Machine learning designs new GCGR/GLP-1R dual agonists with enhanced biological potency.

Several peptide dual agonists of the human glucagon receptor (GCGR) and the glucagon-like peptide-1 receptor (GLP-1R) are in development for the treatment of type 2 diabetes, obesity and their associated complications. Candidates must have high potency at both receptors, but it is unclear whether the limited experimental data available can be used to train models that accurately predict the activity at both receptors of new peptide variants. Here we use peptide sequence data labelled with in vitro potency at human GCGR and GLP-1R to train several models, including a deep multi-task neural-network model using multiple loss optimization. Model-guided sequence optimization was used to design three groups of peptide variants, with distinct ranges of predicted dual activity. We found that three of the model-designed sequences are potent dual agonists with superior biological activity. With our designs we were able to achieve up to sevenfold potency improvement at both receptors simultaneously compared to the best dual-agonist in the training set.

Rapid discovery of high-affinity antibodies via massively parallel sequencing, ribosome display and affinity screening.

Developing therapeutic antibodies is laborious and costly. Here we report a method for antibody discovery that leverages the Illumina HiSeq platform to, within 3 days, screen in the order of 108 antibody-antigen interactions. The method, which we named 'deep screening', involves the clustering and sequencing of antibody libraries, the conversion of the DNA clusters into complementary RNA clusters covalently linked to the instrument's flow-cell surface on the same location, the in situ translation of the clusters into antibodies tethered via ribosome display, and their screening via fluorescently labelled antigens. By using deep screening, we discovered low-nanomolar nanobodies to a model antigen using 4 × 106 unique variants from yeast-display-enriched libraries, and high-picomolar single-chain antibody fragment leads for human interleukin-7 directly from unselected synthetic repertoires. We also leveraged deep screening of a library of 2.4 × 105 sequences of the third complementarity-determining region of the heavy chain of an anti-human epidermal growth factor receptor 2 (HER2) antibody as input for a large language model that generated new single-chain antibody fragment sequences with higher affinity for HER2 than those in the original library.

Tozorakimab (MEDI3506): an anti-IL-33 antibody that inhibits IL-33 signalling via ST2 and RAGE/EGFR to reduce inflammation and epithelial dysfunction.

Interleukin (IL)-33 is a broad-acting alarmin cytokine that can drive inflammatory responses following tissue damage or infection and is a promising target for treatment of inflammatory disease. Here, we describe the identification of tozorakimab (MEDI3506), a potent, human anti-IL-33 monoclonal antibody, which can inhibit reduced IL-33 (IL-33red) and oxidized IL-33 (IL-33ox) activities through distinct serum-stimulated 2 (ST2) and receptor for advanced glycation end products/epidermal growth factor receptor (RAGE/EGFR complex) signalling pathways. We hypothesized that a therapeutic antibody would require an affinity higher than that of ST2 for IL-33, with an association rate greater than 107 M-1 s-1, to effectively neutralize IL-33 following rapid release from damaged tissue. An innovative antibody generation campaign identified tozorakimab, an antibody with a femtomolar affinity for IL-33red and a fast association rate (8.5 × 107 M-1 s-1), which was comparable to soluble ST2. Tozorakimab potently inhibited ST2-dependent inflammatory responses driven by IL-33 in primary human cells and in a murine model of lung epithelial injury. Additionally, tozorakimab prevented the oxidation of IL-33 and its activity via the RAGE/EGFR signalling pathway, thus increasing in vitro epithelial cell migration and repair. Tozorakimab is a novel therapeutic agent with a dual mechanism of action that blocks IL-33red and IL-33ox signalling, offering potential to reduce inflammation and epithelial dysfunction in human disease.

An efficient system for bioconjugation based on a widely applicable engineered O-glycosylation tag.

