Influence of fossoriality on inner ear morphology: insights from caecilian amphibians

Abstract

It is widely accepted that a relationship exists between inner ear morphology and functional aspects of an animal's biology, such as locomotor behaviour. Animals that engage in agile and spatially complex behaviours possess semicircular canals that morphologically maximise sensitivity to correspondingly complex physical stimuli. Stemming from the prediction that fossorial tetrapods require a well-developed sense of spatial awareness, we investigate the hypothesis that fossoriality leads to inner ear morphology that is convergent with other spatially adept tetrapods. We apply morphometrics to otic capsule endocasts of 26 caecilian species to quantify aspects of inner ear shape, and compare these with a sample of frog and salamander species. Our results reveal caecilians (and also frogs) possess strongly curved canals, a feature in common with spatially adept species. However, significantly shorter canals in caecilians suggest reduced sensitivity, possibly associated with reduced reliance on vestibulo-ocular reflexes in this group of visually degenerate tetrapods. An elaboration of the sacculus of caecilians is interpreted as a unique adaptation among amphibians to increase sensitivity to substrate-borne vibrations transmitted through the head. This study represents the first quantitative analyses of inner ear morphology of limbless fossorial tetrapods, and identifies features within a new behavioural context that will contribute to our understanding of the biological consequences of physical stimuli on sensory function and associated morphological evolution.

Introduction

Locomotion is key to the survival of many organisms, and natural selection on the locomotor apparatus has led, at least in part, to the great diversity of organismal form (Dickinson et al. 2000). Correlated adaptation of sensory systems with the evolution of various locomotor behaviours has yielded important insights into the nature of biological responses to the presence of specific physical stimuli (Levine, 1979; Fay & Popper, 2000; Barton, 2006). One such system is the inner ear, where three roughly orthogonally oriented semicircular canals and their associated sensory epithelia sense positional information and angular acceleration as an organism moves.

Morphological aspects of the semicircular canals, including canal length, planar area enclosed by the canal and cross-sectional area of the canal, have been shown to affect their sensitivity (Ten Kate et al. 1970; Oman et al. 1987; Georgi, 2008). Sensitivity, and therefore canal morphology, is expected to reflect adaptations to the demands of specific types of movement an animal is exposed to. A relationship between canal morphology and locomotor behaviour has been demonstrated, wherein attributes that increase sensitivity are present in species considered agile or acrobatic, such as in the well-studied primate and bird groups (e.g. leaping, flying; Hadziselimovic & Savkovic, 1964; Spoor, 2003; Spoor et al. 2007; Walsh et al. 2009). There the canals are longer and more strongly arced with smaller cross-sectional areas in comparison to the canals of relatives considered to be slow or more spatially restricted locomotors. Investigation of behaviours considered spatially complex, such as swimming and arboreality (Georgi & Sipla, 2008; Boistel et al. 2010), have yielded similar patterns of correlation between canal morphology and locomotory behaviour. In these cases the presence of attributes correlated with increased sensitivity of the semicircular canals are considered beneficial as the potential for rotation during movement is great in these environmental contexts.

Fossorial behaviour, in general, refers to a subterranean lifestyle. Fossorial animals not only occupy, but often also navigate within three-dimensionally complex burrow systems. This exposes fossorial animals to stimuli predicted to resemble those experienced during other spatially complex behaviours (Gans, 1973), as movement within the subterranean environment can exploit all three spatial dimensions much like aerial, arboreal and aquatic behaviours do (Georgi & Sipla, 2008). Additionally, it has been suggested that an advanced sense of positional information is increasingly important in the absence or reduction of other sensory systems, such as vision, which typically accompany fossoriality (McVean, 1999). Studies of the inner ear morphology of fossorial mammals yielded results in support of this, where the canals were found to share morphological attributes in common with other spatially adept species (Lindenlaub et al. 1995; McVean, 1999), thus supporting the hypothesis that exposure to similar physical stimuli results in similar (i.e. convergent) sensory modification and morphology of associated structures (Fritzsch & Neary, 1998). However, comparative data from other fossorial taxa are currently sparse (but see Olori, 2010) and, therefore, it remains uncertain whether or not features of the inner ear previously observed in fossorial mammals, such as mole-rats and moles (Lindenlaub et al. 1995; McVean, 1999), represent general responses to fossorial behaviour, or if other independent aspects of their biology are involved.

