Early evidence of sheep lambing de-seasoning in the Western Mediterranean in the sixth millennium BCE

Today, sheep farmers in the Western Mediterranean de-season their ewes to achieve autumnal births. This strategy contrasts sharply with spring lambing further north, and provides benefits in terms of out-of-season availability of animal products. These competences are closely linked to specific sheep physiology and favorable Western Mediterranean climatic conditions. It is not known exactly how far back in the past the ability to de-season Mediterranean sheep breeds extends. This study shows that this practice existed seven millennia ago in Southern France, at an early stage of the major agricultural expansion of the Neolithic into the Western Mediterranean. To achieve this reproductive management regime, three prerequisites were required: (i) the ability of sheep to give birth in autumn, constituting early evidence for the genetic selection of specimens with prolonged reproductive fertility; (ii) intentional management of female and male interactions within the herd, which would have required good knowledge of the timing of the fertility cycle in ewes, and; (iii) adequate pasture resources to support lactation in the autumn, possibly favored by autumnal rains, substantiating previous paleoclimatological data for the existence of a Mediterranean-type precipitation regime at that time. Moreover, we also show that winter foddering of sheep occurred, using forest resources, and that this maintained good body weights for spring mating. These findings add pivotal information about shepherding practices and the socio-economic abilities of Early Neolithic communities, which allowed for the extension of their areas of influence from the Eastern Mediterranean area to the West during the Early Neolithic agricultural expansion in Europe.

Scientific RepoRtS | (2020) 10:12798 | https://doi.org/10.1038/s41598-020-69576-w www.nature.com/scientificreports/ research projects for the last 40 years 33, and have provided key information for understanding the early Neolithization process in the Western Mediterranean region. They supplied important data on Cardial and Epicardial occupations, a well-attested period in other contemporaneous sites, such as Montclus (Montclus, Gard), Camprafaud (Ferrières-Poussarou), Oullins (Le Garn, Gard), Corrège (Leucate, Aude) and Mas de Vignoles VI-X (Remoulins, Gard) 34, 4 . More detailed information about the Taï and Gazel sites is presented in SI Materials and Methods. The analysis focused on the Epicardial archaeological levels at both sites. At the Taï site, the Epicardial phase is dated to 5,270-4,990 cal BCE in levels C1 and C2 5, while at the Gazel site, it is found in Phases II and III, dated to 5,350 to 5,200 cal BCE and 5,200 to 5,050 cal BCE (Guilaine and Manen, in progress), respectively. Both sites were dwelling sites, and were occupied throughout the year. At Taï, the presence of storage pits, evidence for agriculture close to the site, and the nature of the archaeobotanical assemblage suggest a permanent occupation and the presence of domestic animals at the site at least most of the year 35,36 . Gazel seems to have been frequented repeatedly throughout the Early Neolithic (phases I to IV, the latter dated around 4,700 BCE), with phases of permanence more pronounced than others. The functional status of the cave is uncertain. Although its occupations might have been related to pastoralism, it may not be defined as a sheepfold-cave. At Taï, the C1 and C2 levels delivered a faunal spectrum predominated by domestic sheep, goats and cattle (91% of the NISP excluding lagomorph remains), where sheep are clearly the main domestic species (71% of the NISP) (Bréhard et al., in progress and SI). At Gazel, the faunal spectra in Phases II and III were largely dominated by domestic caprines (circa 60% of the NISP) 37,38 . At both sites, the mortality profile for domestic caprines is indicative of non-specialized management strategies, suggesting that both meat and milk could have been exploited (Bréhard et al., in press).
In total, 28 teeth belonging to nine sheep from the Taï site and 12 sheep from the Gazel site were selected for stable isotope analysis (Material and methods and SI). The sequential isotopic analysis of sheep tooth enamel allows for a reconstruction of individual isotopic histories over the period of tooth formation 39 . Oxygen and carbon derived from ingested plants and water are incorporated in the enamel mineral fraction (bioapatite) during tooth enamel formation 40,41 . Their stable isotopic compositions (δ 18 O and δ 13 C) are measured conjointly in bioapatite carbonate.
