Introduction

Maize has great global and national importance, according to production volume, uses and adaptability to different environmental conditions (^{Rocandio-Rodríguez et al., 2014}). Mexico devotes an area of 7.4 million hectares to corn and the state of Morelos devotes a sowing area under rainfed of approximately 26 000 ha, with average yield of 2.8 t ha^-1 (^{SIAP-SAGARPA, 2013}). Based on the rainfed surface, around 80% is planted using seed from native populations (^{Ortega et al., 2010}), so that native maize germplasm is the most important in most of Mexico (^{Ángeles-Gaspar et al., 2010}). On the other hand, corn presents a broad genetic variability that allows species to adapt to different soil types and agro-ecological environments (^{Perales et al., 2003}; ^{Beyene et al., 2005}). Trait characteristic from native varieties have been the basis for the formation of maize races, of which to date have been reported about 59, adapted to specific ecological conditions in Mexico (^{González et al., 2006}).

Native maize is the germplasm that has been generated by continuous selection and adapted to the various regional specific agro-ecological conditions (^{García-Lara and Bergvinson, 2013}). The evolution of this germplasm has survived a no or limited fertilization and low health protection, so that their potential value lies in containing genes for resistance to diseases, pests, nutritional quality and adaptation to adverse environmental conditions (Liu et al., 2009). However, native germplasm has unfavorable traits such as; high lodging level, floral asynchrony and low yields, among others (^{Nava and Mejía, 2002}).

The evolution in native germplasm has its origin in the continuous planting by farmers and selection applied for long periods of time, which has changed agronomic characteristics that allow better adaptation to different ecological niches (^{Bellon and Berthaud, 2004}). However, the diversity of native maize is being lost due to genetic and cultural erosion, use of improved seeds, crop change, migration and occurrence of natural disasters (^{Eschholz et al., 2010}), which forces to keep (ex situ or in situ) and take advantage of this plant resource in breeding programs (^{López-Romero et al., 2005}).

Native maize are heterogeneous-heterozygous populations, which have a high number of allelic combinations, which are a valuable natural source of new alleles and represent an irreplaceable genetic resource bank (^{Smale et al., 2003}). The high genetic variation, characteristic from native maize varieties, has been exploited in a few breeding programs, which has led to improved and adapted germplasm to many environments (Liu et al., 2010); However, due to the large number of existing native populations these have not been fully evaluated and therefore the genetic relationship that may exist between these populations is unknown (^{Martín et al., 2008}).

The increase of genetic base of maize is an important goal in breeding programs (^{Radović et al., 2000}), which can be done by using native germplasm. Evaluation and characterization of native maize, has been based mainly on the expression of morphological traits, so the evaluation across environments is essential to measure the environmental effect on gene expression of the traits (^{Berard et al., 2009}). Characterization has focused to describe plant, spike, grain and cob traits (^{Lucchin et al., 2002}); morphological relationship between populations allows grouping germplasm and identifies promising material for its improvement (^{Andjelkovic and Thompson, 2006}).

Native germplasm evaluation is necessary to understand and determine their agronomic value, besides allows the formation of groups with similar populations morphologically, which makes it easy to maintain and eventually improve agronomic traits of interest (^{Hartings et al., 2008}). In this context, the objectives of this study were to assess the level of morphological variation between native maize populations and their cross F₁, and compare the morphological similarity of native populations with the similarity of their cross F₁.

