Continuous presence of proto-cereals in Anatolia since 2.3 Ma, and their possible co-evolution with large herbivores and hominins

Cereals are a central resource for the human diet and are traditionally assumed to have evolved from wild grasses at the onset of the Neolithic under the pressure of agriculture. Here we demonstrate that cereals may have a significantly longer and more diverse lineage, based on the study of a 0–2.3 Ma, 601 m long sedimentary core from Lake Acıgöl (South-West Anatolia). Pollen characteristic of cereals is abundant throughout the sedimentary sequence. The presence of large lakes within this arid bioclimatic zone led to the concentration of large herbivore herds, as indicated by the continuous occurrence of coprophilous fungi spores in the record. Our hypothesis is that the effects of overgrazing on soils and herbaceous stratum, during this long period, led to genetic modifications of the Poaceae taxa and to the appearance of proto-cereals. The simultaneous presence of hominins is attested as early as about 1.4 Ma in the lake vicinity, and 1.8 Ma in Georgia and Levant. These ancient hominins probably benefited from the availability of these proto-cereals, rich in nutrients, as well as various other edible plants, opening the way, in this region of the Middle East, to a process of domestication, which reached its full development during the Neolithic.

www.nature.com/scientificreports/ populations between Africa and Eurasia. Simultaneous or later signs of early Pleistocene human occupation based on lithic tools were found in Anatolia, in particular in the nearby Gediz river site 5 (120 km NW of Acıgöl) and Dursunlu site (180 km E of Acιgöl). More distant earlier presence of hominins around ~ 1.8 Ma were evidenced in the Levant 6 , Georgia 7 and Ciscaucasia 8 , pointing toward the high probability of hominin occupation of Anatolia, mid-way of Levant and Caucasus, at least sporadic since 1.8 Ma. Evidence from China and Jordan points toward even earlier presence of hominins out of Africa as early as 2.1 9 and 2.5 Ma 10 , respectively. In the Denizli and Burdur basins, adjacent to Acıgöl, abundant fossil remains of large mammals typical of the late Villafranchian have been unearthed, including an extinct species of primitive mammoth (Archidiskodon m. meridionalis), an extinct rhinoceros (Stephanorhinus cf. etruscus), several species of horses (Equus) sp., E. cf. altidens/E. cf. mygdoniensis and E. cf. apolloniensis), small and large-sized deer (Metacervoceros rhenanus, Arvernoceros sp., Cervalces (Libralces) ex gr. minor-gallicus), a large primitive okapi (Palaeotragus sp.), a primitive camel (Paracamelus cf. gigas), a large bovine today extinct (Leptobos cf. etruscus) and several antelopes (Gazellospira torticornis, Gazella sp.) [11][12][13][14] . They indicate that Homo erectus coexisted with a rich and diversified mega-fauna from which they were largely dependent 15 .

