Supporting Information

Piperno et al. 10.1073/pnas.0812525106 SI Materials and Methods structures which are homologous to each other, have been shown Modern Reference Collections and Microfossil Identification. Our to be controlled primarily by tga1, a major domestication gene reference collections of phytoliths and starch grains include with significant effects (10, 17, 18). tga1 underwrites the degree more than 2,000 and about 500 species, respectively, and include of silicification of the glumes and rachids (cupules) of the many wild taxa of economic importance, most of the known fruitcases and cobs. In teosinte, the entire epidermis, consisting domesticated native to Central and South America, and of both the long and short cells, is silicified, whereas in wild progenitors and other close wild relatives of the crop plants. (which requires less natural protection from its herbivores), Investigating the history of maize and squash in the study region silicification is greatly reduced, and only the short epidermal was one of our priorities; therefore, our reference collections cells (which produce the phytoliths called rondels) are filled with include all known species and subspecies of teosinte; 24 maize silica. In addition, the rondels produced in teosinte are more races from Central and South America, including 10 traditional highly decorated than those in maize (a result also of more Mexican land races; and all domesticated and most known wild extensive lignification in teosinte), and the rondel phytoliths in species of Cucurbita, including all those found in Mesoamerica, maize cobs have a more diverse morphology and are in forms not such as C. argyrosperma ssp. sororia, which is native to the study found in teosinte. These differences result in the formation of region and is the wild ancestor of C. argyrosperma (the silver- distinct and identifiable phytoliths in maize and teosinte that seeded squash or cushaw pumpkin) (1). allow them to be distinguished from each other and from With regard to starch grain identification, previous research non-Zea wild grasses, including the (10–16). has demonstrated that starch grains in maize commonly range from about 8 to 26 ␮m in maximum length and from 12 to 16 ␮m Other (Non-Maize) Starch Grains Present on the Stone Tools. Four in mean length (2–8). In non-Zea grasses, grain size typically yam grains (Dioscorea sp.), 3 legume grains, and 1 Marantaceae ranges from about 2 to 18 ␮m in maximum length and from 3 to grain occurred on tool 316d, a large, preceramic grinding stone 11 ␮m in mean length. In many wild species, the maximum grain base recovered from 60–67 cm b.s. of unit 1. The yam grains size is only 6–9 ␮m (2, 4–8). Often these grains are too small to cannot be identified as belonging to either a wild or domesticated allow confident discernment of surface features, but species with species, because considerable work is needed on wild Dioscorea larger grains have been found to have dissimilar morphological species native to Mexico to rule out possible confusion with characteristics to maize (4–7). In this study, we examined cultivated/domesticated taxa. This is the first empirical indica- additional species of non-Zea grasses common in the Mexican tion of yam usage in tropical Mexico during the pre-Columbian flora (Table S1), including a putative early cultivar from high- era, however. The legume grains are similar to some that occur land Mexico, Setaria parvifolia (Poiert) (formerly S. geniculata) in Phaseolus, but they lack some attributes common in P. vulgaris (9). As in other non-Zea grasses, starch grain size is considerably and P. lunatus, such as the presence of lamellae and fissures; thus, smaller than in maize, and morphological characteristics also we cannot unequivocally assign them to a specific legume taxon serve as distinguishing criteria. at this time. Similarly, the Marantaceae grain cannot be assigned With regard to the differentiation of maize and teosinte on the to a specific genus. One unknown grain occurred on tool 365a. basis of starch grain size (see Table S2), mean grain length in teosinte ranges from 9.5 ␮m (Race Balsas) to 11.9 ␮m(Z. Other Types of Phytoliths Present in the Sediments. Marantaceae luxurians, endemic to Guatemala). Maximum grain length varies seed phytoliths, probably from either Maranta or Stromanthe, from 2 ␮mto28␮m; the latter was represented by a single grain were well-represented throughout the sedimentary sequence. observed in a specimen of Chalco teosinte (Z. mexicana), a race These phytoliths are not like those from arrowroot (M. arundi- that commonly hybridizes with maize (7). Maximum grain size nacea L.). A type of phytolith produced in the foliage and does not exceed 20 ␮m in non-Chalco teosintes and 18 ␮min sometimes the wood of various tree species also was common. Balsas teosinte. In contrast, in maize, mean length varies from Also persistently present in lower frequencies were phytoliths 11.4 to 15.8 ␮m, and in most races, mean length is 12.5 ␮m and from palms, Cyperaceae, and Asteraceae. maximum length is 20 ␮m, reaching 26 ␮m in some cases (7). Differences in such features as grain shape and surface Discriminating Phytoliths from Maize and Teosinte Culms (Stalks). sculptoring also provide clear morphological contrasts between Culms or stalks of grasses produce various idiosyncratic forms teosinte and maize (7) (Table S2). For example, nearly every race that do not occur in the leaves and inflorescences of the plants of maize studied has dominant proportions of ‘‘irregular’’ grains (19). Culm phytoliths in maize often are thick and irregularly (those without a clearly describable shape), and many have cross or bilobate in shape with unusual, deeply notched bases. defined (deeply impressed) compression facets, which develop These are distinct from phytoliths produced in leaves and cobs. when the grains are packed together during their formation in A stalk of Z. mays ssp. parviglumis from Guerrero state sampled the cellular organelles called amyloplasts. In contrast, teosinte from the herbarium folders at the U.S. National Museum of exhibits significant percentages of oval and bell-shaped grains, Natural History (NMNH 3123148) produced the same types of which are nearly absent in maize, and has far fewer irregular phytoliths, as well as other phytoliths not seen in the maize stalks grains. Most teosinte grains also lack defined compression facets studied. Although further work is needed to more robustly assess and have different types of fissures (i.e, cracks at the hilum, the whether the maize and teosinte phytoliths are diagnostic to Zea botanical center of the grain). or to the subspecies level, these phytoliths can be used to identify With regard to phytolith identification, criteria for the iden- stalk deposition. To be confident that young maize stalks, which tification of maize and teosinte phytoliths developed by ourselves presumably would have been used because they contain the and other investigators, including with the use of large blind highest quantity of sugar, produce phytoliths, we grew maize studies, are well described elsewhere (10–16). Importantly, the from seed at the Smithsonian Tropical Research Institute in considerable differences in morphological attributes of phyto- Panama. Stalks were harvested 53 days after they were planted liths formed in the fruitcases of teosinte and cobs of maize, and investigated for silica content and phytolith morphological

