De novo design of protein structure and function with RFdiffusion.

There has been considerable recent progress in designing new proteins using deep-learning methods1-9. Despite this progress, a general deep-learning framework for protein design that enables solution of a wide range of design challenges, including de novo binder design and design of higher-order symmetric architectures, has yet to be described. Diffusion models10,11 have had considerable success in image and language generative modelling but limited success when applied to protein modelling, probably due to the complexity of protein backbone geometry and sequence-structure relationships. Here we show that by fine-tuning the RoseTTAFold structure prediction network on protein structure denoising tasks, we obtain a generative model of protein backbones that achieves outstanding performance on unconditional and topology-constrained protein monomer design, protein binder design, symmetric oligomer design, enzyme active site scaffolding and symmetric motif scaffolding for therapeutic and metal-binding protein design. We demonstrate the power and generality of the method, called RoseTTAFold diffusion (RFdiffusion), by experimentally characterizing the structures and functions of hundreds of designed symmetric assemblies, metal-binding proteins and protein binders. The accuracy of RFdiffusion is confirmed by the cryogenic electron microscopy structure of a designed binder in complex with influenza haemagglutinin that is nearly identical to the design model. In a manner analogous to networks that produce images from user-specified inputs, RFdiffusion enables the design of diverse functional proteins from simple molecular specifications.

Structural transitions in influenza haemagglutinin at membrane fusion pH.

Infection by enveloped viruses involves fusion of their lipid envelopes with cellular membranes to release the viral genome into cells. For HIV, Ebola, influenza and numerous other viruses, envelope glycoproteins bind the infecting virion to cell-surface receptors and mediate membrane fusion. In the case of influenza, the receptor-binding glycoprotein is the haemagglutinin (HA), and following receptor-mediated uptake of the bound virus by endocytosis1, it is the HA that mediates fusion of the virus envelope with the membrane of the endosome2. Each subunit of the trimeric HA consists of two disulfide-linked polypeptides, HA1 and HA2. The larger, virus-membrane-distal, HA1 mediates receptor binding; the smaller, membrane-proximal, HA2 anchors HA in the envelope and contains the fusion peptide, a region that is directly involved in membrane interaction3. The low pH of endosomes activates fusion by facilitating irreversible conformational changes in the glycoprotein. The structures of the initial HA at neutral pH and the final HA at fusion pH have been investigated by electron microscopy4,5 and X-ray crystallography6-8. Here, to further study the process of fusion, we incubate HA for different times at pH 5.0 and directly image structural changes using single-particle cryo-electron microscopy. We describe three distinct, previously undescribed forms of HA, most notably a 150 Å-long triple-helical coil of HA2, which may bridge between the viral and endosomal membranes. Comparison of these structures reveals concerted conformational rearrangements through which the HA mediates membrane fusion.

A particle-filter framework for robust cryo-EM 3D reconstruction.

Single-particle electron cryomicroscopy (cryo-EM) involves estimating a set of parameters for each particle image and reconstructing a 3D density map; robust algorithms with accurate parameter estimation are essential for high resolution and automation. We introduce a particle-filter algorithm for cryo-EM, which provides high-dimensional parameter estimation through a posterior probability density function (PDF) of the parameters given in the model and the experimental image. The framework uses a set of random support points to represent such a PDF and assigns weighting coefficients not only among the parameters of each particle but also among different particles. We implemented the algorithm in a new program named THUNDER, which features self-adaptive parameter adjustment, tolerance to bad particles, and per-particle defocus refinement. We tested the algorithm by using cryo-EM datasets for the cyclic-nucleotide-gated (CNG) channel, the proteasome, ß-galactosidase, and an influenza hemagglutinin (HA) trimer, and observed substantial improvement in resolution.

Addressing preferred specimen orientation in single-particle cryo-EM through tilting.

We present a strategy for tackling preferred specimen orientation in single-particle cryogenic electron microscopy by employing tilts during data collection. We also describe a tool to quantify the resulting directional resolution using 3D Fourier shell correlation volumes. We applied these methods to determine the structures at near-atomic resolution of the influenza hemagglutinin trimer, which adopts a highly preferred specimen orientation, and of ribosomal biogenesis intermediates, which adopt moderately preferred orientations.

Computational design of trimeric influenza-neutralizing proteins targeting the hemagglutinin receptor binding site.