Results
Vegetation history of the Acıgöl area. Our 19,20 .
The vast freshwater stretch of Acıgöl, located in a predominantly arid limestone hills environment, seems to have been a crucial resource for the mammalian fauna, which probably concentrated around the site in search Earliest hominin presence in Anatolia [4], Ciscaucasia [8], China plus Jordan [9,10], is indicated in red, blue and green, respectively.   3,5). Unexpectedly, they present the same morphological characteristics as that of modern cereal grains 25,26 , namely an average size of ≥ 40 µm and a large pore + annulus (≥ 8 µm). As by definition cereals are cultivated plants, we will call the corresponding plants "proto-cereals" to highlight that their pollen are identical to those of cereals. This resemblance can be seen clearly in Fig. 5, where we have brought together fossil cereals from Acιgöl (Fig. 5, photos 1-7), from Roman time (Fig. 5, photo 8), not modified by modern agricultural practices, and from the current wheat field of the Lauragais agricultural plain, Gardouch, France (Fig. 5, photo 9). Cerealia-type frequencies reach a maximum of 9% of the PS around 2.2 Ma and can be as abundant as wild Poaceae pollen (Fig. 3).
The Cerealia/Poaceae ratio shows that 24.66% of all Poaceae are proto-cereals from 2.0 to 2.3 Ma (Supplementary  Table 1). Such high proto-cereal rates are almost never reached in pollen records, even in recent periods and in the presence of agriculture, because of the very low pollen dispersal capacity of cereals 27 . A lowering of frequencies down to 2-4% range is recorded in younger periods (Fig. 3), as well as a step like decrease of the Cerealia/ Poaceae ratio (Fig. 6). This change may be related to the Middle-Pleistocene Transition (MPT) cooling and to the mega-mammal fauna change from a Villafranchian to a Galerian type 28 . MPT and faunal changes occurred around 0.9-1.0 Ma, while a decrease in our proto-cereal starts around 1.5 Ma, however signs of cooling and amplified climatic cycles predate the MPT 28 .
The histogram of wild Poaceae and proto-cereal pollen size (Fig. 7a) shows that there are a number of pollen populations modes around 30, 37.5, 45-50, supporting the idea that the larger grain sizes cannot be interpreted Our interpretation is that proto-cereals recorded throughout the Acıgöl sequence derive from wild Poaceae. Their emergence and predominance may have been favoured by the impact of large herbivore herds attracted to Acıgöl lake shores, and through genetic drift. Through the process of trampling, nitrogen enrichment of soils and browsing, large mammal herds could have altered the genotype of proto-cereals naturally present in Acıgöl and   www.nature.com/scientificreports/ thus, favoured the emergence of modern cereals. For genetic reasons, the descendants of these proto-cereals are not represented today among cultivated Poaceae because domestication bottlenecks eliminate genetic variation 29 . Is there a relationship between the size of proto-cereal pollen and climate? To our knowledge, the genetic literature does not show any relationship between the increase in pollen size and temperature. However, there does seem to be a relationship with atmospheric drought 30,31 which is said to have favoured the appearance of polyploidy in certain species of Poaceae. It cannot be excluded that climate has had an influence on the protocereal genome, but only the interaction between herds of large herbivores and proto-cereal steppes can explain why proto-cereal pollen has never been found in such abundance elsewhere in Pleistocene pollen records.
The ancestors of cultivated trees (Olea sp., Juglans sp., Castanea sp., Corylus sp., Prunus t.), typical of the modern Mediterranean agriculture, are also present in the Acıgöl sequence ( Fig. 3 and Supplementary Table 5). Their amount increases after 1.5 Ma, mainly due to Olea (Fig. 6). Other potentially edible plants such as Ephedra, Hippophae, all the Compositae and the Fagaceae have been identified in the pollen assemblages. They correspond to 54.4% of plants identified in the pollen assemblages. Among these plants, there are 72% grasses and 28% trees and, among edible organs, 51% are vegetables and 20% are seeds ( Supplementary Fig. 1a,b). These results testify to the potential wealth of accessible food resources that human and animal populations could feed on. Interestingly, studies carried out in Spain on the present-day consumption of wild plants lead to results close to those obtained at Acıgöl, with 87% grasses and 13% trees 32 .
In recent years, new biological and archaeological data obtained from sites with human occupation have improved our knowledge of the beginnings of agriculture and the modalities of its implementation. In the Levant, the Ohalo II site highlights the presence of proto-cereal seeds, and flint tools to harvest, as early as 23,000 years before the present 33 . Further north, on the archaeological site of Gesher Benot Ya'aqov, proto-cereal seeds (oats, Avena) as well as pollen from cereals and trees currently cultivated, were identified over a period ranging from 750,000 to 820,000 years 34,35 . Moreover, recent genetic data indicate that the emergence of agriculture did not occur at a single location at the onset of the Neolithic (e.g. the "Fertile Crescent" hypothesis) but is, on the contrary, an evolutionary and multi-regional long-term phenomenon [36][37][38] . Alternatively, or simultaneously, are the hominins also partly responsible by having developed episodes of a form of transitory "proto-agriculture"? We already know that this domestication process was discontinuous with shutdown and restart phases 37,39 . Acheulean www.nature.com/scientificreports/ lithic tools, characterised by symmetrically shaped bifaces, testify to the rather advanced cognitive capacities of early Pleistocene populations that may have visited the lakeshore of Acıgöl 5 . Hominin populations may also have benefited from this opportunity to diversify their food regime with easily harvested and nutrient-rich wild plants (Supplementary Table 5), as it is the case today for hunter-gatherer populations in Africa and elsewhere in the world.

