Development of a high cell density transient CHO platform yielding mAb titers greater than 2 g/L in only 7 days.

We developed a simple transient Chinese Hamster Ovary expression platform. Titers for a random panel of 20 clinical monoclonal antibodies (mAbs) ranged from 0.6 to 2.7 g/L after 7 days. Two factors were the key in obtaining these high titers. First, we utilized an extremely high starting cell density (20 million cells/ml), and then arrested further cell growth by employing mild hypothermic conditions (32°C). Second, we performed a 6‐variable Design of Experiments to find optimal concentrations of plasmid DNA (coding DNA), boost DNA (DNA encoding the XBP1S transcription factor), transfection reagent (polyethylenimine [PEI]), and nutrient feed amounts. High coding DNA concentrations (12.5 mg/L) were found to be optimal. We therefore diluted expensive coding DNA with inexpensive inert filler DNA (herring sperm DNA). Reducing the coding DNA concentration by 70% from 12.5 to 3.75 mg/L did not meaningfully reduce mAb titers. Titers for the same panel of 20 clinical mAbs ranged from 0.7 to 2.2 g/L after reducing the coding DNA concentration to 3.75 mg/L. Finally, we found that titer and product quality attributes were similar for a clinical mAb (rituximab) expressed at very different scales (volumes ranging from 3 ml to 2 L).

Keywords: Chinese Hamster Ovary; monoclonal antibody; polyethylenimine; transient gene expression; transient transfection

Monoclonal antibody (mAb) drug discovery and development is a complex, iterative, and lengthy process.1‐6 The initial mAb discovery process typically identifies several mAbs that bind to a specific drug target. However, these initial mAb hits are not typically drug candidates. A mAb must first meet multiple strict selection criteria. For example, the mAb must be efficacious, not be immunogenic or toxic, have a good pharmacokinetic and pharmacodynamic (PK/PD) profile, be easy to express and purify from a Chinese Hamster Ovary (CHO) cell line using standard purification methods, and have a long shelf life in a lyophilized or liquid formulation. In order to simultaneously optimize these very different properties, mAbs are engineered using a variety of techniques.2,3,7,8 Each round of antibody engineering typically generates several variants of a parental mAb. Additionally, several rounds of engineering are typically required before a specific mAb variant is ready to be considered a drug candidate.

Large amounts of protein (up to 1 g of purified protein per molecule) are required after each round of antibody engineering to perform all the necessary characterization. This in turn creates a need for quickly creating large amounts of high‐quality purified protein for a large number of molecules.

Traditionally, the biopharmaceutical industry has used transient transfection of the human embryonic kidney (HEK293) cell line to produce recombinant proteins, such as mAbs, for research purposes.9‐18 The benefits of using a HEK293 cell line include fast cell growth, ease of use, and high transfection efficiencies using multiple transfection methods. Importantly, mAb titers greater than 1 g/L have been reported in 14‐day transient HEK293 processes.19,20

Although recombinant proteins are often expressed in HEK293 cell lines for research purposes, this cell line is not typically used to manufacture biologic drugs. One reason is the increased risk of inadvertently transferring human viruses to patients.21 Therefore, most biologics are manufactured in a rodent cell line (CHO cell lines).22‐26

A major limitation of using HEK293 cell lines to produce protein for research purposes arises due to possible differences in product quality. Expressing recombinant proteins in HEK293 can introduce different product quality attributes (PQAs) compared to expression of the same protein in CHO.20,27‐31 These PQA differences are typically due not only to different N‐linked glycosylation patterns but also due to different levels of C‐terminal lysine cleavage in the mAb heavy chain. Among other notable differences, changes in PQAs can result in different PK/PD properties for recombinant proteins expressed in HEK293 versus CHO.32‐35

Expressing recombinant proteins in CHO cell line for research purposes reduces the likelihood of PQA differences between research material and protein produced from a future manufacturing cell line in a future manufacturing process. Therefore, there is a need to produce recombinant protein for research purposes using CHO cell lines. Unfortunately, recombinant protein titers for transient CHO processes have historically been low. Multiple researchers reported recombinant protein titers in the 60 to 80 mg/L range between 2004 and 2011.36‐40 Despite these modest beginnings, research into increasing transient CHO titers continued, with multiple groups eventually reporting higher transient CHO titers by adopting different approaches.

