Letter | Published:

Early Middle Palaeolithic culture in India around 385–172 ka reframes Out of Africa models

Nature volume 554, pages 97101 (01 February 2018) | Download Citation

Abstract

Luminescence dating at the stratified prehistoric site of Attirampakkam, India, has shown that processes signifying the end of the Acheulian culture and the emergence of a Middle Palaeolithic culture occurred at 385 ± 64 thousand years ago (ka), much earlier than conventionally presumed for South Asia1. The Middle Palaeolithic continued at Attirampakkam until 172 ± 41 ka. Chronologies of Middle Palaeolithic technologies in regions distant from Africa and Europe are crucial for testing theories about the origins and early evolution of these cultures, and for understanding their association with modern humans or archaic hominins, their links with preceding Acheulian cultures and the spread of Levallois lithic technologies2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20. The geographic location of India and its rich Middle Palaeolithic record are ideally suited to addressing these issues, but progress has been limited by the paucity of excavated sites and hominin fossils as well as by geochronological constraints1,8. At Attirampakkam, the gradual disuse of bifaces, the predominance of small tools, the appearance of distinctive and diverse Levallois flake and point strategies, and the blade component all highlight a notable shift away from the preceding Acheulian large-flake technologies9. These findings document a process of substantial behavioural change that occurred in India at 385 ± 64 ka and establish its contemporaneity with similar processes recorded in Africa and Europe2,3,4,5,6,7,8,10,11,12,13. This suggests complex interactions between local developments and ongoing global transformations. Together, these observations call for a re-evaluation of models that restrict the origins of Indian Middle Palaeolithic culture to the incidence of modern human dispersals after approximately 125 ka19,21.

Main

The end of the Lower Palaeolithic Acheulian culture and beginnings of the Middle Palaeolithic, or Middle Stone Age, involved processes that marked substantial changes in hominin behaviour. The legacy of these changes, placed at approximately 300–200 ka2,3,4,5,6,7,8, is expressed primarily through technological transformations that involve a gradual decline in Acheulian large flake and core tools9, including bifaces; a proliferation and diversity of Levallois flake- and point-reduction strategies; and the evolution of blade technologies3,4,5,6,7,10,11. The behavioural processes that underpinned the transition from the Acheulian to the early Middle Palaeolithic or Middle Stone Age were variable and complex through space and time. This is evident at several Middle Palaeolithic and Middle Stone Age sites from the continuation of biface production—characteristic of Acheulian cultures —in small numbers amidst diverse Levallois- and blade-reduction sequences, and from the Acheulian roots of the Levallois concept8,10,13,14,15,16,17 (see Supplementary Information). The co-occurrence of Middle Palaeolithic or Middle Stone Age artefact sequences with not only modern humans2 but also other archaic species—with which modern humans could potentially interact—complicates investigations considerably7,8,14 (see Supplementary Information).

Despite the presence of numerous Middle Palaeolithic sites in South Asia, the age and origin of this cultural phase remain poorly documented8,18 (see Supplementary Information). Important features of the Middle Palaeolithic in India include the continuation of bifaces (albeit occurring less frequently or smaller in size than their Acheulian analogues); a predominance of small flake tools; the presence of Levallois and blade technologies and occasional points; and in some regions, depending on availability, an increased preference for fine-grained cryptocrystalline raw materials8,18 (see Supplementary Information). Radiometric ages have so far placed Indian Middle Palaeolithic cultures at approximately 140–46 ka1,20, potentially overlapping with a possible Late Acheulian occurrence at approximately 140–120 ka19. Regional variants and evolutionary trajectories of the Indian Middle Palaeolithic, and its association with modern humans or other archaic species and with the origins of Levallois technology, continue to be debated8,21. Patterns of hominin dispersals inferred from correlations between genetic, fossil and archaeological records are likewise unclear8. One theory22 links the Middle Palaeolithic in India with modern human dispersals out of Africa during and after Marine Isotope Stage 5 (130–80 ka), with populations surviving the catastrophic Toba volcanic eruptions at around 74 ka, whereas a contrasting theory20,23 associates the Indian Middle Palaeolithic with coexisting archaic species and advocates that the arrival of modern humans—ushering in microlithic blade assemblages and other cultural features—did not occur before Marine Isotope Stage 4 or 3 (71–57 ka). These gaps in our understanding of cultural transformations in South Asia arise from the scarcity of radiometric ages at excavated sites and of hominin fossils.

