Gradual domestication of root traits in the earliest maize from Tehuacán.

Efforts to understand the phenotypic transition that gave rise to maize from teosinte have mainly focused on the analysis of aerial organs, with little insights into possible domestication traits affecting the root system. Archeological excavations in San Marcos cave (Tehuacán, Mexico) yielded two well-preserved 5,300 to 4,970 calibrated y B.P. specimens (SM3 and SM11) corresponding to root stalks composed of at least five nodes with multiple nodal roots and, in case, a complete embryonic root system. To characterize in detail their architecture and anatomy, we used laser ablation tomography to reconstruct a three-dimensional segment of their nodal roots and a scutellar node, revealing exquisite preservation of the inner tissue and cell organization and providing reliable morphometric parameters for cellular characteristics of the stele and cortex. Whereas SM3 showed multiple cortical sclerenchyma typical of extant maize, the scutellar node of the SM11 embryonic root system completely lacked seminal roots, an attribute found in extant teosinte and in two specific maize mutants: root with undetectable meristem1 (rum1) and rootless concerning crown and seminal roots (rtcs). Ancient DNA sequences of SM10—a third San Marcos specimen of equivalent age to SM3 and SM11—revealed the presence of mutations in the transcribed sequence of both genes, offering the possibility for some of these mutations to be involved in the lack of seminal roots of the ancient specimens. Our results indicate that the root system of the earliest maize from Tehuacán resembled teosinte in traits important for maize drought adaptation.

Genetic evidence indicates that maize (Zea mays ssp. mays) populations arose from Balsas teosinte (Zea mays ssp. parviglumis, also named teosinte parviglumis) close to 9,000 y ago (1). This evolutionary transition caused important phenotypic changes in the aerial portion of the plant, including the partial suppression of lateral branching, a decrease in the number of male and female inflorescences per individual, the exposure of the kernel by absence of a cupulate fruitcase, and the transformation of a distichous female inflorescence that disarticulates naturally into a polystichous (3- to 12-ranked) cob with attached grains that require human intervention for dispersal (1–3). A close association has been established between some of these traits and the genes that underlie their developmental control (4, 5), or genomic regions that have lost genetic diversity as a consequence of progressive domestication (6–8). In some cases, paleogenomic analysis of millenary specimens dating to the earliest stages of Mesoamerican cultivation has allowed the establishment of reference time frame for the progression of their genetic diversity and stages of domestication (9–11).

By contrast, and despite their importance for supplying water and nutrients during all stages of growth and development, the influence of domestication on the evolution of root architecture and anatomy has received little attention. A phenotypic analysis and comparison of maize landraces and teosintes concluded that their range of root architectural and anatomical traits was similar, however, a few specific traits permitted some distinction between both subspecies (12). In general, teosintes showed less variation for architectural traits such as root system dry weight, longest nodal root length, nodal system diameter, number of root tips, and number of seminal roots. They also showed smaller mean stele and xylem areas, shorter nodal roots, less frequent lateral root branching, and significantly fewer seminal roots than landraces (12), suggesting they could be important traits affected during domestication. Comparisons of physiological responses to limited nitrogen availability indicates that teosinte parviglumis shows an increase of the shoot:root biomass ratio as compared to maize, as well as an increase in the length of nodal and lateral roots, but also reduced nodal root number (13). A functional decrease in major domestication genes such as Teosinte Branched1 (Tb1) results in an increase of both lateral and nodal roots, although it remains unclear if the effect is direct or indirect (14).

Root architecture is crucial for productivity by determining the temporal and spatial distribution of soil exploration and hence resource capture. Maize root architecture is comprised of embryonic and postembryonic components (15). After seed germination, the emergence of the radicle and the coleoptile is followed by the elongation of the mesocotyl. While the primary root develops from the radicle, the scutellar node gives rise to seminal roots located in a protuberance formed by the remnants of the pericarp and endosperm, located between the mesocotyl and the primary root. Seminal and primary roots are considered components of the embryonic root system. The first node forms the first nodal roots between the mesocotyl and the coleoptile. Subsequent elongation of the main vertical axis of the mesocotyl results in additional subterranean and root nodes (). Aerial nodes will give rise to whorls of brace roots.

