Phytoliths Analysis for the Discrimination of Foxtail Millet (Setaria italica) and Common Millet (Panicum miliaceum)

Foxtail millet (Setaria italica) and Common millet (Panicum miliaceum) are the oldest domesticated dry farming crops in Eurasia. Identifying these two millets in the archaeobotanical remains are still problematic, especially because the millet grains preserve only when charred. Phytoliths analysis provides a viable method for identifying this important crop. However, to date, the identification of millet phytoliths has been questionable, because very little study has been done on their morphometry and taxonomy. Particularly, no clear diagnostic feature has been used to distinguish between Foxtail millet and Common millet. Here we examined the anatomy and silicon structure patterns in the glumes, lemmas, and paleas from the inflorescence bracts in 27 modern plants of Foxtail millet, Common millet, and closely related grasses, using light microscopy with phase-contrast and microscopic interferometer. Our research shows that five key diagnostic characteristics in phytolith morphology can be used to distinguish Foxtail millet from Common millet based on the presence of cross-shaped type, regularly arranged papillae, Ω-undulated type, endings structures of epidermal long cell, and surface ridgy line sculpture in the former species. We have established identification criteria that, when used together, give the only reliable way of distinguishing between Foxtail millet and Common millet species based on their phytoliths characteristics, thus making a methodological contribution to phytolith research. Our findings also have important implications in the fields of plant taxonomy, agricultural archaeology, and the culture history of ancient civilizations.


Introduction
Phytoliths are microscopic silica bodies that precipitate in or between cells of living plant tissues, and are widespread in all types of plants and all their different organs and structures, from roots to culms to inflorescences [1][2][3][4]. They are especially abundant, diverse, and distinctive in the grass family (Poaceae). Phytoliths are released from plant tissues when they are decayed, burned, or digested. Released phytoliths thus become microfossils of the plants that produce them. The development and application of phytolith techniques in archaeology have led to major advances in investigating plant use and subsistence patterns in regions where preservation of macrobotanical remains is poor [2,3,[5][6][7][8][9][10][11][12][13].
Foxtail millet (Setaria italica) and Common millet (also known as Broomcorn millet, Panicum miliaceum), belong to Panicoideae of Poaceae, are considered to be dry farming cereals that form the oldest cultivated foods known to humans [14][15][16]. They were staple foods in the Far East (China, Japan, Russia, and India) and even in the entire Eurasian continent prior to the popularity of rice and wheat, and are still important foods in the semi-arid regions today [14,17,18]. However, the archaeobotanical remains of Foxtail millet and Common millet are difficult to distinguish mainly due to the very small sizes of these grain crops -often less than 2,3 mm in length, and their very similar shapes [8,19,20]. A previous study has also considered at some length how the charred de-husked grains of various native millet species might have been systematically misidentified [21]. Moreover, the identification of millet phytoliths has also been questionable [8,20,22], because very little study has been conducted on millet phytoliths, thus no clear diagnostic feature has been used to distinguish Foxtail millet from Common millet [8].
The inflorescence bracts in Poaceae are characterized by phytoliths with specific morphological characteristics, hence their taxonomic significance [23]. Wynn Parry and Smithson (1966) used light microscopy to record silicification of the various epidermal cell types and cuticle of the inflorescence bracts from various British grass genera [24]. Subsequently, a series of such studies has emerged utilizing the newer techniques of scanning electron microscopy (SEM) focusing on cereals and grasses, including descriptions for barley, oats, rice, rye and species of Panicum, Echinochloa and Dicanthelium [23,25,26]. Pearsall et al. (1995) and Zhao et al. (1998) used morphometric analysis to distinguish between rice and wild Oryza phytoliths [6,12]. Tubb et al (1993) and Ball et al. (1999) developed paradigms using morphometric analysis for distinguishing between inflorescence phytoliths produced by several species of wheat and barley [27,28]. Berlin et al. (2003) used these morphometric paradigms to identify Triticum aestivum in residues taken from ceramics at Tel Kedesh, Israel [29]. This paper reports the first attempt to determine if phytoliths analysis of the inflorescences bracts can be an effective tool for discriminating between Foxtail millet (S. italica) and Common millet (P. miliaceum).

