Abstract
The efficacy of using predators for the biological control of mosquito disease vectors will be reduced if mosquito larvae respond to predator presence. The larvae of two mosquito species were investigated to study whether they responded to predator kairomones by increasing surface filter-feeding, which is a less active and thus less risky feeding strategy than bottom feeding. Culex quinquefasciatus Say is normally found in highly polluted water, where it will have little contact with predators. Except for some third instars, its larvae showed no response to four different types of predators. Culiseta longiareolata Macquart, living in rain-filled rock pools, is frequently attacked by a range of predators. All instars tested (second, third, and fourth instars) strongly responded to chemicals from dragonfly nymphs (Crocothemis erythraea Brullé), damselfly nymphs (Ischnura evansi Morton), and the fish Aphanius dispar Ruppel. However, they did not respond to final-instar water scorpions (Nepa cinerea L.), which would not feed on the mosquito larvae. Second- and third-instar Cs. longiareolata produced the same response to chopped up mosquito larvae as they did to dragonfly nymphs, but fourth instars produced a significantly stronger response to dragonfly nymphs—both those unfed and those fed in situ. Thus, Cs. longiareolata not only identified different predators and responded accordingly, but also responded to conspecific alarm pheromones. Cx quinquefasciatus showed little response to predators or to alarm pheromones from damaged conspecific larvae.
Biological control of mosquito disease vectors using native predators, especially fish, has been extensively used in large pools of water, such as rice fields (Bence 1988) and mangroves (Griffin and Knight 2012). However, in small pools of water, it has been shown that many mosquitoes can detect the presence of a predator and will lay their eggs elsewhere; thus, there is little colonization in pools containing predators. This will affect the efficiency of predator control. Ovipositing female mosquitoes can not only detect fish (Angelon and Petranka 2002), but also insect predators such as dragonfly nymphs (Stav et al. 2000) and notonectids (Eitam et al. 2002, Kiflawi et al. 2003).
After hatching, the mosquito larvae may later find themselves exposed to predators that have colonized the pool. If they are able to detect the predator, they can show phenotypic plasticity in adapting their behavior or physiology to reduce their risk of predation. One strategy is to reduce activity (so they are less visible to a predator) and thus slow development, but this comes at the cost of trait compensation; that is, the adults will be smaller and less competitive. This strategy has been demonstrated in response to dragonfly nymphs (Roberts 2012), notonectids (Knight et al. 2004, Beketov and Liess 2007), and fish (Bond et al. 2005). In Odonata, other antipredator strategies that have been demonstrated include avoiding the location of predators (Pierce 1988) and increasing nocturnal feeding (Koperski 1997). Other predator-mediated indirect effects include morphological changes in Daphnia (Barry 1994) and in tadpoles (Relyea 2001). Most mosquito larvae have two alternative feeding mechanisms, namely, filter-feeding on microorganisms at the water surface and scraping biofilms from underwater vegetation and rocks. The relative importance of each method depends on the food available, but also varies between species. Thus, some species consistently have a much higher proportion of bottom scraping than other species, and because this is much more active than filter-feeding, it puts them a higher risk of being noticed by a predator (Roberts 2012).
In this study, two mosquito species were compared: Culex quinquefasciatus Say is a specialized species largely restricted to urban areas in Oman (Roberts and Irving-Bell 1997) because it mainly breeds in septic tanks (Menon and Rajagopalan 1980), where it is attracted by high levels of ammonia (Sinha 1976), and Culiseta longiareolata Macquart is found in temporary rain-filled pools (Spencer et al. 2002). Therefore, Cx. quinquefasciatus has little exposure to predators because it lives in very anaerobic water, whereas Cs. longiareolata is faced with a high level of predation by a range of different species.
Detection of predators by mosquito larvae may mainly be because of water vibrations, but can also be chemical (Sih 1986). To remove the possibility of water vibrations, in the present experiments, the predator was isolated from the mosquito larvae. Water from the predator container was pumped through the mosquito larvae container, so that chemicals were the only factor involved. Mosquito larvae also respond to chemicals from attacked mosquito larvae (Ferrari et al. 2007); thus, a further set of experiments investigated the relative importance of chemicals from crushed mosquitoes and the actual predator.