In the current study we further investigate the potential influence of fossoriality on inner ear morphology by broadening taxonomic sampling to a group of fossorial amphibians: the caecilians. Caecilians are limbless amphibians that engage in active head-driven tunnel excavation and subterranean navigation (Ducey et al. 1993; Wake, 1993; O'Reilly et al. 1997). The goal of this study is to test the hypothesis that fossorial behaviour is correlated with adaptations of the inner ear that improve sensitivity to spatially complex movement. To do this, we quantitatively analyse the morphology of the inner ear of caecilians, and make comparisons with those of non-fossorial amphibians, frogs and salamanders, which exhibit drastically different locomotory behaviours and none of which are fossorial to the extent of caecilians. Particular attention is given to morphological attributes such as long and strongly curved semicircular canals, which are known to characterise the ears of other spatially adept species (Hadziselimovic & Savkovic, 1964; Spoor & Zonneveld, 1998; Witmer et al. 2008; Boistel et al. 2010). The quantitative analyses presented here represent the first of their kind for fossorial tetrapods, and the results bear implications for the role of physical stimuli in the evolution of sensory systems and associated morphological evolution.

Materials and methods

We sampled 49 species from the three orders of modern amphibians: 26 species of caecilians, 13 frogs and 10 salamanders. Our broad sampling of caecilians includes representatives from nine out of the 10 currently recognised family-level clades (Wilkinson et al. 2011). Frog and salamander species were sampled across the main lineages, taking into account ecological diversity. One specimen per species was sampled as either a dried skull or alcohol-preserved specimen from the collections of The Natural History Museum, London, and the Museum of Comparative Zoology, Cambridge, MA, USA, and other museums (Table S1). Specimens from all three orders were sampled within overlapping size ranges in order to minimise the effect of size-related differences, such as those caused by miniaturisation.

Amphibian species were examined using high-resolution x-ray micro-computed tomography, using machines at four institutions: Nikon Metris X-Tek HMX ST225 system at The Natural History Museum, London, UK; Skyscan 1173 and Scanco μCT35 systems at the University of Calgary, Calgary, AB, Canada; a Skyscan 1173 system in the Museum of Comparative Zoology, Cambridge, MA, USA; and an ACTIS scanner at University of Texas, Austin, TX, USA (see Table S1 for details). Using Amira® v.5 (VGS, Burlington, MA, USA), virtual endocasts of the otic capsules were extracted using the LabelField module to isolate the void space. The SurfaceGen and SurfaceView modules were then sequentially applied to the labelled data in order to generate the endocast used in subsequent morphological analyses (Fig. 1).

Fig. 1.

Volume rendering of the skull and left otic capsule endocast of the inner ear in a representative caecilian, frog and salamander species. Illustrated is the position of the inner ear in situ (a), and lateral (b), dorsal (c) and medial (d) views of the otic endocast for each species. Note that when the horizontal canal is orientated horizontally, the caecilian skull is oriented downwards compared with the flat frog and salamander skulls. Scale bar: 1 mm.

The use of otic capsule endocasts to evaluate morphological aspects of the soft tissues of the inner ear, albeit with limitations such as a slight overestimation of size (see Georgi & Sipla, 2008), is generally accepted for amniotes (Witmer et al. 2008; Walsh et al. 2009; Olori, 2010). However, agreement between otic capsule endocast morphology and the soft tissues within has not been established for amphibians. Validation of the use of endocasts as proxies for inner ear shape in amphibians was conducted here by comparing a virtual reconstruction of the membranous labyrinth derived from histological sections (aligned using the AlignSlices module in Amira) with an endocast derived from the same aligned histological dataset (Fig. S1). The endocast very closely approximates the size and shape of the enclosed soft tissues.