Bioconjugates are an important class of therapeutic molecules. To date, O-glycan-based metabolic glycoengineering has had limited use in this field, due to the complexities of the endogenous O-glycosylation pathway and the lack of an O-glycosylation consensus sequence. Here, we describe the development of a versatile on-demand O-glycosylation system that uses a novel, widely applicable 5 amino acid O-glycosylation tag, and a metabolically engineered UDP-galactose-4-eperimase (GALE) knock-out cell line. Optimization of the primary sequence of the tag enables the production of Fc-based proteins with either single or multiple O-glycans with complexity fully controlled by media supplementation. We demonstrate how the uniformly labeled proteins containing exclusively N-azido-acetylgalactosamine are used for CLICK chemistry-based bioconjugation to generate site-specifically fluorochrome-labeled antibodies, dual-payload molecules, and bioactive Fc-peptides for applications in basic research and drug discovery. To our knowledge, this is the first description of generating a site-specific O-glycosylation system by combining an O-glycosylation tag and a metabolically engineered cell line.

Structural and functional characterization of C0021158, a high-affinity monoclonal antibody that inhibits Arginase 2 function via a novel non-competitive mechanism of action.

Arginase 2 (ARG2) is a binuclear manganese metalloenzyme that catalyzes the hydrolysis of L-arginine. The dysregulated expression of ARG2 within specific tumor microenvironments generates an immunosuppressive niche that effectively renders the tumor 'invisible' to the host's immune system. Increased ARG2 expression leads to a concomitant depletion of local L-arginine levels, which in turn leads to suppression of anti-tumor T-cell-mediated immune responses. Here we describe the isolation and characterization of a high affinity antibody (C0021158) that inhibits ARG2 enzymatic function completely, effectively restoring T-cell proliferation in vitro. Enzyme kinetic studies confirmed that C0021158 exhibits a noncompetitive mechanism of action, inhibiting ARG2 independently of L-arginine concentrations. To elucidate C0021158's inhibitory mechanism at a structural level, the co-crystal structure of the Fab in complex with trimeric ARG2 was solved. C0021158's epitope was consequently mapped to an area some distance from the enzyme's substrate binding cleft, indicating an allosteric mechanism was being employed. Following C0021158 binding, distinct regions of ARG2 undergo major conformational changes. Notably, the backbone structure of a surface-exposed loop is completely rearranged, leading to the formation of a new short helix structure at the Fab-ARG2 interface. Moreover, this large-scale structural remodeling at ARG2's epitope translates into more subtle changes within the enzyme's active site. An arginine residue at position 39 is reoriented inwards, sterically impeding the binding of L-arginine. Arg39 is also predicted to alter the pKA of a key catalytic histidine residue at position 160, further attenuating ARG2's enzymatic function. In silico molecular docking simulations predict that L-arginine is unable to bind effectively when antibody is bound, a prediction supported by isothermal calorimetry experiments using an L-arginine mimetic. Specifically, targeting ARG2 in the tumor microenvironment through the application of C0021158, potentially in combination with standard chemotherapy regimens or alternate immunotherapies, represents a potential new strategy to target immune cold tumors.

Extensive sequence and structural evolution of Arginase 2 inhibitory antibodies enabled by an unbiased approach to affinity maturation.

Affinity maturation is a powerful technique in antibody engineering for the in vitro evolution of antigen binding interactions. Key to the success of this process is the expansion of sequence and combinatorial diversity to increase the structural repertoire from which superior binding variants may be selected. However, conventional strategies are often restrictive and only focus on small regions of the antibody at a time. In this study, we used a method that combined antibody chain shuffling and a staggered-extension process to produce unbiased libraries, which recombined beneficial mutations from all six complementarity-determining regions (CDRs) in the affinity maturation of an inhibitory antibody to Arginase 2 (ARG2). We made use of the vast display capacity of ribosome display to accommodate the sequence space required for the diverse library builds. Further diversity was introduced through pool maturation to optimize seven leads of interest simultaneously. This resulted in antibodies with substantial improvements in binding properties and inhibition potency. The extensive sequence changes resulting from this approach were translated into striking structural changes for parent and affinity-matured antibodies bound to ARG2, with a large reorientation of the binding paratope facilitating increases in contact surface and shape complementarity to the antigen. The considerable gains in therapeutic properties seen from extensive sequence and structural evolution of the parent ARG2 inhibitory antibody clearly illustrate the advantages of the unbiased approach developed, which was key to the identification of high-affinity antibodies with the desired inhibitory potency and specificity.

