Dental cementum virtual histology of Neanderthal teeth from Krapina (Croatia, 130–120 kyr): an informed estimate of age, sex and adult stressors

1. Introduction

The factors that have shaped Homo sapiens' life history are largely still to be understood. Given our significant reliance on material culture and on (pro)social behaviour for the acquisition of energy, care of offspring and many other correlated aspects, it is likely that our life history, cultural and social evolution are intertwined [1]. Life-history variables (LHVs) are those variables that directly record the timing of the several major life-history events, such as gestation length, age at weaning, age at sexual maturity, age at first birth, interbirth intervals, age at menopause and longevity. Nonetheless, tracking the evolution of LHVs in the fossil record has proven to be a significant challenge as it is limited to what can be gleaned and inferred from teeth and bones [2,3]. The current knowledge regarding LHVs in the hominin fossil record is very scant, being limited to estimates of age at death [4], gestation length, onset of weaning [5] and physiological stress occurrence [6]. The present work aims to improve and enrich our understanding of some of the more elusive LHVs using a non-destructive cutting-edge methodology.

Human teeth permanently record physiologically impactful events during their formation, enabling the collection of data about stressors experienced during development in addition to the estimation of the age at death [7]. In Recording structures of mammals, Klevezal [8] thoroughly describes the different types of growth layers pertaining to the mineralized tissues of bone, dental enamel, dental cementum, horn and other mammalian mineralized tissues. All recording structures are periodically layered, and each layer is the visible result of changing micromorphology, which corresponds to changing physiological states of the organism [9]. These structures are extremely conserved across mammalian taxa, from cetaceans to cervids to primates. Recording structures can be categorized according to three parameters [8]: sensitivity, which relates to the frequency with which new layers are added and which can vary across the period of registration [10]; period of registration (when does the structure start and stop recording); and persistence of a record (for how long the recording structure is maintained unchanged) (table 1).

Of the mineralized tissues present in primates, only cementum retains both the lifetime period of registration and lifetime persistence of the record, even if it is with a low sensitivity owing to its yearly periodicity and low growth rate [11]. Physiological events that are strong enough to produce a disruption of organismal mineral homeostasis or organic matrix synthesis leave marks in the corresponding position of the developing cementum, which are visible as changes in tissue refractive index [12–15]. Previous studies already ascertained the possibility of using phase-contrast synchrotron X-ray microtomography (PC-SRµCT) for a non-destructive three-dimensional (3D) visualization of yearly incremental cementum bands in both modern human and fossil animal teeth [16–19], which provide the basis for age-at-death estimation. Furthermore, recent work in classic histology [12,20] has provided evidence, in two distantly related species of primates (H. sapiens and Macaca mulatta), of changes in tissue microstructure in relation to pregnancies, cessation of fertility and illnesses. Fertility data derived from human teeth are scarce, constituting an almost unexplored area of research, particularly when dealing with hominin fossil teeth. The combination of age-at-death estimation provided by the yearly periodicity and the markers associated with physiological stressors can be used to determine their chronology, thus attempting a first informed estimate in fossil remains regarding the timing of possible parturitions [21]. Thus far, studies attempting to investigate them [12,21,22] have relied on destructive methods which are substantially prevented in palaeoanthropological specimens for obvious preservation reasons. To date, non-destructive virtual histology studies based on the use of PC-SRµCT techniques are increasing in number [23–26] but remain scarce in the exploration of dental cementum and non-existent in the investigation of fossil hominin cementum.

Here, we report for the first time the application of this approach to the cementum of Neanderthal teeth, specifically from Krapina (Croatia, 130–120 kyr [27]). Using synchrotron-based high-resolution 3D imaging, we acquired data on 10 Neanderthal specimens, in addition to one Mesolithic, one Neolithic (both from the Danube Gorges area of Serbia) and nine known life-history contemporary humans. In addition to estimating the age at death following previously published methods [16,18], we attempt a first informed estimate of sex and reproductive events and/or other types of stressors for the fossil specimens.

