How do viruses enter cells? How do defined protein complexes assemble on specific sites on chromosomal DNA, either for gene regulation or for chromosome maintenance and segregation? How is the organization of membrane-bound, cytoplasmic compartments maintained? We seek answers to these questions in the language of molecular structure and macromolecular dynamics. In effect, we seek to produce experimentally realistic "molecular movies" of the assemblies we choose to study, using x-ray crystallography, electron cryomicroscopy (cryo-EM), and related biochemical and biophysical approaches.
Viruses and Viral Proteins
Viruses fall into two principal structural classes—those with lipid-bilayer membranes (the so-called "enveloped" viruses) and those without membranes ("nonenveloped"). These classes correspond to different modes of assembly and different modes of entry into a new host cell. Enveloped viruses enter cells by fusion of viral and cellular membranes. The process is facilitated by a viral surface protein—the well-known gp120/gp41 in the case of HIV-1. Work (jointly with the late Don C. Wiley) on gp41 showed that this part of the HIV envelope protein, like related parts of a number of other viral envelope proteins (e.g., the hemagglutinin [HA] of influenza virus), undergoes a striking conformation reorganization, triggered by receptor binding. The conformational change can be inhibited by peptides derived from parts of gp41, and we have now shown that small molecules can also inhibit gp41-mediated membrane fusion and viral entry.
The receptor-binding event that sets off the fusion-inducing conformational change in the HIV envelope protein occurs in gp120. In 2005, we published the structure of gp120 from SIV (simian immunodeficiency virus) in the conformation it adopts prior to receptor binding. When compared with the previously known receptor-bound state, the structure showed that gp120 itself undergoes a striking reorganization, which can be blocked by small-molecule inhibitors of viral infectivity and probably by certain neutralizing antibodies. In collaboration with Bing Chen (Children's Hospital), we are working out further structural details of the multiple conformational changes in gp120/gp41 that accompany viral entry (i.e., trying to produce the full molecular movie), while also asking whether constrained forms of the HIV-1 envelope protein might be useful immunogens.
Not all viral fusion proteins have structures resembling those of HIV gp120/gp41 or influenza HA. For example, the envelope proteins of the flaviviruses, such as dengue and West Nile viruses, undergo dimer-to-trimer rearrangements when suitably triggered by the low pH of an endosome. There are nonetheless common features of the way all known viral fusion proteins carry out their function, as demonstrated by our work on the dengue virus envelope protein and by work of other laboratories on proteins of related viruses. One such feature is the initial exposure, upon triggering, of a hydrophobic "fusion peptide" or "fusion loop," which inserts firmly into the target membrane. A second is refolding of the protein in such a way that its fusion peptide, inserted into a membrane of the cell, and its transmembrane anchor segment, embedded in the membrane of the virus, come together. The conformational change thus draws the two membranes against each other. In collaboration with the laboratory of Antoine van Oijen (Harvard Medical School), we are probing the intimate details of subsequent steps—lipid-bilayer changes that finally lead to a single, continuous membrane where formerly there were two.
Just as efforts to understand the molecular rearrangements in HIV-1 fusion have led to rational strategies for inhibiting viral entry, so we believe that our structural analysis of the dengue virus fusion protein reveals similar opportunities. We can identify at least two stages in the dimer-to-trimer reorganization of dengue virus envelope protein that are potential targets for small-molecule inhibitors. We have used the structural information to design protein-based screens of small-molecule libraries; these screens have yielded promising initial results.
Nonenveloped viruses face a somewhat different barrier to penetration: they have no membrane of their own, and they must therefore perforate a cellular membrane to gain access to the cytoplasm—either by injection of their genomic nucleic acid (the classic mechanism for tailed bacteriophages) or by translocation of a subviral particle that includes the RNA or DNA genome. The double-stranded RNA (dsRNA) viruses, such as reovirus and rotavirus, translocate a subviral particle—the "core" or "inner capsid particle" (ICP)—which remains intact within the cytoplasm. The ICP contains RNA polymerase and capping activities, and it synthesizes viral mRNA without needing to uncoat. Translocation from endosome to cytoplasm of this particle, roughly 700 Å in diameter, is the task of a specific viral outer-coat protein—μ1 of reovirus, VP4 (perhaps with help from VP7) of rotavirus. Structural studies of these proteins during the past five years (collaborations with Max Nibert [Harvard Medical School] and Philip Dormitzer [Children's Hospital]) have shown that like the fusion proteins of enveloped viruses, these penetration proteins undergo large-scale conformational changes, probably coupled to a membrane-insertion event. Our ongoing work is an effort to define these changes more precisely and to work out how they induce perforation of a bilayer. The structures of the reovirus and rotavirus polymerases have revealed additional mechanistic features of the steps in viral replication that directly follow penetration into the cytoplasm.
