• Cyril Charles,

    Corresponding author
    1. Team Evo-Devo of Vertebrate Dentition, Institut de Génomique Fonctionnelle de Lyon, Université de Lyon, CNRS UMR 5242, UCBL 1, Ecole Normale Supérieure de Lyon, 46 Allée d’Italie, 69364 Lyon Cedex 07, France
    Search for more papers by this author
    • These authors contributed equally to the study.

  • Floréal Solé,

    1. Team Evo-Devo of Vertebrate Dentition, Institut de Génomique Fonctionnelle de Lyon, Université de Lyon, CNRS UMR 5242, UCBL 1, Ecole Normale Supérieure de Lyon, 46 Allée d’Italie, 69364 Lyon Cedex 07, France
    Current affiliation:
    1. Royal Belgian Institute of Natural Sciences, Direction Earth and History of Life Vautierstraat 29, 1000 Brussels, Belgium
    Search for more papers by this author
    • These authors contributed equally to the study.

  • Helder Gomes Rodrigues,

    1. Team Evo-Devo of Vertebrate Dentition, Institut de Génomique Fonctionnelle de Lyon, Université de Lyon, CNRS UMR 5242, UCBL 1, Ecole Normale Supérieure de Lyon, 46 Allée d’Italie, 69364 Lyon Cedex 07, France
    Search for more papers by this author
  • Laurent Viriot

    1. Team Evo-Devo of Vertebrate Dentition, Institut de Génomique Fonctionnelle de Lyon, Université de Lyon, CNRS UMR 5242, UCBL 1, Ecole Normale Supérieure de Lyon, 46 Allée d’Italie, 69364 Lyon Cedex 07, France
    Search for more papers by this author


The extant mammals have evolved highly diversified diets associated with many specialized morphologies. Two rare diets, termitophagy and vermivory, are characterized by unusual morphological and dental adaptations that have evolved independently in several clades. Termitophagy is known to be associated with increases in tooth number, crown simplification, enamel loss, and the appearance of intermolar diastemata. We observed similar modifications at the species level in vermivorous clades, although interestingly the vermivorous mammals lack secondarily derived tools that compensate for the dentition's reduced function. We argue that the parallel dental changes in these specialists are the result of relaxed selection on occlusal functions of the dentition, which allow a parallel cascade of changes to occur independently in each clade. Comparison of the phenotypes of Rhynchomys, a vermivorous rat, and strains of mice whose ectodysplasin (EDA) pathway has been mutated revealed several shared dental features. Our results point to the likely involvement of this genetic pathway in the rapid, parallel morphological specializations in termitophagous and vermivorous species. We show that diets or feeding mechanisms in other mammals that are linked to decreased reliance on complex can lead to similar cascades of change.

Over the course of their evolution, mammals acquired an efficient masticatory apparatus that permitted them to strongly diversify their feeding habits (Luo et al. 2001, 2003; Rich et al. 2001). This diversification took place in close relation with modifications of the dentition such as a decrease in the tooth number and an increase in tooth shape complexity (Koussoulakou et al. 2009 and references therein). These general trends of dental evolution led to the acquisition of similar characters in distant species among mammals. Among them, two different diet specializations, vermivory (“worm-eating”) and termitophagy (“termite-eating” or “ant-eating”), are associated with similar morphological traits in mammals, such as a highly reduced dentition. The characteristics and the acquisition of these traits during evolution remain to be detailed, and the genetic bases of their peculiar morphologies have yet to be elucidated.

Within the mammalian feeding habits, termitophagy (termite or ant-eating) and vermivory (worm-eating) are diets scarcely recorded. Termitophageous mammals are known since the late Jurassic (Fruitafossor windscheffeli, Luo and Wible 2005) and their extant counterparts spread over five extant groups: the Monotremata, the Marsupialia, and the placentalian Xenarthra, Afrotheria, and Laurasiatheria (Fig. 1; taxa labeled in black). Several termitophageous mammals are edentulous, but they acquired in return various original traits such as beaks or protractible and spiny tongues (Davit-Béal et al. 2009).

Figure 1.

Phylogeny of the mammalian groups with indication of the phylogenetic position of the termitophageous (black labeled) and vermivorous (red labeled) taxa discussed in this article. Major lineages are colored following Bininda-Emonds et al. (2007): black, Monotremata; orange, Marsupialia; blue, Afrotheria; yellow, Xenarthra; green, Laurasiatheria; and red, Euarchontoglires.

Far less documented are the adaptations to vermivory, notably because very few mammals exclusively feed on worms and larvae. We study here six taxa considered to be strictly or mostly vermivorous (Fig. 1; taxa labeled in red). Four of them belong to the Rodentia (the Philippine shrew rat Rhynchomys, the Sulawesi shrew rats Tateomys and Paucidentomys, and the New-Guinean moss-mice Pseudohydromys), one to the Afrosoricida (the streaked tenrec Hemicentetes), and one to the Carnivora (the falanouc Eupleres). The four vermivorous rodents all belong to the Murinae subfamily and they are Southeast-Asian insular species.

