Introduction

Therapeutic proteins are increasingly important for treating diverse diseases, including cancers, autoimmunity/inflammation, infectious diseases, and genetic disorders. For example, the plasma protein therapeutics market is expected to grow by $36 billion (USD) by 2024. Mammalian cells are the dominant production system due to their ability to perform PTMs that are required for drug safety and function (Jenkins, Murphy, & Tyther, 2008; Matasci, Hacker, Baldi, & Wurm, 2008). However, the complexities associated with the mammalian secretory machinery remains a bottleneck in recombinant protein production (Gutierrez et al., 2020). The secretory pathway machinery includes >575 gene products tasked with the synthesis, folding, PTMs, quality control, and trafficking of secreted proteins (SecPs) (Feizi, Gatto, Uhlen, & Nielsen, 2017; Lund et al., 2017; Novick, Ferro, & Schekman, 1981; Reynaud & Simpson, 2002). The precision and efficiency of the mammalian secretory pathway results from the coordinated effort of these secretory machinery components (SecMs) including chaperones, modifying enzymes (e.g., protein disulfide isomerases and glycosyltransferases), and transporters within the secretory pathway. Overexpression of heterologous proteins in this tightly regulated and complex system could impact its functionality and homeostasis, resulting in adaptive responses that can impair both protein quantity and quality (Hussain, Maldonado-Agurto, & Dickson, 2014; Young, Yuraszeck, & Robinson, 2011). More importantly, variability in the structures and modifications of recombinant proteins could necessitate a customized secretion machinery to handle this diversity, but the secretory machinery of recombinant protein producing cells has not been adapted to facilitate the high titer secretion desired for most recombinant proteins. A previous study also showed human protein secretory pathway genes are expressed in a tissue-specific pattern to support the diversity of secreted proteins and their modifications (Feizi et al., 2017), suggesting that expression of several SecMs is regulated to support client SecPs in the secretory pathway. Unfortunately, the SecMs needed to support any given secreted protein remain unknown. Thus, there is a need to elucidate the SecMs that support the expression of different recombinant proteins with specific features. This can guide mammalian synthetic biology efforts to engineer enhanced cells capable of expressing proteins of different kinds in a client-specific manner.
PPI networks are invaluable tools for deciphering the molecular basis of biological processes. New proximity dependent labeling methods such as BioID (Kim et al., 2014; Roux, Kim, Raida, & Burke, 2012) and APEX (Rhee et al., 2013) can identify weak and transient interactions in living cells, along with stable interactions. Furthermore, BioID offers a high-throughput approach for systematic detection of intracellular PPIs occurring in various cellular compartments and has been used to characterize PPI networks and subcellular organization (VarnaitÄ— & MacNeill, 2016). BioID relies on expressing a protein of interest fused to a promiscuous biotin ligase (BirA) that can biotinylate the proximal interactors in nanometer-scale labeling radius (Kim et al., 2014). For example, this approach has mapped protein interactions at human centrosomes and cilia (Firat-Karalar & Stearns, 2015; Gupta et al., 2015), focal adhesions (Dong et al., 2016), nuclear pore (Kim et al., 2014) and ER membrane-bound ribosomes (Hoffman, Chen, Zheng, & Nicchitta, 2019) . Here we used BioID2, an improved smaller biotin ligase for BioID (Kim et al., 2016; VarnaitÄ— & MacNeill, 2016), to explore how the SecMs involved vary for different secreted therapeutic proteins (Fig. 1). Specifically, BioID2 was employed to identify SecMs that interact with three SERPIN-family proteins (SERPINA1: treatment for Alpha-1-antitrypsin deficiency, SERPINC1: treatment for Hereditary antithrombin deficiency, and SERPING1: treatment for acute attacks of hereditary angioedema) and secreted embryonic alkaline phosphatase (SeAP), which is a truncated form of Alkaline Phosphatase, Placental Type (ALPP). These proteins vary in their PTMs (e.g., glycosylation, disulfide bond and residue modifications) and have different amino acid sequences that consequently form different local motifs. Using a Bayesian statistical model, we identified the critical PPIs that are positively correlated with each protein feature. Identification of these PPIs will refine our understanding of how the secretory pathway functions during the expression of the recombinant proteins and introduce novel targets for secretory pathway engineering in a client specific manner.