Junonia coenia Densovirus-Based Vectors for Stable Transgene Expression in Sf9 Cells: Influence of the Densovirus Sequences on Genomic Integration

Abstract

The invertebrate parvovirus Junonia coenia densovirus (JcDNV) shares similarities with terminal hairpins and nonstructural (NS) protein activities of adeno-associated virus (AAV) despite their evolutionary divergence (B. Dumas, M. Jourdan, A. M. Pascaud, and M. Bergoin, Virology, 191:202-222, 1992, and C. Ding, M. Urabe, M. Bergoin, and R. M. Kotin, J. Virol. 76:338-345, 2002). We demonstrate here that persistent transgene expression in insect cells results from stable integration of transfected JcDNV-derived vectors into the host genome. To assess the integrative properties of JcDNV vectors, the green fluorescent protein (GFP) gfp marker gene was fused in frame into the major open reading frame (ORF1) of the viral sequence under the control of the P9 capsid protein promoter. In addition, the influence of the nonstructural proteins on the posttransfection maintenance of the vectors was examined by interruption of one or all three NS ORFs. Following transfection of Sf9 cells with each of the JcDNV constructs, clones showing persistent GFP expression were isolated. Structural analyses revealed that the majority of the JcDNV plasmid sequence was integrated into the genome of the fluorescent clones. Integration was observed whether or not NS proteins were expressed. However, the presence of NS genes in the constructs greatly influenced the number of integrated copies and their distribution in the host genome. Disruption of NS genes expression resulted in integration of head-to-tail concatemers at multiple sites within the genome. Further analyses demonstrated that the cis JcDNV 5′ inverted terminal repeat region was the primary site of recombination. Sequence analyses of integration junctions showed rearrangements of both flanking and internal sequences for most integrations. These findings demonstrate that JcDNV vectors integrate into insect cells in a manner similar to AAV plasmids in mammalian cells.

Densoviruses (DNVs) are pathogenic, autonomously replicating parvoviruses that have been isolated from arthropod hosts, including insects of medical and economic importance, as well as Crustaceae (5, 14). As with vertebrate parvoviruses, the DNV genome consists of a single-stranded linear DNA, 4 to 6 kb in length, containing two sets of coding sequences, one for capsid proteins (VP) and one for nonstructural proteins (NS). densoviruses also contain self-priming hairpins at both ends of the genome, which enable the host cell DNA polymerase to convert the single-stranded genome into double-stranded DNA, a step required for transcription of NS proteins by host cell RNA polymerases and for viral replication.

Based on their genome size, organization of coding sequences, and structure of extremities, densoviruses are distributed into three genera, Densovirus, Iteravirus, and Brevidensovirus within the subfamily Densovirinae (6, 7). Members of the genus Densovirus along with Junonia coenia densovirus (JcDNV) as the prototype (12) possess a 6-kb genome with an ambisense organization, i.e., the major open reading frame (ORF1) encoding four capsid proteins is located in the 5′ half on one strand, whereas the three nonstructural proteins are encoded by three ORFs (ORF2, ORF3, and ORF4) located in the 5′ half on the complementary strand. These strands have large inverted terminal repeats (ITRs) that exceed 500 nucleotides and include the P9 and P93 viral promoters.

The wide distribution of densoviruses among insect species has triggered interest for their potential use as biocontrol agents against agricultural insect pests and vectors of human diseases such as mosquitoes (5). The potential of recombinant densoviruses for gene transfer in insects has also been investigated, because densoviruses share many characteristics with vertebrate parvoviruses, which are utilized as perennial gene vectors. The cloning of the entire JcDNV genome into pBR322 led to a recombinant construct (pBRJ) that retained the capacity to produce infectious particles when injected into larvae of the sensitive host Spodoptera littoralis or transfected into lepidopteran cell lines (20, 24, 25). The availability of plasmids carrying an infectious sequence of either the JcDNV or the Aedes aegypti densovirus genome has prompted the study of densoviruses as expression vectors (1, 2, 5, 10, 16, 31). The A. aegypti densovirus, a prototype of the Brevidensovirus genus, has been developed as a gene transfer vehicle that is able to transduce genes into mosquito larvae by typical routes of infection, opening the potential for gene introduction into natural populations (1, 2).

To explore the potential of the JcDNV genome as an expression vector for foreign genes, a series of noninfectious JcDNV-derived vectors were constructed expressing either a nonselectable (lacZ) (16) or a dominant selectable marker (neo) (30). These markers were inserted in frame into the VP gene or with their own ATG initiation codon. In these constructs, the foreign gene was under the control of the viral P9 promoter, which regulates expression of the structural polypeptides VP1, VP2, VP3, and VP4 in the wild-type virus. When transfected into S. littoralis and Spodoptera frugiperda cell lines, both marked constructs expressed the respective marker gene.

Following transfection with pJneo, several G418-resistant colonies were obtained which successfully underwent 50 subcultures in the presence or absence of the antibiotic, indicating stable expression of the neomycin phosphotransferase (30). Recently, we reported persistent high expression of β-galactosidase in somatic tissues throughout ontogenesis, from larvae to adult flies, following microinjection of plasmid pJlacZΔNS3 into preblastoderm eggs of Drosophila melanogaster (31). The pattern of β-galactosidase expression in adult tissues strongly suggested that integration of pJlacZΔNS3 into the genome of somatic cells occurred early during the preblastoderm stage of embryogenesis.

The present study examines the genomic status of JcDNV-derived vectors following transfection and selection of Spodoptera frugiperda Sf9 lepidopteran cell clones with the GFP reporter gene. To assess the influence of NS polypeptides on the integration of JcDNV vectors, a series of constructs containing the green fluorescent protein (GFP) gfp gene that included either fully functional (pJGFP, pJGFPH) or partially (pJGFPΔNS3) or entirely (pJGFPΔNSH) deleted NS genes were made.

