Evolutionary and Functional Implications of Incisor Enamel Microstructure Diversity in Notoungulata (Placentalia, Mammalia)

Notoungulates are an extinct clade of South American mammals, comprising a large diversity of body sizes and skeletal morphologies, and including taxa with highly specialized dentitions. The evolutionary history of notoungulates is characterized by numerous dental convergences, such as continuous growth of both molars and incisors, which repeatedly occurred in late-diverging families to counter the effects of abrasion. The main goal of this study is to determine if the acquisition of high-crowned incisors in different notoungulate families was accompanied by significant and repeated changes in their enamel microstructure. More generally, it aims at identifying evolutionary patterns of incisor enamel microstructure in notoungulates. Fifty-eight samples of incisors encompassing 21 genera of notoungulates were sectioned to study the enamel microstructure using a scanning electron microscope. We showed that most Eocene taxa were characterized by an incisor schmelzmuster involving only radial enamel. Interestingly, derived schmelzmusters involving the presence of Hunter-Schreger bands (HSB) and of modified radial enamel occurred in all four late-diverging families, mostly in parallel with morphological specializations, such as crown height increase. Despite a high degree of homoplasy, some characters detected at different levels of enamel complexity (e.g., labial versus lingual sides, upper versus lower incisors) might also be useful for phylogenetic reconstructions. Comparisons with perissodactyls showed that notoungulates paralleled equids in some aspects related to abrasion resistance, in having evolved transverse to oblique HSB combined with modified radial enamel and high-crowned incisors.


Introduction
Notoungulates constitute an extinct clade of South American ungulates showing a wide range of skeletal and cranial shapes that are in some aspects reminiscent of rhinos, horses, and even rabbits or rodents, but for which ecological affinities need to be more accurately understood (Scott 1932(Scott , 1937Simpson 1980;Bond et al. 1995;Reguero and Prevosti 2010;Cassini and Vizcaíno 2012;Giannini and García-López 2013;Croft 2016). Notoungulates evolved in changing geological and environmental contexts in South America from the late Paleocene to the Pleistocene-Holocene transition, a phenomenon that has been abundantly analyzed and commented (e.g., Pascual and Odreman Rivas 1971;Cifelli 1985;Pascual and Ortiz-Jaureguizar 1990;MacFadden et al. 1994;Croft 2001;Ortiz-Jaureguizar and Cladera 2006;Strömberg et al. 2013;Dunn et al. 2015;Kohn et al. 2015). The longest-lived Neogene families of notoungulates, i.e., the Toxodontidae, Interatheriidae, Mesotheriidae, and Hegetotheriidae, repeatedly evolved an impressive series of dental innovations, involving prolonged to continuous growth of the entire dentition (i.e., hypsodonty to hypselodonty; Billet et al. 2009;Madden 2015), and fast eruption of the entire set of cheek teeth (Gomes Rodrigues et al. 2017). The potential ecological and biological correlates of the precocious crown height increases when compared with other euungulates (i.e., more than 10 million years before their appearance in Perissodactyla and Artiodactyla) have received much attention (e.g., Stebbins 1981;Flynn and Wyss 1998;Jacobs et al. 1999;MacFadden 2000;Croft and Weinstein 2008;Townsend and Croft 2008;Billet et al. 2009;Gomes Rodrigues et al. 2017). The hypothesis of a repeated evolution of this dental innovation in increasingly arid environments, especially in Patagonia from the middle Eocene onward, remains intensively discussed (e.g., Madden 2015;Dunn et al. 2015;Kohn et al. 2015) and its functional implication remains to be determined.
Ever-growing and large incisors count among the widespread dental innovations within notoungulates. A significant crown height increase of incisors is also known in some other mammalian clades, such as elephants, hippos, hyraxes, rodents, lagomorphs, or some marsupials (e.g., Koenigswald 2011; Renvoisé and Michon 2014), in which it is often viewed as a major functional innovation. However, the functional and ecological significance of the convergent evolution of highcrowned incisors in notoungulates remains to be more precisely understood. The analysis of incisor enamel microstructure can provide information on biomechanical aspects. As dental enamel microstructure displays highly complex patterns in mammals, it can also provide information on phylogenetic affinities (Korvenkontio 1934;Koenigswald and Clemens 1992;Koenigswald et al. 1993;Clemens 1997). To date, incisor and molar enamel microstructure have been investigated in numerous groups of hoofed mammals (Euungulata and Paenungulata; e.g., Pfretzschner 1993;Koenigswald et al. 2011;Alloing-Seguier et al. 2014;Tabuce et al. 2017). However, microstructural characteristics of the dental enamel in notoungulate incisors remain understudied, because the rare studies have focused almost exclusively on molars (Pfretzschner 1992;Maas 1997;Lindenau 2005).
