Genetic diversity in lowland , midaltitude and highland African maize landraces by morphological trait evaluation

Genetic diversity information is a resource for improvement in crop productivity and trait performance, however, there is scanty information on genetic diversity estimates in the maize landraces covering the major geographical regions of Africa. In the current study, the genetic diversity of 35 landraces originating from lowland, midaltitude and highland regions of Africa and held in the IITA Genetic Resource Center, Ibadan, Nigeria, were evaluated using morphological trait evaluation. The landraces were tested in non-stressed environments in Ghana by evaluation of 27 traits. A large within and between genetic variability was identified which increased from highland to lowland populations and was highest in the midaltitude population. Genetic similarity coefficients ranged from 0.00 to 0.80 with mean of 0.26±0.18 across the three populations, and 0.23±0.16, 0.29±0.18, and 0.38±0.25 in the midaltitude, lowland and highland populations, respectively. A total of 21 discriminant traits were identified from the principal components analysis. A UPGMA cluster analysis and PCA biplot produced four main clusters which provide a sound basis for exploitation of heterosis. Nine distant landraces were identified majority of which produced grain yield exceeding 5.0 Mg ha -1 . In terms of improvement in grain yield, earliness and drought tolerance, TZm-14, TZm-41, TZm-242, TZm-37, TZm-1360, TZm1376, TZm-1367, TZm-4, and TZm-270 would be useful. A large genetic diversity resides in the African maize landraces which could be conserved and exploited for maize improvement.

regions, and 12.3 million hectares in the tropical lowlands (FAOSTAT, 2013), all in smallholder systems except in South Africa.Demand for maize has increased to meet food requirements for population growth, as feed for livestock and poultry and for biofuel ethanol production.At a population growth rate of over 2% per annum, SSA needs to double her maize production to feed an estimated population of 1.5 billion by year 2050 (FAO, 2006).Compared to global increase in maize production of 2.2% per annum at annual yield growth of 1.5 Mg ha -1 , maize yield in SSA has barely increased by little over 1% in 50 years reaching an average yield of 1.8 Mg ha -1 in 2011, about a quarter of the world average yield of 5.5 Mg ha -1 (Fischer et al., 2014).Constraints to maize yield increase in Africa, include low soil fertility (Gibbon et al., 2007), limited availability and low adoption of modern improved varieties which achieve farm yield of only 25% of the potential yield (FAOSTAT, 2013), lack of irrigation facilities and low agricultural input.Other limiting factors include limited labor, and uncertainties with crop success currently arising from climate change effects (Cairns et al., 2013;IPCC, 2007) which drive farmers to have preference for their landraces.
The landraces, being historic and dynamic genotypes, and having evolved from natural and anthropogenic selection system (Bellon and van Etten, 2013) exhibit some buffering effect to climate change effects including extreme heat stress and drought (Mercer et al., 2012;Mercer and Perales, 2010).By their wide genetic base, requirement for low agronomic input, better storage properties, desirable culinary characteristics, and some genotypes possessing superior agronomic characteristics than the improved cultivars (Amanor, 2013;Bellon, 2004), the landraces represent unique class of germplasm for exploitation and development.
Currently, only few attempts have been made to examine the genetic diversity in the SSA landraces.The need to evaluate the entire collection to estimate the breadth of genetic diversity and identify genotypes with important traits such as high grain yield and climate adaptive traits is of prime importance.
Both morphological and molecular methods are employed in estimating genetic diversity in germplasm collections.Although morphological evaluation is limited by effect of environment on trait expression, exhibits low heritability, is time consuming, labour intensive, requires a large population size, and does not cover the genome (Botha and Venter, 2000;Smith and Smith, 1992), it offers an unparalleled means of identification of phenotypic variation.
The research carried out by Sanou et al. (1997), Beyene et al. (2006), Legesse et al. (2007) and Magorokosho (2006)  In the context of maize cultivation, six mega environments are defined in Africa, on the basis of climate, elevation and soil type into the highlands with elevation above 1600 m.a.s.l., upper humid, lower humid, and dry midaltitude elevation of 900 to 1600 m.a.s.l, and humid and dry lowlands below 800 m.a.s.l.It is believed that the landraces that have had long exposure and survival in these conditions have differentiated in their respective environments hence worth examining for their genetic diversity estimates.
The International Institute of Tropical Agriculture (IITA) has in store over 1,000 maize landraces collected from many countries in Africa.The objective of this research was to determine the genetic diversity in maize originating from three environments namely, lowland, midaltitude, and highland regions of SSA using phenotypic characterization.The information will be useful for identifying genotypes for broadening the genetic base in the gene pools of maize improvement programs and provide a guide for conservation and management of maize.

