Resolution of Parvovirus Dimer Junctions Proceeds through a Novel Heterocruciform Intermediate

The minute virus of mice initiator protein, NS1, excises newcopies of the left viral telomere in a single sequence orientation,dubbed flip, during resolution of the junction between monomergenomes in palindromic dimer intermediate duplexes. We examinedthis reaction in vitro using both ³²P-end-labeled linear substratesand similar unlabeled templates labeled by incorporation of[ ${alpha}$ -³²P]TTP during the synthesis. The observed products suggesta resolution model that explains conservation of the hairpinsequence and in which a novel heterocruciform intermediate playsa crucial role. In vitro, NS1 initiates two replication pathwaysfrom OriL_TC, the single active origin embedded in one arm ofthe dimer junction. NS1-mediated nicking liberates a base-paired3' nucleotide to prime DNA synthesis and, in a reaction we call"read-through synthesis," forks established while the substrateis a linear duplex synthesize DNA in the flop orientation, leadingto DNA amplification but not to junction resolution. Nickingleaves NS1 covalently attached to the 5' end of the DNA, whereit can serve as a 3'-to-5' helicase, unwinding the NS1-associatedstrand. In the second pathway, resolution substrates are createdwhen such unwinding induces the palindrome to reconfigure intoa cruciform prior to fork assembly. New forks can then synthesizeDNA in the flip orientation, copying one cruciform arm and creatinga heterocruciform intermediate. Resolution proceeds via hairpintransfer in the extended arm of the heterocruciform, which releasesone covalently closed duplex telomere and a partially single-strandedjunction intermediate. We suggest that the latter intermediateis finally resolved via an NS1-induced single-strand nick atthe otherwise inactive origin, OriL_GAA.

INTRODUCTION

Top

Introduction

The parvovirus minute virus of mice (MVM) has a negative-senselinear DNA genome in which a single-stranded coding region (of4,805 nucleotides) is bracketed between imperfect terminal palindromesof 121 and 246 nucleotides at the left (3') and right (5') ends,respectively (14). These telomeres play a critical role in rollinghairpin replication, the modified rolling circle replication(RCR) mechanism adopted by MVM, both by providing the sequenceelements that serve as viral replication origins and by operatingas "toggle switches," alternately folding back on themselvesand then unfolding, to adapt the RCR strategy for the replicationof a linear chromosome.

MVM DNA is amplified through a series of monomeric and concatemericduplex replicative-form (RF) intermediates by a unidirectional,leading-strand-specific, cellular replication fork, aided andorchestrated by a single multifunctional viral initiator protein,NS1 (13, 18). Like other RCR initiators, NS1 is a site-specificduplex DNA binding protein with site-specific, single-strandnickase activity. It activates replication by binding to itsduplex recognition sequence, (ACCA)_n, present in all viral origins,and introducing a strand-specific, single-strand nick at anadjacent initiation site. This transesterification reactioncreates a base-paired 3' nucleotide to serve as a primer fornew DNA synthesis while leaving NS1 covalently attached to the5' end of the DNA at the nick site, where it likely serves asa 3'-to-5' helicase (9). The available evidence suggests thatNS1 may also operate the toggle-switch mechanisms which unfoldand refold the viral hairpins (22, 23), although these reactionsremain to be elucidated in detail.

The initial rolling hairpin model (21) predicted that any terminuswhich was not a perfect palindrome would be produced with equalfrequency in two sequence orientations, called flip and flop,which are the inverted complements of one another. While thisprediction holds for the MVM right-end hairpin, the left (viral3') end of the genome is present in both packaged virus andintracellular viral RF DNA in only a single orientation, dubbedflip, as shown in Fig. 1 (1, 2). Moreover, while the right endof the genome serves as an NS1-dependent origin in its hairpinconfiguration (3, 15), the origin associated with the left endis silent until the hairpin is extended and copied during replicationto form the fully base-paired, dimer bridge region, which spansadjacent genomes in dimer RF (10, 17).

View larger version (18K):
[in this window]
[in a new window]
FIG. 1. Structure of the MVM dimer bridge. Step i shows how the left-end hairpin, in its normal flip orientation, is unfolded and copied to create the duplex, palindromic junction that spans adjacent genomes in dimer RF. Negative- and positive-sense strands with respect to transcription are denoted by the symbols - and +. Cross-hatched boxes represent internal palindromic sequences that fold to create the ears in the hairpin form. Mismatched nucleotides in the stem are shown, with the bubble sequence referring to the asymmetry which controls the distribution of the active and inactive origins in the bridge arrangement. Step ii indicates the major products of an in vitro resolution reaction using supercoiled plasmid templates containing the junction sequence (10).

Figure 1 details how the sequence of the 3' hairpin is unfoldedand copied in this dimer junction sequence, with particularreference to internal palindromic sequences, designated the"ears," and an asymmetric mismatched "bubble" within the stemof the hairpin. In the flip orientation of the hairpin, thisbubble contains the triplet 5'-GAA on the inboard strand ofthe hairpin stem, aligned with the doublet 5'-GA on the outboardstrand. In the bridge arrangement, these nucleotides are base-pairedto their complementary sequences, TTC and TC, respectively.OriL_TC, the minimal left origin of MVM is a fully duplex, 50-bpsequence derived from the B arm of the bridge as depicted inFig. 1, which contains the dinucleotide bubble sequence. Theequivalent sequence, oriL_GAA, from the inboard A arm, whichis identical except for the bubble trinucleotide, is inactive.We have previously shown that it is predominantly the number,and not the sequence, of nucleotides in this bubble elementwhich regulates initiation (12) and that it acts by modifyingthe distance between the NS1 binding site and the binding sitefor an essential cellular cofactor, called parvovirus initiationfactor, or PIF (7, 8).

