Introduction

The Polish Plio-Pleistocene mammal bearing localities provided a huge amount of fossil talpid remains (Rzebik-Kowalska 2005). Skoczeń (1976, 1980, 1993) described five new species belonging to extant genera currently endemic of North America: Condylura kowalskii Skoczeń, 1976, Condylura izabellae Skoczeń, 1976, Parascalops fossilis Skoczeń, 1993, ?Neurotrichus polonicus Skoczeń, 1980 and ?Neurotrichus minor Skoczeń, 1993. Storch and Qiu (1983) suggested the inclusion of ?Neurotrichus polonicus in the genus Quyania, but, due to the lack of the upper and lower antemolar rows they maintained the generic status given by Skoczeń (1980). They suggested that the Polish species is inserted in an ancestor-descendant lineage in relationship with Quyania chowi Storch and Qiu, 1983, hypothesizing a lineage characterized by a gradual reduction of the precingulid, the strenghtening of the upper molar protoconules and size increase. The description of the small species ?Neurotrichus minor raised the question by the large size as an advanced evolutionary character (Skoczeń 1993). Popov (2004), following the hypothesis of Storch and Qiu (1983), assigned the material from Varshets (Early Pleistocene, Bulgaria) to Quyania aff. Q. polonica. Popov (2004) considered the Polish species as more advanced than Neurotrichus gibbsii by having reduced precingulids and a humerus more adapted to a fossorial lifestyle. Rzebik-Kowalska (2005), maintained the original taxonomic identification provided by Skoczeń (1980). Dalquest and Burgner (1941) described the extant North- American subspecies Neurotrichus gibbsii minor which is still considered valid. Therefore Zijlstra (2010) proposed the new name Neurotrichus skoczeni. Rzebik-Kowalska (2014) pointed out that the generic attribution of ?Neurotrichus polonicus still represents an open question. Her revision showed that the Polish species displays characters shared by both genera Neurotrichus and Quyania. Rzebik-Kowalska (2014) left the generic attribution given by Skoczeń (1980) considering the attribution to the genus Quyania as still im motivated. Although the generic attribution of these species has been questioned (Storch and Qiu 1983; Popov 2004; Rzebik-Kowalska 2014), no new analyses or diagnoses have been provided till now. Here we re-examined the material previously attributed to ?Neurotrichus polonicus and ?Neurotrichus skoczeni and provided a new generic diagnosis in the light of the most recent studies on talpid morphology and evolution (Gambaryan et al. 2003; Sánchez-Villagra et al. 2004, 2006; Piras et al. 2012). We also investigated the patterns of shape and size variation of the humerus by means of Geometric Morphometrics Analysis. The humerus experienced the most remarkable transformations during talpid evolution (Dobson 1882; Freeman 1886; Reed 1951; Yalden 1966; Sánchez-Villagra et al. 2004; Piras et al. 2012). This skeletal element is usually found well preserved and abundant in fossil assemblages. It is thus widely used in systematic studies of extinct Talpidae (Ziegler 2003) and, due to its abundance, allows the use of modern multivariate and univariate statistical methods.

Institutional abbreviations.—LACM, Los Angeles County Museum, Los Angeles, USA; MF, Institute of Systematic and Evolution of Animals, Kraków, Poland; NMNS, National Museum of Nature and Science, Tsukuba, Japan; V, Institute of Zoology, Bulgarian Academy of Sciences, Sofia, Bulgaria; ZPAL, Institute of Paleobiology, Polish Academy of Sciences, Warszawa, Poland.

Other abbreviations.—ANOVA, univariate analysis of variance; bgPCA, between group Principal Components Analysis; CS, Centroid Size; GM, Geometric Morphometrics; MANOVA, multivariate analysis of variance; UPGMA, Unweighted Pair Group Method with Arithmetic Mean.

Material and methods

Specimens collection.—We analyzed a total of 48 left humeri belonging to Urotrichus talpoides Temminck, 1841 (n = 12), Dymecodon pilirostris True, 1886 (n = 8), Urotrichus dolichochir Gaillard, 1889 (n = 5), Quyania chowi Storch and Qiu, 1983 (n = 2), Neurotrichus gibbsii Baird, 1856 (n = 16), Rzebikia polonica Skoczeń, 1980 (n = 6). We included in the analysis all the Late Neogene Neurotrichini and Urotrichini species for which complete humerus was available. For details about specimen codes, localities, and collection storage see SOM 1, Supplementary Online Material available at  http://app.pan.pl/SOM/app61-Sansalone_etal_SOM.pdf.