Here we present chronological and archaeological evidence from Attirampakkam (ATM), a Lower and Middle Palaeolithic site situated on the banks of a tributary stream of the Kortallaiyar River24 (Fig. 1, Extended Data Fig. 1). Excavations to depths of between 4 and 9 m in different trenches have revealed an alluvial sequence deposited by a small stream transporting a sediment load derived from shale, sandstone and laterite outcrops. From the base upwards, layers 8 to 6 are clay-rich and contain exclusively Early Acheulian assemblages (dating to approximately 1.7–1.07 million years ago (Ma))24; the overlying layers 5 to 1 contain the Middle Palaeolithic assemblages and form a sequence of clay-rich silt alternating with ferruginous gravel (Fig. 2, Extended Data Fig. 2). The mineral magnetic record25 indicates a seasonally dry tropical climate that was wetter during the deposition of layers 4 and 3, which are low-energy overbank silt deposits, and drier during the deposition of layers 5 and 2, which are gravel beds, with aridity persisting through layer 1 (see Supplementary Information).

Figure 1: Location of ATM.
Figure 1

In all panels, ATM is indicated by a red star. a, Regional setting. Scale bar (inset), 400 km. b, Drainage pattern, coded by Strahler stream order. c, Regional topography. The Allikulli and Satyavedu Hills are the stumps of Mesozoic alluvial fans. Their quartzite clasts have been redistributed eastward (white arrows) as loose debris by streams and slope processes, and thereby became available for toolmaking. L, laterite. Upward-pointing triangle, spot height of hill (elevation in metres above sea level); downward-pointing triangle, spot height of stream channel (elevation in metres above sea level). Scale bar, 4 km. d, Map highlighting the contrast between the wide ribbons of gravel formed by order-6 braided rivers, and those formed by the smaller order-3 ATM stream. Scale bar, 6 km. e, The meandering ATM stream amidst cultivated fields. Scale bar, 420 m. f, False-colour satellite image (IKONOS) of the excavations. Reprinted from ref. 26, with permission from Elsevier. Red tone: woody vegetation. Green and yellow: crops or bare soil. Scale bar, 80 m. Image sources: ac, Shuttle Radar Topography Mission 90 m Digital Elevation Data (CGIAR-CSI); d, e, Google, CNES/Airbus, DigitalGlobe.

Figure 2: Stratigraphy, cultural phases and chronology of MP layers at ATM.
Figure 2

a, General view of the site. b, Stratigraphic sequence at ATM showing layers 6 to 1. c, d, The stratigraphic section on the west wall of trench T7A that was dated by pIR-IRSL. Age clusters define Middle Palaeolithic cultural phases (error bars: 1σ). See Extended Data Fig. 2 for trench details.

Our description of the composition of the Middle Palaeolithic assemblage is based on the contents of three adjoining trenches (T7A, T7B and T7C) and involves the systematic analysis of 7,261 artefacts excavated from trench T7A (Figs 3, 4, Extended Data Figs 2, 3, 4, 5, 6, 7, 8, Supplementary Table 1). Like their Acheulian predecessors24, Middle Palaeolithic populations used locally available quartzite for making tools: other siliceous rock sources are absent in the region26. Quartzite occurred locally at the site as pebbles or cobbles in layer 5, but clasts suitable for the production of small flakes were absent in the finer-grained sediments of layers 4 to 1. As a result, quartzite clasts were instead collected from within a radius of 5–10 km from ATM, a landscape that is replete with other Middle Palaeolithic sites26 (Extended Data Fig. 1). Manuports in exotic raw material found at ATM included a tool on silicified wood and an unmodified quartz crystal (Fig. 4s).

Figure 3: Representative artefacts from layer 5, trench T7A.
Figure 3

a, b, Blade cores (see Extended Data Fig. 6a for details of blade removals in a). ce, Levallois cores. fi, Levallois points. j, Tanged point (see Extended Data Fig. 5a for details). kn, Retouched points. o, p, Bifacially flaked points. qu, Handaxes (ru are diminutive). v, Diminutive cleaver.

