Adeno-Associated Viruses Can Induce Phosphorylation of eIF2α via PKR Activation, Which Can Be Overcome by Helper Adenovirus Type 5 Virus-Associated RNA

Abstract

Mutants of adenovirus type 5 (Ad5) virus-associated RNA I deficient in inhibiting the activation and subsequent phosphorylation of protein kinase R (PKR) could neither function as helpers for adeno-associated virus type 5 (AAV5) replication nor enhance AAV5 protein accumulation in either the presence or absence of Ad5 E4Orf6 and E2a. Furthermore, a short region of the AAV5 capsid gene RNA leader sequence surrounding the AUG of VP1 could induce the phosphorylation of eIF2α. Both short interfering RNA directed against PKR and the addition of the herpes simplex virus ICP34.5 protein enhanced the accumulation of AAV5 capsid protein in the presence of the AAV5 capsid gene PKR-inducing element, suggesting that VA RNA acted to overcome direct AAV5-induced activation of PKR that led to the phosphorylation of eIF2α. The expression of both the closely related goat-derived AAV and the prototype AAV2 capsid gene transcription units also induced the phosphorylation of eIF2α, suggesting that the induction of the PKR/eIF2α cellular response may be a previously unrecognized general feature of at least the Dependovirus genus of the Parvovirinae.

Adeno-associated viruses (AAVs) are dependoviruses which require the helper virus functions of larger DNA viruses for efficient replication. Adenovirus type 5 (Ad5) has been shown to be the most efficient helper virus for AAV, and the Ad5 factors necessary to support AAV replication are E1A, E1B-55K, E4Orf6, E2a, and virus-associated (VA) RNA. The roles that these gene products play during Ad5 infection have been extensively characterized (1, 2, 16, 26, 28), and their roles in supporting AAV type 2 (AAV2) infection have been the object of study for many years (1, 2, 26, 28).

The Ad5 E1A and E1B-55K proteins have been reported to perform a number of essential roles during AAV gene expression. E1A has a well-defined role in the regulation of the expression of AAV2 P5 and AAV5 P41 (2, 9, 17, 31); however, the role of E1B-55K is less clearly defined. Additional roles that Ad5 may play in other aspects of AAV gene expression are as yet only partially understood (2, 28). Ad5 E2a is a single-stranded DNA (ssDNA) binding protein, and that activity has been shown to function during the strand-displacement phase of AAV2 inverted terminal repeat (ITR)-mediated genome replication (24). The E4Orf6 protein of various Ad serotypes has been shown to be required for the successful completion of second-strand synthesis, both in vitro and in vivo (5, 6); however, the mechanism by which this is accomplished is not yet fully understood. E4Orf6 has been suggested to play other roles during AAV2 replication as well (25, 26, 29). Finally, the Ad VA RNA, a small RNA produced abundantly during Ad infection, had been suggested as enhancing AAV gene expression at the posttranscriptional level, most likely at the level of protein translation (8, 25, 29). VA RNA is not itself translated. It acts as a regulatory RNA during the Ad5 life cycle, and its role in enhancing the translation of Ad5 proteins during Ad5 infection is well characterized (11, 12).

In normal cells, there are a number of innate mechanisms that play important roles in inhibiting productive viral replication. An important participant in this process is the serine-threonine kinase protein kinase R (PKR). In response to activation by double-stranded RNA, PKR undergoes autophosphorylation and brings about the phosphorylation of the translation initiation factor eIF2α (22, 30). Unphosphorylated eIF2α can participate in new rounds of translation when GDP, bound to eIF2α, can be exchanged for GTP in the presence of guanine exchange factor. However, when eIF2α is phosphorylated, it forms a tight complex with guanine exchange factor, so the GDP-GTP exchange cannot occur and translation is blocked (13). Many viruses are known to disturb this system in order to facilitate the protein expression needed for their replication. This is accomplished by a number of mechanisms. For example, the influenza virus protein NS1 inhibits the phosphorylation of PKR by sequestering double-stranded RNA (10); the hepatitis C NS5A protein inhibits PKR phosphorylation by directly binding to PKR (18); the herpes simplex virus (HSV) protein ICP34.5 inhibits protein phosphatase I kinase (PPI kinase), a modification which allows PPI to dephosphorylate eIF2α, preventing its inactivation (27); and the Ad5 VA I RNA binds to the RNA binding domain of PKR, thus competitively inhibiting it (4, 15).

Ad 5 produces two distinct VA RNAs, VA I and VA II. They are 160-nucleotide (nt) RNA molecules which are transcribed by RNA polymerase III and not translated. VA I RNA is the more abundant species (13). It differs considerably at the nucleotide level from VA II, yet both form similar stable secondary structures (11). The secondary structure of VA I is composed of an apical stem-loop, central domain, and terminal stem (11). The regions of VA I RNA which bind to PKR have been identified, and nucleotide substitutions which disturb the integrity of the central domain have been shown to severely hamper the ability of VA I RNA to inhibit the phosphorylation of PKR in vitro (19, 21).