The anatomy of a root transversal section is characterized by the presence of two concentric cellular cylinders: the stele and the cortex. In maize, the central region of the differentiated stele contains xylem vessels responsible for axial transport of water and nutrients. At the periphery of each late metaxylem bundle are the smaller early metaxylem bundles. Phloem vessels, necessary for photosynthate transport, are composed of smaller cells located between late metaxylem bundles. The intersection of the cortex and the stele is composed of two concentric cell files: the pericycle and the endodermis. The cortex is composed of the root epidermis and 6 to 19 files of outer, mid, and inner cortical cells (16, 17). Cortical aerenchyma can be formed via programmed cell death. In some cases, the outer cortex exhibits multiseriate cortical sclerenchyma (MCS) with thick lignified walls, a phenotype recently reported to improve root penetration ability as an adaptation to growth in hard soils (18). Interestingly, the MCS phenotype is present in some modern maize inbreds but not in accessions of teosinte parviglumis and Z. mays ssp. mexicana [teosinte mexicana; (18)], suggesting it might represent an adaptation acquired during domestication.

Two maize genes have been shown to be important for development of seminal roots during the establishment of the embryonic root system. Mutations in ROOT WITH UNDETECTABLE MERISTEM1 (RUM1) result in the absence of seminal and postembryonic lateral roots on the primary root (19–21). RUM1 encodes a monocot specific AUX/IAA protein that can be induced by auxin. Similarly, mutants defective in ROOTLESS CONCERNING CROWN AND SEMINAL ROOTS (RTCS) completely lack seminal roots and the postembryonic shoot-borne root system (21). RTCS encodes a Lateral Organ Boundaries (LOB) domain protein preferentially expressed in roots. Two major quantitative loci contributing to 66% of seminal root number variation comapped with RUM1 and RTCS, suggesting both genes play key regulatory functions in the development of the embryonic root system (22).

Pioneering excavations conducted in rock shelters of the Tehuacán Valley uncovered maize paleobotanical specimens dating back to the earliest stages of agriculture in Mesoamerica (23, 24), including hundreds of cob specimens, but only a few root crowns. Subsequent explorations of San Marcos cave yielded new nonmanipulated specimens dating to a similar age of 5,300 to 4,970 calibrated y B.P., including SM3, a well-preserved root crown that represents the earliest maize root specimen found to date (10). A paleogenomic analysis of SM3 and other specimens of equivalent age showed that the earliest maize from San Marcos genetically diverged from fully domesticated landraces and contained allelic variants absent from extant maize populations (10). Some domestication loci (teosinte branched1, brittle endosperm2) showed reduced nucleotide variability as compared to teosinte parviglumis, but others (teosinte glume architecture1, sugary1) showed conserved levels of nucleotide variability that are absent from extant maize. These temporally similar samples also showed unexpected levels of homozygosity and inbreeding, opening the possibility for Tehuacán maize cultivation evolving from reduced founder populations (10).

To characterize in detail the architecture and anatomy of the earliest maize roots found to date, we conducted laser ablation tomography (LAT) of two paleobotanic specimens (SM3 and SM11) from San Marcos cave, dating at a similar age of 5,280 to 4,956 y B.P. We generated a three-dimensional (3D) reconstruction of a second node root segment for both specimens, confirming the exquisite preservation of their inner cellular organization, and comparing multiple anatomical parameters to extant maize and teosinte accessions. SM3 exhibited MCS proposed to be exclusive to domesticated maize. By contrast, the 3D reconstruction of the scutellar node of SM11 demonstrated the absence of seminal roots, a trait only reported in extant teosintes and two specific maize mutants. Partial sequencing of RUM1 and RTCS alleles present in the genome of SM10—a San Marcos specimen of equivalent age to SM3 and SM11—revealed mutations that could relate to the absence of seminal roots. Our overall results indicate that some of the most important root traits that distinguish extant maize landraces from teosinte were not fully present in the earliest maize from San Marcos.

Results and Discussion

Age and Root Architecture of SM3 and SM11.

Although morphometric analysis of paleobotanical specimens found in archaeological expeditions conducted in the 1960s provided important information to understand the temporal transitions that shaped domestication, high resolution capture of anatomical and architectural parameters has not been explored in ancient maize remains. We concentrated on two well-preserved specimens that were dated by accelerator mass spectrometry (AMS). SM3 is the most ancient specimen, dating to 4,220 to 4,180 14C y B.P. (5,280 to 4,970 2σ calibrated y B.P. at 95% confidence) (10). SM11 dated to 4,470 to 4,410 y B.P. (5280-4,880 2σ calibrated y B.P. at 95% confidence; ). SM3 and SM11 are presented in Figs. 1 and 2, respectively. SM3 is a root crown containing at least five nodes with multiple nodal roots but lacking a primary root system and scutellar node (Fig. 1). SM11 is a larger specimen with at least six nodes and a complete embryonic root system, notably including the scutellar node and primary root (Fig. 2) (23, 24). The presence of at least five nodes in both specimens suggests that the corresponding plant individuals had reached at least the V7-V8 stage (25). Lateral roots were not preserved in either specimen. A general architectural comparison suggested that SM11 had been more drastically affected by burial compression than SM3.