Materials and Methods
We examined modern phytoliths from twenty-seven species of domesticated and wild Paniceae. Domesticated species include nine species of S. italica L. Beauv. and twelve species of P. miliaceum L.; wild species include two species of P. bisulcatum Thunb., S. plicata (Lam.) T. Cooke, S. glauca (Linn.) Beauv., S. viridis (L.) Beauv., and Echinochloa crusgalli (L.) Beauv. Of these 27, four species were sampled from annotated folders at the National Crop Gene Bank of China, Chinese Academy of Agricultural Sciences (CAAS); six species were sampled from the Culture Museum of Cishan, Wuan, Hebei Province, China; fourteen species were sampled from Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing, China; and three species were sampled from East China Normal University, Shanghai, China. The folders contained samples of field collections by many investigators. For passport data on the plants, see Table 1.
In this study, we dissected the spikelet of modern plants into five parts, including lower glume, upper glume, lower lemma (lemma of sterile floret), upper lemma, and palea [30] (see Figure 1) for phytolith analysis. Palea can be divided into ''palea of first floret'' and ''palea of second floret''. However, in both genera Setaria and Panicum, the palea of first floret is atrophied to a very small and membranous organ and sometimes becomes lost in the spikelet. Thus, in this study, we used ''palea'' for the ''palea of second floret''. The five parts of spikelet were prepared in the following manner.
(i) Each part of spikelet was cleaned with distilled water in a water bath to remove adhering particles. (ii) All samples were placed in 20 ml of saturated nitric acid for over 12 h to oxidize organic materials completely. (iii) The solutions were centrifuged at 2000 r.p.m. for 10 min, decanted and rinsed twice with distilled water, and then rinsed with 95% ethanol until the supernatants were clear. (iv) The phytolith sediments were transferred to storage vials. The residual subsamples were mounted onto microscopic slides in Canada Balsam medium for photomicrography and in liquid medium for counting, measuring, and line drawing. (v) Light photomicrography (phase-contrast, and microscopic interferometer) at 4006 magnification was used to determine their anatomy and silicon structure patterns in the glumes, lemmas, and paleas. (vi) Phytolith parameters were measured using computer-assisted image analysis.

Phytolith morphology of the lower lemma and glumes
Micromorphological characters of the lower lemma and glumes are generally similar for each millet sample based on our observation and statistic of all samples. Silicification always occurs in the short cells (silica cells), and occasionally occur in some of the long cells, micro-hairs, macro-hairs, and stomata in the lower lemma and glumes of both Foxtail millet and Common millet.
The shape of the silica bodies formed in the short cells of Foxtail millet is different from Common millet ( Figure 2). Cross-shaped (ratio length: width<1:1) phytoliths are found in Foxtail millet, and tend to increase in size variation (range 4.46-9.98 mm; average 7.5561.17 mm, N = 208) toward the central part of the lower lemma and glumes. However, the Common millet has bilobe-shaped (dumbbell-shaped, ratio length: width<1:2) phytoliths with the two endings distinctly branched. The bilobe-shaped phytoliths are oriented with the bar cross at a right angle to the long cells ( Figure 2D), and tend to increase in size (length) variation (range 8.08-15.05 mm; average 10.8761.43 mm, N = 198) toward the central part of the lower lemma and glumes.
Other silicified cells, including long cells, micro-hairs, macrohairs, and stomata, are without any characteristic shape, and not easily identified in phytoliths. This suggests that the cross-shaped type is formed in the lower lemma and glumes of S. italica, and the bilobe-shaped type is formed in those of P. miliaceum.

Phytolith morphology of the upper lemma and palea
Phytolith morphology of the upper lemma. The presence and form of papillae visible under microscopy on the upper lemmas are important characteristics for identifying S. italica ( Figure 3). Upper lemmas of S. italica have distinct papillae by the silicification of the surface, cell wall, and/or lumen of epidermal papillae cells. The bases of papillae are typically suborbicular with semicircular to sinuous to irregular margins. They typically have a single papillate and tend to decrease in size variation (papillae diameter ranges between 5 mm and 30 mm) from center toward the base and top part of the upper lemma ( Figure 3A), but may also be scutiform or dome-shaped, and lacking a clear projection weak papillae. No papillae area is formed on the surfaces of the upper lemma of some S. italica. (Figure 4).
The upper lemma of P. miliaceum is characterized by a smooth surface without any papillae ( Figure 3B) in every area of all samples. Therefore, the papillae formed on surfaces of the upper lemma are peculiar to S. italica.
