Feeding behaviour in a ‘basal’ tortoise provides insights on the transitional feeding mode at the dawn of modern land turtle evolution

Ecological background

Both extant species of Manouria, the Asian forest tortoise M. emys and the impressed tortoise M. impressa, have a restricted distribution in Southeast Asia. M. emys has a narrow distribution in Bangladesh, India (Assam, Meghalaya, Mizoram, Nagaland), Myanmar, Thailand, Malaysia (East and West), and Indonesia (Kalimantan, Sumatra). The nominate subspecies, M. emys emys—the subject of this study, inhabits the southern part of the species range (Fritz & Havaš, 2007; Stanford et al., 2015).

M. emys inhabits tropical evergreen monsoon forests and exhibits high tolerance for soil moisture. It is commonly found reposing in wet areas, buried in mud or under the leaf litter where it may spend long periods of time. It is active even during rainy weather. Direct sun exposure and basking are not required. Furthermore, this species has a mostly crepuscular and nocturnal lifestyle (Ernst, Altenburg & Barbour, 2000; Vetter & Daubner, 2000; Stanford et al., 2015).

According to the available literature, the diet of M. emys includes plants, fungi, invertebrates, and frogs (Nutphand, 1979; Das, 1995; Lambert & Howes, 1994; Høybye-Mortensen, 2004). It has been reported to feed on plants in shallow mountain streams (Nutphand, 1979).

Experimental setting

Animal husbandry and experiments were in strict accordance with the Austrian Protection of Animals Act. The animals used in the present study were obtained commercially and kept at 12 h dark/light cycles in a large terrarium (150 × 100 cm ground area) with a permanently filtered water basin and spacious terrestrial area. The turtles were fed different fruits, vegetables, commercially obtained tortoise pellets, dead mice, as well as pieces of cattle heart and liver, offered on the terrestrial part of the terrarium. Carapace lengths in the three subadult experimental animals ranged between 109 and 135 mm with body masses between 234 and 236 g. For filming terrestrial feeding, the specimens were put in a dry glass cuvette (24 × 60 × 30 cm). When the food was offered on the floor of the cuvette, the tortoises often twisted their necks and rotated their heads in attempt to grasp the food item. The side movements made the filming of the animals in strict lateral view very difficult and the landmarks were not clearly visible during the sequence. By the use of forceps for food display we completely eliminated these problems and were able to shoot perfect lateral plans of the feeding turtles. The food in the feeding experiments was offered at a position which was similar to the position on which we offered the food in the terrarium where the tortoises were housed. The animals did not extend vastly their necks to reach the food items (see Appendix S1). The position of the offered food was completely “natural”. The tortoises needed to stretch their necks forwards rather than downwards, which did not impacted other kinematic patterns of the feeding cycles.

As food items we used small pieces of cattle heart measuring approximately 5 × 5 × 5 mm. The turtles were filmed from lateral aspect (with a reference grid 1 × 1 cm in the background) via the digital high-speed camera system Photron Fastcam-X 1024 PCI (Photron limited, Tokyo, Japan) at 500 fps with a highly light-sensitive objective AF Zoom—Nikkor 24–85 mm (f/2, 8-4D IF). Two “Dedocool Coolh” tungsten light heads with 2 × 250 W (ELC), supplied by a “Dedocool COOLT3” transformer control unit (Dedo Weigert Film GmbH, München, Germany) were used for illumination. We filmed and analysed the food uptake and the food transport cycles in eight feeding sequences for each specimen.

The setting for filming aquatic feeding of submerged food comprised the experimental aquarium filled with water to a level of 3 cm and presentation of food items in front of the turtle’s snout. In order to reduce the light intensity and for optimisation of the digitising process, the frame rate was reduced to 250 fps. As the tortoises were unable to grasp the food item in a total of 36 trials, the kinematics of the feeding apparatus had been analysed (see below) in nine selected representative feeding trials.