We have recently shown that the Ad5 E4Orf6 protein, together with E1B-55K, can target AAV5 capsid and small Rep proteins for degradation via the proteasome and that at least part of the required help that VA I RNA provides during AAV5 infection is to increase the accumulated levels of viral proteins necessary for replication (16). Our work also showed that the accumulated levels of AAV5 capsid and small Rep proteins were enhanced by VA I RNA when their transcription units were expressed from plasmid expression vectors. These observations led us to suspect that expressed AAV5 RNA might induce cellular PKR, with subsequent phosphorylation of eIF2α, although this had not been previously reported for any parvovirus. In this scenario, Ad5 VA I RNA could increase the accumulated levels of AAV5 proteins by competitively inhibiting PKR.

In this study, we found that VA I RNA mutants deficient in inhibiting the phosphorylation of PKR no longer function as helpers for AAV5 replication or capsid protein accumulation. Also, we report that a short region of the leader sequence of the AAV5 capsid gene RNA activated PKR, resulting in the phosphorylation of eIF2α. Both short interfering RNA (siRNA) directed against PKR and the addition of HSV ICP34.5 enhanced the accumulation of capsid protein in the presence of the PKR-inducing element, suggesting that during AAV5 infection, VA I RNA acts to overcome the direct AAV5-induced activation of PKR that leads to the phosphorylation of eIF2α.

MATERIALS AND METHODS

Cells and viruses.

293 cells and HeLa cells were propagated in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum as previously described.

Plasmid constructions.

The wild-type VA I RNA-expressing plasmid VA RNA wt was constructed by cloning the VA I RNA gene from the standard AAV helper plasmid pHelper (14) into the BamHI-ClaI sites of pBS-SKII plus. Plasmids VA A119G, VA C120A, VAΔIV, VA TT101/102CA, and VA NES KO were generated from VA RNA wt and were constructed to be identical to mutants characterized in references 7, 19, and 21. RNase protection assays (RPAs) were used to determine that all the mutants were expressed to the same levels after transfection. VA A119G contains a replacement of A with G at position 119, VA C120A contains a replacement of C with A at position 120, VAΔIV contains a deletion between nt 92 and 118, VA TT101/102CA contains a replacement of nucleotides T and T at positions 101 and 102 with C and A, and VA NES KO contains replacements of nucleotides G, A, and C between positions 138 and 140 with A, T, and A, all as previously described (7, 19, 21). Plasmids CMV E2a, CMV E4Orf6, the minimal capsid plasmid p41, and the DH probe have been described before (16).

CMV 18 V5 Cap, CMV 19 V5 Cap, CMV 20 V5 Cap, CMV 21 V5 Cap, CMV 23 V5 Cap, CMV 26 V5 Cap, and CMV 27 V5 Cap were generated by inserting AAV5 cap regions spanning nt 1891 to 4448, 1993 to 4448, 2061 to 4448, 2131 to 4448, 2312 to 4448, 2602 to 4448, and 2751 to 4448, respectively, between the HindIII and XbaI sites of pcDNA3. The Cap probe contains the AAV cap region spanning nt 2783 to 3083 of AAV5 between the HindIII and BamHI sites of the PGEM3Z plasmid (Promega, Madison, WI).

The CMV GFP construct was constructed by cloning the green fluorescent protein (GFP) gene from pGET007 into the BamHI and ApaI sites of pcDNA3. Next, the constructs CMV 18-27 V5 GFP, CMV 19-27 V5 GFP, CMV 20-27 V5 GFP, CMV 21-27 V5 GFP, CMV 23-27 V5 GFP, and CMV 26-27 V5 GFP were generated by inserting AAV5 cap regions spanning nt 1891 to 2781, 1993 to 2781, 2061 to 2781, 2131 to 2781, 2312 to 2781, and 2602 to 2781, respectively, between the HindIII and BamHI sites of CMV GFP. The plasmid expressing ICP34.5 was a kind gift from David Lieb, Washington University, St. Louis, MO.

Analysis of intracellular AAV5 DNA forms.

293 cells were plated at 4 × 10⁵ cells/well in 6-well dishes. At 37 h after seeding, the cells were infected with AAV5 (multiplicity of infection, 10) and, 1 h after infection, were transfected with equal amounts of DNA (3 μg/well of a 6-well dish) using Lipofectamine and the Plus reagent (Invitrogen, Carlsbad, CA), as previously described. At 48 h posttransfection, cells were suspended in spent tissue culture medium and aliquoted equally into two separate centrifuge tubes. One aliquot was used for the analysis of intracellular AAV5 DNA, and the second aliquot was used for immunoblot analysis as described below.

Immunoblot analyses.

293 cells seeded in 60-mm dishes were transfected with Lipofectamine and the Plus reagent with 6 μg of plasmid DNA/60-mm-diameter dish as previously described. At 37 h posttransfection, cell lysates were prepared and subjected to immunoblot analysis as previously described (20). Monoclonal antibody 303.9, raised to amino-terminally truncated AAV2 Rep78, and anti-AAV Cap antibody (clone B1) were obtained from American Research Products, Inc., Belmont, MA. Antibody to β-actin (catalog number ab-32800-500) was obtained from Abcam (St. Louis, MO), and antibody to 14-3-3 proteins (catalog number sc-1020) was obtained from Santa Cruz.