Piperno et al. www.pnas.org/cgi/content/short/0812525106 1of10 attributes. The phytolith content was high, and the diagnostic dominate wild fruits (Fig. S5). The incompletely silicified phy- phytoliths produced by mature stalks were commonly present in toliths also commonly form as half-spheres (10, 20). All of these the young stalks. features are linked to the suppression of lignification and We restudied phytolith samples from important sites in Pan- silicification under artificial selection for softer rinds. ama containing starch grain and phytolith evidence for prece- We explored this issue in greater detail by examining 100 ramic maize (4, 10). No Zea-type stalk phytoliths were observed. phytoliths from each of 4 different fruits representing 3 different populations of C. argyrosperma ssp. sororia, the wild ancestor of Presence of Preceramic Phytoliths Indicative of Human Selection at C. argyrosperma. The fruits are homozygous at the Hr locus. In the Hr Genetic Locus. In modern domesticated species and the F 2 of these fruits, only 3 phytoliths with surface features (e.g., 1 marks or holes) characteristic of incomplete silicification were and F progeny of hybrids made between C. sororia and C. 2 recorded. In the other 2 fruits, 1 and 0 phytoliths of this type argyrosperma and between C. sororia and other domesticated occurred. Scans of phytolith preparations made from other fruits species, many phytoliths from plants that are heterozygous at the of C. sororia and other wild species further indicate that these Hr locus (Hr hr), and thus exhibit softer rinds than typically occur characteristics are rare in wild Cucurbita. in wild plants, acquire characteristic surface features, such as In preceramic samples 319d, 325 h, 316c, 318d, and 318e, in incompletely formed and fainter scalloped impressions and even which numerous squash phytoliths occurred, Ͼ 73% of the holes at the surface (Fig. S4B). These patterns may result from phytoliths (a far greater amount than in any wild species) the influence of modifier genes or incomplete dominance of the exhibited surface features like those in modern specimens with Hr locus (20). In any case, the types of phytoliths produced domesticated germ plasm heterozygous for Hr. Half-spheres also significantly outnumber the completely silicified forms that were routinely recorded.

