Impact of Cistron Topology and Bicistronic Vector Design in the piggyBac Transposon System for Improved Antibody Expression

ABSTRACT

The use of transposon‐based technologies in cell line development has gained significant traction over the past decade. This study investigates the piggyBac transposon system combined with bicistronic vectors for monoclonal antibody expression, helping address the complex demands of multi‐specific antibody formats that require multiple cistrons. We systematically evaluated bicistronic vector modifications within the piggyBac framework, including promoter sequences, reporter placements, and cassette configurations, and compared performance to piggyBac/single‐gene vector co‐transfection methods. Our findings demonstrate that double human CMV promoter configurations driving both heavy and light chain genes significantly enhanced pool productivity (2 to 2.5‐fold) and reporter expression compared to separate promoter designs. Among tested cassette arrangements, the Light chain‐GS‐Heavy chain configuration yielded optimal productivity (1.5 to 6‐fold) and superior heavy chain/light chain RNA transcript ratios (1.4 to 3.8‐fold and 1.3 to 6.1‐fold, respectively). While unfed batch conditions showed comparable productivities between optimized piggyBac/bicistronic and piggyBac/single‐gene systems, the bicistronic approach exhibited superior productivity performance under fed‐batch conditions (1.5‐fold), with the resulting clones demonstrating significantly high productivity (top clone 9.6 g/L). This represents the first reported use of LC‐selection marker‐HC topology for monoclonal antibody expression, establishing a foundation for improved therapeutic biologic production through multicistronic vector systems combined with the piggyBac transposon strategy.

Graphical Abstract and Lay Summary

In this study, we systematically evaluated bicistronic vector designs within the piggyBac transposon system and identified that dual CMV promoters driving a Light chain‐GS‐Heavy chain configuration deliver optimal monoclonal antibody productivity. Under fed‐batch conditions, this bicistronic approach significantly outperformed piggyBac single‐gene vector systems, establishing a novel platform for improved therapeutic biologic production.

Abbreviations

CD52

cluster of differentiation 52 gene

CHO

Chinese hamster ovary

CMV

cytomegalovirus promoter

DGV

double‐gene vector

FACS

fluorescence‐activated cell sorting

FITC

fluorescein isothiocyanate

FLARE

flow cytometric attenuated reporter expression

GAPDH

glyceraldehyde‐3‐phosphate dehydrogenase

GMFI

geometric mean fluorescence intensity

GS

glutamine synthetase

GSKO

glutamine synthetase knockout

mAb

monoclonal antibody

MSX

methionine sulfoximine

RFU

relative fluorescence units

SGV

single‐gene vector

Sv40

simian virus 40 promoter

1. Introduction

The generation of recombinant and stable therapeutic cell lines producing monoclonal antibodies (mAbs) has been a cornerstone of biopharmaceutical manufacturing for over 40 years [1]. Early vector systems were typically engineered to contain one or two cistrons, simplifying the vector design and engineering process. However, in response to growing therapeutic demands, particularly for multispecific antibody formats requiring three or more cistrons, the complexity of cell line development has significantly increased over the past two decades [2, 3].

The production of recombinant cell line multispecific antibodies presents several technical challenges [4]. These complex molecules often suffer from improper folding, resulting in mispaired chains or the formation of incorrect aggregated species [5]. The increased genetic load associated with multiple transgenes can overwhelm the host cell's translational machinery, compromising expression efficiency [6]. Long‐term stability is another concern, as productivity may decline, or transgene silencing may occur over time [7]. To address these challenges, the biopharmaceutical industry has explored a range of strategies, including advanced vector design, host cell line optimization, and improvements in culture conditions. These approaches aim to enhance antibody folding, improve expression efficiency, and ensure long‐term expression stability to facilitate the production of more complex and potentially more effective therapeutic antibodies.

Conventional methods for generating recombinant stable cell lines rely on random integration of monocistronic constructs into the host genome via non‐homologous end‐joining (NHEJ) [8]. The location of transgene insertion sites within the genome can significantly influence transgene expression as well as product quality attributes due to the local genomic environment, including adjacent promoters, enhancers, and silencing elements [9, 10]. In recent years, semi‐targeted transposon‐based systems such as piggyBac (PB), Sleeping Beauty (SB), and Tol2 have gained prominence in cell line development due to their ability to support more controlled transgene integration [11, 12, 13]. Among these, the piggyBac transposon system offers several advantages; it preferentially integrates at transcriptionally active TTAA sites within the genome, which are associated with robust gene expression and greater cell line stability [14]. Additionally, the piggyBac system can accommodate large DNA cargoes (> 200 kb), making it particularly suitable for complex expression systems.

