Assessing human weaning practices with calcium isotopes in tooth enamel

Significance

The practice of weaning, the dietary transition from exclusive breastfeeding to exclusive nonmilk food, is a key aspect of development and evolution of hominins, but its study in the fossil record is hampered by a lack of unambiguous biomarkers. Ca stable isotope ratios of skeletal remains are expected to bear information about milk consumption. Here we demonstrate that modern human tooth enamel records a temporal variation of Ca isotope compositions, which is related to breastfeeding duration. Ca isotopes could be used as a biomarker for reconstruction of weaning practices in past human and fossil hominin species.

Abstract

Weaning practices differ among great apes and likely diverged during the course of human evolution, but behavioral inference from the fossil record is hampered by a lack of unambiguous biomarkers. Here, we show that early-life dietary transitions are recorded in human deciduous tooth enamel as marked variations in Ca isotope ratios (δ^44/42Ca). Using a sequential microsampling method along the enamel growth axis, we collected more than 150 enamel microsamples from 51 deciduous teeth of 12 different modern human individuals of known dietary histories, as well as nine enamel samples from permanent third molars. We measured and reconstructed the evolution of ⁴⁴Ca/⁴²Ca ratios in enamel from in utero development to first months of postnatal development. We show that the observed variations of δ^44/42Ca record a transition from placental nutrition to an adult-like diet and that Ca isotopes reflect the duration of the breastfeeding period experienced by each infant. Typically, the δ^44/42Ca values of individuals briefly or not breastfed show a systematic increase during the first 5–10 mo, whereas individuals with long breastfeeding histories display no measurable variation in δ^44/42Ca of enamel formed during this time. The use of Ca isotope analysis in tooth enamel allows microsampling and offers an independent approach to tackle challenging questions related to past population dynamics and evolution of weaning practices in hominins.

The reconstruction of weaning practices, the dietary transition from exclusive breastfeeding to exclusive nonmilk food (1), is fundamental in the study of past populations and in human evolution. Weaning constitutes a major determinant in health and survival of mammals (2–7). On the one hand, breast milk provides offspring with a safe and easily digested source of nutrients and energy together with immunological protection (5, 7–9). On the other hand, transition to nonmilk food, which supplements milk in the course of weaning, possibly exposes infants to exogenous pathogens and energy shortfalls, although its introduction is necessary to meet the growing requirements of offspring (3, 8–11). Hence, the timing of this transition constitutes the biological and behavioral pivot of a trade-off between increased juvenile survival and the recovery of maternal reproductive ability, which is delayed by lactational amenorrhea (5, 8, 9, 12, 13). Study of weaning practices can thus help characterize health, fertility, and demography of present and past human populations (5, 7, 14).

Weaning behavior is also a determinant trait in developmental biology and in evolution of life-history strategies of mammals, and humans in particular (5, 9, 12, 15, 16). Nonindustrialized modern humans are characterized by younger ages at cessation of suckling (i.e., ages at weaning) than those of great apes, namely orangutan (Pongo spp.), gorilla (Gorilla spp.), and their closest relatives, chimpanzees and bonobos (Pan troglodytes and Pan paniscus) (5, 9, 15–20). Contrary to great apes, human infants are fully weaned before independent feeding, which allows provisioning offspring with solid and processed food (5, 16). This early weaning practice is associated with other specific life-history traits, such as a later age at first female reproduction, shorter intervals between births, extended postmenopausal longevity, and a longer lifespan (5, 16, 21).

Study of past human populations including health, demography, and evolution is partly hampered by a lack of direct evidence of weaning behavior in archaeological and fossil settings. Predictions from life-history theory and indirect morphological or histological markers bring little solid insight into past weaning practices (9). Variations in chemical and isotopic composition of bone, tooth enamel, or dentine can bring information on weaning practices. Despite possible effects of dietary transition on carbon and oxygen isotope ratios of skeletal remains (see review in ref. 14), the most widely accepted biomarker for weaning practices is the nitrogen isotope ratio measured in hair, fingernails, bone, or dentine collagen (22–25). The Sr/Ca and Ba/Ca elemental ratios in tooth enamel and dentine have also proved relevant for reconstructing early-life dietary transitions (1, 26, 27). Nevertheless, these various methods are possibly associated with one or several drawbacks. The main concern is that the isotopic and elemental ratios are possibly contaminated or modified during diagenesis, depending on the burial context (28). Regarding nitrogen isotope ratios, the problem lies in the fact that the collagen fraction is not preserved beyond 100,000 y at best (29).

Ca stable isotope ratios from tooth enamel offer new perspectives on the reconstruction of weaning practices:

• i)Mammal milk, especially breast milk, has extreme Ca isotope compositions with ratios significantly lighter than dietary intake, ca. −0.60‰ as measured for cattle, ewes, and human (30–32). The δ^44/42Ca values in breast milk lie between −1.50 and −2.00‰ in modern humans (30), whereas the average Western diet is estimated to lie around −1.00‰ (31–33) (Table S1 for compilation). Thus, the transition from exclusive breast milk consumption to a child’s or an adult’s diet should induce a positive shift in δ^44/42Ca values in dietary Ca of the order of +0.60‰.
• ii)Ca makes up 40% in weight of hydroxylapatite, the major mineral phase of tooth dentine and enamel. This allows faintly destructive microsampling (<100 μg of enamel) and thus increases spatial resolution within incrementally structured dental tissues (34, 35).
• iii)The enamel Ca isotope composition shows little sensitivity to diagenesis, even after several million years (36, 37), given that secondary Ca carbonates are leached accordingly.

