Phenotypic and evolutionary implications of modulating the ERK-MAPK cascade using the dentition as a model

The question of phenotypic convergence across a signalling pathway has important implications for both developmental and evolutionary biology. The ERK-MAPK cascade is known to play a central role in dental development, but the relative roles of its components remain unknown. Here we investigate the diversity of dental phenotypes in Spry2−/−, Spry4−/−, and Rsk2−/Y mice, including the incidence of extra teeth, which were lost in the mouse lineage 45 million years ago (Ma). In addition, Sprouty-specific anomalies mimic a phenotype that is absent in extant mice but present in mouse ancestors prior to 9 Ma. Although the mutant lines studied display convergent phenotypes, each gene has a specific role in tooth number determination and crown patterning. The similarities found between teeth in fossils and mutants highlight the pivotal role of the ERK-MAPK cascade during the evolution of the dentition in rodents.

. The FGF-activated ERK-MAPK cascade. The phosphorylation cascade is activated by the fixation of a growth factor to its receptor tyrosine kinase, and results in the activation of effector kinases and in the transcriptional activation of target genes. Molecular actors of this pathway that are focused on here are depicted in yellow (SPRY) and orange (RSK2).
as their impact on the other teeth of the row. Finally, we have addressed the evolutionary role of the ERK-MAPK cascade by comparing specific dental traits of mutants with dental traits of other extant or extinct rodents.

Results
The mouse (Mus musculus) is a muroid rodent, and like all the members of this superfamily, mice have a simplified dentition composed only of incisors and molars separated by a long toothless gap called a diastema. Three lower (M 1 M 2 M 3 ) and three upper (M 1 M 2 M 3 ) molar teeth are present in each mouth quadrant. The crowns of molar teeth bear a relatively stable number of cusps that are: 8 for M 1 , 6 for M 2 , 4 for M 3 , 7 for M 1 , 5 for M 2 , and 3 for M 3 (Fig. 2, first column; Fig. 3a). Crowns of upper molars are made of rows of 3 cusps arranged in linguo-vestibular chevrons pointing mesially (except the third one, which is incomplete), whereas crowns of lower molars are made of rows of 2 cusps linked by linguo-vestibular and rather straight crests called lophs (Fig. 2, first column; Fig. 3a). The two mesial lophs of the M 1 are also linked together by a mesio-distal connection.
Diversity of the dental phenotypes in Sprouty mutants. We examined the arrangement and shape of the postcanine dentition in four populations comprising 25 Spry1 −/− , 50 Spry2 −/− , 50 Spry4 −/− and 60 WT mice. Area measurements showed that the occlusal surface of molars in Spry1 −/− and Spry4 −/− mice is larger than in the WT mice, whereas the molar occlusal surface in Spry2 −/− mice is smaller than in the WT mice (t test, p value < 0.05, Fig. 4a).
The molar teeth of Spry1 −/− and WT mice are globally similar in shape, and 51% of the Spry1 −/− dental rows display a WT-like phenotype. The changes are numbered as character# (c#) from the mesial to the distal part of the row, and are mildly to moderately represented in the mutant cohorts. The main defects of the Spry1 −/− postcanine dentition are: (c8) the occurrence of a supplementary distal cusp on the M 1 (53%); (c9) the disconnection of the mesio-lingual cusp from the first chevron of the M 2 (40%); and (c12) the absence of the mesio-vestibular cusp of the M 1 (8%) (Figs 2 and 3b, Suppl. Fig. 1a). The postcanine dentition in Spry2 −/− mice had stronger differences compared to WT samples, and only 21% of Spry2 −/− dental rows still display a WT-like phenotype. The main changes of the Spry2 −/− postcanine dentition are: (c7) the connection between the two lingual cusps of the M 1 (38%); (c8) the occurrence of a supplementary distal cusp on the M 1 (46%); (c10) the connection between the two lingual cusps of the M 2 (36%); (c11) the occurrence of a lower ST (27%); and (c13-14) an abnormal shape, number, and/or interconnection of the mesial cusps of the M 1 (29%) (Figs 2 and 3c). Spry4 −/− molar tooth phenotype is the most variable among the 3 Sprouty mutants. These teeth ranged from a WT-like phenotype (24%) to relatively severe anomalies, especially in the M 1 . The main defects of the Spry4 −/− postcanine dentition are: (c1) the occurrence of an upper ST (17%); (c2-3-4) the presence of lingual cusp disconnection, straight mesial cusp and absence of the vestibular cusp affecting the first chevron of the M 1 (66% combined); (c5) the occurrence of a supplementary lingual cusp between the first and the second chevron (14%); (c6) the disconnection of the lingual cusp from the second chevron of the M 1 (20%); (c8) the occurrence of a supplementary distal cusp on the M 1 (36%); (c9) the disconnection of the mesio-lingual cusp from the first chevron of the M 2 (64%); (c11) the occurrence of a lower ST (3%); (c12-14) an abnormal number and/or interconnection of the mesial cusps of the M 1 (16%) (Figs 2 and 3d).