2. Methods

2.2. Image acquisition and processing

The permanent tooth roots were measured via SRμCT at the SYRMEP beamline of the Elettra synchrotron facility (Basovizza, Trieste, Italy) in August 2020. The measurements were performed in propagation-based phase-contrast mode using a filtered white X-ray beam (filter = 1.5 mm Si + 1.0 mm Al) with a mean energy of 27 keV. After the alignment of the μCT set-up, a series of scout scans were first acquired on one of the samples in order to optimize the experimental conditions to be used for the high-resolution measurements. For each tomographic scan, 1800 sample radiographs were recorded during continuous rotation over a total scan angle of 180°. The sample-to-detector distance was set at 150 mm. Samples were imaged with a water-cooled, 16-bit scientific complementary metal–oxide–semiconductor (sCMOS) macroscope camera whose effective pixel size was set at 0.9 × 0.9 µm², yielding a maximum field of view of about 1.85 × 1.85 mm². The detector system is equipped with a motorized quadrupole hosting four different scintillator screens. A 17-μm-thick GGG:Eu scintillator screen (ESRF, Grenoble, France) was selected for our experiment. The exposure time/projection was 3.5 s. For each tooth, we imaged at least one region of interest (ROI) located at approximately one-third of the tooth root from the cervical enamel. We reconstructed the volumes using the filtered back-projection algorithm provided by the SYRMEP Tomo Project (STP) software [33]. Prior to slice reconstruction, a single-distance phase-retrieval algorithm [34] was applied to the sample projections with a d/b parameter (the ratio between the real and imaginary parts of the complex refraction index of the material) in the range of 10–17 depending on the specific sample features.

Of the 21 teeth imaged, one contemporary human (ARF36) and six Neanderthals (KR7, KR25, KR176, KR1, KR137 and KR176) did not present enough cementum preserved in the region selected to offer consistent LHV estimates. Moreover, PC-SRμCT data revealed that extensive diagenetic damage had clearly occurred to the cementum of the Mesolithic specimen (H326/7) so no further analyses were carried out on this tooth. Consequently, we carried out data analyses on the remaining 14 teeth: eight contemporary humans, five Neanderthals and one Neolithic individual (table 2).

Virtual two-dimensional (2D) sections were extracted from the reconstructed 3D volumes using the open-source FIJI ImageJ [35]. To ensure maximum visibility of the tooth cementum annulations (TCAs), using the reslice tool in FIJI ImageJ, the virtual sections were made perpendicular to the root surface [36], rather than to the tooth axis, as, in the latter case, the section would cut the cementum layers in an oblique fashion. To minimize the risk of analysing a non-representative area, and to factually sample in three dimensions, even if by discrete increments, we made five virtual sections for each tooth (figures 1 and 2). For each of the 70 2D virtual sections, we obtained (electronic supplementary material, script S1) between 15 and 25 linear plots (depending on the length of the cementum segment) of the grey-level value of pixels across the entire thickness of the cementum (see electronic supplementary material, methods for a more detailed description). Finally, we saved the data of the plots, with the x-axis representing the distance (in years) from the cementum–dentine junction (CDJ), which was manually identified, and the y-axis the grey-level pixel value [20].

• Download figure
• Open in new tab
• Download PowerPoint

• Download figure
• Open in new tab
• Download PowerPoint

3. Results

3.2. Sexual dimorphism

The results of the recursive partitioning (electronic supplementary material, figure S3) on the known-sex samples indicate that the average width of positive peaks is the best discriminating variable between the sexes regardless of noise threshold (0.3*Std or 1*Std). Based on this parameter, the misclassification rate for females is 4%, while for males it is 40% (electronic supplementary material, figure S4). This difference in misclassification rate is likely to be due to the larger number of female specimens than male specimens. For this variable, the range for females is 22.7–32.15, while for males it is 18.28–26.29 (figure 3). In addition, several other variables (electronic supplementary material, table S2) appear to be significantly different between the two sexes based on the results of the t-tests. Since some of the variables that were significant at 0.3*Std were not significant at 1*Std, we only use the ones significant in both settings to make our informed suggestions regarding the sex of the fossil specimens. We, therefore, used the areas of non-overlap in values between the male and female ranges (thus excluding the range of 22.7–26.29) to suggest a specimen as either male or female. The misclassification rate on the modern specimens based on non-overlapping ranges is 0% for both males and females, with 54.3% of the cases resulting in no estimate being made (electronic supplementary material, figure S5). By suggesting an estimation of sex only if the value fell within the exclusive (non-overlapped) range of either sex our results indicate that the Neanderthal fossils KR172 and KR179 might be females and the Neolithic specimen A2309 is female, while for the remaining three Neanderthal specimens we make no informed suggestion (table 4). Our analyses correctly predicted the sex of A2309.

• Download figure
• Open in new tab
• Download PowerPoint

Table 4. Suggestions of sex of the fossil specimens based on the average width of positive peaks. We report the value for each specimen at two different noise thresholds. The partitioning value derived from the recursive partition is 23.13; the range for known-sex females is 22.70–32.15, while for males it is 18.28–26.29. The ranges of no overlap are 18.28–22.69 for males and 26.30–32.15 for females. For the Neolithic specimen, the sex was known through DNA analysis and our results correctly predict it. For specimens that had a value falling within the overlap male and female range no suggestion is made.