Table S1.

Ca isotope composition of cow dairy products, human breast milk (30, 31), 15 different infant formulas (this study), human whole blood (47), and estimation of average Ca isotope composition of typical Western diet and vegan Western diet (31–33)

Material	δ44/42Ca (relative to ICP Ca Lyon), ‰ ± 2 SD		n	Source
Cow milk and dairy products
Milk	−1.19	0.14	3	Ref. 30
Whey	−1.12	0.12	2	Ref. 30
Curd	−1.14	0.06	1	Ref. 30
Kefir	−1.11	0.20	4	Ref. 30
Yogurt	−1.22	0.04	1	Ref. 30
Commercial milk	−1.00	0.01	4	Ref. 30
Ultrahigh-temperature milk	−1.19	0.28	5	Ref. 31
Cheese (gouda)	−1.11	0.02	3	Ref. 31
Cheese (emmentaler)	−1.05	0.26	1	Ref. 31
Cheese (brie)	−1.21	0.02	2	Ref. 31
BCR-380R (whole milk powder)	−1.10	0.10	16	This study
Average of cow dairy products	−1.13	0.13		(2 SD, n = 11)
Infant formulas
Blédilait, 0–6 mo	−0.71	0.01	2	This study
Gallia, 0–6 mo	−0.62	0.22	3	This study
Gallia, premature/low-weight infant	−0.82	0.19	3	This study
Guigoz, 0–6 mo, antiregurgitation	−0.22	0.09	6	This study
Guigoz, 0–6 mo	−0.51	0.05	3	This study
Guigoz, premature/low-weight infant	−0.01	0.05	2	This study
Milumel, 6–12 mo	−0.62	0.10	3	This study
Modilac, 0–6 mo, antiregurgitation	−0.62	0.17	4	This study
Modilac, premature/low-weight infant	−0.81	0.13	4	This study
Nidal, 0–6 mo	−0.48	0.20	2	This study
Novalac, 0–6 mo, antiregurgitation	−0.25	0.13	3	This study
Nutricia, 0–12 mo, medical nutrition	−0.69	0.11	2	This study
Picot, 0–6 mo, medical nutrition	−0.58	—	1	This study
Picot, 6–12 mo, medical nutrition	−0.20	0.17	4	This study
Picot, 0–12 mo, lactose-free	−0.19	0.10	2	This study
Average of infant formulas	−0.49	0.51		(2 SD, n = 15)
Human breast milk
Individual A (2 mo after birth), United Kingdom	−1.71	0.28	5	Ref. 30
Individual A (4 mo after birth) United Kingdom	−1.97	0.02	3	Ref. 30
Individual B (1 mo after birth), United Kingdom	−1.54	0.26	1	Ref. 30
Individual C (4 mo after birth), United Kingdom	−1.50	0.02	2	Ref. 30
French individual (1 wk after birth)	−1.63	0.13	5	This study
Average of breast milk	−1.67	0.37		(2 SD, n = 5)
Human whole blood
Baseline patient 1, United States	−1.14	0.15	—	Ref. 47
Baseline patient 2, United States	−0.99	0.16	—	Ref. 47
Baseline patient 3, United States	−1.02	0.19	—	Ref. 47
Baseline patient 4, United States	−1.22	0.16	—	Ref. 47
Baseline patient 5, United States	−1.23	0.18	—	Ref. 47
Baseline patient 6, United States	−0.99	0.15	—	Ref. 47
Baseline patient 7, United States	−1.00	0.06	—	Ref. 47
Baseline patient 8, United States	−0.97	0.24	—	Ref. 47
Baseline patient 9, United States	−1.37	0.09	—	Ref. 47
Baseline patient 10, United States	−1.29	0.08	—	Ref. 47
Baseline patient 11, United States	−1.19	0.10	—	Ref. 47
Baseline patient 12, United States	−1.26	0.01	—	Ref. 47
Average of ref. 47 (United States)	−1.14	0.28		(2 SD, n = 12)
25-y-old male, France	−1.41	0.08	5	This study
39-y-old female, France	−1.43	0.12	9	This study
Average of this study (France)	−1.42	0.03		(2 SD, n = 2)
Average whole blood	−1.28	0.39
Average Western diet components
Dairy products	−1.12	—		Ref. 32
Vegetables	−0.86	—		Ref. 32
Fruit	−0.86	—		Ref. 32
Cereal products	−0.80	—		Ref. 32
Meat	−0.56	—		Ref. 32
Fats	−1.02	—		Ref. 32
Water	−0.18	—		Ref. 32
Typical Western diet	−1.02	—		Ref. 32
Typical vegan diet	−0.83	—		Ref. 32

The hypothesis that Ca isotopes allow tracking intake of human or animal milk was formulated earlier (30). In former studies, the focus was on the possible influence of dairy product consumption on bone Ca isotope composition. Unfortunately, results did not allow animal milk intake to be distinguished from intrinsic biological variability (38–40).

