Increased glycoprotein hormone yield in stably transfected CHO cells using human serum albumin signal peptide for beta-chains

Generation of expression constructs with different signal peptides for β-chains

Plasmids previously used for transient transfection experiments (Sinegubova et al., 2024) were employed in this study. All genetic constructs were designed according to the following tricistronic scheme: “β-chain-IRES-α-chain-IRESatt-DHFR”, where IRES is an internal ribosome entry site from the encephalomyocarditis virus, wild-type or attenuated (att), and DHFR is a selection marker murine dihydrofolate reductase (Fig. 1A). Step-out PCR was conducted to replace the NSPs of the β-chains. The pairs of long adapter primers, which encoded the heterologous SPs (HSA-SP, Azu-SP, aSP) and the synthetic Kozak sequence, alongside a reverse primer common to all β-chain ORF areas, were used. After sequencing, the PCR products were subcloned into expression vector plasmids via the AbsI-SpeI/NheI endonucleases.

Cell culture and stable cell line generation

We used the CHO 4BGD host cell line (genotype BAK1^−/− BAX^−/− BCL2⁺ BECN1⁺ GLUL^−/− DHFR^−/−) obtained in-house from the CHO S (Thermo Fisher Scientific, Waltham, Massachusetts, USA) using two successive rounds of genome editing and described in (Orlova et al., 2022; Kovnir et al., 2023). Cells were cultured in a CO₂ incubator (37 °C, 5% CO₂, 155 rpm orbital shaking with amplitude 10 mm) in 125-mL Erlenmeyer flasks (VWR Scientific, Radnor, Pennsylvania, USA) in 30 ml of serum-free ProCHO5 medium (Lonza, Switzerland) supplemented with 8 mM glutamine/alanyl-glutamine, 100 mM hypoxanthine + 16 mM thymidine (HT), all supplements from Paneco, Moscow, Russia. Cell viability and concentration were measured using a Countess automated cell counter (Thermo Fisher Scientific, Waltham, MA, USA) after 0.4% trypan blue (Thermo Fisher Scientific, Waltham, MA, USA) staining.

Ten million cells were electroporated with 50 μg of plasmid DNA using the Neon apparatus and Neon electrotransfection kit (both Thermo Fisher Scientific, Waltham, MA, USA). A single pulse of 1,600 V for 10 ms was applied. Five percent by weight of control plasmid pEGFP-N2 (Addgene#6081-1; Clonentech, Mountain View, California, USA) was added to the target plasmid in each transfection. After 48 h in culture, the cells were centrifuged (300 g, 5 min) and transferred to a selective medium. Selection was performed by culturing the cells in the medium lacking HT, but supplemented with 200 nM MTX (Ebewe, Unterach am Attersee, Austria). The cells were passaged in the selective medium every 3–5 days until viability reached 90%. A higher concentration of MTX (2 μM) was then used to amplify target genes for increased glycoprotein hormones expression. The cells were passaged in the selective medium every 3–5 days until viability reached 90%. Both stably transfected and gene-amplified cell pools were cultured in ProCHO5 medium supplemented with 8 mM glutamine in a simple batch mode for 4 days. Samples of cells and culture supernatant were collected and stored at −20 °C before analysis.

Glycoprotein hormone heterodimer quantification using ELISA

Glycoprotein heterodimeric hormones concentrations in the culture supernatants were determined using a sandwich ELISA, the antibodies used are listed in Table 1. The microplates were coated by capture antibodies at 100 ng per well in 100 mM sodium carbonate buffered solution, incubated overnight at +4 °C. The blocking was performed with 3% bovine serum albumin (BSA) in phosphate-buffered saline (PBS) for 1 h at 37 °C. Fifty microliters of conditioned culture medium were diluted with PBS, added to the wells of the microplates along with 50 μl of anti-α-chain antibodies conjugated with horseradish peroxidase (HRP), dilution 1:20,000. Plates were mixed and incubated for 1 h at 37 °C on a thermostatic shaker. After washing five times with 0.1% PBS-Tween 20 (PBST), 100 μl of 3, 3′, 5, 5′-tetramethylbenzidine (TMB) substrate were added. The reaction was incubated for 10 min at RT, stopped with 100 μl of 5% phosphoric acid, and the optical density at 450 nm was immediately measured using a Feyond-A300 Microplate Reader (Allsheng, Hangzhou, China).

Glycoprotein hormone α-chain quantification using ELISA

Glycoprotein α-chain concentration in the supernatants was determined using a sandwich ELISA, the antibodies toward α-subunit of glycoprotein hormones #K003/1* and a conjugate of monoclonal antibodies against the α-subunit of glycoprotein hormones with horseradish peroxidase #K003 (Diatech LLC, Moscow, Russia) were used. The testing procedure was identical to that employed for heterodimeric hormone quantification.

