Live-cell three-dimensional single-molecule tracking reveals modulation of enhancer dynamics by NuRD.

To understand how the nucleosome remodeling and deacetylase (NuRD) complex regulates enhancers and enhancer-promoter interactions, we have developed an approach to segment and extract key biophysical parameters from live-cell three-dimensional single-molecule trajectories. Unexpectedly, this has revealed that NuRD binds to chromatin for minutes, decompacts chromatin structure and increases enhancer dynamics. We also uncovered a rare fast-diffusing state of enhancers and found that NuRD restricts the time spent in this state. Hi-C and Cut&Run experiments revealed that NuRD modulates enhancer-promoter interactions in active chromatin, allowing them to contact each other over longer distances. Furthermore, NuRD leads to a marked redistribution of CTCF and, in particular, cohesin. We propose that NuRD promotes a decondensed chromatin environment, where enhancers and promoters can contact each other over longer distances, and where the resetting of enhancer-promoter interactions brought about by the fast decondensed chromatin motions is reduced, leading to more stable, long-lived enhancer-promoter relationships.

The Nucleosome Remodeling and Deacetylase Complex NuRD Is Built from Preformed Catalytically Active Sub-modules.

The nucleosome remodeling deacetylase (NuRD) complex is a highly conserved regulator of chromatin structure and transcription. Structural studies have shed light on this and other chromatin modifying machines, but much less is known about how they assemble and whether stable and functional sub-modules exist that retain enzymatic activity. Purification of the endogenous Drosophila NuRD complex shows that it consists of a stable core of subunits, while others, in particular the chromatin remodeler CHD4, associate transiently. To dissect the assembly and activity of NuRD, we systematically produced all possible combinations of different components using the MultiBac system, and determined their activity and biophysical properties. We carried out single-molecule imaging of CHD4 in live mouse embryonic stem cells, in the presence and absence of one of core components (MBD3), to show how the core deacetylase and chromatin-remodeling sub-modules associate in vivo. Our experiments suggest a pathway for the assembly of NuRD via preformed and active sub-modules. These retain enzymatic activity and are present in both the nucleus and the cytosol, an outcome with important implications for understanding NuRD function.

Structure of Cdc42 bound to the GTPase binding domain of PAK.

The Rho family GTPases, Cdc42, Rac and Rho, regulate signal transduction pathways via interactions with downstream effector proteins. We report here the solution structure of Cdc42 bound to the GTPase binding domain of alphaPAK, an effector of both Cdc42 and Rac. The structure is compared with those of Cdc42 bound to similar fragments of ACK and WASP, two effector proteins that bind only to Cdc42. The N-termini of all three effector fragments bind in an extended conformation to strand beta2 of Cdc42, and contact helices alpha1 and alpha5. The remaining residues bind to switches I and II of Cdc42, but in a significantly different manner. The structure, together with mutagenesis data, suggests reasons for the specificity of these interactions and provides insight into the mechanism of PAK activation.

The structure of mouse HP1 suggests a unique mode of single peptide recognition by the shadow chromo domain dimer.

The heterochromatin protein 1 (HP1) family of proteins is involved in gene silencing via the formation of heterochromatic structures. They are composed of two related domains: an N-terminal chromo domain and a C-terminal shadow chromo domain. Present results suggest that chromo domains may function as protein interaction motifs, bringing together different proteins in multi-protein complexes and locating them in heterochromatin. We have previously determined the structure of the chromo domain from the mouse HP1beta protein, MOD1. We show here that, in contrast to the chromo domain, the shadow chromo domain is a homodimer. The intact HP1beta protein is also dimeric, where the interaction is mediated by the shadow chromo domain, with the chromo domains moving independently of each other at the end of flexible linkers. Mapping studies, with fragments of the CAF1 and TIF1beta proteins, show that an intact, dimeric, shadow chromo domain structure is required for complex formation.

Residues in Cdc42 that specify binding to individual CRIB effector proteins.

