Unsuitability of Cnesterodon decemmaculatus (Jenyns, 1842) for mosquito control in Uruguay: evidence from food-preference experiments



We analyzed the feeding preference of Cnesterodon decemmaculatus, a small-bodied poecilid native from the Rio de la Plata and proximate Atlantic Basins in South America. This species has a wide distribution in Uruguayan water bodies but its effectiveness as a predator of mosquito larvae has not been tested. In laboratory trials, five aquatic invertebrates were offered simultaneously as potential prey to fish: Daphnia pulex (Cladocera), copepods, two different instars of mosquito larvae (Culex pipiens), and the 4th instar of Chironomidae larvae. Preference was measured by the Chesson's electivity index (α). In order to determine differences in prey preference according to fish size, individuals ranging from 9.5 mm to 35.3 mm were classified in three different body size classes: small, medium, and large. Small fish showed preference for copepods, while medium-sized fish preferred the smallest mosquito larvae instars and Chironomidae larvae. We conclude that C. decemmaculatus is a zooplankton facultative-feeder fish that prefers large-bodied zooplankton but is a weak predator of mosquito larvae. Thus, the introduction of C. decemmaculatus as a biological-control agent in natural environments is not an effective strategy.


The use of poecilid fish species for biological control of mosquitoes is a world-wide practice (Meisch 1985). Western mosquitofish, Gambusia affinis (Baird and Girard), and the eastern mosquitofish, G. holbrooki (Girard), are commonly used for mosquito control (Wooten et al. 1988). Both species are native to North America and their high tolerance for a wide range of environmental conditions made possible its introduction in different environments (Meisch 1985). Eastern and western mosquitofish are omnivorous, opportunistic generalist species, and their broad diets include insect larvae, crustaceans, algae, and fish fry (Harrington and Harrington 1961, Sublette et al. 1990), depending on prey relative abundance. Mosquito larvae usually represent a small fraction of their diet (Farley 1980, Arthington and Marshall 1999), which is mainly composed of ostracods and copepods (Bence and Murdoch 1986, Bence 1988), particularly cladocerans (Farley 1980, García-Berthou 1999). Because these poecilids have no preference for mosquitoes, their success as biological control agents has been questioned (Courtenay and Meffe, 1989, Blaustein, 1992). In turn, the introduction of Gambusia affinis and G. holbrooki has altered the food webs of native aquatic populations (Arthington and Marshell 1999, Goodsell and Kats 1999, Leyse et al. 2004), and contributed to increased phytoplankton development as a result of their predation impact on large-bodied zooplankton (Hurlbert et al. 1972, Hurlbert and Mulla 1981, Margaritora et al. 2001).

Cnesterodon decemmaculatus (Jenyns 1842) is a small-bodied poeciliid (<5 cm) (Welcomme 1988) broadly distributed from central Argentina to southern Brazil, particularly in the lower Rio de la Plata and Atlantic Basins (Rosa and Costa 1993). This fish species is abundant in low energy and vegetated environments, as well as in degraded systems with low water quality and fish diversity, and can tolerate wide variations in physico-chemical parameters (Bistoni et al. 1999, Hued and Bistoni 2004). C. decemmaculatus is omnivorous, with a diet mainly composed of algae, detritic organic matter, small-sized invertebrates (Escalante 1983) and zooplankton (Quintans et al. 2009). Moreover, laboratory experiments showed that C. decemmaculatus is able to prey on Culex pipiens mosquito larvae when offered as the only choice (Martí et al. 2006).

In Uruguay, governmental authorities have implemented a biological control program for mosquitoes, notably including Culex pipiens and also the increasingly reported Aedes aegypti (Rossi and Martínez 2003, Etchebarne et al. 2006). Through this program, C. decemmaculatus is being spread in natural water bodies close to urban areas, based only on its morphological similarity with Gambusia spp. However, there is actually no evidence that supports the usefulness of C. decemmaculatus as an effective biological control of mosquito larvae. The lack of quantitative information about its diet preferences and the potential direct and indirect effects of its introduction into aquatic communities (e.g., enhancement of phytoplankton biomass following large-bodied zooplankton suppression: see e.g., Hurlbert and Mulla 1981) renders this biological control program as an environmentally risky and scientifically unsupported strategy. Thus, the objective of our study is to quantify the feeding preferences of C. decemmaculatus in order to assess its efficacy as a biological control of larval Cx. pipiens.