PIF and NS1 bind site specifically to both oriL_TC and oriL_GAA,but these interactions are only cooperative on oriL_TC, forminga ternary complex that effectively stabilizes the interactionof NS1 with OriL_TC during ATP hydrolysis (6). Like other RCRinitiators, NS1 can only nick single-stranded forms of its initiationsite (19). Nicking at duplex oriL_TC thus requires ATP hydrolysisto fuel the unwinding activity that melts the duplex substrateand presents the nick site to the nickase as single-strandedDNA. Without this cooperative interaction between NS1 and PIF,the duplex nick site remains silent, as in oriL_GAA. This distinctionbetween the ability of NS1 to nick single-stranded forms ofits nick site unaided by cofactors and the strict requirementfor cofactors to allow nicking at duplex forms is an importantaspect of the resolution model presented here.

This tight control means that NS1 is only able to introduceone single-stranded nick into the duplex bridge sequence, intothe active initiation site in the lower strand of the B arm,as shown in Fig. 1. A new fork initiated at this site on a linearsubstrate would synthesize DNA in the flop orientation, ultimatelyrecreating the linear duplex and displacing one positive-sensesingle strand. We observed this type of reaction in the experimentsdescribed here and designate it "read-through synthesis." Wesuggest that it represents an amplification, rather than a fullresolution, pathway and that such products are not the onlyoutcome of initiation at OriL_TC.

We have previously shown that templates containing the dimerbridge sequence as part of a supercoiled plasmid can be resolvedin vitro using cellular replication extracts and recombinantNS1 to generate two new duplex viral termini, one associatedwith each arm of the junction palindrome. This resolution wasfound to be somewhat asymmetric, as shown in Fig. 1, generatingpredominantly NS1-associated "extended forms" of the A arm,while the B arm was most frequently represented as "turn-around"hairpin forms in which the two strands of the duplex were covalentlycontinuous through the axis of symmetry (10, 17).

Here we reexamine the dimer resolution reaction in vitro, usinglinear substrates and develop a resolution model based on thestructure and temporal expression of three resolution intermediates,which we call MJ1, MJ2, and ${delta}$ J, and on the distribution of newlysynthesized DNA. Evidence is presented that supports the characterizationof MJ2 as a heterocruciform structure. The proposed mechanismrelies on characteristic parvoviral duplex-to-hairpin rearrangementsof the palindrome, mediated by NS1, to explain bridge resolutionand the generation of negative-sense genomes with left-end telomeresexclusively in the flip orientation.

MATERIALS AND METHODS

Top

Materials and Methods

Templates. The pUC18-based plasmid pLEB711, containing a palindromic, 711-bpPstI fragment (MVM nucleotide 415) derived from the left endto the left-end junction of dimeric RF DNA, has been describedpreviously (16). A 2,103-bp AseI fragment containing the dimerbridge was excised, and the ends were filled with unlabeledTTP and dATP by using Sequenase (U.S. Biochemical, Cleveland,Ohio) or they were 3' end labeled with [ ${alpha}$ -³²P]TTP and unlabeleddATP.

Resolution assays. Recombinant histidine-tagged MVM NS1 was expressed in HeLa cellsfrom recombinant vaccinia virus and purified by nickel-chelatechromatography as previously described (20). Recombinant RPA,expressed from baculovirus vectors and purified as describedelsewhere (9), was kindly provided by Jesper Christensen. Replicationextracts were prepared from uninfected A9 cells essentiallyas described by Wobbe et al. (24), except that they were dialyzedagainst 20 mM Tris-HCl (pH 7.5), 0.2 mM EDTA, 25 mM NaCl, 0.5mM dithiothreitol, 10% glycerol, and 10% sucrose, which concentratedthe extracts approximately twofold. Assays were carried outas previously described (11), in a volume of 20 µl thatcontained recombinant NS1 (2.5 µg/ml), deoxynucleotides,MgCl₂, ribonucleotides, ATP, and an ATP-regenerating system,and either ³²P-3'-end-labeled template DNA (1.25 µg/ml)or unlabeled, blunt-ended template DNA plus [ ${alpha}$ -³²P]TTP. Reactionswere terminated by addition of sodium dodecyl sulfate (SDS)and EDTA to final concentrations of 0.5% and 5 mM, respectively,and analyzed on neutral agarose gels (one dimension) or neutralfollowed by alkaline agarose gels (two dimensions), as describedpreviously (11). Samples to be immunoprecipitated were incubatedin buffer containing 2% SDS at 60°C for 20 min to disruptnoncovalent interactions, diluted, and precipitated with rabbitantiserum directed against the N-terminal domain of NS1. Unlessotherwise specified, neutral agarose gels contained 0.2% SDSin both gel and buffer.

RESULTS

Top

Results

Resolution products of linear dimer bridge substrates. Linear substrates containing a 711-bp cloned copy of the MVMdimer bridge positioned asymmetrically in vector sequences areshown in Fig. 2A. Use of an asymmetric substrate allowed usto distinguish resolution products of around 1.6 kb derivedfrom the A arm of the bridge from those of around 0.5 kb derivedfrom the B arm. To identify the synthetic steps associated withresolution, we compared the products of ³²P-3'-end-labeled lineartemplates, where we could monitor the fate of each input strand,with products generated from similar, but unlabeled, substrateswhich were labeled during the course of the reaction by incorporationof [ ${alpha}$ -³²P]TTP into newly synthesized DNA.

