|Home | About | Journals | Submit | Contact Us | Français|
We describe a structural rearrangement that can occur in parvovirus minute virus of mice (MVMp) virions following prolonged exposure to buffers containing 0.5 mM EDTA. Such particles remain stable at 4°C but undergo a conformational shift upon heating to 37°C at pH 7.2 that leads to the ejection of much of the viral genome in a 3′-to-5′ direction, leaving the DNA tightly associated with the otherwise intact capsid. This rearrangement can be prevented by the addition of 1 mM CaCl2 or MgCl2 prior to incubation at 37°C, suggesting that readily accessible divalent cation binding sites in the particle are critical for genome retention. Uncoating was not seen following the incubation of virions at pH 5.5 and 37°C or at pH 7.2 and 37°C in particles with subgenomic DNA, suggesting that pressure exerted by the full-length genome may influence this process. Uncoated genomes support complementary-strand synthesis by T7 DNA polymerase, but synthesis aborts upstream of the right-hand end, which remains capsid associated. We conclude that viral genomes are positioned so that their 3′ termini and coding sequences can be released from intact particles at physiological temperatures by a limited conformational rearrangement. In the presence of divalent cations, incremental heating between 45°C and 65°C induces structural transitions that first lead to the extrusion of VP1 N termini, followed by genome exposure. However, in cation-depleted virions, the sequence of these shifts is blurred. Moreover, cation-depleted particles that have been induced to eject their genomes at 37°C continue to sequester their VP1 N termini within the intact capsid, suggesting that these two extrusion events represent separable processes.
The capsids of nonenveloped viruses have evolved to protect their genomes from hostile extra- and intracellular environments and to undergo a specific program of conformational shifts in response to cellular cues, which sequentially expose trafficking effector structures required to transport the nucleic acid to the appropriate cellular location, at the correct time, for its successful replication (22, 25, 31, 42, 43, 46). In many cases, such shifts are possible because the virion is poised in a metastable configuration, prevented from collapsing to an alternate structure by the energy barrier between the two forms. Cellular or environmental triggers effectively compromise this energy barrier during cell entry, inducing the particle to rearrange. Accordingly, the pattern of particle morphogenesis is in part encrypted within the virion so that probing its structure in vitro, using elevated temperatures or biased ionic conditions, provides an avenue for assessing the types of rearrangements that the virion is capable of sustaining (5, 12, 22, 24, 29, 34, 35, 46, 47, 50). In this study we exploit in vitro manipulation to gain insight into the structure and potential rearrangements programmed into the minute virus of mice (MVM) virion.
As for all members of the family Parvoviridae, virions of MVM are exceptionally small and structurally simple, comprising a single copy of a ~5-kb negative-sense, self-priming, single-stranded, linear DNA enclosed in a nonenveloped protein capsid with an external diameter of 280 Å and T=1 icosahedral symmetry (1, 30). Capsids are assembled from 60 polypeptides encoded by a single structural gene, which are expressed in an approximately 1:5 ratio as two size variants, VP1 (83 kDa) and VP2 (63 kDa) (44). They share 587 amino acids of carboxy-terminal sequence, while VP1 has an N-terminal extension of 142 amino acids that contains a phospholipase A2 (PLA2) domain essential for endosomal escape during entry (19, 53). As can be seen in the X-ray structure, capsid shells are constructed from 549 amino acids common to the C termini of VP1 and VP2, leaving signal-rich N-terminal extensions of 38 residues for VP2 and 180 residues for VP1, whose disposition within the assembled virion remains uncertain (1). The core structure comprises the same eight-stranded, antiparallel, β-barrel motif found in numerous viral capsid structures (2, 36, 39) with individual β-strands separated by large insertion loops that form the surface of the virion. Each of the 60 asymmetric units is constructed by the interdigitation of several neighboring VP polypeptides, intimately wrapped around each other, particularly at the icosahedral 3-fold axes, creating an apparently rugged assembly. Projecting “cylinders” at each icosahedral 5-fold axis appear to be critically involved in particle dynamics. These cylinders are created by the parallel juxtaposition of β-hairpin ribbons from each of the 5-fold related capsid proteins and enclose a small central pore (~8 Å in diameter at its tightest constriction) that penetrates to the particle interior. The cylinders are themselves encircled by highly conserved canyon-like surface depressions of uncertain function. Despite the apparently minimal diameter of the 5-fold pores, genetic and structural approaches implicate the cylinders as the major portals of entry into and egress from the virion (1, 3, 4, 17, 18).
The best-documented example of such transit concerns the N-terminal peptides of VP2, which are sequestered within empty particles but start to become surface exposed early during DNA packaging (12, 14). Twenty amino acids from the N termini of most VP2 molecules can be removed from intact virions by proteolysis, leaving a short, N-terminal, glycine-rich sequence that has been modeled into a partially ordered electron density observed stretching through the 5-fold pores of full, but not empty, particles in X-ray diffraction studies (1, 48, 51). For some crystals this observed density indicates that as many as two-thirds of the 5-fold pores may be thus occupied (9). This represents, at most, 8 of the ~50 VP2 N termini present in the particle, but ultimately, most of these peptides can be proteolytically cleaved from the virion, suggesting that the 5-fold cylinders may open repeatedly during maturation and cell entry to allow the sequential surface exposure of these N termini. Genetic experiments suggest that a 5-fold pore also serves as the entry portal for viral DNA (3, 4, 17, 18). Newly displaced progeny DNA strands are translocated, in a 3′-to-5′ direction (14), into a preformed “empty” capsid in a reaction that is likely driven by the helicase function of the viral replication initiator protein, NS1 (27). Although precisely where the 3′ end of the genome is positioned in the particle remains uncertain, packaging leaves ~22 nucleotides from the 5′ end of the DNA exposed at the particle exterior, still covalently attached to an externally located NS1 molecule (13).
