Enamel–dentine junction morphology reveals population replacement and mobility in the late prehistoric Middle Nile Valley

Significance

This study presents further insights into the population history of North-Eastern Africa during a crucial period when fishing, hunting, and gathering was replaced by food-production and a Neolithic lifestyle. Previous studies attributed variations in cranial and dental anatomy to diet changes accompanying the Neolithic transition in the Nile Valley. This study evaluates population continuity hypotheses by analysis of the morphology of the largest sample of dental remains (n = 122), spanning 14.000 years of deep population history. This study provides firm evidence for population replacement and migrations to this region at the Neolithic transition. It also reveals unexpected settlement and interaction processes of these new populations, and provides direct biological evidence of mobility patterns among Neolithic communities.

Abstract

Transitions from foraging to food-production represent a worldwide turning point in recent human history. In the Middle Nile Valley this cultural shift occurred between the sixth and beginning of the fifth millennium BCE. Significant craniodental morphological differences remain inadequately tested by biometric analyses of ancestry and may reflect population origins or diet change between the last hunter-fisher-gatherers (Mesolithic) and first food-producers (Neolithic). Moreover, with no ancient DNA data for this region and very few morphological studies including large samples of Mesolithic individuals, the late prehistoric population history of the Nile Valley remains unclear. Here, we present enamel–dentine junction (EDJ) morphological analyses (based on X-ray microtomography) and biological affinities for 88 individuals spanning 14,000 y from Sudan and southern Egypt. Significant EDJ morphological differences between the last foragers and first food-producers suggest major biological discontinuity at the Neolithic transition. Nevertheless, the persistence of the earlier forager population in the Sudanese Eastern Sahara indicates settlement and population replacement mainly along the Nile. We also present biological evidence of interaction and mobility between these contemporaneous populations during the middle Holocene in the region. It supports the phylogenetic value of EDJ morphology for investigating population affinities at a microevolutionary scale. These results yield insights into the deep population history of the Nile Valley. They provide firm evidence for population replacement and migration toward the region at the onset of the Neolithic transition, attesting that these key changes were not solely triggered by cultural diffusion and diet change.

In the Middle Nile Valley, the transition from foraging to food-producing societies progressively occurred between the 6th and beginning of the 5th millennium BCE, until its full-fledged adoption during the 5th to mid-4th millennium BCE (1–13). Archeobotanical (4, 12, 14) and archeozoological (15–20) evidence of Near Eastern ancestry for the Neolithic domesticated species prompted us to address the question of whether migration or cultural diffusion yielded these new practices.

In this context of cultural and behavioral changes, the biological record also shows significant skeletal and dental morphological variations between the last hunter-fisher-gatherers and mid-Holocene food-producers (21–30). The interpretations of these morphological differences and their implications regarding Nile Valley settlement processes at the dawn of the Neolithic transition have been debated for decades, especially considering the absence of biomolecular data (i.e., ancient DNA) for the region. While some have proposed an in situ adaptation to the new Neolithic lifestyle and diet (26, 31–35), others interpret this as evidence for population discontinuity and migration (23, 27–29, 36–38).

Recent studies on dental remains from large samples of Late Paleolithic to Neolithic populations from the Nile Valley (24, 36, 39) highlighted complex signals. Dental morphology, crown size, and dental tissue proportions often significantly differ between the last foraging populations and first food-producing societies. Still, individuals from pretransition burial sites such as Al-Khiday, in Central Sudan (40, 41) are exceptions showing greater affinities with later food-producing populations. Moreover, regional biological variation appears to distinguish Upper and Lower Nubian from Central Sudanese Neolithic groups (39, 42).

This complexity in biological relatedness and settlement processes of Middle Nile Valley societies highlights the need for increasing the number of samples analyzed to more precisely characterize population affinities. To do so, we focus on the morphological comparison of the enamel–dentine junction (EDJ) of upper first and second molars (UM1 and UM2). The EDJ is often better preserved than the outer enamel surface (OES), approximates the inner enamel epithelium, and is established early in dental development, thus constituting a strong and reliable genetic proxy for biodistance analyses (43–47). Through three-dimensional geometric morphometric analyses (3D GM), previous studies successfully discriminated both between hominid taxa [e.g., (48–52)] and recent/modern human populations [e.g., (44, 53–55)].

We applied a landmark-based (LM) (53, 54, 56) 3D GM and a landmark-free surface deformation-based (DSM) (52, 57–60) analysis to the UM1 (n = 56) and UM2 (n = 66) EDJ of individuals from Late Paleolithic to Neolithic samples in Nubia, Central Sudan, and the Sudanese Western Desert margins (Fig. 1 and Table 1 and SI Appendix, Supplementary Text S1 and Tables S1-S4). Using both approaches, we investigated shape variation of the dentine horns, marginal crests, and cervical line, and also of the occlusal basin and EDJ lateral aspect. By characterizing, quantifying, and describing EDJ shape variation between the last hunter-fisher-gatherers and first food-producers from the Nile Valley, we aim to test whether these differences support in situ morpho-functional adaptation (i.e., few differences and no sudden change in EDJ morphology at the Neolithic transition) or biological discontinuity (i.e., significant and rapid variation in EDJ morphology between the last foragers and first food-producers). Similarly, we aim to assess and better understand Neolithic regional dental variation observed in previous studies and its implications for understanding Nile Valley population history and settlement processes.

Fig. 1.

Location and dating of the assemblages considered in this study. (A) Location of the sites included in this study. Light gray numbers correspond to the Nile cataracts. (B) Chronology, cultural affiliation, and environmental context of the samples included and discussed in the main text. Further details on the sample dating are given in SI Appendix, Table S1.

Table 1.

Summary of the samples included in this study

		N			References
Sample	Date* (d/i/c)	Ind.	UM1	UM2	Archaeological and chronological context
Late Paleolithic		8	8	5
Tushka	16900–10000 BCEc	2	1	1	(61, 62)
Jebel Sahaba	15000–11300 BCEd	6	7	4	(21, 63, 64)
Mesolithic		21	11	17
El-Barga Mesolithic	7800–6700 BCEi	5	2	4	(8, 65, 66)
Sphinx	6800–5900 BCEd	8	4	6	(67, 68)
Fox Hill Mesolithic	6700–6400 BCEd	7	4	6	(69, 70)
10-W-4	5500–5350 BCEc	1	1	1	(71, 72)
Neolithic		59	37	44
El-Barga Neolithic	6000–5500 BCEi	17	14	7	(8, 65, 66)
Ghaba	5600–4300 BCEi	12	6	9	(73, 74)
Affad	5000–3900 BCEi	3	1	3	(75)
Letti	5th millennium BCEd	2	0	2	(76, 77)
Kadruka	4550–4000 BCEi	18	13	16	(78)
Abu Tabari	3800–2700 BCEi,c	6	3	6	(79–81)
Wadi Shaw	2500–2300 BCEd	1	0	1	(82, 83)

^*d: direct ¹⁴C dating on human remains (bones and/or teeth); ⁱ: indirect ¹⁴C dating of the burials; ^c: dating based on context.

Bold values correspond to the total number of individuals included in the study for each time period.

Results and Discussion

EDJ Shape Comparisons.

We provide EDJ morphological data for the region, comprising 122 teeth from 88 individuals. This dataset spans over 14,000 y, from the Late Paleolithic to Middle/Late Neolithic periods, and covers an expansive area from Sudan and southern Egypt. This includes the largest dental sample of Mesolithic individuals (n = 21) yet analyzed for the Middle Nile Valley. Principal component analyses (PCA), between-group principal component analyses (bgPCA) and canonical variance analyses (CVA) computed on Procrustes landmarks coordinates (LM; SI Appendix, Supplementary Text S2 and Figs. S4–S10) and surface deformation parameters (DSM, SI Appendix, Figs. S11–S15) showed similar results for UM1 and UM2 (SI Appendix, Supplementary Text S2). Surface deformation appears to better identify EDJ shape variation by encompassing significant morphological differences of the occlusal basin and EDJ lateral aspect in both molars.

The bgPCA and CVA computed on EDJ surface deformations (DSM) display similar distinctions between the groups on both UM1 and UM2 datasets (Fig. 2). In all analyses, Late Paleolithic (LP) and Mesolithic (MESO) foragers are distinct from Nubian and Central Sudanese Neolithic food-producers (NEO), with little to no overlap between them. These differences are not entirely related to allometry (SI Appendix, Table S6) and are statistically significant (SI Appendix, Tables S7 and S8).

Fig. 2.

Surface-based shape comparisons of the EDJ of upper first and second molars. Between-group principal component analyses (bgPCAs, A and B) and canonical variance analyses (CVAs, C and D) computed on the surface deformation of 52 UM1s (Left) and 59 UM2s (Right) from the Nile Valley.

The Hunter-Fisher-Gatherer Populations.

Despite a substantial two to three millennia gap between the LP and MESO individuals within our sample, little to no EDJ shape variation appears between them (Fig. 2 and SI Appendix, Figs. S4–S15). For the UM1s, this corresponds to a relatively larger and more buccally positioned metacone, more developed hypocone, higher oblique crest, and slightly higher overall EDJ body for the LP teeth; a slightly deeper occlusal basin and mesial fovea are evident for the MESO individuals. Similar patterns are evident in the UM2s, with the LP individuals having a relatively larger and more buccally located metacone, higher oblique crest, and slightly more developed/higher hypocone dentine horn. Cross-validated classifications of bgPCA and CVA (SI Appendix, Tables S9–S12) show that most forager individuals are consistently classified as LP or MESO (e.g., for DSM, respectively 15/19 and 14/19 for UM1s; and 16/22 and 16/22 for UM2s).

These results further confirm marked homogeneity and affinities between Late Paleolithic and Mesolithic hunter-gatherers (36, 39). Similarities between LP and MESO groups suggest overall spatiotemporal continuity. Nevertheless, we show slight differences in EDJ shape between these samples, with similar patterns of variation for the UM1s and UM2s. Previous studies also recorded slight variation in tooth crown size and dental reduction between Late Pleistocene and Early Holocene foraging populations from this region (36, 39). Similar patterns of EDJ variation are found between populations in contexts of dental reduction from other world regions [e.g., (53) Fig. 4, where PC2 discriminates between Paleolithic to Early Neolithic individuals], with a general trend toward preferential mesiodistal reduction (84). From a developmental perspective, it is known that dental morphology and dimensions are decreasingly stable distally within each tooth type (49, 85, 86). Even within the tooth itself, mesial cusps are considered more stable than distal, as a result of complex interactions between activator and inhibitor processes during odontogenesis (87, 88). In this regard, the distal cusps (i.e., hypocone and metacone for maxillary molars) would be primarily impacted by a reduction of crown size, and greater variability is expected for UM2s than UM1s (86). These assumptions are in line with our observations for the Nile Valley foragers, and therefore suggest that dental reduction led, at a microevolutionary scale, to the variation we observed between LP and MESO samples.

Fig. 4.

Schematic summary of the population history of the Nile Valley (12th to 3rd millennium BCE). (A–C) Maps showing the extension of the different cultural complexes (dashed lines) and population affinities (in color) of the Middle Nile Valley. (D) Inferred population history of the late prehistoric Nile Valley.

The Neolithic Transition along the Nile.

Comparisons between the MESO and NEO samples (Early Neolithic Nubian: ENEON; Early Neolithic Central Sudanese: ENEOSC; Middle Neolithic Nubian: MNEON) display a much different pattern than for the LP/MESO groups. PCA, bgPCA, and CVA consistently show little to no overlap between the last foragers and first food-producers (Fig. 2 and SI Appendix, Figs. S4–S15). For the UM1s, the Neolithic individuals have relatively higher EDJ body and dentine horns, a more developed hypocone cusp, and more rounded aspect of the EDJ wall mesial surface. They also appear to have a well-developed Carabelli tubercle that MESO individuals generally lack, a shallower and narrower expression of the mesial fovea, and a more distally located protocone dentine horn. EDJ shape variation between the forager and food-producer samples for UM2s is characterized by the Neolithic individuals having a relatively more developed paracone and hypocone, higher EDJ body, dentine horns, and oblique crest, but more mesiodistally constricted EDJ. These differences are supported by the PERMANOVAs computed on all PC scores (SI Appendix, Tables S7 and S8) with high R² values and significant P-values for all pairwise comparisons between MESO (and LP) and ENEON, MNEON, or ENEOSC. Cross-validations of bgPCA and CVA (SI Appendix, Tables S9–S12) show that almost all food-producers are consistently classified in one of the Neolithic groups (e.g., for DSM, respectively 28/33 and 30/33 for UM1s; and 27/31 and 26/31 for UM2s).

Contrary to LP vs. MESO, the Mesolithic and Neolithic samples are separated by a few centuries at most, with some chronological overlap (Fig. 1 and Table 1 and SI Appendix, Table S1). Considering the notable evolutionary stability of EDJ morphology (43–48, 89) especially of the UM1s and mesial cusps (85–88), such rapid EDJ shape variation would require substantial evolutionary constraints (i.e., population isolation, dramatic demographic reduction/bottleneck effect, etc.) to occur in a short time. However, archeological (3, 8, 9, 90, 91), biological (39, 68, 73), archeozoological (2, 11, 15), and archeobotanical (4, 12, 14, 92) data show continuous cultural exchanges between different regions of the Nile Valley, little to no change in the health status of these populations, and no indication of a drastic decline in occupation density. Therefore, the EDJ shape differences highlighted between these contemporaneous groups of foragers (MESO) and food-producers (NEO) indicate strong biological discontinuity rather than in situ adaptation. This implies major gene flow and migrations toward the Nile Valley at the dawn of the Neolithic transition, starting in the sixth millennium BCE, with no or limited genetic contribution of Mesolithic foragers to later food-producing populations.

These data also highlight overall stability and homogeneity of the post-Neolithic transition population, with minimal morphological variation between regions and time periods (SI Appendix, Supplementary Text S3). The main EDJ shape variation we observe between Neolithic groups tends to highlight chronological tendencies, especially for UM2s (Fig. 2), with little difference among Early Neolithic individuals from Nubia and Central Sudan or Early Neolithic individuals from Central Sudan and Middle Neolithic individuals from Nubia, but greater ones between both Nubian samples. In the earlier individuals, this consists of having a relatively smaller and more buccally located metacone, lower oblique crest, and slightly less developed/lower hypocone dentine horn than Middle Neolithic individuals. This pattern is analogous to the one highlighted between LP and MESO foragers. Rather than dental reduction, a slight increase in the mesiodistal (MD) dimension of upper molars was recorded (36, 39, 73). These larger diameters might result from the increase in size of the hypocone and metacone cusps. This does not align with previous observations in similar contexts from other regions of the world, where distal cusps (especially the hypocone) gradually reduce through time (93–96). As such, this pattern can provide major insights into the settlement processes of these new Neolithic groups, which may result from various population events, including admixture with other (or earlier) groups, adaptation, or genetic drift. The former hypothesis of admixture with the earlier population (i.e., Mesolithic individuals) is unlikely, considering significant and extreme differences between the last foragers and Middle Neolithic individuals (Fig. 2 and SI Appendix, Tables S7 and S8). The opposite would be expected in an admixture scenario, with the Neolithic populations becoming increasingly more like ancient Mesolithic groups. In situ adaptation is also unlikely considering the high morphological stability of the EDJ and absence of strong evolutionary constraints in the context studied here. On the contrary, genetic drift and/or external gene flow are more probable. Substantial archeological evidence for large-scale communication and exchange within the Nile Valley -from Egypt to the White and Blue Nile Rivers- and with distant regions (especially the Red Sea Hills and Near East) during the Neolithic (3, 13, 65, 76, 90, 97, 98) support the latter scenario. These continuous cultural, and likely biological/genetic, contacts with distant groups, or simply with the native population of new settlers, could result in the slight evolution of the Nile Valley food-producers phenotype highlighted here.

The Western Desert Margins.

Considering their location, recent dating, and small sample size, individuals from Abu Tabari and Wadi Shaw, in the Western Desert margins (i.e., eastern part of the Sahara, west of the Nile River), were considered separately and projected a posteriori on the previously computed bgPCA and CVA (Fig. 3). We then computed typicality probabilities for each individual on CVA and bgPCA (SI Appendix, Table S13).

Fig. 3.

A posteriori projected individuals from the Western Desert (Abu Tabari, Wadi Shaw), the Wadi Howar outlet (Affad, Letti) and Kadruka. (A and B) A posteriori projections of 4 UM1s and 13 UM2s. (C) EDJ surfaces of the a posteriori projected individuals in occlusal and linguo-distal views. Classification of each tooth is based on typicality probabilities (SI Appendix, Table S8).

Considerable heterogeneity is evident among the Western Desert individuals. Indeed, while some (i.e., Abu Tabari 02/28-2; 02/28-8; and 02/1-8) align with the Middle Neolithic Nubian individuals from Kadruka, others (i.e., Abu Tabari 02/28-14; 02/28-23; 03/34-1; Wadi Shaw 83/110-15) fall completely out of the Neolithic variability, and are instead within the MESO range, as confirmed by extremely high typicality probabilities. Archeological evidence (79–81, 91, 99) shows no indication of chronological differences or variation in funerary practices between the Mesolithic-like and Neolithic-like individuals from Abu Tabari. All appear to belong to the same funerary context, dated around the fourth to early third millennia BCE, and all belong to Neolithic pastoral communities (SI Appendix, Supplementary Text S1 and Table S1). In this regard, the observed morphological differences are likely related to actual biological differences, rather than contextual incongruence. More importantly, all of these individuals are the most recent of our entire sample [especially Wadi Shaw 83/110-15: 2,374 ± 85 calBCE (82)], dated three to four millennia after the population replacement we highlighted along the Nile River. Therefore, the preservation of this forager-like signal in the Western Desert is of primary importance and suggests the presence of two biologically different contemporaneous populations: one directly affiliated with the earlier Mesolithic foragers, and the other showing greater affinities with the later Neolithic population along the Nile. We suggest this variation results from complex settlement processes of the new food-producing communities that arrived in the Nile Valley at the Neolithic transition. We hypothesize that the new Neolithic population mainly settled along the Nile River and did not extend further into the Sahara (i.e., not replacing the people living there) and/or that the arrival of these newcomers might have led the previous foraging groups to migrate into the Western Desert margins. In both cases, this would have led to a local biological continuity and preservation of the earlier forager lineage in this region. This would also imply diffusion and adoption of the food-producing practices by earlier local populations in this particular region.

After primary statistical assessment (SI Appendix, Supplementary Note S3), a posteriori projection was also conducted for Affad and Letti (Southern Dongola Reach) individuals. Very similar results to those from the Western Desert samples are found for the fifth millennium BCE pastoralists from this region (75–77) (Fig. 3 and SI Appendix, Table S13). Among them, we have individuals from the same area with extremely different EDJ morphologies (i.e., Affad 132; Grave 1 and Grave 2), suggesting distinct population affinities: one aligning with the MESO and forager-like Western Desert variability and the other falling closer to Early Neolithic groups from the Nile riverbanks. This is of particular interest considering that Abu Tabari and Letti/Affad are located on the Wadi Howar and at its outlet into the Nile River, respectively. In fact, reconstruction of the wadi paleoclimate (9, 79) and archeological evidence (13, 79, 100–102) indicate the continuation of favorable conditions there until the second millennium BCE, with substantial cultural interactions between local populations and those from the Nile Valley, especially Nubia. In that sense, the wadi appears as a natural communication route between regions upstream of the Wadi Howar and the Nile (99, 103). Considering this, we suggest that the presence of individuals with “Nilotic” Neolithic affinities in the Abu Tabari region, and conversely the Letti/Affad area, may represent biological evidence of such exchange and movements along the Wadi Howar. Another possible scenario for the Affad/Letti area is that the presence of forager-like morphologies there may represent some level of a local continuity from the earlier population during the 5th millennium BCE. This would imply adoption of pastoral practices by the local and more ancient population, and, in a way, a “cohabitation” of the two biologically different populations in this region during the 5^th millennium BCE. In both scenarios, these results highlight the importance of the Wadi Howar as a major corridor between the Nile Valley and Sahara Desert in prehistory, and the coexistence and interactions of two different neighboring populations, as well as the need for further analyses and proxies to address the complexity of the Southern Dongola Reach archeological context.

However, we suggest that despite substantial cultural exchanges and/or geographical proximity, little to no biological interactions (i.e., admixture) happened between these two populations. Indeed, both are extremely different based on EDJ morphology and only a few individuals appear to have an intermediate signal (Fig. 3 and SI Appendix, Table S13). Later admixture with the Nilotic groups should be acknowledged, as evidence of more gracile populations in the Wadi Howar was found from the second millennium BCE (80, 104). This could imply delayed absorption of this Western Desert communities into the wider Nilotic food-producing population and the disappearance of the forager-like phenotype in the region. This remains to be assessed through phylogenetically reliable proxies in future analyses.

Finally, these results provide an additional example of the complex Neolithization process of North-Eastern Africa. When looking at adjacent regions, archeological evidence, and ancient DNA results (e.g., from Eastern Africa) suggest: 1) major interactions occurred within the Nile Valley and neighboring areas (91, 103, 105), 2) prolonged migration/admixture events occurred at least until the third millennium BCE (106), and 3) overall complex population history. This highlights the need for more comprehensive and large-scale studies to better understand the settlement processes and biological history of these early food-producers. In all, despite the lack of aDNA data for this region, this study of EDJ morphology provides valuable insights into Nile Valley recent population history and settlement processes (Fig. 4). We present direct evidence for population replacement between the last hunter-fisher-gatherers and first food-producers there at the dawn of the Neolithic transition (starting from the sixth millennium BCE). The implication is that herding and farming practices did not arrive in the Nile Valley solely through cultural diffusion, but rather migration(s). The origin of these settlers is yet to be determined, but we expect a strong Near Eastern component as documented by the appearance of domesticated animals and cereal crops (2–4, 13–16). Evidence was also suggested for prolonged connections and biological interactions between Nile Valley Neolithic groups and extraregional populations. Nonetheless, despite clear overall biological discontinuity along the Nile, we identified the persistence of a forager-like population in the western margins of the Sudanese Eastern Sahara. The presence of this earlier population lineage, that appears to have lasted at least until the third millennium BCE, suggests that the new Neolithic groups settled only along the Nile River. More importantly, we found biological evidence of interactions and mobility along the Wadi Howar, during the Neolithic period and between these two contemporaneous populations. In all, these results further confirm the high phylogenetic value of the EDJ morphology in contexts where ancient DNA is not preserved to provide further insights into the peopling of the Nile Valley. If applied in cross-regional analyses, such an approach might be the key to further understand the population history of North-Eastern African early food-producing communities.

Materials and Methods

The Skeletal Material.

Thirteen assemblages from the Middle Nile Valley were included in this study (Fig. 1, Table 1 and SI Appendix, Supplementary Text S1). These sites cover >14,000 years (~16,900–2,300 BCE) and a vast area, with individuals from Nubia, Central Sudan, and the Western Desert (Eastern Sahara) margins. Late Paleolithic individuals of this study originate from the Nubian sites of Jebel Sahaba [JS, site 117, n = 6, (21, 63, 64, 107)] in Northern Sudan and Tushka in Southern Egypt [TUSH, site 8905, n = 2, (61, 62)] and represent two of the oldest and largest assemblages for this period in the Nile Valley. Mesolithic individuals were sampled from several sites in Nubia and Central Sudan, including five Nubian individuals from El-Barga [EBM, (8, 65, 66)] and 16 Central Sudanese individuals, including 15 from sites in the western part of Jebel Sabaloka, near the sixth cataract of the Nile (Sphinx—SPX, n = 8; Fox Hill—FHM, n = 7) (67–70). The final Central Sudanese individual originates from the Late Mesolithic site 10-W-4 in the region of El-Salha, south of Omdurman (71, 72). Finally, teeth from 52 Neolithic individuals were selected from the Nubian assemblages at El-Barga (EBN, n = 17) (8, 65, 66), Kadruka 1 (KDK, n = 18) (36, 97, 108), and the Affad (AFD, n = 3) and Letti (LTD, n = 2) basins (75). All Central Sudanese Neolithic individuals originate from Ghaba (GHB, n = 12) (73, 74). Teeth were also sampled from sites in the Western Desert: Abu Tabari (sites 02/1, 02/28, and 03/34—ABT, n = 6) in the lower Wadi Howar (79), and Wadi Shaw (83/110—WSH, n = 1), northwest of the Laqiya Arbain oasis (82, 83). In total, 88 individuals are included in this study, represented by 56 UM1 and 66 UM2 (SI Appendix, Tables S1–S4).