Here, we aim at studying the variations of incisor enamel microstructure in notoungulates to better characterize their convergent evolution and their potential ecological significance. The description of the different schmelzmusters (i.e., 3D arrangement of enamel types) observed in the upper and lower incisors of an extensive sample of notoungulate taxa first documents the microstructural diversity and its potential function in relation to the shape and mode of growth of these teeth. An evolutionary scenario for several selected microstructural traits is also proposed and discussed based on parsimonious reconstructions of character evolution.

Material Examined
Fifty-eight samples of incisors representing 21 genera of notoungulates were investigated. They were collected in the MNHN (Paris, France), the MACN (Buenos Aires, Argentina), and the MLP (La Plata, Argentina; Table 1). These samples include genera from the Eocene to Pleistocene, put in a temporal and composite phylogenetical framework (based on data from Nasif et al. 2000;Reguero and Prevosti 2010;Billet 2011;Woodburne et al. 2014;Shockey et al. 2016). For many genera, both upper and lower incisors from different loci were sampled, because of the important size range of incisors observed in many notoungulate genera (see Table 1). Moreover, significant differences of schmelzmuster were reported between upper and lower incisors in mammals, especially rodents (Kalthoff 2000). For some samples, identification at the species level was not possible; therefore, we provide only generic identification for all sampled specimens (referred with Bcf.^for some generic identifications). Additional data for upper incisors of Adinotherium and Mesotherium were taken from Lindenau (2005).

Enamel Microstructure Preparation
Specimens were embedded in polyester or epoxy resin. Longitudinal sections were performed labiolingually from the apex to the root of the incisor, whereas labiolingual transverse sections were perpendicular to the longitudinal plane ( Fig. 1). Some sections were not realized due to limited sizes, intense wear, or absence of the enamel layer (Table 1). As a result, 103 sections were generated. Then, these sections were ground and polished with 1200 Al 2 O 3 grit powder. In order to make enamel prisms and interprismatic matrix visible, we performed phosphoric acid etching for 5 s and rinsed with water combined with ultrasonic cleaning. Specimens were sputter-coated with gold for 2 min and small samples were subsequently mounted on SEM stubs (Kalthoff 2000). The enamel microstructure was studied and documented with a scanning electron microscope Jeol Benchtop JCM-6000 at an acceleration voltage of 15 kV and at magnifications from ×85 to ×2400. Preparations resulting from this study are deposited in the enamel collection of the Steinmann Institute (Bonn, Germany; STIPB) and in the collections of the MNHN (Paris, France).

Parameters Evaluated
The following parameters were evaluated in the description of incisor enamel microstructure ( Fig. 1       layers between rows of prisms, which characterizes a modified radial enamel. Furthermore, the IPM was described as Bintermediate^if there are those two types in the same enamel layer, meaning that IPM is clearly anastomosing. The angle made by the orientation of crystallites forming the IPM with respect to the prisms is sometime difficult to measure, because the orientation of crystallites may not be clearly visible. As a result only the maximal angle was measured when possible.
Notostylops presents enamel only on the labial side of I1, but on both sides of i1. The enamel is~170 μm thick on I1 and 140-200 μm on i1. The enamel microstructure of Notostylops is similar on both I1 and i1, and on both labial and lingual sides concerning i1. The schmelzmuster is onelayered and exclusively represented by radial enamel. HSB are absent, but prisms are slightly oblique, and their diameter is~6 μm. IPM forms closed coats on the entire thickness, but it is slightly anastomosing close to the EDJ. The maximal angle between the IPM and prisms is about 25-30°.
Pyrotherium presents enamel on both sides of the incisor investigated, which display the same enamel microstructure. The enamel measures~1650-1720 μm and is characterized by a two-layered schmelzmuster with ORE (23%) and thick HSB starting at the EDJ (77%). Prism diameter is~6 μm. The HSB show a key hole pattern (see Koenigswald et al. 2015), while IPM forms closed coats in the outer radial enamel. In the keyhole pattern, prisms are densely packed and IPM can be observed in some areas. Furthermore, in longitudinal section, those prisms show an arc-shaped prism sheath that is widely open to one side. HSBs (55-215 μm thick) display a more variable aspect in longitudinal section than in transverse section, where they are more steady. The maximal angle between the IPM and prisms is about 35-40°.  Table 2).
