Illuminating protein space with a programmable generative model

Three billion years of evolution has produced a tremendous diversity of protein molecules1, but the full potential of proteins is likely to be much greater. Accessing this potential has been challenging for both computation and experiments because the space of possible protein molecules is much larger than the space of those likely to have functions. Here we introduce Chroma, a generative model for proteins and protein complexes that can directly sample novel protein structures and sequences, and that can be conditioned to steer the generative process towards desired properties and functions. To enable this, we introduce a diffusion process that respects the conformational statistics of polymer ensembles, an efficient neural architecture for molecular systems that enables long-range reasoning with sub-quadratic scaling, layers for efficiently synthesizing three-dimensional structures of proteins from predicted inter-residue geometries and a general low-temperature sampling algorithm for diffusion models. Chroma achieves protein design as Bayesian inference under external constraints, which can involve symmetries, substructure, shape, semantics and even natural-language prompts. The experimental characterization of 310 proteins shows that sampling from Chroma results in proteins that are highly expressed, fold and have favourable biophysical properties. The crystal structures of two designed proteins exhibit atomistic agreement with Chroma samples (a backbone root-mean-square deviation of around 1.0 Å). With this unified approach to protein design, we hope to accelerate the programming of protein matter to benefit human health, materials science and synthetic biology.

A highly potent CD73 biparatopic antibody blocks organization of the enzyme active site through dual mechanisms

The dimeric ectonucleotidase CD73 catalyzes the hydrolysis of AMP at the cell surface to form adenosine, a potent suppressor of the immune response. Blocking CD73 activity in the tumor microenvironment can have a beneficial effect on tumor eradication and is a promising approach for cancer therapy. Biparatopic antibodies binding different regions of CD73 may be a means to antagonize its enzymatic activity. A panel of biparatopic antibodies representing the pairwise combination of 11 parental monoclonal antibodies against CD73 was generated by Fab-arm exchange. Nine variants vastly exceeded the potency of their parental antibodies with ≥90% inhibition of activity and subnanomolar EC50 values. Pairing the Fabs of parents with nonoverlapping epitopes was both sufficient and necessary whereas monovalent antibodies were poor inhibitors. Some parental antibodies yielded potent biparatopics with multiple partners, one of which (TB19) producing the most potent. The structure of the TB19 Fab with CD73 reveals that it blocks alignment of the N- and C-terminal CD73 domains necessary for catalysis. A separate structure of CD73 with a Fab (TB38) which complements TB19 in a particularly potent biparatopic shows its binding to a nonoverlapping site on the CD73 N-terminal domain. Structural modeling demonstrates a TB19/TB38 biparatopic antibody would be unable to bind the CD73 dimer in a bivalent manner, implicating crosslinking of separate CD73 dimers in its mechanism of action. This ability of a biparatopic antibody to both crosslink CD73 dimers and fix them in an inactive conformation thus represents a highly effective mechanism for the inhibition of CD73 activity.

Trispecific antibody against HIV-1 demonstrate anti-viral activity in vivo and have potent Fc-dependent effector functions

Broadly neutralizing antibodies (bNAbs) against HIV-1 have been suggested as a complementary immunotherapy to current combination small molecule anti-retroviral therapies (cART) for treatment of HIV-1 infection. Monoclonal bNAbs when used as a monotherapy demonstrate high susceptibility for escape in HIV-1 infected individuals, thus a trispecific bNAb targeting three independent HIV-1 envelope determinants was developed to achieve more robust suppression. Upon treatment of viremic simian-human immunodeficiency virus (SHIV)-infected macaques with the trispecific bNAb, plasma viral suppression of up to 1000-fold was observed. To elucidate a mechanism for the observed plasma viremia control, in vitro assays using the trispecific bNAb were conducted to determine Fc-mediated effector functions. First, we determined that the trispecific bNAb bound to HIV-1 infected cell lines with an equal or higher affinity than its parental bNAbs alone. We then conducted antibody-dependent cell cytotoxicity (ADCC) and antibody-dependent cell phagocytosis (ADCP) assays and found that the trispecific bNAb had comparable or better activity compared to its parental bNAbs alone. Similarly, when we measured direct virolysis of HIV-1 virus using an antibody-dependent complement-mediated virus lysis (ADCML) assay, the trispecific bNAb had comparable or better activity compared to its parental bNAbs alone. These studies demonstrate that the trispecific bNAb shows anti-viral activity in vivo and has potent Fc-dependent effector functions that target both free virus and virally infected cells compared to monoclonal bNAbs alone. Thus, the trispecific bNAb is a promising single immunotherapeutic protein for treatment of HIV-1 infection.

