1. Introduction
Amaranths belong to the dicotyledonous genus
Amaranthus L. which is made up of over 70 species [
1], and three subgenera [
2]. The word Amaranthus originated from the Greek word amarantos meaning “one that does not wither” or “never fading” [
3]. About 60 Amaranthus species are native to America while the rest originated from Asia, Africa, Australia and Europe [
4].
The genus
Amaranthus contains both cultivated and wild species. Among the cultivated species, grain amaranths have been grown for more than 8,000 years dating back to before the Pre-Colombian civilization of Central and South America [
5]. The cultivated grain amaranths include
A. caudatus L.,
A. cruentus L. and
A. hypochondriacus L. and their parental wild species are thought to be
A. hybridus L.,
A. quitensis Willd. ex Spreng. and
A. powellii S. Wats. [
6]. Grain Amaranths are important subsistence and commercial food crops for people living in parts of Central and South America [
4,
7]. They are expanding in many regions of Asia as well as Eastern and Southern Africa (
www.amaranthinstitute.org). Other amaranth species like
A. dubious L.,
A. hybridus and
A. tricolor L. are consumed as leafy vegetables [
8]. Meanwhile,
A. retroflexus L. (redroot pigweed),
A. albus L. (tumbleweed),
A. palmeri S. Wats. (Palmer amaranth),
A. spinosus L. (spiny amaranth) represent weed species [
9]. While many of the latter are cosmopolitan in nature; the vegetable amaranths are commonly found in Asia and Africa while grain amaranths are native to Mexico and Peru with recent expansion around the world [
10]. Genetic races have been suggested for grain amaranths with Azteca, Mercado, Mixteca, Nepal and Picos in
A. hypochondriacus; Mexican, Guatemalan and African in
A. cruentus; and finally South American and Edulis in
A. caudatus [
11]. The first of these two species can hybridize to each other as can all the grain amaranth with their immediate wild relatives; however, in general most cultivars tend to be self-pollinating despite being monoecious [
2,
10].
In terms of nutritional content, grain amaranths produce seed with high protein content (17–19% of dry weight) and well-balanced amino acid profiles [
12]. The seeds of grain Amaranths possess double the amount of the essential amino acids (especially lysine, phenylalanine and threonine) and high minerals (calcium, iron and zinc) compared to wheat protein [
13]. As easy to cook grains, the amaranths show promise for amelioration of protein or amino acid deficiencies, supplementing mineral content (Fe, Zn) of foods and providing protein to predominantly or completely vegetarian diets [
14,
15]. Grain amaranths are commonly popped or roasted before milling or mixing with other ingredients; therefore, several flours can be made from this pseudocereal and provide novel organoleptic properties and new tastes and flavors. Chemical composition analysis of grain amaranths confirms their high potential for human nutraceutical uses [
16]. Amaranth seed and amaranth seed oil is high in Vitamin E and squalene, which can be beneficial for people suffering from hypertension or cardiovascular disease [
3,
14]. Regular consumption of grain amaranth can reduce blood pressure, cholesterol levels and improves antioxidant status and immunological parameters [
17]. With increasing demand for food and current malnutrition levels, development of amaranths as an alternative food could be an important boon for people of developing countries suffering from malnutrition and hunger [
10]. In summary, grain amaranth is a healthy and nutritious food crop that could benefit people if it was produced and consumed in greater quantities.
The objective of this research was to assess the morphological diversity of close to 300 cultivated grain amaranths and their wild relatives from two gene banks through field assessments of leaf, flower and grain characteristics. Another goal was to determine if morphological traits could be used for species and population identification. The two gene banks providing germplasm from this study were the United States Department of Agriculture (USDA) through the National Plant Germplasm System (NPGS) with a smaller collection provided by Seed Savers Exchange (SSE). The uncharacterized SSE collection was compared to species in the USDA collection; however, morphological analysis of whole plant traits such as leaf and petiole color blade shape or terminal inflorescent index and branching index did not distinguish species.