Materials and methods

The germplasm evaluated in this study consisted of seven native populations collected and maintained by the Maize Breeding Program from the Escuela de Estudios Superiores of Xalostoc, which were selected for their relatively high grain yield under rainfed and irrigated conditions, and because of their different geographical origin. Three of the seven populations are from the state of Morelos BJM1 (2), BJM2 (3) with similar traits to Tuxpeño Norteño, CB029 (6) similar to Tabloncillo; three populations from the state of Tamaulipas, of which CAB (1) and MOR (7) are similar to Tuxpeño Norteño and to population RAT (4) belonging to Raton race, and a population from the state of Puebla BCP1 (5) which is a mixture of races such as Raton and Crystalline from Caribbean. In addition, the behavior of the 21 diallel crosses F₁ formed from seven native populations was also assessed, and as control included commercial varieties H-515, Costeño Enhanced and VS-535, giving a total of 31 genotypes. The diallel crosses among the seven native populations were formed in the agricultural cycle spring-summer 2012. The agronomic evaluation of 31 genotypes was conducted in experimental fields under three environments; Ayala, Morelos in the agricultural cycle autumn-winter 2012/2013 (AMB-1), Ayala, Morelos in the spring-summer 2013 (AMB-2), and

Tepalcingo, Morelos in the spring-summer 2013 (AMB3). The experiments conducted in AMB-1 (irrigation) and AMB-2 (rainfed) were carried out in the experimental field of the Escuela de Estudios Superiror of Xalostoc, from the Universidad Autonoma from the state of Morelos, and AMB-3 (rainfed) on fields belonging to a cooperating farmer.

The experimental design used in the three experiments was a randomized complete block with three replications. The experimental unit consisted of four rows 5 m long with row spacing of 0.8 m. Two seeds per plant, each 0.25 m were deposited at planting, in the phenological stage V2 a thinning of a plant per plant was performed (50 000 plants ha^-1). The variables measured were, male (FM in days) and female flowering (FF in days), floral synchrony (SF in days), plant height (AP in cm) and cob (AM in cm), stem diameter (DT in cm), ear length (LM in cm), ear diameter (DM in cm), rows per ear (HM), grains per ear (GM), corncob diameter (OD in cm) , 100 grains weight (P100G in grams) and grain yield (RG in t ha^-1).

Data was processed by combined analysis of variance. Comparison of means was performed using the least significant difference (DMS_0.05). Populations were compared against control and populations against crosses through non-orthogonal contrasts. Genetic and phenotypic variances for populations and crosses were estimated, based on the expectation of square means. Population and cross data separately were subjected to principal component and cluster analysis. The principal component analysis was performed with the use of the data matrix X, made with the averages of the 13 variables as columns and rows as native populations or cross.

Cluster analysis was performed based on Mahalanobis statistic distances, same that was used as similarity measure, the matrix was estimated through the formula D_ij² = (µ_i - µ_j )' V^-1 (µ_i - µ_j ), where: D_ij² = distance being the ith and jth population or cross, µ = mean vector of the ith or jth population or crosses, V^-1 = inverse of matrix variance and covariance (^{Manly, 2000}). For the construction of the dendrogram the clustering method UPGMA (Unweighted Pair Group Method with Arithmetic Mean) was used. All statistical analysis were performed with the use of Statistical Analysis System (^{SAS, 1999}] software.

Results and discussion

The analysis of variance showed statistically significant differences (p≤ 0.01) for populations and population × environment interaction for all variables evaluated (Table 1). Regarding to the comparison of populations vs control, statistical test detected differences (p≤ 0.01) in all variables except in DO. In the comparison of native populations vs F1cross were statistical differences (p≤ 0.01) for eleven of the thirteen variables. The coefficients of variation ranged from 2.2 to 9.3%, these values corresponded to flowering and grain yield, respectively. Statistical differences detected in population source are the result of the high degree of genetic variability between the evaluated germplasm, which is effect of the different geographical origin of the populations. Similar results were observed in a characterization study of native maize populations from Puebla, where it was determined that the wide diversity found in local corn, is partly related to the geographical origin of the populations studied (^{Ángeles-Gaspar et al., 2010}).