Collapse

species	specimen	average width of positive peaks at 0.3 * Std	average width of positive peaks at 1 * Std	sex estimated based on range	known sex
H. neanderthalensis	KR162	25.834	25.834	/	/
H. neanderthalensis	KR172	27.394	27.394	F	/
H. neanderthalensis	KR179	34.486	34.486	F	/
H. neanderthalensis	KR27	25.872	25.872	/	/
H. neanderthalensis	KR85	25.268	25.268	/	/
H. sapiens (Neolithic)	A2309	27.038	27.038	F	F

3.3. Age at event occurrence

We found that there is a strong and statistically significant positive correlation (R² = 0.9874; p < 0.05) between known and inferred ages at physiological stressors (pregnancy, illness, etc.) in the modern reference samples. The inferred ages together with the known ages at event occurrence are reported in electronic supplementary material, table S3. The intercept value (y = 0.1063) of the equation of the regression line reported in figure 4 (y = 1.0034x + 0.1063) indicates that there is no systematic error towards over- or underestimating the ages at event occurrence. The average absolute difference between known and inferred ages in the estimates of all known events is 1.9 years; the RMSE is 2.1 years. The false-positive and false-negative rates differ considerably depending on the final parameter used in the analysis. With the cut-off threshold at 40% of the maximum frequency, the false-negative rate is 0%, meaning that all known events are detected; however, the false-positive rate is extremely high at 55.2%. When implementing a stricter cut on the final output (30% of the maximum frequency), the false-negative rate increases to 10.3% while the false-positive one logically decreases to 34.5% (table 5). The inferred ages at event occurrence for the fossil specimens are reported in table 6. Only one event was detected for each specimen.

• Download figure
• Open in new tab
• Download PowerPoint

Table 5. For each of the two settings of the final event detection analysis (see Methods for details): the associated average absolute difference between known age at event occurrence; the RMSE; and the false-negative and false-positive rates.

Collapse

setting	avg absolute difference	RMSE	false negative (%)	false positive (%)
30% of maximum frequency	1.93	2.13	0	55.17
40% of maximum frequency	1.91	2.1	10.34	34.48

Table 6. Results of the inferred ages at event occurrence for the fossil specimens (KR are the Krapina Neanderthals). Ages at cementum initiation and age at death are estimated (see §2.3.1) and reported in table 3. Suggested sex is reported as resulting from our analyses (table 4).

Collapse

specimen	age at event	sex	age at death
KR27	13.94	/	17.03
KR85	24.04	/	24.73
KR162	23.84	/	25.95
KR172	15.61	F	21.38
KR179	23.55	F	26.25
A2309	15.52	F	16.25

4. Discussion

Our age estimates match fairly well with those based on dental wear, even though there are some potential differences from the KDP assignation made by Wolpoff [29]. Nonetheless, as previously reported [16] there is a non-negligible margin of error regarding age estimates based on TCA counts. The difference between known and inferred age, for individuals less than 50 years of age, is reported to be between 6.8 and 6.5 years [16]. Future works could benefit from recently developed semi-automatic TCA count methods [49,50] in order to eliminate inter- and intraobserver error. Additionally, because of the inhomogeneous nature of cementum, tracking the growth increments via 3D virtual reconstruction could help reduce error in age estimates. The age at which cementum begins formation is another source of error, which in the future could be addressed by acquiring scans of whole teeth to enable precise measures of crown and root formation times. Hence, the results of our study are not conclusive enough to add meaningful information to the KDP assignation. It is worth noting that, although our sample size is small, there does not appear to be any systematic over- or underestimation of age, compared with the one derived from dental wear.

The exploratory research regarding sexual dimorphism in cementum microstructure has provided insights into a potentially novel area of research. Based on our small sample and preliminary observations, it appears that the width of peaks in the grey-level profile of cementum transects might be sexually dimorphic, with males presenting on average narrower event oscillations than females. This could reflect the rapidity (frequency of peaks also seems to be sexually dimorphic) in changes in tissue deposition and could be caused by underlying metabolic differences between the sexes, which in turn are likely to be correlated with body size sexual dimorphism since metabolic rate scales with body size [51]. Indeed, research has shown that inner fundamental metabolic rhythms reconcile and regulate tissue formation and organ masses across primates [52,53]. This is reflected, for example, in the sexually dimorphic periodicity of repeat intervals (RIs) in enamel formation such that the average Havers–Halberg oscillation RI values are lower (hence frequency is higher) for males than for females [54]. Accordingly, other works have demonstrated the presence of sexual dimorphism in bone accretion with an increased turnover in males, which suggests a higher metabolic activity in this sex [55,56]. Although our study has a small sample size, we were able to correctly estimate the sex of a Neolithic fossil tooth. Our preliminary results seem to confirm recent research [18], which has reported evidence of sexual dimorphism in the cementum of another primate species (M. mulatta). It is therefore likely that there is sexual dimorphism in the microstructure of this tissue, but systematic exploration on a larger sample is necessary in order to better understand it, especially across different taxa, which in turn have differing degrees and forms of sexual dimorphism (e.g. macaques have sexually dimorphic canines while humans do not).

