business-driven enterprise information management

Non-canonical Interactions in a Kissing LoopComplex: The Dimerization Initiation Site of HIV-1Genomic RNA

Jean-Christophe Paillart, Eric Westhof, Chantal EhresmannBernard Ehresmann and Roland Marquet*

UniteÂ Propre de Recherche duCNRS no 9002, Institut deBiologie MoleÂculaire etCellulaire, 15 rue R. Descartes67084, Strasbourg CedexFrance

Retroviruses encapsidate two molecules of genomic RNA that are non-covalently linked close to their 50 ends in a region called the dimer link-age structure (DLS). The dimerization initiation site (DIS) of humanimmunode®ciency virus type 1 (HIV-1) constitutes the essential part ofthe DLS in vitro and is crucial for ef®cient HIV-1 replication in cell cul-ture. We previously identi®ed the DIS as a hairpin structure, locatedupstream of the major splice donor site, that contains in the loop a six-nucleotide self-complementary sequence preceded and followed by twoand one purines, respectively. Two RNA monomers form a kissing loopcomplex via intermolecular interactions of the six nucleotide self-comp-lementary sequence. Here, we introduced compensatory mutations in theself-complementary sequence and/or a mutation in the ¯anking purines.We determined the kinetics of dimerization, the thermal stabilities andthe apparent equilibrium dissociation constants of wild-type and mutantdimers and used chemical probing to obtain structural information. Ourresults demonstrate the importance of the 50-¯anking purine and of thetwo central bases of the self-complementary sequence in the dimerizationprocess. The experimental data are rationalized by triple interactionsbetween these residues in the deep groove of the kissing helix and are in-corporated into a three-dimensional model of the kissing loop dimer. Inaddition, chemical probing and molecular modeling favor the existenceof a non-canonical interaction between the conserved adenine residues atthe ®rst and last positions in the DIS loop. Furthermore, we show thatdestabilization of the kissing loop complex at the DIS can be compen-sated by interactions involving sequences located downstream of thesplice donor site of the HIV-1 genomic RNA.

# 1997 Academic Press Limited

Keywords: RNA-RNA interactions; retrovirus; AIDS; structure; tertiaryinteractions*Corresponding author

Introduction

A unique feature of retroviruses is that they en-capsidate a dimer of positive genomic RNA. Earlyelectron microscopy studies showed that the twomolecules are held together in a region called the

dimer linkage structure (DLS) located close to their50 end (Bender & Davidson, 1976; Kung et al., 1976;Bender et al., 1978; Murti et al., 1981). The dimericnature of the retroviral genome was proposed tobe linked to key steps of the retroviral life cycle,such as reverse transcription and recombination(Panganiban & Fiore, 1988; Hu & Temin, 1990;Temin, 1991, 1993; Stuhlmann & Berg, 1992;Mikkelsen et al., 1996), translation of the gag gene(Bieth et al., 1990; Baudin et al., 1993), and selectivepackaging of the genomic RNA (for reviews, seeLinial & Miller, 1990; Gorelick et al., 1991; Rein,1994; Darlix et al., 1995). However, efforts to de-monstrate the biological role of the DLS were ham-pered by the fact that the sequences involved in

Present address: J.-C. Paillart, Dana Farber CancerInstitute, Boston, MA 02115, USA.

Abbreviations used: DIS, dimerization initiation site;HIV-1, human immunode®ciency virus type 1; DLS,dimer linkage structure; SD, splice donor; PBS, primerbinding site; CMCT, 1-cyclohexyl 3-(morpholinoethyl)carbodiimide metho-p-toluene sulfonate; DEPC,diethylpyrocarbonate; DMS, dimethylsulfate.

J. Mol. Biol. (1997) 270, 36±49

0022±2836/97/260036±14 $25.00/0/mb971096 # 1997 Academic Press Limited

the intermolecular contacts between the two geno-mic RNA molecules were not identi®ed.

