Chinese hamster ovary (CHO) cells are used to produce almost 90% of therapeutic monoclonal antibodies (mAbs) and antibody fusion proteins (Fc-fusion). The annotation of non-canonical translation events in these cellular factories remains incomplete, limiting our ability to study CHO cell biology and detect host cell protein(HCP) impurities in the final antibody drug product. We utilised ribosome footprint profiling (Ribo-seq) to identify novel open reading frames(ORFs)including N-terminal extensions and thousands of short ORFs (sORFs) predicted to encode microproteins. Mass spectrometry based HCP analysis of eight commercial antibody drug products (7 mAbs and 1 Fc-fusion protein) using the extended protein sequence database revealed the presence of microprotein impurities. We present evidence that micro protein abundance varies with growth phase and can be affected by the cell culture environment. In addition, our work provides a vital resource to facilitate future studies of non-canonical translation and the regulation of protein synthesis in CHO cell lines.

The secretory pathway processes >30% of mammalian proteins, orchestrating their synthesis, modification, trafficking, and quality control. However, its complexity—spanning multiple organelles and dependent on coordinated protein interactions—limits our ability to decipher how protein secretion is controlled in biomedical and biotechnological applications. To advance such research, we present secRecon—a comprehensive reconstruction of the mammalian secretory pathway, comprising 1,127 manually curated genes organized within an ontology of 77 secretory process terms, annotated with functional roles, subcellular localization, protein interactions, and complex composition. Using secRecon to analyze multi-omics data, we identified distinct secretory topologies in antibody-producing plasma cells compared to CHO cells. Genes within proteostasis, translocation, and N-glycosylation are deficient in CHO cells, highlighting them as potential engineering targets to boost secretion capacity. Applying secRecon to single-cell transcriptomics and SEC-seq data, we uncovered secretory pathway signatures underlying secretion diversity among IgG-secreting plasma cells. Different transcriptomic clusters had unique secretory phenotypes characterized by variations in the unfolded protein response (UPR), endoplasmic reticulum-associated degradation (ERAD), and vesicle trafficking pathways. Additionally, we discovered specific secretory machinery genes as new markers for plasma cell differentiation. These findings demonstrate secRecon can identify mechanisms regulating protein secretion and guide diverse studies in biomedical research and biotechnology.

Public protein sequence databases contain samples from the fitness landscape explored by nature. Protein language models (pLMs) pre-trained on these sequences aim to capture this landscape for tasks like property prediction and protein design. Following natural language processing, pLMs have been continuously scaled up. However, the assumption that scale leads to better performance assumes that databases accurately represent the fitness landscape, which is likely false. Assuming that most UniRef sequences are real and not pseudo-genes or the result of sequencing errors, then UniRef100 is a more representative sample of the fitness landscape. This is because redundancy is a sign of observation count, which we should treat as a measure of confidence, and clustering is in fact expected because of the nature of protein evolution - these clusters represent families of related proteins that are common for functional reasons. Removing sequences that show up in large clusters downweights confident sequences and ultimately upweights singleton sequences that are more likely to derive from sequencing errors and other sources of noise. Equipped with this understanding, and thanks to an efficient codebase and modern architecture, we introduce AMPLIFY, a best-in-class pLM that is orders of magnitude less expensive to train and deploy compared to previous models like ESM2 that are trained on clustered versions of UniRef. To support the scientific community and democratize the training of pLMs, we have open-sourced AMPLIFY’s pre-training codebase, data, and model checkpoints.

Human embryonic kidney cells (HEK293) serve as cell factories in viral vector manufacturing, particularly for recombinant adeno-associated virus (rAAV) production. They find widespread utilization in industrial applications, but a comprehensive characterization of the HEK293 genome and epigenome stability is still missing. To address this knowledge gap, the study employs a systematic approach to examine the genetic landscapes of various HEK293 cell lines. The objective is to evaluate their responses to changing environmental conditions and improve the current understanding of how these molecular mechanisms might influence rAAV production processes. Therefore, adherent HEK293 cells were adapted to suspension growth using various commercially available serum-free media formulations. Following successful adaptation, whole-genome deep sequencing was performed on both adapted and parental cell lines. The sequenced reads were then aligned to the human reference genome, enabling the assessment of genome stability, by evaluation of identified structural variants. Comparative analysis of these cell lines along with publicly available genome sequences of different HEK293 derivatives revealed a characteristic genetic signature common to all HEK293 cells, independent of cultivation conditions, phenotypic divergence or phylogenetic distance. Alterations in the distribution of structural variants including insertions and deletions, and of single nucleotide polymorphisms, indicate a continuous accumulation of genetic changes over time in culture, rather than abrupt genomic shifts in response to altered cultivation conditions. In contrast, adenoviral genes integrated into HEK293 cells appear to be highly conserved, as indicated by their stable copy number and consistent integration site. Overall, this work offers novel insights into the cellular response of various HEK293 cells to different cultivation conditions. Furthermore, it lays the groundwork for more comprehensive omics characterization, to support the development of cell lines with higher production efficiency.

The development of cell lines that reliably express large titers of biologics such as monoclonal antibodies (MAbs) is an important step of bioprocess development. After transfection of CHO cells, stable pools are obtained, and clones are isolated. As each clone can behave distinctly when cultivated, clone selection is crucial. This study proposes an adaptable multifactorial data analysis method for clone selection that takes multiple process parameters into account, in addition to the traditionally considered titer and cell growth. We applied our innovative approach on existing in-house data corresponding to the generation of CHO clones producing Omalizumab, an IgG1 monoclonal antibody. Twenty-four clones were chosen for expression stability screening tests conducted in extra deep well plates (18 mL). Among them, eight stable clones were selected for scalability evaluation performed in benchtop stirred-tank bioreactors (0.75-1 L). Considering parameters related to productivity, cell growth and expression stability led to the selection of different clones for bioreactor scalability assessment experiments compared to those originally selected based on a conventional method. As efficient CO2 stripping is challenging in large-scale bioreactors, experiments were conducted with and without addition of air in the bioreactor headspace to modulate the concentration of CO2 in the media. We found that cultures without overlay air addition reached a pCO2 of up to 190 mmHg. Of note, we showed the increased concentration of CO2 to be beneficial. Indeed, on average, we measured 1.31-fold higher final titer, 1.14-fold higher cell specific productivity, and 1.13-fold greater peak viable cell density for cultures exposed to a greater pCO2. In addition, these cultures benefitted from 3 to 5 more days above 80% viability, and their titer kept increasing until the end of the culture (17-21 days). We integrated statistical tools to reliably analyze datasets relating to productivity, cell growth, and key cellular metabolites. This new approach helped selection of a robust clone which performed similarly for low and high pCO2, offering a better potential for subsequent scale-up. These findings underscore the great potential of MVDA to improve bioprocess development.