Simultaneously, the use of multicistronic vectors has become increasingly favored for therapeutic applications requiring co‐expression of multiple transgenes [15, 16]. These vectors enable the expression of multiple proteins from a single DNA vector, offering advantages such as reduced transgene copy number, improved expression consistency, and a lower risk of protein misfolding. Moreover, multicistronic vector expression systems can support better control of relative expression levels and improve protein quality, which is critical for the assembly and functionality of multispecific antibody formats [17]. Although piggyBac transposon systems have emerged as a promising alternative to conventional random integration for cell line development, the impact of multicistronic vector topology and regulatory architecture on transgene expression has not been systematically evaluated. Differences in gene positioning, promoter selection, and regulatory elements can substantially influence transcript balance and overall productivity, making optimization of these design parameters essential for the efficient expression of complex biologics such as multispecific antibodies [18].

In this study, we developed and optimized a glutamine synthetase (GS)‐based piggyBac platform combined with a bicistronic vector strategy for coordinated transgene expression. We systematically evaluated promoter selection, reporter placement, and bicistronic cassette topology to identify vector designs that maximize recombinant protein production.

Our results demonstrate a significant improvement in productivity compared with piggyBac monocistronic controls and conventional vector designs. The optimized light chain–GS–heavy chain topology produced favorable heavy chain‐to‐light chain transcript ratios and superior fed‐batch performance. To our knowledge, this is the first systematic evaluation of bicistronic cassette topologies within a piggyBac transposon platform and the first report of the light chain–GS–heavy chain topology for multispecific antibody expression. Collectively, these findings establish a robust and scalable platform for the development of high‐producing cell lines for next‐generation therapeutic proteins.

2. Materials and Methods

2.1. Construction of DNA Expression Vectors

Each evaluated sequence was inserted into the piggyBac base vector (SPB‐007‐14.7.1) using either restriction enzyme cloning (New England Biolabs) with restriction enzyme multiple cloning sites or, in cases where restriction enzyme sites were unavailable, Gibson Assembly or NEBuilder HiFi DNA Assembly (both from New England Biolabs). For restriction enzyme cloning, NheI and PacI sites were utilized for insertion into the base vector. Gibson Assembly and NEBuilder HiFi Assembly techniques, using 4–7 inserts with 15–20 base pair overlaps, were utilized to construct complex double‐gene vector topologies. All final plasmid maps were sequence‐verified via Plasmid‐EZ, provided by GENEWIZ (Azenta Life Sciences).

2.2. CHO Cell Line and Cell Culture

An internally generated glutamine synthetase knockout (GSKO) CHO‐K1 cell line was used in these studies. Cells were grown and maintained in a controlled shaking incubator (Kuhner) at 37°C, 5% CO₂ and 80% humidity, with shaking speeds of 125 rpm (shake flasks) or 260 rpm (spin tubes).

The parental cell line was cultured in CD‐CHO medium (Life Technologies), supplemented with 8 mM L‐glutamine (Life Technologies) and 50 mg/L dextran sulfate (Sigma Aldrich) prior to selection. Pool transfections were performed via electroporation using a GenePulser Xcell (BioRad Laboratories) in CD‐DG44 medium (Life Technologies) supplemented with 8 mM L‐glutamine and 0.18% Pluronic F‐68 (Life Technologies).

All piggyBac plasmid DNA was transfected as sterile circular DNA with either plasmid DNA or mRNA transposase. During selection, the CHO cells were cultured in CD‐CHO medium supplemented with 50 mg/L dextran sulfate and varying concentrations of methionine sulfoximine (MSX) for stable pool generation. Pools subjected to MSX selection were considered complete once they achieved pool viabilities > 98% and doubling times < 30 h.

2.3. Flow Cytometry Analysis of Reporter Expression After Pool Selection

After pool selection was completed for pools utilizing the FLARE reporter system, reporter expression was evaluated by incubating cells with FITC‐conjugated anti‐human CD52 antibody (Sanofi) for 30 min at 4°C. CD‐CHO media washes at 4°C were performed before and after antibody incubation. For reporter analysis, cells were resuspended in cold paraformaldehyde media at approximately 2 × 10⁶ viable cells/mL and acquired on the FACSCelesta flow cytometer (BD Biosciences). Data analysis was performed using the FlowJo software program. Flow cytometry results were displayed as a measure of the geometric mean fluorescence intensity (GMFI) and quantified in relative fluorescence units (RFU).

Our team previously developed an in‐house FLARE (FLow cytometric Attenuated Reporter Expression) system for the rapid detection and selection of pools or clones that produce high levels of recombinant proteins [19]. In this system, a sequence‐altered human CD52 cell‐surface reporter gene with attenuated expression is co‐expressed upstream of the protein biologic of interest, typically co‐expressed with the heavy chain sequence in a monoclonal antibody format. This arrangement results in correlative low‐level reporter expression and high‐level biologic expression, making the reporter a predictive indicator of protein biologic expression.

Reporter expression was evaluated following pool selection by incubating cells with FITC‐conjugated anti‐human CD52 antibody (Sanofi) for 30 min at 4°C. CD‐CHO media washes at 4°C were performed before and after antibody incubation. For flow cytometric analysis, cells were resuspended in cold paraformaldehyde‐containing media at approximately 2 × 10⁶ viable cells/mL and acquired using a FACSCelesta flow cytometer (BD Biosciences). Data analysis was performed with FlowJo software, and results were reported as geometric mean fluorescence intensity (GMFI) and quantified in relative fluorescence units (RFU).

