Critical Variables affecting clinical-grade production of the self-inactivating gamma-retroviral vector for the treatment of X-linked severe combined immunodeficiency

Abstract

Patients with X-linked severe combined immunodeficiency (SCID-X1) were successfully cured following gene therapy with a gamma-retroviral vector (gRV) expressing the common gamma chain of the interleukin-2 receptor (IL2RG). However, 5 of 20 patients developed leukemia from activation of cellular proto-oncogenes by viral enhancers in the long-terminal repeats (LTR) of the integrated vector. These events prompted the design of a gRV vector with self-inactivating (SIN) LTRs to enhance vector safety. Herein we report on the production of a clinical-grade SIN IL2RG gRV pseudotyped with the Gibbon Ape Leukemia Virus envelope for a new gene therapy trial for SCID-X1, and highlight variables that were found to be critical for transfection-based large-scale SIN gRV production. Successful clinical production required careful selection of culture medium without pre-added glutamine, reduced exposure of packaging cells to cell-dissociation enzyme, and presence of cations in wash buffer. The clinical vector was high titer; transduced 68–70% normal human CD34 ⁺ cells, as determined by colony-forming unit assays and by xenotransplantation in immunodeficient NOD.CB17-Prkdc^scid/J (nonobese diabetic/severe combined immunodeficiency (NOD/SCID)) and NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ (NOD/SCID gamma (NSG))) mice; and resulted in the production of T cells in vitro from human SCID-X1 CD34 ⁺ cells. The vector was certified and released for the treatment of SCID-X1 in a multi-center international phase I/II trial.

INTRODUCTION

X-linked severe combined immunodeficiency (SCID-X1) is caused by a defect in the interleukin-2 receptor gamma chain (IL2RG) that results in a severely compromised immune system.¹ In two independent clinical trials, a combined total of 20 children received autologous bone marrow-derived stem and progenitor cells that had been transduced with a gamma-retroviral (gRV) vector carrying the IL2RG cDNA.^2–4 Although these studies showed successful gene transfer with major therapeutic benefit in 18 patients, 5 patients developed T-cell leukemia between 2 and 6 years following gene therapy.^5–9 All leukemias were found to have been the result of trans-activation of surrounding cellular proto-oncogenes by the enhancer of the viral long-terminal repeat (LTR) in the integrated vector.^7,8,10,11 In response to the insertional mutagenesis events that resulted in genotoxicity in the target cell population,¹² the gRV vector for the treatment of SCID-X1 was re-designed by removing enhancers from the viral LTR that reduced the chance of insertional dysregulation of cellular genes as described.^13–16

We previously published a report on the scale-up and manufacturing of clinical-grade SIN gRV vectors by transient transfection in a closed-system bioreactor in which we identified optimized methodologies for large-scale production of high titer SIN vectors.¹⁷ However, during scale-up for the clinical production of the IL2RG vector, which was to be produced in serum-free X-Vivo media as opposed to the serum-containing media used in our original report,¹⁷ we encountered new problems. An investigation led to the identification of several additional factors that affected production of high titer virus using large-scale transient-transfection-based methodologies. Here, we described the additional process development and subsequent manufacturing of the safety-improved IL2RG gRV vector by transient transfection in support of a new international trial for SCID-X1 and describe the critical factors that affected vector titer.

RESULTS AND DISCUSSION

Human embryonic kidney (HEK)-293 cell-derived 293T cells were expanded in tissue culture flasks in Dulbecco’s modified Eagles medium (DMEM) supplemented with 10% fetal bovine serum (FBS), transfected with a gRV vector carrying the IL2RG gene using calcium–phosphate (Ca-phosphate) and plated in tissue culture flasks or on Fibra-Cel carriers as previously described.^16,17 After transfection, the media was changed to X-Vivo 10 and vector was harvested at 12 h intervals. During process development, vector transfected in DMEM in 10-layer CellSTACKs and vector transfected in X-Vivo 10 on Fibra-Cel carriers showed comparable titer (11.7 × 10⁵ versus 12.6 × 10⁵ IU ml ⁻¹, respectively; Figure 1a). However, after scale-up to 18 l, the titer of the IL2RG vector generated in X-Vivo in the wave bioreactor supplemented with Fibra-Cel, was dramatically reduced and ranged from 4.1 × 10⁴ to 6.0 × 10² IU ml ⁻¹ (harvest 1 and 2, respectively). When attempting to collect end-of-production cells from this run, less than 1% of the number of cells originally seeded (2.1 × 10⁷ of 1.2 × 10¹⁰ cells) could be recovered suggesting that the loss of titer was associated with a loss of cells.

Figure 1.

Effect of ammonia. (a) Infectious titer of the SRS11.EFS.IL2RG.pre* (GALV-pseudotyped) vector (Avg±s.d.; n =2; ns) transfected in DMEM using 10-layer CellSTACKS or X-Vivo 10 (≤3 months old) in roller bottles supplemented with Fibra-Cel. (b) Relative amount of IL2RG vector generated using 3-month- or 9-month-old X-Vivo 10 medium (Avg±s.d.; n =4; **P =0.01). (c) Level of ammonia in X-Vivo 10 medium ranging from 3 to 10 months old as determined on a Roche Diagnostics Hitachi 917 Analyzer. Control is non-ammoniagenic DMEM 10564 (Invitrogen) supplemented with 10% FBS (Avg±s.d., n =2 for 3-month- and 10-month-old samples, ***P =0.001). Dotted line indicates the published 300 μM level of ammonia toxicity in vitro. (d) Infectious titer of the SERS11.GFP.pre* (GALV-pseudotyped) vector generated in 10-month-old X-Vivo 10 medium with or without supplementation with 2 mM glutamine (Avg±s.d., n =2 for Control, n =4 for group supplemented; ns). (e) Relative amount of IL2RG vector generated in DMEM medium spiked with increasing concentrations of ammonium chloride (Avg±s.d., n =3; *P<0.05).

An investigation pointed to two important process variables, the age of the X-Vivo medium used and the use of phosphate-buffered saline (PBS) without cations for the post-transfection rinse of the cells. The same lot of X-Vivo 10 was used for most studies and the medium was within its stated expiration date. Initial development and scale-up had been done with media up to 3 months old. However, at the time of the 18 l run, the available lot of X-Vivo was 9 months old. A direct comparison of 3- and 9-month-old media showed a 65% decrease in the amount of infectious virus when generated in older medium (Figure 1b). Analysis of X-Vivo media ranging from 3 to 10 months old revealed an age-dependent increase in the level of ammonia from 634 to 1442 μM (Figure 1c). As a baseline control, we used DMEM containing a stable ‘non-ammoniagenic’ form of glutamine, GlutaMax. The elevated levels of ammonia in older medium can be explained by the breakdown of L-glutamine into ammonia and pyrrolidone-carboxylic acid, a process that occurs spontaneously over time.^18,19 Our finding is supported by a report on ammonia levels in commercially available cell media up to levels as high as 1115 μM.²⁰ Even at 3 months, the ammonia level found in X-Vivo exceeded by twofold the lowest level at which cell toxicity was observed in HL-60 cells.²⁰ Supplementation of 10-month-old X-Vivo with glutamine did not show an increase in titer (Figure 1d) suggesting that the low titer was not caused by a glutamine deficiency. However, when IL2RG vector was prepared in DMEM medium (containing GlutaMax) supplemented with increasing concentrations of ammonia, there was a clear dose effect between the level of ammonia and vector titer. Indeed, titers were significantly reduced in the presence of 1200 μM of ammonium chloride, which was comparable to the ammonia present in 9-month-old X-vivo (Figure 1e).

The second variable identified was the use of PBS with or without cations. When the decision to use serum-free medium (X-Vivo 10) was made, a rinse with PBS was inserted into the protocol to remove as much residual serum as possible from cells that had been transfected in DMEM with 10% FBS. After the rinse, cells were fed with X-Vivo 10, followed by three harvests 12 h apart. In tissue culture flasks, cells rinsed post-transfection with PBS without cations produced a titer similar to cells rinsed with PBS with cations (Figure 2a). However, when transfection was performed on Fibra-Cel, the combination of 9-month-old X-Vivo and PBS without cations resulted in a dramatic 15-fold decrease in titer (Figure 2b), replicating the loss of vector titer seen in the 18 l run. Together the data suggest that the lower titer in the 18 l pilot run was caused by a loss of cells from the Fibra-Cel matrix as a result of ammonia toxicity which was compounded by use of PBS without cations. Lastly, we observed that extended exposure of cells to TrypLESelect, a non-animal-derived enzyme used to harvest cells before transfection also reduced vector titer (Figure 2c).

Figure 2.

Effect of PBS and TrypLESelect. (a) Infectious titer of IL2RG vector generated in T-75 flasks that had been rinsed prior to media change with PBS with or without cations (Avg±s.d., n =3, ns). (b) Infectious titer of IL2RG vector (research lot) transfected in roller bottles supplemented with Fibra-Cel using 10-month-old X-Vivo 10 media and rinsed at the time of media change with PBS with or without cations (H1-H3; Vector Harvest 1–3). (c) Infectious titer of the SERS11.GFP.pre* (GALV-pseudotyped) vector after exposure of cells to TrypLESelect for 5–10 min (standard, n =4) or 30 min (extended, n =3) at the time of cell harvest immediately before transfection (Avg±s.d.; **P<0.01).

As X-Vivo 10 media that was less than 3 months old was not available at the time of good manufacturing practices (GMP) production, and based on the data above, we modified the manufacturing process to produce the clinical IL2RG vector in DMEM medium containing Glutamax, in the presence of 10% FBS in CellSTACKS rather that in X-Vivo 10 on Fibra-Cel as initially proposed. In addition, the exposure time to TrypLESelect was minimized to 5–10 min and PBS containing cations was used.

To test the level of gene transfer in an animal model prior to production of the clinical lot, a pre-clinical vector lot (vector titer 1.5 × 10⁶ IU ml ⁻¹) was produced and used to transduce primary human mobilized peripheral blood-derived CD34 ⁺ cells. Cells were plated in a colony-forming unit (CFU) assay and injected into NOD.CB17-Prkdc^scid/J (NOD/SCID) and NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ (NSG) immunodeficient mice. The data show 68.1% gene transfer in CFU in vitro and a comparable level of multi-lineage human cell engraftment of transduced and non-transduced cells in vivo indicating that exposure of cells to vector had not altered the cell differentiation potential and engraftment (Supplementary Figure 1). The average vector copy number in human cells in the bone marrow at 6 weeks post-transplantation was 0.871 and 0.598 in NOD/SCID and NSG mice, respectively, showing robust gene transfer (Table 1).

Table 1.

Gene transfer in human CD34⁺ progenitor cells in vitro and in immunodeficient mice in vivo following transduction with a pre-clinical lot of IL2RG vector

Group	Colony-forming units		NOD/SCID mice at 6 weeks		NSG mice at 6 weeks
	Analyzed	% Positive	n	Average copy number	n	Average copy number
Mock	36	0	6	0.004±0.003	8	0±0
IL2RG vector	72	68.1	10	0.871±0.075	14	0.598±0.062

Abbreviations: IL2RG, interleukin-2 receptor gamma chain; SCID, severe combined immunodeficiency. Percent gene transfer in human G-CSF-mobilized peripheral blood (MPB) CD34 ⁺ cell-derived day-14 colony-forming units (CFU) and in bone marrow from NOD/SCID and NSG mice 6 weeks after transplantation with CD34 ⁺ cells transduced with a pre-clinical lot of IL2RG gRV vector (titer 1.5 × 10⁶ IU ml ⁻¹) or exposed to cell expansion media only (Mock). Gene transfer was determined by qPCR using primers specific for wPRE, using human ApoB as loading control. Data are presented as Mean ± s.e.m.; n = number of animals analyzed; animals from which insufficient DNA was isolated were excluded from analysis.

Using the modified manufacturing process outline above, we produced a 20-l clinical lot of IL2RG gRV vector in compliance with GMP guidelines. Vector was collected using three 12-h harvests and each harvest was frozen, aliquoted and certified separately, as the fragility of the GALV envelope does not allow thawing, pooling and re-freezing of the vector without loss of titer. Infectious titers on HT1080 cells were 3.4 × 10⁶, 2.4 × 10⁶ and 8.6 × 10⁵ IU ml ⁻¹ for harvest 1, 2 and 3, respectively (Table 2), with an average titer of 2 × 10⁶ IU ml ⁻¹, as determined by qPCR. This resulted in a 64.7, 74.3 and 73.1% gene transfer in CD34 + cell-derived CFU for harvest 1 through 3. In addition, when tested in vitro, the clinical vector transduced 41% of CD34 + cells from a patient with SCID-X1 resulting in restoration of T-cell differentiation in vitro on OP-9 stromal cells expressing the human Notch ligand Delta1 (Supplementary Figure 2). The vector was subsequently certified for safety, purity and potency as previously described,²¹ and released for clinical use.

Table 2.

Clinical-grade IL2RG gRV Vector for SCID-X1 gene therapy trial

Harvest of clinical lot	H1	H2	H3
Volume (ml)	6800	6800	6800
Titer (IU ml−1)	3.4 × 106	2.4 × 106	8.6 × 105
Gene transfer efficiency in CD34 + cell-derived CFU (%)	64.7	74.3	73.1

Abbreviations: CFU, colony-forming units; IL2RG, interleukin-2 receptor gamma chain; SCID-X1, X-linked severe combined immunodeficiency. Volume, infectious titer and transduction of human G-CSF-mobilized peripheral blood (MPB) CD34⁺ cells as assessed by CFU assay, for each of the three individually certified harvests (H) of the clinical IL2RG vector. Infectious titers were determined by qPCR on transduced HT1080 cells using wPRE primers.

We show that a major variable that affected vector titers in a large-scale production run was the accumulation of high ammonia in X-Vivo 10 serum-free medium that contained glutamine. Indeed, toxic levels of ammonia are present as early as 3 months after manufacture. Based on this finding, our protocol was modified to include the use of DMEM with 10% FBS post-transfection. To comply with the requirements of the regulatory authorities and to meet the standards of the European Pharmacopoeia, the FBS used had been sourced from an approved geographical area with low risk for transmitting bovine spongiform encephalopathy and had been treated with high-dose radiation and triple filtered at 0.1 μm.

In summary, we show the successful manufacturing of a GALV-pseudotyped gRV vector containing the IL2RG gene for clinical use and outline critical factors that influenced high titer vector production. Although we have not directly tested whether the observed effect of ammonia also has a role in the manufacturing of other viral vectors, high levels of ammonia are known to be toxic to cells. Our findings are important to GMP manufacturing of current SIN retrovirus, lentivirus and foamy virus vectors, which are largely dependent on transient-transfection-based methodologies that require adaptation to large-scale production. Optimal GMP production requires careful adjustment of manufacturing variables as process development and scale-up in tissue culture flasks do not directly translate to the bioreactor.

MATERIALS AND METHODS

Plasmids

The gRV vectors SERS11.EGFP.pre* and SRS11.EFS.IL2RG.pre* and gag/pol plasmid pcDNA3.MLVg/p were constructed by Drs C Baum and A Schambach. The envelope plasmid pHCMV-GALVenv was obtained from Dr W Beyer. Plasmids (ccc-grade Classic) used for GMP manufacturing were produced by the Plasmid Factory (Bielefeld, Germany) as previously described.²²

Cell culture and media

Human embryonic kidney-293 cell-derived 293T cells (No. CRL-11268; ATCC, Manassas, VA, USA) were thawed and cultured in DMEM (10564 with GlutaMax; Invitrogen, Carlsbad, CA, USA) supplemented with 10% FBS and 1 m_M Sodium Pyruvate (Invitrogen). 293T cells were cryopreserved in DMEM supplemented with 25% (vol/vol) FBS and 10% (vol/vol) dimethyl sulfoxide (Sigma-Aldrich, St Louis, MO, USA). FBS had been sourced from North America, and had been irradiated at 25–40 kGy and triple 0.1 μm filtered. For clinical production, a GMP Master Cell Bank of 293T cells was prepared as described above and certified. X-Vivo 10 medium was derived from Lonza Walkersville (Walkersville, MD, USA). Ammonia levels were determined on a clinical Roche Diagnostics (Indianapolis, IN, USA) Hitachi 917 Analyzer.

Development and scale-up

293T cells were transfected as previously described.^16,17 Briefly, a mixture of plasmid DNA and 0.25 M CaCl₂ (Sigma-Aldrich) was added drop wise to an equal volume of 2X HEPES-buffered saline, pH 7.05 (Sigma-Aldrich), while slowly bubbling air through the HEPES-buffered saline solution. For transfection in tissue culture plastic (T-75 or CellSTACKs), optimal amounts of plasmid DNA were 10 μg of vector, 9 μg of gag/pol and 4 μg of envelope in 10 ml of media per 75 cm² of surface area. For transfection on Fibra-Cel in ridged 850 cm² roller bottles (Becton Dickinson, Bedford, MA, USA) or in the wave bioreactor (GE Healthcare, Pittsburgh, PA, USA), optimal amounts of plasmid DNA were 4 mg of vector, 3.6 mg of gag/pol and 1.6 mg of envelope per Liter. After 20 min at ambient temperature, the transfection mixture was added to the cells and cells were plated in either tissue culture flasks or on Fibra-Cel in DMEM supplemented with 10% FBS and 1 m_M sodium pyruvate in the presence of 25 μ_M Chloroquine (Sigma-Aldrich). Cells were cultured at 37°C, 5% CO₂. Transfections in the Wave bioreactor were done using a rocking speed of 22 r.p.m. and angle of 6 degrees. Nineteen hours post-transfection, cells were rinsed with PBS, the media was changed to either DMEM with 10% FBS or X-Vivo (as described) and vector was collected at 12 or 24 h intervals, filtered through a 0.45 μm filter (Millipore, Billerica, MA, USA), and stored at or below −70°C.

Clinical production

A total of 1.4 × 10¹⁰ cells from a certified 293T Master Cell Bank were expanded, harvested and transfected as described above using Ca-phosphate in eight (8) 10-layer CellSTACKS in the presence of 850 μg of SRS11.EFS.IL2RG.pre*, 765 μg of pCDNA3.MLVg/p and 340 μg of GALV envelope plasmids per CellSTACK in 850 ml of DMEM supplemented with 10% FBS and 1 m_M Sodium Pyruvate. At 19 h post-transfection, cells were rinsed with PBS with cations and fresh medium was added. Vector was harvested at ~12 h intervals, filtered through a RCQT leukocyte reduction filter (Pall, Ann Arbor, MI, USA), and aliquoted into cryocyte freezing containers (Miltenyi Biotec, Auburn, CA, USA) using a closed-system fluid path. Cryocyte bags were placed in protective freezing cassettes (Biomedical Marketing Associates, Wexford, PA, USA) and stored in quarantine at or below −70°C pending certification and release. As gRV are unstable when thawed, individual vector harvests are never pooled or re-purified, but separately stored, tested and certified.

Titration

Infectious titers of vectors containing the eGFP gene were determined in human HT1080 cells (ATCC No. CCL-121) using flow cytometry. Briefly, cells were seeded at 2 × 10⁴ cells per well in a 24-well plate and exposed for 4 h to serial dilutions of vector in a volume of 0.5 ml in the presence of 8 μg ml⁻¹ polybrene (Sigma-Aldrich). After 48 to96 h, cells were analyzed by flow cytometry for the expression of eGFP and titers (infectious units (IU) ml ⁻¹) were calculated based on the number of cells at the time of infection, the dilution and percent GFP ⁺ cells. For titration of vectors without GFP, transduced cells were carried and split for an additional 12–14 days, DNA was harvested and titer was determined by qPCR using wPRE primers as described below.

Quantitative real-time PCR

Genomic DNA was isolated from cells 12–14 days post-transduction using the Gentra Puregene Cell and Tissue Kit (Qiagen, Valencia, CA, USA) per the manufacturer’s protocol. CD34⁺ cell-derived colonies were picked and the percentage transduced colony-forming units (CFU) was determined as previously described.²³ qPCR was performed on an ABI 7900 HT sequence detection system (Applied Biosystems, Foster City, CA, USA) in a multiplexed reaction detecting wPRE and human ApoB sequences. The wPRE sequence present in the vector was detected using oligonucleotides 5′-TGTATAAATCCTGGTTGCTGTC-3′ and 5′-GCACACCACGCCACGTTG-3′, and 5′-6-FAM/TTATGAGGAGTTGTGGCCCGTTGT/3BHQ1-3′ as a probe. Human ApoB genomic sequences were detected using 5′-TGAAGGTGGA GGACATTCCTCTA-3′ and 5′-CTGGAATTGCGATTTCTGGTAA-3′ and 5′-VIC/CG AGAATCACCCTGCCAGACTTCCGT/Tamra. A linear regression was made from each standard curve to calculate the average number of vector copies per cell. Serial dilutions of genomic DNA derived from clone A5, a HT1080 cell line-derived single-copy clone containing the wPRE target sequence, was used as reference.

Transduction and progenitor assays

Human CD34⁺ cells isolated from G-CSF-mobilized peripheral blood were pre-stimulated for 2 days in X-Vivo 10 containing human recombinant interleukin-3 (20 ng ml ⁻¹), human recombinant stem cell factor (300 ng ml ⁻¹), human recombinant Flt-3 ligand (300 ng ml⁻¹) and human recombinant thrombopoietin (100 ng ml⁻¹). All cytokines were purchased from Peprotech (Rocky Hill, NJ, USA). Briefly, vector supernatant was preloaded twice onto tissue culture-treated vessels coated with Retronectin (Clontech Laboratories, Madison, WI, USA), as recommended by the manufacturer, and cytokine-stimulated CD34⁺ cells were added to the pre-loaded plates. After overnight transduction, cells were harvested and exposed to a second transduction. Cells were harvested after 4–16 h, treated with 100 U ml⁻¹ of Benzonase (Novagen, Gibbstown, NJ, USA) to remove residual plasmid,²⁴ and plated in Methocult GF (Stem Cell Technologies, Vancouver, BC, USA). Individual colonies were isolated after 14 days of culture in a humidified incubator at 5% O₂, 5% CO₂ and 37°C and analyzed by qPCR to evaluate the percent of the colonies containing integrated vector sequences as previously described.²³

Engraftment of transduced CD34 ⁺ cells in immunodeficient mice

NOD.CB17-Prkdc^scid/J (NOD/SCID) and NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ (NSG) immune deficient mice were sublethally irradiated at 300 and 280 cGy, respectively, using a Mark I-68 Cesium irradiator (JL Shepherd, San Fernando, CA, USA) and transplanted with 1.6–3.0 × 10⁶-transduced CD34⁺ cells by tail vein injection. At 6 weeks post-transplant, a femoral bone marrow aspirate was collected, red blood cells were lysed using PharmLyse Buffer (BD Pharmingen, San Diego, CA, USA), and cells were analyzed by 6-color flow cytometry after staining with anti-human CD45-APC, anti-human CD11b PE/Cy7, anti-human CD34-FITC, anti-human CD19-APC/Cy7 following the manufacturer instructions (BD Pharmingen). Before analysis, cells were stained with 7-Aminoactinomycin D (7-AAD) for dead cell exclusion. Data were acquired on a FACScanto flow cytometer (BD, Immunocytometry Systems, San Jose, CA, USA) and analyzed using FACSDiva (BD, Immunocytometry Systems) software.

Correction of SCID-X1 progenitor cells in vitro

BM-derived CD34⁺ cells from a patient with SCID-X1 were transduced with the clinical IL2RG gRV vector, and analyzed by FACS for expression of the IL2RG receptor (CD132). Transduced cells were then differentiated to T cells on OP-9 stromal cells expressing the human Notch ligand Delta1, as previously described,²⁵ and analyzed for co-expression of CD4 and CD8. CD34-APC, CD132-PE, CD4-APC, CD8-PE and Isotype controls were obtained from BD.

Statistical analysis

P-values were determined by Student’s t-test.