Thus, these experiments compare two mosquito species, one with little natural exposure to predators and another that has to cope with high levels of predation, to determine whether they alter their feeding behavior in the presence of chemicals from four different types of predators, namely, dragonfly nymphs, damselfly nymphs, water scorpion nymphs (Nepa), and the predacious fish Aphanius, and also to compare the response to crushed mosquito larvae with these predator chemicals.
Materials and Methods
Source of Mosquitoes and Predators.
Cx. quinquefasciatus egg rafts were collected from Sultan Qaboos University Botanic Garden (Oman) using water-filled plastic bowls 70 cm diameter and 20 cm deep (77 liters). These had been seeded with both rabbit food pellets and the seed pods of Prosopis cineraria Druce and left for 3 wk to develop microbial flora. Egg rafts were collected daily and each raft was kept separately in the laboratory at 24°C until they hatched.
Cs. longiareolata egg rafts were collected weekly from rain-filled rock pools in Wadi Qurai at Sumail in the Jebel Akhdar Mountains, 60 km from the university campus. Between rains, the rock pools were artificially filled each week from the perennial stream that flowed a few meters away down the wadi. Egg rafts were also kept separately in the laboratory.
Dragonfly nymphs (Crocothemis erythraea Brullé) and damselfly nymphs (Ischnura evansi Morton) were collected from very small (<2 m) fish-free pools in Wadi Al-Khod, ≈5 km from the University. The Aphanius fish were also collected from Wadi Al-Khod. Nepids (possibly Nepa cinerea L.) were collected from Wadi Qurai.
During the experiments, only final instars of the dragonfly nymphs, damselfly nymphs, and nepid nymphs were used. Aphanius dispar Ruppel used were all mature females.
Between experiments, the dragonfly nymphs, damselfly nymphs and Aphanius were fed on the final (fourth) instars of the appropriate mosquito species currently being investigated. However, the nepid nymphs ignored the mosquito larvae and were fed on damselfly nymphs (one every 2 d).
Response of Mosquito Larvae to Predator Chemicals.
The predator being tested was kept in 4 liters of conditioned tap water in a 16-cm-diameter plastic container (i.e., predator bottle). Aphanius and nepids were allowed to move about freely in the bottle (although the nepids were very inactive and held on to the tubing pumping out the water). The dragonfly nymphs and the damselfly nymphs were individually put in translucent polythene tubes (2.5 cm diameter and 4 cm length) that had both ends covered in netting to allow water to pass through. The tubes were each weighted with a stone, so they remained on the bottom. Water from the predator bottle was pumped using Welco peristaltic pumps (WPX1 from Welco Ltd, Tokyo, Japan) that had been calibrated to a flow rate of 70 ml/h into the mosquito jar. This was a polystyrene, 8-cm-diameter jar containing 400 ml of water that overflowed in to a bucket. The overflow spout was covered with insect netting to prevent escape of the mosquito larvae. Four mosquito jars, each with a separate pump, were connected to each bottle. There were three bottles and thus 12 mosquito containers. The first bottle was a control (only water and no predator); while the second and third bottles contained the Predator 1 (e.g., Aphanius) and Predator 2 (e.g., nepid), respectively. Although the relatively large Aphanius and nepids had a single individual in each bottle, the much smaller dragonflies and damselflies consisted of three individuals, each in their own tube, in the bottle. All predators were changed daily from a large stock of individually reared specimens. Nonfeeding specimens (about to molt) were not used.
