![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Previous Article | Next Article 
Journal of Virology, September 2007, p. 9976-9989, Vol. 81, No. 18
0022-538X/07/$08.00+0 doi:10.1128/JVI.00630-07
Copyright © 2007, American Society for Microbiology. All Rights Reserved.
Department of Gene and Cell Medicine, Mount Sinai School of Medicine, New York, New York 10029
Received 23 March 2007/ Accepted 3 July 2007
 |
ABSTRACT |
|---|
 Top ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
A distinguishing characteristic of AAV is its inability to replicate autonomously. Wild-type AAV (wtAAV) productively infects cells only when these cells are coinfected with a helper virus, normally adenovirus or herpes virus (1). In the absence of helper virus coinfection, wtAAV can persist in the cell indefinitely by integrating its genome into a specific site on human chromosome 19 (16). This site-specific integration requires the presence of the AAV rep gene which, however, is not present in typical rAAV vectors. Consequently, the persistence of the transgene in rAAV infection is by another mechanism, presumably by extrachromosomal persistence of the vector DNA (12, 13, 28).
rAAV vectors are most commonly produced by transfection of a plasmid containing the gene of interest inserted into an AAV vector cassette. This vector cassette need only contain the two terminal regions from the AAV genome, flanking the transgene. These two terminal inverted repeats, which are 145 bases in length (29), serve as the viral origins of DNA replication and as the packaging signals. In addition to the vector genome, the AAV genes rep and cap as well as adenovirus helper factors are required for rAAV production. A limited set of adenovirus helper genes is sufficient and can be provided by plasmid cotransfection (23, 26, 35). In some cases the AAV rep and cap genes have been added to this helper plasmid, resulting, for example, in the widely used plasmid pDG (11). Through these modifications, rAAV production has been simplified to a cotransfection of two plasmids.
It is noteworthy that, in all variations of this production strategy, recombinant AAV genomes yield less virus than does the wild-type AAV equivalent. At present, the reason for the much greater yield of wild-type virus compared to recombinant virus remains unclear.
Attempts to improve the yield of rAAV have included, most notably, that of Urabe et al., who developed a method of producing large batches of recombinant vectors using baculovirus in insect cell cultures. Their method achieves higher yields principally by enabling greater numbers of producer cells per volume of medium (31).
In mammalian cells, several reports have noted that the production of recombinant AAV can be increased by the provision of more Cap protein (4, 11, 14, 20). These approaches were based on the hypothesis that the amount of capsid might be limiting because, in contrast to wtAAV, during the production of recombinant virus the number of Cap-expressing genomes remains constant throughout vector replication. Inoue and Russell generated cell lines that contain stably integrated simian virus 40 (SV40) origins linked to the helper genes, while Chiorini et al. used SV40 origin-containing replicons. With the expression of large T-antigen, higher yields of recombinant viruses were achieved. Attempts to produce greater amounts of vector in mammalian cells have also included the development of Rep- and Cap-expressing cell lines, which, when successful, seemed to involve amplification of the rep/cap genes (2, 3, 6, 8, 22).
Kleinschmidt and colleagues reported, however, that in cells in which recombinant genomes were produced using pDG to supply Cap proteins, the particle titer was not substantially lower than for wild-type genomes (10). In recombinant production, between 80 and 98% of their capsid particles were empty, compared with less than 50% for wild-type virus. In several elegant experiments, they showed that the provision of additional Cap protein did increase the amount of genome-containing particles, but to an amount disproportionately less than the increase in empty particles. More recently, Li and Samulski showed a dramatic (10- to 20-fold) increase in vector yield when Cap proteins were supplied by an AAV type 5 (AAV5) Rep-driven construct that could replicate in cells transfected with a recombinant AAV. However, in a detailed analysis they noted that it was not possible to correlate the production of virus with Cap expression or, indeed, with any of the several variables they measured (20).
Tullis and Shenk have observed that the absence of a cis-acting element, present in the wild-type genome between nucleotides 194 and 1882, leads to lower levels of virus production. They correlated a deletion of this region to a sixfold reduction in the amount of monomer-length replicative form (i.e., double-stranded) DNA (30).
Here we present a further study asking which elements of the wtAAV genome are required in cis for the high yield of the wild-type virus and whether these elements affect primarily DNA replication or encapsidation of the replicated genome. We produced virus in a manner identical to that by which recombinant genomes are commonly produced, i.e., cotransfection with the helper plasmid pDG. By employing deletion mutants of AAV, we have avoided possibly confounding effects due to recombinant DNA sequences. To examine the relationship of DNA replication and packaging, we developed reliable assays to distinguish encapsidated genomes from unencapsidated genomes in the cell.
We found that the amount of unencapsidated AAV or rAAV DNA in cells transfected with wild-type AAV, deletions which produced abundant virus, deletions which produced much less virus, and a representative recombinant genome (pTRUF11) were remarkably similar despite great variations in the amount of encapsidated genomes that were produced. We further found that deletions from the left portion of the AAV genome had only a modest effect on the production of packaged genomes. In contrast, constructs with deletions from the right portion of the AAV genome produced markedly less encapsidated virus. Our observations are most consistent with a model in which the constructs package poorly neither due to lack of sufficient Cap nor to disruption of a genomic packaging signal but rather to some problem associated with the provision of cap "in trans."
|
MATERIALS AND METHODS |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
PTRUF11 is derived from pSM620 (27) in which the internal sequences have been replaced by a green fluorescent protein (GFP) gene under the control of a chicken ß-actin -cytomegalovirus promoter and a neo gene under the control of a thymidine kinase promoter (a kind gift of Sergei Zolotukhin).
