Living mammals show remarkably consistent patterns of growth and morphology in their dentitions. Most have four readily identifiable dental types: incisors, canines, premolars (collectively known as antemolars) and molars (Maier 1984). Groups such as rodents and lagomorphs lack canines and anterior premolars. Monotremes, xenarthrans, aardvarks, pangolins, cetaceans and aquatic carnivorans further lose some or all tooth types. With very few exceptions (Domning and Hayek 1984; Gomes-Rodrigues et al. 2011), and documented by an extensive fossil record (O’Meara and Asher 2016), mammalian teeth are replaced just once, if at all. In marsupials, antemolar replacement is limited to the last premolar; other species lack premolar replacement altogether (Van Nievelt and Smith 2005), either because the deciduous (e.g., soricids) or replacement (e.g., proboscideans) generation never fully mineralizes or breaks the gum. Tooth types and deciduous vs. permanent generations are usually distinguishable in most species based on their anatomy and eruption patterns (Ungar 2010).

When postnatal anatomy is not sufficient to recognize positional or ontogenetic dental homologies, embryonic development has played a key role (Luckett and Maier 1982; Luckett 1993a, b; Van Nievelt and Smith 2005). Nineteenth and early 20th century studies of dental ontogeny extensively utilized juvenile and embryonic specimens (e.g., Tomes 1874, 1897; Leche 1907; Butler 1937), and indeed our understanding of dental development arises from such material. As summarized by Williams and Evans (1978), Luckett (1993a), Järvinen et al. (2009), and others, mammalian teeth of both the deciduous and permanent dentitions undergo the same ontogenetic stages during development: bud, cap, and bell, each subdivided into early, middle and late. The primary dental lamina in mammals has a restricted capacity to generate successional teeth, and its regression limits the dentition to two generations (Popa et al. 2016). The deciduous dentition and molars have long been regarded as arising from the embryonic primary dental lamina, connected to the oral epithelium (Owen 1845; Butler 1937). While molars are functionally part of the permanent dentition, their apparent origin from the primary dental lamina would make them, developmentally speaking, late-erupting “deciduous” teeth (Luckett 1993a; Tucker and Fraser, 2014). Replacement teeth arise from the lingual successional lamina of their corresponding deciduous precursors (Butler 1937), although origins from a primary or successionary lamina is not always obvious in cases when one or the other is vestigial (see Van Nievelt and Smith 2005, discussed further below).

The first premolar locus has regressed in nearly all placental mammals to just one generation or none at all. First premolar replacement has been documented in some fossil groups, including hyracoids (Gheerbrant et al. 2007; Asher et al. 2017) and cetaceans (Uhen 2000). Among living groups, first premolar replacement consistently occurs only in the upper dentition of Tapirus (Ziegler 1971), but may also occur in some individuals of Procavia (Asher et al. 2017). The xenarthran genus Dasypus typically exhibits seven or eight replaced, antemolar teeth in each quadrant, without obvious homologies of any of them to the antemolars of other mammals (Ciancio et al. 2012). A single generation at the first premolar characterizes many carnivorans, artiodactyls, perissodactyls, talpids, and macroscelidids. Otherwise, the first premolar locus is absent in living mammals.

Kindahl (1957, 1967) argued that the macroscelidid first premolar was a retained deciduous tooth. She collected data from five histologically prepared embryos of Elephantulus myurus of 16mm, 17mm, 19mm, 25mm, and 40mm greatest length, as well as osteological specimens. She observed in her embryos that the anterior-most premolar arose from the primary dental lamina, and never observed a successional lamina (or in later stages a replacement tooth) arising from that locus. Kindahl (1957:29) therefore concluded that “the first premolar [of Elephantulus] belongs to the lacteal set of teeth”, or dp1.

Williams and Evans (1978) suggested the opposite for domesticated dogs (beagles) based on 172 specimens of known age. Of these, 29 were histologically sectioned and 143 were cleared and stained; sizes ranged from 12–95mm CRL (histology) and 20–166mm (cleared and stained), representing developmental ages of approximately 25–63 days post coitus, corresponding roughly to the last 5.4 weeks of a 9 week pregnancy. In their sample, the primary dental lamina between tooth buds of the dp2 and canine remained undeveloped throughout most of prenatal ontogeny. Only at later stages, from 42–47 days, did signs of bud initiation corresponding to the first premolar locus appear, temporally coincident with the initiation of the permanent replacements of deciduous teeth. Williams and Evans (1978: 161) tentatively concluded that “the first premolars develop in the late fetus and are regarded here as teeth of the permanent dentition without deciduous predecessors”.