Recently we identi®ed the dimerization initiationsite (DIS) of the genomic RNA of the MAL isolateof the human immunode®ciency virus type 1(HIV-1) by combining chemical interference exper-iments and site-directed mutagenesis in vitro(Paillart et al., 1994; Skripkin et al., 1994). The DIS

consists of a short sequence located between theprimer binding site (PBS) and the major splicedonor site (SD) that folds into a stem-loop struc-ture (Figure 1A and B). In vitro, the DIS constitutesthe essential part of the DLS of HIV-1 MAL (Pail-lart et al., 1994, 1996a; Skripkin et al., 1994) andHIV-1 LAI (Laughrea & JetteÂ, 1994, 1996a; Muriauxet al., 1995). When mutations are introduced in the

Figure 1. Wild-type and mutant dimerization initiation site (DIS) of HIV-1 RNA. A, Schematic representation of the50 end of HIV-1 genomic RNA. R, repeat sequence; U5, unique sequence at the 50 end of the genome; PBS, primerbinding site, DIS, dimerization initiation site; SD, splice donor site; AUG, start codon of gag translation. B, Analysisof the DIS loop sequence of 29 HIV-1 and two related SIVCPZ isolates shown on the stem-loop structure of the DIS.The isolates included in the analysis were: U455 (accession number M62320), IBNG (L39106), SF2 (K02007), LAI(K02013), HXB2R (K03455), NY5 (K03346), NL43 (M19921), MN (M17449), JH31 (M21137), JRCSF (M38429), JRFL(M74978), OYI (M26727), CAM1 (D10112), CDC41 (M13136), HAN (U43141), P896 (M96155), D31 (U43096), RF(M17451), YU2 (M93258, BCSG3C (L02317), YU10 (M93259), 320A12 (U34603), 320A21 (U34604), WEAU160 (U21135),ELI (K03454), Z2Z6 (M22639), NDK (M27323), ANT70 (M31171), MAL (K03453), CPZGAB (X52154) and CPZANT(U42720) (Myers et al., 1995). C, Schematic representation of the kissing loop complex formed by the HIV-1 MALRNA. D, Variant RNAs used in this study. The sequence corresponding to the wild-type DIS loop is written in whitein the black box. Deletions are indicated by broken lines, substitutions by bold underlined letters, and the insertion/substitution by underlined italic.

HIV-1 RNA Dimerization 37

DIS of an infectious clone, the replication rate ofthe viral clones is reduced (Berkhout & vanWamel, 1996; Haddrick et al., 1996; Paillart et al.,1996b) and a signi®cant fraction of the genomicRNA is monomeric (Haddrick et al., 1996). Indeed,substitutions or deletions in the DIS stem-loop re-duce infectivity of the mutant virions up to 1000-fold (Paillart et al., 1996b). Mutations of the DISaffect at least two steps of the HIV-1 life cycle:encapsidation of the genomic RNA (Berkhout &van Wamel, 1996; McBride & Panganiban, 1996;Paillart et al., 1996b) and synthesis of the proviralDNA (Paillart et al., 1996b). No effect of these mu-tations is detected on the synthesis of the Gag-Polpolyprotein in vivo (Berkhout & van Wamel, 1996;Paillart et al., 1996b).

The dimerization mechanism of fragments ofHIV-1 genomic RNA containing parts or the entire50 untranslated region has been extensively investi-gated in vitro. It was initially proposed that dimeri-zation of the HIV-1 genome could involveformation of purine quartets (Marquet et al., 1991;Awang & Sen, 1993; Sundquist & Heaphy, 1993;Weiss et al., 1993). However, it is now well estab-lished that this mechanism takes place only whenusing RNA fragments lacking the 50 end of theviral genome, and is irrelevant to the dimerizationof HIV-1 genomic RNA in vivo (Fu et al., 1994;Marquet et al., 1994, 1996; Skripkin et al., 1994;Haddrick et al., 1996). Identi®cation of the DIS asthe essential element of the DLS allowed us to pro-pose another dimerization mechanism, the kissingloop complex (Skripkin et al., 1994), that has re-ceived strong experimental support (Laughrea &JetteÂ, 1994, 1996a,b; Paillart et al., 1994, 1996a;Skripkin et al., 1994, 1996; Muriaux et al., 1995,1996; Clever et al., 1996; Haddrick et al., 1996). Wehave proposed that dimerization of HIV-1 RNA ismediated by a symmetric intermolecular inter-action involving the six nucleotide self-complemen-tary sequence located in the DIS apical loop(Figure 1C; Skripkin et al., 1994). Indeed, any mu-tation that perturbs the self-complementarity ofthis sequence prevents dimerization (Paillart et al.,1994; Skripkin et al., 1994), while introduction ofcompensatory mutations restores the process (Pail-lart et al., 1994). Furthermore, two RNAs withpoint mutations in the six nucleotide self-comp-lementary sequence that prevent formation ofhomodimers are able to heterodimerize when themutated sequences are complementary to eachother (Paillart et al., 1994; 1996a). Contrary to thekissing loop complexes formed by antisense regu-latory RNAs with their sense RNA targets, the kis-sing loop complex formed by the HIV-1 DIS is notconverted into an extended duplex in vitro (Paillartet al., 1996a), unless non-physiological conditions,such as high temperature, low ionic strength or ab-sence of divalent cations, are used (Laughrea &JetteÂ, 1996b; Muriaux et al., 1996).