Glycoprotein hormone heterodimer and free chains quantification using Western blotting

Samples of conditioned media (from 4-day batch culture) were concentrated approximately 30-fold using Vivaspin 500 ultrafiltration devices (Sartorius, Goettingen, Germany) with a 5 kDa MWCO polyethersulfone membranes. Gel samples were normalized based on heterodimeric hormone concentrations determined using ELISA, the same amount of each hormone (150 or 30 ng) was applied to each lane. Samples containing 150 ng of heterodimers were loaded on the gel without pretreatment for analyzing the distribution of the target hormone between the heterodimeric state and the free chain state. Samples containing 30 ng of heterodimers were heated at 95 °C for 10 min before loading for analysis of the total chain content. For the TSH samples stained with anti-β-antibodies the loading was 30 ng in all cases. Separation was performed using sodium dodecyl-sulfate polyacrylamide gel electrophoresis (SDS–PAGE, 12.5% acrylamide in separating gel) at 120 V for 20 min followed by 180 V for 90 min. Semi-dry transfer of proteins was performed on the nitrocellulose membrane (GVS, #1215471) using TE 70 PWR Transfer Unit (GE Lifesciences, Chicago, Illinois, USA) at 60 V for 2 h in a modified Towbin solution containing 5.76 g/L Tris-base, 2.95 g/L glycine, 20% ethanol. In the case of TSH samples, the membrane was additionally heated at 95 °C for 20 min in 0.1% PBST solution to improve the antibodies binding to βTSH. Blocking was performed in 5% BSA-PBST solution at room temperature for 1 h or overnight at 4 °C. Membranes were incubated with primary antibodies in 1% BSA-PBS solution for 1 h at room temperature with rocking, washed twice, and incubated with anti-species antibodies-HRP conjugate when necessary. The antibodies used were as follows: for βFSH and βLH - XF2 and XL1 (ХЕМА) + Goat Anti-Mouse-HRP (ab6789; Abcam, Cambridge, UK), for βTSH – r-mAb to TSH beta (ab8155959; Abcam) + Goat Anti-Rabbit-HRP (ab6721, Abcam), for βCG – К002 (Diatech). The membrane was washed with 0.1% PBST solution three times for 15/5/5 min. Pierce ECL Western reagent (Thermo Fisher Scientific,Waltham, MA, USA) was used for detection. The membrane image was captured using a FUSION Solo 6X chemidocumentation visualization system (Vilber Lourmat, France). Band quantification was performed using the TotalLab TL120 Software v2009 (TotalLab, Newcastle upon Tyne, England), absolute peak volumes method, baseline subtracted.

Transgene copy number analysis using qPCR

Target gene copy number was estimated using quantitative real-time polymerase chain reaction (qPCR). Genomic DNA (gDNA) was isolated from 1.5–2.5 × 10⁶ cells using the ExtractDNA Blood & Cells genomic DNA extraction kit (Eurogen) according to the manufacturer’s instructions. Quantitative PCR was performed using BioMaster HS-qPCR Lo-ROX SYBR 2× mix (Biolabmix, Novosibirsk, Russia) on an Applied Biosystems 7500 Real-Time PCR System (Thermo Fisher Scientific, Waltham, MA, USA). The amplification protocol was as follows: pre-denaturation for 10 min at 95 °C, 40 cycles of amplification (10 s denaturation at 95 °C, 15 s annealing at 58 °C, 15 s elongation at 72 °C). Each sample was analyzed in at least three replicates. The single-copy rab1 gene was taken for normalization. The concentration of the plasmids for calibration was calculated using the Endmemo online calculator (ENDMEMO, 2025). Primers (Table 2) were selected using the Benchling platform (Benchling, 2024) and tested for specificity using the Primer-BLAST tool (Ye et al., 2012). Threshold cycles, PCR efficiency, calibration curves, and copy number calculations were performed using the Applied Biosystems software.

RNA secondary structure analysis

Secondary structure predictions and free energy calculations for the SP-hormone-encoding mRNAs were performed using the RNAfold algorithm (ViennaRNA Web Services, 2024). The analyzed sequences included the complete pre-translational area of the spliced mRNA, the SP ORF (approximately 60 nucleotides), and 120 nucleotides downstream, encoding the N-terminal part of the corresponding β-chains. The length of the mRNA fragment used for calculations was sufficient for folding the nucleotides of the varied SP area into hairpins and removal of the downstream hairpins.