Transfection procedure. 293T cells were plated in six-well plates. One day later, when cells were approximately 70% confluent, medium was replaced by low-glucose Dulbecco's modified Eagle's medium. Four h later, cells were transfected by the CaCl2 method. Each well received 3.0 µg of pDG and either 1.2 µg, 0.01 µg, or 0.001 µg of pAV2 (1x, 0.01x, or 0.001x, respectively). Other test plasmids were added at equivalent molar amounts. Plates were incubated 2 days. Harvesting was by scraping cells into phosphate-buffered saline. Cells were resuspended in phosphate-buffered saline, divided into three equal aliquots, and then pelleted at 5,000 rpm for 5 min in a microcentrifuge. Supernatant was removed and pellets were stored at –20°C until processing by procedure 1 or procedure 2 (Fig. 1).
|
Procedure 1. The pellet was resuspended in 100 µl TE and frozen and thawed three times. A mixture of 5.0 µl 3.0 M NaCl, 2.0 µl 100 mM MgCl2, and 1.0 µl 500 mM Tris (pH 8.0) was added. This was followed by the addition of 1.0 µl benzonase (5 µ/µl) and 0.5 µl DNase (10 µ/µl). Incubation was at 37°C for 2 h. Then, 5 µl of proteinase K (10 mg/ml), 5 µl 10% sodium dodecyl sulfate (SDS), and 10 µl 250 mM EDTA were added. Incubation was at 37°C for a further 2 h or until the solution was clear and there was no precipitate. The mix was either stored at –20°C or processed directly for gel electrophoresis. Samples (10.0 µl) were either directly loaded onto a gel or processed first by adding 1.5 µl of 1.5 N NaOH and incubating at 37°C for 30 min prior to loading.
Procedure 2. The pellet was resuspended in 100 µl TE and frozen and thawed three times. A mixture of 5.0 µl 3.0 M NaCl and 5.0 µl proteinase K (10 mg/ml) was added. The suspension was mixed. Then, 10.0 µl 250 mM EDTA and 5.0 µl 10% SDS were added. Incubation was at 37°C for 2 h. The mixture was then passed through a 16-gauge needle at least 15 times, followed by gel electrophoresis or storage at –20°C. Samples (10.0 µl) were directly loaded onto a gel or in some cases processed first by adding 1.5 µl of 1.5 N NaOH and incubating at 37°C for 30 min prior to loading.
Analysis. Samples were analyzed by electrophoretic separation on 0.8% agarose gels with Tris-borate-EDTA (TBE) buffer. Gels were dried, rehydrated, stained in water with SYBR Gold (Molecular Probes) for 90 min at a 1/10,000 dilution, and analyzed by using a PhosphorImager (Molecular Dynamics) with ImageQuant 1.1 software.
Hybridization of gels. Dried gels were rehydrated in water and incubated briefly in a prehybridization solution of 6x SScP (0.72 M NaCl, 0.09 M sodium citrate, 0.12 M sodium phosphate), 0.3% SDS, and 0.2 mg/ml salmon sperm DNA. Gels were hybridized for 16 h with either a random-primed probe of 32P-radiolabeled AAV fragment from the plasmid pAV2 or a T4 kinase-treated oligonucleotide representing the AAV D-region. Hybridized gels were washed briefly in 6x SScP with 0.5% SDS. Gels were redried and analyzed by phosphorimager.
|
RESULTS |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
293T cells were transfected with the helper plasmid pDG and with either pAV2, deletion mutants of pAV2, or pTRUF11, followed by harvesting as described in Materials and Methods. In order to determine the relative amounts of encapsidated and unencapsidated viral DNA, we employed the procedures shown in Fig. 1. During this processing, all material in the tube was rendered soluble. In addition, nothing was removed (i.e., no precipitation or extraction steps) until analysis by gel electrophoresis. In this way we could ensure that there would not be uncontrolled selective depletion of any species of nucleic acids.
Procedure 1 was designed to eliminate all genetic material not protected by the viral capsid. Upon the subsequent addition of NaOH, capsids were disrupted, allowing a determination of capsid-protected DNA. Procedure 2 was designed to preserve all genetic material but not disrupt the viral particles, thereby allowing a determination of unencapsidated material. If NaOH was added to procedure 2 as a final step, the total amount of both capsid-protected and unencapsidated DNA could be determined.
In the context of cellular debris and multiple freeze-thaw cycles, it proved difficult to disrupt the capsids. Digestion with proteinase K performed at 50°C in the presence of 2% SDS disrupted only approximately 15% of the particles. Digests with proteinase K at 37°C with 0.5% SDS, the conditions employed in procedure 1, failed to disrupt a detectable amount of particles. However, NaOH treatment appeared to readily disrupt the capsids, as was determined by the appearance of full-length AAV DNA. More-extensive NaOH digestion and higher temperatures as well as disruption by boiling did not result in greater amounts of AAV genomes (data not shown), implying that all encapsidated genomes had been released.
