A candidate antibody drug for prevention of malaria.

Over 75% of malaria-attributable deaths occur in children under the age of 5 years. However, the first malaria vaccine recommended by the World Health Organization (WHO) for pediatric use, RTS,S/AS01 (Mosquirix), has modest efficacy. Complementary strategies, including monoclonal antibodies, will be important in efforts to eradicate malaria. Here we characterize the circulating B cell repertoires of 45 RTS,S/AS01 vaccinees and discover monoclonal antibodies for development as potential therapeutics. We generated >28,000 antibody sequences and tested 481 antibodies for binding activity and 125 antibodies for antimalaria activity in vivo. Through these analyses we identified correlations suggesting that sequences in Plasmodium falciparum circumsporozoite protein, the target antigen in RTS,S/AS01, may induce immunodominant antibody responses that limit more protective, but subdominant, responses. Using binding studies, mouse malaria models, biomanufacturing assessments and protein stability assays, we selected AB-000224 and AB-007088 for advancement as a clinical lead and backup. We engineered the variable domains (Fv) of both antibodies to enable low-cost manufacturing at scale for distribution to pediatric populations, in alignment with WHO's preferred product guidelines. The engineered clone with the optimal manufacturing and drug property profile, MAM01, was advanced into clinical development.

Mobilization of innate and adaptive antitumor immune responses by the RNP-targeting antibody ATRC-101.

Immunotherapy approaches focusing on T cells have provided breakthroughs in treating solid tumors. However, there remains an opportunity to drive anticancer immune responses via other cell types, particularly myeloid cells. ATRC-101 was identified via a target-agnostic process evaluating antibodies produced by the plasmablast population of B cells in a patient with non-small cell lung cancer experiencing an antitumor immune response during treatment with checkpoint inhibitor therapy. Here, we describe the target, antitumor activity in preclinical models, and data supporting a mechanism of action of ATRC-101. Immunohistochemistry studies demonstrated tumor-selective binding of ATRC-101 to multiple nonautologous tumor tissues. In biochemical analyses, ATRC-101 appears to target an extracellular, tumor-specific ribonucleoprotein (RNP) complex. In syngeneic murine models, ATRC-101 demonstrated robust antitumor activity and evidence of immune memory following rechallenge of cured mice with fresh tumor cells. ATRC-101 increased the relative abundance of conventional dendritic cell (cDC) type 1 cells in the blood within 24 h of dosing, increased CD8+ T cells and natural killer cells in blood and tumor over time, decreased cDC type 2 cells in the blood, and decreased monocytic myeloid-derived suppressor cells in the tumor. Cellular stress, including that induced by chemotherapy, increased the amount of ATRC-101 target in tumor cells, and ATRC-101 combined with doxorubicin enhanced efficacy compared with either agent alone. Taken together, these data demonstrate that ATRC-101 drives tumor destruction in preclinical models by targeting a tumor-specific RNP complex leading to activation of innate and adaptive immune responses.

Long-Acting IL-33 Mobilizes High-Quality Hematopoietic Stem and Progenitor Cells More Efficiently Than Granulocyte Colony-Stimulating Factor or AMD3100.

Mobilization of hematopoietic stem and progenitor cells (HSPCs) has become increasingly important for hematopoietic cell transplantation. Current mobilization approaches are insufficient because they fail to mobilize sufficient numbers of cells in a significant fraction of patients and are biased toward myeloid immune reconstitution. A novel, single drug mobilization agent that allows a more balanced (myeloid and lymphoid) reconstitution would therefore be highly favorable to improve transplantation outcome. In this present study, we tested commercially available IL-33 molecules and engineered novel variants of IL-33. These molecules were tested in cell-based assays in vitro and in mobilization models in vivo. We observed for the first time that IL-33 treatment in mice mobilized HSPCs and common myeloid progenitors more efficiently than clinical mobilizing agents granulocyte colony-stimulating factor (G-CSF) or AMD3100. We engineered several oxidation-resistant IL-33 variants with equal or better in vitro activity. In vivo, these variants mobilized HSPCs and, interestingly, also hematopoietic stem cells, common lymphoid progenitor cells, and endothelial progenitor cells more efficiently than wild-type IL-33 or G-CSF. We then engineered an IL-33-Fc fusion molecule, a single dose of which was sufficient to significantly increase the mobilization of HSPCs after 4 days. In conclusion, our findings suggest that long-acting, oxidation-resistant IL-33 may be a novel approach for HSPC transplantation. IL-33-mobilized HSPCs differ from cells mobilized with G-CSF and AMD3100, and it is possible that these differences may result in better transplantation outcomes.