1. Sanjur OI, Piperno DR, Andres TC, Wessel-Beaver L (2001) Phylogenetic relationships 13. Mulholland SC (1993) in Current Research in Phytolith Analysis: Applications in among domesticated and wild species of Cucurbita (Cucurbitaceae) inferred from a Archaeology and Paleoecology, eds Pearsall DM, Piperno DR (MASCA, University mitochondrial gene: Implications for crop evolution and areas of origin. Proc Natl Museum of Archaeology and Anthropology, Univ. of Pennsylvania, Philadelphia), pp. Acad Sci USA 99:535–540. 131–145. 2. Reichert ET (1913) The Differentiation and Specificity of Starches in Relation to 14. Pearsall DM, Chandler-Ezell K, Chandler-Ezell A (2003) Identifying maize in Neotropical Genera, Species, etc. (Carnegie Institution of Washington, Washington, DC). sediments and soils using cob phytoliths. J Archaeol Sci 30:611–627. 3. Seidemann J (1966) Sta¨ rke-Atlas (Paul Parey, Berlin). 15. Piperno DR, et al. (2007) Late Pleistocene and Holocene environmental history of the 4. Piperno DR, Ranere AJ, Holst I, Hansell P (2000) Starch grains reveal early root crop Iguala Valley, Central Balsas watershed of Mexico. Proc Natl Acad Sci USA 104:11874– horticulture in the Panamanian tropical forest. Nature 407:894–897. 11881. 5. Pearsall DM, Chandler-Ezell K, Zeidler JA (2004) Maize in ancient Ecuador: Results of 16. Iriarte J (2003) Assessing the feasibility of identifying maize through the analysis of residue analysis of stone tools from the Real Alto site. J Archaeol Sci 31:423–442. cross-shape size and three-dimensional morphology of phytoliths in the grasslands of 6. Zarillo S, Kooyman B (2006) Evidence for berry and maize processing on the Canadian southeastern South America. J Archaeol Sci 29:1085–1904. plains from starch grain analysis. Am Antiq 71:473–499. 17. Dorweiler JE, Stec A, Kermicle J, Doebley J (1993) Teosinte glume architecture 1: A 7. Holst I, Moreno JE, Piperno DR (2007) Identification of teosinte, maize, and Tripsacum major locus controlling a key step in maize evolution. Science 262:233–235. in Mesoamerica by using pollen, starch grains, and phytoliths. Proc Natl Acad Sci USA 18. Dorweiler JE, Doebley J (1997) Developmental analysis of teosinte glume architecture 104:17608–17613. 1: A key locus in the evolution of maize (). Am J Bot 84:1313–1322. 8. Zarillo S, Pearsall DM, Raymond JS, Tisdale MA, Quon DJ (2008) Directly dated starch 19. Piperno DR, Pearsall DM (1998) The silica bodies of tropical American grasses: Mor- residues document Early Formative maize (Zea mays L.) in tropical Ecuador. Proc Natl Acad Sci USA 105:5005–5011. phology, , and implications for grass systematics and fossil phytolith identi- 9. Austin DR (2006) Fox-tail millets (Setaria:Poaceae)–abandoned food in two hemi- fication. Smith Cont Bot 85: 1–40. spheres. Econ Bot 60:143–158. 20. Piperno DR, Holst I, Wessel-Beaver L, Andres TC (2002) Evidence for the control of 10. Piperno DR (2006) Phytoliths: A Comprehensive Guide for Archaeologists and Paleo- phytolith formation in Cucurbita fruits by the hard rind (Hr) genetic locus: Archaeo- ecologists (AltaMira, Lanham, MD). logical and ecological implications. Proc Natl Acad Sci USA 99:10923–10928. 11. Piperno DR, Pearsall DM (1993) Phytoliths in the reproductive structures of maize and 21. Piperno DR, Weiss E, Holst I, Nadel D (2004) Processing of wild cereal grains in the Upper Teosinte: Implications for the study of maize evolution. J Archaeol Sci 20:337–362. Paleolithic revealed by starch grain analysis. Nature 430:670–673. 12. Bozarth SR (1993) Maize (Zea mays) cob phytoliths from a central Kansas Great Bend 22. Henry A, Piperno DR (2008) Using plant microfossils from dental calculus to recover Aspect archaeological site. Plains Anthropol 38:279–286. human diet: A case study from Tell al-Raqí, Syria. J Archaeol Sci 35:1943–1950.