The equipment was set up in the afternoon beforehand at a temperature of 23°C. The predators were put in their 4-liter water bottles to give time for predator kairomones to accumulate. Thirty mosquito larvae of the same instar were put into each jar (0.075 larvae per ml of water), which is well below the levels of crowding that might affect the mosquito behavior (Roberts and Kokkinn 2010). Each egg raft hatches around 150 larvae, and because it is known that there is great variation in behavioral and physiological response between the cohort of larvae from one egg raft and the cohort from another (Kokkinn et al. 2012), each egg raft cohort was used for one replicate (control + Predator 1 + Predator 2 = total of 90 larvae). The larvae were then fed with yeast powder at a dosage of 0.09 mg/larva. The pumps were started at 0900 hours the next morning, but no readings were taken to allow time for the predator chemicals to disperse through the mosquito larval jar and to allow the larvae to acclimate to the disturbance of the water dripping in and draining out of their container. The water in the predator bottles was topped up around 1200 hours and the readings started at 1400 hours. Fifteen readings were taken at 5-min intervals for each of the 12 mosquito containers, by counting the number of mosquito larvae that were filter feeding at the surface and those that were scraping biofilms on the sides and bottom. These were repeated on subsequent days with different sets of mosquito larvae, so that there were 12 replicates (each with 15 readings) for each mosquito instar for each predator and the controls.
Mosquito larvae that were second, third, and fourth instars were tested for a response against each of the four predators. First instars were not tested because of the short duration of their stage and the difficulty of reliably seeing them in the jars.
Comparison of Chemicals from Feeding.
Using the same setup as described in Response of Mosquito Larvae to Predator Chemicals section, a comparison was made between alarm kairomones from killed mosquito larvae, chemicals from the predator only, and chemicals from the predator + alarm kairomones from eaten mosquito larvae. The following were thus compared: 1) eight live, fourth-instar mosquito larvae were chopped into small pieces using dissecting needles and washed into the 4-liter bottle. No predator was present. The chopped larvae were added when pumping started at 0900 hours and repeated when the water was topped up at 1200 hours. 2) Unfed dragonfly nymphs were used, as in the first experiment. 3) Each in situ dragonfly nymph was fed with four fourth-instar mosquito larvae put inside the dragonfly tube cage at 0900 hours, and then a further two mosquito larvae were added at 1200 hours. As in the first experiment, there were 12 replicates, each with 15 readings of the proportion of mosquito larvae that were no surface filter-feeding.
The data were arcsine transformed and then analyzed using a nested univariate analysis of variance (ANOVA) with SPSS software (SPSS for Windows 10.0.0, 1999; SPSS, Chicago, IL). Significant effects were separated by Tukey's honestly significant difference (HSD) test. The graphs used back-transformations of the data used in the ANOVAs to calculate the mean number of bottom-feeding larvae.
Results
Response of Mosquito Larvae to Predator Chemicals.
Cs. longiareolata showed highly significant responses in all three instars to dragonfly nymphs, damselfly nymphs, and the fish Aphanius. Thus, when exposed to chemicals from either dragonfly or damselfly nymphs, all three instars had a highly significant reduction in bottom feeding (Fig. 1), but with no difference between the effect of the dragonflies and the damselflies. The second instars controls had a bottom-feeding mean ± SE of 13.4 ± 1.2 larvae, while that for those exposed to damselflies was 8.6 ± 1.2 and to dragonflies was 8.3 ± 1.4 (F = 6.47; df = 2, 25; P < 0.005); the third-instar control mean was 6.7 ± 0.8, and exposed to damselflies was 5.0 ± 0.5 and to dragonflies was 4.1 ± 0.7 (F = 16.05; df = 2, 25; P < 0.0001); the fourth-instar control mean was 9.6 ± 0.4, and exposed to damselflies was 7.2 ± 0.5 and to dragonflies 6.9 ± 0.4 (F = 17.55; df = 2, 25; P < 0.0001). When exposed to chemicals from the fish Aphanius, all three instars also had a highly significant reduction in bottom feeding (Fig. 2). The second-instar control mean was 5.8 ± 1.0 and exposed to fish was 3.1 ± 0.5 (F = 21.67; df = 1, 17; P < 0.0001); the third-instar control mean was 8.4 ± 0.1 and exposed to fish was 4.5 ± 0.8 (F = 19.37; df = 1, 17: P < 0.0001); the fourth-instar control mean was 9.1 ± 0.7 and exposed to fish was 6.0 ± 0.7 (F = 16.77; df = 1, 17; P < 0.001).