The δ 18 O values in tooth enamel are mainly derived from ingested water. In temperate latitudes, δ 18 O values in the skeletons of large herbivores are strongly correlated to the values of local meteoric water 40,[42][43][44] available to herbivores from surface water and plant water. At high and middle latitudes, the δ 18 O composition of precipitation varies with ambient temperature, with the highest values occurring during the warmest months and the lowest during the coldest months [45][46][47] . Once retrieved from the dental enamel, these seasonal trends display cyclical variation, which can be used to reconstruct seasonal birth patterns. As the timing of tooth eruption and growth is fixed within a species, inter-individual variability in the δ 18 O sequences retrieved from the enamel is related to variability in the birth season [48][49][50] . In order to reconstruct birth seasonality, inter-individual variability is described through the observation of the position of the highest δ 18 O value in the tooth crown. A quantitative estimation of variability involves a normalization procedure through modeling 51 (see Material and methods). The lambing season of the archaeological specimens can then be determined by comparison with modern reference sets with known dates of birth 12,52,53 . This approach has been applied successfully to sheep second (M2) and third molars (M3) in previous studies 12,53 .
The δ 13 C values in tooth enamel bioapatite are derived from ingested plants 41,54 . Carbon stable isotope ratios in plants vary according to the photosynthetic pathway and prevailing growing conditions 55,56 . In the Early Neolithic, C 3 plants dominated vegetal landscape composition in Southern France and no cultivated C 4 plants may be expected at that time 36,57 . A diet of C 3 plants from open areas should produce δ 13 C values around − 11.8‰ in archaeological sheep enamel (Material and methods), ± 1‰ to account for seasonal variations 58 . If sheep were foddered, which implies the use of forest-tree leaves, this would lead to lower δ 13 C values due to the canopy effect, where the reduction of light intensity causes a depletion of 13 C in plants 59,60 . In Southern France, the Late Mesolithic mature oak-dominated forests persisted during the Early Neolithic with only very moderate clearances 29 . Still, the magnitude of the canopy effect on plant δ 13 C values differs according to forest type and is therefore difficult to predict. 13 C-depletions of 2-5‰ in comparison with plants in neighboring open settings have been reported 58,60 . In archaeological sheep enamel, we estimate that winter δ 13 C values significantly different to − 12.8‰ and closer to − 14.5‰ could reflect a contribution of plant resources gathered in the forest (Material and Methods).

Results
Intra-tooth sequences of δ 18 O and δ 13 C values are shown in Figures S1 to S3. The results are summarized in Table S2.
Lambing season. In all the sampled teeth, a pattern of sinusoidal variation in δ 18 O values reflects the seasonal cycle. Birth distribution and lambing seasons were evaluated from the δ 18 O sequences observed in the M2 molar, for which modern reference sets for lambing seasons are available. The modeling procedure was applied to all M2s where a minimum and maximum δ 18 O value could be identified (this excludes specimens Ovis 11 and Ovis 18 at Taï, and specimens Ovis 10 and Ovis 11 at Gazel). The modeling results are shown in Table S3. Inter-individual variability in the position of the maximum δ 18 O value in the tooth crown (x 0 /X), reflecting variability in the birth date, is presented in Fig. 2. These results were then compared to modern reference data sets from sheep specimens with a known birth season (see SI Materials and Method). At both sites, the results suggest seasonal lambing, occurring partly in the autumn (Ovis 21 and 30 from Taï and Ovis 01, 02, 03, 06 and 07 from Gazel), and partly in late winter until spring (Ovis 09, 19 and 20 from Taï and Ovis 04, 05, 08, 09 and 12 from Gazel).   www.nature.com/scientificreports/ two cases, winter δ 13 C values reach − 14‰ (TAÏ Ovis 10 and 40). At Gazel, 11 out of the 14 analyzed teeth also show winter δ 13 C values below − 12.8‰, with the lowest reaching − 13.6‰ (Table S2 and Figure S1-S3). These δ 13 C values fall within the threshold defined as potentially indicating a significant dietary contribution from forest resources.

