Portal de Periódicos da CAPES

Structure of the chromatin binding (chromo) domain from mouse modifier protein 1

1997; Springer Nature; Volume: 16; Issue: 9 Linguagem: Inglês

10.1093/emboj/16.9.2473

1460-2075

Linda Ball, Natalia V. Murzina, R. William Broadhurst, Andrew R. C. Raine, Sharon J. Archer, Francesca J. Stott, Alexey G. Murzin, Prim B. Singh, Peter J. Domaille, Ernest D. Laue,

Protein Degradation and Inhibitors

Article1 May 1997free access Structure of the chromatin binding (chromo) domain from mouse modifier protein 1 Linda J. Ball Linda J. Ball Cambridge Centre for Molecular Recognition, Department of Biochemistry, University of Cambridge, Tennis Court Road, Cambridge, CB2 1QW UK Search for more papers by this author Natalia V. Murzina Natalia V. Murzina Cambridge Centre for Molecular Recognition, Department of Biochemistry, University of Cambridge, Tennis Court Road, Cambridge, CB2 1QW UK Search for more papers by this author R.William Broadhurst R.William Broadhurst Cambridge Centre for Molecular Recognition, Department of Biochemistry, University of Cambridge, Tennis Court Road, Cambridge, CB2 1QW UK Search for more papers by this author Andrew R.C. Raine Andrew R.C. Raine Cambridge Centre for Molecular Recognition, Department of Biochemistry, University of Cambridge, Tennis Court Road, Cambridge, CB2 1QW UK Search for more papers by this author Sharon J. Archer Sharon J. Archer Du Pont Merck Pharmaceutical Company, Box 80328, Wilmington, DE, 19880-0328 USA Search for more papers by this author Francesca J. Stott Francesca J. Stott Cambridge Centre for Molecular Recognition, Department of Biochemistry, University of Cambridge, Tennis Court Road, Cambridge, CB2 1QW UK Search for more papers by this author Alexey G. Murzin Alexey G. Murzin Centre for Protein Engineering, MRC Centre, Hills Road, Cambridge, CB2 2QH UK Search for more papers by this author Prim B. Singh Prim B. Singh The Babraham Institute, Babraham Hall, Babraham, Cambridge, CB2 4AT UK Search for more papers by this author Peter J. Domaille Peter J. Domaille Du Pont Merck Pharmaceutical Company, Box 80328, Wilmington, DE, 19880-0328 USA Search for more papers by this author Ernest D. Laue Corresponding Author Ernest D. Laue Cambridge Centre for Molecular Recognition, Department of Biochemistry, University of Cambridge, Tennis Court Road, Cambridge, CB2 1QW UK Search for more papers by this author Linda J. Ball Linda J. Ball Cambridge Centre for Molecular Recognition, Department of Biochemistry, University of Cambridge, Tennis Court Road, Cambridge, CB2 1QW UK Search for more papers by this author Natalia V. Murzina Natalia V. Murzina Cambridge Centre for Molecular Recognition, Department of Biochemistry, University of Cambridge, Tennis Court Road, Cambridge, CB2 1QW UK Search for more papers by this author R.William Broadhurst R.William Broadhurst Cambridge Centre for Molecular Recognition, Department of Biochemistry, University of Cambridge, Tennis Court Road, Cambridge, CB2 1QW UK Search for more papers by this author Andrew R.C. Raine Andrew R.C. Raine Cambridge Centre for Molecular Recognition, Department of Biochemistry, University of Cambridge, Tennis Court Road, Cambridge, CB2 1QW UK Search for more papers by this author Sharon J. Archer Sharon J. Archer Du Pont Merck Pharmaceutical Company, Box 80328, Wilmington, DE, 19880-0328 USA Search for more papers by this author Francesca J. Stott Francesca J. Stott Cambridge Centre for Molecular Recognition, Department of Biochemistry, University of Cambridge, Tennis Court Road, Cambridge, CB2 1QW UK Search for more papers by this author Alexey G. Murzin Alexey G. Murzin Centre for Protein Engineering, MRC Centre, Hills Road, Cambridge, CB2 2QH UK Search for more papers by this author Prim B. Singh Prim B. Singh The Babraham Institute, Babraham Hall, Babraham, Cambridge, CB2 4AT UK Search for more papers by this author Peter J. Domaille Peter J. Domaille Du Pont Merck Pharmaceutical Company, Box 80328, Wilmington, DE, 19880-0328 USA Search for more papers by this author Ernest D. Laue Corresponding Author Ernest D. Laue Cambridge Centre for Molecular Recognition, Department of Biochemistry, University of Cambridge, Tennis Court Road, Cambridge, CB2 1QW UK Search for