For both terrestrial and “aquatic feeding” sequences, horizontal (X-axis) and vertical (Y-axis) coordinates of relevant landmarks (see Fig. 2) were digitised frame by frame using “SIMI-MatchiX” (SIMI Reality Motion Systems, Unterschleißheim, Germany). Based on the displacement of the markers, we were able to calculate the gape amplitude (distance between the tip of the upper and lower beak), head movement (distance between the anterior tip of the carapace and the point “P” on Fig. 2), tongue movements (distance between the most ventral point on tympanum and the tip of the tongue when visible), and hyoid movements (distance between the point “P” on Fig. 2 and the basis of the posterior ceratobranchial). To compare the kinematic feeding pattern of M. emys to those of other studied turtles and to understand the coordination between the elements of the feeding apparatus, these data were used for calculation of the following kinematical variables: duration of Slow open phase (SO); duration of Slow open phases I and II (SOI and SOII) when present; duration of fast open (FO); duration of maximum gape phase (MG); duration of fast close (FC); time to peak gape (TPG); total cycle duration (TCD); duration of hyoid protraction (HyDD); duration of hyoid retraction (HyVD); duration of the total hyoid cycle (THC); hyoid retraction velocity (HyRV); duration of head protraction (HP); duration of head retraction (HR); duration of tongue protraction (TP); tongue retraction velocity; delay of the start of hyoid retraction relative to the tongue retraction start; delay of reaching peak gape relative to start of the hyoid retraction; delay of reaching peak gape relative to tongue retraction start (see Table 1).

Statistics

We tested for any differences among the frequency of occurrences of defined patterns both in food uptake (FU) and food transport (T), i.e., sequences with: missing split of the jaw opening in SO and FO; without detectable split of discrete SOI and SOII slow gape phase; lacking MG phase. In order to provide the comparisons, Chi-square test with Yates’ correction was performed. Then we tested for possible existence of differentiation in kinematical variables in both feeding stages (FU and T). All variables were tested with the Shapiro–Wilk test for normal distribution. When the p-value was less than the chosen alpha level (p < 0.05), the null hypothesis was rejected and data were excluded from further analyses. In addition, all variables included in Table 1 were tested with Levene’s and Brown–Forsythe tests and then processed with Welch’s ANOVA for heteroscedastic data. Tukey’s honest significant difference test (HSD) was performed for post-hoc analyses when applicable.

Furthermore, in order to express the degree of individual differentiation among the studied specimens, a Canonical discriminant analysis (CDA) was performed. Standard descriptive statistics including mean, range, standard deviation ((SD) and confidential interval at 95% CI) were presented.

Evolution of food uptake among turtles

In general, most aquatic turtles combine a fast acceleration of the head towards the food or prey item and a suction feeding mechanism is induced by fast oropharyngeal volume expansion. In some extant turtles, a strong suction flow can be generated and prey is directly sucked into the oropharynx without contact with the jaws (e.g., Chelus fimbriatus (Lemell et al., 2002), Apalone spinifera (Anderson, 2009), Pelodiscus sinensis (N Natchev & I Werneburg, 2013, unpublished data)). However, most extant turtles cannot generate such strong suction flows and only compensate (“gulp”) the bow wave that otherwise would push small to moderately sized food items away from the fast approaching head. These species finally fix and grasp prey with the jaws (see Lauder & Prendergast, 1992; Lemell, Beisser & Weisgram, 2000; Aerts, Van Damme & Herrel, 2001; Natchev et al., 2009; Natchev et al., 2011). We consider the latter plesiomorphic behaviour for extant turtles.

Among extant turtles, the ability to complete the whole feeding process (including food uptake, food manipulation and transport, esophageal packing, and swallowing) on land has been tested and documented for only six species so far. All of them were members of Testudinoidea (Fig. 1; see also Summers et al., 1998; Bels et al., 2008; Natchev et al., 2009). The terrestrial mode of food uptake differs dramatically among and within the three testudinoid subgroups (see Bels, Davenport & Delheusy, 1997; Bels et al., 2008; Summers et al., 1998; Wochesländer, Hilgers & Weisgram, 1999; Natchev et al., 2009, present study). Correspondingly, it appears as if terrestrial feeding re-evolved several times independently amongst turtles. Unfortunately, only limited experimental data are available on feeding mechanisms in emydids and geoemydids. Further functional and palaeontological investigations may help to sort out the issues on the evolution of the feeding behaviour and the morphology of the feeding apparatus in testudinoids.