The data provide the first demonstration that transfected JcDNV-based vectors integrate into the Sf9 genome, which represents a convenient means of integrating foreign DNA sequences into an insect cell genome.

MATERIALS AND METHODS

Cell line and plasmids.

The insect cell line Sf9 (ATCC CRL 1711) derived from S. frugiperda ovaries (36) was maintained in TC100 (Gibco Invitrogen, Cergy Pontoise, France)-derived medium supplemented with 10% heat-inactivated (56°C for 30 min) fetal calf serum, penicillin (125 μg/ml), and streptomycin (50 μg/ml). Cells were grown in 25-cm² flasks (BD Biosciences Discovery Labware, Bedford, Mass.) at 28°C. Stock cells were maintained in exponential growth by appropriate dilutions on a weekly basis. Cells were transfected with DOTAP transfection reagent according to the manufacturer's protocol (Roche Diagnostics, Indianapolis, Ind.).

Plasmid construction.

All JcDNV-based constructs (Fig. 1) were derived from pBRJ (20) and contained the GFP marker. The pEGFP-N1 SmaI-SspI fragment (BD Biosciences Clontech, Palo Alto, Calif.) encompassing the GFP coding sequence and simian virus 40 polyadenylation signal was inserted into the PvuII site of ORF1 (nucleotide 1864) of pBRJ, so as to place the enhanced green fluorescent protein (EGFP) in frame, 25 codons downstream of the ATG initiation codon for VP4 polypeptide, under the transcriptional control of the P9 viral promoter. pJGFPH was generated by substitution of the ClaI-BclI fragment of pBRJH (F. Rolling, personal communication) with that of pJGFP, restoring the full-length right ITR sequence deleted of 89 5′-terminal bases in pBRJ (12). pJGFPΔNS3 was derived from pJGFP by an NsiI frameshift deletion that disrupts ORF4 (NS-3). pJGFPΔNSH was constructed from pJGFPH by a BclI-NsiI deletion in the NS genes.

FIG. 1.

Schematic organization of recombinant JcDNV plasmids. All plasmids derive from pBRJ, a plasmid encompassing an infectious sequence of the JcDNV genome (20). Restriction sites used to construct or analyze recombinant plasmids are located on the map and designated as follows: A, AflII; B, BclI; N, NsiI. Plasmid sizes are given in brackets. Only relevant plasmid features are shown. The P9 and P93 viral promoters (dotted arrows) drive the expression of the VP-GFP gene fusion and of the NS genes, respectively. Plasmid pJGFPΔNS3 contains an NsiI frameshift deletion from nucleotides 5166 to 5277 in pBRJ. Interruption of the NS protein sequences in pJGFPΔNSH was performed by a BclI-NsiI deletion from positions 4519 to 5277 in pBRJ. Plain thick arrows represent sequences that form self-priming hairpins at the end of the viral genome. t indicates the location of the 3′ ITR truncation on plasmid pJGFP and pJGFPΔNS3, which is restored in pJGFPH and pJGFPΔNSH. VP-GFP fusion proteins are shown above their coding sequence. Dotted boxes show JcDNV regions that are no longer transcribed. Plain thick lines represent the NS proteins, and dark dotted lines show NS proteins that are no longer expressed. PCR primers used for structural analysis (see Table 1) are represented by small open arrows.

DNA electrophoresis and Southern hybridization.

Genomic DNA was prepared from cells with the Wizard Genomic DNA purification kit (Promega, Madison, Wis.). About 30 μg of DNA was digested with the appropriate restriction enzymes and separated on a 1% agarose gel. Very large DNA fragments were separated by pulsed-field gel electrophoresis (PFGE; CHEF Mapper DR III; Bio-Rad, Hercules, Calif.) in Seakem Gold Agarose (FMC, Philadelphia, Pa.). Sample preparation and restriction digestion of DNA in agarose were performed according to Birren (8). Samples were separated on a 0.8% agarose gel at 6 V/cm for 20 h in 0.5× TBE (45 mM Tris, 45 mM boric acid, 1 mM EDTA) at 14°C with alternate pulses of 40 and 50 s. Following electrophoresis, DNA was stained with ethidium bromide, blotted onto a nylon membrane (Roche Diagnostics, Indianapolis, Ind.), and fixed by UV (312 nm). Hybridization probes were labeled with [³²P]dCTP by random priming (Roche Diagnostics, Indianapolis, Ind.). Hybridizations were performed according to the manufacturer's protocol. The membranes were autoradiographed on Kodak X-Omat film at −80°C.

PCR and inverse PCR.

PCRs were carried out with cell DNA as the template and the primers described below (see also Table 1 and primer positions in Fig. 1 and 5). The JcDNV VP-GFP fusion was detected with forward 547F and reverse 2600R primers. pBR322-JcDNV 5′ ITR junctions were examined with forward −647F and reverse 592R primers. For the JcDNV 3′ ITR junctions with the plasmid backbone, we used the forward 6437F and reverse primer 7188R for plasmids lacking the end of the 3′ ITR (pJGFP and pJGFPΔNS3) or reverse primer 7232R for plasmids in which the end was restored (pJGFPH and pJGFPΔNSH). Primers 4008F and 6359R were used to assess the status of the NS regions in the stable fluorescent cell clones.

TABLE 1.

Primers used for PCR and inverse PCR analyses

Region analyzed	Primer	Sequence (5′-3′)
JcDNV VP-GFP region	547F	TAGTCAGTATGTCTTTCTACACGGC
	2600R	CTTGTACAGCTCGTCCATGCCGAG
JcDNV 5′ ITR-pBR322	647F	GGCGTCAACACGGGATAATACCGCG
junction	592R	CGCGCACGATGTATTAACCCGGCCG
JcDNV 3′ ITR-pBR322 junction	6437F	CTGCTCGTAGACTGATGATGGCGC TCTGCTGT
	7188R	CCATCGGTGATGTCGGCGATATAG
	7232R	CCGGCTCGTATGTTGTGTGGAATTG
NS genes	4008F	GTATCCCATAACCATTTATTAAAATG
	6359R	GCCAGGTTTCTTTGTGTTCGTCGTT
Inverse PCR 5′	1370F	CCCTTCTGTTTCAGGTATGTCCCGT
junction	535R	CGTCAGCCTTTATATAGTCAATGGT
Cell DNA-viral DNA	US1F	CTAAAGAGGATTCAAAGACC
junction	US1R	CATCTAGACAAATAGGAGTG
	535R	CGTCAGCCTTTATATAGTCAATGGT

FIG. 5.