Here, we test this hypothesis by measuring δ^44/42Ca along enamel of human deciduous teeth of modern individuals that were weaned at various known ages. Using a sequential microsampling method along the enamel growth axis, we collected more than 150 enamel microsamples from 51 deciduous teeth of 12 different modern human individuals of known dietary histories, as well as nine enamel samples from permanent third molars. The deciduous teeth set of samples stemmed from healthy individuals with various diet histories, covering three main scenarios: exclusive breastfeeding from birth, exclusive formula feeding from birth, and a breastfeeding period with subsequent formula feeding (Table 1).

Table 1.

Description of individuals’ early life, dietary histories, and sampled teeth

ID	Sex	Year of birth	Gestation length, mo	Breastfeeding		Formula feeding		Age at nonmilk food introduction, mo	Deciduous teeth		Permanent teeth
				Yes or no	Age at end, mo	Yes or no	Age at end, mo		No.	Types
A	Male	1997	9	No	—	Yes	12	4	4	m2 - m1 - c' - i1	—
G	Female	1999	8.5	No	—	Yes	12	6	5	m2 - m1 - c' - i2 - i1	—
L	Male	1971	9	Yes	0.5	Yes	12	3	3	m2 - m1 - i1	—
C	Female	1983	9.6 ± 0.4	Yes	1.7	Yes	>4	3	5	m2 - m1 - m1 - c' - i1	M3*
F	Male	1991	9	Yes	2	Yes	12	4	5	m2 - m1 - c' - i2 - i1	—
K	Male	1993	9	Yes	2.5	Yes	12	4	5	m2 - m1 - c' - i2 - i1	—
H	Female	1997	9	Yes	3.75	Yes	12	4	5	m2 - m1 - c' - i2 - i1	—
B	Female	1992	8.75	Yes	4	No	—	3.5	4	m2 - m1 - c' - i1	—
I	Female	1979	9	Yes	4	Yes	12	5	5	m1 - m1 - c' - i2 - i1	M3*
D	Female	2011	NA	Yes	24	NA	NA	NA	3	m1(L) - m1(R) - m1(L)	—
E	Male	2011	9	Yes	24	No	—	6	4	m2 - m1 - m1 - i	—
J	Male	2009	9	Yes	36	No	—	6	3	m1 - m1 - c,	—
M	Female	1948	—	—	—	—	—	—	—	—	M3*
N	Female	1981	—	—	—	—	—	—	—	—	M3*
O	Male	1985	—	—	—	—	—	—	—	—	M3*
P	Female	1983	—	—	—	—	—	—	—	—	M3*
Q	Female	NA	—	—	—	—	—	—	—	—	M3*
R	Female	1947	—	—	—	—	—	—	—	—	M3†
S	Female	1990	—	—	—	—	—	—	—	—	M3†

Individuals are sorted according to age at end of breastfeeding. NA, not available.

^*Described in ref. 59.

^†Described in ref. 34.

Results

All 163 enamel δ^44/42Ca values vary around a median value of −1.75‰ and range from −2.28‰ to −1.30‰, representing the very lower end of the natural accounted-for variability of Ca isotope compositions (Figs. S1 and S2 and Dataset S1).

The lowest δ^44/42Ca values were measured in the group of prenatal enamel samples, with a median of −1.87‰ and values ranging from −2.28‰ to −1.51‰ (n = 51). This is in significant contrast with the higher values of the wisdom teeth group (from −1.73 ‰ to −1.34‰, average value of −1.58‰, n = 9) and the postnatal enamel group that covers a wider range of values (from −2.15‰ to −1.30‰, median value of −1.70‰, n = 84). The total range of values from enamel sampled on the neonatal line is indistinguishable from prenatal enamel, with values ranging from −2.11‰ to −1.61‰ (n = 16).

We observe significant differences among these groups (Fig. 1A; Welch’s ANOVA, P < 0.001; Kruskal–Wallis, P < 0.001). More precisely, we observe a transition of Ca isotopic compositions toward ⁴⁴Ca-enriched values, from prenatal to postnatal development stages (Welch’s t test, P < 0.001; Wilcoxon–Mann–Whitney, P < 0.001). For each measurement in a given individual, we can define a ∆^44/42Ca value, given by the difference between the δ^44/42Ca value of a considered spot and the average δ^44/42Ca value of an individual’s prenatal enamel. We observe for infants with no or short breastfeeding histories (less than or equal to 4 mo) an increase of the ∆^44/42Ca value from pre- to postnatal development stages (Fig. 1B; Welch’s ANOVA, P < 0.001; Kruskal–Wallis, P < 0.001). Conversely, we do not observe for infants with long breastfeeding histories (more than 12 mo) any significant increase of the ∆^44/42Ca value from pre- to postnatal development stages (Fig. 1C; Welch’s ANOVA, P = 0.76; Kruskal–Wallis, P = 0.66). At the individual level (Fig. 2 and Fig. S3), a systematic and significant increase of the δ^44/42Ca value is observed for individuals with short or no breastfeeding history except potentially for one individual (C). No significant difference is observed for infants that were breastfed longer than 12 mo. Parametric and nonparametric statistical analyses of differences between aforementioned categories for both δ^44/42Ca and ∆^44/42Ca values were performed using R software (41) and are summarized in Table S2 and Fig. S3.