Very limited information is available on feeding mechanisms employed by amphibious non-testudinoid turtles that occasionally exploit terrestrial food sources. Weisgram (1985a) and Weisgram (1985b) documented a kinosternid (Claudius angustatus) that caught prey on land and dragged it into water for transport and swallowing. Natchev et al. (2008) documented another kinosternid (Sternotherus odoratus) catching food on land, but failing to transport it through the oropharynx. Among extant turtles, successful food transport on land seems to be restricted to testudinoids. The development of enlarged and muscular tongues within this group (Von Bayern, 1884; Werneburg, 2011) represents adaptation to improved terrestrial food manipulation.

Based on experimental data, Natchev et al. (2009) described and summarised three categories of terrestrial food uptake modes among Testudinoidae: (A) Jaw prehension with retracted hyolingual complex, as observed in the geoemydid genus Cuora (Natchev et al., 2009); (B) Jaw prehension with slightly protracted hyolingual complex, as observed in emydids (Bels, Davenport & Delheusy, 1997; Stayton, 2011); (C) Lingual prehension—the tongue touches the food item prior to food uptake, as documented in all tortoises studied so far (Wochesländer, Hilgers & Weisgram, 1999; Bels et al., 2008). The food uptake mode of M. emys, however, differs substantially from that of all remaining tortoises (category C). In fact, the hyolingual complex in M. emys remained fully retracted during the food prehension on land, and the first contact with the food item was by the jaws. Accordingly, the feeding mechanism of M. emys should be assigned to category A, along with that of semi-aquatic geoemydids.

We now aim to construct a theoretical scenario on the evolution of terrestrial feeding mechanisms in turtles. Given the aquatic origin of all living turtles, the functional transition from aquatic to terrestrial feeding mechanisms could hypothetically have involved four stages, beginning with an exclusively aquatic feeding ancestor. In different lineages and stages, turtles may have left their aquatic environments for various reasons e.g., for exploiting new food niches. The species that retained predominantly aquatic life styles may grasp food by the jaws on land, but have to drag it into the water for further intraoral (hydrodynamic based) transport. In recent turtles, such behaviour was documented in the kinosternids C. angustatus (Weisgram, 1985a; Weisgram, 1985b) and S. odoratus (Natchev et al., 2011), as well as in the emydid Trachemys scripta (Weisgram, 1985b; Weisgram, Dittrich & Splechtna, 1989) and other emydids (see Stayton, 2011). Turtles of the second hypothetical evolutionary stage grasped food by the jaws, while the tongue was used for intraoral food transport on land. Such species would still have retained their underwater feeding ability by using hydrodynamic mechanisms. When grasping food on land, the tongue remained retracted or was protracted without touching the food item. Among extant turtles, such a feeding mode is found in the geoemydid genus Cuora (Heiss, Plenk & Weisgram, 2008; Natchev et al., 2009; Natchev et al., 2010) and in some emydids (Bels, Davenport & Delheusy, 1997; Summers et al., 1998; Stayton, 2011). In the next theoretical evolutionary step (stage three), behavioural and morphological adaptations for terrestrial feeding were further advanced, increasing the efficiency of terrestrial food transport at the expense of the ability to use effective hydrodynamic mechanisms in water. Such species still grasped food items with their jaws on land (as typical for aquatic or semiaquatic turtles), but were no longer able to take up dispersed food if submerged, which features prominently in the present case of M. emys. Finally, in a fourth stage, turtles became fully terrestrial and their tongue was obligatorily involved in food uptake as documented in the tortoises Testudo (Eurotestudo) hermanni boettgeri (Weisgram, 1985b; Wochesländer, Hilgers & Weisgram, 1999), Kinixis belliana, Geochelone elephantopus and G. radiata (Bels et al., 2008).

Our investigations demonstrate that the ‘basal’ tortoise M. emys does not contact food with the tongue prior to jaw prehension on land. This shows that tongue to food contact is characteristic of advanced tortoises only. We consider the terrestrial feeding behaviour of M. emys as plesiomorphic and potentially inherited from its semiaquatic ancestors. On that basis, M. emys can be considered a transitional turtle in regard to secondary terrestriality.