Comparative dental anatomy reveals that vermivorous and termitophageous mammals display many similar dental traits. The aim of this study was to precisely identify the particular dental characters associated to termitophageous and vermivorous diets. From this framework and focusing on the peculiar case of Rhynchomys, we tried to understand how such traits can be acquired during evolution, and decipher potential genetic modification associated with these peculiar diet specializations.

Material and Methods


Despite their similar vernacular names (shrew rat), Rhynchomys, Paucidentomys, and Tateomys are not closely related. Rhynchomys is a medium-sized shrew rat endemic to the Island of Luzon, Philippines (Musser and Freeman 1981). Rhynchomys is a ground-dwelling rodent close to the genera Archboldomys, Chrotomys, and Apomys (Fig. 1), and it differs from these related genera by a diet mostly based on earthworms and soft insects (Balete et al. 2007). Indeed, it is considered that Rhynchomys forage over the forest floor, looking at the soil surface for worms and soft-bodied invertebrates (Balete et al. 2007). Paucidentomys is endemic to Sulawesi Island (Indonesia) and is considered as a strict vermivorous rodent (Esselstyn et al. 2012). Tateomys is also endemic to Sulawesi Island; this genus encompasses strictly vermivorous species which possess digging claws used for extracting worms from moss, soil, and rotten wood (Musser, 1982). Its closest relative genus, Melasmothrix, also encompasses worm-eating species, but part of their diet includes adult dipterans (Musser, 1982). Tateomys is closely related to Rattus and referred to the Tribe Rattini (Lecompte et al. 2008) (Fig. 1). Pseudohydromys, Neohydromys, Mayermys, Microhydromys, and Mirzamys are known as “moss-mice” (Fig. 1). They belong to the tribe Hydromyini. The Hydromyini are all endemic to New Guinea, except the false water rat Xeromys. Helgen and Helgen (2009) distinguished five groups among Pseudohydromys. The most intriguing is the ellermani species group. This group contains Pseudohydromys ellermani, P. germane, P. carlae, and P. pumehanae. Among these species, the diet of P. ellermani is probably similar to that of the Rhynchomys (Helgen and Helgen 2009). The two other vermivorous mammals studied here are Eupleres and Hemicentetes (Fig. 1); they are endemic to Madagascar. The genus Eupleres is monospecific (E. goudotii), whereas Hemicentetes encompasses two species: H. nigriceps and H. semispinosus. The Sumatran Hog-badger (Arctonyx hoeveni), a vermivorous species belonging to Carnivora, also feeds on ants but the relative proportion of ants in the diet is still unclear (Helgen et al. 2008).


We analyzed the dentition of the aardvark, aardwolf, sloth bear, pangolins, anteaters, armadillos, numbats, and echidnas (Fig. 1). Specimens came from the collection of the National Museum of Natural History of Paris (MNHN) and from the collections of the University of Lyon Dental School. Table 1 provides information concerning systematics and distribution of those taxa.

Table 1. General data concerning the termitophageous and vermivorous mammals discussed in this study


Molars are symbolized by Mn and Mn, respectively, referring to the nth upper and nth lower molar tooth, and Mn for both. The murine postcanine dentition comprises three molar teeth (i.e., M1-M3 on lower rows, M1-M3 on upper rows). A supplementary upper tooth located mesial to the M1 occasionally occurs. This tooth was named P4? in reference to its position which corresponds to the only known premolar in the oldest known muroids (Gomes Rodrigues et al. 2011). Homology of structures observed in studied murines has been established following Miller (1912) for the upper molars and Michaux (1971) for the lower molars (Fig. 2).

Figure 2.

(A and B): Terminology used to name the main cusps of upper molars (A) and lower molars (B). The molars here illustrated are from wild-type mice. (C and D): Illustration of the main cusps of upper (C) and lower (D) molars of Rhynchomys isarogensis (FMNH 183555; reversed views).


Rhynchomys specimens came from the Field Museum of Natural History (FMNH), Chicago, and from the Smithsonian National Museum of Natural History (USNM), Washington. Measurements of maximum length and width of Rhynchomys molars were performed using ImageJ. Potential modifications of the M1 occlusal morphology in case of supernumerary tooth development were evaluated using the ratio (length of t2)/(length of M1). A low ratio indicates that the mesial part of the M1, constituted by the t2, is short relatively to the length of the tooth.