Characterization of fluorescent pJGFPΔNSH clones. (A) Southern blot analysis of genomic DNA isolated from cell clones E1 and E2. Cell DNA was digested with EcoRV (lanes 1, 3, 6, and 8) or AflII (lanes 2, 4, 7 and 9), separated by gel electrophoresis, and subjected to Southern hybridization as described in Materials and Methods. Lanes 5 and 10, AflII-digested plasmid pJGFPΔNSH. Hybridization was performed with a JcDNV or a pBR322 probe as indicated. (B) Pulsed-field gel analysis of clone E2 cell DNA. (a) Undigested (nd) and BglII- or SfiI-restricted cell DNA samples were separated on a 0.8% agarose gel (see conditions of migration in Materials and Methods). Molecular size standards are from the Saccharomyces cerevisiae chromosome marker II kit (MkII; Roche Diagnostics, Indianapolis, Ind.). (b) Southern blot of gel a after hybridization with a pJGFPΔNSH probe. Relative BglII and SfiI band intensities presented on each side were obtained by quantification through a PhosphorImager and analysis with ImageQuant software (Amersham Biosciences, Piscataway, N.J.). (C) Metaphase chromosome spreads prepared from clone E2 were analyzed by primed in situ synthesis. The yellow fluorescence indicates labeling with the pJGFPH probe. Chromosomes were stained with propidium iodide. On the right are two magnified images showing the bivalent labeling of paired chromosomes. (D) Southern blot of BglII-digested DNA from clone E2. Hybridization was performed with mixed hsp90 and pBR322 probes. Quantitative assessment of plasmid pJGFPΔNSH copy number in this clone was obtained by comparing the band intensities. nd, nondigested DNA as a control.

For inverse PCR (27), we digested 1 to 3 μg of DNA from positive clones with AseI or XmnI for 4 h and then proceeded to ligation at 25°C for 3 h (Rapid DNA ligation kit; Roche Diagnostics, Indianapolis, Ind.). PCR was performed on the circularized DNA with the forward 1370F and reverse 535R primers. PCR products were run on a 1.5% agarose gel, purified (PCR purification kit; Roche Diagnostics, Indianapolis, Ind.), and sequenced (Eurogentec, Seraing, Belgium). Sequence analysis was performed with MACAW software (NCBI, Bethesda, Md.). Chromosomal insertion sequences were subjected to BlastN analysis (3). PCR analysis of the integration junction isolated from clone pJGFPΔNS3-C1 was performed with primers US1F and 535R for amplification of the junction and primers US1F and US1R for amplification of the flanking genomic sequence.

Primed in situ synthesis.

Cultured cells were prepared as described in Gerbal et al. (15), and primed in situ synthesis was conducted following a protocol adapted from Gosden et al. (19; P. Barry and P. Fournier, unpublished data) with primers defined on sequences of the vector pJGFPH. Microscopic examination of labeled chromosomes was achieved on a 2× Leica DM RA epifluorescence microscope, and images were treated by deconvolution on an Octan Silicon Graphics computer with Metamorph software (Universal Imaging Corporation, Downingtown, Pa.).

RESULTS

Persistence of transgene expression in lepidopteran cells transfected with JcDNV vectors.

A series of pBRJ-derived plasmids that express the gfp marker gene were constructed to study the maintenance of the JcDNV vectors in cultured insect cells (see Materials and Methods). The gfp gene was inserted in frame 3′ of the VP4 ATG translational initiation codon of ORF1 in the JcDNV sequence. This results in the expression of chimeric GFP fusion proteins fused with the four truncated structural polypeptides VP1, VP2, VP3, and VP4 (see Fig. 1). To assess the role of the nonstructural proteins in the persistence of the JcDNV vectors, we compared the fate of constructs expressing all NS genes (pJGFP and pJGFPH) to that of vectors in which expression of NS-3 (pJGFPΔNS3) or of all NS proteins (pJGFPΔNSH) was abolished.

Sf9 cells were transfected with pJGFPH, and the number of GFP-expressing cells was scored by fluorescence microscopy for 23 generations (approximately 50 to 55 days). Shortly after transfection, GFP expression regulated by the P9 viral promoter was detected. Maximum number of transfected cells was observed 48 h later with approximately 31% of the cells showing GFP fluorescence as estimated by manually counting the number of GFP-fluorescent cells per microscope field. The number of GFP-fluorescent cells declined rapidly over the next 15 days to finally stabilize to 7 to 8% of the population (Fig. 2A).

FIG. 2.

Persistence of GFP fluorescence in cells transfected with pBRJ-derived constructs. (A) Nonclonal and clonal cell populations were monitored for GFP fluorescence after transfection. The names of clones and of the plasmids from which they were derived are shown on the right. (B) pJGFPH-transfected cells observed under epifluorescence microscopy. A colony of 24 cells at the 82nd subculture (approximately 6 months) that shows the punctate nuclear localization of the GFP fluorescence is presented.

As a control, plasmid pA3lacZ (26), in which expression of the lacZ reporter gene is regulated by the Bombyx mori cytoplasmic actin 3 promoter, was transfected into Sf9 cells. Following transfection with this construct, the number of β-galactosidase-positive cells decreased rapidly to an undetectable level after 12 generations (data not shown). Although differential expression of the transgenes cannot be ruled out, these observations strongly suggested that the persistence of the GFP phenotype after multiple rounds of cell division resulted from the maintenance of JcDNV sequences of the vector and did not require positive selection pressure for its maintenance.