Fig. 1.

(A) Box plot of δ^44/42Ca values (per mil, ICP Ca Lyon) in enamel sorted by development stage categories for all individuals, corresponding to prenatal development period, birth, postnatal development period, and wisdom teeth enamel formation (i.e., 7–15 y of age). Error bar represents average 2 SD. (B and C) Box plots represent ∆^44/42Ca values (δ^44/42Ca_sample − δ^44/42Ca_{prenatal average}, for a given individual) calculated for individuals with short breastfeeding histories (B, less than or equal to 4 mo, n = 9) and longer breastfeeding histories (C, longer than 12 mo, n = 3). P values of Welch’s unequal variance t tests, Wilcoxon–Mann–Whitney test, Welch’s ANOVA, and Kruskal–Wallis tests are given here and summarized in Table S2. NS, nonsignificant P value; *P = 0.01–0.05; **P = 0.001–0.01; and ***P < 0.001.

Fig. 2.

(A–L) Temporal evolution of δ^44/42Ca values (per mil, ICP Ca Lyon) in each individual. Individuals are sorted by increasing age at breastfeeding cessation (months). Color scale corresponds to age at cessation of breastfeeding. Vertical error bars are 2 SD and horizontal error bars are estimated time envelopes. Black dashed vertical lines mark birth; red shaded areas cover periods of introduction of nonmilk food (age at diversification ± 0.5 mo).

Table S2.

Summary of statistical analyses performed on dataset between main groups

Discussion

Ca Isotope Composition of Adult and Prenatal Diet.

The estimated average δ^44/42Ca value of the Western diet lies around −1.00‰ (31–33), and the physiological processing of Ca results in a shift of δ^44/42Ca from diet to bone of −0.60‰ (∆_bone-diet) on average (30). The ∆_bone-diet value is well conserved among adult mammals, including horses, seals, mice, deer, sheep, and Göttingen minipigs (30, 42–45) and is assumed to be comparable in humans (46, 47). The average δ^44/42Ca value of wisdom tooth enamel that we measured in nine adult individuals (−1.58 ± 0.26 ‰, 2 SD, n = 9) is thus in good agreement with a hypothetical average Western diet composition of −1.00‰. An individual 7–15 y of age has a diet near or identical to that of an adult and this is thus compatible with our observations in wisdom tooth enamel, known to grow during this period.

Enamel that initiates formation early, that is, during the second and third trimesters of in utero development (48), is characterized by a ⁴⁴Ca-depleted isotope composition compared with third molar enamel (Welch t test, P < 0.001; Wilcoxon–Mann–Whitney, P < 0.001) and has a mean δ^44/42Ca lower than that of third molars by −0.31 ± 0.11 ‰ (Welch t test 95% confidence interval). Such a ⁴⁴Ca-depleted isotope composition in fetus enamel has several possible and likely combined causes. First, compared with the estimated −1.00‰ mean value of diet, the mother’s blood has lower δ^44/42Ca values: a compilation of available data in mammals yields a Ca isotopic shift value from diet to blood, denoted ∆_blood-diet and given by the difference between δ^44/42Ca_blood and δ^44/42Ca_diet, of −0.30 ± 0.13 ‰ (1 SD, Table S3). Second, increased bone turnover and possible transient bone loss in pregnant women (49, 50) could also involve a decrease in blood δ^44/42Ca values (46, 47). Third, the transfer of Ca from maternal to fetal blood involves an active transport of Ca through the placenta (51), hypothetically responsible for preferential transport of light Ca isotopes (32, 33). Fourth, metabolism of the fetus itself, notably involving mineralization, could explain a further decrease in δ^44/42Ca from source Ca to mineralized tissues (42).

Table S3.

Compilation of measured ∆^44/42Ca_blood-diet shifts (δ^44/42Ca_blood − δ^44/42Ca_diet) in mammals (32, 42, 44, 46, 47)

Mammal	∆44/42Cablood-diet, ‰	n	Source
Horse	−0.30	1	Ref. 42
Seal	−0.16	1	Ref. 42
Minipig	−0.20	2	Ref. 32
Pig (sow)	−0.52	1	Ref. 46
Sheep	−0.34	4	Ref. 44
Human*	−0.26	14	Ref. 47 (baselines only) and this study
Average	−0.30 ± 0.13 (SD)

^*Estimated in humans using −1.02‰ estimated diet and −1.28‰ blood average composition.

