Unsupervised evolution of protein and antibody complexes with a structure-informed language model.

Large language models trained on sequence information alone can learn high-level principles of protein design. However, beyond sequence, the three-dimensional structures of proteins determine their specific function, activity, and evolvability. Here, we show that a general protein language model augmented with protein structure backbone coordinates can guide evolution for diverse proteins without the need to model individual functional tasks. We also demonstrate that ESM-IF1, which was only trained on single-chain structures, can be extended to engineer protein complexes. Using this approach, we screened about 30 variants of two therapeutic clinical antibodies used to treat severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection. We achieved up to 25-fold improvement in neutralization and 37-fold improvement in affinity against antibody-escaped viral variants of concern BQ.1.1 and XBB.1.5, respectively. These findings highlight the advantage of integrating structural information to identify efficient protein evolution trajectories without requiring any task-specific training data.

Engineering a SARS-CoV-2 vaccine targeting the RBD cryptic-face via immunofocusing.

The receptor-binding domain (RBD) of the SARS-CoV-2 spike protein is the main target of neutralizing antibodies. Although they are infrequently elicited during infection or vaccination, antibodies that bind to the conformation-specific cryptic face of the RBD display remarkable breadth of binding and neutralization across Sarbecoviruses. Here, we employed the immunofocusing technique PMD (protect, modify, deprotect) to create RBD immunogens (PMD-RBD) specifically designed to focus the antibody response towards the cryptic-face epitope recognized by the broadly neutralizing antibody S2X259. Immunization with PMD-RBD antigens induced robust binding titers and broad neutralizing activity against homologous and heterologous Sarbecovirus strains. A serum-depletion assay provided direct evidence that PMD successfully skewed the polyclonal antibody response towards the cryptic face of the RBD. Our work demonstrates the ability of PMD to overcome immunodominance and refocus humoral immunity, with implications for the development of broader and more resilient vaccines against current and emerging viruses with pandemic potential.

Combining Cell-Intrinsic and -Extrinsic Resistance to HIV-1 By Engineering Hematopoietic Stem Cells for CCR5 Knockout and B Cell Secretion of Therapeutic Antibodies.

Autologous transplantation of CCR5 null hematopoietic stem and progenitor cells (HSPCs) is the only known cure for HIV-1 infection. However, this treatment is limited because of the rarity of CCR5 -null matched donors, the morbidities associated with allogeneic transplantation, and the prevalence of HIV-1 strains resistant to CCR5 knockout (KO) alone. Here, we propose a one-time therapy through autologous transplantation of HSPCs genetically engineered ex vivo to produce both CCR5 KO cells and long-term secretion of potent HIV-1 inhibiting antibodies from B cell progeny. CRISPR-Cas9-engineered HSPCs maintain engraftment capacity and multi-lineage potential in vivo and can be engineered to express multiple antibodies simultaneously. Human B cells engineered to express each antibody secrete neutralizing concentrations capable of inhibiting HIV-1 pseudovirus infection in vitro . This work lays the groundwork for a potential one-time functional cure for HIV-1 through combining the long-term delivery of therapeutic antibodies against HIV-1 and the known efficacy of CCR5 KO HSPC transplantation.

Efficient evolution of human antibodies from general protein language models.

Natural evolution must explore a vast landscape of possible sequences for desirable yet rare mutations, suggesting that learning from natural evolutionary strategies could guide artificial evolution. Here we report that general protein language models can efficiently evolve human antibodies by suggesting mutations that are evolutionarily plausible, despite providing the model with no information about the target antigen, binding specificity or protein structure. We performed language-model-guided affinity maturation of seven antibodies, screening 20 or fewer variants of each antibody across only two rounds of laboratory evolution, and improved the binding affinities of four clinically relevant, highly mature antibodies up to sevenfold and three unmatured antibodies up to 160-fold, with many designs also demonstrating favorable thermostability and viral neutralization activity against Ebola and severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pseudoviruses. The same models that improve antibody binding also guide efficient evolution across diverse protein families and selection pressures, including antibiotic resistance and enzyme activity, suggesting that these results generalize to many settings.