2.4. Reporter Bulk Enrichment, Clone Generation, Screening, and Expansion

After pools completed selection and were chosen for clone generation, 20–80 × 10⁶ cells were incubated with FITC‐conjugated anti‐human CD52 antibody for 30 min at 4°C, with CD‐CHO media washes at 4°C before and after antibody incubation. These cells were resuspended in cold CD‐CHO/DMEM:F12 medium (Life Technologies) at a concentration of 2 x 10⁷ viable cells/mL and sorted for bulk enrichment on the BDInflux flow cytometer (BD Biosciences) using an argon laser emitting at 488 nm and detecting FITC emission with a 530/30 filter. Selection of cells was based on viable cell scatter plots (FSC‐A and SSC‐A) with doublet discrimination and cells exhibiting the highest FITC fluorescence. Cells were then sorted at the top 5%–10% FITC expression as a bulk collection, plated into a 6‐well static plate and allowed to recover for 24 h in CD‐CHO media supplemented with 50 mg/L dextran sulfate and MSX (if pool previously contained).

After 24 h of recovery, the bulk pool was prepared for clone generation using the Beacon Optofluidic System (Bruker Cellular Analysis). The cells were counted and resuspended in EX‐CELL CHO Cloning medium (Sigma‐Aldrich) supplemented with InstiGRO CHO (Advanced Instruments) at a cell concentration of 4 × 10⁶ viable cells/mL. Single‐cell imports were then performed onto the Beacon using OptoSelect 1750 chips for clone generation. On Days 3–4 post‐import and after clones had expanded to 6–30 total cells within each NanoPen chamber, all clones were screened using a fluorescent diffusion gradient assay to detect secreted protein biologic targets. The top‐expressing clones were then exported into each well of a static 96‐well plate. Twice per week, clones were scanned using the CellMetric (Advanced Instruments) for well confluence. Once a well‐achieved > 45% confluence, clones were FACS‐screened for the reporter as described above after pool selection. Utilizing the reporter expression, the top clones were then expanded from 96‐well plates → 24‐well static plates → 24‐deep well plates → spin tubes in CD‐CHO medium supplemented with 50 mg/L dextran sulfate and MSX (if pool previously contained). Clones were considered finished with expansion after achieving clone viabilities in spin tubes > 98% and doubling times < 30 h.

2.5. Analysis of Pool and Clone Protein Production

After pool selection or clone expansion was completed, cells were inoculated into shake tubes at 5 × 10⁵ viable cells/mL for batch culture analysis without feeding. On Day 7 of batch culture, conditioned media was harvested, clarified by centrifugation, and analyzed for titer by Octet Protein A methods.

Pools and clones were also analyzed on the Ambr‐15 (Sartorius) for fed‐batch analysis. The day before inoculation (Day −1), Ambr‐15 reactors were prepared by adding 11 mL of in‐house medium and 20 µL of anti‐foam solution (20% diluted in WFI) to each reactor.

Each pool or clone was inoculated as follows: An aliquot of the seed train culture was counted on the Vi‐CELL XR (Beckman Coulter) and, based on the cell count result, 30 × 10⁶ viable cells were harvested from the culture and transferred into a 15‐mL tube. The tube was centrifuged, the supernatant was removed, and the cells were resuspended in 4 mL of in‐house medium pre‐equilibrated to 37°C. An Ambr‐15 reactor was then inoculated with the 4 mL cell suspension to achieve a target seed density of 2 × 10⁶ viable cells/mL in a total working volume of 15 mL.

Daily feeds with in‐house neutral feeds and separate in‐house basic feeds were initiated on Day 1. Starting on Day 3, 200 µL of culture was sampled daily for IgG and LDH measurements using the Cedex Bio HT (Roche). Additionally, 450 µL of culture was sampled daily for cell counts, viability, cell size, glucose, lactate, glutamine, glutamate, ammonium, potassium, calcium, sodium, pH, pCO₂, pO₂, and osmolality using a BioProfile FLEX2 (Nova Biomedical).

On Days 7 and 12, conditioned media was harvested, clarified by centrifugation, and analyzed for titer using Octet Protein A methods.

2.6. Transcript Level Analysis

To quantify transcript levels, cells were harvested through centrifugation after finishing selection. RNA was extracted using Qiagen's AllPrep DNA/RNA kit according to the manufacturer's protocol (#80204). Reverse transcription was performed using the SuperScript IV Reverse Transcriptase kit (ThermoFisher, cat# 18090010) to synthesize complementary DNA (cDNA).

The QX200 Droplet Digital PCR System (Bio‐Rad Laboratories) was used to quantify the heavy chain and light chain in a duplex reaction with GAPDH as a reference. ddPCR supermix for probes (no dUTP) (Bio‐Rad Laboratories, cat# 1663025) was used based on the manufacturer's protocol. Primers and probes, including fluorescence dyes and quenchers, were designed based on parameters suited for ddPCR. Droplets were generated using the AutoDG Droplet generator (Bio‐Rad Laboratories).