The feeding habits of C. decemmaculatus were determined in laboratory experiments. C. decemmaculatus specimens were caught from the wild and acclimatized to laboratory conditions over a three-week period in a 60 liter aquarium with natural lighting and at 19–21º C. Meanwhile, fish were fed fish food flakes. To assess differences in prey preference according to fish size, individuals were categorized in three different body-size classes (total length) as in Quintans et al. (2009): small (< 25 mm), medium (25–29 mm), and large (≥30 mm). Within-class differences in body size were controlled, resulting in mean sizes (± SD) for each class: 12.70 ± 1.73 mm (small fish class), 26.1 ± 0.54 mm (medium fish class), and 31.60 ± 1.57 mm (large fish class). Our experimental design followed a factorial-randomized block design, where size classes and food items were main factors in triplicate aquariums.

Fish of each size class (n = 30) were exposed to five food items: cladocerans (Daphnia pulex: mean size ± SD = 1.52 ± 0.21 mm), Calanoid copepods (Notodiaptomus incompositus: 0.45 ± 0.15 mm), 1st instar larvae (2.62 ± 0.37 mm) and 4th instar larvae of the mosquito Cx. pipiens (6.52 ± 0.35 mm), and Chironomidae larvae (Chironomus sp.: 12.10 ± 1.07 mm). Cladocerans and copepods species are representative of planktonic fauna that occurred in freshwater environments, whereas Chironomidae larvae typify the benthic fauna. All of these constitute food items of freshwater fish (Gerking, 1994), and have been observed in the digestive tract of C. decemmaculatus (Quintans et al., 2009). Cladocerans and copepods were obtained from crops, while instars of mosquito and Chironomidae larvae were caught immediately before the trials. The abundance of each item was inversely proportional to their mean body weight. Thus, some 972 copepods, 324 cladocerans, 180 1st instars and 72 4th instars of mosquito larvae and 36 Chironomidae larvae, following a ratio of, 27, 9, 5 and 2 individuals, respectively, per Chironomidae larvae. These abundances assured that each item was provided ad libitum, so that prey choices were not influenced by the availability of a particular food item. Fish specimens selected for each trial were isolated and starved for 24 h and acclimated in aquariums for 30 min. The experiments began by adding known abundances of the five food items and ended after one hour. Immediately afterward, fish were overexposed to a solution of 2-phenoxyethanol anaesthetic, and their guts were excised and preserved in 5% formalin solution until their examination using an optical microscope at 40x magnification. All individuals of each prey item in the gut content were counted.

Feeding preferences were evaluated according to the Chesson's electivity index (Chesson 1978):


where m is the number of food items, ri the numeric proportion of the item i in the fish gut content, and pi the numeric proportion of the item i in the environment. When ai= m-1 (with i= 1,…, m), selective predation does not occur; when ai > m-1, the item occurs in the diet more frequently than expected by random feeding and indicates preference for that food item; and when ai < m-1, occurrence of the item i is lower than expected, indicating avoidance. This index is independent of the relative abundance of each food item and is therefore appropriate for comparisons between samples with different number of food items (Lazzaro 1987). Selective predation occurs when the relative frequencies of consumed prey items differ from the relative frequencies in the environment.

A two-way ANOVA was performed to compare the Chesson's electivity index, using food items and size classes of C. decemmaculatus as main factors. Data were transformed (Y1/8) in order to fulfil the requirement of homoscedasticity. When significant differences were detected, multiple comparisons were performed with the post-hoc least significant difference (LSD) test (Zar 1996).