Where or how the viral genome is uncoated in vivo remains uncertain, but the protein chains of the virion are intricately interwoven and largely resistant to proteolytic assault, except for the short VP2 N-terminal peptides. As part of an unrelated study, we recently removed these VP2 peptides from large batches of MVM virions by trypsin digestion at 37°C and then rebanded the digested product on density gradients. Remarkably, virions in a subset of virus stocks appeared to have undergone a reaction that caused them to band at a lower density, indicating that a major structural rearrangement had likely occurred. In this paper we characterize the nature of this rearrangement, explore the conditions that allow it to occur, and discuss the implications of these findings for virion structure and the genome-uncoating process.
A9 ouabr11 cells were grown in spinner culture in Dulbecco's modified Eagle's medium (DMEM) containing 5% fetal bovine serum and antibiotics to a density of ~6 × 105 cells/ml. Following infection with predetermined titers of transfection-derived parvovirus MVM (MVMp) stocks (GenBank accession number J02275), these cultures were expanded until cell counts indicated a progressive rise in numbers of dead cells. The remaining cells were harvested by centrifugation, washed in Dulbecco's phosphate-buffered saline without Ca2+ or Mg2+ (PBS; Invitrogen, Carlsbad, CA), pelleted, and resuspended in 10 ml of 50 mM Tris-HCl (pH 8.7) containing 0.5 mM EDTA (TE8.7) per liter of infected cells. Pellets were extracted by three cycles of freezing and thawing at 37°C, followed by repeated centrifugation at 2,000 rpm (800 × g) to clarify extracts, and were stored at −20°C.
For purification, 6-ml aliquots were further clarified by centrifugation at 11,000 rpm in a Sorvall SS34 rotor at 4°C and then applied to step gradients containing 1 ml of 55% and 2 ml of 45% iodixanol (OptiPrep; Axis-Shield, Oslo, Norway) buffered with PBS (pH 7.2) or PBS supplemented with 2.5 mM KCl and 1 mM MgCl2 (KM), as indicated, and 2 ml of 35% and 1 ml of 15% iodixanol buffered in TE8.7 and centrifuged at 35,000 rpm for 18 h at 18°C in a Beckman SW41 rotor. Fractions were collected from the bottom of the gradient, and viral particles were identified by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and staining with brilliant blue R (referred to throughout as Coomassie blue; Sigma).
For large-scale analysis, ~1013 virions purified and stored in iodixanol containing PBS or PBS-KM were diluted with 3 volumes of PBS or PBS-KM, as indicated, and digested with tosylsulfonyl phenylalanyl chloromethyl ketone (TPCK)-trypsin at a final concentration of 2 ng/μl in the presence of 5 μM CaCl2 for 1 h at 37°C. Defined trypsin inhibitor (Cascade Biologics, Portland, OR) was added to 10 times the predicted effective dose, and samples were diluted to 5.9 ml with PBS before sedimentation through iodixanol step gradients similar to those used in their initial purification except that all fractions were buffered with PBS or PBS-KM, as indicated. Thirteen ~350-μl fractions were collected from the bottom of the gradient, and the remainder was pooled.
For subsequent miniaturized assays, ~1012 virions purified in iodixanol buffered with 50 mM HEPES and 120 mM NaCl (pH 7.5) were diluted with 3 volumes of the buffer indicated in the text, added from 21 days to immediately before trypsin digestion and stored at 4°C in the interim. Some samples also received low concentrations of specific cations at various times prior to incubation at 37°C, as specified. Samples were digested, or not, with trypsin at 2 ng/μl for 1 h at 37°C, and trypsin inhibitor was added before they were diluted to 1 ml and rebanded through a 4-ml step gradient containing 0.75 ml of 55%, 1.5 ml of 45%, 1 ml of 35%, and 0.75 ml of 15% iodixanol buffered with 50 mM HEPES and 120 mM NaCl (pH 7.5). Twelve ~350-μl fractions were collected from the bottom of the gradient, while the 13th fraction contained all of the residual aqueous sample.
Samples were separated by SDS-PAGE using 7.5% acrylamide-0.2% bis-acrylamide gels, and the proteins were stained with Coomassie blue or electroblotted onto polyvinylidene difluoride membranes. Membranes were blocked, incubated with a rabbit antibody against the “allo-p” peptide spanning amino acids 311 to 327 in the MVMp VP2 sequence, and developed using the ECL system according to the manufacturer's instructions (Amersham, Uppsala, Sweden). Low-intensity exposures were photographed, using a Kodak digital camera driven by MDS 290 software, and boxed densities were determined by using Kodak 1D software.
As indicated, prior to electrophoresis some samples were digested with micrococcal nuclease to quantitate packaged genomes, as described previously (17), following alkaline gel electrophoresis, Southern blotting, and hybridization to a random-primed MVM DNA EcoRV-XbaI fragment spanning nucleotides 381 to 4342 (GenBank accession number J02275) or with 32P-5′-labeled positive-sense oligonucleotide probes (15) representing MVM nucleotides 1165 to 1190 (1.2-kb probe), 2199 to 2228 (2.2-kb probe), or 2872 to 2897 (2.9 kb) or the 65-bp repeat at nucleotides 4738 to 4767 and 4803 to 4832 (4.8-kb probe). DNA was detected by autoradiography or quantified, using a Molecular Dynamics PhosphorImager SI.