PCR reactions were performed on a C1000 Touch Thermal Cycler (Bio‐Rad Laboratories) under the following conditions: initial denaturation at 95°C for 10 min, followed by 40 cycles of denaturation at 94°C for 30 s and annealing/extension at 58°C for 1 min. The reaction was finalized with a step at 98°C for 10 min before being held at 4°C (Table 1).

TABLE 1.

Sequences of primers and probes used for ddPCR.

Sequence name	Sequence
Heavy chain forward primer	5’‐CTCCAAAGCCTAAGGACACC‐3’
Heavy chain reverse primer	5’‐CGTACCAATTGAACTGCACC‐3′
Heavy chain probe	5’‐CGGACCCCTGAAGTGACCTGCGTGG‐3′
Light chain forward primer	5’‐CCCAAGAGTCTGTGACCGA‐3’
Light chain reverse primer	5’‐GTACACCTTGTGCTTCTCGT‐3’
Light chain probe	5’‐CCACACTGACCCTGTCCAAGGCCGA‐3’
GAPDH forward primer	5’‐TCAAGAAGGTGGTGAAGCAG‐3’
GAPDH reverse primer	5’‐CAAAGGTGGAAGAGTGGGAG‐3’
GAPDH probe	5’‐TGGGCTACACCGAGGACCAGGTTGT‐3’

Following amplification, droplets were measured in the QX200 ddPCR Droplet Reader, and relative transcript levels were quantified using QX manager software (standard edition, Bio‐Rad Laboratories).

2.7. Product Quality Analysis

2.7.1. Non‐Reduced Capillary Electrophoresis (NR‐CE)

Sample purity and integrity were assessed by non‐reduced capillary electrophoresis using the Caliper (LabChip GXII) and PA800+ platforms. Samples were analyzed under denaturing conditions to determine percent purity, calculated from the relative peak area of intact monomer versus total detected species including, low‐ and high‐molecular‐weight species (LMWS, HMWS) and fragments.

2.7.2. Capillary Isoelectric Focusing (cIEF)

Charge heterogeneity was evaluated by capillary isoelectric focusing using a Sanofi Platform and Release methods. Samples were focused within a pH gradient, resolving acidic, main, and basic charge variants according to isoelectric point (pI). Results were expressed as the percent distribution of each charge‐variant region.

2.7.3. Size Exclusion Chromatography (SEC)

Aggregate content was determined by size exclusion chromatography using high‐throughput and release methods. Samples were separated based on hydrodynamic volume under native conditions and detected by UV absorbance (280 nm). Results were expressed as percent high molecular weight species (HMWS) and percent monomer (purity).

3. Results

3.1. Expression Construct Vector Sequences #1–10

Using our GSKO CHO‐K1 host cell line and an optimized piggyBac transfection and pool selection protocol, this study evaluated the impact of vector topology and various vector elements on transgene expression and antibody production. Expression vector sequences (#1–10), shown in Figure 1, were constructed with a human CMV promoter upstream of the light chain (LC) gene and either a human CMV or hamster beta‐actin promoter upstream of the heavy chain (HC) gene. In all vector configurations, a truncated SV40 promoter (Δ3 variant) was included upstream of the glutamine synthetase (GS) gene to attenuate GS expression and facilitate a stringent selection for improved expression in pools [20]. In several constructs, a modified human CD52 reporter was inserted immediately upstream of the heavy chain gene to enable screening of both pool and clone production.

FIGURE 1.

Orientation alignment of the monoclonal antibody double‐gene expression vector sequences #1–10. Referred in this study as [Position #1–Position #2–Position #3]. Note: CMV‐p, human cytomegalovirus promoter; Hamβ actin‐p, hamster beta‐actin promoter; delta 3 Sv40‐p, truncated version (delta 3) of simian virus 40 promoter; Ham GS, hamster glutamine synthetase; CD52, altered human CD52 cell surface receptor; ITR, inverted terminal repeat sequences; HC, heavy chain; LC, light chain; SGV, single‐gene vector.

For vectors #1–4, the gene cassette configuration followed a GS‐LC‐HC topology. For vectors #5–10, the arrangement of the GS, LC, and HC sequences was varied to determine the most productive vector configuration. Sequence #5 featured an LC‐GS‐HC configuration with a CD52 reporter, while sequence #6 utilized a GS‐HC‐LC arrangement. Additional topology variations were introduced in sequences #7–10: sequences #8–10 placed the heavy chain (HC) in position #1, and sequences #9–10 included the glutamine synthetase (GS) selection gene in position #3. Sequence #7 was structurally identical to #5 except for the absence of the CD52 reporter gene.

Each piggyBac vector construct was flanked by a core insulator and an inverted terminal repeat sequence. The bicistronic/piggyBac transposon expression system formats were evaluated against the co‐transfected, single‐gene vector (SGV)/piggyBac expression system.