3. Results
3.1. Morphological Variability in the SSE Collection
The seed scanning of the 33 genotypes from the SSE collection showed 11 white seeded accessions, 8 cream, 12 black and 2 red brown in color (
Supplementary Figure S1). Once germinated, the majority of seedling and growing transplants showed green leaves, a few with marginal or vein pigmentation, some with red leaves and a very few were with dark green leaves. In all cases we evaluated the lamina instead of the petiole and amaranthine is used inter-changeably with red as the pigmentation type. Stem color were solid red, solid green, orange, amaranthine striped or pink based with green stem. Leaf shape included oval, oblong, elliptical and ovate. Petiole pigmentation varied from dark amaranthine, light amaranthine, green to yellow among different genotypes.
At flowering, plant architecture varied and included branched all along the stem, only few branches at top or without any branches. Flower color ranged from dark amaranthine to green, yellow, orange or mixed. The inflorescence could be at the terminal part of plants but many had long or short side branch. Inflorescence density were found to be high, intermediate and low among different accessions. Many of them had erect inflorescence and some of them had drooping type or arched shape inflorescence. One genotype, SSE39, did not flower and was not included in the rest of the study, making a total of 32 genotypes for dendogram construction.
Significant differences in blade pigmentation (BP), blade shape (BS) and flower color (FC) were found across different clusters of SSE germplasm (
Table 1). Other morphological characteristics like Petiole pigmentation (PP), Branching index (BI), Inflorescence shape (IS), Inflorescence density (ID), Terminal inflorescence attitude (TAI) and stem color (SC) were not significantly different across different clusters.
Based on the variation among different morphological characteristics, six distinct clusters were seen for the SSE germplasm (
Figure 1). The first cluster (group I) included SSE 1, SSE3, SSE34, SSE112, SSE115 and SSE117. The second cluster (II) included SSE40 and SSE108. The third cluster (III) included SSE4, SSE5, SSE6, SSE24, SSE29, SSE30 and SSE35. The fourth cluster (IV) included SSE7, SSE10, SSE15, SSE22, SSE31, SSE38, SSE42, SSE86, SSE92, SSE93, SSE99, SSE104, and SSE132. The fifth cluster (V) included SSE8, SSE79 and SSE119. Cluster six (VI) included SSE80 alone.
The distances between different clusters were calculated (
Table 2). Chi square test showed the Mahalanobis
D2 distance of 14.06 as significant distance between two clusters. We did not find any significant distance between any two clusters.
3.2. Morphological Variability in the USDA Collection
Among 260 accessions obtained from USDA we found flowering and non-flowering genotypes. In total only 208 accessions producing noticeable inflorescence and the 52 accessions not producing noticeable inflorescence were not considered further or used for cluster analysis. Among the genotypes included, we found blade pigmentation from dark green to green to amaranthine (red) lamina with marginal or vein pigmentation. Different stem colors included solid red, pink base and green stem, green, orange or amaranthine striped. Most genotypes had oval leaves but some had oblong, elliptical and ovate leaves. Different types of petiole pigmentation were observed like amaranthine, dark amaranthine, green or yellow. At full flowering stage, the observed flower color ranged from dark amaranthine flower color, to pink, orange or mixed flower color. Both erect and drooping inflorescence shapes were seen among different genotypes.