For example, Agrawal et al. reported a transient CHO titer of 200 mg/L for a human–mouse chimeric IgG1 antibody in a 10‐day process (30 ml scale) in 2013.41 Agrawal et al. optimized the DNA to polyethylenimine (PEI) ratio in addition to evaluating different growth media, feeds, temperatures, and cell densities. Rajendra et al. reported a transient CHO titer of 250 mg/L for a mAb in a 14‐day process (500 ml scale) in 2011.42 Rajendra et al. employed a very high starting cell density (4 million cells/ml) and mild hypothermic conditions (cells were incubated at 31°C for 14 days). Cain et al. reported a mAb titer of 875 mg/L in a 14‐day process (200 ml scale) in 2013.43 Cain et al. created a new CHO host cell line by engineering a CHO‐K1 host cell line to overexpress X‐box binding protein (XBP1S) and endoplasmic reticulum oxidoreductase (ERO1‐Lα). Stuible et al. published a mAb titer of 894 mg/L in a 14‐day process (25 ml scale) in 2018.44 Stuible et al. created a new CHO host cell line expressing the Epstein–Barr virus (EBV) nuclear antigen‐1 (EBNA‐1). DNA plasmids required the addition of the OriP DNA element. Additionally, these researchers optimized DNA and PEI concentrations. Steger et al. published mAb titers of 1.2 g/L from a 14‐day process (2.8 L scale) in 2015.45 Steger et al. used a combination of flow electroporation and daily feeding to obtain high titers in a CHO‐S cell line. Daramola et al. published a mAb titer of 2 g/L titer in a 21‐day process (6 L scale) in 2014.46 Daramola et al. created a new CHO host cell line by coexpressing the GS gene and EBNA‐1 in a CHO‐K1 host cell line. Plasmid DNA required the introduction of the OriP DNA element.

In addition to processes developed by biopharmaceutical companies and academic groups, ThermoFisher's ExpiCHO™ process is currently widely utilized throughout the biopharmaceutical industry. Titers as high as 3 g/L from a 14‐day process have been reported.20,47 The ExpiCHO™ process utilizes a high expressing CHO clone (ExpiCHO‐S™) that was isolated from a CHO‐S cell line. The system also requires ExpiFectamine™ (a proprietary CHO transfection reagent), ExpiCHO™ Expression Medium, ExpiCHO™ Feed, and ExpiFectamine™ CHO Transfection Enhancer.

We have previously reported the development of a simple transient expression system based on our existing CHO host cell line, existing CHO media, and existing CHO feeds. We reported a mAb titer of 0.35 g/L in a 7‐day process and 1 g/L in a 16‐day process (2 ml–2 L scale) in 2015.48 This high titer was obtained by using a high starting cell density (4 million cells/ml), optimizing PEI and DNA concentrations, and introducing a polar solvent into the medium (dimethylacetamide [DMA]). We further increased transient CHO titers to 0.625 g/L in a 7‐day process (2 ml–2 L scale) in 2015.49 The titer increase was due to the addition of DNA encoding the CHO XBP1S transcription factor at the time of transfection (boost DNA).

The objective of this study was to further improve transient CHO titers by substantially increasing cell densities while still employing our existing CHO cell line, media, and feeds. In this specific study, we combined several methods to increase transient CHO titers. First, we adopted an extremely high starting cell density of 20 million cells/ml. In contrast, other transient CHO processes utilized a starting cell density ranging between 0.5 and 6 million cells/ml.20,41‐46 In order to produce large amounts of untransfected CHO cells required for this new high cell density transient CHO platform, we employed an 8 L perfusion bioreactor to grow and maintain CHO cells at high cell densities (35–50 million cells/ml).

