Design for Solubility May Reveal Induction of Amide Hydrogen/Deuterium Exchange by Protein Self-Association.

Structural heterogeneity often constrains the characterization of aggregating proteins to indirect or low-resolution methods, obscuring mechanistic details of association. Here, we report progress in understanding the aggregation of Adnectins, engineered binding proteins with an immunoglobulin-like fold. We rationally design Adnectin solubility and measure amide hydrogen/deuterium exchange (HDX) under conditions that permit transient protein self-association. Protein-protein binding commonly slows rates of HDX; in contrast, we find that Adnectin association may induce faster HDX for certain amides, particularly in the C-terminal ß-strand. In aggregation-prone proteins, we identify a pattern of very different rates of amide HDX for residues linked by reciprocal hydrogen bonds in the native structure. These results may be explained by local loss of native structure and formation of an inter-protein interface. Amide HDX induced by self-association, detected here by deliberate modulation of propensity for such interactions, may be a general phenomenon with the potential to expose mechanisms of aggregation by diverse proteins.

A Selection of Macrocyclic Peptides That Bind STING From an mRNA-Display Library With Split Degenerate Codons.

Recent improvements in mRNA display have enabled the selection of peptides that incorporate non-natural amino acids, thus expanding the chemical diversity of macrocycles beyond what is accessible in nature. Such libraries have incorporated non-natural amino acids at the expense of natural amino acids by reassigning their codons. Here we report an alternative approach to expanded amino-acid diversity that preserves all 19 natural amino acids (no methionine) and adds 6 non-natural amino acids, resulting in the highest sequence complexity reported to date. We have applied mRNA display to this 25-letter library to select functional macrocycles that bind human STING, a protein involved in immunoregulation. The resulting STING-binding peptides include a 9-mer macrocycle with a dissociation constant (KD ) of 3.4 nM, which blocks binding of cGAMP to STING and induces STING dimerization. This approach is generalizable to expanding the amino-acid alphabet in a library beyond 25 building blocks.

Using yeast surface display to engineer a soluble and crystallizable construct of hematopoietic progenitor kinase 1 (HPK1).

Hematopoietic progenitor kinase 1 (HPK1) is an intracellular kinase that plays an important role in modulating tumor immune response and thus is an attractive target for drug discovery. Crystallization of the wild-type HPK1 kinase domain has been hampered by poor expression in recombinant systems and poor solubility. In this study, yeast surface display was applied to a library of HPK1 kinase-domain variants in order to select variants with an improved expression level and solubility. The HPK1 variant with the most improved properties contained two mutations, crystallized readily in complex with several small-molecule inhibitors and provided valuable insight to guide structure-based drug design. This work exemplifies the benefit of yeast surface display towards engineering crystallizable proteins and thus enabling structure-based drug discovery.

Synthesis and Preclinical Evaluation of a 68Ga-Labeled Adnectin, 68Ga-BMS-986192, as a PET Agent for Imaging PD-L1 Expression.

Blocking the interaction of the immune checkpoint molecule programmed cell death protein-1 and its ligand, PD-L1, using specific antibodies has been a major breakthrough for immune oncology. Whole-body PD-L1 expression PET imaging may potentially allow for a better prediction of response to programmed cell death protein-1-targeted therapies. Imaging of PD-L1 expression is feasible by PET with the adnectin protein 18F-BMS-986192. However, radiofluorination of proteins such as BMS-986192 remains complex and labeling yields are low. The goal of this study was therefore the development and preclinical evaluation of a 68Ga-labeled adnectin protein (68Ga-BMS-986192) to facilitate clinical trials. Methods:68Ga labeling of DOTA-conjugated adnectin (BXA-206362) was performed in NaOAc-buffer at pH 5.5 (50°C, 15 min). In vitro stability in human serum at 37°C was analyzed using radio-thin layer chromatography and radio-high-performance liquid chromatography. PD-L1 binding assays were performed using the transduced PD-L1-expressing lymphoma cell line U-698-M and wild-type U-698-M cells as a negative control. Immunohistochemical staining studies, biodistribution studies, and small-animal PET studies of 68Ga-BMS-986192 were performed using PD-L1-positive and PD-L1-negative U-698-M-bearing NSG mice. Results:68Ga-BMS-986192 was obtained with quantitative radiochemical yields of more than 97% and with high radiochemical purity. In vitro stability in human serum was at least 95% after 4 h of incubation. High and specific binding of 68Ga-BMS-986192 to human PD-L1-expressing cancer cells was confirmed, which closely correlates with the respective PD-L1 expression level determined by flow cytometry and immunohistochemistry staining. In vivo, 68Ga-BMS-986192 uptake was high at 1 h after injection in PD-L1-positive tumors (9.0 ± 2.1 percentage injected dose [%ID]/g) and kidneys (56.9 ± 9.2 %ID/g), with negligible uptake in other tissues. PD-L1-negative tumors demonstrated only background uptake of radioactivity (0.6 ± 0.1 %ID/g). Coinjection of an excess of unlabeled adnectin reduced tumor uptake of PD-L1 by more than 80%. Conclusion:68Ga-BMS-986192 enables easy radiosynthesis and shows excellent in vitro and in vivo PD-L1-targeting characteristics. The high tumor uptake combined with low background accumulation at early imaging time points demonstrates the feasibility of 68Ga-BMS-986192 for imaging of PD-L1 expression in tumors and is encouraging for further clinical applications of PD-L1 ligands.

