Food habits of three non-native cichlid fishes in the lowermost Chao Phraya River basin, Thailand

Abstract The food habits of three non-native cichlid fishes, Mayan cichlid (Mayaheros urophthalmus), Mozambique tilapia (Oreochromis mossambicus) and Nile tilapia (O. niloticus), in the lowermost Chao Phraya River basin, Thailand, was examined by stomach contents analysis. The index of preponderance, an index of the importance of prey items, was calculated from two relative metrics of prey quantity: percent frequency and percent volume. The index of niche breadth and the overlap coefficient were calculated to compare the breadth of food habits among the size classes and species groups. The M. urophthalmus mainly preyed on fish scales, detritus and aquatic invertebrates (molluscs and crustaceans). The O. mossambicus and O. niloticus fed mostly on detritus. The diets of the latter two species overlapped almost completely; however, green filamentous algae mixed with detritus was observed in the diet of the O. niloticus only. The observation that fish scales were a predominant food source in the stomach of M. urophthalmus (high importance value 45.48%) was specific to this study area.


Introduction
Nile tilapia, Oreochromis niloticus, and Mozambique tilapia, O. mossambicus, are common African cichlid fishes that have been introduced widely all over the world for aquaculture (Canonico et al. 2005;De Silva 2006). They now can be found throughout most of Southeast Asia (Arthington et al. 1984;Ang et al. 1989;Chou and Lam 1989;Eidman 1989;Juliano et al. 1989;Piyakarnchana 1989;Welcomme and Vidthayanon 2003;Canonico et al. 2005;Peterson et al. 2005;Weyl 2008;Grammer et al. 2012;Lowe et al. 2012;Russell et al. 2012;Welcomme et al. 2016). In addition, the Mayan cichlid, Mayaheros urophthalmus, (and hybrids of this species), which is endemic to the Atlantic slope of Central America, has been traded as an aquarium fish worldwide (Nico et al. 2007). It has been documented in some parts of Southeast Asia, including Singapore, Sulawesi in Indonesia, around Manila Bay in the Philippines and the lowermost Chao Phraya River basin in Thailand (Nico et al. 2007;Herder et al. 2012;Kwik et al. 2013;Ordoñez et al. 2015).
In Thailand, all three of these fish species were introduced into the wild in the middle of the last century due to escaped aquarium trade (Welcomme and Vidthayanon 2003). The lowermost Chao Phraya River basin is a sympatric region where all three fishes can be found in Thailand. Recently, the Thai government banned the rearing, release and transportation of four alien cichlid species in Thailand, including O. mossambicus and M. urophthalmus (The Nation, 2018). While reporting the establishment of non-native fishes in new region is a critical first step for reducing ecological impacts of invasion (Corsini-Foka and Economidis 2007;Galil 2008), it also is important to determine the role the non-native organisms play in their new habitat. This provides a clear picture of the issues necessary to mitigate the ecological problems associated with the invasion. One of these important aspects is their diets (Hughes 1986;Getabu 1994;Kadye and Booth 2012;Kalogirou et al. 2012).
Some preliminary studies of diet in these three non-native cichlid fishes showed that exotic O. niloticus and indigenous O. mossambicus mostly consumed plant detritus in the Limpopo River system, South Africa (Zengeya et al. 2011). Gut contents analysis suggested that the main food source of O. niloticus was plant material in the lower Nile River basin (Abdelghany 1993). However, Njiru et al. (2004) showed that this species consumed various prey items, such as insects, plants, fishes and small aquatic invertebrates. Oreochromis mossambicus has been known as an opportunistic omnivore that consumes terrestrial vegetation, aquatic plant materials, zooplankton, aquatic and terrestrial macrovertebrates, tadpoles and fish (Wager and Rowe-Rowe 1972;Bruton and Boltt 1975;Mathavan et al. 1976;De Silva et al. 1984;DeMoor et al. 1986;Jameson 1991;Arthington and Bluhdorn 1994;Fuselier 2001;Komarkova and Tavera 2003). One study in Asia found that O. mossambicus acted as a detrivore, herbivore and carnivore in different lakes in Sri Lanka . The diet of M. urophthalmus ranged from herbivore, to omnivore, and carnivore depending on sites in their native range in Mexico and Belize (Caso Ch avez et al. 1986;Martinez-Palacios and Ross 1988;Ch avez-L ovez et al. 2005;Vaslet et al. 2012). Bergmann and Motta (2005) reported that M. urophthalmus was an omnivore in south Florida, where this species had been established as an exotic species. From these previous studies, it is clear that the food habits of these three cichlid fishes can vary widely depending on the location where they were introduced.