We propose that hyolingual protrusion evolved in the lineage forming to advanced tortoises (Fig. 1). Manouria emys has a large tongue with massive intrinsic and extrinsic musculature (see Heiss et al., 2011). The advanced and complex lingual musculo-skeletal architecture allows the turtle to protrude the tongue outside the margins of the rhamphothecae (see Fig. 2). However, M. emys does not use lingual food prehension as typical for all other tortoises studied so far. In fact, it seems that the Manouria (and perhaps Gopherus (N Natchev, pers. obs., 2015)) “lineage” has retained the jaw prehension mechanism inherited from earlier aquatic ancestors. It seems that the tortoises, in general, evolved fleshy tongues which improve the food transport performance. The advanced tortoises only refined the behaviour of food uptake on land via lingual food contact prior to jaw closure (see Wochesländer, Hilgers & Weisgram, 1999; Bels et al., 2008).

Function of the protruded tongue in the testudinid’s food uptake

What would be the potential advantage of the obligatory lingual protrusion, found in the more derived tortoises? One possible explanation is that the tongue is used as a prehensile organ for food ingestion analogous to that found in other tetrapod groups (for overview see Schwenk, 2000; Schwenk & Wagner, 2001). However, for tortoises such interpretation might be put into question. By examining all published data available (Wochesländer, Hilgers & Weisgram, 1999; Bels et al., 2008), we were not able to find any convincing evidence that tortoises collect food with their tongues—they just touch it. In all published feeding sequences, the contact between the food and the tongue is clearly demonstrated—yet, in all cases, after the initial contact of the tongue with the food, the head moves forward and the food item is not dragged into the mouth by tongue retraction, but is grasped by the jaws during the fast jaw closing (FC gape phase). Initial food ingestion in tortoises might not be considered “lingual prehension” (see Schwenk, 2000; Bels et al., 2008) in the strict sense, but should be regarded as “jaw prehension following lingual contact”. This prompts the question: why is an obligatory contact of the tongue to the food present in extant tortoises (except in Manouria and also conceivably in Gopherus) during food uptake? In other words: why do tortoises apply a more complex and presumably more energetically expensive food uptake mechanism by including movements of the hyolingual complex in addition to the movements of the neck and jaws alone?

We propose that the lingual contact provides tactile information on the position of the food item and helps the advanced tortoises to compensate the “information gap” which occurs when the food is approached to a distance where it is out of sight. The eyes of tortoises are positioned laterally on the head (Pritchard, 1979) and the turtles are not able to permanently observe the position of the food item when the neck is protracted and the gape is positioned around it. The prolonged maximum gape (MG) phase found in most ingestion cycles of M. emys (see Table 1, Figs. 3 and 6) might be the result of lack of lingual contact with the food surface. In all published sequences and kinematical profiles on food uptake in tortoises, there is a clear tendency toward a split of the gape cycle into slow open (SO) and fast open (FO) gape phases (see Wochesländer, Hilgers & Weisgram, 1999; Bels et al., 2008). The lack of tongue protrusion might explain the lack of slow open (SO) and fast open (FO) split in the gape cycle of food uptake in the geoemydid Cuora (see Natchev et al., 2009). In most food uptakes analysed in M. emys, SO phases are not present and the gape increases gradually (see Table 1 and Fig. 6). Similar as in Cuora, the lack of SO phases might be explained by the lack of tongue protrusion in food uptake (see Natchev et al., 2009).

Intraoral food transport on land

The execution of the transport cycles require coordination of the activities of contractile elements such as the jaw opening and closing muscles, head protracting and retracting muscles, intrinsic and extrinsic lingual muscles, as well as muscles that protract and retract the hyolingual complex as a whole unit (Jones et al., 2012; Werneburg, 2011; Werneburg, 2013). In contrast to this complicated choreography, the mode of food prehension in M. emys suggests less complex neuromotoric coordination between neck and jaw movements. Yet, the execution of the transport cycles is often more than twice shorter in duration (see Table 1). A possible explanation for the longer duration of food uptake cycles relative to transport cycles might be that during transport, the coordination centres of the muscle activities execution are permanently supplied with information concerning the position of the food item within the oropharynx and the proper movements can be executed precisely in a shorter time.