Except for three characters, the dentition of Spry1 −/− mice thus resembles that of WT mice, whereas Similarities and differences of Rsk2 dental phenotype as compared to Sprouty dentitions.
We next compared a cohort of 45 Rsk2 −/Y mice with a cohort of 45 WT littermates. The postcanine occlusal surface area in Rsk2 −/Y mice is smaller than in WT mice (M 1 and M 1,2 being significantly smaller, p-value < 0.05, Fig. 4a). Although many Rsk2 −/Y specimens display relatively severe dental defects, 45% of the examined dental rows display a WT-like phenotype. The main defects of the Rsk2 −/Y postcanine dentition are: (c1) the occurrence of a upper ST (14%); (c2-3-4) the presence of many defects on the first chevron of the M 1 (71%); (c11) the occurrence of a lower ST (14%); (c12 + 14 + 16) an abnormal number and/or interconnection of the mesial cusps of the M 1 (19%); (c17) abnormal mesio-distal connections between the second and the third lophs of the M 2 (9%) (Figs 2 and 3e). The frequency of ST occurrence is lower than what has been previously reported 8 , and this may be explained by the examination of a larger sample or by shifts in the genetic background over time.
The comparison of the postcanine dental phenotypes between Rsk2 −/Y and the three Sprouty mutant mice shows that  dentitions share many similar defects affecting the mesial parts of both the upper and lower first molar. The occurrence of defects on mesial parts of the M 1 is much more frequent than the occurrence of an upper ST, but the defects in the mesial parts of the M 1 and M 1 are also associated with a shortening in  the Spry4 −/− upper row, in which the area of the M 3 seems to increase in compensation for the decrease of M 1 area (Fig. 3).
In the most severely impacted phenotypes, the mesial contour of the M 1 becomes rounded, whereas it is triangular in WT mice (e.g. Fig. 4e without ST). In addition, the two most mesial cusps of the first chevron become extremely reduced into a simple crest (53% Spry4 −/− and in Rsk2 −/Y ), and the mesio-vestibular cusp may completely disappear (20% Spry4 −/− and 4% in Rsk2 −/Y ). Interestingly, the mesial shrinkage of the M 1 is associated with a change in the tilt angle of both cusps and roots of the first chevron (Suppl Fig. 1b). In WT mice, the slope of the M 1 mesial root is in continuity with the tilt of 50° that makes the central cusp of the first chevron with the dental neck. In mutants having a ST, as well as in some mutants that do not display any ST, the tilt between the root axis and the tooth neck tends to be more vertical (about 70°), as does the slope of the first chevron central cusp (about 60°). The same type of defects can be seen when M 1 is preceded by a ST in the three mutants: the first loph of the M 1 is shortened and cusps appear as crushed by the presence of the ST.