To date, comprehensive investigation of the food habits of the three non-native cichlid fishes is lacking in the lowermost Chao Phraya River basin. Nico et al. (2007) conducted a limited survey of stomach contents of M. urophthalmus and suggested that this species was an omnivore, which was consistent with most previous studies. In the current study, we performed quantitative stomach contents analysis of these three non-native cichlid fishes to determine diet composition in the lowermost Chao Phraya River basin.

Collections
The Chao Phraya System is one of the largest river systems in Southeast Asia, (approximately 372 km in length and 160,000 km 2 in area) and serves as the origin of four rivers (the Pin, Wan, Yom and Nan Rivers) that start in the highlands of northern Thailand (Van BeeK 1995). This system is divided into an upper and lower basin by a narrow section at the confluence of these four rivers in Nakhon Sawan Province (Van Beek 1995; Komori et al. 2012). The Chao Phraya System joins with the Sakae Krang River and the Pa Sak River, and the Tha Chin River branches off at Chai Nat Province in the lower basin of this river system. The so-called 'Chao Phraya River' normally is the mainstream part of the lower basin of the river system (Van Beek 1995; Komori et al. 2012). The present study was conducted at the lowermost Chao Phraya River basin which is defined here as a delta area expanding along the main stream from the point where this river and the Pa Sak River meet to the mouth of the river. The lower Chao Phraya River basin previously formed a vast delta area, an amphibious environment that was not suitable for human life. The delta was developed when European colonials arrived and a number of canals were constructed in the area (Takaya 1987). These canals, which provide native aquatic animals with artificial refuges and habitats, were the main locations of investigation in this study (Figure 1).
The study sites were selected from sampling locations where M. urophthalmus, O. mossambicus or O. niloticus had been collected in previous work from 2007 to 2010 by the Research Laboratory of Ichthyology, Faculty of Fisheries, Kasetsart University (RLIKU) (Prachya Musikasinthorn, unpublished data). When adequate sampling sites were encountered or found from observations during fieldwork or interviews with local fishermen, additional sampling was conducted at these sites.
All fish were collected using cast nets and scoop nets in small canals (number of sites = 24; Figure 1), 50 to 130 cm in depth, in the lower Chao Phraya River basin from September 2014 to October 2015. The collected fish were anaesthetised in ice water and immediately fixed in 10% formalin. For large specimens, formalin stock liquid was injected into the stomach directly to prevent digestion. After 10 days in formalin, the specimens were washed in fresh water and then stored in 75% ethanol. The number of specimens varied from 1 to 33 per collection, depending on study location.

Stomach contents analysis
The stomachs were extracted from the specimens after measurement of standard length (SL) and wet weight (WW) using calipers and a gravimeter. SL was measured from the tip of the upper jaw to the posterior end of the midlateral point of the hypural plate. The stomach contents were removed and sorted into 12 major functional categories (detritus, mixture of detritus and filamentous algae, filamentous algae, plant matter, benthic crustaceans, molluscs, adult fishes, fish bones, fish larvae, fish eggs, fish scales and artificial matter). For comparison among growth stages, specimens were divided into three size classes in each species group when all analyses were performed ( Table 1).