Otic capsule endocasts were measured in two- and three-dimensions using ImageJ v.1.45s (Schneider et al. 2012), Landmark Editor v. 3.6 (Institute for Data Analysis and Visualization, University of California, Davis, 2007) and 3D-Tool v.10 (3D-Tool GmbH, 2013). Because divergent skull and body form morphologies exist among modern amphibians, traditional proxies for head size, such as skull length and body mass measures, are not comparable among orders. Therefore, we took a measurement that is more conservative across taxa, the width between right and left otic capsule endocasts within each specimen, which was measured at the boundary zone between the pars superior (dorsal region including the semicircular canals) and pars inferior (ventral region including the sacculus). Gans (1974) suggested the lower limit of the width between otic capsules is constrained by functional aspects of the inner ear, meaning head size or body mass would be overestimated by this measure in the smallest taxa. The specimens measured in our sample do not appear to approach this lower limit, as a regression of skull length against otic capsule width accurately predicts skull length (i.e. smallest taxon, Idiocranium, falls directly on the regression line; data not shown).

To quantify the variation of the pars superior, six measurements were taken, determining the degree of curvature and the length of each of the anterior (ASC), posterior (PSC) and horizontal semicircular canals (HSC; Figs 1 and 2a). These two features are distinct and considered here separately, but they have often been compounded as a single measure of canal size (discussed below). Three-dimensional datasets offer greater accuracy in quantifying the size and shape of semicircular canals. Unfortunately, current methods using a central streamline approach (e.g. David et al. 2010; Gunz et al. 2012) cannot be applied to our dataset because definition of the medial surfaces of the canals is too often poor or completely lacking, and so precludes the establishment of centralised points (this problem is not restricted to amphibians but occurs also in other taxa, such as turtles; Georgi, 2008). Instead, we took all measurements from the outer surface of the endocast. How strongly arced a canal is (i.e. the degree of curvature) may be defined as the inverse of the geometric radius of curvature. We estimated the geometric radius of curvature for a circle defined by the arc of the canal, using a three-point system implemented in 3D-Tool v.10 (3D-Tool GmbH, 2013). The three points were placed along a curve at ends of each canal and a midway point on the canal surface before the ampulla (Fig. 2a). A more strongly curved canal defines a smaller circle, which will be represented by a lower radius of curvature value. This method is not without its own limitations, including the assumption that the canal forms part of a circle; however, we are able to achieve at least a rough estimate of degree of curvature of the canals using this approach. Note that this geometric radius of curvature differs from the similarly named measurement that is traditionally used in the inner ear literature (Jones & Spells, 1963; Spoor & Zonneveld, 1998; Spoor et al. 2002), which actually estimates semicircular canal size (as half the average of canal width and height) and therefore is often used as a proxy for canal length (Georgi, 2008).

Fig. 2.

An example endocast showing the definitions of the measurements. Clockwise from top left: lateral; planar view of horizontal canal (dorsal); planar view of posterior canal; and planar view of anterior canal (a). Abbreviations as follows: ASCl, anterior semicircular canal length; ASCr, anterior semicircular canal radius; HSCl, horizontal semicircular canal length; HSCr, horizontal semicircular canal radius; PIh, pars inferior height; PIw, pars inferior width; PSCl, posterior semicircular canal length; PSCr, posterior semicircular canal radius; see Materials and methods for details. Bivariate double logarithmic plots between log₁₀ head size (the width between the inner ear) and three noteworthy measurements: average semicircular canal length (b); anterior semicircular canal radius (c); and pars inferior height (d). Caecilian (black line and points), frog (dashed line and white points) and salamander (grey line and points).

Here we measured the length of the ASC, PSC and HSC directly using the summed Euclidean distance between 10 landmark points placed along the external surface of the canal, implemented in Landmark Editor v. 3.6. Because of the problem of resolving the end of the canal and the beginning of the ampulla, canal length measures taken here include the ampulla. As such, all length measures are slight overestimates of actual canal length, but are universally so within our sample. All measures were taken on the left endocast twice and an average was used for subsequent analyses.