Engineering enzyme specificity using computational design of a defined-sequence library.

Engineered biosynthetic pathways have the potential to produce high-value molecules from inexpensive feedstocks, but a key limitation is engineering enzymes with high activity and specificity for new reactions. Here, we developed a method for combining structure-based computational protein design with library-based enzyme screening, in which inter-residue correlations favored by the design are encoded into a defined-sequence library. We validated this approach by engineering a glucose 6-oxidase enzyme for use in a proposed pathway to convert D-glucose into D-glucaric acid. The most active variant, identified after only one round of diversification and screening of only 10,000 wells, is approximately 400-fold more active on glucose than is the wild-type enzyme. We anticipate that this strategy will be broadly applicable to the discovery of new enzymes for engineered biological pathways.

Pairwise selection assembly for sequence-independent construction of long-length DNA.

The engineering of biological components has been facilitated by de novo synthesis of gene-length DNA. Biological engineering at the level of pathways and genomes, however, requires a scalable and cost-effective assembly of DNA molecules that are longer than approximately 10 kb, and this remains a challenge. Here we present the development of pairwise selection assembly (PSA), a process that involves hierarchical construction of long-length DNA through the use of a standard set of components and operations. In PSA, activation tags at the termini of assembly sub-fragments are reused throughout the assembly process to activate vector-encoded selectable markers. Marker activation enables stringent selection for a correctly assembled product in vivo, often obviating the need for clonal isolation. Importantly, construction via PSA is sequence-independent, and does not require primary sequence modification (e.g. the addition or removal of restriction sites). The utility of PSA is demonstrated in the construction of a completely synthetic 91-kb chromosome arm from Saccharomyces cerevisiae.

Creation of a type IIS restriction endonuclease with a long recognition sequence.

Type IIS restriction endonucleases cleave DNA outside their recognition sequences, and are therefore particularly useful in the assembly of DNA from smaller fragments. A limitation of type IIS restriction endonucleases in assembly of long DNA sequences is the relative abundance of their target sites. To facilitate ligation-based assembly of extremely long pieces of DNA, we have engineered a new type IIS restriction endonuclease that combines the specificity of the homing endonuclease I-SceI with the type IIS cleavage pattern of FokI. We linked a non-cleaving mutant of I-SceI, which conveys to the chimeric enzyme its specificity for an 18-bp DNA sequence, to the catalytic domain of FokI, which cuts DNA at a defined site outside the target site. Whereas previously described chimeric endonucleases do not produce type IIS-like precise DNA overhangs suitable for ligation, our chimeric endonuclease cleaves double-stranded DNA exactly 2 and 6 nt from the target site to generate homogeneous, 5', four-base overhangs, which can be ligated with 90% fidelity. We anticipate that these enzymes will be particularly useful in manipulation of DNA fragments larger than a thousand bases, which are very likely to contain target sites for all natural type IIS restriction endonucleases.

Rotamer optimization for protein design through MAP estimation and problem-size reduction.

The search for the global minimum energy conformation (GMEC) of protein side chains is an important computational challenge in protein structure prediction and design. Using rotamer models, the problem is formulated as a NP-hard optimization problem. Dead-end elimination (DEE) methods combined with systematic A* search (DEE/A*) has proven useful, but may not be strong enough as we attempt to solve protein design problems where a large number of similar rotamers is eligible and the network of interactions between residues is dense. In this work, we present an exact solution method, named BroMAP (branch-and-bound rotamer optimization using MAP estimation), for such protein design problems. The design goal of BroMAP is to be able to expand smaller search trees than conventional branch-and-bound methods while performing only a moderate amount of computation in each node, thereby reducing the total running time. To achieve that, BroMAP attempts reduction of the problem size within each node through DEE and elimination by lower bounds from approximate maximum-a-posteriori (MAP) estimation. The lower bounds are also exploited in branching and subproblem selection for fast discovery of strong upper bounds. Our computational results show that BroMAP tends to be faster than DEE/A* for large protein design cases. BroMAP also solved cases that were not solved by DEE/A* within the maximum allowed time, and did not incur significant disadvantage for cases where DEE/A* performed well. Therefore, BroMAP is particularly applicable to large protein design problems where DEE/A* struggles and can also substitute for DEE/A* in general GMEC search.