Second, we used a Design of Experiments (DOE) approach to simultaneously optimize six key process parameters in a single experiment. Various DOE approaches have been used by other researchers to improve expression in transient CHO systems.50‐52 During our DOE, we optimized the amount of plasmid DNA (coding DNA), inert filler DNA, XBP1S DNA (boost DNA), transfection reagent (PEI), nutrient feed addition, and glucose addition to maximize titer. As described above, most other researchers have included some of these variables to increase transient CHO titer. Our aim was to simultaneously optimize numerous process variables using a DOE approach in a single experiment.

Our DOE identified very high coding DNA amounts as optimal. We typically perform transfections at the 1 and 2 L scale to generate gram quantities of protein. At this scale, coding DNA is the most expensive component of our process. In order to reduce cost, we used cheap, inert filler DNA to reduce the amount of coding DNA. Rajendra et al. previously showed that high transient titers were still possible after replacing a large percentage of the plasmid DNA (coding DNA) with inert filler DNA.53 In this study, we substantially reduced the amount of coding DNA while maintaining high titers.

After identifying a process with reduced coding DNA amounts, we benchmarked our new transient CHO process relative to our previously published transient CHO process and to ThermoFisher's ExpiCHO™ process using a panel of 20 clinical mAbs.

Finally, we performed an in‐depth analysis using a reference mAb. We selected rituximab as the reference mAb. We expressed rituximab at five very different scales: 3 ml in a 24 deep well plate (24‐DWP), 100 ml in a 500 ml shake flask, 500 ml in a 1 L shake flask, 1 L in a 2.8 L shake flask, and 2 L in a 3 L shake flask. We purified the protein, and then measured the PQAs of rituximab.

In summary, by leveraging a perfusion bioreactor to grow very large amounts of untransfected CHO cells and by performing a multivariable DOE to optimize our process, this current investigation represents a novel approach to increase mAb titers by leveraging existing platforms.

All DNA plasmids encoding a random panel of 20 clinical mAbs were synthesized and sequence verified by ATUM (Newark, CA). The corresponding amino acid sequences of the 20 mAbs were obtained from Jain et al.54 and the Therapeutic Structural Antibody Database (Thera‐SAbDab; http://opig.stats.ox.ac.uk/webapps/therasabdab).

Construction of the DNA plasmid encoding the XBP1S transcription factor has been previously described.49

Inert herring sperm DNA (Promega, Cat. D1816) was diluted in water at 1 mg/ml and filter sterilized prior to use.

Lilly 2015 method: CHO K1SV GS KO cells were maintained in a proprietary Dulbecco's Modified Eagle Medium (DMEM)‐based medium with 8–12 mM L‐Glutamine (Cat. 59202C‐100, SAFC, St. Louis, MO) in shake flask vessels at 37°C and 6% CO2. Cells were passaged at a seeding density of 0.5–1.5 million cells/ml, every 2–3 days. Untransfected cells were maintained in shake flasks under conditions of 37°C, 6% CO2, and 160 rpm at a 25 mm shaking diameter in an 80% humidified incubator as previously described.48

ExpiCHO‐S™ method: ExpiCHO‐S™ cells were cultured in ExpiCHO™ Expression medium (Thermo Fisher, Cat. A2910001) in a humidified 8% CO2 incubator at 37°C. Cells were split every 3–4 days at a density of 0.15–0.3 million cells/ml in polycarbonate, disposable, sterile, vented Erlenmeyer shake flasks as described in the user guide.