Discussion
Autumnal lambing: sheep physiology and human agency. The occurrence of autumnal lambing in Early Neolithic contexts at Taï and Gazel constitutes the earliest evidence of this reproductive behavior in sheep in the Western Mediterranean region, where extended lambing periods and de-seasoning without artificial treatments are current practices today 8,11,17,18 . Autumnal births have been reported during the Chalcolithic period elsewhere in Europe but only as isolated cases: these are exceptions in assemblages where spring lambing appears to have been the general rule 12,53 . In the present investigation, where the limited sample size is nevertheless similar to previous studies, the relative proportion of autumnal lambing is high, suggesting a well-established practice.
Autumnal lambing demonstrates an extended breeding season for Early Neolithic sheep in Southern France, in comparison with present-day and Neolithic European sheep living in higher latitudes 12, but similar to presentday western Mediterranean breeds. Exactly how far back this physiological capability goes is a central question for understanding the reasons for its emergence, including heritage from the wild ancestor (Ovis orientalis), modifications driven by domestication due to selection and animal diet management, as well as the climatic framework. Consolidated data on birth seasonality in the early Holocene mouflon-the ancestor of domestic sheep -is currently lacking. Following the same analytical procedure as the one applied in the present study, stable isotope analyses of Ovis orientalis teeth at the Epigravettian hunting campsite of Kalavan (north-eastern Armenia) demonstrated a reduced birth period for mouflons around 12,000 BCE 61 . However, both the environmental framework (Late Glacial) and wild sheep ethology (vertical mobility) may have engendered specific pressure on their reproductive behavior. The lambing season of the present-day mouflon in the Near East (Ovis orientalis gmelini) is very short (1-2 months), and occurs in the spring 62 . Early domestic sheep at Tell Halula in the Middle Euphrates Valley, dated to ca 7,500 cal BCE (Late Pre-Pottery Neolithic B) were shown to have a similar short period of birth (2.5 months) 63 . By contrast, an extended lambing period was confirmed in the Pre-Pottery Neolithic of Cyprus (7,600-6,900 cal BCE), which included autumn/winter births 64 . Both contexts are located at a latitude below 35°N, where present-day sheep tend to have a short period of sexual rest 65 . It is interesting to note that in feral domestic sheep (i.e., the European mouflon) currently found in Cyprus, Corsica and Sardinia, births occur over a short time period in spring [66][67][68] . Experiments on European mouflon kept under natural photoperiod conditions in Madrid, Spain (40°N), involving the monitoring of hormonal profiles in females, demonstrated extended ovulatory activity, starting in October and terminating in the spring 69 . The apparent discrepancy between a prolonged fertility season and a short lambing period may be explained by the role of social interaction in regulating seasonal reproduction in the mouflon. In the wild, males and females are segregated when females are not sexually active, and female and male herds are reunited at the start of the mating season 70 . If mating occurs within a short time after reuniting, this would explain a restricted lambing season. This implies that autumnal lambing in domestic herds not only involves the physiological capacity for a prolonged fertility period, but also adequate management of interactions between females and males, most likely in the form of the separation of females and males, and reuniting them towards the end of the fertility period.
Assuming that the observed extended fertility season in Early Neolithic Western Mediterranean sheep differed from that of their wild ancestors, this physiological character would result from domestication, but the driving mechanisms-i.e., the role of environmental factors and human selection-remain difficult to determine. Although the photoperiod is generally accepted as the primary environmental cue influencing seasonal reproductive patterns in sheep, a genetic basis for photoresponsiveness has been argued for and demonstrated in both sheep and European mouflon 10,71-73 . Indeed, sheep breeds in the Mediterranean area show different responses to photoperiod cues 74 . For example, Manchega and Lacaune ewes appear to be much more sensitive to photoperiod cues than Ripollessa ewes or present-day European mouflon ewes raised in the same locations in Spain 18 . This induces a longer duration of seasonal ovulatory activity in breeds with higher sensitivity to changes in the photoperiod. Human-controlled genetic selection is assumed to contribute to the longer mating season and greater prolificacy in Mediterranean sheep, as well as other aspects of the breeds such as, docility, color, productive rates and milk 18,75,76 . On the other hand, genetic origins confer phenotypical reproductive capacity, while environmental conditions influence their expression 77 . Although the heritability of fertility traits is low in sheep 78,79 , the onset, offset and duration of the breeding season can still theoretically be selected 80 . The simplest way for herders to operate selection for autumnal lambing would have been to keep the autumn offspring. Although both young females and males could have been selected for this purpose, operating selection on the females would probably have been more intuitive to Neolithic herders to produce autumn lambing females. Even under these circumstances, the process would thus require many generations, but could still be envisaged on a long-term scale.