Briefly, cells were washed with ice-cold phosphate-buffered saline prior to being scraped into the same buffer and collected by centrifugation. The cells were then lysed in four packed cell volumes of lysis buffer (20 mM HEPES, pH 7.5, 150 mM NaCl, 1% Triton X-100, 10% glycerol, 1 mM EDTA, 10 mM tetrasodium pyrophosphate, 100 mM NAF, 17.5 mM beta-glycerophosphate, and 1 mM phenylmethylsulfonyl fluoride) on ice. The lysates were then centrifuged at 14,000 rpm for 30 min at 4°C. After transfer to a nitrocellulose membrane, the samples were immunoblotted with anti-Rep, anti-Cap, or anti-phospho-eIF2α antibody.

RNA isolation and RPAs.

Total RNA was isolated using guanidine isothiocyanate, and RPAs were performed as previously described (23). The homologous RNase protection probe “cap” (AAV5 nt 2783 to 3083) was generated from linearized templates by in vitro transcription using SP6 polymerase, as previously described (20). RNA hybridizations for RPAs were done with a substantial excess of probe, and RPA signals were quantified with a molecular imager FX and Quantity One software (version 4.2.2 image software; Bio-Rad, Hercules, CA). The relative molar ratios of individual species of RNAs were determined after adjusting for the number of ³²P-labeled uridines in each protected fragment as previously described (20).

siRNA transfections.

293 cells, seeded at 60 to 80% confluence, were transfected with the shortcut siRNA to PKR (catalog number N2008S; New England Biolabs, Beverly, MA) using Transpass R1 reagent (QIAGEN, Valencia, CA). At 24 h after siRNA transfection, plasmid transfection with Lipofectamine and Plus reagent was carried out.

RESULTS

VA RNA mutants deficient in inhibiting the activation of PKR were unable to support full AAV5 replication or overcome the degradative effects of E4Orf6 to increase accumulated levels of AAV5 protein.

VA RNA is a necessary helper factor for AAV5 replication. We have previously shown that in its absence, wild-type levels of the AAV5 double-stranded monomer replicative form (mRF) could be produced; however, the levels of progeny ssDNA were significantly diminished (16). One of the functions of VA RNA during AAV5 infection is to increase the accumulated levels of AAV5 capsid and small Rep proteins (16). This effect helps counteract a degradative effect of E4Orf6 on these viral proteins (16), which is necessary for complete viral replication and the generation of ssDNA. VA RNA has a well-characterized role in suppressing innate cellular antiviral responses by binding to and competitively inhibiting PKR during Ad5 infection (22). Therefore, we chose to investigate whether this property was required for the enhancement by VA RNA of AAV5 protein accumulation and genome replication.

We first introduced into a cloned VA I RNA construct a set of extensively characterized PKR activation mutants first generated by Rahman et al. (21). The ability of these mutants to both support AAV5 replication and enhance viral protein accumulation in the presence of the other Ad helper factors, E1A, E1B-55K, E2a, and E4Orf6, was then compared to that of wild-type VA RNA. We have confirmed the expression and stability of each of the mutant VA RNAs following their transfection into 293 cells by performing RPAs using two specific probes to different portions of the target molecule (data not shown).

As demonstrated previously, the combination of the cytomegalovirus (CMV)-expressed Ad5 E2a and E4Orf6 proteins, together with wild-type VA RNA, supported full AAV5 replication in Ad5 E1A- and E1B-55K-expressing 293 cells (Fig. 1A, lane 1 [although the levels of ssDNA isolated in this particular experiment were low]), and high levels of both Rep and Cap proteins accumulated (Fig. 1B, top and bottom gels, lane 1). As expected, in the absence of E2a, E4Orf6, and VA RNA, neither viral replication (Fig. 1A, lane 8) nor AAV5 protein production (Fig. 1B, lane 8) was apparent. As previously observed (16), when only VA RNA was withheld from such experiments, the AAV5 double-stranded mRFs were generated at essentially wild-type levels; however, the relative levels of progeny ssDNA were dramatically reduced (Fig. 1A, lane 2). As previously argued (16), the lack of ssDNA accumulation was likely due to the paucity of AAV5 protein produced from this cotransfection, in spite of the presence of the substantial amount of the double-stranded mRF template that was generated (Fig. 1B, lane 2). When wild-type VA was replaced in these assays with VA RNA mutants containing single-nucleotide changes in the PKR binding domain (VA A119G and VA C120A) which had been previously characterized as reducing, but not totally eliminating, the inhibition of PKR activation and subsequent phosphorylation (21), an intermediate reduction in the accumulation of ssDNA forms was observed, while mRF levels remained high (Fig. 1A, lanes 3 and 4). This was accompanied by an intermediate level of accumulation of AAV5 Rep and Cap proteins (Fig. 1B, lanes 3 and 4). When VA RNA mutants were added that were shown to be unable to inhibit the activation and subsequent phosphorylation of PKR, either due to the lack of the PKR binding central domain (VAΔIV) or because the mutant was unable to be transported to the cytoplasm (VA NES KO), substantial levels of mRF were produced (Fig. 1A, lane 5); yet, as in the absence of VA, significantly less AAV5 ssDNA was detected (Fig. 1A, lane 7) and little Rep and Cap accumulated, in spite of the high mRF levels (Fig. 1B, lanes 5 and 7). A mutant of VA RNA with mutations in the central domain (VA TT101/102CA) shown to inhibit the activation and subsequent phosphorylation of PKR like the wild type (19) was able to support full AAV5 replication (Fig. 1A, lane 6) and sustained near-wild-type levels of AAV5 protein (Fig. 1B, lanes 6).