Cdc42 is a member of the Rho family of small G proteins. Signal transduction events emanating from Cdc42 lead to cytoskeletal rearrangements, cell proliferation, and cell differentiation. Many effector proteins have been identified for Cdc42; however, it is not clear how certain effectors specifically recognize and bind to Cdc42, as opposed to Rac or Rho, or in many cases, which effector controls what cellular events. Mutations were introduced into Cdc42 at residues: Met1, Val8, Phe28, Tyr32, Val33, Thr35, Val36, Phe37, Asp38, Tyr40, Val42, Met45, Ile46, Glu127, Ala130, Asn132, Gln134, Lys135, and Leu174. Measurements were made of their equilibrium binding constants to the Cdc42 binding domains of the CRIB effectors ACK, PAK, and WASP and to the GTPase-activating protein Rho GAP. Generally, mutations in the effector loop have an equally deleterious effect on binding to all CRIB proteins tested, though the F37A mutation resulted in significant selectivity. Residues outside the effector loop were found to be important for binding of Cdc42 to CRIB containing proteins and also to contribute to selectivity. Mutations such as V42A and L174A resulted in large, selective changes in binding to specific CRIB effectors. Neither mutation resulted in alteration in PAK binding, whereas both severely disrupt binding to ACK and only L174A disrupted binding to WASP. These mutations are interpreted using the structures of the Cdc42/ACK and Cdc42/WASP complexes to give insight into how effectors can specifically recognize Cdc42. Those mutations in Cdc42 that inhibit certain interactions, while retaining others, should aid investigations of the role of specific effectors in Cdc42 signaling in vivo.

Structure of the small G protein Cdc42 bound to the GTPase-binding domain of ACK.

The proteins Cdc42 and Rac are members of the Rho family of small GTPases (G proteins), which control signal-transduction pathways that lead to rearrangements of the cell cytoskeleton, cell differentiation and cell proliferation. They do so by binding to downstream effector proteins. Some of these, known as CRIB (for Cdc42/Rac interactive-binding) proteins, bind to both Cdc42 and Rac, such as the PAK1-3 serine/threonine kinases, whereas others are specific for Cdc42, such as the ACK tyrosine kinases and the Wiscott-Aldrich-syndrome proteins (WASPs). The effector loop of Cdc42 and Rac (comprising residues 30-40, also called switch I), is one of two regions which change conformation on exchange of GDP for GTP. This region is almost identical in Cdc42 and Racs, indicating that it does not determine the specificity of these G proteins. Here we report the solution structure of the complex of Cdc42 with the GTPase-binding domain ofACK. Both proteins undergo significant conformational changes on binding, to form a new type of G-protein/effector complex. The interaction extends the beta-sheet in Cdc42 by binding an extended strand from ACK, as seen in Ras/effector interactions, but it also involves other regions of the G protein that are important for determining the specificity of effector binding.

Nuclear magnetic resonance and molecular dynamics studies on the interactions of the Ras-binding domain of Raf-1 with wild-type and mutant Ras proteins.

The Ras protein and its homolog, Rap1A, have an identical "effector region" (residues 32-40) preceded by Asp30-Glu31 and Glu30-Lys31, respectively. In the complex of the "Ras-like" E30D/K31E mutant Rap1A with the Ras-binding domain (RBD), residues 51-131 of Raf-1, Glu31 in Rap1A forms a tight salt bridge with Lys84 in Raf-1. However, we have recently found that Raf-1 RBD binding of Ras is indeed reduced by the E31K mutation, but is not affected by the E31A mutation. Here, the "Rap1A-like" D30E/E31K mutant of Ras was prepared and shown to bind the Raf-1 RBD less strongly than wild-type Ras, but slightly more tightly than the E31K mutant. The backbone 1H, 13C, and 15N magnetic resonances of the Raf-1 RBD were assigned in complexes with the wild-type and D30E/E31K mutant Ras proteins in the guanosine 5'-O-(beta,gamma-imidotriphosphate)-bound form. The Lys84 residue in the Raf-1 RBD exhibited a large change in chemical shift upon binding wild-type Ras, suggesting that Lys84 interacts with wild-type Ras. The D30E/E31K mutant of Ras caused nearly the same perturbations in Raf-1 chemical shifts, including that of Lys84. We hypothesized that Glu31 in Ras may not be the major salt bridge partner of Lys84 in Raf-1. A molecular dynamics simulation of a model structure of the Raf-1 RBD.Ras.GTP complex suggested that Lys84 in Raf-1 might instead form a tight salt bridge with Asp33 in Ras. Consistent with this, the D33A mutation in Ras greatly reduced its Raf-I RBD binding activity. We conclude that the major salt bridge partner of Lys84 in Raf-1 may be Asp33 in Ras.