View larger version (23K):
[in this window]
[in a new window]
FIG.2. Resolution products of ³²P-3'-end-labeled linear templates and of similar unlabeled templates, labeled by incorporation of [ ${alpha}$ -³²P]TTP into newly synthesized DNA. (A) Diagram of the 2,103-bp linear AseI template containing an asymmetrically placed 711-bp MVM dimer bridge sequence (grey box). The axis of symmetry separates the A arm of the bridge, containing the inactive duplex origin sequence OriL_GAA, from the B arm of the bridge containing the active origin, OriL_TC. The active and inactive NS1 nick sites in these sequences are marked by a dark and light arrow, respectively. Asterisks mark the position of ³²P label present in 3'-end-labeled substrates. The predicted sizes of extended and turn-around forms of each arm are specified. (B) Autoradiograph of a 1.4% neutral agarose gel containing 0.2% SDS. ³²P-3'-labeled substrates are shown before (0.5x concentration, lane 1) and after (lane 2) incubation with cell extract alone or cell extract supplemented with recombinant NS1 (lanes 3 and 4). Unlabeled substrates incubated with [ ${alpha}$ -³²P]TTP in the presence of cell extract alone are shown in lane 5, and those with extract supplemented with NS1 are shown in lanes 6 and 7. Samples in lanes 4 and 7 were digested with proteinase K prior to electrophoresis. The single-stranded form of the³²P-3'-labeled substrate, melted by incubation at 95°C for 5 min, is shown in lane 8. "A ext" and "B ext" denote extended forms of the A and B arms of the template (as depicted in panel A), while "A ta" and "B ta" denote turn-around forms. MJ1 and MJ2, for modified junction fragments 1 and 2, and ${delta}$ J denote putative resolution intermediates. Fragments denoted "+NS1" are still covalently associated with NS1 and so show retarded migration relative to their proteinase-treated neighbors. (C) Autoradiograph of a 1.4% neutral agarose gel containing 0.2% SDS. The resolution products of ³²P-3'-labeled templates are shown in lanes 1 to 4, and the products labeled by incorporation of [ ${alpha}$ -³²P]TTP into newly synthesized DNA are shown in lanes 5 to 8, before (lanes 1 and 5) and after (lanes 4 and 8) immunoprecipitation with nonspecific rabbit serum or anti-NS1-specific serum (lanes 3 and 7). Unbound materials from the anti-NS1 precipitates are shown in lanes 2 and 6. All samples were digested with proteinase K prior to electrophoresis. Fragments are labeled as detailed for panel B.

(i) Resolution products of 3'-end-labeled linear substrates. When incubated with A9 cell extracts in the absence of NS1,³²P-3'-end-labeled linear fragments containing the dimer bridgeshowed little evidence of modification (Fig. 2B, compare lanes1 and 2), although some dimeric and trimeric ligation productswere generated. Such ligation products are the inevitable consequenceof incubating blunt-ended substrates in cell extracts, but theyare likely irrelevant for MVM replication in vivo, since viralreplication intermediates do not have exposed blunt ends. Incontrast, when NS1 was included in the reaction mixture (lane3), much of the input DNA was modified, generating the expectedduplex resolution products (extended and turn-around forms ofthe A and B arms), together with three higher-molecular-weightspecies (labeled MJ1 and MJ2, for modified junction 1 and 2,and ${delta}$ J) which are potential resolution intermediates. In thefollowing analyses we show, in detail, the structure of eachof these products and intermediates.

As observed previously using supercoiled dimer bridge substrateslabeled by [ ${alpha}$ -³²P]TTP incorporation (10), resolution of the linearsubstrates appeared asymmetric, in that extended forms of theA arm greatly outnumbered its turn-around form. However, inthe present analysis, extended and turn-around forms of theB arm often appeared approximately equimolar and were greatlyoutnumbered by duplex forms of the A arm.

The native agarose gel shown in Fig. 2B contained SDS, allowingfragments that were covalently associated with NS1 to enterthe gel. When a sample equivalent to the one shown in lane 3was digested with proteinase K prior to electrophoresis (lane4), potential intermediates MJ1, MJ2, ${delta}$ J, and extended formsof the A and B arms showed enhanced mobility, indicating thatthey were initially covalently complexed with NS1. In contrast,unit-length input molecules and turn-around forms of the A andB arms were not NS1 associated and remained in the unbound fractionfollowing immunoprecipitation (Fig. 2C, lane 2).

Since MJ1 and MJ2 forms migrated more slowly than input DNAeven after proteinase K digestion, they likely contained moreDNA than the input, while ${delta}$ J migrated faster than input, suggestingthat it had lost DNA. None of the observed forms comigratedwith denatured template (Fig. 2B, lane 8), confirming that theywere not simply melted-out forms of the substrate.

