Systematic use of synthetic 5'-UTR RNA structures to tune protein translation improves yield and quality of complex proteins in mammalian cell factories

Abstract

Predictably regulating protein expression levels to improve recombinant protein production has become an important tool, but is still rarely applied to engineer mammalian cells. We therefore sought to set-up an easy-to-implement toolbox to facilitate fast and reliable regulation of protein expression in mammalian cells by introducing defined RNA hairpins, termed 'regulation elements (RgE)', in the 5'-untranslated region (UTR) to impact translation efficiency. RgEs varying in thermodynamic stability, GC-content and position were added to the 5'-UTR of a fluorescent reporter gene. Predictable translation dosage over two orders of magnitude in mammalian cell lines of hamster and human origin was confirmed by flow cytometry. Tuning heavy chain expression of an IgG with the RgEs to various levels eventually resulted in up to 3.5-fold increased titers and fewer IgG aggregates and fragments in CHO cells. Co-expression of a therapeutic Arylsulfatase-A with RgE-tuned levels of the required helper factor SUMF1 demonstrated that the maximum specific sulfatase activity was already attained at lower SUMF1 expression levels, while specific production rates steadily decreased with increasing helper expression. In summary, we show that defined 5'-UTR RNA-structures represent a valid tool to systematically tune protein expression levels in mammalian cells and eventually help to optimize recombinant protein expression.

Figure 1.

Schematic representations of regulation element (RgE) actions. (A) The RgE is cloned between the transcription start site (TSS) and the coding sequence of the gene to be regulated. When this gene is transcribed the RgE will fold and form an RNA secondary structure. The efficiency of the RgE can be influenced by changing the minimum free energy (MFE), the GC-content of the RNA stem and the relative position to the 5′-cap of the mRNA. (B) The underlying hypothesis is that using such RgE to tune expression levels of either the recombinantly introduced product itself or a helper factor can increase secretion of the desired product by mammalian cell factories by improving assembly or processing.

Figure 2.

Proof-of-principle and evaluation of RgEs in mammalian cell lines. (A) Cloning procedure of RgE insertion into 5′-UTRs. GOI = gene of interest. (B) Schematic representation of experimental proof-of-principle. (C andD) Calculated fold change (FC) of RFP/BFP expression ratio mediated by RgEs screened in CHO (C) and HEK293 (D) cells in reference to the unregulated CMV sample. Bars show average, colored points show individual values of samples (n = 3 independent samples each; CMV: n = 6) and error bars show SEM. Color, shape and GC-content indicated on the right side. Elements are ordered according to their average fold change. The horizontal black line shows average, red dotted lines show SEM of CMV samples. All samples, except the ones marked with ‘n.s.’, showed a significant difference with P < 0.001 in comparison to the CMV sample. (E) Average fold changes of CHO versus HEK cells. Gray area shows standard error. (F) Average fold changes (FC) of CHO (left) and HEK (right) cell plotted against the MFE of the respective elements. Gray area as in (E). (G) Differences in fold change observed between high and low GC content elements of similar MFE at the same relative position. Y-axis depicts the calculated FC difference. X-axis shows the RgE pairs that were compared. Red full line shows averages (RgE 12:13 not included). Light red dotted line depicts SEM.

Figure 3.

mRNA expression level analysis. (A and B) Observed average fold changes in RFP expression levels on the protein (taken from Figure 2C and D, full dots; n = 3 independent samples each) and mRNA level (open dots; n = 2 independent samples each; see Supplementary Figure S6 for individual values) for CHO (top) and HEK (bottom) cells ordered in decreasing order for protein expression levels (A) and mRNA levels (B). RNA to protein fold change ratio is depicted in Supplementary Figure S7. The red lines highlight specific fold changes that should simplify evaluation of fold change and impact of mRNA levels. (C) Average fold changes of mRNA levels plotted against fold changes on protein level for CHO (top) and HEK (bottom) cells. (D) Average mRNA fold changes of CHO versus HEK cells. Gray area shows standard error.

Figure 4.

Using RgEs to tune HC expression levels to optimize recombinant production of an IgG. (A) Schematic overview of the expression construct. (B) Viable cell densities (VCD) and viabilities (%) of the transfected ExpiCHO™ cells (n = 3 independent samples each; RgE 11: n = 2). Columns represent different RgE constructs introduced. (C) Measured IgG titers on different days post transfection (pT) by either protein A- (top) or L-based (bottom) sensors. Samples are ordered by decreasing expression strength of the HC (n = 3 independent samples each; RgE 11: n = 2). Shapes of the data points indicate the respective replicates as depicted by the shapes in (B).

Figure 5.

Analysis of product quality of the produced IgG. (A) Non-reducing sodium dodecyl sulphate-polyacrylamide gelelectrophoresis (SDS-PAGE) of protein-A beads purified IgG from the supernatant on day 6 post-transfection. fs = full size, 2 HC = heavy chain dimer. An image of the reduced plots can be found in Supplementary Figure S9b. (B) Calculated fold change (FC) of the ratio of fs to 2 HC from purified IgG from the supernatant on day 6 post-transfection referenced to sample CMV (n = 3 independent samples each; RgE 11: n = 2). Shapes of the data points indicate the respective replicates as indicated in Figure 4B. Gel pictures from all three replicates are shown in Supplementary Figure S9a. (C) Selected SEC chromatograms of protein A column purified IgGs produced. HMW = high molecular weight non-native fractions, LMW = low molecular weight non-native fractions, mAb = monoclonal antibody (wanted, native form of trastuzumab). All SEC chromatograms are shown in Supplementary Figure S10. (D) Calculated population distribution for respective RgE-produced IgGs based on peak areas shown as percentage of total peak areas for each chromatogram.

Figure 6.

Tuning of helper protein SUMF1 levels impact ASA expression and activity. (A) Schematic overview of the expression strategy. (B) Western blot detection of SUMF1 levels intracellularly (from cell lysates) and secreted (supernatant). GAPDH was used as loading control. (C) Calculated relative SUMF1 expression levels (to CMV sample) based on intracellular and secreted SUMF1 signals form (B). (D) Measured and calculated SUMF1 levels based on the western blot versus the expected levels from Figure 2C. Gray area shows standard error. (E) Measured ASA titer (top) and calculated qP (bottom) values (n = 2 independent samples). Error bars show 95% confidence interval. (F) Measured ASA activity (top) and calculated protein specific activity (bottom) (n = 2 independent samples). Error bars as in (E). (G) ASA qP versus relative SUMF1 expression. R² calculated by linear regression of low (blue) and high (brown) relative SUMF1 expressions. Colored areas show standard error. (H) ASA specific activity versus relative SUMF1 expression.