The viral serpin, crmA, is distinguished by its small size and ability to inhibit both serine and cysteine proteases utilizing a reactive loop shorter than most other serpins. Here, we characterize the mechanism of crmA inhibition of serine proteases and probe the reactive loop length requirements for inhibition with two crmA reactive loop variants. P1 Arg crmA inhibited the trypsin-like proteases, thrombin, and factor Xa, with moderate efficiencies (approximately 10(2)-10(4) M(-1)sec(-1)), near equimolar inhibition stoichiometries, and formation of SDS-stable complexes which were resistant to dissociation (k(diss) approximately 10(-7) sec(-1)), consistent with a serpin-type inhibition mechanism. Trypsin was not inhibited, but efficiently cleaved the variant crmA as a substrate (k(cat)/K(M) of approximately 10(6) M(-1) sec(-1)). N-terminal sequencing confirmed that the P1 Arg-P1'Cys bond was the site of cleavage. Altering the placement of the Arg in a double mutant P1 Gly-P1'Arg crmA resulted in minimal ability to inhibit any of the trypsin family proteases. This variant was cleaved by the proteases approximately 10-fold less efficiently than P1 Arg crmA. Surprisingly, pancreatic elastase was rapidly inhibited by wild-type and P1 Arg crmAs (10(5)-10(6) M(-1)sec(-1)), although with elevated inhibition stoichiometries and higher rates of complex dissociation. N-terminal sequencing showed that elastase attacked the P1'Cys-P2'Ala bond, indicating that crmA can inhibit proteases using a reactive loop length similar to that used by other serpins, but with variations in this inhibition arising from different effective P2 residues. These results indicate that crmA inhibits serine proteases by the established serpin conformational trapping mechanism, but is unusual in inhibiting through either of two adjacent reactive sites.