Lilly 2020 perfusion bioreactor method: CHO K1SV GS KO cells were maintained in an 8 L stirred tank perfusion bioreactor (Chemglass Life Sciences CLS‐1406‐08). The bioreactor was maintained at 37°C with a dissolved oxygen concentration of 60% and a pH range of 6.8–7.2 using a Celltrol II controller (Chemglass Life Sciences CLS‐1438‐108). Tangential flow filtration (TFF) was used for perfusion. A laboratory and pilot scale microfiltration cartridge (GE Healthcare CFP‐4‐E‐5A) with 0.45 μm pore size, 1 mm lumen, 30 cm nominal flow path, 1,200 cm2 filter area, and 170 polysulfone fibers. The cells were continuously circulated through the lumen side of an external, vertically oriented cartridge entering the lower end of the TFF. The bioreactor was initially seeded at 1–1.5 million cells/ml in a proprietary DMEM‐based medium with 16 mM L‐Glutamine. The recirculation rate was set at 650 ml/min with a perfusion rate of 0.5 reactor volumes per day (RV/day). The perfusion rate was increased daily, to a maximum of 2.25 RV/day, based on cell density. The recirculation rate was increased to 725 ml/min 24 hr prior to transfection. On the day of transfection, cell density ranged from 35 to 50 million cells/ml. A dual head Masterflex® L/S® pump (Cole Parmer, Vernon Hills, IL) was used for perfusion, and a single head Masterflex® L/S® pump (Cole Parmer, Vernon Hills, IL) was used for recirculation.

Polyethylenimine Max (Polysciences, Cat. 24,765–1, hydrochloride salt of the linear 25 kDa PEI) was dissolved in water at 1 mg/ml based on the free base concentration (1.6 g in 1 L H2O) and filter sterilized prior to use.

Three milliliters of transfections were carried out in 24‐DWP (Axygen, Cat. PDW10ML24CS). Hundred milliliters of transfections were carried out in 500 ml flat bottom flasks (Corning, Cat. 431,145). Five hundred milliliters and 1 L transfections were carried out in 1 L and 2.8 L baffled flasks, respectively (Fisher Scientific Cat. BBV1000 and BBV2800). Two liters of transfections were carried out in a 3 L baffled flask (Corning Cat. 431253).

Lilly 2015 method: One, two, or three days prior to transfection, CHO cells were seeded at 0.5–2 million cells/ml in a proprietary DMEM‐based medium with 8–12 mM L‐Glutamine (Cat. 59202C‐100, SAFC). On the day of transfection, an appropriate amount of the cells were recovered by centrifugation at 450g for 3 min and resuspended in a proprietary blend of media hereafter referred to as CHO‐TNX media containing 0.125% N, N DMA (Acros Organics, Geel, Belgium) to a density of 4 million cells/ml. For transfection, 1.6 μg/ml coding DNA, 0.8 μg/ml XBP1S, 0.8 μg/ml filler DNA (Herring sperm DNA Promega, Cat. D1815), and 8 μg/ml PEI were sequentially added to the cells and culture was incubated at 32°C in an incubator with 6–8% CO2 and 80–90% humidity. Four hours posttransfection, proprietary feed solutions were added, and cultures were returned to the incubator. Cultures were harvested 7 days posttransfection as previously described.48

ExpiCHO‐S™ method: ExpiCHO™ transfections were performed using the ExpiCHO™ Expression System Kit (Thermo Fisher, Cat. A29133) according to the manufacturer's protocol. ExpiFectamine™ CHO transfection reagent and plasmid DNA were separately diluted in OptiPRO™ SFM (Thermo Fisher, Cat. 12,309,019). ExpiFectamine™ CHO and DNA mixtures were immediately complexed for 1–5 min. The ExpiFectamine™ CHO‐DNA‐OptiPRO™ complex was then added to the cells. For ExpiCHO™ Standard Protocol transfections, enhancer and 30% volume/volume (vol/vol) feed were added 18–22 hr posttransfection. For transfections completed with the Max Titer Protocol, Enhancer and 16% vol/vol feed were added 18–22 hr posttransfections; cultures were then temperature shifted to 32°C and an additional 16% vol/vol feed was added on Day 5 posttransfection.