Adnectin-drug conjugates for Glypican-3-specific delivery of a cytotoxic payload to tumors.

Tumor-specific delivery of cytotoxic agents remains a challenge in cancer therapy. Antibody-drug conjugates (ADC) deliver their payloads to tumor cells that overexpress specific tumor-associated antigens-but the multi-day half-life of ADC leads to high exposure even of normal, antigen-free, tissues and thus contributes to dose-limiting toxicity. Here, we present Adnectin-drug conjugates, an alternative platform for tumor-specific delivery of cytotoxic payloads. Due to their small size (10 kDa), renal filtration eliminates Adnectins from the bloodstream within minutes to hours, ensuring low exposure to normal tissues. We used an engineered cysteine to conjugate an Adnectin that binds Glypican-3, a membrane protein overexpressed in hepatocellular carcinoma, to a cytotoxic derivative of tubulysin, with the drug-to-Adnectin ratio of 1. We demonstrate specific, nanomolar binding of this Adnectin-drug conjugate to human and murine Glypican-3; its high thermostability; its localization to target-expressing tumor cells in vitro and in vivo, its fast clearance from normal tissues and its efficacy against Glypican-3-positive mouse xenograft models.

Conformational Assessment of Adnectin and Adnectin-Drug Conjugate by Hydrogen/Deuterium Exchange Mass Spectrometry.

Higher-order structure (HOS) characterization of therapeutic protein-drug conjugates for comprehensive assessment of conjugation-induced protein conformational changes is an important consideration in the biopharmaceutical industry to ensure proper behavior of protein therapeutics. In this study, conformational dynamics of a small therapeutic protein, adnectin 1, together with its drug conjugate were characterized by hydrogen/deuterium exchange mass spectrometry (HDX-MS) with different spatial resolutions. Top-down HDX allows detailed assessment of the residue-level deuterium content in the payload conjugation region. HDX-MS dataset revealed the ability of peptide-based payload/linker to retain deuterium in HDX experiments. Combined results from intact, top-down, and bottom-up HDX indicated no significant conformational changes of adnectin 1 upon payload conjugation. Graphical Abstract ᅟ.

Synthesis and Biologic Evaluation of a Novel 18F-Labeled Adnectin as a PET Radioligand for Imaging PD-L1 Expression.

The programmed death protein (PD-1) and its ligand (PD-L1) play critical roles in a checkpoint pathway cancer cells exploit to evade the immune system. A same-day PET imaging agent for measuring PD-L1 status in primary and metastatic lesions could be important for optimizing drug therapy. Herein, we have evaluated the tumor targeting of an anti-PD-L1 adnectin after 18F-fluorine labeling. Methods: An anti-PD-L1 adnectin was labeled with 18F in 2 steps. This synthesis featured fluorination of a novel prosthetic group, followed by a copper-free click conjugation to a modified adnectin to generate 18F-BMS-986192. 18F-BMS-986192 was evaluated in tumors using in vitro autoradiography and PET with mice bearing bilateral PD-L1-negative (PD-L1(-)) and PD-L1-positive (PD-L1(+)) subcutaneous tumors. 18F-BMS-986192 was evaluated for distribution, binding, and radiation dosimetry in a healthy cynomolgus monkey. Results:18F-BMS-986192 bound to human and cynomolgus PD-L1 with a dissociation constant of less than 35 pM, as measured by surface plasmon resonance. This adnectin was labeled with 18F to yield a PET radioligand for assessing PD-L1 expression in vivo. 18F-BMS-986192 bound to tumor tissues as a function of PD-L1 expression determined by immunohistochemistry. Radioligand binding was blocked in a dose-dependent manner. In vivo PET imaging clearly visualized PD-L1 expression in mice implanted with PD-L1(+), L2987 xenograft tumors. Two hours after dosing, a 3.5-fold-higher uptake (2.41 ± 0.29 vs. 0.82 ± 0.11 percentage injected dose per gram, P < 0.0001) was observed in L2987 than in control HT-29 (PD-L1(-)) tumors. Coadministration of 3 mg/kg ADX_5322_A02 anti-PD-L1 adnectin reduced tumor uptake at 2 h after injection by approximately 70%, whereas HT-29 uptake remained unchanged, demonstrating PD-L1-specific binding. Biodistribution in a nonhuman primate showed binding in the PD-L1-rich spleen, with rapid blood clearance through the kidneys and bladder. Binding in the PD-L1(+) spleen was reduced by coadministration of BMS-986192. Dosimetry estimates indicate that the kidney is the dose-limiting organ, with an estimated human absorbed dose of 2.20E-01 mSv/MBq. Conclusion:18F-BMS-986192 demonstrated the feasibility of noninvasively imaging the PD-L1 status of tumors by small-animal PET studies. Clinical studies with 18F-BMS-986192 are under way to measure PD-L1 expression in human tumors.

