Extreme insular dwarfism evolved in a mammoth: Supplementary Information
Victoria L. Herridge & Adrian M. Lister
S1. Taxonomic history and synonymy of Cretan dwarf elephants
The taxonomic history of Cretan dwarf elephants is summarized in Figure S1, showing the synonyms of Mammuthus creticus to be Elephas creticus, Elephas antiquus race melitensis, Palaeoloxodon creticus, Loxodonta creticus and Hesperoloxodon antiquus falconeri. There are two distinct issues in the generic attribution of dwarf elephants: ancestry and nomenclature. For Sicily, Malta, Cyprus and Tilos, nomenclatorial issues predominate, reflecting the debate over the validity and usage of the genus-name Palaeoloxodon. Elephas was used synonymously with Mammuthus and Palaeoloxodon well into the 20th Century, and continues to be used for European Palaeoloxodon by many researchers to this day (e.g. Ambrosetti 1968, Bonfiglio et al. 2002, Palombo 2001a, 2007). However, Palaeoloxodon has been shown to be a monophyletic clade, including P. antiquus, P. namadicus, P. naumanni and the P. recki group (Ferretti 2008), and is accepted here as a valid genus (following Inuzuka & Takahashi 2003, Shoshani & Tassy 2005).
For Cretan taxa, putative ancestry has played a much larger role in recent taxonomic revisions (Figure S1). With the exception of Kuss (1965, 1966), who believed the African elephant genus Loxodonta should also be considered, the majority of authors have considered Mammuthus or Palaeoloxodon as the likely progenitor of Cretan dwarf elephants (e.g. Mol et al 1996, Poulakakis et al. 2002a). These taxa are the only elephantid genera present in Pleistocene Europe. P. antiquus has been generally accepted as the ancestor to all Mediterranean dwarf taxa since Pohlig (1893) hypothesized the shared ancestry of Maltese and Sicilian dwarf taxa, synonymizing the three existing taxa (‘E’. melitensis[1] Falconer 1867, ‘E.’ falconeri Busk 1867 and ‘E.’ mnaidriensis Adams 1874) to ‘E. antiquus melitae’. Taxonomic affinity with Palaeoloxodon for Maltese, Sicilian, Cypriot and Tilos dwarf elephants, and for P. antiquus creutzburgi from Crete, has been established on morphological grounds, with these taxa exhibiting diagnostic Palaeoloxodon features (e.g. relatively narrow, high crowned molars; lozenge or ‘cigar-shaped’ enamel loops with mesial expansions; distinctive early occlusal wear pattern, and – in Sicilian ‘P. mnaidriensis’ – expanded parietal bosses and a well developed parietal-occipital crest, or frontal torus) (Bate 1905, Osborn 1942, Ambrosetti 1968, Theodorou 1983, Poulakakis et al. 2002a, Palombo 2003; Ferretti 2008).
In contrast, evidence for the affinity of ‘E.’ creticus to P. antiquus has been circumstantial. Bate (1907) documented the morphological similarity of ‘E.’ creticus to Mammuthus meridionalis, noting in particular the low crowned nature of the teeth, but nevertheless assigned ancestry to P. antiquus (Figure S1). This was in part due to a contemporary belief that P. antiquus was characterised by ‘adaptability’ and a propensity to dwarf on islands (Bate 1907). Bate had also found material referable to full-sized P. antiquus on Crete, and believed that it was not possible for the island to have maintained two contemporaneous species of elephant: thus with evidence of P. antiquus on the island, she reasoned it must be the ancestor of ‘E.’ creticus.
Recently, there has been a resurgence of interest in the possibility of Mammuthus ancestry for some dwarf elephants. Lister and Bahn (1994) posited a possible Mammuthus ancestry for Sicilian ‘E.’ falconeri, based on skull morphology. This suggestion was later dismissed by the authors (and removed from the 2nd and 3rd editions of the same book: Lister & Bahn 2000, 2007), but was re-iterated by Mol et al. (1996), who also proposed Mammuthus ancestry for ‘E.’ creticus (Figure S1) on grounds of its antiquity (see main text). Poulakakis et al. (2002a, 2006) re-stated Mol et al.’s arguments in support of Mammuthus ancestry for ‘E.’ creticus and ‘E.’ falconeri. Chronological arguments are not sufficient to merit taxonomic revision, even without reference to the large errors and uncertainties surrounding the dating of dwarf taxa (particularly on Crete; see main text).