FIG. 1.

(A) Replication of AAV5 viral DNA in 293 cells in the presence of Ad5 E2a, Ad5 E4Orf6, and different combinations of Ad5 VA RNA mutants. Southern blot analysis of AAV5 replication in 293 cells together with wild-type helper E2a and E4Orf6 and mutant VA RNA, as indicated above the gel (lanes 1 to 7), or empty vector (lane 8). Monomer-length (mRF), dimer-length (dRF), and ssDNA replicative forms are indicated. (B) Analysis of AAV5 Rep protein and capsid protein accumulation in 293 cells in the presence of Ad5 E2a, Ad5 E4Orf6, and various combinations of VA RNA mutants. The results of immunoblot analysis to determine AAV5 protein expression during the same experiment in which viral replication was determined as described above, using antibody against either the viral Rep proteins (top gel) or capsid proteins (bottom gel), are shown. Helper plasmids cotransfected together with AAV5 infection are indicated above the gels. The locations of the AAV5 proteins, as well as the endogenously expressed 14-3-3 and actin proteins which serve as loading controls, in response to the various transfection combinations are shown. +, present; −, absent.

These results suggested that, in order to fully support AAV5 replication in concert with the other Ad helper functions, VA RNA needed to retain its ability to inhibit the activation and subsequent phosphorylation of PKR. In the absence of this activity, mRF, but not ssDNA, was produced. The ability of these mutants to support the accumulation of AAV5 ssDNA and protein was inversely proportional to the predicted deficiency of these mutants in inhibiting the phosphorylation of PKR (Fig. 1A), and the deficiency in the ability of these mutants to support AAV5 protein accumulation was apparent even though abundant amounts of the presumed transcription template (mRF) were generated. Because this phenotype was similar to the results obtained when VA RNA was left out altogether, these results suggested that the role of VA RNA in supporting AAV5 protein accumulation was dependent on its ability to inhibit the phosphorylation of PKR. These results further suggest that the ability of VA RNA to inhibit the activation and subsequent phosphorylation of PKR may have been required for the proper accumulation of the AAV5 proteins required to generate progeny DNA.

Expression of the AAV5 P41 transcription unit induced phosphorylation of eIF2α in both 293 and HeLa cells.

If the interaction between VA RNA and PKR was necessary for the effect of VA RNA in enhancing AAV5 protein accumulation and, as a result, for its help in genome replication, we considered whether perhaps AAV5 could activate PKR and its subsequent downstream effects, although such an effect has not been previously reported for members of the Parvovirinae. A well-characterized consequence of PKR activation is the phosphorylation of eIF2α, which leads to the inhibition of translation (22). Although the experiment whose results are shown in Fig. 1 examined helper effects during viral infection, we have previously shown that protein expression from the AAV5 P41 transcription unit alone (which, unlike the AAV2 P40 transcription unit, is active in 293 cells in the absence of Rep or Ad) was enhanced by VA RNA (16). As can be seen in the results shown in Fig. 2, the AAV5 P41 transcription unit strongly induced the phosphorylation of eIF2α in 293 cells at the earliest time point tested posttransfection, well before the detectable accumulation of the AAV5 capsid proteins. (In the presence of VA RNA, the accumulation of viral capsid proteins was apparent much earlier [see Fig. 3].) In addition, the P41 transcription unit also induced the phosphorylation of eIF2α in HeLa cells, albeit with somewhat delayed kinetics. These results suggested that transfection with the AAV5 P41 transcription unit induced the activation of PKR. The phosphorylation of eIF2α did not require virus infection; however, P41 plasmids in which the promoter was inactivated could not induce phosphorylation of eIF2α, indicating that RNA expression was required for this induction (data not shown). In addition, the expression of both the closely related goat-derived AAV (Go-AAV) and the prototype AAV2 capsid gene transcription units induced the phosphorylation of eIF2α (Fig. 2C).

FIG. 2.

