Specificity and Phenetic Relationships of Iron- and Manganese-containing Superoxide Dismutases on the Basis of Structure and Sequence Comparisons*

The iron- and manganese-containing superoxide dismutases (Fe/Mn-SOD) share the same chemical function and spatial structure but can be distinguished according to their modes of oligomerization and their metal ion specificity. They appear as homodimers or homotetramers and usually require a specific metal for activity. On the basis of 261 aligned SOD sequences and 12 superimposed x-ray structures, two phenetic trees were constructed, one sequence-based and the other structure-based. Their comparison reveals the imperfect correlation of sequence and structural changes; hyperthermophilicity requires the largest sequence alterations, whereas dimer/tetramer and manganese/iron specificities are induced by the most sizable structural differences within the monomers. A systematic investigation of sequence and structure characteristics conserved in all aligned SOD sequences or in subsets sharing common oligomeric and/or metal specificities was performed. Several residues were identified as guaranteeing the common function and dimeric conformation, others as determining the tetramer formation, and yet others as potentially responsible for metal specificity. Some form cation-π interactions between an aromatic ring and a fully or partially positively charged group, suggesting that these interactions play a significant role in the structure and function of SOD enzymes. Dimer/tetramer- and iron/manganese-specific fingerprints were derived from the set of conserved residues; they can be used to propose selected residue substitutions in view of the experimental validation of our in silico derived hypotheses.

The iron-and manganese-containing superoxide dismutases (Fe/Mn-SOD) share the same chemical function and spatial structure but can be distinguished according to their modes of oligomerization and their metal ion specificity. They appear as homodimers or homotetramers and usually require a specific metal for activity. On the basis of 261 aligned SOD sequences and 12 superimposed x-ray structures, two phenetic trees were constructed, one sequence-based and the other structurebased. Their comparison reveals the imperfect correlation of sequence and structural changes; hyperthermophilicity requires the largest sequence alterations, whereas dimer/tetramer and manganese/iron specificities are induced by the most sizable structural differences within the monomers. A systematic investigation of sequence and structure characteristics conserved in all aligned SOD sequences or in subsets sharing common oligomeric and/or metal specificities was performed. Several residues were identified as guaranteeing the common function and dimeric conformation, others as determining the tetramer formation, and yet others as potentially responsible for metal specificity. Some form cation-interactions between an aromatic ring and a fully or partially positively charged group, suggesting that these interactions play a significant role in the structure and function of SOD enzymes. Dimer/tetramer-and iron/manganese-specific fingerprints were derived from the set of conserved residues; they can be used to propose selected residue substitutions in view of the experimental validation of our in silico derived hypotheses.
Aerobic organisms have developed mechanisms to protect against reactive oxygen intermediates arising from oxidative processes. The superoxide dismutase (SOD) 1 metalloenzymes (EC 1.15.1.1) constitute an example of such a defense against oxidative damage (1,2). They catalyze the degradation of toxic superoxide radicals to oxygen and hydrogen peroxide (3)(4)(5)(6)(7). They are subdivided into two structurally distinct families: (i) copper/zinc-containing SODs (Cu/Zn-SODs) that use copper and zinc simultaneously in their active sites and are found in eukaryotes and bacteria; and (ii) iron/manganese-containing SODs (Fe/Mn-SODs) that bind specifically either Fe or Mn (8). Fe-SOD is found in prokaryotes, chloroplasts, and protozoans, and Mn-SOD is in both prokaryotes and mitochondrial matrices.
Iron and manganese SODs exhibit a high degree of sequence and structure similarity, strongly suggesting that these enzymes originate from a common ancestry. Each monomer adopts a similar ␣/␤-fold, which combine to form a dimeric or tetrameric structure in solution (9,10). Despite the similarity of the three-dimensional molecular environment around the metal (7,11,12), these proteins generally require a specific metal ion for activity (13)(14)(15); only a small number of SODs, called cambialistic, are functional with both iron or manganese. Despite several attempts to explain the strict metal specificity (10, 12, 16 -21), its precise sequence and structural basis remain to be identified.