Despite these possible explanations for a ⁴⁴Ca-depleted isotope composition of fetus enamel, the calculated ∆_blood-diet value (ca. −0.3‰) perfectly matches the observed difference between the δ^44/42Ca values of wisdom tooth enamel (−1.58‰), representative of the adult diet, and the δ^44/42Ca values of prenatal enamel (−1.87‰), representative of the mother’s blood. This result supports the interpretation that the observed long-term δ^44/42Ca shift from in utero enamel to wisdom tooth enamel mainly reflects a dietary transition in Ca uptake from mother’s blood to adult diet.

Ca Isotope Composition of Postnatal Diet and Influence of Breastfeeding.

The drift in Ca isotope compositions is related to the duration of breast milk intake. Introduction of human milk at birth involves a source of Ca with a highly ⁴⁴Ca-depleted isotope composition (∼−1.6‰; see Table S1). The explanations for such low δ^44/42Ca values in breast milk are multiple. First, as discussed above, mother’s blood has a Ca isotope composition lower than that of the diet by the order of −0.30‰. Second, the transfer of Ca to milk involves active transportation through mammary epithelium (52), which is thought to account for a preferential secretion of light Ca isotopes (30, 32). Third, lactation is known to involve an increased mobilization of light skeletal Ca in the mother (49) that could induce a decrease in mother’s blood δ^44/42Ca (46, 47). The transition from prenatal diet (i.e., mother’s blood) to breastfeeding should thus not be accompanied by a significant isotopic shift toward ⁴⁴Ca-enriched compositions, the human δ^44/42Ca_blood value being quite ⁴⁴Ca-depleted, somewhere around −1.3‰ (Table S1). This assumption is matched in the three infants breastfed for a longer period. Individuals that were breastfed more than 12 mo (24, 24, and 36 mo for individuals D, E, and J, respectively; Figs. 1C and 2 and Fig. S3) do not display significant positive deviations in Ca isotope compositions, either at birth or in postnatal enamel (Welch’s ANOVA, P = 0.76; Kruskal–Wallis, P = 0.66). Postnatal enamel in these individuals is indistinguishable from prenatal enamel (Welch’s t test, P = 0.89, shift of + 0.00 ± 0.07‰, 95% confidence interval; Wilcoxon–Mann–Whitney, P = 0.98), whereas the long-term amplitude of isotopic deviation between postnatal enamel of these individuals and third molar enamel yields a value of + 0.33 ± 0.11‰ (Welch’s t test, P < 0.001, 95% confidence interval). The postnatal duration that was sampled in tooth enamel for each of these individuals is lower than 10 mo and thus was markedly shorter than their ages at cessation of suckling, known to be 24 and 36 mo. This confirms the hypothesis first formulated 10 y ago (30) that breast milk consumption is recorded within human deciduous tooth enamel.

With the exception of individual C, postnatal enamel of all other briefly breastfed or not breastfed individuals displays significantly higher δ^44/42Ca values than enamel contemporary to birth and prenatal enamel. For each of these individuals, the sampled postnatal estimated time period equals or exceeds the first 5 mo after birth. This is in agreement with a dietary change of Ca intake from placental nutrition to an infant breast milk-free diet within a timeframe of 0–4 mo (Table 1).

Residual Variability of Ca Isotope Compositions.

In briefly breastfed individuals (all except D, E, and J) no clear and precise relationship appears between the duration of the breastfeeding period (from 0 to 4 mo) and the value of Δ^44/42Ca in postnatal enamel. In other words, the value of the slope is not correlated with the duration of the breastfeeding period. Several explanations can be put forward.

First, all non-breastfed or briefly breastfed individuals were subsequently fed with various infant formulas, except for individual B. Substitute milk or infant formulas have high Ca levels resulting from various mixtures of animal milk and whey, inorganic Ca, and, possibly, vegetables. Their average isotope compositions are variable (15 different infant formulas analyzed for δ^44/42Ca range between −0.82 and −0.01‰) and ⁴⁴Ca-enriched compared with breast milk considering the origin of Ca in these ingredients (average δ^44/42Ca of −0.49 ± 0.51‰, 2 SD, Table S1). The same holds true for transition alimentation that is composed of various dairy and vegetable components with average ⁴⁴Ca-enriched compositions. The amplitude of the shift between the prenatal period and the period of milk-free food intake is thus likely variable depending on transition food types and on formula compositions. The study of more simple dietary histories, such as in captive macaques (1), would facilitate discerning patterns with finer time and amplitude resolutions.

Second, the spatial resolution that the sampling method allows is about 400 µm, which corresponds to 2.4- to 4.4-mo time envelopes, depending on enamel secretion rates. This temporal resolution likely induces a dampening of steep variations, such as experienced by individual B.

Third, the maturation of newly formed enamel (53, 54) possibly generates elemental and isotopic mixing between initially secreted enamel and secondary matured enamel, which could participate in a dampening and a phase shifting of the recorded signal (55). The improvement of the estimation of the timing of dietary transition would thus benefit from a comprehensive investigation of the transduction of Ca isotope signal from dietary intake to enamel such as in hypsodont herbivores.

