Certification Assays for HIV-1-Based Vectors: Frequent Passage of Gag Sequences without Evidence of Replication-Competent Viruses

ARTICLE doi:10.1016/j.ymthe.2003.08.003 Certification Assays for HIV-1-Based Vectors: Frequent Passage of Gag Sequences without Evidence of Replication-Competent Viruses Lakshmi Sastry, 1 Yi Xu, 2 Terry Johnson, 2 Kunal Desai, 2 David Rissing, 3 Jonathan Marsh, 2 and Kenneth Cornetta 1,2,3, * 1 Department of Microbiology and Immunology, 2 Department of Medical and Molecular Genetics, and 3 Department of Medicine, Indiana University School of Medicine, Indianapolis, Indiana 46202 *To whom correspondence and reprint requests should be addressed at Indiana University School of Medicine, IB-130, 975 W. West Walnut Street, Indianapolis, IN 46202. Fax: (317) 274-2387. E-mail: kcornett@iupui.edu. A principal concern regarding the safety of HIV-1-based vectors is replication-competent lentivirus (RCL). We have developed two PCR assays for detecting RCL; the first detects recombination between gag regions in the transfer vector and the packaging construct (sensitivity of detection 10 100 copies of target sequence). The second assay uses real-time PCR to detect vesicular stomatitis virus glycoprotein (VSVG) envelope DNA (sensitivity 5 50 VSVG sequences). In an attempt to amplify any RCL, test vectors were used to transduce C8166 and 293 cells, which were then screened weekly for 3 weeks. Psi gag recombinants were routinely detected (20 of 21 analyses) in four transductions using the RRL-CMV-GFP vector. In contrast, VSVG sequences were detected only once in 21 analyses. Interestingly, p24 levels (as measured by ELISA) were occasionally detectable after 3 weeks of culture. To determine if a true RCL was present, 21-day cell-free medium was used to transduce naïve cells. No evidence of psi gag or VSVG transfer was detected, indicating that the recombination events were insufficient to reconstitute a true RCL. These findings have important implications for the design and safety of HIV-1-based vectors intended for clinical applications. Key Words: RCL, lentivirus, PCR, p24 ELISA, VSVG, psi gag, recombination, safety INTRODUCTION Lentiviral vectors derived from HIV-1 have great potential as gene therapy vectors for the treatment of inherited and acquired diseases due to their ability to integrate vector sequences into nondividing cells [1,2]. Another advantage of lentiviral vectors is their ability to be pseudotyped with the vesicular stomatitis virus G protein (VSVG), which allows for concentration of the virus to high titer [3]. While this envelope protein enables the vector to transduce a wide variety of cells at high efficiency, the tendency of VSVG to form syncytia when expressed at high levels has complicated the generation of stable packaging cell lines that have been used effectively in the production of oncoretroviral vectors [4 6]. Therefore, lentiviral vectors are generally produced by transient transfection of three or more plasmids: a plasmid expressing the transfer vector (containing the gene of interest in a lentiviral vector backbone), the packaging plasmid(s) (containing gag, pol), and an envelope-expressing plasmid (e.g., VSVG) [7 10]. While the potential of lentiviral vectors has been shown for a variety of diseases, the clinical application of this vector system has been hindered by lack of effective assays for excluding inadvertent contamination of vector preparations with replication-competent lentivirus (RCL). A number of factors complicate the design of RCL testing. RCL arising from lentiviral vector systems being developed for clinical use has yet to be reported. Therefore, the RCL detection system must anticipate a currently theoretical virus. When considering the structure of a RCL, a replicating virus might arise through recombination between transfer vector and packaging construct sequences. While areas of homology between the transfer and the packaging plasmids will greatly increase the likelihood of such recombination, nonhomologous recombination is also possible. Another concern is recombination with human endogenous retroviral (HERV) sequences [11]. In terms of providing envelope function, VSVG would appear to be the most likely candidate to pseudotype a potential RCL but a number of other viral envelopes have 830 1525-0016/03 $30.00

doi:10.1016/j.ymthe.2003.08.003 ARTICLE been shown to pseudotype lentiviral vectors [12 14] and must be considered. In addition to predicting how RCL might arise, there are certain technical issues that must be considered when designing RCL assays. RCL testing is complicated by the transient transfection methods of lentiviral vector generation, which result in substantial contamination of supernatants with packaging plasmid DNA [1,15], the same viral sequences likely to be present in a RCL. Currently screening for RCL can be conducted using a variety of assays [16]. The p24 gag detection assay is most commonly used for RCL associated with HIV-1 vectors. Though this assay is simple and commercially available, it has a relatively low sensitivity ( 3 pg/ml capsid p24 protein). Furthermore, vector particles contain p24 so detection of RCL requires passage on permissive cells or some other step that will eliminate the large amount of p24 normally present in vector particles [4]. PCR assay to detect tat sequences from a packaging construct in cultures infected with lentiviral vectors has been described for bovine JDV vectors [17]. As tat is not present in many later generation lentiviral packaging constructs, this assay has limited applicability for RCL testing. Syncytia formation assays for testing vector-transfected cells, producer cells, or transduced cells have been developed using cell lines permissive for HIV-1 infection [18]. However, these assays detect a fully competent, env-containing lentivirus whose pathogenicity results in syncytia formation and may not detect the type of RCL generated with the later generation HIV-1 vectors. Marker rescue assays involving mobilization of an integrated marker provirus following infection of an indicator cell line with RCL have also been used