Ribosome profiling-guided depletion of an mRNA increases cell growth rate and protein secretion

Recombinant protein production coopts the host cell machinery to provide high protein yields of industrial enzymes or biotherapeutics. However, since protein translation is energetically expensive and tightly controlled, it is unclear if highly expressed recombinant genes are translated as efficiently as host genes. Furthermore, it is unclear how the high expression impacts global translation. Here, we present the first genome-wide view of protein translation in an IgG-producing CHO cell line, measured with ribosome profiling. Through this we found that our recombinant mRNAs were translated as efficiently as the host cell transcriptome, and sequestered up to 15% of the total ribosome occupancy. During cell culture, changes in recombinant mRNA translation were consistent with changes in transcription, demonstrating that transcript levels influence specific productivity. Using this information, we identified the unnecessary resistance marker NeoR to be a highly transcribed and translated gene. Through siRNA knock-down of NeoR, we improved the production- and growth capacity of the host cell. Thus, ribosomal profiling provides valuable insights into translation in CHO cells and can guide efforts to enhance protein production.


Supplementary Figure S1
Read length distribution of ribosome profiling. All three replicates from day 3 and day 6 presented. RPF: Ribosomal profiling fragments.

Supplementary Figure S2 Global codon periodicity.
To test if the Ribo-Seq data demonstrated a pattern in which pausing occurs on the first base of each codon, we superimposed the ribosomal occupancy for the first, second and third bases in each codon. We included all genes whose protein codon median coverages were greater than 0 in at least one replicate. First, we calculated the percent coverage at each reading frame by summing the coverage at all bases of each frame, divided by total coverage of the whole coding sequence (CDS). Then for each gene, the average frame percentage coverage of biological replicates was used to represent its final frame percentage coverage at day3 and day6.
To substantiate that the codon periodicity of the recombinant mRNAs are equal to that of eEF1a1, we performed an ANOVA analysis. For each gene, ribosomal occupancy, at each base pair of coding region, was normalized as the percentage of total coverage of the codon it belongs to. Following this, an ANOVA test was performed to test if 3 coding frames have the same mean percentages. The ANOVA results show that ribosomal occupancy is significantly different between 3 coding frames. To test if the recombinant mRNAs showed increased or decreased ribosomal stalling, we quantified pausing on all genes whose protein codon median coverage were greater than 0 in at least one replicate. For each gene, we calculated the median codon coverage excluding the first 15 and last 10 codons to avoid the bias introduced at those regions. Then coverage at each codon is normalized by the median. The number of stall sites were then counted for each gene, with a normalized density of greater than 25 being considered as a stall sites.
*Note: A recent study has shown that CHX can bias the translation rate of codons 1 . But here we are looking at the overall stall sites instead of codon specificity, therefore the stall sites is comparable.
Supplementary Figure S4. Increased BiP expression at late growth phase. Western blot of BiP across the batch culture. Vinculin presented as loading control.

Comment to Figure 4
The NeoR gene was utilized as an initial selection marker before switching to the DHFR selection system to drive up gene copy numbers 2 . Once the transfected cells are exposed to MTX, the cell will be forced to amplify the DHFR cassette to survive the MTX treatment. Because NeoR was encoded from both expression vectors, used to generate the CS13-1.0 cells line, it was most likely co-amplified together with the heavy-and light chain genes. Since NeoR sequesters a substantial amount of translational capacity, it is interesting that the DHFR mRNA only sequestered 0.0002% and 0.0001% of the RNA-Seq reads on day three and day six, respectively (Table S6). This is probably due to the fact that the DHFR gene was is under control of a crippled promoter instead of a strong fully functional promoter as the NeoR gene 2,3 . Since the cells were not grown in the presence of MTX in this experiment, transcription of the crippled DHFR gene would probably decrease significantly. This could explain the low mRNA levels of DHFR observed in our study.

Supplementary statistical analysis to Fig. 3
To our claim that the HC and LC have significantly more reads in the 5'UTR, ribosome occupancy at each base pair of 5'UTR was normalized by the median coverage of coding region, excluding first 45 and last 30 nucleotides. Base pairs with no coverage were excluded from downstream analysis. The Mann-Whitney test was performed to compare coverage at 5'UTR region between recombinant genes and endogenous genes, based on the assumption they have similar median coverage. The results show recombinant median coverages are significantly larger than eef1a1.  Supplementary table S3 --Ribosome profiling  alignment offset   Length  Offset  22  7  23  8  24  9  25  10  26  11  27  12  28  15  29  15  30  15  31  15  32  15  33  16  34  16  35  16  36  16  37 16