The bias of widely used degenerate nifH-specific primer sets was first tested using denaturing gradient gel electrophoresis (DGGE), and their application for profiling of complex communities assessed for roots of Oryza longistaminata. When primers (P) with mismatches at nondegenerate positions were used on genomic DNA of Azotobacter vinelandii, which harbors three single divergent nifH genes, template-to-product ratios were highly skewed. In contrast, we obtained no evidence for a large PCR bias when we used highly degenerate primers with no mismatches (Z). Similar results were obtained for reverse transcription (RT)-PCR amplifications from root RNA from O. longistaminata grown at the river bed of the Okavango, where Z-primers detected a more complex nifH pool, corroborating that the P-primers are quite biased in the nifH sequences they amplify. In microcosms of O. longistaminata grown in the phytotron in the presence or absence of constant low nitrogen input (25 kg NH4NO3 ha(-1) year(-1)), roots of nitrogen-treated plants showed similar, slightly higher levels of nifH-mRNA. However, nitrogen treatment had a strong effect on the composition and diversity of expressed nifH pools that shifted towards methylotroph-related nitrogenases. Thus the active population of diazotrophs was not resistant towards low rates of nitrogen input and decreased significantly in richness, as also observed for plant species richness in grasslands by others.