Implications for Trophic-Level Reconstruction Using δ^44/42Ca of Mammalian Tooth Enamel.

Trophic-level studies in modern and past environments using Ca isotopes are promising but are confronted with poorly understood residual variability both in terrestrial and marine environments (37, 56). These studies are based on observed isotopic offsets in δ^44/42Ca varying between −0.14‰ and −0.65‰ from one trophic level to another (37, 42, 57). This is of a magnitude comparable to the shift observed here from prenatal or exclusive breast milk to a breast milk-free diet (∼ +0.30‰). The consumption of breast milk is likely to induce a difference in the Ca isotope composition of enamel that could be confused with a signature of a superior trophic level.

Perspectives.

The present approach allows distinguishing weaning practices in modern humans as recorded in deciduous tooth enamel. In this case, it allows distinguishing a cessation of suckling occurring within the first year, representative of weaning practices in Europe (6), from a behavior resembling early-weaning of nonindustrialized modern human, occurring between 2 and 3 y old (5, 9, 16).

We emphasize that potential applications to past human populations and extinct hominins could help in studying their weaning practices. Provided a good knowledge of enamel crown development in studied individuals, Ca isotope compositions could help determine age at significant reduction of suckling with a temporal resolution of the order of 6–12 mo. Present-day humans wean their infants earlier (2–3 y) than do great apes (3–7 y) (5, 9, 16). Whether the common ancestor of hominins was characterized by an older age at cessation of suckling than modern humans remains a matter of debate (5, 16, 19, 20) based on rather indirect inferences (9). Ca isotope studies offer an independent approach to test such hypotheses.

Materials

A total of 51 deciduous teeth from 12 healthy European children born between 1971 and 2011 were used in this study (Table 1). For each individual, three to five deciduous teeth were selected depending on available teeth, to cover the widest time span of enamel crown formation. This period corresponds to the timing of tooth crown formation in human deciduous teeth, which initiates on average at about 5 mo before birth (48) and concludes at around 1.5 y of postnatal age (58). All teeth were naturally shed or extracted for surgical purposes in accordance with the World Medical Association’s Declaration of Helsinki. In each case, the informed consent of the patients or their parents was collected. Information concerning early diet was provided retrospectively by the parents when possible. We also used nine permanent third molars, initially described elsewhere (34, 59), for which crown enamel forms between 7 and 15 y of age (58), to assess the long-term trend of Ca isotope composition evolution of enamel. All teeth were collected without identifying data. Details about all individuals’ early-life diet and sampled teeth are given in Table 1.

Methods

Sampling.

Each permanent and deciduous tooth was halved longitudinally along the buccolingual plane using a low-speed rotating diamond saw. One half of each tooth was then embedded in araldite resin and the cut surface was polished using sandpaper with decreasing grain sizes. The sampling was performed using a precise position drilling MicroMill device allowing sampling of 60–80 µg hydroxylapatite and drilling holes of 350–400 µm in diameter and 200–300 µm in depth, as described in Tacail et al. (34) and in Supporting Information.

The sampling strategy consisted of drilling a series of spots with the widest possible time span available on enamel surface of the buccal side in general. Sampling was thus performed in deciduous teeth at regular intervals along the crown height (i.e., from enamel cusp to cervix). Teeth displaying substantial enamel thickness such as deciduous first and second molars allowed in some cases sampling of more than one sample from the enamel–dentine junction to the outer surface. A single enamel sample per wisdom tooth was obtained likewise (Supporting Information).

Location of Spots and Estimation of Mean Formation Ages of Sampled Enamel.

The neonatal line is used to distinguish between enamel formed prenatally from enamel formed postnatally (60–63) (see drawings on pictures using Adobe Photoshop software, Fig. S4 and Supporting Information). On this basis, we split samples into three categories according to their position relative to the neonatal line: (i) more than 60% of the sampling spot surface lies in prenatal enamel, (ii) more than 60% of the sampling spot surface lies in postnatal enamel, and (iii) less than 60% of the sampling spot surface lies in either of the pre- or postnatal enamel, referred to here as birth category. We also measured the distance of each sampling spot to neonatal line along the main prism orientation and thus propose a first-order age model for sampled enamel assuming an average enamel secretion rate of 4 µm⋅d⁻¹ for all teeth together (48, 64, 65) (Supporting Information).

Sample Preparation and δ^44/42Ca Measurement.

After collection, each powder sample was chemically purified following method described elsewhere (44) and in Supporting Information. The measurement of Ca isotope compositions was performed at the Laboratoire de Géologie de Lyon, France, on a Neptune Plus multicollector induced coupled plasma mass spectrometer (MC-ICP-MS) from Thermo Scientific using a previously described protocol (34, 44) (Fig. S1, Table S4, and Supporting Information). All Ca isotope compositions are expressed in per mil units, using the “delta” notation for the ⁴⁴Ca/⁴²Ca isotope ratios defined as follows:

$${\delta }^{44/42}\text{Ca}=\left(\frac{{\left({}^{44}\mathrm{Ca}/{}^{42}\mathrm{Ca}\right)}_{\mathrm{sample}}}{{\left({}^{44}\mathrm{Ca}/{}^{42}\mathrm{Ca}\right)}_{\text{ICP}\hspace{0.17em}\text{Ca}\hspace{0.17em}\text{Lyon}}}-1\right)\times \mathrm{1,000},$$

Table S4.