(A) Induction of phosphorylated eIF2α in 293 cells following transfection with an AAV5 capsid gene-expressing construct. Immunoblot analysis of protein from 293 cells was performed at successive times following transfection with P41Cap, as indicated. After transfer, filters were probed with either anticapsid antibody and antiactin antibody (top) or antitubulin antibody and antibody against phosphorylated eIF2α. (B) Induction of phosphorylated eIF2α in HeLa cells following transfection with an AAV5 capsid gene-expressing construct. Immunoblot analysis of protein from HeLa cells was performed at successive times following transfection with P41Cap, as indicated. After transfer, filters were probed with either anticapsid antibody and antitubulin antibody (top) or antiactin antibody and antibody against phosphorylated eIF2α. (C) Induction of phosphorylated eIF2α in 293 cells following transfection with an AAV5, Go-AAV, or AAV2 capsid gene-expressing construct. The results of immunoblot analysis of protein from 293 cells 48 h posttransfection with the AAV5 P41Cap construct, a Go-AAV P41Cap construct, a CMV-driven AAV2 cap construct, or the empty vector parent SK construct, with or without a VA RNA-expressing plasmid, as indicated, are shown. After transfer, filters were probed with either anticapsid antibody and antibody against phosphorylated eIF2α (top) or antiactin antibody (bottom).

FIG. 3.

Accumulation of AAV5 capsid proteins and RNA and phosphorylated eIF2α generated in 293 cells 24 h (A) or 48 h (B) following transfection with P41Cap in the presence of either wild-type VA RNA or VA RNA mutants. The results of immunoblot analysis of protein accumulation in 293 cells following transfection with the P41Cap construct together with various VA RNA mutants are shown. The complete set of samples was run on two separate gels, and after transfer, they were probed with antibodies against either AAV5 capsid proteins and actin (top) or tubulin and phospho-eIF2α (middle). The lower panels show the results of quantitative RPAs, using the AAV5 DH probe as described in Materials and Methods, of RNA generated from the same experiment. Bands protected by P41 spliced RNAs utilizing small intron donors A1 or A2 are indicated on the left.

AAV5 P41-induced phosphorylation of eIF2α, as well as accumulation of AAV5 protein, could be reversed by wild-type VA RNA but not VA RNA mutants exhibiting reduced inhibition of the activation of PKR.

As shown in Fig. 2, following transfection with the P41 transcription unit alone, capsid protein expression was not detectable until after 24 h. However, in the presence of VA RNA, the accumulation of capsid proteins from P41Cap was substantial by 24 h, dramatically greater than following transfection with P41 alone (Fig. 3A, compare lanes 1 and 7). Concomitantly, VA RNA expression reduced the accumulation of phosphorylated eIF2α compared to the effect of P41Cap alone (Fig. 3A, compare lanes 1 and 7). VA RNA also reduced the levels of phosphorylated eIF2α induced by both Go-AAV and AAV2 (Fig. 2C). VA RNA mutants which were either shown to be deficient in inhibiting PKR activation and subsequent phosphorylation (VA A119G, VA C120A, and VAΔIV) or could not be transported to the cytoplasm (VA NES KO) were, at this time point, unable to support accumulated levels of AAV5 capsid proteins (Fig. 3A, top panel, lanes 3, 4, 5, and 2, respectively) and were unable to prevent the accumulation of phosphorylated eIF2α (Fig. 3A, middle panel). Cotransfection with the VA RNA mutant with a mutation in the central domain (VA TT101/102CA) which has been shown to retain the ability to inhibit PKR activation and subsequent phosphorylation still inhibited P41Cap induction of eIF2α (Fig. 3A, middle panel, lane 6) and supported increased levels of capsid protein accumulation (Fig. 3A, top panel, lane 6). The levels of RNA which accumulated in the cytoplasm during each cotransfection were similar, demonstrating that the effects were at the posttranscriptional level (Fig. 3A, bottom panel).

By 48 h posttransfection, capsid protein accumulation from P41Cap alone had become detectable even in the presence of phosphorylated eIF2α; however, the levels of accumulated protein still lagged significantly behind the levels which accumulated with P41Cap in the presence of wild-type VA RNA (Fig. 3B, compare lanes 1 and 7). Cotransfection with VA RNA mutants VA C120A and VAΔIV, which have been shown to be deficient in inhibiting PKR activation and subsequent phosphorylation, as well as VA NES KO, which prevented the accumulation of VA RNA in the cytoplasm, were unable to prevent the phosphorylation of eIF2α even at this later time, and these cotransfections showed no increase in capsid protein accumulation over the level which accumulated with P41Cap alone (Fig. 3B, lanes 4, 5, and 2, respectively). The mutant VA TT101/102CA, shown to remain proficient in inhibiting PKR activation and subsequent phosphorylation, was also able to prevent the phosphorylation of eIF2α and supported an increased level of capsid protein accumulation (Fig. 3B, lane 6). Mutant VA A119G, whose help was not apparent at 24 h posttransfection, was effective by 48 h in decreasing the phosphorylation of eIF2α and supporting increased capsid protein accumulation, suggesting that this mutation was somewhat leaky (Fig. 3B, lane 3).

The analysis of VA RNA mutants with various abilities to inhibit PKR activation and subsequent phosphorylation reinforces the correlation between phosphorylation of PKR, phosphorylation of eIF2α, and AAV5 capsid protein production. The augmentation of capsid proteins by VA RNA occurred most prominently when the phosphorylation of eIF2α was inhibited by VA RNA, strongly suggesting that VA RNA brought about the translation of AAV5 capsid proteins by inhibiting the activation of PKR. The total levels of RNA produced in the various transfections remained similar (Fig. 3A and B, bottom panels).