The economic outcomes of autumnal lambing. A prolonged lambing period is preferred by modern sheep farmers in order to extend the availability of sheep products throughout the year 10,13 . Early Neolithic farmers who settled in Southern France could also have pursued these economically attractive practices. These seasonal bi-modulated reproductive patterns might have brought important advantages, enabling the group to balance the timing of production and consumption throughout the year, and may also have reduced risks for the survival of the herd by scheduling the harvesting of animal resources throughout a longer period in the year. Newborn lambs and dairy products with high fat and animal protein content could be planned on a longer Scientific RepoRtS | (2020) 10:12798 | https://doi.org/10.1038/s41598-020-69576-w www.nature.com/scientificreports/ time scale, even though this probably resulted in a lower production rate for lactation during late autumn and winter 80 . Nevertheless, at the Taï and Gazel sites, the caprines mortality profiles suggest that both milk and meat could have been exploited by Early Neolithic groups (Bréhard et al. in press). Regarding herd demographic dynamics, modern studies have demonstrated that autumn lambs reach puberty later than lambs born in the spring, which is the period during which plant resources are optimal 69,81-83 . In consequence, autumnal lambing delays the incorporation of new sexually mature individuals into the herd. Today, the spring lamb reaches puberty during its first breeding period, whereas in autumn lambs, the length of the prepubertal period is longer and they may not reach puberty before the second breeding season. However, in traditional husbandries, sheep are not mated until their second year, when they are 18 to 24 months old 84-86 , whereas they are mated during their first year in present-day industrial husbandry. Consequently, it is unlikely that this delay curbed herd growth rate in Neolithic husbandries.
Winter foddering. The data presented here provide early evidence of sheep winter foddering with a significant contribution of resources from forested areas. At the time of the Early Neolithic occupation at Gazel, mixed deciduous oak and maple tree woods dominated the surrounding landscape 87,88 . Carpological remains include hazelnuts, acorns and grapes (L. Bouby, pers. comm.). The archaeobotanical record at Taï also attests to a nonhomogeneous forest component, including open Mediterranean evergreen oak woods, but also a potentially denser alluvial forest near the Gardon River 36 . The archaeobotanical assemblage at the Taï site delivered a relatively high number of wild plant species showing great diversity 36 . The generally better state of preservation of these remains compared to those of cereals and pulses suggests that the wild plant assemblage did not follow the same formation process. Given that this assemblage also includes a fair number of species not considered to be edible for humans, the presence of wild plants at the site could be linked to animal husbandry. In particular, the seeds and whole fruits of the hazel tree, elder, juniper, pistachio tree and evergreen oak could have been brought to the site with leafy branches and given to animals as fodder or litter 36 .
In this context, it is appropriate to combine the isotopic ratios measured in the Taï sheep tooth enamel with the archaeobotanical evidence. First of all, the low δ 13 C values attest to the presence of a woodland component in the landscape, dense enough to create a "canopy effect" 60 . Secondly, the δ 13 C values suggest that sheep fed on resources from the forest, but cannot distinguish between herding under forest canopy and provisioning with culled fodder. Archaeobotany demonstrates that leafy branches were brought to the site. Therefore, these two indications suggest that these branches were actually used as fodder (but could still also have been used as litter). Lastly, the presence of fruit suggests that leafy branches were collected in late summer and possibly until early winter 36, while the stable isotope evidence demonstrates that fodder -or at least the forest component, was mostly fed to sheep during winter. This arboreal fodder would have necessarily been stored between collecting and provisioning to sheep.