more papers by this author Author Information Linda J. Ball1, Natalia V. Murzina1, R.William Broadhurst1, Andrew R.C. Raine1, Sharon J. Archer2, Francesca J. Stott1, Alexey G. Murzin3, Prim B. Singh4, Peter J. Domaille2 and Ernest D. Laue 1 1Cambridge Centre for Molecular Recognition, Department of Biochemistry, University of Cambridge, Tennis Court Road, Cambridge, CB2 1QW UK 2Du Pont Merck Pharmaceutical Company, Box 80328, Wilmington, DE, 19880-0328 USA 3Centre for Protein Engineering, MRC Centre, Hills Road, Cambridge, CB2 2QH UK 4The Babraham Institute, Babraham Hall, Babraham, Cambridge, CB2 4AT UK The EMBO Journal (1997)16:2473-2481https://doi.org/10.1093/emboj/16.9.2473 PDFDownload PDF of article text and main figures. ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinked InMendeleyWechatReddit Figures & Info The structure of a chromatin binding domain from mouse chromatin modifier protein 1 (MoMOD1) was determined using nuclear magnetic resonance (NMR) spectroscopy. The protein consists of an N-terminal three-stranded anti-parallel β-sheet which folds against a C-terminal α-helix. The structure reveals an unexpected homology to two archaebacterial DNA binding proteins which are also involved in chromatin structure. Structural comparisons suggest that chromo domains, of which more than 40 are now known, act as protein interaction motifs and that the MoMOD1 protein acts as an adaptor mediating interactions between different proteins. Introduction The chromatin organization modifier (chromo) domain has been defined as a 30–70 residue protein module found in a large number of proteins. So far more than 40 examples of the chromo domain are known (Aasland and Stewart, 1995; Koonin et al., 1995) and the list is growing rapidly. Chromo domains were first identified in two Drosophila proteins, heterochromatin associated protein 1 (HP1) and Polycomb (Pc) (Paro and Hogness, 1991). HP1 has been shown to be involved in repression of gene expression in heterochromatin (Eissenberg et al., 1990, 1992) and Pc is a member of the Polycomb group of transcriptional repressors which suppress expression of homeotic genes (Paro, 1990, 1993). Studies of both Pc (Messmer et al., 1992) and HP1 (Platero et al., 1995) have shown that chromo domains are responsible for targeting their respective proteins to their sites of action on chromatin. Similarly, the chromo domain in human retinoblastoma binding protein (RBP1) may deliver retinoblastoma protein (Rb), a global transcription factor, to particular sites (Szekely et al., 1991; Koonin et al., 1995). In addition, Drosophila male-specific lethal (MSL-3) protein, an activator, might also be localized to chromatin via its chromo domain (Koonin et al., 1995). It is now apparent that the chromo domain family of proteins comprises at least two groups. The HP1-like proteins contain two conserved homologous domains (Clark and Elgin, 1992): an N-terminal chromo domain and a C-terminal ‘shadow’ chromo domain (Epstein et al., 1992; Aasland and Stewart, 1995). Mutational analysis has shown that the N-terminal chromo domain of HP1 is sufficient to direct heterochromatin binding of a β-galactosidase fusion protein in vivo, whereas the C-terminal shadow domain is necessary for nuclear localization and heterochromatin binding (Powers and Eissenberg, 1993; Platero et al., 1995). The two chromo domains of the HP1-like proteins may thus serve as adaptors bringing together different proteins in a multi-protein complex (Aasland and Stewart, 1995). In other cases, such as Pc, CHD-1 (Delmas et al., 1993) and SWI6 (Lorentz et al., 1994; Ekwall et al., 1995), the chromo domain is a component of a much larger protein and is typically fused to other protein modules (Aasland and Stewart, 1995). The HP1 gene was identified as a suppressor of position-effect variegation (PEV) (Eissenberg et al., 1990). PEV occurs when a gene normally residing in euchromatin is placed next to a heterochromatic boundary after a chromosomal rearrangement. The gene becomes specifically silenced in some cells, but not in others, leading to a variegated phenotype. Suppressors or enhancers of PEV are likely to encode proteins that are involved in the formation or modification of chromatin structure (for a review