35Keywords 36 Metabolic network, secretory pathway, biotherapeutic production, systems biotechnology 37Indeed, recent studies have incorporated portions of the secretory pathway in metabolic models of yeast 55 3-5 . Furthermore, Lund and colleagues reconstructed a genetic interaction network of the mouse secretory 56 pathway and the unfolded protein response and analyzed it in the context of CHO cells 6 . However, such a 57 network does not encompass a stoichiometric reconstruction of the biochemical reactions involved in the 58 secretory pathway nor it is coupled to existing metabolic networks of mammalian cells. 59Here we present the first genome-scale stoichiometric reconstructions and computational models of 60 mammalian metabolism coupled to protein secretion. Specifically, we constructed these for human, 61 mouse, and CHO cells, called RECON2.2s, iMM1685s, and iCHO2048s, respectively. We first derive an 62 expression for computing the energetic cost of synthesizing and secreting a product in terms of molecules 63 of ATP equivalents per protein molecule. We use this expression and analyze how the energetic burden 64 of protein secretion has led to an overall suppression of more expensive secreted host cell proteins in 65 mammalian cells. Given its dominant role in biotherapeutic production, we further focus on the 66 biosynthetic capabilities of CHO cells. We then demonstrate that product-specific secretory pathway 67 models can be built to estimate CHO cell growth rates given the specific productivity of the recombinant 68 3 product as a constraint. We identify the features of secreted proteins that have the highest impact on 69 protein cost and productivity rates. Finally, we use our model to identify proteins that compete for cell 70 resources, thereby presenting targets for cell engineering. Through this study we demonstrate that a 71 systems-view of the secretory pathway now enables the analysis of many biomolecular mechanisms 72 controlling the efficacy and cost of protein expression in mammalian cells. We envision our models as 73 valuable tools for the study of normal physiological processes and engineering cell bioprocesses in 74 biotechnology. All models and data used in this study are freely available at 75 https://github.com/LewisLabUCSD/MammalianSecretoryRecon. 76 77 RESULTS 78 A stoichiometric expression of protein secretion energetics 79In any cell, the secretory machinery is concurrently processing thousands of secreted and membrane 80 proteins, which all compete for secretory pathway resources and pose a metabolic burden. To quantify 81 this burden, we estimated the energetic cost of synthesizing and/or secreting 5,641 and 3,538 82 endogenous proteins in the CHO and human secretome and membrane proteome in terms of total 83 number of ATP equivalent molecules consumed (see Methods). These protein costs were compared to 84 the cost of five recombinant proteins commonly produced in CHO cells (Fig. 1a). To refine estimates, we 85 predicted signal peptides 7 , GPI anchor attachment signals 8 , an...

In mammalian cells, >25% of synthesized proteins are exported through the secretory pathway. The pathway complexity, however, obfuscates its impact on the secretion of different proteins. Unraveling its impact on diverse proteins is particularly important for biopharmaceutical production. Here we delineate the core secretory pathway functions and integrate them with genome-scale metabolic reconstructions of human, mouse, and Chinese hamster ovary cells. The resulting reconstructions enable the computation of energetic costs and machinery demands of each secreted protein. By integrating additional omics data, we find that highly secretory cells have adapted to reduce expression and secretion of other expensive host cell proteins. Furthermore, we predict metabolic costs and maximum productivities of biotherapeutic proteins and identify protein features that most significantly impact protein secretion. Finally, the model successfully predicts the increase in secretion of a monoclonal antibody after silencing a highly expressed selection marker. This work represents a knowledgebase of the mammalian secretory pathway that serves as a novel tool for systems biotechnology.

No abstract

Mass spectrometry is being used to identify protein biomarkers that can facilitate development of drug treatment. Mass spectrometry-based labeling proteomic experiments result in complex proteomic data that is hierarchical in nature often with small sample size studies. The generalized linear model (GLM) is the most popular approach in proteomics to compare protein abundances between groups. However, GLM does not address all the complexities of proteomics data such as repeated measures and variance heterogeneity. Linear models for microarray data (LIMMA) and mixed models are two approaches that can address some of these data complexities to provide better statistical estimates. We compared these three statistical models (GLM, LIMMA, and mixed models) under two different normalization approaches (quantile normalization and median sweeping) to demonstrate when each approach is the best for tagged proteins. We evaluated these methods using a spiked-in data set of known protein abundances, a systemic lupus erythematosus (SLE) data set, and simulated data from multiplexed labeling experiments that use tandem mass tags (TMT). Data are available via ProteomeXchange with identifier PXD005486. We found median sweeping to be a preferred approach of data normalization, and with this normalization approach there was overlap with findings across all methods with GLM being a subset of mixed models. The conclusion is that the mixed model had the best type I error with median sweeping, whereas LIMMA had the better overall statistical properties regardless of normalization approaches.

Derivitization of peptides with isobaric tags such as iTRAQ and TMT is widely employed in proteomics due to their compatibility with multiplex quantitative measurements. We recently made publicly available a large peptide library derived from iTRAQ 4-plex labeled spectra. This resource has not been used for identifying peptides labeled with related tags with different masses, because values for virtually all masses of precursor and most product ions would differ for ions containing the different tags as well as containing different tag-specific peaks. We describe a method for interconverting spectra from iTRAQ 4-plex to TMT (6- and 10-plex) and to iTRAQ 8-plex. We interconvert spectra by appropriately mass shifting sequence ions and discarding derivative-specific peaks. After this "cleaning" of search spectra, we demonstrate that the converted libraries perform well in terms of peptide spectral matches. This is demonstrated by comparing results using sequence database searches as well as by comparing search effectiveness using original and converted libraries. At 1% FDR TMT labeled query spectra match 97% as many spectra against a converted iTRAQ library as compared to an original TMT library. Overall this interconversion strategy provides a practical way to extend results from one derivatization method to others that share related chemistry and do not significantly alter fragmentation profiles.