The amphibian inner ear is morphologically very different from that of mammals or reptiles, and therefore there are no standard measurements for the pars inferior, which mainly comprises the sacculus (Fig. S1). The size of the pars inferior was characterised using height and length measured from the lateral plane in two-dimensions (Fig. 2a): width was measured at the widest point, parallel to the horizontal canal; height was measured from the boundary zone between the pars superior and inferior, below the horizontal canal, and the most ventral point on the pars inferior.

Quantitative comparisons are restricted among the amphibians, as quantitative comparison beyond was not considered a viable option due to the drastically different patterns of scaling among distantly related species and unique modifications to various organs independently acquired in many lineages (Fritzsch & Beisel, 2001; Spoor et al. 2002). As such, comparison with distantly related species is largely qualitative and derived from the literature.

We compared the measures of inner ear morphology among caecilians, frogs and salamanders using analysis of covariance (ancova). This analysis is used to test differences among groups for a given measured trait while controlling for the effect of size. The null hypothesis states that the measured trait scales with head size in the same way across all taxa, as shown by regression slopes that have the same slope and intercept. We expect a significant size term for all traits because of allometry. A significant ‘group’ term indicates the regression lines among groups differ in intercept, suggesting group-specific allometry but shared scaling. A significant ‘interaction’ term (group*size) indicates the regression lines among groups differ in slope and intercept, suggesting a shift in allometric scaling among groups. ancovas were performed in JMP Pro v.10 (SAS institute, Cary NC, 1989–2012).

Results

Among species of the three orders of amphibians, there exist clear morphological differences in the curvature and length of the semicircular canals, as well as the size of the pars inferior (Figs 4). In general, the pars superior appears widest with the longest canals in salamanders, and narrowest with the shortest canals in caecilians (frogs are intermediate). The canals of frogs appear the most highly curved, and those of salamanders the most oblong (Figs 3 and 4). Many caecilians also appear to have strongly curved canals; however, they look much shorter than highly curved canals of frogs. The sacculus appears larger and more bulbous in caecilians than in either frogs or salamanders – the pars inferior of frogs looks large, but this is because the lateral perilymphatic chamber is enlarged and extends more ventrally, whereas the extent of the sacculus (visible on the medial surface) is much smaller.

Fig. 4.

Left otic capsule endocasts of the frog and salamander species used in this study, arranged in phylogenetic order. Salamanders at the top (black branches) and frogs below (grey branches). The topology for visualisation purposes is based upon Pyron & Wiens (2011). Endocasts are not shown to scale.

Fig. 3.

Left otic capsule endocasts of the caecilian species used in this study, arranged in phylogenetic order. The topology for visualisation purposes is based upon Zhang & Wake (2009), supplemented with Wilkinson & Nussbaum (1999) and Maddin et al. (2012). Endocasts are not shown to scale.

With the quantitative comparison of eight variables of inner ear morphology among the three orders (Fig. 2a) using an analysis of covariance (ancova), we find seven measurements differ significantly among orders (Table 1; Fig. 2). All ear measures covary significantly with head size (Table 1). There were no significant differences in these slopes between orders for any of the seven variables, so the interaction term was not included in the model (Table S2).

Table 1.

Results of analysis of covariance (ancova) for each of the eight measurements of otic shape

	Head size		Group
Variable	F (df = 1)	P-value	F (df = 2)	P-value
Pars inferior width	81.08	< 0.0001	4.36	0.0186
Pars inferior height	116.52	< 0.0001	13.38	< 0.0001
Horizontal semicircular canal radius	81.09	< 0.0001	16.01	< 0.0001
Anterior semicircular canal radius	116.75	< 0.0001	16.09	< 0.0001
Posterior semicircular canal radius	85.18	< 0.0001	2.04	0.1420
Horizontal semicircular canal length	100.72	< 0.0001	22.57	< 0.0001
Anterior semicircular canal length	88.28	< 0.0001	39.92	< 0.0001
Posterior semicircular canal length	92.76	< 0.0001	29.09	< 0.0001

The interaction term was not significant for any measurement; therefore, it was not included in the model. For partial regression coefficients, see Table S2. Significant values in bold. The test statistic (F) is given with the degrees of freedom (df) and significance test value (P).