Plant material
Thirty-five lowland, midaltitude and highland landraces originating from 12 countries in SSA were sampled from the IITA maize collection (Table 1).An open-pollinated lowland genotype, "Obatanpa GH" was included as a check.Accessions were planted in the wet season in two consecutive years, March to July 2011 and April to August 2012 in the Agricultural Experiment Station of the Kwame Nkrumah University of Science and Technology, Kumasi, Ghana, in randomized complete block design with three Fertilizer application was equivalent to 120:60:40 kg ha -1 of N-P2O5-K2O plus 50 kg ha -1 sulphate of ammonia applied at 21 days after planting and at ear emergence.Irrigation was carried out as and when needed.Maize stem borers (Busseola fusca, Sesamia calamistis) and cutworms (Agrotis spp.) were controlled using Conpyrifos 48% (1.0 to 1.5 L ha -1 ) and Cymethoate Super (1.0 to 1.5 L ha -1 ).

Morpho-phenological trait evaluation
Morpho-phenological traits were evaluated by employing the IBPGRI and CIMMYT (1991) (Felsenstein, 1985) using the PAST software (Hammer et al., 2001).Finally, a principal components analysis (PCA) and biplots determined the discriminatory power of the traits and revealed relationships among traits and accessions.The NTSYS pc 2.2 software (Rohlf, 2009) was employed for all multivariate analysis.

Within and between population differences
Ear characteristics included 57% pale yellow silks, 92% white cobs and 62% dent kernels in predominantly regular kernel arrangement.Mean values, minimum and maximum, standard deviation, mean squares and coefficient of variation of the 22 phenotypic traits are presented in Table 2.The populations exhibited a large within population differentiation except for anthesis-silking interval and a large between population variation for all traits but tassel length, ear leaf width, plant height, and stalk diameter.A trend of increasing variability based on the number of significant mean squares was identified in 16 traits in highland, 20 traits in lowland, and 21 in midaltitude populations (Table 2).Low to moderate correlation coefficients were found between traits.Anthesis-silking interval showed predominantly low nonsignificant negative correlations with all other traits whereas ear leaf width showed moderate significant positive correlations with all traits.Similarly, grain yield showed low to moderate significant correlation with all traits except anthesis date, silking date, ear leaf length, ear length, and kernel thickness (Table 3).
Variations in traits on the basis of accession means are compared to a check yield of 6.29 Mg ha -1 .Kernel characteristics were highly variable.

Cluster analysis
A UPGMA cluster analysis displayed four clusters (Figure 1) differentiated predominantly by geographic origin and confirmed seven accessions to be distant TZm-42, TZm-20, TZm-5, TZm-2, TZm-270, TZm-41, and "Obatanpa GH".Cluster I had eight accessions, six of which originated from Midaltitude regions and two, including the check "Obatanpa GH" from lowland regions.The genotypes exhibited early maturity, long ear leaf length, large ear and cob diameters, and long anthesis-silking interval.Cluster II, a mixed group with nine accessions, were mostly collected from all three mega environments.All highland accessions except one belonged to this cluster most of which exhibited shortest anthesis-silking interval, large tassels, large ear leaf widths, largest stalk diameter, tall plants, and highest mean grain yield of 5.20 .Cluster IV was predominated by seven members of Midaltitude origin which were late maturing, short plants with small tassels, small ear leaf width, and least mean grain yield of 0.14 Mg ha -1 below the overall average yield (Table 5).Numbers in parenthesis represent differences between the cluster and overall mean; the legends for traits and their units are as given in Table 2.