Dentitions associated with termitophageous and vermivorous diets display several morphological trends, such as drastic modifications in the tooth number (hypodontia or hyperdontia), crown shape simplification, acquisition of dental ever-growth, intermolar diastemata, and enamel loss (Table 2 and Fig. 3). Hereafter, we detail these trends for both diet specializations.

Table 2. Summary of the features discussed for the termitophageous and vermivorous mammals. In black: termitophageous taxa; in red: vermivorous taxa. The mutant mice are included for comparison
Figure 3.

Summary of the dental features discussed for the termitophageous and vermivorous taxa. Because the cheek teeth of the sloth bear are poorly modified, its dental evolution is not presented in this figure. See text for discussions concerning the evolutionary trends. Gray lines: features unrelated to termitophageous or vermivorous adaptations; black lines: features related to termitophageous adaptations; red lines: features related to vermivorous adaptations.


Decrease in tooth number

The adaptation to termitophagy is often related to hypodontia, leading in extreme cases to the total loss of teeth (Fig. 3). Compared to other Ursidae, sloth bears (genus Melursus) lack the two first upper incisors (Erdbrink 1953). Aardwolf dentition significantly differs from that of other hyenas as it only has two to four postcanines, but it retains three incisors and one canine per quadrant. Aardvarks (genus Orycteropus) possess neither incisors nor canines, but two premolars and three molars remain per jaw quadrant. Extreme cases of decrease in tooth number are found in echidnas, pangolins, and anteaters, which are toothless. The fossil record concerning echidnas and anteaters does not bring any information on the evolutionary dynamics of dental reduction because oldest known fossils already are toothless. On the contrary, evolutionary trends in pangolins are better illustrated within the Palaeanodonta fossil record. Indeed, the Palaeanodonta are considered as possible ancestors of present pangolins (Rose et al. 2005). Escavadodon, considered as the oldest known Paleocene palaeanodont (Rose and Lucas 2000), appears to have an almost complete dentition. As most primitive eutherians, it appears to have four premolars and three molars. During the late Paleocene and Eocene, the evolution of palaeanodont dentition displays marked reduction of the number of teeth (Rose et al. 2005; Rose 2008), so that postcanine dentition of Eocene palaeanodonts only retains one to five postcanine teeth. Thus, the possible lineage that links Eocene palaeanodonts to the modern pangolins is characterized by a long and progressive reduction of the entire dentition.

The study of the evolution of the dentition of the termitophageous mammals thus permits to hypothesize that a progressive decrease of the cheek teeth (premolars and molars) occurred before its complete loss.

Occurrence of supernumerary teeth

Some termitophageous mammals, such as numbats and armadillos, have more teeth than basal mammals (Fig. 3). Although the maximal tooth number for marsupials is 50 including four molar teeth, numbats can possess up to five upper and six lower molars. The number of teeth on opposite sides of the same jaw may differ, suggesting a low canalization of this character. Several armadillos also have supernumerary teeth. The only tooth-replacing xenarthran, Dasypus (an armadillo), presents a multiplication of premolars (up to six or seven) and the occasional development of a supernumerary tooth in front of the tooth row, both on the upper and lower dentitions (Ciancio et al. 2012). The giant armadillo can have a dentition composed of 100 teeth. The homology of supernumerary teeth remains problematic for both numbats and giant armadillos (Dependorf 1898; Bensley 1903; Tate 1951).

The presence of supernumerary teeth in numbats and armadillos shows that termitophagy can be related to a decrease or an increase in the number of teeth. In sum, termitophageous diet appears to be related to important variations in the number of teeth.

Crown shape simplification and reduction in number of roots

Mammals usually possess postcanine teeth with complex crown shapes and multirooted molars. Evolution of pangolins illustrates the simplification of the postcanine crown shape (Fig. 3). The oldest palaeanodont, Escavadodon, has four molariform premolars and three molars, with multiple roots, recalling the primitive eutherians. Slightly younger palaeanodonts from the late Paleocene maintain seven postcanine teeth, but the shape of the crowns becomes highly simplified. In Eocene palaeanodonts, the remaining postcanines are generally of cylindrical (peg-like) or elliptical shape and they are single rooted.

With the exception of the anteater, which is toothless, the remaining xenarthrans (sloths, armadillos) have peg-like and single-rooted postcanines. Similar peg-like and single-rooted postcanine teeth also characterize the dentition of aardwolves. Numbats are characterized by the presence of minute and simple-shaped teeth. However, numbat postcanine teeth are still more complex than those of palaeanodonts and xenarthrans, they are small, transversely compressed, and cusps are aligned mesio-distally on a crown that evokes that of the Eutriconodonta (i.e., group of early mammals), which mostly lived during the Mesozoic. Ursidae have complex premolars and molars in relation with their omnivorous diet. Postcanine teeth of the sloth bear are complex compared to other termitophageous mammals, but they are smaller than in other ursids. Dented termitophageous mammals differ from their ancestors in that they possess less complex cheek teeth, these being eutriconodont-like in numbats and peg-like in armadillos, aardwolves, and aardvarks. Interrelated with the shape simplification, the number of roots also tends to reduce. Moreover, the evolution of the Palaeanodonta allows us to hypothesize that both crown simplification and reduction in the number of root number occurred before the complete loss of the cheek teeth in the Pholidota evolution. Similar trends were probably set up during the evolution of echidnas and anteaters.

