Enhanced Production of HCV E1E2 Subunit Vaccine Candidates via Protein-Protein Interaction Identification in Glycoengineered CHO cells

Abstract

Hepatitis C Virus (HCV) is a pervasive bloodborne virus and the leading cause of chronic liver disease and cancer. Thus, development of an HCV vaccine is of great importance. Prior work has developed candidate vaccines, including more potent glycoengineered viral proteins and secreted forms of the E1E2 envelope heterodimer (sE1E2). However, efforts to express them recombinantly in Chinese hamster ovary (CHO) cells have resulted in very low titers. To address this challenge, here we employed a multi-omics approach to identify protein interactors that may enhance the secretion of an sE1E2 vaccine candidate. We detected protein-protein interactions (PPIs) using the Biotinylation by Antibody Recognition (BAR) assay and integrated these data with RNA-Seq. Through this we identified and overexpressed proteins that interact with sE1E2 in CHO cells. Among these, CUL4A and YWHAH enhanced sE1E2 secretion in our glycoengineered CHO cells. The integration of omics techniques and genetic engineering in this study provides valuable insights into the host cell proteins that interact with the HCV E1E2 heterodimer, and how they may be harnessed to improve protein secretion in CHO cells to enable more affordable and accessible biotherapeutics.

Graphical Abstract

E1E2 glycoprotein from Hepatitis C virus was expressed in glycoengineered CHO cells to generate stable clones. Omics techniques including BAR assay and bulk RNA-seq identified protein interactors associated with E1E2 secretion. Two proteins, CUL4A and YWHAH, were identified whose overexpression enhances E1E2 secretion.

Introduction

Hepatitis C Virus (HCV) is the leading cause of chronic liver disease such as cirrhosis and hepatocellular carcinoma (Khullar & Firpi, 2015). Given that it affects more than 57 million people globally, there is a great need for a vaccine. HCV can be treated with direct-acting antivirals (DAAs) such as sofosbuvir/velpatasvir over the course of a few months; however, by the time of viral clearance, damage has often already occurred. Moreover, DAAs do not prevent reinfections or transmissions, since infections often go unnoticed (Manns et al., 2017). An effective vaccine for HCV has yet to be achieved due to several factors. First, the error-prone RNA replication process of HCV leads to high mutation rates, resulting in at least 7 genotypes and over 90 subtypes of HCV (Duncan et al., 2020; Guest & Pierce, 2021). Second, viral glycans mask important neutralizing epitopes on the E1E2 heterodimer complex surface (Hedskog et al., 2019; Peña et al., 2022). Third, immunodominant, non-neutralizing epitopes distract the immune response from generating broadly neutralizing antibodies (bNAbs) (Toth et al., 2021). Fourth, it remains challenging to produce a homogeneous E1E2 vaccine at scale, given the difficulty of generating large quantities of subunit vaccine protein (Toth et al., 2021; Lavie et al., 2007). Therefore, there remain multiple challenges to vaccine development.

Recent advances in new technology platforms exist that can be applied to develop more efficacious vaccines. For example, changes in glycosylation may enhance vaccine efficacy (Deng et al., 2022; Liu et al., 2016). These effects likely occur because glycosylation can impact protein stability and half-life, which in turn can enhance efficacy and immune responses (Rocamora et al., 2023). In addition, removing glycosylation sites on the protein can expose important epitope sites for neutralizing antibody (NAb) binding. Specifically, by adding glycans to cover up known neutralization binding sites while simultaneously removing other specific N-glycan sites, new binding regions can be exposed (Ringe et al., 2019). Thus, tuning glycosylation on vaccine proteins with glycoengineered CHO (geCHO) cells could help elicit NAbs (Kulakova et al., 2025).

Consistent with the evidence that glycosylation impacts vaccine efficacy, we previously found that geCHO cells can enhance neutralization of a secreted E2 (sE2) vaccine candidate (Kulakova et al., 2025), particularly one glycoform we called geCHO.sE2.1, produced in a geCHO cell referred to here as CL1. CL1 was genetically modified to produce simple glycans that are monoantennary, asialylated and fucosylated on glycoproteins with glycosyltransferase knockouts (KO) for Mgat2, St3gal3/4/6, B3gnt2, and Sppl3. In contrast, our control (CTRL) CHO cells produce branched and complex glycans, as they only harbor a Sppl3 KO (to increase glycan size).

Aside from glycoengineering, other protein engineering strategies can further enhance vaccine design. For example, an improved HCV vaccine candidate was developed by including the E1 subunit alongside E2, forming an E1E2 heterodimer vaccine that mimics the glycoprotein on the HCV envelope (Guest et al., 2021; Wang et al., 2022; Metcalf et al., 2023). The native membrane-bound E1E2 (mbE1E2) heterodimer on the virus surface has been targeted for vaccine designs to induce bNAbs and stimulate B cell immunity (Lavie et al., 2007; Peña et al., 2022; Toth et al., 2021). The E1E2 complex vaccine presents conserved antigenic epitope sites on E2, including domains A, B/AR3, D, and E; and antigenic epitope sites on E1 including the N-terminal residues 192–202 and a conserved alpha-helix 314–327 (Toth et al., 2023). In addition, the E1E2 heterodimer structure presents additional antigenic regions such as AR4 and AR5, which require intact E1E2 for effective antibody recognition (Pierce et al., 2024). Notably, AR4A is associated with viral clearance (Frumento et al., 2022; Kinchen et al., 2019). Studies in chimpanzees have shown that vaccination with the E1E2 complex elicits a stronger immune response compared to E2 alone (Frey et al., 2010; Kundu et al., 2024).

With mbE1E2 being a target for a protein-based vaccine, there is a clear need for a soluble form amenable to mass production in high quantities. This is a challenge since the native E1E2 heterodimer is a membrane-bound protein, making current production of the mbE1E2 in CHO cells difficult. Indeed, it yields only 1 mg of purified mbE1E2 per 100g of CHO cells (Logan, et al. 2016) and 2–5 mg per liter in suspension HEK cells (Toth, et al. 2021). In addition, non-uniformity of mbE1E2 can complicate downstream processing and make quality control difficult (Guest et al., 2021). Fortunately, secreted versions of E1E2 have been created by replacing the transmembrane domain (TMD) with natural and synthetic scaffolds such as leucine zipper and SynZip (SZ), respectively (Guest, et al., 2021; Metcalf et al., 2023). Therefore, a secreted version of E1E2 (sE1E2.SZ) was created with a furin cleavage site containing six arginines between the E1 and E2 ectodomains. This facilitates proper cleavage and encourages native-like assembly of the E1E2 by forming coiled-coil interactions between SZ1 and SZ2 (Metcalf et al., 2023). Additionally, a stabilizing mutation (H445P) in Domain D was incorporated into the sE1E2.SZ design expressed in this study, sE1E2.SZ-H445P. This mutation is located in a region lacking secondary structure near key antibody-binding sites and has been shown to improve antibody binding stability by reducing the epitope site’s mobility (Pierce et al., 2020).

Building upon the promising results of E1E2-based vaccine development, we attempted to express the sE1E2.SZ-H445P construct in our CTRL and CL1 geCHO cell lines. However, initial expression trials resulted in insufficient protein yields to support downstream purification and subsequent binding affinity and antigenicity testing protocols. To address this production limitation, we hypothesized that identifying and manipulating the protein-protein interaction (PPI) network surrounding sE1E2 could reveal cellular bottlenecks limiting secretion efficiency (Wu et al., 2025).

We measured sE1E2 PPIs using the biotinylation by antibody recognition (BAR) method (Bar et al., 2018; Masson et al., 2024). For this, sE1E2 cells were fixed and permeabilized, then incubated with anti-E1E2 antibodies. A secondary antibody conjugated with horseradish peroxidase was then applied, along with H₂O₂ and tyramide-biotin. This process biotinylated the proteins interacting with sE1E2. The biotinylated proteins were then purified and identified via mass spectrometry (MS). We also obtained RNA-Seq from the sE1E2-producing geCHO to complement our proteomic findings.