The contribution of each prey functional category to the overall stomach contents was assessed using two indices: percent frequency of occurrence (% F) and percent composition by volume (%V) (Hyslop 1980;Bowen 1996) (Table 2). %V was estimated using a modified method of Platter and Potter (2001) by evenly spreading the contents of the stomach on waterproof paper ruled into 1 mm squares in approximately 1 mm thickness and examining them under a microscope. The number of squares occupied by each food category on the waterproof paper was later converted to the percent volume for each food category.
For the assessment of the importance of each prey category, the index of preponderance (IOP a ) (Natrajan and Jhingran 1962;Sreeraj et al. 2006) was calculated: where S is the number of prey types, F a is the frequency of occurrence of species a and V a is the percent composition by volume of species a. To facilitate comparisons among species, IOP a was converted into percent IOP a (%IOP a ).
To estimate prey item diversity of the three non-native cichlids, the index of niche breadth (INB) (Smith and Zaret 1982) was calculated for each individual stomach as where n is the number of prey items and p i is the proportion of the total samples belonging to the ith prey item group (Smith and Zaret 1982).
To express dietary overlap between cichlids of different species and sizes, the overlap coefficient (O kj ) was calculated using volume (%V) following Schoener (1970): where p ik and p ij are the proportions of the total samples belonging to the ith prey item group of species k and species j, respectively. The value of O kj may range from zero (indicating a specialised diet or no overlap) to 1.0 (indicating an equal use of food resources or complete overlap), with values >0.6 considered to indicate a biologically significant overlap (Wallace 1981;Langton 1982).
To test for significant differences in diet between cichlids of different species and sizes, analysis of similarity (ANOSIM) was performed on the Bray-Curtis similarity matrix using the vegan package in R (Oksanen et al. 2018) on statistic software R version 3. 4. 4 (R Core Team 2017). ANOSIM is a non-parametric test that uses a permutation procedure applied to a ranked similarity matrix to test whether there is a significant difference between two or more groups (Doi and Okamura 2011). Though this statistic was developed for testing differences of community composition among study sites, it also has been used to test differences of food items among taxonomic group and size classes (i.e. Platter and Potter 2001; Maddern et al. 2007;Zengeya and Marshall 2007;Muñoz et al. 2011). This study area was located in a tropical monsoon climate, and the mean monthly temperature during the study period was 29.1 C and ranged from 26.5 to 31.1 C (Japan Meteorological Agency 2008). There likely were differences in rainfall and water depth between dry and rain seasons. However, most of this study area was located in coastal areas, where the environment was affected by more tides rather than rainfall (Takaya 1987).

Results
In M. urophthalmus, the SL ranged from 53.6 to 134.7 mm, with a mean (± SE) of 79.5 ± 18.7 mm; and WW ranged from 6.4 to 98.1 g, with a mean of 24.6 ± 20.1 g. In O. mossambicus, the SL ranged from 66.2 to 164.2 mm, with a mean of 115.5 ± 23.2 mm; and WW ranged from 8.6 to 138.3 g, with a mean of 57.4 ± 32.9 g. Finally, in O. niloticus. SL ranged from 76.5 to 183.5 mm, with a mean of 119.6 ± 27.9 mm, while WW ranged from 15.7 to 232.5 g, with a mean of 74.5 ± 52.4 g. On the whole, specimens of M. urophthalmus were smaller and no individuals of this species were placed in the third size class (151-200 mm) ( Table 1).
The IOP value revealed that overall size classes, the most important dietary item was detritus (Table 3). In O. niloticus there were no significant differences in diet among size classes (ANOSIM, R = -0.09, P > 0.1), and the overlap coefficients ranged from 0.99 to 1.00, indicating an almost complete overlap in diet (Table 4). There was a tendency for the INB to decrease through the size classes (Table 3).