(ii) Products labeled by incorporating [ ${alpha}$ -³²P]TTP into newly synthesized DNA. In the absence of NS1, unlabeled linear fragments containingthe dimer bridge were barely labeled in vitro (Fig. 2B, lane5). Introducing NS1 into the reaction mixture (lane 6) resultedin the appearance of newly synthesized DNA in a variety of species,including unit-length linear molecules that comigrated withthe 3'-end-labeled input DNA shown in lane 1. Digestion withproteinase K did not affect the mobility of this product (lane7), and it did not immunoprecipitate with anti-NS1 serum (Fig.2C, compare lanes 6 and 7), indicating that it was not covalentlyassociated with NS1 even though NS1 was required for its synthesis(Fig. 2B, compare lane 5 with lanes 6 and 7). We suggest thatthese labeled unit-length molecules are the products of read-throughreplication (discussed below), which allows DNA amplificationwithout junction resolution.

Surprisingly, the resolution product which contained most ofthe newly synthesized DNA was the turn-around form of the Barm. This product was identified as a B turn-around form becauseit migrated at $~$ 0.5 kb under neutral conditions but migratedat $~$ 1 kb under denaturing conditions (see below), indicatingthat the strands of the duplex were covalently continuous atone end. In addition, this DNA species was not directly associatedwith NS1, as confirmed by its presence in the unbound fractionof the anti-NS1 immunoprecipitate shown in Fig. 2C, lane 6.

All fully resolved forms of the bridge (i.e., extended and turn-aroundforms of both the A and B arms) contained at least some newlysynthesized DNA, as did the NS1-associated MJ1 and MJ2 intermediates(Fig. 2C, lane 7). In contrast, the ${delta}$ J intermediate was not significantlylabeled by in vitro synthesis (compare Fig. 2B, lane 4 with7, and 2C, lane 3 with 7), suggesting that it did not containsignificant stretches of newly synthesized DNA.

Read-through synthesis. Unit-length duplex molecules labeled by incorporation of [ ${alpha}$ -³²P]TTPinto newly synthesized DNA in the NS1-dependent reaction (Fig.2B, lanes 6 and 7, and C, lanes 5 and 6) are products of a typeof reaction we call read-through synthesis, which does not leaddirectly to the generation of duplex resolution products. Inthe amplification pathway depicted in Fig. 3A, a leading-strand-specificcellular fork, which can be assembled in vitro with recombinantRPA, RFC, PCNA, and DNA polymerase ${delta}$ (S. F. Cotmore, J. Christensen,and P. Tattersall, unpublished results), forms rapidly on thenewly available 3' hydroxyl (step i), allowing strand displacementsynthesis to occur on the linear template (step ii). This displacesan NS1-associated, single-stranded, positive-sense form of theA arm, while replacing it with labeled, newly synthesized DNA(step iii). Displaced single-stranded forms of the A arm wereidentified in the resolution products of 3'-end-labeled substrates,but they migrated as a diffuse band in the SDS-agarose gelsshown here and were minor products, perhaps because single-strandedDNA is more susceptible to degradation than duplex DNA in complexcell extracts. In Fig. 3B, lane 1, we show proteinase K-digestedproducts of an NS1-driven reaction displayed on a non-SDS gel.Single-stranded forms of the A arm can be identified in theseproducts, but they become particularly apparent when comparedto products shown in lane 2, where the release of single-strandedforms has been promoted by the inclusion of a large molar excessof recombinant RPA. Their identity as NS1-associated 1.6-kbsingle strands was confirmed by two-dimensional gel electrophoresisand by their specific immunoprecipitation with anti-NS1 serum(data not shown).

View larger version (20K):
[in this window]
[in a new window]
FIG. 3. Read-through synthesis. (A) Diagram of read-through synthesis, as discussed in the text. "A s-s" denotes the single-stranded form of the A arm, with its 5' hairpin; "axis" is the symmetry axis of the junction; other symbols are as detailed in the legend for Fig. 1. (B) Autoradiograph of a 1.4% neutral agarose gel (without SDS) showing resolution products of ³²P-3'-end-labeled templates incubated in a standard, NS1-supplemented, resolution mix (lane 1) or in a mix supplemented with recombinant RPA (25 µg/ml). Fragments are labeled as detailed for Fig. 2B.

Since the dimer bridge remains in the linear configuration throughoutthis read-through reaction, new DNA would be synthesized inthe flop orientation on the negative-sense primer strand (Fig.3A, step ii). However, since this reaction does not appear tolead directly to resolution of the resulting duplex product,no termini in the flop orientation would be excised.

Time course of the resolution reaction. Products generated and labeled by incorporation of [ ${alpha}$ -³²P]TTPinto newly synthesized DNA were analyzed, at intervals duringthe course of a 2-h reaction, by electrophoresis through neutralSDS-agarose gels, both before (Fig. 4A, lanes 2 to 8) and after(Fig. 4A, lanes 9 to 12) proteinase K digestion. During thefirst 90 min of the reaction, the rate of [ ${alpha}$ -³²P]TTP incorporationinto DNA was approximately linear, but declined thereafter.The clear redistribution of incorporated label with time, fromhigher-molecular-weight structures at the earliest time pointsto the final resolution products after 1 to 2 h, suggests thatthis reaction was relatively synchronous and that initiationwas most common early during synthesis, perhaps because activeNS1 concentrations became limiting thereafter.