Based on the integrated BAR and RNA-Seq analysis, we selected a panel of significant PPIs for functional validation through targeted overexpression experiments in geCHO cells. Among the candidates tested, two proteins emerged as particularly effective enhancers of sE1E2 secretion: CUL4A (Cullin-4A) and YWHAH (14-3-3 protein eta). Overexpression of CUL4A or YWHAH resulted in significant improvements in sE1E2 secretion levels compared to control conditions, validating our hypothesis that manipulating the cellular protein interaction network can enhance difficult-to-express protein production. These findings demonstrate the practical utility of combining proteomic interaction mapping with transcriptomic analysis to identify targets for cellular engineering.

Results

Generating stable cell clones expressing sE1E2.SZ-H445P

Plasmids harboring genes for furin protease and sE1E2.SZ-H445P were transfected into the CTRL and CL1 cell lines (Figure 1A) by electroporation to generate stable pools expressing sE1E2 via cloning. After transfection, we screened ~20 individual clones for sE1E2 secretion for each cell line via dot blot analysis to identify functional expressing variants (Supplementary Figure S1). Two clones from each cell line were selected for PPI analysis and transcriptomics: CTRL(H445P:Furin) #1, CTRL(H445P:Furin) #3, CL1(H445P:Furin) #17, and CL1(H445P:Furin) #21. These specific clones were chosen based on their consistent sE1E2 secretion profiles and represented suitable biological replicates for downstream experimental applications.

Figure 1. sE1E2.SZ-H445P was expressed in two geCHO cell lines and neighboring proteins were biotinylated using BAR.

(A) A schematic depicting the design of the sE1E2.SZ-H445P plasmid and its transfection into glycoengineered CHO-S (geCHO) cells. The sE1E2.SZ-H445P plasmid was designed for secretion by replacing the transmembrane domains (TMD) with a synthetic scaffold, SynZip (SZ), which lacks human sequence homology. Furin was co-expressed to cleave the 6xArginine sites, facilitating native-like assembly of E1E2 by forming coiled-coil interaction between SZ1 and SZ2. The construct includes a key proline substitution at position H445 (H445P) in domain D of E2. This plasmid was co-expressed with human furin in two geCHO cell lines: CTRL (Sppl3 knockout) and CL1 (Sppl3, Mgat2, St3gal3/4/6, B3gnt2 knockout). (B) Stable clones of CTRL and CL1 were obtained expressing sE1E2.SZ-H445P. Two clones from each geCHO cell line were selected for biotinylation by antibody recognition (BAR) as depicted in the schematic. Cells were fixed, permeabilized, and stained with the primary HCV1 antibody targeting domain D of E2. A secondary anti-human horseradish peroxidase (HRP)-conjugated antibody was used to bind HCV1. Upon addition of biotin and hydrogen peroxide, proteins proximal to E1E2 were biotinylated. (C) BAR was performed on technical triplicates of each clone, with biotin added for 5 minutes. Biotinylation was confirmed by western blot with 25 μg of lysates loaded per sample. Streptavidin-HRP staining was used to visualize biotinylated proteins. Non-transfected parental cell lines served as negative controls.

To confirm sE1E2 secretion, we performed western blot analysis under reducing conditions for each clone, probing with an anti-E1 antibody. Both the supernatant and cell lysate were analyzed to assess protein secretion levels and intracellular protein accumulation, respectively (Supplementary Figure S2A). There was no significant difference in titer performance between clones from the two cell lines. However, western blot analysis revealed a broader sE1E2 band in the CTRL cell line compared to CL1, likely due to the more complex and heterogeneous glycans produced by CTRL. In contrast, the narrower band in CL1 is consistent with its production of simpler glycoforms. In addition, both cleaved forms of sE1E2 (processed by furin) and uncleaved forms of sE1E2 (not processed by furin) were observed in the supernatant and the lysate, indicating partial proteolytic processing. The cleaved sE1E2 appeared as either E1 (~30 kDa) or E2 (~65 kDa), depending on the antibody used for detection, while the uncleaved sE1E2 was detected at ~100 kDa. This pattern of both processed and unprocessed forms suggested that furin cleavage was occurring but not proceeding to completion under the experimental conditions.

To enhance furin-mediated cleavage of sE1E2, we tested several strategies, including codon-optimizing both E1E2 and furin for CHO cells, co-expressing them on the same plasmid, and testing different promoter combinations (Supplementary Figure S3). The codon optimization approach was designed to enhance E1E2 and furin expression levels and proper protein folding in the CHO cell system, while the single plasmid strategy aimed to ensure coordinated expression of both the enzyme and substrate. Additionally, the colocalization of both genes on one construct was expected to facilitate more efficient substrate-enzyme interactions within the cellular secretory pathway. Despite these attempts, uncleaved forms of sE1E2 persisted, suggesting either a furin functionality issue or inaccessibility of the furin cleavage site for efficient proteolytic processing. These results indicate that the incomplete processing may stem from inherent limitations in the experimental system or structural constraints that prevent optimal furin-substrate interactions.

BAR targets proximal proteins of sE1E2 produced in geCHO cells

We used BAR to detect and quantify PPIs in stable clones expressing sE1E2.SZ-H445P, given its ability for proximity labeling in cells in their native state. Other proximity labeling techniques, such as BioID (Kim & Roux, 2016; Pfeiffer et al., 2022; Samoudi et al., 2021; Sears et al., 2019) require fusion of the protein of interest with another enzyme (e.g., BirA) and/or lengthy labeling periods that may compromise protein functions. However, BAR preserves cell integrity and a protein’s native state. This approach requires only fixation and treatment with readily available antibodies, making it more accessible and less disruptive to cellular processes. BAR employs an HRP-conjugated antibody specific to the target protein to biotinylate nearby proteins, effectively capturing proximal proteins at a specific moment during harvesting. The technique’s ability to provide a temporal snapshot of protein interactions while maintaining cellular architecture makes it particularly suitable for studying secreted proteins like sE1E2 in their native cellular environment.

To perform BAR on the selected clones, cells were harvested during the mid-exponential phase, fixed, permeabilized, and labeled with HCV1 primary antibody targeting domain E of the E2 protein (Figure 1B), which is a highly conserved region sensitive to bNAbs responses (Toth, et. al., 2021). Following primary antibody binding, we applied a secondary anti-human HRP-conjugated antibody to bind to HCV1, establishing the enzymatic labeling system. Upon biotin addition, HRP transfers biotin to proteins in proximity to sE1E2, creating biotinylated protein complexes that can be subsequently isolated and identified. Western blot analysis with HRP-streptavidin confirmed successful biotinylation, showing substantially more labeled proteins at much higher stoichiometries in sE1E2-expressing clones compared to non-transfected controls (Figure 1C), validating the specificity and effectiveness of the labeling approach. We then isolated biotinylated samples using streptavidin-coated beads, performed trypsin digestion, and analyzed the results by liquid chromatography-tandem mass spectrometry (LC-MS/MS) to identify the complete repertoire of proteins in proximity to sE1E2 during expression and processing.

The BAR proteomics spectra were analyzed using MaxQuant software (Tyanova, Temu, & Cox, 2016), and the MS/MS spectra were searched against the Cricetulus griseus Uniprot protein sequence database. Proteomic data were analyzed using the Differential Enrichment analysis of Proteomics data (DEP) package in RStudio (Zhang et al., 2018) to process and analyze the biotinylated proteins, including data normalization, handling of missing values, and data visualization. We found the total biotinylated proteins in each sample showed consistent levels of protein detection across replicates, with the nonproducer cell lines containing fewer proteins (Figure 2A). However, a substantial number of biotinylated endogenous proteins were detected in the nonproducer cell line, albeit at lower stoichiometries (Figure 2A), consistent with prior BioID studies reporting high background signal and considerable overlap between control and experimental samples (Samoudi et al., 2021). While biotin is not synthesized natively in mammalian cells, it is found in foods and media, and it is a cofactor for carboxylases, which can cause false positive signals in streptavidin-based protein detection (Tytgat et al., 2015). Given this expected background, we accounted for nonspecific binding and endogenous biotinylation by including control cell lines not expressing the E1E2 to distinguish true interactions in comparison to a nonproducer control to filter out background signals. These steps ensured that only high-confidence interactors were retained in our final dataset for analysis.

Figure 2: BAR proteomic analysis of E1E2-expressing geCHO cell clones: differential protein expression and pathway enrichments.