Crystal structure of the complex of the cyclin D-dependent kinase Cdk6 bound to the cell-cycle inhibitor p19INK4d.

The crystal structure of the cyclin D-dependent kinase Cdk6 bound to the p19 INK4d protein has been determined at 1.9 A resolution. The results provide the first structural information for a cyclin D-dependent protein kinase and show how the INK4 family of CDK inhibitors bind. The structure indicates that the conformational changes induced by p19INK4d inhibit both productive binding of ATP and the cyclin-induced rearrangement of the kinase from an inactive to an active conformation. The structure also shows how binding of an INK4 inhibitor would prevent binding of p27Kip1, resulting in its redistribution to other CDKs. Identification of the critical residues involved in the interaction explains how mutations in Cdk4 and p16INK4a result in loss of kinase inhibition and cancer.

Dual amino acid-selective and site-directed stable-isotope labeling of the human c-Ha-Ras protein by cell-free synthesis.

We developed two methods for stable-isotope labeling of proteins by cell-free synthesis. Firstly, we applied cell-free synthesis to the dual amino acid-selective 13C-15N labeling method, originally developed for in vivo systems by Kainosho and co-workers. For this purpose, we took one of the advantages of a cell-free protein synthesis system; the amino acid-selective stable-isotope labeling is free of the isotope scrambling problem. The targets of selective observation were Thr35 and Ser39 in the effector region (residues 32-40) of the Ras protein complexed with the Ras-binding domain of c-Raf-1 (Raf RBD) (the total molecular mass is about 30 kDa). Using a 15-mL Escherichia coli cell-free system, which was optimized to produce about 0.4 mg of Ras protein per 1-mL reaction, with 2 mg each of DL-[13C']proline and L-[15N]threonine, we obtained about 6 mg of Ras protein. As the Pro-Thr sequence is unique in the Ras protein, the Thr35 cross peak of the Ras.Raf RBD complex was unambiguously identified by the 2D 1H-15N HNCO experiment. The Ser-39 cross peak was similarly identified with the [13C']Asp/[15N]Ser-selectively labeled Ras protein. There were no isotope scrambling problems in this study. Secondly, we have established a method for producing a milligram quantity of site-specifically stable-isotope labeled protein by a cell-free system involving amber suppression. The E. coli amber suppressor tRNATyrCUA (25 mg) was prepared by in vitro transcription with T7 RNA polymerase. We aminoacylated the tRNATyrCUA transcript with purified E. coli tyrosyl-tRNA synthetase, using 2 mg of L-[15N]tyrosine. In the gene encoding the Ras protein, the codon for Tyr32 was changed to an amber codon (TAG). This template DNA and the [15N]Tyr-tRNATyrCUA were reacted for 30 min in 30 mL of E. coli cell-free system. The subsequent purification yielded 2.2 mg of [15N]Tyr32-Ras protein. In the 1H-15N HSQC spectrum of the labeled Ras protein, only one cross peak was observed, which was unambiguously assigned to Tyr32.

Structure and properties in solution of PsaD, an extrinsic polypeptide of photosystem I.