Discussion
Challenging the paradigm of domestication: how and when? The question surrounding the Neolithic emergence with plant and animal domestication has been debated for decades and has been the subject of countless studies, in most cases carried out by archaeologists and geneticists focusing either on plant macrorests from archaeological sites or on the genome history of cultivated plants 40,41 . The study of natural environments (wetlands, lakes), such as at Acıgöl, and of their microbiological content has so far been largely neglected to tackle this question. www.nature.com/scientificreports/ What happened in the Neolithic, when humans went from a hunter-gatherer to a farmer lifestyle? Did they reproduce conditions that existed two million years ago? Has there been a new stage of cereal speciation linked to humans? Vaughan 42 emphasises that "the time scale of domestication of 10,000 years or less is a very short evolutionary time span" (p. 893). The proto-cereal pollen of Acıgöl appears to indicate that the genetic modification of cereals could have also been a natural process that appeared long before agriculture emerged, and that the conditions were already present when human populations shifted from hunter-gatherer to agricultural societies. Our results enable an important enigma to be reformulated in relation to human evolution: when did cereals appear and are humans solely responsible? Our study appears to challenge the long-held paradigm that humans were the progenitors of cereal grasses, when it seems in fact possible that cereal grasses may have appeared naturally, with humans simply accelerating their expansion. If this were confirmed by the presence of proto-cereal pollen in sediments of the lower Pleistocene or in older sediments from other regions, this would necessitate a fundamental revision of our overall vision of the history of human nutrition. To substantiate this, we can report our initial observation of similar proto-cereals in the Kocabaş travertine sequence (Fig. 8)

Material and methods
Chronostratigraphy. 188 oriented standard cylinders were regularly subsampled in Acıgöl core C3, from 100 to 600 m depth. Natural remanent magnetization (NRM) were measured with the Superconducting Rock Magnetometer (SRM 560R, 2G Enterprises) of the Rock Magnetic Laboratory in CEREGE. Demagnetization was mostly performed using alternating magnetic fields. Magnetostratigraphy 1 allowed identifying the temporal tie points listed in Supplementary Table 6. Bruhnes age cyclostratigraphy based of Xray fluorescence analyses was proposed by Akcer-On 43 , leading to an estimated age of 700 ka at 170 m, coherent with the magnetostratigraphy. 566 samples were used to build a magnetic susceptibility curve versus depth, pointing toward a regular climatic cyclicity record throughout the core, in agreement with sedimentological and palynological proxies 1 .

