The Role of Experience in the Visual and Non-Visual Prey Recognition of Fire Salamander Populations from Caves and Streams

Abstract

The study of foraging behaviour is crucial for understanding several ecological and adaptive processes, as well as for developing conservation measures. While extensive research has been completed on birds and mammals, few studies have been conducted on the learning capabilities of amphibians, particularly those pertaining to foraging behaviour. Amphibians may detect potential prey through distinct sensory systems including visual detection, chemoreception, and mechanoreception. In this study, we tested whether fire salamander larvae shift their prey recognition depending on the prey stimulus typology. We performed behavioural assays to better understand the roles of visual and chemical cues in prey recognition and how a continuative visual stimulus may change behavioural patterns. For this assessment, larvae from different habitats (cave and stream) were reared under laboratory conditions and fed while exposed to accompanying sensory stimuli. Their responses to visual and olfactory cues were measured before and after rearing. Both visual and chemical cues significantly affected the time of approach to the stimulus. The period of rearing significantly interacted with the time of approach for both cues. After rearing, when visual cues occurred, the time of approach was much lower than before rearing. These findings provide a basis for further studies on the role of plasticity in the predator–prey interactions of fire salamander larvae.

1. Introduction

The study of foraging behaviour may serve as a useful stepping stone in the comprehension of more general processes, such as how species and populations move in their environments, distribute themselves, or colonise a particular habitat [1]. Furthermore, as foraging for food is the initiation of an interspecific interaction, studying this activity may provide fundamental information for ecologists and conservationists alike. The recognition of food items and the display of a proper response to obtain such items are basic components of animal survival. Studying these features may illuminate the general processes and variations underlying an animal’s cognitive abilities [2]. This is especially advantageous for historically understudied taxa, which lack established baselines for laboratory cognition experiments [3]. Despite the few pre-existing studies on amphibians’ learning capabilities, these animals are increasingly being recognised as vertebrates with sophisticated communicative and problem-solving skills [4]. Most amphibians, and especially salamanders, are efficient predators both in terrestrial and aquatic environments [5,6,7]. One of the most important components of predator foraging behaviour is prey detection efficiency [8]. Understanding the factors that enable predators to successfully detect their prey may reveal fundamental insights into their natural history and ecology. Amphibians, including salamander larvae, can employ a relatively diverse set of sensory systems to recognise and detect potential prey. These systems involve chemoreception, vision, mechanoreception, and even electroreception [9,10].

Typically, amphibians inhabiting murky, silted habitats or underground waters—in which the use of visual stimuli is generally limited—have a tendency to depend on chemoreception to detect potential prey [11,12,13]. On the contrary, species that live in clear, open freshwater habitats where visualization is not limited are inclined to rely primarily on visual cues [14].

In amphibians, prey-catching behaviour generally starts with an orienting response towards a potential prey item; successively, there is an approaching phase that brings the individual more or less close to the prey [15]. Finally, the predator performs a snap to complete the catch and collect the prey in its mouth [16]. While most species need a careful approach when moving closer to prey, others have developed snapping techniques that may be performed across relatively far distances, as evidenced by the extendible tongues of Plethodonthids [17].

Different studies have suggested that this behavioural sequence is automatically displayed toward objects fitting a prey structure, so that even relatively simple objects that move and are neither too large nor too small will be snapped at [15]. In the past, researchers have viewed amphibians’ behaviour as primarily driven by instinct (e.g., [18,19]). However, different research studies have shown that perhaps this inference was presumptuous, and that there is clear evidence that both chemical and visual prey detection are dependent on experience and learning [4,15]. For example, among Anura, tadpoles learn to display effective responses towards predators’ chemical cues following different experiences during the early stages of larval development [20]. Moreover, adult frogs, as in the case of Oriental fire-bellied toads (Bombina orientalis), are able to learn to discriminate between prey-rich and -poor sites [21]. Studies performed on salamanders have demonstrated that both chemical and visual cues are important for prey location and may be connected to experience [4]. Plethodon cinereus adults are able to discriminate chemically rich nutritional prey items from poorly nutritional ones [22], so that males actively guard sites containing high-quality food [22]. Recent research has shown that tiger salamanders (Ambystoma tigrinum) can learn to respond to different visual cues [23]. Such studies suggest that amphibian responses are not limited to instinctual components, but rather that experience and learning modulate the relative importance of the different sensory systems related to prey detection.