Computational design of antibody-affinity improvement beyond in vivo maturation.

Antibodies are used extensively in diagnostics and as therapeutic agents. Achieving high-affinity binding is important for expanding detection limits, extending dissociation half-times, decreasing drug dosages and increasing drug efficacy. However, antibody-affinity maturation in vivo often fails to produce antibody drugs of the targeted potency, making further affinity maturation in vitro by directed evolution or computational design necessary. Here we present an iterative computational design procedure that focuses on electrostatic binding contributions and single mutants. By combining multiple designed mutations, a tenfold affinity improvement to 52 pM was engineered into the anti-epidermal growth factor receptor drug cetuximab (Erbitux), and a 140-fold improvement in affinity to 30 pM was obtained for the anti-lysozyme model antibody D44.1. The generality of the methods was further demonstrated through identification of known affinity-enhancing mutations in the therapeutic antibody bevacizumab (Avastin) and the model anti-fluorescein antibody 4-4-20. These results demonstrate computational capabilities for enhancing and accelerating the development of protein reagents and therapeutics.

Numerical integration techniques for curved-element discretizations of molecule-solvent interfaces.

Surface formulations of biophysical modeling problems offer attractive theoretical and computational properties. Numerical simulations based on these formulations usually begin with discretization of the surface under consideration; often, the surface is curved, possessing complicated structure and possibly singularities. Numerical simulations commonly are based on approximate, rather than exact, discretizations of these surfaces. To assess the strength of the dependence of simulation accuracy on the fidelity of surface representation, here methods were developed to model several important surface formulations using exact surface discretizations. Following and refining Zauhar's work [J. Comput.-Aided Mol. Des. 9, 149 (1995)], two classes of curved elements were defined that can exactly discretize the van der Waals, solvent-accessible, and solvent-excluded (molecular) surfaces. Numerical integration techniques are presented that can accurately evaluate nonsingular and singular integrals over these curved surfaces. After validating the exactness of the surface discretizations and demonstrating the correctness of the presented integration methods, a set of calculations are presented that compare the accuracy of approximate, planar-triangle-based discretizations and exact, curved-element-based simulations of surface-generalized-Born (sGB), surface-continuum van der Waals (scvdW), and boundary-element method (BEM) electrostatics problems. Results demonstrate that continuum electrostatic calculations with BEM using curved elements, piecewise-constant basis functions, and centroid collocation are nearly ten times more accurate than planar-triangle BEM for basis sets of comparable size. The sGB and scvdW calculations give exceptional accuracy even for the coarsest obtainable discretized surfaces. The extra accuracy is attributed to the exact representation of the solute-solvent interface; in contrast, commonly used planar-triangle discretizations can only offer improved approximations with increasing discretization and associated increases in computational resources. The results clearly demonstrate that the methods for approximate integration on an exact geometry are far more accurate than exact integration on an approximate geometry. A MATLAB implementation of the presented integration methods and sample data files containing curved-element discretizations of several small molecules are available online as supplemental material.

Progress in computational protein design.

Current progress in computational structure-based protein design is reviewed in the areas of methodology and applications. Foundational advances include new potential functions, more efficient ways of computing energetics, flexible treatments of solvent, and useful energy function approximations, as well as ensemble-based approaches to scoring designs for inclusion of entropic effects, improvements to guaranteed and to stochastic search techniques, and methods to design combinatorial libraries for screening and selection. Applications include new approaches and successes in the design of specificity for protein folding, binding, and catalysis, in the redesign of proteins for enhanced binding affinity, and in the application of design technology to study and alter enzyme catalysis. Computational protein design continues to mature and advance.