Fig. 1.

Ancient maize root specimens from San Marcos cave. (A) Maize root stalk corresponding to specimen SM3 and dating 5,280 to 4,970 y B.P. (B) Transversal section (10 µm thick) of a nodal root belonging to specimen SM3. (C) Maize root stalk corresponding to specimen SM11 and dating 5,280 to 4,880 y B.P. (D) Transversal section of a second node root from SM3 used for lignin content determination. (E) Red staining of the outer cortical cells by phloroglucinol-HCl, indicating presence of lignin, a distinctive component of the MCS phenotype.  Abbreviations: EP, epidermis; C, cortex; EN, endodermis; PE, pericycle; P, phloem; LMX, late metaxylem; EMX, early metaxylem. (Scale bars: A, 1 cm; B, 500 µm; C and D, 500 µm.)

Fig. 2.

LAT of the SM11 scutellar node and primary root. (A) Root architecture of SM11 depicting the presence of consecutive nodes. (B) Embryonic root system of SM11. (C) Scutellar node of SM11 showing the plane of transversal sections depicted in micrographs D–G. (D) Transversal section of the SM11 scutellar node top region. (E) Transversal section of the mid portion of the SM11 scutellar node. (F) Transversal section of the bottom portion of the SM11 scutellar node. (G) Transversal section of the primary root adjacent to the bottom part of the SM11 scutellar node. (Scale bars: 500 µm.)

We measured root architecture and anatomy parameters in SM3 and SM11, comparing their value to those previously reported for 195 accessions of maize landraces from the Americas and Caribbean islands, 36 accessions of teosinte parviglumis, 16 accessions of teosinte mexicana, and a small group of 9 additional accessions that included Z. mays ssp. huehuetenangensis, Zea luxurians, Zea nicaragüensis, and Zea perennis (12). The results of these comparisons are summarized and illustrated in Table 1 and Fig. 3, and . With the exception of the individual ancient specimens, as reported in ref. 12, all parameters were measured in three individuals per accession selected at the V6 to V7 stage, which precedes the V7 to V8 stage that had been reached by SM3 and SM11. In the case of extant landraces and teosintes, measurements were taken from 30- to 50-µm-thick cross-section segments collected 5 to 9 cm from the base of a second whorl nodal root in three different individuals per accession (12). In the case of SM3 and SM11, measurements were taken from three random 10-µm-thick cross-section LAT images of a 1-cm segment collected 5 to 9 cm from the base of a second whorl nodal root. Although the diameter of maize nodal roots can significantly vary between samples corresponding to different nodes, their characteristics tend to remain constant within the same node (12). Maize nodal roots do not undergo secondary growth, i.e., their diameter remains constant during development, from emergence to full maturity, allowing a valid comparison between previously reported parameters for extant Zea accessions and parameters measured at equivalent root segments and nodes from ancient specimens.

Table 1.

Root architectural and anatomical trait values in extant teosintes, extant maize landraces, and 5,300 to 4,970 y B.P. maize from San Marcos cave

Description	Landraces (n = 195), mean ± SD	SM11	SM3	Teosinte parviglumis (n = 36), mean ± SD	Teosinte mexicana (n = 16), mean ± SD	Other teosintes* (n = 9), mean ± SD
Architectural traits
No. of nodal roots	20.60 ± 4.2	21	25	23.2 ± 6.54	25.9 ± 6.10	23.4 ± 9.58
No. of seminal roots	3.90 ± 1.3	0	ND	0.47 ± 0.48	0.49 ± 0.45	0.44 ± 0.69
Stem diameter, mm	22.80 ± 4.5	14.3	11	15.5 ± 3.97	18.8 ± 3.1	12.4 ± 4.65
Nodal root system diameter, mm	60.90 ± 14.5	32.3	23.5	43.8 ± 11.8	49.4 ± 11.5	43 ± 23.2
Angle		45.67	60.08
Anatomical traits
Cross-section area, mm2	0.966 ± 0.258	ND	0.95 ± 0.0051	0.997 ± 0.27	0.933 ± 0.27	0.900 ± 0.352
Total stele area, mm2	0.258 ± 0.079	0.19 ± 0.015	0.22 ± 0.0018	0.252 ± 0.076	0.235 ± 0.062	0.195 ± 0.084
Xylem vessel area, mm2	0.053 ± 0.019	0.041 ± 0.0008	0.03 ± 0.0004	0.040 ± 0.009	0.041 ± 0.008	0.028 ± 0.011
Aerenchyma area, mm2	0.051 ± 0.047	ND	0	0.052 ± 0.044	0.054 ± 0.045	0.076 ± 0.111
Percent of cortex as aerenchyma, %	6.49 ± 5.25	ND	0	6.46 ± 5.02	6.28 ± 4.05	7.55 ± 8.44
Total cortical area, mm2	0.708 ± 0.187	ND	0.74 ± 0.0068	0.745 ± 0.199	0.698 ± 0.211	0.705 ± 0.27
Cortical cell traits
No. of cortical cells	596 ± 141	ND	1,272 ± 151	602 ± 144	582 ± 101	595 ± 157
No. of cortical cells files	9.96 ± 1.09	ND	13.67 ± 0.47	10.1 ± 1.0	9.89 ± 1.25	10.1 ± 1.25
Inner cortical cell size, µm2	ND	ND	530.32 ± 72.78	ND	ND	ND
Middle cortical cell size, µm2	ND	ND	1,045.45 ± 91.12	ND	ND	ND
Outer cortical cell size, µm2	ND	ND	94.05 ± 30.12	ND	ND	ND