Samples for immunoprecipitation were diluted into RIPA buffer (50 mM Tris-HCl [pH 7.5], 150 mM NaCl, 1% NP-40, 1% sodium deoxycholate, 0.1% sodium dodecyl sulfate, and 2 mM EDTA) and incubated at 4°C for 2 h with previously described polyclonal rabbit antisera (12) that recognize either the VP1 N-terminal peptide (anti-VP1SR), assembled capsids only (PN1), or variable ratios of assembled and dissociated capsids (Tatt-1 and Tatt-3). Note that samples were not “precleared” by incubation with nonspecific antiserum and immunoadsorbent before each specific immunoprecipitation step, because we wished to sample the entire extract, but this approach does result in slightly higher background levels in some lanes. Complexes were collected on formalin-fixed Staphylococcus aureus cells (Pansorbin; Calbiochem) and washed three times with a solution containing 5% sucrose, 1% NP-40, 500 mM NaCl, 50 mM Tris-HCl, and 5 mM EDTA (pH 7.5) and once with 10 mM Tris-HCl (pH 7.5). Some samples were then digested with restriction enzymes for 40 min at 37°C, and some of these digests were fractionated into antigen-Pansorbin-associated and nonassociated fragments by centrifugation prior to proteinase K digestion and gel electrophoresis through neutral agarose gels. DNA was detected by Southern transfer and hybridization with a probe derived from MVM nucleotides 381 to 4342. For transition analysis, iodixanol samples were first diluted 1:2 into PBS and incubated at temperatures specified in the text for 10 min before the addition of 10 volumes of RIPA buffer and immunoprecipitation at 4°C as described above. After washing, some samples were digested with micrococcal nuclease, as described above, followed by proteolysis and Southern analysis.
Purified single-stranded MVM DNA or fraction 11 material from the −KM gradient, each containing ~50 ng of DNA, was incubated with the modified T7 DNA polymerase Sequenase, version 2.0 (5 units; USB, Cleveland, OH), in Sequenase buffer supplemented with 5 mM dithiothreitol (DTT); 200 μM dCTP, dGTP, and dTTP; 50 μM unlabeled dATP; and 10 μCi of [32P]dATP for 1 h at 37°C. Reactions were stopped by the addition of EDTA to 10 mM and SDS to 0.5%, and aliquots were used for immunoprecipitation, as described above, or purified by phenol extraction and spin column separation following incubation with proteinase K at 60°C for 3 h, dropping to room temperature overnight.
High-titer viral stocks were extracted from cell pellets by repeated freezing and thawing in 50 mM Tris-HCl (pH 8.7) containing 0.5 mM EDTA (TE8.7) and clarified by repeated centrifugation at 11,000 rpm. Adapting a method from Zolotukhin and colleagues (54), aliquots of this extract were then fractionated through step density gradients of the inert, nonionic medium iodixanol. Low-density fractions in these gradients were buffered with TE8.7, which enhances viral yield by minimizing virion interactions with residual cell debris, but in the high-density fractions where virions accumulate, phosphate-buffered saline (PBS) was used, either alone or in the presence of an additional 2.5 mM KCl and 1 mM MgCl2 (KM), as described for the original protocol (54).
Although newly released virions contain only VP1 and VP2, most VP2 N termini eventually become exposed at the virion surface, where they are highly susceptible to proteolysis; thus, virions extracted using the TE8.7 extraction method contain various amounts of particles in which the VP2 termini are truncated to a form called VP3 (44). As can be seen in Fig. Fig.1,1, purified virions showed similar VP1, VP2, and VP3 profiles whether they were recovered in iodixanol containing PBS-KM or PBS only, and when digested with trypsin after dilution into PBS (pH 7.2) or TE8.7, most of the remaining VP2 polypeptides in all samples were converted to VP3 without generating additional proteolytic species or causing the degradation of the VP1 N termini. Thus, the capsids of viruses purified with or without KM appeared similarly resistant to exogenous proteases encountered before or after gradient purification. However, subsequent analyses showed that for some virus extracts, such as those shown in Fig. Fig.11 and and2,2, the structures of the particles after trypsin digestion was very different in these two groups.
When purified virions were digested with trypsin at 37°C for 1 h and rebanded through iodixanol, we observed two different outcomes. All virions originally purified in the presence of PBS supplemented with KM (PBS-KM) rebanded as a single peak at the expected density for full virions (the 45%-to-55% iodixanol interface), as illustrated by the profile in Fig. Fig.2A.2A. In contrast, an equivalent aliquot of the same virus extract purified without KM and then digested and repurified without KM rebanded as two distinct capsid protein peaks, one at the 45%-to-55% iodixanol interface and the other in the 35% iodixanol zone (Fig. (Fig.2B),2B), indicating that this second group had likely undergone a structural rearrangement. If this same −KM virus stock was digested at 37°C without KM but then rebanded through iodixanol-containing KM, the split profile was again observed (Fig. (Fig.2C),2C), but if KM was introduced into the −KM stock prior to the 37°C trypsin incubation, the rearrangement was suppressed, and the virus again migrated as a single peak at the 45-to-55% iodixanol interface (Fig. (Fig.2D).2D). This analysis was repeated, using proteinase K instead of trypsin, and a similar level of protease resistance was observed for virions in both peaks (data not shown). Thus, heating virions at 37°C with protease in the absence of KM induced a physical rearrangement in many particles that caused them to shift in a density gradient, while still remaining largely resistant to protease digestion.