Poulakakis et al. (2006) reported aDNA evidence for a Mammuthus, rather than Palaeoloxodon, affinity for ‘E.’ creticus (Figure S1). The credibility of this study was called in to question owing to ‘serious theoretical and methodological flaws’ (Binladen et al. 2007, p.56): (i) two of the three nucleotides identified by Poulakakis et al. as ‘diagnostic’ for Mammuthus lay within the original primer-binding site used to amplify the aDNA fragment (Binladen et al. 2007), (ii) all three ‘diagnostic’ sites were found within GenBank sequences of the African elephant Loxodonta spp., and are not Mammuthus autapomorphies (Binladen et al. 2007, Orlando et al. 2007), and (iii) Poulakakis et al.’s phylogenetic analysis identified Mammuthus and Loxodonta as sister-taxa, at odds with the elephant phylogeny based on whole mitochondrial and nuclear genome data (in which Elephas and Mammuthus are sister-taxa; e.g. Krause et al. 2006; Rogaev et al. 2006; Capelli et al. 2006; Miller et al. 2008), suggesting their results would not be robust to the addition of more aDNA data (Orlando et al. 2007). The low likelihood of aDNA being recovered from 800 ka material preserved in a warm environment also cast doubt on the legitimacy of the claim (Smith et al. 2003, Binladen et al. 2007).
Further criticisms of Poulakakis et al. (2006) can be made: (i) DNA was extracted from a non-diagnostic fragment of rib bone collected in the vicinity of the area that ‘E.’ creticus was excavated (Bate 1907, Poulakakis et al. 2006), but not definitely referable to ‘E.’ creticus, and (ii) in responding to criticisms of the likelihood of DNA amplification from 800 ka material, Poulakakis et al. (2007, p. 60) stated “the bone was retrieved from a cave, where environmental conditions remain relatively constant”. If this is the case, then the sampled material, attributable to ‘E’. creticus only on the basis of its provenance, was not excavated from the type locality of ‘E.’ creticus, which is now open coast line (Bate 1907; Figure S2). In light of all these criticisms, the ancestry of ‘P’ creticus remains an open question.
Figure S1. Systematic history of dwarf elephant taxa from Crete. [A] Hypothesized mainland ancestor used for generic attribution of M. creticus. Before Osborn (1942) Elephas was widely used for all Elephantinae taxa. S denotes a change in genera/ancestral taxa based on systematic opinion, N denotes a nomenclatorial change, e.g. author-preference for Elephas vs Palaeoloxodon. [B] Year and Author of publication. [C] Synonymy of dwarf elephant taxa: lines connect Linnaean binomials that refer to the same taxon. Brackets indicate lumping and splitting of taxa by different authors. [D] Locality of material used in new description of taxa or systematic revision
Figure S2. Cape Malekas, Crete. View east from lower level of fossiliferous deposit identified by us (VH and George Iliopoulos, U. Patras) as probable type locality for M. creticus. Photocredit: Adrian Glover, 2007
S2. Methodological details & discussion
Morphometric data were taken with callipers accurate to 0.05mm (measurements <150mm) or 1mm (<30cm), or a measuring tape/osteometric board accurate to 1mm (measurements >30cm, e.g. larger humeri).
Molar Width (W)
Molar width (W) excluded cement, and was calculated from the total maximum width along the tooth minus the observed cement thickness at that point. This contrasts with earlier protocols followed by AL, where reported molar width includes an estimate of cement lost or abraded based on known average cement thickness for that population or taxon. We could not make these assumptions for dwarf taxa a priori, and most dwarf M3s lacked cement. Previous data of AL have been revised accordingly, ensuring equivalence across our tooth width measures.
Maglio (1973) specified molar width should be inclusive of cement, but protocols for other published data do not include any discussion on the inclusion or exclusion or cement in molar width, nor degree of correction applied to those widths, and thus we cannot be sure of the equivalence between our measures and the literature data for insular mammoths. This may also mean that the lower widths in our data may exaggerate size-differences between our measurements and those from the literature. However, we consider this effect not to prejudice our results (effects on HI and RLI dealt with below). First, cement thickness tends to be low in smaller elephant taxa (pers. obs.). Second, in relation to molar size in M. lamarmorai and M. exilis, their likely ancestors are M. trogontherii and M. columbi, respectively. Mean cement thickness in M. trogontherii is 6.15 mm, and M. columbi is very similar in molar morphology. Thus, even if it is assumed that literature data on these dwarfs includes cement thickness equivalent to the mean value of their larger-sized ancestor (which we consider unlikely), subtracting that thickness shows that M. creticus mean W (minus cement) is still smaller than the M. exilis mean and the single data point for M. lamarmorai.
Enamel Thickness (ET)
Enamel thickness (ET) is either the mean of 10 measurements taken at different points across the occlusal surface (for P. antiquus, P. falconeri, P. cypriotes and ‘P’. creticus, collected by VH), or a modal value of 3-10 measurements (M. meridionalis and M. trogontherii, collected by AL).
Lamellar Frequency (LF)
Lamellar frequency (LF) is calculated following Maglio (1973) as modified by Albayrak & Lister (2011): LF=[(NBI/LBI)+(NBO/LBO)]/2, where NBI and NBO are the number of plates counted in a well preserved, central portion of the inner and outer sides of the tooth, respectively; and LBI and LBO the length (in mm) across those plates, taken at the base of the tooth.