(A) Pre-processing of the BAR data involved plotting total biotinylated proteins across all samples and their replicates to ensure consistent proteins across replicates. Nonproducer cell lines showed a lower number of biotinylated proteins compared to the producer cell lines, indicating that the BAR method successfully labeled numerous proximate proteins of E1E2. Replicate samples contained relatively consistent protein numbers within their respective groups. (B) Principal Component Analysis (PCA) of the normalized proteomic dataset demonstrated clear separation between nonproducer (parental) cell lines and E1E2 producer clones. PC1 captured 70% of the total variation, effectively distinguishing between the producer and nonproducer cell lines. (C) Welch’s t-test was performed to compare proteomic data from BAR between producer and nonproducer cell lines, using respective parental cell lines as controls. Volcano plots visualize the statistical analysis, revealing proteins that are differentially expressed in the producer cell lines (red dots) vs the parental cell lines (blue dots) with statistical significance (p-adjusted < 0.05 and Log₂(fold-change) > 0). (D) Biological pathway enrichment analysis was performed using STRING on the approximate 600 protein interactors that were either observed exclusively in the producer cell lines or differentially expressed, focusing on biological processes. The analysis revealed a comprehensive profile of cellular mechanisms, highlighting proteins involved in RNA processing, cell cycle regulation, and protein folding/transport/regulatory pathways.

To visualize the relative similarity of the samples, we performed principal component analysis (PCA) of the normalized data, which revealed clear separation between producer and nonproducer cell lines, with PC1 accounting for 70% of the variance (Figure 2B), demonstrating that sE1E2 expression creates a distinct proteomic signature. A preliminary analysis of the data (Supplementary Figure S2B) revealed proteins with higher relative abundances in producer cells. These distinct protein abundance patterns demonstrated that the BAR method successfully labels numerous proteins, presumably those in proximity to sE1E2 during its production and secretion pathway.

To prioritize candidate proteins that support sE1E2 production for further study, we first filtered for proteins that were biotinylated and detected in at least 2 of 3 producer cell lines and not in the nonproducer cell lines. This resulted in 914 and 874 unique interactors in the CTRL and CL1 producer cell lines, respectively. This captures proteins that appear exclusively in the presence of sE1E2, representing interactions of particular interest or cellular responses specific to viral protein expression. Second, for proteins that were detected in at least two-thirds of the replicates in the parental cell lines and in the producers, we performed Welch’s t-test comparing each producer cell line against its own control parental cell line (Figure 2C). Welch’s t-test identified 60 and 29 statistically significant interactors (Benjamini-Hochberg p-adjusted < 0.05 and Log₂(fold-change) > 0) in CTRL and CL1, respectively. This dual analytical approach ensured comprehensive detection of both exclusive protein associations and statistically significant abundance changes related to sE1E2 expression.

In total, we combined the two lists of proteins, and detected 974 and 903 unique potential protein interactors in CTRL and CL1 producer cell lines, respectively (Supplementary Table S1). This analysis provided an extensive repertoire of proteins potentially involved with sE1E2 across both cell line backgrounds. Notably, 604 potential interactors were shared across all sE1E2-producing cell lines, representing a core set of proteins that consistently associate with E1E2 production regardless of cellular background. Pathway enrichment analysis of these shared proteins revealed involvement in protein transport, RNA processing, cell cycle regulation, and quality control pathways (Figure 2D). These findings suggest that a conserved cellular machinery underlies sE1E2 production, potentially indicating essential host pathways co-opted during recombinant protein expression. This provides a valuable starting point for validation studies to uncover critical host factors that aid in improving sE1E2 production.

BAR and RNA-Seq analysis identifies potential protein interactors for improving sE1E2.SZ-H445P secretion

To refine our candidate gene list, we performed whole-transcriptome analysis to identify genes that showed BAR enrichment and significant correlation with secretion efficiency in RNA-Seq data. We conducted differential gene expression analysis using DESeq2 (Love, et al., 2014) and identified >2500 significant genes (p-adjusted < 0.05 and Log₂(fold-change) > 0) in CTL and CL1 producers individually (Figure 3A) when compared to their parental cell lines (nonproducers) (Supplementary Table S2). Gene set over-representation analysis revealed that upregulated genes were enriched in biological processes compared to the nonproducers primarily related to DNA replication, mRNA/DNA processing, mitotic cell cycle processes, and intracellular transport (Figure 3B). These results indicate that sE1E2 expression triggers broad changes in fundamental cellular processes related to protein production and trafficking.

Figure 3. Transcriptomic analysis of bulk RNA-Seq in geCHO cells expressing sE1E2.

(A) Bulk RNA-Seq was performed on geCHO cell clones expressing sE1E2, with respective parental cell lines serving as negative controls. Using the DESeq2 package in RStudio, we compared producer cell lines to their parental cell lines. The transcriptomic analysis revealed approximately >2,500 differentially expressed genes in CTRL and CL1 individually (red dots) compared to their respective parental cell lines (blue dots) (p-adjusted < 0.05 and Log₂(fold-change) > 0) as visualized in the volcano plots. (B) Gene Ontology enrichment analysis focused on biological processes across statistically significant genes in all producer cell lines. The analysis uncovered enriched pathways predominantly associated with cellular processes involving mitotic cell cycle process, mRNA/DNA processes, and intracellular transport mechanisms.

Among the approximate 600 shared PPIs identified through BAR analysis, >100 genes were both significant (p-adjusted < 0.05) and upregulated, overlapping between proteomic and transcriptomic datasets. These proteins with both increased interactions and mRNA abundance participate in various cellular processes, particular interest in protein synthesis and processing (Eif2b1/2b2/3j/4e,5, Utp2/4, Nip7, Farsa, Dnajb1, Spata5), RNA processing and regulation (Sf3a3/b2, Ncbp1, Rbm8a, Spats2, Dhx20/36, Safb), protein degradation and quality control (Psmc3/Psmd14, Psma5, Sugt1, Cops2/4, Ubap2, Ufd1, Cul4a, Bag6), intracellular transport and trafficking (Sec24b, Mia3, Arfip1/2, Osbp, Ap1m1, Ist1, Nup35/160), cell cycle and DNA metabolism (Rfc3/4/5, Msh2, Rad23b, Zw10, Ywhah, Plk1, Cdc16). The diversity of these cellular processes reflects the complex network of molecular changes observed in the cells we engineered to produce the sE1E2 heterodimer, so we used these for selection of candidates for improving sE1E2 secretion efficiency.

Overexpression of PPI proteins can increase sE1E2 secretion

Here, we focus specifically on proteins associated with cell cycle regulation, cell growth, proteostasis, and transport mechanisms, as these processes are directly relevant to protein production and secretion. To validate our findings, we selected 12 candidate interactors (p-adjusted < 0.05 and Log₂(fold-change) > 0.2) that exhibited strong statistical support and potential involvement in protein secretion pathways.

The 12 proteins selected for overexpression were AP1M1, CUL4A, DNAJB1, IST1, MIA3, PSMA5, PSMD14, SPATA5, SUGT1, UFD1, YWHAH, and ZW10. These proteins participate in transport, quality control, co-chaperone functions, and cellular regulation of replication and apoptosis. Each protein was overexpressed in the sE1E2-producing CL1(H445P:Furin) #21 cell line, which was engineered to provide optimal glycosylation (Kulakova et al., 2025). This targeted approach allowed us to focus experimental resources on the cell line most likely to demonstrate functional improvements while maintaining biological relevance for broader applications.

We transfected CL1(H445P:Furin) #21 cells in triplicate with the selected genes using Polyethylenimine (PEI) MAX for transient transfection, where RNA was extracted 48 hours post-transfection for cDNA synthesis and subsequent qPCR analysis. Gene expression was analyzed by qPCR with the empty vector control normalized to 1. Relative fold-change analysis confirmed successful overexpression of all target genes compared to the empty vector control (Figure 4A). This validation step was critical to ensure that observed effects on protein secretion could be attributed to successful gene overexpression rather than transfection variability. On day 4 post-transfection, cell viability was assessed via cell counting, and supernatants were collected for protein analysis via quantitative western blot. We observed a significant difference in viable cell count (cells/ml) (p < 0.05, two-sided paired t-test) in cells transfected with Ap1m1, Ist1, and Ywhah compared to the empty vector (Supplementary Figure S4A). However, we did not observe significant difference in cell viability percentage between the overexpression groups and the control (p > 0.05, two-sided paired t-test; Supplementary Figure S4B).