Geometric morphometrics.—The humeri have been photographed in caudal view at a fixed distance of 50 cm with a Nikon D100 camera with a Micro-Nikkor 105mm lens. We digitized 21 landmarks and 15 semi-landmarks (Fig. 1) using the tpsDig2 software (Rohlf 2006). Semi-landmarks are a useful tool to capture the morphology of complex outlines due to the lack of homologous anatomical points. They assume that curves or contours are homologous among specimens (Adams et al. 2004; Perez et al. 2006). Thus, semi-landmarks are useful to depict the shape of curved lines where landmarks cannot be detected. Successively, a Generalized Procrustes Analysis (GPA; Bookstein 1991; Goodall 1991) implemented in the procSym() function from R-package “Morpho” (Schlager 2014) was used to rotate, translate and scale landmark configurations to the unit centroid size (CS = the square root of the sum of squared distances of a set of landmarks from their centroid; Bookstein 1986). Rotation of the scaled and translated landmark sets starts by comparison with a reference configuration (usually the first specimen in the dataset). Once the first rotation is completed, a mean shape is calculated and the rotation process is repeated using the mean shape as the reference configuration for the sample (including the reference-specimen configuration). This mean shape/rotation procedure is iterated to minimize rotation differences between subsequent iterations through a least-square procedure (Rohlf and Slice 1990). The residual differences correspond to real shape differences plus measurement error. In order to visualize the ordination of the aligned specimens we performed a between group PCA (bgPCA), using the function groupPCA() included in the R-package “Morpho”. The bgPCA provides a projection of the data onto the principal components of the group means, leading to an ordination of the shape variables between the group means. The new axes are orthogonal and can be computed even when the data are not of full rank, such as for Procrustes shape coordinates (Mitteroecker and Bookstein 2011). This method offers a good performance when the number of observations is smaller than the number of variables (Boulesteix 2005), which is often the case for GM analyses. The significance of the observed shape differences among species was evaluated by performing a permutational multivariate analysis of variance (perMANOVA) on Procrustes coordinates using adonis() function included in the “vegan” R package (Oksanen 2013). The significance of shape differences between species was then evaluated performing a pairwise permuted MANOVA using the pwpermanovac() wrapper function, available in SOM 4. Size variation was visualized using a boxplot. The significance of size differences has been evaluated by performing a permutational univariate analysis of variance (perANOVA) on CS using the function adonis(). Between species size differences were evaluated performing a pairwise permuted ANOVA using the wrapper function pwperanovac(), available in SOM 4. All p-values were corrected using “Holm” correction. We excluded Quyania chowi from all the pairwise permuted comparisons due to its small sample size (n = 2). The phenetic relationships among the taxa included in this study have been visualized performing an UPGMA on the Euclidean distance matrix computed on per-species mean shape variables.

Fig. 1.

Landmarks (black circles) and semilandmarks (white circles) digitized on the humerus in caudal norm: 1, lateral end of greater tuberosity; 2, articular facet for clavicula; 3, proximal edge of the articular facet for clavicula; 4, bicipital notch; 5, proximal end of lesser tuberosity; 6, medial edge of the minor tuberosity; 7, lateral edge of the lesser tuberosity; 8, bicipital ridge; 9, middle point of the bicipital tunnel; 10, lateral end of the scalopine ridge; 11, proximal end of the teres tubercle; 12–14, surface of the teres tubercle; 15, distal end of the teres tubercle; 16–18, minor sulcus; 19, posterior margin of the lateral epicondyle; 20–22, lateral epicondyle; 22–24, trochlear area; 25–27, medial epicondyle; 28, posterior margin of the medial epicondyle; 29–32, greater sulcus; 33–36, humeral head.

Systematic paleontology

Fig. 2.

Photographs showing the different conditions of the bicipital tunnel (arrowed) in talpid mammals. A. Rzebikia polonica (Skoczeń, 1980) gen. nov. (MF/1020/1), Early Villanyian (MN 16) of Poland, Rębielice Królewskie 1A, frontal view with partially unfused bicipital tunnel. B. Urotrichus talpoides Temminck, 1841 (NMNS 28207), Recent, frontal view with completely open bicipital tunnel. C. Neurotrichus gibbsii Baird, 1856 (LACM 93944), Recent, lateral view with completely fused bicipital tunnel.

Fig. 3.

A. Scatterplot of the first two axes of the bgPCA. Deformation grids refer to axes extremes (positive and negative values). B. Scatterplot of the first and third axes of bgPCA. Deformation grids refer to axes extremes (positive and negative values).