for RCL screening but the sensitivity of these assays has not been ascertained rigorously [19,20]. To begin to evaluate RCL generation in lentiviral vector production, we chose to monitor transfer of viral gene sequences to transduced cells. In addition, the p24 ELISA was used to monitor transfer and expression of the HIV-1 gag gene. Recombination between the transfer vector and packaging plasmids was monitored using a PCR assay with primers within the transfer vector and gag sequences in the lentiviral packaging plasmid. Transfer of genetic sequences from the VSVG envelope was monitored by quantitative PCR. This three-pronged approach is designed specifically to address concerns that analysis of a single gene or gene product may miss mutant viruses that arise through recombination with endogenous retroviral sequences or result from the high mutation rates that are characteristic of lentiviruses. Using the three detection assays, we have screened for RCL in cells transduced with third-generation HIV-1 vectors [7 10,21]. Recombination between transfer vector and packaging sequences was detected consistently in cells transduced with third-generation HIV-1 vectors. Despite frequent recombinations in the psi and gag regions, sufficient recombination events required for reconstitution of a true RCL was not FIG. 1. RCL testing scheme for HIV-1 vectors. Vector supernatants are used to transduce an amplification cell line, which is maintained in culture for 21 days. The 21-day medium is then used for a second round of transduction on an indicator cell line, which is cultured for an additional 7 days. Amplification cultures at 7, 14, and 21 days and indicator cultures at 7 days are subsequently analyzed by PCR for psi gag and VSVG sequences and by ELISA for p24 gag. detected. Our findings have important implications for the safety of HIV-1 vectors for clinical applications and the combinatorial approach for RCL detection described here will help understand the nature and occurrence of RCL associated with HIV-1 vectors. RESULTS For developing RCL testing, we modeled our detection assay on FDA guidelines for detection of replication-competent retroviruses that use a permissive cell line to amplify virus over a 3-week period (amplification phase) and then analyze the culture media on an appropriate indicator cell line (indicator phase) [22]. We conducted RCL testing of HIV-1 vectors according to Fig. 1. We defined a true RCL as the mobilization of viral sequences from amplification cells to indicator cells. We also hypothesized that the shared homology of the gag start region present in both the transfer vector and the packaging plasmids would increase the likelihood of recombination and serve as a potential first step for RCL formation. Mobilization of VSVG sequences would detect both a VSVG-pseudotyped RCL and recombinant endogenous or exogenous viruses that could become pseudotyped with the VSVG envelope. The initial challenge was the development and characterization of assays used to detect transfer of viral sequences. For detecting transfer and expression of the HIV-1 gag gene we utilized a commercially available ELISA method with a sensitivity of 3 pg/ml ( 36,000 viral particles). To detect transfer of VSVG sequences and to monitor recombination between transfer vectors and gag sequences, we evaluated a variety of PCR-based assays for their sensitivity and reproducibility. PCR Assays for Detection of VSVG Sequences To develop a sensitive and reproducible method for detecting VSVG sequences in cells we compared a variety of PCR-based assays. Primers for standard PCR (VSVF2/ VSVR2 and VSVF3/VSVF4) were designed to amplify 668-831

ARTICLE doi:10.1016/j.ymthe.2003.08.003 FIG. 2. Design of PCR primers for detection of VSVG and psi gag sequences. (A) Forward and reverse primers (VSVF2, VSVF3/VSVR2,VSVR3) for normal and real-time PCR (VSVF1/VSVR1/VSVP) of VSVG sequences in the plasmid pmdg, which was used as a positive control for developing VSVG detection methods. (B) Schematic of the transfer vector and packaging constructs used in the four-plasmid lentiviral vector system. Dotted areas represent homologous regions of gag and hatched regions are homologous RRE sequences. Recombination in gag and/or RRE regions may occur to generate RCL. The forward primer GrecF1 is designed to recognize the primer binding site of the transfer vector and the reverse primers (GrecR1, GrecR2) are designed to be complementary to gag sequences specific for the packaging construct that are absent in the transfer vector. (C) phiv-gfp contains the intact psi gag region found in wild-type HIV and was used as a positive control for psi gag recombinants between transfer vectors and packaging plasmids. and 528-bp bands, respectively, from the VSVG-expressing plasmid, pmdg (Fig. 2A). We optimized PCR conditions by varying MgCl 2 concentrations and annealing and extension temperatures and PCR products were analyzed by three methods: agarose gel electrophoresis and staining with ethidium bromide, agarose gel electrophoresis and staining with SYBR green, and Southern blot analysis of the PCR product using a radiolabeled probe. The VSVF2/VSVR2 primer pair was chosen for standard PCR as it gave maximal signal and had minimal crossreactivity with genomic DNA obtained from nontransduced cells (negative control). As a positive control, 10- fold dilutions of the VSVG expression plasmid, pmdg, corresponding to 5 10 5 to 5 copies of VSVG, were prepared in a background of 0.1 g of 293 DNA. While we initially hypothesized that analysis by Southern blot would increase the sensitivity of our assay, this did not prove to be the case (Fig. 3A). The sensitivity of detection of VSVG sequences was equivalent by all three methods ( 5 copies), indicating that the kinetics of the PCR in this instance is sufficient to provide detectable signal in all three methods of detection. In addition to standard PCR, we also evaluated realtime PCR for VSVG detection using probe/primers (VSVP, VSVF1/VSVR1) designed to amplify a 120-bp region within the VSVG sequence. PCR primers were designed and chosen to obtain optimal product formation using the Primer Express software guidelines. Sensitivity of VSVG detection was also 5 copies by real-time PCR analysis, which was equivalent to that noted with standard PCR (Fig 3B). To evaluate the reproducibility of the various methods, we generated two additional independent sets of controls, which were amplified by the respective PCR methods (standard versus quantitative) and analyzed by each of the detection methods. All methods were highly reproducible in three independent experiments (Table 1). As all methods are comparable in sensitivity and reproducibility (P value 0.05 indicating insignificant variation in methods), we chose to use real-time PCR for VSVG sequence 832