ND: not determined.

*Z. perennis (5); Z. luxurians (1); Z. huehuetenangensis (1); Z. nicaragüensis (1); F1 B73 X Z. diploperrenis hybrid (1).

Fig. 3.

Comparison of root architectural and anatomical parameters in extant teosintes, extant maize landraces, and 5,300 to 4,970 y B.P. maize from San Marcos cave. (A) Architectural parameters. (B) Anatomical parameters. Letters indicate significant (a-b) or nonsignificant (a-a) differences between extant teosinte and maize distributions following both one-way ANOVA and Tukey’s honest significant test. SM3 values are represented in green and SM11 values in brown.

Although stem diameter (11 mm for SM3 and 14.3 mm for SM11; as compared to 15.5 ± 3.97 mm for teosinte parviglumis, 18.8 ± 3.1 mm for teosinte mexicana, and 22.8 ± 4.2 mm for mean value all maize landrace accessions) and nodal root system diameter (23.5 mm for SM3 and 32.3 mm for SM11; as compared to 43.8 ± 11.8 mm for teosinte parviglumis, 18.8 ± 3.1 mm for teosinte mexicana, and 60.9 ± 14.5 mm for maize landrace accessions) were likely affected by compression in both ancient specimens, we obtained a reliable assessment of the total number of nodal roots included in the first four nodes, making our values comparable to mean estimations previously reported for extant teosintes and maize landraces (12). By taking into consideration all root segments and scars, SM3 and SM11 contained 25 and 21 nodal roots in the first four nodes, respectively, a range within mean values previously reported for all teosinte (23.2 ± 6.5 for teosinte parviglumis and 25.9 ± 6.1 for teosinte mexicana) and maize landrace accessions (20.6 ± 4.2). Although both one-way ANOVA and Tukey’s statistical tests suggested significant differences between teosinte and maize accessions in the total number of nodal roots (Fig. 3), the SM3 and SM11 values did not permit a clear distinction of teosinte versus maize for this trait in any of the two specimens. The embryonic root system was not present in SM3, whereas the scutellar node of SM11 did not show any remnants of seminal roots. Previously reported results showed significant differences in the average number of seminal roots between teosinte (0.47 ± 0.48 for teosinte parviglumis, 0.49 ± 0.45 for teosinte mexicana, 0.44 ± 0.69 for the small group of distinct teosinte subspecies and species) and maize accessions (3.9 ± 1.3) (12), suggesting that this architectural trait in SM11 is reminiscent of teosinte and not extant maize landraces (Fig. 3).

Root Anatomy of SM3 and SM11.

LAT allows detailed quantitative and qualitative morphometric analysis of root anatomy and 3D reconstruction of selected root portions, providing new opportunities to assess and compare traits related to domestication. LAT consists of a pulsed ultraviolet (UV) laser that oscillates in an ablation plane as a sample is moved into with a mechanical stage, permitting a camera focused on the ablation plane to capture images resulting from UV fluorescence emission of the ablated tissue (26). To determine the degree of inner preservation of both specimens, we conducted LAT of a 7- mm segment of a second node root of both SM3 and SM11, capturing a cross-sectional image every 10 µm (for a total of 700 images), and allowing comparison with current standard measurements conducted in extant teosinte and maize accessions. In both samples the inner cell contour and organization remained intact when soil compression did not affect root structure (Fig. 1 and ). In SM3, the central cylinder, stele, and cortex were clearly distinguishable, showing detailed preservation of xylem, phloem, pericycle, and endodermal cells, as well as several files of cortical cells and a well-defined epidermis (Fig. 1). The total transversal cross-sectional area of the analyzed SM3 nodal root segment (RXSA) was 0.95 mm2 (Table 1), a value that is similar to the mean RXSA previously reported for teosinte and maize (0.997 ± 0.27 mm2 for teosinte parviglumis, 0.933 ± 0.27 mm2 for teosinte mexicana, and 0.966 ± 0.280 mm2 for maize landraces). By contrast, the cortex of SM11 was partially collapsed (), but the stele remained intact, allowing a reliable measurement of xylem and stele area, as described below.