3.2. Stable Pool Expression Analysis of Various Double‐Gene Vector Topology Orientations

A comprehensive stable pool expression analysis was performed for vector sequences #1‐10 to assess the impact of gene cassette topology on pool productivity, CD52 reporter expression, and heavy/light chain RNA transcript levels. Stable pools were generated in duplicate for each vector configuration, and expression performance was evaluated in both batch and fed‐batch formats.

As shown in Figure 2A, pool selection kinetics varied across constructs and appeared to correlate with the position of the GS gene. Constructs with GS located in position #2 or #3 (sequences #5, #7–10) exhibited greater selection stringency and slower recovery relative to those with GS in position #1 (sequences #1–4, #6). All constructs completed selection within 25 days. Batch productivity measured on Day 7 post‐selection (Figure 2B) indicated that sequences #5 and #7, both containing the LC‐GS‐HC configuration, achieved the highest titers. Sequence #5 included the CD52 reporter, while sequence #7 did not. To benchmark performance, piggyBac/double‐gene vector pools were compared to control pools generated by piggyBac/single‐gene constructs of heavy and light chains both driven by the hamster beta‐actin promoter. Sequence #5 yielded productivity comparable to the control, whereas sequence #7 exceeded control titers.

FIGURE 2.

Stable pool productivity and RNA transcript analysis after selection. Pools were transfected in duplicate, and all pools completed selection within 25 days. (A) Pool viability profiles during selection for sequences #1–10. Viability profiles were benchmarked against the piggyBac/single‐gene vector pool condition (black). (B) Stable‐pool batch titer production after selection for sequences #1–10. Batch productivities were assessed on Day 7 using an in‐house media formulation. All pools were compared with the piggyBac/single‐gene vector pool condition (black lines), in which co‐transfection of the mAb was performed. (C) Stable‐pool Ambr‐15 fed‐batch titer production after selection for sequences #1–7. Conditioned media was harvested on variable days up to Day 12 and analyzed for pool productivity. Sequence #2 was contaminated and was eliminated from analysis. (D) Heavy‐ and light‐chain RNA transcript levels after pool selection for sequences #1–10. All transcript levels were normalized to GAPDH. All pools were performed in duplicate.

These results highlight the role of vector topology, particularly the positioning of the GS gene, in influencing both the selection kinetics and overall expression performance. The LC‐GS‐HC configuration supports robust pool productivity.

Fed‐batch productivity (Figure 2C) was assessed for piggyBac‐mediated double‐gene vector constructs #1, #3, and #4, as well as orientation variants of construct sequences #5‐7. Cultures were maintained in the Ambr 15 bioreactor system, and conditioned media was collected on variable days up to Day 12 for titer analysis.

Among the constructs tested, sequences #5 (with reporter) and #7 (no reporter), both featuring the LC‐GS‐HC configuration, exhibited the highest fed‐batch productivity, outperforming all other piggyBac/double‐gene vector configurations. Notably, sequences #5 and #7 also demonstrated higher productivity than the control pools generated through piggyBac/single‐gene vectors driven by the human CMV promoter. These findings support the conclusion that the LC‐GS‐HC orientation, as represented by sequences #5 and #7, is the most effective configuration among the vector topologies tested for maximizing recombinant antibody expression under fed‐batch conditions.

To further investigate the impact of vector configuration on transgene expression, RNA transcript levels for the heavy and light chains were quantified among sequences #1‐10. This analysis aimed to assess the correlation between transcript abundance and pool productivity and to inform future vector design by identifying optimal gene positioning within double‐gene constructs.

As shown in Figure 2D, sequence #5 (LC‐GS‐HC) exhibited significantly elevated transcript levels, with a 3.5‐fold increase in heavy chain and a 2‐fold increase in light chain expression compared to sequence #2 (GS‐LC‐HC). This enhancement appears to result from repositioning of the GS gene to position #2 and the light chain to position #1, supporting the notion that upstream gene placement promotes higher transcriptional output. In contrast, sequence #6, in which the light chain was moved to position #3, showed a 3.5‐fold decrease in light chain transcript levels relative to sequence #2. Sequence #6 also demonstrated a 2‐fold increase in heavy chain transcript levels, corresponding to the relocation of the heavy chain gene from position #3 to #2.

RNA expression data from sequences #7–10 further reinforced the influence of gene position on transcriptional output. Constructs with the heavy chain gene placed in position #1 (sequences #8‐10) consistently showed high heavy chain transcript levels. Similarly, the highest light chain expression was observed in constructs where the light chain gene occupied the first position (sequences #5 and #7). Notably, sequence #9 exhibited the lowest light chain transcript levels, despite placing the light chain in position #2. Differences in GS gene placement (positions #2 or #3), as well as extended selection durations, may have contributed to these observed discrepancies in gene expression.

Overall, sequences #5 and #7 demonstrated the highest heavy and light chain RNA expression levels among all constructs, correlating with their superior pool productivity. These findings support the presence of a positional hierarchy in gene position within the vector cassette, following the general trend: position 1 > position 3 > position 2. This positional effect underscores the critical role of vector topology in optimizing RNA transcript levels for recombinant protein expression.