In the present paper, we concentrate on the known Fe/Mn-SOD structures. A phenetic analysis of the Fe/Mn-SOD family is achieved on the basis of three-dimensional structure superimpositions and compared with sequence-based phylogenetic investigations. The conservation of structure and sequence patterns within subsets of SOD sequences is systematically explored with the aim of identifying the reasons underlying the oligomerization characteristics and metal specificity. The results of our research directly apply to function identification in the context of genome sequencing and to knowledge-based modeling of Fe/Mn-SOD enzymes in the framework of structural genomics.

MATERIALS AND METHODS
Sequence and Structure Data-Our analysis is based on 261 Fe/Mn-SOD sequences listed in Table II in the supplemental material (available in the on-version of this article). The oligomerization state and the type of catalytic metal ion of each SOD were assigned according to the □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental material in the form of Fig. 4 (a depiction of the r.m.s. deviation of heavy main chain atoms after coordinate superimposition of the 66 pairs of SOD structures) and data base annotations and the literature. We identified 156 dimers (71 iron, 72 manganese, and 13 cambialistic SODs) and 105 tetramers (20 iron, 82 manganese, and 3 cambialistic SODs). We would like to stress that this assignment is quite a delicate point. Indeed, neither the metal specificity nor the oligomeric state exhibits all-or-none behavior. For example, some SOD enzymes are dimeric or tetrameric, depending on the experimental conditions (22). Other SODs can bind either iron or manganese according to the metal available in the culture medium (23)(24)(25)(26)(27). The existence of such SODs, named cambialistic, render unambiguous metal ion attribution difficult, especially considering that they may be cambialistic to various degrees. As a matter of fact, only a few SODs have had their preferred metal experimentally identified. Most attributions were made on the basis of sequence similarity using BLAST (28) and pairwise sequence comparisons.
Sequence-and Structure-based Similarity Trees-Two distance matrices were computed on the basis of the sequence and structure similarities, respectively, of the 12 known SOD structures. The former was obtained with the ProtDist program of the PHYLIP package (29) using the JTT substitution model (30). For the latter, we performed pairwise superimpositions of three-dimensional structures of apoenzymes with the SoFiSt program (31). This program is designed to yield optimal superimpositions with respect to the root mean square (r.m.s.) deviation of heavy main chain atoms after coordinate superimposition, generally considered to be the most reliable measure for comparing closely related structures. It was applied in two steps. First, the subset of secondary structure elements that are the most similar in the two structures, i.e. that superimpose with, at most, 2-Å r.m.s. deviation, was identified. The aligned segments were then extended in such a way that all residues except those corresponding to insertions and deletions are aligned. The r.m.s. deviations of the pairwise superimpositions so obtained were used as entries of the second distance matrix.
A hierarchical classification (tree) was obtained from each distance matrix using the unweighted pair-group method with arithmetic averages (UPGMA), the most widely utilized algorithm for generating hierarchical classifications (32). The reliability of this classification was assessed on the basis of a jackknife test (33) so as to identify the reliable nodes (34,35). The trees were drawn using the TreeView program (36).
Sequence Conservation-From the multiple sequence alignment of 261 SOD proteins, we identified residues that were conserved at a given position in at least 90% of all SOD protein sequences. Simultaneously, we searched for residues that are specific to the oligomeric state (dimer or tetramer) and/or type of metal (iron or manganese). More precisely, we defined four subgroups (iron dimer, iron tetramer, manganese dimer, and manganese tetramer) and looked for residues that occur in at least 80% of the proteins of one or several subgroups and in, at most, 20% of the proteins of each of the remaining subgroups. A gap in the sequence was considered to be a 21st residue. We also searched for conserved residue properties, e.g. aromatic, aliphatic, and charged, with a stricter threshold of 90%.