Cortical cell number, files, and size of SM3.

Cortical phenotypes were only observed in SM3, since SM11 exhibited extensive collapse of its cortex. The total cortical area (TCA) of a transversal section of a SM3 second node root was 0.74 ± 0.006 mm2, a value slightly higher than the mean TCA previously reported for teosinte and maize (0.727 ± 0.211 mm2 for all teosintes included in the analysis; 0.745 ± 0.199 for teosinte parviglumis; 0.698 ± 0.211 for teosinte mexicana; and 0.708 ± 0.187 for maize landraces; Table 1). Strikingly, it contained an average of 1,272 ± 151 cortical cells (#CC), a number significantly greater than mean values for teosinte and maize reported to date (595.7 ± 134.5 for all teosintes included in the analysis; 602 ± 144 for teosinte parviglumis; 582 ± 101 for teosinte mexicana; 596 ± 141 for maize landraces). The mean number of cortical cell files (#CF) of SM3 was also significantly greater than those previously reported for extant accessions: 13.67 ± 0.47, as compared to 10.1± for teosinte parviglumis, 9.89 ± 1.25 for teosinte mexicana, and 9.96 ± 1.09 for maize landraces. Cortical cells of SM3 were heterogeneous in size. The average inner cortical cell size (CCS) and middle CCS cortical cell size was 530.32 ± 72.78 and 1045.45± 91.12 µm2, respectively, whereas the average outer CCS was 94.04 ± 30.12 µm2. In general, CCS in maize is substantially variable across the root cortex, with cells presenting the largest cross-sectional area located in the center and reducing size toward the periphery. Previous studies reported 6 to 19 cortical cell files in maize in second node roots (16), and maximum values of cross-sectional area of cortical cells comprised between 514.6 and 533.9 µm2 (27), a range close to 50% smaller than values obtained for mean middle CCS size in SM3. A CCS within the range of those shown by SM3 has not previously been reported in extant maize.

Stele and xylem area of SM3 and SM11.

Whereas extant teosintes and maize landraces exhibit equivalent values of RXSA (Table 1) on the basis of previously reported results (12), total stele area (TSA) tend to be greater in extant landraces as compared to teosinte accessions (0.258 ± 0.079 mm2 for landraces vs. 0.252 ± 0.076 mm2 for teosinte parviglumis and 0.235 ± 0.062 mm2 for teosinte mexicana). The same is true for mean xylem vessel area (XVA; 0.053 ± 0.019 mm2 for landraces vs. 0.04 ± 0.009 mm2 for teosinte parviglumis and 0.041 ± 0.008 mm2 for teosinte mexicana) We measured both parameters in SM3 and SM11. TSA values were 0.22 ± 0.0018 mm2 and 0.19 ± 0.015 mm2, respectively, whereas XVA values were 0.03 ± 0.0004 mm2 and 0.041 ± 0.0008 mm2. Although one-way ANOVA and Tukey’s test suggested significant differences in XVA between teosinte and maize populations (Table 1), the SM3 and SM11 values tend to be smaller than the average value for maize, however, the values did not allow a clear discrimination between the two groups.

Presence of MCS in SM3.

The exquisite preservation of the epidermis and cortex of the SM3 second node crown root segment allowed the identification of outer cortical cell files with substantial reduction in cell size as compared with the rest of the cortex, a phenotype indicative of MCS (Fig. 1 and ). These cells had an average area of 94.04 µm2, whereas middle and inner cortical cells had an average area of 530.32 µm2 and 1045.45 µm2, respectively. Maize lines with MCS have a wall-to-lumen ratio ranging from 2.1 to 4.4, whereas the same ratio ranges from 0.4 to 2.2 in non-MCS maize lines (18). The average wall-to-lumen ratio of the outer SM3 sclerenchyma was 4.1 ± 0.11 µm2, a high value within the MCS range. The presence of outer cortical cells with enlarged cell walls and small cell lumen was fully confirmed by cryoscanning electron microscopy of a second node crown root segment containing MCS cell walls with a thickness comprised between 3.5 and 4 µm (). Histological staining of SM3 root segments with phloroglucinol-HCl demonstrated that thickening of cell walls in MCS is primarily due to the presence of lignin (Fig. 1 ). The nonrandom specific deposition of lignin in the outer cortical cells corresponds to the previously reported MCS phenotype found in extant maize but not in teosinte parviglumis or mexicana accessions (18).