A direct role for SNX9 in the biogenesis of filopodia.

Filopodia are finger-like actin-rich protrusions that extend from the cell surface and are important for cell-cell communication and pathogen internalization. The small size and transient nature of filopodia combined with shared usage of actin regulators within cells confounds attempts to identify filopodial proteins. Here, we used phage display phenotypic screening to isolate antibodies that alter the actin morphology of filopodia-like structures (FLS) in vitro. We found that all of the antibodies that cause shorter FLS interact with SNX9, an actin regulator that binds phosphoinositides during endocytosis and at invadopodia. In cells, we discover SNX9 at specialized filopodia in Xenopus development and that SNX9 is an endogenous component of filopodia that are hijacked by Chlamydia entry. We show the use of antibody technology to identify proteins used in filopodia-like structures, and a role for SNX9 in filopodia.

A receptor for the complement regulator factor H increases transmission of trypanosomes to tsetse flies.

Persistent pathogens have evolved to avoid elimination by the mammalian immune system including mechanisms to evade complement. Infections with African trypanosomes can persist for years and cause human and animal disease throughout sub-Saharan Africa. It is not known how trypanosomes limit the action of the alternative complement pathway. Here we identify an African trypanosome receptor for mammalian factor H, a negative regulator of the alternative pathway. Structural studies show how the receptor binds ligand, leaving inhibitory domains of factor H free to inactivate complement C3b deposited on the trypanosome surface. Receptor expression is highest in developmental stages transmitted to the tsetse fly vector and those exposed to blood meals in the tsetse gut. Receptor gene deletion reduced tsetse infection, identifying this receptor as a virulence factor for transmission. This demonstrates how a pathogen evolved a molecular mechanism to increase transmission to an insect vector by exploitation of a mammalian complement regulator.

Structure of the trypanosome transferrin receptor reveals mechanisms of ligand recognition and immune evasion.

To maintain prolonged infection of mammals, African trypanosomes have evolved remarkable surface coats and a system of antigenic variation1. Within these coats are receptors for macromolecular nutrients such as transferrin2,3. These must be accessible to their ligands but must not confer susceptibility to immunoglobulin-mediated attack. Trypanosomes have a wide host range and their receptors must also bind ligands from diverse species. To understand how these requirements are achieved, in the context of transferrin uptake, we determined the structure of a Trypanosoma brucei transferrin receptor in complex with human transferrin, showing how this heterodimeric receptor presents a large asymmetric ligand-binding platform. The trypanosome genome contains a family of around 14 transferrin receptors4, which has been proposed to allow binding to transferrin from different mammalian hosts5,6. However, we find that a single receptor can bind transferrin from a broad range of mammals, indicating that receptor variation is unlikely to be necessary for promiscuity of host infection. In contrast, polymorphic sites and N-linked glycans are preferentially found in exposed positions on the receptor surface, not contacting transferrin, suggesting that transferrin receptor diversification is driven by a need for antigenic variation in the receptor to prolong survival in a host.

Antibodies binding the head domain of P2X4 inhibit channel function and reverse neuropathic pain.

P2X4 is a ligand-gated ion channel implicated in neuropathic pain. Drug discovery efforts targeting P2X4 have been unsuccessful largely because of the difficulty in engineering specificity and selectivity. Here, we describe for the first time the generation of a panel of diverse monoclonal antibodies (mAbs) to human and mouse P2X4, capable of both positive and negative modulation of channel function. The affinity-optimised anti-P2X4 mAb IgG#151-LO showed exquisite selectivity for human P2X4 and induced potent and complete block of P2X4 currents. Site-directed mutagenesis of P2X4 revealed the head domain as a key interaction site for inhibitory mAbs. Inhibition of spinal P2X4 either by intrathecal delivery of an anti-P2X4 mAb or by systemic delivery of an anti-P2X4 bispecific mAb with enhanced blood-spinal cord barrier permeability produced long-lasting (>7 days) analgesia in a mouse model of neuropathic pain. We therefore propose that inhibitory mAbs binding the head domain of P2X4 have therapeutic potential for the treatment of neuropathic pain.