For further insight into amphibian cognition regarding food capture, the fire salamander (Salamandra salamandra) provides a remarkable case study. This widespread European species typically breeds in streams and small pools [24]; however, some populations are able to breed in subterranean environments where their larvae successfully undergo development [25]. On the one hand, subterranean environments can provide advantages, such as more stable environmental conditions and limited predation risk [26,27]. On the other hand, freshwater subterranean habitats have much lower productivity than surface streams, and their ecology strongly depends on the input of external trophic resources [28]. Salamander larvae laid in caves, despite being predators, may themselves constituting a major trophic subsidy in these poorly productive environments [29]. Because the lack of light in these habitats indirectly limits the populations of primary consumers (e.g., insect larvae and crustaceans), food scarcity is among the major challenges faced by subterranean fire salamander larvae [30]. Consequently, successful prey detection is likely an essential prerequisite for the growth of fire salamander larvae. The olm Proteus anguinus—the only troglobitic European vertebrate species—can distinguish the chemical cues of food and non-food items, as well as those of dead and live prey, while living in total darkness. It does so by utilizing mechanically and chemically guided approaches [31,32,33]. The Pyrenean brook newt, Calotriton asper, which has cave-dwelling populations, demonstrates visually dominated behaviour when in light conditions, but can detect prey using chemoreception when in total darkness [34]. For fire salamander larvae, recent studies have shown that individuals from both caves and streams display plastic behaviour: under light conditions, they adopt a sit-and-wait predation strategy, whereas in complete darkness, they perform an active search for prey during which they continuously explore the environment, often travelling long distances [35]. However, this plastic response is not constant among populations. For instance, cave-born larvae show higher behavioural plasticity than stream-born larvae, and are superior in exploiting the available space within test environments [34]. The role of early experience, as pertaining to the development of visual prey recognition by fire salamander larvae, has been identified in an experiment by Luthardt-Laimer [36]. Larvae were found to preferentially snap at slow, horizontally moving objects. The author suggests that this preference was experience-dependent, indicating that the larvae learned to recognise stimuli similar to the prey with which they were fed during the rearing period [36]. However, a more recent study performed on eight larvae found that kinematic and foraging behaviour during feeding remained unchanged throughout the larval ontogeny [37]. That study suggests that foraging behaviour in S. salamandra is developmentally fixed (innate) and uninfluenced by learning or experience, thus supporting the hypothesis that aquatic salamander feeding is highly stereotyped [37].

The aims of our paper are to (1) disentangle the role that experience may play in fire salamander foraging behaviour, and to (2) determine whether experience drives a shift between non-visual and visual sensory systems in controlling foraging behaviour.

2. Materials and Methods

2.1. Larvae Collection and Rearing

We collected 52 fire salamander larvae at developmental stage 1 (newborns: well-developed tail fin and the tip of the fin bluntly rounded [38]) from seven caves and six streams situated in the Italian Prealps of Lombardy (NW Italy; around 45°48′ N, 9°02′ E). Larvae were individually contained in perforated plastic boxes (diameter of the boxes = 8 cm) for 55 days at a mean temperature of 18 °C, exposed at natural photoperiods. Each plastic box was randomly assigned to one of four independent water-filled blocks (40 × 50 cm, water depth: 7 cm). During the rearing period (between the pre- and post-tests), all larvae were assigned the same feeding treatment: larvae were fed ad libitum with defrosted Chironomus sp. larvae using tweezers that, while approaching the box before water touching, were continuously shaken. This feeding modality established a case of classical conditioning, consistently introducing food in a visually and olfactorily stimulating manner. The collection and rearing of fire salamander larvae was allowed by the Region Lombardy authority, permit number: T1.2016.0052349.