The differences between F₁ cross is explained by the genetic divergence among native populations; it has been observed that as genetic diversity in parent germplasm increases, genetic variation increases in their progeny (^{Pecina-Martínez et al., 2009}). The presence of genotype ×environment interaction found in all variables measured, suggests that the genotypes have a differential behavior from one environment to another, this type of interaction has been reported in native maize from Puebla valley where the main objective was to know the level of morphological diversity in the germplasm studied (^{Hortelano et al., 2008}). The results of the comparison between native populations and F₁ crosses are explained by allele combination that come from different population and to native populations per se have high genetic variability. Similar results were observed in a study of genetic divergence between native maize populations from the highlands of Mexico and their crosses (^{Caraballoso-Torrecilla et al., 2000}).

Estimates of genetic and phenotypic variance for native populations and F₁ cross (Table 2) show that phenotypic variance in all variables, was expressed in greater magnitude than genetic variance; which is in concordance with expectations, because the phenotypic variance involves not only genetic component, the environmental effects and genotype-environment interaction (^{Rocandio-Rodriguez et al., 2014}). Overall amplitude range was higher in F₁ crosses than in native populations, which can be explained by the latter for being heterogeneous populations and in large amount heterozygotes, showed a greater degree of allelic combinations which generated a higher level of genetic variability (^{Hortelano et al., 2008}).

The results of principal component analysis for native populations (Table 3) identified five principal components that cumulatively accounted for 91.2% of the total phenotypic variation among populations. Considering that the visualization of inter-population relationships is appropriate graphically when considering three dimensions; discussion of results focuses on the first three components, which jointly accounted for 72.5% of the inter-population total phenotypic variation. Regarding the eigenvectors of the principal components, considering CP1, which explained most of phenotypic variation component (33.3%), it was found that this component was largely determined by six of the thirteen measured variables such as: male (FM) and female (FF) flowering with values of 0.8 and 0.87 respectively, length and ear diameter with values of -0.73 and -0.80, P100G and RG with values of 0.77 and 0.78.

Regarding morphological similarity between native populations (Figure 1), the principal component analysis (Figure 1A) shows a wide spatial dispersion of the seven populations in the four two-dimensional quadrants formed by CP1 and CP2. The CB029 populations (from Morelos) and CAB (from Tamaulipas), were located in quadrant one, considering the third dimensional axis (CP3) this populations are in a different spatial plane. In the second quadrant are the BJM1 (Morelos), RAT (Tamaulipas) and BCP1 (Puebla) populations. The native population BJM2 from the state of Morelos was located in the third quadrant and finally the MOR population (Tamaulipas), was located in the fourth quadrant. Cluster analysis (Figure 1B) for native populations detected similarities in interpopulation morphological similarity, such as those identified in the principal component analysis.

Figure 1 Relationships between populations based on principal component analysis (A) and morphological similarities dendrogram generated by cluster analysis (B) for native maize populations. 

The dendrogram distinguished six groups (Figure 1B), where the largest number of genotypes was located in group (III), grouping the BJM1, RAT populations, and improved varieties control, Costeño Enhanced and VS535. This group was mainly characterized by having an intermediate flowering (66 days), an LM of 18.5 cm and grain yield of 5.10 t ha^-1. In the other five groups, were distributed the rest of the native populations, which is indicative of the high genetic and phenotypic variability, present between populations evaluated, these being panmictic nature have also broad genetic base. The above results indicate the existence of a high inter-population genetic variation in the 13 measured morphological traits, which also has been reported in other studies conducted with native germplasm, particularly in the eastern part from the State of Mexico, which concluded that vegetative variables related to spike and corn grain traits were relatively less affected by the environment, which could be used more efficiently to explain the differences between native maize populations (^{Herrera et al., 2000}).

The dispersion of native populations through the four dimensional quadrants and the third axis represented by CP3, confirm the existence of high levels of variation in the traits evaluated, which was expected due to the different origins of the germplasm. On the other hand, variables related to flowering and cob size identified by this research, such as those that determined most CP1 (33.3% of phenotypic variation), agree with the most important in a characterization study of native maize populations in the state of Puebla (^{Ángeles-Gaspar et al., 2010}).