View larger version (48K):
[in this window]
[in a new window]
FIG. 4. Time course of resolution reactions. (A) Autoradiographs of 1.4% neutral agarose gels containing 0.2% SDS. Unlabeled templates, replicated in the presence of [ ${alpha}$ -³²P]TTP, are shown before (lanes 2 to 8) and after (lanes 9 to 12) digestion with proteinase K. For size comparison, ³²P-3'-labeled template is shown in lane 1. Samples were harvested after incubation for 10 min (lane 2), 20 min (lane 3), 30 min (lanes 4, 5, and 9), 60 min (lanes 6 and 10), 90 min (lanes 7 and 11), or 2 h (lanes 8 and 12). Fragments are labeled as detailed previously (Fig. 2B). (B) Autoradiograph of a 1.4% neutral agarose gel containing 0.2% SDS, showing the products of ³²P-3'-end-labeled substrates before (0.5x concentration, lanes 1 to 3) and after (lanes 4 to 6) immunoprecipitation with antiserum directed against NS1, or the unbound material from these precipitates (lane 7 to 9). Reaction products were harvested after incubation for 30 min (lanes 1, 4, and 7), 60 min (lanes 2, 5, and 8), or 90 min (lanes 3, 6, and 9).

The major labeled species observed in samples incubated for30 min or less (Fig. 4A, lanes 2 to 5) were the MJ intermediatesand unit-length linear molecules, presumably labeled by read-throughsynthesis as discussed above. Notably, while the MJ2 form appearedearly, its concentration remained fairly constant with time,while the more heterogeneous MJ1 forms increased in concentrationduring the first 30 min but declined significantly thereafter,as extended and turn-around forms of both arms, but most notablyturn-around forms of the B arm, accumulated. This suggests thatMJ1, in particular, is an active intermediate in the generationof new termini.

Generation and processing of ³²P-3'-end-labeled substrates werealso analyzed as a function of time in an NS1-dependent reaction(Fig. 4B). Analysis of total reaction products harvested at30-min intervals (lanes 1 to 3) and species immunoprecipitatedfrom reactions with anti-NS1 serum (lanes 4 to 6) showed thatthe heterogeneous MJ1 species appeared early and declined withtime, while the ${delta}$ J intermediate progressively accumulated. Residualinput DNA remained in the unbound fraction of these precipitates(lanes 7 to 9), together with turn-around forms of the resolutionproducts.

Structures of MJ1, MJ2, and ${delta}$ J intermediates. Products generated and labeled by incorporation of [ ${alpha}$ -³²P]TTPinto newly synthesized DNA from the 60-min time point were digestedwith proteinase K (as in Fig. 4A, lane 10) and further analyzedby two-dimensional electrophoresis under native and then denaturingconditions (Fig. 5A). This type of analysis allowed us to distinguishextended forms of the telomeres, where the two strands of theduplex were free to separate under denaturing conditions, fromturn-around forms, where the two strands were held togethercovalently at one end. Thus, extended forms of the B arm migratedas $~$ 0.5-kb duplexes in the first dimension and as $~$ 0.5-kb singlestrands in the second, while turn-around forms of the B armmigrated as somewhat smaller duplexes than extended forms butappeared almost twice as long (approximately 1 kb) when denatured.

View larger version (31K):
[in this window]
[in a new window]
FIG.5. Two-dimensional analysis of resolution products on native and denaturing gels. Samples were digested with proteinase K and subjected to electrophoresis under native conditions, as shown for a parallel track positioned along the left side of each panel. The sample lane was then subjected to electrophoresis, at 90° to the original direction, under alkaline, denaturing conditions. (A) Autoradiograph of a two-dimensional 1.4% agarose gel showing resolution products labeled by incorporation of [ ${alpha}$ -³²P]TTP into newly synthesized DNA, extracted after a 60-min reaction. Fragments are labeled as detailed in the legend for Fig. 2B. (B) Autoradiograph of a two-dimensional 1.4% agarose gel showing total resolution products of ³²P-3'-end-labeled templates, extracted after a 60-min reaction. Arrows indicate both of the 3'-labeled species originating from the single MJ1, MJ2, and ${delta}$ J bands present in the neutral dimension. (C) Autoradiograph of a two-dimensional agarose gel showing resolution products of ³²P-3'-end-labeled templates, as in panel B but immunoprecipitated with anti-NS1 serum prior to proteinase K digestion. Regions containing the labeled resolution intermediates are enlarged relative to those shown in panels A and B, to allow discrimination of both the 2.1-kb and 1.6-kb spots derived from MJ1, MJ2, and ${delta}$ J intermediates. (D) The probable configuration of MJ2 intermediates labeled by incorporation of [ ${alpha}$ -³²P]TTP into newly synthesized DNA (thick dahsed line). Dotted lines with arrows indicate the labeled fragment of 557 nucleotides. (E) The probable configuration of ³²P-3'-end-labeled MJ2 intermediates. Dotted lines with arrows indicate labeled fragments of 2,103 and 1,666 nucleotides.

These gels also allowed us to dissect high-molecular-weightproducts and intermediates and thus to identify what types ofDNA fragment contained newly synthesized DNA. Unit-length linearmolecules, presumably labeled by read-through synthesis, migratedas labeled $~$ 2-kb duplexes in the first dimension and as $~$ 2-kbsingle strands in the second, indicating that the new DNA wascontained within full-length DNA strands. In contrast, noneof the new DNA in the MJ1 and MJ2 intermediates was associatedwith a full-length strand. Instead, most of the new DNA in theMJ2 intermediate was present in a single strand of approximately0.5 kb which comigrated in the alkaline dimension with extendedforms of the B arm. In Fig. 5D we suggest a heterocruciformstructure for this MJ2 intermediate, in which new DNA is confinedto an extended form of the B arm. This newly synthesized DNAis predicted to be in the flip orientation, but this has notbeen formally confirmed.