Two of the three semicircular canals are significantly different in the degree of curvature among caecilians, frogs and salamanders (Table 1). The HSC and ASC have a smaller radius of curvature in caecilians and frogs compared with salamanders, and therefore are more strongly curved (e.g. Fig. 2c). Post hoc Tukey tests for the HSC and ASC showed that the regression slopes of frogs and caecilians are not significantly different at the 0.05 level. The PSC does not differ in degree of curvature among the orders.

Semicircular canal length is significantly different among the orders (Table 1). Caecilians have proportionally shorter canals compared with frogs and salamanders. Because each canal presented the same pattern, the average length is presented in Fig. 2b. The length of all three canals is significantly shorter in caecilians, while frogs and salamanders are not different from each other (post hoc Tukey tests, at 0.05 level).

The pars inferior, encompassing the sacculus, is significantly different in both width and height among caecilians, frogs and salamanders (Fig. 2d; Table 1). All three groups differ in pars inferior height, with caecilians having proportionally the tallest, frogs the shortest, and salamanders intermediate (Fig. 2d). In width, the salamanders are proportionally narrower compared with frogs and caecilians, yet caecilians and frogs are not statistically different (post hoc Tukey test at 0.05 level).

Discussion

Morphological variation of the inner ear, particularly the length and degree of curvature of the semicircular canals, can inform of the mode of locomotion and spatial behaviour of vertebrates (e.g. Hadziselimovic & Savkovic, 1964; Spoor et al. 2002; Alonso et al. 2004). Our results reveal several important differences in the morphology of the inner ear of caecilians, a group of limbless and predominantly fossorial vertebrates, compared with those of frogs and salamanders. Caecilians possess much shorter semicircular canals, but they are equally as curved as those of frogs (which are also much longer). Caecilians also possess a substantially larger pars inferior, which houses the sacculus. These features are discussed below as they relate to aspects of fossorial locomotion and associated adaptations.

Size and shape of semicircular canals in a group of fossorial amphibians

Canal length is one of the most-cited aspects of the semicircular canals thought to correlate with behaviour (Jones & Spells, 1963; Spoor & Zonneveld, 1998; Spoor, 2003; Spoor et al. 2007; Walker et al. 2008). Previous studies have found canal length to be positively correlated with canal sensitivity, and it has been hypothesised that animals that engage in agile behaviour benefit from increased sensitivity, and therefore possess longer canals (Spoor, 2003; Spoor et al. 2007; Walker et al. 2008; Ryan et al. 2012; but see Jones & Spells, 1963 for an alternative interpretation). Interestingly, animals that are not considered particularly agile but instead engage in spatially complex behaviour possess similar attributes of canal morphology (e.g. arboreal vs. terrestrial chameleons; Boistel et al. 2010). This suggests greater length provides increased sensitivity to spatial perception in addition to resolving rapid movement (McVean, 1999; Georgi, 2008).

We hypothesised, based on predictions of the types of spatial stimuli experienced during tunnel navigation, and from previous studies on fossorial mammals, that caecilians would possess long canals as an adaptation to increase sensitivity to spatial information. Our results, however, are at direct odds with this hypothesis and conflict with previous interpretations for fossorial mammals (Lindenlaub et al. 1995; McVean, 1999). Caecilians have very short semicircular canals, which consequently positions the ampullae of the ASC and PSC in a more dorsal location than those of frogs or salamanders. The significantly shorter canals of caecilian inner ears imply reduced sensitivity in comparison to those of frogs and salamanders. However, this inference is confounded by our finding that caecilian semicircular canals are, despite their length, relatively highly curved.

