1. Introduction 286

2. Membrane protein assembly inE. coli286

2.1. Role of the SRP 287

2.2. YidC – a translocon component devoted to membrane proteins? 287

2.3. The TAT pathway 288

2.4. ‘Spontaneous’ membrane protein insertion 288

3. Membrane protein assembly in the ER 289

3.1. How TM segments exit the translocon 289

3.2. Proteins with multiple topologies 290

3.3. Stop-transfer effector sequences 291

3.4. Non-hydrophobic TM segments? 291

3.5. ‘Frustrated’ topologies 291

3.6. N-tail translocation across the ER 292

4. Membrane protein assembly in mitochondria 292

4.1. The Oxa1p pathway 292

4.2. The TIM22/54 pathway 293

5. Evolution of membrane protein topology 293

5.1. RnfA/RnfE – two homologous proteins with opposite topologies 293

5.2. YrbG – duplicating an odd number of TMs 294

6. Genome-wide analysis of membrane proteins 295

6.1. Prediction methods 295

6.2. How many membrane proteins are there? 295

6.3. The positive-inside rule 296

6.4. Dominant classes of membrane proteins 296

7. The structure of transmembrane α-helices 296

7.1. What TM helices look like 297

7.2. The ‘helical hairpin’ 297

7.3. Prolines in TM helices 297

7.4. Charged residues in TM helices: the ‘snorkel’ effect 298

7.5. The ‘aromatic belt’ 298

8. Helix–helix packing in a membrane environment 298

8.1. Lessons learnt from glycophorin A 298

8.2. Genetic screens for helix–helix interactions 299

8.3. Statistical studies 299

8.4. Membrane protein folding 299

9. Recent 3D structures 300

9.1. KcsA – the first ion channel 300

9.2. MscL – sensing lateral pressure changes 300

9.3. The cytochrome bc 1 complex 300

9.4. Fumarate reductase 301

9.5. Bacteriorhodopsin – watching a membrane protein at work 301

10. Concluding remarks 301

11. Acknowledgements 302

12. References 302

For a variety of reasons – not the least biomedical importance – integral membrane proteins are now very much in focus in many areas of molecular biology, biochemistry, biophysics, and cell biology. Our understanding of the basic processes of membrane protein assembly, folding, and structure has grown significantly in recent times, both as a result of new methodological developments, more high-resolution structure data, and the possibility to analyze membrane proteins on a genome-wide scale.

So what is new in the membrane protein field? Various aspects of membrane protein assembly and structure have been reviewed over the past few years (Cowan & Rosenbusch, 1994; Hegde & Lingappa, 1997; Lanyi, 1997; von Heijne, 1997; Bernstein, 1998); here, I will try to bring together a number of exciting recent developments. Particularly noteworthy are the discoveries related to the mechanisms of membrane protein assembly into the inner membrane of E. coli, the inner membrane of mitochondria, and the way transmembrane segments are handled by the ER translocon.

Other advances include detailed studies of the interaction between transmembrane helices and the lipid bilayer, and of helix–helix packing interactions in the membrane environment. The availability of full genomic sequences have made it possible to study membrane proteins on a genome-wide scale. Finally, a handful of new high-resolution 3D structures have appeared.

This review will deal only with helix bundle proteins, i.e. integral membrane proteins where the transmembrane segments form α-helices. For reviews on the other major class of integral membrane proteins – the β-barrel proteins – see Schirmer (1998) and Buchanan (1999). For readers who prefer a more ‘literary’ introduction to the membrane protein field, may I suggest von Heijne (1999).