Analysis of Structure Features-Secondary structure assignments were defined according to the DSSP program (37), and hydrogen bonds were determined using HBPLUS (38). Cation-and amino-interactions were defined geometrically by a distance and an angle criterion (39). For simplicity, both interactions are here referred to as cation-. Amino acids in the vicinity of the metal ion were identified as having a side chain atom at less than 6.5 Å from the metal. Buried residues were defined as having Ͻ10% side chain solvent accessibility in the dimeric form. Solvent accessibilities were computed with NACCESS (40). Residues at the dimeric or tetrameric interface were defined as having a side chain solvent accessibility of 25% at least in the monomeric or dimeric form, and losing Ͼ60% of their solvent accessibility upon complex formation.

RESULTS
Phenetic Classifications-The 12 SOD structures present a pairwise structural similarity ranging from 0.4 to 5.2 Å r.m.s. deviation, with a sequence identity between 25.4% and 85.4%. On the basis of the distance matrices defined from pairwise sequence and structure similarities (Table I), two phenetic trees were inferred and are depicted in Fig. 1, a and b. Note that the sequence-based tree can be viewed as phylogenetic, as its construction involves an evolutionary model (41). To test the reliability of the trees, a jackknife procedure was applied (see "Materials and Methods"), which showed that all of the nodes of the three-dimensional structure-based tree ( Fig. 1b) are reliable, whereas there are two unreliable nodes in the sequencebased tree (Fig. 1a).
The connectivities of the two trees show marked differences, which can be considered as significant despite the limited number of SOD structures. First, the structure-based tree makes a clear-cut distinction between dimer and tetramer SODs. Only the Mn-SOD of T. thermophilus (1mng) is grouped with dimers, but it is quite a special kind of tetramer. Indeed, structure superimpositions of complete tetramer units reveal three types of tetramer organization characterized by a solvent accessible surface area buried by the dimer-dimer interface of ϳ10,500 Å 2 (in the iron or cambialistic SODs 1b06, 1sss, 1ids, 1coj and 1bsm), 4500 Å 2 (in the Mn-SODs 1n0j and 1kkc), and Ͻ3000 Å 2 (in 1mng only). In fact, 1mng exhibits very few dimer-dimer contacts (Fig. 2a) and is a totally atypical tetramer that will be shown to possess all the sequence characteristics of the dimers.
The dimer/tetramer distinction is less clear-cut in the sequence-based tree, where the dimeric Mn-SOD 1d5n appears merged with the tetramers. Some dimer/tetramer distinction is nevertheless perceptible, considering that 1b06 and 1sss proteins are dimeric at room temperature (22) and that the node linking them to the main tree is unreliable according to the jackknife test.
The comparison of the two trees reveals the existence of significant structural variations between dimers and tetramers induced by less marked sequence alterations. These variations involve a shorter H1 helix in the dimers and a H2 helix sepa-  (Fig. 2a). They are accompanied by a modification of the solvent-accessible surface area buried upon dimer formation, which is Ͻ8000 Å2 in dimeric SODs and Ͼ8000 Å2 in tetrameric forms. A second noticeable feature of the structure-based tree is the segregation between manganese and iron enzymes among the dimeric SODs, with the cambialistic SOD (1qnn) located in between. A structural difference explaining this segregation implicates the H2a-H2b loop, which is longer in manganese than in iron dimers. The manganese/iron differentiation is less manifest in the tetramer subgroup. The sequence-based tree also distinguishes between Fe-and Mn-SODs to a certain extent, but the dimer/tetramer and iron/manganese classifications overlap.
In the sequence-based tree, the iron tetramer from A. pyrophilus (1coj), an extremely heat-stable enzyme occurring in a hyperthermophilic bacterium (42), appears very distant from all other SODs. This suggests that the sequence has evolved to ensure optimal heat resistance while maintaining the functional SOD structure.