While previous research has confirmed that physiologically impactful events, such as those relating to reproduction, are permanently recorded in cementum microstructure and can be accurately timed [12–14,20], this is the first study to demonstrate that it is possible to obtain such information in a non-destructive manner. This is of tremendous importance for its applicability to fossil specimens. Indeed, the results of our analyses support the hypothesis that physiologically impactful events are detectable using the PC-SRµCT technique. However, the high false-positive rates of the known-event sample indicate that most events are detected but not all detected events are necessarily real. Nonetheless, it is important to note that we cannot be completely certain that false positives are not actually real events, which the anonymous participant omitted or forgot to include in their questionnaire, as documented by previous research that has addressed the problem of retrospective reporting [57,58]. Additionally, it is possible that events subjectively deemed insignificant and not directly related to health and reproduction (which is what our questionnaire assessed) actually have significant effects on cementum deposition. Indeed, recent research [59] has shown that in mandrills (Mandrillus sphinx) events such as changes of alpha male within a group correlate with an accentuated line in enamel.

When applying our event detection algorithms to our sample of fossils, our results (table 6) indicate only one physiologically impactful event per individual. While our known-event specimens had a high false-positive rate, they also had a 0% false-negative one, meaning that we should expect all known events to be detected. It is likely that the physiology of modern urban-dwelling people is more perturbed and less tightly correlated to seasonal rhythms than that of Neanderthals, leading to an increased amount of perturbations in the cementum that are not correlated to any known stressors. Additionally, the results of our sex, age and event estimations combined could suggest that the physiologically impactful events reported for A2309 (15.5 years) and KR172 (16.6 years) could reasonably relate to menarche, which in contemporary hunter–gatherer populations has a mean of 16.6 and a median of 17.1 years [60], and is expected to be earlier for Neanderthals, who had an overall accelerated life-history scheduling compared with H. sapiens. As for KR179, for whom we report only one physiologically significant event at 23.6 years, we are inclined to suggest the possibility of it relating to a reproductive event or severe illness. However, we lack an explanation for the absence of a menarche signal, if this is indeed a female.

However, it is necessary to acknowledge that all inferred events are between 0.7 and 5.4 years before death, making it possible that the physiologically stressful events detected close to death were possibly correlated to it. Notwithstanding, as already acknowledged by several recent works [12,21], it is not possible to infer the cause of stressor from the histological analyses of cementum alone. It is likely that elemental analyses would provide more event-specific information, in that each different type of stressor alters the mineral homeostatic balance in a different way. However, that type of data is not yet available using non-destructive methods. Future research aiming at collecting elemental data and teasing apart the cause of the stressor is necessary, such as synchrotron X-ray fluorescence mapping [61] on the cementum of fossil teeth having natural broken roots.

In conclusion, the present work provides tentative evidence that sex can be recovered from cementum microstructure, and robust evidence that the chronology of physiologically impactful events is both preserved in cementum microstructure and detectable using non-destructive 3D imaging methods. Future research will hopefully build on the present work and further advance our knowledge.

Ethics

The research was carried out under the auspices of the College of Medicine Research and Ethics Committee (COMREC) (protocol no. P.05/06/373). All methods were carried out in accordance with relevant guidelines and regulations and all experimental protocols were approved by the NYU College of Dentistry.

Data accessibility

The data are provided in the electronic supplementary material [62].

Authors' contributions

P.C.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, software, validation, visualization, writin—original draft, writing—review and editing; A.N.: data curation, formal analysis, funding acquisition, investigation, methodology, project administration, writing—review and editing; D.R.: data curation, investigation, resources, writing—review and editing; D.B.: resources, writing—review and editing; L.C.: data curation, formal analysis, methodology, software; T.B.: data curation, formal analysis; G.R.: resources, writing—review and editing; D.W.F.: supervision, writing—review and editing; A.P.K.: data curation, investigation, methodology; L.B.: formal analysis, investigation, methodology, project administration, supervision, writing—review and editing; L.M.: data curation, investigation, methodology, project administration, supervision, writing—review and editing; T.G.B.: data curation, formal analysis, funding acquisition, investigation, methodology, project administration, supervision, validation, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed herein.

Competing interests

We declare we have no competing interests.

Funding

The research was supported by a National Science Foundation Graduate Dissertation Research Improvement Grant (to P.C., grant no. 2018357), by the NYU MacCracken Fellowship (to P.C.) and by the Marie Skłodowska-Curie Actions Individual Fellowship (to A.N., grant no. H2020-MSCA-IF-2018-842812).