see Reuter and Spierer, 1992). PEV is similar in many ways to other mechanisms of gene silencing, such as those involving the homeotic genes (Paro, 1993), mating type control in yeast (Laurenson and Rine, 1992), X-chromosome inactivation (Gartler and Riggs, 1983) and imprinting in mammalian cells (Solter, 1988). In this paper we describe experiments that determined the domain structure of HP1-like heterochromatin-associated mouse chromatin modifier protein 1 (MoMOD1) (Singh et al., 1991). The N-terminal chromo domain (MoMOD1-N, see Figure 1) was expressed in Escherichia coli, purified to homogeneity and its solution structure determined using nuclear magnetic resonance (NMR) spectroscopy. Unexpectedly, the structure is similar to that of the Sac7d and Sso7d DNA binding proteins from the archaebacteria Sulfolobus acidocaldarius and S.solfataricus (Baumann et al., 1994; Edmondson et al., 1995). The structural comparisons suggest how the chromo domain may function as a protein interaction motif. Figure 1.Amino acid sequence of the MoMOD1 chromo domain showing its homology with other representative chromo domains. Selected conserved residues are coloured yellow (core hydrophobic), green (Gly and Pro) and blue (basic). See text for details. Download figure Download PowerPoint Results Expression of the intact MoMOD1 protein The full-length MoMOD1 protein (residues 1–185) was cloned into pET16b and expressed in E.coli BL21(DE3) as a histidine-tagged fusion protein. The protein was purified using nickel affinity chromatography, where it co-purified with small amounts of two shorter fragments of the protein. These may result from truncated transcription or translation or, perhaps, degradation by endogenous proteases. The shorter fragments did not interfere with interpretation of the limited proteolysis experiments presented below. Limited proteolysis of the intact MoMOD1 protein Before the proteolysis experiments were carried out, the stability of the protein was assessed under the conditions of the digest. This showed that the protein was not degraded. Digestion conditions were then varied over a wide range to determine which regions of the molecule are most susceptible to proteolysis. The proteolytic fragments were then separated using either tricine–SDS–PAGE (Figure 2a and b) or reversed phase HPLC (not shown) and were identified using N-terminal amino acid sequencing and electrospray ionization mass spectroscopy. The N-terminus of each fragment was determined directly by N-terminal sequencing, whereas the C-terminus was identified indirectly using mass spectroscopy and the statistical program WEIGHT (W.Boucher, unpublished data); this program assigns a series of possible protein sequences to each mass spectral fragment. A summary of the results is presented in Figure 2C. Figure 2.Limited proteolysis of MoMOD1: 16% tricine–SDS–PAGE separation of a 0.5% trypsin or 1% subtilisin digest of the MoMOD1 protein. (a) Lane 1, intact protein; lane 2, after 5 min incubation; lane 3, after 15 min incubation; lane 4, after 30 min incubation; lane 5, after 60 min incubation with subtilisin. (b) Lane 1, intact protein; lane 2, after 5 min incubation; lane 3, after 30 min incubation; lane 4, after 60 min incubation with trypsin. On the right of these panels, the results of the microsequencing and mass spectrometric analysis are summarized. The numbers represent the N- and C-terminal amino acids of the major components identified from the bands. (c) Amino acid sequence of the MoMOD1 protein showing the protease cut sites. The sequences of the protease-resistant N- and C-terminal domains are underlined. Up arrow, trypsin cleavage site; down arrow, subtilisin cleavage site. Download figure Download PowerPoint Expression of the MoMOD1-N chromo domain fragment Based on the sequence alignment (Figure 1) and limited proteolysis data (Figure 2), a fragment containing residues 10–80 of MoMOD1 was expressed in E.coli in milligram quantities. The N-terminal sequence was HMVEEVL . . ., which is as expected with the addition of a histidine and methionine residue from the expression vector. Amino acid analysis showed the expected amino acid composition and electrospray ionization mass spectra identified one major species with a