Linear Discriminant Analysis for SM3 and SM11.

To determine if root architecture and anatomy could be used as phenotypes to separate teosintes from extant maize landraces, we trained a linear discriminant analysis (LDA) model with a set of parameters that included number of nodal roots (#Nod), diameter of the root system (SysDia), diameter of the stem (StemDia), RXSA, TCA, TSA, percentage of cortex aerenchyma (%A), cortical cell area, XVA, number of cortical cells (#CC), and number of cortical cell files (#CF) (model 1), using previously reported values for 41 teosintes (parviglumis and mexicana), 172 maize landraces, and two inbred lines (12, 28, 29, 30). The resulting model was independently tested with 20 teosintes, 20 maize landraces, and SM3 as an outgroup. The results are presented in the and Fig. S5. In the case of teosintes and landraces, Model 1 correctly predicted the corresponding group in 90% of the cases (18 out of 20 teosintes; 18 out of 20 landraces; ). On the basis of this level of confidence, SM3 was predicted to belong to teosinte class with a probability close to 1. A second LDA model included #Nod, number of seminal roots (#Sem), SysDia, TSA, and XVA (model 2, ), and was trained with values from the same collection of germplasm as model 1. Model 2 correctly predicted the corresponding group in 95% of the cases (19 out of 20 teosintes; 19 out of 20 landraces). On the basis of this second level of confidence, SM11 was predicted to belong to teosinte class with a probability of 0.99. These results suggest that root phenotypes in both maize specimens tend to resemble teosinte rather than maize landraces.

Three-Dimensional Reconstruction of the SM11 Scutellar Node.

In SM11, high-resolution imaging of the scutellar node surface suggested that seminal roots did not develop as part of the embryonic root system. To confirm the absence of seminal roots in SM11, we generated and analyzed a 3D reconstruction of the complete scutellar node. As shown in the , transversal sections of the scutellar node in teosinte parviglumis and extant maize landraces show that seminal roots emerge from inner cell files, at the boundaries of central cylinder. Seminal roots emerge from primordia extending from the central core across dozens of cell files before reaching the node surface. In SM11, root primordia or root extensions were completely absent from the scutellar node (Fig. 2 ). The only root extension identified in the 3D reconstruction was a lateral root emerging from the primary root, outside the bottom part of the scutellar node (Fig. 2). This evidence indicates that SM11 corresponds to a plant that did not develop seminal roots, a phenotypic feature that is specific to extant teosinte and not maize.

Molecular Analysis of RUM1 and RTCS.

In addition to SM3 and SM11, excavations at San Marcos cave uncovered specimen SM10, a cob previously dated to 4,240 ± 30 14C y B.P. (5,300 to 5,040 2σ calibrated y B.P. at 95% confidence). Whereas SM3 and SM11 had a poor representation of reads mapping to unique genomic regions, analysis of the SM10 aDNA yielded 1.26 Gb coverage of the nonrepetitive maize genome (10). To indirectly test the possibility that RUM1 (GRMZM2G037368) or RTCS (GRMZM2G092542) could be involved in the lack of seminal roots characterizing the SM11 embryonic root system, we analyzed ancient DNA sequences from SM10. We aligned SM10 reads to the RUM1 and RTCS transcribed sequence in the B73 genome and 26 recently reported de novo assembled genomes (ref. 19; and Figs. S7–S10). Among more than 19.7 million reads mapping to unique genomic regions, we recovered 52 corresponding to the transcribed sequence of RUM1, and 14 corresponding to the transcribed sequence of RTCS, with an average length of 88.46 (RUM1) and 79.78 (RTCS) nucleotides (nt) per read, respectively (). These sequences resulted in a segmental coverage of 62.47% of the RUM1 and 43.64% of the RTCS transcribed region. Although the average coverage depth (1.8×) was not sufficient to ensure a reliable determination of the SM10 diploid genotype, several sequences included variants with respect to the B73 reference sequence. We identified 42 single nucleotide variants (SNVs) included in the ancient sequence of RUM1 (), with six located within exons and seven within untranslated regions (UTRs). We also identified a deletion of eleven nucleotides located 33 nt upstream of the 3′UTR end. This deletion was also found in the RUM1 sequence of four additional genomes (Zm-Oh43, Zm-M37W, Zm-Ky21, and Zm-Ki3; ), suggesting its presence prevailed in some extant lineages. In the case of RTCS, we identified 13 SNVs; 3 are located in exons and 10 in the UTRs (). Interestingly, a SNV located in the second exon generates a STOP codon that was not found in any of the 25 extant genomes analyzed (), suggesting possible detrimental effects subsequently purified by negative selection. None of the teosinte and maize diversity datasets analyzed (HapMap3, MaizeSNP50 BeadChip, and maize GBS2.7; (27)) reports SNPs at the RUM1 deletion or RTCS STOP codon site. Although our limited coverage is unable to confirm the homozygous nature of the mutations described above, their identification opens the possibility for these genes to be involved in the lack of seminal roots shown by SM11.

