The origin of soil protease in field soil was estimated using culture-dependent and independent approaches. Overall soil protease activity was much higher in field soils with an annual application of liquid livestock feces (120 t ha−1 year−1 and 600 t ha−1 year−1) compared with the activity recorded in other field soils, and the character of the soil proteases became highly homogeneous (approximately 70% metalloprotease in a 600 t field). Selective incubation studies suggested that bacteria were the most important source of soil protease. There were significantly higher correlations between serratial metalloprotease and the overall soil protease in both feces-applied fields in terms of the effect of inhibitors, and the bacteria, which produced serratial metalloprotease, were suggested to proliferate in both the 120 t and 600 t fields. The gene homologous to serratial metalloprotease gene was amplified in directly extracted DNA from field soils using selective DNA primer and proteolytic Serratia marcescens was certified to be one source of soil protease in these field soils. Proteolytic S. marcescens and its metalloprotease gene have occasionally been isolated and detected in field soils applied with raw feces, and have rarely been isolated or detected from other field soils. Proteolytic S. marcescens is believed to be introduced in the raw feces and subsequently colonizes the field soil and replaces the indigenous bacteria in the soil.