To further investigate this phenomenon, fluorescent cells were cloned. Sf9 cells transfected with plasmid pJGFP, pJGFPH, pJGFPΔNS3, or pJGFPΔNSH were diluted 5 days posttransfection, a stage at which integration events have already occurred and the status of individualized cells is stabilized. Therefore, the separated fluorescent colonies that were picked up 2 to 3 weeks later corresponded to homogeneous populations with regard to integration events. Isolated clones were propagated by subculturing every 5 days. Several stable fluorescent cell lines were recovered for the pJGFPH plasmid (clones B1, B4, B5, B6, B7, and B8), which contained intact NS genes and a restored 3′ hairpin, as well as for each of the JcDNV constructs that contained deletions for these sequences: pJGFP (clone A1), pJGFPΔNS3 (clones C1, C2, and C7), and pJGFPΔNSH (clones E1, E2, and E3) (see also Table 2 and Fig. 2A). The population of cells showing GFP fluorescence within an isolated colony almost reached 100% and remained constant in subsequent passages for over 165 generations (approximately 12 months; see Fig. 2A), indicating that the GFP sequence was stably maintained and expressed within these cell lines. This percentage was lower in cells transfected with the pJGFPΔNSH construct, but this phenotype was reproducible even after subculturing of these clones. This observation will be analyzed in a forthcoming paper.

TABLE 2.

PCR analysis of JcDNV plasmid-transfected cell clones^a

Plasmid or clone	Fragment size (kb) or amplification
	pBR322-5′ ITR	VP-GFP	3′ ITR-pBR322	NS region
pJGFP	1.2	2	0.75	2.3
A1	−	+	+	+ (and smaller products)
pJGFPH	1.2	2	0.8	2.3
B1	−	+	+	+ (and 1.2)
B4	−	+	+	0.7
B5	−	+	+	0.7
B6	−	+	+	0.7
B7	−	+	+	+
B8	−	+	+	+
pJGFPΔNS3	1.2	2	0.75	2.2
C1	+	ND	+	+
C2	+	ND	+	+ (and 2)
C7	+	+	+	+
pJGFPΔNSH	1.2	2	0.8	1.6
E1	+	ND	+	+
E2	+	+	+	+
E3	+	ND	ND	ND

^aThe DNA of each JcDNV-transfected clone was PCR tested to study the persistence of viral sequences. −, no PCR amplification; +, successful PCR amplification of regions that are intact. ND, not determined.

Analysis of JcDNV sequences in GFP-fluorescent Sf9 clones.

To identify which regions of the viral sequence were maintained in the cell clones, genomic DNA was extracted from each of the fluorescent isolates and subjected to PCR analyses. The structural integrity of four regions was assessed (see Fig. 1 and the primers described in Table 1). The VP-GFP fusion coding sequence, the junctions of the 5′ and 3′ JcDNV ITRs with the pBR322 sequence, and the sequence encoding the NS polypeptides were characterized with the sets of primers listed in Table 1. As shown in Table 2, two regions of the recombinant viral genome were unaffected following transfection and cell clone isolation. The 2-kb VP-GFP region could be amplified from all clones, indicating that no major rearrangements had occurred in this portion of the plasmid sequence following transformation. This was corroborated by assessing the location of the GFP fluorescence in transformed cells (Fig. 2B), which established the correct nuclear targeting of the VP-GFP chimeric proteins. Similarly, the JcDNV 3′ ITR-pBR322 junction amplified correctly from genomic DNA of all the clones, generating the expected 751-bp (pJGFP, pJGFPΔNS3) or 795-bp (pJGFPH, pJGFPΔNSH) amplicons.

PCR amplification of the pBR322-JcDNV 5′ ITR junctions resulted in two different situations depending upon the vector used. Attempts to amplify the 5′ ITR region with genomic DNA from Sf9 clones transformed with pJGFP or pJGFPH, two constructs expressing all NS proteins, failed repeatedly. These results strongly suggested that in the presence of NS polypeptides, specially NS-1, the pBR322-5′ ITR junction is resolved, preventing amplification of the region spanning from −647 (pBR322 sequence) to +592 (viral sequence). In contrast, Sf9 clones transformed with pJGFPΔNS3 (no NS-3 expression) or with pJGFPΔNSH (no NS genes expression) contained an intact pBR322-JcDNV 5′ ITR region as the expected 1.2-kb amplicons were produced by PCR, indicating that no resolution of the vector-viral DNA junction or rearrangements had occurred in this region.

The structure of the viral NS region in clones transfected with each of the constructs was examined. Amplification of genomic DNA from pJGFPΔNSH-transfected clones with primers encompassing the entire NS mutated region resulted in PCR products of the expected 1.6 kb. However, significantly smaller amplicons were frequently observed in the Sf9 clones transfected with pJGFP, pJGFPH, or pJGFPΔNS3 (see Table 2). This showed that while being maintained, JcDNV vectors supporting NS expression presented in some cell lines recurrent deletions in the NS genes.

Plasmids pJGFP and pJGFPH integrate at low copy numbers into the host cell genome.

Genomic DNA of GFP-fluorescent clones was digested with either EcoRV, which has no recognition site in the JcDNV constructs, or AflII, an enzyme that has a single recognition site within the transgene sequence (Fig. 1). Restricted DNA was then subjected to Southern blot analysis with either the entire JcDNV DNA or the pBR322 or pJGFPH random-primed plasmid as the probe.

Figure 3A shows the analysis of B7, a fluorescent cell clone isolated after pJGFPH transfection. Following EcoRV digestion, hybridization with the JcDNV probe revealed a single band at approximately 14 kb, which was larger than the entire 9.8-kb pJGFPH sequence, and after AflII digestion, two smaller bands were detected. These observations strongly suggest that the JcDNV sequences were integrated into the genome of the host cell and most likely at a single locus. Hybridization of the Southern blot with the pBR322 probe after AflII digestion revealed a band of the same size as the one detected by the JcDNV probe, demonstrating that the pBR322 sequence of the vector was integrated along with the JcDNV sequence into the genome of the host cell.

