ABSTRACT
Antibodies are one of the predominant treatment modalities for various diseases. To improve the characteristics of a lead antibody, such as antigen-binding affinity and stability, we conducted comprehensive substitutions and exhaustively explored their sequence space. However, it is practically unfeasible to evaluate all possible combinations of mutations owing to combinatorial explosion when multiple amino acid residues are incorporated. It was recently reported that a machine-learning guided protein engineering approach such as Thompson sampling (TS) has been used to efficiently explore sequence space in the framework of Bayesian optimization. For TS, over-exploration occurs when the initial data are biasedly distributed in the vicinity of the lead antibody. We handle a large-scale virtual library that includes numerous mutations. When the number of experiments is limited, this over-exploration causes a serious issue. Thus, we conducted Monte Carlo Thompson sampling (MTS) to balance the exploration-exploitation trade-off by defining the posterior distribution via the Monte Carlo method and compared its performance with TS in antibody engineering. Our results demonstrated that MTS largely outperforms TS in discovering desirable candidates at an earlier round when over-exploration occurs on TS. Thus, the MTS method is a powerful technique for efficiently discovering antibodies with desired characteristics when the number of rounds is limited.
Subject(s)
Antibodies , Protein Engineering , Bayes Theorem , Monte Carlo Method , Antibodies/chemistry , Protein Engineering/methodsABSTRACT
Molecular evolution is an important step in the development of therapeutic antibodies. However, the current method of affinity maturation is overly costly and labor-intensive because of the repetitive mutation experiments needed to adequately explore sequence space. Here, we employed a long short term memory network (LSTM)-a widely used deep generative model-based sequence generation and prioritization procedure to efficiently discover antibody sequences with higher affinity. We applied our method to the affinity maturation of antibodies against kynurenine, which is a metabolite related to the niacin synthesis pathway. Kynurenine binding sequences were enriched through phage display panning using a kynurenine-binding oriented human synthetic Fab library. We defined binding antibodies using a sequence repertoire from the NGS data to train the LSTM model. We confirmed that likelihood of generated sequences from a trained LSTM correlated well with binding affinity. The affinity of generated sequences are over 1800-fold higher than that of the parental clone. Moreover, compared to frequency based screening using the same dataset, our machine learning approach generated sequences with greater affinity.
Subject(s)
Algorithms , Antibodies/immunology , Antibody Affinity/immunology , Cell Surface Display Techniques , Protein Engineering , Amino Acid Sequence , Databases, Protein , High-Throughput Nucleotide Sequencing , Humans , Likelihood Functions , Machine Learning , Reproducibility of ResultsABSTRACT
Agonistic antibodies targeting CD137 have been clinically unsuccessful due to systemic toxicity. Because conferring tumor selectivity through tumor-associated antigen limits its clinical use to cancers that highly express such antigens, we exploited extracellular adenosine triphosphate (exATP), which is a hallmark of the tumor microenvironment and highly elevated in solid tumors, as a broadly tumor-selective switch. We generated a novel anti-CD137 switch antibody, STA551, which exerts agonistic activity only in the presence of exATP. STA551 demonstrated potent and broad antitumor efficacy against all mouse and human tumors tested and a wide therapeutic window without systemic immune activation in mice. STA551 was well tolerated even at 150 mg/kg/week in cynomolgus monkeys. These results provide a strong rationale for the clinical testing of STA551 against a broad variety of cancers regardless of antigen expression, and for the further application of this novel platform to other targets in cancer therapy. SIGNIFICANCE: Reported CD137 agonists suffer from either systemic toxicity or limited efficacy against antigen-specific cancers. STA551, an antibody designed to agonize CD137 only in the presence of extracellular ATP, inhibited tumor growth in a broad variety of cancer models without any systemic toxicity or dependence on antigen expression.See related commentary by Keenan and Fong, p. 20.This article is highlighted in the In This Issue feature, p. 1.
Subject(s)
Adenosine Triphosphate , Neoplasms , Animals , Antibodies, Monoclonal/pharmacology , Antigens, Neoplasm , Immunotherapy , Mice , Neoplasms/drug therapy , Tumor Microenvironment , Tumor Necrosis Factor Receptor Superfamily, Member 9ABSTRACT
TSC1 acts coordinately with TSC2 in a complex to inhibit mTOR, an emerging therapeutic target and known promoter of cell growth and cell cycle progression. Perturbation of the mTOR pathway, through abnormal expression or function of pathway genes, could lead to tumorigenesis. TSC1 and TSC2 expression is reduced in invasive breast cancer as compared with normal mammary epithelium. Because single nucleotide polymorphisms (SNPs) in regulatory genes have been implicated in risk and age at diagnosis of breast cancers, systematic SNP association studies were performed on TSC1 and TSC2 SNPs for their associations with clinical features of breast cancer. TSC1 and TSC2 haplotypes were constructed from genotyping of multiple loci in both genes in healthy volunteers. SNPs were selected for further study using a bioinformatics approach based on SNP associations with drug response in NCI-60 cell lines and evidence of selection bias based on haplotype frequencies. Genotyping for five TSC1 and one TSC2 loci were performed on genomic DNA from 1,137 women with breast cancer. This study found that for TSC1 rs7874234, TT variant carriers had a 9-year later age at diagnosis of estrogen receptor positive (ER+), but not ER-, ductal carcinomas (P = 0.0049). No other SNP locus showed an association with age at diagnosis, nor any other breast cancer phenotype. TSC1 rs7874234 is hypothesized to be functional in ER+ breast cancer because the T allele, but not the C allele, may create an estrogen receptor element (ERE) site, resulting in increased TSC1 transcription and subsequent inhibition of mTOR.
Subject(s)
Breast Neoplasms/genetics , Breast Neoplasms/metabolism , Gene Expression Regulation, Neoplastic , Polymorphism, Genetic , Tumor Suppressor Proteins/genetics , Alleles , Carcinoma, Ductal/genetics , Computational Biology/methods , Estrogen Receptor alpha/biosynthesis , Female , Genotype , Haplotypes , Humans , Phenotype , Time Factors , Tuberous Sclerosis Complex 1 Protein , Tuberous Sclerosis Complex 2 ProteinABSTRACT
The motile mechanism of Mycoplasma mobile remains unknown but is believed to differ from any previously identified mechanism in bacteria. Gli349 of M. mobile is known to be responsible for both adhesion to glass surfaces and mobility. We therefore carried out sequence analyses of Gli349 and its homolog MYPU2110 from M. pulmonis to decipher their structures. We found that the motif "YxxxxxGF" appears 11 times in Gli349 and 16 times in MYPU2110. Further analysis of the sequences revealed that Gli349 contains 18 repeats of about 100 amino acid residues each, and MYPU2110 contains 22. No sequence homologous to any of the repeats was found in the NCBI RefSeq non-redundant sequence database, and no compatible fold structure was found among known protein structures, suggesting that the repeat found in Gli349 and MYPU2110 is novel and takes a new fold structure. Proteolysis of Gli349 using chymotrypsin revealed that cleavage positions were often located between the repeats, implying that regions connecting repeats are unstructured, flexible and exposed to the solvent. Assuming that each repeat folds into a structural domain, we constructed a model of Gli349 that fits well the shape and size of images obtained with electron microscopy.