Lilly 2020 method (Full DNA): On the day of transfection, the cells were harvested from the bioreactor, leaving enough cells to reseed the bioreactor at 1.5–3.5 million cells/ml. The appropriate number of harvested cells were recovered by centrifugation at 700g for 4 min, resuspended, and washed with cold CHO‐TNX (without 0.125% N,N DMA) media, centrifuged at 700g for 4 min, and resuspended at 20 million cells/ml in prewarmed CHO‐TNX media containing 0.125% N,N DMA (Acros Organics, Geel, Belgium). DNA and PEI were sequentially added to the cells, unless stated otherwise. All small‐scale transfections were performed in square shaped, pyramid bottom 24‐DWP with vented lids to minimize evaporation (Deutz, 2007) at a final volume of 3 ml. The transfected cultures were maintained in an incubator at 32°C with 6–8% CO2 and 80–90% humidity and agitated at 200 rpm on a 2‐in. diameter shaker. All small‐scale transfections were performed in duplicate, unless otherwise stated. All the large‐scale transfections (culture volume of up to 2 L), were performed in appropriately sized shake flasks on a 2‐in. diameter shaker.

Lilly 2020 method (Reduced coding DNA/ filler DNA): On the day of transfection, the cells were harvested from the bioreactor, leaving enough cells to reseed the bioreactor at 1.5–3.5 million cells/ml. The appropriate number of harvested cells were recovered by centrifugation at 700g for 4 min, resuspended, and washed with cold CHO‐TNX (without 0.125% N,N DMA) media, centrifuged at 700g for 4 min, and resuspended at 20 million cells/ml in prewarmed CHO‐TNX media containing 0.125% N,N DMA (Acros Organics, Geel, Belgium). For transfection, 3.75 μg/ml coding pDNA, 8.75 μg/ml of filler DNA, 2 μg/ml of XBP1S, and 27 μg/ml PEI were sequentially added to the cells and the culture was incubated at 32°C in an incubator with 6–8% CO2 and 80–90% humidity. Four hours posttransfection, proprietary feed solutions were added, and cultures were returned to the incubator. Cultures were harvested 7 days posttransfection.

Protein concentration (titer) was measured via analytical protein A affinity chromatography using an Applied Biosystems POROS A 20 μm, 2.1 × 30 mm2 column (Applied Biosystems, Foster City, CA) on an Agilent 1100 high performance liquid chromatography (HPLC) with UV detection (Agilent Technologies, Santa Clara, CA). All mAbs were loaded using 20 mM sodium phosphate and 500 mM NaCl pH 7.0 mobile phase and eluted using 50 mM sodium phosphate and 500 mM NaCl pH 2.0 mobile phase.

Size exclusion chromatography (SEC) was used to assess the formation of high and/or low molecular weight variants. Samples were analyzed using an Agilent 1100 HPLC with UV detection on a TSKgel G3000SWxl SEC column (7.8 × 300 mm2, 5 μm, 250 Å; Tosoh Bioscience, #08541). The column was eluted isocratically with a mobile phase consisting of 50 mM sodium phosphate, 300 mM NaCl, pH 7 at a flow rate of 0.5 ml/min at ambient temperature. The chromatography was detected by UV absorbance at 214 nm wavelength. About 10 μg of protein was injected onto the column in each analysis.

Charge variants were determined using imaged capillary isoelectric focusing on a Protein Simple Maurice instrument (San Jose, CA) with a Protein Simple Maurice cIEF Cartridge (Cat. PS‐MC02‐C). All samples were analyzed in a fluorocarbon coated capillary (50 cm × 100 μm ID) and prepared in a mixture of Pharmalyte pH 3–10 (GE Healthcare), 0.5% methyl cellulose, urea (Sigma Aldrich), and pI markers 4.05 and 10.17. The separation was achieved using a prefocus period of 1,500 V for 1 min and a focus period of 3,000 V for 10 min. Anolyte and catholyte solutions were 80 mM phosphoric acid in 0.1% methyl cellulose and 100 mM sodium hydroxide in 0.1% methyl cellulose, respectively. UV absorbance was detected at 280 nm. Data analysis was carried out with the software Compass for iCE (ProteinSimple).