Shape and size analyses

The bgPCA performed on the procrustes coordinates shows a neat separation between the urotrichine and neurotrichine shrew moles in particular across the PC1 (Fig. 3A). At positive values of the PC1 (62.80% of the total variance) the humeral shape shows an enlargement of the teres tubercle, an enlargement of the medial epicondyle and an expansion of the greater tuberosity, while at negative values the humerus shows a contraction of these regions. Along the PC2 (17.56% of the total variance) it is possible to observe a separation between Neurotrichus gibbsii and Rzebikia polonica At positive values the humeral morphology shows a reduction of the teres tubercle, a lengthening of the greater sulcus and a contraction of the lesser tuberosity, while at negative values the humerus shows an enlargement of the teres tubercle and of the lesser tuberosity while the greater sulcus becomes shorter. Along the PC3 (10.84% of the total variance) it is possible to appreciate the separation between Urotrichus talpoides and Dymecodon pilirostris (Fig. 3B). At positive values the humeral shape shows an enlargement of the medial epicondyle and an increase of the greater tuberosity, while at negative values it is possible to observe a contraction of the regions previously described.

Fig. 4.

Boxplot of the centroid sizes. Bottom and top of the boxes are the first and third quartiles, the horizontal black lines represent the median, the whiskers represent the minimum and maximum values.

Permutational MANOVA returned an overall highly significant difference (p-value < 0.001) among species and pairwise permutation MANOVA returned significant values (Table 1) for all the comparisons.

The boxplot computed for the CS (Fig. 4) showed a significant size variation (permutational ANOVA p-value < 0.001) among species. R. polonica was significantly different from all other taxa by means of pairwise permuted ANOVA (Table 2).

The UPGMA computed on the Euclidean distance matrix calculated on the shape variables (Fig. 5A) evidenced a neat morphological difference between the Urotrichini, where Urotrichus talpoides and Dymecodon pilirostris showed close similarities, and Neurotrichini, where Neurotrichus gibbsii and R. polonica showed the closest morphological affinities.

Table 1.

Pairwise permuted MANOVA results. All p-values are corrected using “Holm” correction.