Figure 4: Representative artefacts from layers 2, 3 and 4 in trenches T7A, T7B and T7C.
Figure 4

ai, Representative artefacts from layer 2: Levallois flake core (a), blade cores (bd), tanged point (e) (see Extended Data Fig. 5b for enlarged images showing details), Levallois point with broken tip (f), scraper on Levallois flake (g) and blades (h, i). jn, Representative artefacts from layer 3: blade cores (j, k), diminutive cleaver (l), Levallois point (m) and blade fragment (n). os, Representative artefacts from layer 4: point (o), tanged point (p), Levallois point (q), Levallois flake core (r) and quartz crystal manuport (s).

Sediment samples from layers 5 to 1 in trench T7A were dated using post-infrared infrared-stimulated luminescence (pIR-IRSL) (Supplementary Tables 2–4, Extended Data Figs 9, 10; see Supplementary Information for methodological details of luminescence dating). The age sequence indicates three main chronological phases (Fig. 2, Supplementary Tables 2–4, Extended Data Figs 9, 10); of these, phases I and II correlate with changes observed in the cultural sequence. The technology associated with phase I (385 ± 64 ka) is confined to layer 5. A key feature of this technology is the almost complete abandonment of Acheulian large-flake strategies; these earlier strategies produced large cutting tools, including handaxes and cleavers, of greater than 10 cm maximum dimensions (Fig. 3, Extended Data Fig. 3). The sporadic occurrence of bifaces (including diminutive examples) and of a few large flakes suggests the persistence of Acheulian technological skills among the early Middle Palaeolithic hominin groups at ATM. This phenomenon has also been encountered at sites in Africa and parts of Europe10,13,15 (see Supplementary Information). A few handaxes that display a preferential flake-removal scar (Extended Data Fig. 3i) provide a hint that the Levallois technique may have possibly been derived from bifacial knapping strategies17. Small cores, including preferential and recurrent Levallois cores aimed at the production of small flakes, abound in layer 5 (Fig. 3, Extended Data Figs 3, 7b, c, 8) and—along with Levallois points—indicate that proficiency in Middle Palaeolithic knapping skills was well-established during phase I. These features were entirely absent from the earlier Acheulian assemblages at ATM24. Phase I was also associated with a high frequency of small retouched flake tools, which included scrapers, points on flakes, bifacially flaked points and a tanged point (see Supplementary Information). Incipient blade cores and associated debitage were also present (Extended Data Fig. 6). These features suggest that the behavioural changes that led to the establishment of an early Middle Palaeolithic culture at ATM were occurring during phase I.

The ATM Middle Palaeolithic phase II comprises layers 4 (268 ± 68 ka) and 3 (210 ± 64 ka). In this phase, Levallois strategies for the production of small flakes, blades, points and scrapers were a feature that continued from phase I (Fig. 4, Extended Data Fig. 4). A greater proficiency in blade detachment was evidenced, with the presence of uni- or bi-directional blade removals and associated debitage. In addition, Middle Palaeolithic assemblages in layer 2 (172 ± 41 ka) display up-sequence continuity in terms of the Levallois and blade techniques (Fig. 4, Extended Data Fig. 4).

The low artefact density in layer 1 (see Supplementary Information) precludes any robust definition of a phase III (which, if present, would date to approximately 74 ± 10 ka). However, the sharp drop in artefact densities in this layer suggests a decline in hominin occupation. Despite the absence of ash deposits at ATM, the fact that this sharp decline coincides with dates for the Toba volcanic super-eruption27 could suggest an environmental cause for site abandonment.

The silt-dominated sedimentary sequence at ATM throughout layers 8–124 suggests that the site was situated in a relatively low-energy floodplain environment compared to the substantially wider and more energetic neighbouring rivers such as the Kortallaiyar and Arani, which rank much higher in the stream-order hierarchy (Fig. 1).

Despite the antiquity of the Middle Palaeolithic artefacts from layer 5 (385 ± 64 ka; Fig. 2), their burial age nonetheless suggests that there was a long hiatus between Early Acheulian occupation (1.7–1.07 Ma)24 and the processes that led to the Middle Palaeolithic technological changes. This is inferred from the stratigraphic unconformity that separates layer 6 from layer 5. A Late Acheulian phase has been documented elsewhere in the region and in India more generally (see Supplementary Information and references therein); because this phase is absent at ATM, we infer that the site was temporarily abandoned during the Late Acheulian period —probably as result of local rather than regional causes, whether environmental or otherwise.