2.2. Behavioural Tests

Both before and after the feeding treatment, we tested the efficiency of the larvae’s foraging behaviour. The first set of tests commenced 3 days after salamander collection and were repeated following the 55 days of feeding treatment. Because, at collection, the starvation statuses of the larvae were unknown, we fed them during the collection day and then left them without food for 3 consecutive days before beginning the trials. This adjustment period was necessary because satiety is known to reduce aggressiveness [39]. During the behavioural tests, each larva was individually placed in a 13.5 × 18.3 cm plastic container filled with 5 cm water under daylight conditions, with an average lux intensity of 500 lux, and was allowed to acclimatise for 5 min. The box was divided into two areas: an arena for larvae introduction and an area for prey introduction. The area of prey introduction was divided from the arena by a line which, if required, it was possible to arrange into a barrier of the same length of the box border and 4.8 cm in height. For the olfactory stimulus-only scenario, this barrier served to prevent larvae vision while permitting the spread of chemical cues. During tests, the food item was always placed in the same position.

To assess foraging behaviour, we recorded, for each larva (hereafter named “focal”), the time of approach to a potential prey item under four sets of experimental conditions, including chemical and visual stimuli presented together, presented separately, or both being absent (Figure 1). For each of these subtests, we performed two non-successive trials. Trials lasted until the focal larva reached, with the mouth, the divisor between the arena and the prey introduction area, or until the time exceeded 5 min. If the focal larva did not reach the divisor, we recorded the time as 5 min. For each focal larva, the same four subtests were performed in a randomized order, twice before and twice after rearing. See Figure 2 for more information.

Testing conditions were as follows:

(a)

Concurrent occurrence of chemical and visual stimuli (O+V+): A defrosted Chironomus sp. larva (red) was placed into the water in the area of prey introduction using shaking tweezers. Tweezer shaking continued throughout the trial. The defrosted larva released chemical cues into the water, and shaking gave visual cues.

(b)

Chemical stimulus only (O+V−): after dividing the arena from the area of prey introduction, a defrosted Chironomus sp. larva was quickly introduced by gravity.

(c)

Visual stimulus only (O−V+): A decoy (not coloured) constituting an odourless plastic string was placed into the area of prey introduction using shaking tweezers. Tweezer shaking continued throughout the trial.

(d)

Test control conditions, with the absence of both visual and chemical stimuli (O−V−): after dividing the arena from the area of prey introduction, a decoy constituting an odourless plastic string was placed by gravity.

2.3. Statistical Analyses

We analysed the relationship between feeding treatments and foraging behaviour with generalized linear mixed models (GLMMs). For this analysis, the dependent variable was the time employed by the focal larvae to reach the area of prey introduction. The independent variables included visual stimulus (present or absent); olfactory stimulus (present or absent); origin of larvae (cave or stream); test period (before or after rearing treatment); size difference between focal and target larvae; and the interaction between feeding stimuli and test period. We included larval identity, rearing block, and trial number as random factors in the model. Finally, we assessed the significance of effects using the Wald F test [40]. All analyses were performed in R version 3.2 [41].

3. Results

Both cave and stream larvae successfully approached the divisor between the arena and the area of prey introduction when presented with visual or chemical stimuli. No differences between cave and stream larvae were observed. The time of approach was generally lower for test conditions in which one or more stimuli occurred, compared to subtests without any stimuli (

Table 1. Relationships between the time taken to approach the area of prey introduction and larvae origin (cave/stream); period of testing (before or after visual conditioning); occurrence of a visual cue during the test; and occurrence of a chemical cue during the test, as well as the interactions between the test period and the two different cues used during behavioural tests. NumDF = degrees of freedom at numerator. DenDF = degrees of freedom at denominator. Significant relationships are reported in bold.