PsaD is a small, extrinsic polypeptide located on the stromal side (cytoplasmic side in cyanobacteria) of the photosystem I reaction centre complex. The gene from the cyanobacterium Nostoc sp. PCC 8009 was expressed in Escherichia coli and the structure of the recovered protein in solution investigated. Size-exclusion chromatography, dynamic light scattering and measurement of 15N transverse relaxation times showed that the protein is a stable dimer in solution, whereas in the reaction centre complex it is a monomer. NMR experiments showed that the dimer is symmetrical and that there are at least two domains, one structured and the remainder unstructured. The structured domain contains a small amount of beta-sheet. Three-dimensional heteronuclear NMR spectra of [13C, 15N]PsaD showed that the structured domain is associated with the central part of the sequence while the N- and C-terminal regions are mobile. Evidence was obtained for a shift in equilibrium between two slightly different conformational states at about pH 6, and the protein was shown to bind to PsaE preferentially at neutral pH. Addition of trifluoroethanol was shown to induce the formation of a small amount of alpha-helix, and the form present in 30% trifluoroethanol appears to be more closely related to the in situ structure, which has been reported to contain one short helix in crystals [Schubert, W.-D., Klukas, O., Krauss, N., Saenger, W., Fromme, P. & Witt, H. T. (1997) J. Mol. Biol. 272, 741-769]. The significance of these findings for the assembly of the complex is discussed.

Structure of the cyclin-dependent kinase inhibitor p19Ink4d.

In cancer, the biochemical pathways that are dominated by the two tumour-suppressor proteins, p53 and Rb, are the most frequently disrupted. Cyclin D-dependent kinases phosphorylate Rb to control its activity and they are, in turn, specifically inhibited by the Ink4 family of cyclin-dependent kinase inhibitors (CDKIs) which cause arrest at the G1 phase of the cell cycle. Mutations in Rb, cyclin D1, its catalytic subunit Cdk4, and the CDKI p16Ink4a, which alter the protein or its level of expression, are all strongly implicated in cancer. This suggests that the Rb 'pathway' is of particular importance. Here we report the structure of the p19Ink4d protein, determined by NMR spectroscopy. The structure indicates that most mutations to the p16Ink4a gene, which result in loss of function, are due to incorrectly folded and/or insoluble proteins. We propose a model for the interaction of Ink4 proteins with D-type cyclin-Cdk4/6 complexes that might provide a basis for the design of therapeutics against cancer. The sequences of the Ink4 family of CDKIs are highly conserved

Regional polysterism in the GTP-bound form of the human c-Ha-Ras protein.

The backbone 1H, 13C, and 15N resonances of the c-Ha-Ras protein [a truncated version consisting of residues 1-171, Ras(1-171)] bound with GMPPNP (a slowly hydrolyzable analogue of GTP) were assigned and compared with those of the GDP-bound Ras(1-171). The backbone amide resonances of amino acid residues 10-13, 21, 31-39, 57-64, and 71 of Ras(1-171).GMPPNP, but not those of Ras(1-171).GDP, were extremely broadened, whereas other residues of Ras(1-171).GMPPNP exhibited amide resonances nearly as sharp as those of Ras(1-171). GDP. The residues exhibiting the extreme broadening, except for residues 21 and 71, are localized in three functional loop regions [loops L1, L2 (switch I), and L4 (switch II)], which are involved in hydrolysis of GTP and interactions with other proteins. From the temperature and magnetic field strength dependencies of the backbone amide resonance intensities, the extreme broadening was ascribed to the exchange at an intermediate rate on the NMR time scale. It was shown that the Ras(1-171) protein bound with GTP or GTPgammaS (another slowly hydrolyzable analogue of GTP) exhibits the same type of broadening. Therefore, it is a characteristic feature of the GTP-bound form of Ras that the L1, L2, and L4 loop regions, but not other regions, are in a rather slow interconversion between two or more stable conformers. This phenomenon, termed a "regional polysterism", of these loop regions may be related with their multifunctionality: the GTP-dependent interactions with several downstream target groups such as the Raf and RalGDS families and also with the GTPase activating protein (GAP) family. In fact, the binding of Ras(1-171).GMPPNP with the Ras-binding domain (residues 51-131) of c-Raf-1 was shown to eliminate the regional polysterism nearly completely. It was indicated, therefore, that each target/regulator selects its appropriate conformer among those presented by the "polysteric" binding interface of Ras. As the downstream target groups exhibit no apparent sequence homology to each other, it is possible that one target group prefers a conformer different from that preferred by another group. The involvement of loop L1 in the regional polysterism might suggest that the negative regulators, GAPs, bind to the polysteric binding interface (loops L2 and L4) of Ras and cooperatively select a conformer suitable for transition of the GTPase catalytic center, involving loops L1 and L4, into the highly active state.