The positive evidence used by both Williams and Evans (1978) and Kindahl (1957) concerned the morphology of the primary dental lamina and timing of mineralization. On the other hand, neither Kindahl (1957) nor Williams and Evans (1978) observed specimens in possession of a rudimentary, partly formed dp1 or p1 simultaneously with its cognate. In principle, one or few observations of a developing successional tooth connected to a dp1 could alter their conclusions on the homology of the tooth at the first premolar locus.

Here, we build on the work of past authors to examine homology of the first premolar locus and test the hypothesis that the functional, first premolar locus in each belongs to separate generations: the dp1 in macroscelidids and the p1 in canids. We focus primarily on these two groups given the contrasting interpretations of their first premolar homologies, and for comparison examine other carnivorans and afrotherians, in particular hyracoids, as they are among the only mammals with documented replacement at the first premolar locus (Gheerbrant et al. 2007; Asher et al. 2017). We use microCT scans and histologically-prepared specimens to try and identify replacement at the p1 locus in growth stages that previous authors may have missed.

We furthermore consider the inhibitory cascade hypothesis (IC, Kavanagh et al. 2007; Polly 2007; Wilson et al. 2012; Evans et al. 2016) to determine if its predictions of size can distinguish between deciduous and replacement teeth at the first-premolar locus. There are many instances in which inhibitory cascade predictions depart from observed molar areas (Renvoisé et al. 2009; Asahara 2013; Evans et al. 2016; Roseman and Delezene 2019). Its applicability to deciduous premolar sizes is less well known. If it does predict areas of known deciduous premolars, then it may be useful to distinguish deciduous vs. replacement teeth at loci in which replacement has not yet been observed, such as the first premolar.

We employed two methods to test the developmental homology of teeth at the p1 locus. First, we assembled a dataset of skeletal (Table 1) and soft-tissue stained (Table 2) CT scanned prenatal and juvenile specimens that could in theory show evidence for replacement at this locus, to which we added four histologically prepared macroscelidids, one canid, and made comparisons to six hyracoids (Table 3). Second, we measured linear dimensions of deciduous and permanent generations from the lower canine to last molar to determine if tooth area can be predicted based on the inhibitory cascade model (Kavanagh et al. 2007; Evans et al. 2016). This was articulated by Evans et al. (2016) as B=0.5 (A+C), algebraically equivalent to A=2B–C and C=2B–A, where A, B, and C represent two-dimensional areas of each of three adjacent teeth (in their case, molars) derived from the primary dental lamina. If tooth area follows the inhibitory cascade, then the middle tooth of a triplet would have an area approximating the average of the areas of the first and third (B=0.5 (A+C)); the first tooth would have an area about twice the second minus the third (A=2B–C); the last tooth would have an area about twice the second minus the first (C=2B–A). Therefore, following Kindahl’s (1957) interpretation, the first premolar of Elephantulus (dp1) should have an area approximating twice that of dp2 minus dp3. We would similarly expect a dp1 of Canis to correspond more than a p1 to an area approximating twice that of dp2 minus dp3.

In order to be consistently measurable, tooth crowns have to be at or close to full mineralization. It would also be ideal to have measurements across all loci to be able to test area predictions for any tooth arising from the primary dental lamina. In reality, a full complement of molars rarely co-occurs with all deciduous teeth, and area estimates for conical, trenchant teeth are more prone to error than those from rectangular, molariform teeth. Therefore, we did not test predictions for loci anterior to the first premolar. Of the three equations noted above to predict tooth area using the inhibitory cascade, we used whichever applied to one individual’s available teeth without assuming first premolar homology. For example, to predict size of an m1, we used C=2B–A (2×dp4–dp3) unless dp3 or dp4 were missing, in which case we used B=(A+C)/2 ((dp4+m2)/2) unless dp4 or m2 were missing, in which case we used A=2B–C (2×m2–m3). For predicting dp2 area we used only A=2B–C and for dp3, either A=2B–C or B=(A+C)/2 in order to avoid using the first premolar to predict areas of other loci. In addition, because incisors lack clear landmarks with which to define length and width, we omitted incisors from our sample, using instead teeth between (and including) the canine and the last molar. We made length and width measurements based on 3D rendered CT scans (and did not try to measure teeth from histological specimens) using Drishti (versions 2.6.4 to 2.7, Limaye 2012) as per the landmarks shown in Figure 1.