The specimens referred as cf. Pleurostylodon and cf. Thomashuxleya present enamel on both labial and lingual sides of upper and lower incisors. On the transverse section of upper incisors, the enamel is~310-580 μm thick in cf. Thomashuxleya and~150-180 μm in cf. Pleurostylodon. On lower incisors, the enamel is~280-410 μm thick in cf. Pleurostylodon. Lower incisors of cf. Thomashuxleya were not available for analysis.
The enamel microstructure is very different between the two taxa. Cf. Pleurostylodon presents slightly different enamel microstructure on dI1 and lower incisors, according to the IPM arrangement, but the microstructure pattern is similar on both lingual and labial sides. dI1 is characterized by a onelayered schmelzmuster formed by radial enamel. HSBs are absent, but prisms are oblique except near the EDJ where they present a high angle with respect to the longitudinal plane. Prism diameter is~5-7 μm. The IPM forms closed coats, but it is slightly anastomosing close to the EDJ. On the longitudinal section, prism orientation is 60°throughout the entire thickness. The enamel microstructure of i1/i2 and i2/i3 is similar. The schmelzmuster is one-layered and exclusively represented by radial enamel. The IPM forms closed coats near the OES and is intermediate near the EDJ. The maximal angle between the IPM and prisms is about 40-60°.
The enamel microstructure of cf. Thomashuxleya is similar on I1 and I3 and on their both sides. We noticed a threelayered schmelzmuster. There is radial enamel close to EDJ and the OES (10-19% and 15-25%), and HSBs in the middle layer (56-73%). Prisms diameter is between~6-7 μm. The HSBs are generally thin (35-80 μm thick), oblique, and steady. Moreover, HSB are discernible on both longitudinal and transverse sections. The IPM forms closed coats in the HSB layer and in the inner and outer radial enamel. The maximal angle between the IPM and prisms is about 60-80°.
Leontinia presents enamel on both labial and lingual sides of I1 and i3, but only on the lingual side of I2. On the transverse section of upper incisors, the enamel is~350-800 μm thick and on i3 it is~270-370 μm thick. The enamel microstructure is similar in the sampled upper incisors, which are slightly different from the lower incisor. Furthermore, the enamel microstructure is also different between labial and lingual sides for upper incisors, while it is similar on both sides of i3.
The schmelzmuster of the labial side of the upper incisors is three-layered with radial enamel near the EDJ and OES (10-16% and 12-28%) and HSBs in the middle layer (62-72%). Prism diameter is~8-9 μm. The HSBs are 35-100 μm thick, and they are oblique with variable aspect. The IPM forms closed coats in the HSB and in the ORE. We observed intermediate IPM near the EDJ in the IRE. The schmelzmuster of the lingual side of I1 is one-layered and exclusively presents oblique HSBs (45-105 μm thick), which have a variable aspect. IPM forms closed coats.
The maximal angle between the IPM and prisms is about 40-65°.
Eurygenium presents enamel on both sides of the lower incisors but only on the labial side of upper incisors. On the transverse section of upper incisors, the enamel is~470-550 μm thick and on lower incisors the thickness is~370-580 μm. All upper incisors display identical schmelzmuster, but it is different from lower incisors. The enamel on lower incisors presents the same schmelzmuster on both sides.
The schmelzmuster of upper incisors is three-layered with radial enamel near the EDJ and OES (20-24% and 21-23%) and HSBs in-between (53-59%). Prism diameter is~7 μm. The HSBs are thin (20-65 μm) and transverse. On the transverse sections, prisms present a slightly wavy pattern; the HSBs are not clearly distinguishable. The IPM forms closed coats in the HSBs and near the OES. We observed intermediate IPM near the EDJ.
Lower incisors are characterized by a two-layered schmelzmuster represented exclusively by radial enamel. Prism diameter is~7 μm. Prisms are oblique near the EDJ and less inclined near the OES. The IPM is intermediate in the IRE and forms closed coats in the ORE.
The maximal angle between the IPM and prisms is about 55-65°.
The sampled toxodontids present different enamel microstructures, but the schmelzmuster is homogeneous within each genus. Nesodon and Toxodon present enamel only on the labial side of upper incisors, Adinotherium has enamel on the labial side of I1, but both on the labial and lingual side of I2 (all information for upper incisors of Adinotherium is from Lindenau 2005). Adinotherium and Nesodon present enamel on both sides of lower incisors whereas Toxodon present enamel only on the labial side.