Viral DNA in selected fractions from the gradients shown in Fig. 2A and B (+KM and −KM, respectively) was then analyzed by electrophoresis through denaturing agarose gels and Southern blotting either before or after digestion with micrococcal nuclease (Fig. (Fig.3A).3A). As expected, the vast majority of viral DNA in gradient a, containing KM, resisted digestion with nuclease and comigrated with the capsid protein peak, centered around fraction 4 (Fig. (Fig.3A,3A, left). In gradient b, genomes without KM remained largely intact before nuclease digestion but were split between the 45-to-55% iodixanol interface (fraction 4) and the shifted capsid peak centered on fraction 11 (Fig. (Fig.3A,3A, top right). Remarkably, most of the unit-length genomes were in fraction 11, whereas subgenomic species were heavily enriched in fraction 4, suggesting that virions containing a single 5-kb genome, rather than one or more subgenomic DNA molecules, were predisposed to undergo capsid rearrangement. When the 5-kb DNA in the shifted peak (fractions 10 to 12) was expressed as a percentage of all the 5-kb DNA in both peaks (fractions 3 to 5 plus 10 to 12), this conformational shift was estimated to involve 83% of the population. In contrast, a similar calculation for the capsid proteins, based on the intensity of Coomassie blue-stained VP3 in these fractions, suggested that only 49% of the capsid protein had shifted (Fig. (Fig.2B),2B), supporting the idea that virions with full-length genomes were more likely to rearrange than the subpopulation of defective particles with submolar genomes. This observation suggests that internal pressure exerted by the full-length, 5-kb genome may be involved in driving the rearrangement.
Most of the 5-kb DNA associated with the shifted peak in gradient b (−KM) was not protected from nuclease digestion by the associated capsids (Fig. (Fig.3A,3A, bottom right) and thus had effectively become uncoated. This DNA is discussed in more detail below. However, there were residual submolar fragments of ~2 to 3 kb in fractions 10 and 11 that resisted nuclease digestion. To determine which region of the genome remained protected in these rearranged virions, we reanalyzed the nuclease-digested material from fractions 4 to 5 and 9 to 12, using a set of oligonucleotide probes that hybridize to different regions of the negative-sense progeny DNA strand. As shown in Fig. Fig.3B,3B, the protected fragments were all derived from the right end of the genome. For example, a major band of ~2.5 kb (marked b) hybridized with probes that detected MVM nucleotides at ~2.9 and 4.8 kb but not with probes hybridizing at 2.2 or 1.2 kb, indicating that this fragment encompasses the right half of the genome. These residual protected DNA fragments thus suggest that uncoating begins at the 3′ end of the strand and proceeds in a 3′-to-5′ direction.
Virions from fraction 4 of both the +KM and −KM gradients and the capsid-DNA mixture from fraction 11 of the −KM gradient were immunoprecipitated with a range of antibodies directed against the viral capsid, and the extent to which viral DNA was coprecipitated was assessed by agarose gel electrophoresis and Southern blotting. In each case, antibodies specific for intact capsids precipitated the DNA essentially quantitatively, as assessed by PhosphorImager comparison with input DNA (Fig. (Fig.4A,4A, compare lane 4 with lanes 1 and 7), indicating that the shifted capsids and genomes in the −KM gradient were still physically associated with one another. Significantly, the N termini of VP1 were not accessible to antibodies in either the intact fraction 4 virion samples from the +KM or −KM gradients or the partially uncoated fraction 11 sample from the −KM gradient (Fig. (Fig.4A,4A, lane 3). Thus, we conclude that the shifted fraction 11 peak contains particles in which much of the genome has become uncoated but the capsid remains intact, with its VP1 N termini still sequestered within the particle.
We have previously shown that VP1 N termini, which are typically sequestered within intact virions, become progressively exposed at the surface of intact particles following incremental heating between 45 and 65°C (12), but viral genomes also become exposed during these heating steps. To analyze the connection between these two events at the level of the individual virion and to ask whether divalent cations influence the outcome, virions from fractions 4 of the +KM and −KM gradients were heated to various temperatures for 10 min prior to immunoprecipitation with antibodies directed against the N terminus of VP1 (anti-VP1SR), and the precipitated DNA was analyzed by Southern blotting before and after digestion with micrococcal nuclease. As can be seen in Fig. Fig.4B4B (top left), incremental heating produced a progressive exposure of the VP1 N termini at the surface of +KM virions. Importantly, when an equivalent set of immunoprecipitates was digested with nuclease prior to electrophoresis (Fig. (Fig.4B,4B, bottom left), all of the DNA in particles incubated at 45°C and some of the DNA in samples incubated at 55°C remained protected during VP1SR exposure, indicating that the structural transition(s) leading to VP1 N-terminal extrusion and genome exposure is at least a component of a two-step reaction, which could represent two quite independent processes, each with different energy thresholds.