Phytolith morphology of the palea. Regularly arranged papillae are consistently found in center surfaces of the palea of S. italica, and tend to decrease in size variation (papillae diameter ranges between 5 mm and 25 mm) toward the base and top of the palea ( Figure 5A). P. miliaceum does not have any papillae in the entire area of palea in all samples ( Figure 5B). The phytolith morphology of S. italica and P. miliaceum can be clearly distinguishable based on the presence or absence of papillae. Regularly arranged papillae on the surface of the upper lemma and palea are peculiar to S. italica. However, it should be cautioned that the identification of P. miliaceum cannot be confirmed based solely on the absence of papillae, because papillaes may sometimes vanish into a smooth surface on the surface of upper lemma and palea in S. italica.
The undulated patterns of epidermal long cell in the upper lemma and palea. By means of light microscopy with phase-contrast and microscopic interferometer the surface undulated patterns of epidermal long cell walls in the upper lemma and palea from both S. italica and P. miliaceum can be divided into two distinctly different types by means of particularity analysis ( Figure 6). The epidermal long cell walls are V-undulated (undulations rounded, wider toward the apex and narrower at the base) in S. italica, and are g-undulated in P. miliaceum. V-undulated types can produce branching subordinate V subtypes based on the degree of undulations as V I, II, III. Similarly, g-undulated types can also produce branching subordinate g-undulated subtypes, including gI, II, III ( Figure 6).
It is noteworthy that there are also >-undulated types ( Figure 6), which only occur at the narrow margin part of the lemmas and palea. It is very difficulty or impossible to distinguish them, because their very simple and similar morphology occurs in both Foxtail millet and Common millet.
The undulations tend to increase in highly sinuous variation toward the central part of the lemmas and palea, where the undulations of the long cell walls produce branching subordinate V (V II, III) or g (gII, III) sinuous margins that join the margins across the cells. The different V-undulated regular patterns occur at different parts by gradual change in a general way from base and top (V I), to side (VII), to center (V III) of the lemmas and palea in Foxtail millet (Figure 7). Similarly, the different gundulated patterns also occur at different parts of the lemmas and palea of Common millet, by gradual change from base and top (g I), to side (gII), to center (g III) (Figure 8).
This suggests that the surface undulated patterns of epidermal long cell walls in the upper lemma and palea can be used to distinguish between Foxtail millet and Common millet. The epidermal long cell walls are V-undulated (V-I, II, III) in S. italica, and g-undulated (g-I, II, III) in P. miliaceum.  Table 2). Cross finger type, dendriform epidermal long cell endings joining others in a deeply digital pattern, is formed in the upper lemma and palea of P. miliaceum. However, the average width of endings interdigitation of dendriform epidermal long cells is longer (8.9562.02 mm, N = 3303) in the Cross finger type of P. miliaceum than that in S. italica. (Figure 10). Figure 10 shows the bivariate biplot by s-coordinates of the 3303 measurements from epidermal long cells of P. miliaceum and 2774 measurements from those of S. italica, plotted along axis W and axis R, and their classification into two groups corresponding to the two species (P. miliaceum and S. italica). The R-value is higher (0.7960.12, N = 3303) in P. miliaceum than in S. italica (0.3360.11, N = 2774) ( Figure 10) ( Table 2).
The surface sculpture of epidermal long cells in the upper lemma. Diverse silicon deposits can occur at different cell layers, including extracellular sheet (keratose layer), outer epidermis, hypoderm fibres, vascular bundle, and occasional silicification of internal spongy mesophyll in the transection of lemma and palea [23].
Surface ridgy line sculpture of the upper lemma is important for the identification of S. italica, which is characterized by having an adnate silicon extracellular sheet and outer epidermis, forming a very heavy silicon layer ( Figure 11). P. miliaceum have a smooth spotted sculpture with adnate silicon extracellular sheet and outer epidermis, or a surface saw-toothed sculpture with adnate silicon outer epidermis and hypoderm fibres. This is a reliable feature in distinguishing them from S. italica (Figure 12).
Based on our observation of surface characteristic with different adnate silicon layers in different V-types or g-types, we found that the surface ridgy line sculpture of the upper lemma is peculiar to S. italica ( Figure 13).