A peculiar IclR family transcription factor regulates para-hydroxybenzoate catabolism in Streptomyces coelicolor

In Streptomyces coelicolor, we identified a para-hydroxybenzoate (PHB) hydroxylase, encoded by gene pobA (SCO3084), which is responsible for conversion of PHB into PCA (protocatechuic acid), a substrate of the β-ketoadipate pathway which yields intermediates of the Krebs cycle. We also found that the transcription of pobA is induced by PHB and is negatively regulated by the product of SCO3209, which we named PobR. The product of this gene is highly unusual in that it is the apparent fusion of two IclR family transcription factors. Bioinformatic analyses, in vivo transcriptional assays, electrophoretic mobility shift assays (EMSAs), DNase I footprinting, and isothermal calorimetry (ITC) were used to elucidate the regulatory mechanism of PobR. We found that PobR loses its high affinity for DNA (i.e., the pobA operator) in the presence of PHB, the inducer of pobA transcription. PHB binds to PobR with a KD of 5.8 μM. Size-exclusion chromatography revealed that PobR is a dimer in the absence of PHB and a monomer in the presence of PHB. The crystal structure of PobR in complex with PHB showed that only one of the two IclR ligand binding domains was occupied, and defined how the N-terminal ligand binding domain engages the effector ligand.

Structure-based engineering to restore high affinity binding of an isoform-selective anti-TGFβ1 antibody

Metelimumab (CAT192) is a human IgG4 monoclonal antibody developed as a TGFβ1-specific antagonist. It was tested in clinical trials for the treatment of scleroderma but later terminated due to lack of efficacy. Subsequent characterization of CAT192 indicated that its TGFβ1 binding affinity was reduced by ∼50-fold upon conversion from the parental single-chain variable fragment (scFv) to IgG4. We hypothesized this result was due to decreased conformational flexibility of the IgG that could be altered via engineering. Therefore, we designed insertion mutants in the elbow region and screened for binding and potency. Our results indicated that increasing the elbow region linker length in each chain successfully restored the isoform-specific and high affinity binding of CAT192 to TGFβ1. The crystal structure of the high binding affinity mutant displays large conformational rearrangements of the variable domains compared to the wild-type antigen-binding fragment (Fab) and the low binding affinity mutants. Insertion of two glycines in both the heavy and light chain elbow regions provided sufficient flexibility for the variable domains to extend further apart than the wild-type Fab, and allow the CDR3s to make additional interactions not seen in the wild-type Fab structure. These interactions coupled with the dramatic conformational changes provide a possible explanation of how the scFv and elbow-engineered Fabs bind TGFβ1 with high affinity. This study demonstrates the benefits of re-examining both structure and function when converting scFv to IgG molecules, and highlights the potential of structure-based engineering to produce fully functional antibodies.

Trispecific broadly neutralizing HIV antibodies mediate potent SHIV protection in macaques

The development of an effective AIDS vaccine has been challenging because of viral genetic diversity and the difficulty of generating broadly neutralizing antibodies (bnAbs). We engineered trispecific antibodies (Abs) that allow a single molecule to interact with three independent HIV-1 envelope determinants: the CD4 binding site, the membrane-proximal external region (MPER), and the V1V2 glycan site. Trispecific Abs exhibited higher potency and breadth than any previously described single bnAb, showed pharmacokinetics similar to those of human bnAbs, and conferred complete immunity against a mixture of simian-human immunodeficiency viruses (SHIVs) in nonhuman primates, in contrast to single bnAbs. Trispecific Abs thus constitute a platform to engage multiple therapeutic targets through a single protein, and they may be applicable for treatment of diverse diseases, including infections, cancer, and autoimmunity.

The MqsR/MqsA toxin/antitoxin system protects Escherichia coli during bile acid stress

Toxin/antitoxin (TA) systems are ubiquitous within bacterial genomes, and the mechanisms of many TA systems are well characterized. As such, several roles for TA systems have been proposed, such as phage inhibition, gene regulation and persister cell formation. However, the significance of these roles is nebulous due to the subtle influence from individual TA systems. For example, a single TA system has only a minor contribution to persister cell formation. Hence, there is a lack of defining physiological roles for individual TA systems. In this study, phenotype assays were used to determine that the MqsR/MqsA type II TA system of Escherichia coli is important for cell growth and tolerance during stress from the bile salt deoxycholate. Using transcriptomics and purified MqsR, we determined that endoribonuclease toxin MqsR degrades YgiS mRNA, which encodes a periplasmic protein that promotes deoxycholate uptake and reduces tolerance to deoxycholate exposure. The importance of reducing YgiS mRNA by MqsR is evidenced by improved growth, reduced cell death and reduced membrane damage when cells without ygiS are stressed with deoxycholate. Therefore, we propose that MqsR/MqsA is physiologically important for E. coli to thrive in the gallbladder and upper intestinal tract, where high bile concentrations are prominent.

McbR/YncC: implications for the mechanism of ligand and DNA binding by a bacterial GntR transcriptional regulator involved in biofilm formation

MqsR-controlled colanic acid and biofilm regulator (McbR, also known as YncC) is the protein product of a highly induced gene in early Escherichia coli biofilm development and has been regarded as an attractive target for blocking biofilm formation. This protein acts as a repressor for genes involved in exopolysaccharide production and an activator for genes involved in stress response. To better understand the role of McbR in governing the switch from exponential growth to the biofilm state, we determined the crystal structure of McbR to 2.1 Å. The structure reveals McbR to be a member of the FadR C-terminal domain (FCD) family of the GntR superfamily of transcriptional regulators (this family was named after the first identified member, GntR, a transcriptional repressor of the gluconate operon of Bacillus subtilis). Previous to this study, only six of the predicted 2800 members of this family had been structurally characterized. Here, we identify the residues that constitute the McbR effector and DNA binding sites. In addition, comparison of McbR with other members of the FCD domain family shows that this family of proteins adopts highly distinct oligomerization interfaces, which has implications for DNA binding and regulation.