Several studies linking canal morphology to behaviour focused on the degree of curvature as an important predictor of locomotor ability (Gray, 1908; Hadziselimovic & Savkovic, 1964; Alonso et al. 2004). Species considered acrobatic or spatially adept, such as aerial species (Hadziselimovic & Savkovic, 1964) and those that leap (Matano et al. 1986) and climb (Boistel et al. 2010), tend to possess more strongly curved canals. However, it is unclear if a more highly-curved canal improves sensitivity itself, or if a more highly-curved canal is simply a byproduct of increasing canal length in these species. One could predict that a strongly curved canal (i.e. circular) is more sensitive than a flatter, more oblong canal as a circular canal traverses all degrees of space of its planar axis more evenly and is therefore sensitive to a broader range of angular movements. This is supported by the observation that species considered highly spatially perceptive possess canals that more closely approach circularity, in addition to being long, rather than being ovoid and long. Additionally, sensitivity has been shown to correlate with the planar area enclosed by the canal, whereby sensitivity decreases with smaller areas, which is caused by deviation from circularity (i.e. increasingly oblong; McVean, 1999).

The degree of curvature of the canals of caecilians is similar to that of frogs (small radii of curvature), and in contrast to that of salamanders (large radii of curvature). This parallels the results of comparisons made previously between spatially adept and slow-moving or ground-dwelling members of groups like primates and birds (Hadziselimovic & Savkovic, 1964; Spoor et al. 2007). It is uncertain if the highly curved canals of caecilians actually enclose a greater planar area or if such a small portion of the curve is present that the area is actually quite small. If the former, the highly curved canals may represent an adaptation towards maximising sensitivity to spatial perception for canals of a certain length. Similarity with frogs would be consistent in this scenario because an adaptation to increased spatial perception is also expected for frogs due to their rapid and spatially complex hopping behaviour (Fritzsch & Neary, 1998). If the latter, sensitivity is not actually increased even though curvature is. The result that the semicircular canals of caecilians are significantly shorter than those of frogs and salamanders supports this latter scenario. However, in the absence of a quantitative analysis of form (reasons discussed below), it is difficult to assess whether or not the aspect of the curve relevant to canal sensitivity is represented by the actual canal, and not just by the circle that it is defining.

Taken together, it appears as though aspects of the semicircular canals of caecilians associated with increasing sensitivity to complex movement are poorly developed in comparison to frogs and salamanders. This suggests that the canals of caecilians may actually be less well adapted to 3D spatial perception than are those of frogs and salamanders. This points to a possible degenerate condition, and one or more potential causes may be involved.

A major driver of increased canal sensitivity in agile and spatially adept species is thought to be an increased reliance on gaze stabilisation (Spoor et al. 2007; Witmer et al. 2008). Stabilisation of gaze is achieved through compensatory movements of the extraocular muscles, and these vestibulo-ocular reflexes are driven by signals from the canals. It has been shown that animals that rely more heavily on gaze stabilisation possess features of the canals that improve sensitivity (i.e. long, strongly curved canals; Spoor et al. 2007; Witmer et al. 2008). The drastically reduced visual system of caecilians may be related to a reduction in canal sensitivity in the group.

Visual reduction has been interpreted as having very different consequences on canal morphology and sensitivity. On the one hand visual reduction was interpreted as leading to increased sensitivity in fossorial rodents (Lindenlaub et al. 1995; McVean, 1999). On the other hand, in whales visual reduction was suggested to have driven decreased canal sensitivity as a functional adjustment to prevent overstimulation during spatially complex movement where reliance on gaze stabilisation is minimal (Spoor et al. 2002). It is fair to note that comparing rodents with whales makes a large number of generalisations, which can easily lead to conflicting observations. It seems, however, that even attempting to find features that universally describe sensory responses within a single behavioural category may be equally challenging, as evidenced by the results obtained here for fossorial caecilians that conflict with those for fossorial mammals.

Animals within a single behaviour category, such as fossoriality, likely interact with the environment in dramatically different ways and therefore experience different pressures on sensory system sensitivity. For example, fossorial mammals may be much more active than caecilians, and so may experience different demands on the sensory apparatus. This would in turn lead to different morphologies of the semicircular canals, like the pattern observed by the comparisons made here. The same may be true of other behaviour categories, such as arboreal, aquatic and aerial behaviours, further complicating attempts to find universal patterns to describe what may in fact be unique combinations of factors specific to each taxon within a category.