Preliminary contrast of phytolith morphology between millets and related grasses
The phylogenetic relationship of Eurasian Panicum species is currently unknown, and the wild ancestor of P. miliaceum, if it still exists, has not been conclusively identified. Panicum bisulcatum   Thunb., a species of wild grass in China potentially related to P. miliaceum, can be distinguished from P. miliaceum based on its phytolith characteristics, because it typically has simple obvious silica skeleton (gI type) ( Figure 14A, B, C) that is distinct from the well-defined gII-III type in P. miliaceum.
The wild ancestor of Foxtail millet (S. italica) is presumed to have originated from S. viridis (green foxtail), a ubiquitous weed from the Eurasian continent [17]. We examined the silicon structure patterns in the glumes, lemmas, and paleas from the inflorescence bracts in modern Foxtail millet, and closely related grasses, including S. viridis, S. plicata (Lam.) T. Cooke. Figure 14 shows that foldaway 'undulated pattern occurs in S. viridis ( Figure 14E) and the multiped worm sculpture pattern occurs in S. plicata ( Figure 14F). A speciesspecific identification of phytoliths is possible for S. italica ( Figure 14D) and P. miliaceum because they have typically welldefined silica skeletons that are distinguishable from those in P. bisulcatum, S. viridis, and S. plicata, which have no such demonstrable patterns, additional studies are needed to confirm the observations.

Discussion
Early investigators have reported the potential of using grain shapes (expressed as the length-to-breadth ratio, and morphological variations) for discriminating between P. miliaceum and S. italica [19], and between wild and domesticated S. italica [30][31][32]. However, the grain shape analysis is ineffective in discriminating between Foxtail millet and Common millet, because their grains are very small in size compared to wheat or barley, and their chaff is more delicate and similar to each other. Moreover, the overlapping ranges of the length-to-breadth ratios between S. italica and P. miliaceum make the identification of at least charred de-husked grains of the domesticated species difficult [8,20,21]. Other studies also reported on the variation of bilobes/crosses phytolith within leaves from Panicum sp. and Setaria sp. [33][34][35], however, the morphological characteristics of bilobes/crosses are not sufficient to distinguish S. italica from P. miliaceum.
The inflorescence bracts are well known sites of heavy silicification in cereals, due to the hot and arid habitat of the cereals, conditions that promote intensive transpiration and water loss and lead to the close formation of phytoliths in inflorescence bract cell [2]. Most of the silicon in the inflorescence bracts has been concentrated in the outer (abaxial) epidermis, but the precise location of the heaviest deposits varies with the species. As mentioned above, early investigators have recognized the potential of using inflorescence phytoliths for discriminating between wheat and barley [10,28], wild and domesticated Oryza [6,12], Avena sativa and A. strigosa [36], and many other grasses [23,[37][38][39].
Foxtail millet and Common millet are vitally important food crops for people living in the Far East and even in the entire Eurasian continent prior to the popularity of rice and wheat [15,40,41]. However, the previous studies show that there is no valid method for separating Foxtail millet and Common millet based on inflorescence phytolith analysis [8,20,22]. These are several reasons for rendering the previous identifications questionable. One of the main obstacles in inflorescence phytolith systematics is that perplexing variations in morphology occur from tissue to tissue in spikelet, and from apex to base within lemma or  glume, and in specific tissues. Another reason is that diverse silicon deposits can occur at different cell layers in the transection of lemma and palea. The complication surface sculptures of upper lemmas with adnate different silicon layers have not been discussed in detail. As a result of very little work conducted on the phytoliths, no clear diagnostic feature has been found and used to distinguish between Foxtail millet and Common millet.
Because of this, we had to dissect the spikelet of modern plants into five parts, including lower glume, upper glume, lower lemma (lemma of sterile floret), upper lemma, and palea, to examine the variation of anatomy and silicon structure patterns in different parts, from base to center to apex, and margins, cover with whole glumes, lemmas, and paleas. Our observation and statistics of the micromorphology throughout each glume, lemma, and palea reveal well-regulated variations in phytoliths between Foxtail millet and common millet, particularly regarding the presence or absence of papillae, undulated patterns of epidermal long cell, and surface ridgy line sculpture. Based on a large sample, the average width of endings interdigitation of dendriform epidermal long cells examined in a large number is found to be consistently and significantly different between the two species examined and therefore can be considered a diagnostic feature to distinguish between S. italica and P. miliaceum. This character is as significant in the taxonomy of each genus as other quantitatively important characters.