For O. niloticus, there there were no significant differences in diet among the size classes (ANOSIM, R = -0.07, P > 0.1), and the overlap coefficients ranged from 0.96 to 1.00 (Table 4). The most important food source was detritus over all the size classes ( Table 3). The INB values of O. mossambicus were nearly the same for all the size classes (Table 3). There were no significant differences in the diets between O. mossambicus and O. niloticus (ANOSIM, R = -0.02, P > 0.1). However, the overlap coefficient in diets between the two species was high overall size classes, ranging from 0.96 to 1.00 (Table 4). The secondary prey items of O. mossambicus were crustaceans, fish eggs and fish larvae. In contrast, the diet of O. niloticus also contained filamentous algae and no aquatic  invertebrates overall size classes (Table 3). According to additional observations under the binocular stereomicroscope, the larvae in the diet of O. mossambicus were larvae of cichlid fishes. No significant differences in important prey items among size classes were observed in M. urophthalmus (ANOSIM, R ¼ 0.03, P > 0.1). The IOP results showed differences in important prey items among size classes in M. urophthalmus. In addition to detritus as a main prey item, other important food resources for smaller-sized fish (MU1) were fish scales, whereas crustaceans and molluscs were the main food resources for the larger-size class (MU2) ( Table 3). The scales from stomachs of MU1 probably were eaten from live fish, because other fish body fractions, such as eyes, skins and muscle, were not found in the stomach. The INB value of M. urophthalmus in the smaller-sized fish (MU1) was smaller than in the larger-sized fish (MU2) ( Table 3). Additional comparisons of scale morphology among the one removed from stomachs and the one from fish specimens under a microscope showed that the scales in the diet of M. urophthalmus came from all species of three non-native cichlid fishes. All adult fish found in the diet of M. urophthalmus were identified as non-native mosquito fish (Gambusia affinis).
There were significant differences among the three cichlid fishes in their prey items (ANOSIM, R ¼ 0.25, P < 0.01). The INB value of M. urophthalmus was the highest of the three cichlid fishes, indicating that M. urophthalmus utilised more widely varied prey resources, particularly fish in the smaller-size class, compared with the other two cichlid fishes.
The overlap coefficient was higher than 0.6 in all the size classes of O. niloticus and O. mossambicus as well as among larger-sized M. urophthalmus (Table 4), indicating that prey items of these groups overlapped significantly. The overlap coefficient was significantly higher in smaller (MU1) than in larger M. urophthalmus. No significant overlap was observed with the other size groups of the other two tilapiine fishes (Table 4).

Discussion
Previous work has shown that the food habits of the non-native cichlid fishes can be hihgly variable depending on where they are sampled. For example, M. urophthalmus acted as an herbivore, omnivore or carnivore in three different lagoons in Mexico (Caso Ch avez et al. 1986;Martinez-Palacios and Ross 1988;Ch avez-L ovez et al. 2005). Thus, in order to determine the food habits of these three species in the lowermost Chao Phraya River basin, it must be established in their new environment.
The present study found that the main food resource of O. niloticus and O. mossambicus was detritus, with filamentous algae and small aquatic animals as secondary prey items of these species, respectively. By contrast, the diet of M. urophthalmus was more variable and changed with size classes. The main food item of small fish (MU1) was fish scale, but Bold characters indicate significant (>0.60) overlapping among groups (Wallace1981; Langton1982). a Size classes are defined in Table 2.
larger fish (MU2) tended to consume various aquatic animals, such as crustaceans and molluscs. Prey items of the three cichlid fishes mostly overlapped, except that only small M. urophthalmus ate fish scale.
The main food resources of O. niloticus in the lowermost Chao Phraya River basin were consistent with previous studies, that examined the proportion (%) by volume of food items (Zengeya et al. 2011(Zengeya et al. , 2013. However, other studies conducted in its native range found that O. niloticus was omnivorous and consumed diverse prey items, including various insects, algae and crustaceans, depending on the surroundings (Njiru et al. 2004;Canonico et al. 2005;Oso et al. 2006). Oreochromis mossambicus are normally recognised as opportunistic omnivores in their native and introduced ranges (Wager and Rowe-Rowe 1972;Bruton and Boltt 1975;Mathavan et al. 1976;De Silva et al. 1984;DeMoor et al. 1986;Jameson 1991;Arthington and Bluhdorn 1994;Fuselier 2001;Komarkova and Tavera 2003). For example, by examining the percentage of dietary contribution by volume and occurrence, Maddern et al. (2007) documented that this species utilised various food items, like algae, vegetable matter, terrestrial insects, diptera, ephemeroptera and silt/biofilm … etc. in Western Australia.
