Retrodeformation of the Steinheim Cranium: Insights into the Evolution of Neanderthals

: A number of different approaches are currently available to digitally restore the symmetry of a specimen deformed by taphonomic processes. These tools include mirroring and retrodeformation to approximate the original shape of an object by symmetrisation. Retrodeformation has the potential to return a rather faithful representation of the original shape, but its power is limited by the availability of bilateral landmarks. A recent protocol proposed by Schlager and colleagues (2018) overcomes this issue by using bilateral landmarks and curves as well as semilandmarks. Here we applied this protocol to the Middle Pleistocene human cranium from Steinheim (Germany), the holotype of an abandoned species named Homo steinheimensis . The peculiar morphology of this fossil, associated with the taphonomic deformation of the entire cranium and the lack of a large portion of the right side of the face, has given rise to different hypotheses over its phylogenetic position. The reconstruction presented here sheds new light on the taphonomic origin of some features observed on this crucial specimen and results in a morphology consistent with its attribution to the Neanderthal lineage.


Introduction
The study of fossil specimens has been revolutionised by the foundation of modern morphometrics [1]. Symmetry is one prominent feature of biological objects, and possibly the one affected the most by taphonomic processes [2][3][4][5]. However, symmetry also offers the possibility to restore the original shapes of fossil remains that are found broken or incomplete [6,7]. This is key to the interpretation of these specimens, since taphonomic alteration affecting diagnostic features may lead to incorrect taxonomic attributions and dubious phylogenetic reconstructions [7][8][9]. Digital methods for the reconstruction and restoration of broken fossil remains are nowadays available thanks to an ensemble of techniques that commonly fall under the heading of 'virtual anthropology' [10][11][12]. Specimens can be handled in a safe, virtual environment [7] and undergo restoration protocols that can include the realignment of dislocated fragments [13][14][15][16] or the digital removal of the plaster from traditional reconstructions [8,17] without the risk of damaging the original material. These protocols can be associated with symmetrisation, which helps to recreate missing portions or 'undo' the effects of plastic deformation. In the former case, symmetrisation 'fills the gaps' (i.e., missing portions) in one half of the fossil by mirroring the preserved counterparts [7,[18][19][20][21][22]. In the latter case, referred to as retrodeformation, the plastic distortion of the original shape is corrected by relying on biological symmetry, as calculated by the acquisition of bilateral landmarks, curves, or patches of semilandmarks [4,7,8,[23][24][25][26][27][28]. Mardia and colleagues [3] defined two types of bilateral symmetry: one referring to structures present as two separate copies on both sides of the specimen as mirror images (matching symmetry), the other defined (in three-dimensional objects) by the midsagittal plane passing through the specimen and thus determining an internal left-right symmetry (object symmetry) [2,3]. One key difference between matching and object symmetry is that genuine asymmetry is ignored by the former, but still apparent under the latter. In the case of the vertebrate skull, which provides an example of object symmetry [3], this implies that retrodeformation preserves genuine asymmetry, whereas mirroring does not. Moreover, mirroring can generate artefacts, or a biased morphology, if the only preserved portion is itself distorted [7]. On the other hand, the application of retrodeformation can be affected by the state of preservation of the object [4].
A perfect example of the combination of missing parts and plastic deformation affecting a single specimen is given by the cranium from Steinheim (hereafter, Steinheim), which is the holotype of the abandoned species Homo steinheimensis (Berckhemer, 1936) [29]. This human fossil was found in July 1933 in a gravel pit 70 km north of the town of Steinheim an der Murr, Baden-Württemberg, Germany [30,31] (Figure 1). It was recovered from Pleistocene fluvial deposits along the Murr river, which were well known at the time of the discovery for having yielded well-preserved fossils of Pleistocene mammals [30,32]. Since the discovery came from a well-studied area, the fossil received proper geological contextualisation. It was therefore possible to estimate the specimen's age based on the biochronological dating of the faunal assemblage, roughly corresponding to OIS 9 (i.e., 300-320 ka to 250 ka) [30,[32][33][34][35]. The complex pattern of deformation that affected Steinheim, as well as its incomplete status and the presence of extensive incrustations, made it difficult to discern whether its peculiar morphology represents the original shape of the individual, or it is the product of taphonomic deformation [36,37]. This uncertainty contributed to a longstanding debate concerning the Steinheim phylogenetic position [36,[38][39][40]. The cranium is characterised by a peculiar mixture of archaic and derived traits, which originated different proposals about its position within or close to the Neanderthal lineage-as representing a 'pre-Neanderthal stage' along the so-called process of accretion-or even as a specimen somehow related to the origin of Homo sapiens [41][42][43][44][45]. However, not only most of the left side of the facial skeleton in Steinheim is missing, but also the cranium presents a peculiar plastic deformation, further complicating the recognition of its features. For example, the highly diagnostic infraorbital plate and orbitomaxillary region are preserved only on the left side. This part of Steinheim's face shows an angled transverse profile, which was interpreted in the past as 'anticipating' the modern human morphology to some degree [30,43], but has been conversely interpreted as the result of the retention of archaic facial morphology, also observed in some Western European earlier taxa (i.e., Homo antecessor) [46][47][48]. The relatively low and long neurocranium of Steinheim, possessing a rather vertical occipital plane, also shows a slightly angled coronal profile, or a 'roofed' appearance [36], with the maximum cranial width occurring in the lower portion [35].
A specific name was initially proposed for this specimen (Homo steinheimensis Bereckhemer, 1936) [29], but it is currently considered invalid [40,49], despite that this name has been resurrected at the taxonomic rank of subspecies [50,51]. Steinheim is now generally considered as belonging to the Neanderthal lineage [45,[52][53][54] and possibly related to other Middle Pleistocene populations (e.g., Atapuerca Sima de los Huesos, SH), with which it shares several derived traits in addition to its the geographical and chronological attributions [45,53,54].

