Among metallo-dependent phosphatases, ADP-ribose/CDP-alcohol diphosphatases form a protein family (ADPRibase-Mn-like) mainly restricted, in eukaryotes, to vertebrates and plants, with preferential expression, at least in rodents, in immune cells. Rat and zebrafish ADPRibase-Mn, the only biochemically studied, are phosphohydrolases of ADP-ribose and, somewhat less efficiently, of CDP-alcohols and 2´,3´-cAMP. Furthermore, the rat but not the zebrafish enzyme displays a unique phosphohydrolytic activity on cyclic ADP-ribose. The molecular basis of such specificity is unknown. Human ADPRibase-Mn showed similar activities, including cyclic ADP-ribose phosphohydrolase, which seems thus common to mammalian ADPRibase-Mn. Substrate docking on a homology model of human ADPRibase-Mn suggested possible interactions of ADP-ribose with seven residues located, with one exception (Cys253), either within the metallo-dependent phosphatases signature (Gln27, Asn110, His111), or in unique structural regions of the ADPRibase-Mn family: s2s3 (Phe37 and Arg43) and h7h8 (Phe210), around the active site entrance. Mutants were constructed, and kinetic parameters for ADP-ribose, CDP-choline, 2´,3´-cAMP and cyclic ADP-ribose were determined. Phe37 was needed for ADP-ribose preference without catalytic effect, as indicated by the increased ADP-ribose Km and unchanged kcat of F37A-ADPRibase-Mn, while the Km values for the other substrates were little affected. Arg43 was essential for catalysis as indicated by the drastic efficiency loss shown by R43A-ADPRibase-Mn. Unexpectedly, Cys253 was hindering for cADPR phosphohydrolase, as indicated by the specific tenfold gain of efficiency of C253A-ADPRibase-Mn with cyclic ADP-ribose. This allowed the design of a triple mutant (F37A+L196F+C253A) for which cyclic ADP-ribose was the best substrate, with a catalytic efficiency of 3.5´104 M-1s-1 versus 4´103 M-1s-1 of the wild type.