Type II toxin/antitoxin MqsR/MqsA controls type V toxin/antitoxin GhoT/GhoS

Toxin endoribonucleases of toxin/antitoxin (TA) systems regulate protein production by selectively degrading mRNAs but have never been shown to control other TA systems. Here we demonstrate that toxin MqsR of the MqsR/MqsA system enriches toxin ghoT mRNA in vivo and in vitro, since this transcript lacks the primary MqsR cleavage site 5'-GCU. GhoT is a membrane toxin that causes the ghost cell phenotype, and is part of a type V TA system with antitoxin GhoS that cleaves specifically ghoT mRNA. Introduction of MqsR primary 5'-GCU cleavage sites into ghoT mRNA reduces ghost cell production and cell death likely due to increased degradation of the altered ghoT mRNA by MqsR. GhoT also prevents cell elongation upon the addition of low levels of ampicillin. Therefore, during stress, antitoxin GhoS mRNA is degraded by toxin MqsR allowing ghoT mRNA translation to yield another free toxin that forms ghost cells and increases persistence. Hence, we show that GhoT/GhoS is the first TA system regulated by another TA system.

The Escherichia coli toxin MqsR destabilizes the transcriptional repression complex formed between the antitoxin MqsA and the mqsRA operon promoter

Bacterial biofilms are complex communities of cells containing an increased prevalence of dormant cells known as persisters, which are characterized by an up-regulation of genes known as toxin-antitoxin (TA) modules. The association of toxins with their cognate antitoxins neutralizes toxin activity, allowing for normal cell growth. Additionally, protein antitoxins bind their own promoters and repress transcription, whereas the toxins serve as co-repressors. Recently, TA pairs have been shown to regulate their own transcription through a phenomenon known as conditional cooperativity, where the TA complexes bind operator DNA and repress transcription only when present in the proper stoichiometric amounts. The most differentially up-regulated gene in persister cells is mqsR, a gene that, with the antitoxin mqsA, constitutes a TA module. Here, we reveal that, unlike other TA systems, MqsR is not a transcription co-repressor but instead functions to destabilize the MqsA-DNA complex. We further show that DNA binding is not regulated by conditional cooperativity. Finally, using biophysical studies, we show that complex formation between MqsR and MqsA results in an exceptionally stable interaction, resulting in a subnanomolar dissociation constant that is similar to that observed between MqsA and DNA. In combination with crystallographic studies, this work reveals that MqsA binding to DNA and MqsR is mutually exclusive. To our knowledge, this is the first TA system in which the toxin does not function as a transcriptional co-repressor, but instead functions to destabilize the antitoxin-operator complex under all conditions, and thus defines another unique feature of the mqsRA TA module.

Demonstration of ignition radiation temperatures in indirect-drive inertial confinement fusion hohlraums

We demonstrate the hohlraum radiation temperature and symmetry required for ignition-scale inertial confinement fusion capsule implosions. Cryogenic gas-filled hohlraums with 2.2 mm-diameter capsules are heated with unprecedented laser energies of 1.2 MJ delivered by 192 ultraviolet laser beams on the National Ignition Facility. Laser backscatter measurements show that these hohlraums absorb 87% to 91% of the incident laser power resulting in peak radiation temperatures of T(RAD)=300 eV and a symmetric implosion to a 100 μm diameter hot core.

Multiple modes of hepatitis A virus transmission among methamphetamine users

Methamphetamine users are at increased risk of hepatitis A, but modes of transmission are unclear. The authors conducted a case-control study among methamphetamine users during an outbreak in Iowa in 1997. Twenty-eight reported, laboratory-confirmed, hepatitis A cases did not differ from 18 susceptible controls with respect to age, sex, or number of doses used. When compared with controls in multivariate analysis, case-patients were more likely to have injected methamphetamine (odds ratio (OR) = 5.5, 95% confidence interval (CI): 1.1, 27), to have used methamphetamine with another case-patient (OR = 6.2, 95% CI: 0.95, 41), and to have used brown methamphetamine (OR = 5.5, 95% CI: 0.51, 59). Receptive needle sharing was reported by 10 of the 20 case-patients who injected. Methamphetamine use with another case-patient was also associated with hepatitis A in an analysis restricted to noninjectors (OR = 17, 95% CI: 1.0, 630). During this outbreak, hepatitis A may have been transmitted from person to person among methamphetamine users through the fecal-oral and the percutaneous routes. Methamphetamine users should be vaccinated against hepatitis A and should be given immune globulin if they used methamphetamine with a case-patient in the last 2 weeks. Persons who intend to continue using methamphetamine should be advised about safer practices.