Experimentally guided computational antibody affinity maturation with de novo docking, modelling and rational design.

Antibodies are an important class of therapeutics that have significant clinical impact for the treatment of severe diseases. Computational tools to support antibody drug discovery have been developing at an increasing rate over the last decade and typically rely upon a predetermined co-crystal structure of the antibody bound to the antigen for structural predictions. Here, we show an example of successful in silico affinity maturation of a hybridoma derived antibody, AB1, using just a homology model of the antibody fragment variable region and a protein-protein docking model of the AB1 antibody bound to the antigen, murine CCL20 (muCCL20). In silico affinity maturation, together with alanine scanning, has allowed us to fine-tune the protein-protein docking model to subsequently enable the identification of two single-point mutations that increase the affinity of AB1 for muCCL20. To our knowledge, this is one of the first examples of the use of homology modelling and protein docking for affinity maturation and represents an approach that can be widely deployed.

A single dose of antibody-drug conjugate cures a stage 1 model of African trypanosomiasis.

Infections of humans and livestock with African trypanosomes are treated with drugs introduced decades ago that are not always fully effective and often have severe side effects. Here, the trypanosome haptoglobin-haemoglobin receptor (HpHbR) has been exploited as a route of uptake for an antibody-drug conjugate (ADC) that is completely effective against Trypanosoma brucei in the standard mouse model of infection. Recombinant human anti-HpHbR monoclonal antibodies were isolated and shown to be internalised in a receptor-dependent manner. Antibodies were conjugated to a pyrrolobenzodiazepine (PBD) toxin and killed T. brucei in vitro at picomolar concentrations. A single therapeutic dose (0.25 mg/kg) of a HpHbR antibody-PBD conjugate completely cured a T. brucei mouse infection within 2 days with no re-emergence of infection over a subsequent time course of 77 days. These experiments provide a demonstration of how ADCs can be exploited to treat protozoal diseases that desperately require new therapeutics.

Interactions Between the Gravitostat and the Fibroblast Growth Factor System for the Regulation of Body Weight.

Both fibroblast growth factors (FGFs), by binding to FGF receptors (FGFRs), and activation of the gravitostat, by artificial loading, decrease the body weight (BW). Previous studies demonstrate that both the FGF system and loading have the capacity to regulate BW independently of leptin. The aim of the current study was to determine the possible interactions between the effect of increased loading and the FGF system for the regulation of BW. We observed that the BW-reducing effect of increased loading was abolished in mice treated with a monoclonal antibody directed against FGFR1c, suggesting interactions between the two systems. As serum levels of endocrine FGF21 and hepatic FGF21 mRNA were increased in the loaded mice compared with the control mice, we first evaluated the loading response in FGF21 over expressing mice with constant high FGF21 levels. Leptin treatment, but not increased loading, decreased the BW in the FGF21-overexpressing mice, demonstrating that specifically the loading effect is attenuated in the presence of high activity in the FGF system. However, as FGF21 knockout mice displayed a normal loading response on BW, FGF21 is neither mediating nor essential for the loading response. In conclusion, the BW-reducing effect of increased loading but not of leptin treatment is blocked by high activity in the FGF system. We propose that both the gravitostat and the FGF system regulate BW independently of leptin and that pharmacologically enhanced activity in the FGF system reduces the sensitivity of the gravitostat.

Exosomal delivery of doxorubicin enables rapid cell entry and enhanced in vitro potency.