To confirm that the migration of AAV DNA in an agarose gel would not be affected by the presence of either total genomic DNA or proteinase K digestion products, procedure 1 or procedure 2 material was added to BglII-digested pAV2 (BglII digests pAV2 to release the full-length double-stranded [ds]AAV2 genome). There was little effect on the migration of the added viral genome in the gel (data not shown). To ensure that there could be no benzonase activity in the procedure 1 material at the time of capsid disruption, BglII-digested pAV2 was added to procedure 1 material. Additional benzonase and MgCl2 were then added to the mix. Additional benzonase was unable to digest the added plasmid in the presence of procedure 1 material, demonstrating that there can be at most only a minimal amount of nuclease activity at the time of capsid disruption (data not shown).
With these procedures, sufficient viral DNA was produced for direct visualization by SYBR Gold staining. Figure 2A demonstrates the results of procedure 1 and procedure 2 (each with and without NaOH) that were applied to cells transfected with pAV2. Lanes 1 and 2 show a BglII digest of the plasmid pAV2, i.e., the same plasmid with which the cells were transfected. (The AAV segment is identical to the ds intermediate in AAV replication except for the presence of one or five extra bases at the end of each strand consequent to BglII digestion.) In lane 1 the two pAV2 fragments produced by BglII digestion of pAV2 migrated at their expected positions. In lane 2 the same material was boiled for 5 min just prior to the loading of the gel in order to show the relative migration of the single-stranded (ss) AAV genome. Lanes 3 through 6 each contain the entire material from approximately 20,000 cells processed by either procedure 1 or 2, with or without NaOH. Lane 3, in which no SYBR Gold staining material is detectable, shows the contents of the cell when all free nucleic acids have been digested by benzonase. Lane 4, in which the same material was subsequently treated by NaOH, reveals a band that migrates indistinguishably from ss AAV. Lane 5, containing procedure 2 material, shows the entire contents of the cell untreated by any reagent which would destroy nucleic acids or release the contents of AAV capsids. A band which migrates at the position of ds AAV was detected. In lane 6, this procedure 2 material was additionally treated with NaOH, which would release encapsidated DNA. As with lane 4, lane 6 reveals a band migrating indistinguishably from ss AAV. Figure 2B demonstrates that the expected full-length ss AAV is only seen in cells transfected with both a construct containing a full-length AAV genome and a helper plasmid.
|
It should be noted that it cannot be determined from the data in lane 5 that all of the unencapsidated AAV DNA in the cell is in the ds form ss DNA may have annealed during the somewhat lengthy processing time of these samples. Similarly, after capsid digestion the ratio of ds and ss material changes with time; DNA released from capsids by NaOH treatment is initially completely ss but apparently undergoes a gradual transition from the ss form to the ds form despite the relatively high pH of the sample.
Effects of different levels of transfected plasmid. In order to determine whether different amounts of template would affect the amount of virus produced, we transfected three different amounts of pAV2, equimolar with pDG (i.e., 1x), 0.01 times as much as pDG, and 0.001 times as much as pDG. In all cases the amount of transfected pDG remained constant. We also transfected pTRUF11 (an ITR-containing plasmid with transgenes in place of rep and cap) at the 1x and 0.01x levels as an example of a recombinant plasmid and the deletion mutant pDA which, in cell-free systems, demonstrated high levels of DNA replication but which, due to its small size, would not be expected to encapsidate its DNA well (5, 30).
In order to determine the amount of capsid-protected material, we processed equivalent amounts of cells from each transfection mixture by procedure 1 with NaOH. The results are shown in Fig. 3A. As expected, pAV2 produced abundant virus. Somewhat surprisingly, the highest level of encapsidated genomes was derived from transfection at the 0.01x level, while encapsidated genome amounts at 0.001x were at least as high as at the 1x level. As expected, the production of encapsidated genomes by pTRUF11 was much lower and was not at a level detectable by SYBR Gold staining. (Infectious center assays as well as hybridization with pTRUF11-derived probes demonstrated that encapsidated TRUF11 is being produced [data not shown].) In conclusion, production of wild-type virus is not limited by the amount of initial template over a wide range of transfection levels.
|
Persistence of plasmid DNA. This assay also demonstrates the persistence of transfected plasmid despite benzonase digestion. In Fig. 3A, a band which migrates at approximately 9,000 nt is detected in lane 6 but not in the other lanes. Additionally, in every lane a pair of bands which only slightly migrate into the gel are also detected. The DNA species represented by these bands are too large to have been encapsidated. Nevertheless, they have persisted despite benzonase treatment that was sufficient to completely digest the entire genomic content of the cell. (Figure 3D shows by comparison the amount of cellular nucleic acid present in the same material not treated with benzonase and NaOH.) The migration of the high-molecular-weight bands is consistent with helper plasmid pDG, while the size of the 9,000-base band is consistent with linear pAV2. The lack of a 9,000-nt band in lanes 4 and 5, in which 1/100 and 1/1,000 as much pAV2 was added to the cells, reinforces the notion that this band is pAV2. Figure 3B, which shows hybridization of this gel with an AAV D-region probe, further supports the notion that the 9,000-nt band is pAV2. It also demonstrates that plasmid is undetectable in transfections performed at the 0.01x level. Figure 3C, which shows a rehybridization of the same gel with a pBR probe, indicates that the 9,000-base band contains pBR sequences, further supporting that this band represents the pAV2 with which the cells had been transfected. Interestingly, in lane 3 little benzonase-resistant pTRUF11 is detected. The cells of this transfection received the same molar amount of pTRUF11 as the cells of lane 6 received of pAV2.