The overall analysis of our data (all class sizes pooled) showed that cladocerans were the most consumed food item by Cnesterodon decemmaculatus. In fact, this item was the only one with a Chesson's electivity index of α > 0.2 (Figure 1).

Figure 1.

Mean (±SE) values of the Chesson electivity Index (α) (all size classes pooled). α values > 0.2 denotes preference. Mosquito larvae 1= 1st instar mosquito larvae; Mosquito larvae 4= 4th instar mosquito larvae; Ch. Larvae = Chironomidae larvae.

The electivity index differed significantly among fish size classes and food items, as well as in their interaction (Table 1). Multiple comparisons (LSD test) showed that 1) for small fish, the Chesson's Index was significantly higher for cladocerans when compared with the three larval items (P<0.001), but did not differ with copepods; 2) medium fish significantly preferred cladocerans and Chironomidae larvae compared to other items (P<0.001); and 3) large fish significantly preferred cladocerans over all other items (P<0.001) (Figure 2). Multiple comparisons also showed significant differences in preference values among size classes (LSD test: p<0.001), with the exception of 4th instar mosquito larvae (Table 2). All fish sizes showed preference for cladocerans (α>0.2), but α values were significantly higher for large fish than for small or medium fish (Figure 2). Preference values for copepods differed significantly between size classes and were inversely proportional to fish size, small fish being the only size class with preference values >0.2. Medium fish was the only size class that showed preference for the 1st instar mosquito larvae, but all the three fish sizes did not differ significantly in their preferences for the 4th instar mosquito larvae and none of them showed preference values. Only medium fish showed a preference value >0.2 for chironomid larvae, which was significantly higher than those estimated for small and large fish size classes. The significant size class by prey item interaction was clearly reflected in the dissimilar prey preference between size classes. Whereas preference for cladocerans increased with fish size, copepods and 1st instar mosquito larvae showed the reverse pattern. Moreover, medium fish only consumed 4th instar mosquito larvae at negligible levels (Figure 2).

Table 1. Cnesterodon decemmaculatus. Two-way ANOVA results on Chesson's electivity index, using fish size classes (small, medium, and large) and food items (cladocerans, copepods, 1st instar mosquito larvae, 4th instar mosquito larvae, and Chironomidae larvae) as main factors. *** P < 0.001.
Source of variationdfMSF
 Size classes (A)22.94937.60***
 Food items (B)47.73898.67***
   A × B81.72822.01***
Figure 2.

Mean (± SE) values of the Chesson's electivity Index (α) discriminated by size class. α values > 0.2 denotes preference. M. larvae 1= 1st instar mosquito larvae; M. larvae 4= 4th instar mosquito larvae; Ch. Larvae = Chironomidae larvae. Zero values are noted as 0.

Table 2. Cnesterodon decemmaculatus. Multiple comparisons (post-hoc LSD test) on preferences for each food item between different size classes (MS = 0.7842, df = 435). Significant p values (p<0.05) are highlighted in italics. Small size fish class= s; medium size class= m; large size class= l. Mosquito larvae 1= 1st instar mosquito larvae; Mosquito larvae 4= 4th instar mosquito larvae; Ch. Larvae = Chironomidae larvae.
ItemsFish sizeP
Cladoceranss – m0.2024
m – l0.0040
s – l0.0000
Copepodss – m0.0002
m – l0.0000
s – l0.0000
M. larvae 1.s – m0.3734
m – l0.0342
s – l0.2183
M. larvae 4.s – m0.1364
m – l0.9776
s – l1.0000
Ch. larvaes – m0.0000
m – l0.0000
s – l0.4295

Individual analyses showed that 21 small fish, 29 medium fish, and all large-size fish preferred cladocerans (α>0.2). In addition, 16 small fish, 10 medium fish, and 25 large fish only consumed this item (Figure 3). Copepods were only preferred by small fish; all individuals consumed this item and nine of them did it exclusively (Figure 3). First instar mosquito larvae were consumed by all fish sizes, but more often by medium ones, which preferred them (Figure 2). Nevertheless, all small individuals that consumed 1st instars did it exclusively, whereas large fish consumption for this item was less important (Figure 3). Only three individual medium fish consumed the 4th instar larvae, and α values for the entire size class were <0.2 (Figure 2). Chironomidae larvae were consumed almost exclusively by medium fish (Figure 2), with 22 individuals showing a marked preference for this item (Figure 3).

