Circumventing surface tension: tadpoles suck bubbles to breathe air

Abstract

The surface tension of water provides a thin, elastic membrane upon which many tiny animals are adapted to live and move. We show that it may be equally important to the minute animals living beneath it by examining air-breathing mechanics in five species (three families) of anuran (frog) tadpoles. Air-breathing is essential for survival and development in most tadpoles, yet we found that all tadpoles at small body sizes were unable to break through the water's surface to access air. Nevertheless, by 3 days post-hatch and only 3 mm body length, all began to breathe air and fill the lungs. High-speed macrovideography revealed that surface tension was circumvented by a novel behaviour we call ‘bubble-sucking’: mouth attachment to the water's undersurface, the surface drawn into the mouth by suction, a bubble ‘pinched off’ within the mouth, then compressed and forced into the lungs. Growing tadpoles transitioned to air-breathing via typical surface breaching. Salamander larvae and pulmonate snails were also discovered to ‘bubble-suck’, and two insects used other means of circumvention, suggesting that surface tension may have a broader impact on animal phenotypes than hitherto appreciated.

So still the Tadpole cleaves the watery vale

With balanc'd fins, and undulating tail;

New lungs and limbs proclaim his second birth,

Breathe the dry air, and bound upon the earth.

     —Erasmus Darwin in The Temple of Nature (1803) [1, pp. 31–32]

1. Introduction

Philosopher and social critic, Marshall McLuhan, famously observed, ‘The medium is the message’ [2, p. 7]. This may be as true in biology as in culture. Aquatic organisms, for example, are profoundly shaped by the physical characteristics of water, including its viscosity, thermal conductance and high freezing point, to name a few. Less well appreciated, however, is its surface tension. Owing to the strong cohesion of water molecules via hydrogen bonds and their weaker adhesion to the air molecules above the water, the molecules at the water's surface are especially tightly bound together. This strong binding causes the surface layer to contract, minimizing its area. As a consequence, the water's surface behaves functionally like a taut, elastic membrane. For large animals with substantial mass, hydrogen bonds are negligible and surface tension, inconsequential. For very small animals, however, surface tension is a tangible reality that can be exploited adaptively. Epineuston, for example, are small invertebrates, including springtails, insects and spiders, that spend much of their lives moving on top of, or within, the water's surface [3]. Rarely considered, however, is the ‘flip side’ of this phenomenon—aquatic organisms for which surface tension represents a roof rather than a floor, and a barrier instead of an opportunity. This circumstance should apply, for example, to small, aquatic animals that must regularly access air above the water to breathe.

Most air-breathing aquatic vertebrates are large enough to merely ‘breach’ the water's surface to obtain air. However, many amphibians have fully aquatic larvae that begin life at very small body sizes. Frog larvae (tadpoles) are often particularly minute at hatching and are among the smallest, free-living vertebrates in the world. Furthermore, despite their use of gills and cutaneous gas exchange, most tadpoles also develop lungs and frequently surface to breathe air (reviewed in [4]). Air-breathing is essential for survival in the hypoxic waters many tadpoles experience for some, if not most of their lives [5]—a problem that is exacerbated by the presence of predators in open, more oxygenated waters [6–8]. Air-breathing may also be necessary for normal lung development in some species [9]. These factors suggest that there is likely to be strong selection on lunged tadpoles for accessing air efficiently. Given their tiny body size, we wondered if surface tension is, indeed, an impediment to air-breathing in young tadpoles and if so, how they are able to circumvent it.

Although the occurrence of air-breathing in tadpoles has been widely noted [4,8,10–12] and its physiology well characterized in a few species (reviewed in [13]), the mechanics of air-breathing are virtually unstudied [14]. When referred to at all, tadpole air-breathing is described, unhelpfully, as ‘gulping’ [15] or ‘swallowing’ [16] air. We hypothesized that either (i) air-breathing would not be necessary for tadpoles until they were large enough to breach the surface, or that (ii) tadpoles would exhibit adaptations that allow them to access air some other way. If the latter, we predicted that, (iii) as tadpoles gain mass and swimming speed during growth, they should transition to typical breach-breathing.

We used high-speed macro-videography to visualize air-breathing throughout the larval period in five tadpole species and found that surface tension does, indeed, represent a barrier for small tadpoles attempting to reach air. We describe a novel vertebrate breathing mechanism that tadpoles employ to circumvent this barrier and present preliminary evidence from several unrelated taxa suggesting that the impact of surface tension on animal phenotypes may be more widespread than previously considered.