Presence of intermolar diastemata

Another feature shared by several termitophageous mammals is the presence of important interdental spacing between postcanine teeth, termed intermolar diastemata (Fig. 3). Eutriconodont-like teeth of numbats are separated by long intermolar diastemata. Although smaller than those of numbats, aardwolves and aardvarks also have intermolar diastemata separating their remaining teeth. The evolution of the palaeanodonts provides clear evidence that intermolar diastemata appeared together with the shape simplification of the postcanines (Fig. 3).

Enamel loss and dental ever-growth

Enamel loss and dental ever-growing often occur together in termitophageous extant species. Aardvarks, sloths, and many armadillos have ever-growing postcanines lacking enamel. Conversely to the Tubulidentata (aardvarks), the fossil record of the Xenarthra gives clues on evolutionary trends in postcanine teeth toward enamel loss and acquisition of ever-growth. The fossil record supports the hypothesis of multiple loss of enamel among Xenarthra and indicates that the loss of enamel clearly predates dental ever-growth (Rose 2006). Compared to strict termitophageous species, retaining cheek teeth is functionally important for armadillos because they possess a more omnivorous diet than edentulous species. Aardvark teeth have peculiar fused dentine columns surrounded by cement. Their ever-growing postcanines probably appeared secondarily, as in Xenarthra, but no fossil documents this transition.


As for the termitophageous mammals, several dental characters are shared by the four vermivorous rodents (Rhynchomys, Tateomys, Paucidentomys, and Pseudohydromys), the falanouc, and the streaked tenrec (Fig. 3). These features have been acquired independently (Fig. 1).

Decrease in tooth number

Murine rodents all lack premolars, but they usually conserve three molars in each dental quadrant. Conversely to the other murine rodents, Rhynchomys is distinguished by the loss of the third lower and upper molars, Pseudohydromys has only one molar in each jaw quadrant, and Paucidentomys has no molar at all (Fig. 3). The dental formulae of Rhynchomys, Pseudohydromys, and Paucidentomys, thus, exemplify decreases in tooth number.

Occurrence of supernumerary teeth

Abnormal occurrences of supernumerary teeth have been observed in Rhynchomys and Pseudohydromys. One of the very few known specimens of Rhynchomys tapulao (FMNH 183553) exhibits a mesial supernumerary cheek tooth (Balete et al. 2007) (Fig. 4A). This tooth is simple and unicuspidate. Its location corresponds to that of the P4 among eutherian mammals. However, because the P4 had been lost among Muroidea since at least the Middle Eocene (Gomes Rodrigues et al. 2010), the homology between the P4 of eutherian mammals and the supernumerary tooth is uncertain. We thus chose to note this supernumerary tooth “P4?.” We also observed a specimen of Rhynchomys soricoides (FMNH 198787) which has three teeth on the right tooth row, but only two on the left tooth row. As for the R. tapulao specimen (FMNH 183553), the mesialmost tooth on the right tooth row is unicuspidate, whereas the two other teeth correspond to M1 and M2. We, thus, refer to this mesialmost tooth as P4?.

Figure 4.

Comparison of the upper molars of Rhynchomys and EdaTa mutants. (A) R. isarogensis, FMNH 147182. (B) R. tapulao, FMNH 183553. (C) Homozygotous EdaTa. (D) Heterozygotous EdaTa.

The tooth row of the abnormal R. tapulao is 10% longer than those of the two other specimens of R. tapulao analyzed (see Table 3). However, conversely to the abnormal R. tapulao, the right tooth row is not longer than in other specimens of R. soricoides, despite the presence of P4?. The presence of supernumerary tooth thus disturbs the length of the M1 and M2 in abnormal R. soricoides without increasing the tooth row length, which is highly variable in this species (Table 3).

Table 3. Dental measurements of Rhynchomys tapulao and Rhynchomys soricoides, including abnormal specimens with supernumerary teeth; n indicates the number of measured specimens. Note that we observed two abnormal R. soricoides, but one of them lost its P4?, such that only one tooth raw length measurement was possible

On the two abnormal specimens of Rhynchomys, the occurrence of P4? seems to affect the morphology of M1. Gomes Rodrigues et al. (2011) showed that the loss of P4 in Muroidea permitted the lengthening of M1 and notably the growing of mesial cusps. To evaluate the respective proportions of M1, we calculated the ratio (t2 length)/(M1 length) in which t2 length evaluates the size of the mesial part of the tooth in regards of the whole tooth size. This ratio ranges from 0.19 to 0.25 in abnormal Rhynchomys specimens and from 0.29 to 0.33 in other Rhynchomys specimens. These results indicate that the presence of the supernumerary tooth clearly leads to a decrease of the mesial part of the M1.