1. Introduction 100

2. Molecular architecture 107

2.1 Primary structure 108

2.1.1 Homologous regions 109

2.1.2 Chain typing 115

2.1.3 Post-translational modifications 117

2.2 Secondary structure 118

2.2.1 Central rod domain 118

2.2.2 Head and tail domains 119

2.3 Tertiary structure 123

2.3.1 Coiled-coil rod domain 123

2.3.1.1 Specificity through salt bridges 124

2.3.1.2 Specificity through apolar interactions 127

2.3.1.3 A consensus trigger sequence for two-stranded coiled-coils 128

2.3.2 Discontinuities in the rod domain 128

2.3.2.1 Links 129

2.3.2.2 Stutter 131

2.3.3 Head and tail domains 131

2.4 Electron microscope observations 133

3. Assembly 136

3.1 Role of the coiled-coil rod domain 137

3.2 Role of end domains 141

3.3 Experimentally induced crosslinks and modes of assembly 145

3.4 Naturally occurring crosslinks for tissue coordination 154

3.5 STEM data 154

4. Quaternary structure 160

4.1 Protofilaments and protofibrils 160

4.2 Head and tail domains 163

4.3 Surface lattice structure 164

4.4 Crystal studies on intermediate filament fragments 168

5. Polymorphism 169

5.1 Variations on a theme 170

5.1.1 Axial structure 170

5.1.2 Lateral structure 171

6. Keratin intermediate filament chains in diseases 172

7. Concluding remarks 175

8. Acknowledgments 176

9. References 176

Three types of intracellular filament have been identified in eukaryotic cells, and together they constitute the key elements of the cytoskeleton. They are the microtubules, the actin-containing microfilaments and the intermediate filaments. The uniqueness of the former two types of filament in cells has been well known for a long time but, in contrast, the intermediate filaments have been a relative new-comer to the scene. The microtubules and the microfilaments have always been easy to distinguish from one another on the grounds of their respective sizes (microtubules are about 25 nm in diameter and microfilaments are about 7–10 nm in diameter). Additionally, microtubules were capable of being disaggregated by the action of colchicine, and microfilaments could be disassembled by other drugs or be decorated with heavy meromyosin to generate arrowhead-like structures. Importantly, in both microtubules and microfilaments the constituent protein subunits were arranged to give the filaments a directionality, and the ability of these filaments to function in vivo depended crucially on this feature of their structure. Microtubules, for example, are involved in mitosis, motility and transport within the cell: each of these functions is clearly a ‘directional’ one. With this background the discovery and characterization of the intermediate filaments can begin.

Nucleoside triphosphates are crucial mediators of life. The high energy phosphoanhydride bond of (usually) ATP is used to drive unfavorable chemical reactions, to fuel biological machines, and to regulate a vast number of processes via phosphorylation of proteins. GTP, in turn, is used almost exclusively for the regulation of signal transduction and transport processes, whereas the other nucleotides play a less important role, except in synthesis pathways involving sugars (UTP) and phospholipids (CTP) and as building blocks of polynucleotides such as RNA and DNA. Proteins that bind and use these nucleotides for enzymatic reaction and regulation are very diverse. Although some of them constitute the largest protein superfamilies known (e.g. protein kinases), others seem to be far less conserved in evolution.

Many basic events which concern management and manipulation of genes within a cell – for instance transcription, replication and recombination – rely on specific interactions between proteins and nucleic acids. Such interactions are also essential for many house-keeping functions, like packing and unpacking of DNA in chromatin and assembly of ribosomes. Moreover, the details of protein–nucleic acid interplay is essential for understanding the action of viruses. The list of functional mechanisms in biology that rely on protein–DNA and protein–RNA interactions can be made much longer, but these examples represent some of the topics which motivated structural biologists to study complexes between proteins and nucleic acids as a first step beyond structure determinations of individual biomolecules.

1. Introduction 189

2. Channel properties 191

2.1 Voltage-dependent gating 191

2.2 Ion permeability 193

2.2.1 Selectivity between potassium and chloride 193

2.2.2 Permeability to large cations and large anions 193

2.3 Single-channel characteristics 194

2.4 Molecularity of the channel 195

3. Colicin Ia channel topology and protein translocation 195

3.1 Channels formed by whole colicin Ia 195

3.1.1 General channel topology 196

3.1.2 The translocated region 199

3.1.3 The nonuniqueness of the upstream membrane-inserted segment 199

3.2 Channels formed by the C-terminal domain of colicin Ia 200

4. Concluding remarks 202

5. Acknowledgement 203

6. References 203

Colicins are plasmid-encoded proteins, produced by some strains of E. coli, that kill other strains lacking the specific immunity protein encoded by the same plasmid. Most of the colicins have a three-domain structure: a central domain that binds to a receptor in the outer membrane of the target cell; an N-terminal domain that interacts with target cell proteins to move the C-terminal domain across the outer membrane and periplasmic space to the inner membrane; and a C-terminal domain that carries the toxic activity. In some colicins the C-terminal domain is an enzyme that kills the cell by entering the cytoplasm and attacking its DNA (e.g. colicin E2), its ribosomal RNA (e.g. colicin E3), or another target (Schaller et al. 1982; Ogawa et al. 1999). In other colicins, the C-terminal domain forms an ion-conducting channel in the inner membrane that ultimately leads to cell death by allowing essential solutes to leak out of the cell. These colicins, or their isolated C-terminal domains, can also form voltage-dependent channels in planar phospholipid bilayers. (For a review of the E colicins, including enzymatic colicins, see James et al. 1996; for a review of channel-forming colicins, see Cramer et al. 1995; and for a review of colicin import into E. coli, see Lazdunski et al. 1998.) The channel-forming colicins are the subject of this review, with particular emphasis on one member of this group, colicin Ia, and the protein translocation associated with the gating of its channel.