In the upper incisors in Nesodon and Toxodon, the enamel is thick (~960-1120 μm). On their lower incisors, the enamel is thinner (between~320-1000 μm). The enamel microstructure is similar on upper and lower incisors of Adinotherium and Nesodon, whereas the schmelzmuster in Toxodon is different from the latter. When enamel is present on the lingual side of lower incisors, the enamel microstructure is similar to the labial side.
Upper and lower incisors of the sampled toxodontids present a three-layered schmelzmuster with radial enamel near the EDJ and OES (7-18% and 5-29%) and HSBs in-between (58-85%). Prism diameter is~7 μm. Prisms orientation is different in the HSB compared to the IRE and ORE. Variation in the thickness of HSBs is important among Toxodontidae (Adinotherium: 20-90 μm, Nesodon: 25-65 μm, Toxodon: 45-105 μm). Aspect of HSB is generally steady in Adinotherium and Nesodon, but rather variable in Toxodon. Toxodontidae have transverse HSB, given that prisms have a wavy pattern on the transverse sections, which makes HSB difficult to discern. The IPM generally forms closed coats but tends to anastomose in the HSB and near the EDJ. More precisely, the IPM is intermediate near the EDJ in the lower incisors of Adinotherium and Toxodon, while it also forms interrow sheets in the HSB and near the EDJ in upper incisors of Nesodon. The EDJ is undulated (Fig. 3d-h).
The maximal angle between the IPM and prisms is quite similar between incisors of Adinotherium and Nesodon (about 80-90°, and 85-90°respectively), while it is slightly slower in Toxodon (about 65-80°).
Cf. Oldfieldthomasia presents enamel on both sides of the upper and lower incisors. In the upper incisors, the enamel is 110-410 μm thick and in the lower incisors between~190-360 μm. The enamel displays the same microstructure on both upper and lower incisors and on both sides.
The schmelzmuster is one-layered and exclusively represented by radial enamel. Prism diameter measures 6 μm. Prisms are inclined in the radial enamel of I2/I3 and of the lingual side of I1 compared to both planes, while they are only slightly inclined on the labial side of I1. Prisms are inclined in the radial enamel near the EDJ of i1 and i2/i3. They are less inclined regarding the transverse plane near the OES. The IPM forms closed coats to intermediate along the entire thickness. The maximal angle between the IPM and prisms is about 55-65°.
Interatheriidae present enamel on both sides of lower incisors but only on the labial side of the upper incisors (except in Interatherium). The enamel is thicker on the transverse section of the upper incisors in Interatherium and Protypotherium (between~90-320 μm) than in cf. Notopithecus (~70 μm).
The enamel microstructure varies within the family. The enamel shows different microstructures between the upper and lower incisors (except in cf. Notopithecus), but it is similar between the labial and lingual sides (except in Interatherium). Notopithecus present a one-layered schmelzmuster with radial enamel. Prisms are inclined, and their diameter is~7-8 μm.
The IPM forms closed coats on the entire thickness, but it is slightly anastomosing near the EDJ.
The same microstructure pattern is observed on the lingual side of I1 and on the lower incisors in Interatherium. Prism diameter is~6 μm. The labial side of I1 of Interatherium is characterized by a three-layered schmelzmuster represented by IRE and ORE (16% and 17%, respectively) and prisms present a slightly wavy pattern in the intermediate layer (67%) on the transverse section, which indicates the presence of incipient HSB. The IPM forms closed coats in the entire thickness, but it is also slightly anastomosing near the EDJ.
The upper incisors of Protypotherium are characterized by a three-layered schmelzmuster represented by ORE (14-21%), HSBs (76-86%), and IRE (5%). Prism diameter is 6 μm. The HSBs are thin (20-50 μm thick) and oblique. The IPM forms closed coats near the OES and interrow sheets near the EDJ, and is intermediate to modified in the HSBs. Lower incisors of Protypotherium are characterized by a onelayered schmelzmuster with HSBs. Prism diameter is~6 μm. HSBs are generally oblique and steady, even if they present a low decussation on both sections and are less discernible on some transverse sections. The IPM forms closed coats in the entire thickness, but it is also slightly anastomosing near the EDJ.
If the maximal angle between the IPM and prisms is about 45-55°in cf. Notopithecus, it is higher and quite similar in Interatherium and Protypotherium (about 60-70°and 65-75°, respectively).