The findings from the current study supported the work of Zengeya et al. (2011), which was conducted in a non-native habitat that also included O. niloticus and O. mossambicus. Zengeya et al. (2011) suggested that O. mossambicus and O. niloticus in the Limpopo River in South Africa switched their primary food resource in favour of more detritus due to the recession of the river in the dry season. Thus, it also may be possible that the preferred prey of these two species may be lacking in the lowermost Chao Phraya River basin, a hypothesis that will need to be investigated further.
There is some evidence that O. mossambicus uses fish eggs and larvae as food resources (Berrios-Hernandez and Snow 1983;Macintosh and de Silva 1984;Panastico et al. 1988). However, in our study, only mature females consumed fish eggs and larvae (n ¼ 12, SL = 115.5 ± 13.3 mm), suggesting that the eggs and the larvae found in their stomach was that probably were swallowed by mouth-brooding individuals accidentally when they were captured in the cast net. The prey items of M. urophthalmus differed among size classes. The main prey items were detritus with fish scales consumed by the smaller-size class (MU1), and detritus with crustaceans and molluscs in fish of the larger-size class (MU2). The life history of M. urophthalmus is diverse, depending on the environment, and its food habits can vary widely (Loftus 1987;Martinez-Palacios et al. 1990;Martinez-Palacios and Ross 1992;Faunce and Lorenz 2000;Bergmann and Motta 2005) (Table 5). For example, there are clear differences in the diet of M. urophthalmus in their native and introduced range. In their native range in Belize, their main food items were crustaceans, detritus and plant material. However, in non-native Florica, they primarily consumed fish, ostracods and filamentous algae (Bergmann and Motta 2005;Vaslet et al. 2012) (Table 5). In the current study, M. urophthalmus fed as an omnivore. Some previous studies also found fish scales in the M. urophthalmus diet, but they made up a small contribution to overall prey (e.g. Martinez-Palacios and Ross 1988;Ch avez-L ovez et al. 2005). In the study of Martinez-Palacios and Ross (1988), fish scales were found (%V ¼ 0.45 -5.24) in the larger fish (>150 mm), whereas they were found in the diet of smaller fish in this study. In our study, fish scales were an important secondary prey item, whereas in Martinez-Palacios and Ross's study, fish scales were one of many options.
To the best of our knowledge, this study is the first study to investigate the food habits of non-native fishes in the lowermost Chao Phraya River basin. The capital, Bangkok, is located in the centre of this area and the physical environment and the biota of aquatic ecosystems in this area are continually changing. Cichlid fishes change their food habits depending on environmental changes in the region (e.g. Canonico et al. 2005). The continuous investigation of the food habits of cichlid fishes in this area is necessary to evaluate the ecological impacts on native ecosystems. Because there are prey items of animal origin in the diet of M. urophthalmus and this species has become established relatively recently, M. urophthalmus should be monitored carefully. Additionally, in the future, an extension of the methodology is needed to evaluate the broader impact of non-native cichlid fishes on native ecosystems because it is difficult to determine the actual prey items that are absorbed and assimilated in the body and their effects on food web dynamics (Okuda 2012). It will be important to undertake more detailed descriptions of the microhabitats, to determine the stomach contents of native fishes and other aquatic animals and to utilise more powerful ecological tools that can investigate trophic levels and food webs.
This study has shown that the ecology of non-native fishes is not always the same in an introduced region, even for the same species. Thus, research on non-native fish ecology should be performed site-specifically and multidimensionally, taking into account ecological, geological and human cultural factors. Site-specific research will be essential in the future, considering the possible impacts of non-native fishes on the environment, particularly in areas where non-native fishes are introduced frequently and their actual status is not clear.