Discussion
Similar yet distinct dental phenotypes in Sprouty and Rsk2 mutants. Despite some variations in the penetrance of the dental defects in the mutants that we studied, which are likely to be accounted for by differences in the genetic background of our cohorts, some trends stand out. Morphological comparisons show that the occurrence of a supplementary cusp at the distal extremity of the M 1 can be considered as a phenotypic signature of the Sprouty mutant dentitions, and Rsk2 mutants never develop this supplementary cusp. These comparisons also highlight that Spry4 −/− and Rsk2 −/Y mutants share many phenotypic features, such as the occurrence of upper and lower ST as well as abnormal arrangements in the mesial parts of the first molars. Although mesial abnormalities in these teeth may result from the occurrence of ST, a question remains concerning the higher frequency of modifications of the M 1 first chevron in Spry4 −/− and Rsk2 −/Y mutants (66-71%) when compared to the frequency of occurrence of ST (14-17%). It has already been suggested in Eda Ta mutant mice that a supplementary dental germ can develop mesially to the M 1 until advanced stages of odontogenesis without necessarily giving rise to a mineralized tooth 32 . In these cases, the development of a supplementary dental germ may be sufficient to cause disorders in the development of the mesial part of the following tooth. We can thus infer that about 70% of the specimens develop an upper ST germ that will impact on the phenotype of the M 1 .

Roles of Sprouty and
Rsk2 genes within the ERK-MAPK cascade. Sprouty genes are thought to encode proteins that inhibit signalling through the FGFR pathway 18,20,33 . Examination of the dental phenotypes in mutant mice led to the discovery of shared features. All three Sprouty mutants share the presence of a supernumerary distal cusp, arguing in favour of shared functions. But distinct characters were also observed in each mutant, pointing to some specific functions of each of these genes in tooth morphology. The phenotypic diversity in the Sprouty mutants may be ascribed to differences in spatial and temporal expression patterns of these genes. Our data complement previous findings concerning the generation of similar phenotypes with the loss of function of Spry2 and Spry4 genes. Spry4 −/− phenotype penetrance was previously reported as lower than the one in Spry2 null mutants 7 . Our results show that the differences between the two proportions are even larger than previously reported. This trend is reversed for the upper rows, for which we did not observe ST in Spry2 −/− mice, whereas these were found in 17% in Spry4 −/− mice. The proportion of ST in mutant dental rows was comparable to what has been recently reported 22 , but lower than the first publication on the topic 7 . Such a change in penetrance could be explained by a shift in the genetic background over generations. Finally, neither of the previous reports 7, 22 studied the effect of loss of function of Spry1. Our work shows that Spry1 −/− mice also display modifications in the shape of the M 1 and M 2 , as well as in overall size of the dental row.
RSK2 is an effector kinase that acts downstream of the MAPK cascade, whereas Sprouty genes modulate the cascade through their inhibitory effect on the pathway. The possible negative feedback regulation from Rsk2 on the adaptor protein SOS, which recruits Ras 27 , may lead to similarities in the phenotypes of Sprouty and Rsk2 mutants. Both Spry4 and Rsk2 are expressed during odontogenesis in the dental papilla 8,21,34,35 , which may explain the high degree of similarity in the phenotypes associated with their loss of function. Our study thus provides an example of molecular and phenotypic convergence of various steps in the same signalling pathway.
Mutations in Sprouty genes and Rsk2 lead to more teeth and more cusps. This is the opposite trend of mutations in the Fgf genes (e.g. mutations in Fgf3 36,37 lead to fewer cusps). This contrast in phenotypic impacts of the mutations between Rsk2 and Fgf3 is concordant with the current knowledge of regulatory feedbacks acting on the whole pathway. Overall, in terms of phenotype penetration, one must note that the frequencies depend on the defect considered, and that the penetrance is never complete. Variations in the initial genetic background of each line might be responsible for some of the differences observed. Structural conservation and functional redundancy of the actors of the pathway 38 can however account for the shared dental defects. The Apc and Ctnnb1 null mutants display many more ST located in the vicinity of the incisor region 41 . These authors hypothesized that incisor stem cells might be migrating, thus generating odontogenesis-competent foci. The difference between one ST being displayed in a mesial position (compared to the molar row), and multiple ST being displayed in a disordered manner is likely influenced by the direct interaction with Ras and ERK. Indeed, such interactions have been demonstrated in the context of multiple cancer models in the mouse 43,44 . Potential roles of Sprouty and Rsk2 genes in the evolution of rodent dentition. The most striking phenotypic trait of Spry2 −/− , Spry4 −/− and Rsk2 −/Y mutants is the occurrence of ST located in front of the first molars at both the lower and the upper jaws. It has long been known that the rodent dentition has evolved towards a reduction in the number of teeth over the course of evolution in parallel with a specialization of the molar teeth 45 . Basal rodents had lower and upper premolars, and it is only since the rise of muroid rodents that premolars have been completely lost in the lineage leading up to the murine rodents. It has however been demonstrated that rudimentary dental germs are maintained in both the upper and the lower diastema during mouse early dental development 24,25 . Because these rudimentary germs are located in the vicinity of the first molar germs, they have been considered to be rudiments of premolar germs lost over evolution 26 . These rudimentary germs rapidly abort by apoptosis and do not become autonomous mineralized teeth. It has been shown that they can contribute to the development of the M 1 by merging with the M 1 germ 26 . Although it has not yet been formally proved, it is likely that the same mechanism occurs with the upper rudimentary germ immediately adjacent to the M 146 for participation in the development of the first chevron.