FIG. 3.

Characterization of fluorescent pJGFPH and pJGFPΔNS3 clones. (A and C) Southern blot analysis of genomic DNA isolated from cell clones B7 (A) and C7 (C). Cell DNA was digested with either EcoRV (e), AflII (a), BamHI (b), or AflII-BamHI (a+b), separated by gel electrophoresis, and subjected to Southern hybridization as described in Materials and Methods. Hybridization was performed with either JcDNV, pBR322, or pJGFPH probes as indicated below the pictures. Sizes are shown in kilobases. (B and D) Metaphase chromosome spreads prepared from clones B7 (B) and C7 (D) were analyzed by primed in situ synthesis. The yellow fluorescence indicates labeling with the pJGFPH probe. Chromosomes were stained with propidium iodide.

Metaphase chromosome spreads from cells of clone B7 labeled by primed in situ synthesis with GFP-specific primers and observed by epifluorescence microscopy (Fig. 3B) provided additional evidence that integration of the pJGFPH vector into the genomic DNA of the B7 clone had occurred. A single spot was consistently detected among the cell chromosomes, indicating that pJGFPH had inserted in this fluorescent clone at a single site in the Sf9 polyploid genome.

As stated above, the failure to amplify the pBR322-JcDNV 5′ ITR region by PCR in all the pJGFP and pJGFPH fluorescent clones indicated systematic rearrangements of these vectors in this region. Based on this observation, we hypothesized that the JcDNV 5′ ITR may be involved in the integration of both constructs. To explore putative virus-cell DNA junctions, inverse PCR was performed with primers 1370F and 535R (Table 1) in opposite orientation and located 3′ of the 5′ ITR within the JcDNV AseI or XmnI restriction fragment (Fig. 4A). PCR products were successfully amplified from the genomic DNA of clones A1, B1, B5, and B7 after digestion with XmnI. The PCR product generated by inverse PCR from clone A1 showed a junction site between a 327-nucleotide-long non-pJGFP sequence and the first nucleotide of the JcDNV 5′ ITR (Fig. 4A). We were unable to obtain the sequence of the junction beyond nucleotides 21 of the viral sequence because the secondary hairpin structure formed by the terminal palindrome of the 5′ ITR is known to be resistant to sequencing by the Sanger method (33).

FIG.4.

Sequence analysis of recombination events. (A) Schematic representation (not to scale) of the junction breakpoints found by inverse PCR in the 5′ ITR of plasmids pJGFP, pJGFPH, and pJGFPΔNS3. PCR primers (see Table 1) are shown. Vertical arrows with the clone name indicate the positions of the breakpoints relative to the 5′ ITR. The sequences of the junctions aligned with the JcDNV sequence are shown below. Diagonal lines indicate the crossover points. (B) Schematic representation of the recombination events found in plasmids rescued from chromosomal DNA of clone C7. The 6.8-kb region delimited by the pJGFPΔNS3 NdeI sites and the HindIII (H), BstEII (Bs), and BamHI (Ba) sites used to analyze the rescued plasmids are shown. The breakpoint regions are underlined. The rescued plasmids (RP) in which rearrangements were detected are listed underneath according to the location of the recombination event (nucleotide position of the recombination points numbered from the left of the 5′ ITR are given when sequenced). Below are shown sequences of rescued plasmid junctions aligned with the sequences of parental molecules.

Because the sequence of the region flanking the 5′ ITR had no homology with the pJGFP sequence, we concluded that this region corresponded to genomic DNA of the host cell. The presence of this viral-cellular DNA junction in the genome of clone A1 was confirmed by PCR amplification with primer US1F (Table 1) located 5′ to the junction point and primer 535R (Table 1) internal to the JcDNV DNA sequence. Amplification of the expected 0.8-kb amplicon validated the presence of an integration event between the JcDNV vector and cellular DNA in this clone (data not shown). However, attempts to identify the flanking cellular DNA sequence in nontransformed Sf9 genomic DNA by nested PCR were unsuccessful. This result suggests that the integration of the JcDNV vector into the genome of clone A1 very likely induced rearrangement of the flanking cellular sequence, a phenomenon similar to complex cellular DNA rearrangements observed following integration of wild-type, recombinant (37) and plasmid adeno-associated virus (AAV) vectors in human cells (34). The inverse PCR analysis of clone B1 revealed a junction between sequences of the vector and the viral DNA with a breakpoint at position 280 nucleotides on the JcDNV 5′ ITR (Fig. 4A). This rearrangement corresponds to a deletion of about half of the ITR with a flanking sequence consisting in a complex rearrangement involving JcDNV sequences in both direct and reverse orientations.

Integration of plasmid pJGFPΔNS3 occurs at low copy numbers in several sites into the host cell genome.

The genomic DNA isolated from the stable fluorescent clone C7 was subjected to Southern blot analysis. Following digestion with AflII, which has a single recognition site in the plasmid, hybridization to the entire pJGFPH probe revealed a strong signal corresponding to high molecular weight DNA (Fig. 3C, lane 2). After digestion with BamHI, which has four recognition sites, hybridization with the pJGFPH probe resulted in a complex hybridization pattern (Fig. 3C, lane 3). In addition to the four expected pJGFPΔNS3 fragments of 4.7, 3.3, and 1.6 kb (the 0.3-kb fragment migrated out of the gel), four additional fragments of 3.7, 2.4, 1.8, and 0.9 kb were also detected. The presence of large intact regions of the plasmid within the genomic DNA was confirmed by the resolution of the 4.7-kb BamHI fragment containing both pBR322 and JcDNV sequences into two expected 3.8-kb and 1-kb bands following AflII and BamHI digestion (Fig. 3C, lane 1). The bands that were not of predicted sizes following BamHI or AflII-BamHI digestion could correspond to fragments containing genomic-vector DNA junctions or to rearrangements of plasmid sequences. Taken together, these data suggested that several copies of viral and plasmid sequences had integrated into the cell genome.