Figure 4. Validation of Candidate Genes Identified from Proteomic and Transcriptomic Datasets with Overexpression to Improve E1E2 Secretion.

(A) Polyethylenimine (PEI) transient transfection was used to overexpress candidate genes in CL1(H:F) #21 identified from our proteomic and transcriptomic analyses. RNA was extracted 2 days post-transfection, and qPCR was performed to verify overexpression using the 2^(-ΔΔCT) method with the empty vector control normalized to 1. All selected candidate genes demonstrated successful overexpression following transient transfection compared to the empty vector control. (B) Quantitative western blot analysis of supernatant collected on day 4 post-transfection was performed to assess total E1E2 secretion (encompassing both furin-cleaved and uncleaved versions). Using ImageJ for band quantification and normalizing to cell counts, we analyzed total E1E2 protein secretion (ng/cell). Cul4a showed statistically significant improvements in E1E2 secretion compared to the empty vector control (p-adjusted < 0.05, one-sided paired t-test), with fold changes of 1.30 relative to the empty vector control, which was set to 1.0. Ywhah trended toward significance (p-adjusted = 0.07). (C) Focusing on the two genes, Cul4a and Ywhah that were significant (p < 0.05 in B) for total sE1E2 secretion by one-sided paired t-test against the empty vector backbone, we also observed a significant increase in furin-cleaved sE1E2 secretion for both (ng/cell) (p < 0.05).

The sE1E2 secretion was analyzed by western blot, and band intensities were quantified using ImageJ to obtain quantitative measurements of secretion efficiency. Proteins were analyzed on separate western blot gels, each including its own empty vector control to account for gel-to-gel variation. We utilized a one-sided paired t-test to analyze the selected genes, based on our proteomic and transcriptomic data hypothesizing that the tested genes elevate sE1E2 levels compared to the empty vector control. Resulting p-values were adjusted using the Benjamini–Hochberg method to control for multiple hypotheses with an adjusted p-value < 0.05 (Supplementary Table S3A). Based on viable cell counts and protein secretion analysis, CUL4A overexpression significantly increased the total secretion yield of sE1E2 (ng/cell), encompassing both furin-cleaved and uncleaved forms, compared to the empty vector control (adjusted p < 0.05; Figure 4B; Supplementary Figure S4C), with approximately a 1.3-fold increase relative to the empty vector control (set to 1.0). In contrast, YWHAH overexpression showed a trend toward significance (adjusted p = 0.07).

While total secreted sE1E2 levels are important, proteins that enhance secretion must also maintain proper furin cleavage of sE1E2. Therefore, we further assessed whether genes that increased total sE1E2 secretion also significantly increased properly cleaved sE1E2. Since CUL4A and YWHAH showed significant p-values (p < 0.05) in the one-sided paired t-test, we focused on these two genes to evaluate their effect on furin-cleaved sE1E2 secretion. Using a one-sided paired t-test, overexpression of both genes significantly increased cleaved sE1E2 secretion yield (ng/cell) compared to the empty vector control (p < 0.05), with each showing an approximately 1.3-fold increase relative to the empty vector control (set to 1.0) (Figure 4C; Supplementary Figure S4D; Supplementary Table S3B).

Based on the viable cell count, YWHAH-overexpressing cells exhibited reduced viable cell count but no changes in cell viability suggesting that the increase in protein secretion reflect enhanced per-cell yield (ng/cell). Unlike YWHAH, CUL4A-overexpressing cells maintained similar cell counts and viability compared to the control. Nonetheless, both genes showed increased per-cell yield (ng/cell), indicating that enhanced secretion was primarily due to improved secretion efficiency per cell. These two proteins were only detected in cells that expressed sE1E2 in the BAR proteomic data and not in the control cell lines, indicating their specificity to E1E2-expressing cells. The involvement of CUL4A and YWHAH in cell cycle regulation and ubiquitination suggests that enhanced quality control checkpoints may benefit CHO cells expressing HCV proteins. Among all tested candidates, CUL4A and YWHAH emerged as the most promising for improving the secretion of both total sE1E2 protein and correctly folded.

Discussion

The development of an sE1E2 subunit vaccine, engineered by replacing the TMD with a synthetic scaffold, SZ, has demonstrated promising scalability and purification potential. Although a secreted form of E1E2 could be produced in geCHO cells, the yields were insufficient for viable vaccine production. To address this limitation, we employed the BAR method to systematically analyze PPIs of sE1E2, aiming to identify and overcome potential bottlenecks in secretion and yield, leading to the identification of CUL4A and YWHAH as productivity-enhancing targets for further cell line development. However, this proof-of-concept study presented further insights, implications, and future work to pursue.

First, we focused on PPIs involved during the mid-exponential phase to ensure active growth and robust protein synthesis, maximizing the capture of E1E2 interacting proteins before secretion. Given that titers increase the most as cell growth slows in the stationary phase (González-Hernández & Perré, 2024; Templeton et al., 2013), future studies can optimize protocols to assay interactions during the production phase to gain further insights into the host cell machinery driving production. Other productivity-enhancing conditions can further expand insights by using temperature shifts (Bollati-Fogolín et al., 2005; Hossler et al., 2009; Yoon et al., 2003), thereby providing a more dynamic view of host-cell proteins that enhance protein production.

Second, we focused on overexpressing targets from a subset of pathways identified through our proteomic and transcriptomic analysis. The BAR proteomic PPI enrichment analysis (Figure 2D) revealed several cellular pathways that may support efficient E1E2 secretion, including vesicle-mediated transport, protein localization to organelles, RNA processing, and regulation of cell cycle. In addition, enrichment analysis revealed significant over-representation of pathway terms related to response to ER stress, ERAD pathway, and regulation of proteolysis—processes associated with the quality control and degradation of misfolded proteins within the ER. Both the proteomic (Figure 2D) and transcriptomic analyses (Figure 3B) showed convergence of enrichment in regulation of cell cycle and DNA/RNA processes, indicating substantial cellular stress associated with recombinant protein production. This cellular burden was evident when producer cell lines exhibited slower growth during scale-up compared to non-producer controls in reaching mid-exponential phase. Particularly, the upregulation of RNA processes via RNA splicing/transport, DNA processes via DNA replication/DNA repair, and cell cycle process in the transcriptomic data suggests that producer cells systematically enhance their transcriptional and translational machinery to meet E1E2 production demands while reallocating cellular resources from growth processes to protein production and stress management systems.

Third, based on these comprehensive omics findings, we strategically focused our engineering efforts on overexpressing proteins directly involved in or closely associated with the secretory pathway, particularly those involved in transport and quality control, as well as proteins involved in cell cycle regulation, especially those that could mitigate the observed growth defects in producer cell lines. This targeted approach is well-supported by previous successful cellular engineering studies that have demonstrated significant improvements in recombinant protein production through secretory pathway enhancement. For instance, overexpression of the transcription factor ATF4, involved in the unfolded protein response, can enhance IgG production (Haredy et al., 2013) and improve the production of recombinant human antithrombin III by twofold (Ohya et al., 2008). Building upon this established framework and guided by our integrated BAR proteomic and RNA-Seq analysis, we identified two promising candidate proteins, CUL4A and YWHAH, that demonstrated significant potential for enhancing sE1E2 secretion when overexpressed in our experimental system. Although viable cell counts were lower for YWHAH compared to the empty vector control, no difference in cell viability percentage was observed between the two groups. When normalizing for cell count, we found YWHAH overexpression significantly increased E1E2 secretion on a per-cell basis relative to the control. In contrast, overexpression of AP1M1 and IST1 did not improve per-cell secretion despite also exhibiting lower viable cell counts compared to the control.

