1. Introduction

Metal atoms or ions occur widely in association with proteins and have a variety of functions. In some cases the metal is part of the active site for a catalytic process; in others the metal appears to play a role in maintaining structure. Knowledge and understanding of the architecture of protein molecules (see, for example, Lesk, 2001) play a key role in understanding their function. In a similar way, knowledge of the architecture of different metal coordination groups within proteins is important in addition to an understanding of the different chemical behaviour of the metals. The Biological Chemistry of the Elements (Frausto da Silva & Williams, 1991) provides an excellent account of both the chemical behaviour of the different metals and the biological significance of metal coordination groups.

Two metal coordination groups are illustrated in Fig. 1. The first aspects of interest are the number and nature of the donor groups around the metal atom or ion, the metal-to-donor atom distances and the angles between metal-donor bonds. In very many cases the protein molecule is, in the coordination chemist's terminology, a multidentate ligand, so we are also interested in the number and nature of the amino-acid donor groups from the protein chain, their relative positions in the amino-acid sequence and the size and conformation of the resulting chelate rings. How does the protein-chain conformation adapt to the requirements of the metal coordination? We want to know what generalizations, if any, can be made about these different properties and how far they can be predicted. Vallee & Auld (1990) commented on the significance of the spacing between donor residues in 12 zinc enzyme structures and suggested how the observed long and short spacings contributed to effectiveness in catalytic function; a wealth of additional data is now available.

Figure 1 Schematic illustration of coordination groups; see text for definitions.

This paper presents a set of tables which allow comparisons of donor groups, chelate-loop sizes and conformations in 600 metal coordination groups. They are for the metals Ca, Mg, Mn, Fe, Cu, Zn, Na, K, the most commonly occurring metals in biological chemistry and the commonest in the Protein Data Bank [PDB (Bernstein et al., 1977); available through the RCSB (Berman et al., 2000)], which is the primary source of the information used here. For Na and K the borderline between a `coordination compound' and an electrostatic association of ions is certainly debatable, but regardless of the description of the bonding it is useful to describe the geometric situation around these ions as found in protein crystals. This study concentrates on coordination groups where amino-acid side chains or the main-chain carbonyl group provide donor atoms: haem groups, iron-sulfur clusters and chlorophyll derivatives have been excluded (there are some specialized articles about these, e.g. Parisini et al., 1999; Maher et al., 1999; Chong et al., 1999; see also Huber et al., 2001). In previous articles (Harding, 2000, 2001), data on coordination numbers and on metal-to-donor atom distances and angles in proteins have been gathered for these eight metals. For all the coordination groups in the tables presented here, the distances, angles and coordination group shape may be found at http://tanna.bch.ed.ac.uk/ ; for a more extensive range of metal coordination groups the Metalloprotein Database (MDB) is valuable (Castagnetto et al., 2002).

The definition of a coordination group used here requires the donor atoms to be within specified target distances of the metal atom; this definition is objective, but it is narrow and it excludes the second and third coordination shells around a metal atom that are generally considered to be important in enzyme activity (see, for example, Duda et al., 2003). Similarly, in building a library of structural motifs of metal coordination sites with catalytic activity, MacArthur & Thornton (2002) include functional groups substantially further from the metal than a simple bond distance; their motifs, including the three-dimensional coordinates of donor atoms, can be used as templates or probes for a systematic classification of sites.

The results and discussion given here are based entirely on a `representative set' of proteins, a set within which none has more than 30% sequence identity with any other. There are various difficulties in comparing the frequency of occurrence of different coordination groups or donor patterns in proteins using the PDB. The proteins whose structures have been deposited in the PDB are far from a random sample. The use of a `representative set' of proteins is a simple expedient to obtain a fairly diverse sample, but is based on sequence similarity of the whole protein chain, not just the part in the immediate vicinity of and more relevant to the metal coordination group. Furthermore, many protein crystals contain two or more copies of the protein molecule in the crystal asymmetric unit; allowance for this has been made in several different ways at different stages of this project. Even with these allowances, the set of proteins in the PDB is by no means a random selection; small differences in statistics of distributions should not be thought to be significant, only broad trends.

2. Some definitions

`Target distances' for different types of metal-donor atom bond were based on the distances observed in accurately determined small-molecule crystal structures (Harding, 2001, 2002). The metal coordination number is the number of donor atoms within the target distance + 0.75  Å; some of these are normally donor atoms from amino-acid side chains within the protein or the O atoms of main-chain carbonyl groups, but they may also include water-molecule O atoms or atoms from other non-protein small molecules present at the metal site.

Metal coordination groups can be drawn very schematically, as in Fig. 1. In this work, a donor atom is defined entirely on geometric criteria: it must be within the already established target distance + 0.75  Å of the metal atom. Single-letter amino-acid codes are used to specify the donor groups (of the protein) and O indicates main-chain carbonyl O atom as a donor. We thus describe the coordination group shown in Fig. 1(a) as CHCC Zn 2 18 3, since Zn is coordinated to the sulfur of cysteine (n) with sequence number n, N of histidine (n + 2), S of cysteine (n + 2 + 18) and S of cysteine (n + 2 + 18 + 3). The total coordination number (CN) is 4. The sequence differences, seqdif, in the three chelate loops are 2, 18 and 3. In the first chelate loop there are seven backbone atoms of the donor residues and the residue between, as well as atoms from both side chains, making a ring of 14 atoms altogether (if N of histidine coordinates). The relative sequence number of each donor amino acid in the coordination group is given by relseq; in this example there are cysteines at relseq = 0, 20 and 23, and histidine at relseq = 2. nspan is the sequence-number difference between the last and first amino-acid donors, which is the sum of all the seqdifs between them; nspan is 23 in this example. Many metal coordination groups also include water molecules or donor atoms from small non-protein molecules, for example as in Fig. 1(b). It has also been useful to look at the chelate loops, which are the building blocks of coordination groups, i.e. the adjacent pairs of donors, such as CH 2, HC 18, CC 3 in this example. For full identification of a particular coordination group or a chelate loop, the protein name and the residue number and chain letter of the first amino acid must be given, e.g. the above group occurs in 1a1i at A137 (and another at A165). In comparing the compositions of metal coordination groups it has been necessary to treat the carboxylate group as one donor whether it is monodentate or bidentate, since the distinction between these is unreliable in structures determined at lower resolutions.