Capillary electrophoresis‐sodium dodecyl sulfate (CE‐SDS) was used to assess the formation of fragments and low molecular weight variants. Analysis was performed using the Protein Simple Maurice instrument (San Jose, CA) and a ProteinSimple Maurice CE‐SDS Size Application Kit (Part # PS‐MAK02‐S). All samples were analyzed in a bare‐fused silica capillary (15 cm length × 50 μm ID) and prepared using 1X Sample Buffer (ProteinSimple), β‐mercapto‐ethanol (βME) or iodoacetamide (IAM). Reduced samples were denatured with 2.5 μl of 14.2 M βME and nonreduced samples were alkylated with 2.5 μl of 250 μM IAM. All samples contained 2 μl of 10 kDa 25X CE‐SDS Internal Standard (ProteinSimple) that was previously reconstituted in 240 μl of 1X Sample Buffer (ProteinSimple). The prepped samples were heated in a Bio‐Rad T‐100 Thermal Cycler for 10 min at 70°C. For both reduced and nonreduced methods, samples were electrokinetically injected into the cartridge capillary at 4,600 V for 20 s. Reduced samples were separated at 5,750 V for 25 min and nonreduced samples were separated at 5,750 V for 35 min. UV absorbance was detected at 220 nm. Data analysis was carried out with the software Compass for iCE (ProteinSimple).

Six parameters (coding DNA, XBP1S DNA, PEI, 4‐hour feed, Day 3 feed, and glucose supplementation) were simultaneously tested and optimized using the fractional factorial design available in the JMP v14 software package. The experiment consisted of 16 runs with eight midpoints at 3 ml scale in a 24‐DWP. We chose to test two molecules, elotuzumab (low expressing) and rituximab (high expressing). The coding DNA ranged from 10 to 15 mg/L, and the boost DNA ranged from 1 to 3 mg/L. The PEI ranged from 16 to 38 mg/L, with a 4–6% vol/vol feed at 4 hr. An additional 2.5–4.5% vol/vol feed was added on Day 3 with the addition of 0–100 mM glucose.

Of the six parameters tested, five were found to be statistically significant (PEI, coding DNA, boost DNA, 4‐hour feed, and Day 3 feed), with only the glucose supplementation having no significant impact on expression (Figure 1a). PEI was the most important parameter tested with a calculated optimal concentration of 26 mg/L. Interestingly, even though four other variables were found to be statistically significant, the impact of PEI concentration on titer was far greater relative to the other four variables. The maximum titer achieved with elotuzumab was 1.3 g/L in a 7‐day process.

Of the six parameters tested, three were found to be statistically significant (Figure 1b). Again, PEI was the most important parameter tested, with coding DNA and the 4‐hour feed showing statistical significance. Similar to experimental results with elotuzumab, coding DNA and the amount of 4‐hour feed were found to be statistically significant, but the impact on titer in the experiment was small relative to the impact of PEI. The calculated optimal PEI concentration was 28 mg/L (similar to 26 mg/L obtained in the elotuzumab experiment). The maximum titer achieved with rituximab was 2.7 g/L in a 7‐day process.

Based on the results of our DOE experiments, we adopted the midpoint conditions for all further experiments. Therefore, all subsequent transfections were done at a transfection density of 20 million cells/ml with coding DNA concentrations of 12.5 mg/L, XBP1S DNA concentration of 2 mg/L, and PEI concentration of 27 mg/L.

With coding DNA being a critical yet expensive part of the transient process, we wanted to minimize the amount of coding DNA needed. Previous publications have indicated that coding DNA can be substituted with inert filler DNA while maintaining high expression levels.53 Here, we titrated in inert herring sperm DNA to replace coding DNA. The optimal amount of coding DNA was 12.5 mg/L. Using elotuzumab and rituximab, we reduced coding DNA in 10% increments while adding filler DNA in 10% increments, keeping the total concentration of DNA fixed at 12.5 mg/L (Table 1). We found that we could reduce the amount of coding DNA by 70% (12.5–3.75 mg/L) with no loss in expression for elotuzumab and only a 28% reduction in expression for rituximab (Figure 2). Based on these results, our new optimal condition for transfection was 3.75 mg/L coding DNA, 8.75 mg/L filler DNA, 2 mg/L XBP1S DNA, and 27 mg/L PEI.