molecular mass of 8684.01; the calculated molecular mass is 8683.54. NMR studies of 15N relaxation in the backbone amides (data not shown) show that the structured part of the protein has an overall rotational correlation time of 6.8 ns. In conjunction with the results of sedimentation analysis (data not shown), this is consistent with MoMOD1-N being monomeric in solution. NMR assignment strategy and location of the secondary structure elements The triple resonance assignment strategy employed for MoMOD1-N is similar to that outlined in Clowes et al. (1995). Assignments of the chemical shifts of 1H, 13C and 15N nuclei in the protein backbone were made using a combination of three-dimensional CBCA(CO)NNH and CBCANNH experiments (Grzesiek and Bax, 1992a,b) and confirmed using three-dimensional 15N-separated TOCSY-HMQC, NOESY-HMQC (Marion et al., 1989a,b) and HNHB (Archer et al., 1991) spectra. Aliphatic side chain 1H and 13C assignments were identified from four-dimensional HCC(CO)NNH (Clowes et al., 1993) and three-dimensional 13C-separated HCCH-TOCSY (Bax et al., 1990a) experiments. Aromatic side chain signals were assigned by reference to two-dimensional CBHD, CBHE (Yamazaki et al., 1993), DQF-COSY (Rance et al., 1983) and NOESY (Kumar et al., 1980) spectra. Through-space nuclear Overhauser effect (NOE) connectivities from backbone amide to side chain protons were identified from a four-dimensional 13C/15N-separated NOESY experiment (Kay et al., 1990). Amide–amide and side chain–side chain connections were obtained from three-dimensional 15N- and 13C-separated NOESY spectra (Marion et al., 1989b; Ikura et al., 1990; Zuiderweg et al., 1990). Vicinal coupling constants between amide and 1Hα spins, 3JαN, were determined by least squares fitting to the intensities of crosspeaks from a set of two-dimensional J–modulated 15N-1H-COSY experiments (Billeter et al., 1992), leading to estimates of the backbone angle φ by reference to a Karplus curve (Pardi et al., 1984). Slowly exchanging amide protons were identified by recording 1H-15N HSQC (Bax et al., 1990b; Norwood et al., 1990) spectra on a sample immediately after the H2O solvent had been exhanged for D2O. The elements of secondary structure were then identified from 1Hα, 13Cα and 13Cβ chemical shifts (Wishart and Sykes, 1994), patterns of short and intermediate range inter-residue NOEs, 3JαN couplings and amide exchange rates (Wüthrich, 1986), as illustrated by the connectivity summary in Figure 3. The presence of strong dαN NOEs, retarded amide exchange rates, positive values for the consensus chemical shift index (CSI) and large values of 3JαN indicate that there are β-strands from residues 23 to 31, 36 to 42 and 51 to 54. Between residues 60 and 71 the dαN(i, i + 3) and dαβ(i, i + 3) NOEs, the strong dNN and weak dαN NOEs, the slow amide exchange, the negative values for the consensus CSI and the small values of 3JαN reveal a C-terminal α-helix. Figure 3.Location of the secondary structure in MoMOD1-N. Below the classification of the elements of secondary structure, the CSI line indicates the consensus chemical shift index for 1Hα, 13Cα and 13Cβ nuclei. The stars below the amino acid sequence represent the location of slowly exchanging amide protons. In the next rows, filled and empty circles represent residues with 3JαN > 9 Hz and < 4Hz respectively and squares represent residues for which χ1 has been determined. Following this, three rows of solid bars represent the observed sequential dαN, dβN and dNN NOE connectivities; the thickness of the bars indicates the intensities of the crosspeaks in the NOESY spectra. Further bars represent dαN(i, i + 2), dNN(i, i +2), dαN(i, i + 3), dαβ(i, i + 3), dNN(i, i +3) and dαN(i, i + 4) NOE connectivities between the residues shown. Download figure Download PowerPoint Determination of the tertiary structure In addition to the 329 sequential and 218 medium range inter-residue NOEs that were used to locate the elements of secondary structure in the previous section, 910 intra-residue and 329 long-range NOEs were identified. For certain side chains it was possible to identify the most populated χ1 rotamer and make stereospecific assignments of 1Hβ or γ-methyl resonances. 