Doxorubicin is a chemotherapeutic agent that is commonly used to treat a broad range of cancers. However, significant cardiotoxicity, associated with prolonged exposure to doxorubicin, limits its continued therapeutic use. One strategy to prevent the uptake of doxorubicin into cardiac cells is the encapsulation of the drug to prevent non-specific uptake and also to improve the drugs' pharmacokinetic properties. Although encapsulated forms of doxorubicin limit the cardiotoxicity observed, they are not without their own liabilities as an increased amount of drug is deposited in the skin where liposomal doxorubicin can cause palmar-plantar erythrodysesthesia. Exosomes are small endogenous extracellular vesicles, that transfer bioactive material from one cell to another, and are considered attractive drug delivery vehicles due to their natural origin. In this study, we generated doxorubicin-loaded exosomes and demonstrate their rapid cellular uptake and re-distribution of doxorubicin from endosomes to the cytoplasm and nucleus resulting in enhanced potency in a number of cultured and primary cell lines when compared to free doxorubicin and liposomal formulations of doxorubicin. In contrast to other delivery methods for doxorubicin, exosomes do not accumulate in the heart, thereby providing potential for limiting the cardiac side effects and improved therapeutic index.

Anti-PrPC antibody rescues cognition and synapses in transgenic alzheimer mice.

Objective: Amyloid-beta oligomers (Aßo) trigger the development of Alzheimer's disease (AD) pathophysiology. Cellular prion protein (PrPC) initiates synaptic damage as a high affinity receptor for Aßo. Here, we evaluated the preclinical therapeutic efficacy of a fully human monoclonal antibody against PrPC. This AZ59 antibody selectively targets the Aßo binding site in the amino-terminal unstructured domain of PrPC to avoid any potential risk of direct toxicity. Methods: Potency of AZ59 was evaluated by binding to PrPC, blockade of Aßo interaction and interruption of Aßo signaling. AZ59 was administered to mice by weekly intraperitoneal dosing and brain antibody measured. APP/PS1 transgenic mice were treated with AZ59 and assessed by memory tests, by brain biochemistry and by histochemistry for Aß, gliosis and synaptic density. Results: AZ59 binds PrPC with 100 pmol/L affinity and blocks human brain Aßo binding to PrPC, as well as prevents synaptotoxic signaling. Weekly i.p. dosing of 20 mg/kg AZ59 in a murine form achieves trough brain antibody levels greater than 10 nmol/L. Aged symptomatic APP/PS1 transgenic mice treated with AZ59 for 5-7 weeks show a full rescue of behavioral and synaptic loss phenotypes. This recovery occurs without clearance of plaque pathology or elimination of gliosis. AZ59 treatment also normalizes synaptic signaling abnormalities in transgenic brain. These benefits are dose-dependent and persist for at least 1 month after the last dose. Interpretation: Preclinical data demonstrate that systemic AZ59 therapy rescues central synapses and memory function from transgenic Alzheimer's disease pathology, supporting a disease-modifying therapeutic potential.

Discovery and characterisation of an antibody that selectively modulates the inhibitory activity of plasminogen activator inhibitor-1.

Plasminogen activator inhibitor-1 (PAI-1) is a serine protease inhibitor (serpin) that regulates fibrinolysis, cell adhesion and cell motility via its interactions with plasminogen activators and vitronectin. PAI-1 has been shown to play a role in a number of diverse pathologies including cardiovascular diseases, obesity and cancer and is therefore an attractive therapeutic target. However the multiple patho-physiological roles of PAI-1, and understanding the relative contributions of these in any one disease setting, make the development of therapeutically relevant molecules challenging. Here we describe the identification and characterisation of fully human antibody MEDI-579, which binds with high affinity and specificity to the active form of human PAI-1. MEDI-579 specifically inhibits serine protease interactions with PAI-1 while conserving vitronectin binding. Crystallographic analysis reveals that this specificity is achieved through direct binding of MEDI-579 Fab to the reactive centre loop (RCL) of PAI-1 and at the same exosite used by both tissue and urokinase plasminogen activators (tPA and uPA). We propose that MEDI-579 acts by directly competing with proteases for RCL binding and as such is able to modulate the interaction of PAI-1 with tPA and uPA in a way not previously described for a human PAI-1 inhibitor.