Analysis of the morphological traits among 208 accessions of USDA resulted in 10 clusters, five of which were major and five of which were minor (
Supplemental Table S4). The minor clusters contained few accessions. Overall, a variable number of genotypes were found in each cluster (
Figure 2). Major Cluster 1 (V on
Figure 2) contained 29 accessions; 8 from
A. caudatus, 8 from
A. cruentus, 4 from
A. hybridus, 5 from
A. hypochondriacus and 4 from
A. quitensis. This cluster represented 12 accessions from South America, 4 accessions from North America, 2 accessions from Africa, 4 accessions from Asia and 7 accessions from Central America. They were predominantly cultivated except for the
A. hybridus and
A. quitensis genotypes which are direct ancestors of the cultivated species.
The other clusters similarly had a mix of cultivated and wild accessions. Major Cluster 2 (II in
Figure 2) included 85 accessions representing 1 from
A. caudatus, 61 from
A. cruentus, 4 from
A. hybridus, 14 from
A. hypochondriacus and 5 from
A. quitensis; and being predominantly of cultivated germplasm. Based on geographical origin, this cluster represented 13 accessions from Africa, 6 from Asia, 41 from Central America, 2 from Europe, 11 from North America, 11 from South America and 1 with unknown origin. This cluster had the highest number of accessions. Major Cluster 3 (III) had 31 accessions with 3 accessions from
A. caudatus, 11 accessions from
A. cruentus, 8 accessions from
A. hybridus and 9 accessions from
A. hypochondriacus, all of which were cultivated. The accessions of this cluster represented 3 accessions from Africa, 8 from Asia, 7 from Central America, 2 from Europe, 5 from North America and 6 from South America. Major Cluster 4 (IV) contained 49 accessions; among which 31 belonged to
A. cruentus, 4 belonged to
A. hybridus, 8 to
A. hypochondriacus, 1 to
A. palmeri, 2 to
A. powellii, 2 to
A. quitensis and 1 to
A. retroflexus, showing this cluster to be based on weedy and cultivated accessions. Based on geographical origin, this cluster consisted of 5 accessions from Africa, 5 from Asia, 27 from Central America, 2 from Europe, 5 from North America and 5 from South America. Major Cluster 5 (or I) was found to have 8 genotypes including 3 cultivated accessions of
A. hypochondriacus and 5 wild accessions (1
A. palmeri, 3
A. powellii, and 1
A. quitensis). Based on geographical origin, the cluster had 1 accession from Central America, 2 from Asia, 3 from Europe, 1 from North and 1 from South America.
Among the minor groupings, Cluster 6 (VI) comprised only 2 accession from A. cruentus and A. hypochondriacus which originated in Asia and Africa. Cluster 10 (X) also consisted of 2 accessions this time A. hypochondriacus which originated in Asia and Central America. Cluster 8 (VIII) represented one accession of A. cruentus from Europe. Cluster 9 (IX) showed only 1 accession of A. cruentus which originated in South America. Cluster 7 (VII) represented only 1 accession of A. hybridus which originated in North America.
The significance of different morphological traits among ten clusters of USDA accessions is shown in
Table 3. The main morphological traits to vary between clusters were blade pigmentation, blade shape, petiole pigmentation, flower color and flower shape had significant contribution on differentiating clusters. However, inflorescence shape, inflorescence density and terminal inflorescence attitude had no significant contributions on distinguishing different clusters. The Mahalanobis
D2 distances between different USDA clusters are shown in
Table 4; however, no significance distances were found. Relative distance measures showed that distance between Cluster 3 and Cluster 9 was highest followed by distance between Cluster 2 and Cluster 10. The minor clusters 7 and 8 were also distant from the major clusters 1, 2 and 3.
3.3. Cluster Analysis of Full Collection from Both USDA and SSE
The morphological data analysis of all the accessions representing USDA and SSE collections together showed seven distinct clusters. Clustering was done primarily to determine if the SSE genotypes clustered apart from the USDA ones or not. If they clustered together, we were interested in seeing if there were species or morphotype associations of the SSE genotypes with a subset of the USDA collections. In general we found morphological traits to be shared among species and phenotypic analysis to constitute cross species clusters without the ability to identify unknown accessions.