Ensemble Modeling and Intracellular Aggregation of an Engineered Immunoglobulin-Like Domain.

The production of recombinant proteins in Escherichia coli frequently results in the formation of insoluble protein aggregates called inclusion bodies (IBs). The determinants of IB formation remain poorly understood and are of much interest for biotechnological and research applications, as well as offering insight into disease-related in vivo protein aggregation. Here we investigate a set of engineered target-binding proteins based upon the fibronectin type III domain, and we find that variations in sequence at just three positions in a solvent-exposed loop greatly alter the extent of IB formation. The loop is analogous to the third complementarity-determining region of immunoglobulin variable domains and has been shown to be conformationally mobile. In contrast to studies of other proteins, the extent of IB formation is not explained by differences in thermal stability measured by differential scanning calorimetry. Instead, IB formation is correlated with the average local stability of the FG loop, as modeled by an ensemble of structures generated using Rosetta's kinematic closure loop reconstruction method. This correlation suggests that loop instability may promote local unfolding, exposing aggregation-prone surfaces. Consistent with this mechanism, sequence-based predictions of aggregation propensity produced by Zyggregator are also correlated with IB formation, though not with modeled loop stability. The combination of average model energy scores with sequence-based aggregation predictions accounts for the variation in IB formation remarkably well (R(2)=0.8). The comparison with experimental data validates the ensemble modeling approach, which may be applicable to dynamic protein loops involved in a wide range of phenomena.

Critical features for biosynthesis, stability, and functionality of a G protein-coupled receptor uncovered by all-versus-all mutations.

The structural features determining efficient biosynthesis, stability in the membrane and, after solubilization, in detergents are not well understood for integral membrane proteins such as G protein-coupled receptors (GPCRs). Starting from the rat neurotensin receptor 1, a class A GPCR, we generated a separate library comprising all 64 codons for each amino acid position. By combining a previously developed FACS-based selection system for functional expression [Sarkar C, et al. (2009) Proc Natl Acad Sci USA 105:14808-14813] with ultradeep 454 sequencing, we determined the amino acid preference in every position and identified several positions in the natural sequence that restrict functional expression. A strong accumulation of shifts, i.e., a residue preference different from wild type, is detected for helix 1, suggesting a key role in receptor biosynthesis. Furthermore, under selective pressure we observe a shift of the most conserved residues of the N-terminal helices. This unique data set allows us to compare the in vitro evolution of a GPCR to the natural evolution of the GPCR family and to observe how selective pressure shapes the sequence space covered by functional molecules. Under the applied selective pressure, several positions shift away from the wild-type sequence, and these improve the biophysical properties. We discuss possible structural reasons for conserved and shifted residues.

Target-binding proteins based on the 10th human fibronectin type III domain (¹°Fn3).

We describe concepts and methods for generating a family of engineered target-binding proteins designed on the scaffold of the 10th human fibronectin type III domain ((10)Fn3), an extremely stable, single-domain protein with an immunoglobulin-like fold but lacking disulfide bonds. Large libraries of possible target-binding proteins can be constructed on the (10)Fn3 scaffold by diversifying the sequence and length of its surface loops, which are structurally analogous to antibody complementarity-determining regions. Target-binding proteins with high affinity and specificity are selected from (10)Fn3-based libraries using in vitro evolution technologies such as phage display, mRNA display, or yeast-surface display. (10)Fn3-based target-binding proteins have binding properties comparable to those of antibodies, but they are smaller, simpler in architecture, and more user-friendly; as a consequence, these proteins are excellent building blocks for the construction of multidomain, multifunctional chains. The ease of engineering and robust properties of (10)Fn3-based target-binding proteins have been validated by multiple independent academic and industrial groups. In addition to performing well as specific in vitro detection reagents and research tools, (10)Fn3-based binding proteins are being developed as therapeutics, with the most advanced candidate currently in Phase II clinical trials.