CUL4A (Cullin 4A), a positive correlator with increased sE1E2 secretion, is a member of the Cullin-RING E3 ligases (CRLs) family that forms the CRL4Cdt2 complex by interacting with Cdt2 (DNA replication licensing factor), a protein crucial for cell proliferation with the CUL4A/B scaffold (Dar et al., 2014). This complex ubiquitinates proteins such as p21 (cyclin-dependent kinase inhibitor), Set8 (histone methyltransferase), and Cdt1 (chromatin licensing and DNA replication factor 1) for degradation during S phase, ensuring cell cycle progression, DNA damage repair, prevention of re-replication, and proper replication initiation (Abbas et al., 2008, 2010; Senga et al., 2006; Sharma & Nag, 2014). The 14-3-3ε and 14-3-3γ proteins positively regulate CRL4CDt2, where it interacts with phosphorylated Cdt2 during the S phase, protecting it from degradation, and in the absence of 14-3-3γ, destabilized Cdt2 can lead to increased levels of Set8, resulting in G2/M arrest (Dar et al., 2014). In addition to its cell cycle functions, CUL4A undergoes neddylation, a modification that activates its ubiquitin ligase activity through binding to NEDD8, a ubiquitin-like protein involved in protein modification (Osaka et al., 1998; Sharma & Nag, 2014). Neddylation activates Cullins by inducing conformational changes that enhance the flexibility of the RING–E2 ubiquitin-transfer module (Duda et al., 2008). This increased flexibility promotes efficient substrate polyubiquitination and subsequent proteasomal degradation (Duda et al., 2008).

Our second identified enhancer of sE1E2 secretion, specifically for furin cleaved sE1E2 is YWHAH (14-3-3η), a member of the 14-3-3 protein family. These proteins bind phosphoserine/phosphothreonine residues and regulate diverse cellular processes such as cell cycle progression, signal transduction, apoptosis, and vesicular and protein trafficking (Darling et al., 2005; Muslin et al., 1996; Rubio et al., 2004). Specifically, the 14-3-3 family promote cell survival by inhibiting pro-apoptotic proteins such as Bad (BCL2 associated agonist of cell death), Bax (BCL2 associated X, apoptosis regulator), and ASK1 (Apoptosis signal-regulating kinase 1) (Nomura et al., 2003; Zha et al., 1996; Zhang et al., 1999). Previous studies engineered cells to improve cell viability and titer by knocking down pro-apoptosis genes such as BAX and BAK (Lim et al., 2006), which provides a mechanistic explanation for our observed increase in protein production, as YWHAH a member of the 14-3-3 family inhibits these same pro-apoptotic genes. The 14-3-3 proteins have diverse cellular localizations and are found in the plasma membrane, ER, Golgi apparatus, nucleus, and mitochondria (Kongsamut & Eishingdrelo, 2023), and their presence in the ER and Golgi apparatus may directly influence the secretory pathway relevant to E1E2 processing. For example, 14-3-3 proteins interact with phosphorylated Nedd4-2, a HECT E3 ubiquitin ligase, inducing structural changes that may regulate substrate ubiquitination (Pohl et al., 2021).

Collectively, our findings suggest that CUL4A potentially positively regulates the cell cycle and proper replication initiation, while YWHAH likely positively influences cell viability by inhibiting pro-apoptotic proteins. Notably, both proteins belong to families that engage with the Nedd E3 ubiquitin ligase to promote ubiquitination. Together, these proteins create a more favorable cellular environment for sE1E2 production and secretion by addressing different aspects of cellular function that can limit recombinant protein production. Additionally, simultaneously overexpressing both proteins would be of interest to determine if they further increase protein secretion compared to individual overexpression, since each protein targets a different pathway and their combined effects could potentially be synergistic.

Future work would benefit from systematic screens that test which interactors in the enriched pathways in the proteomic and transcriptomic data are positive or negative regulators through experiments using CRISPRa/i, siRNA, knockouts, or overexpression. It would be beneficial to test whether silencing some of the proteins we overexpressed that failed to increase sE1E2 secretion. These include PSMD14, PSMA5, UFD1, DNAJB1/HSP40, and SUGT1 which are involved in proteasome function, protein folding, and quality control. Similarly, overexpression of transport-related proteins such as AP1M1, IST1, MIA3/TANGO1, and ZW10 did not enhance E1E2 secretion, suggesting that overexpressing trafficking components could have neutral or even negative effects. While overexpression of these proteins did not enhance secretion of E1E2 in our system, previous studies have shown that co-expression of the co-chaperone SUGT1 using polycistronic expression enhances protein production (Snead et al., 2022). The failure of overexpression of the validated proteins to enhance E1E2 secretion in our system could be due to several factors: the target protein may already be present at saturating levels; excessive expression might lead to pathway imbalance or misregulation; or the protein may require cofactors or binding partners that were not co-expressed at appropriate levels.

While identifying CUL4A and YWHAH as positive regulators of protein secretion is valuable, further optimization of other cellular pathways is likely needed to enhance product quality, specifically by eliminating the secretion of the undesired uncleaved form of sE1E2 in the supernatant. Attempts to purify the correctly folded version for testing were unsuccessful due to the low yield of secreted E1E2. Future studies should investigate how sE1E2 is assembled intracellularly and the accessibility of the furin cleavage site in CHO cellular environments to improve cleavage efficiency. Enhancing cleavage efficiency would likely increase the proportion of correctly processed sE1E2, facilitating downstream purification and improving overall vaccine yield.

Finally, the BAR method provides a valuable approach for identifying CHO cell proteins that support sE1E2 secretion. However, since this method labels all proteins in proximity, it cannot distinguish between direct and indirect interactions and may include nonspecific interaction that can lead to false positives and negatives. To mitigate these limitations, we (1) included appropriate negative controls, (2) performed triplicate experiments to filter out noise, and (3) incorporated RNA-Seq as a mostly independent data type. This helped us to increase confidence in the identified PPIs. Future work could account for predicted PPIs, e.g., using structural biology approaches.

Despite these challenges, the BAR method remains a valuable and user-friendly technique for studying both intracellular and secreted proteins, offering critical insights into the CHO secretory machinery. By identifying key components and mechanisms involved in protein secretion, it enhances our understanding of and provides a pathway to optimizing protein production. This approach demonstrates the potential of applying BAR to improve the secretion of other difficult-to-express proteins. When combined with other optimization strategies—such as plasmid design with enhanced promoters (Sou et al., 2023), site-specific integration using recombinase-mediated cassette exchange to prevent potential disruption of host genes (Shin et al., 2022), and bioprocess optimization (including fed-batch cultivation, pH control, aeration adjustment, media formulation, and temperature modulation) (Yang, 2019)—such cellular and process engineering approaches together create a comprehensive optimization regime that substantially enhances protein production.

Materials and Methods

Cell line engineering and protein expression

CHO-S cells (Life Technologies, Carlsbad, California, USA) were glycoengineered to create two variants: CTRL, which produces more complex glycans resembling the wild type, and CL1, which produces simpler glycoforms. The geCHO cells were generated using the gene editing technique CRISPR/Cas9, specifically using “CRISPy” to knock out Sppl3 in CTRL cell line and Sppl3, Mgat2, St3gal3/4/5, and B3gnt2 in CL1. The single cell cloning was done as described previously (Amann et al., 2019).

Cell culture

The geCHO cells were grown in 125 ml Erlenmeyer shaker flasks with baffled bottom (Fisher Scientific, USA) at 130 rpm in 37°C at 5% CO2 with maintenance media containing CD CHO Medium (Gibco, USA) supplemented with 8 mM L-Glutamine (Gibco) and 1x anti-clumping agent (Gibco). The geCHO cells expressing E1E2 and furin were grown in media containing 1 mg/ml Geneticin Selective Antibiotic (G418 sulfate; Gibco) and 0.75 mg/ml Hygromycin B (Gibco) for selection.

CellTiter-Blue Assay-Kill Curve

CellTiter-Blue assay was used to measure the viability of CHO-S cell lines to determine the concentration of Geneticin Selective Antibiotic (Gibco) and Hygromycin B (Gibco) to be used for selection after transfection. The cells were grown in 125 ml shake flasks in maintenance media at 5% CO2 and were > 95% viable before the experiment was conducted. Cells were seeded at 0.2 × 10⁶ cells/mL in non-treated 12-well plates (Greiner Bio-One, Kremsmunster, Austria) and incubated with different concentrations (0–1 mg/mL final concentration) of G418 or Hygromycin B at a final volume of 1 ml. Each concentration was measured in technical duplicates. The assay was conducted for 10 days and the media was changed every 2 days. To measure viability and cell count, 100 μL was taken every other day from each well and mixed with 20 μL of the CellTiter-Blue reagent in a 96 well plate (Sarstedt, Numbrecht, Germany). The reaction was incubated at 37°C for 2 hours. Then the plate was read using the BioTek Synergy MX microplate reader (Agilent, Santa Clara, CA, USA) with excitation wavelength at 530 nm and emission wavelength at 590 nm. The plate was shaken at medium speed for 1 minute before reading.

