Blood specimens from a random sample of 981 South African Negroid females were typed electrophoretically inter alia for their G-6-PD phenotypes, The allele frequency for GdB and GdnonB was found to be 0.8126 and 0.1874 respectively. Calculating the number of individuals expected for each phenotypic class, a highly significant deviation from the Hardy-Weinberg equilibrium became manifest, i.e. there was a deficit of 24.6% of heterozygotes and an excess of 12.3% of each of the two classes of homozygotes. Several possible reasons for this discrepancy e.g. the effects of pooling sub-samples, selection and misclassifications due to insufficient staining were examined and were found not be likely explanations for the observed phenomenon. Instead, the result is interpreted as due to only 3--4 stem cells which give rise to the haematopoetic system in man.

In a recent population study, we observed a striking deficit of G6PD heterozygotes among Southern African Negroid females. This finding was interpreted tentatively as evidence for a small number of hematopoetic stem cells in man. In a follow-up study we examined peripheral blood and cord blood in 547 mothers and in their newborn offspring. In mothers and sons, the frequencies of the G6PD alleles are apparently quite different. When the allele frequencies determined in sons are used for calculation of the expected phenotype frequencies in mothers and daughters, there is a large deficit of maternal G6PD AB phenotypes, and an equivalent surplus of G6PD homozygotes. However, no relevant heterozygote deficit is observed in newborn daughters. This discrepancy may be explained by the assumption that in peripheral blood of heterozygotes carrying the GdA- allele, G6PD-deficient cells progressively become eliminated during development from birth to adulthood. In other words, the large heterozygote deficit observed in adult females may be due to somatic selection rather than to a small pool of hematopoetic cells at the time of X differentiation.

The causes of genotype interaction with the environment was investigated using AMMI (Additive Main Effects and Multiplicative Interaction) analysis. AMMI quantified environments by means of E_IPCA 1 (first principal component analysis for environment) such that it was highly significantly correlated (r = -0.79) with accumulated growing degree units (GDUs) during the growing season. G_IPCA 1 (first principal component analysis for genotype) scores were highly significantly correlated with growth rate related observations, such as days to 50% pollen shed (-0.77), days to physiological maturity (-0.81), and grain moisture at harvest (-0.87). Multiple linear regression analysis of data led to the conclusion that G _IPCA 1 scores can be explained with R2 = 0.80 accuracy from a combination of grain moisture at harvest and either days to physiological maturity, or days to pollen shed. The AMMI required that E_IPCA1 and GJPCA1 scores shouldrboth be either positive or negative in order to increase AMMI1 yield estimates. This means that hybrids with a slower growth rate will be better adapted to environments with more GDUs during the growing season, and hybrids with a faster growth rate will be better adapted to environments with less GDUs during the growing season. This conclusion suggests that temperature, especially minimum night temperature, which determines the rate of development between physiological stages in the maize plant, may be a major contributor to genotype by environment interaction.

Individual maize cultivars may react differently to fertilisation. Thus differential fertilisation levels can potentially have a marked influence on profitability of maize production and also on optimal cultivar selection. Fertilisation trials are not normally cultivardirected. The effect of fertilisation on a cultivar can thus not be directly determined. Adaptation regression lines of cultivars were used to determine economic optimum fertilisation levels for different maize cultivars at different locations. The conditions for such determinations are that the reaction of at least one cultivar on fertiliser must be known, and that the relationships between yields of different cultivars must be available. The variances between observed and predicted cultivar yields are relatively small. The proposed method should thus be satisfactory for the prediction of yields. * Based on a D.Sc. (Agric) dissertation by J. van Zyl of the University of Pretoria. The research was funded by a BP-NAMPO bursary that was awarded to Van Zyl

No abstract