Materials and Methods
The description of Steinheim's morphology is influenced by the extensive deformation of the skull [32,37]. A representation of the major directions of the deformation has been obtained by observations on the CT scan of the fossil and a review of literature [32,37,55] and is shown in Figure 2. Prossinger and colleagues [37] performed the first digital segmentation of the cranium, resulting in a model cleared from the encrustations but still heavily affected by taphonomic distortions. Such distortions are observed in the internal structure of the cranium, including a shift to the right of the midsagittal plane of the splanchnocranium, an inward 'inflation' of the left orbital roof, and a rightward rotation of the axis of the crista galli in the anterior endocranial surface [37]. Since the left portion of the face is missing, it is difficult to assess how much of this morphology is determined by the deformation [37]. The right orbit is characterized by an angled shape with a sloped inferior margin. The preserved infraorbital plate shows an angled transverse profile with a point of bending roughly corresponding to the infraorbital foramen [33]. Curiously, this is associated with a moderate inflation of the anterior portion of the infraorbital plate, whereas the lateralmost portion appears flattened [35]. Through investigations conducted via CT scan and digital imaging, it was possible to assess the relative size of the frontal sinuses inside the well-developed supraorbital torus [36] and to diagnose a possible meningioma located in the upper part of the neurocranium [56].
To obtain a reconstruction consistent with object symmetry (sensu Mardia and colleagues [3]), we started by applying retrodeformation [4]. The choice of landmarks ( Figure 3  It was possible to define only a few landmarks on the small preserved portion of the left side of the face, comprising the nasomaxillary region ( Figure 3). In defining surface semilandmarks, we excluded the preserved portion of the temporal squama because it is affected by local breakage and subsequent reconstruction (see Figure 1) [36][37][38]. Similarly, in defining the patches of semilandmarks, the upper part of the left parietal was excluded, as this portion of the neurocranium is more affected by breakage and surface damage ( Figure 3). The basioccipital is also damaged, cracked, and partly shifted inside the neurocranium itself, and therefore, no landmarks were placed on this region.
The high-resolution CT scan of Steinheim was kindly provided by Prof. Dr. Christoph P.E. Zollikofer (Department of Anthropology, University of Zurich). The CT data, obtained in the form of a DICOM stack, were processed in Amira [57] to obtain a 3D mesh, subsequently converted into the .ply format. We defined 52 bilateral landmarks on the skull and 8 curves. The curves were later processed in R by the function equidistantCurve (Morpho R package) [58] to sample evenly spaced semilandmarks along each curve. The coordinates of 500 semilandmarks were obtained by applying a k-means clustering algorithm to the vertex coordinates from a portion of the mesh corresponding to the left part of the cranium, from which we excluded the temporal squama and the damaged area of the coronal suture ( Figure 3). The set of surface semilandmarks built this way was rotated and projected on the right side. In sum, we defined 8 curves (120 points), 52 bilateral landmarks, and 1000 surface semilandmarks for a total of 1172 paired coordinates (Figure 3). After the retrodeformation, applied according to the protocol in Schlager and colleagues [4], we calculated and visualized the local displacement between the starting and retrodeformed meshes using the function localmeshdiff (Arothron R package) [59].