Changes in the cultural sequence that are associated with phase I establish that processes that signal the end of the Terminal Acheulian culture and transitions that mark the beginning of the Indian Middle Palaeolithic were occurring between approximately 450 and 320 ka. Notably, phase I includes Marine Isotope Stage 1128, which corresponds to a time period in which the global climate was similar to that during Marine Isotope Stage 5e and the Holocene. The warmer and wetter conditions associated with Marine Isotope Stage 11 would have been conducive to long-distance hominin dispersals, minimally obstructed by greener deserts between Sub-Saharan Africa and South Asia. Phase II spanned a succession of global Middle Pleistocene climatic changes, with little discernible influence on the intensity or continuity of site occupation. The overlap between ages at ATM and the few dated Late Acheulian sites elsewhere in India19 also suggests that spatial variability among Palaeolithic cultural sequences is larger than previously thought. The assemblage structure at ATM thus adds not only antiquity but also diversity to the mosaic of younger Middle Palaeolithic sites that have previously been described22.

The behavioural transformations that mark the advent of the Indian Middle Palaeolithic at ATM are summarized by the following diagnostic features: the obsolescence of Acheulian large-flake reduction sequences, with a directional shift towards smaller tool components; the adoption and continuance of Levallois recurrent and preferential strategies; a gradual intensification of blade reduction; and an increased use of finer-grained quartzite during phase II than during earlier occupations. A gradual discontinuation of biface use—which becomes definite at ATM after approximately 172 ± 41 ka—has been reported at other Middle Palaeolithic and Middle Stone Age sites worldwide (see Supplementary Information and references therein). Accordingly, and given the well-recognized complexity of cultural transitions7,8,10,11,12,13, where bifaces occasionally occur amidst reduction sequences that overwhelmingly suggest new technical preferences and behavioural strategies, it would be inappropriate to use sporadic bifaces as supporting evidence for the persistence of a separate Acheulian culture (see Supplementary Information).

Conclusive correlations between the Middle Palaeolithic assemblages at ATM and a specific hominin species2,8,21—whether modern humans2 or archaic hominins29,30—cannot be established because India currently lacks fossil or genetic evidence for this time period other than the Narmada fossil cranium, which could signal the late survival of an archaic species (see Supplementary Information). Evidence of distinct behavioural changes is nonetheless provided by the assemblage structure in the form of new technological strategies, which retain minor components of an archaic nature at around 385 ± 64 ka but depart considerably from Acheulian strategies, and which evolved at ATM for approximately another 200 ka. These processes thus establish the presence of a fully fledged Middle Palaeolithic culture in India at around 385–172 ka, which long pre-dates any previous evidence that suggests Middle Palaeolithic technologies were disseminated out of Africa by modern humans from around 125 ka or later1,8,20,21,22,23. The respective parts played by local rather than external influences in the early rise of Middle Palaeolithic culture in India remain uncertain. However, when set in a global context (see Supplementary Information and references therein), the sequence at ATM suggests a succession of population dispersals across South Asia during the Middle Pleistocene, which perhaps involved interactions with other archaic species.

Methods

Optically stimulated luminescence (OSL) provides the burial age of archaeological sediments on the premise that at the time of deposition there was a minimal OSL signal in the constituent mineral grains. The reduction of OSL before burial occurs as a consequence of sediment grains being exposed to daylight during their transport and/or by human activity. On burial, exposure to daylight ceases and the re-accumulation of a luminescence signal in the mineral grains begins as a result of irradiation from ambient radioactivity. This continues until subsequent re-exposure or excavation for sampling. The measurement of total luminescence and its conversion to radiation dose units is possible using standard protocols (see Supplementary Information and references therein), and the corresponding radiation dose is called the palaeo-dose. When divided by the annual radiation dose, this gives the age (that is, time spent in the dark since the last exposure to daylight). The annual radiation dose is computed by the measurement of radioactivity concentrations produced in the burial environment by uranium, thorium, potassium and cosmic rays.