1 TABLEFiller optimization layout

Up to this point, all transfections had been carried out in a 24‐DWP. In order to generate large quantities of mAb, we must also be able to scale the process up to 2 L. We transfected rituximab at 3 ml, 100 ml, 500 ml, 1 L, and 2 L scales. We monitored daily viable cell count, viability, and nutrient levels. Viable cell count, viability, and nutrient levels were similar for all scales transfected (data not shown). Expression was similar for all scales tested and ranged from 1.7 to 2.0 g/L (Figure 3a). Seven days posttransfection, samples were centrifuged and purified using a plate‐based protein A method.55 Purified protein samples were analyzed by a series of analytical methods in order to examine the PQAs of the antibody. The amount of monomer and aggregation were quantified by analytical SEC (aSEC). No noteworthy differences were observed in monomer percentages or aggregate percentages when comparing purified protein produced at different scales (Figure 3b). Capillary electrophoresis under both reduced and nonreduced conditions (CE‐SDS R and CE‐SDS NR) was used to quantify the amount of fragmentation in each sample. Fragmentation was similar for each culture (Figure 3c,d). CE‐iCE was used to determine the level of charged species for each culture (Figure 3e). There was ~10% difference observed in the main peak across all scales tested.

We then wanted to compare expression of our new high‐density transient transfection process with our previous transient process published in 201548 (Lilly 2015). We also wanted to compare our new transient process with a commercially available CHO transient transfection kit. Therefore, we chose to compare our new process with Thermo Fisher Scientific's ExpiCHO™ kit. Previous publications report that titers up to 3 g/L are achievable in the ExpiCHO™ expression kit.20 In addition to comparing our new transient CHO process to the ExpiCHO™ expression system and our previous process, we wanted to compare expression of the new full DNA process (Lilly 2020 Full DNA) to the new filler DNA process (Lilly 2020 Filler DNA) using a panel of 20 clinical mAbs (Table 2). For the Lilly 2020 Full DNA process, 12.5 mg/L of coding DNA was used. For the Lilly 2020 Filler DNA process, we used 3.75 mg/L of coding DNA and 8.75 mg/L of filler DNA.

2 TABLETransfection process comparison

1 Abbreviations: CHO, Chinese Hamster Ovary; PEI, polyethylenimine.

Using the previous transfection method (Lilly 2015) (24‐DWP, 4 ml), expression of the 20 clinical mAbs ranged from 0.1 to 0.3 g/L in a 7‐day expression, with an average of 0.2 g/L (Figure 4). Using the ExpiCHO™ standard titer protocol, the expression ranged from 0.1to 0.7 g/L, with an average of 0.4 g/L (7‐day process). The ExpiCHO™ Max titer protocol yielded titers ranging from 0.2 to 0.6 g/L, also with an average of 0.4 g/L, in a 14‐day expression. It is worth noting that the 14‐day process versus the standard 7‐day protocol yielded a lower titer. This was unexpected, and future research should interrogate why that might be.

The average expression of the panel of clinical mAbs in our new transient transfection process (Lilly 2020) (24‐DWP, 3 ml) with full DNA was 1.6 g/L with a range of 0.5–2.7 g/L in a 7‐day process. The average expression of the transfections with filler DNA was also 1.6 g/L with a range of 0.7–2.2 g/L. The expression of the lower titer mAbs (0.5– 1 g/L) saw an increase in expression of ~25%, while the highest expressing mAbs (>2 g/L) saw an average decrease in expression of ~15% with the introduction of filler DNA. This is consistent with what were observed during the filler optimization experiments. Reducing plasmid DNA concentrations had a slight negative impact on the expression of high expressing mAbs, while it had little to no impact on expression for low expressing mAbs.