The enamel microstructure is quite similar among hegetotheriids (except regarding the lower incisors of Paedotherium). They have a general pattern characterized by different enamel schmelzmusters between the upper and lower incisors, as well as between the labial and lingual sides (mostly for the upper incisor). Prism diameter is generally~4-5 μm, but 6 μm was measured for the I1 of Hegetotherium. HSBs are only observed on the labial side of I1 in all hegetotheriids, and on the lingual side of lower incisors in Paedotherium. They are thin (10-45 μm) steady and oblique, given they are clearly discernible in longitudinal and transverse sections. The IPM generally forms closed coats, it is sometimes intermediate on the entire thickness especially regarding the labial side of the lower incisors, and interrow sheets are only observed on the labial sides of the lower incisors in Paedotherium.
More generally, the labial side of I1 of all hegetotheriids, and the lingual side of lower incisors of Paedotherium are characterized by a two-layered schmelzmuster with ORE (18-49%) and HSBs (54-82%). The lingual side of both I1 and lower incisors, except in Paedotherium, are characterized by a one-layered schmelzmuster represented exclusively by radial enamel. The labial side of the lower incisors is characterized also is represented by radial enamel, but prisms present different orientations near the EDJ compared to the OES.
Mesotheriidae present enamel on both sides of the upper and lower incisors, except on the lingual side of I1 in Trachytherus. The investigated mesotheriids also display the same enamel microstructure. On the transverse section of the upper incisors, the enamel is thicker in Trachytherus (~650 μm) than in Plesiotypotherium (~330-350 μm) and Pseudotypotherium (~300-380 μm). On the lower incisors, the enamel is~330-520 μm thick in Trachytherus,~260-440 μm in Plesiotypotherium, and~270-280 μm in Mesotherium. The investigated Mesotheriidae are characterized by a three-layered schmelzmuster represented by IRE and ORE (13-54% and 8-24%) and HSBs (27-77%). Prism diameter is~6-8 μm. Prism orientation is different in the central parts of the enamel band compared to the outer parts of the bands. The variation of the thickness of HSBs is high among Mesotheriidae (20-130 μm thick). HSBs are oblique except in Trachytherus and in the lingual side of I1 of Plesiotypotherium and Mesotherium (cf. Lindenau 2005) where they are slightly inclined compared to the transverse plane. HSBs are also discernible on the transverse section, except in Plesiotypotherium on the lingual side of I1, Mesotherium on I1, and Trachytherus, which presents prisms with a slightly wavy pattern in the middle layer. When present, HSB have a steady aspect, except for lower incisors of Trachytherus and Mesotherium. The IPM forms interrow sheets near the EDJ and in HSBs and is variably intermediate regarding the labial sides of lower incisors. We only observed closed coats IPM near the OES.
The maximal angle between the IPM and prisms is about 50-65°in Trachytherus. This angle is higher in Mesotherium ( a b o u t 8 0 -8 5°) , a n d i n P l e s i o t y p o t h e r i u m a n d Pseudotypotherium (both about 85-90°).

Evolutionary History of Incisor Enamel Microstructure in Notoungulates
Notoungulates present a wide diversity of incisor shapes and crown heights, which is accompanied by diverse enamel microstructure patterns (Fig. 6, Table 2), as already noticed for molars (Pfretzschner 1992;Maas 1997;Lindenau 2005). This diversity (e.g., presence of HSB, IPM characteristics) is also observed between the different incisor loci in many notoungulates. These differences are especially obvious between upper and lower incisors of the same taxon (Figs. 6 and 8), and far less between adjacent teeth, despite frequent differences in size (see Tables 1 and 2). The family Interatheriidae provides a good example of the spectrum of microstructural evolutionary traits that exists in notoungulates. Great differences of schmelzmusters can indeed be found within the Interatheriidae, which occur over a long time span between the earliest (i.e., Notopithecus) and latest sampled genera (i.e., Interatherium; Fig. 7). Many characters are different between these taxa: the tooth morphology, the distribution of enamel, the number of layers, and the orientation of HSBs. Minor differences are also observed between the Miocene interatheriine taxa Protypotherium and Interatherium, but more important differences are also observed between upper and lower incisors (Fig. 8).
Most of the Eocene notoungulates incisors studied here present only radial enamel (Fig. 7). Cf. Thomashuxleya is the only Eocene taxon presenting a derived schmelzmuster with HSB, also combined with a high angle between the  ., 80°, Fig. 7). This kind of schmelzmuster involving HSB is observed in all investigated notoungulate families from the Oligocene onwards, even if it does not concern all incisors (Fig. 7). Parsimonious reconstruction suggests that HSB have evolved convergently within Typotheria (i.e., Interatheriidae, Hegetotheriidae, Mesotheriidae), Toxodontia, and Pyrotheria, on lower incisors (Fig. 8).