Mean (±SE) bottom-feeding larvae (as a proportion of 30 produced by back-transforming the ANOVAs) of Cs. longiareolata for second, third, and fourth instars when they are controls (exposed to clean water) or exposed to chemicals from damselfly larvae, or exposed to chemicals from dragonfly larvae. Means with the same letter are not significantly different (Tukey's HSD test, P > 0.05).
Mean (±SE) bottom-feeding larvae (as a proportion of 30 produced by back-transforming the ANOVAs) of Cs. longiareolata for second, third, and fourth instars when they are controls (exposed to clean water) or exposed to chemicals from the fish Aphanius. Means with the same letter are not significantly different (Tukey's HSD test, P > 0.05).
However, the mosquito larvae did not respond to chemicals from Nepa (Fig. 3). The second-instar control mean was 0.7 ± 0.1 and exposed to Nepa was 0.4 ± 0.2 (F = 3.77; df = 1, 17; P = 0.08); the third-instar control mean was 5.2 ± 0.5 and exposed to Nepa was 6.0 ± 0.6 (F = 1.28; df = 1, 17; P = 0.28); the fourth-instar control mean was 9.5 ± 0.8 and exposed to Nepa was 7.9 ± 1.3 (F = 1.04; df = 1, 17; P = 0.33).
Mean (±SE) bottom-feeding larvae (as a proportion of 30 produced by back-transforming the ANOVAs) of Cs. longiareolata for second, third, and fourth instars when they are controls (exposed to clean water) or exposed to chemicals from the water scorpion Nepa. Means with the same letter are not significantly different (Tukey's HSD test, P > 0.05).
Cx. quinquefasciatus showed no response to any of the predators, except among the third instars. Thus, when exposed to chemicals from damselfly nymphs and the fish Aphanius, both second and fourth instars showed no significant response (Fig. 4), with the second-instar control mean being 0.5 ± 0.2, while exposed to damselflies was 0.1 ± 0.03 and to fish was 0.3 ± 0.3 (F = 0.82; df = 2, 25; P = 0.45). The fourth-instar control mean was 3.7 ± 0.2, while exposed to damselflies was 3.2 ± 0.5 and to fish was 4.0 ± 0.4 (F = 1.46; df = 2, 25; P = 0.25). However, the third instars had a significant reduction in bottom feeding in response to both predators, with the control mean of 2.0 ± 0.5, and exposed to damselflies was 0.2 ± 0.1 and to fish was 0.5 ± 0.2 (F = 10.71; df = 2, 25; P < 0.0001). Similarly, when exposed to chemicals from dragonfly nymphs and Nepa, both second instars and fourth instars showed no significant response (Fig. 5). The second-instar control mean was 3.4 ± 0.5, while exposed to dragonflies was 1.8 ± 0.6 and to Nepa was 2.1 ± 0.8 (F = 2.53; df = 2, 25; P = 0.1). The fourth-instar control mean was 3.7 ± 0.3, while exposed to dragonfly were 3.6 ± 0.4 and to Nepa was 2.8 ± 0.3 (F = 1.45; df = 2, 25; P = 0.25). However, the third instars had a significant reduction in bottom feeding in response to both predators, with a control mean of 5.4 ± 3.1; exposed to dragonflies being 3.1 ± 0.4 and to Nepa being 0.5 ± 0.2 (F = 15.94; df = 2, 25; P < 0.0001).
Mean (±SE) bottom-feeding larvae (as a proportion of 30 produced by back-transforming the ANOVAs) of Cx. quinquefasciatus for second, third, and fourth instars when they are controls (exposed to clean water), exposed to chemicals from damselfly larvae, or exposed to chemicals from the fish Aphanius. Means with the same letter are not significantly different (Tukey's HSD test, P > 0.05).
Mean (±SE) bottom-feeding larvae (as a proportion of 30 produced by back-transforming the ANOVAs) of Cx. quinquefasciatus for second, third, and fourth instars when they are controls (exposed to clean water), exposed to chemicals from dragonfly larvae, or exposed to chemicals from the water scorpion Nepa. Means with the same letter are not significantly different (Tukey's HSD test, P > 0.05).