The high-molecular-weight bands in Fig. 3A did not hybridize to a probe made from the AAV ITR D-region (Fig. 3B), indicating that these bands are not higher-molecular-weight forms of AAV DNA replication intermediates. They did, however, hybridize to a probe containing Rep and Cap sequences (data not shown), which is consistent with the notion that they represent the migration of pDG.
Figure 2B further supports the hypothesis that the high-molecular-weight bands and the band at 9,000 nt are derived from input plasmid. The high-molecular-weight bands were not detected in the lanes which did not receive pDG, lanes 1 and 2, but were detected in all other lanes. The putative pAV2 band was not detected in cells transfected with either no pAV2 (lane 1), lesser amounts of pAV2 (lane 4), or with pAV2DA, a mutant of pAV2 in which almost all internal AAV sequences are deleted but which replicates its DNA well if Rep is supplied in trans (lane 5). In addition, the pAV2 band was not detected in virus-producing cells which were not transfected with pAV2 but rather infected with wild-type AAV particles (lane 6). In Fig. 2 and 3, the presence of small amounts of a species consistent with a double-stranded AAV genome was detected. This apparently resulted from annealing of the single-stranded form that had been released from the capsids. In these two assays, the samples were not separated by electrophoresis immediately after capsid disruption.
An approximate quantification of the pAV2 band in lane 6 of Fig. 3A, B, and C indicates that it represents approximately 50% of the transfected plasmid. Apparently, transfected plasmids can occupy a privileged niche in the cell such that when the cells are disrupted solely by three freeze-thaw cycles, these plasmids remain resistant to digestion by benzonase. In Fig. 3B, lane 6, which shows material derived from cells transfected with pAV2 at the 1x level, a substantial percentage of the D-region-hybridizing signal is derived from intact input plasmid and lower-molecular-weight material (most likely digested plasmid; see below). Consequently, in hybridization assays to determine the amount of genome-containing particles, a considerable portion of the signal may in some cases inadvertently be derived from the transfected plasmid.
The hybridization of Fig. 3B shows an abundant smear of faster-migrating material that hybridizes to the D-region probe. The relative absence of this smaller material in lanes 4 and 5, which show cells transfected with substantially less pAV2 but which show greater amounts of AAV DNA synthesis and encapsidation than did lane 6, suggests that this faster-migrating material represents partial digestion of the transfected pAV2 rather than the encapsidation of less-than-full-length genomes. The smear at the bottom of all lanes in Fig. 2B, which must in some lanes represent free genomic DNA rather than encapsidated viral DNA, is consistent with the notion that the smear at the bottom of lane 6 (Fig. 3B) represents incomplete digestion by benzonase. This point is reinforced by Fig. 2B, lane 7, which shows no low-molecular-weight species in encapsidated material found in the cell culture medium. In this assay, pAV2/pDG-transfected cells were spun down and the medium was removed. Virus in this medium was subsequently concentrated by high-speed centrifugation. This cell-free material was then processed according to procedure 1. In summary, Fig. 3A and B and 2B together make the point that, within the limits of the size resolution of this gel, essentially all capsid-protected AAV genomes are full length. Limiting dilution assays of the produced virus showed that there was at least one infectious unit per 40 genome-containing particles.
Unencapsidated genomes. To gain insight into whether the much greater production of virus with pAV2 in comparison with pTRUF11 was due to a failure of DNA replication in the latter, we measured the amount of unencapsidated genomes. Figure 3D, lanes 1 to 6, shows material treated according to procedure 2. Each lane represents approximately 20,000 cells and corresponds to lanes 1 to 6 in Fig. 3A. The gel of Fig. 3D shows all nucleic acids from the cells that do not require NaOH treatment for detection. This includes all the nucleic acids that are not sequestered by viral particles. In the pAV2 lanes and the pTRUF11 lanes, the principle viral bands migrate consistent with ds forms. (As mentioned above, this should not be taken as indicating that all nonencapsidated viral DNA in the cell is ds, since ss genomes might have annealed during processing. The amount of rRNA is seen to vary from lane to lane. The rRNA quickly but variably degraded in our samples, since no efforts were made to prevent RNA degradation.) As expected for a successful replication, there was only a relatively modest amount of dimer-length material in all cases. The dimer-length band for 1x pAV2 was essentially undetectable, which correlates with the somewhat lesser viral production in this assay. (The higher-molecular-weight bands in lane 6 are compatible in migration with linear and open circular pAV2.) Somewhat surprisingly, there was little difference in unencapsidated genomes for pAV2 or for pTRUF11, whether cells were transfected with 1x or 0.01x plasmid or, in the case of pAV2, 0.001x. More surprising is the observation that there was as much unencapsidated TRUF11 DNA as there was AAV DNA. This last result is consistent with the notion that the failure of the pTRUF11 plasmid to produce abundant virus is not a failure of DNA replication.