The principal component analysis for F₁ crosses (Table 4) showed the same behavior as in native populations, the five first components as the most important. Together, they accounted for 83% of the total phenotypic variation observed between F₁ crosses. Based on the first three components; these explained independently, CP1= 27.5, CP2= 21.1, and CP= 15% of total observed variation between F₁ crosses. Most phenotypic variation was explained by CP1, the variables that contributed most to this component were: AP, AM, DT, LM and GM, with eigenvalues of -0.76, -0.64, -0.46, 0.81 and 0.79, respectively.

When comparing the degree in which the five and first three principal components explain the inter-population morphological variation (91.2 and 72.5%, respectively) and the inter-cross F₁ variation (83 and 63.6%, respectively), the decrease in the values of F₁ crosses is due to when hybridizing panmictic populations greater genetic variability is generated in their progeny (^{Romero et al., 2002}), so it requires a greater number of principal components to explain morphological variation generated in F₁ cross. The results of principal component analysis (Figure 2A) found greater spatial dispersion between crosses in the two-dimensional quadrants formed with CP1 and CP2 components. Six crosses were located in quadrant I, which showed the highest values in ear length (19 cm), ear diameter (5.0 cm) and rows per ear (16). In quadrant II eight crosses that had the ears with the shortest length (17 cm) and ear diameter (3.2 cm) and larger cob diameter (2.8 cm) were grouped. In quadrant III six crosses, related by cob length (17 cm), grains per ear (506 to 659) and grain weight of 57 g were located. In quadrant IV, C17 and C37 cross, defined by its late flowering (70 days to FM) and plant height of 274 cm, also had the highest number of grains per ear (622) were located.

Figure 2 Relationship between crosses based on principal component analysis (A) and morphological similarities dendrogram generated by cluster analysis (B) for F₁ maize crosses. 

The dendrogram generated by the cluster analysis for F₁ crosses (Figure 2B) defines the formation of five groups. Only C17 and C47 crosses were separated as independent groups. The eight crosses in group I, are characterized by having the highest values in ear length (19 cm), intermediate values in the number of grains per ear (578), higher grain weight (56 g) and intermediate yields (5.2 t ha^-1). This group I includes the largest number of crosses and all native populations of the study are involved. Group VI consists of four crosses, was characterized by exhibiting greater plant height (AP= 292 cm), reduced ear length (LM= 17 cm) and the lowest number of grains per ear (GM= 532), as that in group I seven populations are involved, however, shows greater morphological inter-cross variation.

In group IV three crosses were located, which are distinguished by a high number of GM (592) and intermediate values in P100G (55 g). Groups constituted by only one cross were groups II and VII, group II was the cross RAT × Mor, both populations from Tamaulipas, while group VII was constituted by CAB × Mor where CAB population comes from Tamaulipas, which means that despite having the same geographical origin shows that there is great genetic variation in maize germplasm from different regions of Mexico. The greater dispersion of the cross in the four two-dimension quadrants supports the fact that genetic variability in maize can be increased by crossing native populations, both from the same and different geographical and genetic origin, the same effect has been reported in the morphological characterization study with germplasm from northwestern Mexico (^{Martín et al., 2008}).

Conclusions

Native maize populations showed a high degree of interpopulation genetic variability in all variables, which caused that F₁ crosses showed greater genetic and phenotypic variance than native populations. In the native populations and their F₁ crosses, were identified the five most important principal components, which accounted for 91.2 and 83% of the total phenotypic variation respectively. The spatial dispersion of populations and F₁ cross confirmed the high level of this phenotypic variation among germplasm evaluated. Cluster analysis revealed a high degree of genetic divergence among native populations, by locating in different group to five of the seven native populations and only in group III were located the BJM1 and RAT populations from races Tuxpeño Norteño and Raton. For crosses seven groups were identified, where RAT × MOR (C₄₇) and CAB × MOR (C₁₇) constituted own groups, revealing high genetic variability among crosses.