Cluster 1 was comprised of 39 accessions representing 11 accessions of A. caudatus, 6 accessions of A. cruentus, 5 accessions of A. hybridus, 4 accessions of A. hypochondriacus, 4 accessions of A. quitensis and 9 accessions of SSE. Based on geographical origin the cluster showed 4 accessions from Africa, 2 accessions from Asia, 5 accessions from Central America, 2 accessions from North America, 17 accessions from South America and 9 accessions from SSE with unknown origin. Cluster 2 consisted of 102 accessions with 14 accessions from A. caudatus, 58 accessions of A. cruentus, 5 accessions of A. hybridus, 50 accessions of A. hypochondriacus, 2 accessions of A. quitensis and 1 accession of A. retroflexus. Among the accessions of this cluster, 12 originated in Africa, 13 in Asia, 36 in Central America, 1 in Europe, 11 in North America, 20 in South America and 9 were from SSE and of unknown origin. Cluster 3 consisted 92 accessions. The cluster represented 3 accessions of A. caudatus, 44 accessions of A. cruentus, 6 accessions of A. hybridus, 17 accessions of A. hypochondriacus, 1 accession of A. palmeri, 7 accessions of A. quitensis and 1 accession of A. retroflexus. Based on geographical origin, it was found that 9 accessions were from Africa, 9 accessions were from Asia, 39 accessions were from Central America, 6 accessions were from Europe, 6 accessions were from North America, 15 accessions were from South America and 8 accessions were from SSE and had unknown origin. Cluster 4 was comprised of 39 accessions representing 5 accessions of A. caudatus, 12 of A. cruentus, 10 of A. hybridus, 9 of A. hypochondriacus and 3 of SSE. The cluster showed that 4 accessions originated in Africa, 8 in Asia, 8 in Central America, 2 in Europe, 6 in North America and 8 in South America. Cluster 5 revealed 2 accessions of A. quitensis which originated in South America. Cluster 6 represented 1 accession of A. quitensis from South America. Cluster 7 comprised 1 accession of amaranth from SSE.
Supplementary Table S4 depicts significance of different traits among various clusters. It was found that there was significant differences of petiole pigmentation, stem color, blade pigmentation, blade shape and flower color among different clusters among the 276 accessions. Mahalanobis D2 distance between different clusters is shown in
Supplementary Table S5, where distance between cluster 4 and cluster 7 was found to be highly significant at the threshold Chi-square value of 14.06.
4. Discussion
Information about genetic diversity and clustering among and within crop species is important for effective utilization of plant genetic resources [
18]. Analysis of genetic diversity and development of population structure have direct benefits in research related to evolution, population structure and plant breeding [
19]. Clustering can also indicate phylogenetic relations. However, the clusters of amaranth accessions shown in our analysis were based on morphological traits and for the most part were either mostly cultivated accessions or mostly wild accessions; but clustering did not agree with species identification and several species accessions were found in each cluster.
Different morphological characteristics were evaluated across all the amaranths both separately by collection and together. The genotypes consisted in a total of nearly 300 accessions from two collections, SSE and USDA. Overall, the plasticity in major morphological traits like leaf and flower color did not allow us to identify species as has been discussed before [
10,
11]. We could not correlate the SSE collection with species identified in the USDA collection and these genotypes clustered together across species identifications in the GRIN database. Despite this, by overall race type morphology as discussed by Espitia [
11], it appeared that the majority of SSE genotypes were likely to be of the cultivated species
A. cruentus or
A. hypochondriacus of cultivated races, given their predominantly upright architecture. The few exceptions to this were vegetable types potentially from
A. hybridus, and one labeled as
A. gangeticus [
16].
Among the morphological traits, there was significant effects of blade pigmentation, blade shape and flower color among different clusters of amaranths accessions from SSE. In addition to these morphological traits, petiole pigmentation and stem color had significant contribution among different clusters of amaranths accessions from USDA. This showed that there was more variation in USDA collection than in SSE collection validating the points made by Brenner et al. [
10] about landraces held by NPGS.