A 200-nt region of AAV5 P41-generated RNA induced the phosphorylation of eIF2α and could confer increased accumulation of capsid proteins in the presence of VA RNA.

Replacing the AAV5 P41 promoter with the CMV promoter had little effect on the ability of the AAV5 capsid transcription unit to induce phosphorylated eIF2α or on the ability of this transcription unit to be responsive to VA RNA suppression of such inhibition (Fig. 4A and B, lanes 1 and 2). This parent construct was then used to identify the region(s) within the capsid gene responsible for these effects. Successive deletion from the 5′ end of this construct, up to nt 2131, generated constructs which were indistinguishable from the wild type; i.e., these constructs induced the phosphorylation of eIF2α, and the addition of wild-type VA RNA suppressed this effect, resulting in the increased accumulation of AAV5 capsid proteins (Fig. 4A and B, CMV 18 V5 cap, CMV 19 V5 cap, CMV 20 V5 cap, and CMV 21 V5 cap in lanes 3 to 8, Fig. 4D). The enhancement of capsid protein accumulation was somewhat less in this particular experiment, for reasons that are currently unclear. When the deletions extended downstream past nt 2312, however, the constructs were no longer able to induce the phosphorylation of eIF2α (Fig. 4A and B, CMV 23 V5 cap, CMV 26 V5 cap, and CMV 27 V5 cap in lanes 9 to 14, Fig. 4D). In the absence of PKR activation, the capsid protein accumulation from these constructs was high in the absence of VA RNA and not further enhanced by its addition. Similar steady-state levels of total RNA were generated by all the mutants tested (Fig. 4C).

FIG. 4.

A 200-nt region of AAV5 P41-generated RNA induced the phosphorylation of eIF2α and could confer increased accumulation of capsid proteins in response to VA RNA. Panels A and B show the results of immunoblot analysis of protein accumulation in 293 cells following transfection with CMV-driven AAV5 capsid gene constructs with 5′ deletions in their leader sequence, as described in the text, in the absence (odd-numbered lanes) and presence (even-numbered lanes) of VA RNA as indicated. The complete set of samples were run on two separate gels and after transfer were probed with antibodies to AAV5 capsid proteins and actin (A) and antibodies to tubulin and phosphorylated eIF2α (B). Panel C shows the results of quantitative RPAs, using the AAV5 Cap RPA probe as described in Materials and Methods, of total RNA generated in the same experiment. Panel D is a diagram showing the deletions in the capsid protein gene leader region of the various constructs used, along with genetic landmarks and the probe used for RPA depicting the protected RNA region. Panels E and F show the results of an immunoblot analysis of protein accumulation in 293 cells following transfection with CMV-driven GFP constructs with different regions of the AAV5 leader sequence inserted between the CMV promoter and the GFP reporter gene, as described in Materials and Methods. The complete set of samples was run on two separate gels which, after transfer, were probed with antibodies raised against either GFP and tubulin (E) or tubulin and phosphorylated eIF2α (F).

It was important to rule out the possibility that the inability of VA RNA to enhance the synthesis of capsid proteins from constructs CMV 23 V5 cap, CMV 26 V5 cap, and CMV 27 V5 cap was due merely to the use of the strong CMV promoter upstream of the VP3 ATG start codon. It was conceivable that translation by VA RNA from these constructs became rate limiting due to increased amounts of capsid gene RNA generated by this strong heterologous promoter. Therefore, we created a control construct in which a CMV promoter, whose enhancer region had been removed such that its strength was compromised, was placed at nt 2312, similar to construct CMV 23 V5 cap. Although this construct generated significantly less RNA product than the wild-type parent, VA RNA still had little effect on the accumulated protein levels, suggesting that the strength of the promoter was not a determinant in the ability of VA RNA to augment translation (data not shown).

These results identified an RNA element between nt 2131 and 2312 that could both induce the activation of PKR and be responsive to VA RNA suppression of this effect. To determine if this element was both necessary and sufficient for these effects, the AAV5 capsid gene leader sequence was cloned into a heterologous construct in which the GFP gene was driven by the CMV promoter. As can be seen in the results shown in Fig. 4E and F, the CMV-driven GFP parent construct neither induced the phosphorylation of eIF2α nor was responsive to VA RNA enhancement of protein accumulation (Fig. 4E and F, CMV GFP in lanes 13 and 14). However, when the 5′ region of the AAV5 VP1 capsid gene, from nt 1891 to 2781, containing the putative PKR-activating region, was inserted downstream of the CMV promoter, its expression led to the phosphorylation of eIF2α, and there was a consequent reduction of GFP protein accumulation which could be suppressed by the expression of VA RNA (Fig. 4E and F, CMV 18-27 V5 GFP in lanes 1 and 2). Furthermore, successive 5′ deletions of this element in this heterologous context yielded results similar to those seen above, i.e., the activation of PKR could be localized to the same region of RNA identified above (nt 2131 to 2312; Fig. 4E and F; compare CMV 21-27 GFP, containing AAV5 sequences 2131 to 2781, in lanes 7 to 8 to CMV 23-27 GFP, containing AAV5 sequences 2312 to 2781, in lanes 9 to 10), and under conditions where PKR was not induced, the expression of GFP was again independent of VA RNA (Fig. 4E and F, lanes 9 to 12).