Discussion and concluding Remarks

The continuous humeral shape variation evidenced by the GM analysis was congruent with the qualitative morphological differences observed in the specimens included in this study. In particular, the neat separation between Urotrichini and Neurotrichini observed along the PC1 is due to major modifications of the regions mainly involved in the digging process (Gambaryan et al. 2003; Piras et al. 2012), such as the expansion of the teres tubercle. Rzebikia polonica has a partially unfused bicipital tunnel (Fig. 2A) while the Urotrichini have it open (Fig. 2B). Field observations on the extant Neurotrichus gibbsii (Campbell and Hochachka 2000: 578; Stone 1995: 57) and recent Finite Elements Analysis (Piras et al. 2012) suggest that Neurotrichini are more adapted to a fossorial lifestyle than Urotrichini. The UPGMA (Fig. 5A) confirmed the distinction between Neurotrichini and Urotrichini. Rzebik-Kowalska (2014) pointed out that Quyania chowi and R. polonica should be included in the Neurotrichini tribe. Our results support the inclusion of R. polonica in Neurotrichini tribe and exclude any Urotrichine affinity. Along the PC2 R. polonica sets apart from N. gibbsii and from Q. chowi. According to Storch and Qiu (1983) and Popov (2004), R. polonica descends from Q. chowi. In fact, the Polish genus has a more robust humerus, a bicipital tunnel showing a higher fusion degree between the pectoral crest and the lesser tuberosity, and a teres tubercle larger and more rounded. The phenetic relationships support R. polonica being more advanced than Quyania chowi and hence justify its different generic allocation. Storch and Qiu (1983) hypothesized a parallel evolution of Neurotrichus gibbsii and Rzebikia polonica, suggesting that the Polish species is more andvanced than N. gibbsii by having a relatively larger size. Popov (2004), following Storch and Qiu (1983), considered R. polonica as more advanced than N. gibbsii. Here we reject such hypothesis because Rzebikia gen. nov. shows many primitive features of the humerus when compared with the North American forms. The most striking features are the partially unfused bicipital tunnel (Fig. 2A), that is completely fused (Fig. 2C) in N. gibbsii (Reed 1951; Sánchez-Villagra et al. 2004), the reduced teres tubercle, and the widened minor sulcus. The enlargement of the teres tubercle is an important character of talpids evolution (Gambaryan et al. 2003; Piras et al. 2012). This humeral region allows the insertion of the muscles Teres major and Latissimus dorsi, two of the main muscles involved during burrowing (Gorman and Stone 1990; Gambaryan et al. 2003; Piras et al. 2012). A larger teres tubercle would allow the insertion of larger and more powerful digging muscles. N. gibbsii and R. polonica are separated along the PC2 and the humeral shape changes associated with this axis are in good agreement with our qualitative observations about the humeral morphological differences between these two taxa. Moreover, N. gibbsii and R. polonica are significantly different under pairwise permutational MANOVA. These evidences suggest that N. gibbsii is better adapted to digging than R. polonica and in a more derived evolutionary state. Nevertheless, R. polonica shows some derived features on teeth such as the reduced precingulid of m1 and more reduced protoconules (Storch and Qiu 1983; Popov 2004), not equally advanced in N. gibbsii. Rzebik-Kowalska (2014) noted that, in R. polonica, the protoconules are absent and vestigial only in one specimen. Moreover, the upper and lower teeth of R. polonica are wider than those of N. gibbsii and more similar to those of Quyania (Rzebik-Kowalska 2014). All of these evidences well support a new generic allocation. The UPGMA (Fig. 5A) shows close similarities with the phylogenetic hypothesis (Fig. 5B) proposed by Storch and Qiu (1983). We follow them in considering Q. chowi as the probable ancestor to N. gibbsii, R. polonica, and R. skoczeni. According to Storch and Qiu (1983) Q. chowi can be considered the ancestor of the neurotrichine lineage. N. gibbsii could represent a derived form that colonized North America during the Early Pliocene, while one or two colonization events towards Eastern Europe could have occurred. A colonization event could have involved the ancestor of Quyania europaea during the Early Pliocene. Another colonization wave from Asia, that involved the Urotrichini, during the Miocene-Pliocene boundary, is testified by the presence of Urotrichus sp. (Maramena locality; see Doukas et al. 1995). In this scenario it is possible to hypothesize Rzebikia gen. nov. being derived from the European Q. europaea. This represents the most parsimonious explanation, although we note that Rzebikia gen. nov. is more similar to Quyania chowi (Storch and Qiu 1983; Popov 2004). If we consider Rzebikia gen. nov. directly derived from Q. chowi we should hypothesize a subsequent colonization event during the late Early Pliocene. Q. europaea is clearly distinct from Rzebikia gen. nov. by its slender humerus and relative smaller size (Rzebik-Kowalska 2014), suggesting a different digging capability and ecological adaptation. Rzebikia skoczeni and Rzebikia polonica are both larger than Q. europaea (Skoczeń 1993; Rzebik-Kowalska 2014). R. skoczeni has been found in the MN15 locality of Węże 2 only (see SOM 3), where no other neurotrichine moles are present. R. polonica first appearence is in the MN16 Rębielice Królewskie 1A locality (see SOM 3). This species could be descended from R. skoczeni anagenetically by an increase in size. However, due to the scarcity of the R. skoczeni fossil record it is not possible to test this hypothesis. R. polonica have been found in sympatry with Q. europaea (MN16, Rębielice Królewskie 1A and MN17, Kadzielnia localities). Size differences have been documented for sympatric species belonging to genera Talpa and Mogera (Abe 1996; Loy et al. 1996; Loy and Capanna 1998; Cleef-Roders and Hoek Ostende 2001; Yokohata 2005; Bego et al. 2008; Loy 2008), this phenomenon has been documented also in the extinct genus Geotrypus (Hoek Ostende 2001). Moreover, we found a significant size difference between Urotrichus talpoides and Dymecodon pilirostris which has been reported to live in sympatry in Honshu and Shikoku regions (Abe 1967). Following this evidence, the size displacement between the species of Rzebikia and Quyania europaea could have occurred in response to eco-evolutionary constraints, such as inter-specific competition and the ability to exploit low productive habitats. Size character displacement between pairs of ecologically close and geographically overlapping species is a common pattern in mammals (Simberloff and Boecklen 1981; Dayan and Simberloff 1998) and could represent a rapid response to strong inter-specific competition in talpids (Loy and Capanna 1998; Loy et al. 2001).

Fig. 5.

A. UPGMA calculated on the Euclidean distance matrix computed on the shape variables. B. Phylogenetic hypothesis of Storch and Qiu (1983).

Finally, recent contributions highlighted that humeral morphology possesses a taxonomic value at the genus level and in some cases at the species level as well (Hoek Ostende 1997; Ziegler 2003; Klietmann et al. 2014). In the present paper the highly significant values reported by pairwise permutational MANOVA confirm the chance to consider the humerus as a diagnostic element. Moreover, our results suggest that the landmark based shape analysis is useful in supporting systematics in palaeontological investigations where only skeletal elements are avilable.