The cutting, transport and storage of leafy fodder would have required labor and specific organization, which may potentially have overlapped with other tasks, including the summer harvest or the autumn sowing of winter cereals (as shown at the Taï site 36 . It is difficult to evaluate just how much labor was invested as this would primarily have been a matter of scale. The number of sheep, the length of the foddering period, the proportion of leafy fodder in the sheep diet and the distance at which fodder was harvested, are key parameters 25, none of which can be accurately determined. Forest resources were fed to sheep in significant quantities and over a sufficiently long period to affect the stable isotope record. Additionally, at both sites, winter foddering appears to be a recurrent seasonal practice. On an individual scale, it is mirrored in both sampled teeth (M2 + M3), providing records for two consecutive winters (Fig. 3). On the scale of the sampled population, the practice is reflected in several, not necessarily contemporaneous specimens over the time span when the sites were settled. Overall, this would suggest a more systematic and intensive practice rather than opportunistic activity during brief periods of severe winters. These conclusions echo previous findings in other Neolithic contexts in Southern France where the archaeobotanical evidence attests to tree leaf supplies to stalled domestic herds [89][90][91] . In terms of the lambing pattern at Taï and Gazel, providing additional fodder to sheep over winter would have helped maintain good body weight throughout the winter, promoting good conditions for spring mating leading to autumnal lambing. At the same time, it would have been beneficial to ewes in the last period of gestation for late winter/ early spring lambing.

conclusion
The results of our study are threefold. Firstly (i), they relate to animal physiology, as they demonstrate the capacity of Early Neolithic sheep from Southern France to give birth in autumn, at a time of year currently considered to be "off-season" i.e., contrasting sharply with the general pattern of spring lambing at higher latitudes in Europe. This would constitute the earliest evidence for the genetic selection of specimens with prolonged reproductive fertility 18, which may therefore have been practiced since the earliest presence of sheep in the Western Mediterranean, at the least. These ancient systems may or may not eventually have led to the origins of present-day Mediterranean breeds with specific photoresponsive aptitudes. Although regional particularisms in the Mediterranean area, due to geographic compartmentalization, and the history of human migrations and trade, have favored the consistency of genetic specificity in local animal breeds 7, it is nonetheless beyond the scope of this study to establish a direct link between present-day and Early Neolithic sheep in Southern France. Secondly (ii), these results relate to zootechnical knowledge: autumnal lambing is the outcome of the intentional management of female and male interactions within the herd, probably by means of separating both sexes and reuniting them towards the end of the fertility period. Managing sheep socio-sexual cues in this manner would have required good knowledge of the timing of the fertility cycle in ewes. Autumnal lambing may have been supported by the winter foddering of sheep, in order to maintain a good body weight for spring mating. Forest resources were used Scientific RepoRtS | (2020) 10:12798 | https://doi.org/10.1038/s41598-020-69576-w www.nature.com/scientificreports/ for this purpose, involving potentially important-although difficult to evaluate, task investment. Thirdly (iii), these results relate to paleoclimatology, as autumnal lambing would support the assumption that a Mediterranean type precipitation regime existed at this time 19 . Indeed, Early Neolithic ewes could not have given birth in the autumn without a secure availability of pasture resources at that time of the year in order to support lactation. The autumn lambing observed at Taï and Gazel occurs in both cases alongside more classical lambing at the end of winter to early spring. The subsequent expansion of the sheep birthing period should be taken into account when establishing the farming calendar, as the period of births and initiation of lactation necessitate increased monitoring by herders. It is very unlikely that females would have lambed twice a year. Rather, a system involving two groups of females giving birth at different times of the year should be envisaged, with sufficiently long periods of post-partum sexual rest to allow females to move from one group to the other, in order to enhance flexibility. Such a system would also enable empty ewes to lamb. In return for this investment, herders benefited from the distribution of seasonal animal resources over a longer period in the year, even though milk exploitation does not seem to have been substantial enough to induce specific demographic management. Caprine mortality profiles at the Taï and Gazel settlements suggest mixed exploitation for both meat and milk. In conclusion, the establishment of sheep husbandry systems involving autumnal lambing in the Early Neolithic of Southern France attests to a strong capacity to adapt to environmental contexts, which may have been a key to the successful expansion of agropastoral socio-economies across the wide range of climatic and ecological zones 92 of the vast European continent. The specificities described in this study in terms of demographic and food management may also explain the multi-directional 5 and arrhythmic 6,93 character of the expansion of husbandry systems in the Western Mediterranean region. This scenario fits well with those proposed by Guilaine 94, who considered that these Early Neolithic communities successfully adapted to the advantages of new ecological niches in the Western Mediterranean.