Although the complete pJGFPΔNS3 hybridization pattern was recovered following BamHI digestion, the lack of a 10-kb band corresponding to the size of pJGFPΔNS3 after AflII digestion suggests that no concatemeric copies of the pJGFPΔNS3 vector have integrated in the genome of clone C7. Similar results were observed with clone C1 (data not shown).

As shown above by PCR and Southern blot analysis, these data confirmed that the majority of the vector sequences, both viral and plasmidic, are conserved in the clones and that nonconcatemeric copies of the vector are integrated into the host cell genome. To further investigate the possibility of multiple integration events of the pJGFPΔNS3 vector into the clone C7 genome, metaphase chromosome spreads were labeled by primed in situ synthesis with GFP primers. The C7 chromosome spreads repeatedly showed four to six labeled spots (Fig. 3D). Because DNA hybridization suggested that the entire plasmid, including the bacterial origin of replication and β-lactamase selectable marker, was integrated into the clone C7, a plasmid rescue experiment was conducted to recover integrated vectors.

Digestion of C7 genomic DNA with NdeI, was followed by dilution, ligation and transformation of Escherichia coli. NdeI has two recognition sites in the vector, one positioned 5′ of the bacterial origin of replication and ampicillin resistance gene sequences and the second located within the NS-1 sequence of JcDNV in pJGFPΔNS3 (Fig. 4B). This allows the isolation of a selectable circular fragment containing the 3′ end junctions of integrated plasmids. This method allowed scoring of the recombinational events that had occurred in a region of pJGFPΔNS3 comprised of sequences between position −2065 (pBR322 sequence) and position +4620 (viral sequence) from the extremity of the 5′ ITR, a region representing 68% of the total plasmid sequence.

Nine out of 292 plasmids rescued following NdeI digestion were analyzed in detail by restriction enzyme digestion and sequencing of critical regions (Fig. 4B). One plasmid retained BamHI, BsteII, HindIII, and NdeI restriction maps identical to those of the circularized 6.8-kb NdeI fragment of pJGFPΔNS3 (Fig. 4B), confirming that in this case a large portion of the vector had integrated into the genomic DNA. Plasmid RP05 contained a recombination breakpoint within the pBR322 sequence whereas the digestion patterns for the seven other plasmids showed that the recombination occurred within the viral sequence. For RP06 and RP18 the recombination occurred more than 1 kb, 3′ from the extremity of the 5′ ITR. In the five other plasmids, breakpoints were mapped within the 5′ ITR. Plasmids RP04 and RP10 contained a rearrangement beyond the distal 96 nucleotides of the 5′ ITR and only the first 20 nucleotides of the ITR from these two plasmids was successfully sequenced. For plasmids RP14, RP17, and RP22, where recombination occurred within the hairpin structure, none of the flanking sequences corresponded to cell DNA but rather to rearrangements of the ITR with either viral or pBR322 DNA sequences. These results strongly support the possibility that the JcDNV 5′ ITR is the primary site for recombination.

To further examine the integration sites of pJGFPΔNS3 in transformed cells, putative viral-cellular junctions were identified by inverse PCR. The same set of primers was used that permitted characterization of the integration junctions in pJGFP and pJGFPH transformed clones. The results for clone C1 are presented as representative for pJGFPΔNS3-transformed clones. Although the presence of intact 5′ ITR sequences in this clone resulted in preferential amplification of the expected 1.5-kb PCR product following AseI digestion and ligation, an additional 0.6-kb amplicon was also recovered. The sequence revealed a junction breakpoint at nucleotides 146 of the JcDNV 5′ ITR followed by a 195-nucleotide sequence that had no homology with the pJGFPΔNS3 sequence (Fig. 4A).

A Blast search for the sequence flanking the JcDNV ITR in GenBank revealed no significant homologies with referenced sequences. Nested PCR of clone C1 DNA with a primer distal to the crossover point and a primer internal to the viral sequence resulted in successful amplification of a 547-bp product, confirming the presence of a cellular-viral DNA junction in this clone. The successful amplification of a 157-bp product with primers designed at both ends of the flanking sequence and nontransfected Sf9 genomic DNA as a template confirmed the cellular nature of the flanking sequence (data not shown). This result also revealed that no rearrangements had occurred within the flanking genomic region in clone C1 following integration of the vector at this particular site.

Deletion of NS genes results in integration of head-to-tail concatemers of the JcDNV vector.

Genomic DNA from clones E1 and E2 resulting from pJGFPΔNSH transfection was subjected to Southern blot analysis following digestion either with EcoRV that has no recognition site within the JcDNV construct or with AflII that has a single recognition site in pJGFPΔNSH. The Southern blot analysis with the entire JcDNV DNA as probe showed two large intense bands in the EcoRV-restricted DNA of each clone (Fig. 5A, lanes 1 and 3). Following AflII digestion, the same probe revealed a single intense band (Fig. 5A, lanes 2 and 4) identical in size to that of pJGFPΔNSH (Fig. 5A, lane 5). Hybridization of the Southern blot with a pBR322 probe revealed similar patterns (Fig. 5A, lanes 6 through 10).

These results demonstrate the presence of several intact copies of the pJGFPΔNSH plasmid in both clones. While the hybridization profiles following EcoRV digestion suggested that copies of pJGFPΔNSH could be distributed in several large clusters integrated within the genome, the AflII patterns were consistent with the maintenance of pJGFPΔNSH as episomal molecules within the fluorescent clones. To further examine the status of pJGFPΔNSH DNA, the presence of episomal circular concatemeric structure of the plasmid was investigated by two-dimensional gel electrophoresis. DNA from clone E2 was digested with EcoRV (no recognition sites within pJGFPΔNSH) and subjected to two-dimensional gel electrophoresis according to the procedure of Cohen and Lavi (9). After gel transfer and hybridization with the pJGFPΔNSH probe, no spots other than genomic linear ones were ever evidenced (data not shown).