A direct comparison of encapsidated versus unencapsidated genomes for AAV2 and TRUF11 is shown in Fig. 3E. Lanes 3 and 6 show relative amounts of encapsidated genomes in 20,000 transfected cells. The amount of encapsidated TRUF11 genomes produced by approximately 20,000 cells is, as expected, below the limit of detection by SYBR Gold staining. Lanes 2 and 5 demonstrate that there is a comparable amount of unencapsidated genomes for AAV2 and TRUF11. A comparison of lanes 1 and 4 (which represent the total nucleic acids in the cell, i.e., including both encapsidated and unencapsidated genomes) shows that there was substantially more DNA synthesis of the AAV2 genomes than of the TRUF11 genomes. Figure 3 demonstrates that most of the viral DNA synthesized in the pAV2-transfected cells has been encapsidated. The much lesser amount of viral DNA produced in the pTRUF11-transfected cells has mostly not been encapsidated. In conclusion, the difference in the total viral genomic content between pAV2- and pTRUF11-transfected cells is almost entirely accounted for by the encapsidated genomes.
Deletion mutants of AAV. Previous data from other groups as well as the above results demonstrating equivalent amounts of unencapsidated DNA when comparing pTRUF-11- and pAV2-transfected cells suggest that the lower titers of rAAV compared to wild-type AAV are principally a failure in encapsidation. It has been shown that in the autonomous parvovirus minute virus of mice (15), the viral genome contains internal cis-acting packaging signals. It has been noted that in the production of rAAV, more empty capsids are produced than in synthesis of wild-type virus (10). This latter observation suggests the possibility of an encapsidation signal in the wild-type AAV genome.
In order to determine if any internal, i.e., non-ITR sequences, played a significant role in AAV DNA encapsidation and to further explore the possible role of such sequences in viral replication, we constructed a series of deletion mutants (Fig. 4). We reasoned that constructs with deletions but without replacement sequences would give the most interpretable results. Differences in virus production between the deletion mutants and wild type could be due only to a missing specific sequence, a decrease in length, or to a change in the relative positioning between remaining sequences of importance.
Effects of deletions on DNA synthesis and encapsidation. Viral production of several deletion constructs as well as of the recombinant pTRUF11 are shown in Fig. 5A. This material was generated by transfection of equimolar amounts of the plasmid into 293T cells followed by procedure 1 with NaOH treatment and as such showed only the encapsidated DNA in the cell. Figure 5A shows that the deletion mutants fall into two classes. SS, DS, and SB produce abundant virus, though somewhat less than the wild-type construct. SX, SD, and DD produce barely detectable amounts of encapsidated genomes.
|
Interestingly, deletion DS, which encompasses the p5 promoter or CARE element that has been implicated in Rep-dependent DNA replication of constructs with absent or deficient ITRs (7, 25, 32), does not seem to produce substantially less virus than deletion constructs from the left side of AAV that did not encompass the element.
Next, we processed equivalent aliquots of cells from the same transfection mixture according to procedure 2 in order to determine the amount of unencapsidated genomes. The results, shown in Fig. 5B, make a simple point, namely, that cells transfected with constructs that produced very low amounts of encapsidated genomes, i.e., SX, SD, and DD, contained as much unencapsidated genomes on a molar basis as cells with abundant encapsidated genomes. Indeed, for all seven constructs, the amount of unencapsidated genomes was remarkably similar. Figure 3D and E had previously demonstrated that cells transfected by the recombinant construct pTRUF11 also contain equivalent amounts of unreplicated genomes to pAV2. As with pTRUF11, the failure of constructs SX, SD, and DD to produce encapsidated genomes does not seem due to a failure to replicate their genomes. (The absence of detectable rRNA in this assay is due to the storage of the samples at –20°C for several months between procedure 2 and gel electrophoresis.)
Table 1 shows relative values for total viral genome synthesis (as determined by phosphorimager analysis) and encapsidated genomes comparing the wild-type construct pAV2 with two left-side deletion constructs, SS and SB. The left-side deletions show a lesser amount of DNA replication and production of encapsidated genomes compared to the wild-type construct. For both replication and encapsidation, the results suggest that the construct with the larger deletion, SB, is not less efficient than the construct with the smaller deletion, SS. It appears that the difference between the wild type and mutants with respect to encapsidated genomes is greater than the difference with respect to total DNA replication, i.e., in the case of the wild-type construct a greater proportion of the total synthesized viral DNA is encapsidated. However, Fig. 5 shows that the amount of unencapsidated genome for each construct is remarkably constant. Therefore, since apparently for each construct a relatively fixed amount of genome remains unpackaged, it is to be expected that the proportion of synthesized genomes which become encapsidated would be somewhat less for the SS and SB genomes.
|
Grieger and Samulski observed that packaging efficiencies of transgenes that ranged from 4.4 to 6.0 kb were similar (9). Previously it had been reported, however, that constructs below a certain minimal length package poorly (5). Tullis and Shenk also observed less viral production below a certain length but correlated this poorer encapsidation to a difference in DNA synthesis (30). They noted that with shorter genomes terminal resolution occurred less readily, leading to the production of more concatemers, which apparently remained unencapsidated. However, in the case of SB and BX, the proportions of concatemers, which because of their size must remain unencapsidated, were essentially equivalent.