The sequence-and structure-based trees thus yield distinct phenetic views of SOD enzymes, which is not surprising considering the limited correlation between sequence and structure similarity scores (as monitored by a correlation coefficient of 0.4; see Fig. 4 in the supplemental material that can be found in the on-line version of this article). Comparison of these trees allows us to identify features, such as the oligomeric state or the metal ion specificity in dimers, that are obtained by means of a minimum of sequence modifications but that induce sig-nificant structural changes. It also allows us to detect properties, such as hyperthermophilicity, that require, on the contrary, few structural changes but drastic sequence alterations.
Conserved Sequence and Structure Characteristics-Residues conserved in at least 90% of all 261 available SOD sequences are indicated in Fig. 2b. Six are perfectly conserved, among which four ligand the metal ion (His 26 , His 73 , Asp 156 , and His 160 ; we use the PDB numbering of 1isa) and three occur near the dimer interface (His 160 , Glu 159 , and Tyr 163 ). The latter three residues ensure the conservation of two interchain interactions at the dimer interface, namely a 100% conserved double salt bridge between Glu 159 and His 160 (43) and an almost conserved double H-bond between His 30 and Tyr 163 (44).
Some other highly conserved residues are in the immediate environment of the metal cofactor (His 30 (45). Still other conserved residues are part of the protein core or can be expected to be structurally important. In particular, Asn 72 -Trp 122 form a cation-interaction, and so do Lys 107 -Trp 178 , whose level of conservation is, however, slightly lower than 90%. Residues Ile 96 , Leu 125 , Leu 133 , and Trp 183 contribute to hydrophobic packing. The conserved proline Pro 16 , always in cis-conformation, is located at the end of the N-terminal arm. This arm adopts, in all SODs, the same extended structure packed against the helices H1 and H2, ensured by the conserved residues Leu 7 , Ala 13 , Leu 14 , and Asn 39 . Some residues are conserved to maintain particular types of turns between successive secondary structures. Gly 101 has a positive value and is situated in the two-residue H3-H4 turn. In general, this turn is of type ␣GB␣ (G denotes left-handed helical conformation, B is an extended ␤-type conformation, and ␣ is an ␣-helix) with the sequence GS. In the thermophilic proteins 1mng and 1coj, however, it is an ␣GE␣ turn (E denotes a positive extended conformation) (46) with sequence GG. The conserved glycines Gly 119 and Gly 121 adopt a positive E-type extended conformation and are required for the particular ␣BBEBE␤ turn (47) between helix H4 and ␤-strand B1. Finally, the ␣EAB␣ turn linking the ␣-helices H5 and H6 FIG. 2. Structure alignments of SOD proteins, obtained with the SoFiSt program (31), using a maximum r.m.s. threshold value of 2 Å. The proteins are labeled by their protein code with annotations about the organism source, the oligomeric state in solution, and the metal cofactor specificity. The 1mng SOD is an atypical tetramer (designated 4Ј-mer) that has the sequence specificity of dimers. The secondary structure elements are indicated above the alignments. a, structural features and cationinteractions in the aligned SOD sequences. The ␣-helices are underlined in red, and ␤-strands are underlined in green. The symbol # marks the metal liganding residues, and the symbol $ marks residues in the close vicinity of the active center. In the last line, an asterisk indicates the consensus buried residues whose side chain accessibility in the dimeric form is Ͻ10%. Positions that are not structurally aligned with the 1isa structure are indicated in italics, residues involved in the dimer-dimer interface are in boldface, prolines in cis conformation are in red, residues with a positive left-handed helical conformation are in purple, and residues in extended conformation with a positive dihedral angle are in blue. Residues on colored backgrounds are involved in cation-interactions. Yellow background, Lys 107 -Trp 178 cation-interaction