The Retrodeformation
Most of the retrodeformation procedure intervened on the anteroposterior shift of the two sides of the skull (Figure 2c). In the frontal view (Figure 4a), the shift produces a relative enlargement of the piriform aperture, mainly on the left side, associated with a forward shift of the left rim and a slight retraction of the medial portion of the right rim. A slight 'relaxation' of the nasal profile in the superoinferior direction is apparent, as well as the symmetrisation of the general profile of the neurocranium, which is even more evident in the posterior view (Figure 4b). Symmetrisation of the occlusal plane of the teeth eliminates the unnatural downward displacement of the right maxilla along the midsagittal plane, which is present in the original specimen (Figure 4a,b).
The correction of the anteroposterior shift of the face along the midsagittal plane is also evident from the inferior view (Figure 4c), where the reduction of the slight clockwise rotation of the palate becomes apparent, accompanied by a deflation of the right postorbital portion of the neurocranium. In addition, the basicranium regained a more natural position, appearing straighter and medially placed in comparison with the original specimen, even though a slight deformation remains due to the lack of landmarks to be placed on this badly preserved portion. Preservation similarly affects the retrodeformation process of the flexion of the basicranium and the anteroposterior compression along the coronal suture (Figure 4d). In the lateral view (Figure 4d), the general profile of the neurocranium does not show any major changes. However, it is evident that the retrodeformation produces a retraction of the right portion of the face. Corresponding to the frontal squama, it is possible to see in transparency the previous position of the right side (Figure 4d), which was originally shifted forward according to the deformation directions illustrated in Figure 2.

Local Displacement
The local displacement between the starting and the retrodeformed meshes ( Figure 5) indicates the areas of maximum expansion, which is apparent on almost the entire left side of the skull, with the highest values recorded in the preserved portions of the left maxilla, left parietal, and left orbit. Conversely, the right side of the face is affected almost entirely by contraction, with the highest values recorded at the level of the anterior portion of the maxilla and nasofrontal suture. However, the inferior portion of the right maxilla is expanded in the retrodeformed mesh. A somewhat balanced pattern of deformation occurs on the basicranium. The areas of maximum contraction on the right side are associated with areas of maximum expansion on the opposite side. Similarly, the moderate expansion recorded on the right side of the occipital squama corresponds to an almost symmetrical contraction of the left side. Lastly, along the midsagittal plane an area of contraction appears evident.

Principal Component Analysis
The results of the PCA are reported in Figure 6. The first three PCs explain 62.74% of the total variance in the sample, weighting 41.51%, 12.27% and 8.96%, respectively. In the plot, it is possible to discern a clear separation between modern (Sap) and fossil humans both along PC1 and PC3. Along PC2 is visible the separation between the Neanderthals sensu stricto (Nea) and a small group of Middle Pleistocene humans (Mph). The two models of Steinheim (Ste) fall within an intermediate position along PC1, between the Sap cluster and the fossil human group. While the original model of Steinheim (Ori) clearly diverges from the rest of the sample along both PC1 and PC2, the retrodeformed model of Steinheim (R.D.) approaches the fossil human group along the PC1, reaching the limit of the Nea cluster along PC2.