As quantitative studies broaden in taxonomic scope beyond mammals and birds, it is becoming increasingly apparent that the form of the semicircular canals is far more complex than traditional measures can accommodate. For example, how the canal deviates from a circular outline and how the canal twists and bends from planarity likely bears a critical influence on the sensitivity and perceptual capabilities of the inner ear. Landmark-based shape analysis has attempted to capture this variation (see Georgi, 2008; Bradshaw et al. 2010; Gunz et al. 2012); however, limitations still remain, such as identifying homologous landmarks on the endocasts. Studies of the geometric form of the semicircular canals are in their infancy (Georgi, 2008; Bradshaw et al. 2010; Gunz et al. 2012) compared with the decades of studies using methods applying assumptions of circularity and linear approximations (e.g. Jones & Spells, 1963). As a result, comparisons between taxa using complementary, updated methods and measures are currently limited. Studies implementing these modern techniques are revealing aspects of canal form not previously considered that may be important predictors of behaviour (e.g. Boistel et al. 2011; Malinzak et al. 2012). Therefore, it is hoped that as these methods are further developed, some of the limitations identified here (i.e. lack of definition, preclusion of central streamlines) are considered so that broader comparisons across taxa can be made. This will enable a better understanding of the relationship between complex aspects of semicircular canal morphology and locomotory behaviour.

Unique adaptations to fossoriality in the inner ear of caecilians

Of the three amphibian groups examined, the size of the pars inferior was found to be tallest in caecilians, corroborating earlier qualitative observations (Retzius, 1881; Sarasin & Sarasin, 1890; Wever, 1985; Fritzsch & Wake, 1988). From histological sections this region can be seen to be dominated by the sacculus (Fig. S1). This endolymph-filled sac houses the saccular macula, which is also particularly large in caecilians (Wever, 1985; Fritzsch & Wake, 1988). In frogs and salamanders, for which the responsiveness of inner ear maculae has been investigated, the saccular macula has been shown to play a major role in the detection of substrate-borne vibrations (Capranica & Moffat, 1974; Ross & Smith, 1980, 1982; Koyama et al. 1982; Lewis & Narins, 1985; Lewis & Lombard, 1988). Its large otoconial mass is thought to behave as an inertial sensor, and it is oriented to maximise movement in the dorso-ventral direction (Ross & Smith, 1979; Mason, 2007; Georgi, 2008). The limbless condition of caecilians means that the head is almost always in contact with the substrate, and enlargement of the sacculus may represent an adaptation to maximise sensitivity to an increased vibratory component of their sensory spectrum. Whole-body anaesthetised salamanders placed on their backs can perceive substrate-borne vibrations, presumably via transmission through the skull (Ross & Smith, 1979), and the results of earlier studies of inner ear responses in caecilians showed that sensitivity is limited to low-frequency air-borne sound, with the most sensitivity reported when vibrations are applied directly to the head (Wever, 1975).

The mode of sensory perception in caecilians may also help explain why drastic enlargement of the sacculus took place. In some limbless tetrapods in which vibrations are an important source of information, such as in snakes, the integument contains specialised nerve endings that serve in mechanoreception (Proske, 1969). Caecilians appear to lack such mechanosensory neuromasts as adults (but note their presence in larvae; Fritzsch & Wake, 1986), and elaboration of the sacculus may represent an alternative strategy to increase sensitivity to vibration stimuli. In other taxa, such as burrowing snakes, similar enlargement of the saccular macula has been observed (Olori, 2010), suggesting a potentially convergent strategy to improve vibration sensation in these distantly related taxa.