Piperno et al. www.pnas.org/cgi/content/short/0812525106 2of10 Fig. S1. Starch grains from modern kernels of Z. mays ssp. parviglumis (A) and maize races Pepetilla (B) and Harinoso de Ocho (C). Maize starch is larger and mostly irregularly shaped, with well-defined compression facets and various types of fissures. In contrast, grains in ssp. parviglumis are predominantly oval to round, with fewer compression facets and other surface features typical of maize.

Piperno et al. www.pnas.org/cgi/content/short/0812525106 3of10 Fig. S2. Rondel phytoliths diagnostic of maize cobs. (A and B) Ruffle-top rondels from sediments directly associated with grinding stone 318d and from the needle probe analysis of a used facet of 318d, respectively. (C and D) Wavy-top rondels from the 60–65 cm b.s. level of the unit 1 column sample.

Piperno et al. www.pnas.org/cgi/content/short/0812525106 4of10 Fig. S3. Rondels typical of maize cobs from the needle probe analysis of a used facet of grinding stone 318d.

Piperno et al. www.pnas.org/cgi/content/short/0812525106 5of10 Fig. S4. Cucurbita phytoliths from sediments associated with grinding stone 318 (A), from a modern hybrid of C. sororia and C. argyrosperma that is heterozygous at the Hr locus (Hr hr)(B), and from the 65–75 cm b.s. level from the unit 1 column sample (C and D). The phytoliths exhibit surface features (surface cavities and marks, faint scalloped impressions) typical of fruits that have undergone human manipulation at the Hr locus.

Piperno et al. www.pnas.org/cgi/content/short/0812525106 6of10 Fig. S5. Phytoliths from C. argyrosperma ssp. sororia. Unlike in domesticated fruits and hybrids with domesticated germ plasm that are heterozygous at the Hr locus, these phytoliths exhibit no signs of incomplete silicification, such as surface cavities and faint scalloped impressions.

Piperno et al. www.pnas.org/cgi/content/short/0812525106 7of10 Table S1. Starch grain size in panicoid grasses common in the Mexican flora Length, ␮m Range, ␮mSD n

Setaria geniculata (NH1501561) 5.4 2–9 1.7 50 (NH2551512) 5.4 3–8 1.7 50 Heteropogon melanocarpus 10.0 2–18 4.1 50 (NH 2378339) granularis (NH 2461435) 9.1 4–19 3.4 50 (NH 734955) 4.3 2–19 3.3 28 Schizachyrium condensatus 4.4 3–7 1.2 25 (NH1386960) Aristida schiediana 4.8 3–10 1.6 25 (NH2118693)

Our starch research focused primarily on the , the subfamily to which maize belongs. Various studies have shown that other subfamilies of grasses are characterized by different types of starch grains that would not be confused with maize; for example, the Pooideae have simple round lenticular grains, sometimes with lamellae, or compound (and thus more highly angled) grains (2, 21, 22), and the Chloridoideae have mostly compound grains, whereas maize has simple grains (2).

