African humid savannas are highly productive ecosystems, despite very low soil fertility, where grasses and trees coexist. Earlier results showed that some perennial grass species are capable of biological nitrification inhibition (BNI) while trees likely influence differently on nitrogen cycling. Here we assessed the impact of the dominant grass and tree species of the Lamto savanna (Ivory Coast) on soil nitrifying and denitrifying enzyme activities (NEA and DEA, respectively) and on the abundances of archaeal and bacterial ammonia oxidizers (AOA and AOB, respectively) and nitrite reducers. This is one of the first studies linking nitrifying and denitrifying activities and the abundances of the involved groups of microorganisms in savanna soils. NEA was 72-times lower under grasses than under trees while AOA and AOB abundances were 34- and 3-times lower. This strongly suggests that all dominant grasses inhibit nitrification while trees stimulate nitrification, and that archaea are probably more involved in nitrification than bacteria in this savanna. While nitrite reducer abundances were similar between locations and dominated by nirS genes, DEA was 9-times lower under grasses than trees, which is likely explained by BNI decreasing nitrate availability under grasses. The nirS dominance could be due to the ferruginous characteristics of these soils as nirS and nirK genes require different metallic co-enzymes (Fe or Cu). Our results show that the coexistence of grasses and trees in this savanna creates a strong heterogeneity in soil nitrogen cycling that must be considered to understand savanna dynamics and functioning. These results will have to be taken into account to predict the feedbacks between climate changes, nitrogen cycling and tree/grass dynamics at a time when savannas face worldwide threats.