3. Methods and procedures

The basis for generating the coordination-group tables is the program MP (Harding, 2001), which reads a PDB file, extracts the coordinates and occupancy of each metal atom and of all atoms within 3.6  Å of the metal atom and summarizes all the coordination information. Lists of PDB codes were obtained using the Jena Image Library search facility (http://www.imb-jena/ImgLibPDB/pages/hetDir/PSE2HET.shtml ) for structures containing each of the metals. From these lists protein and protein-nucleic acid complexes were selected with structures determined by diffraction to a resolution 2.5  Å and the program MP run for all that were available in the RCSB release of July 2001 (except that the July 2002 release was used for potassium proteins in order to augment the very small number of available structures). Additional smaller programs then gave information on coordination group descriptions for the full lists or for selections from them. One such selection is a `representative set' which excludes any structure which has more than 30% sequence identity with any other in the set; the culled PDB files of Dunbrack (2001) were used to make this selection.

4. Results and discussion

Tables for all eight metals are deposited as supplementary Table 1D. Table 1 illustrates some of the data stored for a small selection of Zn coordination groups; not shown here but also stored in these files are (i) the increase in coordination number corresponding to an increase in coordination sphere radius of 0.2  Å, (ii) water molecules and other non-protein donors in the coordination group, (iii) the EC enzyme number when it is given in the PDB, (iv) part of the header name from the PDB file, (v) the names in the PDB file of the metal and the first donor atom and (vi) the sequence of residue conformations in each of the chelate loops (in full when the loops contain up to five residues; abbreviated for larger loops). The tables can be downloaded and searched for particular coordination groups or other features and sorted or otherwise manipulated in, for example, Microsoft EXCEL. For each coordination group the metal-donor atom distances and bond angles can be found at http://tanna.bch.ed.ac.uk (or at http://metallo.scripps.edu/ ).

There is much diversity in the coordination groups and different metals have very different characteristics. The preferences of different metals for different amino-acid donors are shown in Table 2(a) and 2(b). Oxygen donors (carboxylate, amide, water etc.) are almost never found in the same coordination group as cysteine, although either may occur alongside histidine. Tables 2(c), 2(d) and 2(e) summarize, for Ca and Zn coordination groups, metal coordination numbers and chelate-loop sizes and Table 2(f) lists the most commonly occurring chelate loops for each; fuller details are deposited for these and all the other metals (supplementary Table 2D).

In Ca proteins the EF-hand (see Pidcock & Moore, 2001; Nelson & Chazin, 1998; see also http://structbio.vanderbilt.edu/cabp_database/ ) is a very dominant structural motif, with 27 examples of the coordination group DDDOE 2225 or its close relatives in this set of representative proteins, and for Zn the pattern CCCC 3 n 3 with n = 10-20 is common in zinc fingers and related proteins. Apart from these, identical coordination groups (same donors, same residue separation) do not often recur in these tables; when they do the proteins usually have related folds and functions, but even then the conformations may differ, especially in the larger chelate loops. Fig. 2 shows an example. A detailed study was made of all the recurring Ca and Zn coordination groups, their conformations, amino-acid sequences etc. and these are available in Table 3W (at http://tanna.bch.ed.ac.uk/arch/ ).

Figure 2 The coordination group Zn CCCC 3 3 8 showing its conformation in 1het_A 97 (green) and in 1e3j_A 96 (blue). The coordinating cysteines are labelled C96 in the blue chain, C97 in the green chain etc. 1het is an alcohol dehydrogenase; 1e3j is a ketose reductase. The backbone atoms of residues of the first two chelate loops, CCC 3 3, have been superposed using the program LSQKAB from the CCP4 suite (Collaborative Computational Project, Number 4, 1994); their conformations are the same, r.m.s. displacement 0.23  Å, whereas there are marked differences in the larger chelate loop, CC 8. This figure was prepared using MOLSCRIPT (Kraulis, 1991) and RASTER3D (Merritt & Murphy, 1994).

While whole coordination groups are not often repeated here, except for the Ca EF-hands, some chelate loops that are their components occur frequently in different unrelated proteins, although other chelate loops are found only once or a few times. Small chelate loops, particularly seqdif = 2, are very common for calcium, whereas for zinc seqdif = 3 and larger loops are much more common. In coordination groups with only two protein donors these donors are rarely more than ten residues apart, which is understandable on simple stability grounds. When there are three or more protein donor groups it is common for there to be at least one large loop. Large chelate loops will usually serve the function of holding two parts of the polypeptide chain close to each other; this may be at the active site or simply to provide stability for the whole structure. It is common for long and short chelate loops to alternate in the protein-chain sequence and uncommon for a long loop to follow another long loop; a short loop following another short loop is uncommon in zinc coordination groups, but common in calcium groups.