In order to answer the question of why SX failed to produce much encapsidated product while SB did, we tested several additional mutants. Figure 6A shows material from transfected cells processed by procedure 1 with NaOH, showing the relative amounts of encapsidated genomes. It should be noted that the difference between encapsidated pAV2 and encapsidated SS (4.8-fold as determined by phosphorimager analysis) is the largest we observed in numerous replication assays. As in Fig. 5A, a large difference in encapsidated genomes between SX and SB was apparent. In order to determine whether the failure of SX to encapsidate was due to the increase in size of the deletion or to some specific element present in the 530 bases between the B and X sites, we tested a construct which was deleted only between B and X, the BX construct. This construct did not encapsidate despite being longer than SB, which does encapsidate. This suggests that the absolute length of the construct is not the reason for the failure of BX and SX to encapsidate, but rather that it is the deletion of the specific sequence between B and X. Two alternative possibilities suggest themselves. The first is that there is a specific packaging sequence contained within the B-to-X segment. The second possibility is that deletions to the right of the BstE2 site fail to encapsidate well because the cap gene and the consequent production of Cap have been disturbed. (The B-to-X region encompasses the p40 promoter and the ATG for VP1.) To distinguish between these alternatives, the construct BsW was tested. It contains a 3-base insertion at nucleotide 3255 (Fig. 6C), introducing a stop codon into the cap gene. As can be seen in lane 3 of Fig. 6A, the encapsidation of this construct is low. Since the BsW 3-base insertion minimally disturbs the genomic sequence and is remote from the SX deletion, it seems unlikely that both constructs are affecting some specific packaging signal. Figure 6B, which shows unencapsidated DNA, demonstrates that these two nonencapsidating mutants, lanes 3 and 4, produce as much unencapsidated DNA as the constructs that do encapsidate. It appears that the constructs which encapsidate poorly, SX, SD, DD, BX, and BsW, do so because they are unable to produce Cap protein.
|
It can be determined from Fig. 6A that, due to the shorter length of BX compared to AV2, in the cotransfected cells there is little encapsidated BX. Since the lengths of AV2 and BsW are indistinguishable, it was not possible to directly determine whether the BsW construct was being encapsidated. Therefore, we digested the encapsidated and unencapsidated DNA obtained from these cells with BsiWI. BsiWI would digest only the AV2 construct and leave the BsW construct intact due to the 3-base insertion into the BsiWI restriction site. In order to digest this material, it was necessary to phenol-chloroform extract and ethanol precipitate it to remove the SDS and EDTA. Extraction, precipitation, and resuspension also have the benefit, in this case, of promoting the annealing of the single-stranded viral genomes. To ensure that BsiWI digestion was complete, a small amount of full-length linearized pAV2 (band M) was added to each tube immediately prior to the addition of restriction enzyme. This substrate contains the same sequence as the wild-type AAV virus, which was produced from pAV2. Digestion of this added substrate (band M) produces two species (bands M1 and M2). Figure 7 shows the results of this assay. Lanes 3, 4, and 5 show encapsidated and unencapsidated material from cotransfected, BsW-transfected, and AV2-transfected cells, respectively. The ratio of band A to bands A1 and A2 shows how much material was resistant to BsiWI digestion, i.e., how much material was the BsW construct. From lanes 3 and 5 it can be determined that there was little or no encapsidated or encapsidated plus unencapsidated BsW genomes in the cotransfected cells. The small amount of BsiWI-resistant genome in lanes 3 and 5 most likely represents a less-than-perfect annealing of the single-stranded genomes. The presence of a small amount of BsiWI digestion product in lane 4 may represent recombination between the BsW and pDG constructs in the transfected cells. Therefore, in the cotransfection experiments in which pAV2 was packaged efficiently, BX and BsW were not. A comparison of the mutant alone and the cotransfected lanes in Fig. 7, most easily visualized with the BX mutant because of its size difference from the wt, illustrates that not only is the absolute amount of encapsidation higher for the wt genome but the ratio of encapsidated to unencapsidated is higher for the wild type than for the mutant. This suggests that failure to encapsidate the non-Cap-producing constructs is not likely to be a function of insufficient capsids present in the cell, since in both cotransfection experiments there is ample capsid to result in substantial production of wild-type virus. Apparently, even when there is abundant capsid present in the cell that can efficiently support encapsidation in cis, this same capsid does not support efficient encapsidation of genomes for which it was produced in trans.
|
|
DISCUSSION |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
In this report we made use of an assay which unambiguously distinguishes unencapsidated from encapsidated DNA. It proved possible to release the cell's nucleic acids in such a manner as to enable them to be separated by agarose gel electrophoresis, without disrupting the AAV capsids. The convenient size of the AAV genome, larger than 28S rRNA but smaller than sheared genomic DNA, allowed direct visualization of the unencapsidated viral DNA. The stability of the AAV capsid made it possible to digest the cellular nucleic acids (except for transfected plasmids, as discussed) while leaving the capsid protected genomes intact.