This hypothesis is further corroborated by the relatively smaller size of the sacculus in the secondarily aquatic taxa Typhlonectes and Potomotyphlus, where contact with the substrate is reduced or minimal in the aquatic realm (although these taxa do exhibit some head-first burrowing behaviour into the wet sediment; Moodie, 1978). The role of vibrations in the activities caecilian engage in, such as prey capture, is not adequately understood at this time, but may have been an important selective pressure to increase sensitivity to vibrations in the dark, subterranean environment. Together these factors suggest a head–substrate transmission pathway may be an important sensory pathway not previously identified in caecilians. A similar transmission pathway is established with specialised anatomical modifications in amphisbaenian squamates. There, an expanded extracolumella of the middle ear serves to transmit vibrations collected in the anterior part of the head to the inner ear (Gans & Wever, 1972). Empirical testing of saccular sensitivity in caecilians is needed to confirm or refute this hypothesis.

Evolution of inner ear traits within caecilians

The effects of phylogeny on the evolution of traits associated with fossoriality needs consideration. Because all caecilians are fossorial, it is difficult to distinguish between the presence of morphological features due to adaptation to fossoriality and those that are present simply due to common ancestry. Potential variation within the group supports the former, and is discussed here.

Inspection of the morphology of the inner ear endocasts, within a phylogenetic context, reveals potential patterns in the evolution of inner ears within the caecilian clade (Fig. 3). The morphology of the pars superior of early diverging caecilians (i.e. rhinatrematids) more closely aligns with that of salamanders (Fig. 4). These caecilian species possess longer, more oblong canals in comparison to later diverging taxa. Interestingly, this pattern of morphological evolution parallels a trend in behavioural evolution. Early diverging caecilian species are considered to be predominantly leaf litter inhabitants, rather than deep-tunnel dwellers (e.g. they are encountered in pitfall traps; Gower et al. 2010). These species possess the most well-developed visual systems, with exposed eyes and maintenance of visual innervation (Wake, 1985; Fritzsch & Wake, 1988). Subsequently diverging caecilian species exhibit a trend towards increased tunnel excavation and subterranean life. Reduction of the visual system takes place, such that eyes are frequently covered by skin or bone, and degradation of visual innervation is common (Wake, 1985; Fritzsch & Wake, 1988). Based on these observations we can, therefore, hypothesise that as the degree of fossoriality increases, so too do the features that were found here to characterise inner ear shape for the group. Unfortunately, adequate behavioural data do not exist for each species sampled here, precluding the formal testing of this hypothesis.

In contrast, the morphology of the pars inferior of virtually all caecilians – even the earliest diverging species – is found to significantly diverge from the morphology seen in frogs and salamanders (Fig. 2d). We suggested above that expansion of the sacculus may be correlated with a head–substrate vibration transmission pathway associated with the loss of limbs. All caecilians, even the earliest diverging members, differ from frogs and salamanders in completely lacking limbs. It is therefore likely that pressure to expand this organ took place on the caecilian stem, while limb reduction was taking place. This is a hypothesis that could be tested if adequately preserved fossils existed. These data are also not available at this time (Jenkins et al. 2007; Maddin et al. 2012).

Conclusion

Organismal responses to physical stimuli experienced during locomotion have provided important insights into sensory function, structure and morphological evolution. We interpret a different morphological, and thus sensory, response of the inner ear to the conditions imposed by fossoriality in caecilian amphibians in comparison to interpretations made for fossorial mammals. These results suggest that establishment of universal sensory responses within categories of behaviour may be oversimplifications and more challenging than traditionally thought. Two aspects of canal morphology previously considered correlated (length and curvature) are found to vary independently in caecilians. We also observe a potentially novel sensory response to fossoriality: elaboration of the sacculus, likely correlated with increased sensitivity to substrate-borne vibrations through a head–substrate transmission pathway. Together our study generates a new suite of possible morphological correlates of a widespread behaviour, fossoriality, that can be used to understand sensory system responses to movement within the physical environment. Our observations represent a first survey of inner ear morphology in a large number and diversity of amphibian taxa. Further work to parse out the significance of variation among and within all three orders of amphibians should be the subject of future study.

Author contributions

Both authors contributed equally to this work. Both authors designed the project, acquired and interpreted the data, and wrote the manuscript.

Supporting Information

Additional Supporting Information may be found in the online version of this article: