Insights into the palaeobiology of an early Homo infant: multidisciplinary investigation of the GAR IVE hemi-mandible, Melka Kunture, Ethiopia

Childhood is an ontogenetic stage unique to the modern human life history pattern. It enables the still dependent infants to achieve an extended rapid brain growth, slow somatic maturation, while benefitting from provisioning, transitional feeding, and protection from other group members. This tipping point in the evolution of human ontogeny likely emerged from early Homo. The GAR IVE hemi-mandible (1.8 Ma, Melka Kunture, Ethiopia) represents one of the rarely preserved early Homo infants (~ 3 years at death), recovered in a richly documented Oldowan archaeological context. Yet, based on the sole external inspection of its teeth, GAR IVE was diagnosed with a rare genetic disease–amelogenesis imperfecta (AI)–altering enamel. Since it may have impacted the child’s survival, this diagnosis deserves deeper examination. Here, we reassess and refute this diagnosis and all associated interpretations, using an unprecedented multidisciplinary approach combining an in-depth analysis of GAR IVE (synchrotron imaging) and associated fauna. Some of the traits previously considered as diagnostic of AI can be better explained by normal growth or taphonomy, which calls for caution when diagnosing pathologies on fossils. We compare GAR IVE’s dental development to other fossil hominins, and discuss the implications for the emergence of childhood in early Homo.


Estimating the age of the stress events in LRP4 enamel. (Suppl. Text S1)
1

.1. Estimating GAR IVE's age at death using modern standards
Virtual developmental sections were recorded in each tooth to assess its stage of formation (e.g., 1 ). Since the inner structure of the teeth was obscured by taphonomic alterations, the EDJ was drawn in colour after close inspection of the scans to facilitate the assessment and the scoring of the developmental stages (Suppl. Fig. S1).
Supplementary Figure S1 -Virtual developmental 2D sections through the GAR IVE deciduous and permanent teeth. Due to the lack of contrast, the EDJ was delineated in pink to facilitate the visualisation of the enamel and dentine thickness.
To note that the I2 has developed 1.5 mm of root on the lingual aspect, and ~2 mm of root on the buccal aspect, the P3 has formed 0.6 mm on both sides, and the M1 has formed 2.6 mm of root on its lingual aspect, for 3.8 mm on the lingual side of its mesial section.

Recent modern humans
Based on radiographs, Condemi 2 had estimated GAR IVE's age to fall between 3 and 4 years. Zanolli et al. 3 refined this estimation at 2.5-3.5 years.
After inspection of the PPC-SR µCT data (Suppl. Fig. S1), the calcification stages were scored and corresponding median ages were listed per tooth type after the modern human atlas built by AlQahtani et al. 4 Table 4 ; †: from Table 5; ‡: from Table 6 in AlQahtani et al. 4 .
Supplementary Table S1  Where "Dm" is the sum of the individual scores for a quadrant of the mandibular dentition (in this equation, omitting the M3).
This scoring system is often used to estimate a chimpanzee-equivalent age for fossil hominins (e.g., 7,8 ).
In GAR IVE, the LRI1 is not preserved but is assumed to be at least at a stage equivalent to that of the LRI2. The LRM2 is also missing, but since its initiation is likely very close in time to that of the LRP4, their developmental stage is assumed to be the same.  Kuykendall (1996) 5 .
The sum of these scores is Dm = 29 (Suppl . Table S2). This yields a "chimpanzeeequivalent" age of 3.87 years with a range of 2.88 -4.86 years.
Yet, since Kuykendall 5 's scoring was designed on radiographs and that we score on µCT data, the regression underestimates the stage of development of the germs. We apply the conversion system designed in Gunz et al. 8 Kuykendall (1996) 5 scores from the use of µCT data after Gunz et al. (2020) 8 .
The new sum of these converted scores is Dm = 25 (Suppl . Table S3), leading to a new age and range estimate : 3.13 [2.14 -4.12] years.