A first question was whether the procedure 1 material represents genuinely encapsidated genomes or genomes that are merely protected from benzonase/DNase digestion by capsid particles. The reasons for proposing the former are severalfold. First, we do not observe free genomes in procedure 1-treated material (without NaOH treatment) that has been separated by gel electrophoresis, despite the addition of 0.5% SDS to the gel loading buffer. Second, we do not detect any dimer-length material after procedure 1 treatment. Procedure 2 gels demonstrate that the infected cells have significant amounts of dimer-length forms. It seems unlikely that if monomer-length forms of the viral genome could bind to the capsid proteins or some cellular component in such a way as to become nuclease resistant that dimer-length forms could not also bind in a similar manner. Third, the deletion mutants which replicate but show little capsid-protected DNA also argue that replicated but unencapsidated viral DNA in the cell is unprotected. Fourth, the great majority of the unencapsidated viral genomes in our transfected cells are double stranded, with one terminus in the closed hairpin conformation. In contrast, all the genomes revealed by procedure 1 NaOH treatment have, as expected, termini which are not closed hairpins.
We did, however, observe transfected plasmid which survived benzonase/DNase digestion. This plasmid material was detectable in equal amounts with or without NaOH treatment. That the survival of this plasmid material is not dependent on the presence of AAV capsids is shown in Fig. 2B, in which pAV2 is detected in cells that do not contain helper factors and therefore do not produce virus. It is unclear why the plasmids were nuclease resistant when digestion of both genomic DNA and the abundant unencapsidated AAV genomes produced by AAV DNA replication was complete. Apparently much of the transfected plasmid was sequestered in a cellular compartment not sufficiently disrupted by the freeze-thaw cycles to make it accessible to the nucleases. Presumably, the SDS, EDTA, and proteinase K digestion subsequent to nuclease digestion disrupted these compartments. Interestingly, the rep- and cap-containing plasmids, pDG and pAV2, seemed more resistant than the plasmids pTRUF11 and pAV2-DA.
A corollary of this observation is the possibility that survival of input plasmids can distort the results of standard genome titrations. More importantly, if the helper plasmid remains with the viral particles during subsequent purification it may be delivered to future cells along with the recombinant virus, leading to the possible expression of Rep and Cap.
In the cells transfected by the various constructs, an unexpected observation was that the amount of unencapsidated DNA does not vary greatly between constructs (Fig. 3, 5, and 6). This is despite great differences in the viral production by these constructs. We also observed that the amount of unencapsidated genomes in cells infected by AAV was equivalent to the amount in cells transfected by the various plasmid constructs (data not shown). The fact that the unencapsidating genomes have as much, or more, unencapsidated DNA than do encapsidating genomes argues that a failure in DNA replication is not an explanation for why some constructs and the recombinant pTRUF11 demonstrate little encapsidation.
One explanation for the unexpected similarity in the amounts of unencapsidated DNA in each case might be that encapsidation commences only when a certain level of viral DNA is reached. This explanation, however, cannot explain why different nonencapsidating genomes would each reach approximately the same DNA level and, moreover, why this level is approximately equivalent to the level of unencapsidated genomes maintained by constructs that efficiently encapsidate. A simple alternative explanation is that, rather than encapsidation being limited by DNA replication, DNA replication is limited by encapsidation. In this alternative explanation, genomes must be continually removed from the replication complex by encapsidation, in order to allow DNA synthesis to proceed. A mechanistic explanation for consideration might be that, as free genomes accumulate in the cell, the increasing number of potential replication substrates dilutes essential DNA replication factors to the point at which DNA replication stalls.
We observed, as have Tullis and Shenk, that short substrates, such as DA and DD, showed larger proportions of multimers than did longer constructs. Tullis and Shenk concluded that the most likely explanation for this is that the replication complex traverses the distance between the ITRs in these shorter substrates too quickly for efficient terminal resolution by Rep (30). In that sense, DNA replication can proceed on these templates to a greater extent without the creation of additional terminal origins of DNA replication. The relatively longer nonencapsidating constructs in this report, however, did not show increased levels of replication or a higher proportion of multimers, implying that their failure to encapsidate was not due to the problem with terminal resolution, particular to very short constructs.
We observed a modest decrease in viral production with deletions from the left side of the AAV genome. When Tullis and Shenk tested a deletion in their wild-type (pSUB201) construct from nt 194 to 1882, they noted a decrease in viral replication of about sixfold when compared with wild type (30). It is likely that our report is describing the same phenomenon noted by Tullis and Shenk. Salvetti et al. described an element (CARE) which is found between nucleotides 190 and 540 which can play a cis-acting role in DNA replication of the wild-type genome (25). This element was later localized specifically to nucleotides 250 to 304 (and is therefore p5 associated) (7). However in this report a deletion from nt 810 to 1060 resulted in a two- to threefold reduction in DNA synthesis. This deletion does not overlap with the element characterized by Salvetti and colleagues and so presumably exerts its effect by a separate mechanism. The implication is that the decrease in replication described in this report as well as originally by Tullis and Shenk is a separate phenomenon from the effects on DNA replication of the p5-associated element described by Salvetti and her colleagues. They have reported that in preliminary assays the p5 element does not affect viral production by transfected plasmids containing two intact ITRs (7). They suggest that the replication effect of the p5 element is masked in the context of the ITRs or that its role may be in other steps of the AAV life cycle, such as integration. Correspondingly, when we tested the DS construct, which contains a deletion from nt 235 to 810 and which therefore encompasses the element described by Salvetti and her colleagues, we cannot detect an additional significant decrease in replication or encapsidation over the decrease produced by left-side deletions which do not contain this element.