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.

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.

Selection of horseradish peroxidase variants with enhanced enantioselectivity by yeast surface display.

We report a method for in vitro selection of catalytically active enzymes from large libraries of variants displayed on the surface of the yeast S. cerevisiae. Two libraries, each containing approximately 2 x 10(6) variants of horseradish peroxidase (HRP), were constructed; one involved error-prone PCR that sampled mutations throughout the coding sequence, whereas the other involved complete combinatorial enumeration of five positions near the active site to non-cysteine residues. The enzyme variants displayed on the yeast surface were allowed to modify it with a fluorescently labeled substrate. A combination of positive and negative selection applied to the active-site-directed library resulted in variants with up to an 8-fold altered enantioselectivity, including its reversal, toward L/D-tyrosinol. In contrast, the library constructed by using error-prone PCR yielded no HRP variants with a significantly improved enantioselectivity.

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.

In-vitro protein evolution by ribosome display and mRNA display.

In-vitro display technologies combine two important advantages for identifying and optimizing ligands by evolutionary strategies. First, by obviating the need to transform cells in order to generate and select libraries, they allow a much higher library diversity. Second, by including PCR as an integral step in the procedure, they make PCR-based mutagenesis strategies convenient. The resulting iteration between diversification and selection allows true Darwinian protein evolution to occur in vitro. We describe two such selection methods, ribosome display and mRNA display. In ribosome display, the translated protein remains connected to the ribosome and to its encoding mRNA; the resulting ternary complex is used for selection. In mRNA display, mRNA is first translated and then covalently bonded to the protein it encodes, using puromycin as an adaptor molecule. The covalent mRNA-protein adduct is purified from the ribosome and used for selection. Successful examples of high-affinity, specific target-binding molecules selected by in-vitro display methods include peptides, antibodies, enzymes, and engineered scaffolds, such as fibronectin type III domains and synthetic ankyrins, which can mimic antibody function.

Directed evolution of high-affinity antibody mimics using mRNA display.

We constructed a library of >10(12) unique, covalently coupled mRNA-protein molecules by randomizing three exposed loops of an immunoglobulin-like protein, the tenth fibronectin type III domain (10Fn3). The antibody mimics that bound TNF-alpha were isolated from the library using mRNA display. Ten rounds of selection produced 10Fn3 variants that bound TNF-alpha with dissociation constants (K(d)) between 1 and 24 nM. After affinity maturation, the lowest K(d) measured was 20 pM. Selected antibody mimics were shown to capture TNF-alpha when immobilized in a protein microarray. 10Fn3-based scaffold libraries and mRNA-display allow the isolation of high-affinity, specific antigen binding proteins; potential applications of such binding proteins include diagnostic protein microarrays and protein therapeutics.

Identification of epitope-like consensus motifs using mRNA display.

The mRNA display approach to in vitro protein selection is based upon the puromycin-mediated formation of a covalent bond between an mRNA and its gene product. This technique can be used to identify peptide sequences involved in macromolecular recognition, including those identical or homologous to natural ligand epitopes. To demonstrate this approach, we determined the peptide sequences recognized by the trypsin active site, and by the anti-c-Myc antibody, 9E10. Here we describe the use of two peptide libraries of different diversities, one a constrained library based on the trypsin inhibitor EETI-II, where only the six residues in the first loop were randomized (6.4 x 10(7) possible sequences, 6.0 x 10(11) sequences in the library), the other a linear-peptide library with 27 randomized amino acids (1.3 x 10(35) possible sequences, 2 x 10(13) sequences in the library). The constrained library was screened against the natural target of wild-type EETI, bovine trypsin, and the linear library was screened against the anti-c-myc antibody, 9E10. The analysis of selected sequences revealed minimal consensus sequences of PR(I,L,V)L for the first loop of EETI-II and LISE for the 9E10 epitope. The wild-type sequences, PRILMR for the first loop of EETI-II and QKLISE for the 9E10 epitope, were selected with the highest frequency, and in each case the complete wild-type epitope was selected from the library.