As this technique has only been shown to be efficient for a relatively small circular molecule (10 to 50 kb), a larger concatemer might have escaped detection. We therefore embedded transfected cells within agarose blocks and prepared samples suitable for pulsed-field gel electrophoresis (PFGE). Chromosomal DNA was digested with either BglII or SfiI, which do not have recognition sites within pJGFPΔNSH, and then resolved on PFGE. Southern blot analysis was performed with the entire pJGFPΔNSH plasmid as a probe. Following digestion with BglII, 9 to 10 bands were observed, while 3 to 4 bands were present in the SfiI digest (Fig. 5B). A different band pattern was obtained with DNA digested by EcoRV (data not shown). The differences in hybridization patterns unambiguously demonstrate that the plasmid DNA was integrated into host cell genomic DNA, since episomal concatemeric molecules should have generated similar patterns with enzymes that do not cut within the monomer.

In order to confirm these observations, we performed primed in situ DNA labeling on the E2 cell clone. Multiple labeled spots were clearly observed for the different karyotypes examined (Fig. 5C). In some cases, bivalent labeling could be detected on the paired chromosomes (Fig. 5C, magnified observations), a situation difficult to observe when examining spreads of holocentric chromosomes. These observations further confirm that the pJGFPΔNSH vector was integrated into the host cell genomic DNA and suggest that multiple plasmid clusters are distributed throughout the genome. Furthermore, we noticed strong variations of the label intensity from one chromosome to another, which reflect differences in the number of integrated copies of pJGFPΔNSH at each locus. This is consistent with the variability in labeling intensities of the banding patterns of genomic DNA observed in Southern blots (Fig. 5B).

pJGFPΔNSH-transfected cell line contains about 60 copies of the vector.

In order to estimate the number of pJGFPΔNSH plasmid copies integrated in the genomic DNA of the E2 cell line, we quantified the hybridization signal from the pJGFPΔNSH inserts by comparison with the signal from the endogenous hsp90 gene. To normalize the signals, a probe that anneals to both templates over equivalent lengths of sequence was used, i.e., the 2.5-kb pBR322 portion of a pUC-hsp90 probe hybridized to the 7.9-kb HindIII fragment of pJGFPΔNSH, while the 2.2-kb hsp90 portion of the same probe annealed to the 2.5-kb HindIII fragment of the hsp90 gene. Quantitative analysis of the hybridization signal from these two bands showed that the pJGFPΔNSH inserts of clone E2 retained 6.9-fold more label than the heat shock protein gene sequence (compare signal intensities for the 7.9-kb and 2.5-kb bands in Fig. 5D, lane 3). Previously reported quantitative analysis established that there are 8 copies of the hsp90 gene present in the genome of the Sf9 cell line (23). Based on this copy number, we estimated that 55 copies of the pJGFPΔNSH vector were present per E2 cell. This result is in accordance with an independent estimate based on quantitative PCR (Françoise-Xavière Jousset, personal communication).

In light of these quantitative measurements, the PFGE results described above were reexamined to estimate the relative intensities of the bands observed in Fig. 5B. Assuming that the less intense band (relative intensity = 1.0) corresponds to one copy of the vector, results show that the number of copies of the pJGFPΔNSH plasmid varied from 1 to 11 copies per BglII band and from 5 to 48 copies per SfiI band. For each digestion pattern, the total number of pJGFPΔNSH copies adds up to an estimated 50 to 60, which lends additional support to the previous estimate.

DISCUSSION

In this study, we investigated the properties of JcDNV-derived plasmid constructs to promote stable expression of the gfp reporter gene in Sf9 lepidopteran cells. To this end, the gfp coding sequence was inserted into the cloned JcDNV genome, downstream of and in frame with the ATG translation initiation codon of VP4 capsid protein. We previously reported that insertion of the lacZ gene in this position promoted the highest levels of VP1-β-galactosidase, VP2-β-galactosidase VP3-β-galactosidase, and VP4-β-galactosidase fusion proteins in the transfected SPC-SL 52 lepidopteran cell line (16). To assess the role of the nonstructural proteins on the modalities of vector maintenance, a series of plasmids were constructed that contained a viral genome with fully functional NS genes (pJGFP, pJGFPH), with a deletion abolishing expression of NS-3 (pJGFPΔNS3), or with a deletion abolishing expression of the three NS genes (pJGFPΔNSH).

Transfection of each of the four JcDNV constructs into Sf9 cells conferred a persistent GFP⁺ phenotype to a high proportion of cells, exceeding at least 10 times the frequency of random insertion events of plasmids that did not contain densovirus DNA. This allowed direct isolation of GFP-fluorescent clones in the absence of selective pressure. GFP expression remained remarkably stable in these clones for over a year (Fig. 2A). As for persistent β-galactosidase activity in SPC-SL 52 cells (16) and throughout development in somatically transformed Drosophila cells with JcDNV-derived constructs (31), the stability of the GFP phenotype in all Sf9 cell clones appears to reflect a property, specific of the JcDNV genomic sequences, to persist in insect cells.

The localization of fluorescence in the nuclear compartment (Fig. 2B) indicated that, similar to the situation observed in cells transfected with JcDNV-lacZ constructs (16, 31), the insertion of the GFP gene did not disturb the nuclear targeting of the chimeric proteins.