Conversely, HSB is regarded as plesiomorphic for upper incisors when mapped on the tree of sampled genera. This scenario is however questionable given that HSB are widely considered as a derived condition in mammals (e.g., Koenigswald et al. 1987). A greater diversity of sampled Eocene genera might have allowed to obtain this expected evolutionary pattern for notoungulate upper incisors. These results at least clearly suggest that HSBs could be first achieved for upper incisors in comparison to lower ones (e.g., Eurygenium, Interatheriidae, Hegetotheriidae; Figs. 6 and 8). In addition, HSB are more developed in incisors of medium to large n o t o u n g u l a t e s ( e . g . , L e o n t i n i a , To x o d o n t i d a e , Mesotheriidae), where they are present in all investigated loci and incisor sides. All high-crowned taxa investigated in notoungulates presented HSBs on their upper incisors, but not always on their lowers. HSBs evolved in a large number of placental clades (Koenigswald et al. 1987 and not exclusively in hypsodont or hypselodont taxa, and this is also the case in notoungulates (e.g., the brachydont Thomashuxleya). These widespread structures are recognized as an effective crack-stopping mechanism in enamel (Pfretzschner 1992;Koenigswald et al. 2011). Most importantly, the common occurrence of HSB in notoungulate incisors adds to the growing body of evidence that the clade Euungulata, comprising Perissodactyla, Artiodactyla, and potentially Litopterna and Notoungulata (Welker et al. 2015;Buckley 2015;Westbury et al. 2017), may be characterized by a frequent development of HSB (Stefen 1999;Lindenau 2005;Alloing-Seguier et al. 2014;Tabuce et al. 2017).
Most post-Oligocene notoungulates convergently acquired at least intermediate levels of modified radial enamel (anastomosing IPM) notably near the EDJ, whereas Mesotheriidae were among the only ones to evolve a fully modified radial enamel (IPM forming interrow sheets; Figs. 7 and 8). Surprisingly and conversely to HSB, radial enamel is more often modified in lower than in upper incisors in our sample. More generally, the development of modified enamel is observed in most hypsodont to hypselodont incisors but not in the taxa with brachydont incisors (see interrow sheets on Fig.  6). As noticed in many large ungulates, the development of modified radial enamel might be associated with an increase in crown height (Pfretzschner 1992), and this innovation is assumed as a structural adaptation to radial stresses on highcrowned dentitions (Pfretzschner 1992). Nonetheless, this association does not represent an obligate condition in notoungulates given its absence in high-crowned or evergrowing incisors of Pyrotherium, Interatherium, Hegetotheriidae, and Toxodon (only for upper incisors in the two latter taxa). The angle between the IPM and prisms also tends to increase with most taxa showing low values during the Paleogene, while maximal values (i.e., 90°) were reached during the Miocene, in most late-diverging families except interatheriids (Fig. 7).

Phylogenetic Implications of Incisor Enamel Microstructure in Notoungulates
Our study shows that there is a high number of convergences in the enamel microstructure (e.g., presence of HSB, pattern and orientation of IPM) and crown height of incisors in Notoungulata (Figs. 6 and 8). These convergences are also observed for molars regarding both enamel microstructure and crown height (Maas 1997;Lindenau 2005). The distribution of enamel microstructural features could even make it easier to discriminate early diverging from late diverging notoungulates (i.e., Paleogene versus Neogene taxa) rather than to discriminate between late diverging families (i.e., To x o d o n t i d a e , I n t e r a t h e ri i d a e , H e g e o t h e r i i d a e , Mesotheriidae). This high homoplasy in the investigated characters could cast doubts on the ability of enamel microstructure traits to help clarify uncertain phylogenetic relationships among notoungulates (Cifelli 1993;Billet 2010;Reguero and Prevosti 2010), as previously noted by Maas (1997) for molars.