We sampled postnatal and soft-tissue stained embryonic specimens across 13 species in 9 genera of afrotherians and carnivorans. Our sample of CT-scanned Canis specimens consists of multiple species and subspecies, including two coyotes (C. latrans), two Mexican wolves (C. lupus baileyi), two arctic wolves (C. lupus arctos), a wolf (C. lupus) native to Lebanon, and seven domestic dogs (C. lupus familiaris). Six of the latter are pointers, five of which have known ages. Age post-birth was also known for 16 specimens of Macroscelides (Table 1 and Asher and Olbricht 2009). While the anterior-most premolariform cheektooth in Procavia has sometimes been interpreted as a canine (Luckett 1993b), based on data from Asher et al. (2017) and Gomes-Rodrigues et al. (2019) we regard it as a first premolar. All of the other genera are uncontroversially known to typically possess four permanent premolars. Most also have three molars, except for Otocyon (with four) and Macroscelides, Nandinia, Nasua and Viverra (with two). For soft-tissue stained CT scans (Table 2), we followed the PTA (phosphotungstic acid) protocol described by Metscher (2009: table 2). Supplementary appendix S1 provides measurements and details on individual specimens in csv format.

Our sample is limited by the number of specimens that simultaneously possess a measurable first premolar along with adjacent deciduous teeth that allow for calculation of dp1 area according to the inhibitory cascade. This has the advantage of not averaging across specimens that may vary in size due to unrelated factors (e.g., variation among breeds), but on the other hand reduces our sample size. For example, Canis is a common species, well-represented in museum collections and the literature, but even when developmental series are available, they rarely exhibit specimens with a fully mineralized first premolar crown along with measurable dp2 and dp3. Hence, we have just one Canis specimen in our sample providing both predicted and observed areas for the first premolar, as detailed below.

Throughout the text, we use dental abbreviations that correspond with our results and those of Kindahl (1957) and Williams and Evans (1978). That is, we describe macroscelidids with a retained dp1/dP1 and Canis with an unreplaced p1/P1. Individual teeth are abbreviated with “i” for incisor, “c” for canine, “p” for premolar, “m” for molar, the prefix “d” to indicate deciduous antemolars, and (unless specified otherwise) upper and lowercase for upper and lower teeth (respectively). Numbers follow serial homologies as used by Butler (1937) and illustrated by Asher (2019).

Homology determination based on origin of a given locus from the primary dental lamina has been considered sufficient evidence for identification as a deciduous locus. Indeed, as originally proposed (see Luckett 1993a), origin from the primary and not successional dental lamina is the definition of a deciduous tooth. Based on this criterion alone, the first premolar locus in Canis should be a dp1. Two lines of ontogenetic evidence arising from our dataset, also reflected in the conclusions of Kindahl (1957) and Williams and Evans (1978), lead us to regard this definition of deciduous teeth as insufficient: 1) the relatively early appearance of an embryonic first premolar toothbud in macroscelidids compared to Canis and 2) eruption of the functional tooth at the first premolar locus after mineralization of other functional teeth in Canis (Fig. 9A) and before in macroscelidids (Fig. 9B).

Van Nievelt and Smith (2005) identified permanent antemolars in the didelphid marsupial Monodelphis as successional, not deciduous. As in other marsupials, its only mineralized deciduous tooth is the last premolar or dp3, from which a successional lamina develops into the replacement p3. Like the tooth at the first premolar locus in Canis, other antemolars in Monodelphis appear to develop from the primary dental lamina without obvious successional laminae. Van Nievelt and Smith (2005: 338) wrote that “the apparent absence of even transitory, vestigial teeth in [Monodelphis] makes unambiguous identification of the generational homology of these permanent teeth difficult.” A lack of an obvious origin from a successional lamina in Monodelphis “would imply that the permanent teeth in [Monodelphis] are not homologous with those of [other] marsupials.” Yet marsupials such as Dasyurus (Luckett and Wooley 1996) and Caluromys (Van Nievelt and Smith 2005) show bud-stage rudiments in the primary dental lamina of at least some loci, from which successional laminae develop to eventually form the functional teeth. Hence, rather than identify permanent antemolars in Monodelphis as deciduous and Dasyurus as successional, Van Nievelt and Smith (2005:338) argued that “it is far more parsimonious … to assume that in some marsupial taxa … the vestigial first generation seen in other marsupials has been lost entirely. Therefore, the functional, permanent teeth observed in [Monodelphis] can most reasonably be interpreted as being homologous with the functional permanent teeth observed in other marsupials.” This implies that as-yet unobserved clusters of transient odontogenic cells in Monodelphis actually do appear during ontogeny, from which their replacements arose as a successional lamina. We agree with this interpretation.