3JCγC′ and 3JCγN couplings were determined for six Val, Ile and Thr residues possessing resolved γ-methyl signals from 15N and 13CO spin-echo difference experiments (Grzesiek et al., 1993; Vuister et al., 1993). Stereospecific 1Hβ assignments for a further nine residues could be made by analysis of peak intensities in three-dimensional 15N-separated HMQC-TOCSY, HMQC-NOESY and HNHB spectra (Powers et al., 1993). Three dimensional structures of the chromo domain were produced essentially as described before (Kraulis et al., 1994). Initial structures were computed using simulated annealing calculations from 1207 1H–1H distance restraints that could be determined unambiguously. These computations used the program X-PLOR (Brünger, 1992), starting from random globular initial structures (Nilges et al., 1988, 1991). This set of preliminary structures was used to assist in assignment of the remaining NOE crosspeaks for which the assignments were ambiguous and in selection of acceptor carbonyl oxygen atoms for hydrogen bonds to the slowly exchanging backbone amide protons. The final round of calculations used a restraint list comprising 1937 distance restraints, of which 1786 were unambiguous and 151 semi-ambiguous (Nilges, 1995), 22 φ and 15 χ1 dihedral angle restraints and 17 hydrogen bonds; the selected ensemble contained all of the 27 converged structures out of 40 calculated. The structures are well defined (see Figure 4A) except for the first 11 residues at the N-terminus and the last nine residues at the C-terminus, which are genuinely flexible in solution, as judged by NMR studies of 15N relaxation in the backbone amides (data not shown). The chromo domain therefore corresponds to residues 21–71 of MoMOD1. The average r.m.s. differences from the mean structure are 0.66 Å (± 0.14) for the backbone atoms and 1.14 Å (± 0.17) for all the non-hydrogen atoms in the well-defined core of the molecule. If the less well-ordered loop region (residues 43–47) is included, these values increase to 0.96 Å (± 0.27) and 1.52 Å (± 0.29) respectively. No distance restraint was violated by >0.5 Å and no dihedral angle restraint by >5.0°. The structures have good covalent geometry as judged by PROCHECK (Laskowski et al., 1993) and non-bonded contacts (see Table I). Figure 4.(a) Stereoviews of the backbone (N, Cα and C′) of the 27 superimposed structures of MoMOD1-N (residues 20–73) that converged out of 40 computed from the NMR data. (b) Structure closest to the mean, showing the side chains of the conserved hydrophobic residues in the core and hydrophobic groove of the protein. The side chains are colour coded according to their r.m.s. deviation from the mean structure of all the non–hydrogen atoms in the set of calculated structures: red < 0.75 Å; light blue < 1.5 Å. Download figure Download PowerPoint Table 1. Structural statistics for the final ensemble of 27 refined structures of MoMOD1-N Root mean square deviations from restraints and idealized geometry a (SA)cb NOE distances (Å) 0.010 ± 0.003 0.009 Dihedral angles (°) 0.61 ± 0.29 0.53 Bonds (Å) 0.0016 ± 0.0003 0.0016 Angles (°) 0.38 ± 0.08 0.39 Impropers (°) 0.22 ± 0.04 0.22 Final energy EL−J (kJ/mol) −119.2 ± 26.5 −119.8 Assessment of backbone quality according to the Ramachandran plot All structures Closest to mean Most favoured region 63.9% 66.7% Additionally allowed region 28.7% 25.0% a Represents the average r.m.s. deviations for the ensemble. b Represents values for the final structure that is closest to the mean. Analysis of the tertiary structure The structure of the chromo domain is shown in Figures 4 and 5. The protein consists of an N-terminal three-stranded anti-parallel β-sheet, which packs against the C-terminal α-helix (Figure 4A). A hydrophobic core is formed by residues 23, 26, 38, 40, 42, 58, 60, 63, 64 and 67 (see Figure 4B), all of which are highly conserved in different chromo domains (Figure 1). Several of these residues (23, 40, 42, 58, 60 and 63) also lie at the bottom of a hydrophobic groove (see below) on the surface of the β-sheet below the α-helix. The first strand of the β-sheet is rather irregular and contains two bulges at residues 24 and 27 (Figure 6). The insertions and deletions in the sequences of the different chromo domains are all found in the loops. Insertions and