New DNA associated with the more heterogeneous MJ1 intermediatemigrated as an extended diagonal in the alkaline dimension,ranging from the position of an extended form of the B arm tothe size of a B turn-around form in the more slowly migratingduplex MJ1 forms (Fig. 5A). Extrapolating from the MJ2 structuredepicted in Fig. 5D, we suggest that MJ2 forms are likely convertedto MJ1 forms by a hairpin transfer event that melts out thestrands of the lower, extended-form, duplex and allows themto fold back on themselves, repositioning the 3' nucleotideso that it primes synthesis back along the B arm. Accordingly,MJ1 forms would be heterogeneous, because they include a spectrumof molecules in which the returning fork has progressed to differentlengths back along the B arm.

Two-dimensional analysis of 3'-end-labeled products (Fig. 5B)indicated that MJ1 and MJ2 intermediates were relatively minorspecies. Although the 3'-end-labeled forms associated with theseintermediates were faint, they clearly contained one strandof $~$ 1.6 kb, corresponding to a 3'-end-labeled form of the lowerstrand that had been nicked in the B arm at OriL_TC, and onefull-length strand of 2.1 kb. This full-length strand is difficultto see for the MJ2 intermediate in Fig. 5B, since it migratescoincidentally with the residual input substrate; however, itcan be readily detected in two-dimensional gels of 3'-labeledproducts immunoprecipitated by virtue of their covalent associationwith NS1, as shown at higher magnification in Fig. 5C, sinceresidual input is not present in these precipitates. Thus, MJ2(and MJ1) and products are in accord with those detailed inthe proposed heterocruciform structure of the MJ2 intermediateas shown for 3'-end-labeled substrates in Fig. 5E, having one3'-labeled intact strand and one 3'-labeled strand that hasbeen nicked at OriL_TC, creating an extended form of the A arm.

Most notably, labeled forms of the B arm (0.5 to 1 kb) werenot apparent in the 3'-end-labeled MJ intermediates shown inFig. 5B, indicating that the B forms labeled so heavily by invitro synthesis in Fig. 5A were most likely produced by nickingand extension of the lower strand of the B arm, which wouldnot carry a 3' end label, as illustrated in Fig. 5E.

Figures 5B and C also reveal the structure of the ${delta}$ J intermediate.This structure was not visible in Fig. 5A because it containedno newly synthesized DNA, but in the products of the 3'-end-labeledsubstrates it could be seen to contain an intact 2.1-kb strandand a 1.6-kb extended-form A arm. Since it is also present inFig. 5C it must carry a covalently associated NS1 molecule,but it migrates faster than substrate DNA in the neutral dimension,so that it must have lost some DNA, presumably the unlabeledlower strand of the B arm. We therefore suggest that ${delta}$ J molecules(together with free turn-around forms of the B arm) are theproducts of MJ1 intermediates in which the returning fork hascompleted synthesis of the B turn-around structure, causingit to be released.

DISCUSSION

Top

Discussion

The products and intermediates of the in vitro dimer junctionresolution reaction are complex, in part because two differenttypes of reactions appear to be going on simultaneously. Oneof these, read-through synthesis, would provide an efficientmechanism for DNA amplification but would regenerate the duplexdimer intermediate rather than resolve the bridge structureinto separate duplex telomeres. The second type of reaction,which provides the resolution pathway as described below, reliesupon the ability of NS1 to induce the palindrome to undergoa duplex-to-hairpin transition, creating a resolution substratein which the nick site in the A arm is ultimately exposed asa single strand.

In Fig. 6 we suggest a heterocruciform resolution model to explainhow NS1 is able to bring about resolution of the dimer bridgeinto separate duplex telomeres while conserving the flip configurationof the hairpin on all negative-sense strands. This model takesinto account the structures and apparent kinetic relationshipsbetween the various resolution intermediates and products observedabove.

View larger version (24K):
[in this window]
[in a new window]
FIG. 6. The resolution pathway. After nicking the initiation site in the B arm of the dimer bridge (step i), NS1 associates with RPA to function as a 3'-to-5' helicase (step ii), unwinding the lower strand of the palindrome and allowing the exposed single strands to fold back on themselves, creating a cruciform intermediate (step iii). Branch migration proceeds (step iv), eventually passing the inactive initiation site in the A arm. At this point the exposed 3' nucleotide can switch templates and anneal to its complement in the lower cruciform arm (step v). A replication fork assembling at this time will copy and unwind the cruciform arm, synthesizing a palindrome in the flip orientation on the end of the negative-sense B strand (step vi). This heterocruciform structure corresponds to the MJ2 intermediate. In a second duplex-to-hairpin transition, the palindromic heterocruciform arm of MJ2 is then melted out and both strands fold back on themselves (step vii), allowing the exposed 3' end to base-pair with inboard sequences in the B arm. A replication fork established at this 3' end would copy the lower strand of the B arm (step viii), creating the MJ1 intermediate and progressively displacing the upper strand, leading to the eventual release of a newly synthesized B turn-around form (step ix). The residual ${delta}$ J intermediate is partially single stranded, having an intact upper strand paired to an NS1-associated lower strand from the A arm. Since this complex carries the active helicase, it is presumed to be a dynamic structure in which the bridge palindrome is periodically reconfigured into a cruciform structure, as shown.