Figure 3.

Numerical frequencies of Chesson's electivity Index for each item at each size class. M. larvae 1= 1st instar mosquito larvae; M. larvae 4= 4th instar mosquito larvae; Ch. Larvae = Chironomidae larvae.


Our results demonstrated that C. decemmaculatus markedly preferred cladocerans to other food items. Thus, this fish can be categorized as a zooplankton facultative-feeder fish with preference for large-bodied zooplankton. The body size, mouth morphology, and teeth structure of C. decemmaculatus are typical of an omnivorous fish (Loureiro, unpublished data). These fish characteristics, together with the structural, morphological, and behavioral characteristics of their prey, could explain the greater preference of the three size classes for Cladocera and the lowest preference for larvae. While mosquito larvae have defense structures and display sharp movements, cladocerans have virtually no evasive movements and thus were more easily captured and swallowed (Brooks and Dodson 1965).

According to optimal foraging theory (MacArthur and Pianka 1966), predators will prefer smaller, more profitable prey that are directly swallowed without an energy-costly handling time, in a high prey density environment (Werner and Hall 1974). Prey size is thus a determinant factor in food choice (Guma'a 1978) and fish growth is often concomitant with an increase in prey size (Mills et al., 1985; Lazzaro, 1987). Cnesterodon decemmaculatus was capable of consuming all available prey, but cladocerans were the most preferred item by all fish size classes. The higher preference for cladocerans by large fish could indicate that this prey item has an optimal size for this size class, whereas the lower preferences shown by small fish for this item might be due to gape-limitation. Moreover, the preference of small fish for copepods could be explained by their smaller mouth size.

From an ecological point of view, our experiment suggests that the introduction of Cnesterodon decemmaculatus in aquatic ecosystems of Uruguay is questionable. Marti et al. (2006) showed that C. decemmaculatus could be successfully introduced into artificial environments (e.g., water tanks for agricultural or industrial uses, ditches) with impoverished invertebrate and fish fauna, thus stimulating predation of early mosquito larval instars. These authors also showed that C. decemmaculatus preys on Cx. pipiens larvae when they are the only available prey (Marti et al. 2006). However, our experiments demonstrated that predation on Cx. pipiens larvae decreases significantly if other food sources are simultaneously available. Therefore, the efficacy of C. decemmaculatus as biological control in natural environments is compromised. Besides, by reducing cladocerans populations (i.e., Daphnia spp.) and indirectly releasing phytoplankton from grazing pressure, C. decemmaculatus would enhance phytoplankton biomass, thus reducing water transparency (Hrbáček et al. 1961, Brooks and Dodson 1965, Carpenter et al. 1987, Kitchell and Carpenter 1993). This cascade effect has been demonstrated for Gambusia spp. (Hurlbert et al. 1972, Hurlbert and Mulla, 1981; Margaritora et al., 2001) and for C. decemmaculatus in outdoor mesocosms (Lacerot and Kruk, unpublished data). Thus, the introduction of C. decemmaculatus in natural environments should not be considered to be an effective strategy towards a biological control of mosquito larvae.


This work is part of the M.Sc. thesis of Federico Quintans, who was partially supported by the Environmental Sciences Master Program at the Faculty of Sciences (Uruguay). The authors thank the staff of the Limnology Section at the Faculty of Sciences, Universidad de la República. We also acknowledge Dr. Matías Arim, Dr. Marcelo Loureiro and María Martinez for their help with the experimental design, laboratory trials and taxonomic larvae identification. Special thanks to Dr. Gissell Lacerot and Dr. Xavier Lazzaro for their help improving the manuscript.