in 1isa numbering; pink, interchain Asn 65 -Phe 118 interaction; green, interchain Gln 118 -Phe 65 ; mauve: His 141 -Tyr 34 ; turquoise blue, Asn 72 -Trp 122 ; sky blue, His 27 -Trp 77 ; light brown, Gln 141 -Trp 158 ; dark brown, Gln 69 -Trp 158 . b, specific residue conservation in SOD sequences. The first line contains the PDB sequence numbering from the 1isa PDB file. Dark yellow background, residues conserved in 100% of SOD sequences; light yellow background, residues conserved in at least 90% of SOD sequences; blue background, residues specific to dimers (Thr 22 in 1isa numbering, the gap of at least seven residues following residue 51, and Asn 65 , Phe 118 , and Pro 144 ); purple background, tetramer-specific (Phe 65 ); orange background, manganese-specific (Met 23 , Gly 68 , Gln 141 , and Asp 142 ); violet background, iron dimer-specific (Phe 64 , Ala 68 , Gln69, Phe 75 , and Ala 141 ); violet letters, specific for all but iron dimers (Gly 69 ); red background, manganese dimer-specific (Asp 19 , Arg 64 , Arg 117 , and Ser 137 ); red letters, specific for all but manganese dimers (Ser 19 , the gaps of at least 11 residues following residue 59, and the other proteins having gaps of, at most, four residues); green background, iron tetramer-specific (Leu 72 and His 141 ); green letters, specific for all but iron tetramers (Asn 72 and aromatic position 124). Note that the results are derived from the 261 SOD sequences and represented on the 12 SOD structures; therefore, some residues appearing as conserved in the structures are not colored, as they are not conserved enough among all sequences. contains two conserved residues, namely Asn 168 close to the dimer interface and, at position 170, a positively charged residue located near the substrate funnel entry (45). This particular loop seems thus to be necessary for the enzymatic function.
Dimer/Tetramer-specific Sequence and Structure Characteristics-Several residues are conserved in subsets of the 261 SOD sequences, characterized by specific oligomeric and/or metal binding properties. They are listed in Fig. 2b and depicted in Fig. 3.
Residues Thr 22 , Asn 65 , Phe 118 , and Pro 144 are systematically encountered in dimers and never in tetramers, whereas Phe 65 is conserved among tetramers. Tetramers also exhibit a shorter loop, by one residue at least, at the end of helix H1. The larger number of dimer-specific residues compared with tetramerspecific residues is not surprising, as several kinds of tetramers are merged in the tetramer subset (see above).
In dimers, Asn 65 and Phe 118 are linked by an interchain cation-interaction (Fig. 2a) across the dimer interface situated not far from the entrance of the main substrate channel. In tetramers, these two positions are mutated simultaneously, with Asn 65 always substituted by Phe and Phe 118 often by Gln; in the latter case, a cation-interaction is formed again. The conserved Pro 144 in dimers introduces a short 3 10 helix located in the loop between the ␤-strands B2 and B3. Thr 22 is situated between the metal-liganding residue His 26 and residue Ser 19 or Asp 19 , which has been suggested as forming the entrance of an alternative pathway, allowing the substrate to reach or leave the functional site (44).
Manganese-specific Sequence and Structure Characteristics-Four manganese-specific residues are conserved in both dimers and tetramers. Met 23 is near the entry of the potential alternative channel to the functional site; Gly 68 , Gln 141 , and Asp 142 are spatially close and situated near the dimer interface. Note that Asp 142 frequently forms a salt bridge with residue 64, which is usually positively charged, especially in manganese dimers. Gly 68 is flanked by Gly 69 (also present in iron tetramers) and often by Gly 67 . This GGG pattern occurs in the middle of helix H2, an unusual feature that can be expected to locally weaken the structure and is required by packing constraints. The last manganese-specific residue, Gln 141 , forms a cation-interaction along the dimer interface with Trp 158 , which is conserved in almost all SODs and situated close to the metal ion.