A self-complementary sequence in the DIS loopis an absolute requirement for dimerization ofHIV-1 RNA, but it is not a suf®cient condition.

Sequence comparison of more than 30 HIV-1 andrelated simian immunode®ciency virus (SIVCPZ)isolates reveals that only a tiny fraction of the poss-ible six nucleotide self-complementary sequencesare found naturally (Figure 1B; Berkhout, 1996;Paillart et al., 1996c). Futhermore, the threenucleotides ¯anking the self-complementary se-quence in the loop are highly conserved: these aremost often adenine and almost always purines(Figure 1B). The importance of the purine residuesis supported by recent experimental evidence. In-deed, antisense DIS RNA, which contains the sameself-complementary motif in the loop as the senseRNA, does not dimerize (Clever et al., 1996;Skripkin et al., 1996). Likewise, a sense DNA ver-sion of the DIS is also unable to dimerize (Skripkinet al., 1996), suggesting that the intrinsic structureof the DIS stem-loop is a key determinant of thedimerization process.

Here we focus on the role of the purine residuessurrounding the self-complementary sequence ofthe DIS loop on the dimerization process. SeveralRNA variants were constructed by site-directedmutagenesis. We determined the kinetics of dimer-ization of the RNAs, and the apparent dissociationconstants and the thermal stabilities of the dimers.These data, combined with detailed chemical prob-ing information, were integrated into a three-di-mensional model of the RNA dimer. The threepurine residues are directly involved in the for-mation of the kissing loop dimer via the formationof non-canonical interactions. The DIS representsthe ®rst example of a kissing loop complex invol-ving such interactions.

Results

We used site-directed mutagenesis to study theimportance of the purine residues ¯anking the self-complementary sequence of the DIS in the dimeri-zation process. We also introduced mutations inthe self-complementary sequence to test the poss-ible formation of base triples with these purines.For each mutant RNA, we analyzed the dimeriza-tion yield and, when signi®cant dimerization wasobserved, the dimerization kinetics and the thermalstability of the RNA dimer. First, all mutationswere introduced in an RNA corresponding tonucleotides (nt) 1 to 615 of HIV-1 MAL. In additionto the DIS, it contains sequences, named 30 DLS,located downstream of the major splice donor site(SD, nucleotides 305-306) which are not requiredfor dimerization, but participate in the stabilizationof the dimer of HIV-1 MAL RNA (Marquet et al.,1994; Paillart et al., 1994; Skripkin et al., 1994). Inorder to assess the relative roles of the DIS and 30DLS, several of the mutations were also introducedin an RNA corresponding to nucleotides 1 to 311of HIV-1 MAL.

38 HIV-1 RNA Dimerization

Kinetics of dimerization and stability of the1-615 RNA dimers

Sequence analysis of 29 HIV-1 and 2 SIVCPZisolates indicates that the three purine residues ofthe DIS loop are conserved in all but four isolates(Figure 1B). In two of them (HIV-1 OYI andSIVCPZ GAB), one adenine located 50 to the self-complementary sequence is deleted, while in twoothers (HIV-1 ANT70 and SIVCPZ ANT)nucleotides 272 and 280 form a Watson±Crickbase-pair extending the DIS stem (Figure 1B). TwoRNA mutants were constructed in order to test theeffects of deletion and of base-pairing of these pur-ines, which are normally unpaired in the kissingloop complex (Figure 1C). In the ®rst one,nucleotides 272, 273 and 280 were deleted(DIS�R), while in the second one, nucleotide 280was replaced by the dinucleotide CU, which iscomplementary to nucleotides 272-273 (DISCU;Figure 1D). The absence of unpaired purines in thekissing loop strongly affects the kinetics of dimeri-zation of both RNAs and the stability of the corre-sponding dimers (Figure 2 and Table 1).Dimerization of these RNAs is three- to ®vefoldslower than that of wild-type HIV-1 MAL RNA.The thermal stabilities of the DIS�R and DISCURNA dimers are decreased by 11�C compared tothe wild-type dimer and their apparent Kd valuesare sixfold higher than that of HIV-1 MAL RNA.