A structural blueprint for interleukin-21 signal modulation.

Interleukin-21 (IL-21) plays a critical role in generating immunological memory by promoting the germinal center reaction, yet clinical use of IL-21 remains challenging because of its pleiotropy and association with autoimmune disease. To better understand the structural basis of IL-21 signaling, we determine the structure of the IL-21-IL-21R-γc ternary signaling complex by X-ray crystallography and a structure of a dimer of trimeric complexes using cryo-electron microscopy. Guided by the structure, we design analogs of IL-21 by introducing substitutions to the IL-21-γc interface. These IL-21 analogs act as partial agonists that modulate downstream activation of pS6, pSTAT3, and pSTAT1. These analogs exhibit differential activity on T and B cell subsets and modulate antibody production in human tonsil organoids. These results clarify the structural basis of IL-21 signaling and offer a potential strategy for tunable manipulation of humoral immunity.

Broad and Durable Humoral Responses Following Single Hydrogel Immunization of SARS-CoV-2 Subunit Vaccine.

Most vaccines require several immunizations to induce robust immunity, and indeed, most SARS-CoV-2 vaccines require an initial two-shot regimen followed by several boosters to maintain efficacy. Such a complex series of immunizations unfortunately increases the cost and complexity of populations-scale vaccination and reduces overall compliance and vaccination rate. In a rapidly evolving pandemic affected by the spread of immune-escaping variants, there is an urgent need to develop vaccines capable of providing robust and durable immunity. In this work, a single immunization SARS-CoV-2 subunit vaccine is developed that can rapidly generate potent, broad, and durable humoral immunity. Injectable polymer-nanoparticle (PNP) hydrogels are leveraged as a depot technology for the sustained delivery of a nanoparticle antigen (RND-NP) displaying multiple copies of the SARS-CoV-2 receptor-binding domain (RBD) and potent adjuvants including CpG and 3M-052. Compared to a clinically relevant prime-boost regimen with soluble vaccines formulated with CpG/alum or 3M-052/alum adjuvants, PNP hydrogel vaccines more rapidly generated higher, broader, and more durable antibody responses. Additionally, these single-immunization hydrogel-based vaccines elicit potent and consistent neutralizing responses. Overall, it is shown that PNP hydrogels elicit improved anti-COVID immune responses with only a single administration, demonstrating their potential as critical technologies to enhance overall pandemic readiness.

HIV-1 prehairpin intermediate inhibitors show efficacy independent of neutralization tier.

HIV-1 strains are categorized into one of three neutralization tiers based on the relative ease by which they are neutralized by plasma from HIV-1-infected donors not on antiretroviral therapy; tier-1 strains are particularly sensitive to neutralization while tier-2 and tier-3 strains are increasingly difficult to neutralize. Most broadly neutralizing antibodies (bnAbs) previously described target the native prefusion conformation of HIV-1 Envelope (Env), but the relevance of the tiered categories for inhibitors targeting another Env conformation, the prehairpin intermediate, is not well understood. Here, we show that two inhibitors targeting distinct highly conserved regions of the prehairpin intermediate have strikingly consistent neutralization potencies (within ~100-fold for a given inhibitor) against strains in all three neutralization tiers of HIV-1; in contrast, best-in-class bnAbs targeting diverse Env epitopes vary by more than 10,000-fold in potency against these strains. Our results indicate that antisera-based HIV-1 neutralization tiers are not relevant for inhibitors targeting the prehairpin intermediate and highlight the potential for therapies and vaccine efforts targeting this conformation.

Structure-guided stabilization improves the ability of the HIV-1 gp41 hydrophobic pocket to elicit neutralizing antibodies.