Molecular basis of ST development. In addition to Sprouty and
Here we deliver new insights into the role of the MAPK signalling pathway members beyond what has been previously reported 7,8,22 . Interesting, this signalling pathway is conserved in mammalian species, and its signal transduction properties are even shared more broadly 38,47 , adding its plausible role in the evolution of mammalian dentition. Here, the loss of function of Spry2, Spry4 or Rsk2 allows the continuation of the development of some rudimentary premolar germs until the formation of autonomous mineralized teeth. The segmentation of the postcanine dental row is thus impacted, because it comprises four teeth instead of three and the mesial parts of the first molars are accordingly reduced (Fig. 4). We can thus assign the ST located immediately in front of the first molars to deciduous fourth premolar teeth (dP4 and dP 4 ) that will not be replaced. Teeth located at the fourth premolar position are known in the earliest muroid rodents 48 as well as in some extant dipodoids such as Sicista and Zapus 49,50 , the sister group of muroids. Fourth premolars are present in the squirrel-and in the guinea pig-related clades, Scientific RepoRts | 5:11658 | DOi: 10.1038/srep11658 which arose earlier in the history of rodents than the mouse-related clade 51 . The loss of function of Spry2, Spry4 or Rsk2 involves developmental mechanisms that revitalize autonomous premolar teeth and that simplify accordingly the mesial parts of the first molars. The dental phenotype in Spry2 −/− , Spry4 −/− , and Rsk2 −/Y mutants could thus be considered as a partial reversal in the evolutionary trends of the dental phenotype through the transition from pre-muroid to muroid rodents (Fig. 6a), which is documented within the fossil record to have probably occurred between 50 and 45 million years ago (Ma) 48,52 .
As discussed above, many other mouse mutants display ST mesially to first molars, but both the phenotypic and the developmental aspects of the postcanine row segmentation have solely been scrutinized in mutants of the EDA pathway 32 . From these studies, it is clear that dental phenotypes in Eda Ta (Tabby) and Edar dl-J (Downless) mice radically differ from those of Spry2 −/− , Spry4 −/− , and Rsk2 −/Y mice. Eda Ta and Edar dl-J mice have postcanine teeth with simplified occlusal patterns characterized by the absence of many cusps, as well as by abnormal arrangements of the lophs and the chevrons respectively in the lower and upper molars 32 . In contrast, Spry2, Spry4, or Rsk2 losses of function provoke rearrangements of the M 1 and M 1 mesial parts, but minimally impact the rest of the dental rows. From these arguments, we propose that Spry2, Spry4 or Rsk2 are candidate genes for a major role in the evolutionary trend towards reduction of the dental formula in muroid rodents 45 Ma.
Interestingly, a character that stands out from the mutant mice is the occurrence of a supplementary distal cusp, named the posterocone, on the M 1 of the three Sprouty mutants (arrowhead in Fig. 6b). The posterocone is rarely present in extant species of murine rodents, whereas this cusp or the equivalent crest located at the same location, the posteroloph, were almost always present in the basal fossil murine genera such as Potwarmus, Antemus, and Progonomys. However, neither the modern mouse (Mus musculus) nor its ancestor (Mus auctor) displays an individualized posterocone (Fig. 6b). In the lineage leading to the modern mouse, the posterocone has been lost during the transition from Progonomys debruijni to Mus auctor, and this transition is documented to have occurred between 9.2 and 6.5 Ma by the fossil record in the Siwaliks of Pakistan 53 . Apart from the presence of a posterocone, Spry1 −/− , Spry2 −/− , and Spry4 −/− mice share other dental features with these basal murine genera such as the trend of the lingual cusps of the M 1 to be less connected to the central cusps. All of these similarities indicate that Spry1, Spry2, or Spry4 could have played a major role during the transition from Progonomys to Mus, which constitutes a major transition in the history of murine rodents.