Surprisingly, there are only M1 and M2 on the left upper row in the abnormal R. soricoides (FMNH 198787). However, the M1 and M2 are 20% shorter than on the other specimens of R. soricoides and the mesial part (t2) of the left M1 is reduced as on the right M1. No evidence of tooth loss is visible on the specimen and we did not see unerupted supernumerary tooth during X-ray examination. We could however hypothesize that the modification of molar proportions are linked to the development of a tooth germ in front of the M1, and that this mesial germ completely regressed before its mineralization.

The other vermivorous murine rodent, Pseudohydromys, also sometimes exhibits supernumerary teeth. Helgen and Helgen (2009) reported two specimens of P. ellermani with M2, whereas the general dental formula of Pseudohydromy includes only an M1. From an interspecific perspective, we can also note that the relative size of M1 is variable between the four species referred to the Pseudohydromys ellermani group. P. carlae and P. ellermani have large M1, whereas P. pumehanae and P. germani have relatively small M1 (Helgen and Helgen 2009).

Rhynchomys and Pseudohydromys, thus, exemplify both a decrease in number of cheek teeth and the presence of supernumerary teeth. Their dentitions appear more variable than those of the falanouc, streaked tenrecs, and the shrew rat Tateomys.

Crown shape simplification and reduction in number of roots

Murine molars usually have complex crown shapes. The general morphotype of the upper molars are characterized by the presence of three transversal cusp rows (Fig. 2A).

The genus Rhynchomys is characterized by a simplification of its molar crown shape, with M1 lacking vestibular cusps (t3 and t9), and showing a distinct reduction of the lingually located t1 (Fig. 4B). The t1 on Rhynchomys M1 is absent in 69% of the specimens, crestiform in 14%, and small in 17%. Thus, only five cusps remain generally present and the t6 is only slightly individualized. Only t4, t5, t6, and t8 are present on M2. The t1, t3, and t9 have thus been lost compared to the basic murine crown pattern. Among Rhynchomys, lower molars are less modified than the upper molars. The M1 differs from the basic murine pattern by the absence of tE. Among Murinae, M1 and M2 primitively have three roots, whereas M1 and M2 have two roots. Bienvenu et al. (2008) showed that there is a trend in increasing the upper molar root number in many murine lineages. We observed in Rhynchomys that the number of roots on M1 varies from two to three, and that M2 is always single-rooted. Concerning lower molars, M1 has two roots, whereas M2 is single-rooted. However, M1 roots are very close and tend to be partially fused. Rhynchomys thus shows a reduction of the number of molar roots conversely to what has been observed in many murine lineages. The reduction in the number of roots occurred in parallel with the simplification of the crown pattern.

In Pseudohydromys, the remaining molars are excessively reduced in size. Their morphology is simplified to almost peg-like crown topping a single root.

The case of the Sulawesi shrew rat Tateomys is less clear than those of Rhynchomys and Pseudohydromys. Tateomys indeed distinguishes itself from primitive Murinae by the presence of supernumerary cusps on both upper and lower molars (Musser 1982), but Tateomys molars have less cusps than those of Melasmothrix, its closest genera. Thus, the presence of supernumerary cusps in Tateomys probably represents a relict, and by comparison with its close relatives, Tateomys exemplifies a crown shape simplification.

The falanouc, similarly to murine rodents, has simplified molars compared to their relatives. The teeth are transversally compressed and mesiodistally elongated. They are reminiscent of termitophageous numbat molars in that they are almost eutriconodont-like.

Both tenrec species, H. semispinosus and H. nigriceps, display simplified teeth compared to primitive mammals. The teeth of H. semispinosus in particular are almost eutriconodont-like, as in the falanouc.

Our observations, thus, indicate that vermivorous mammals exhibit a simplification of the tooth crown pattern with a decrease in the number of cusps.

Intermolar diastemata

Among vermivorous mammals, only the falanouc and streaked tenrecs display intermolar diastemata. Diastemata occur between all cheek teeth, and they are longer between the premolars. All rodents, including the vermivorous species studies here, have diastemata between incisors and cheek teeth. This diastema is thus unrelated to feeding habits and preceded the adaptation to vermivorous diet in rodents.


The comparison of evolutionary trends among the various termitophageous and vermivorous mammals allows us to highlight several similarities. These adaptations are linked to drastic modifications in the shape and number of teeth. The case of the Sumatran Hog-badger (A. hoeveni) is interesting as its dental simplification (in terms of size and shape, Helgen et al. 2008) allows for a diet based on both worms and ants.