The hydrophobic pocket found in the N-heptad repeat (NHR) region of HIV-1 gp41 is a highly conserved epitope that is the target of various HIV-1-neutralizing monoclonal antibodies. Although the high conservation of the pocket makes it an attractive vaccine candidate, it has been challenging to elicit potent anti-NHR antibodies via immunization. Here, we solved a high-resolution structure of the NHR mimetic IQN17, and, consistent with previous ligand-bound gp41 pocket structures, we observed remarkable conformational plasticity of the pocket. The high malleability of this pocket led us to test whether we could improve the immunogenicity of the gp41 pocket by stabilizing its conformation. We show that the addition of five amino acids at the C terminus of IQN17, to generate IQN22, introduces a stabilizing salt bridge at the base of the peptide that rigidifies the pocket. Mice immunized with IQN22 elicited higher avidity antibodies against the gp41 pocket and a more potent, albeit still weak, neutralizing response against HIV-1 compared with IQN17. Stabilized epitope-focused immunogens could serve as the basis for future HIV-1 fusion-inhibiting vaccines.

Inverse folding of protein complexes with a structure-informed language model enables unsupervised antibody evolution.

Large language models trained on sequence information alone are capable of learning high level principles of protein design. However, beyond sequence, the three-dimensional structures of proteins determine their specific function, activity, and evolvability. Here we show that a general protein language model augmented with protein structure backbone coordinates and trained on the inverse folding problem can guide evolution for diverse proteins without needing to explicitly model individual functional tasks. We demonstrate inverse folding to be an effective unsupervised, structure-based sequence optimization strategy that also generalizes to multimeric complexes by implicitly learning features of binding and amino acid epistasis. Using this approach, we screened ~30 variants of two therapeutic clinical antibodies used to treat SARS-CoV-2 infection and achieved up to 26-fold improvement in neutralization and 37-fold improvement in affinity against antibody-escaped viral variants-of-concern BQ.1.1 and XBB.1.5, respectively. In addition to substantial overall improvements in protein function, we find inverse folding performs with leading experimental success rates among other reported machine learning-guided directed evolution methods, without requiring any task-specific training data.

A Derivative of the D5 Monoclonal Antibody That Targets the gp41 N-Heptad Repeat of HIV-1 with Broad Tier-2-Neutralizing Activity.

HIV-1 infection is initiated by the viral glycoprotein Env, which, after interaction with cellular coreceptors, adopts a transient conformation known as the prehairpin intermediate (PHI). The N-heptad repeat (NHR) is a highly conserved region of gp41 exposed in the PHI; it is the target of the FDA-approved drug enfuvirtide and of neutralizing monoclonal antibodies (mAbs). However, to date, these mAbs have only been weakly effective against tier-1 HIV-1 strains, which are most sensitive to neutralizing antibodies. Here, we engineered and tested 11 IgG variants of D5, an anti-NHR mAb, by recombining previously described mutations in four of D5's six antibody complementarity-determining regions. One variant, D5_AR, demonstrated 6-fold enhancement in the 50% inhibitory dose (ID50) against lentivirus pseudotyped with HXB2 Env. D5_AR exhibited weak cross-clade neutralizing activity against a diverse set of tier-2 HIV-1 viruses, which are less sensitive to neutralizing antibodies than tier-1 viruses and are the target of current antibody-based vaccine efforts. In addition, the neutralization potency of D5_AR IgG was greatly enhanced in target cells expressing FcγRI, with ID50 values of <0.1 µg/ml; this immunoglobulin receptor is expressed on macrophages and dendritic cells, which are implicated in the early stages of HIV-1 infection of mucosal surfaces. D5 and D5_AR have equivalent neutralization potency in IgG, Fab, and single-chain variable-fragment (scFv) formats, indicating that neutralization is not impacted by steric hindrance. Taken together, these results provide support for vaccine strategies that target the PHI by eliciting antibodies against the gp41 NHR and support investigation of anti-NHR mAbs in nonhuman primate passive immunization studies. IMPORTANCE Despite advances in antiretroviral therapy, HIV remains a global epidemic and has claimed more than 32 million lives. Accordingly, developing an effective HIV vaccine remains an urgent public health need. The gp41 N-heptad repeat (NHR) of the HIV-1 prehairpin intermediate (PHI) is highly conserved (>90%) and is inhibited by the FDA-approved drug enfuvirtide, making it an attractive vaccine target. However, to date, anti-NHR antibodies have not been potent. Here, we engineered D5_AR, a more potent variant of the anti-NHR antibody D5, and established its ability to inhibit HIV-1 strains that are more difficult to neutralize and are more representative of circulating strains (tier-2 strains). The neutralizing activity of D5_AR was greatly potentiated in cells expressing FcγRI; FcγRI is expressed on cells that are implicated at the earliest stages of sexual HIV-1 transmission. Taken together, these results bolster efforts to target the gp41 NHR and the PHI for vaccine development.