Comparison of Chemicals from Feeding.
For Cs. longiareolata, second and third instars responded equally to chopped mosquito larvae, unfed dragonfly nymphs, and fed dragonfly nymphs (Fig. 6). The second instars exposed to chopped mosquitoes had 5.0 ± 1.1 larvae bottom-feeding, while response to unfed dragonflies was 5.6 ± 1.0 and to fed dragonflies was 6.3 ± 1.2 (F = 0.65; df = 2, 25; P = 0.53). The third instars exposed to chopped mosquitoes had 8.0 ± 1.1 larvae bottom-feeding, while response to unfed dragonflies was 6.9 ± 1.0 and to fed dragonflies was 8.1 ± 0.6 (F = 2.11; df = 2, 25; P = 0.14). The fourth instars responded more strongly to both fed and unfed dragonflies than they did to the chopped mosquitoes. Thus, larvae exposed to chopped mosquitoes maintained bottom feeding at 7.8 ± 1.0 larvae, while those exposed to unfed dragonflies reduced bottom feeding to 4.7 ± 0.3, while those exposed to fed dragonflies was 5.5 ± 0.7 larvae (F = 7.31; df = 2, 25; P = 0.003).
Mean (±SE) bottom-feeding larvae (as a proportion of 30 produced by back-transforming the ANOVAs) of Cs. longiareolata for second, third, and fourth instars when they are exposed to chopped up fourth-instar mosquito larvae, exposed to chemicals from unfed dragonfly larvae, or exposed to chemicals from dragonfly larvae that are fed in situ on fourth-instar mosquito larvae. Means with the same letter are not significantly different (Tukey's HSD test, P > 0.05).
For Cx. quinquefasciatus, all instars responded equally to chopped mosquito larvae, unfed dragonfly nymphs, and fed dragonfly nymphs (Fig. 7). The second instars exposed to chopped mosquitoes had 0.5 ± 0.2 larvae bottom-feeding, while response to unfed dragonflies was 0.7 ± 0.2 and to fed dragonflies was 0.8 ± 0.4 (F = 0.32; df = 2, 25; P = 0.73). The third instars exposed to chopped mosquitoes had 1.0 ± 0.3 larvae bottom-feeding, while response to unfed dragonflies was 1.3 ± 0.4 and to fed dragonflies was 0.6 ± 0.2 (F = 1.17; df = 2, 25; P = 0.33). The fourth instars exposed to chopped mosquitoes had 2.8 ± 0.4 larvae bottom-feeding, while response to unfed dragonflies was 3.3 ± 0.5 and to fed dragonflies was 3.1 ± 0.4 (F = 0.16; df = 2, 25; P = 0.86).
Mean (±SE) bottom-feeding larvae (as a proportion of 30 produced by back-transforming the ANOVAs) of Cx. quinquefasciatus for second, third and fourth instars when they are exposed to chopped up fourth-instar mosquito larvae, exposed to chemicals from unfed dragonfly larvae, or exposed to chemicals from dragonfly larvae that are fed in situ on fourth-instar mosquito larvae. Means with the same letter are not significantly different (Tukey's HSD test, P > 0.05).
Discussion
A successful prey species should be able to detect and avoid its predators, but not all potential predators are equally dangerous, and thus the prey should respond differently (showing phenotypic plasticity) to different predators depending on the risk that they pose to the prey. First, the prey needs to chemically recognize the predator. Sih (1986) showed that Aedes aegypti L., which is not normally exposed to Notonecta predators, showed a weak predator response in comparison with Cx. pipiens L. (a normal prey of Notonecta), whereas Kesavaraju and Juliano (2004) showed that native Ochlerotatus triseriatus Say strongly responded to the predator Toxorhynchites rutilus Theobald, but introduced Aedes albopictus Skuse did not identify and respond to the predator. Similarly, in these experiments, Cx. quinquefasciatus showed little response (except among the third instars) to any of the predators to which it was exposed, but it normally lives in largely predator-free very anaerobic organically polluted water. In contrast, Cs. longiareolata is in nature exposed to a range of predators and showed strong responses.