Engineering of a GLP-1 analogue peptide/anti-PCSK9 antibody fusion for type 2 diabetes treatment.

Type 2 diabetes (T2D) is a complex and progressive disease requiring polypharmacy to manage hyperglycaemia and cardiovascular risk factors. However, most patients do not achieve combined treatment goals. To address this therapeutic gap, we have developed MEDI4166, a novel glucagon-like peptide-1 (GLP-1) receptor agonist peptide fused to a proprotein convertase subtilisin/kexin type 9 (PCSK9) neutralising antibody that allows for glycaemic control and low-density lipoprotein cholesterol (LDL-C) lowering in a single molecule. The fusion has been engineered to deliver sustained peptide activity in vivo in combination with reduced potency, to manage GLP-1 driven adverse effects at high dose, and a favourable manufacturability profile. MEDI4166 showed robust and sustained LDL-C lowering in cynomolgus monkeys and exhibited the anticipated GLP-1 effects in T2D mouse models. We believe MEDI4166 is a novel molecule combining long acting agonist peptide and neutralising antibody activities to deliver a unique pharmacology profile for the management of T2D.

Computational Design of Epitope-Specific Functional Antibodies.

The ultimate goal of protein design is to introduce new biological activity. We propose a computational approach for designing functional antibodies by focusing on functional epitopes, integrating large-scale statistical analysis with multiple structural models. Machine learning is used to analyze these models and predict specific residue-residue contacts. We use this approach to design a functional antibody to counter the proinflammatory effect of the cytokine interleukin-17A (IL-17A). X-ray crystallography confirms that the designed antibody binds the targeted epitope and the interaction is mediated by the designed contacts. Cell-based assays confirm that the antibody is functional. Importantly, this approach does not rely on a high-quality 3D model of the designed complex or even a solved structure of the target. As demonstrated here, this approach can be used to design biologically active antibodies, removing some of the main hurdles in antibody design and in drug discovery.

Haplosaurus computes protein haplotypes for use in precision drug design.

Selecting the most appropriate protein sequences is critical for precision drug design. Here we describe Haplosaurus, a bioinformatic tool for computation of protein haplotypes. Haplosaurus computes protein haplotypes from pre-existing chromosomally-phased genomic variation data. Integration into the Ensembl resource provides rapid and detailed protein haplotypes retrieval. Using Haplosaurus, we build a database of unique protein haplotypes from the 1000 Genomes dataset reflecting real-world protein sequence variability and their prevalence. For one in seven genes, their most common protein haplotype differs from the reference sequence and a similar number differs on their most common haplotype between human populations. Three case studies show how knowledge of the range of commonly encountered protein forms predicted in populations leads to insights into therapeutic efficacy. Haplosaurus and its associated database is expected to find broad applications in many disciplines using protein sequences and particularly impactful for therapeutics design.

High affinity single-chain variable fragments are specific and versatile targeting motifs for extracellular vesicles.

Exosomes are extracellular vesicles that mediate cell-to-cell communication by transferring biological cargo, such as DNA, RNA and proteins. Through genetic engineering of exosome-producing cells or manipulation of purified exosomes, it is possible to load exosomes with therapeutic molecules and target them to specific cells via the display of targeting moieties on their surface. This provides an opportunity to exploit a naturally-occurring biological process for therapeutic purposes. In this study, we explored the potential of single chain variable fragments (scFv) as targeting domains to achieve delivery of exosomes to cells expressing a cognate antigen. We generated exosomes targeting the Her2 receptor and, by varying the affinity of the scFvs and the Her2 expression level on recipient cells, we determined that both a high-affinity anti-Her2-scFv (KD≤ 1 nM) and cells expressing a high level (≥106 copies per cell) of Her2 were optimally required to enable selective uptake. We also demonstrate that targeting exosomes to cells via a specific cell surface receptor can alter their intracellular trafficking route, providing opportunities to influence the efficiency of delivery and fate of intracellular cargo. These experiments provide solid data to support the wider application of exosomes displaying antibody fragments as vehicles for the targeted delivery of therapeutic molecules.