doi:10.1016/j.ymthe.2003.08.003 ARTICLE FIG. 3. PCR for analysis of VSVG sequences. (A) Sensitivity of detection of VSVG sequences by normal PCR. Serial dilutions of the VSVG plasmid, pmdg, corresponding to 0 5 10 5 copies in a background of 0.1 g of uninfected DNA were amplified using normal PCR primers and products were analyzed by ethidium bromide and SYBR green gels and by probing with radiolabeled probe. Arrows indicate the predicted 668-bp band. (B) Amplification plot of the pmdg plasmid. Serial dilutions of the VSVG plasmid, pmdg, corresponding to 0 5 10 5 copies in a background of 0.1 g of uninfected DNA were amplified using VSVG probe/primers. The amplification plot shows the change in fluorescence (Rn) as a function of the PCR cycle. analysis as it is a quantitative method that provides rapid results. PCR Assay for Detection of Psi Gag Sequences Recombination of homologous sequences in the gag and rev response element (RRE) regions of the transfer vector and the packaging construct may represent a first step toward generation of RCL (Fig. 2B). We therefore designed PCR oligos that would specifically amplify sequences generated by recombination of homologous gag sequences in the transfer vector and the packaging construct (psi gag). TABLE 1: Reproducibility of psi gag and VSVG PCR Test sequence Detection method Sensitivity (No. HIV-GFP or pmdg mol/0.1 g of 293 DNA) Expt 1 Expt 2 Expt 3 psi gag PCR Ethidium bromide 10 100 10 SYBR green 10 100 10 Labeled oligo 10 100 10 VSVG Real time PCR 5 50 5 PCR Ethidium bromide 5 50 5 SYBR green 5 5 5 Labeled oligo 5 5 5 The sensitivity of detection of psi gag and VSVG sequences was determined in three independent experiments to ascertain reproducibility of the PCR methods. The forward primer (GrecF1) for psi gag PCR was designed to recognize sequences in the primer binding site of the HIV-1 vector, RRL-CMV-GFP [10]. The reverse primers (GrecR1, GrecR2) were complementary to gag sequences in the packaging construct and were not present in the transfer vector (Fig. 2B). Two reverse primers were designed to analyze the effect of amplicon size on sensitivity. The PCR primers are homologous with sequences present in most HIV-1-derived transfer vectors as well as with wild-type gag/pol packaging sequences (as determined by sequence comparison using the molecular analysis program DNA STAR). The GrecF1/GrecR1 and GrecF1/GrecR2 primer pairs amplify 537- (428-bp vector 109-bp packaging gag) and 953- (428-bp vector 525-bp packaging gag) bp bands, respectively. Real-time PCR for psi gag analysis was not conducted as the region prone to homologous recombination is too large for detection by this method. For optimizing PCR conditions and determining sensitivity of psi gag PCR, we used the plasmid phiv-gfp (Fig. 2C), containing the psi gag sequence, as a positive control. We amplified 10-fold dilutions of phiv-gfp in a background of 0.1 g genomic DNA from 293 cells corresponding to 10 5 to 1 copy using the GrecF1/GrecR1 primer pair and analyzed the products on agarose gels stained with either ethidium bromide or SYBR green. Following gel analysis, products were transferred to nitrocellulose membranes and probed with a radiolabeled gagspecific probe, GagP2 (Fig. 4). The results indicate that the sensitivity of detection of psi gag sequences is approximately 10 copies/0.1 g of 293 DNA by either ethidium bromide or SYBR green staining or by probing with a 833

ARTICLE doi:10.1016/j.ymthe.2003.08.003 FIG. 4. PCR analysis of psi gag sequences. Sensitivity of detection of psi gag sequences by ethidium bromide, SYBR green, and labeled oligo methods. Tenfold serial dilutions of phiv-gfp corresponding to 0 to 10 5 copies mixed with 0.1 g of uninfected 293 DNA were amplified with GrecF1/GrecR1 as described under Materials and Methods. Products were analyzed on ethidium bromide and SYBR green gels and by probing with radiolabeled GagP2. Arrows indicate the expected 537-bp band. Low amounts of nonspecific amplification bands at 400 bp are from genomic DNA and are not recognized by Southern blot analysis (labeled oligo lanes). radiolabeled oligo. As probing with a radiolabeled oligo also determines the specificity of the PCR product, we chose to use this method for analysis of all psi gag PCR. Amplification of the transfer vector (prrl-cmv-gfp) and packaging plasmid (pmdl) with GrecF1/GrecR1 primers showed no detectable bands, indicating no cross-reactivity with either of these plasmids. We also determined the sensitivity of psi gag PCR using the second primer pair (GrecF1/GrecR2) and it was approximately10 copies of HIV-GFP/0.1 g of 293 DNA (data not shown). Detection of psi gag sequences by PCR was reproducible in three independent experiments (Table 1). Amplification Cell Line for RCL To enhance the sensitivity of our detection assays further, we investigated the possibility of amplifying potential RCL prior to analysis by PCR and p24 detection assays. Our first goal was to identify a cell line capable of amplifying a small number of viral particles to an easily detectable amount. When adding an amplification step, we recognized that the sensitivity of RCL detection would be dependent on the infectability of the amplification cells and the rate at which RCL replication occurs. To address infectivity, we titered a stock preparation of RRL-CMV- GFP vector on a variety of cell lines (C8166, Sup-T1, CEM-SS, H9, 293, and HeLa-T4). We hypothesized that the cell with