deletions are found in the first and at the end of the second loop in the β-sheet (residues 33–34 and 47–48). Insertions are also found in the loop connecting the β-sheet to the α-helix (residues 59–60) (Aasland and Stewart, 1995). It is therefore likely that all the chromo domains will have similar three-dimensional structures. Figure 5.Comparison of the structure of the MoMOD1-N chromo domain with that of the DNA binding domain of Sac7d (Edmondson et al., 1995). (a) MOLSCRIPT (Kraulis, 1991) plots, in the same orientation as those shown in Figure 4, showing a comparison, of the structures closest to the mean, of the chromo domain from MoMOD1 (left) with that of the DNA binding domain from Sac7d (right); the side chains of residues where mutation affects gene silencing and/or chromatin binding (Messmer et al., 1992; Platero et al., 1995) are shown. (b) Electrostatic potential at the protein surfaces contoured and colour coded at −4.8 kT (red) and +4.8 kT (blue). The potential was calculated and displayed with the program GRASP (Gilson et al., 1988; Nicholls et al., 1991). Compared with the plots in Figures 4 and 5a and c, the structures in (b) have been rotated by ∼130° so that the β-sheet is now facing out of the page. (c) A similar surface, but now showing the putative peptide binding site in the chromo domain and binding of the N-terminal hairpin in Sac7d. Download figure Download PowerPoint Figure 6.A comparison of the secondary structure of the chromo domain (left) with that of the DNA binding domain of Sac7d (Edmondson et al., 1995) (right). Dashed lines represent hydrogen bonds, thick lines conserved secondary structure and circles hydrophobic residues in the protein core. A conserved arginine that interacts with residues 64 and 67 in the α-helix (MoMOD1 numbering) is also shown. Download figure Download PowerPoint Discussion Limited proteolysis of MoMOD1 revealed two regions that are resistant to digestion; these correspond to the conserved sequences in the HP1-like proteins (Clark and Elgin, 1992). The results suggest that the MoMOD1 protein consists of two structural domains, residues 10–78 and 104–171, connected by an exposed linker. Our NMR studies of the MoMOD1-N chromo domain show that the structured region lies between residues 21 and 71, although it is possible that, in the intact MoMOD1 protein, residues 10–20 and 71–78 are protected. The structure of the MoMOD1-N chromo domain explains the results of mutations in the chromo domains of both HP1 (Platero et al., 1995) and Pc (Messmer et al., 1992), which have been shown to disrupt gene silencing and/or chromatin binding. Three of these mutations, V23M in HP1, I26F and removal of residues 64/65 in Pc (all MoMOD1 numbering), are to residues in the hydrophobic core and they may therefore disrupt the structure of the chromo domain (Figure 4a). One mutation, Y21F in HP1, should not disrupt the structure, but this residue lies to one side of the hydrophobic groove, which we suggest mediates protein–protein interactions (see below). This tyrosine may therefore be important for the function of at least the HP1-like chromo domains. We note that the V23M mutation may additionally disrupt function, because this residue is also exposed in the hydrophobic groove. Comparisons with other protein structures from the Brookhaven Protein Databank were carried out by eye and using the computer program DALI (Holm and Sander, 1993). Homologies to two different classes of protein were found. The first group includes the α and β chemokines, e.g. IL-8, MGSA/GRO and MIP-1β, RANTES and MCP1, all of which are involved in protein–protein interactions with receptors (Leonard and Yoshimura, 1990). Most interestingly, the second group comprises small histone-like DNA binding proteins found in the archaebacteria S.acidocaldarius and S.solfataricus. These proteins are also involved in the formation of chromatin structure (Dijk and Reinhardt, 1986) and they are more similar to the MoMOD1-N chromo domain. The r.m.s. deviations for the 31 Cα atoms, which comprise all of the conserved secondary structure, are 1.44 and 2.01 Å, when the chromo domain is compared with the DNA binding domains of Sac7d and Sso7d respectively (Baumann et al., 1994; Edmondson et al., 