Abstract
We studied symbiotic performance of factorial combinations of diverse rhizobial genotypes (G(R)) and East African common bean varieties (G(L)) that comprise Andean and Mesoamerican genetic groups. An initial wide screening in modified Leonard jars (LJ) was followed by evaluation of a subset of strains and genotypes in pots (contained the same, sterile medium) in which fixed nitrogen was also quantified. An additive main effect and multiplicative interaction (AMMI) model was used to identify the contribution of individual strains and plant genotypes to the G(L) × G(R) interaction. Strong and highly significant G(L) × G(R) interaction was found in the LJ experiment but with little evidence of a relation to genetic background or growth habits. The interaction was much weaker in the pot experiment, with all bean genotypes and Rhizobium strains having relatively stable performance. We found that R. etli strain CFN42 and R. tropici strains CIAT899 and NAK91 were effective across bean genotypes but with the latter showing evidence of positive interaction with two specific bean genotypes. This suggests that selection of bean varieties based on their response to inoculation is possible. On the other hand, we show that symbiotic performance is not predicted by any a priori grouping, limiting the scope for more general recommendations. The fact that the strength and pattern of G(L) × G(R) depended on growing conditions provides an important cautionary message for future studies.IMPORTANCE The existence of genotype-by-strain (G(L) × G(R)) interaction has implications for the expected stability of performance of legume inoculants and could represent both challenges and opportunities for improvement of nitrogen fixation. We find that significant genotype-by-strain interaction exists in common bean (Phaseolus vulgaris L.) but that the strength and direction of this interaction depends on the growing environment used to evaluate biomass. Strong genotype and strain main effects, combined with a lack of predictable patterns in G(L) × G(R), suggests that at best individual bean genotypes and strains can be selected for superior additive performance. The observation that the screening environment may affect experimental outcome of G(L) × G(R) means that identified patterns should be corroborated under more realistic conditions.