For the USDA collection, we found wide variation in morphological characteristics among and within species of amaranths. In some cases the accessions from cultivated species were clustered together however geographical origin was not important in determining clusters. In most cases, the same species was found to have variable morphological traits. This difficulty in the phenotypic identification of amaranth species has been observed before by various authors [
19,
20].
Variability in morphology was a widely observed factor in the accessions evaluated. The result was that morphological states were shared between species and it was hard to divide the accessions into morphotypes or to find correlation of traits with species. The one exception was level of branching in the USDA collection found to be high for wild accessions from
A. hybridus,
A. powellii,
A. quitensis and
A. retroflexus compared to the single stem of cultivated types from
A. caudatus,
A. cruentus and
A. hypochondriacus in most cases. Difficult weed species identification was observed before [
20,
21].
In another observation, we discovered that the grain amaranth accessions from SSE were already adapted to growing condition in the Southeastern USA, even if their respective species were unknown. SSE is a non-governmental organization that works to preserve America’s gardening heritage, so it would be important to determine the species through molecular means. In most studies [
1,
2,
22,
23,
24,
25,
26,
27,
28,
29,
30], molecular markers have been able to distinguish South American (
A. caudatus,
A. quitensis) from Central American (
A. cruentus,
A. hypochondriacus) species of the subgenus
Amaranthus as well as outgroups from other subgenera. The other subgenera of the genus are subgenus
Acnida (includes weedy amaranths such as
A. palmeri and
A. spinosus) and
Albersia (includes wild and vegetable species such as
A. tricolor and
A. viridis).
The use of molecular markers such as Simple Sequence Repeat (SSR) or Single Nucleotide Polmorphism (SNP) seems better than morphological analysis for distinguishing species of grain amaranths [
1,
2,
22,
23,
24,
25,
26,
27,
28,
29,
30]. SSR markers in the study by Oo and Park [
26] did find clear clustering pattern of geographically close accessions and related species but Suresh et al. [
1] did not. Vegetable species have been less well studied [
31,
32,
33,
34,
35,
36,
37] at least by molecular means than grain amaranths while weedy amaranths have been well studied, especially with recent outbreaks of herbicide resistant amaranths in countries with genetically modified crops and those developing countries transitioning to mechanical weed control.
Other marker types based on isozymes, seed protein patterns or better yet next generation sequencing have proven to be effective for species identification but the first two of these methods are time consuming while the last of these methods is cost intensive. Seed protein analysis would also require producing the seed of the genotypes in either field conditions or in a greenhouse where day length and photoperiod could be controlled and allow the evaluation of seed protein variability [
38] as well as provide clean tissues for isozyme analysis [
39]. Perhaps most promising, the recent use of next generation sequencing technology of Genotyping by sequencing (GBS), performed by Wu and Blair [
40] and Stetter and Schmid [
2], was successful at differentiating cultivars from wild accessions and different species from each other.
The latter studies show that molecular marker studies can complement botanical or morphological descriptors for Amaranth species. Species separation is usually based on time consuming and growth-phase specific, reproductive traits such as bract and tepal sizes of female flowers as the traditional methods for evaluating species differences [
10,
11]. Meanwhile, our study showed that field assessment of major morphological traits can be successful in the grain amaranths. However, as was previously observed in species and race characterization, there is a tremendous plasticity of plant size and branching within each species [
10]. Some major morphological differences, like flower and leaf color segregate across species, and most other traits like plant size depend on photoperiod and soil conditions in the site used for evaluation. A lack of flowering or seed production in many short-day photoperiod sensitive
A. caudatus and
A. quitensis genotypes found in the USDA collection [
10], but less so in the SSE collection, prevented some morphological traits from being evaluated and is a drawback of phenotyping that would not be present in DNA studies. Common cross species traits and phenotypic plasticity in the grain amaranths make species identification difficult in an open air, field setting as compared to a greenhouse.