As noted, all the left-side deletions tested gave similar results, on average a threefold reduction in viral production as determined by procedure 1. Therefore, the most likely explanation is that the reduction is due not to the absence of particular cis-acting sequences but rather to an absence of a functional Rep protein produced from the replicating construct. Since pDG produces sufficient amounts of Rep (11) and since an excessive amount of Rep has been shown to lessen AAV production (21), the reason for this observation remains unclear.
The reduction in viral production observed when Cap sequences were disturbed, however, was much greater. It is unlikely that the insertion of three nucleotides in the BsW construct or the deletion of nucleotides 1700 to 2233 in the BX construct could have each disrupted some essential cis-acting packaging signal. Except for pTRUF11 and the BsW mutant, which has three added nucleotides, none of the nonencapsidating constructs contain non-AAV sequences, implying that failure to encapsidate need not be based on disruptive exogenous sequences. The insertion of a stop codon in the cap gene at nucleotide 3255 suggests that it is expression of Cap proteins from the construct to be packaged that is responsible for the much higher yield of virus from wild-type compared with recombinant constructs. Also striking was the failure to encapsidate the BsW and BX constructs in the cotransfection experiments. Since a further serial dilution of the amount of pAV2 on these cells showed that total viral production did not fall significantly until amounts of plasmid per cell were below 100-fold less than was used in these assays (data not shown), it is likely that all cells which could be transfected were transfected. In addition, since the differences in sequence between pAV2 and the two mutants are not large, it is unlikely that the plasmids are being taken up by different subpopulations of 293T cells. The optimal encapsidation of the pAV2 construct in both these assays demonstrates the presence of substantial amounts of functional capsid in these cells; nevertheless, the BX and BsW constructs are not efficiently encapsidated.
In the case of the wild-type construct, Cap proteins are in greater abundance than recombinant constructs because they are expressed from a replicating template. The observation of previous investigators (10), in which empty capsids were found in much greater excess in cells transformed with recombinant genomes, argues that reduced capsid production cannot be the whole explanation. The above cotransfection experiments further support the argument by showing that even in the presence of plentiful capsid, this capsid material is preferentially encapsidating the genomes expressed in cis.
It is not clear what the provision of Cap "in cis" entails in this system. The results seem to suggest that there are localizations within the cell between replication/transcription centers and the protein products of those transcription complexes.
The presence in all cells of approximately equivalent amounts of unencapsidated genomes whether virus production/encapsidation is high or low lends support to the notion of a DNA replication-encapsidation link. (It has been noted previously that DNA replication and encapsidation appear to be a coupled process [24, 36].) In a model for this coupling, a stall in either mechanism might terminate production in a manner not easy to restart. In this scenario DNA replication might be unable to recommence before encapsidation has reduced the pool of genomes capable of initiating replication, but encapsidation might be unable to resume without an active DNA replication to present genomes in an appropriate fashion to the packaging complex. The suggestion is that the provision of Cap in trans does not efficiently support the DNA replication link.
The possibility that an interruption in encapsidation might enforce a subsequent limit on DNA replication suggests caution about the interpretation of data regarding DNA replication and viral production. We observed a modest decrease in total DNA synthesis with Rep mutants and a corresponding decrease in viral production. Is this lower viral production due to reduced viral DNA synthesis or to reduced encapsidation, which thereby enforces a limit on total DNA replication? From these data it cannot be determined which is the primary cause.
While this report supports the notion that the provision of Cap in trans is a reason for the lower yields of non-wild-type constructs, it does not eliminate the possibility that there is in addition a Cap region packaging signal whose effect is masked by the failure of the cap deletion mutants to package. For example, the N terminus of VP1 and VP2, which has been shown to be sequestered within the virion (18), has also been shown to interact with the encapsidated genome (17). It remains a possibility that this interaction is in part sequence specific.
The observations of this report suggest that, to a first order, the failure of recombinant genomes to produce virus with the efficiency of wild-type genomes would seem not due to the presence of inhibiting sequences or to the absence of a required sequence in the recombinant genome, but rather to the seemingly more tractable issue of providing Rep and Cap in an optimal manner. Determining how to do this in the case of recombinant genomes is likely to be a key to both increasing the yield of recombinant virus per cell and to reducing the quantity of empty particles.
|
ACKNOWLEDGMENTS |
|---|
We thank Susan Cotmore, Nathalie Dutheil, Jurgen Kleinschmidt, Robert Kotin, Patricia Rebollo, and Peter Tattersall for helpful discussions.
|
FOOTNOTES |
|---|
Published ahead of print on 11 July 2007.
|
REFERENCES |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
| ||||