These results suggested that AAV5 RNA nt 2131 to 2312, in the capsid gene leader sequence surrounding the VP1 ATG, contains a signal which activated PKR, with the subsequent phosphorylation of eIF2α, and played a crucial role in the responsiveness of AAV5 to VA RNA. Interestingly, this signal does not reside in those (majority) AAV5 P7- and P19-generated RNAs which are polyadenylated at the (pA)p site in the center of the genome.

The induction of the phosphorylation of eIF2α by P41-generated RNA could be reversed by either siRNA knockdown of PKR or the expression of the HSV protein ICP34.5.

That VA RNA mutants deficient in inhibiting PKR activation did not suppress the phosphorylation of eIF2α induced by P41-generated RNA suggests that the induction of PKR was a critical intermediate step in this effect. If this was the case, the knockdown of PKR levels by siRNA should mimic the activity of VA RNA. Figure 5A shows that this was the case. Transfection with CMV 18 V5 cap, which contains the AAV5 PKR induction element, resulted in the phosphorylation of eIF2α and produced low levels of accumulated AAV5 capsid protein (Fig. 5A, lane 1). When siRNA to PKR was added by cotransfection, the phosphorylation of eIF2α was inhibited, and capsid protein accumulation was regained (Fig. 5A, lane 2). As seen previously, capsid expression from CMV 23 V5 cap, which lacked the PKR activation element and did not induce the phosphorylation of eIF2α, was high (Fig. 5A, lane 3) and was not further enhanced by siRNA to PKR (Fig. 5A, lane 4). Similar results were also seen following AAV5 viral infection. The introduction of siRNA directed against PKR (Fig. 5C, lanes 3 and 4) acted similarly to the addition of VA RNA (Fig. 5C, lane 1) in suppressing the inhibitory effect of E4Orf6 on the accumulation of AAV5 capsid and Rep proteins generated during infection (Fig. 5C, lane 2).

FIG. 5.

The induction of the phosphorylation of eIF2α by P41-generated RNA could be reversed by either siRNA knockdown of PKR or expression of the HSV protein ICP34.5. Results are shown of an immunoblot analysis of protein accumulation in 293 cells following transfection with CMV-driven AAV5 capsid constructs with deletions in their leader sequence (as described in the text) in the absence (odd-numbered lanes) or presence (even-numbered lanes) of an siRNA construct directed against PKR (A) or a plasmid expressing the HSV protein ICP34.5 (B). The complete set of samples was run on two separate gels which, after transfer, were probed with antibodies raised against the AAV5 capsid proteins and actin (A and B, top panels), against tubulin and phosphorylated eIF2α (A and B, middle panels), and against total PKR (A and B, bottom panels). Panel C shows the results of an immunoblot analysis of protein accumulation in 293 cells following infection with AAV5 and cotransfection with plasmids expressing Ad E2a and Ad E4Orf6 or siRNA directed against PKR, as indicated. The complete set of samples was run on three separate gels which, after transfer, were probed with antibodies raised against AAV Rep proteins and cellular 14-3-3 (top panel), against the AAV5 capsid proteins and actin (middle panel), and against total PKR (bottom panel). +, present; −, absent. Panel D shows a diagram showing the deletions in the Cap leader region present in the constructs used in the experiments whose results are shown in panels A and B, along with genetic landmarks of the capsid gene.

To determine whether the phosphorylation of eIF2α was directly responsible for the reduction in AAV5 capsid protein accumulation in our system, we utilized the HSV protein ICP34.5. ICP34.5 inhibits PPI kinase, an activity which allows PPI to effectively dephosphorylate eIF2α, thereby suppressing its inactivation. The reduction of the phosphorylation of eIF2α by ICP34.5 acted similarly to both VA RNA and the siRNA knockdown of PKR in overcoming the effects of the AAV5 capsid gene leader sequence and allowing an increase in the levels of accumulated capsid protein produced by CMV 18 V5 cap (Fig. 4B, compare lanes 1 and 2) while having little effect on the level of protein generated by CMV 23 V5 cap (Fig. 4B, compare lanes 3 and 4). Taken together, these observations indicate that the inhibition of AAV5 capsid protein accumulation resulted from the induction of PKR by AAV5 P41-generated RNA, leading to the phosphorylation of eIF2α. The suppression of this effect is thus one of the roles that VA RNA plays in helping AAV5 replication.