Read-through synthesis versus resolution. According to this model, NS1 initially introduces a site-specific,single-strand nick into OriL_TC, in the B arm of the bridge,becoming covalently attached to the DNA at the 5' side of thenick and exposing a base-paired 3' nucleotide (Fig. 6, stepi). This reaction can be reproduced in vitro with 3'-labeleddimer bridge substrates in standard nicking assays using onlyrecombinant NS1 and PIF (data not shown). NS1 would then beexpected to bind RPA (9) and recruit it to the nick site, buttwo possible outcomes would ensue, depending upon the speedwith which a replication fork was assembled on the exposed 3'nucleotide. As illustrated in Fig. 3A, if a fork assembled rapidlywhile the bridge remained in the linear configuration, read-throughsynthesis would copy the upper strand, displacing a positive-sensesingle strand from the A arm and regenerating the duplex substrate.In vivo, in the context of an MVM infection, we would expectthat the displaced positive-sense single strand would be convertedrapidly into an extended-form duplex telomere by a fork arisingfrom the origin at the right end of the viral genome. However,since our in vitro substrates lack right-end origins, the duplexresolution products we have observed must be generated by analternative mechanism.

Formation of the cruciform resolution substrate. In order for resolution to occur, we suggest that the initialnicking event in OriL_TC must be followed by melting and rearrangementof the bridge palindrome into a cruciform intermediate priorto fork assembly, allowing synthesis of new DNA in the fliporientation. Such duplex-to-hairpin rearrangements could beaccomplished by a simple helicase mechanism, as illustratedin Fig. 6, steps ii to iv. In this pathway assembly of the forkis delayed, so that in the presence of RPA the NS1 3'-to-5'helicase activity begins to unwind the duplex, as recently documented(9). As the helicase passes through the axis of symmetry, thetwo separated strands begin to fold back on themselves, andbranch migration continues until the cruciform arms extend pastthe nick site (steps iii and iv).

Alternatively, NS1 may be able to melt out and reconfigure theMVM dimer bridge palindrome in an RPA-independent reaction.Willwand and colleagues (23) observed NS1-dependent, hairpin-primedreplication at both termini of a duplex MVM DNA template usingcell extracts in vitro, and they went on to show that purifiedNS1 was able to melt out extended forms of the MVM right-endtelomere in the absence of cellular replication factors (22).This latter reaction was dependent upon the presence of bothcopies of the paired NS1 binding sites which surround the symmetryaxis of the right-end telomere. Although plausible, at presentthere is no direct experimental evidence for comparable reactionsinvolving extended-form left-end telomeres of MVM or the dimerbridge.

Synthesis of the heterocruciform. Once the cruciform extends to include sequences beyond the nicksite in the A arm, the primer exposed at the nick site in OriL_TCcan undergo a template switch by annealing with its complementin the lower cruciform arm (Fig. 6, step v). If a replicationfork assembles after this point, the resulting synthesis willunfold and copy the flop sequence of the lower cruciform arm,with NS1 perhaps again serving as the 3'-to-5' helicase requiredfor strand displacement. This would create a heterocruciformintermediate that contains a newly synthesized telomere in theflip orientation attached to the lower (-) strand of the B arm(step vi). Such a structure corresponds exactly to that of theMJ2 intermediate, as determined in Fig. 5.

We had previously suggested that a cellular or viral recombinasecould resolve heterocruciform intermediates of this type, withdirectional specificity, into the expected termini (13, 14).However, such a reaction would transfer the labeled, newly synthesizedDNA (depicted in step vi) onto an extended form of the A arm,leaving the B turn-around arm unlabeled. Since this is not compatiblewith our experimental data, in which turn-around forms of theB arm are seen to accumulate most of the newly synthesized DNA,we now conclude that a directional recombinase is not responsiblefor subsequent resolution.

Hairpin transfer and release of the B turn-around arm. We suggest that synthesis pauses after completion of the MJ2intermediate because the fork reaches the end of its templatestrand (Fig. 6, step vi). Although evidence of a precursor-productrelationship has not been established for the subsequent transition,we suggest that MJ2 forms are converted to MJ1 forms by a hairpintransfer event that allows the two strands of the lower heterocruciformarm to melt out and fold back on themselves, repositioning theexposed 3' nucleotide on the B arm (Fig. 6, step vii) and allowinga new fork established on this primer to synthesize a B turn-aroundform (step viii). This hairpin transfer reaction could be mediatedby either of the NS1-driven mechanisms discussed previouslyin relation to dimer bridge substrates. Thus, according to thenew model, both creation of the heterocruciform and its subsequentresolution would be driven by the characteristic NS1-mediatedduplex-to-hairpin rearrangements of the palindrome that arethe hallmark of parvovirus DNA amplification.

Upon completion of the B turn-around structure, it would bereleased from the MJ1 intermediate (Fig. 6, step ix), leavinga ${delta}$ J intermediate which consists of one intact DNA strand base-pairedto an extended form of the A arm. Since the A arm of this intermediatehas an NS1-associated, recessed 5' end, the 3'-to-5' helicaseactivity of NS1 would again tend to unwind the lower strandof the palindrome, allowing the single strands to fold backon themselves and creating the branched intermediate depictedin Fig. 6.