1995). For the DNA binding domains of Sac7d and Sso7d themselves, which are 70% identical, the r.m.s. deviation is 1.72 Å over the same region. (By comparison, the Cα r.m.s. deviations over the same region are 2.25 and 1.96 Å when the chromo domain is compared with the two chemokines IL-8 and MCP1 respectively.) In Figure 5a, a comparison of the structure of Sac7d with that of the chromo domain is shown. The structures are very similar, but the chromo domain lacks the N-terminal hairpin which binds in a hydrophobic groove below the α-helix in Sac7d. We believe that there may be a distant evolutionary relationship between the chromo domain and Sac7d, for the following reasons. First, in addition to conservation of the three-dimensional folds, there are detailed structural similarities as well. For example, the first bulge at residue 24 in the chromo domain is conserved in Sac7d, to facilitate wrapping of the β-strand around the α-helix (Figure 6). The interaction of a basic residue in the β-sheet (residue 29) with residues 64 and 67 in the α-helix is also conserved (Figure 6), as is the loop between the third β-strand and the α-helix (Figure 5a). Second, both the HP1-like chromo domains and the archaebacterial DNA binding proteins have a cluster of negatively charged residues adjacent to the N-terminus of the structured part of the chromo domain. Finally, distinct similarities in the control of transcription in archaebacteria and eukaryotic cells have been found (Baumann et al., 1995b) and both proteins are involved in the formation of chromatin structure (Dijk and Reinhardt, 1986). The surface charge of MoMOD1-N and Sac7d is compared in Figure 5b. It will be seen that the charge distribution on the exterior of the three-stranded β-sheet of Sac7d, which has been shown to be involved in binding non-specifically to the major groove of DNA (Baumann et al., 1995a), is not conserved in the chromo domain. There are two clusters of basic residues on the surface of the chromo domain (residues 25, 41 and 43 and residues 29, 30, 33 and 35) but overall the charge is negative, suggesting that this class of chromo domain would not bind to DNA or RNA by itself. In support of this, MoMOD1-N shows no appreciable affinity for DNA–cellulose (data not shown). It is also interesting to note that mutations in one of the clusters (R29QR30Q and K31Q, both MoMOD1 numbering) had no influence on the activity of HP1 (Platero et al., 1995). Instead, we believe that the structural homologies suggest that the chromo domain is a protein interaction motif and that the structure of Sac7d provides a model for how the chromo domain might bind other proteins. In the structure of the chromo domain there is a hydrophobic groove around the protein, where the N-terminal hairpin would bind in Sac7d, and we speculate that protein–protein interactions occur via this groove (Figure 5c). This hypothesis is supported by previous studies which suggest that mutation of residues at the edge of this groove (Y21F) and within the groove itself (V23M) disrupt gene silencing in HP1 (Platero et al., 1995). It has been proposed that gene silencing might be controlled by tyrosine phosphorylation at this site (Platero et al., 1995; Madireddi et al., 1996) and phosphorylation should also affect protein–protein interactions in this groove. Strong support for the notion that the chromo domain mediates protein–protein interactions also comes from the results of two-hybrid screens, where both Le Douarin et al. (1996) and ourselves (N.Murzina, K.Johnson, A.De Smet and E.D.Laue, submitted for publication) have isolated different proteins that interact with the chromo domains of either the HP1α or MoMOD1 proteins respectively. It has been proposed that the C-terminal domain may be related to the N-terminal chromo domain in HP1-like proteins (Epstein et al., 1992) and the term ‘shadow’ domain has been suggested (Aasland and Stewart, 1995). Sequence comparisons based on the structure of the MoMOD1-N chromo domain provide direct experimental support for this hypothesis. If the N- and C-terminal domains have a similar fold we would expect that residues in the hydrophobic core and key residues required to make the turns in the β-sheet would be conserved between the N-term