Variables	B	NumDF	DenDF	F. Value	p
Cave vs. stream	−0.01	1	65.06	0.03	0.85
Test period	−0.021	1	636.76	0.52	0.47
Visual cue	−0.47	1	605.63	271.76	<0.001
Chemical cue	−0.17	1	605.63	38.22	<0.001
Test period x visual cue	−0.93	1	605.49	475.52	<0.001
Test period x chemical cue	0.12	1	605.45	7.86	<0.01

). Moreover, while the test period was not significant, its interactions with both visual and chemical stimuli were. However, the trends were opposite (Table 1). After rearing, for the tests including a visual stimulus, there was a reduction in the average approach time by more than 170 s (Figure 3,

Table 2. Time (in seconds) of larvae approach to the border of the area of prey introduction. O+V+ = contemporary occurrence of visual and chemical stimuli; O+V− = chemical stimulus only; O−V+ = visual stimulus only; O−V− = no visual nor chemical stimuli occurred (control conditions).

	O+V+	O−V+	O+V−	O−V−
Before conditioning	87 ± 13.06 s	183 ± 17.11 s	250 ± 12.64 s	276 ± 8.2 s
After conditioning	11 ± 0.69 s	10 ± 0.74 s	217 ± 11.83 s	279 ± 7.81 s

). For tests including a chemical stimulus, a reduction occurred after rearing, but was lower.

4. Discussion

Our study demonstrates that both chemical and visual perception are important for the prey detection of fire salamander larvae. Individuals readily approached the area of prey introduction when presented with chemical, visual, or both types of stimuli, with significantly faster responses compared to trials without cues. This underscores the importance of both senses for prey recognition in aquatic environments. Both vision and chemical perception are used by aquatic predators to detect prey. However, olfactory cues are thought to be particularly relevant in several freshwater habitats because of the poor visibility available in biotopes with high turbidity levels and reduced light incidence during the nighttime [42,43]. Nevertheless, for several aquatic salamanders or larvae, vision is considered a key aspect of prey detection [14,15]. Fire salamander larvae may be active during both day and night, even if encounter probability is much higher at night, especially when predators are present in the same habitat [44]. Even though recent studies highlight that night vision in amphibians may be much more sophisticated than originally thought [45], S. salamandra usually breeds in small streams flowing in wooded narrow valleys, during which low illuminance levels are likely to occur at night and prey olfactory detection may be a key aspect to allow effective foraging by the larvae [46,47]. For example, the Western lesser siren Siren intermedia nettingi, which is a nocturnal forager with noticeably reduced eyes, uses chemical cues over visual cues when foraging [11]. For the salamander Plethodon cinereus, visual cues are the primary mode of prey detection used during daytime and are linked to an ambush foraging strategy; while during nighttime, these salamanders can switch to a more active foraging behaviour that involves the use of chemical cues to detect prey in total darkness [48]. A similar situation has recently been described in fire salamander larvae that are able to switch from ambush mode to an active foraging strategy depending on light conditions; larvae born in caves are, however, more capable of performing this switch [34]. The active foraging strategy, probably mediated by chemical olfactory sensing, is displayed by the cave-adapted salamander Proteus anguinus and by the cave populations of the Pyrenean brook newt, Calotriton asper [13,49]. It is worth noting that fire salamanders are an ovoviviparous species in which females keep larvae in their uterus before deposition. In the mother’s uterus, adelphophagy may occur [50], so chemical and mechanical perception may be important for the detection of conspecifics in the early stages of larval development.

We did not detect any effect of larvae origin on the time of approach with the different stimuli, meaning that both cave-born and stream-born larvae used both visual and chemical detection to locate prey. This may be because both of the tested sensory systems are highly plastic during the early stages of larval development and their use is mediated by environmental stimuli, determined by foraging conditions and prey typology availability. However, there exists a notable caveat—we tested larvae that were only reared in normal day–night light conditions, not in the total darkness that occurs in caves. Thus, olfactory conditioning may have been masked, due to the over-availability of visual stimuli for food recognition. To further investigate this, future studies could replicate the same experimental plan, but with two distinct lighting conditions for rearing: complete light, with normal day–night light variation, and total, continuous darkness. This would help verify whether olfactory conditioning through chemical stimuli has a different effect on cave-born salamander larvae compared to surface-born ones.

