scaleable purification process for gene therapy retroviral vectors

THE JOURNAL OF GENE MEDICINE R E S E A R C H A R T I C L EJ Gene Med 2007; 9: 233–243.Published online in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/jgm.1021

Scaleable purification process for gene therapyretroviral vectors

Teresa Rodrigues1

Andreia Carvalho1

Marlene Carmo1

Manuel J. T. Carrondo1,2

Paula M. Alves1

Pedro E. Cruz1,3*

1ITQB/IBET, Av. da Republica (EAN),P-2781-901 Oeiras, Portugal2FCT/UNL, P-2825 Monte daCaparica, Portugal3ECBIO, Lab 4.11, Ed. ITQB,Apartado 98, P-2781-901 Oeiras,Portugal

*Correspondence to: Pedro E. Cruz,ITQB/IBET, Av. da Republica (EAN),P-2781-901 Oeiras, Portugal.E-mail: [email protected]

Received: 17 November 2006Revised: 9 January 2007Accepted: 24 January 2007

Abstract

Background Retroviral vectors (RVs) constitute one of the preferred genetherapy tools against inherited and acquired diseases. Development ofscaleable downstream processes allowing purification under mild conditionsand yielding viral preparations with high titer, potency and purity is criticalfor the success of clinical trials and subsequent clinical use of this technology.

Methods A purification process for murine leukaemia virus (MLV)-derivedvector supernatants was developed based on membrane separation and anion-exchange chromatography (AEXc). Initial clarification of the vector stockswas performed using 0.45 µm membranes followed by concentration with500 kDa molecular weight cut-off (MWCO) membranes; further purificationwas performed by AEXc using a tentacle matrix bearing DEAE functionalligands. Finally, concentration/diafiltration was performed by 500 kDaMWCO membranes. To validate final product quality the process was scaledup 16-fold.

Results Optimization of microfiltration membrane pore size and ultrafiltra-tion transmembrane pressure allowed the recovery of nearly 100% infectiousparticles. Further purification of the RVs by AEXc resulted in high removal ofprotein contaminants while maintaining high recoveries of infectious vectors(77 ± 11%). Up-scaling of the process resulted in high titer vector prepara-tions, 3.2 × 108 infectious particles (IP)/ml (85-fold concentration), with anoverall recovery reaching 26%. The process yielded vectors with transduc-tion efficiencies higher than the starting material and more than 99% pure,relative to protein contamination.

Conclusions The combination of membrane separation and AEXc processesresults in a feasible and scaleable purification strategy for MLV-derivedvectors, allowing the removal of inhibitory contaminants thus yielding purevectors with increased transduction efficiencies. Copyright 2007 John Wiley& Sons, Ltd.

Keywords retroviral vectors; purification; gene therapy; anion-exchangechromatography; membrane separations

Introduction

The potential for gene therapy to cure a wide range of diseases has led tohigh expectations and intensive research efforts in this field [1]. Viral vectorsare the most efficient means of delivering a corrective gene into humancells; from these, retroviral derived vectors are among the most widelyused in gene therapy clinical trials [2]. The increasing demand for highlypure retroviral vectors for application in clinical trials strengthens the needfor development of efficient and high-throughput production and purification

Copyright 2007 John Wiley & Sons, Ltd.

234 T. Rodrigues et al.

methods [3]. Here, several challenges arise, mainly dueto the inherent instability of retroviral vectors [4,5]and to the low titers produced by packaging cell lines(usually between 106 and 107 infectious particles (IP)per ml) [6]. Since clinical protocols usually requirehigh titer retroviral stocks (from 107 IP/ml for ex vivotrials up to 109 IP/ml for in vivo trials), this involvesconcentration and purification of large volumes ofsupernatant [3]. Traditional methods used for purificationand concentration of retroviral vectors (RVs), based onultracentrifugation, do not meet these challenges as theyare time-consuming, difficult to scale up and greatlyimpair biological activity [7,8].

Reports on complete purification processes, fromharvest to final polishing, are scarce. Within thecurrently available methods, membrane filtration andchromatography constitute promising candidates for theestablishment of scaleable purification processes forretroviral vectors [3]. Initial processing of producedretroviral supernatants involves the removal of cells andcellular debris; here, centrifugation and microfiltrationare frequently the methods of choice. Centrifugation hasscale limitations and low efficiency in the removal ofthe cellular debris, therefore, other methods to removethese contaminants are necessary [9]. Removal of cellsand debris can be performed using membrane filterswith 0.2 to 0.45 µm pore size [10,11]. Clarification ofretroviral supernatants using these filters is challengingas the pores often become obstructed with cellular debris,hindering further passage of the vectors and resulting inlow recoveries [12,13]. Step filtration using decreasingpore size filters (150 µm − 40 µm − 20 µm followed by aleucocyte reduction filter) has been proposed to increaseclarification recoveries up to 39–66%, compared with26–41% obtained with 0.45 µm filters [12]. However,these recoveries are still low; thus better understanding ofoperational conditions and improved clarification systemsare necessary.

Membrane ultrafiltration (UF) is currently used forconcentration, buffer exchange and partial purification ofretroviral particles. Molecular weight cut-off (MWCO)filters in the range of 20–750 kDa in combinationwith several geometries have been tested. Membranegeometries where the retentate flows tangentially to themembrane, like hollow fibers or flat-sheet cassettes, areconsidered more advantageous due to reduced foulingunder mild operational conditions allowing high permeatefluxes while maintaining viral infectivity. Althoughpromising preliminary results have been published, theinfluence of process parameters (ex. transmembranepressure and temperature) on final recoveries and vectortiters has not been addressed and UF has not beenincluded in an appropriate downstream process (DSP)for MLV-derived vectors [3]. In addition, the use ofUF membranes with high MWCO (500–750 kDa) allowsthe removal of both larger molecular weight proteinsand transduction inhibitors [14] while permitting higherpermeate fluxes. Alternative concentration methodsinclude flocculation of RVs with charged polymers, which

has also been shown to increase transduction efficiency,but has not been properly integrated in a DSP [15–17].

Concentrated retroviral stocks still require furtherpurification steps to remove contaminating proteinsand DNA from producer cells, and this can beaccomplished by using chromatographic methods [3].Several chromatographic techniques, exploring differentcharacteristics of the RVs, have been studied, namelyheparin affinity [11], streptavidin–biotin affinity (usingbiotinylated vectors) [18,19], immobilized metal affinity(using his-tagged RVs) [20], gel filtration [10,21] andanion exchange [22–24]. From these, we have studiedthe suitability of anion-exchange chromatography (AEXc)as a purification step [22]. Our results suggest thatAEXc allows milder conditions for the purification ofRVs resulting in high recoveries of infectious particlesand high purity without the need to modify the vectoritself, either by biotinylation or by his-tagging of theENV proteins. Moreover, tag removal is necessary priorto clinical use, which is usually achieved under harshdenaturing conditions or by protease cleavage introducingundesirable contaminants and process steps [25].

In spite of all these efforts, the development of acomplete DSP strategy for the purification of retroviralvectors has been poorly addressed until now. Thisstudy describes the development of a robust, scaleablepurification process based on membrane separationsand AEXc allowing the recovery of high-quality MLV-based gene therapy vectors with enhanced transductionefficiencies.

Materials and methods

Cell lines and culture conditions

Retroviral vectors were obtained from the supernatantof the human TE FLY A7 packaging cell line, derivedfrom TE 671 cells (ECACC no. 89 071 904) transformedwith pMFGSnlsLacZ plasmid [26]. This cell line producesMLV-based vectors with the 4070A envelope proteinencoding the LacZ marker gene. The HCT 116 adherentcell line was used for quantitation of infectious particles(ATCC no. CL-247). All cells were cultured in Dulbecco’smodified Eagle’s medium (DMEM, Gibco, Paisley, UK)supplemented with 4.5 g/l of glucose (Merck, Darmstadt,Germany), 6 mM of glutamine (Gibco), 5% fetalbovine serum (FBS) (Gibco) at 37 ◦C in a humidifiedatmosphere of 5% CO2 in air. For production of retroviralsupernatants DME base (Gibco) supplemented in thesame way as DMEM was used. Additionally, the HCT116 cell line was cultured in the presence of 0.1 mg/mlpenicillin/streptomycin (Gibco).

Production of retroviral vectors

For the small-scale studies retroviral supernatantswere produced in 175 cm2 T-flasks (NUNC, Roskilde,

Copyright 2007 John Wiley & Sons, Ltd. J Gene Med 2007; 9: 233–243.DOI: 10.1002/jgm

Scaleable Purification Process for Retroviral Vectors 235

Denmark). TE FLY A7 cells were seeded at a concentrationof 2 × 104 viable cells/cm2 (determined by trypanblue exclusion) in DMEM. After 48 h of growth, thesupernatant was discarded, the cells rinsed with 4 mlof phosphate-buffered saline (PBS) and the mediumreplaced by 25 ml of supplemented DME base. Twenty-four hours after medium exchange, the supernatant wascollected, filtered through 0.45 µm pore sterile filter(Starstedt, Newton, USA) to remove cells and cellulardebris, and frozen at −85 ◦C or immediately used. For thelarger-scale studies retroviral supernatants were producedusing cell factories (NUNC) with 6300 cm2 area. TEFLY A7 cells were seeded at the same concentrationused for the production in the T-flask using 1.5 l ofsupplemented DMEM. After 48 h growth the supernatantwas discarded and replaced by 900 ml of supplementedDME base. Twenty-four hours after medium exchange thesupernatant was collected and used immediately.

Quantitation of infectious particles

The concentration of retroviral infectious particles wasmeasured by the β-gal assay as described elsewhere [22].Viral titers were determined by counting between 20–200LacZ-positive (blue) cells in each well at the highestdilution. Average counts of at least three replicates wereused and errors were calculated as the standard deviationof the mean of the replicate counts.

Measurement of transductionefficiency of retroviral process samples

Transduction efficiency assays (ONPG assay) wereperformed as described elsewhere [27]. Briefly, 96-wellplates were seeded with 3.6 × 104 HCT 116 cells/well24 h before infection. Infection was performed using 20 µlof 5–250-fold dilutions of the samples in supplementedDMEM containing 8 µg/ml of Polybrene [22]. Cells ofthree non-infected wells were trypsinized and counted forfurther calculation of the multiplicity of infection (MOI).Forty-eight hours post-infection the wells were washedwith 100 µl of PBS with 1 mM MgCl2. After removingthe wash solution, the cells were incubated with 50 µlof lysis buffer (PBS with 1 mM MgCl2 and 0.5% NonidetP-40) per well for 30 min at 37 ◦C. Afterwards 50 µl oflysis buffer with 6 mM ONPG (Sigma-Aldrich, Steinheim,Germany) warmed to 37 ◦C were added per well, and theplates incubated at 37 ◦C for 15 min. The reaction wasstopped by adding 20 µl of stop buffer (1 M Na2CO3). Theoptical density at 420 nm (OD420) was measured and thenon-specific background read at 650 nm was subtracted.Values for each point are the averages of triplicate wells.

Clarification and concentration ofretroviral supernatants

Several microfiltration membranes were tested for theclarification of retroviral supernatants. Tangential flow

microfiltration of the viral supernatants was performedusing polysulfone hollow-fiber cartridges with 26 cm2 fil-tration area and 0.45 or 0.65 µm pores (MidGee CrossFlow cartridges; GE Healthcare, Buckinghamshire, UK).These cartridges were coupled to an Advanced MidJetsystem (GE Healthcare). Clarification of retroviral super-natants using Sartopore 2 polyethersulfone membranecapsules, with a pre-filter of 0.8 µm followed by a 0.45 µmfilter, was also performed.

Ultrafiltration (UF) of the viral supernatants wasperformed by tangential flow filtration using polysulfonehollow-fiber cartridges (GE Healthcare) coupled to theAdvanced MidJet system. On the small scale, 200 mlof supernatant were concentrated using a constant feedflow rate of 72 ml/min with continuous recirculationof the concentrate to the feed, using cartridges with26 cm2 filtration area. For supernatants to be used inAEXc, discontinuous diafiltration was performed whenthe volume of concentrate reached less than 10% of theinitial volume using 20 ml of 20 mM sodium phosphatebuffer with 150 mM NaCl at pH 7.5. The recirculation flowrate used with the hollow-fiber cartridges was constantand equal to 72 ml/min, with continuous recirculation ofthe concentrate to the feed until its volume reached lessthan 5% of the initial supernatant volume.

For the 16-fold scale-up study micro- and ultrafiltrationcartridges with the same configuration and length, anda total of 420 cm2 filtration area, were used with arecirculation flow rate of 624 ml/min.

Anion-exchange chromatography

Batch chromatographyDetermination of optimum pH, buffer and temperatureconditions for adsorption of the RVs was performedin batch mode, as described elsewhere [22]. Briefly,retroviral supernatants diluted with buffer, either 20 mMsodium phosphate (at pH 7.0, 7.5 and 8.0) or20 mM tris-base (at pH 7.5 and 8.0), containing60 mM NaCl and 0.5 M sucrose, were incubated with1 ml of Fractogel DEAE EMD 650 (M) media (MerckkAG, Darmstadt, Germany), previously equilibrated withbuffer, in polypropylene beakers with orbital agitation.For a period of 120 min, 200 µl samples of the supernatantof each chromatographic media and of the control weretaken, subjected to a spin-down, and the supernatantstored at −85 ◦C for later quantitation. The controlconsisted of diluted retroviral supernatant in buffersolution with no adsorbent. The results were evaluatedusing a model published by our group [22].

Column chromatographyColumn chromatography was carried out using an AktaExplorer 10 or 100 (GE Healthcare) equipped with UV andconductivity detectors and a FRAC-950 fraction collector(GE Healthcare).

For the AEXc small-scale experiments, Fractogel DEAEEMD 650 (M) media was packed into a XK16/20 column



(GE Healthcare) to a bed height of 2.8 cm and atotal volume of 5.6 ml. Before each run, the columnwas equilibrated with loading buffer (20 mM sodiumphosphate with 150 mM NaCl at pH 7.5). Then 120 mlof 2-fold diluted supernatants with loading buffer wereinjected into the XK16 column at a flow rate of 5 ml/min(2.5 cm/min). After injection, the column was washedwith 100 ml of loading buffer at the same flow rate;afterwards, elution of the retroviral vectors was performedby: (1) linear gradient, from 150 to 1500 mM NaCl at5 ml/min during 35 min; (2) step elution, with the firststep at 582 mM NaCl, second step at 1136 mM NaCl andthird step at 1500 mM NaCl.

A 16-fold scale-up of the AEXc step was performedusing a XK26/20 column (GE Healthcare) packed withFractogel DEAE EMD 650 (M) media to a bed height of20 cm and a total volume of 106 ml. The buffer volumesused in the wash and elution steps were calculated basedon the 16-fold linear scale-up factor. Similar step elutionconditions as in the small-scale experiments were usedand the column operated at a maximum flow rate of15 ml/min (2.8 cm/min). At the end of the process allfractions and the initial diluted supernatant were titeredusing the β-gal assay for determination of the infectiousparticles titer.

Gel filtration chromatography (GFC) was performedwith a Superdex 200 HR 10/30 column (GE Healthcare).Elution buffers composed by 20 mM sodium phosphate(pH 7.5) or tris-base (pH 7.8) in 150 mM NaCl wereused at a flow rate of 0.7 ml/min. The void volume ofthe column was determined using blue dextran 2000 (GEHealthcare) to be 7.4 ml. All the column chromatographyexperiments were performed at 4–6 ◦C.

Protein and DNA analysis

Protein concentration was determined with BCA andmicro BCA kits (Pierce, Rockford, USA) using bovineserum albumin (BSA) as standard protein. DNA con-centration was determined using the Quant-iT DNAassay kit (Molecular Probes Inc., Eugene, USA). Proteinanalysis by sodium dodecyl sulfate/polyacrylamide gelelectrophoresis (SDS-PAGE) was performed with 4–12%NuPAGE gradient pre-cast gels (Invitrogen, Carlsbad CA,USA) using MOPS running buffer (Invitrogen). Reducedsamples containing 3 µg total protein were resolved for50 min at a constant voltage of 200 V. Silver staining ofthe gels was performed using the PlusOne silver stainingkit (GE Healthcare). For Western blotting analysis the gelswere blotted onto a PVDF membrane for 1 h and 30 min at35 V. Detection of the viral protein bands was performedusing a swine polyclonal anti-amphoteric MLV antibody(77S000445; Viromed, Minnetonka MN, USA) diluted1 : 400 and a secondary goat HRP-conjugated anti-swineIgG antibody diluted 1 : 20 000 (Jackson ImmunoRe-search, West Grove PA, USA), followed by developmentwith the ECL Plus Western blotting detection reagents(GE Healthcare). BSA quantitation was performed with an

enzyme-linked immunosorbent assay (ELISA) kit (CygnusTechnologies Inc., Southport, NC, USA).

Negative stain electron microscopy

Negative stain electron microscopy was used to verifythe integrity of the final lot of purified vectors. Briefly,3 µl of the sample were adsorbed onto a Formvar-coated400-mesh copper grid (Electron Microscopy Sciences,USA) for 1 min. The grid was then soaked in 1% uranylacetate (Sigma) for three cycles of 10 s and dried at roomtemperature. The grids were examined with a JEM-100CXII electron microscope (JEOL, Sweden).

Results

Clarification of retroviral supernatants

Clarification of retroviral supernatants following produc-tion is necessary to remove producer cells and cellulardebris. To evaluate suitable and scaleable clarificationtechniques polysulfone hollow-fiber membranes with apore size of 0.45 µm were tested. Initial tests with thesemembranes showed very poor recoveries of infectious par-ticles (approximately 40%) that could be increased up to60–80% after several rounds of filtration with the samemembrane or after a 4-fold increase in the membranearea to supernatant volume ratio. A combination of twoevents may justify these poor recoveries of infectious par-ticles namely, adsorption of the retroviral vectors to themembrane and/or clogging of the pores with the vectorsor cell debris. Spontaneous inactivation of the vectors hasbeen excluded from these mechanisms as the processingtimes are very low (less than 10 min for 200 ml clarifi-cation tests) and mild recirculation conditions are beingused. This problem was solved using larger pore diametermembranes (0.65 µm) achieving a recovery of infectiousparticles of 116 ± 22%, or using polyethersulfone mem-brane capsules with a pre-filter of 0.8 µm and a filter with0.45 µm pores resulting in a recovery of infectious parti-cles of 97 ± 21%. This membrane was previously coatedwith 4 g/l BSA solution to prevent unspecific adsorptionof the RVs to the membrane material. Thus, entrapmentof the vectors in 0.45 µm membrane pores is negligibleand introduction of a pre-filter prevents clogging of themembranes with cell debris resulting in higher recoveriesof RVs in the filtrate.

Concentration of retroviralsupernatants

Concentration of RVs is essential to achieve high titervector stocks and decrease the volume of significantlydiluted supernatants currently produced in static cultureconditions. For this purpose the performance of polysul-fone hollow-fiber membranes with 500 kDa MWCO for



Table 1. Effect of temperature and transmembrane pressure(TMP) in the ultrafiltration process performance

21 ◦C 6 ◦C

TMPa (bar) 0.5 1.1 1.8 1.1 1.5 1.7

Duration (min) 60 54 60 156 118 165Concentrate volume (ml) 8.2 8 9 7.5 15 11.5Recoveryb (%) 41 58 70 87 101 100CFc 11 15 16 22 15 20Protein removal (%) 76 50 46 39 30 38

aTMP is defined as the difference between the average pressure of theconcentrate and the pressure of the filtrate side of the membrane.bRecovery was calculated based on the infectious particle titers of theinitial supernatant and final concentrate.cCF is the concentration factor calculated as the ratio between theconcentrate titer and initial titer.

the concentration/diafiltration and partial purification ofclarified viral supernatants was evaluated. Initially, theeffect of applied transmembrane pressure (TMP) in theUF process was studied at 21 ◦C. Evaluation of the influ-ence of the TMP in the final recovery of RVs is importantas it has been demonstrated to affect the recovery of RVsin the range of 2.5–5 psig inlet pressure using hollow-fiber membranes [28] and 3–4 bar applied pressure inother UF geometries [29]. Herein, clarified supernatantswere concentrated up to 25-fold using three differentTMPs with the volume and titer of the concentrate beingmonitored during the process. The results indicate thatthe final recovery after the UF process is influenced to agreat extent by the applied TMP (Table 1). In fact, therecovery of infectious particles, and consequently the con-centration factor of infectious particles (IPCF), increasedwith the applied TMP when comparing the three assaysperformed at a similar temperature (21 ◦C) and processduration. Analysis of the recovery of infectious particlesin the concentrate during the process indicates that inac-tivation of the RVs is not influenced by the process length,after an initial steep decay (Figure 1). As expected, titra-tion of the filtrate using the β-gal assay shows that thereis no passage of infectious particles through the mem-brane pores. Results indicate that both thermal and shearimposed inactivation of the vectors are not occurring inthis system; thus, the initial loss of retroviral infectivitycan only be explained by possible adsorption of the RVsonto the membrane. This adsorption seems to be favoredat low TMPs, where the driving force pushing virusesinto contact with the membrane and cross it is lower.Under low TMPs the build-up of the protein film at thesurface of the membrane occurs slowly; consequently, theadsorption of the retroviral vectors is favored. However,the build-up of a thicker protein film at high TMPs actsalso as a barrier to protein passage to the filtrate resultingin lower contaminant protein removal (Table 1).

Although thermal inactivation does not seem to affectthe recovery of infectious particles during the UFprocess, the influence of process temperature on retroviralrecovery, process time and protein removal was alsoevaluated (Table 1). Low process temperatures allowedvery high recoveries of infectious particles, reaching 100%

Figure 1. Influence of transmembrane pressure in the recoveriesof infectious particles during ultrafiltration of retroviralsupernatants using 500 kDa MWCO hollow-fiber membranes

at high TMPs, but at the cost of increased process timeand lower protein removal.

Purification by chromatographicmethods

Anion-exchange chromatographyWe have previously evaluated AEXc as a very promisingcandidate technique for the purification of MLV-derivedvectors [22]. In these previous studies several AEXcmatrices in batch and column mode were evaluated;diethylaminoethyl (DEAE) ligands were best suited forthe purification of these vectors. Here an AEXc matrix,Fractogel DEAE, was evaluated where the functionalligand is coupled to spacer arms composed of dextranchains, making it more appropriate for large moleculesthat cannot enter the matrix pores. Using a similarapproach, the best adsorption conditions were determinedusing retroviral supernatants buffered with tris-baseand sodium phosphate buffers at pH values between7.0–8.0, in batch mode. The kinetics of viral adsorptionwas followed by measuring the non-adsorbed infectiousparticles in the supernatant incubated with the matrix; thepercentage of infectious particles adsorbed with time wasdetermined using the model previously reported by ourgroup [22] fitted to the experimental results (Figure 2).The resulting curves show that the sodium phosphatebuffer promotes faster adsorption of retroviral particlesand results in a higher capacity of the matrix to adsorb thevectors than the tris-base buffer. The pH of the buffer alsohas a significant influence in the kinetics of adsorption ofthe vectors to Fractogel DEAE; overall sodium phosphatebuffer at pH 7.5 represents the best condition for viraladsorption to the AEXc matrix (Figure 2). Subsequently,packed column chromatography tests were performedunder these conditions and low temperature (6 ◦C), as this



Figure 2. Simulations of the percentage of adsorbed infectiousviral particles to Fractogel

TMDEAE media, obtained after fitting

the experimental results with the adsorption kinetics model[22]. The experimental data was obtained under different bufferand pH conditions at 21 ◦C

increases vector stability and promotes faster adsorptionof the RVs to the AEXc matrices [22].

To determine the elution profiles for this matrix,linear gradient elutions were performed with increasingsalt concentrations (NaCl). These profiles allow thedetermination of the separation capability of the mediaand the conductivity at which the viral vectors elute fromthe column. A very good separation between the infectiousparticles and the contaminant proteins was achieved asinfectious particles eluted in the range of 57–80 mS/cmconductivity (750–1100 mM NaCl) and the majorityof the bound protein contaminants at 15–40 mS/cm(150–530 mM NaCl) (Figure 3A). The majority of proteincontaminants did not bind to the AEXc matrix and wereseparated into the flowthrough of the column. Based onthis profile, elution of the retroviral vectors using NaClsteps was further studied to increase viral concentrationand reduce process time. Three steps were defined:582 mM NaCl to elute contaminant proteins; 1136 mM

Table 2. Results obtained for the purification of retroviral vectorsby anion-exchange chromatography using FractogelTM DEAEmatrix with gradient and step elution

Gradient Step

Fraction poola Peak fraction Fraction poola

Initial titerb (106 IP/ml) 3.4 ± 0.2 4.0 ± 0.5Pool titer (107 IP/ml) 1.25 ± 0.06 5.4 ± 0.3 1.9 ± 0.2CFc ∼4 ∼14 ∼5Volume (ml) 25 5 20Recovery (%) 71 ± 8 56 ± 7 77 ± 11Total recovery (%) 89 ± 11 81 ± 12

aPool of fractions that yield titers above 1 × 107 IP/ml.bThe initial titer corresponds to the titer of the supernatant 2-fold dilutedin buffer.cCF is the concentration factor calculated as the ratio between the pooland initial titer.

NaCl to obtain the retroviral vectors; and a third stepwith 1500 mM NaCl to wash the column (Figure 3B).Additional analysis showed that the column flowthroughdoes not contain infectious particles indicating thatthe capacity of the media is not exceeded under thestudied conditions. The results obtained are summarizedin Table 2. Purification of RVs with this matrix allowshigher recoveries (at least 56 ± 7%, depending on thedesired viral concentration) in comparison to otherchromatographic processes used for MLV-derived vectors[19,20,30]. Determination of the maximum capacity ofthe matrix to adsorb RVs was also performed by loadingan excess volume of 2-fold diluted supernatant in 20 mMsodium phosphate buffer with 150 mM NaCl at pH 7.5.The determined capacity, 2.4 × 108 IP/ml matrix, is alsohigher than for other reported matrices [3]. Variations inthe feed conductivity up to 32 mS/cm do not significantlyaffect the recovery of infectious particles after elution, asdetermined after loading retroviral supernatants 2-folddiluted with 580 mM of NaCl.

Gel filtration chromatographyGFC using a Superdex 200 column was studied as acandidate process for further purification of both UF-concentrated retroviral supernatants and AEXc-purified

Figure 3. Gradient (A) and step (B) elution profiles obtained after loading a Fractogel EMD DEAE anion-exchange column with120 ml of 2-fold diluted retroviral supernatants in 20 mM sodium phosphate with 150 mM NaCl at pH 7.5. Loading of thesupernatant and elution were performed at flow rates of 150 cm/h. Columns correspond to the infectious particle titer of eachfraction determined with the β-gal assay



Figure 4. Comparison of the S200 gel filtration elution profilesof process samples: (A) comparison of the curves using differentscales (left axis corresponds to the AEXc viral peak, and rightaxis to the concentrated supernatant by UF and the AEXccontaminants peak); (B) amplification of the shaded area of(A) and comparison of the three chromatograms using the samescale. Samples with a volume of 0.5 ml were injected into thecolumn and eluted at a flow rate of 0.7 ml/min using 20 mMsodium phosphate buffer with 150 mM NaCl at pH 7.5

vectors. This specific matrix was chosen because the viralparticles are excluded from the matrix pores, elutingin the void volume of the column (around 8 ml afterinjection of the sample); thus, separation from proteinswith less than 669 kDa (exclusion limit for globularproteins as specified by the manufacturer) is possiblewithout excessive dilution of the vectors. S200 elutionprofiles were generated after injection of 0.5 ml of peakfractions at 0.7 ml/min, using 20 mM sodium phosphate

buffer with 150 mM NaCl at pH 7.5 (Figure 4). AEXc-purified vectors elute as one peak starting at the columnvoid volume indicating that the RVs purified by AEXchave already very high purity. Recovery of AEXc-purifiedvectors after S200 gel filtration is in the range of 45–60%of the injected infectious particles but with a dilutionfactor in the range of 5–12. The buffer used for elutionin gel filtration was either 20 mM tris-base or sodiumphosphate with 150 mM NaCl at the pH used in AEXc.The type of buffer had no significant effect on the recoveryof the retroviral vectors after gel filtration.

Process development

Downstream process (DSP) development involves thestudy of a combination of purification steps having in mindthe optimization of the most important process variables,e.g. final recovery, infectious particle titers and processcontaminant removals. Membrane concentration, AEXcand GFC have been presented as potential candidates forthe development of a DSP strategy for retroviral vectors.Potential combinations of these processes are presentedin Figure 5. As discussed above, purification by GFC doesnot add significant improvement to the quality and titer ofthe vectors. Thus, process 1 is not an interesting option; itwould result in a lower final recovery without significantaddition to product quality. The downstream processes 2and 3 were analyzed in detail and the results obtainedare shown in Table 3. The difference between these twooptions is the introduction of a concentration/diafiltrationstep of the clarified vector supernatant prior to an AEXcpurification step in process 3. The final recovery ofinfectious particles for process 3 is 20% higher thanfor process 2, at equal final concentration factor. Thisdifference occurs due to a decrease in the recovery afterAEXc in process 2, compared with recoveries previouslydemonstrated. This decrease is probably related with anincreased loading time in the AEXc step (around 80 minfor process 2 and 10 min for process 3) that might haveaffected the infectivity of the retroviral vectors given

Figure 5. Possible combinations of the processes studied



their short half-life. Neither of the two processes attestedposed capacity limitation effects for the AEXc column sinceno infectious particles are detected in the flowthrough.The concentration step in process 3 reduced the proteincontamination by a further 30%. Thus, inclusion of thisconcentration step prior to AEXc is advantageous as itallows working with smaller columns and lower flow ratesprotecting retroviral vectors from high backpressures andshear stresses that might affect their infectivity uponscaling. In addition, the high MWCO of the membranesused resulted in the removal of around 63% of the totalDNA contamination present in the supernatants leading tohigher purification factors when using process 3 (Table 3).

A 16-fold scale-up of the proposed downstream strategy(process 3, Figure 5) was performed in order to obtaina sufficient volume of AEXc-purified vectors to beconcentrated/diafiltered by UF and further analyzed interms of final quality. A total of 3.6 l of vector supernatantwas produced in four NUNC cell factories and processedto a final volume of 10 ml of purified vectors in 6 h. Thecriteria used for scale-up are described in the ‘Materialsand methods’ section for each process step; optimumprocess conditions were maintained during scale-up.

Table 3. Comparison of the combination of different down-stream processes for retroviral vectors

Productiona Ultra/diafiltrationb AEXc Final

Process 2IP (106 IP/ml) 2.3 ± 0.2 – 20 ± 1 20 ± 1CF 1 – 8.8 8.7Recovery (%) 100 – 43 ± 5 43 ± 5Volume (ml) 200 – 10 10103 IP/µg protein 0.97 – 1100c 1100c

106 IP/µg DNA 4.9 – 7.9c 7.9c

Process 3IP (106 IP/ml) 5.7 ± 0.3 82 ± 9 50 ± 2 50 ± 2CF 1 14.4 0.6 8.8Recovery (%) 100 101 ± 12 61 ± 7 62Volume (ml) 200 10 10 10103 IP/µg protein 2.5 2.6 2200c 2200c

106 IP/µg DNA 8.5 23 53 53

aData relative to clarified supernatants using 0.45 µm membranes.bThe supernatant was concentrated to 14 ml final volume but only 10 mlwere further purified by AEXc.cResults presented are relative to the peak fraction.

Table 4. Results obtained for the scale-up of process 3

Process stepTiter

(106 IP/ml) CFaRecovery

(%)103 IP/µg

protein106 IP/µgDNA

Production 3.7 1 – 1.6 4.9Microfiltrationb 2.8 1 73 ± 6 1.3 3.51st Ultra/diafiltration 64 17 77 ± 4 3.7 11AEXc 88 24 71 ± 3 1400 122nd Ultra/diafiltration 320 85 66 ± 4 1800 10Final product 320 85 26 ± 6 1800 10

aCF is the concentration factor calculated as the ratio between thetiter of each step in terms of infectious particles and the titer of theproduction lot.bMicrofiltration was performed using 0.45 µm pore hollow-fibermembranes.

The results obtained are in agreement with thesmall-scale results, in terms of titers, recoveries andpurity regarding contaminant protein removal (Table 4).Regarding DNA removal, the up-scaled results show thatthe final DNA specific titer is somewhat affected by theinitial contaminant DNA concentration, comparing theresults obtained for process 3 in Tables 3 and 4. Thisissue is currently being addressed by our group.

Quality assessment of purified vectors

To analyze the vectors purified after the scale-up ofprocess 3 several techniques were used, including gelfiltration, electron microscopy, SDS-PAGE with silverstaining and Western blotting. Analysis of the final productby S200 GFC shows no contamination of the vectorswith small molecular weight contaminant proteins, as thevectors elute in the void volume of the column as one peak(Figure 6). Contaminant proteins with molecular weightsbelow 669 kDa would elute at elution volumes higherthan 10.2 ml (determined by calibration of the column).Electron microscopy photographs confirm the integrity ofthe purified vectors (Figure 7).

Figure 6. Analysis of 0.5 ml of the final purified vectors using aS200 gel filtration column

Figure 7. Negative staining electron microscopy photograph ofthe final lot of purified vectors by process 3 showing completelyenveloped viruses



Figure 8. Analysis of process samples by SDS-PAGE with silver staining (left) and Western blotting using a polyclonal antibodyagainst amphoteric MLV (right). Each gel was loaded with 3 µg of total protein per well. Legend: A, MW markers (kDa); B, retroviralsupernatant; C, 1st ultrafiltration concentrate; D,1st ultrafiltration filtrate; E, AEXc contaminant peak; F, AEXc viral peak; G, finalproduct; H, ultracentrifugation purified vectors

Protein patterns of process samples and purifiedproduct were further analyzed by SDS-PAGE and Westernblotting. RVs have many proteins in their composition,most of them in very small amounts, some of whichderived from the producer cell line [31]. Nevertheless,silver staining of samples recovered throughout theprocess (Figure 8) shows that major protein contaminantbands, present in lanes B to E, are not present orare greatly reduced for vectors purified by AEXc andfinal product (lanes F and G, respectively). Westernblotting analysis using a polyclonal amphoteric murineleukaemia virus (MLV) antibody (that reacts with severalviral proteins and their precursors) supports the resultsobtained with silver staining as it is possible to see theviral protein pattern in lanes F and G as opposed tothe starting material (lane B) heavily contaminated withserum proteins. Furthermore, the concentration of BSAin the final purified vector preparation was determinedby ELISA to be 0.9 µg/ml (∼0.5% of the total protein)indicating very low residual contamination levels in thefinal purified vectors. Since this is a non-lytic productionsystem, the contamination with cell-derived proteins isrelatively low and the major contaminant proteins areserum proteins, mainly BSA. Using BSA concentration asan indicator of absolute purity, we can infer that thepurified RVs are more than 99% pure.

Finally, the transduction efficiency of the purified viralvectors was assessed using HCT 116 cells (Figure 9).The final purified vectors show increased transductionefficiency as compared with process samples obtainedafter the first UF and after purification by AEXc.

Discussion

Downstream processing of complex macromolecularstructures, like viruses, is currently one of the mainchallenges in this field [32]. Most of these structuresare very fragile, as in the case of MLV-derived vectors,reducing the range of applicable processes. In this studywe have investigated the applicability of a combination ofmembrane and chromatographic methods to purify MLV-derived retroviral vectors. Suitable operational conditionswere initially tested and optimized for each method,

Figure 9. Effect of the multiplicity of infection (MOI) in thetransduction efficiency of process samples and final prod-uct: (�) retroviral supernatants concentrated by ultrafiltra-tion; (°) retroviral vectors after purification by AEXc; (�)AEXc-purified vectors after concentration/diafiltration. Trans-duction efficiencies were measured using the β-gal ONPG assay

having as goals the optimization of the recovery ofinfectious particles, titer and product quality.

Clarification of the supernatants using membranesshould be straightforward, as RVs are secreted byanchorage-dependent cells, thus few cells and cell debrisare present. However, significant loss of titer occurredwhen using scaleable membrane geometries (hollow-fibermembranes) with the widely used 0.45 µm pores. Thepotential causes for these low recoveries were investigatedand it was found that they mainly occur due to clogging ofthe 0.45 µm membrane pores with cellular debris. Thus,increasing the membrane pore size or using membraneswith a higher pore pre-filter should result in completerecovery of the retroviral vectors.

Ultrafiltration, using large MWCOs, was also evaluatedas a candidate process for concentration and partialpurification of retroviral supernatants and purified vectorsafter anion-exchange chromatography; the effect oftransmembrane pressures assumed during UF upon therecovery of infectious viral vectors was studied. Theincrease in filtration pressure has been described to affectretroviral infectivity during concentration using stirred



cells, where working pressures in the range of 3–4 barwere used under a nitrogen atmosphere [29]. Paul et al.[28] also demonstrated that the recovery of infectiousparticles after UF decreases significantly from around80% to 8–9% by increasing the hollow-fiber inlet pressurefrom 2.5 psig (0.17 bar) to 5 psig (0.34 bar), at roomtemperature. Here we have found that higher recoveriesare achieved when increasing the inlet pressure of thehollow fibers, in the 0.5–2 bar range (Table 1); theserecoveries are further increased when working at lowtemperature. There is not enough information to discusswhy this inverse effect with inlet pressure occurs in thestudied range, as the reported systems and conditionsare slightly different from those presented here. Highapplied TMPs and low process temperature have thedisadvantage of increasing process time and decreasingcontaminant protein removal to the filtrate; however,they provide higher recoveries of the target product.Unspecific adsorption of the vectors to the membranesmay be reduced by pre-coating the membranes with aconcentrated solution of BSA. The recirculation flow ratesmay have a more significant effect on the infectivity of thevectors than the applied inlet pressures, but no studies onthis subject have been reported yet.

AEXc has been widely used for the purification ofseveral types of viruses. However, till recently it was notconsidered a good purification method for oncoretroviralvectors, due to the low recoveries of infectious particlesattributed to the high salt concentrations required forelution of the vectors [3]. Our group has previouslyreported the interaction of retroviral vectors with severalAEXc matrices and found that weak anion exchangers,like DEAE ligands, offer the best binding and elutionconditions for these vectors, with recoveries reaching53 ± 13% [22]. Thereby, an AEXc matrix with superiorbinding capacity due to the use of dextran tentaclesas spacer arms between the matrix and the DEAEligands was used. Herein, recovery of infectious particlesusing this Fractogel DEAE matrix is higher (up to77 ± 11%, depending on the desired concentrationfactor, Table 2) than other chromatographic processespublished to date for oncoretroviral vectors, ex. heparinaffinity (61.1 ± 2.1%) [11], immobilized metal affinitychromatography (IMAC, 56%) [20], streptavidin–biotininteraction (16.7%) [33], CHT hydroxyapatite (18%)[30], or other AEXc matrices tested by our group [22].

The inclusion of an additional step for bufferexchange and concentration is usually necessary foreach of the mentioned chromatographic methods. Bufferexchange is usually accomplished by gel filtration orultra/diafiltration. After assessing the purity of AEXc-purified vectors it was confirmed that gel filtration wouldnot be necessary for further polishing of the vectors(Figure 4). Moreover, undesirable dilution of the viralpool occurs after gel filtration. A higher concentrationcould be obtained by using UF hollow fibers with 500 kDaMWCO to concentrate the viral pool obtained afterscaling up the initial steps of the process (clarification,

concentration and purification by AEXc) and removingthe salt from the AEXc eluate.

The results for the up-scaling (Table 4) show similarstep recoveries when compared with the small-scaleexperiments (Table 3), proving the scalability of theproposed process. The difference in the final recoveryis due to the inclusion of the microfiltration andsecond ultrafiltration step recoveries. Final recoveriesare comparable with the best process results publishedto date [10,11]; moreover, higher concentration factorscould be achieved. Additionally, the implementation of amore efficacious clarification step (using microfiltrationmembrane capsules having a 0.8 µm pre-filter followedby a 0.45 µm filter) would further increase the finalrecovery and titer of the purified retroviral vectors, byincreasing the recovery of this step from 73 ± 6% upto nearly 100% (Table 4), resulting in a final recoveryof approximately 36%. In principle, the process describedherein can be used to purify several types of RVs. However,some variability in the surface charge and composition ofthe vectors (lipid and envelope pseudotyping) derivedfrom the different cell lines used to produce the vectorsis expected [5]. These variations are also expected toinfluence vector stability [5,34] potentially having a greatimpact on the final recoveries and titer.

The quality of the vectors purified by the proposedprocess is very high, comparable or superior to thoseobtained with other DSP strategies for these vectors [3].Confirmation of the very low concentration of a proteincontaminant indicator (BSA) in the purified vectorssupports the applicability of this process in clinical humantrials. According to the limit of 100 µg complementingcell protein per dose specified by the FDA [35], up to100 ml of the purified vectors could be administered to apatient, based on the BSA content of the final preparation(0.9 µg/ml).

The high final vector titer and transduction efficiencyalso allow the use of up to 100-fold lower volumes of thepurified vectors, compared with vector doses currentlyused in clinical trial protocols [3]. The resulting purifiedvectors show increased transduction efficiency, comparedto the vectors just concentrated by UF or even afterpurification by AEXc. This suggests that the presentedprocess is also able to remove transduction efficiencyinhibitors that are usually secreted by the producer cells[15,16,36]. The increase in transduction efficiency isquite significant at low MOI (<5, Figure 9). The MOIhas been demonstrated to be directly correlated withthe occurrence of multiple integration events in thetarget cells [37]. Therefore, the purified vectors areable to transduce the target cells more efficiently underconditions where the risk of malignant transformation islower.

Overall the scaleable purification process describedherein enables the obtention of high quality retroviralvectors exhibiting a titer, purity and potency compatiblewith clinical trial requirements.



Acknowledgements

We wish to thank Cristina Peixoto for technical support andhelpful discussions during the production of this manuscript,Ana Alves for her help with the quality analysis, and Sarto-rius A.G. for offering the microfiltration capsules. The authorsalso acknowledge the financial support received from thePortuguese Foundation for Science and Technology (THER-AVECT, FCT/POCI/A003/2005, POCTI SFRH/BD/13771/2003)and from the European Commission (Clinigene Network of Excel-lence, LSHB-CT-2006-018933).

References

1. Mountain A. Gene therapy: the first decade. Trends Biotechnol2000; 18: 119–128.

2. Gene Therapy Clinical Trials Worldwide, 2006; Available:http://www.wiley.co.uk/genmed/clinical/, last accessed Octo-ber 2006.

3. Rodrigues T, Carrondo MJ, Alves PM, Cruz PE. Purification ofretroviral vectors for clinical application: Biological implicationsand technological challenges. J Biotechnol 2007; 127: 520–541.

4. Layne SP, Merges MJ, Dembo M, et al. Factors underlyingspontaneous inactivation and susceptibility to neutralization ofhuman immunodeficiency virus. Virology 1992; 189: 695–714.

5. Beer C, Meyer A, Muller K, Wirth M. The temperature stabilityof mouse retroviruses depends on the cholesterol levels of virallipid shell and cellular plasma membrane. Virology 2003; 308:137–146.

6. Merten OW. State-of-the-art of the production of retroviralvectors. J Gene Med 2004; 6(Suppl 1): S105–124.

7. Powell SK, Kaloss MA, Pinkstaff A, et al. Breeding of retrovirusesby DNA shuffling for improved stability and processing yields.Nat Biotechnol 2000; 18: 1279–1282.

8. Burns JC, Friedmann T, Driever W, Burrascano M, Yee JK.Vesicular stomatitis virus G glycoprotein pseudotyped retroviralvectors: concentration to very high titer and efficient genetransfer into mammalian and nonmammalian cells. Proc NatlAcad Sci U S A 1993; 90: 8033–8037.

9. Kempken R, Preissmann A, Berthold W. Clarification of animalcell cultures on a large scale by continuous centrifugation. J IndMicrobiol 1995; 14: 52–57.

10. Transfiguracion J, Jaalouk DE, Ghani K, Galipeau J, Kamen A.Size-exclusion chromatography purification of high-titervesicular stomatitis virus G glycoprotein-pseudotypedretrovectors for cell and gene therapy applications. Hum GeneTher 2003; 14: 1139–1153.

11. Segura MM, Kamen A, Trudel P, Garnier A. A novel purificationstrategy for retrovirus gene therapy vectors using heparin affinitychromatography. Biotechnol Bioeng 2005; 90: 391–404.

12. Reeves L, Cornetta K. Clinical retroviral vector production: stepfiltration using clinically approved filters improves titers. GeneTher 2000; 7: 1993–1998.

13. Przybylowski M, Hakakha A, Stefanski J, Hodges J, Sadelain M,Riviere I. Production scale-up and validation of packaging cellclearance of clinical-grade retroviral vector stocks produced incell factories. Gene Ther 2006; 13: 95–100.

14. Le Doux JM, Morgan JR, Yarmush ML. Removal ofproteoglycans increases efficiency of retroviral gene transfer.Biotechnol Bioeng 1998; 58: 23–34.

15. Landazuri N, Le Doux JM. Complexation with chondroitinsulfate C and Polybrene rapidly purifies retrovirus frominhibitors of transduction and substantially enhances genetransfer. Biotechnol Bioeng 2006; 93: 146–158.

16. Landazuri N, Gupta M, Le Doux JM. Rapid concentration andpurification of retrovirus by flocculation with Polybrene. JBiotechnol 2006; 125: 529–539.

17. Zhang B, Xia HQ, Cleghorn G, Gobe G, West M, Wei MQ. Ahighly efficient and consistent method for harvesting large

volumes of high-titre lentiviral vectors. Gene Ther 2001; 8:1745–1751.

18. Hughes C, Galea-Lauri J, Farzaneh F, Darling D. Streptavidinparamagnetic particles provide a choice of three affinity-basedcapture and magnetic concentration strategies for retroviralvectors. Mol Ther 2001; 3: 623–630.

19. Williams SL, Nesbeth D, Darling DC, Farzaneh F, Slater NKH.Affinity recovery of Moloney Murine Leukaemia Virus. JChromatogr B Biomed Sci Appl 2005; 820: 111–119.

20. Ye K, Jin S, Ataai MM, Schultz JS, Ibeh J. Tagging retrovirusvectors with a metal binding peptide and one-step purificationby immobilized metal affinity chromatography. J. Virol 2004;78: 9820–9827.

21. McGrath M, Witte O, Pincus T, Weissman IL. Retroviruspurification: method that conserves envelope glycoprotein andmaximizes infectivity. J Virol 1978; 25: 923–927.

22. Rodrigues T, Carvalho A, Roldao A, Carrondo MJT, Alves PM,Cruz PE. Screening anion-exchange chromatographic matricesfor isolation of onco-retroviral vectors. J Chromatogr B AnalytTechnol Biomed Life Sci 2006; 837: 59–68.

23. Richieri SP, Bartholomew R, Aloia RC, et al. Characterizationof highly purified, inactivated HIV-1 particles isolatedby anion exchange chromatography. Vaccine 1998; 16:119–129.

24. Yamada K, McCarty DM, Madden VJ, Walsh CE. Lentivirusvector purification using anion exchange HPLC leads toimproved gene transfer. Biotechniques 2003; 34: 1074–1078,1080.

25. Arnau J, Lauritzen C, Petersen GE, Pedersen J. Currentstrategies for the use of affinity tags and tag removal for thepurification of recombinant proteins. Protein Expr Purif 2006;48: 1–13.

26. Pizzato M, Merten OW, Blair ED, Takeuchi Y. Development ofa suspension packaging cell line for production of high titre,serum-resistant murine leukemia virus vectors. Gene Ther 2001;8: 737–745.

27. Morgan JR, Lee J, Tompkins RG, Yarmush ML. Rapidquantitation of recombinant retroviruses. Biotechnol Prog 1994;10: 441–446.

28. Paul RW, Morris D, Hess BW, Dunn J, Overell RW. Increasedviral titer through concentration of viral harvests from retroviralpackaging lines. Hum Gene Ther 1993; 4: 609–615.

29. Cruz PE, Goncalves D, Almeida J, Moreira JL, Carrondo MJ.Modeling retrovirus production for gene therapy. 2. Integratedoptimization of bioreaction and downstream processing.Biotechnol Prog 2000; 16: 350–357.

30. Kuiper M, Sanches RM, Walford JA, Slater NK. Purification ofa functional gene therapy vector derived from Moloneymurine leukaemia virus using membrane filtration and ceramichydroxyapatite chromatography. Biotechnol Bioeng 2002; 80:445–453.

31. Hammarstedt M, Wallengren K, Pedersen KW, Roos N,Garoff H. Minimal exclusion of plasma membrane proteinsduring retroviral envelope formation. Proc Natl Acad Sci U S A2000; 97: 7527–7532.

32. Morenweiser R. Downstream processing of viral vectors andvaccines. Gene Ther 2005; 12(Suppl 1): S103–110.

33. Williams SL, Eccleston ME, Slater NK. Affinity capture of abiotinylated retrovirus on macroporous monolithic adsorbents:towards a rapid single-step purification process. BiotechnolBioeng 2005; 89: 783–787.

34. Carmo M, Faria TQ, Falk H, et al. Relationship betweenretroviral vector membrane and vector stability. J Gen Virol2006; 87: 1349–1356.

35. O’Keeffe RS, Johnston MD, Slater NK. The affinity adsorptiverecovery of an infectious herpes simplex virus vaccine. BiotechnolBioeng 1999; 62: 537–545.

36. Le Doux JM, Morgan JR, Snow RG, Yarmush ML. Proteoglycanssecreted by packaging cell lines inhibit retrovirus infection. JVirol 1996; 70: 6468–6473.

37. Kustikova OS, Wahlers A, Kuhlcke K, et al. Dose finding withretroviral vectors: correlation of retroviral vector copy numbersin single cells with gene transfer efficiency in a cell population.Blood 2003; 102: 3934–3937.


scaleable purification process for gene therapy retroviral vectors

Documents