Purified protein samples were analyzed by a series of analytical methods in order to examine the PQAs of the antibodies. The amounts of monomer and aggregation were quantified by aSEC. The amount of monomer for the panel of clinical mAbs ranged from 81.5 to 98.5% in the Lilly 2015 transfection method (Figure 5a,b). The amount of monomer for the panel of clinical mAbs ranged from 92.0 to 98.7% in the ExpiCHO™ standard titer protocol and 83.8–98.3% in the ExpiCHO™ Max titer protocol. Monomer percentages ranged from 79.2 to 97.4% in the Lilly 2020 full DNA transfection protocol and 84.1–97.4% in the Lilly 2020 protocol with filler DNA. Interestingly, the monomer percentages were similar for the low titer Lilly 2015 and ExpiCHO™ platforms. Additionally, we observed that the monomer percentages from the high titer Lilly 2020 processes were slightly lower than the monomer percentages from the low titer Lilly 2015 and ExpiCHO processes. The differences in monomer content were not unexpected as we typically observe slightly lower monomer percentages from material generated from high titer stable CHO pool processes (data not shown).

Antibody drug discovery is a complex and highly iterative process, as each individual candidate molecule must satisfy multiple criteria to become a safe and efficacious drug. Material requirements for each step of antibody engineering can range from a few milligrams of material up to several grams in a drug discovery setting. Generating a stable CHO pool or clonally derived cell line is a process that can take anywhere from weeks to months, increasing the amount of time it takes for a molecule to reach milestones. Therefore, transient CHO expression is a powerful tool to quickly express a large number of antibodies.

In this study, we wanted to improve transient CHO yields from milligram quantities to multiple gram quantities without the use of a genetically engineered host cell line or plasmid DNA modifications. By increasing cell density, and using a DOE to optimize transfection conditions, we were able to generate 4 g of rituximab from a simple 2 L shake flask in only 7 days.

When benchmarking our new Lilly 2020 transfection process using a panel of 20 clinical mAbs, we found that the Lilly 2020 process yielded titers ~eightfold higher than our Lilly 2015 process and ~ fourfold higher than the ExpiCHO™ expression system. Titers ranged from 0.1 to 0.3 g/L in the Lilly 2015 transfection process, and 0.1–0.7 g/L in the ExpiCHO™ expression system. In our new and improved transfection process (Lilly 2020 filler DNA), titers ranged from 0.7 to 2.2 g/L. Interestingly, when we tested 20 clinical mAbs with and without filler DNA, the average expression remained unchanged, allowing us to greatly reduce the amount of plasmid DNA used per transfection with minimal impact on expression and product quality.

Intriguingly, rituximab was the highest expressing mAb in the panel of 20 clinical mAbs. This was true in all the transient CHO processes tested (Figure 3). Also, note that infliximab was a low‐expressing mAb according to Figure 3. This observation was consistent with the literature reports citing infliximab as a difficult to express mAb.56,57

In summary, we developed a new transient CHO platform that yielded titers ranging from 0.7 to 2.2 g/L for a random panel of 20 clinical mAbs in a simple 7‐day process. Titers were substantially higher for all 20 molecules compared to our previously published process and to ThermoFisher's ExpiCHO™ process. Titers as high as 3.2 g/L for rituximab were measured during our titer optimization experiments.

Protein product quality characterization was performed on rituximab using various analytical methods. After expressing and purifying rituximab in volumes ranging from 3 mL to 1 L, no large differences in PQAs were observed. In addition to generating high titers, no CHO host cell line engineering or changes to plasmid DNA were required. This simple platform therefore allows us to generate gram quantities of high quality mAb in only 7 days for a large number of molecules. Typically, stable CHO pools are required to produce multigram quantities of representative quality protein. Given that stable CHO pool generation and subsequent scale‐up requires several weeks to produce protein, this new process greatly speeds up the mAb drug discovery and development process.

Although the goal of the current work was to develop a new system for the expression of large quantities of representative mAb, future work should focus on evaluating this system for the expression of more complex modalities (e.g., Fc‐fusions and bispecific antibodies).

We would like to thank Dr. Chetan Patel, Dr. Jonathan Day, Mr. Jeffrey Boyles, and Mr. John Herrington for carefully reviewing this manuscript and for providing valuable feedback.