Concluding Remarks
Spry1, Spry2, Spry4, and Rsk2, which encode components of the ERK-MAPK signalling pathway are identified here as potentially major actors in the evolution of the dentition in rodents. These genes may have played a pivotal role in the reduction of the dental formula and in the rearrangement of the mesial parts of the first molars during the transition from pre-muroid to muroid rodents prior to 45 Ma. They also may have acted in the disappearance of the posterocone and in the connections of the lingual cusps to the chevrons during the transition from basal murines to Mus auctor between 9.2 and 6.5 Ma. Both morphological transformations happened independently, separated by about 35 Ma, suggesting that the premolar loss might have been triggered by Rsk2 gene dosage augmentation. The posterocone would then have been lost because of an increase in Sprouty gene expression. An interesting issue remains open. Indeed the potential role of Sprouty or Fgf genes in the morphogenesis of the connections between lingual cusps and central cusps of the M 1 chevrons needs further investigations. Here we report that Sprouty mice, especially Spry2 −/− and Spry4 −/− mice, often have M 1 with disconnected lingual cusps. As Sprouty and Fgf genes are known to play antagonist roles during tooth morphogenesis, it is interesting to note that the loss of function of the two antagonist results in an equivalent impact on the lingual cusps of the M 1 . The role of this central pathway could be further studied using other genetically engineered mice, notably over-expressing the same molecules to understand the plausible evolutionary implications of dosage modulations. Focusing on the molecular redundancy allowing vestigial tooth buds to develop into fully erupted teeth is likely to bring new insights into the molecular networks involved in both tooth development and evolution.

Material And Methods
Sprouty mutant mice. All the Sprouty mutant mice were generated in backgrounds resulting from crossing between several lineages. The three mutants and the wild type mice were generated by breeding at UCSF. The sample set was composed of homozygous mice as indicated: 25 Spry1 −/− , 50 Spry2 −/− , 50 Spry4 −/− , and 60 WT individuals (littermates of the various mutants). For each specimen, left and right, upper and lower tooth rows were studied independently. The age of the specimens ranged from 1 month to 2.5 months. Animal experimentation was carried out in compliance with the policies and procedures established by the UCSF Institutional Animal Care and Use Committee.
Rsk2 −/Y mice. The Rsk2 mutant mouse line was generated as previously described 54 . Since Rsk2 is located on the X chromosome, analyses were performed on Rsk2 −/Y males, on a C57BL/6J background. Observation and imaging of dental rows. All the heads were prepared in order to remove all non-mineralized tissues to allow good observation and measurement of the dental rows. They were examined and photographed using a Leica stereomicroscope. The measurements were obtained by following the outline of each tooth from the photos of occlusal view of the row. Thus, the length, width, and area of the tooth were produced by Leica software. Some Sprouty mutant dental rows were imaged using X-ray-synchrotron Radiation Facility (ESRF, Grenoble, France), beamline ID 19 and BM5, with a monochromatical beam at energy of 25 keV. X-ray synchrotron microtomography has been demonstrated to bring high-quality results for accurate imaging of small teeth 55 . A cubic voxel of 7.46 μ m was used. Some Rsk2 −/Y mutant dental rows were imaged using the X-ray cone-beam computed microtomography with a Nanotom machine (GE) with a source tension of 100 kV with a cubic voxel of 3 μ m. All 3D renderings were performed using VGStudiomax software.
Statistics. Student's t-tests were used to verify the significance of differences in tooth size between the mutant mice but also between the mutant and the WT mice. A threshold value of 0.05 (p-value) was used to assess the significance of the observed differences.