Principal components analysis
components with eigenvalues exceeding 1.0 explained 86.67% of the total phenotypic variance (Table 6).In the first PC which accounted for 39.97%, the variance was attributed to the weighted sum of tassel length, ear leaf length, ear leaf width, plant height, ear position, stalk diameter, stay green, ear length, ear diameter, hundred kernel weight, kernel length, and grain yield.This indicates a direct relationship between plant architectural traits, stay green, grain yield and yield components.
Accessions having this combination of positively loading traits represent positive associations that are likely to share common alleles.Crossing within this gene pool would accumulate alleles for large values of tassel size, ear leaf length and width, stay green, yield components and grain yield.The PC2 explained 18.70% of the total variance attributed to the weighted sum of number of kernels per row, kernel length, kernel width and ear weight, and a weighted difference of cob diameter, number of rows per ear and kernel thickness.A variance contribution in PC3 of 16.74% was attributed to weighted sum of anthesis and silking dates and anthesis-silking interval but a weighted difference in ear diameter and grain yield, indicating an inverse relationship between maturity date, anthesis-silking interval and grain yield.In other words, the late maturing genotypes with long anthesis-silking interval exhibited low grain yield.Being landraces, it was not unexpected to find genotypes exhibiting undesirable traits such as large anthesis-silking interval and low grain yield.The contribution of PC4 of 11.06% of the total variance was attributed to the weighted difference between anthesis-silking interval and ear height (Table 6).
A trait biplot of PC1 and PC2 (Figure 2A) accounted for a cumulative variance of 58.87% and identified four major correlation groupings, namely, a group based on earliness, plant architectural traits and grain yield; a second group based on ear-related traits; a third group based on kernel characteristics; and lastly, an uncorrelated group of traits comprising number of rows per ear, anthesis-silking interval, cob diameter and number of kernels per row.The tight angles between anthesis and silking dates, kernel width and kernel length, stay green and kernel thickness, ear leaf length and ear leaf width, plant height and grain yield signified direct association among these traits.In contrast, anthesis-silking interval, cob diameter, number of rows per ear, and number of ears per plant were uncorrelated.Strong associations were exhibited between stay green and kernel thickness.Plants with longer stay green also had large kernel thickness.In the accession biplot (Figure 2B), three uncorrelated groups and eight accessions that were separated from the core groups were identified, confirming that TZm-270, TZm-2, TZm-5, TZm-251, TZm-42, TZm-385, TZm-41 and TZm-1358 were distant.Hybridization within clusters I and III should be favorable in accumulating alleles for early maturity, low to moderate anthesis-silking interval, stay green, and high grain yield.On the contrary, accessions of cluster IV had predominance of small plant architectural traits and low yield, low values of stay green and longest anthesissilking interval.
In the context of maize improvement, a breeding strategy that would exploit the existing variation within and between the clusters and accumulate desirable characteristics is to create a cross-pollinated population with a high frequency of high grain yield, early maturing, low anthesis-silking interval and high stay green genotypes and improve this group via recurrent selection to obtain highly heterozygous populations from which superior heterozygous genotypes may be selected for inbred line development.

DISCUSSION
The African maize landraces constitute an important class of genotypes characterized by wide diversity in phenology, plant growth, grain yield, and leaf photosynthesis most of which connote a diversity of farmer preferences and adaptive characteristics to a wide range of environments in which they have evolved.This diversity awaits to be exploited for maize improvement in an efficient manner.Noteworthy was the identification of some genotypes that outperformed the improved cultivar in grain yield, leaf photosynthesis and phenology.
The simultaneous large within and large between population variations is the likely result of large rate of gene flow between midaltitude and lowland populations but somehow restricted between the geographically discrete highland genotypes.The variation highlights a naturally conserving potential which has been shaped by evolutionary factors including a large effective population size with a large proportion of gene flow which does not disrupt patterns of local adaptation of the taxon.Marshall (1977) reported that the landraces possess both within and between variations.A PCR-based study on genetic diversity, as well as estimation of F ST , F SC , and G ST are needed to confirm the wide within and among diversity in the African maize landraces.
The larger variability in the midaltitude and lowland accessions was not surprising considering that the major entry of maize to Africa was through the west coast rather than the east or north.In fact, besides midaltitude Tanzania and Zambia, many of the lowland and midaltitude accessions in current study originated from West African countries such as, Togo, Congo, Guinea, Equatorial Guinea, and Chad.The fewer representations of the highland genotypes may have influenced the low variation observed.
The overall mean anthesis dates of 58.62±5.67 days and silking date of 62.68±6.10days were similar to the 61.5±0.2 days and silking date of 62.7±0.7 days (Salami et al., 2007), but shorter than anthesis and silking dates of 65.1±3.2 and 71.5±3.0 days (Beyene et al., 2006), 83.0±0.5 days and 85.9±0.55 days (Azad et al., 2012), 84.3±1.7 and 86.6±12.0days of 498 maize accessions of Asia, Latin America and U.S.A. (Weiwei et al., 2012).Early maturing genotypes that may be useful to breeding programs in marginal regions in tropical Africa were TZm-8, TZm-2, TZm-1521, and TZm-385.The short anthesissilking interval of TZm-1376 (2.3 days) would benefit breeding for drought tolerance through escape and avoidance.The positive association of ear leaf width with grain yield was noteworthy and would be relevant in selection based on correlated response.
Genetic distance within the accessions confirmed a large variability which was further validated by cluster analysis and principal components analyses.TZm-42, TZm-2 and TZm-5 exhibited least values of grain yield and yield-related traits of hundred kernel weight, stay green, and stalk diameter, ear leaf width and tassel length.In contrast, TZm-270 demonstrated largest values of these traits, unequalled with all other accessions and would be worthy to incorporate into breeding programs.
Existence of both within and among population variation presents a special opportunity for in situ conservation which in Africa is a common traditional farming practice to check genetic erosion and ensure maintenance of the evolutionary processes in the taxon while adopting modern agricultural technology (Brush,  1995).The corollary of this conservation is environmental resilience and creation of novel genotypes (Altieri and Merrik;1987;Friis-Hansen, 1994).Additionally, the partitioning of the populations into clusters suggests that benefits would be accrued from intercrossing between clusters to exploit heterosis for grain yield, earliness, and the desirable low values of anthesis-silking interval which indicate less sensitivity to drought stress.A within cluster crossing involving a large number of ears should preserve the rare alleles in the collection (Crossa, 1989).The combined high yield and early-maturity traits present novel genes at these loci and may be beneficial for broadening the genetic base of elite gene pools.