Evolution of an interloop disulfide bond in high-affinity antibody mimics based on fibronectin type III domain and selected by yeast surface display: molecular convergence with single-domain camelid and shark antibodies.

The 10th human fibronectin type III domain ((10)Fn3) is one of several protein scaffolds used to design and select families of proteins that bind with high affinity and specificity to macromolecular targets. To date, the highest affinity (10)Fn3 variants have been selected by mRNA display of libraries generated by randomizing all three complementarity-determining region -like loops of the (10)Fn3 scaffold. The sub-nanomolar affinities of such antibody mimics have been attributed to the extremely large size of the library accessible by mRNA display (10(12) unique sequences). Here we describe the selection and affinity maturation of (10)Fn3-based antibody mimics with dissociation constants as low as 350 pM selected from significantly smaller libraries (10(7)-10(9) different sequences), which were constructed by randomizing only 14 (10)Fn3 residues. The finding that two adjacent loops in human (10)Fn3 provide a large enough variable surface area to select high-affinity antibody mimics is significant because a smaller deviation from wild-type (10)Fn3 sequence is associated with a higher stability of selected antibody mimics. Our results also demonstrate the utility of an affinity-maturation strategy that led to a 340-fold improvement in affinity by maximizing sampling of sequence space close to the original selected antibody mimic. A striking feature of the highest affinity antibody mimics selected against lysozyme is a pair of cysteines on adjacent loops, in positions 28 and 77, which are critical for the affinity of the (10)Fn3 variant for its target and are close enough to form a disulfide bond. The selection of this cysteine pair is structurally analogous to the natural evolution of disulfide bonds found in new antigen receptors of cartilaginous fish and in camelid heavy-chain variable domains. We propose that future library designs incorporating such an interloop disulfide will further facilitate the selection of high-affinity, highly stable antibody mimics from libraries accessible to phage and yeast surface display methods.

Improved mutants from directed evolution are biased to orthologous substitutions.

We have engineered human epidermal growth factor (EGF) by directed evolution through yeast surface display for significantly enhanced affinity for the EGF receptor (EGFR). Statistical analysis of improved EGF mutants isolated from randomly mutated yeast-displayed libraries indicates that mutations are biased towards substitutions at positions exhibiting significant phylogenetic variation. In particular, mutations in high-affinity EGF mutants are statistically biased towards residues found in orthologous EGF species. This same trend was also observed with other proteins engineered through directed evolution in our laboratory (EGFR, interleukin-2) and in a meta-analysis of reported results for engineered subtilisin. By contrast, reported loss-of-function mutations in EGF were biased towards highly conserved positions. Based on these findings, orthologous mutations were introduced into a yeast-displayed EGF library by a process we term shotgun ortholog scanning mutagenesis (SOSM). EGF mutants with a high frequency of the introduced ortholog mutations were isolated through screening the library for enhanced binding affinity to soluble EGFR ectodomain. These mutants possess a 30-fold increase in binding affinity over wild-type EGF to EGFR-transfected fibroblasts and are among the highest affinity EGF proteins to be engineered to date. Collectively, our findings highlight a general approach for harnessing information present in phylogenetic variability to create useful genetic diversity for directed evolution. Our SOSM method exploits the benefits of library diversity obtained through complementary methods of error-prone PCR and DNA shuffling, while circumventing the need for acquisition of multiple genes for family or synthetic shuffling.

Isolating and engineering human antibodies using yeast surface display.

This protocol describes the process of isolating and engineering antibodies or proteins for increased affinity and stability using yeast surface display. Single-chain antibody fragments (scFvs) are first isolated from an existing nonimmune human library displayed on the yeast surface using magnetic-activated cell sorting selection followed by selection using flow cytometry. This enriched population is then mutagenized, and successive rounds of random mutagenesis and flow cytometry selection are done to attain desired scFv properties through directed evolution. Labeling strategies for weakly binding scFvs are also described, as well as procedures for characterizing and 'titrating' scFv clones displayed on yeast. The ultimate result of following this protocol is a panel of scFvs with increased stability and affinity for an antigen of interest.