Although many phenotypic traits related to the architecture and anatomy of the root system are not distinguishable between teosintes and maize, large-scale phenotypic analysis of extant accessions suggests that XVA and the #Sem are significantly different between extant maize landraces and their teosinte ancestors (12). The value of these parameters is remarkably conserved among teosinte accessions, even within different taxa. The presence of MCS also distinguishes teosintes from maize (18). XVA values for both ancient specimens did not allow to establish a tendency to resemble extant teosinte or maize accessions. By contrast, the presence of MCS in the outer cortex of SM3, and the absence of seminal roots developing in the scutellar node of SM11, suggest that some of the maize root phenotypes found in extant accessions were present in 5,300 to 4,790 y B.P. maize, but others were not. Although MCS is absent from seven previously analyzed teosinte parviglumis and three teosinte mexicana accessions, a larger sampling is required to confirm its eventual condition of a maize specific root trait. Extant maize lines with MCS are better adapted to mechanical impedance by increasing root depth by 22% and producing 39% more biomass in soils compacted by vehicle traffic (18). Whereas the large size of SM3 cortical cells could represent an ancestral adaptation causing a reduction of metabolic costs (28), a combination of small XVA (31), large mCCS (32), and the presence of MCS (18), could contribute to establish an integrated phenotype adapted to drought. An equivalent integrated phenotype is associated with superior drought tolerance in extant maize (31). Taking aside mutations in RUM1 and RTCS, the absence of seminal roots has been rarely reported in extant maize accessions. Seminal roots are beneficial for phosphorus capture in maize (33). A recent in silico study suggests that although seminal roots are indeed beneficial for capturing both phosphorus and nitrogen, teosinte cannot form seminal roots due to limited seed carbohydrate reserves (34). In addition to teosintes, a reduced number of seminal roots has also been reported in wild wheat and barley, suggesting a possible adaptation to water stress (35, 36), or a possible consequence of a reduced seed endosperm that restricts seminal root formation in these wild taxa (34). Earlier studies indicated that, under nonlimiting soil phosphate availability, the probability for extant maize individuals to lack seminal roots is close to null; under low soil phosphate conditions, the probability is close to 0.02 (33). Under these assumptions, the absence of seminal roots in SM11 is unlikely due to nutritional adaptation, but rather caused by seed phenotypic traits—and their genetic control—that brought the embryonic root system of the earliest Tehuacán maize to closely resemble extant teosintes. Our overall results suggest that selection affected some maize root traits late during domestication, with human selection progressing at different temporal rates in the aerial and subterranean organs. While X-ray microscopy 3D technologies promise to refine the nondestructive internal analysis of paleobotanical remains (37), our study demonstrates the value of LAT for the phenotypic analysis of paleobotanical root specimens, opening new possibilities for the identification of domestication traits selected during the transition from teosinte to maize.

Materials and Methods

Archaeobotanical Specimens and Dating.

SM3 and SM10 are specimens discovered during the 2012 Langebio Cinvestav–Instituto Nacional de Antropología e Historia (INAH) expedition to San Marcos cave, as previously reported (10). SM11 is a maize root stalk found during MacNeish expeditions (1961 to 1962) and curated by the INAH. Using the service provided by Beta Analytic, 10 to 20 mg of each specimen was dated with AMS.

Estimation of Root Architecture Parameters.

Specimens were observed, analyzed, and photographed using a Keyence VHX high resolution digital microscope with a 20 and 50× magnification lens. General specimen images were captured using a Nikon D3300. A ruler in centimeters was included in each image as a scale reference. Stem diameter was measured at the most basal node of the brace roots, while the diameter of the root system was measured at the widest section of the root network. The number of nodal roots per node was counted manually. Root angle was estimated by measuring the distance from the most basal part of the stem to the maximum diameter of the root network. These two parameters were subsequently used to determine the main vertical axis of the specimen and the external point to the periphery of the root network. The root angle is the angle comprised between the main vertical axis with the axis emerging from the intersection of the main vertical axis and the widest point intersecting with the external root network. Image analysis was performed using RSAJ (38) and ObjectJ (39) plugins.