Introduction of a single-strand nick and resolution of the ${delta}$ J intermediate. In the absence of a large excess of RPA, the helicase activityof NS1 rarely manages to fully displace the lower strand ofthe ${delta}$ J intermediate (Fig. 3B), but it likely creates a dynamicstructure in which the nick site in the inactive A arm (OriL_GAA)is transiently, but repeatedly, exposed in a single-strandedform. We speculate that under these conditions OriL_GAA can benicked by free NS1 in a PIF-independent reaction (Fig. 7, stepsi and ii). This type of single-stranded nicking reaction couldalso occur on MJ1 and MJ2 intermediates, perhaps bypassing releaseof the ${delta}$ J form. However, ${delta}$ J forms do accumulate, suggesting thatthe postulated single-strand nicking reaction may be inefficientor that NS1 becomes limiting at late times in vitro.

View larger version (17K):
[in this window]
[in a new window]
FIG. 7. Introduction of a single-strand nick and resolution of the ${delta}$ J intermediate. The initiation site in the A arm of the palindrome is periodically exposed as a single strand during duplex-to-hairpin rearrangements of ${delta}$ J (step i). This allows NS1 to attack the initiation site in OriL_GAA without the help of a cofactor (step ii). Nicking leads to the release of a positive-sense B strand and leaves a base-paired 3' nucleotide on the A arm (step iii) to prime assembly of a fork which will copy the hairpin, creating an extended form of the A arm (step iv).

This final nick would leave NS1 covalently attached to a positive-senseB arm (Fig. 7, step ii), which would be released from the complexby strand displacement synthesis from the new 3' nucleotideexposed at the nick site, generating a new extended form ofthe A arm (Fig. 7, step iv). In vivo, in the context of theMVM genome, this single-strand nick may be a rare event. Thisis because both ends of dimer RF bear efficient "right-end"hairpin origins and, in vivo, forks may proceed back towardsthe dimer bridge from the right end of the genome, along thetop strand of the B arm, at the same time that a turn-aroundform of the B arm is being released from MJ1-like intermediates.This would effectively bypass dimer bridge resolution, recyclingthe top strand into a replicating duplex dimer pool. Since weshow here that there is a mechanism which will allow NS1 tobring about the full resolution of dimer bridge structures intoduplex forms of the two arms, we suggest that, in vivo, whichprocess dominates may depend, to a surprising extent, upon theefficiency of the origins at the right end of the genome.

Although all other termini generated in this resolution pathwayare in the flip orientation, the final single-stranded nickwould generate a positive-sense strand with a flop hairpin inwhich the three-nucleotide bubble sequence, GAA, is positionedin the outer arm. Although such telomeres were not observedin vivo for MVM (1), they have been identified in the closelyrelated virus LuIII (5). Unlike MVM, LuIII packages DNA strandsof both senses with equal efficiency (4), and while the left-endhairpins of its negative-sense strands are all in the flip orientation,in positive-sense strands flip and flop orientations are expressedwith equal frequency (5). We suggest that there could be twodifferent mechanisms for processing ${delta}$ J intermediates and thatthe balance between these that is achieved in vivo is determinedby the rate of hairpin transfer at the right end of the viralgenome. Since the orientation of termini excised by these twomechanisms is somewhat different, one way to test this predictionexperimentally would be to impair the efficiency of the MVMright-end origin by genetic manipulation.

Preliminary experiments indicate that nature has already exploredthis avenue. We are currently mapping the sequences in the LuIIIgenome which allow it to package all of the resolution productspredicted from the heterocruciform model, and we find that thisability maps to a 2-nucleotide insertion in the NS1 nick siteof the right-end origin. This insertion has the effect of makingthe LuIII origin refractory to nicking when compared in vitroto its MVM counterpart. Thus, when the right-end hairpin ofLuIII is substituted for the equivalent MVM sequence, a virusis generated which has a slow right-end origin and which packages,in addition to negative-sense strands, positive strands whoseleft-end termini exist in both flip and flop orientations (S.F. Cotmore and P. Tattersall, unpublished data).

This observation suggests that any genome-length single strandwhich is excised from replicative-form concatemers will ultimatelybe packaged, and that it is only the failure to excise certaintermini efficiently which precludes their encapsidation. Hence,it predicts that there are no specific sequences in MVM or LuIIIDNA that allow the selection of negative- versus positive-sensestrands for packaging. In this respect the heterocruciform modelpotentially provides a mechanistic perspective from which toview the earlier kinetic hairpin transfer model of parvovirusDNA replication (5). This mathematical model was based purelyon physical measurements of terminal sequence heterogeneityand the ratio of plus to minus strands packaged into virionsby different parvoviruses, and it suggested that all such heterogeneitycould be explained simply by the relative rates of hairpin transferat the right and left ends of each genome. Although this modelhad many interesting implications, it was difficult to reconcilewith our knowledge of viral biochemistry, since we have longbeen aware that left-end termini of viruses such as MVM andLuIII are excised by resolution of the dimer intermediate andnot simply by hairpin transfer reactions. The data presentedin this paper may therefore provide a conceptual basis for reconcilingthe earlier mathematical model with specific biochemical mechanisms.

ACKNOWLEDGMENTS

This work was supported by U.S. Public Health Service grantnumber AI26109 from the National Institute of Allergy and InfectiousDiseases.

FOOTNOTES

* Corresponding author. Mailing address: Department of Laboratory Medicine, Yale University School of Medicine, 333 Cedar St., New Haven, CT 06510. Phone: (203) 785-4586. Fax: (203) 688-7340. E-mail: peter.tattersall{at}yale.edu.

REFERENCES

Top