the highest titer was most susceptible to infection with lentiviral particles. Of the six cell lines tested, the highest titer was obtained in the C8166 and 293 cell lines (2.9 10 6 and 1.3 10 6 TU/ml, respectively) (Table 2). As sensitivity is likely to decrease if cell lines are relatively resistant to infection, C8166 and 293 cells appear to be the best candidates of the six cell lines tested. Since the sensitivity of RCL detection will also depend on the rate of virus replication, we selected two cell lines (C8166, HeLa-T4) susceptible to infection by an attenuated replication-competent HIV-1 virus, R7-GFP [23]. We monitored amplification by infecting the cell lines with 100-fold dilutions of the virus (10 1,10 3,10 5 ) corresponding to 18, 0.18, and 0.0018 ng/ml p24 equivalents of the virus and measuring p24 gag at 2-day intervals for a total of 8 days. The p24 values indicate an 100-fold increase in virus by day 8 (389 pg/ml) for C8166 with the 10 5 dilution of the virus (Fig. 5), whereas HeLa-T4 had a minimal increase at this viral dilution (1.0 pg/ml). Amplification was also evident with the 10 3 viral dilution in the C8166 cell line, though the 10 1 dilution showed saturating p24 levels (Fig. 5). By day 8, the p24 values at all dilutions were close to baseline levels in the HeLa-T4 cell line, indicating absence of amplification (data not shown). Based on the optimal infection and amplification kinetics, we selected the C8166 cell line for the amplification phase of our RCL assay. Having established a method for RCL amplification, we next compared the sensitivity of the PCR assay with that of the p24 ELISA in detecting R7-GFP virus. We infected C8166 cells with 1 ml of R7-GFP (TCID 50 10 5 IU/ml) at 10-fold dilutions (from 10 4 to 10 7 corresponding to 10 to 0.01 IU) and analyzed each culture by psi gag PCR and p24 ELISA after 2 weeks. At the 10 4 dilution, 5/5 replicates were positive by PCR and 6/6 were positive by ELISA. At the 10 5 dilution 3/6 and 1/6 replicates were positive for psi gag and p24, respectively. At the 10 6 dilution, 1/6 and 0/6 replicates were positive by PCR and p24, respectively, indicating that psi gag PCR can detect approximately 0.1 IU/ml of the R7-GFP virus, whereas p24 can detect 1 IU/ml of the virus. No replicates were detected at the 10 7 dilution by either method. These TABLE 2: GFP titers of cell lines transduced with RRL-CMV-GFP vector Cell line GFP titer (No. of GFP cells/ml) 293 1.30 0.01 10 6 C8166 2.90 0.05 10 6 Sup-T1 1.04 0.09 10 5 CEM-SS 4.20 0.06 10 5 H9 1.05 0.12 10 5 HeLa-T4 1.20 0.12 10 5 Cells were transduced with the RRL-CMV-GFP vector for 4 h at 37 C and analyzed by FACS analysis 72 h posttransduction. GFP titers were calculated from the % GFP-positive cells and are shown here as No. of GFP-positive cells/ml of viral supernatant. 834

doi:10.1016/j.ymthe.2003.08.003 ARTICLE FIG. 5. Amplification of R7-GFP virus. R7-GFP virus was generated by transient transfection of 293T cells and the physical titer was determined by p24 ELISA (2 10 9 viral particles/ml). C8166 cells were infected with 100-fold serial dilutions (10 1,10 3,10 5 ) of the virus corresponding to 2 10 8,2 10 6, and 2 10 4 viral particles, respectively, for 8 days, and p24 gag was measured in culture supernatants every 2 days. The p24 values at 0, 2, 4, 6, and 8 days at each infecting dilution are shown in pg/ml for the C8166 cell line. Amplification was seen in the C8166 cell line at the 10 3 and 10 5 dilutions. results suggest, not unexpectedly, that psi gag PCR appears to be as sensitive as or more sensitive than p24 ELISA at detecting R7-GFP virus. RCL Screening of HIV-1 Vectors by PCR and P24 gag Detection To evaluate recombination between vector and the gag packaging constructs and generation of RCL, we used PCR and p24 antigen detection to test two batches of unconcentrated CMV-GFP vector (CMV-GFP-1, CMV-GFP-2, titer 2 3 10 6 TU/ml) and CMV-GFP-1 after concentration (CMV-GFP-1C, titer 3.0 10 8 TU/ml). To assay for RCL, we used two dilutions of each vector corresponding approximately to 100 and 500 ng/ml p24 ( 10 5 to 5 10 5 TU/ml, respectively) to transduce C8166 or 293 cells and cultured them for 21 days. We conducted RCL testing in both cell lines, as the exact structure and infectivity of a potential RCL is unknown. The 21-day supernatant was used for a second round of transduction of C8166 or 293 cells. We analyzed the cultures by PCR and p24 ELISA at 7, 14, and 21 days into the amplification phase and 7 days into the indicator phase (Fig. 1). As shown in Table 3, cells transduced with CMV-GFP vector were consistently positive for psi gag sequences (3/3 replicates at 7, 14, and 21 days in CMV-GFP vectors) in both cell lines tested (Table 3, Fig. 6A). DNA sequence analysis of the psi gag PCR material from vector-transduced 293 cells showed vector and packaging gag hybrid sequences, confirming that these sequences were indeed generated by recombination of transfer vector and gag packaging constructs (Fig. 6B). VSVG sequences were detected very rarely in transduced cells. Analysis of cells 7 days into the indicator phase were negative for psi gag sequences, sug- Cell line Vector TABLE 3: Screening HIV-1 vectors for RCL Amplification phase Indicator phase Psi gag VSVG p24 ( 10) Psi gag VSVG p24 ( 10) 7 14 21 7 14 21 7 14 21 293 CMV-GFP-1 (1:10) cv cv cv cv (1:2) 1 0.03 1 0.4 2 0.2 cv CMV-GFP-1C (1:10,000) 1 0.04 cv cv cv (1:1000) 14 2 1 0.09 3 0.3 cv CMV-GFP-2 (1:10) 5 1 14 2 23 6 cv (1:2) 35 7 114 15 177 50 cv C 8166 CMV-GFP-2 (1:2) 104 7 350 8 65 8 8 2 293 and C8166 cells were transduced for 4 h at 37 C with test vectors at two different dilutions in triplicate and cultured for 21 days (amplification phase). Genomic DNA from transduced cells was analyzed at 7, 14, and 21 days for psi-gag and VSVG sequences by PCR. The supernatants at these time points were also analyzed by ELISA for p24 gag. The 21-day supernatants from the amplification phase were used for second round of transduction of 293/C8166 cells (indicator phase) and DNA and supernatants were analyzed 7 days posttransduction by PCR (for psi-gag, VSVG) and ELISA (for p24 gag). The p24 values are shown in pg/ml and were meaured in triplicate. The cut off value (cv) for p24 in the ELISA assay was 5-8 pg/ml. CMV vectors scored positive (33) for psi-gag. A positive VSVG signal was detected at 14 days with the psi-gag. A positive VSVG signal was detected at 14 days with the concentrated CMV-GFP vector (CMV-GFP-1C, 10-2) but was negative at 21 days and in the second transduction. 835