The luminescence age of a sedimentary deposit provides the time at which any archaeological artefacts contained within it were discarded and buried. In order to date the sedimentary deposits containing Middle Palaeolithic assemblages at ATM, pIR-IRSL was preferred from among the range of existing luminescence methods, because of the anticipated antiquity of the stratigraphy and because the use of quartz OSL was hampered by the saturation of its luminescence signal (see Supplementary Information). Given the high solar irradiance, low sediment accumulation rate and indications from independent measurements on modern samples in similar environments in India (see Supplementary Information and references therein), it was reasonable to assume that the pIR-IRSL signal was bleached to low residual levels of less than a few Grays. This assumption is supported by the tight distribution of subsample ages in each sample (see Supplementary Information).

Data availability

All relevant data are included in the article and its Supplementary Information.

References

  1. 1.

    et al. A ~200 ka record of climatic change and dune activity in the Thar Desert, India. Quat. Sci. Rev. 29, 3095–3105 (2010)

  2. 2.

    et al. The age of the hominin fossils from Jebel Irhoud, Morocco, and the origins of the Middle Stone Age. Nature 546, 293–296 (2017)

  3. 3.

    ‘Early’ Middle Stone Age lithics of the Kapthurin Formation (Kenya). Curr. Anthropol. 47, 367–375 (2006)

  4. 4.

    The Acheulian to Middle Stone Age Transition: Tephrostratigraphic Context for Archaeological Change in the Kapthurin Formation, Kenya (Univ. Connecticut Press, 2003)

  5. 5.

    , , , & Chronological and behavioral contexts of the earliest Middle Stone Age in the Gademotta Formation, Main Ethiopian Rift. Quat. Int. 331, 6–19 (2014)

  6. 6.

    et al. New radiometric ages for the Fauresmith industry from Kathu Pan, southern Africa: implications for the Earlier to Middle Stone Age transition. J. Archaeol. Sci. 37, 269–283 (2010)

  7. 7.

    First Neanderthal settlements in northern Iberia: the Acheulean and the emergence of Mousterian technology in the Cantabrian region. Quat. Int. 326–327, 288–306 (2014)

  8. 8.

    & Modern human origins and the evolution of behavior in the Late Pleistocene record of South Asia. Curr. Anthropol. 46, S3–S27 (2005)

  9. 9.

    Acheulian Large Flake Industries: Technology, Chronology, and Significance (Archaeopress, 2007)

  10. 10.

    & in Transitions Before the Transition: Evolution and Stability in the Middle Palaeolithic and Middle Stone Age (eds & ) 257–277 (Springer, 2006)

  11. 11.

    Patterns of technological change at the origin of Homo sapiens. Before Farming 3, 1–5 (2003)

  12. 12.

    et al. The southern route ‘out of Africa’: evidence for an early expansion of modern humans into Arabia. Science 331, 453–456 (2011)

  13. 13.

    et al. Early Levallois technology and the Lower to Middle Paleolithic transition in the Southern Caucasus. Science 345, 1609–1613 (2014)

  14. 14.

    et al. Stratigraphic, chronological and behavioural contexts of Pleistocene Homo sapiens from Middle Awash, Ethiopia. Nature 423, 747–752 (2003)

  15. 15.

    , & The Early Middle Palaeolithic in Britain: archaeology, settlement history and human behaviour. J. Quat. Sci. 21, 525–541 (2006)

  16. 16.

    ., & Levallois lithics from the Kapthurin Formation, Kenya: Acheulian origin and Middle Stone Age diversity. Afr. Archaeol. Rev. 22, 199–229 (2005)

  17. 17.

    & Note on a link between Acheulian handaxes and the Levallois method. J. Israel Prehist. Soc. 31, 9–23 (2001)

  18. 18.

    A Re-examination of the Palaeolithic Archaeological Record of Northern Tamil Nadu, South India (BAR International Series, 2001)

  19. 19.

    et al. Late Acheulean hominins at the Marine Isotope Stage 6/5e transition in north-central India. Quat. Res. 75, 670–682 (2011)

  20. 20.

    , & Continuity of microblade technology in the Indian Subcontinent since 45 ka: implications for the dispersal of modern humans. PLoS ONE 8, e69280 (2013)

  21. 21.

    & in Southern Asia, Australia and the Search for Human Origins (eds & ) 64–75 (Cambridge Univ. Press, 2014)

  22. 22.

    et al. Middle Paleolithic assemblages from the Indian subcontinent before and after the Toba super-eruption. Science 317, 114–116 (2007)

  23. 23.