Termitophageous and vermivorous mammals that still have teeth underwent a simplification of their dental crown shape with a reduction in the number of cusps. This simplification can reach a peg-like stage characterized by single-cusped teeth. This shape simplification is often accompanied by the appearance of intermolar diastemata. The study of the evolution of the Palaeanodonta–Pholidota shows that the crown simplification, the reduction in the number of roots, and the intermolar diastemata all appear before the complete loss of the cheek teeth. Finally, another major trend exemplified by Xenarthra and Tubulidentata is the loss of enamel, which is secondarily compensated by the acquisition of dental ever-growth.

Termitophagy and vermivory are cases of relaxed selection

The existence of similar dental trends in termitophageous and vermivorous mammals is not surprising because termitophageous and vermivorous diets both imply soft-food ingestion and are assumed to little require shearing, mastication, and food comminution in contrast to the main mammalian feeding habits (e.g., carnivory, herbivory). Davit-Béal et al. (2009) proposed that the reduction of the dentition in termitophageous mammals was possible only because of the presence of a preadapted secondary tool, which substituted for the cheek teeth. These secondary tools may correspond to beaks or protrusive tongues in termitophageous mammals (Table 2). Davit-Béal et al. (2009) also proposed that the dental continuous growth is a secondary tool, which developed to compensate the loss of enamel. The presence of a secondary tool would thus have relaxed the functional constraints on cheek teeth in termitophageous mammals, allowing shape simplification. The lack of significant tooth wear and the important plasticity of the molars in both number and morphology in numbats (Calaby 1960) support the hypothesis of decrease of the selection pressure on teeth. The decrease of the selection pressure is termed relaxed selection (often termed “degenerative” or “regressive” evolution). The model of relaxed selection notably permits to explain the case of trait loss or vestigialization. Because the source of selection was important for the maintenance of particular traits, the weakening of the selection pressure results generally in the regression or loss of the latter traits (Lahti et al. 2009). This regression can concern morphological, physiological, or behavioral characters.

Three evolutionary fates have been defined for the structures which are no longer under strong selective pressure: persistence as a relict, vestigialization, or loss (Lahti et al. 2009). The dentitions of the numbat, falanouc, and streaked tenrecs can be referred to the first group (persistence as a relict). This persistence can be due to cryptic functionality or buttressing pleiotropy (Lahti et al. 2009) and does not equate as having no function. However, the authors indicate that “increased variance of a structural trait, and particularly in its fluctuating asymmetry, can suggest a lack of function.” This variance and fluctuating asymmetry is marked in the numbat dentition (for both tooth number and morphology), pointing to a potential lack of function. Dentitions of the aardwolf, Rhynchomys and Pseudohydromys correspond to the second group (vestigialization). Paucidentomys in particular can be seen as an extreme case of vestigialization. Indeed, the two known specimens are toothless but they still have upper molar alveoli (Esselstyn et al. 2012) suggesting either a loss of teeth at a juvenile stage or aborted tooth development. In both cases, Paucidentomys must have vestigial teeth or tooth germs during its development. Anteaters, echidnas, and pangolins are clearly cases of complete loss of the dentition. As indicated above, the evolution of Palaeanodonta and Pholidota showed that teeth decreased in number and simplified in shape complexity before their disappearance. Dentition loss is thus clearly preceded by relict and vestigialization states, and they successively illustrate the three classic states of evolutionary fate under relaxed selection.

The phenomenon of relaxed selection is powerful for explaining the reduction of tooth number, the occurrence of intermolar diastemata, and shape simplification among the vermivorous and termitophageous mammals. However, less clear is the increase of the number of teeth, a feature observed in numbat, some armadillos, Rhynchomys, and Pseudohydromys specimens. As noted above, the dentition of numbats does not appear to be under selective pressure. It is worth mentioning that all their teeth are importantly simplified and that there is a frequent asymmetry in number of teeth between opposite sides of each jaw, suggesting a poor canalization of tooth number during development. In sum, because the teeth are simplified compared to the primitive pattern and have weak occlusal contacts, the variability of number of teeth does not impact the animal feeding abilities. Thus, relaxed selection allows important variations in the setting of tooth number (increase or decrease). We can consider that relaxed selection on dentition of termitophageous and vermivorous mammals leads to the phenotypic plasticity of the dentition where morpho-functional constraints are weaker.