Early hominins of similar dental developmental stage
Several early hominins show a stage of dental development similar to that of GAR IVE, some of which have a published histological age at death 7,9 . The stage of development of the M1 roots and of the P4 are of special interest.
KNM-ER 1477 (Suppl. Fig. S3, and Fig. 1 in Dean, 1987 9 ) is slightly less advanced than GAR IVE as its M1 has achieved crown formation but has yet not initiated root formation. The C, P3 and P4 are still developing their crown, with P3 being more advanced than P4.
-P. boisei KNM-ER 812: 2.5-3.0 years 9 . KNM-ER 812 (Suppl. Fig. S4) seems to overall follow the same pattern and timing of dental development. Besides the preserved roots of the deciduous canine and molars, I1 has completed its crown and has already formed ~2mm of root. The I2 is slightly less advanced. The canine crown is still developping. P3 and P4 are still forming their cuspal enamel. M1 has complete crown and started root formation.
-P. boisei KNM-ER 1820: 2.5-3.1 years 9 . In the left mandibular corpus of KM-ER 1820 (Suppl. Fig. S5), the I1 has formed >2mm of root, the canine and P4 crowns are developing, while the P3 has completed its cuspal enamel. The M1 has started forming its roots although the furcation level has not been attained yet. This makes it slightly more advanced than GAR IVE.
-P. boisei OH 30: 2.7-3.2 years 9 . OH 30 consists in isolated teeth 9 : 3 deciduous teeth and 13 developing permanent teeth with virtually no root formed yet (no photo nor radiographs). The I1 is just complete while the I2 and C crowns are still forming. The M1s have completed crown formation and, in some places, 1 mm of root can be observed. The P3 and P4 have finished forming their cuspal enamel.
-A. africanus Sts 24: 4.35 years 7 . GAR IVE is developmentally less advanced than Sts 24 (Suppl. Fig. S8), especially regarding its M1 root which has not formed until the furcation level. In Sts 24, the C, P3 and P4 are at similar stages than in GAR IVE. To note that GAR IVE's I2 is more advanced in having initiated its root formation. -Early Homo KNM-ER 820 : 5.3 years 11 , and KNM-ER 1507 9 . For both early Homo specimens, no histological age at death was estimated, yet, base don perikymata counts on an I2 and an estimated age of initiation, KNM-ER 820 was estimated to be 5.3 years 11 . The following descriptions follow the data provided by Dean (1987) 9 .
KNM-ER 820 (Suppl. Fig. S9) is a fairly complete mandible with a mixed dentition. The four deciduous molars are preserved although worn because in functional occlusion. The deciduous canines are slightly less well preserved. Among permanent teeth, 6 are erupted, 8 are still in their crypt. The central incisors are worn to the dentine with root apices still open. Both I2 show minimal wear and also have their root apex open. The canines and the premolars are at a similar calcification stage, with 3.6 mm and 4-5 mm of root developped, respectively. The M1s are in occlusion, with the dnetine horns of the mesio-buccal cusp exposed, and the root apices open. The M2 are not emerged although they have completed their crown and started root formation. The M3 are not formed yet, i.e., the crypt is visible but no cusp could be identified on the radiographs. KNM-ER 1507 (Suppl. Fig. S10) is a fragment of left mandible preserving 3 erupted teeth (dm1, dm2, and M1) and 4 developping teeth in their crypt (C, P3, P4, M2). The canine has 2.5 mm of root formed, the M1 is in occlusion, already slightly worn and its root apices are still open. The P3 has completed crown formation and has developped 2 mm of root. The P4 is at a slightly less advanced stage of development. The M2 is at a stage of calcification similar to the P3, with 1.5-2 mm of root formed. KNM-ER 1507 and 820 have their M2 and premolars developping at the same pace, although KNM-ER 820 has a slightly more advanced M2. Dean (1987) 9 also highlight that both early Homo are developmentally similar to four Paranthropus specimens which are OH30, KNM-ER 1477 (Suppl. Fig. S3), KNM-ER 812 (Suppl. Fig. S4), and KNM-ER 1820 (Suppl. Fig. S5). Yet, Dean (1987) noticed that the early Homo infants are both 'dentally more mature than [these] specimens of P. boisei described [as] being closer to 5 years of age' 9 . (Words in square brackets are our modifications).
Both early Homo specimens are developpementally older than GAR IVE (Suppl. Figs S9 and S10).