Fossil pollen.
Pollen analyses were conducted on 87 samples, of which 72 were polliniferous. Non-polliniferous samples correspond to micro-tephra layers (markers of volcanic eruptions or redeposition of detrital tephra) or levels rich in microcharcoals (fire indicators) or algae (blooms of Botryococcus or Chrysophyceae in response to a disturbance of the lacustrine trophic system). At Kocabaş, pollen analyses were conducted on three samples taken from fluviatile conglomerates, on both sides of the travertine deposit in which the remains of the Kocabaş Man were discovered and dated to 1.2-1.6 Ma 4 .
The sporo-pollinic material was extracted from the sediment using chemical treatment (HCl acid, NaOH 10% and acetolysis) followed by flotation in heavy liquid (d = 2) and 160 + 10 µm sievings 44 . The identification of pollen, spores and Non-Pollen Palynomorph (NPP) was carried out with a photonic microscope Biomed Leitz (500× magnification). Standard palynological identifications were based on the pollen reference collection of IMBE (CNRS, Aix-en-Provence, France), pollen photographic atlases [45][46][47][48] and articles on NPP [49][50][51] . The pollen percentages were calculated on a Pollen Sum (PS) including all plants except Filicophytes, Bryophytes, Algae, and NPP. For each sample, pollen counting was carried out up to 300 to 400 grains (phanerogams alone) and then continued over the whole slide for rare taxa 52 . At Acıgöl, the mean PS varies from 3025 grains (all pollen, spores and NPP included) to 558 grains per sample (all pollen, without spores and NPP). The average pollen concentration (weighting method 53 ) is of 7829 pollen/g sediment (minimum: 38, maximum: 451,556). The average weight of the samples is 11.24 g (minimum: 3.5 g, maximum: 16.6 g). The total number of taxa identified (pollen, spores, NPP, algae) is 201. At Kocabaş, the mean PS varies from 356 grains (all pollen, spores and NPP included) to 314 grains per sample (all pollen without spores and NPP). The average pollen concentration is 6.5 www.nature.com/scientificreports/ pollen/g sediment (minimum: 2.9, maximum: 9.9). The weight of the samples is 250 g each. The total number of taxa identified (pollen, spores, NPP, algae) is 69. The identification of pollen of cereals relies on morphological characteristics 25,53 , especially the longest axis of the grain which is a reliable means to discriminate cereal grains, and the pore + annulus diameter which can be used as a complementary criterion. According to Andersen 25 , pollen of Poaceae with a longest diameter larger than 37 µm are pollen of cereals. For Bastin 54 and Leroi-Gourhan 55 , the diameter of cereals should be ≥ 40 µm and the external diameter of the annulus should be at least 10 µm. Detailed information can be found in Emery-Barbier and Thiébault 56 . In our work, we have followed the proposal of Bastin 54 and Leroi-Gourhan 55 and designated as cereals the pollen of Poaceae with (1) a longest diameter ≥ 40 µm, (2) a large and protuberant annulus + pore, (3) and scabrate to verrucate surface sculpturing of the exine. We thus call proto-cereals (or Cerealia type) the Poaceae pollen ≥ 40 µm and wild Poaceae the others, although we acknowledge that some rare wild Poaceae may show diameter ≥ 40 µm, such as Aegylops sp., Glyceria sp. 57 . The measurements made on 991 pollen grain of wild Poaceae and proto-cereals of 10 samples of Acıgöl (core 3) chosen among the richest in pollen of cereals (Supplementary Table 1) show that the distribution is polymodal (Supplementary Fig. 1a). The mean pollen diameter of wild Poaceae and cereal is respectively 31.01 and 45.07 µm. The mean pore diameter is 4.28 µm (minimum: 2.5 µm; maximum: 5 µm) and the mean pore + annulus diameter is 10.12 µm (minimum: 6.25 µm; maximum: 13.75 µm). In the literature, the mean size of the cereal pore is 2.7 µm 57 . There is no relationship between the diameter of the wild Poaceae and proto-cereal and the size of the pore (R 2 = 0.046) or the pore + annulus (R 2 = 0.097).
Modern pollen rain. Sampling of moss pollsters were carried out in May 2017 along the Acıgöl lakeshores and surroundings, in a wide variety of vegetal landscapes, altitudes, and anthropisation levels (Supplementary Table 2), in order to reconstruct the modern pollen rain and to improve the accuracy of palaeoenvironmental reconstructions based on fossil pollen analyses. In this article, we show the results of the pollen analysis of eight samples. The samples were taken on the lakeshores, in specific ecosystems with halophytic or unsalted vegetation. The objective was to evaluate the amount of cereals in the current pollen rain and to find out whether or not cereal pollen found in Acıgöl sediments may have originated from the hygrophilous vegetation of the lake shoreline. The results show that the pollen of cereal is underrepresented in the current pollen rain of Acıgöl (range of values between 0 and 0.97% of the PS, Cerealia/Poaceae ratio 4.52% against 16.9% (n = 72) in the Acıgöl core Supplementary Tables 3 and 4). This indicates that there is not a biological specificity among the Poaceae community of the Acıgöl lakeshore and that there is a good chance that the proto-cereal pollen of the Acıgöl sedimentary archives come from the surrounding steppe communities and not from hygrophilous wild Poaceae.

Data availability
All data generated or analysed during this study are included in this published article (and its Supplementary Information files) and are available in zenodo at https:// zenodo. org/ record/ 59126 16. www.nature.com/scientificreports/