Engineering a Rugged Nanoscaffold To Enhance Plug-and-Display Vaccination.

Nanoscale organization is crucial to stimulating an immune response. Using self-assembling proteins as multimerization platforms provides a safe and immunogenic system to vaccinate against otherwise weakly immunogenic antigens. Such multimerization platforms are generally based on icosahedral viruses and have led to vaccines given to millions of people. It is unclear whether synthetic protein nanoassemblies would show similar potency. Here we take the computationally designed porous dodecahedral i301 60-mer and rationally engineer this particle, giving a mutated i301 (mi3) with improved particle uniformity and stability. To simplify the conjugation of this nanoparticle, we employ a SpyCatcher fusion of mi3, such that an antigen of interest linked to the SpyTag peptide can spontaneously couple through isopeptide bond formation (Plug-and-Display). SpyCatcher-mi3 expressed solubly to high yields in Escherichia coli, giving more than 10-fold greater yield than a comparable phage-derived icosahedral nanoparticle, SpyCatcher-AP205. SpyCatcher-mi3 nanoparticles showed high stability to temperature, freeze-thaw, lyophilization, and storage over time. We demonstrate approximately 95% efficiency coupling to different transmission-blocking and blood-stage malaria antigens. Plasmodium falciparum CyRPA was conjugated to SpyCatcher-mi3 nanoparticles and elicited a high avidity antibody response, comparable to phage-derived virus-like particles despite their higher valency and RNA cargo. The simple production, precise derivatization, and exceptional ruggedness of this nanoscaffold should facilitate broad application for nanobiotechnology and vaccine development.

Treatment outcome of creatine transporter deficiency: international retrospective cohort study.

To evaluate the outcome of current treatment for creatine transporter (CRTR) deficiency, we developed a clinical severity score and initiated an international treatment registry. An online questionnaire was completed by physicians following patients with CRTR deficiency on a treatment, including creatine and/or arginine, and/or glycine. Clinical severity score included 1) global developmental delay/intellectual disability; 2) seizures; 3) behavioural disorder. Phenotype scored 1-3 = mild; 4-6 = moderate; and 7-9 = severe. We applied the clinical severity score pre- and on-treatment. Seventeen patients, 14 males and 3 females, from 16 families were included. Four patients had severe, 6 patients had moderate, and 7 patients had a mild phenotype. The phenotype ranged from mild to severe in patients diagnosed at or before 2 years of age or older than 6 years of age. The phenotype ranged from mild to severe in patients with mildly elevated urine creatine to creatinine ratio. Fourteen patients were on the combined creatine, arginine and glycine therapy. On the combined treatment with creatine, arginine and glycine, none of the males showed either deterioration or improvements in their clinical severity score, whereas two females showed improvements in the clinical severity score. Creatine monotherapy resulted in deterioration of the clinical severity score in one male. There seems to be no correlation between phenotype and degree of elevation in urine creatine to creatinine ratio, genotype, or age at diagnosis. Combined creatine, arginine and glycine therapy might have stopped disease progression in males and improved phenotype in females.

Prospective cohort study for identification of underlying genetic causes in neonatal encephalopathy using whole-exome sequencing.