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

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 beta-sheet which folds against a C-terminal alpha-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.

An approach to global fold determination using limited NMR data from larger proteins selectively protonated at specific residue types.

A combination of calculation and experiment is used to demonstrate that the global fold of larger proteins can be rapidly determined using limited NMR data. The approach involves a combination of heteronuclear triple resonance NMR experiments with protonation of selected residue types in an otherwise completely deuterated protein. This method of labelling produces proteins with alpha-specific deuteration in the protonated residues, and the results suggest that this will improve the sensitivity of experiments involving correlation of side-chain ((1)H and (13)C) and backbone ((1)H and (15)N) amide resonances. It will allow the rapid assignment of backbone resonances with high sensitivity and the determination of a reasonable structural model of a protein based on limited NOE restraints, an application that is of increasing importance as data from the large number of genome sequencing projects accumulates. The method that we propose should also be of utility in extending the use of NMR spectroscopy to determine the structures of larger proteins.

Structure of the A-domain of HMG1 and its interaction with DNA as studied by heteronuclear three- and four-dimensional NMR spectroscopy.

HMG1 has two homologous, folded DNA-binding domains ("HMG boxes"), A and B, linked by a short basic region to an acidic C-terminal domain. Like the whole protein, which may perform an architectural role in chromatin, the individual boxes bind to DNA without sequence specificity, have a preference for distorted or prebent DNA, and are able to bend DNA and constrain negative superhelical turns. They show qualitatively similar properties with quantitative differences. We have previously determined the structure of the HMG box from the central B-domain (77 residues) by two-dimensional NMR spectroscopy, which showed that it contains a novel fold [Weir et al. (1993) EMBO J. 12, 1311-1319]. We have now determined the structure of the A-domain (as a Cys-->Ser mutant at position 22 to avoid oxidation, without effect on its DNA-binding properties or structure) using heteronuclear three- and four-dimensional NMR spectroscopy. The A-domain has a very similar global fold to the B-domain and the Drosophila protein HMG-D [Jones et al. (1994) Structure 2, 609-627]. There are small differences between A and B, in particular in the orientation of helix I, where the B-domain is more similar to HMG-D than it is to the A-domain; these differences may turn out to be related to the subtle differences in functional properties between the two domains [Teo et al. (1995) Eur. J. Biochem. 230, 943-950] and will be the subject of further investigation. NMR studies of the interaction of the A-domain of HMG1 with a short double-stranded oligonucleotide support the notion that the protein binds via the concave face of the L-shaped structure; extensive contacts with the DNA are made by the N-terminal extended strand, the N-terminus of helix I, and the C-terminus of helix II. These contacts are very similar to those seen in the LEF-1 and SRY-DNA complexes [Love et al. (1995) Nature 376, 791-795; Werner et al. (1995) Cell 81, 705-714].

Backbone dynamics of the A-domain of HMG1 as studied by 15N NMR spectroscopy.

The HMG-box sequence motif (approximately 80 residues) occurs in a number of abundant eukaryotic chromosomal proteins such as HMG1, which binds DNA without sequence specificity, but with "structure specificity", as well as in several sequence-specific transcription factors. HMG1 has two such boxes, A and B, which show approximately 30% sequence identity, and an acidic C-terminal tail. The boxes are responsible for the ability of the protein to bend DNA and bind to bent or distorted DNA. The structure of the HMG box has been determined by NMR spectroscopy for the B-domain of HMG1 [Weir et al. (1993) EMBO J. 12, 1311-1319; Read et al. (1993) Nucleic Acids Res. 21, 3427-3436) and for Drosophila HMG-D (Jones et al. (1994) Structure 2, 609-627]. It has an unusual twisted L-shape, suggesting that the protein might tumble anisotropically in solution. In this paper we report studies of the A-domain from HMG1 using 15N NMR spectroscopy which show that the backbone dynamics of the protein can be described by two different rotational correlation times of 9.0 +/- 0.5 and 10.8 +/- 0.5 ns. We show that the relaxation data can be analyzed by assuming that the protein is a rigid, axially symmetric ellipsoid undergoing anisotropic rotational diffusion; the global rotational diffusion constants, D parallel and D perpendicular, were estimated as 2.47 x 10(7) and 1.49 x 10(7) s-1, respectively. By estimating the angle between the amide bond vectors and the major axis of the rotational diffusion tensor from the family of structures determined by NMR spectroscopy [see accompanying paper, Hardman et al. (1995) Biochemistry 34, 16596-16607], we were able to show that the ellipsoid spectral density equation can reproduce the major features of the 15N T1 and T2 profiles of the three helices in the HMG1 A-domain. The backbone dynamics of the A-domain were then compared with those of the B-domain and the HMG box from HMG-D. This comparison strongly supported the differences observed in the orientation of helix I in the three structures, where the B-domain appears to be more similar to HMG-D than it is to the A-domain. These differences may turn out to be related to subtle differences in the DNA-binding properties of the A- and B-domains of HMG1.