Plasmid Design

The sE1E2.SZ plasmid was generated as described previously (Metcalf et al., 2023) with a mutation in Domain D in E2 forming sE1E2.SZ-H445P (Pierce et al., 2020), and the furin plasmid was generated as described previously (Guest et al., 2021). The furin plasmid was redesigned with a new backbone, pcDNA4-TO-Hygromycin-sfGFP-MAP purchased from Addgene. It contains a CMV promoter expressing the gene for furin of 2385 bp with a bGH poly(A) signal, an ampicillin resistance marker and a hygromycin resistance marker for post-transfection selection.

USER Cloning plasmid design

The E1E2+Furin expression vectors were constructed using the pcDNA3.1(+) backbone with uracil-containing primers (IDT DNA, USA) and Phusion Hot Start Flex polymerase (New England Biolabs, Ipswich, MA, USA). The E1E2 and furin sequences were codon-optimized for CHO cells (IDT DNA, USA). Three different plasmid constructs were designed: (1) CMV-E1E2-BGHpA-EF1α-Furin-BGHpA, (2) EF1α-E1E2-BGHpA-CMV-Furin-BGHpA, and (3) CMV-Furin-BGHpA-CMV-E1E2-BGHpA. All constructs included the selection marker SV40-NeoR-SV40pA. CTRL or CL1 CHO-S suspension cells were transfected with the E1E2+Furin plasmids using PEI-based transient transfection and cultured for 6–7 days in 6-well plates. Culture supernatants were collected daily for western blot analysis.

Stable Transfection

The geCHO cells were initially transfected with the furin plasmid and cloned. The cloned furin cell lines were then transfected with sE1E2.SZ-H445P plasmid and cloned again. At 24 hours before transfection, the cells were passaged at 0.8–1.0 × 10⁶ cells/mL with CD CHO Medium (Gibco) supplemented with 8 mM L-Glutamine (Gibco). On the day of transfection, the cells were filtered with a 40 μm nylon mesh cell strainer (Falcon) before cell count. Twenty-five million cells were harvested, which was enough to transfect one negative control and two samples with plasmid. Cells were spun down at 200g for 4 minutes and washed once with warm PBS. Cells were resuspended in Buffer R from the Neon Transfection System 100 μL Kit (Invitrogen) with 12 μg of furin plasmid. Subsequently, DNA/cell suspension containing 5 × 10⁶ cells for each transfection were transfected using the Neon Transfection System with the following parameters: Voltage: 1550 V, pulse width: 10 ms, pulse number: 2 pulses. The transfected cells were transferred to a 6-well plate (GenClone; Genesee Scientific, El Cajon, CA, USA) with pre-warmed mixture media (2 mL/well). The plate(s) were placed on a shaker at 90 rpm at 37°C with 5% CO2. The next day, the viability of the cells were measured and the shaker speed was increased to 120 rpm. On day 2, the cells was expanded to a shake flask in media containing Hygromycin B. When the cells expanded, cells were passaged at 0.2 × 10⁶ cells/mL. The total cell count and viability were measured every other day, and media was changed every 3 days until the cells recovered and had > 95% viability. Afterward, cells were harvested for western blot for confirmation. The stable furin polyclonal cells were then cloned as described and the stable furin clones were transfected with 12 μg of sE1E2.SZ-H445P plasmid to generate stable clones expressing sE1E2.SZ-H445P and furin.

Cloning stable polyclonal cell lines

To form a stable cell line, cloning was done for each transfection. Cloning media contained CD CHO Medium (Gibco), 80% Ex-Cell CHO Cloning Medium (Sigma-Aldrich, St Louis, MO, USA), 8 mM L-Glutamine (Gibco), 1× ClonalCell-CHO ACF Supplement (Stemcell Technologies, Vancouver, Canada), 1 mg/mL Geneticin Selective Antibiotic (G418 sulfate; Gibco) and 0.75 mg/mL Hygromycin B (Gibco). Cells needed to be > 95% viable to be used for cloning. Cells were serially diluted at 1:10 dilution in media with 1 mg/mL of G418 and 0.75 mg/mL Hygromycin B containing no anti-clumping until the cells were at 1,000 cells/mL in 15 mL falcon tubes at a volume of 10 mL. Non-treated 96-well plates for suspension (N=3) (Genesee Scientific) were filled with 180 μL of cloning media into each well except column 1. Column 1 was plated with 200 μL of the 1,000 cells/mL suspension. Then the cells were serially diluted down to 10 cells (column 3) by taking 20 μL from previous wells. Onward, 20 μL was taken from the 10 cells well (column 3). This gave 1 cell/well from column 4–12. The plates were incubated at 37°C in 5% CO2 with no shaking. After 2 weeks, the plates were checked for single clones under microscopy. After 3 weeks, cells were expanded to non-coated 24-well plates (Genesee Scientific) prefilled with 400 μL of maintenance media containing antibiotics with no shaking. Cells were grown for a week before transferring to a non-coated 12-well plate (Genesee Scientific) prefilled with 600 μL of maintenance media containing antibiotics with shaking. After ~1 week or when cells expanded, half of the cells were taken for dot blot while the other half was expanded to 3 mL in a 6-well non-coated plate (Genesee Scientific) shaken at 130 rpm. Viability was performed on the cells once they grew and were expanded to 125 mL shake flasks.

Dot blot for screening clones

To screen the clones in 12-well plates, dot blot was used to ensure furin expression in lysate and secretion of E1E2 in the supernatant of these clones before further expansion. Cells and supernatants were taken in 12-well plates. The cells were lysed using 1× RIPA buffer (Thermo Fisher Scientific, USA) and 1× cOmplete Protease Inhibitor Cocktail (Roche, Basel, Switzerland) in distilled water. The 0.2 μm Nitrocellulose membrane (BioRad) was loaded with 10 μL of lysate or supernatant. The membrane(s) were dried for ~1 hour, and then blocked for 1 hour in the Intercept Blocking Buffer (LI-COR) in a rocker at room temperature. Primary antibodies used were: 1:2,000 Furin Monoclonal antibody mouse (67481–1-Ig; Proteintech, Rosemont, IL, USA); 1:500 E2 (HCV1); or 1:400 HepC E1 Mouse Monoclonal IgG 100 μg/mL (sc-65459, Santa Cruz Biotechnology, Dallas, TX, CA). The primary antibody was diluted in Intercept antibody diluent (LI-COR, Lincoln, NE, USA) and stained overnight in the rocker at 4°C. The membrane was washed three times with 0.1% TBST for 10 minutes each. Secondary antibody was diluted in the Intercept Dilution Buffer (LI-COR): 1:15,000 IRDye 800CW Goat anti-Human IgG Secondary Antibody (LI-COR) or 1:15,000 IRDye 680RD Goat anti-Mouse IgG Secondary Antibody. The membrane was stained with the secondary antibody for 1 hour at room temperature in the rocker and was washed with 0.1% TBST 3× for 10 minutes each. The membrane was imaged using the LI-COR 9120 Odyssey Infrared Imaging System (LI-COR).

Western blot for furin and E1E2 transfection

Western blot was performed to analyze protein secretion of E1E2 using a precast polyacrylamide gel, 4–15% Mini-Protean TGX Gels (Bio-Rad, Hercules, CA, USA). The samples were run under reducing conditions by mixing with loading dye (4x Laemli + 10% β-Mercaptoethanol) (Bio-Rad) and incubating at 95°C for 5 minutes. The gel was run on a Mini-PROTEAN Tetra Vertical Electrophoresis Cell (Bio-Rad) at 160V for 40 minutes with the Precision Plus Protein standards ladder (Bio-Rad). The gel was transblotted onto 0.2 μm nitrocellulose membranes (Bio-Rad) with the Trans-Blot Turbo Transfer System (Bio-Rad). Primary antibodies used were: 1:2,000 Furin mouse Monoclonal antibody (67481–1-Ig; Proteintech); 1:1000 Anti-HCV E2 mouse, clone AP33 25 MG (MABF2820; Sigma-Aldrich); or 1:400 HepC E1 Mouse Monoclonal IgG 100 μg/mL (sc-65459, Santa Cruz Biotechnology). The primary antibody was diluted in Intercept antibody diluent (LI-COR) and stained overnight in the rocker at 4°C. Secondary antibody was diluted in the Intercept Dilution Buffer (LI-COR): 1:15,000 IRDye 800CW Goat anti-Human IgG Secondary Antibody (LI-COR) or 1:15,000 IRDye 680RD Goat anti-Mouse IgG Secondary Antibody. The membrane was stained with the secondary antibody for 1 hour at room temperature in the rocker and was washed with 0.1% TBST 3× for 10 minutes each. The membranes were washed 3x at 10 minutes with 0.1% TBST and imaged using the LI-COR 9120 Odyssey Infrared Imaging System (LI-COR). Overexpression of validation genes in CL1(H:F) #21 and their effect on E1E2 secretion was determined by western blot quantification using ImageJ protocol by Hossein Davarinejad.