By all accounts, our research indicates that five key diagnostic characteristics in phytolith morphology could be used to distinguish Foxtail millet from Common millet (Table 3): (i) Cross-shaped type is formed in the lower lemma and glumes of S. italica, whereas Bilobe-shaped type is formed in those of P. miliaceum. (ii) Regularly arranged papillae on the surface of the upper lemma and palea are peculiar to S. italica. (iii) The epidermal long cell walls are V-undulated (V-I, II, III) in S. italica, and gundulated (g-I, II, III) in P. miliaceum. (iv) Cross wavy type     Table 1; y = 1, 2, 3 correspond to gI, gII, and gIII types (see Figure. 6), respectively. ** Sx-z; S = Setaria italica (L.) Beauv.; Sx (x = 1-9) corresponding to No. in Table 1; z = 1, 2, 3 correspond to VI, VII, and VIII types (see Figure.    identification of S. italica, which is characterized by having an adnate silicon extracellular sheet and outer epidermis, forming a very heavy silicon layer that is a reliable feature in distinguishing them from P. miliaceum. These five diagnostic characteristics used together give the only reliable way of distinguishing Foxtail millet from Common millet. A species-specific identification of phytoliths is possible for S. italica and P. miliaceum because they have typically well-defined silica skeletons that are distinguishable from those in P. bisulcatum, S. viridis, and S. plicata, which have no such demonstrable patterns.
The results of this study have revealed distinct differences between Foxtail millet and Common millet. Nevertheless, several caveats exist in the morphological characteristics of phytoliths, which should be mentioned. A number of factors may influence the within-individual variations in phytolith morphology, including the stage of plant maturity [42], intraspecific variation within the plant taxa [1,2,13,43], the amount of soluble silica in local ground water [33], the rate of tissue transpiration [44], the tissue within which the phytoliths form [43,45], the location of phytoliths in leaf blades [43,46], genetic variation among plants, and geographic location where the plants grew [43,47]. Additional phytolith studies, particularly those that concentrate on variation within a single species in different seasons and regions, will help provide more efficient and more accurate methods.
It is not the focus of this paper to discuss in detail how phytoliths can be used to distinguish millets from other related grasses. Although the phytolith production patterns revealed in our preliminary research give encouraging results that may point the way to distinguishing the millets from related grasses in China, more research is needed, especially the study of more wild species and landraces of domesticated millet species. To be of practical use to investigators, further morphometric analysis of a wide variety of millet species is required. Future work could use this methodology to develop classification paradigms, and to gain an understanding of the effect of domestication and polyploidization on phytolith morphometries.

Conclusions
In discussions of the origin of early agriculture, Foxtail millet and Common millet have received particular attention, since they were the dominant traditional crops in the Far East and even in the entire Eurasia. However, until now, the identification and taxonomic distinction between Foxtail millet and Common millet in archaeobotanical remains had been problematic, especially because the crop grains preserve only when charred, or where their preservation is poor. This is the first study of the variation of the anatomy and silicon structure patterns in the glumes, lemmas, and paleas occurring among 27 modern millets and related grass species collected from different regions in China. We found that five key diagnostic characteristics in phytolith morphology could be used to distinguish between Foxtail millet and Common millet, as follows. (i) Crossshaped and Bilobe-shaped are formed in S. italica and P. miliaceum, respectively; (ii) Papillae on the upper lemma and pales are peculiar to S. italica; (iii) The epidermal long cell walls are V-undulated in S. italica, and g-undulated in P. miliaceum; (iv) The endings structures of epidermal long cells are Cross wavy type in S. italica, and Cross finger type in P. miliaceum; (v) Surface ridgy line sculpture of the upper lemma are peculiar to S. italica. Collectively, these five diagnostic characteristics provide the only reliable way of distinguishing Foxtail millet from Common millet. A species-specific identification of phytolith is possible for S. italica and P. miliaceum because they have typically well-defined silica skeletons that are distinguishable from those in P. bisulcatum, S. viridis, and S. plicata, which have no such demonstrable patterns. This practical protocols, if supported by additional studies of phytoliths derived from more millets and related grass species, can be give the only reliable way of separating the Common millet, Foxtail millet, and other related grass species based on their phytoliths.