δ^44/42Ca values of standards as measured in this study (in per mil relative to ICP Ca Lyon standard) compared with previously published values (34, 37, 44, 46, 47)

Standard	Source	n	δ44/42Ca, ‰, ±2 SD relative to ICP Ca Lyon
SRM1486 (cow bone meal)	This study, July 2015	17	−1.04	0.13
	This study, November 2015	27	−1.03	0.12
	This study, December 2015	22	−1.03	0.10
	This study, August 2016	29	−1.05	0.12
	This study, February 2017	31	−1.00	0.09
	This study, March 2017	21	−1.02	0.11
	This study, average of all sessions	147	−1.03	0.12
	Ref. 44	17	−0.96	0.14
	Ref. 34	120	−1.03	0.13
	Ref. 37	25	−1.04	0.11
SRM915b ( Ca carbonate)	This study	4	−0.16	0.04
	Ref. 44	11	−0.12	0.07
	Ref. 34	4	−0.14	0.06
	Ref. 37	13	−0.15	0.11
SRM915a ( Ca carbonate)	This study	5	−0.52	0.10
	Estimation from literature (37)	—	−0.52	0.08
SRM1400 (cow bone ash)
Powder	Ref. 34, SRM1400	26	−1.24	0.13
Microsampled on sinter	This study, SRM1400 SPS 1	3	−1.23	0.08
	This study, SRM1400 SPS 2	5	−1.27	0.06
	Ref. 34, SRM1400 SPS	11	−1.18	0.16
BCR-380R (cow whole milk powder)	This study	16	−1.10	0.10
ICP1 Ca solution (46, 47)	This study	10	−0.20	0.10

where (⁴⁴Ca/⁴²Ca)_sample and (⁴⁴Ca/⁴²Ca)_{ICP Ca Lyon} are Ca isotope abundance ratios measured in sample and in ICP Ca Lyon bracketing standard, respectively.

Sampling

The sampling of enamel was performed using the method described in Tacail et al. (34). Briefly, it consisted of drilling the enamel surface using a tungsten carbide drill mounted on a precise position drilling MicroMill device. Drill holes were typically 350–400 µm in diameter and 200–300 µm in depth. Small powder heaps accumulated on the rims of the holes were collected using razorblades and transferred to trace-level clean Savillex vials. Before each sampling, enamel surface, drill bits and razorblades were washed and wiped using 99% pure ethanol and blown off using a compressed air duster. Each typical drill spot allowed recovery of ∼60–80 µg hydroxylapatite corresponding to about 22–30 µg of Ca. Depending on available enamel thickness relative to drill size, we performed sampling by drilling one spot or two smaller and shallower contiguous spots.

We checked for the absence of any sampling method bias on Ca isotope composition measurement. We performed microsampling under the same conditions as for teeth using SRM1400 SPS, a sintered powder bone standard (34). The two recovered powder samples were chemically purified and analyzed as unknown samples during Ca isotope composition measurement sessions.

With the exception of individuals R and S, all wisdom teeth were sampled using the same drilling method as for deciduous teeth. We performed two drilling holes in thickness of enamel cusp (near the dentin–enamel junction and close to the outer surface) and pooled the two sampled powder heaps for each individual to assess an average composition of tooth enamel. Individuals R and S (previously described in ref. 34 as BMM3 and HPME) were sampled by recovering chips of broken enamel crown.

Location of Spots and Measurement of Distance to Neonatal Line

After sampling, teeth were further smoothly polished using alumina suspensions of decreasing particle sizes, up to 0.1 µm. Direct observations were performed together with pictures taken with the help of a computer-assisted binocular microscope. The neonatal line, which is a marked stria that forms at birth in all deciduous teeth and can occur in some permanent first molars, separates pre- from postnatal enamel (60–63). The neonatal line is used to distinguish enamel formed prenatally from enamel formed postnatally (see drawings on pictures using Adobe Photoshop software, Fig. S4). On this basis, we split samples into three categories according to their position relative to the neonatal line: (i) more than 60% of the sampling spot surface lies in prenatal enamel, (ii) more than 60% of the sampling spot surface lies in postnatal enamel, and (iii) less than 60% of the sampling spot surface lies in either of the pre- or postnatal enamel, referred here to as birth category.

We also identified and mapped the main direction of enamel prisms and located observable accentuated incremental markings, often corresponding to postnatal stress events (61, 63, 66). We measured the distance along the main prism orientation from neonatal line to sampling spot center. Distances of sampling spot centers to neonatal lines were either measured directly or with the help of accentuated markings to estimate cumulated distances to neonatal line for sampling spots with remote locations. Accentuated growth lines were not always observed, which necessitated inferring isochronous lines according to the general geometry of the sectioned tooth. In the cases of six strictly postnatal sampling spots, distinguished in Dataset S1, their location was too imprecise or remote from the neonatal line to allow a confident estimation of distance. These samples were rejected for temporal evolution of δ^44/42Ca presented in Fig. 2 but kept in the postnatal enamel category. Annotated pictures of all sampled teeth are presented in Fig. S4.

