The WAS transgene expression plasmid (pTL20c_MND_WAS_650) is a 3rd generation self-inactivating (SIN) LV expression plasmid. LV production of this HIV-1 based SIN LV expression plasmid (derived from the pCl20c plasmid backbone (Throm et al., 2009) is tetracycline-regulated with a mini cytomegalovirus (CMV) promoter and 7 tet operators (7tetO). The LV uses a MND (MPSV LTR, NCR deleted, dl587 PBS) promoter for WAS protein expression. The WAS cDNA was modified with two silent point mutations, C417A and C1098T, to remove SfiI restriction sites without altering the encoded amino acids. Downstream of the WAS cDNA is a Woodchuck Hepatitis Virus (WHV) Post-transcriptional Regulatory Element (WPRE) containing six point mutations to abrogate woodchuck hepatitis virus X (WHX) truncated protein expression while maintaining transgene expression (Zanta-Boussif et al., 2009). For WAS construct-1 (plasmid size: 9.6 kb, Supplementary Figure S1A), a 650 bp segment of the chicken hypersensitivity site 4 (cHS4) β-globin chromatin insulator (Urbinati et al., 2008) is inserted in the 3′LTR in reverse orientation to the viral transcript, serving as an additional safety and anti-silencing element. Variants of this LV construct were utilized in this study and differ in the sequence of the cHS4 insulator: WAS construct-2 (plasmid size: 9.6 kb, Supplementary Figure S1B) contains a mutated version of the 650 bp segment, WAS construct-3 (plasmid size: 9.3 kb, Supplementary Figure S1C) contains a 400 bp segment (Aker et al., 2007). An additional Enhanced Green Fluorescent Protein (EGFP) reporter transgene expression plasmid (pTL20c_MND_GFP_400, plasmid size: 8.3 kb) was utilized that differs in the encoded transgenic sequence, where EGFP replaces WAS. The antibiotic-resistance plasmid (pPGK_ble, plasmid size: 3.8kb, Supplementary Figure S1D) contains a Zeocin selection marker driven by PGK promoter.

Plasmids (pTL20c_MND_WAS_650 or its variants, pTL20c_MND_GFP_400, pPGK_ble) were transformed into NEB^® Stable Competent E. coli Cells (a derivative of the DH10B strain) in LB Broth, Vegitone. Concatemers were generated with the respective transgene plasmid (pTL20c_WAS/ pTL20c_GFP) and the bleomycin resistance plasmid (pPGK_ble). SfiI and PfIMI restriction sites on pTL20c and pPGK_ble respectively, were used to linearize plasmids, which were then purified using Qiaquick gel extraction. The linearized plasmids of transgene plasmid and antibiotic resistance plasmid were directionally ligated in a molar ratio of 25:1. The concatemer mixture was purified using a Qiagen Genomic-tip 500/G column, eluted, precipitated in isopropanol, air-dried for 60 min, and resuspended in TAE buffer, to reach a usable concentration for experiments.

Three transfection reagents and varying amounts of concatemer were tested to identify the optimal condition for stable producer cell line development (CLD). The experiment first focused on selecting the best transfection reagent by comparing ViaFect™, Lipofectamine^® 3,000 and calcium phosphate by using a constant DNA concentration of 5 µg. GPRG PCL was seeded in 6-well plates at 1.33E+05 cells/cm², 24 h before transfection. The growth medium used was Dulbecco’s modified Eagle’s medium with high glucose and Glutamax (DMEM, Gibco) with 10% Fetal Bovine Serum (FBS, USA origin Premium Corning, Gamma Irradiated), referred to as D10. The DMEM was supplemented with 1 ng/mL DOX (Doxycycline hyclate, MP Biomedicals; prepared in distilled water, Fisher Scientific) and is referred as D10+DOX. Transfection mixtures containing the GFP concatemer (5 µg per well) along with a transfection reagent (Lipofectamine™ 3,000 (kit with p3000™ enhancer reagent), ViaFect™ or CalPhos™) were preprared in OptiMEM without DOX. The transfection mixtures were added to cells and incubated for 4 h at 37 °C in OptiMEM (Reduced-Serum Media, Thermofisher) with 1 ng/mL DOX. Post-incubation, transfection media was replaced with D10+DOX. Transfection efficiency was determined by measuring % positive GFP cells and median fluorescence intensity (MFI) using a flow cytometer (MACSQuant Analyzer 16) at 72 h and 2 weeks post-transfection. Non-transfected cells were maintained as a negative control, to monitor the viability decline on host cell line during the selection phase. Zeocin (Zeocin selection antibiotic, Gibco; at 50 μg/mL) was introduced in cell culture media starting 72 h post-transfection for selection of stably transfected cells. Later different concatemer amounts such as 5 μg, 7.5 µg and 10 µg per well were tested with Lipofectamine 3,000™ transfection reagent. Transfection efficiency and median GFP fluorescence were examined 96 h and 1-month post-transfection to ensure stability of inserted DNA copies.

GPRG and GPRTG PCLs were transfected in quadruplicates with WAS construct-1 concatemer and subjected to Zeocin selection for 21 days, resulting in stable polyclonal producer cells (pools) (four GPRTG-WAS pools and GPRG-WAS pools each). Similarly the two PCLs were transfected with the GFP concatemer, resulting in one GPRTG-GFP pool and GPRG-GFP pool each. These stable pools were seeded at a cell density of 2E+04 cells/cm² in a single-layer CellSTACK (CS1) (Corning) with 100 mL of D10+DOX (growth phase). At 96 h post-seeding (hps), LV production was induced with a rinsing step of the cell layer with D10 to remove doxycycline, followed by addition of fresh D10. At 24 h post-induction (hpi), the supernatant was discarded and replaced with fresh D10, and the 1st harvest was collected 48 hpi, after which harvests were collected once every 24 h. Depending on extent of cell layer detachment, 6 to 10 LV harvests could be performed. In case of high cell detachment, harvests were centrifuged at 1,200 rpm for 3 min to have a cell-free harvest. Harvest samples were aliquoted and stored at −80 °C for subsequent analytical assays, including infectious titer (TU/mL), physical viral particle titer by p24 (vp/mL), RNA content (RNA copies/mL), RNA truncation analysis to check integrity and length of RNA sequences by ddPCR and VCN from CD34⁺ transduction studies.

The WAS construct-3 underwent comprehensive monoclonal cell line development utilising GPRTG packaging cells and transfection via the Lipofectamine™ 3,000 protocol. Cells derived from the GPRTG Master Cell Bank (MCB) were expanded and transfected with the WAS construct-3_zeocin concatemer using Lipofectamine™ 3,000, under two DNA input conditions (8.17 µg and 12.2 µg per 60 mm dish), resulting in the generation of eight polyclonal pools (L-1.1, L-1.2, L-1.3, L-1.4, L-2.1, L-2.2, L-2.3, L-2.4). Zeocin selection (80 μg/mL) commenced 72 h post-transfection, for all transfected pools and one non-transfected control. Following selection, stable polyclonal pools were expanded and cryopreserved as a Research Cell Bank (RCB) in a cryoprotectant solution of 90% FBS and 10% DMSO. To assess LV productivity, these stable polyclonal pools were expanded and induced in CS1 stacks, producing six harvests per pool, analyzed by ddPCR for infectious titer (TU_[ddPCR]) quantification.

All stable polyclonal pools were subjected to a single-cell sorting workflow to capture cellular heterogeneity for monoclonal cell line derivation. Sorting was performed using the Cytena F.SIGHT™ system into 96 well-plates containing InstiGRO^® HEK in D10 medium supplemented with DOX and Zeocin. Plates were imaged using the Solentim™ imager on days 0, 3, 7, 11 and 13 post-sorting to confirm monoclonality. Clones were scaled from 96 well plate to CS5 (Cell Stack 5-layer) for monoclonal RCB cryopreservation. For LV productivity screening, 51 monoclonal clones were seeded into 6-well plates and the top four clones were selected based on monoclonality verification and TLA analyses of the RCB. The long-term stability of the four LV producer monoclonal cell clones was assessed over a cultivation period of 60–67 days. During this timeframe, each clone underwent 8 to 9 independent LV production runs (P1 to P8/P9), with three sequential harvests per production, to evaluate consistency in infectious titer output. TLA sequencing was performed on the top performing clone to confirm vector integrity, identify integration sites, detect sequence variants and estimate the integrated vector copy number. Non-induced cells from the seed train of the stability study were analyzed by TLA sequencing at time point 0 (from RCB vials, referred to as TLA day0), and day 43 (TLA day43). LV genome integrity and variant detection were further assessed by Next-Generation Sequencing (NGS) using harvest 2 from LV production 1 (P1) and production 6 (P6) during the stability study. The RNA was reverse transcribed into cDNA and sequenced by NGS to evaluate consistency across the production timeline.

One GPRTG-WAS pool was used for process optimizations and technological evaluations. LV production optimization was conducted in CS1, with the best conditions validated in 10-layer CellSTACK (CS10). Variables such as seeding densities, induction with or without a wash step, surface coating, and harvest media volume (0.24 mL/cm² vs. 0.16 mL/cm²) were examined. For coating experiments, CS1 was coated with Poly-L-lysine solution of high and low molecular weight (P4832 Sigma Aldrich and A-005-M Merck). For a seeding density of 8.5E+04 cells/cm², induction was performed 24 h post-seeding (hps). Lower seeding densities (1.3E+04 and 2E+04 cells/cm²) required 72–96 hps before induction.

To examine upscaling differences, four conditions (1.3E+04 and 2E+04 cells/cm² with and without wash at induction) were performed in both CS1 and CS10 for yield comparison. The first at-scale LV production in 16 CS10 units was performed at a seeding density of 8.5E+04 cells/cm², with LV production induced at 24 hps. Following CS1 optimization, the process was applied to another 16 CS10 batch for comparison where CS10 were seeded at 2E+04 cells/cm², with LV production induced at 96 hps without wash. Harvest was collected every 24 h, starting at 48 h post-induction, and stored at −80 °C for analysis.

The Scale-X™ Carbo 10 m² (R&D grade) versions V1, V1.4, and V2 were utilized for pilot-scale LV production. This fully automated system is equipped with pH, DO, and temperature probes. Initial runs (SC01, SC02) used WAS construct-1 producer cells, followed by runs with WAS construct-2 and 3 to assess robustness.

The R&D grade allows for fixed-bed carrier sampling of 11.5 cm² each, and cell count by cell lysis. V1 and V1.4 were single-use vessels assembled with calibrated pH and DO probes for autoclaving, while vessels of V2 were gamma-irradiated ready-to-use with single-use pH and DO probes. All vessels were filled with 1.6 L of media post-sterilization The pH was set to 7.2 ± 0.05, temperature to 37 °C, and stirrer speed to achieve 1.1–1.3 cm/s linear speed. DO was maintained at 50% with 200 mL/min air or O₂ supply.

After media equilibration and pH confirmation (7.2–7.3), the bioreactor was inoculated at 1.3E+04 to 2.2E+04 cells/cm² with a final volume of 1.8 L. Recirculation of 19 L of media at 40 mL/min was initiated for 4-day growth phase after confirmation of >95% of cell attachment (4–24 hps), to ensure adequate nutrient supply for cells. Induction for LV production was performed, without wash, at 96 hps or upon reaching >1.80E+05 cells/cm². The perfusion process included continuous harvest collection at 4 °C and constant media feed, similar to small-scale perfusion process. Target perfusion volume during induction was 23L (26L for V2), 19L (for all versions) at Harvest 1 and 23L (26L for V2) for subsequent harvests. Media and harvest bags were changed every 24 h, with LV aliquots frozen at −80 °C for analysis.

HEK293T/17 cells were cultured in DMEM medium supplemented with 10% FBS and 1% Penicillin-Streptomycin (Thermo Fisher). Prior to transduction, cells were seeded at a density of 1E+05 cells/well into 24-well plates using DMEM medium supplemented with 12.5 μg/mL polybrene (Merck). LV supernatants were serially diluted and 200 µL of each dilution was added to seeded cells. Four days post-transduction, cells were trypsinized, stained with LIVE/ DEAD Fixable Aqua Dead Cell Stain Kit (Thermo Fisher Scientific), fixed and permeabilized with BD Cytofix/Cytoperm™ Fixation/Permeablization Kit (BD Biosciences). The WAS protein (WASp) staining was performed by incubating cells with mouse anti-WAS 5A5 antibody (BD Biosciences) at 1:100 dilution and goat anti-mouse APC conjugated (Thermo Fisher Scientific) at dilution of 1:1,000. The WASp-positive cells were determined using MACSQuant^® 16 Flow Cytometer (Miltenyi Biotec) and infectious titers were calculated in TU/mL.

Samples were treated with DNase (RNAse free DNAse from Qiagen™) to remove host cell genomic DNA contamination. Viral RNA was purified using the QIAamp Viral RNA Mini Kit and reverse-transcribed into cDNA using the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific). Samples were analysed by ddPCR using specific primers and probes according to the lentivirus genome region to be analyzed. The physical titer is reported in copies/mL which allows the deduction of the number of lentiviruses as each virus contains two RNA copies. To understand potential differences between packaging cell lines based on incomplete or truncated sequences, RNA truncation was evaluated and here results are reported as a ratio between the 3′ end (insulator region) and the 5′ end (U5 psi region).

The p24 ELISA was performed using the QuickTiter™ LV Titer Kit (Cell Biolabs) that quantifies only the LV-associated p24. Samples were subjected to a pull-down step to collect LV particles, followed by serial dilution and sandwich ELISA with FITC-conjugated anti-p24 and HRP-conjugated anti-FITC antibodies at 1:1,000 dilution. Results were analyzed using a Tecan Spark reader and reported as ng/mL of p24, with physical viral particle titers calculated with the formula from the kit manual [1 ng/mL p24 = 1.25E+07 LV particles (vp/mL)].

Two harvests each from GPRG-WAS pools and GPRTG-WAS pools from small scale CS1 were processed individually: GPRG-WAS Pool 1 (harvests 3 and 4), GPRG-WAS Pool 2 (harvests 3 and 4), GPRG-WAS Pool 3 (harvests 3 and 4), GPRG-WAS Pool 4 (harvests 3 and 4), GPRTG-WAS Pool 1 (harvests 3 and 4), GPRTG-WAS Pool 2 (harvests 2 and 3), GPRTG-WAS Pool 3 (harvests 3 and 4), GPRTG-WAS Pool 4 (harvests 2 and 3). Harvest 2 and 3 were taken for GPRTG-WAS Pool 2 and 3 due to noticeable cell detachment in CS1. Harvest material from CS1s was purified by 20% sucrose density gradient ultracentrifugation at 25,000 rpm for 2 h at 4 °C. The viral pellet was resuspended in X-VIVO10 (Lonza) supplemented with 2% human serum albumin (CSL Behring), and incubated at 4 °C for 2 h. Cryopreserved G-CSF Mobilized Human Peripheral Blood CD34⁺ Cells (Cat# M34C-GCSF-2, Hemacare Bioresearch Products, US) from a male donor were thawed and cultured in a prestimulation media containing serum-free X-VIVO10 medium (Cat# LZ-BE04-743Q, Lonza, Basel, Switzerland), 1X GlutaMax CTS (Cat# A12860-01, Gibco, Thermo Fisher Scientific, US), 2% Human Serum Albumin (CSL Behring, Switzerland), 100 ng/mL rhTPO (Cat# 1017–050), 300 ng/mL rhSCF (Cat# 1018–050), 200 ng/mL rhFlt-3 (Cat# 1015–050) (Cell Genix Technologies, Germany) for 24 h at 37 °C and 5% CO₂ at 2E+06 cells/mL. After 24 h, cells were seeded at 2E+05 cells/0.1 mL in a 96-well plate, and the prestimulation media was changed to a fresh medium containing the same cytokines and the following Transduction Enhancers (TEs): 1 mg/mL LentiBOOST (Cat#SB-A-LF-901–01, SIRION Biotech, Germany), 10 μM PGE2 (Cat# 72372, Stem Cell Technologies, Canada), 8 μg/mL Protamine Sulfate (Cat# P3369-10G, Sigma Aldrich, US). Following 1 h pre-incubation with TEs, CD34⁺ cells were transduced with LV at a MOI of 1, 5 and 10 and returned to the incubator for 24 h. Post-transduction, cells were washed and resuspended in expansion media (X-VIVO10, GlutaMax, 2% human serum albumin, rhTPO, rhSCF, rhFlt-3) for 24 h. Cells were then transferred to a 12-well plate with fresh expansion media. VCNs were evaluated by ddPCR at 8 days after transduction.

For pilot-scale comparison, LV from Scale-X™ Carbo 10 m² (SC01) and 16 CS10 R2 were processed in Sublot 1 (combined harvests 1 and 2) and Sublot 2 (combined harvests 3 and 4). They were purified using Mustang Q anion exchange membrane and concentrated by tangential flow filtration (TFF) (Vogel et al., 2024) and VCN was analysed by ddPCR.

Vector copy Number (VCN) was assessed by ddPCR using QX200™ Droplet Digital PCR (Bio-Rad Laboratories, Germany) according to manufacturer’s instructions. In brief, the reaction mixture (20 μL) containing 50 ng template DNA, 1X ddPCR Supermix for Probes [(No deoxyuridine triphosphate (dUTP) (Bio-Rad)], 50 units/μL of HaeIII restriction enzyme (Cat# R0108M, New England Biolabs, US) 900 nM of each primer, and 250 nM of each probe (Bio-Rad) was loaded into the sample well in the QX100 Droplet Generator (Bio-Rad). A total of 40 μL of oil–water emulsion containing approximately 20,000 droplets was generated with the droplet generator. Post-PCR amplification, droplets were analyzed using the QX200™ Droplet Reader and QuantaSoft software (Bio-Rad). The following primer-probe sets were used to amplify lentiviral psi (Ѱ) and the human ribonuclease P protein subunit p30 (RPP30) as a control for normalization (Ѱ: fwd 5′-TAG​TGT​GTG​CCC​GTC​TGT​TG-3′, rev 5′-CCT​CTG​GTT​TCC​CTT​TCG​CT-3′ and probe 5′-FAM-TCTCTAGCAGTGGCGCCCGA-3’; RPP30: fwd 5′-TGT​AAG​TGG​TAG​TGC​ATA​GAC​TTT​A-3′, rev 5′- GTC​AAG​AGT​AGG​AGG​ACA​TTT​GA-3′ and probe 5′-HEX- AGG​CAG​ACT​GAC​ACT​AGA​GTT​CAC-3′). VCNs were determined by calculating the number of copies of Ѱ to every two copies of RPP30.

The graphical workflow for the generation of stable producer cell lines is illustrated in Figure 1A showing concatemer generation and transfection and selection of cells with Zeocin antibiotic. To determine the optimal transfection conditions for generating stable producer cell lines, GPRG PCLs were transfected with GFP concatemer (long continuous DNA containing 25 copies of GFP and one copy of Zeocin resistance gene) using ViaFect™, Lipofectamine™ 3,000 (with p3000 enhancer reagent), and CalPhos™. Flow cytometry analysis at 72 h post-transfection showed that cells transfected with ViaFect™ and Lipofectamine™ 3,000 had >95% GFP positive cells, while the CalPhos™ condition only achieved 50% (Figure 1B). Afterwards, cells were cultured with Zeocin for 2 wks to select for stable integration reaching a cell viability above 90% and all three transfection reagents resulted in a 100% GFP-positivity (Figure 1C). The median fluorescence intensity (MFI) increased following the 2-wks selection period, with the highest intensity observed in cells transfected with Lipofectamine™ 3,000 (Figure 1C). It was observed that selection period of 2 weeks were not sufficient (upon observing non-transfected control cells) and in upcoming experiment, cells were monitored upto 1 month in Zeocin selection.

Different concentrations of concatemer (5 μg, 7 μg, and 10 µg) were analyzed to determine the optimal concentration that enhances transfection efficiency and LV titer while ensuring it is not toxic to cells. The optimal amount of concatemer for Lipofectamine™ 3000-mediated transfection was determined to be between 7.5 µg and 10 µg. This range was identified based on the high fluorescence intensities observed at both 96 h and 1-month post-transfection and Zeocin selection (Figure 1D). To this end, cells were seeded and induced for lentiviral production. Elevated GFP LV titers were observed at higher concatemer DNA concentrations, reaching up to 4.45E+06 TU/mL in pooled LV harvests for 10 μg and 4.385E+06 TU/mL for 7.5 µg, compared to 1.105E+06 TU/mL for the 5 µg condition. Lipofectamine™ 3,000 was used in all upcoming experiments to generate stable producer cell lines with WAS constructs 1, 2 and 3.

Prior to generating a stable producer cell line for ex-vivo gene therapy applications, it was imperative to select one suitable PCL, for producing good quality LV expressing the transgene. For this, stable producer pools namely GPRG-WAS pools (n = 4), GPRTG-WAS pools (n = 4), GPRG-GFP pool (n = 1) and GPRTG-GFP pool (n = 1) were generated by transfection with WAS construct-1 concatemer (long continuous DNA containing 25 copies of WAS and one copy of Zeocin resistance gene) or GFP concatemer (long continuous DNA containing 25 copies of GFP and one copy of Zeocin resistance gene) respectively. Cells were subsequently selected with Zeocin and seeded in CS1 for LV production. Overall, 10 harvests were achieved for GPRG-WAS pools and the GPRG-GFP pool, while 6 and 8 harvests were obtained for GPRTG-WAS pools and the GPRTG-GFP pool, respectively, due to cell detachment.

GPRTG-WAS pools yielded more infectious LV, with an average total TU of 1.25E+09 (calculated by total TU from all harvests per pool per CS1 production, then average of all 4 pools) compared to 3.25E+08 average total TU for GPRG-WAS pools. The average TU/mL per harvest (average of TU/mL across all harvest and 4 pools) was 6.6-fold higher for GPRTG-WAS pools (2.17E+06 Avg TU/mL, n = 4) compared to GPRG-WAS pools (3.26E+05 Avg TU/mL, n = 4). Similarly, a 5.5-fold difference was observed between GPRTG-GFP pools (1.16E+06 Avg TU/mL per harvest, n = 1) and GPRG-GFP pools (6.39E+05 Avg TU/mL per harvest, n = 1) (Figure 2A).

The infectious titer showed the same trend as the genomic titer. The genomic titer for the GPRTG-WAS pools measured by ddPCR was 3-fold higher (1.54E+10 RNA copies/mL per harvest, n = 4) compared to GPRG-WAS pools (5.28E+09 RNA copies/mL, n = 4). The same tendency was identified also for the GFP candidate with a 1.7-fold higher genomic titer for GPRTG-GFP in comparison to GPRG-GFP pools (n = 1) (Figure 2B).

The physical viral particle titer (vp/mL) derived from virus-associated p24 measurement is also having the same trend. It was 1.8-fold higher by GPRTG-WAS pools (3.9E+10 vp/mL, n = 4), compared to GPRG pools (2.15E+10 vp/mL, n = 4). A similar fold change was observed for the GFP candidate (1.7 fold change) (Figure 2C). The comparison of GOI-containing versus GOI-deficient LV particles showed a 3.6-fold increase in GPRTG-WAS pools compared to GPRG-WAS pools (n = 4), and a 4.6-fold increase in GPRTG-GFP pools compared to GPRG-GFP pools (n = 1) (Figure 2D). As shown in Figure 2E, the average RNA truncation [ratio between the 3′ end (insulator region) and the 5′ end (U5 psi region)] evaluated by ddPCR for both cell lines did not significantly change between GPRG and GPRTG pools (WAS and GFP constructs).

Transduction experiments on CD34⁺ cells at MOI of 1, 5, and 10 demonstrated a dose-dependent response for LV material produced by GPRTG-WAS pools (n = 4). A VCN of 1 was achieved at an MOI of 1, and subsequent higher VCNs of up to 3 and above were achieved at MOIs of 5 and 10. This dose-response trend was not observed with LV produced from GPRG-WAS pools (n = 4), where VCN of 1 was achieved at MOI of 5, and no further increase at higher MOIs (Figure 2F). These results highlight the superior performance of GPRTG-WAS pools compared to GPRG-WAS pools, emphasizing the importance of selecting the appropriate PCL for Cell Line Development (CLD) projects. Infectious titers were determined by transducing HEK293T cells with serial dilutions and analyzing transgene expression via flow cytometry. The MOI was calculated based on infectious titer (TU/mL) using the formula: MOI = [(volume of LV added in mL) × (infectious titer in TU/mL)]/ (number of cells seeded).

Monoclonal stable producer cell lines were derived from WAS-stable polyclonal pools via lipofectamine method using WAS-Zeocin concatemer. To establish zeocin-resistant polyclonal stable pools, both transfected and non-transfected control cells were subjected to continuous zeocin selection for 21 days. Complete cell death in non-transfected controls by day 21 validated the efficacy of antibiotic selection. Cell proliferation and viability were rigorously monitored throughout the selection period. A pronounced decline in viable cell density, termed the crisis point, was observed at day 10 post zeocin exposure, followed by a recovery in cell viability in all stable pools (Figure 3A). In contrast, non-transfected controls showed no recovery. Total viable cell counts and viability percentages are shown in Figure 3A. Infectious titers from the stable polyclonal pools were quantified via droplet digital PCR (ddPCR). As shown in Figure 3B, the titers of individual harvests are reported as TU_[ddPCR]/mL for each pool. All pools except L1.1 and L2.4 exhibited high lentiviral titers, exceeding 2.0E+07 TU_[ddPCR]/mL at harvest 3. Polyclonal pools were subjected to single-cell sorting to derive monoclonal cell lines. 51 monoclonal clones exhibiting robust proliferation were assessed for lentiviral vector productivity. The lentiviral titers among the clones were heterogeneous, as illustrated in the Figure 3C, with the highest titer observed in clone 66, reaching 1.33E+08 TU/mL.

To verify stability and integrity of the genomic integration, Targeted Locus Amplification (TLA) analysis was performed on the generated RCBs derived from the monoclonal clones. Clones 23 and 173, which lacked sequence variants within the vector genome, were prioritized for further analysis. Clones 16 and 3 were also included, as no mutations were identified in the WAS coding region or the insulator. Based on TLA results, lentiviral titers, and cell growth profiles, clones 23, 173, 16, and 3 were chosen for long-term stability assessment. Throughout a 60/67 days stability study, all five evaluated monoclonal producer cell lines exhibited comparable morphological characteristics, with no significant differences in confluency or detachment observed across LV production runs. Cell growth remained robust and consistent, with surface cell densities ranging from 4E+05 to 6E+05 cells/cm² and viabilities exceeding 95% throughout the culture period. Infectious titer analysis shown in Figure 3D) revealed minor inter-clonal variability. Clone 173 maintained stable performance across eight production runs and yielded the highest mean titer at 2.1E+07 infectious particles across eight productions. In contrast, Clone 214 exhibited the lowest mean titer at 7.03E+06 infectious particles over nine productions. Three of the five clones demonstrated stable LVV production, with approximately ≤25% titer reduction over the study duration. The remaining two clones showed a decline of approximately 50% in infectious titer, indicating reduced production stability.

TLA samples (non-induced cells) were collected at Days 0 (RCB), 43, and 60 from non-induced cells, corresponding to theoretical timelines for MCB (43 days) and WCB (60 days) generation including Univercells 10 m² manufacturing processes. TLA sequencing of clone 173 at Days 0 and 43 confirmed a single integration site on chromosome X with no sequence variants. In parallel, lentiviral vectors (LVs) harvested from productions P1 and P6 were subjected to next-generation sequencing (NGS) to evaluate sequence fidelity over time. The analysis revealed high sequence identity across the full-length LV genome, confirming genetic stability between early (P1) and late (P6) production stages. Figure 3E summarizes the NGS analyses, including the annotated vector regions and corresponding amino acid substitutions for variants with a frequency above 5%.

Two nucleotide variants were identified at positions 3,869 and 5,919 in both P1 and P6 samples, each with a frequency exceeding 5%. These variants were determined to be either silent or located outside the WAS coding region and, therefore, are not expected to affect the functionality or integrity of the LV product. Overall, the results indicate consistent genetic stability of the LV genome during the extended culture period, supporting the robustness of the selected monoclonal clones for long-term lentiviral vector production.

For the optimization of lentiviral vector (LV) production in CellSTACKS^® various seeding densities, induction methods, and chemical coatings of the flatware layer were evaluated using producer cell lines of WAS construct-1. The uncoated CellSTACK (CS1), seeded at a higher density of 8.5E+04 cells/cm² and induced 24 h post-seeding (hps) with a wash and 100 mL volume, served as the “control” and achieved over 50% cell layer confluency at the induction time point. To achieve similar cell layer confluency at induction, lower seeding densities of 1.3E+04 and 2E+04 cells/cm² with cell growth phases of 72 or 96 h were investigated. Figure 4A demonstrates that the uncoated control yielded higher number of transducing units compared to Poly-L-lysine (high or low molecular weight) coated flatware conditions. The anticipated benefits of coating with Poly-L-lysine, such as reduced cell detachment, were not observed in the number of harvests and total TU yield. Results also indicated that the lower seeding density of 2E+04 cells/cm² (with induction at 72 hps and 150 mL perfusion volume) on uncoated CS1 yield higher total TU compared to the control (8.5E+04 cells/cm²). To investigate the benefits of seeding density and the impact of process scale-up with the aim to reduce laborious steps, optimization by varying seeding density and influence of wash step at induction was explored in CS1 and CS10. Figure 4B shows that conditions without a wash step performed the best in both systems. All conditions with a wash step showed high variation in TU yields, potentially due to high cell detachment and cell layer disturbance. Additionally, the wide error bars on the graph indicate that CS1 was highly prone to handling issues and operator-related cell detachment during washing steps. Figure 4B shows that, on average, CS10 yields 20%–30% lower TUs per unit area compared to CS1, potentially due to slower equilibration of nutrients and gas exchange in the multiple layers of CS10.

Results in Figure 4 indicate that the lower seeding density of 1.3 E+04 and 2E+04 cells/cm² and a growth phase of 96 h was optimal for LV production. Thus, the process that yielded high TUs and decently reproducible results, CS10 with 2E+04 cells/cm² seeding density, 96 hps induction without a wash step, was chosen for the final pilot-scale head-to-head comparison. Overall, we have concluded that LV production processes using flatware systems have limitations in reproducibility due to operator-related process variability and are labor-intensive, especially for large-scale processes requiring the use of multiple CellSTACKS^®.

Adherent bioreactors can be utilized to produce LV in a highly controlled environment compared to CellSTACKS^®. Therefore, LV production was compared across CellSTACKS^® and two adherent bioreactors (iCELLis™ and Univercells Technologies™) to evaluate and compare the LV process yield, process length and robustness. The small-scale technological evaluation for a stable LV production platform comprised of six different processes of iCELLis™ Nano 4 m² bioreactor (iC01, iC02, iC03, iC04, iC05 and iC06), two processes of Scale-X™ Hydro 2.4 m² bioreactor (S05, S08) and one at-scale CellSTACK run with 16 units of CS10 (16CS10 with surface area 10.17 m²) seeded at 8.5E+04 cells/cm² and induced with wash at 24 hps for LV production. All runs were seeded with the same producer cell line of WAS construct-1 for comparability.

Figures 5A–C presents a comparative analysis of various iCELLis™ (iC) processes and a CellSTACKS^® run (16CS10), with a graphical overview shown in Supplementary Figure S2.

iC01: This process showed a three-fold decline in infectious titer (TU/mL) from the third to the fourth harvest, coinciding with the addition of a base for pH control (setpoint of 6.9 ± 0.15 during the LV production phase) starting from the second harvest day (Figures 5A,B; Supplementary Figure S3A). iC02 and iC03: These runs were operated with the same process parameters except for the pH setpoints, 7.2 for iC02 and 7.4 for iC03. Induction was performed without vessel volume exchange, and the perfusion volume was moderately increased for iC03 to prevent the pH from dropping below 6.75. Both processes showed consistent LV productivity and yields (Figures 5A,B) and exhibited low cell detachment (Figure 5C). iC04: This run was conducted in recirculation mode with a higher perfusion volume of 0.19 mL/cm² from the second harvest onwards to reduce LV retention on cells and improve oxygenation. There was no increase in TU productivity per unit area (Figures 5A, 6A). The amount of LV produced per unit volume (TU/mL) was lower, suggesting dilution of the produced LV (Figure 5B). iC05 and iC06: These were technical replicates with induction involving vessel volume exchange to instantly remove residual doxycycline. Technical replicate runs showed good reproducibility and peak productivity was observed at the second and third harvests (Figures 5A,B), after which productivity drastically reduced. The peak titer at the second harvest was higher than in any other process. Cell detachment from the fixed bed was gradual (Figure 5C), with peak detachment at 2E+05 cells/mL, compared to high detachments in the 16CS10 batch of up to 7E+05 cells/mL. The TU/mL per harvest was higher for all iCELLis™ processes (except iC04) compared to CS10s.

Figures 5D–F summarizes the results from Scale-X™Hydro processes, Runs S05 and S08.

Run S08: Operated in perfusion mode with vessel volume exchange, this run showed a similar trend to iC05 and iC06, with peak titers at the second and third harvests before declining seven-fold. It exhibited the lowest cell detachment (Figure 5F) compared to other runs. Run S05: This run operated in recirculation without vessel wash at media exchange rate of 0.19 mL/cm² achieved higher total TU and maintained high productivity (TU/cm² per harvest, Figure 5D) over the process duration compared to all other processes. This is attributed to a higher media exchange volume of 0.19 mL/cm², higher oxygen mass transfer coefficients (kLa) and steady, constant LV production in daily harvests. The glucose trend (Supplementary Figure S4C) shows that glucose levels were generally lower for S05 than for iC02, which had 0.16 mL media/cm² in perfusion, while lactate trends (Supplementary Figure S4D) remained similar. This suggests that more glucose may have been metabolized in S05 towards LV production. Further pH and pCO₂ were lower in SCO5 (Supplementary Figures S4A,F). All these led to a 20% higher total TU yield/cm² (Figure 6A) in S05 compared to the best iCELLis™ Nano 4 m² run (iC02) in total TU yield (seven harvests). Additionally, Figure 6A shows that the total TU yield/cm² for both S05 and S08 is higher than for 16CS10. In conclusion, the S05 process performed better than 16CS10, iC01-iC06, and S08.

A recirculation process such as S05, is complex for scale-up and has a contamination risk due to open manipulations of the media container for harvesting. Hence it was hypothesized, if the Scale-X™ Hydro bioreactor was operated in perfusion with induction performed without a vessel wash, similar or better process yields than iC02 could be expected, potentially resulting in the appropriate process design for large scale-up for stable LV production. Due to the positive outcomes with Scale-X™ Hydro, including higher total TU yield per unit area, better mixing characteristics, and the availability of a pilot-scale size, the subsequent scale-up of Scale-X™ Hydro technology led to the investigation of the Scale-X™ Carbo 10 m². Three processes in the Scale-X™ Carbo 10 m² system (SC01, SC02, SC03) were operated in perfusion mode, with induction performed without a wash step. The comparative analysis of these runs with flatware (16CS10R2, seeded at 1.3E+04 cells/cm²) and Scale-X™ Hydro (2.4 m²) systems is illustrated in Figures 6B–D.

SC01: The initial Scale-X™ Carbo process (run SC01) demonstrated a similar TU/cm² per harvest as run S05 and a higher TU/cm² compared to 16CS10R2, although it had a significantly longer process duration. The cell detachment profiles for runs SC01, SC02, and SC03 were notably lower than in the 16CS10R2 batch (Figure 6C).

Total TU Yield: Figure 6D shows the total TU yield per run normalized by area (cm²). Using the yield from 16CS10R2 as a reference (1-fold), the Scale-X™ Carbo 10 m² SC01 achieved a 2.7-fold higher TU yield, a result consistently reproduced in SC02 and SC03. The results from runs S05 and S08 were averaged and extrapolated to a 10 m² scale, yielding a theoretical 1.7-fold higher TU yield than 16CS10R2, compared to the 2.7-fold increase observed for SC01 at pilot scale. Cell Detachment and Nuclei Counts: SC02 had more cell detachment compared to other runs in Scale-X™ Carbo, resulting in a lower count of attached cells as determined by nuclei counts at the last harvest (Supplementary Figure S5B). Nevertheless, Supplementary Figure S5B shows reproducible nuclei counts from SC01, SC02, and SC03, indicating healthy cell attachment and growth on the fixed bed throughout the process duration. Oxygenation and CO2 Trends: The pO2 trend indicates adequate oxygenation in this system (Supplementary Figure S5E). The pCO2 trends were generally higher than those of the small-scale counterpart (Supplementary Figure S5F vs. Supplementary Figure S4F), likely due to the approximately four-fold higher total cell count in the vessel at pilot scale. Glucose Availability: The higher glucose availability in all three runs, approximately 2 g/L (Supplementary Figure S5C), can be attributed to the higher media perfusion rate of 0.23 mL/cm², which is 20% higher than that in S05 (0.19 mL/cm²). In summary, the Scale-X™ Carbo 10 m² system demonstrated superior performance in terms of TU yield and cell attachment compared to the flatware and Scale-X™ Hydro systems.

This study aimed to assess the suitability of various versions of Scale-X™ Carbo 10 m² consumables as a manufacturing platform for LV production, using different stable producer cell lines generated at different times using WAS constructs 1, 2 and 3. The Scale-X™ Carbo 10 m² versions (V1, V1.4, and V2) were assessed, each featuring minor differences in stirrer configuration, manifold and head plate design, and consumable readiness for use. The process parameters remained similar, to SC01, SC02, and SC03, with minor differences in volume of media perfused during induction and harvest two onwards (26L for V2 instead of 23L for V1 and V1.4) and various seeding densities employed to ensure an induction density of >1.80E+05 cells/cm² is achieved with the various producer cell lines.

Figures 7A,B summarizes the outcomes showing comparable TU yields and similar cell detachment profiles (Figure 7C), across all runs with the three WAS constructs, confirming overall process robustness across different bioreactor versions. Processes utilizing monoclonal cell lines (Runs SC06-1′, SC06-2′, SC09″, SC10″) yielded comparable or higher TUs with respect to their corresponding polyclonal cell line-based counterparts (Runs SC03′, SC04′, SC05′, SC07″, SC08″). Metabolite profile trends for pH, glucose and lactate were consistent across all runs (Supplementary Figure S6).

Additionally, the transduction capacity of LV originating from different pilot-scale production systems was evaluated using CD34⁺ HSCs. LV produced in CellSTACKS^® (16CS10 R2) and in Scale-X™ Carbo 10 m² (SC01) show comparable CD34⁺ HSCs transduction efficiency, demonstrating a dose-response (Figure 7D). Remarkably high VCNs were achieved, ranging from 2 to 5 VCN/cell at MOI of 0.3 and 1. This outcome is highly relevant for clinical ex-vivo cell-based therapies. While LV produced from harvest 1-2 of SC01 achieved exceptionally high VCN of 4 at MOI of 0.3, for harvest 3-4, the VCN achieved was very similar to that of 16CS10 R2 harvest 1-2 and harvest 3-4 at MOIs of 0.3 and 1. This suggested that LV produced in Flatware has comparable LV quality to that in Scale-X™ Carbo 10 m², despite higher cell detachment at harvests 3-4 and potential host cell-related impurities. The Scale-X™ Carbo 10 m² process discussed in this study achieved a total of 1.13E+12 TU per run (7 harvests). With a 30% DSP yield post-purification, approximately 3.4E+11 TUs could be achieved, sufficient to treat about 140 patients with WAS at a dose of 10 MOI and 5E+06 CD34⁺ cells per kg body weight for a 50 kg individual (Tables 2, 3).