DISCUSSION

In this study, we have demonstrated that a small element in AAV5 P41-generated RNA can activate PKR, leading to the phosphorylation of the essential translation initiation factor eIF2α and a reduction in the accumulation of de novo-generated AAV5 protein. Furthermore, we show that VA RNA's known property of binding to and competitively inhibiting PKR is at least one of the functions necessary for its support of AAV5 replication. VA RNA suppresses the AAV5 induction of this cellular inhibitory response, allowing the accumulation of viral proteins necessary for replication. Although a posttranscriptional role for VA RNA in AAV replication has long been suggested (8, 29), an interaction of parvovirus with the cellular PKR/eIF2α system has not been previously reported. One reason for our ability to observe these effects was that, as we have previously shown (16) and in contrast to the situation with AAV2, the promoters of the AAV5 small Rep and capsid genes are active in 293 cells in the absence of Rep. We have also shown that the expression of the AAV5 capsid transcription unit activates PKR and the phosphorylation of eIF2α in HeLa as well as 293 cells and that both the AAV2 and Go-AAV capsid genes can activate PKR and the phosphorylation of eIF2α in 293 cells. These latter observations suggest that the activation of the PKR/eIF2α cellular response may be a general feature of at least the Dependovirus genus of the Parvovirinae, although the role this plays during AAV infection is not yet clear.

That AAV5 RNA could activate both PKR and its subsequent phosphorylation of eIF2α and that this induction had an inhibitory effect on the accumulation of AAV5 protein was verified by the reversal of these effects by siRNA directed against PKR and by the ICP34.5 protein, which allows the dephosphorylation of eIF2α. In addition, we were able to show that the activation of PKR was due to the expression of a small RNA element in the capsid gene leader sequence surrounding the VP1 AUG and that this activity was transferable upon the introduction of this element to a heterologous gene. In preliminary experiments, mutation of this AUG did not diminish the ability of this element to induce the phosphorylation of eIF2α (R. Nayak and D. J. Pintel, unpublished data).

It is not yet known how widespread or complete is the inhibition of the translation of cellular RNAs in response to AAV5-induced PKR activation and eIF2α phosphorylation. How PKR induction factors into the Ad-dependent AAV viral life cycle is also not yet clear. The activation of PKR and inhibition of AAV5 protein synthesis in the absence of VA RNA may be a mechanism that helps maintain very low levels of expression so as to sustain AAV persistence in the absence of helper virus.

We have recently shown that the Ad5 E4Orf6 protein can target AAV5 protein for degradation (16). When VA RNA was left out of the set of helper factors, AAV5 replicated through its double-stranded intermediates but did not generate progeny virus, due at least in part to the paucity of both capsid proteins and small Rep. VA RNA suppressed this effect but apparently did so by increasing the translation of AAV5 protein in parallel rather than directly interfering with the degradative effects of E4Orf6. It is not yet clear whether the supportive effect of VA RNA in the replication of AAV5 is due exclusively to its role in overcoming the effects of E4Orf6 or whether other, more-generalized effects on translation are also important. As can be seen in Fig. 1, significant amounts of AAV5 monomer and dimer replicative DNA forms accumulate in the presence of only small amounts of large Rep protein. This has been observed before (16) and suggests that only low levels of the large Rep protein may be required to support these types of replication in the absence of capsid proteins.

In both the presence and the absence of VA RNA, the expression of the AAV5 capsid proteins from a P41 expression vector increased over time, suggesting that the inhibition of translation due to the phosphorylation of eIF2α is moderating rather than complete. At late times posttransfection, the P41Cap plasmid directed the expression of the capsid proteins even in the presence of phosphorylated eIF2α, yet a much more significant effect was seen at earlier times (Fig. 3), suggesting that the role of VA RNA in suppressing PKR-activated translation inhibition may be greatest during the early stages of infection. Interestingly, while the expression of protein from AAV5 P41-generated RNA was detected substantially earlier in the presence of VA RNA than in its absence, the total levels of intact viral RNA present seem to be similar. Where untranslated RNA resides in the absence of VA RNA and why it is not being translated are important undetermined features of this system.

VA RNA could overcome the PKR-induced inhibition of the accumulation of AAV5 capsid proteins; however, we have previously shown that CMV-driven Ad5 E2a and E4Orf6, when cotransfected with the PKR-inducing AAV5 P41 transcription unit, were expressed in the absence of VA RNA. Thus, the inhibitory effect of PKR induction and the VA RNA suppression of this effect are at least somewhat discriminatory. The degree of specificity of the inhibitory effect of the AAV5 induction of PKR and the enhancing effect of VA RNA, both following PKR induction and in its absence, is likely to be complex and is currently being investigated. However, as mentioned above, for AAV5 the great majority of the Rep-encoding P7- and P19-generated mRNAs are polyadenylated at the internal (pA)p site. They share only very limited overlap with the P41-generated capsid gene mRNAs and do not harbor the element within the capsid gene RNA that induces PKR. Consistent with this model, a construct expressing only the P19-generated cDNA by itself does not induce PKR (data not shown).

It is important to note that HSV can also act as a helper virus for AAV2 (3) and AAV5 (Nayak and Pintel, unpublished) replication. Thus far, however, an inhibitor of the PKR/eIF2α system has not been identified in the minimal set of HSV gene products that support AAV replication. Whether one of the identified HSV helper gene products has an as-yet-unidentified effect on this pathway or whether HSV sustains AAV replication in a qualitatively different manner is yet to be determined.