Hyposodonty Index (HI)
To increase the hypsodonty index sample size (where observed cement thickness was not recorded for 5/6 specimens of M. trogontherii with overall width and crown height data collected by AL), an average cement thickness (6.15mm) for that species was calculated from the total dataset (n=17), and subtracted from measured overall W of those for which it had not been recorded. In any case, intraspecific distribution of HI, and comparisons of M. trogontherii with other full-sized taxa, were similar using uncorrected and corrected W.
M. rumanus width measures are taken from the literature and are not corrected for cement thickness. However, cement thickness can be estimated from published images: ~6mm (max) for the Rendlesham Red Crag specimen (Lister & Van Essen 2003); ~11mm for the Felixstowe Reg Crag specimen (Pontier 1913; only included here for LF and PC); <10mm for the Bossilkovtsi specimen (Markov & Spassov 2003). Furthermore, the Tuluceşti ̧type specimen (not included in metric data set owing to advanced wear state/incompleteness) has a cement thickness of up to 7mm (H. Van Essen, pers. comm. to AL). This equates to a mean cement thickness of 8.5mm (n=4), consistent with the average cement thickness in M. meridionalis (8.30mm, n=22). We chose not to adjust the published measures using this average figure given the uncertainty in our estimates, and as such M. rumanus HI and RL may appear artificially low in comparison to other taxa. However, recalculating width measures using the average estimated cement thickness (table S1), shows that M. rumanus still falls within the M. meridionalis range for both HI (M. m. range: 116-160) and RLI (M. m. range 280-356). Thus we would obtain similar results using both uncorrected and corrected width measures: M. creticus tooth shape resembles that of both M. rumanus and M. meridionalis.
Specimen / W+C (mm) / W (mm) / CH (mm) / L (mm) / HI / RLI / HI(w-c) / RLI (w-c)R.955-12.10A-B / 75 / 66.5 / 90 / 226 / 120 / 301 / 135 / 340
594 / 95 / 86.5 / - / 260 / - / 274 / - / 301
OF 37797 / 106 / 97.5 / 142 / 306 / 134 / 297 / 146 / 314
Table S1. Hypsodonty and relative length indices of M. rumanus calculated for both width measures inclusive of cement (W+C), and width minus av. estimated cement thickness of 8.5mm (W). R.955-12.10A-B from Rendlesham Estate, UK (Lister & Van Essen 2003); 594 from Bossilkovtsi, Bulgaria (Markov & Spassov 2003); OF 37797 from Novotroitsk, Russia (Baygusheva & Titov 2012).
Distinguishing full-sized Palaeoloxodon & Mammuthus M3s
Mammuthus and Palaeoloxodon differ in the following non-metric characters of the occlusal surface: the shape of the enamel loop visible on the occlusal surface, the early wear pattern of that loop (Maglio 1973), and the existence and shape of any enamel ‘expansions’ in the mesial region of the enamel figure (“mesial expansions”). Metrically, continental Mammuthus are distinguished from P. antiquus on the basis of molar size and shape (Osborn 1942, Maglio 1973), with P. antiquus having narrow M3s which are relatively tall and long, especially in relation to M. rumanus and M. meridionalis (table 1; figure 2; S5), but these characters do not provide reliable genus identifiers (especially when trying to delineate Palaeoloxodon from more derived Mammuthus species). The number of molar plates (Plate Count, PC), also known as lamellae, can delineate P. antiquus from M. rumanus and M. meridionalis, but not from M. trogontherii (figure 2; S5). Lamellar frequency (LF: average no. plates per 100mm of molar length) overlaps considerably among Palaeoloxodon and the three Mammuthus species, although P. antiquus has a statistically lower mean LF value than M. trogontherii (S5). All four taxa also overlap in enamel thickness (ET), although mean differences in ET are apparent (P. antiquus ET < M. trogontherii ET < M. meridionalis ET , M. rumanus ET; figure 2). LF and ET are thus less useful as genus identifiers. Within Mammuthus, M. trogontherii can be distinguished from M. rumanus and M. meridionalis on the basis of PC, molar shape (relative height or hypsodonty index, and relative length) and, to a lesser extent, ET and LF, while M. rumanus and M. meridionalis differ significantly only in PC and ET (P<0.05, see ESM; figure 2).
Non-metric characters of the occlusal surface thus provide the most robust data for genus identification, given the clear distinction between Mammuthus and Palaeoloxodon in these traits. Metric characters (plate count, relative molar length, hypsodonty index and enamel thickness) are useful for species level identification and are thus potentially informative on Cretan to mainland elephant taxon affinity, although the impact of allometric change on the characters is not well understood. Lamellar Frequency (LF) is considered for completeness (it is commonly used in species-level identification of elephants), but is expected to be of limited taxonomic value in dwarf taxa: full-sized elephants have been shown to exhibit intraspecific size-related trends in M3 teeth (Lister & Joysey 1992), and through the molar ontogenetic series (dP2-M3) (Aguirre 1969).