ARTICLE doi:10.1016/j.ymthe.2003.08.003 FIG. 6. Detection of psi gag sequences in HIV-1 vector-transduced cells. (A) 293 cells were transduced with two dilutions (1:2, 1:10) of CMV-GFP-2 vectors in triplicate and cultured for 21 days (Amplification Phase). The 21-day supernatants were used for a second round of transduction of 293 cells and cells were cultured for an additional 7 days (Indicator Phase). Genomic DNA was isolated from transduced cells at 7, 14, and 21 days following the first transduction and 7 days after the second transduction and analyzed for psi gag sequences by PCR using the GrecF1/GrecR2 primers followed by detection on ethidium bromide gels as described under Materials and Methods. Results of psi gag PCR amplification of 21-day DNA from the first transduction and 7-day DNA from the second transduction are shown. The arrow indicates the expected 953-bp band in the PCR. Genomic DNA from CMV-GFP-2-transduced cells was consistently positive for psi gag sequences. Negative controls (DNA from untransduced cells) were negative for the psi gag PCR product. (B) DNA sequence analysis of psi gag PCR. DNA from psi gag PCR was purified, cloned into a TA vector, and subjected to automated DNA sequencing. Partial sequences of parental transfer vectors (CMV-GFP) and psi gag PCR material from CMV-GFP-transduced cells at 14 (CMV-GFP psi gag 14) and 21 (CMV- GFP psi gag 21) days are shown. The psi gag sequences are a hybrid of vector (in italics) and packaging gag sequences (underlined sequences). The CMV- GFP psi gag sequences at 14 and 21 days are nearly identical. gesting that these sequences are not mobilized and therefore the signal detected in the amplification phase does not represent a true RCL. To evaluate transfer and expression of gag sequences in vector-transduced cells, we analyzed supernatants at 7, 14, and 21 days posttransduction by p24 ELISA. P24 values after the amplification phase were below cut-off range of the assay for CMV-GFP-1 (1:10 dilution) vectors, whereas CMV-GFP-1 (1:2 dilution) and the concentrated CMV-GFP-1 vectors showed higher p24 values. However, cultures from the indicator phase were generally negative for p24 (Table 3). The p24 values increased over time when cultures from CMV-GFP-2-transduced cells were tested at 7, 14, and 21 days, but 7-day cultures from the indicator phase had minimal p24 levels again, suggesting that there was no true RCL present in this test vector (Table 3). Also, during testing for RCL by p24 in the C8166 cell line, the p24 values in the indicator phase remained significantly higher than the cut-off values (Table 3, CMV-GFP2). As there were no psi gag sequences in cells from this culture, the high p24 values are most probably due to high input of p24 antigen associated with the vector particles. DISCUSSION A major safety concern associated with HIV-1 vectors is the generation of RCL. As RCL associated with HIV-1 vectors is yet to be described in experimental systems, we designed RCL assays modeled after existing methodologies used for replication-competent retroviruses. We have used a combination of PCR and p24 antigen detection methods to analyze for RCL. Our findings indicate that recombination between the transfer vector and packaging sequences occurs at low but consistently detectable levels and results in transfer of gag sequences to target cells. The PCR assays developed for this analysis proved to be sensitive and reproducible in detecting target sequences in genomic DNA. To increase the sensitivity of RCL detection further, we identified cell lines that were highly infectable and able to amplify potential RCL to high titer. Both the nonadherent C8166 T cell line and the adherent 293 cell line were identified as highly susceptible to infection by VSVG-pseudotyped lentiviral particles. C8166 was also shown to amplify an attenuated HIV-1 virus to high titer and should be suitable for detecting RCL with properties similar to those of wild-type HIV. Unfortunately, it is not known how well C8166 will amplify RCL that does not have properties similar to those of wild-type HIV. For example, it is known for oncoretroviruses that the amplification kinetics of RCRs varies with the cell type and T cells in particular amplify RCRs poorly. Given the removal of many HIV genes and regulatory regions from most current lentiviral packaging systems, the potential contribution of HERV sequences, and the unknown envelope that may be incorporated into a RCL, it may be prudent to use a second cell line when screening for RCL. As 293 cells have been shown to be highly infectable by VSVG-pseudotyped lentiviral vectors, and by a variety of other retroviral envelopes [24,30], and also have been shown to amplify other retroviruses rapidly [24,30], this cell line appears to be a suitable second cell line for use in lentiviral amplification assays. Our findings of recombination using this SIN vector are similar to those of a previous study that described recombination between a non-sin vector and the packaging construct without the formation of a fully reconstituted RCL [25]. In contrast to the frequent detection of psi gag recombinants, VSVG sequences were rarely detected in transduced cells. Unlike the VSVG plasmid, psi gag sequences contain the viral packaging (psi) sequences, which greatly increase the uptake of RNA molecules into virions. This suggests that the psi gag sequences detected were transferred to target cells by incorporation of recombinant RNA molecules in lentiviral particles (as opposed to transfer of plasmid molecules). The infrequent detection of VSVG sequences in transduced cells also suggests that HERV or other endogenous viruses are not pseudotyped with the VSVG envelope in the 293T cells used in vector production. 836