2. Material and methods

(a). Animals

We examined tadpoles of five frog species representing three families (table 1). Eggs of two ranid and one hylid species were collected in the field under permit within 5 km of the University of Connecticut campus (Storrs, CT, USA), March to June, 2018–2019. Eggs of a second hylid, Pseudacris crucifer, are minute, laid individually on vegetation and difficult to find. We, therefore, collected three amplexing pairs of frogs and placed each in a 19 l bucket filled with marsh water and vegetation. Buckets were covered with nylon mesh and left in situ overnight. The adults were then released and vegetation with adherent eggs was returned to the laboratory. Finally, 100 fertilized eggs of albino African clawed frogs (Xenopus laevis) were purchased from a commercial dealer (Nasco, Fort Atkinson, WI, USA). All live animal use was approved by the University of Connecticut Institutional Animal Care and Use Committee (IACUC protocol A18-032).

Table 1.

Species studied and bubble-suck (BS) duration.

species	common name	estimated no. of individuals a	body length b (mm)	BS duration (s) mean ± s.d.) c
Ranidae
Rana clamitansd	green frog	300	4–35	0.345 ± 0.093
Rana sylvaticad	wood frog	400	5–16	0.299 ± 0.069
Hylidae
Hyla versicolor	grey tree frog	500	2.5–15	0.460 ± 0.211
Pseudacris crucifer	spring peeper	200	5–10	0.326 ± 0.086
Pipidae
Xenopus laevise	African clawed frog	100	3–11	0.321 ± 0.181

^aIn all cases except X. laevis, multiple clutches were collected over 2 years, making a precise count impracticable. Xenopus laevis were purchased as fertilized eggs from a dealer.

^bFrom hatch to start of forelimb emergence.

^cn = 10.

^dWe use the generic name Rana rather than Lithobates, as recommended by the most recent and authoritative work on ranid systematics and taxonomy [17].

^eCommercially bred albinos.

Eggs were maintained in 38 l aquaria filled with 19 l of untreated water kept at room temperature (20°C). Each aquarium held multiple clutches of eggs. Tadpoles were distributed to additional aquaria to maintain approximately 50–200 tadpoles per aquarium, depending on size. They were fed boiled (10–15 min) green-leaf lettuce ad libitum except for X. laevis, which were fed commercial food (Nasco ‘frog brittle’™) every 2 days. Aquaria were regularly cleaned with hot water and soap.

As part of a separate study, eggs of spotted salamanders, Ambystoma maculatum (Amphibia, Urodela), were also collected and the larvae raised in the laboratory for videography. Larvae were maintained on Daphnia and mosquito larvae while small, and tadpoles when larger. In addition, several small, invertebrate species were collected as bycatch. Respiratory behaviour in these taxa was filmed for comparison to tadpoles.

(b). Videography

Breathing behaviour was visualized with an Edgertronic SC1 monochrome high-speed video camera shooting at 30–1000 frames per second (fps). One to a dozen tadpoles were placed in a small glass or plastic filming chamber; occasionally, tadpoles were filmed in their home tanks. Three large LED video light banks provided illumination. Tadpoles rose to the surface to breathe regularly, thus we were able to capture hundreds of breathing events on video. Videos of spotted salamander larvae and several invertebrates were obtained similarly.

3. Observations

(a). Onset of air-breathing

Tadpoles of all species inflated their lungs and began air-breathing at minute body sizes (as small as 3 mm body length) within three or four days of hatching. However, histological data for Hyla versicolor show that the lungs do not become vascularized until sometime later in development [18] and this may be true for other species, as well. As such, air-breathing/lung filling at very early stages may be non-respiratory.

(b). Surface tension as a barrier to air-breathing

As predicted, all tadpoles were initially unable to breach the water's surface, despite many attempts to do so. Tadpoles were sometimes observed to swim rapidly upward, hitting the surface and bouncing off (electronic supplementary material, video S1). At impact, the water's surface was pushed upward and stretched elastically. Even when a tadpole continued to swim actively upwards, the surface quickly returned to its equilibrium position. Tadpoles then either swam away or initiated a novel breathing behaviour that does not require breaching the surface to access air.

(c). ‘Bubble-sucking’—breathing without breaching

At small body sizes, while unable to breach, all tadpoles employed an alternative breathing mechanism we call ‘bubble-sucking’ (figure 1; electronic supplementary material, videos S2–S4). Bubble-sucking consists of several discrete mechanical events, described next. This description is based largely on our data for X. laevis tadpoles because their transparency permitted definitive observations of lung emptying and filling. Exceptions or variations on the X. laevis pattern are noted. Bubble suck duration for each species is given in table 1.

Figure 1.

Bubble-sucking in three species of anuran tadpoles, ventral view, from high-speed videos. (a–d) Xenopus laevis (Pipidae). (e–h) Rana sylvatica (Ranidae). (i–l) Hyla versicolor (Hylidae). Note that in all photos, the view is slightly upward so that the mirrored undersurface of the water is visible at the top of the frame, reflecting the tadpole's image. The albino X. laevis tadpoles are nearly transparent, except for the opaque gut, which appears as a white ball near the vent. The first panel in each row (a,e,i) shows attachment, which is followed very quickly by initiation of the bubble-suck (b,f,j), during which lung emptying usually occurs. The third column (c,g,k) shows each tadpole after pinch-off and detachment from the surface when the buccal air bubble is compressed and forced into the lungs (not visible in these sequences). Finally, remaining buccal air is expelled from the mouth (d,h,l). Not to scale.

Tadpoles of the two hylid species examined exhibited a variant form of bubble-sucking at larger body sizes characterized by a second suction event immediately following the first. We treat this phenomenon in a separate paper [18]. Here, we restrict our attention to ‘single bubble-sucks’, the form of bubble-sucking that occurred in all species.

(i). Attachment

A bubble-suck cycle is initiated by attachment of the tadpole to the undersurface of the water with its mouth (figure 1; electronic supplementary material, videos S2–S4). The tadpole pushes up against the surface of the water, typically rocking from side-to-side until a suitable position is attained (figure 2). The tadpole then opens it mouth wide and uses the water's surface tension to adhere to the surface, as well as buccal suction (see next section). In hylids and ranids, the well-developed oral disc is shaped into a circular cup with its rim fitted to the undersurface, forming a seal, possibly aided by the disc's fimbriated margin, which increases its circumference (figure 3a). Xenopus laevis tadpoles lack an oral disc and attach only briefly.

Figure 2.

Kinematic plots of breach-breathing (a) and bubble-sucking (b) in large and small specimens of R. clamitans, respectively. In both cases, the tadpoles move from bottom to top, and from right to left. During bubble-sucking, the water surface is typically stretched upwards as the tadpole pushes against the surface, attempting to attach (dotted line in (b)). (Online version in colour.)

Figure 3.

A comparison of bubble-sucking and breach-breathing. (a) Hyla versicolor (Hylidae) showing attachment to the water's undersurface via the oral disc, which is formed into a circular cup. Although air is drawn into the mouth, the surface tension is not broken. (b) Rana clamitans (Ranidae). This large, 2nd year tadpole has no trouble breaking through the surface to draw gaseous air directly into the buccal cavity. Not to scale. (Online version in colour.)

Approach and attachment angle vary widely among species. Xenopus laevis mouths are near terminal, but open dorsally and tadpoles sometimes attach obliquely, dorsal side up. The sub-terminal mouths of the hylids and ranids, however, promote oblique attachment with the ventral side up. Near perpendicular attachment is also possible in all species.

(ii). Bubble-suck, lung emptying and pinch-off

Once attached, the tadpole depresses the floor of its mouth and pharynx, rapidly expanding the buccal cavity. The surface of the water is drawn into the buccal cavity, creating a pocket of air, which briefly maintains a connection to the air above the surface (figure 1; electronic supplementary material, videos S2–S4). Buccal water presumably is displaced by the air, exiting through the spiracle, but some water remains as the bubble is drawn in, its surface tension unbroken. In most tadpoles, the patent connection to the water's surface is narrow and stalk-like, with a tiny, circular opening, but in the broad-mouthed X. laevis, the mouth forms a wide, circular connection to the surface (figure 4; electronic supplementary material, videos S2–S4).

Figure 4.

Mouth form during attachment and bubble-sucking in two tadpole species shown from above and from the side. (a,b) Xenopus laevis, with a broad round mouth lacking mouthparts. The buccal bubble has a correspondingly wide attachment to the surface air. (c,d) Hyla versicolor, with a narrow mouth opening, an oral disc formed into a cup that seems to facilitate attachment, and a narrow, stalk-like connection of the buccal bubble to the surface. Not to scale.

The transparency of the albino X. laevis tadpoles makes lung emptying directly visible (figure 5), but in other species it is frequently evident externally on the dorsomedial surfaces where the lungs bulge or are visible through unpigmented patches of skin (electronic supplementary material, videos S3–S4). Lung emptying is explosive (3–6 ms) and typically occurs while the buccal bubble is open to the surface air. Lung air is injected directly into the buccal bubble. In early tadpoles, however, lung emptying often occurs after mouth closure, or rarely, not at all. In only one of many such sequences observed in X. laevis, lung emptying resulted in the formation of a second, posterior buccal bubble. This smaller, second bubble was immediately compressed back into the lungs (see next).

Figure 5.

A bubble-suck in X. laevis showing lung emptying and lung filling. (a) Initiation of a bubble-suck. The bubble is highly reflective and appears as a silver-white shape. The tadpole is shown in dorsal view so that the brain blocks the bubble in the midline. The arrows indicate the inflated lungs. (b) In the frame immediately before lung emptying, the buccal bubble has expanded posteriorly. (c) Just 3 ms later, the lungs are have completely emptied into the buccal bubble while the mouth is still open to the air. (d) Following detachment, the buccal bubble is compressed; the lungs have just begun to fill (arrow head). In dorsolateral view, the full extent of the buccal bubble is visible. The bubble seems to be larger in X. laevis than in other species, apparently because it fills most of the pharyngeal/branchial space as well as the buccal cavity. (e) The lungs are now fully filled (arrow) and the excess air remaining in the buccal cavity will soon be expelled.

Finally, mouth closing, or ‘pinch-off’, severs the buccal bubble's surface connection. Pinch-off finally breaks the surface tension of the water, probably aided by keratinous beaks in the ranid and hylid tadpoles. Xenopus laevis tadpoles typically jerk their heads laterally when breaking away from the surface, probably relying on shear to break the surface connection (electronic supplementary material, video S2).

(iii). Compression, lung-filling and bubble expulsion

Following pinch-off, the floor of the mouth is elevated, compressing the air bubble within the buccal cavity. The air is forced through the glottis and into the now empty lungs. Although complete lung emptying prior to compression and filling was observed in X. laevis tadpoles of all sizes, in the smallest tadpoles of other species the lungs sometimes appeared to remain partially full. Lung filling is slower than emptying (50–70 ms) (electronic supplementary material, videos S2–S5). In all tadpoles, the volume of the buccal bubble exceeds the volume of the lungs. Thus, after lung-filling, air remains within the buccal cavity. Buccal compression expels the remaining air from the mouth.

(d). Transitioning to breach-breathing

As predicted, tadpoles of X. laevis and the two ranid species transitioned from bubble-sucking to breaching as they grew. In breaching, the body is thrust upward, mouth first, with enough force to rupture the elastic surface tension, exposing the mouth to gaseous air. While the mouth is open to the air, the lungs empty and fresh air is drawn in via buccal expansion (figure 3b; electronic supplementary material, video S5). Subsequent breathing stages are identical to bubble-sucking. Ranid tadpoles (particularly Rana clamitans) exhibited a distinct behaviour at intermediate body sizes (jaw breaching) in which the upper jaw is snapped open from just beneath the surface, breaking the surface tension as the mouth is pushed into the air. Although larger individuals nearly always breach-breathed, they continued to bubble-suck at low frequencies until metamorphosis.

Contrary to our prediction, H. versicolor tadpoles never transitioned to breach-breathing as they grew. Rather, they continued to bubble-suck throughout the larval period and metamorphic climax (electronic supplementary material, video S4). Larger tadpoles developed the physical capacity to breach, which we observed occasionally during spontaneous bursts of explosive swimming, but we never observed breathing while tadpoles were above the water. Almost fully metamorphosed tadpoles continued to bubble-suck when submerged, until metamorphic transformation of the mouth apparently made attachment to the undersurface of the water impossible. Tadpoles of the second hylid species, P. crucifer, also bubble-sucked well into metamorphosis, but unlike H. versicolor, some larger individuals also breach-breathed like Rana and Xenopus.

(e). Surface tension and respiration in other small aquatic organisms

We observed respiratory behaviour in salamander larvae and three invertebrate taxa: (i) tiny spotted salamander larvae, like small tadpoles, were unable to breach the water's surface to breathe and like the tadpoles, they circumvented this problem by bubble-sucking (figure 6a). Despite external gills with patent gill slits, they compressed air into the lungs successfully most of the time, but occasionally, air was forced out the gill slits instead, and like the tadpoles, they transitioned to breach-breathing as they grew; (ii) mosquito larvae hung inverted from the surface by narrow breathing tubes (siphons). The siphons penetrate the surface and attach to the surface tension above, suspending the larva. The larvae were observed to detach actively from the water's surface and later, to reattach; (iii) predaceous diving beetles (Acilius sp.) repeatedly rose to the water's surface to refill a supply of air held beneath their elytra. At the surface they floated oblilquely, tail up. As the pygidium (a posterior extension of the terminal abdominal tergite) was elevated, the water's surface layer was attracted to its dorsal side, pulling the surface down so that it spanned the gap between the abdomen and the raised elytra. This connected the bubble beneath the elytra to air above the surface for replenishment (figure 6d,e); (iv) unexpectedly, pulmonate pouch snails (Physa heterostrophus) were also discovered to bubble-suck, using their pulmonary siphons to draw air from the surface without breaching it (contra [19]), presumably via pulmonary sac expansion. Snails bubble-sucked while moving along an aquarium side or while inverted, gliding beneath the water's undersurface (figure 6b,c).

Figure 6.

Three non-anuran, freshwater species exhibiting respiratory adaptations related to circumventing the physical constraint of surface tension. (a) A hatchling spotted salamander (Ambystoma maculatum) bubble-sucks at the surface (note its reflection on the undersurface of the water and the large buccal bubble. (b) A pulmonate snail (Physa heterostrophus) using its pulmonary siphon to bubble-suck while it moves along the aquarium side (length ≈ 17–20 mm). (c) Another snail bubble-sucking while moving on the undersurface of the water. (d) A predaceous diving beetle (Acilius sp.) uses a combination of hydrophilic and hydrophobic materials to maintain the appropriate angle at the water's surface while drawing a bubble into the space beneath its elytra. (e) Another diving beetle as it begins to swim away from the surface. Scale bar in millimetres.

4. Discussion

(a). Breathing by bubble-sucking

To our knowledge, air-breathing by bubble-sucking has not previously been described for any aquatic vertebrate. Given that small tadpoles attempt to breach the water's surface, but cannot, and that the same tadpoles when larger do breach to breathe, we conclude that bubble-sucking is a larval adaptation to circumvent the physical constraint of surface tension.

In both bubble-sucking and breach-breathing, buccal expansion serves to fill the buccal cavity with air that is then compressed into the lungs. In both cases, lung emptying (usually) occurs during this expansion phase while the mouth is open to the air. There are no functional differences between breathing modes thereafter.

The transition from bubble-sucking to breach-breathing occurs as a consequence of growth manifested in greater mass and swimming speed. Tadpoles regularly ‘test’ the surface tension by accelerating rapidly towards the surface, but most often they drift slowly upwards, gently attaching to the surface to bubble-suck. As such, bubble-sucking is not merely a default behaviour invoked when breaching initially fails, but often appears to reflect the tadpoles initial intent.

If small tadpoles are unable to produce sufficient force to break the water's surface tension, how are they able to pull the surface into the mouth during a bubble-suck? We cannot answer this question directly, but note that the force generated by negative pressure within a tadpole's buccal cavity is exerted over a very small area, which would significantly increase its mechanical advantage, presumably enough to overcome the elastic resistance of the surface tension.

(b). Lung emptying and filling

Like virtually all other anamniotes, tadpoles use a buccal pump to force air into the lungs [18]. Diving and hydrostatic pressure do not drive lung filling (contra [20]). Several mechanisms, however, could underlie lung emptying: negative pressure within the buccal cavity [21], or positive pressure within the lungs caused by either intrinsic or extrinsic muscle contraction, or by recoil of stretched elastic elements within the lungs [8,22,23]. Our observations of emptying in small tadpoles following mouth closure, sometimes during buccal compression, demonstrate unambiguously that air is expelled under positive pressure, but we cannot say by what mechanism.

In X. laevis, lung air was often seen to extend anteriorly towards the glottis as surface air expanded the buccal cavity. Lung emptying typically occurred when the two air bubbles met at the glottis. The adjacency of the bubbles is probably important to insure their fusion when the lungs empty. The jetting of lung air through a narrow glottis would insure that it exerts high pressure over a small area of the buccal bubble to break its surface tension. Only once was lung emptying seen to generate a second bubble within the buccal cavity; this occurred when a narrow zone of water intervened between the glottis and the buccal bubble. The water apparently dissipated the pressure of the jetted lung air, preventing it from breaking the buccal bubble's surface tension.

It has been suggested that adult bullfrogs expel depleted air from the lungs in a dorsal ‘jet-stream’ that bypasses fresh air within the buccal cavity, minimizing mixing before lung filling [24], a conclusion that has been both supported [25,26] and challenged [23,27]. In tadpoles, explosive lung emptying suggests the possibility of a coherent air stream, but their much smaller buccal cavity compared to adults renders the avoidance of turbulence and mixing unlikely. Lung emptying typically occurs at the same time that fresh air is drawn into the mouth so that incurrent and excurrent airstreams are in opposition, also promoting mixing. Furthermore, jet-stream models depend on expulsion of air through the external nares, which remain closed throughout much of tadpole development. Thus air-breathing in most tadpoles probably results in the lungs being filled with mixed air. This inefficiency, however, should be mitigated by the greater volume of the buccal cavity and fresh air relative to the lungs and depleted air.

(c). Surface tension as a physical constraint

It is widely understood that physical laws and material properties impose limits on phenotypic evolution, circumscribing the universe of realized phenotypes [28,29]. For example, body size in early metazoans would have been severely limited by the slow rate of oxygen diffusion. Nevertheless, this physical constraint was eventually overcome by the evolution of circulatory and respiratory systems. Similarly, the physical constraints of fluid viscosity and drag limit the maximum air speed a bird can attain. In this case, however, a bird can largely overcome this constraint behaviourally by streamlining its shape (wing-tucking) and diving (e.g. peregrine falcons fly at a maximum of 150 km h⁻¹, but dive at 320 km h⁻¹ [30]). These examples illustrate that physical constraints are not absolutes, but apply only to a particular body plan that is potentially modifiable [29] and that phenotypic changes in response to such constraints are manifested at two temporal scales—evolutionary adaptation and/or within an organism's lifetime (phenotypic plasticity). Tadpole bubble-sucking represents a plastic response to a physical constraint, like the falcon's, by means of a behavioural shift. However, the inclination to behave this way may, itself, reflect evolutionary adaptation, as might modification of the mouthparts that improve surface attachment in some species.

Our preliminary observations of salamander larvae and pulmonate snails demonstrate that bubble-sucking is not unique to tadpoles. It is limited, however, to organisms that can generate suction—something arthropods, for example, cannot do. Regardless, aquatic insects employ diverse adaptations to circumvent the problem of surface tension in other ways (e.g. [31–35]). Our observations of mosquito larvae and diving beetles, for example, highlight two common themes among air-breathing aquatic insects: the use of narrow-diameter structures to penetrate the surface tension to access air, and the use of hydrophilic and hydrophobic body parts to manipulate the water surface or to trap air bubbles for storage within cuticular hairs [32,33].

In summary, our observations suggest that surface tension may have far-reaching consequences for phenotypic evolution in small, aquatic animals. As a physical constraint, surface tension could be viewed as a limiting factor in evolution. However, it may be more usefully cast as a source of selection pressure driving diverse and often remarkable adaptive responses, as well as convergent phenotypic solutions in unrelated taxa.

Ethics

Work with live tadpoles was approved by the University of Connecticut Animal Care and Use Committee, protocol A18-032.

Data accessibility

Sample videos are available in the electronic supplementary material and from the Dryad Digital Repository: https://doi.org/10.5061/dryad.2ngf1vhjj [36]. Additional videos are part of ongoing studies and are available upon request from the authors.

Authors' contributions

K.S. conceived the study and wrote the first draft of the manuscript. J.R.P. edited the manuscript. K.S. and J.R.P. collected, analysed and interpreted the data, and produced the figures.

Competing interests

We declare we have no competing interests.

Funding

Funding was provided by the Department of Ecology and Evolutionary Biology to K.S. and the EEB Graduate Student Research Fund to J.R.P.