Materials and method
Sampling, pre-treatment and IRMS analysis. The studied sample consists of nine sheep from the Taï site and 12 sheep from the Gazel site. One second (M2) or one third molar (M3), or both molars, were taken from each sheep, totalizing 28 teeth (Table S1). Sequential sampling of enamel along the tooth growth axis was conducted by drilling from the tooth surface 39 . As the timing of the formation of second and third molars is different, the signals in both teeth are complementary. Sampling combining the slightly worn M2 with the unworn but fully mineralized M3 molars from a given specimen provides a record of the first two years of the life of the specimen 53,95,96 .
Enamel powders were treated with 0.1 M acetic acid (0.1 ml solution/mg enamel) for 4 h at room temperature to eliminate diagenetic carbonates. Pretreated powders were analyzed individually on a Thermo Kiel IV device interfaced with a Thermo Delta V Advantage isotope ratio mass spectrometer. The δ 13 C and δ 18    x is the distance from the enamel-root junction; X is the period (in mm), or the length of tooth crown potentially formed over a whole annual cycle; A is the amplitude (= max-min/2) (in ‰); x 0 is the delay (mm) ; δ 18 O attains maximum value when x = x 0 ; M is the mean (= (max + min)/2) expressed in ‰. The best match of the measured data with the parameters of the model was determined using an iterative method, and a minimization of the sum of the square of the difference between the model and the measurements (the method of least squares). Calculations were carried out using Microsoft Excel software.
The tooth size factor is eliminated through normalization of distances using the period X of the δ 18 O cycle. The position of the maximum δ 18 O values in the tooth crown is therefore expressed as x 0 /X. Inter-individual variability in (x 0 /X) reflects variability in the birth season 51 . Following Balasse et al. (in press), we use a circular representation of the data to reflect the cyclical nature of seasonality: January follows December, i.e. when (x 0 /X) reaches 1, it also reaches 0.
Threshold δ 13 C values in sheep tooth enamel. The typical enamel δ 13 C value reflecting feeding on plants from open settings was determined using a mean δ 13 C value of − 27‰ for modern C 3 plants 97,98 , corrected by + 1.5 ‰ to account for the fossil fuel effect 99,100, and by applying an enamel-diet 13 C-enrichment factor (e*) of + 14.1‰ 101 . The effect of the consumption of forest tree leaves as winter fodder on enamel δ 13 C values was estimated as follows: in the modern dense deciduous forest of Dourdan (France), δ 13 C values measured in roe deer bone collagen, reflecting δ 13 C values for ingested plants over a life span, average − 24.7‰ 102 . An average value of − 29.7‰ may be estimated for plants 103, approximately 3‰ lower than the average values of − 27‰ expected for C 3 plants in open areas. When corrected for the fossil fuel effect, this would lead to a value of − 28.2‰ in preindustrial plants, or − 14.5‰ in the enamel of animals feeding on them. Therefore, significantly different δ 13 C values to the average value of − 11.8‰ expected for feeding in open settings, lowered to − 12.8‰ in winter (to account for a seasonal variation of approximately ± 1‰ 58, closer to − 14.5‰, could therefore reflect a contribution of plant resources gathered in the forest.
SI "Materials and Methods" contains a more detailed description of the archaeological sites, osteological analyses and stable isotope methods.