No residues are specific to manganese tetramers, whereas Asp 19 , Arg 64 , Arg 117 , and Ser 137 are specific to manganese FIG. 3. Ribbon views of typical SOD structures. Dimer Fe-SOD 1isa (a), dimer Mn-SOD 1d5n (b), tetrameric Fe-SOD 1b06 (c), and tetrameric Mn-SOD 1n0j (d) are depicted. Iron and manganese atoms are represented in red and orange, respectively. Ribbons in the foreground dimeric unit are in sable color, whereas, for the tetrameric protein the background ribbons are gray. In the right part of the dimeric unit, the residues involved in conserved cationinteractions are labeled and depicted according the color convention of Fig. 2a; in the left part, the specific residue conservations in SOD sequences are shown and colored as in Fig. 2b. See the Fig. 2 legend for more details.
dimers. Arg 64 , in a salt bridge with the manganese-specific residue Asp 142 , faces Arg 117 across the dimer interface and flanks the dimer-specific Asn 65 -Phe 118 cation-interaction. Ser 137 and Arg 117 are spatially close but do not seem to be involved in structure or function. Asp 19 is situated at the entry of the possibly alternative funnel toward the active site (44); all other SODs possess a serine at that position. Finally, manganese dimers display a large insertion between helices H2a and H2b, which is located in the vicinity of the tetramer interface in the tetrameric SODs.
Iron-specific Sequence and Structure Characteristics-Fe-SODs have no conserved features, but iron dimers and iron tetramers do. The iron dimer-specific residues are Phe 64 , Ala 68 , Gln 69 , Phe 75 , and Ala 141 . Phe 64 is next to Asn 65 , which forms the dimer-specific interchain cation-interaction Asn 65 -Phe 118 ; it occupies the same position as Arg 64 in manganese dimers. Residue Phe 75 is in helix H2b and points in the direction opposite to the metal; it is stacked against an aromatic residue at position 71, located one helix turn ahead. Gln 69 is near the metal ion and forms a cation-interaction with Trp 158 , which is known to be involved in the reactivity (51); note that, in manganese enzymes, Trp 158 forms an alternative cation-interaction with Gln 141 , whose side chain fills the place occupied by the Gln 69 side chain in iron dimers. Steric reasons impose the presence of small residues at positions 68 and 141 when the cation-interaction Gln 69 -Trp 158 is formed; this explains the conservation of Ala 68 and Ala 141 .
There are only two conserved residues in iron tetramers, Leu 72 and His 141 ; moreover, these proteins never present an aromatic residue at position 124, probably due to structural constraints. Leu 72 is positioned just before one of the metalliganding His residues; in all but the iron tetramers this position is occupied by an Asn residue. His 141 forms a cationinteraction with the conserved Tyr 34 located at the entry of the main substrate access funnel. In addition, His 141 is hydrogen bonded with the OH of Tyr 34 and with a metal-bound water molecule (H 2 O or OH Ϫ , depending on the oxidation state of the metal ion). The same two H Ϫ bonds are observed for Gln 141 in Mn-SODs and Gln 69 in iron dimers. Not surprisingly, therefore, residues at positions 69 and 141 are largely described as influencing the metal specificity (11, 20, 48 -50). The nature of the residues at these positions is thought to play an important role in the catalytic fine tuning of the enzyme, and their mutation induces large effects on the catalytic activity. DISCUSSION Several explanations have been proposed to understand the strict metal ion specificity of SOD proteins (10, 12, 16 -20). The most convincing is from Vance and Miller (19,21), who showed that the active site environment of E. coli SOD induces redox potential tuning that is appropriate for reactions with one type of metal ion only. However, the chemical groups responsible for the redox tuning effects have not yet been fully identified. This task is quite complicated, given that metal selectivity and specificity is not overruled by a single residue but by the synergetic effects of several key groups.