The comparative anatomy and developmental timing of odontogenesis in macroscelidids and Canis supports the identification of their first premolars as, respectively, dp1 and p1. Our histologically prepared and CT-scanned specimens of macroscelidids, at the relevant developmental stages, consistently show a tooth developing from the primary dental lamina that corresponds to the dp1, also evident in Procavia. As Kindahl (1957) noted, at no point do we observe a successional lamina arising from the macroscelidid first premolar locus. As Williams and Evans (1978) noted among Canis specimens, at no point do we observe a toothbud arising from the primary dental lamina that likely corresponds to a first deciduous premolar. Moreover, there is a substantial difference in the timing of post-birth mineralization in sengis compared to Canis, differences which correspond to the homology inferences made by Kindahl (1957) and Williams and Evans (1978). The first premolar in sengis mineralizes only slightly after the other deciduous teeth, and the mineralized apex of the dp1 crown is evident in CT scans of newborn specimens (Fig. 4A). No replacement teeth appear adjacent to their deciduous precursors during the first two weeks of development in Macroscelides (Figs 4, 9). At 16 days, Macroscelides with a nearly-erupted first premolar shows no sign of mineralized antemolars and just the apices of the m1 are evident in its crypt (Fig. 9B). Even at 38 days post-birth (Fig. 4D), when the dp1 is fully erupted, replacement antemolars are still not fully mineralized and at 69 days they are still not fully erupted (Fig. 5A).

Canis, in contrast, shows no sign of any mineralized tooth at the p1 locus until over two-weeks post-birth (Fig. 8A). At 18 days, the p1 and permanent canine have begun to mineralize. When it is finally mineralized and partly-erupted at 33 days, the first premolar is surrounded by mineralized crowns of most permanent teeth in a pointer specimen that is of uncertain age but likely more than 33 days post-birth (Figs 8C, 9A). These patterns of mineralization and eruption support the conclusions of Kindahl (1957) that the first premolar in macroscelidids belongs to the deciduous generation and is never replaced, and of Williams and Evans (1978) that the first premolar in Canis is a successional tooth without a deciduous predecessor.

While the anatomical data are reasonably clear, there is nonetheless still a possibility that a key ontogenetic stage is missing that might overturn these conclusions about homology. The inhibitory cascade hypothesis has previously been discussed in terms of the posterior-most teeth of the primary dental lamina, the molars, and has been able to predict molar areas in a number of cases (Kavanagh et al. 2007; Evans et al. 2016). Our application of the IC model to more anterior loci is novel and based on the possibility that IC area estimates could help resolve the question of p1 homology. At least as quantified here, it does not. Neither Macroscelides nor Canis exhibited first premolar crown areas expected by the inhibitory cascade model; indeed, for some specimens of the former, predicted areas based on 2B-C (i.e., twice the area of dp2 minus dp3 to infer predicted area for dp1) yielded values at or below zero (Fig. 12).

With the linear representations of crown size used here, hyracoids deviate less from inhibitory cascade predictions across their toothrow compared to macroscelidids and carnivorans (Figs 13, S1). The largest hyracoid in our sample, the fossil Thyrohyrax litholagus, exhibits particularly close matches of predicted to observed areas over most of its toothrow (Fig. 13). This is not simply due to reduced measurement error in larger vs. smaller specimens, as taxa with comparably-sized or larger teeth (e.g., some loci in Saghatherium and Canis) depart more from predicted areas than does T. litholagus. However, even in Thyrohyrax, most individuals have dp1s outside of the range predicted by the inhibitory cascade (Figs 11, 12). Of course, areas calculated here using simple rectangles do not do justice to the diversity of crown size and shape, and improved methods for representing this diversity may yield different results. Until such methods are forthcoming, and at least for our sample, the inhibitory cascade hypothesis does not reliably distinguish first premolar homology based on predicted areas.

It is still worth noting that the four of the five taxa with the closest matches of observed to predicted first premolar area (Thyrohyrax, Procavia, Nasua, and Saghatherium; see Fig. 13) show, on average, steadily increasing areas from anterior to more posterior deciduous and molar cheek teeth. The two most widely divergent estimates (Viverra and Macroscelides) correspond to taxa with a substantial change in the midst of the toothrow, with dp4<m1>m2 in Viverra and dp3<dp4>m1 in Macroscelides (as noted in hominins by Evans et al. 2016). The possibility that selection acting on such size differences in the midst of a toothrow could override IC-predictions is worth further investigation.

In summary, and in contrast to our application of the IC model to infer generational homologies of mammalian teeth, anatomical data over the course of ontogeny support the first premolar homologies previously inferred for macroscelidids and Canis. Our data are consistent with the conclusions of Kindahl (1957) that macroscelidids retain a functional dp1 into adulthood which is never replaced. They also support the interpretation of Williams and Evans (1978) that canids erupt a successional tooth at the p1 locus which lacks a deciduous precursor.

RJA conceived of the project, collected and analyzed data, and wrote the paper. CJM contributed to project conception, collected and analyzed data, and helped to write the paper. CW, ES, HS, JL, SK, and NB contributed data, analysis and helped edit the text.