Transcriptomic Insights Into Serum-Free Medium Adaptation and Temperature Reduction in Chinese Hamster Ovary Cell Cultures

Abstract

Chinese hamster ovary (CHO) cells are widely used in recombinant biopharmaceutical production; yet, yields remain low, leading to high market prices. Improving product yield and quality has heavily relied on empirical characterization with limited insight into internal molecular dynamics. RNA-seq offers a powerful alternative to understand intracellular responses to process changes through gene expression measurement. In this study, three RNA-seq datasets across three CHO cell lines and four industrially relevant treatments were integrated to characterize the global transcriptome changes, construct a weighted gene co-expression network, assess the impact on recombinant anti-interleukin 8 (anti-IL8) immunoglobulin heavy and light chain transcript abundance, and expression of glycosylation genes. Treatments included adaptation to serum-free medium, low temperature, low pH, and low glucose concentration in the medium. The findings suggest upregulation of cholesterol biosynthesis is critical for serum-free medium adaptation, and the rate-limiting enzymes in the sterol regulatory element-binding protein pathway (Insig1 and Srebf2) could be targeted to accelerate adaptation. Temperature-induced cell cycle suppression was likely mediated by p53 activation, consistent with previous reports, with the p53-targets, Zmat3 and Btg2, identified as key hub genes. Conversely, glucose and pH were observed to have negligible impacts on the transcriptome. This study uniquely identifies novel genes mediating temperature-induced cell cycle arrest, distinct glycosylation-related gene responses impacting product quality, and new stable housekeeping genes for accurate gene expression normalization in CHO cells.

Keywords: RNA‐Seq; cholesterol metabolism; co‐expression analysis; gene expression; sterol regulatory element‐binding proteins.

FIGURE 1

Culture growth and metabolite profiles for (a) serum‐free adaptation dataset, (b) glucose dataset, (c) DP‐12 temperature/pH dataset, (d) K1‐PF temperature/pH dataset. Circled data represent RNA extraction timepoints. Error bars represent the standard deviation of biological replicates (N = 2). To facilitate visual comparisons, the same symbols and colors are used in both Figures 1 and 2.

FIGURE 2

Principal component analysis (PCA) for the three datasets representing 14 conditions (N = 2). To facilitate visual comparisons, the same symbols and colors are used in both Figures 1 and 2.

FIGURE 3

Module eigengene plots of the three datasets. The y‐axes represent the eigengene expression levels. (a) serum‐sensitive modules, (b) cell line modules, (c) temperature‐sensitive modules, (d) differences in CHO dhFr‐ expression, (e) differences in glucose samples expression. Dashed lines delineate the serum adaptation, glucose, and temperature datasets, respectively. Diagonal filled bars represents CHO dhFr‐, solid filled bars represents rCHO DP‐12, and cross‐hatched bars represents CHO K1‐PF. Number of genes within respective module provided in lower right corner. Error bars represent standard deviation of module eigengenes (N = 2).

FIGURE 4

Relative gene expression levels for the product genes and the selection gene. Normalized counts of the transfected heavy chain (HC), light chain (LC), and dhfr genes in the rCHO DP‐12 clone, shown for the three CHO cell lines across the 11 unique conditions. Dashed lines delineate the serum adaptation, glucose, and temperature datasets, respectively. Error bars represent standard deviation of biological replicates (N = 2).

FIGURE 5

Identification of new housekeeping genes for CHO cells. The coefficient of variation (CV) of gene expression was compared across the 14 conditions. Dashed lines represent average CV. Ten newly identified genes (red circles) had significantly lower CV than previously identified genes in the CHO cell literature (blue squares) or traditional (green diamonds) housekeeping genes (HKGs) within this dataset. (a) Bahr et al. [98], (b) Brown et al. [99], and (c) Ma et al. [97].