The systematic investigation of conserved sequence and structure characteristics of SOD enzymes performed in this paper led us to recover and specify known features, but it also revealed new ones. The relevance of these characteristics in explaining oligomeric states or metal specificities should first, of course, be confirmed computationally using semi-empirical approaches and, ultimately, by experimental means through, for instance, site-directed mutagenesis. Our results lead us to propose a concrete list of single-site or concerted mutations to be tested, which can be deduced from Fig. 2b and involve residues that either appear to guarantee the common SOD structure or function or to modulate the quaternary structure or metal specificity.
This list of residues is not meant to be exhaustive; other residues also influence structure or function. For example, position 154, located ϳ10 Å away from the active site, has recently been shown to affect metal specificity (52). It is occupied by a Gly in 74% of the Mn-SODs and by a Thr in 72% of the Fe-SODs. According to our criteria, which require 80% conservation at least, these residues are not considered as representing typical manganese/iron characteristics. The fact that they, nevertheless, play a role demonstrates that specificity is achieved by an ensemble of residues of which some are situated far from the active site. The latter observation supports our prediction of specificity influencing residues not situated in the neighborhood of the metal ion.
Seven cation-interactions between aromatic and (partially) charged groups were identified as being particularly well conserved (Fig. 2a) and are thus suspected to play important structural and functional roles. Besides their obvious structural role, it can be argued that these cation-interactions play a functional role by fixing the exact positions of residue side chains in the vicinity of the metal ion and probably by tuning the redox potential of the metal ion by exploiting the electronic properties of aromatic amino acids. For example, the His 141 -Tyr 34 pair, which only forms when the histidine is protonated and increases the pK a (53), is especially likely to play an active role in the function. Indeed, it is situated at the entry of the main pathway to the active site, where it can be suggested to play the role of a gate (17,45,49,54,55) and even to be involved in the catalysis as a proton donor.
The dimer interface generates two symmetrical substrate funnels that lead from the bulk solvent to the metal ions (45). There are several conserved side chains found to line this funnel; others close its bottom and seem to prevent substrates from reaching the metal ion of the other monomer. Note that the metal-specific residues are situated near the metal and along the channel entrance, not at its bottom, which is consistent with their role in tuning specificity.
The existence of an alternative substrate access channel has been proposed with the aim of explaining the rapid turnover, which is incompatible with a single pathway toward the active site (44). This alternative channel has been suggested as being situated at the interface between the N-terminal helical domain and the C-terminal ␣/␤ domain of each monomer. The conserved residues Asp 19 in manganese dimers, Ser 19 in all but manganese dimers, Thr 22 in dimers, and Met 23 in manganese enzymes are situated along this alternative pathway and, hence, tend to support its very existence.
The present analysis also reveals that SOD dimers, whether binding iron or manganese, have common sequence and structure characteristics that differentiate them from tetramers. In contrast, tetramers only present common features after subdivision into iron-and manganese-specific enzymes. This can be taken to mean that iron and manganese dimers have recently evolved from a common ancestor, whereas the common ancestor of iron and manganese tetramers dates from the very remote past.
The T. thermophilus SOD (1mng) presents quite an atypical behavior. Despite the experimental evidence indicating its tetrameric behavior in solution (56 -61), it has all the sequence and structural features of typical dimers (see Figs. 1 and 2). The few contacts across the tetramer interface, moreover, indicate quite a loose tetrameric packing. This apparent contradiction probably reflects the possibility that the oligomeric state of 1mng depends on the experimental conditions such as temperature, pH, protein concentration, or ionic strength.
More generally, the ensemble of residues that we identified as ensuring the metal specificity and/or oligomeric state of SOD enzymes, summarized in Fig. 2b, can be used to define manganese/iron-and dimer/tetramer-specific fingerprints. If a given sequence presents some deviations from the typical fingerprints, it can be thought to adopt alternative oligomeric states or to have cambialistic tendencies. Large deviations can even be taken to indicate possible misannotations in the databases.