Estimating the age of stress events in the enamel of the LRP4 1.3.1. Time of P4 initiation in early Homo
In hominids, both M2 and P4 initiate together ( 4,5,7 ; see Sts 24 in Suppl. Fig. S8).
Dean and Smith 13 show that the Sangiran S7-37 P 4 and the KNM-WT 15000 M 2 initiate at about the same age, shortly after 2 years. Dean 9 described the stage of development of postcanine teeth in juvenile early Homo KNM-ER 820 and KNM-ER 1507 (see Suppl. Figs S9 and S10), and observed that, in each specimen, both premolars seem to parallel their M2 in terms of developmental stage. Dean 14 further notice that M1 and I1 have coinciding calcification stages in Paranthropus and early Homo (See Fig. 2.5 in 14 ). Interestingly in Paranthropus robustus, M2 also initiates shortly after 2 years of age (735 days for ML cusp of SK 62, and 823-888 days for the ML cusp of DNH 108; 7 ). To note that Australopithecus africanus STS 2 has just fused both cusps in its ULP4 at the time of death at 2.52 years 7 . DNH 35, a South African early Homo, has just started the formation of its LRP4 and is at the stage of 3 unfused cusps (See Suppl. Fig. S11). Its age at death was histologically estimated to be 795 days, or 2.18 years 7 . Ramirez Rozzi 15,16 investigated the enamel microstructure of Plio-Pleistocene hominid teeth and concluded that premolars crowns form in 2.54 ± 0.35 years, while molars crowns form in 2.56 ±0.41 years. He further demonstrates that the ratio between the molar and the premolar crown formation times is ~1.01 in Plio-Pleistocene hominids, while it is at 0.77 in modern humans and ranges from 0.6 to 0.75 in great apes 15 . In StW 151, both P4 and M2 seem to reach crown completion at the same time (See Fig. Q in file S1 in 7 ). In Paranthropus robustus DNH 108, both teeth also seem to have competed their crown at roughly the same time, at ~4.6 years for M 2 (See Fig. N in file S1 in 7 ). In DNH 84 (P. robustus), the ULP4 has just fused its two cusps and likely completed its cuspal enamel by the time of death at 816 days or 2.2 years (See Fig. L in file S1 in 7 ).
In the absence of any histological data for GAR IVE, one can assume an initiation of its LRP4 shortly after 2 years of age. Four accentuated lines (S1 to S4) were identified in the cuspal enamel of the LRP4. In spite of thorough attempts to optimize their visibility by tuning rotation in x,y,z and slice thickness, these accentuated lines could not be identified on a slice passing through the true dentine horn tip. Instead, the virtual thick slice passes through an EDJ ridge (Suppl. Fig. S13). The distance of these lines from the EDJ crest was measured. Keeping in mind the assumed achieved cuspal thickness observed in StW151, S1 to S3 would be in the inner cuspal enamel, while the remainder (including S4) would belong to the middle cuspal enamel.

Linear measures stress to dentine horns
Supplementary Figure S13 -Virtual section (200 µm thick) though the lingual cusp of the GAR IVE LRP4. The orientation of the 2D slice is shown as a blue line on the 3D model. The cusp of interest is marked by an asterisk. Four accentuated lines (S1 to S4) were identified in the cuspal enamel, and their distance to the EDJ crest (EDJc) was measured.

Time of formation of the 4 accentuated lines
Abbreviations used to the following sections: EDJc: EDJ crest. S1, S2…, S4: Linear distance in µm between the EDJc and each stress event.  The stress events occur on average every ~1.4 months (Suppl . Table S4), whatever their cause (e.g., illness, food deficiencies).

b) Published enamel daily secretion rates (DSR) for early
Homo In their appendix 1, Lacruz et al. 18 published enamel daily secretion rates (DSR) for two African early Homo specimens. Because the stress in the GAR IVE LRP4 are located in the cuspal enamel, we consider only their cuspal DSR (Suppl. Considering that S1 to S3 are in the inner cuspal enamel, and S4 to death in the middle cuspal enamel, one can apply the local rates provided by Lacruz et al. 18

c) Cross-validation and comparison of the two approaches
The overall mean difference between formation times calculated using Dean et al. 17 's equation and Lacruz et al. 18 's average cuspal DSR remains acceptable, at ~20 days (Suppl. Table S8). The difference gets even smaller when using the local rates ~9 days, the best result being when using inner cuspal enamel DSR.

Time of formation for the 4 stress events (in days)
Dean et al. We choose to further use the times calculated using Dean et al. 17 's equation.