CD52 reporter expression was evaluated in vector sequences #1‐10 that included the reporter gene (sequences #1–2, #5–6, and #10). In these constructs, the reporter gene was positioned immediately upstream of the heavy chain gene, enabling a tightly coupled expression relationship between the two (Figure 1). This design allowed CD52 cell surface expression to serve as a reliable surrogate for evaluating monoclonal antibody production, at both the pool and clonal levels.

As shown in Figure 3, sequence #5 (LC‐GS‐HC with the reporter) exhibited the highest level of reporter expression among the constructs tested. Sequences #6 (GS‐HC‐LC) and #10 (HC‐LC‐GS) also showed elevated reporter expression relative to the original sequence #2. These findings highlight the impact of vector configuration and gene positioning on transgene expression.

FIGURE 3.

Stable pool reporter expression after pool selection is shown for sequences #1–2, #5–6, and #10. Sequence #5 showed the highest reporter expression compared to the other vector configurations tested, which included the reporter gene sequence (#1–2, #5–6, and #10). A mock control pool not expressing the reporter was used as a negative control. Note: RFUs, relative fluorescent units. All pools were performed in duplicate.

3.3. Impact of Heavy Chain Promoter Selection on Pool Productivity and RNA Transcript Levels

Vector constructs #1 and #2 (both containing the CD52 reporter) and sequences #3 and #4 (both lacking the reporter) were evaluated to assess the impact of promoter selection, human CMV or hamster beta‐actin, on heavy chain expression. In all constructs, the light chain gene was positioned upstream and driven by the human CMV promoter, while the heavy chain gene was positioned downstream and driven either by the human CMV or hamster beta‐actin promoter.

As shown in Figure 2A, all four constructs exhibited comparable pool viability profiles during selection. However, Day 7 batch productivity revealed marked differences based on promoter choice (Figure 2B). Sequence #2 (contains reporter, human CMV‐driven heavy chain) produced approximately a 2‐fold higher titer than sequence #1 (contains reporter, hamster beta‐actin‐driven heavy chain). Similarly, sequence #4 (no reporter, human CMV‐driven heavy chain) yielded a 2.5‐fold increase in titer compared to sequence #3 (no reporter, hamster beta‐actin‐driven heavy). These results indicate that employing dual human CMV promoters to drive both the heavy and light chain genes in a GS‐LC‐HC configuration enhances productivity in this vector orientation.

Interestingly, constructs lacking the CD52 reporter gene (sequences #3 and #4) outperformed their reporter‐containing counterparts (sequences #1 and #2), suggesting that inclusion of the reporter may impose a burden on overall pool productivity. Notably, the magnitude of this productivity difference was promoter‐dependent: the human CMV promoter exhibited a more pronounced decrease in productivity when the reporter was present, whereas the hamster beta‐actin promoter showed a comparatively smaller impact. These observations indicate that reporter‐associated expression limitations may interact with promoter strength or regulatory characteristics to influence overall expression output.

Reporter expression analysis (Figure 3) further confirmed that sequence #2 showed higher CD52 expression than sequence #1, consistent with the productivity differences between the CMV promoter and the hamster beta‐actin promoter‐containing constructions.

RNA transcript analysis in Figure 2D revealed that sequence #4 (no reporter, CMV‐driven heavy chain) exhibited increased heavy chain transcript levels relative to sequence #3 (no reporter, hamster beta‐actin‐driven heavy chain). In contrast, sequences #1 and #2 had comparable heavy chain transcript levels, suggesting that the presence of the reporter may attenuate the transcriptional advantage conferred by the CMV promoter. Sequences #1 and #3 exhibited higher light chain transcript levels than sequences #2 and #4, despite the only difference being the promoter driving the heavy chain. This observation may reflect transcriptional competition or broader vector configuration effects that indirectly influence light chain expression.

Collectively, these findings highlight the superiority of the human CMV promoter for driving heavy chain expression and support the use of dual CMV promoters for maximizing productivity in this vector architecture.

3.4. Clone Expression for Pool Sequences #5 and #7

Based on the pool productivity screens in Section 3.2, sequences #5 and #7, both featuring the LC‐GC‐HC vector configuration (with sequence #7 lacking the reporter gene), demonstrated the highest pool productivity and optimal heavy/light chain (HC/LC) RNA transcript ratios. To determine whether these vector configurations would yield productive clones, clonal analysis was conducted using pool sequences #5 and #7 for two separate mAb molecules to evaluate the productivity gain from the resulting clones. Sequence #5 was selected due to its inclusion of the reporter sequence. The reporter enabled enrichment of high‐expressing populations through bulk sorting prior to cloning and has the potential to compensate for the slight reduction in pool‐level expression (Figure 2B).

Pools generated with sequence #5 were subjected to bulk sorting via the reporter and enriched at the top 5% prior to clone generation. The resulting bulk‐sorted cells were then loaded onto the Optofluidic Beacon System the following day for clone generation. Pools generated with sequence #7 (lacking the reporter gene) were simply loaded live onto the Optofluidic Beacon System for clone generation. Individual pens were screened for the expression of the target biologic, and the top 50 clones from each clone set were selected for export and expansion into 96‐well plates. These clones were subsequently expanded through standard cell line development stages, progressing from 96‐well plates to 24‐well plates, 24‐deep well plates, and finally spin tubes via our cell line development procedure. The majority of the 50 clones for both mAb molecules exported successfully, completed spin tube expansion, and were evaluated for batch and fed‐batch productivity.

Figure 4 shows the Day 7 batch titer results for the two mAb molecules, with the top‐performing clone from mAb2 reaching a titer of 1.42 g/L. For comparison, the original sequence #5 and #7 pools are also shown, demonstrating that the majority of the resulting clones from both clone sets exhibited a substantial 2 to 3‐fold productivity improvement (mAb1) and a 1.5 to 2‐fold productivity improvement (mAb2) compared to their respective original pools.

FIGURE 4.

Enriched sequence #5 (light and dark green) and unenriched sequence #7 (light and dark blue) clone batch titer productivities after clone expansion, with sequence #5 and #7 presort pool productivities (black) shown as comparators. The sequence #5 pools were combined and sorted via reporter expression at the top 5%, loaded onto the Beacon Optofluidic System for clone generation, screened for the secreted antibody target, and the top clones (50) were exported for clone expansion. The top surviving clones are shown and were evaluated for batch titer productivities. All batch clone collections were performed on Day 7 with an in‐house media formulation.

To further assess productivity under conditions more representative of bioreactor manufacturing, the top‐performing mAb1 clones from the batch screening were evaluated in a fed‐batch process. The top 20 clones from the batch titer analysis from each clone set were thawed and inoculated in a 12‐day fed‐batch process using the Ambr‐15 system. Figure 5 shows the fed‐batch clone growth and titer productivity profiles of the clones derived from sequences #5 and #7 pools. The highest‐producing clone reached a peak titer of 9.6 g/L on Day 12. Sequence #5 clones exhibited slightly lower cell counts (Figure 5A) and culture viabilities (Figure 5B) during the fed‐batch process; however, both clone sets observed similar Day 12 titers (Figure 5C). Notably, the majority of clones outperformed their respective original pools, indicating successful enrichment (+ reporter) and clone screening (+/‐ reporter).

FIGURE 5.

Enriched sequence #5 clones (green) and unenriched sequence #7 clones (blue) Ambr‐15 fed‐batch clone titer productivities after clone expansion, with sequence #5 and #7 presort pools (black) shown as comparators. The top 20 clones are shown and were evaluated for fed‐batch growth (A, B) and titer productivities (C). The top clone achieved 9.6 g/L in the fed‐batch process. All batch clone collection harvests were performed on Days 3, 4, 7, 10, and 12 with in‐house media formulations. Note: VCD, viable cell density.

3.5. Clone Product Quality and Stability Evaluations for Sequences #5 and #7

To evaluate the influence of the LC‐GS‐HC vector topology on product quality attributes in the piggyBac transposon system, sequence #5 enriched clones selected via reporter gene expression were compared against sequence #7 unenriched clones derived without reporter selection for mAb1. Both clone populations were characterized across key product quality metrics, including molecular purity, monomer content, and main peak distribution.

As illustrated in Table S1, both enriched and unenriched clone populations maintained consistently high product quality profiles. Molecular purity was equivalent between the two populations (unenriched clones: 90.55% ± 0.96%; enriched clones: 90.80% ± 0.94%), as was monomer content (unenriched clones: 96.60% ± 0.97%; enriched clones: 96.44% ± 1.01%), demonstrating that reporter‐based selection did not adversely impact these critical quality attributes. A modest difference was observed in main peak distribution, with unenriched clones exhibiting a slightly higher main peak percentage (59.45% ± 2.93%) relative to enriched clones (57.65% ± 2.17%). However, given the limited effect size, the biological relevance of this difference remains minimal. Collectively, the high product purity, monomer content, and main peak percentages observed demonstrate that the LC‐GS‐HC vector topology supports robust antibody expression with equivalent product quality for mAb1 in the piggyBac transposon system, indicating that this vector topology does not significantly influence product quality outcomes.

To evaluate the long‐term expression stability of enriched and unenriched clones utilizing the LC‐GS‐HC topology in the piggyBac transposon system, representative high‐performing clones from both populations were cultured continuously for up to 40 generations. Batch titers and percent titer changes were monitored at regular intervals to assess transgene stability and cellular performance over extended passaging (Figure S1).

Long‐term stability was assessed by continuously culturing five sequence #5‐enriched and five sequence #7‐unenriched clones for both molecules over 40 generations. For mAb1, both enriched and unenriched populations maintained stable (within 70% of the original starting titer at generation 0) Day 7 titers (enriched: 0.8–1.1 g/L; unenriched: 0.8–0.9 g/L) across all generations. Both clone groups were within 80% of their original generation 0 titer values after generation 40 with the exception of a single enriched clone which showed a slight unstable titer profile by generation 40 (66% of original generation 0 titer value). Similar strong stability was observed for mAb2, with enriched clones maintaining titers of 0.8‐1.1 g/L and unenriched clones exhibiting titers of 0.8‐1.2 g/L (all within 80% of their original generation 0 titer values).