For equivalent experiments using −KM virions, we included additional temperature increments (Fig. (Fig.4B,4B, top right) in order to check that the transition that exposes VP1 N termini occurred within the usual temperature range (12). Although trace exposure occurred at 37°C, major N-terminal peptide exposure was still first apparent at 45°C, but it became maximal by 55°C. This process occurred equally well in particles that contained 5-kb unit-length DNA or submolar genomes. This suggests that although it occurs in full virions but not in empty capsids, the transition leading to VP1 N-terminal extrusion is not driven predominantly by the presence of the full-length genome. When an equivalent set of precipitates was digested with nuclease prior to electrophoresis (Fig. (Fig.4B,4B, bottom right), all of the 5-kb DNA, but less of the subgenomic species, were found to be exposed, even after the minimum supraphysiological heat pulse. Thus, fraction 4 virions purified in the absence of KM did not respond in exactly the same way as their +KM counterparts and, in particular, failed to show a biphasic response to incremental heating.
Purified single-stranded DNA and the shifted material from fraction 11 of the −KM gradient were incubated with the modified T7 DNA polymerase Sequenase in the presence of [α-32P]dATP and cold deoxynucleoside triphosphates (dNTPs). The resulting duplex products were analyzed by restriction endonuclease digestion, as diagrammed in Fig. Fig.5,5, before and after immunoprecipitation with an antiserum that recognizes intact viral capsids. Both purified single-stranded DNA and the uncoated DNA from fraction 11 primed extensive second-strand synthesis, but the latter was subgenomic, migrating on a neutral gel as a broad band of around 4.5 kb (Fig. (Fig.5B,5B, compare lanes 1 and 2). When digested with restriction enzymes, the resulting left-end fragments were comparable in intensity to those generated from the replication product of purified single-stranded DNA. In contrast, the right-end XhoI and BglI fragments of DNA generated from fraction 11 DNA were mostly absent and were replaced by a DNA smear starting approximately 0.5 kb below the expected band (Fig. (Fig.5B,5B, lanes 4 and 6), even though the analysis shown in Fig. Fig.3A3A showed the template DNA in fraction 11 to be genome length. Thus, although the 3′ hairpins of uncoated genomes were accessible to T7 DNA polymerase, a significant region from the 5′ end of the strand resisted strand progression. This block to synthesis did not occur at a specific nucleotide, as might be expected for site-specific DNA binding, but rather suggested that synthesis became increasingly difficult through this region, presumably because it was impeded by the associated capsids. Immunoprecipitates obtained from the duplex products of partially uncoated virions with antiserum specific for intact capsids received the same extensive proteinase K digestion at 60°C in SDS as the purified products shown in Fig. Fig.5B,5B, but we were never able to remove all of the associated material, and the DNA migrated as a series of retarded smears (compare lane 4 of Fig. Fig.5C5C with lanes 1 and 2 of Fig. 5B and C). Nevertheless, this DNA clearly precipitated specifically with the anticapsid serum and, when digested with XhoI or BglI, released the expected left-end fragment, while the partially replicated right-end fragment was again represented by a series of retarded smears, indicating that it remains associated with some of the residual products of proteolysis. Parallel restriction digests were centrifuged prior to analysis to separate material directly associated with the immunoadsorbent-antigen complex from fragments that were not directly capsid associated. In each case, the fragment from the left end was quantitatively released, while all species from the right end remained capsid associated (Fig. (Fig.5C,5C, lanes 5 to 10). Thus, after complementary-strand synthesis in vitro, these genomes remained quantitatively and strongly associated with the capsid through interactions with a region in their right-hand ends.
Similar syntheses and immunoprecipitations were performed by using HeLa cell in vitro replication extracts in place of T7 DNA polymerase (11). In these experiments, the fraction 11 products were somewhat longer, although still shorter than the products of purified genomic DNA. Right-end fragments again remained quantitatively attached to the capsid-immunoadsorbent complex (data not shown).
To further explore conditions that control this uncoating reaction, we used a miniaturized iodixanol gradient-based assay and simplified the buffer from PBS to one containing 120 mM NaCl in either 50 mM HEPES (pH 7.5) or 50 mM MES (morpholineethanesulfonic acid) (pH 5.5). As described above, full virions and uncoated particles band at the 45-to-55% iodixanol interface and in the 35% iodixanol step, respectively. The products were analyzed by Western or Southern blotting, and for quantitation purposes, particles that migrated in fractions 2 to 4 in these gradients were scored as being intact virions, while those in fractions 6 to 8 were scored as having uncoated. As shown in Fig. Fig.6A,6A, 21% of the virions in this particular virus stock rearranged when incubated with trypsin at pH 7.5 for 1 h at 37°C in the absence of KM, as assessed by Western blotting with a polyclonal antipeptide serum and VP3 quantitation by densitometry. Remarkably, under otherwise similar conditions, only 2% of particles incubated at pH 5.5 underwent a similar rearrangement (Fig. (Fig.6B).6B). Uncoating was also substantially suppressed at pH 7.5 if KM was added to the samples 2 h prior to incubation (to 9%) (Fig. (Fig.6C).6C). To mimic the KCl concentration present in PBS-KM, we then tested the prior addition of 5 mM KCl for its ability to suppress rearrangement, but ~19% of the virus uncoated, indicating that KCl has a minimal, if any, effect on virion instability (Fig. (Fig.6D).6D). In contrast, Fig. Fig.6E6E shows that the addition of 1 mM MgCl2 reduced uncoating to 5%, while the addition of 1 mM CaCl2 appeared even more effective, reducing uncoating to 1% (Fig. (Fig.6F).6F). We conclude that virions do not uncoat at pH 5.5 and that at pH 7.5, magnesium is the critical component of KM that stabilizes the particle, although calcium ions perform this function equally well or better.