Age of the five stress events
Assuming the initiation time of P4 (~2 years) and that the EDJ crest where the virtual slice was recorded formed ~3 months after the dentine horn tip, these were added to the times of formation of each stress line will yield an age of formation (Suppl . Table  S9). Dean  When superimposed with the lingual cusp of the GAR IVE LRP4, the most developed cusp of the DNH 35 LRP4 matches with the purple stress "S3" in GAR IVE (see 1.3.2. and Suppl. Fig. S12), which, following the DNH 35 growth pattern, would have then occurred at ~the 796 th day of life of GAR IVE (i.e., 2.18 years). Death occurred~77 days later in GAR IVE, which correspond to ~873 days or ~2.4 years in the DNH 35 growth pattern. "S3" shows that the calculated estimation is slightly ahead of the developmental stage observed in other early Homo specimens (here by 6.5 months at most), yet this is may be accounted for by the variability which is not captured here, and by the uncertainties related to our measurements (e.g., the virtual section does not pass through the dentine horn tip).

Surface features
Supplementary Figure S15 -Bone surface modifications on the lingual aspect of the GAR IVE mandible.
Supplementary Figure S16 -Bone surface modifications on the buccal aspect of the GAR IVE mandible showing the bone perforation in place of the mental foramen (red framed insert), slight weathering on the posterior aspect of the corpus. Note the damages (blue arrows, see Table S12) referred to as "Surface pit, Class 1" by Parkinson 19 , showing striations radiating from the outer circumference of the depression. See also Courtenay et al. 20 .

Investigating the role of necrophagous insects in the taphonomic damages observed on GAR IVE.
Dermestids beetles marks on bones represents a particular point in the history of remains, and provides information on the taphonomic processes, including possible season of death 21 . Their occurrence noticeably indicates carcass exposure under warm conditions for several weeks to months and the presence of desiccated tissues during this period 21 .
These insects produce various traces, like surface tunnels, pits, and bore holes 19,22 . Such traces have been reported on dinosaur, rhinoceros and Stegomastodon bones 22-26 , Bison latifronshorns 21 and human bones from Bronze Age 27 .
Larder beetles (Coleoptera: Dermestidae) are small (0.5 to 1 cm long) necrophagous beetles with a worldwide distribution. Hundreds of species have been reported worldwide, but less than 10 are currently observed on human remains 28 . They are mostly observed on dry corpses, particularly in indoor cases 29,30 or dry outdoor environment 28,31 . On the contrary, dermestid activity does not occur on submerged or buried carcasses 28 . Both larvae and adults are negatively phototrophic and most activity takes place in shaded places.
The larvae feed on dry necrotic tissues, yet bury themselves in a safe place before each molt. After 7 to 8 molts, they stop feeding and dig pupation chambers for nymphosis 32 . These chambers are excavated into adjacent compact surfaces including hard materials such as wood and bone 33 . Two kinds of dermestid traces on bones have been observed: osteophagy and burying of pupation chambers; both can occur at the same time and places 32 .
Martin & West 21 defined the following criteria to diagnose dermestid pupation chambers: (1) The creation of pupation chambers is always associated with presence of carrion.
(2) Pupation chambers on a single bone and necessarily of a given class are uniform in shape and fall within a certain size range (do not vary greatly in size).
(3) Pupation chambers should generally be found in closely associated groups.
(4) Because chambers are usually formed under desiccated tissue that later decays, they are usually found as half casts rather than completely enclosed burrows.
(5) Chambers are excavated by the mandibles and their surface tends to be covered with microscopic depressions representing mandible bites. These are expressed as small bumps on the casts (these are best studied as casts of the original chamber).
(6) Pupation chambers usually are not much longer than the length of the pupa itself. They are flask shaped with the opening slightly constricted and the main chamber expanded slightly to fit the pupa.
(7) Dermestids, within a species, are fairly uniform in size and the girth of the chamber is a reliable clue for determining if traces found on bone are produced by dermestids.
A close examination of the GAR IVE early Homo mandible allows scoring the following features (Suppl. Presence of a pre-existing vascular hole (mental foramen).
They may have been removed by taphonomic processes or cleaning. 6 The hole is quite deep, not bottle-shaped.
Presence of a pre-existing vascular hole (mental foramen).

7
Internal width of the chamber: 3.4-5 mm Internal width of the chamber compliant with previous observations (Supplementary Fig.  S17), but the outer width is larger than expected (8.02 mm). However, there was a pre-existing vascular hole (mental foramen). Parkinson 19 also stressed that only the co-occurrence of varying damage types allows for dermestids to be addressed as a variable for bone modification in a fossil faunal assemblage.
Accordingly, and to determine the agent at the origin of the bone perforation seen on the GAR IVE early Homo mandible (Figs. 2, 4), a taphonomic study of the associated faunal remains was undertaken. The small sample associated with the early Homo mandible originates from layer E and was uncovered during field seasons 2005 and 2009 as well as 1974 and 1975. It was analysed by Fiore and Tagliacozzo 34 (Suppl . Table S11) and is best described as a palimpsest.
However, based on very characteristic damages, 22 specimens can be attributed to the same time-slice within the chain of taphonomic events. Following the nomenclature outlined by Parkinson 19 , one can observe surface tunnels (on 9 specimens), bore-holes (on 3 specimens) and surface pits (on 18 specimens). Three specimens showed all three damage-types represented (in red font in Suppl. Table S12).
To note that three specimens show gnawing damage by a small carnivore. Clear indication for bone modification by hominins is not evident.