These results demonstrate that the piggyBac transposon system with the vector topology configuration of LC‐GS‐HC supports robust, stable antibody expression over extended culture periods regardless of enrichment strategy, making it suitable for industrial cell line development where long‐term production stability is critical.

4. Discussion

In this study, we explored an expression platform utilizing bicistronic vectors with a semi‐targeted integration via the piggyBac transposon system. This approach simplified transfection workflows while providing enhanced control of gene expression and stability. We compared various bicistronic vector designs within this framework, assessing the impact of promoter sequences, reporter gene placement, and expression cassette configurations on pool productivity and gene expression, all using the semi‐targeted piggyBac transposon system. Incorporation of a double human CMV promoter to drive both the heavy and light chain genes, significantly increased pool productivity (2 to 2.5‐fold) and reporter expression as compared to vectors using a combination of hamster beta‐actin and human CMV promoters together (Figures 2B and 3). However, the presence of the reporter gene was found to moderately reduce heavy chain RNA transcript levels (Figure 2D) and overall monoclonal antibody production (Figure 2B), demonstrating that the reporter gene, while useful for expression monitoring, may negatively impact productivity in its current configuration. In the HC‐LC‐GS cassette configuration, an improvement in pool productivity was observed, with only a modest increase in RNA transcript levels being shown, which indicates a context‐dependent effect of the enhancer on gene expression.

Among the various vector variations tested, the rearrangement of gene cassette configurations had the most pronounced effect on pool expression and productivity. Specifically, the LC‐GS‐HC configuration consistently yielded the highest titers and the most favorable heavy‐chain‐to‐light‐chain RNA transcript ratios. Vector designs #5 and #7, both of which contained the LC‐GS‐HC configuration, demonstrated superior pool performance, in both pool productivity (1.5 to 6‐fold) and heavy/light RNA transcript ratios (1.4 to 3.8‐fold and 1.3 to 6.1‐fold, respectively) compared to the other topology conditions. This finding may be due to a variety of factors working independently or together affecting differential expression across these topologies. These factors include, but are not limited to, potential promoter interference with downstream expression, the use of an attenuated SV40 promoter for GS expression, and/or positioning the light chain in the first position to promote its excess expression. Furthermore, excess light chain expression from the first position may also help with enhanced antibody folding and assembly, as has been previously reported [21], although excessively high light chain levels may negatively affect overall pool productivity. In our study, optimal pool productivity was associated with light chain to heavy chain RNA transcript ratios of approximately 2:1 to 4:1, although this could be molecule‐specific.

Using the optimized LC‐GS‐HC double‐gene vector configuration and piggyBac‐based integration strategy, we compared this platform against the piggyBac/single‐gene vector platform with a standard monoclonal antibody. While pool titers were comparable under unfed batch conditions (Figure 2B), significantly higher pool titers (1.5‐fold) were observed under fed‐batch conditions in Ambr‐15 bioreactors with the piggyBac/double‐gene vector system (Figure 2C).

Reporter‐based enrichment did not significantly influence product quality (Table S1) or long‐term expression stability (Figure S1) for the optimized LC‐GS‐HC vector configuration in the piggyBac transposon system. Both enriched and unenriched populations maintained equivalent purity and monomer content, with negligible differences in main peak distribution. Over 40 generations of continuous culture, both clone groups demonstrated stable batch titer productivity, indicating that the vector topology configuration of LC‐GS‐HC has no detrimental impact on long‐term clone performance.

Collectively, these results support the use of a multi‐cistronic vector system with semi‐targeted integration and a more efficient and scalable approach for the development of cell lines producing complex biologic molecules. This positional expression trend was observed exclusively within this fixed piggyBac backbone with dual CMV promoters and an attenuated SV40‐driven GS cassette and may not extend to unrelated vector frameworks. Ongoing work aims to further refine this platform for broader application across diverse therapeutic modalities.

Author Contributions

Jason Vitko: conceptualization, methodology, investigation, formal analysis, supervision, project, administration, resources, writing – original draft, writing – review and editing. Thomas King: data curation, investigation. Efecan Aral: data curation, investigation. Steven McLellan: data curation, investigation. Hope Divello: data curation, data curation, investigation. Sara Maimouni: conceptualization, data curation, investigation. Tiffany McLamarrah: data curation, investigation. Jennifer Tedstone: data curation, investigation. Victor Cairns: conceptualization, methodology, supervision, resources, writing – original draft, writing – review and editing. Christine DeMaria: conceptualization. John J. Scarcelli: conceptualization, methodology, supervision, resources, writing – original draft, writing – review and editing.

Funding

This work was funded by Sanofi.

Conflicts of Interest

All the authors are/were Sanofi employees and may hold shares and/or stock options in the company.

Supporting information

Supporting File: biot70296‐sup‐0001‐SuppMat.docx.

Data Availability Statement

The data used in this study are not publicly available due to confidentiality and proprietary restrictions.