1. Introduction 310

2. Protein-only hypothesis 312

3. The scrapie prion protein PrPSc313

3.1 Purification of PrP 27–30 313

3.2 Proteinase K resistance 314

3.3 Scrapie-associated fibrils 314

3.4 Smallest infectious unit 316

3.5 Conformational properties 316

3.6 Dissociation and stability 319

4. The cellular prion protein PrPC321

4.1 Prnp expression 321

4.2 Biosynthetic pathway 322

4.3 NMR structures 324

4.4 Copper binding 326

5. Post-translational PrP conversion 327

5.1 Conformational isoforms 327

5.2 Location of propagation 329

5.3 Minimal PrP sequence 330

5.4 Prion species barrier 331

5.5 Prion strains 332

6. Effect of familial TSE mutations 333

6.1 Thermodynamic stability of PrPC 334

6.2 De novo synthesis of PrPSc 335

6.3 Transmembrane PrP forms 337

7. Physical properties of synthetic PrP 337

7.1 Amyloidogenic peptides 337

7.2 Folding intermediates 339

8. Hypothetical protein X 340

8.1 Two species-specific epitopes 340

8.2 Mapping the protein X epitope 341

9. Chaperone-mediated PrP conversion 343

9.1 Hsp60 and Hsp10 chaperonins 343

9.2 GroEL promoted PrP-res formation 345

9.3 Membrane-associated chaperonins 345

9.4 Preference of GroEL for positive charges 347

9.5 Potential GroEL/Hsp60 epitopes on PrP 347

9.6 Conformations of chaperonin-bound PrP 349

9.7 Conserved Hsp60 substrate binding sites 349

9.8 Requirement of ATP-hydrolysis 351

9.9 Hsp60-mediated prion propagation 354

10. Template-assisted annealing model 355

11. Acknowledgments 357

12. References 357

Although the central paradigm of protein folding (Anfinsen, 1973), that the unique three-dimensional structure of a protein is encoded in its amino acid sequence, is well established, its generality has been questioned due to two recent developments in molecular biology, the ‘prion’ and ‘molecular chaperone’. Biochemical characterization of infectious scrapie material causing central nervous system (CNS) degeneration indicates that the necessary component for disease propagation is proteinaceous (Prusiner, 1982), as first outlined by Griffith (1967) in general terms, and involves a conversion from a cellular prion protein, denoted PrPC, into a toxic scrapie form, PrPSc, which is facilitated by PrPSc acting as a template for PrPC to form new PrPSc molecules (Prusiner, 1987). The ‘protein-only’ hypothesis implies that the same polypeptide sequence, in the absence of any post-translational modifications, can adopt two considerably different stable protein conformations (Fig. 1). Thus, in the case of prions it is possible, although not proven, that they violate the central paradigm of protein folding. There is some indirect evidence that another factor, provisionally named ‘protein X’, might be involved in the conformational conversion process (Prusiner et al. 1998), which includes a dramatic change from α-helical into β-sheet secondary structure (Fig. 1). This factor has not been identified yet, but it has been proposed that protein X may act as a molecular chaperone. The idea that molecular chaperones play a critical role in the generation of PrPSc is appealing also from a theoretical point of view, because PrPSc formation involves changes in protein folding and possibly intermolecular aggregation (Fig. 1), processes in which chaperones are known to participate (Musgrove & Ellis, 1986). The discovery and functional analysis of more than a dozen molecular chaperones made it clear that these proteins do not complement folding information that is not already contained in the genetic code (Ellis et al. 1989); rather they facilitate the folding and assembly of proteins by preventing misfolding and refolding misfolded proteins (Hartl, 1996). Whether a molecular chaperone or another type of macromolecule is identified as the conversion factor, therefore, the molecular chaperone concept is likely to contribute to the understanding of the molecular nature of PrPC to PrPSc conversion.

In this review I consider the prion concept from the view of a structural biologist whose main interest focuses on spontaneous and chaperone-mediated conformational changes in proteins.