Two mutations in the HMG-box with very different structural consequences provide insights into the nature of binding to four-way junction DNA.

Mutation of the highly conserved tryptophan residue in the A-domain HMG-box of HMG1 largely, but not completely, destroys the protein tertiary structure and abolishes its supercoiling ability, but does not abolish structure-specific DNA binding to four-way junctions. Circular dichroism shows that the protein has some residual alpha-helix (< 10%) and does not re-fold in the presence of DNA. Structure-specific DNA binding might therefore be a property of some primary structure element, for example the N-terminal extended strand, which even in the unfolded protein would be held in a restricted conformation by two, largely trans, X-Pro peptide bonds. However, mutation of P5 or P8 of the A-domain to alanine does not abolish the formation of the (first) complex in a gel retardation assay, which probably arises from binding to the junction cross-over, although the P8 mutation does affect the formation of higher complexes which may arise from binding to the junction arms. Since mutation of P8 in the W49R mutant has no effect on structure-specific junction binding, we propose that some residual alpha-helix in the protein might be involved, implicating this element in the interactions of HMG-boxes generally with DNA.

Three-dimensional structure of a lipoyl domain from the dihydrolipoyl acetyltransferase component of the pyruvate dehydrogenase multienzyme complex of Escherichia coli.

The structure of a lipoyl domain from the pyruvate dehydrogenase multienzyme complex of Escherichia coli has been determined by means of nuclear magnetic resonance spectroscopy. A total of 549 nuclear Overhauser effect distance restraints, 52 phi torsion angle restraints and 16 slowly exchanging amide protons were employed as input for the structure calculations. These were performed using a combined distance geometry-simulated annealing strategy. The domain is a hybrid between the N and C-terminal halves of the first and third lipoyl domains, respectively, of the dihydrolipoyl acetyltransferase component of the E. coli multienzyme complex, representing residues 1 to 33 and 238 to 289 (wild-type numbering). The lipoyl-lysine residue was also replaced by glutamine. Nonetheless, its structure, two four-stranded beta-sheets forming a flattened beta-barrel, closely resembles that of the lipoyl domain from the pyruvate dehydrogenase multienzyme complex from Bacillus stearothermophilus determined previously. As before, the lipoylation site is physically exposed in a tight turn in one of the beta-sheets, and the N and C-terminal residues are close together at the other end of the molecule in adjacent strands of the other beta-sheet. Another prominently conserved feature of the structure is the 2-fold axis of quasi-symmetry relating the N and C-terminal halves of the domain. Consistent with the high level of sequence similarity between lipoyl domains of 2-oxo acid dehydrogenase multienzyme complexes from many different sources, these results confirm that all lipoyl domains are likely to have closely related structures.

Improved methods for structural studies of proteins using nuclear magnetic resonance spectroscopy.

The past few years have seen the development of three- and four-dimensional heteronuclear nuclear magnetic resonance methods. Increased sophistication in labelling strategies, use of pulse-field gradients and the application of these methods at higher magnetic fields has, in combination with improved software, allowed studies of the structure, interactions and dynamics of significantly larger proteins (now up to approximately 270 amino acid residues).