Multiprotein Assemblies on DNA
The switching mechanisms that underlie transcriptional regulation in eukaryotes embody relatively complex logical circuits. They achieve this complexity by relying on cooperative assembly of "generic" transcription factors into specific superstructures, guided by recognition of binding sites on chromosomal DNA. One such superstructure, the association of six distinct proteins (two of them in double copy) on a 55-bp segment of DNA (the interferon-β enhancer), is sometimes called the interferon-β enhanceosome, to emphasize that it functions as an organized entity. We have determined a high-resolution structure of the complete DNA-proximal part of this enhanceosome—that is, of the eight DNA-binding domains arrayed on the enhancer—by crystallographic analysis of three overlapping parts. (This project is a collaboration with Tom Maniatis, Harvard University.) Binding of one protein can influence the registration of another on the enhancer—that is, a particular transcription factor may "prefer" to bind in one way on its own, but in a "second choice" way (displaced by a base pair, for example) as part of the assembly. Thus, the enhancer is a concerted unit of information. Although the DNA binding domains are densely arrayed along the DNA, their association is not cooperative. Functional cooperativity (synergy) is probably provided at a different level, by interaction with coactivators such as CBP/p300.
The protein/DNA complex known as the kinetochore links chromosomes to the microtubules of the mitotic spindle. The kinetochores of budding yeast cells appear to be elementary, unimodular versions of the multimodule kinetochores of metazoan cells. Most of the 50–80 distinct protein species that make up a yeast kinetochore are organized into well-defined subcomplexes. Together with the laboratory of Peter Sorger (Harvard Medical School), we have been studying the structures of these subcomplexes. Members of one class—e.g., the four-protein complex known as CBF3 and the homodimeric protein Mif2p—bind directly to centromeric DNA. Structures for parts of each of these have emerged from our crystallographic work during the past year. An initial analysis suggests that these components resemble transcription factors subverted for recognizing a centromere instead of an upstream binding site.
Members of another class of kinetochore subcomplexes associate directly or indirectly with microtubules. The DASH/Dam1 complex contains 10 distinct proteins; coexpression of all 10 chains in bacterial cells yields a homogeneous, heterodecameric assembly, which binds microtubules by forming rings or spirals around them. We have proposed that this unusual mode of binding may reveal the mechanism by which kinetochore microtubules can grow or shrink without losing the attached chromosome.
A third class of kinetochore components link the DNA- and microtubule-binding layers. The heterotetrameric Ndc80 complex is one such linker. It contains two heterodimers (Ndc80p/Nuf2p and Spc24p/Spc25p), joined end-to-end. Each of the heterodimers has an α-helical coiled-coil shaft and a globular end. Crystal structures of the globular ends suggest how the Ndc80p end interacts with microtubules, and the Spc24p end, with centromere-associated proteins.
Carriers of Membrane Traffic
Vesicular membrane traffic transports proteins and lipids from one membrane-bound compartment to another, while maintaining the functions and biochemical heterogeneity of the donor and acceptor membranes. In collaboration with Tomas Kirchhausen (CBR Institute and Harvard Medical School), we have studied for many years the structures of proteins associated with clathrin-mediated endocytosis, the pathway by which many viruses and various receptors enter a cell. These are particularly favorable objects for linking structure and biochemistry with intracellular dynamics, because the time course of their formation, the timing of their dissolution, and the fate of their cargo can be followed by sensitive live-cell imaging. Clathrin organizes these miniorganelles by forming a cage-like lattice around the membrane vesicle it engulfs. We have combined our earlier crystallographic analysis of fragments of clathrin with interpretation of cryo-EM images of in vitro assembled clathrin coats, to obtain a full molecular model of the lattice (in collaboration with Thomas Walz [Harvard Medical School] and Nikolaus Grigorieff [HHMI, Brandeis University]). The structure suggests mechanisms for regulation of assembly and disassembly of clathrin coats. We are now using cryo-EM to determine additional structures, with the short-term goal of understanding the mechanism by which the Hsc70 ATPase, recruited to coats by the protein auxilin, actively takes apart their clathrin lattice.