For staining of intracellular biotinylated proteins, 20 μg of total protein from labeled cells was loaded and transblotted. The membrane was blocked by 3% BSA in 0.1% TBST for 1 hour and probed with HRP-conjugated streptavidin (Abcam, Cambridge, UK) diluted in blocking buffer at 1:2000 for 40 minutes. For visualizing the proteins’ bands, the Clarity Western ECL Substrate (Bio-Rad) was used.

Quantify PPIs of E1E2 in geCHO cells using the BAR method

BAR was utilized to target sE1E2 in 2 clones from each geCHO cell lines expressing sE1E2 and identify its PPIs in each clone. To do this, the geCHO clones expressing E1E2 with the non-producing geCHO cells as control, each done in technical triplicates were harvested at the mid-exponential phase, fixed with 4% PFA in PBS (Thermo Fisher Scientific), and permeabilized with 0.4% PBST. Peroxidases were inactivated with 0.4% H₂O₂ (Fisher Scientific) and cells were blocked with 5% goat serum (Gibco) in 1% BSA in PBST for 1 hr. The cells were then incubated with E2 (HCV1, human) in blocking buffers overnight with rotation at 4°C. Next day, the samples were incubated with Goat anti-human IgG HRP (Invitrogen) for 1 hour at room temperature with rotation. Proximal protein biotinylation occurred for 5 minutes with treatment of H₂O₂ and tyramide-biotin using the TSA Biotin Reagent Pack (SAT700001EA; Akoya Biosciences, Marlborough, MA, USA), resulting in tyramide-biotin radicalization and deposition onto proximal proteins. The reaction was stopped with 0.5 M sodium ascorbate in PBS (Sigma-Aldrich). Lysate was extracted using 1.5% SDS (G Biosciences, St Louis, MO) and 1% Sodium Deoxycholate (bioWorld, Dublin, OH, USA) in 0.1% PBST and heated at 99°C for 1 hour with mild shaking. The samples were spun down at max speed for 5 minutes, and supernatant was taken for quantification using BCA Protein Assay (Lamda Biotech, Inc., St Louis, MO, USA). The samples were stored at −80°C until being sent to Sanford Burnham Prebys Proteomic Core (San Diego, CA, USA) for LC-MS/MS.

LC/MS-MS on biotinylated samples

The BAR samples were sent to Sanford Burnham Prebys Proteomic Core for LC-MS/MS analysis. Biotinylated proteins were affinity-purified using the Bravo AssayMap platform (Agilent) with AssayMap streptavidin cartridges (Agilent). Briefly, cartridges were first primed with 50 mM ammonium bicarbonate, and then proteins were slowly loaded onto the streptavidin cartridge. Background contamination was removed with 8M urea, 50 mM ammonium bicarbonate. Finally, cartridges were washed with Rapid digestion buffer (Promega, Rapid digestion buffer kit) and on-cartridge digestion of the bound proteins were performed using mass spectrometry-grade Trypsin/Lys-C Rapid digestion enzyme (Promega, Madison, WI) at 70°C for 1 hour. The resulting peptides were desalted on the Bravo platform using AssayMap C18 cartridges. Organic solvents were removed using a SpeedVac concentrator prior to LC-MS/MS analysis. The dried peptides were reconstituted in 2% acetonitrile with 0.1% formic acid and quantified using a NanoDropTM spectrophotometer at A280 nm (Thermo Fisher Scientific). Samples were analyzed via LC-MS/MS using a Proxeon EASY nanoLC system (Thermo Fisher Scientific) coupled with a Q-Exactive Plus mass spectrometer (Thermo Fisher Scientific).

Peptide separation was conducted on an analytical C18 Aurora column (75μm x 250 mm, 1.6μm particles; IonOpticks) at a flow rate of 300 nL/min. The 75-minute gradient applied was: 2% to 6% B in 1 minute, 6% to 23% B in 45 minutes, 23% to 34% B in 28 minutes, and 34% to 48% B in 1 minute (A: 0.1% FA; B: 80% ACN with 0.1% FA). The mass spectrometer was operated in positive data-dependent acquisition mode. MS1 spectra were measured in the Orbitrap in a mass-to-charge (m/z) of 375 – 1500 with a resolution of 60,000. Automatic gain control target was set to 4 × 10^5 with a maximum injection time of 50 ms. The instrument was set to run in top speed mode with 1-second cycles for the survey and the MS/MS scans. After a survey scan, the most abundant precursors (with charge state between +2 and +7) were isolated in the quadrupole with an isolation window of 0.7 m/z and fragmented with HCD at 30% normalized collision energy. Fragmented precursors were detected in the ion trap as rapid scan mode with automatic gain control target set to 1 × 10⁴ and a maximum injection time set at 35 ms. The dynamic exclusion was set to 20 seconds with a 10 ppm mass tolerance around the precursor.

Mass spectra were analyzed using MaxQuant software (Tyanova, Temu, & Cox, 2016), version 1.5.5.1. MS/MS spectra were searched against the Cricetulus griseus Uniprot and trEMBL protein sequence database (downloaded in September 2024) and GPM cRAP sequences (common protein contaminants). Precursor mass tolerance was set to 20 ppm for the initial search, which included mass recalibration, and 4.5 ppm for the main search. Product ions were searched with a mass tolerance of 0.5 Da. The maximum precursor ion charge state for the search was set to 7. The enzyme specificity was set to trypsin, allowing up to two missed cleavages. The target-decoy-based false discovery rate (FDR) filter for spectrum and protein identification was set to 1%.

BAR analysis

The BAR proteomic mass spectrometry dataset was preprocessed and analyzed using the DEP package in R Studio to identify differentially expressed proteins. Proteins were assigned unique names, and samples were filtered to remove proteins with excessive missing values. Specifically, proteins were retained only if they were identified in at least 2 out of 3 replicates for any given condition. To address variations in protein abundance and improve data quality, we normalized the samples using variance stabilizing normalization (VSN) (Huber et al., 2002; Karp et al., 2010), which reduces the dependence of variance on mean protein abundance levels. VSN transforms the data such that variance becomes approximately constant across all levels of protein abundance, making comparisons across samples more reliable and improving the performance of statistical analyses that assume equal variance. This normalization helps ensure that downstream interpretation is not dominated by highly expressed proteins alone, allowing for more balanced detection of both abundant and low abundance interacting partners. Following quality control assessment, missing values in the proteomics dataset were imputed using a left-shifted Gaussian distribution, based on the assumption that the missing values were not randomly missing (MNAR) but instead reflect low-abundance proteins that were below the detection limit of the instrument (Pereira et al., 2024). This imputation method fills in missing values with small numbers drawn from a Gaussian distribution that is shifted toward the lower end of the intensity range. By doing so, it mimics the cells’ condition that proteins not detected in some samples are likely present at low levels, rather than being completely absent from the cellular environment. This approach helps preserve the biological variability in the dataset while allowing statistical analyses, such as PCA or differential expression, to be performed without distortion caused by missing values. The imputation strategy ensures that subsequent comparative analyses between producer and nonproducer cell lines accurately reflect the true biological differences in protein abundance rather than technical limitations of the detection method. The data were then visualized using Principal Component Analysis (PCA) of the normalized and imputed data. Proteins were filtered using the following criteria: detected in two-thirds of the replicates for each producer clone and present in < 2 replicates in control or present in 2 out 3 replicates in control but having a Log₂(fold-change) > 0 and p-adjusted < 0.05. These were marked as potential hit proteins. Welch’s t-test was performed comparing producer cell lines versus non-producer cell lines to identify proteins that were significantly expressed with p-adjusted < 0.05 and Log₂(fold-change) > 0 in the producer cell lines. The STRING database (https://string-db.org/) (Szklarczyk, et al., 2023) was used to perform functional enrichment analysis on proteins observed exclusively in the producer cell lines or differentially expressed. This analysis focused on biological pathways, aiming to understand the roles and interactions of these proteins within cellular processes.