Segmentation.

Upper first and second molars with limited occlusal wear degree (less than Molnar’s (109) grade 3), a complete crown and well-preserved cervix were selected for microtomographic acquisitions (µCT scans) with the following parameters: 78 to 260 µA intensity; 100 to 180 kV voltage; a 0.1 to 1.6 mm copper, or 1.0 mm aluminum, or copper + aluminum filter. The final volumes were reconstructed with an isotropic voxel size ranging from 11.6 µm to 79.2 µm, depending on the acquisition parameters and size of the skeletal piece scanned. See SI Appendix, Table S5 for more details on each sample acquisition parameters, scan resolution, device, and reconstruction software. Segmentation of the reconstructed images was performed on Avizo 9.5.0 (Thermo Fisher Scientific) through a semiautomatic process followed by manual corrections. Slightly worn dentine horn apices were manually reconstructed following the protocol of Zanolli et al. (110) (SI Appendix, Tables S3-S4). Error test for the reconstruction of the dentine horns was performed by N.M. and C.Z., both reconstructing five UM1s randomly selected. It showed reliable and consistent reconstruction (SI Appendix, Fig. S2). All EDJ surfaces were then extracted and computed using a constrained smoothing and a 0.4 padding value (111). Finally, all right antimeres surfaces were mirrored to allow comparison between right and left teeth. Two shape comparison approaches were then applied to this set of EDJ surfaces.

Landmark-Based Analysis (LM).

Using the B-spline function in Avizo 9.5.0, two sets of landmarks were manually digitized by N.M. on the EDJ surfaces (SI Appendix, Fig. S1). The first included four landmarks, placed at the tip of each dentine horn (in order: paracone, protocone, hypocone, metacone). The second set consisted of equidistant semilandmarks: curves were digitized on the marginal ridges, connecting the landmarks at the tip of the dentine horns (starting from the paracone and moving lingually); on the oblique crest, from the tip of the protocone to the metacone; and at the cervical line (starting from the center and most buccal point of the lingual aspect of the cervix—i.e., between the protocone and hypocone, proceeding mesially) [adapted from (56)]. A set of equally distributed semilandmarks were then projected on each of these curves: 26 semilandmarks for the EDJ marginal ridges (eight between the paracone and protocone; six between the protocone and hypocone; six between the hypocone and metacone; six between the metacone and paracone); eight on the oblique crest; and 25 on the cervical line.

In cases where the hypocone horn was too diminutive or absent and/or when the starting point of the cervical line semilandmark curve was not clearly identifiable (i.e., lingual aspect too “rounded” with no clear distinction between the protocone and hypocone at the cervix level), we chose not to approximate their positions and to exclude the tooth from landmark-based analyses (SI Appendix, Tables S3 and S4).

Landmarks and semilandmarks coordinates were then imported in R (v4.2.2; R Core Team 2022) (112) and shapes v1.2.7 package (113) was used for data preprocessing (i.e., Procrustes superimposition). To account for possible intraobserver error, 10 UM1s were randomly selected, and landmarks and semilandmarks were digitized again after several months. Principal component analysis (PCA) computed on the original and control Procrustes coordinates (SI Appendix, Fig. S3) show that each pair of measurements clusters together. This analysis also shows that the distance between control and original measurements of an individual is always lower than the distance with any other individual. These results indicate good repeatability of the method and low intraobserver error.

Surface Registration-Based Analysis (DSM).

To provide finer-grained characterization of the EDJ shape variation, not only on the position and height of the dentine horns and cervical line, but the occlusal basin topography and morphology of the EDJ lateral aspect, a surface deformation-based approach was applied. Contrary to the LM approach, which relies on the 3D coordinates of a set of homologous points, the DSM method is based on the mathematical principle of diffeomorphism and relies on the computation of a sample-average surface model (i.e., here for UM1s and UM2s separately) and its deformations to each investigated surfaces (52, 57, 59, 60). In other words, it does not focus on a point-to-point correspondence as in LM method, but on the distances and local orientations between two meshes, which allows comparison of entire surfaces rather than predefined points.