Piperno et al. www.pnas.org/cgi/content/short/0812525106 8of10 Table S2. Starch grain characteristics in a representative sample of maize and teosinte Teosinte Maize race

Balsas Chalco Central Plateau Bolita Reventador Naltel Tabloncillo Pepetilla

Shape Round 44 (32–55) 33 (16–58) 11 (2–20) 5 2 14 3 15 Oval 15 (11–19) 6 (2–10) 12 (12–13) 0 0 0 0 0 Bell 10 (2–16) 3 (0–5) 7 (5–8) 0 3 0 0 1 Irregular 30 (27–33) 58 (40–73) 70 (61–78) 94 95 86 97 83

Hilum Cavity 24 (15–33) 12 (6–18) 9 (9–10) 12 1 7 5 3

Compression facets Slight 58 (54–63) 47 (35–64) 44 (42–45) 49 5 50 17 66 Defined 29 (21–35) 54 (36–63) 56 (55–58) 52 95 50 83 32

Fissures Transverse 16 (12–19) 14 (12–18) 20 (11–30) 15 41 11 22 15 Total with fissure 38 (37–40) 47 (41–54) 37 (25–48) 25 55 31 31 32

Range of mean size, ␮m 8.8–9.5 5–10.7 6.5–9 11.1 15.3 11.4 12.6 12.9 Range of individual grain sizes, ␮m 4–18 2–22 2–20 6–20 6–24 6–18 6–20 6–24

Numbers are percentages of the different types of grains and surface features. Numbers in parentheses are the ranges for percentages of each attribute found in different populations studied. The range of mean size in teosinte represents the mean size in the different populations studied. (Data from ref. 7.)

Piperno et al. www.pnas.org/cgi/content/short/0812525106 9of10 Table S3. Number and distribution of starch grains recovered from each step of the stone tool analyses Tool catalogue number Sediment Needle probe First wash Ultrasound Total number

Ground stone tools 310a unit 1, layer B, 30–35 cm b.s. NA NNP 3 3 6 312a unit 1, layer B, 40–45 cm b.s. NA NNP 1 0 1 314c unit 1, layer C, 50–55 cm b.s. 1 NNP 7 4 11 315c unit 1, layer C, 57 cm b.s. 0 7 11 0 18 316c unit1, layer D, 60–65 cm b.s. 0 NNP 2 0 2 316d unit 1, layer D, 60–67 cm b.s. 0 22 41 13 76 318d unit 1, layer E, 77 cm b.s. 0 2 15 5 22 318e unit 1, layer E, 70–75 cm b.s. 0 12 43 25 80 319d unit 1, layer E, 78 cm b.s. 0 3 0 5 8 322c unit 1, layer E, 85–90 cm b.s. 0 3 7 1 11 361a unit 2, layer B, 10–20 cm b.s. NA NNP 3 0 3 362a unit 2, layer B, 20–30 cm b.s. NA NNP 6 2 8 364a unit 2, layer B, 45 cm b.s. 0 NNP 1 2 3 365 unit 2, layer B, 45–50 cm b.s. NA NNP 0 2 2 365a unit 2, layer C, 49 cm b.s. 0 8 9 8 25 365c unit 2, layer C, 51 cm b.s. NA NNP 0 0 0 365b unit 2, layer C, 54 cm b.s. 0 NNP 5 0 5 366a unit 2, layer C, 57 cm b.s. 3 NNP 0 2 2 367 unit 2, layer C, 55–60 cm b.s. NA NNP 0 0 0 367a unit 2, layer D, 63 cm b.s. 0 4 3 2 9 368a unit 2, layer D, 63 cm b.s. 1 NNP 3 0 3 Chipped stone tools 308a unit 1, layer B, 20–25 cm b.s. NA NNP 0 1 1 310a unit 1, layer B, 30–35 cm b.s. NA NNP 0 0 0 362 unit 2, layer B, 20–30 cm b.s. NA NNP 0 0 0 370 unit 2, layer E, 70–75 cm b.s. NA NNP 0 8 8 322a unit 1, layer E, 85–90 cm b.s. NA NNP 0 4 4

NA, sediment was not analyzed; NNP, no needle probes were carried out. On tool 316d, the distribution of different types of grains was as follows: 68 from maize, 4 from Dioscorea, 3 from the Fabaceae, and 1 from the Marantaceae. On tool 365a, one unknown grain occurred. On tool 365a, 24 of the starch grains were from maize. Sediment is that sampled from immediately below and around the tools.

Piperno et al. www.pnas.org/cgi/content/short/0812525106 10 of 10