We had noticed substantial variation in the extent to which individual virus stocks purified in the absence of KM became uncoated when heated with trypsin. Since this variability seemed likely to reflect differences in their exposures to TE8.7 during extraction, aliquots of a stock in which very few particles typically uncoated under standard conditions were diluted with three volumes of TE8.7 and stored at 4°C for 21 days before processing. As can be seen in the Southern analysis of viral DNA shown in Fig. 7A and B, prior dilution of the virus into TE8.7 had a profound effect. Without reexposure to TE8.7, this virus stock sustained ~3% rearrangement, but ~80% of the particles containing 5-kb genomes uncoated following reexposure. Remarkably, as shown in Fig. Fig.7C,7C, the addition of CaCl2 to 4 mM 3 h prior to the 37°C incubation with trypsin almost completely suppressed rearrangement (to 4%), even though the virus had already been exposed to TE8.7 for 21 days. This finding indicates that the virions did not rearrange at 4°C, although they presumably did lose divalent cations, which promoted their subsequent rearrangement. Instead, the uncoating reaction proceeded only when the sample was exposed to physiological temperature, and the reintroduction of earth metals before this step suppressed the rearrangement.
The dilution of this same virus stock in TE8.7 immediately prior to incubation with trypsin (Fig. (Fig.7D)7D) versus 3 h prior to incubation (Fig. (Fig.7E),7E), induced 34% versus 41%, respectively, of virions to shift in the gradient. Thus, although prior exposure to TE8.7 for 3 weeks did drive the percentage of virions in this stock that rearrange up to ~80%, a substantial fraction of these virions appeared to be predisposed to undergo the shift immediately upon reexposure to TE8.7 and incubation at 37°C. Finally, Fig. Fig.7F7F shows a sample that was diluted in TE8.7 3 h prior to incubation at 37°C but where the trypsin was omitted. Here we observed 28% rearrangement, suggesting that VP2-to-VP3 cleavage may not be needed to potentiate uncoating. Since MVM stocks contain a mixture of particles with variable numbers of intact VP2 N termini, it was still possible that only particles in which most of the VP2 had already been cleaved to VP3 had rearranged. However, Western analysis of the gradient in Fig. Fig.7F7F shows that this was not the case, since both unaffected and rearranged virion populations had similar proportions of intact VP2 termini (Fig. (Fig.7G,7G, compare lane 7 with lanes 2 to 4). Thus, for calcium or magnesium ion-depleted virions, incubation for 1 h at physiological temperature, rather than proteolytic cleavage, appears to drive the uncoating reaction.
Finally, we asked what components of TE8.7 enhance rearrangement. For this, the virus stock used in the experiment shown in Fig. Fig.77 was diluted 1:3 into various buffers, in the absence of divalent cations, 3 h before incubation with trypsin at 37°C, as shown in Fig. Fig.8.8. While prior dilution in 50 mM HEPES-120 mM NaCl had essentially no effect (Fig. (Fig.8A),8A), the addition of 0.5 mM EDTA to this buffer elevated the uncoating reaction to 17% (Fig. (Fig.8B).8B). However, this was substantially less than the 47% uncoating seen for this virus preparation following dilution into TE8.7 (Fig. (Fig.8C)8C) or the 46% uncoating seen when TE8.7 was supplemented with 120 mM NaCl (Fig. (Fig.8D),8D), indicating that pH plays a significant role in potentiating the reaction, while the salt concentration has a negligible effect. This conclusion was supported by the gradients shown in Fig. 8E and F, where buffer comprising either 50 mM Tris-HCl (pH 7.5)-0.5 mM EDTA or 50 mM Tris-HCl (pH 8.7) alone enhanced rearrangement only slightly, to 13% and 9%, respectively. Thus, exposure to EDTA at high pH appears optimal, perhaps suggesting that local basic charges in the capsid structure resist cation loss or that global changes induced by high pH promote uncoating.
In response to cellular and environmental cues, virions can utilize only a limited range of conformational rearrangements that are programmed into their structure. In this study we varied the ionic environment of the MVM particle in vitro and in so doing unleashed a potent genome-uncoating reaction, the details of which are surprising and potentially provide new insight into virion structure. This uncoating reaction occurs following divalent cation depletion, but the depleted particles do not rearrange at 4°C, even after months of storage. However, upon heating to physiological temperature, they undergo a structural shift that ejects much of the viral genome in a 3′-to-5′ direction, while the extreme 5′ end of the DNA remains tightly associated with the otherwise intact capsid. This suggests that during packaging the leading, 3′, end of the viral strand is not deeply buried within the DNA but rather occupies a position proximal to an exit portal and that the rest of the strand is configured in a way that can be easily unraveled in a 3′-to-5′ direction. Moreover, the uncoating reaction can be efficiently inhibited by the addition of 1 mM CaCl2 or MgCl2 2 to 3 h prior to incubation at 37°C, suggesting that readily accessible divalent cation binding sites in the virion induce conformational constraints that somehow ensure genome retention at physiological temperatures. The importance of divalent cations in potentially modulating virion structure was unexpected. Unlike many other viruses, capsid-bound earth metals were not observed in the MVM crystal structure (1, 28), but these may have simply escaped detection because they are present in disordered density and/or are substoichiometric. Notably, almost one-third of the MVM genome appears ordered in these studies, anchored by bases from two short stretches of DNA in each subunit that project into a cavity in the inner capsid wall and establish noncovalent interactions with specific amino acid side chains (1, 10). Two phosphate-chelated magnesium ions per nucleotide were modeled into this density, so it is possible that the modest (0.5 mM) but protracted EDTA extraction conditions employed here might influence this DNA-capsid organization. However, except for their influence on virion rigidity (7, 8), the significance of these DNA-capsid interactions for viral biology remains largely undetermined, and it is not clear whether such cations could be readily removed from the DNA and then simply reintroduced without compromising particle integrity or infectivity (discussed below). Thus, the distribution and biological significance of earth metal binding sites in the MVM virion need to be reexamined.