Statistical analysis

Independent two-tailed Student’s t-tests were performed using the GraphPad Prism software 10.3.1 (GraphPad Software, San Diego, CA, USA), available at GraphPad (2024). One-way ANOVA with a post-hoc Tukey HSD test was performed using the LibreOffice software (LibreOffice, Berlin, Germany), available at LibreOffice, 2024.

Genetic constructs encoding hormone chains with variable signal peptides

In this study, we used the same plasmids used to obtain transiently transfected cell pools (Sinegubova et al., 2024). Briefly, the tricistronic constructs p1.1-Tr2-Gon-BIA, based on the p1.1-Tr2 plasmid vector (Sinegubova, Orlova & Vorobiev, 2023), encoded the β-subunits of the glycoprotein hormones with various SPs in the first cistron, the common α-subunit with the NSP in the second cistron, and the selection marker DHFR in the third cistron, as shown in Fig. 1A.

Strong hairpins in mRNA can cause lower protein expression level in a predictable manner (Eisenhut et al., 2020; Weenink et al., 2018). Using the RNAfold algorithm, we calculated values of free energy of the thermodynamic ensemble for the optimal mRNA secondary structures. Correct 5’-parts of all transcripts were determined as the minimal part of mRNA, which gave all hairpin structures, involving the nucleotides from variable parts, i.e., nucleotides encoding SPs. Centroid secondary structures of the SP-gonadotropin-encoding mRNA variants are shown in Fig. S1. For all four hormones, the most energetically stable structures (low energy) were identified when using Azu-SP (Table 3, in bold).

Development of stably transfected cell pools

To estimate whether the data obtained for transiently transfected cell culture would be relevant for stably transfected cell culture, we obtained 16 corresponding stably transfected cell pools. The experiment outline is shown in Fig. 1B. The 16 plasmids (four SPs for each of the four hormones) were transfected into CHO 4BGD cells using electroporation. The plasmids contained the DHFR selection marker in the expression cassette and therefore could be used for the development of stable cell lines when transfected into the DHFR-deficient (DHFR^−/−) CHO 4BGD cells (Orlova et al., 2022; Kovnir et al., 2023). MTX is a selective inhibitor of the DHFR enzyme expressed by an encoded selection marker gene. During selection, only cells containing the DHFR selection marker in their genome can survive in the MTX-supplemented medium. The initial selection was performed with a low concentration of MTX (200 nM in all cases). Specific productivities of cells were increased by the target gene amplification step, performed as culturing of stably transfected cell pools in the presence of 2 μM MTX for 2–3 weeks until the cell viabilities reached 90%. The resulting gene-amplified cell pools were passaged simultaneously at 300,000 cells/mL and cultured in batch mode for 4 days, viable cell densities were determined and culture samples were collected. Cell supernatants were analyzed using ELISA and Western blotting; cell pellets were used for analysis of genome-integrated plasmid copy numbers.

Secretion levels of heterodimeric hormones and free chains estimated using ELISA and Western blotting

Hormone concentrations (heterodimers and total a-chains) in the cell culture supernatants were measured by sandwich ELISA. The cell-specific productivities (Qp) were calculated as the titer of the target protein secreted into the culture medium divided by the cell number (Figs. 2A–2D).

For FSH, a small (30%) but statistically significant increase in specific cell productivity by heterodimer was observed only when using the Azu SP (2.7 ± 0.08 pg/cell) compared to the NSP (Fig. 2A).

For LH, replacing the signal peptide with heterologous ones in all three cases resulted in a significant (2–4 times) increase in hormone titers. The specific productivity was 2.5 ± 0.4 pg/cell for HSA-LH, 1.8 ± 0.1 pg/cell for Azu-LH, 1.2 ± 0.1 pg/cell for aSP-LH vs. 0.6 ± 0.1 pg/cell for NSP-LH (Fig. 2B). Previously, we have obtained a polyclonal LH producer cell line using a different approach: co-transfection with two separate plasmids encoding α- and β-chains, both chains with their respective NSP. The productivity of this cell pool was comparable to that obtained in the present study—about 0.3 pg/cell (Orlova et al., 2017) but inferior to all three heterologous SPs. In the case of LH, the use of heterologous SPs was found to be an effective way to increase the specific productivity without complex experiments with pairs of co-transfected plasmids and variable selection pressures for two markers.

For CG, in the case of HSA-CG, we achieved a 60% increase in specific productivity (1.0 ± 0.1 pg/cell compared to 0.6 ± 0.01 pg/cell for NSP-CG). In contrast, in the case of Azu-CG, a 60% drop in productivity was observed (Fig. 2C).