Prior to this, all EDJ meshes were decimated to 50,000 triangles, manually oriented, and then aligned and rigidly scaled using an Iterative Closest Points (ICP) algorithm through the Align Surfaces function of Avizo 9.5.0 (with the “rigid and uniform scale” option). After conversion to VTK files on ParaView v5.6., the EDJ surfaces were imported in Deformetrica v.4.3.0 (https://www.deformetrica.org/) (60) to compute the sample average shape, called global mean shape (GMS). With this software, we also computed the deformations between the GMS and each individual mesh, summarized by sets of control points associated with vector fields (called momenta). These shape parameters (n_parameters = 1,260 for UM1s and n_parameters = 1,287 for UM2s) were then imported in R v.4.2.2, using the RToolsForDeformetrica package (114).

A comparison was made between LM and DSM methods, to account for possible inaccuracies (or “noise”) introduced by comparing whole EDJ surfaces (115) in such a microevolutionary context.

Statistical Analyses.

All statistical analyses were computed in R v.4.2.2. With the ade4 v.1.7.22 (116–118) and Morpho v2.11 packages (119), we computed principal component analyses (PCA) on the GMS-to-individual diffeomorphisms and Procrustes coordinates of all individuals. Between-group principal component analyses (bgPCA) were then computed, using all PCs and the individuals from five a priori defined groups: Late Paleolithic (LP), Mesolithic (MESO), Nubian Early Neolithic (ENEON), Central Sudanese Early Neolithic (ENEOSC), and Nubian Middle Neolithic (MNEON). The individuals from the Western Desert and the Affad/Letti region were projected a posteriori onto the bgPCA morphospace.

We also computed cross-validated bgPCA (cv-bgPCA) using Morpho v2.11 package, to assess possible spuriousness effect (118, 120). The results yield a similar distribution of groups between the bgPCA and cv-bgPCA for both UM1s and UM2s (SI Appendix, Figs. S10 and S15). Canonical variance analyses (CVA) and their cross-validation (cv-CVA; SI Appendix, Figs. S10 and S15) were also computed using the Morpho package and based on the same five a priori groups as bgPCA (i.e., LP, MESO, ENEON, ENEOSC, MNEON). The number of PC scores included in the computation of the CVA (ranging from 6 to 9 and representing ca. 70% of the total variance) were determined following Hastie et al. (121) recommendations.

Allometry was tested using permutational multivariate analyses of covariance (MANCOVA, 10,000 permutations) (122), with the EDJ area and groups as explanatory variables (SI Appendix, Tables S6).

We also performed permutational multivariate analyses of variance (PERMANOVA), using the pairwiseAdonis v.0.4.1 package (123) [1,000 permutations; based on Euclidean distance and with no correction (124)], to assess the structure of data and results, and to estimate the morphological distances between the a priori defined groups and their significance. Finally, classification resulting from cross-validated bgPCA and CVA are listed in SI Appendix, Tables S9-S12. All analyses show consistent distribution and distinction of the a priori defined groups and confirm that no or few spurious effects influenced the groups observed in the bgPCA and CVA (118, 120, 125).

Ethics Statement

The material included in this study was excavated in the framework of several international archeological missions in Sudan and Egypt, with the authorization and support of the National Corporation for Antiquities and Museums (NCAM) of Sudan and the Egyptian government. The material recovered from these excavations, including the human remains studied in this work, are currently on temporary loan or permanent deposit in various European institutions for study, documentation and restoration, with all necessary licenses and permits issued by the NCAM of Sudan and the Egyptian government. For further details on each sample excavation and export history, see SI Appendix, Supplementary Note S1.

Access to the collections was granted by the institutions in charge of the remains. Precautions were taken not to damage the remains during the microtomographic acquisitions. A copy of all data were deposited in each curating institution and are fully available to the Sudanese and Egyptians authorities.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Some study data available (The micro-tomographic scans/data are available from the corresponding authors upon reasonable request and, for some samples, pending the agreement of the institutions in which the remains are curated. A copy of all micro-CT data is also available at the curating institutions of the remains included in this work. Access to the original material should be addressed to the curators of these institutions. More specific information on the access to the collection can be obtained from the corresponding authors. All other data underlying the results presented in this study are included in the manuscript or SI Appendix file).