We then calculated a first-order estimate of chronological age assuming an average enamel secretion rate of 4 µm⋅d⁻¹ for all teeth together (48, 64, 65). Each spot was thus associated with an estimate of the age at enamel formation with respect to birth. Also, we measured maximum sampling spot width along the prism direction and calculated for each spot the time envelope that each sampling spot encompasses.

We thus propose a first-order age model for each individual, sufficient to discuss main features of Ca isotope variation provided the sampling spatial resolution is of 400 µm on average, reflecting 2.4- to 4.4-mo periods depending on enamel secretion rates of a given tooth or crown sector (48, 64, 65).

Chemical Processing of Samples

Briefly, every sample was dissolved in subboiled distilled 1 N HCl acid and processed through AG50X-W12 cation exchange resin in 1 N HCl medium to dispose of sample matrix (i.e., phosphates, sulfates, alkali elements, and Mg). Ca and Sr fractions were collected in 6 N HCl medium. Ca fractions were then separated from Sr by loading samples onto columns filled with Sr-specific resin (Eichrom Sr-Spec) in subboiled distilled 2 N HNO₃ medium (34, 44). Blanks for the whole procedure did not exceed 100 ng Ca (44). This is 200 times smaller than smallest processed Ca samples (about 20 µg) and could not affect the measured isotopic compositions beyond the measurement precision.

A series of 15 different infant formulas from various trademarks (Table S1) were also sampled. One breast milk sample was obtained from a 36-y-old lactating French woman. Whole milk powder standard is BCR 380-R provided by Institute for Reference Materials and Measurements. The whole blood samples were collected from two French individuals residing in Lyon, France (described in ref. 67). The chemical preparation of milk and whole blood samples was performed as described elsewhere (44). Briefly, after mineralization using concentrated subboiled distilled HNO₃ acid and H₂O₂ 30% Suprapur, samples were processed through AG50W-X12 cationic resin for recovery of Ca, Sr, and Fe. Ca was then separated from Fe by processing samples through AG1-X8 anionic resin before removing Sr using Sr-specific resin.

MC-ICP-MS Analysis

A standard-sample bracketing measurement method was used with the ICP Ca Lyon standard (44) as bracketing standard. Measurements of all samples and standards were performed during six sessions, between 2015 and 2017. When δ^43/42Ca values of all measured materials are plotted as a function of their δ^44/42Ca values, compositions fall on a line in good agreement with the 0.507 slope predicted by the linear approximation of exponential mass-dependent fractionation (Fig. S1).

The measurements were systematically checked for long-term precision and accuracy using SRM1486 bone meal NIST secondary standard previously described and analyzed for Ca isotope compositions (34, 36, 37, 44, 68). SRM1486 yielded constant values across the six different analysis sessions, with an average δ^44/42Ca value of −1.03 ± 0.01 ‰ (2 SE, n = 147), in very good agreement with previously published values (as listed in Table S4), notably −1.03 ± 0.01 ‰ (2 SE, n = 120) (34). We also analyzed the commonly used SRM915a and SRM915b clinical-grade carbonate standards, as well as BCR-380R cow whole milk powder standard and ICP1 Ca solution used as standard in former studies (46, 47, 69). All measured δ^44/42Ca values of standards and previously published compositions are given for comparison in Table S4. Long-term external precision was estimated using the SRM1486 standard and yields a 2 SD value of 0.12‰ for δ^44/42Ca for 147 analyses, over the six different sessions.

As was previously shown (34), the microsampling method did not affect measured isotope compositions of Ca. The two SRM1400 SPS microdrilled samples did not show significant differences compared with the previously published δ^44/42Ca value of −1.24 ± 0.13 ‰ (2 SD, n = 26) (34): SRM1400 SPS 1 and 2 yielded indistinguishable values of −1.23 ± 0.08 ‰ (2SD, n = 3) and −1.27 ± 0.06 ‰ (2 SD, n = 5), respectively (Table S4).

Conversions of literature values expressed relative to SRM915a reference standard were performed as follows. All δ^44/40Ca data were converted to δ^44/42Ca values by dividing by a 1.9996 factor, calculated using the power fractionation law. The δ^44/42Ca values were then converted to δ^44/42Ca relative to ICP Ca Lyon using a value of SRM915a relative to ICP Ca Lyon of −0.52‰ as measured in the present study and estimated in Martin et al. (37) (Table S4). Data stemming from Channon et al. (47) were converted to ICP Ca Lyon using the measured −0.20‰ value of ICP1 Ca solution used as reference material by these authors (Tables S1 and S4). This −0.20‰ value of ICP1 Ca solution is in good agreement with the +0.05‰ published SRM915b value against ICP1, which corresponds, once converted to ICP Ca Lyon, to −0.15‰, indistinguishable from our measured composition of SRM915b.