The significance of the observations presented here for viral morphogenesis during cell entry and uncoating remains uncertain, but these observations provide a model for the previously intractable step of genome release from the rugged, protease-resistant, parvoviral capsid (37) that should be amenable to in vivo analysis. However, the experimental conditions that induce this rearrangement in vitro are unlikely to directly mimic in vivo cues. Thus, although our data indicate that the capsid can retain its 5-kb genome only because of a structural conformation that is dependent on the presence of divalent cations, uncoating in vivo may not be triggered by low ambient calcium concentrations, such as those typically encountered in the cytoplasm, or by a wave of cation chelation, as might occur in the nucleus following a surge in the concentration of deoxynucleotide triphosphates at the start of S phase. Instead, a specific interaction with an unknown cellular trigger could be involved, which induces, or conformationally mimics, cation release. Similarly, we show that uncoating is promoted in vitro by elevated pH, but we do not suggest that the virion encounters pH 8.7 during cell entry, only that this pH either weakens basic interactions that preserve a metastable capsid structure or has global effects, e.g., on the DNA structure, that stress the particle. Finally, most of the experiments described here involve proteolytic digestion, which removes VP2 N termini and perhaps also cleaves the particle at other, undetected, site(s). However, in Fig. Fig.7,7, we show that proteolytic digestion is not absolutely required for uncoating and that the 37°C incubation step is a necessary and sufficient stimulus.
If a similar mechanism is recapitulated, or approximated, in vivo during infectious entry, it would mean that members of the Parvoviridae, like members of the Microviridae (23, 32, 33, 38, 41, 52), employ a genome ejection mechanism rather than the disassembly processes currently known to be used by other virus groups that replicate in eukaryotic nuclei (31, 42). This would perhaps explain why it is possible for parvoviral particles to be so exceptionally robust. The reported mechanism also suggests that after uncoating, the viral DNA might remain physically associated with its parental capsid, which would allow the particle to play an additional role(s), in cis, in trafficking, the selection of a specific nuclear compartment, and/or the initiation of replication or transcription.
Since this reaction became apparent only when virions were analyzed for density or DNA exposure, rather than parameters reflecting capsid integrity such as protease sensitivity or epitope exposure, and because the reaction required limited Ca2+/Mg2+ depletion to be coupled with exposure to 37°C, it is perhaps not surprising that similar reactions have previously gone undetected. We encountered the rearrangement in a limited number of virus stocks while pursuing a different goal and, recognizing that such density shifts could not easily be explained by known mechanisms, sought to track its origin in our virus purification procedure. Since MVM binds to its host cell efficiently, high-titer viral stocks are typically generated by allowing ~75% of the cells in an infected culture to lyse, releasing virus, which mostly binds back on the residual cells and cell debris. This cell-associated virus can then be harvested, and virus can be extracted by repeated freezing and thawing. The extraction buffer used for MVM is unusual in that it relies on the effects of relatively high pH combined with low levels of EDTA, a combination that was selected because it gave optimal recovery of viral particles (44), but exactly why this was the case was unclear for many years. However, with the advent of iodixanol gradient purification, one potential reason for its efficiency became apparent. If PBS was substituted for TE8.7 in the top layers of the iodixanol gradient, the particle recovery of virions at their expected density was greatly reduced, but these particles could be recovered if the extract was digested with neuraminidase prior to centrifugation, suggesting that the special potency of TE8.7 likely depends upon its ability to release MVM from sialic acid residues associated with cell debris. Whether this property is in any way linked to the efficiency with which uncoating is induced at high pH remains to be assessed. The particle-to-infectivity ratios of the resulting stocks are low, typically estimated to be between 1:300 and 1:1,000 (16), but not conspicuously different from those of other members of the genus Parvovirus (21). Infectivity assays invariably failed to detect differences between the virions prepared with KM and those prepared without KM used in this study (data not shown), presumably because the extracted divalent cations were rapidly replaced by dilution in medium. Moreover, cation-depleted stocks behaved exactly like their KM-supplemented counterparts in single-round initiation assays (40), whether newly purified or after ~6 months of storage at 4°C, likely reflecting the fact that uncoating does not occur at 4°C. However, once the particles had rearranged, infectivity was reduced by several orders of magnitude (data not shown).