Many viral surface glycoproteins and cell surface receptors are homo-oligomers, and thus can potentially be targeted by geometrically matched homo-oligomers that engage all subunits simultaneously to attain high avidity and/or lock subunits together. The adaptive immune system cannot generally employ this strategy since the individual antibody binding sites are not arranged with appropriate geometry to simultaneously engage multiple sites in a single target homo-oligomer. We describe a general strategy for the computational design of homo-oligomeric protein assemblies with binding functionality precisely matched to homo-oligomeric target sites. In the first step, a small protein is designed that binds a single site on the target. In the second step, the designed protein is assembled into a homo-oligomer such that the designed binding sites are aligned with the target sites. We use this approach to design high-avidity trimeric proteins that bind influenza A hemagglutinin (HA) at its conserved receptor binding site. The designed trimers can both capture and detect HA in a paper-based diagnostic format, neutralizes influenza in cell culture, and completely protects mice when given as a single dose 24 h before or after challenge with influenza.

CryoEM Structure of an Influenza Virus Receptor-Binding Site Antibody-Antigen Interface.

Structure-based vaccine design depends on extensive structural analyses of antigen-antibody complexes.Single-particle electron cryomicroscopy (cryoEM) can circumvent some of the problems of x-ray crystallography as a pipeline for obtaining the required structures. We have examined the potential of single-particle cryoEM for determining the structure of influenza-virus hemagglutinin (HA):single-chain variable-domain fragment complexes, by studying a complex we failed to crystallize in pursuing an extended project on the human immune response to influenza vaccines.The result shows that a combination of cryoEM and molecular modeling can yield details of the antigen-antibody interface, although small variation in the twist of the rod-likeHA trimer limited the overall resolution to about 4.5Å.Comparison of principal 3D classes suggests ways to modify the HA trimer to overcome this limitation. A closely related antibody from the same donor did yield crystals when bound with the same HA, giving us an independent validation of the cryoEM results.The two structures also augment our understanding of receptor-binding site recognition by antibodies that neutralize a wide range of influenza-virus variants.

Cryo-electron Microscopy Structures of Chimeric Hemagglutinin Displayed on a Universal Influenza Vaccine Candidate.

UNLABELLED: Influenza viruses expressing chimeric hemagglutinins (HAs) are important tools in the quest for a universal vaccine. Using cryo-electron tomography, we have determined the structures of a chimeric HA variant that comprises an H1 stalk and an H5 globular head domain (cH5/1 HA) in native and antibody-bound states. We show that cH5/1 HA is structurally different from native HA, displaying a 60° rotation between the stalk and head groups, leading to a novel and unexpected "open" arrangement of HA trimers. cH5/1N1 viruses also display higher glycoprotein density than pH1N1 or H5N1 viruses, but despite these differences, antibodies that target either the stalk or head domains of hemagglutinins still bind to cH5/1 HA with the same consequences as those observed with native H1 or H5 HA. Our results show that a large range of structural plasticity can be tolerated in the chimeric spike scaffold without disrupting structural and geometric aspects of antibody binding. IMPORTANCE: Chimeric hemagglutinin proteins are set to undergo human clinical trials as a universal influenza vaccine candidate, yet no structural information for these proteins is available. Using cryo-electron tomography, we report the first three-dimensional (3D) visualization of chimeric hemagglutinin proteins displayed on the surface of the influenza virus. We show that, unexpectedly, the chimeric hemagglutinin structure differs from those of naturally occurring hemagglutinins by displaying a more open head domain and a dramatically twisted head/stalk arrangement. Despite this unusual spatial relationship between head and stalk regions, virus preparations expressing the chimeric hemagglutinin are fully infectious and display a high glycoprotein density, which likely helps induction of a broadly protective immune response.

Actin mediates the nanoscale membrane organization of the clustered membrane protein influenza hemagglutinin.

The influenza viral membrane protein hemagglutinin (HA) is required at high concentrations on virion and host-cell membranes for infectivity. Because the role of actin in membrane organization is not completely understood, we quantified the relationship between HA and host-cell actin at the nanoscale. Results obtained using superresolution fluorescence photoactivation localization microscopy (FPALM) in nonpolarized cells show that HA clusters colocalize with actin-rich membrane regions (ARMRs). Individual molecular trajectories in live cells indicate restricted HA mobility in ARMRs, and actin disruption caused specific changes to HA clustering. Surprisingly, the actin-binding protein cofilin was excluded from some regions within several hundred nanometers of HA clusters, suggesting that HA clusters or adjacent proteins within the same clusters influence local actin structure. Thus, with the use of imaging, we demonstrate a dynamic relationship between glycoprotein membrane organization and the actin cytoskeleton at the nanoscale.

Importance of viral genomic composition in modulating glycoprotein content on the surface of influenza virus particles.