LAT and Estimation of Root Anatomy Parameters.

To precisely adjust sampling procedures to those described in ref. 12, an ∼1-cm-long segment collected 5 to 9 cm from the base of a second node root was sectioned from SM3 and SM11 specimens. LAT was used to obtain images of transversal root sections. LAT consists of a pulsed UV laser (s-Pulse HP, 343 nm THG, Amplitude Systems) that oscillates in the focal plane of a camera (α7R III digital camera, Sony. MP-65 mm F/2.8 Lens photo macro 1–5, Canon). The transversal root segment is perpendicularly placed in front of the laser beam that vaporizes the sample while the camera captures the cross-sectional image. For each sample, 7-mm-wide transversal segments were captured, capturing an image every 10 µm, for a total of 700 images. The 3D animations were created by stacking all 700 images with Avizo 9 lite software (VSG Inc). Root anatomy parameters were measured by selecting three random images for each sample using the RootScan 2.3v software (38, 41) and calculating the average values for each parameter presented in Table 1. The same procedure was used for estimating cortical cell area from inner, middle, and outer cortical regions with imageJ (42). To phenotype MCS, the ratio of cell wall to lumen area in the outer cortex was determined using MIPAR software (43). In specimen SM11, the scutellar node and a fraction of the primary root and the mesocotyl were also analyzed by LAT.

Histology.

For determining the cellular localization of lignin, 400-µm-thick transversal sections of a second node SM3 root were generated using LAT, and directly stained with a 3% (wt/vol) phloroglucinol solution in ethanol for 10 min. After rinsing with deionized water, sections were mounted in conventional slides and observed under a Nikon SMZ1500 microscope using bright field illumination.

Cryoscanning Electron Microscopy.

A second node root segment of specimen SM3 was directly mounted, embedded in liquid nitrogen, and transferred to a cryopreservation chamber to withdraw the holder under vacuum. The sample was subsequently transferred to a SEM chamber and analyzed with a Zeiss Sigma VPFESEM microscope at temperature and voltage of 195 °C and 10 kV, respectively.

Statistical Analysis.

Mean and SD values for architectural and anatomical root parameters in extant maize landraces and teosintes were obtained from data previously reported in ref. 12 and corresponding to 195 maize landrace accessions from a wide diversity of countries in North, Central and South America, as well as the Caribbean islands, and 61 teosinte accessions that included 36 of Zea mays ssp. parviglumis, 16 of Zea mays ssp. mexicana, 1 of Zea mays ssp. huehuetenangensis, 5 of Z. perennis, 1 of Z. luxurians, 1 of Z. nicaragüensis, and 1 hybrid (B73 inbred line per Z. mays ssp. diploperennis). For extant landraces and teosintes, values correspond to overall measurements of a 30- to 50-µm-thick cross-section segment collected 5 to 9 cm from the base of a second whorl nodal root in three different individuals per accession. In the case of SM3 and SM11 ancient specimens, values correspond to measurements in three random 10-µm-thick cross-section images of a 1-cm segment collected 5 to 9 cm from the base of a second whorl nodal root. One-way ANOVA and Tukey’s honest significant tests were conducted to compare extant teosinte to extant maize distributions using R v. 4.0.0 (43). Graphs were created with the ggplot2 package (43, 44). An LDA was conducted using MASS (44). Since available parameters for SM3 were not identical than parameters available for SM11, an independent model was developed for each specimen. In the case of SM3, the LDA model was trained with stem diameter, root system diameter, number of nodal roots, nodal root transversal area, total cortical area, total stele area, percentage of aerenchyma, cortical cell area, xylem vessel area, number of cortical cells, and number of cortical cell files. In the case of SM11, the LDA model was trained with stem diameter, number of seminal roots, number of nodal roots, root system diameter, TSA, and XVA. In both cases, parameter values to train the model were obtained from ref. 12 by including 172 maize landraces, 41 teosintes, and 2 inbred lines (12, 29, 30).

Ancient DNA Analysis.

All SM10 genomic reads corresponding to nonrepetitive regions are reported in ref. 10. Filtered quality reads corresponding to the transcribed sequence of RUM1 (RMZM2G037368) and RTCS (GRMZM2G092542) were aligned to the B73 reference genome (MaizeGDB 3.0) using Burrows-Wheeler Aligner (45). Translational alignment visualization was generated with Tablet (46). The RUM1 eleven nucleotide deletion of specimen SM10 initiates in coordinate Chr3:209176295; the RTCS SNV causing a STOP codon is located in coordinate Chr1:10825088. Both were searched as SNPs in datasets HapMap3, MaizeSNP50 BeadChip, and maize GBS2.7 (27), but neither was included in those analyses.