal & Braum 1978). The relatively lower regression coefficient as well as weaker correlation between fecundity and standard length (Fig. 6) for both species in this study as well as for G. rufa (Abedi et al. 2011) andG. ceylonensis (de Silva 1991) indicates that fecundity could vary much within each size class. According to our findings, the coefficient of correlation for the relationship between fecundity and fish weight was 0.4 and 0.7, and for the relationship between fecundity and gonad weight 0.48 and 0.87 for G. regressus and G. tana, respectively. Thus, it is clear that the gonad weight has a better correlation with reproductive capacity than the body weight, as is the case in G. rufa (Abedi et al. 2011).
The high variability in absolute fecundity observed in these species could also be the result of either genetic differences among the females, or environmental conditions, or a combination of both (Wootton 1998;Rideout & Morgan 2007). The reproductive potential as reflected in the increase in absolute fecundity with length indicates that Garra allocates more energy to reproduction as it grows (Abedi et al. 2011).
The relative fecundity in Garra species of Lake Tana is low in comparison with other species of Garra; G. rufa from Iran (mean = 86.84 eggs/ gram) (Esmaeili et al. 2005) and G. ceylonenesis from Sri Lanka (mean = 866.7 eggs/gram) (de Silva 1991). However, G. tana has a comparatively higher relative fecundity (102 eggs/ gram) than G. regressus. This might be a strategy for this abundant and highly predated prey species (Nagelkerke et al. 1995;de Graaf et al. 2000) in the deep off-shore part of the lake or possibly a difference in the reproductive investments or genetic variability (Rideout & Morgan 2007). The effect of predation on one of the most abundant pelagic species in Lake Tana, B. tanapelagius was earlier reported in which the fish showed a larger relative fecundity at smaller length class (Dejen 2003). The piscivores in Lake Tana have their optimum prey size at 40-50 mm (de Graaf 2003), which is smaller than the modal length classes of Garra species (Geremew 2007). This might then indicate the lesser predation pressure on Garra species by the piscivores in the Lake. Thus, it is likely that less variability at different size classes in relative fecundity of Garra observed might be a different strategy to the one reported in B. tanapelagius (Dejen 2003).
The result of egg size frequency distribution showed that G. tana is a single spawner similar to G. ceylonensis (de Silva 1991;Sundrabaranthy et al. 2005). However, G. regressus with an extended breeding season showed two distinct peaks in their ova size frequency distribution, suggesting multiple spawning. Multiple spawning helps to reduce competition for nursery and spawning sites by partitioning their use in time, helps in assuring the survival of some portion of the total spawn to recruitment, and provides more adaptability in highly variable environments (Rinchard & Kestemont 1996;Dejen 2003). Lower fecundity estimates in G. regressus reported here for a comparable size class may be related to the multiple spawning nature of this fish, or may be due to the larger mean size of its eggs as compared to G. tana (Table 3). It is known that smaller sized fishes with large sized eggs have smaller fecundities and produce multiple clutches of eggs in a season (Kramer 1978;Paugy 2002).
Female specimens predominated over males in G. regressus. However, G. tana showed no significant variation in the proportion of females over males from unity. It should be noted that this result does not necessarily indicate the sex ratio of the total population in the lake. This is because several factors may have an interactive effect as discussed in Admassu (1994) and Tadesse (1997). Sex based differences in activity associated with breeding behavior might be the reason for sex biased catches in some of the breeding months. For instance, G. tana showed preponderance of females in March and April which are periods at the start of peak breeding activity. On the other hand, males dominated in August, which is after the end of the peak breeding months. In G. regressus, although not statistically significant to support a deviation from theoretical one-to-one value, females were numerous than males in the catch during the peak breeding months of June and July. So without additional information on the spawning behavior of these species, it would be difficult to eliminate this factor.
The present study provides vital information on the main spawning period and spawning area to inform conservation plans and fisheries management plans for G. tana and G. regressus. Thus, the proposed smallmeshed fishery for small barbs in the lake should also consider Garras, especially in the pelagic habitat as this is shared with B. tanapelagius (Dejen et al. 2009). But prior to any management decision on small-meshed fisheries, studies on growth, mortality, biomass and potential yield of Garra stock is required.