doi:10.1016/j.ymthe.2003.08.003 ARTICLE Despite the detection of psi gag recombinants, there was no evidence of transfer of psi gag or VSVG sequences from the amplification cells to indicator cells in the assay. This indicates that despite frequent recombination between the transfer vector and the packaging plasmid, the subsequent recombination events required to generate a functional virus are not occurring at detectable levels. The significance of psi gag recombination remains to be determined. Given the low but persistent detection of p24 in certain cell cultures, the possibility that the transferred gag is being expressed must be considered. While we continue to investigate this possibility, the use of SIN LTRs would predict minimal expression of the transferred sequences, although recently low amounts of residual activity from the SIN 3 LTR have been observed [26,27]. Most likely the persistently positive p24 levels represent carryover of p24 antigen given the large amounts ( 900 ng/ml) in the initial vector supernatant. The finding in one experiment of low levels of p24 in indicator cultures, without evidence of viral sequences by PCR, supports this contention. Experiments in which cultures were spiked with p24 antigen at levels similar to those found in vector did not result in persistent p24 levels (data not shown), but it is possible that the transduction process introduces p24 intracellularly, which may prolong the half-life of residual p24 antigen. If the p24 antigen detected after transduction represents carryover, there may be concerns regarding the suitability of the ELISA for detecting RCL. While attenuated wild-type HIV was shown to be rapidly amplified in C8166 cells, it is possible that a recombinant RCL may not have similar kinetics. For example, the relatively minor expansion of an attenuated HIV-1 virus we observed in HeLa-T4 might be interpreted as nonspecific carryover of p24, rather than the presence of a RCL. Since the p24 assay cannot distinguish between carryover and a RCL with slow growth kinetics, the inclusion of PCR analysis in RCL assays will decrease the likelihood of false positives while maintaining the ability to detect RCLs that expand slowly in culture. As we design an assay to detect a theoretical virus, the selection of a positive control virus remains problematic. We believe that an attenuated HIV is a reasonable measure of amplification but we are assuming it will have kinetics similar to that of a RCL. One potential criticism is that this virus contains the wild-type envelope and may act different from a VSVG-pseudotyped virus. We have chosen the attenuated HIV for two reasons. First, we believe a VSVG-pseudotyped HIV presents an unacceptable risk to laboratory workers. Second, there is no evidence that the RCL will have kinetics similar to that of a wildtype or VSVG-pseudotyped HIV-1. Therefore, one must merely set a detection standard that defines the assay. This is a reasonable approach given that amplification generally increases HIV many orders of magnitude above the detection limits of our PCR and p24 assays. Therefore, even a RCL with a much slower amplification potential is predicted to rise above the detection level at the end of the 3-week amplification phase. As laboratories determine the sensitivity of their RCL assays they must also consider the effect the final products (e.g., vector particles and media) have on the sensitivity of RCL detection. For screening clinical grade vectors, the assay should include dilution of the positive control in the final vector product when determining the sensitivity of RCL detection. In conclusion, we have used a combination method that detects multiple viral sequences for RCL testing of HIV-1 vectors. Test vectors have been negative for RCL though we show transfer of gag packaging sequences to vector-transduced cells. To certify HIV-1 vectors for clinical applications, rigorous RCL testing should be conducted using the three-pronged approach in two different cell lines described here, as the exact nature and structure of RCL associated with these vectors are unknown. It is likely that the assays will be further refined, and potentially simplified, as experience is gained with screening lentiviral vectors for RCL. Therefore these assays represent a first step toward RCL screening. The RCL certification method described here can also be adapted to certify new HIV-1 vector systems that may be designed in the future for better safety performance in the clinic. MATERIALS AND METHODS Cell lines and constructs. Human 293, Sup-T1, and CEM-SS cell lines were obtained from ATCC. C8166-45, H9, and HeLa-T4 cells were obtained from the AIDS Research and Reference Reagent Program (Rockville, MD). 293T cells used for vector production were received from Cell Genesys (Foster City, CA). The four-plasmid system containing the self-inactivating vector construct prrl-cmv-gfp, the packaging construct pmdl, the rev construct prsv-rev, and the VSVG-expressing construct pmdg was obtained from Cell Genesys. The attenuated HIV-1 construct, pr7-gfp, was obtained from the AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, NIH, from Drs. Kathleen Page, Teri Leigler, and Mark Feinberg and contains all accessory genes except nef and is replication competent [23]. The plasmid phiv-gfp was supplied by Dr. Johny J. He s laboratory (Indiana University). Vector preparation. All vectors were prepared by transient transfection of 293T cells using calcium phosphate transfection. All vectors were collected 48 h posttransfection, filtered, and stored at 70 C. Concentrated vector was prepared by ultracentrifugation for 1hat45,000 rpm using a fixed angle rotor (45 Ti) in a Beckman centrifuge (Optima XL 100K). GFP titers. 293 and HeLa-T4 cells were plated at 10 5 cells/well in six-well plates. The suspension cells C8166, H9, CEM-SS, and Sup-T1 were plated at 10 6 cells/well in a six-well plate. Serial dilutions (0, 10 1,10 2 ) of RRL- CMV-GFP-1 vector supernatant were prepared in D10 medium and incubated with cells for 4hinthepresence of Polybrene at 8 g/ml (5% CO 2, 37 C). GFP titer was determined using a FACScan (Becton Dickinson Immunocytometry Systems, San Jose, CA) as previously described [15,28]. p24 ELISA. P24 ELISA was performed using a commercially available p24 gag ELISA kit (Beckman Coulter, sensitivity 3 pg/ml). For measuring p24 in supernatants, 10 4 /10 5 dilutions of the supernatants were used. For assaying supernatants from transduced cells, 200 l was used directly. All ELISA measurements were done in triplicate. Amplification of R7-GFP. C8166 (2 10 6 /well) and HeLa-T4 (1 10 5 / well) cells were plated in six-well plates in appropriate medium and in- 837