Conclusion
The findings of this work suggest that the African maize landraces are rich in genetic diversity which increases from highland to lowland and to midaltitude genotypes.The large difference between population differentiation and the rich diversity suggest a historic formation from a large effective population size.At the same time, the large within population variability indicates a gene flow among the populations which is consistent with an ex situ conservation of maize.The morphological variability can provide basis for creating core subsets especially from the two highly populated clusters.Identification of the large variabilities among the traits bring to fore the richness in alleles and the urgency to conserve and incorporate the landraces into regional breeding programs to widen the genetic base of the genepool.These genotypes have adapted to the environmental conditions of Africa and are expected to contribute alleles for both trait improvement and to withstand environmental stress factors in time and space.
In terms of world average maize yield of 5.15 Mg ha -1 , the important genotypes and their respective grain yields were TZm-14 (5.2 Mg ha ), all of which were midaltitude and highland accessions.

Figure 1 .
Figure 1.A UPGMA dendrogram of 22 quantitative agro-phenological traits based on Pearson correlation coefficients on 35 highland, midaltitude, and highland maize landraces and a check.Bootstrap values are shown at the nodes.

.
Five members of cluster II were distant TZm-1095, TZm-242, TZm-1360, TZm-1424, and TZm-270.Cluster III had 11 accessions mostly of lowland and few Midaltitude genotypes with intermediate maturity, high stay green and intermediate grain yield exceeding mean grain yield by 0.12 Mg ha -1 Principal components analysis delimited 21 important discriminatory descriptors.The first four principal

Figure 2 .
Figure2.Plots of PC1 against PC2 for (A) traits and (B) 35 accessions and "Obatanpa GH", the check.The labels are as defined in Table2footnote.

Table 1 .
The landraces of lowland, midaltitude and highland origins in SSA sampled from the IITA maize collection.
Data were analyzed by analysis of variance using the General Linear Model ofSAS 9.3 (SAS Institute, 2011).Pairwise genetic distances based on Pearson correlation coefficient were computed on the standardized data matrix.Mean genetic distance of the entire population was calculated.The hierarchical Unweighted Pair Group Method with Arithmetic Averages (UPGMA) clustering of the distance matrix was carried out to generate a dendrogram.Statistical significance of the dendrogram was determined by bootstrap analysis provided.Plant height varied from 139.95 to 207.95 cm and ear height from 55.00 to 112.98 cm.Tassel length of 42.69 to 53.41 cm, stalk diameter of 15.38 to 22.97 mm and stay green from 33.33 to 96.25%.The accessions showed average ear leaf length and ear leaf width of 79.16 and 8.50 cm, respectively, compared to the check with 84.16 and 9.86 cm, respectively.The largest ear leaf width of 11.11 cm was exhibited by TZm-270.

Table 2 .
Mean, standard deviation, minimum and maximum, mean squares and coefficient of variation of morpho-phenological traits on 35 lowland, midaltitude, and highland maize populations including a check evaluated in Kumasi in 2011 and 2012 major rainy seasons.Arrangement of data in a cell is in the order of mean, standard deviation, range in parenthesis, mean square, and coefficient of variation; Letters represent across mega-environment differences; Significant mean squares are represented as *P<0.05;** P<0.01, *** P<0.001.

Table 3 .
Pearson correlation coefficients of 14 selected morpho-phenological traits performance of 35 lowland, midaltitude and highland maize populations including a check evaluated in Kumasi in 2011 and 2012 major rainy seasons.

Table 5 .
Overall mean, cluster means, and standard deviation of the 35 highland, midaltitude and lowland African maize landraces and a check evaluated in Ghana by morphological trait measurement.

Table 6 .
Eigenvalues, eigenvectors, and cumulative percentage of variation explained by the first four principal components (PC) on 22 morphological traits in 35 maize landraces from lowland, midaltitude, and highland region of Africa including a check.