Concerning the enamel-free ever-growing teeth, it can be mentioned that several species of armadillos are omnivorous, and that their closest related species—the extinct glyptodonts—were probably grazers. This character does not appear to be directly related consumption of ants and termites and can thus be inherited. Moreover, the ecological diversity of the other xenarthrans (sloths and anteaters) reveals an important range of adaptative possibilities. As noted by Vizcaino (2009), the peculiar features of xenarthran dentition (continuous growth, enamel-free crown) represent key innovations. They should not be underestimated as they allowed the ecological diversification of the Xenarthra. The presence of ever-growing teeth in several armadillos is explained by the history of Xenarthra, and not only by relaxed selection. Concerning the case of aardvarks, Prinz et al. (2003) concluded that chewing insects and insect larvae before swallowing is more nutritionally advantageous than swallowing them whole. Foraging and eating induce the fortuitous ingestion of dust and grit, leading to important abrasion linked with the presence of functional ever-growing crushing molars. This character would thus still be under selection.

Interestingly, there is an important difference between vermivorous and termitophageous mammals as the former lack a recognized secondary tool (Table 2), indicating that the acquisition of secondary tools such as adhesive tongue or beaks compensating tooth and/or enamel loss is not a necessity to allow loss of dentition.

We can mention that all the specialized vermivorous mammals are insular species. This observation raises the question of absence of such specialized mammals in continents. In insular environments, the specialization toward vermivory could either represent an ecological niche refuge for invasive species, which are in competition with autochthonous species, or, on the contrary, be favored by to the absence of competition or predation. The analysis of the relationships between insular environments and selection pressure on such dental traits clearly needs further investigations.

Some of the evolutionary trends we point out concerning dentitions of vermivorous and termitophageous species also exist in other mammalian species. For instance, some Delphinidae exhibit similar dentition pattern with increase in the number of teeth, diastemata, and shape simplification. Interestingly, these dolphin species use their teeth to catch prey but not for food processing (Ungar 2010), we can thus hypothesize a weakening of this source of selection on their dentition similarly to our observations on vermivorous and termitophageous species. The same is true for other mammals that do not chew their food such as the nectarivorous bats in which teeth are smaller and simplified compared to their relatives (Freeman 1995). Our interpretations of relaxed selection on dentition of vermivorous and termitophageous species could thus be extended to other diets or feeding mechanisms linked with a decrease selective pressure on teeth.

Possible genetic basis for dental evolution in vermivorous and termitophageous mammals

Lahti et al. (2009) noted that the rapidity of trait loss is affected by the mutations that happened, and their role in the trait development. The family Murinae includes some of the vermivorous mammals together with the laboratory mouse, used as a model of dental development. Vermivorous murines and the mouse are likely to share similar dental developmental mechanisms driven by the expression of similar sets of genes. To explore possible involvement of a peculiar genetic pathway in its evolution, we compared the dentition of one vermivorous rodent, Rhynchomys, and the dentitions of mutant mice (e.g., Klein et al. 2006; Ohazama et al. 2008; Charles et al. 2009a, 2009b; Ahn et al. 2010; Michon et al. 2010). We found similarities between Rhynchomys and mutants of the ectodysplasin (EDA) pathway, which is involved in the development of organs with an ectodermal origin, such as teeth (Grüneberg, 1971). Mice with mutations on the EDA pathway (e.g., EdaTa and EdardlJ) produce abnormal postcanine dentition compared to the wild-type mice (WT). Tooth number and crown shape are strongly modified (Kristenova et al. 2002; Kangas et al. 2004). As indicated above, one major feature of Rhynchomys is the loss of the distal molars (M3 and M3). EdaTa and EdardlJ mouse mutants are also characterized by modifications in the number of teeth (Charles et al. 2009a). The absence of M3 is observed in both EdaTa and EdardlJ null mutants. The M3 is usually present but simplified compared to its normal morphology. The occasional presence of a supernumerary tooth has been evidenced for the mutants of the EDA pathway (Sofaer 1969; Kangas et al. 2004; Charles et al. 2009a), but the homology of the teeth remains difficult to establish. However, as observed for abnormal Rhynchomys specimens, the first tooth in mutants is shorter and has a simpler shape than the distally located tooth. Moreover, the t2 of the second tooth of the row in EdaTa and EdardlJ mutants is reduced as in the two abnormal specimens of Rhynchomys, which possess a P4?. As in Rhynchomys, specimens with a supernumerary tooth, there are size compensations in Eda mutants with supernumerary teeth (Charles et al. 2009c). Thus, we hypothesize that the first tooth is a P4?, whereas the second, third, and fourth teeth are respectively M1, M2, and M3 in EdaTa and EdardlJ mutants (Fig. 4,C). In sum, the dentitions of EdaTa and EdardlJ mutants with supernumerary teeth share many similarities with those observed in the two abnormal Rhynchomys specimens. Concerning crown shape simplification, Charles et al. (2009a) noted that the EdardlJ mutants and homozygous EdaTa are characterized by the presence of a single lingual cusp. Among the EdaTa mutants, the t1 is absent in 69% of the homozygous mutants. This percentage is similar to that observed for Rhynchomys. As a consequence, the variability of the t1 in Rhynchomys recalls observations made on mutants for the EDA pathway. The M1 of EdaTa and EdardlJ null mutants are also characterized by the reduction of t3 and t9, and only a small t6 is present on the vestibular side of the M1. This overall M1 morphology in EdaTa and EdardlJ mutants resembles that of M1 of the Rhynchomys species (Fig. 4B, D). Concerning the lower molars, the presence of a single anteroconid on M1 is also a character shared by mutant mice of the EDA pathway (Kristenova et al. 2002; Charles et al. 2009a) and Rhynchomys species. The EdaTa mice display an important variability of the number of roots on upper molars (e.g., from one to three on M2), indicating a direct or indirect implication of the Eda gene in the dental root patterning. The variations observed in Rhynchomys (two or three roots on M1) are also a morphological convergence between Rhynchomys and EDA pathway mutant mice. In sum, the same morphological characteristics are observed in mutants for the EDA pathway and Rhynchomys: occurrence of supernumerary teeth, loss of distal molars, modifications in number of cusps and roots. Moreover, it appears that the dentition of upper jaw is more modified than this of the lower jaw when comparing EdaTa to WT mice and Rhynchomys to its sister taxa (e.g., Archboldomys). This comparison shows that losses of function in the EDA pathway occurring in EdaTa and Edardl mutants lead to dental similarities with the Rhynchomys specimens. The presence of numerous shared features between the EdaTa/Edardl mutants and Rhynchomys conduct us to hypothesize that Rhynchomys represents a possible natural mutant of the EDA pathway. Further studies will be needed to localize precisely the likely mutations by comparing sequences of Rhynchomys and sister taxa.