ARTICLE doi:10.1016/j.ymthe.2003.08.003 fected in triplicate with 100-fold dilutions of R7-GFP (10 1,10 3,10 5 ) for 4 h in the presence of 8 g/ml Polybrene (5% CO 2,37 C) and cultured for 8 days. Cells were spun down and medium was changed during the 8-day period and supernatants at 0, 2, 4, 6, and 8 days were tested by p24 ELISA for amplification. For comparing psi gag PCR and p24 ELISA, C8166 cells (2 10 6 /well) were infected with 10-fold dilutions of R7-GFP (10 4 to 10 7 ) in multiples of6 and cultured for 2 weeks. Supernatants were then tested by ELISA for p24 and genomic DNA was isolated from 10 6 infected cells for psi gag PCR using the Puregene kit (Gentra, Minneapolis, MN). Psi gag PCR was performed using the primer pair GrecF1/GrecR1 using conditions described below and analyzed by transferring products to a nitrocellulose filter and probing with the radiolabeled oligo GagP2. The TCID 50 for R7-GFP virus was calculated using the Reed Muench formula [29]. PCR for detecting psi gag and VSVG sequences. Psi gag detection: PCR oligos GrecF1 (5 -CAGGACTCGGCTTGCTGAA-3 ), GrecR1 (5 -GGT- GATATGGCCTGATGTACCA-3 ), and GrecR2 (5 -TGTCTTATGTCCA- GAATGCT-3 ) and the oligo probes GagP2 (5 -GCAGCTGACACAGGACA- CAGC-3 ), GrecP (5 -AAGATTTAAACACCATGCTA-3 ) were designed using the Primer Express Software (Applied Biosystems, Foster City, CA). For determining sensitivity of psi gag detection by PCR, 10-fold serial dilutions of phiv-gfp corresponding to 10 5 to 1 copies mixed with 0.1 g of uninfected 293 DNA were amplified with GrecF1/GrecR1 and GrecF1/ GrecR2 primer pairs. The PCR consisted of 5 l of phiv-gfp and 45 l of a master mix consisting of 1 PCR buffer II, 1.5 mm MgCl 2, 200 M each dntp, 150 ng of each forward (GrecF1) and reverse primer (GrecR1/ GrecR2), and 0.025 units/ l of AmpliTaq Gold. For negative controls, 5 l of 10 mm Tris with 0.1 mm EDTA (TE), uninfected 293 genomic DNA, the vector (prrl-cmv-gfp), and the packaging plasmid (pmdl) each at 10 5 copies was used. Amplification was performed using 1 cycle of 95 C for 10 min and 40 cycles of 95 C for 1 min, 55 C for 1 min, 72 C for 1 min, and a final extension step of 72 C for 7 min. A portion of the PCR product (10 l) was analyzed on ethidium bromide (Continental Lab Products, San Diego, CA) and SYBR green gels (Roche Biochemicals, Indianapolis, IN) and visualized by UV. The PCR products on the ethidium bromide gel were then transferred to a nitrocellulose membrane overnight using 0.4 M NaOH and the membrane was probed using - 32 P-labeled GagP2 (for GrecF1/GrecR1 products) or GrecP (for GrecF1/GrecR2 products) and results were visualized by autoradiography. VSVG detection: Oligos for normal VSVG PCR, VSVF2 (5 -TTATTGC- CCGTCAAGCTCAG-3 ), VSVR2 (5 -ACTCCCCAATGCTTGCAGTAT-3 ), VSVF3 (5 -GGATGCCGAAGCAGTGATT-3 ), and VSVF4 (5 -TGGAAGAC- CCGCTCTGATTT-3 ), and real-time PCR, VSVF1 (5 -TGCAAGGAAAG- CATTGAACAA-3 ), VSVR1 (5 -GAGGAGTCACCTGGACAATCACT-3 ), and the probe VSVP (6FAM AGGAACTTGGCTGAATCCAGGCTTCC TAMRA) were designed using the Primer Express software and were obtained from ABI. For determining sensitivity of VSVG detection by PCR, 10-fold serial dilutions of pmdg corresponding to 5 10 5 to 5 copies mixed with 0.1 g of uninfected 293 DNA were amplified with VSVF2/ VSVR2. The PCR consisted of 5 l of appropriate pmdg dilution and 45 l of a master mix consisting of 1 PCR buffer II, 1.5 mm MgCl 2, 200 M each dntp, 320 nm each forward (VSVF2) and reverse primer (VSVR2), and 0.025 units/ l of AmpliTaq Gold. For negative controls, 5 lofteand uninfected genomic DNA from 293 cells was used. Amplification was performed using 1 cycle of 95 C for 10 min and 40 cycles of 95 C for 1 min, 55 C for 1 min, 72 C for 1 min, and a final extension step of 72 C for 7 min. A portion of the PCR product (10 l) was analyzed on ethidium bromide and SYBR green gels, visualized by UV, and transferred to membranes and probed with - 32 P-labeled VSVF1 as described above. For realtime PCR, 5 l of appropriate pmdg dilution was mixed with 45 l ofa master mix consisting of 1 TaqMan buffer A, 3.5 mm MgCl 2, 200 M each dntp, 320 nm each forward primer (VSVF1) and reverse primer (VSVR1), 200 nm lentiviral probe (VSVP), and 0.025 units/ l of AmpliTaq Gold. For negative controls, 5 l of TE and genomic DNA from uninfected cells were used. All reactions were carried out in triplicate and amplifications were performed using 1 cycle of 95 C for 10 min and 35 cycles of 95 C for 15 s and 60 C for 2 min. The statistical significance of the reproducibility of the methods was determined by using the one-way ANOVA method. RCL screening of HIV-1 vectors. For RCL screening, vector supernatants were used to transduce 293 or C8166 cells. For transduction, 10 5 293 cells or 10 6 C8166 cells were plated in six-well plates and 1 ml of the appropriate vector supernatant was added together with 8 g of Polybrene/ml. All transductions were performed in triplicate at 37 C. After 4 h, supernatants were removed and replaced with appropriate media and cells were cultured for 72 h and a portion of the cells was analyzed by FACS for GFP expression. Remaining cells were transferred to 75-cm 2 flasks, cultured for 21 days with weekly collection of cells (for PCR analysis) and cell-free medium (for ELISA) and cells were split when confluent. The 21-day supernatants from the amplification stage were subsequently used for a second round of transduction of 293 or C8166 cells and cultured for an additional 7 days (indicator phase). Cells and cell-free media were analyzed by PCR and