In his work on the feeding mechanisms in domestic animals, Bels (2006) established that a pre-programmed Generalise Cyclic Model (GCM)—very similar to those proposed from Bramble & Wake (1985)—is universally valid among the different groups of tetrapods. The feeding kinematics of the Asian forest tortoise differs in some aspects from those proposed by the GCM. The kinematics of the feeding system in M. emys seems to be pre-programmed, but under permanent feedback control. The values of the gape and hyoid/hyolingual cycle patterns in the three specimens studied here show high degrees of variation, both concerning food uptake and food transport (see Table 1 and Figs. 5–7). The slow open phases (SO) are not obligatory. The gape cycle often includes a phase of retaining maximum gape (see Table 1, Figs. 3 and 5). In turtles, the maximum gape phase (MG) was described for the gape cycle in Cuora sp. (Natchev et al., 2009; Natchev et al., 2010) and was confirmed for kinosternidae (Natchev et al., 2011). The presence of a prolonged maximum gape phase (MG) can be easily overlooked when the frame rate of the film sequence is not high enough (i.e., step between successive frames over 10 ms). Thus, it may be present in other turtles, but was not taken into account by the calculations of the kinematical profile (see Stayton, 2011; Nishizawa et al., 2014).

The GCM presupposes that the start of hyoid retraction coincides with the start of fast open phase (FO). However, our calculations (see Table 1 and Fig. 5) demonstrate that in M. emys the hyoid retraction in the food transport cycle starts shortly prior reaching peak gape. The same pattern was detected by the investigation of aquatic, semi-aquatic, but also predominantly terrestrial cryptodirans (Natchev et al., 2008; Natchev et al., 2009; Natchev et al., 2010; Natchev et al., 2011).

Relations between the habitat preferences and the diet in turtles

In the evolution of the testudinids there is a clear shift not only in the habitat preferences (from aquatic to terrestrial), but corresponding shifts are also seen in dietary preferences. In that taxon it manifests in a tendency toward herbivory. Most of the recent tortoises rely on diets mainly composed of plant material (for overview see Pritchard, 1979; Ernst, Altenburg & Barbour, 2000; Bonin, Devaux & Dupre, 2006). The partly carnivorous lifestyle of Manouria sp. (Bonin, Devaux & Dupre, 2006) may be a relic of the carnivorous diet of the ancestor of the tortoises and supports the transitional status of the genus.

Apparently the feeding media (air vs. water) and the dietary shift had a large influence on the overall feeding behaviour of the testudinids (see Bels et al., 2008). The suction mechanism was lost and replaced by a jaw food prehension system (this study) or by “lingual prehension” (Wochesländer, Hilgers & Weisgram, 1999; Bels et al., 2008). Probably, the switch to herbivory determined the reorganisations in the morphology of the jaw muscle system and the proportions of the skull in tortoises (see Werneburg, 2011; Werneburg, 2012; Werneburg, 2013). By feeding predominantly on plants (immobile items), the tortoises may prolong the duration of the food uptake cycles and have more time to adjust their prehension kinematics to every single feeding situation.

We propose that the habitat preferences and the diet change in turtles are firmly correlated (e.g., Bels et al., 2008; Werneburg, 2014). The overall rigid design of the turtle ‘body plan’ hinder the animals to actively hunt for agile prey in terrestrial environments (King, 1996). We hypothesise that this statement is also valid for the terrestrial stem turtles and that these animals were predominantly herbivorous (sensu King, 1996). In aquatic turtles, in contrast, the buoyancy of the water overrides the shell-caused restrictions of mobility and also suction feeding can be applied for carnivorous feeding. Those advantages might have been the key factor for the turtles to become aquatic (and carnivorous) in the Jurassic. There may be several reasons for the secondary terrestriality in tortoises and presumably one of the main reasons was the inter- and intraspecific concurrence for food resources.