SPR-1 from S. proteamaculans (UniProt code: A8GGW7) was used as query sequence for a Conserf protein database search [http://consurf.tau.ac.il /verify.php].

Multiple sequence alignments including representatives from the B1, B2 and B3 subgroups of the MBL family were carried out using ClustalW2.1 [http://www.ebi.ac.uk/Tools/msa/clustalw2]. The alignment also includes SPR-1 and its only close homologue, CSA-1, which was discovered in this study. Sequence identities and similarities were also determined from pairwise sequence comparisons using ClustalW software.

A phylogenetic tree based on the multiple sequence alignment was generated using the program PhyML 3.0 [16]. Selected bootstrap numbers, which indicate the statistical reliability of the branching, with 1000 bootstrap replicates, are shown. The sequence distances can be estimated from the bar underneath the tree.

Using SPR-1 as query the database search revealed only one close relative, a sequence from the Gram-negative pathogenic bacterium Cronobacter sakazakii [17]. C. sakazakii can cause meningitis and necrotizing enterocolitis in particular in infants with a mortality rate estimated as high as 80% [18]. To our knowledge no cases of antibiotic resistance has yet been reported. This putative MBL from C. sakazakii, CSA-1, shares approximately 60% sequence identity with SPR-1 (Table 1).

Among the MBLs of the well-established subgroups the B3-type ones are the closest relatives with ~30% sequence identity (Table 1). This is not surprising since SPR-1 was initially identified due to its homology to B3 MBLs [15]. In contrast, the homology between CSA-1 and members of the B1 and B2 subfamilies is generally less than 20% sequence identity (Table 1). Also included in the comparison are two more distantly related enzymes with functions that differ greatly from that of MBLs. Methyl-paraoxonhydrolase (MPH) is an organophosphate pesticide-degrading enzyme [19], while glyoxalase II is an enzyme involved in the detoxification

process of methylglyoxal and other reactive aldehydes produced in metabolism [20]. Both share less than 15% sequence identity with CSA-1. In Figure 3 a portion of a multiple sequence alignment involving all the enzymes listed in Table 1 is shown. Focus is on amino acid residues that are relevant for binding of the catalytically essential metal ions (i.e. Zn²⁺ ions). These metal ion-coordinating residues are spread across four short sequence motifs. Within the three subgroups a high degree of conservation is observed. For the B1-type MBLs residues His116, His118 and His196 form one metal binding site (i.e. the M1 site), while Asp120, Cys221 and His263 form the M2 site. In B2-type MBLs His116 in the M1 site is replaced by an asparagine while the M2 site is unaltered. In contrast, B3-type MBLs have an M1 site identical to that of B1 MBLs but Cys221 is replaced by a histidine (i.e. His121) [5]. MPH and glyoxalase II have metal binding sites identical to those in B3 MBLs plus an additional metal ion-bridging ligand (i.e. Asp221; Figure 3). In contrast, SPR-1 was found to differ significantly from all known MBLs [15]. In its M1 site an arginine is proposed to replace His118, while the M2 site is even more diverse. No crystal structure of SPR-1 is yet available, but based on a homology model only Asp120 may be conserved. The remaining ligands may be Gln121 and Asn262, thus replacing Cys221/His121 and His263 observed in B1 and B3 MBLs, respectively. However, Ser221 instead of Cys221/His121 may also be a possible alternative. The unusual composition of the metal ion

binding site in the active center of SPR-1 is also reflected in its metal ion requirement. In the resting form only one Zn²⁺ may be present; the second metal ion may bind only when substrate is added [15]. It was thus speculated that SPR-1 activity is tightly controlled [15], possibly as a means to minimize undesired side reactions. This hypothesis also suggests that the in vivo function of SPR-1 may be diverse or even different from that of common MBLs. Insofar, the discovery of an SPR-1-like protein in only one other microorganism is significant; the C. sakazakii enzyme has active site residues identical to those of SPR-1 (Figure 3), indicating that these unusual MBLs may have evolved from a common ancestor separate from that of other MBLs.

In order to assess the evolutionary origin of SPR-1 and CSA-1 a phylogenetic analysis was carried out (Figure 4). MPH and in particular Glyoxalase II, the two enzymes that have a MBL overall fold but have biological functions different from MBLs are the most diverged sequences. Especially in the case of Glyoxalase II the fold similarity may be a reflection of convergent evolution rather than sequence divergence from a common ancestral gene. It is also evident that B1 and B2 MBLs are closely related (indicated by the 96% statistical (i.e. bootstrap) reliability). In contrast, B3 MBLs form a separate cluster. As expected from the pairwise sequence comparisons (Table 1) SPR-1 and CSA-1 cluster with B3 MBLs but form a separate group with high statistical reliability (bootstrap score of 90%). In fact, the separation of these unusual MBLs from the B3 MBLs is of a similar statistical reliability as that of the B2 MBLs from the B1 group (91%; Figure 4). It should also be noted

that the sequence distance between SPR-1/CSA-1 and the B3 subgroup is similar to that between the B1 and B2 subgroups. Hence, it is very likely that SPR-1 and CSA-1 originated from the same ancestral sequence as the well characterized MBLs from the B1, B2 and B3 subgroups, but based on the variations in both the overall sequence homology and the active site (i.e. metal binding site) residues (Figure 4) it is plausible to propose that these two unusual MBLs constitute a separate subgroup. We thus suggest the introduction of a new subgroup, B4, to encompass MBLs such as SPR-1 and CSA-1.