Despite progress in our knowledge of the internal organisation of influenza virus particles, little is known about the determinants of their morphology and, more particularly, of the actual abundance of structural proteins at the virion level. To address these issues, we used cryo-EM to focus on viral (and host) factors that might account for observed differences in virion morphology and characteristics such as size, shape and glycoprotein (GP) spike density. Twelve recombinant viruses were characterised in terms of their morphology, neuraminidase activity and virus growth. The genomic composition was shown to be important in determining the GP spike density. In particular, polymerase gene segments and especially PB1/PB2 were shown to have a prominent influence in addition to that for HA in determining GP spike density, a feature consistent with a functional link between these virus components important for virus fitness.

Source of high pathogenicity of an avian influenza virus H5N1: why H5 is better cleaved by furin.

The origin of the high pathogenicity of an emerging avian influenza H5N1 due to the -RRRKK- insertion at the cleavage loop of the hemagglutinin H5, was studied using the molecular dynamics technique, in comparison with those of the noninserted H5 and H3 bound to the furin (FR) active site. The cleavage loop of the highly pathogenic H5 was found to bind strongly to the FR cavity, serving as a conformation suitable for the proteolytic reaction. With this configuration, the appropriate interatomic distances were found for all three reaction centers of the enzyme-substrate complex: the arrangement of the catalytic triad, attachment of the catalytic Ser(368) to the reactive S1-Arg, and formation of the oxyanion hole. Experimentally, the--RRRKK--insertion was also found to increase in cleavage of hemagglutinin by FR. The simulated data provide a clear answer to the question of why inserted H5 is better cleaved by FR than the other subtypes, explaining the high pathogenicity of avian influenza H5N1.

Why does avian influenza A virus hemagglutinin bind to avian receptor stronger than to human receptor? Ab initio fragment molecular orbital studies.

Influenza A viruses attach to alpha-sialosides on the target cell surface by their hemagglutinins, which strictly recognize the difference in sialic acid-galactose linkage. Why does avian virus H3 subtype bind to avian receptor Neu5Ac(alpha2-3)Gal stronger than to human receptor Neu5Ac(alpha2-6)Gal? Why does avian H3 mutated Gln226 to Leu preferentially bind to human receptor? In this paper, we theoretically answer the questions by molecular mechanics and ab initio fragment molecular orbital (FMO) calculations. The binding energy between avian H3 and avian receptor is 8.2kcal/mol larger than that of the avian H3-human receptor complex estimated at the FMO-HF/STO-3G level, which is a reason that avian H3 binds to avian receptor stronger than to human receptor. Avian Leu226 H3 clashes to Gal unit on the avian receptor to quite decrease its binding affinity. In contrast, Gal unit on the human receptor forms intermolecular hydrophobic interaction with avian Leu226 H3 to afford moderate binding affinity.

Electrostatic energy calculation on the pH-induced conformational change of influenza virus hemagglutinin.

The pH-induced conformational change of influenza virus hemagglutinin (HA) has been investigated by calculating the change of electrostatic energy of the fragment of HA2 upon pH change. The average charge and electrostatic free energy are calculated as a function of pH for the fusion peptide (residues 1-20 of HA2) and the polypeptide of residues 54-77 of HA2 by using the finite difference Poisson-Boltzmann method. It is found that as pH decreases from 8 to 5, the electrostatic free energy of the fusogenic state is lowered by approximately 2 kcal/mol and the fusogenic state is less ionized compared to that of the native state for both polypeptides. For the fusion peptide at the fusogenic state, most of ionizable residues are neutral at acidic pH except Glu-11. For the polypeptide of residues 54-77 at the fusogenic state, most of residues except Glu-74 and His-64 are fully charged between pH 5 and pH 8.

Full-length influenza hemagglutinin HA2 refolds into the trimeric low-pH-induced conformation.

The influenza virus uses hemagglutinin (HA) to fuse the viral and cellular membranes. As part of an effort to study the membrane-interacting elements of HA, the fusion peptide, and the C-terminal transmembrane anchor, we have expressed in Escherichia coli the full-length HA(2) chain with maltose-binding protein fused at its N-terminus. The chimeric protein can be refolded in vitro in the presence of specific detergents to yield stable, homogeneous trimers, as determined by analytical ultracentrifugation. The trimers have the so-called "low pH" conformation-the rearranged HA(2) conformation obtained when intact HA(1)/HA(2) is induced to refold by exposure to low pH-as detected by electron microscopy and monoclonalantibody reactivity. These results provide further evidence for the notion that the neutral-pH structure of intact HA is metastable and that binding of protons lowers the kinetic barriers that prevent rearrangement to the minimum-free-energy conformation. The refolded chimeric protein described here is a suitable species for undertaking studies of how the fusion peptide inserts into membranes and assessing the nature of possible intermediates in the fusion process.