In the case of TSH, the productivity of cells in the heterodimeric form of the hormone increased by 7–13 times when using all three heterologous signal peptides, with the maximum productivity of 1.6 ± 0.05 pg/cell for HSA-TSH (Fig. 2D).

As previously described (Sinegubova et al., 2024), in transiently transfected cultures, the relative levels of secreted α-subunit decreased when the NSPs of the β-subunit were replaced with heterologous ones. For the stable cell pools, we also measured α-chain concentrations using ELISA (Figs. 2E–2H). This analysis allowed us to accurately measure the total α-chain concentration, i.e., the sum of the β-chain-bound α-chain and the free α-chain. Since the signal peptide of the α-chain was the same in all plasmids studied, we expected the same level of total α-chain in all 16 cases. In fact, the α-chain level varied with changes in the SPs for the β-chains. Statistically significant differences in total α-chain were observed for LH (Azu vs. NSP, Fig. 2F) and TSH (Azu vs. NSP, Fig. 2H).

Therefore, for the stably transfected pools, we performed Western blotting to further investigate the distribution of the α- and β-chains between the freely secreted form and the heterodimeric form. We studied the samples in two ways: without any pretreatment to see the distribution of the chains secreted as the αβ-heterodimer and the freely secreted form, and after heating samples in sample buffer at 95 °C for 10 min—as fully dissociated chains. Blotting bands intensities of the total heat-dissociated chains were considered as the total levels of corresponding chains in the culture medium. The relative intensities of the free chains bands in the non-treated samples were considered as the level of free chain secretion. Medium samples were not treated with DTT because all reduced chains were not recognized by the antibodies used (data not shown). According to the control lanes with the purified hormones, no heterodimer dissociation was visible without heat treatment and no heterodimers remained after heat treatment in all cases.

For FSH, the increase in titer for Azu-FSH corresponded to a decrease in the free β-chain secretion level. This level was reduced almost twofold (from 32% to 13%, Figs. 3A, 3I). The secretion level of the free α-chain remained less than 2% in all cases. For FSH, CG, and TSH, very small amounts (1–9% of the total) of free α-chains were observed in the culture medium (Figs. 3A, 3C, 3D) in all cases except for the NSP-TSH cell pool, which secreted 66% free α-chain. In contrast, for LH, significant amounts (22–64%) of free α-chain were observed for all four signal peptides (Figs. 3B, 3J). In the case of CG (Fig. 3G), the free β-chain level remained relatively constant (36–41%) when varying their signal peptides. It is noteworthy that LH and CG possess nearly identical β-subunits, with CG containing an additional C-terminal domain with O-glycosylation sites, called CTP. This study suggests that this additional domain significantly alters the behavior of the pre-β-chain within cellular compartments prior to secretion. The β-chain of LH appears to be unstable, leading to a large amount of the pre-α-chain remaining in the free state in cellular secretory pathways. In the case of NSP-TSH, Western blotting data (Fig. 3D) and ELISA results (Fig. 2D) revealed unusually low levels of secreted TSH heterodimer. Simultaneously, a significant excess () of free α-chain was observed (66%, Figs. 3D, 3I). The excess was significantly reduced to 6–10% when heterologous SPs were used, coinciding with low levels of free β-chains (1–3%, Figs. 3H, 3I) and a substantial increase in heterodimeric hormone titer (ELISA, Fig. 2D).

We hypothesize that the NSPs of the TSH and LH β-chains are inefficient, and this “ineffective” biosynthesis of their β-chains may contribute to the fine regulation of TSH and LH production in the anterior pituitary.

Copy number of integrated into genome plasmids estimated using qPCR

To investigate whether increased hormone titers were due to higher copy numbers of genome-integrated plasmids, we performed quantitative PCR (qPCR) analysis of cellular DNA (Fig. 4). For FSH and LH, the integrated plasmid copy number ranged from 10 to 33 copies per haploid genome when measured by the β-chain ORF. No significant differences were observed between various plasmid variants when comparing copy numbers for both β- and α-chains. However, in the case of NSP-CG, the copy number of both β and α-chains was significantly higher than other plasmids, which aligns with the enhanced level of secreted β-chain observed by Western blotting.

Despite the increased copy number of β-chain ORF copies, no corresponding increase in β-chain synthesis or heterodimer production was observed. This discrepancy may be attributed to genetic cassette fragmentation during genomic amplification, potentially leading to transcriptionally inactive regions within the integrated plasmid. Furthermore, the copy numbers for the two pairs of plasmids exhibiting the highest protein titers, NSP-LH–HSA-LH and NSP-TSH–HSA-TSH, were nearly identical. This finding suggests that increased copy number alone cannot fully explain the significant increase in hormone heterodimer secretion levels.