Second, the prey needs to be able to chemically separate different predator species, so that their response will depend on the risk that the predator poses to the prey. This has been shown in damselfly nymphs(Chivers et al. 1996),dragonfly nymphs (Stoks et al. 2003), and in dragonfly nymphs responding to different predacious fish (Hopper 2001). This differential response to chemicals from different predators has also been shown in mosquito larvae (Roberts 2012). In the present experiment, second to fourth instars of Cs. longiareolata strongly responded to chemicals from dragonfly and damselfly nymphs and to fish, but did not respond to final-instar nepids. Between the two Odonata, the mosquito larvae consistently had a greater response to dragonfly nymphs. This means that the Cs. longiareolata must be able to chemically identify the type of predator and their response depends upon their perceived risk. Several studies have shown that the prey alters its response to a predator that has been fed on conspecifics of the prey (so if the predator is fed on some other food, the response of the prey to the predator is much less, as the perceived risk of the predator is reduced). Examples are tadpoles responding to newts (Wilson and Lefcort 1993) and damselflies responding to fish (Chivers et al. 1996). Thus, the low response by Cs. longiareolata to nepids may have been influenced by the nepid diet, which would not eat mosquito larvae and hence were fed on damselfly nymphs. The prey need to show phenotypic plasticity in its response because the perceived risk may change. Ferrari et al. (2007, 2008) showed that prey response changes as the chemical concentration increases and that the mosquito larva can learn to respond to a novel (abnormal) prey.
A mosquito larva can respond to a predator in a number of ways. The most commonly studied is to reduce activity because an inactive prey is more difficult for a predator to see and feel through water vibrations. This has been shown in mosquito larvae as a response to dragonfly nymphs (Roberts 2012), notonectids (Knight et al. 2004, Beketov and Liess 2007), and fish (Bond et al. 2005). However, inactivity reduces feeding efficiency and thus has a cost in trait compensation, such as extending development time, thereby extending exposure to the predator. The larvae of most mosquito species show a mixture of two feeding strategies, namely, surface filter feeding and bottom feeding by scraping biofilms and an individual larva rapidly changes from one strategy to the other. Cs. longiareolata spends a greater proportion of its time bottom feeding compared with other species (Roberts 2012), but this particularly puts it at risk from dragonfly nymphs, which mainly sit on the bottom of a pool, unlike damselfly nymphs, which prefer to attach to vegetation (Corbet 1980). This would explain the greater response to dragonfly nymphs by Cs. longiareolata larvae. Bottom feeding also involves more activity on the part of the larvae, and thus attracts the attention of predators. Juliano and Gravel (2002) found that Aedes triseriatus reduced bottom feeding in the presence of the predatory mosquito Toxorhynchites rutilus.
Mosquito larvae may detect a predator through the water vibrations that it produces during feeding (Sih 1986), but in the current study, predator detection was experimentally limited to chemical kairomones coming from the predator. However, under natural conditions, the larvae could also be responding to alarm pheromones coming from injured conspecifics. Alarm pheromones have been shown to have a significant effect on mosquito larval behavior (Ferrari et al. 2008). In the second part of the current study, alarm pheromones where found to have a great effect on second and third-instar Cs. longiareolata, but had a lesser effect on fourth instars than dragonfly kairomones. However, Cx. quinquefasciatus was unaffected by both alarm pheromones and dragonfly kairomones.
Cx. quinquefasciatus, living in a specialized microhabitat where predators are probably of little importance, showed little response to the presence of predator kairomones. Cs. longiareolata larvae, inhabiting a microhabitat where a variety of predators are frequently present, showed highly significant responses to damselfly and dragonfly nymphs and to the predatory fish Aphanius, but did not respond to final-instar predatory nepids.
Acknowledgments
I thank Sultan Qaboos University for providing research facilities, Michael Barry for assistance in field collection of predators, and Prof. Reg Victor for reading the manuscript.
References Cited