ELISA at the conclusion of the indicator phase (7 days). For analyzing genomic DNA from vector-transduced and untransduced cells for VSVG and psi gag, PCR was conducted with 1 3 g of DNA. The VSVG sequences in transduced cells were compared with those in the untransduced cells by using the pair-wise t test, and sample pairs with P 0.05 were scored positive for VSVG. Samples were scored positive for psi gag by comparison with negative control untransduced DNA following Southern blotting of PCR products with appropriate probes. ACKNOWLEDGMENTS We thank Karen Pollok and Ghalib Alkhatib from IU School of Medicine for helpful comments on the manuscript and Ben Smucker for experimental help. L. Sastry is supported by the National Institutes of Health, National Research Service Award (1T32 HLO7910), Basic Science Studies on Gene Therapy of Blood Disease. The Indiana University Vector Production Facility is a NIH-designated National Gene Vector Laboratory (U42 RR11148) and this work was supported, in part, by a Core Centers of Excellence in Molecular Hematology (CCEMH) grant (PHS P50 DK49218) and a core laboratory supporting PHS P01 HL53586. RECEIVED FOR PUBLICATION MAY 26, 2003; ACCEPTED AUGUST 5, 2003. REFERENCES 1. Naldini, L., et al. (1996). In vivo gene delivery and stable transduction of nondividing cells by a lentiviral vector. Science 272: 263 267. 2. Naldini, L. (1998). Lentiviruses as gene transfer agents for delivery to non-dividing cells. Curr. Opin. Biotechnol. 9: 457 463. 3. Bartz, S. R., and Vodicka, M. A. (1997). Production of high-titer human immunodeficiency virus type 1 pseudotyped with vesicular stomatitis virus glycoprotein. Methods 12: 337 342. 4. Farson, D., et al. (2001). A new-generation stable inducible packaging cell line for lentiviral vectors. Hum. Gene Ther. 12: 981 997. 5. Kafri, T., et al. (1999). A packaging cell line for lentivirus vectors. J. Virol. 73: 576 584. 6. Rohll, J. B., et al. (2002). Design, production, safety, evaluation, and clinical applications of nonprimate lentiviral vectors. Methods Enzymol. 346: 466 500. 7. Kim, V. N., et al. (1998). Minimal requirement for a lentivirus vector based on human immunodeficiency virus type 1. J. Virol. 72: 811 816. 8. Mochizuki, H., et al. (1998). High-titer human immunodeficiency virus type 1-based vector systems for gene delivery into nondividing cells. J. Virol. 72: 8873 8883. 9. Gasmi, M., et al. (1999). Requirements for efficient production and transduction of human immunodeficiency virus type 1-based vectors. J. Virol. 73: 1828 1834. 10. Dull, T., et al. (1998). A third-generation lentivirus vector with a conditional packaging system. J. Virol. 72: 8463 8471. 11. Urnovitz, H. B., and Murphy, W. H. (1996). Human endogenous retroviruses: nature, occurrence, and clinical implications in human disease. Clin. Microbiol. Rev. 9: 72 99. 12. Landau, N. R., Page, K. A., and Littman, D. R. (1991). Pseudotyping with human T-cell leukemia virus type I broadens the human immunodeficiency virus host range. J. Virol. 65: 162 169. 13. Kobinger, G. P., et al. (2001). Filovirus-pseudotyped lentiviral vector can efficiently and stably transduce airway epithelia in vivo. Nat. Biotechnol. 19: 225 230. 14. Beyer, W. R., et al. (2002). Oncoretrovirus and lentivirus vectors pseudotyped with lymphocytic choriomeningitis virus glycoprotein: generation, concentration, and broad host range. J. Virol. 76: 1488 1495. 15. Sastry, L., et al. (2002). Titering lentiviral vectors: comparison of DNA, RNA and marker expression methods. Gene Ther. 9: 1155 1162. 16. Delenda, C., Audit, M., and Danos, O. (2002). Biosafety issues in lentivector production. Curr. Top. Microbiol. Immunol. 261: 123 141. 838

doi:10.1016/j.ymthe.2003.08.003 ARTICLE 17. Metharom, P., et al. (2000). Novel bovine lentiviral vectors based on Jembrana disease virus. J. Gene Med. 2: 176 185. 18. Chang, L. J., et al. (1999). Efficacy and safety analyses of a recombinant human immunodeficiency virus type 1 derived vector system. Gene Ther. 6: 715 728. 19. Mautino, M. R., et al. (2000). Modified human immunodeficiency virus-based lentiviral vectors display decreased sensitivity to trans-dominant Rev. Hum. Gene Ther. 11: 895 908. 20. Srinivasakumar, N., and Schuening, F. G. (1999). A lentivirus packaging system based on alternative RNA transport mechanisms to express helper and gene transfer vector RNAs and its use to study the requirement of accessory proteins for particle formation and gene delivery. J. Virol. 73: 9589 9598. 21. Zufferey, R., et al. (1998). Self-inactivating lentivirus vector for safe and efficient in vivo gene delivery. J. Virol. 72: 9873 9880. 22. Wilson, C. A., Ng, T. H., and Miller, A. E. (1997). Evaluation of recommendations for replication-competent retrovirus testing associated with use of retroviral vectors. Hum. Gene Ther. 8: 869 874. 23. Page, K. A., Liegler, T., and Feinberg, M. B. (1997). Use of a green fluorescent protein as a marker for human immunodeficiency virus type 1 infection. AIDS Res. Hum. Retroviruses 13: 1077 1081. 24. Chen, J., Reeves, L., and Cornetta, K. (2001). Safety testing for replication-competent retrovirus associated with gibbon ape leukemia virus-pseudotyped retroviral vectors. Hum. Gene Ther. 12: 61 70. 25. Wu, X., et al. (2000). Development of a novel trans-lentiviral vector that affords predictable safety. Mol. Ther. 2: 47 55. 26. Xu, K., et al. (2001). Generation of a stable cell line producing high-titer self-inactivating lentiviral vectors. Mol. Ther. 3: 97 104. 27. Logan, A. C., Lutzko, C., and Kohn, D. B. (2002). Advances in lentiviral vector design for gene-modification of hematopoietic stem cells. Curr. Opin. Biotechnol. 13: 429 436. 28. White, S. M., et al. (1999). Lentivirus vectors using human and simian immunodeficiency virus elements. J. Virol. 73: 2832 2840. 29. Payment, P. and Trudel, M. (1993). Isolation and identification of viruses. In Methods and Techniques in Virology. Dekker, New York. 30. Duffy, L, Koop, S., Fyffe, J. and Cornetta, K. (2003). Extended S /L assay for detecting replication-competent retroviruses pseudotyped with the RD114 viral envelope. Preclinica 53 59. 839