Ramanujan, S. | Search bioRxiv

bioRxiv2024

Molecular characterization of response to etrolizumab and anti-TNF reveals treatment resistance in ulcerative colitis is associated with the abundance of residual neutrophil subsets and inflammatory fibroblast populations

Ulcerative colitis (UC) is a chronic inflammatory large bowel disease characterized by immune cell infiltration and continuous erosion of intestinal crypts, causing severe ulceration and abdominal pain. In the etrolizumab Phase 3 studies, transcriptional analyses of colonic biopsies revealed reductions in genes associated with aEb7+ intraepithelial lymphocytes with etrolizumab but not adalimumab. Both treatments significantly reduced stromal and myeloid cell-associated genes, with changes associated with MCS remission status. Generation of a single-cell atlas from inflamed and uninflamed colonic biopsies from UC patients led to the identification of thirty-six discrete cell populations, including cells of the myeloid compartment. The UC atlas was used to generate cell-specific signatures, allowing for cellular deconvolution of the Phase 3 datasets. It revealed significant reductions in neutrophil subsets, monocyte-derived macrophages, and inflammatory fibroblasts, as well as increases in colonic epithelial cells common to both etrolizumab and adalimumab. Pseudo-time trajectory analyses identified four unique neutrophil subsets with unique cell phenotypes reflecting changes in cell state or differentiation from PADI4hi, OSMhi, MX1hi, and ultimately to CXCR4hi populations. PADI4hi and OSMhi neutrophils exhibited high levels of proteases (MMP9, LYZ), inflammatory cytokines (CXCL1, IL1B, OSM), and abundant cytokine or chemokine receptors (CXCR1, CXCR2). MX1 populations expressed markers indicating prior IFN exposure (MX1, IFIT1). In contrast, more differentiated or mature neutrophils exhibited high levels of CXCL2, TNF-a, and CXCR4, as well as angiogenic factors like VEGFA. PADI4hi and OSMhi neutrophils, we predict, have abundant cytokine and chemokine interactions with inflammatory fibroblasts within the inflamed colon, such as OSM: OSMR and IL1B: IL1R1 interactions. Changes in PADI4hi and OSMhi neutrophils were closely associated with MCS remission in both etrolizumab and adalimumab-treated patients. In contrast, only minor changes in CXCR4hi neutrophils were observed and not associated with clinical outcomes. Our results suggest that neutrophils are not only heterogeneous in phenotype but have abundant cell-cell interactions in inflamed colonic tissue that are likely implicated in maintaining chronic disease activity. We hypothesize that limiting the interactions between neutrophils and other myeloid cells with resident cells such as inflammatory fibroblasts may reduce the production of inflammatory mediators and limit activation and infiltration of neutrophils, which may be necessary for achieving greater rates of clinical remission in response to interventional agents.

immunology↗

bioRxiv2024

Leveraging protein language and structural modelsfor early prediction of antibodies with fast clearance

Monoclonal antibodies (mAbs) with long systemic persistence are widely used as therapeutics. However, antibodies with atypically fast clearance require more dosing, limiting their clinical usefulness. Deep learning can facilitate using sequence-based modeling to predict potential pharmacokinetic (PK) liabilities before antibody generation. Assembling a dataset of 103 mAbs with measured nonspecific clearance in cynomolgus monkeys (cyno), and using transfer learning from large protein language models, we developed multiple machine learning models to predict mAb clearance as fast/slow clearing. Focusing on minimizing misclassification of potentially promising molecules as fast clearing, our results show that using physicochemical properties yielded up to 73.1+/-1.1% classification accuracy on hold-out test data (precision 65.2+/-2.3%). Using only sequence-based features from deep learning protein language models yielded a comparable performance of 71+/-1.4% (precision 65.5+/-2.5%). Combining structural and deep learning derived features yielded a similar accuracy of 73.9+/-1.1%, and slightly improved precision (68.3+/-2.4%). Features important for classifying fast/slow clearance point to charge, moment, and surface area properties at pH 7.4 as well as deep learning derived features. These results suggest that the protein language models provide comparable information and predictive performance of clearance as physicochemical features. This work provides a foundation for in silico prediction of protein pharmacokinetics to inform antibody candidate generation and early deprioritization of designs with high risk of fast clearance. More generally, it illustrates the value of transfer learning-based application of protein language models to address characteristics of importance for protein therapeutics.

pharmacology and toxicology↗