Discussion
We used a retrodeformation protocol to produce a restoration of the Steinheim cranium. The application seemingly restored object symmetry to the specimen [3], despite the poor starting conditions of this incomplete and severely deformed fossil. We relied on the preserved portions, mainly on the left side of the cranium, to drive the reshaping of its counterpart. Even though some directions of the taphonomic deformations were not addressed, our application minimises their effect. The reconstruction (Figure 7), hence, allows us to better contextualise Steinheim among the coeval-or at least chronologically close-Middle Pleistocene Homo specimens. The neurocranial shape in the posterior view appears intermediate in morphology between the populations of Sima de los Huesos and early Neanderthals (e.g., Saccopastore 1), in keeping with the slight lateral expansion of the parietal walls after the reconstruction as compared with the more vertical and 'compressed' profile in Ori. In this respect, the neurocranial morphology of R.D. seems to approach the morphology of the early Neanderthal from Altamura [60]. In the posterior view, the original 'roofed' appearance (as described by Schwartz and Tattersall [36]) weakens in R.D. neurocranium, appearing close in morphology to penecontemporaneous individuals such as Skull SH5 from Atapuerca [40,54,61], except for the further laterolateral enlargement of the parietals. This trait, difficult to discern before restoring symmetry, places the maximum width of the skull in a slightly lower position relative to that of the original specimen, and roughly at the level of the temporal squama, similar to the typical Neanderthal condition [62] (Figure 4a,b). It is also possible to see a change in the relative position of the two mastoid processes, which, although partly damaged, after retrodeformation show reduced development compared with those of SH5. Their slight rotation can be interpreted as a trait anticipating the Neanderthal condition of tapering [44,45,62], although high variability in this feature among the Middle Pleistocene humans has been observed [63]. An almost symmetrical pattern of contraction and expansion is visible at the level of the glenoid fossae ( Figure 5), associated with a change in the relative size of the postorbital portion of the neurocranium. This contributes, in turn, to the slight shortening and laterolateral enlargement of the neurocranium. On the other hand, the 'strip' recorded along the midsagittal axis ( Figure 5) corresponds to an almost continuous area of contraction, which is a clear indication of the taphonomic deformation that occurred along this axis (see Figure 2c). This 'strip' can probably be traced back to a local expansion along the midsagittal line due to the two 'halves' moving in opposite directions.
As can be seen from Figure 5, the retrodeformation was not able to address the anteroposterior vectors of deformation. This is because such vectors acted in a single straight line, rather than bilaterally. Thus, it is not possible to reach evidence-based assumptions on whether the flexion of the basicranium reflects the original condition or it is the result of taphonomic deformation. Nevertheless, the restored midsagittal profile of the cranium suggests that the anteroposteriorly elongated profile of the neurocranium could possibly be associated with a less flexed basicranium. As we proposed in Figure 2b, the present flexion could be related to a deformation operating along the sagittal axis on the upper midface.
Unfortunately, the almost completely missing left portion of the face made it difficult to correct for some local modifications in this area. Nonetheless, it is still possible to carefully evaluate whether some features are due to taphonomic deformation. As mentioned above, the retrodeformation resulted in a 'proper' midsagittal profile (Figure 4d) by undoing the rotational deformation caused by the anterolateral crushing (Figure 2d). By examining the lateral view of the reconstruction, it is more evident how the 'plica' obliterating the frontonasal suture-which is not found in any other hominin from Middle to Late Pleistocene-is consistent with an anterior crushing of the upper part of the nasal portion ( Figure 2a). This, in turn, can be associated with the 'notch' found along the lowerright orbital rim, corresponding to a point of weakness represented by the zygomaxillary suture. We suggest that the peculiar facial morphology of Steinheim is mostly a result of the crushing that occurred in the upper portion of the midface (Figure 2b). In our opinion, the reconstruction showed that the infraorbital plate was in origin possibly less flexed than Ori suggests.
As evidenced by the PCA (Figure 6), Steinheim is distinguished from the rest of the sample, and this 'uniqueness' can be traced back to its complex pattern of taphonomic deformation. Nonetheless, when a part of this is corrected by retrodeformation, it is possible to see how the new model approaches the fossil human subsample, towards the Neanderthal cluster. We hypothesize that since some of the deformation vectors-namely, those operating on the anteroposterior axis-cannot be intercepted by the retrodeformation, Steinheim still presents itself with a 'unique' morphology, distinguished from other Middle Pleistocene specimens.

Conclusions
The 'mosaic' evolution of the typical Neanderthal cranial morphology (i.e., 'classic' Neanderthal cranial shape [62]) seems to have included an earlier development of some facial traits, combined with the retention of a more primitive condition for the neurocranium [51,61]. The full development of the typical Neanderthal en-bombe shape must thus be considered a derived trait. Consequently, the moderate expansion of the parietals of the retrodeformed Steinheim (posterior view, Figure 4b) suggests association with a greater expression of midfacial prognathism than that observed in this individual, consistent with other specimens from the Middle Pleistocene of Europe. In our opinion, the present facial morphology of Steinheim is influenced by the deformation caused on the upper midface by taphonomy. Even though the reconstruction presented here did not correct the whole pattern of deformation, nor did it provide the exact original shape of Steinheim, it contributes to shedding new light on the morphology of this specimen and concurrently to placing Steinheim more firmly in the Neanderthal evolutionary lineage.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/sym13091611/s1: Table S1: Comparative samples used in the analysis; Figure S1: Landmark configuration used for the PCA.