Reconstruction of the palaeoenvironment. (Suppl. Text S3)
Melka Kunture (8°42'N; 38°35'E) is located 60 km south of Addis Ababa, in a tectonic depression of the Upper Awash Valley of Ethiopia. This is a cluster of paleontological and archaeological sites within a volcanic area active all over the Early and Middle Pleistocene. During Quaternary times, the meandering paleo-Awash River transported and deposited sediments with variable grainsizes. Tephra generated by nearby volcanic eruptions accumulated in the area, clogging up the streams, contributing to the formation of shallow ponds and pools. Ephemeral streams currently erode the Pleistocene deposits and cut gullies, such as the Garba gully, where the site named Garba IV is located, exposing the stratigraphic sequence in natural incisions and allowing identification of archeological deposits which are the focus of excavations. Sedimentological facies suggest a fluvial setting where sedimentation was controlled by the reworking of volcanic material eroded upstream. Fluvial sedimentation of the paleo-Awash alternated with primary volcanic distal fallouts of ashes from the Early Pleistocene Melka series 36 .
In Early Pleistocene times, the meandering upper paleo-Awash flowed through a gently undulating landscape 37 . The fluvial channel was slightly further south than the modern river course. The exposed stratigraphic sequence of the Garba IV site starts with silts and sands interbedded with rhyolitic volcanic ash layers and pumice lapilli providing evidence of low-energy fluvial sedimentation in an active volcanic area (Fig. 1c). Mottles suggest periodic backwater. Above, a deposit of pebbles within a sandy matrix (Layer E) is interpreted as the lag deposit of a fluctuating seasonal river. Abundant lithic implements and faunal remains were discovered within this layer. The hominin mandible, which laid on the margin of the river bank, is also part of the record (Fig. 1d). A thin sand deposit capped Layer E and was overlain, in turn, by a primary, partially eroded rhyolitic ash layer, dated at ⁓1.7 Ma 38 .
The fluvial-volcanic context of the Melka area, with its high sedimentation rates provided essential conditions for fast burying of the GAR IVE remains. Acid rains and water could also have affected fossilization processes. In active volcanic regions, ash falls affect both the chemistry of rain and ground waters 39,40 . Waters contaminated by ash leaching are typically enriched in chloride, sulphate and fluoride anions 41 . In modern volcanic settings, ground waters can reach pH as low as 2 42 . In analogy to modern processes, when the late Oldowan layer containing the infant mandible was immersed in a humid environment, it was permeated by water with reduced pH due to leaching through the volcanic ash layers and contamination by volcanic aerosol derived from degassing and eruptions of the volcanoes at short distance from Melka Kunture 40,41 .

PPC-SRµCT at the ESRF of GAR IVE.
The GAR IVE mandible was scanned using PPC-SR-μCT on the beamline ID 19 at the ESRF (Grenoble, France). First, an overview scan of the whole specimen was performed at ~25 µm, then the dentition was also scanned at ~6 µm (Suppl. Volumes were reconstructed using a filtered back-projection algorithm (PyHST2 software, ESRF), and Paganin's approach ( 43 , see 44 for an application at the ESRF). Original 32-bit stacks were converted into 16-bit tiff stacks. The subscans were then concatenated, ring artifact correction was applied as required, and the concatenated subscans were cropped to define the final size (bounding box) of the dataset. Data were saved as JPEG 2000 with a compression factor of 10 (this format minimizes the loss of data quality). Volumes were reconstructed using a filtered back-projection algorithm, beamhardening and ring artifacts correction were applied as needed. Data were saved as a stack of tiff files. The scans were resampled at 15 µm and filtered using a median and a Kuwahara filter (kernel size of 3 for each; 45 ) to facilitate the segmentation in Avizo 6.3.1 (VSG). Enamel thickness was visualized in 3D by computing distances points by point between the enamel and dentine surfaces.