Abstract Although elevation of CO 2 has been reported to impact soil microbial functions, little information is available on the spatial and temporal variation of this effect. The objective of this study was to determine the microbial response in a northern Colorado shortgrass steppe to a 5-year elevation of atmospheric CO 2 as well as the reversibility of the microbial response during a period of several months after shutting off the CO 2 amendment. The experiment was comprised of nine experimental plots: three chambered plots maintained at ambient CO 2 levels of 360 μmol mol −1 (ambient treatment), three chambered plots maintained at 720 μmol mol −1 CO 2 (elevated treatment) and three unchambered plots of equal ground area used as controls to monitor the chamber effect. Elevated CO 2 induced mainly an increase of enzyme activities (protease, xylanase, invertase, alkaline phosphatase, arylsulfatase) in the upper 5 cm of the soil and did not change microbial biomass in the soil profile. Since rhizodeposition and newly formed roots enlarged the pool of easily available substrates mainly in the upper soil layers, enzyme regulation (production and activity) rather than shifts in microbial abundance was the driving factor for higher enzyme activities in the upper soil. Repeated soil sampling during the third to fifth year of the experiment revealed an enhancement of enzyme activities which varied in the range of 20–80%. Discriminant analysis including all microbiological properties revealed that the enzyme pattern in 1999 and 2000 was dominated by the CO 2 and chamber effect, while in 2001 the influence of elevated CO 2 increased and the chamber effect decreased. Although microbial biomass did not show any response to elevated CO 2 during the main experiment, a significant increase of soil microbial N was detected as a post-treatment effect probably due to lower nutrient (nitrogen) competition between microorganisms and plants in this N-limited ecosystem. Whereas most enzyme activities showed a significant post-CO 2 effect in spring 2002 (following the conclusion of CO 2 enrichment the previous autumn, 2001), selective depletion of substrates is speculated to be the cause for non-significant treatment effects of most enzyme activities later in summer and autumn, 2002. Therefore, additional belowground carbon input mainly entered the fast cycling carbon pool and contributed little to long-term carbon storage in the semi-arid grassland.