Generally, our results show that, during larval development, experience can be an important trait in affecting foraging behaviour of fire salamanders. During the second test period, when chemical or visual stimuli occurred, larvae had a much faster approach to the area of prey introduction so that the difference between the first and the second test period was highly significant for the occurrence of visual stimuli. This indicates that the rearing experience involving sight could have affected larvae foraging behaviour.

We observed a limited effect of the size of the individuals tested. Although larval body lengths were longer in the second test period, thus enabling quicker movements generally, the test period alone did not result in significantly quicker approach times.

The decreased velocity of approach acquired by fire salamander larvae during the second test period when visual stimuli occurred is consistent with the findings of [35]. They found that, prior to metamorphosis, fire salamander larvae were much more stimulated when catching objects similar in shape to the prey with which they were fed during rearing than they were with objects shaped similarly to other prey. Moreover, in our study, the larvae also learned to recognise the motion performed by the operator with the tweezers, meaning that the recognition of prey movement may be an experience-dependent trait. The role of experience has been increasingly studied in amphibians, being appreciated for bolstering prey recognition and discrimination in different salamander and frog species [4,51].

The most important result of our study is a preliminary indication that sensory conditioning influences chemical and visual perception. While incorporating chemical or visual stimuli significantly reduced the time of approach, this effect was different following sensory-conditioned rearing. After rearing (with mostly visual conditioning), the time taken to respond to visual stimuli decreased more than that taken to respond to chemical stimuli. This result highlights a possible shift during larval conditioning with both visible light and olfactory cues, where sight overrides chemical perception in driving foraging behaviour. Nevertheless, our results show significantly reduced approach times when olfactory cues were provided during feeding compared to without them, indicating that olfactory detection remains important, even if it was not relied upon during rearing. A comparative study between chemical and visual sensory systems has been carried out on the salamander Eurycea multiplicata griseogasteris which, in natural streams, is more efficient in detecting its predators chemically than visually, but no information is available on the role that experience played in this efficiency [52].

Encounter history has been reported to play a major role in detection for predator–prey interactions. Experience affects the learning memory of predator encounters [50] and enhances the prey’s ability to escape predatory attacks [53]. Fish that are visually and/or chemically trained with predator cues are more likely to survive compared to individuals with no pre-existing experience with such cues [50]. However, there are no behavioural experiments on how an individual’s feeding history may affect the interactions of different sensory systems to produce a behavioural response, especially in predator amphibian species. Our results suggest that when individuals learn that a stimulus represents a trophic resource through one sensory mode (e.g., vision), they may display a different response to that feeding opportunity compared with a predator that has experienced the same trophic resource through another sensory mode (e.g., olfaction chemical sensing).

Understanding how predators detect the presence of suitable prey is fundamental to the comprehension of the dynamics of predator–prey interactions [54,55,56] and of the pressures affecting predator fitness [57]. Our research investigates the effects of feeding history on the relative roles of visual and olfactory sensory systems in experience, and shows how the learning gained through visual conditioning may affect the foraging behaviour of fire salamander larvae. In particular, under daylight conditions, visual perception overcame olfactory–chemical sensing by becoming predominant in driving the fire salamanders’ approach to food items.

In natural conditions, the heterogeneity of trophic resource availability or the Hunters’ Horizon (HuHo) variation across the interface between groundwater and surface water can determine strong intraspecific variation of foraging behaviour, with multiple stimuli and different pressures on the sensory systems being used to detect food [58,59,60,61]. Studying the responses of both typical cave organisms exploiting interfaces and typical surface animals colonizing caves can allow us to understand which pressures are driving such exploitations. This study provides a basis for further research aiming to investigate plastic responses in amphibian species capable of foraging in markedly varied environments.