By definition, relaxed selection allows mutations in some parts of the genome. Mutations affecting tooth morphology could occur in coding regions or regulatory elements and the relative importance of each mechanism is unclear (Hoekstra and Coyne 2007; Stern and Orgogozo 2008). We can hypothesize that the tooth morphology of Rhynchomys is due to mutations in the coding sequences or regulatory elements of the EDA pathway. Due to pleiotropic effects of the genes from the EDA pathway, mutations in the coding regions could also impact other organs such as hair or glands (Grüneberg 1971). Mutations in cis-regulatory elements could modify dental morphology without deleterious effects on other organs and would thus seem more likely to be conserved. Such mutations have been reported in morphological evolution of various vertebrates (Cretekos et al. 2008; Chan et al. 2010). Moreover, Eda haplotypes have already been related to morphological modifications during vertebrate evolution (Colosimo et al. 2005; Barrett et al. 2008). These authors noted a reduction of the armor plate of sticklebacks due to mutations in the Eda region. In sticklebacks, Jones et al. (2012) found out that coding changes contribute for 17% to changes in the parallel marine-freshwater adaptation and the regulatory changes for at least 41%. The remaining 42% are probably regulatory changes but it still has to be demonstrated. Thus, although they appear less likely, there is evidence that mutations in coding regions can drive morphological evolution. Further studies are still needed to decipher the localization of mutations leading to tooth reduction and loss in Rhynchomys and other termitophageous and vermivorous mammals.

It is thought that Rhynchomys catches worms with its incisors, and does not masticate them. As a consequence, molars play no role in food processing, and thus seem to be useless. Thus, the adaptation to a vermivorous diet in these species leads to a decrease of the selection upon the molar morphology. The molar morphology being under weak physical constraints, it could be altered by important mutations, such as those affecting the EDA pathway. Rhynchomys, thus, clearly represents a natural case of relaxed selection upon its dentition. Our results illustrate the fact that a relaxed selection upon the postcanine dentition allows an important phenotypic plasticity in the number of teeth, suggesting a decrease in dental development canalization and possible conservation of genetic mutations impacting the dental development.

The relaxed selection model, thus, explains the original dentition of both termitophageous mammals (as noted by Davit-Béal et al. 2009) and the vermivorous mammals. However, conversely to the cases of termitophageous mammals, vermivorous mammals do not develop secondary tools to compensate the decrease function of the teeth. Thus, the relaxed selection upon dentition is not related to the positive selection of a second feature in vermivorous mammals.


The authors want to thank L. Heaney for access to the Field museum collection of mammals and for discussions on Philippines shrew rats. They also thank J. Cuisin and V. Nicolas for their assistance in the collections of the Museum National d’Histoire Naturelle (Paris), all the members of the team ‘Evo-Devo of Vertebrate Dentition’ of the Institute of Functional Genomics of Lyon for fruitful discussions, D. Polly and anonymous reviewers for their insightful comments, and J. Burden for language revision. This work has been supported by the French National Research Agency (Agence Nationale de la Recherche) (ANR, “Quenottes” program).