    , , , & Genetic and archaeological perspectives on the initial modern human colonization of southern Asia. Proc. Natl Acad. Sci. USA 110, 10699–10704 (2013)

  24. 24.

    et al. Early Pleistocene presence of Acheulian hominins in South India. Science 331, 1596–1599 (2011)

  25. 25.

    et al. A rock magnetic record of Pleistocene rainfall variations at the Palaeolithic site of Attirampakkam, southeastern India. J. Archaeol. Sci. 38, 3681–3693 (2011)

  26. 26.

    , , & Applications of satellite remote sensing for research and heritage management in Indian prehistory. J. Archaeol. Sci. 37, 2316–2331 (2010)

  27. 27.

    et al. Environmental impact of the 73 ka Toba super-eruption in South Asia. Palaeogeogr. Palaeoclimatol. Palaeoecol. 284, 295–314 (2009)

  28. 28.

    et al. Palaeoclimatology: the record for marine isotopic stage 11. Nature 436, 39–40 (2005)

  29. 29.

    et al. Southern African ancient genomes estimate modern human divergence to 350,000 to 260,000 years ago. Science 358, 652–655 (2017)

  30. 30.

    , , , & Homo naledi and Pleistocene hominin evolution in subequatorial Africa. eLife 6, e24234 (2017)

  31. 31.

    , , & Preliminary report on excavations at the Palaeolithic site of Attirampakkam, Tamil Nadu (1999–2004). Man and Environment XXXIX, 1–17 (2004)

Download references

Acknowledgements

S.P. and K.A. thank the Sharma Centre for Heritage Education, the L. S. B. Leakey Foundation, the Earthwatch Institute, the Homi Bhabha Fellowships Council (S.P.: 2000–2002; K.A.: 2014–2016) and the ISRO-GBP program for funding various aspects of the research project, and the Archaeological Survey of India and Department of Archaeology, Government of Tamil Nadu, for issuing licenses. Y.G. benefited from an Institut Universitaire de France grant for field and analytical work. A.K.S. acknowledges the Department of Science and Technology and the Department of Atomic Energy, India, for a J. C. Bose national fellowship and for Raja Ramanna fellowships, respectively. H.M.R. was supported by the contingency grant of the J. C. Bose fellowship awarded to A.K.S. S.P. and K.A. thank M. Taieb for his encouragement.

Author information

Affiliations

  1. Sharma Centre for Heritage Education, 28 Ist Main Road, C.I.T. Colony, Mylapore, Chennai 600004, Tamil Nadu, India

    • Kumar Akhilesh
    •  & Shanti Pappu
  2. Department of Physics, Electronics and Space Science, Gujarat University, Navrangpura, Ahmedabad 380009, India

    • Haresh M. Rajapara
  3. AMOPH Division, Physical Research Laboratory, Navrangpura, Ahmedabad 380009, India

    • Haresh M. Rajapara
    •  & Ashok K. Singhvi
  4. Université de Lyon, Department of Geography, UMR 5600 Environnement Ville Société, 5 Avenue Pierre Mendès-France, F-69696 Bron, France

    • Yanni Gunnell
  5. Geosciences Division, Physical Research Laboratory, Navrangpura, Ahmedabad 380009, India

    • Anil D. Shukla

Authors

  1. Search for Kumar Akhilesh in:

  2. Search for Shanti Pappu in:

  3. Search for Haresh M. Rajapara in:

  4. Search for Yanni Gunnell in:

  5. Search for Anil D. Shukla in:

  6. Search for Ashok K. Singhvi in:

Contributions

K.A. and S.P. direct the project, are researching ATM and neighbouring sites and analysed the lithic artefacts; H.M.R., A.D.S. and A.K.S. were responsible for the luminescence sampling and dating; Y.G. analysed the geomorphology and palaeoenvironmental evidence at the site. All authors contributed to the writing of the manuscript.

Competing interests

The authors declare no competing financial interests.

Corresponding author

Correspondence to Shanti Pappu.

Reviewer Information Nature thanks K. Fitzsimmons, M. Petraglia and E. Rhodes for their contribution to the peer review of this work.

Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Supplementary information

PDF files

  1. 1.

    Supplementary Information

    This file contains Supplementary Text, Supplementary Tables 1-4 and Supplementary References.

  2. 2.

    Life Sciences Reporting Summary

About this article

Publication history

Received

Accepted

Published

DOI

https://doi.org/10.1038/nature25444

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.