Membrane fusion mechanisms: the influenza hemagglutinin paradigm and its implications for intracellular fusion.

The mechanism of membrane fusion induced by the influenza virus hemagglutinin (HA) has been extensively characterized. Fusion is triggered by low pH, which induces conformational changes in the protein, leading to insertion of a hydrophobic 'fusion peptide' into the viral membrane and the target membrane for fusion. Insertion perturbs the target membrane, and hour glass-shaped lipidic fusion intermediates, called stalks, fusing the outer monolayers of the two membranes, are formed. Stalk formation is followed by complete fusion of the two membranes. Structures similar to those formed by HA at the pH of fusion are found not only in many other viral fusion proteins, but are also formed by SNAREs, proteins involved in intracellular fusion. Substances that inhibit or promote HA-induced fusion because they affect stalk formation, also inhibit or promote intracellular fusion, cell-cell fusion and even intracellular fission similarly. Therefore, the mechanism of influenza HA-induced fusion may be a paradigm for many intracellular fusion events.

The mechanism for low-pH-induced clustering of phospholipid vesicles carrying the HA2 ectodomain of influenza hemagglutinin.

Homotrimeric hemagglutinin (HA) is one of the major spike membrane glycoproteins of the influenza virus. Initial pH-triggered conformational changes in the target membrane-interacting HA2 domain are necessary for a preliminary step in membrane fusion. Using spin-labeling electron paramagnetic resonance (EPR) spectroscopy, we examined subsequent pH-dependent changes of a membrane-bound HA2 construct (FHA2, aa 1-127). Residues 91-94, 108-115, 122, and 125 were mutated to cysteine and spin-labeled. Low solvent accessibility and side chain mobility were observed by EPR at positions 91-94, 122, and 125. Spin-labels at residues 108-115 were solvent-exposed and highly mobile, revealing the presence of a flexible loop. These results are consistent with the low-pH crystal structure of a truncated HA2 domain, particularly the unusual kink loop at residues 108-115 [Bullough et al. (1994) Nature (London) 371, 37-43]. Most interestingly, at endosomal pH, spin-labels at 108-115 become immobile and no longer solvent-exposed, and this change is reversible upon reneutralization. However, little change in the EPR line shape and accessibility of spin-labels was observed in other regions. This observation implies that the FHA2 trimers interact reversibly via this specific loop, most likely in an intermolecular fashion. Furthermore, this interaction correlates well with a reversible pH-dependent clustering of FHA2-bearing vesicles evidenced by the reversible increase in turbidity and further confirmed in detail by electron microscopy. The implications of this reversible, pH-dependent interaction between FHA2 trimers are discussed in light of recent fusion models.

Structure-based identification of an inducer of the low-pH conformational change in the influenza virus hemagglutinin: irreversible inhibition of infectivity.

Past efforts to employ a structure-based approach to design an inhibitor of the fusion-inducing conformational change in the influenza virus hemagglutinin (HA) yielded a family of small benzoquinones and hydroquinones. The most potent of these, tert-butyl hydroquinone (TBHQ), inhibits both the conformational change in HA from strain X:31 influenza virus and viral infectivity in tissue culture cells with 50% inhibitory concentrations in the micromolar range (D. L. Bodian, R. B. Yamasaki, R. L. Buswell, J. F. Stearns, J. M. White, and I. D. Kuntz, Biochemistry 32:2967-2978, 1993). A new structure-based inhibitor design search was begun which involved (i) the recently refined crystal structure (2.1-A resolution) of the HA ectodomain, (ii) new insights into the conformational change, and (iii) improvements in the molecular docking program, DOCK. As a result, we identified new inhibitors of HA-mediated membrane fusion. Like TBHQ, most of these molecules inhibit the conformational change. One of the new compounds, however, facilitates rather than inhibits the HA conformational change. Nonetheless, the facilitator, diiodofluorescein, inhibits HA-mediated membrane fusion and, irreversibly, infectivity. We further characterized the effects of inhibitors from both searches on the conformational change and membrane fusion activity of HA as well as on viral infectivity. We also isolated and characterized several mutants resistant to each class of inhibitor. The implications of our results for HA-mediated membrane fusion, anti-influenza virus therapy, and structure-based inhibitor design are discussed.