RNA-Seq prep and analysis

Each of the geCHO cell clones expressing sE1E2 along with its parental cell line, each in triplicates, were harvested during the same time as the sample used for BAR. Total RNA was extracted from each sample using the RNeasy Kit (Qiagen, Hilden, Germany) following the manufacturer’s instructions and quantified using nanodrop at A260 nm. Samples were sent to UCSD IGM Genomics Center for mRNA library preparation and sequencing. RNA quality was analyzed using the Agilent Tapestation 4200, and only samples with an RNA Integrity Number (RIN) exceeding 8.0 were selected for library construction. Libraries were prepared using the Illumina Stranded mRNA Sample Prep Kit with Illumina RNA UD Indexes (Illumina, San Diego, CA), following the manufacturer’s guidelines. The finalized libraries were multiplexed and sequenced on an Illumina NovaSeq X Plus platform, generating 150 base pair (bp) paired-end reads (PE150) with an approximate depth of 25 million reads per sample. The samples were demultiplexd with bcl2fastq v2.20 Conversion Software. Adapter sequence trimming and quality control (FastQC) of the sequences were performed using trimgalore. Reads were mapped to the CHO GCF_003668045.3 CriGri-PICRH-1.0 genome, aligned and quantified with kallisto. The samples are analyzed using R studio using the DESeq2 package (Love et al., 2014) comparing each cell line against its parental cell line (nonproducer). Differential expressed genes are identified as p-adjusted < 0.05 and Log₂(fold-change) > 0.

Over-representation Analysis

We performed the Gene Ontology (GO) enrichment of the transcriptomic differentially expressed genes in producer cell lines using the enrichGO function from the clusterProfiler package in R with genome annotation provided by the org.Hs.eg.db package (version 3.18.0), using all genes with expression counts ≥10 as the background to reflect the detectable transcriptome.

Generating plasmids for overexpression

The Ap1m1, Dnajb1, Cul4a, Ist1, Mia3, Psma5, Psmd14, Spata5, Sugt1, Ufd1, Ywhah, Zw10, and the empty vector (pcDNA3.1(+)) plasmids were designed from Chinese hamster (Cricetulus griseus) and purchased from GenScript (Piscataway, NJ, USA). The plasmids were transformed in DH5alpha Competent Cells (Invitrogen, USA) and expanded in Miller’s LB Broth (Corning Inc., Corning, NY, USA) supplemented with 100 μg/ml of ampicillin (Sigma-Aldrich). The plasmids were extracted using QIAprep Spin Miniprep Kit (Qiagen).

Transient Transfection

Polyethylenimine (PEI) Max transient transfection was used for overexpression of the validation genes. PEI MAX-Transfection Grade Linear Polyethylene Hydrochloride (MW 40,000) (Kyfora Bio, Warrington, PA, USA) was dissolved at 1 mg/ml in distilled water and pH was adjusted to 7.0 using 1M of NaOH. The dissolved PEI was sterile filtered in 0.22 μm steritop (MilliporeSigma, Burlington, MA, USA). A day before transfection, cells were passaged at 0.5–0.8 × 10⁶ cells/ml in media containing no anti-clumping. On the day of transfection, cell counts were done and were at least > 95% viable for transfection. Cells were harvested and resuspended at 1.0 × 10⁶ cells/ml in media containing no anti-clumping. Cells were transfected with 1 μg plasmid and 7 μg of PEI Max solution per ml of cell culture (5% of total culture volume). The DNA/PEI mixture was incubated in OptiPRO SFM (Life Technologies) for 20 minutes at room temperature before adding to the cells in 6-well non-treated, polystyrene, flat bottom wells plate (Genesee Scientific). The cells were shaken at 130 RPM at 37°C in 5% of CO2. The day post transfection, 1× of anticlumping agent (Gibco) and 0.6 mM valproic acid sodium salt (Sigma-Aldrich) was added to each well. Some cells were harvested on day 2 post transfection for RNA extraction. Viability was performed on day 4 post transfection and supernatants were harvested for western blot analysis.

qPCR to confirm gene overexpression

RNA extraction was done following the RNeasy Mini kit protocol (Qiagen). RNA was converted to cDNA following the SuperScript II RT protocol (Thermo Fisher Scientific) with 1 μg of RNA and RNAseOUT Recombinant Ribonuclease Inhibitor (Life Technologies). qPCR was performed using iTaq^™ Universal SYBR Green Supermix (Bio-Rad) using primers targeting the validating genes purchased from IDT DNA (USA) (Table 1) with Gnb1 as the housekeeping gene in 96-well PCR Plates (Bio-Rad). The samples were run as follows: 95°C for 2 min; 40×: 95°C for 10 sec, 60°C for 30 sec; 65°C for 5 sec, 95°C for 5 sec. Gene expression levels were analyzed using the ΔΔCt method, first normalized to the reference gene (ΔCt) and then to the empty vector control group (ΔΔCt) to determine relative fold-changes. Each experiment included no-template controls and was performed using technical duplicates.

Table 1:

Primers of validating genes for qPCR.

Primers	Sequence (5’−3’)
Ap1m1_Fw	TGTGGTCCATCAAGTCCTTTC
Ap1m1-Rv	GGATGCCAGAGGTAGTGAAATAG
Cul4a_Fw	CAGTCTCTTGCCTGTGGTAAA
Cul4a_Rv	TGTCTGTCCTGGAATACTCTCT
Dnajb1_Fw	CTACGGAGAAGAAGGCCTAAAG
Dnajb1_Rv	CCACCAAAGAACTCAGCAAAC
Ist1_Fw	GCTTCAAAGCTGAACGCTTAC
Ist1_Rv	AAGGTAGTCCTCCCGGATAAT
Mia3_Fw	TCCACCACTAGGCATAAGAGA
Mia3_RV	TAAACGGGCACCAGGAATAAA
Psma5_Fw	TCCAGACATCAGAGGGTGTAT
Psma5_RV	GTGTCTCCACTCTGGCTTTATC
Psmd14_Fw	GGAACTGGTGTCAGTGTAGAAG
Psmd14_Rv	CAGAAAGCCAACAGCCAAAG
Spata5_Fw	CAGGACCGGCTTGATATTCTTC
Spata5_Rv	GCATACAGACCTGCCTCATTAC
Sugt1_Fw	TGGTGAGATCAAAGAGGAAGAG
Sugt1_Rv	CTGTACCACCAGACTCCATAAA
Ufd1_Fw	CCTGGATATCACCAACCCTAAAG
Ufd1_Rv	CATCACCCGCAGTTCATAGAT
Ywhah_Fw	CTTACCGGGAGAAGATTGAGAAG
Ywhah_RV	CTGCCAAGTAGCGGTAGTAATC
Zw10_Fw	CTCAACTGAAGATGGCGATAGA
Zw10_RV	CTGGAACCTCTTCCTGGTATTT

Statistical Analyses

Statistical analysis of BAR proteomic data was performed using the DEP package (version 1.24.0) in RStudio (version 4.3.3) and Microsoft Excel. Bulk RNA-seq data was analyzed using the DESeq2 package (version 1.42.1) in RStudio. Data visualization was performed using volcano plots, PCA plots, and enrichment pathway plots, all generated with the ggplot2 package (version 3.5.0). BAR proteomic PCA plots were created with the DEP package. Bar graphs for qPCR and validated gene expression data were generated using GraphPad Prism (version 10.4.0). Statistical analysis was performed using one-sided paired t-test, and multiple testing correction by the Benjamini–Hochberg method was applied in RStudio. All figures were designed and assembled using BioRender.com.

Data availability

Mass spectrometry proteomics data, raw and processed, has been uploaded to the MassIVE (member of the proteome Xchange consortium) repository at https://doi.org/10.25345/C5ZZ0V, reference number PXD063086. Raw and processed transcriptomic data are available at GEO [GSE294982]. The code for the BAR proteomic and transcriptomic analysis is available at [https://github.com/lewiscelllabs/E1E2_BAR_Proteomic_and_RNASeq].