Although subunit interactions in the MVM capsid appear generally very stable, during virion maturation and cell entry, structural shifts that lead to the progressive extrusion, and then cleavage, of the VP2 N termini occur. Indirect evidence also suggests that the N termini of VP1 become exposed in an unknown cellular entry vacuole (19), since it is essential for the penetration of the cell's lipid bilayer. The trigger that induces this rearrangement in vivo and its vacuolar location remain unknown. In vitro, the exposure of VP1 N termini can be induced by incremental heat pulses, but these heat pulses also result in genome exposure (12, 29). However, in the current study, we have reexamined the response of the virion to supraphysiological heat pulses and shown that in individual particles, the extrusion of VP1 N termini can precede genome exposure. We interpret these observations to mean that these two processes may not necessarily be linked and may represent two quite separate reactions, both of which can be induced by heating. Whether the cation-modulated (CM) uncoating reaction that we describe is the same reaction that occurs during genome exposure following supraphysiological heating remains uncertain and could perhaps best be evaluated by determining the direction in which the genome exits from the heat-treated particle. Certainly, the extrusion of VP1 N termini and CM uncoating are physically separate processes because CM uncoated virions generated at 37°C do not have exposed VP1 N termini. Moreover, whereas VP1 N-terminal extrusion can occur whether or not virions have full-length genomes, CM uncoating appears to be limited to particles with a 5-kb genome. If this reaction does mimic the equivalent step in the viral life cycle, it might explain why parvoviral vectors with approximately unit-length genomes transduce encoded genes more effectively than their sub-unit-length counterparts (6, 26).
The requirement for a 5-kb genome to induce uncoating is interesting because it suggests that internal forces that develop during packaging of full-length DNA provide part of the energy required to drive uncoating, as was suggested previously for other viruses (20). Newly displaced MVM progeny strands are translocated into preformed “empty” capsids in a two-phase reaction that proceeds efficiently through the first (3′) one-third of the genome but is prone to stall thereafter (14). Although a specific packaging complex has yet to be identified in MVM, in adeno-associated virus type 2 (AAV2), the superfamily 3 helicase activity of one of the small Rep proteins is required for the efficient encapsidation of full-length DNA strands (27). This finding indicates that energy is required to pump at least the 5′ half of the strand into the capsid. It was previously suggested that basic amino acids in the VP1 N-terminal peptides may play a role in limiting the electrostatic self-repulsion of the DNA (45), which could in part explain why the heat-pulse-induced extrusion of these termini was closely linked to uncoating in vitro, as discussed above. However, if VP1 N termini do not play such a role and merely occupy space inside the virion, they should add to the internal pressure and hence be expected to promote the uncoating reaction. This is difficult to assess experimentally because currently we can induce VP1 extrusion in vitro only using supraphysiological heat pulses, which invariably induce CM uncoating in cation-depleted particles. Within the context of the viral life cycle, VP1 extrusion is predicted to precede uncoating (discussed in reference 16) so that the latter would need to occur in the absence of internalized VP1 N termini, as might be simulated in vitro by using VP2-only virions (49). However, if basic clusters in the VP1 N termini do play a role in internal-charge neutralization, such VP2-only particles may not package full-length DNA efficiently, so this is clearly an area that needs further study.
In Fig. Fig.9,9, we suggest a “pass-through” or “second-portal” model linking the packaging and uncoating processes for parvoviral genomes. In this model, uncoating does not simply reverse or nullify the effects of an energy-dependent 3′-to-5′ packaging process but instead releases the internalized 3′ telomere, indicating that this palindrome is not deeply buried within the stacked DNA but rather occupies a specific, pore-proximal position. Since the majority of the genome must follow the left-hand end through the exit portal while the 5′ end may still occupy the entry portal, this suggests that the 3′ end of the genome may be positioned near, and ultimately exit the particle through, a different vertex, as shown. This “first-in, first-out” mechanism of uncoating stands in contrast to the “last-in, first-out” unraveling process that was previously observed for the extrusion of DNA from 5-fold cylinder mutants, such as VP2 L172T, which are unable to retain their genomes following the cleavage of VP2 to VP3 (17). While little is presently known about the arrangement of the viral DNA within the virion, the encapsidation of the genome in the “concertina-like” fashion shown in Fig. Fig.99 most easily satisfies this “first-in, first-out” mode of packaging and uncoating, although other topological arrangements are certainly possible.
At present, we do not have experimental data that categorically identify a 5-fold pore as the exit portal, although this seems plausible. The “tether” sequence from the extreme 5′ end of the viral DNA remains outside the virion after packaging (13), so that the ejection of its 3′ telomere through a different portal would leave the right-hand end of the genome threaded through the capsid, as shown in Fig. Fig.9.9. This would provide a topological explanation for the observed 5′ association with the empty particle following uncoating and the inability of DNA polymerase to complete complementary-strand synthesis in vitro. This model is intended only to illustrate the reaction products that we have observed in vitro and does not incorporate any other modifications to virion components that may occur during entry, such as the cleavage of VP2 N termini or the extrusion of VP1 N termini (16). The observed rearrangement provides a potential mechanism for uncoating DNA that could, in theory, be specifically triggered at the required time and locale for replication and that would optimally present the 3′ primer-template end of the negative-sense viral genome to the cellular DNA replication and transcription machinery while the genome remained associated via its 5′ end with the viral capsid.
We thank Tony D'Abramo, Jr., for his valuable assistance in many phases of this study.
This work was supported by Public Health Service grants CA029303 and AI26109 from the National Institutes of Health. S.H. was supported by postdoctoral fellowship AI060155 from the National Institutes of Health and by award AI011219 to Michael G. Rossmann from the National Institutes of Health.
Published ahead of print on 2 December 2009.