Nonetheless, some characters may still constitute interesting synapomorphies of several notoungulates clades. For example, Mesotheriidae represent the only clade having modified radial enamel in all incisors. Hegetotheriidae depart from the rest of notoungulates in showing distinct enamel microstructure patterns on the lingual side and the labial side of incisors, especially on upper incisors. Toxodontidae and Leontiniidae present an unusual pattern involving furcated HSB and strong undulation of the EDJ. Consequently, the different levels of complexity of incisor enamel (Koenigswald and Clemens 1992;Koenigswald et al. 1993), in connection with results from molar enamel (Pfretzschner 1992;Lindenau 2005), should be carefully considered when searching for phylogenetically informative characters in

Comparisons with Closest Extant and Extinct Relatives
Notoungulata, together with Litopterna (and possibly Astrapotheria; Billet 2010), were reported to be more closely related to Perissodactyla than to any other extant taxon of placentals (Welker et al. 2015;Buckley 2015;Westbury et al. 2017). Among perissodactyls, there are three different configurations of HSB observed for incisors : transverse HSB (the HSB are parallel to the occlusal surface and the base of enamel crown), vertical HSB (the HSB are perpendicular to the occlusal surface and the base of enamel crown), and compound HSB (transverse HSB in an inner layer and vertical HSB in an outer layer). Unlike perissodactyls, notoungulates having HSB only showing the transverse to oblique configurations. Koenigswald et al. (2011) stated that the transverse HSB is the ancestral configuration when HSB appeared in perissodactyls, and this pattern is still present in incisors of Equidae, as observed in notoungulates. Unfortunately, no data on incisor enamel microstructure are available for Litopterna to date, but a few data on Astrapotheria incisors are presented in Lindenau's thesis (Lindenau 2005). The investigated lower incisors of Astrapotheria present a different condition in having vertical HSB (Lindenau 2005), a configuration defined as derived in perissodactyls, and only found in molars and incisors of some Rhinocerotoidea (Koenigswald et al. 2011).
The transverse HSBs have the optimal configuration to withstand the tensile stresses and reduce crack propagation in being parallel to the occlusal surface, but they are less resistant to abrasion (Pfretzschner 1988;Koenigswald et al. 2011). Transverse HSBs are observed in equids (for both incisors and molars ;Pfretzschner 1993;Koenigswald et al. 2011), in which acquisition of high-crown dentitions combined with modified radial enamel probably allowed reduction of the effect of intense abrasion (Pfretzschner 1994). Interestingly, this combined acquisition of high-crowned incisors and modified radial enamel is convergently observed in many notoungulates (for both incisors and molars, Lindenau 2005). Moreover, this combination associated with a reorientation of prisms (i.e., oblique HSBs, meaning that they tend to be vertical) as observed in many incisors of notoungulates, could represent an improved resistance to abrasion. In fact, when HSBs are vertical, they present an optimum prism orientation perpendicular to the occlusal surface, which is considered to be an advantage for reducing the effect of abrasion .
While aridity and volcanism increased in Patagonia and probably in other parts of South America during the mid-Cenozoic, South American ungulates, and especially those eating close to the ground and/or with a strong chewing effort, probably ingested high quantities of exogenous abrasive particles (Billet et al. 2009;Strömberg et al. 2013;Madden 2015;Gomes Rodrigues et al. 2017; see also Janis 1988Janis , 1995Mainland 2003). In that context, various dental configurations with different strategies against abrasion emerged among these ungulates: (1) notoungulates with transverse to oblique HSB and very high-crowned incisors and molars, as partly observed in equids, (2) and astrapotheres with vertical HSB with lower-crowned incisors and molars, as observed in rhinocerotoids (at least for molars).

Comparison with Extant Analogues and Morpho-Functional Hypotheses
Certain enamel structural characteristics may have evolved in response to high forces applied during the mastication of specific food items or exogenous particles. For example, Shimizu and Macho (2007) showed that the development of an undulated EDJ in some extant primates evolved in response to high masticatory forces. However, these authors also suggested that the equivocal relationship between undulation and presumed bite force could in fact represent exaptations within primates (see Shimizu and Macho 2007). In some toxodontian notoungulates, the EDJ undulates and is accompanied by prolonged growth of incisors and a three-layered schmelzmuster (IRE-HSB-ORE; Fig. 6). However, most notoungulates (Pyrotherium, Eurygenium, Protypotherium, Interatherium, Mesotheriidae, and Hegetotheriidae) present a prolonged to continuous growth of incisors and a threelayered schmelzmuster without the undulation of the EDJ. In our sample, the undulated EDJ was found only within the families Leontiniidae and Toxodontidae. This may represent a single evolutionary event given the possible close relationship between these taxa (Billet 2011) and this undulation may as in primatesbe an exceptional feature. The functional importance of this undulation is difficult to evaluate and it could also be investigated whether the development of this structure could be linked to large body masses in notoungulates, such as found in Leontiniidae and Toxodontidae (body mass estimates by Croft 2000;Gomes Rodrigues et al. 2017; Leontiniidae: 50-210 kg and Toxodontidae: 10 kg -1650 kg).