To determine the status of the JcDNV constructs in transformed cells, DNA from the GFP-fluorescent clones was analyzed after a 30-generation passage (approximately 10 weeks) (Fig. 2A). PCR analyses (see Table 2) revealed that the persistence of the GFP fluorescence in all transformed clones resulted from the maintained VP-GFP chimeric proteins expression under the control of the P9 promoter. In addition, pBR322 plasmid sequences are maintained along with the JcDNV sequence in all the clones. The 3′ ITR-pBR322 junction was apparently intact in all clones whereas the 5′ ITR-pBR322 junction was subject to frequent rearrangements in cell clones pJGFP-A and pJGFPH-B transformed with constructs expressing the NS proteins. The presence in Southern blots obtained from cell DNA restricted with EcoRV, which has no recognition site in the JcDNV constructs, of at least one band larger than the plasmid is a good argument in favor of the integration of all the constructs into the cell genome. Furthermore, the variable intensity of bands suggested that cell DNA of clones A and B potentially carry a small number of copies of plasmid pJGFP or pJGFPH, clones C several nonconcatemeric copies of pJGFPΔNS3, and clone E concatemeric copies of pJGFPΔNSH.

Consistent with these data, primed in situ synthesis analysis of metaphase chromosome spreads (Fig. 3B, 3D, and 5C) clearly showed a single integration event in clone B7 and several integrations in clones C7 and E2. Further evidence for integration of the constructs was obtained by sequencing the products of inverse PCR assumed to contain integration junctions. In addition to extensive rearrangements including deletions, inversions or translocations of both viral and plasmid sequences, we identified in two clones (A1 and C1, Fig. 4) cellular DNA sequences flanking the 5′ JcDNV ITR region. This is to our knowledge the first documented evidence of integration of a densovirus sequence into an insect cell genome. This result provides a clear explanation to the in vivo persistence through all developmental stages of Drosophila melanogaster of lacZ gene expression following microinjection of the pJlacZΔNS3 plasmid into syncytial embryos (31).

Taken together, our results show great similarities with those recently reported on the integration process of AAV vectors following their transfection into human cells. Analyses of cell DNA-AAV junctions have clearly established that the 5′ ITR as well as the P5 promoter AAV regions are hot-spots for integration (17, 21, 29, 32, 34, 35, 37). Similarly, PCR analyses and plasmid rescue experiments revealed referential rearrangements of the JcDNV sequences in the 5′ ITR and therefore suggested that this region is also the primary site of recombination events (Fig. 4A and B). In contrast, the absence of detectable rearrangements at the 3′ end of the viral sequence suggested that prior excision of the JcDNV genome from the plasmid vector is not required for integration. While the precise excision of the wild-type JcDNV sequence from pBRJ leads to the production of infectious virions in pBRJ-transfected insect cells or S. littoralis larvae (20, 24), no clear explanation can be given for the apparent lack of excision of the recombinant JcDNV sequences prior to their integration into the insect cell DNA. Interestingly, the integration of the plasmid backbone sequence together with recombinant AAV sequence into the genome of human cells transfected with AAV plasmids is a phenomenon also frequently observed (34, 35).

Transfection of the pJGFP and pJGFPH constructs lead to their integration at a low, probably single copy number into the host chromosomes (no evidence of head-to-tail tandem repeats was found). Southern blot analysis of clone B7 revealed integration at a single site (Fig. 3A), a result also supported by the observation of a single spot after primed in situ synthesis labeling of metaphase chromosomes spreads (Fig. 3B). This observation raises the question as to whether or not JcDNV vectors expressing NS proteins exhibit site-specific integration properties in insect cells similar to those reported in Rep-expressing AAV plasmids in vertebrate cell lines (17, 18, 21, 22, 32). A recent in vitro study clearly demonstrated that JcDNV NS-1 protein shares the same biochemical properties as AAV Rep78 protein (11). Based on these functional similarities, we looked for homologies between the two cell DNA sequences identified at the integration junction with the viral DNA. No significant homologies were found that could provide the basis for homologous recombination.

All integrated pJGFP, pJGFPH, and pJGFPΔNS3 sequences displayed extensive internal rearrangements such as deletions or inversions. This is reminiscent of observations from integration studies of wt AAV with an episomal model system (13, 17) or of AAV plasmids (29, 34). As reported for AAV Rep and MVMp NS-1 expression in mammalian cells (4, 28, 35, 37), expression of JcDNV NS proteins may interfere with the Sf9 cell cycle. In this respect, it is likely that in the present study, fast-growing colonies resulting from spontaneous deletions in the NS genes, which could impair or stop NS expression, were preferentially selected.

Transfection with pJGFPΔNS3 construct resulted in multiple integration events but no head-to-tail tandem arrays were detected by Southern blot analysis. Because of the polyploidy of the Sf9 genome and the fragmentation of the chromosomes, a precise assessment of the total number of integration loci per given chromosome was not possible. Our results clearly indicate that NS-3 is not essential for the integration of the plasmid vector but may directly or indirectly play a role that affects recognition of integration sites within the host cell genome.

In the absence of NS-1 expression (pJGFPΔNSH), i.e., of terminal hairpin resolution, clusters containing a high number of copies of the plasmid (50 to 60 per cell) were shown to integrate as head-to-tail concatemers into various loci of the genome (Fig. 5B and 5C). The complex mechanisms leading to the synthesis of head-to-tail concatemeric molecules of plasmid DNA and to their integration at different loci in the cell chromosomes are totally unknown.

Because information on the status of JcDNV vector in Sf9 cell DNA were obtained from stably transformed cell clones, they do not provide a clear understanding of the dynamics of the JcDNV vector stabilization, which occurs well before characterization. While this study demonstrates that JcDNV sequences can mediate plasmid vector integration into a lepidopteran genome, our results only revealed partial aspects of the underlying mechanisms, and many questions about this system remain unanswered. Further experiments will more specifically address the nature of the sites in the cell genome in which JcDNV vectors integrate. NS trans-complementation of pJGFPΔNS3 and pJGFPΔNSH will also be necessary to better understand the regulatory role played by the nonstructural proteins of JcDNV vectors in chromosomal integration. With the present demonstration of their integrative properties, JcDNV-derived vectors appear to be promising tools for the production of transgenic insects.