1. Introduction 211

2. Screening methods 213

2.1 Chemical shifts 213

2.2 Diffusion 214

2.3 Transverse relaxation 218

2.4 Nuclear Overhauser effects 218

3. Strategies for drug discovery and design 221

3.1 Fragment-based methods 221

3.1.1 Linked-fragment approach 221

3.1.2 Directed combinatorial libraries 222

3.1.3 Modification of high-affinity ligands 223

3.1.4 Solvent mapping techniques 223

3.2 High-throughput NMR-based screening 224

3.3 Enzymatic assays 226

4. Discovery of novel ligands 227

4.1 High-affinity ligands for FKBP 227

4.2 Potent inhibitors of stromelysin 229

4.3 Ligands for the DNA-binding domain of the E2 protein 233

4.4 Discovery of Erm methyltransferase inhibitors 233

4.5 Phosphotyrosine mimetics for SH2 domains 236

5. Conclusions 237

6. References 237

A critical step in the drug discovery process is the identification of high-affinity ligands for macromolecular targets. Traditionally, the identification of such lead compounds has been accomplished through the high-throughout screening (HTS) of corporate compound repositories. Conventional HTS methodology has enjoyed widespread application and success in the pharmaceutical industry and, through recent technological advances in screening (Fernandes, 1998; Oldenburg et al. 1998; Silverman et al. 1998) and combinatorial chemistry (Fauchere et al. 1998; Fecik et al. 1998), these programs will continue to have a prominent role in drug discovery. However, suitable leads cannot always be found using conventional methods. This is not surprising since typical corporate libraries contain fewer than 106 compounds compared with the estimated 1050–1080 universe of compounds (Martin, 1997). In addition, most conventional assays are limited to screening libraries of compounds against proteins with known function, excluding the large number of targets becoming available from genomics research.

Recently, a number of NMR-based screening methods have been employed to identify and design lead ligands for protein targets (see Table 1). These NMR-based strategies can augment ongoing conventional HTS for identifying leads and can be used to aid in lead optimization. All of these techniques take advantage of the fact that upon complex formation between a target molecule and a ligand, significant perturbations can be observed in NMR-sensitive parameters of either the target or the ligand. These perturbations can be used qualitatively to detect ligand binding or quantitatively to assess the strength of the binding interaction. In addition, some of the techniques allow the identification of the ligand binding site or which part of the ligand is responsible for interacting with the target. In this article, the current state of NMR-based screening is reviewed.

1. How do ribozymes work? 241

2. The hammerhead RNA as a prototype ribozyme 242

2.1 RNA enzymes 242

2.2 Satellite self-cleaving RNAs 242

2.3 Hammerhead RNAs and hammerhead ribozymes 244

3. The chemical mechanism of hammerhead RNA self-cleavage 246

3.1 Phosphodiester isomerization via an SN2(P) reaction 247

3.2 The canonical role of divalent metal ions in the hammerhead ribozyme reaction 251

3.3 The hammerhead ribozyme does not actually require metal ions for catalysis 254

3.4 Hammerhead RNA enzyme kinetics 257

4. Sequence requirements for hammerhead RNA self-cleavage 260

4.1 The conserved core, mutagenesis and functional group modifications 260

4.2 Ground-state vs. transition-state effects 261

5. The three-dimensional structure of the hammerhead ribozyme 262

5.1 Enzyme–inhibitor complexes 262

5.2 Enzyme–substrate complex in the initial state 264

5.3 Hammerhead ribozyme self-cleavage in the crystal 264

5.4 The requirement for a conformational change 265

5.5 Capture of conformational intermediates using crystallographic freeze-trapping 266

5.6 The structure of a hammerhead ribozyme ‘early’ conformational intermediate 267

5.7 The structure of a hammerhead ribozyme ‘later’ conformational intermediate 268

5.8 Is the conformational change pH dependent? 269

5.9 Isolating the structure of the cleavage product 271

5.10 Evidence for and against additional large-scale conformation changes 274

5.11 NMR spectroscopic studies of the hammerhead ribozyme 278

6. Concluding remarks 280

7. Acknowledgements 281

8. References 281

1. How do ribozymes work? 241

The discovery that RNA can be an enzyme (Guerrier-Takada et al. 1983; Zaug & Cech, 1986) has created the fundamental question of how RNA enzymes work. Before this discovery, it was generally assumed that proteins were the only biopolymers that had sufficient complexity and chemical heterogeneity to catalyze biochemical reactions. Clearly, RNA can adopt sufficiently complex tertiary structures that make catalysis possible. How does the three- dimensional structure of an RNA endow it with catalytic activity? What structural and functional principles are unique to RNA enzymes (or ribozymes), and what principles are so fundamental that they are shared with protein enzymes?