The development of continuous growth in notoungulate incisors, especially in late diverging families, is a major morphological innovation, which was often accompanied by changes in the enamel microstructure (HSBs, modified radial enamel). It can also be accompanied by a reduction of the number of incisors, especially for uppers (e.g., Hegetotheriidae, Mesotheriidae) or the loss of enamel on the lingual side of upper incisors (e.g., Trachytherus, Toxodon). The reduction in the number of incisors toward one prominent pair of upper incisors is a trend also observed in extant mammals that present hypselodont incisors (e.g., rodents, rabbits, hyraxes). This pattern can confer a higher bite force at the incisor level as demonstrated for rodents (Druzinsky 2010;Cox et al. 2012), and this may have been the case for rodent-like notoungulates (Hegetotheriidae, Mesotheriidae). However, the complexity of incisor enamel microstructure observed in rodents (Wahlert and Koenigswald 1985;Martin 1997;Kalthoff 2000Kalthoff , 2006Koenigswald 2004;Koenigswald and Kalthoff 2007;Gomes Rodrigues et al. 2013;Gomes Rodrigues 2015) is much higher than that observed in notoungulates. If we except modifications associated with crown height increase (i.e., modified radial enamel), the loss of lingual enamel observed in rodents is only reported in a few notoungulates (Notostylops, Toxodontidae, Protypotherium, Trachytherus), and this does not involve rodent-like species (i.e., Hegetotheriidae, Mesotheriinae). Consequently, these important microstructural differences hamper comparisons of notoungulate incisors with those of rodents from a functional viewpoint.
The loss of lingual enamel could be associated with the category Benamel-band hypsodonty^(Koenigswald 2011), which comprises hypsodont teeth that have enamel only on one or two sides, while the dentine surface covers more than one third of the circumference of the tooth. Stiff enamel is less favored in incisors that need to be more elastic to prevent breakage, as observed in rodents and a few primates (Druzinsky et al. 2012;Kupczik and Chattah 2014). The selection of the loss of lingual enamel is Fig. 8 Comparison of the distributions of incisor crown height states, body mass (Croft 2000;Gomes Rodrigues et al. 2017), and parsimonious reconstruction of evolution of prism decussation and IPM characteristics on the composite tree of notoungulates (Billet 2011;Shockey et al. 2016). Black branches represent an absence of data for the respective character associated with balanced wear (Koenigswald 2011), and is in turn probably correlated to the specific functional roles of incisors (e.g., gathering food, foraging, digging, fighting). It is also known that the loss of enamel on one side of incisors has evolved independently in various mammalians lineages, and this character is sporadically present in notoungulates. This raises questions on why some species kept enamel on both sides or on only one side.
As mentioned above, incisors may have a large variety of functional roles. Previous investigations have shown that some behavior for food intake (e.g., cutting, shearing, gnawing) or digging can shape the morphological characteristics of incisors in mammals (the incisors are protruding anteriorly, they have long Broots^and are thick; Becerra et al. 2012). In notoungulates, while fossorial habits have been proposed for rodent-like Hegetotheriidae  and Mesotheriidae (Shockey et al. 2007), it is not established whether or not their incisors could have served as digging tools. Both closely related families present derived morphologies of their incisor enamel microstructure, with the presence of HSB and inner modified enamel. These features are also convergently found within non-digging notoungulates such as toxodontids. Thus, no clear link exists between a putative fossorial lifestyle and such specialization of the enamel microstructure within our sample.

Conclusions
This study shows that notoungulates exhibit diverse enamel microstructural patterns for their incisors, with several features being convergently acquired throughout their evolutionary history. More precisely, the following patterns along with the incisor crown height (regardless of its position) could be highlighted: most brachydont taxa only present radial enamel, which generally correspond to early diverging taxa; most hypsodont to hypselodont taxa present inner and outer radial enamel with intermediate IPM or completely modified radial enamel and transverse to oblique HSBs; incisors with modified radial enamel are found only in post-Eocene taxa. The latter characters often represent convergent acquisitions among late-diverging families, and may have evolved in response to various biomechanical stresses, such as increased abrasion.
The diversity of enamel patterns observed in notoungulate incisors at different levels of microstructural complexity will be of interest for phylogenetic reconstructions and characterization of early to late-diverging taxa, including also cheek teeth. Further analyses are necessary to comment on the potential roles that other factors such as development, body size, and masticatory function may have on enamel microstructure evolution in mammals and in notoungulates in particular.