PurposeNeonatal encephalopathy, which is characterized by a decreased level of consciousness, occurs in 1-7/1,000 live-term births. In more than half of term newborns, there is no identifiable etiological factor. To identify underlying genetic defects, we applied whole-exome sequencing (WES) in term newborns with neonatal encephalopathy as a prospective cohort study.MethodsTerm newborns with neonatal encephalopathy and no history of perinatal asphyxia were included. WES was performed using patient and both parents' DNA.ResultsNineteen patients fulfilling inclusion criteria were enrolled. Five patients were excluded owing to withdrawal of consent, no parental DNA samples, or a genetic diagnosis prior to WES. Fourteen patients underwent WES. We confirmed a genetic diagnosis in five patients (36%): epileptic encephalopathy associated with autosomal dominant de novo variants in SCN2A (p.Met1545Val), KCNQ2 (p.Asp212Tyr), and GNAO1 (p.Gly40Arg); lipoic acid synthetase deficiency due to compound heterozygous variants in LIAS (p.Ala253Pro and p.His236Gln); and encephalopathy associated with an X-linked variant in CUL4B (p.Asn211Ser).ConclusionWES is helpful at arriving genetic diagnoses in neonatal encephalopathy and/or seizures and brain damage. It will increase our understanding and probably enable us to develop targeted neuroprotective treatment strategies.

Phenotypic and genotypic spectrum of congenital disorders of glycosylation type I and type II.

BACKGROUND: Congenital disorders of glycosylation (CDG) are inborn defects of glycan metabolism. They are multisystem disorders. Analysis of transferrin isoforms is applied as a screening test for CDG type I (CDG-I) and type II (CDG-II). We performed a retrospective cohort study to determine spectrum of phenotype and genotype and prevalence of the different subtypes of CDG-I and CDG-II. MATERIAL AND METHODS: All patients with CDG-I and CDG-II evaluated in our institution's Metabolic Genetics Clinics were included. Electronic and paper patient charts were reviewed. We set-up a high performance liquid chromatography transferrin isoelectric focusing (TIEF) method to measure transferrin isoforms in our Institution. We reviewed the literature for the rare CDG-I and CDG-II subtypes seen in our Institution. RESULTS: Fifteen patients were included: 9 with PMM2-CDG and 6 with non-PMM2-CDG (one ALG3-CDG, one ALG9-CDG, two ALG11-CDG, one MPDU1-CDG and one ATP6V0A2-CDG). All patients with PMM2-CDG and 5 patients with non-PMM2-CDG showed abnormal TIEF suggestive of CDG-I or CDG-II pattern. In all patients, molecular diagnosis was confirmed either by single gene testing, targeted next generation sequencing for CDG genes, or by whole exome sequencing. CONCLUSION: We report 15 new patients with CDG-I and CDG-II. Whole exome sequencing will likely identify more patients with normal TIEF and expand the phenotypic spectrum of CDG-I and CDG-II.

Phenotype, biochemical features, genotype and treatment outcome of pyridoxine-dependent epilepsy.

We report treatment outcome of eleven patients with pyridoxine-dependent epilepsy caused by pathogenic variants in ALDH7A1 (PDE-ALDH7A1). We developed a clinical severity score to compare phenotype with biochemical features, genotype and delays in the initiation of pyridoxine. Clinical severity score included 1) global developmental delay/ intellectual disability; 2) age of seizure onset prior to pyridoxine; 3) current seizures on treatment. Phenotype scored 1-3 = mild; 4-6 = moderate; and 7-9 = severe. Five patients had mild, four patients had moderate, and two patients had severe phenotype. Phenotype ranged from mild to severe in eight patients (no lysine-restricted diet in the infantile period) with more than 10-fold elevated urine or plasma α-AASA levels. Phenotype ranged from mild to moderate in patients with homozygous truncating variants and from moderate to severe in patients with homozygous missense variants. There was no correlation between severity of the phenotype and the degree of α-AASA elevation in urine or genotype. All patients were on pyridoxine, nine patients were on arginine and five patients were on the lysine-restricted diet. 73% of the patients became seizure free on pyridoxine. 25% of the patients had a mild phenotype on pyridoxine monotherapy. Whereas, 100% of the patients, on the lysine-restricted diet initiated within their first 7 months of life, had a mild phenotype. Early initiation of lysine-restricted diet and/or arginine therapy likely improved neurodevelopmental outcome in young patients with PDE-ALDH7A1.