The HIV-1 MAL isolate we used has a G residueat position 273. However, most HIV-1 isolatessequenced to date have an A at that position(Figure 1B). Careful sequence analysis indicatesthat all isolates with a G273 form a U275 �A0278 (0 indi-cates a nucleotide on the second molecule in theRNA dimer) or U275 �G0278 base-pair in the kissingloop complex, while almost all isolates having anA273 form a C275 �G0278 base-pair (Figure 1B). Thesecovariations raise the possibility that G �U-A orA �C-G base-triples may be involved in the for-mation of the kissing loop complex. Three mutantRNAs were synthesized to test the existence of abase-triple between nucleotides 273, 275 and 2780(or 273, 2750 and 278).

In the mutant DISComp, U275 and A278 werereplaced by C and G, respectively (Figure 1D).Thus, it contains two compensatory mutations thatmaintain the self-complementarity of the DIS loopand maintain its ability to dimerize (Paillart et al.,1994). No signi®cant difference is found betweendimerization of wild-type RNA and DISCompRNA on the basis of tm, Kd and ka (Figure 2, Table 1and Paillart et al. (1994)). Since the kissing loopcomplex formed with wild-type RNA is main-tained by four G �C and two A �U base-pairs andthe dimer of DISComp RNA is stabilized by sixG �C base-pairs (Figure 1C), it is rather unexpectedthat both dimers have the same stability. Thissuggests that the stabilization provided by the G �Cversus A �U base-pairs is compensated by the desta-bilization of other interactions. Interestingly, whenG273 of RNA DISComp is replaced by an A (RNA

DISCompA, Figure 1D), the thermal stability of thedimer increases by 8�C (Figure 2A and Table 1).The increased stability of this dimer correlates withan increase of the dimerization rate constant (ka,Figure 2B and Table 1). However, stabilization ofthe RNA dimer by A273 is independent of thenature of the 275 �2780 base-pair. Indeed, when G273

is replaced by an A in the context of the wild-typeself-complementary sequence (RNA DISA;Figure 1D), the dimerization rate is increased by2.4-fold and the melting temperature of the RNAdimer is increased by 7�C. Taken together, our re-sults with wild-type RNA and mutant RNAs DIS-Comp, DISCompA and DISA indicate that thepurine at position 273 (R273) is involved in non-ca-nonical interactions. However, if R273 forms a base-triple with nucleotides 275 and 2780, as suggested

Figure 2. Physico-chemical parameters of wild-type andmutants HIV-1 1-615 RNA dimerization. A, Thermalstability of the dimers. After dimer formation at 30�Cfor 30 minutes, the temperature was gradually increasedby 7�C steps, incubated ®ve minutes at the desired tem-perature and an aliquot was loaded on 1.2% agarosegel. To facilitate comparison, the dimerization yields arenormalized according to the dimerization yield at 30�C.B, Analysis of the dimerization kinetics. The experimen-tal dimerization data of wild-type and mutated RNAswere introduced into equation 1/Md � 1/Md

� �2 � kdim � t corresponding to a second-order kinetics(Marquet et al., 1994).


by sequence comparison, one would have expecteda destabilization of the dimer of RNA DISA com-pared to that of the wild-type RNA.

Preliminary molecular modeling indicated thatR273 is close to the central base-pairs of the kissingcomplex. Consequently, we constructed two RNAvariants with an inversion of the two central nts ofthe DIS self-complementary sequence (CG276 in-stead of GC276) and either a G (RNA DISInv) or anA (RNA DISInvA) at position 273, in order to testthe possible interactions between these nucleotides(Figure 1C and D). The association rate constantsand the equilibrium dissociation constants of thesevariant RNAs are similar to those of the wild-typeRNA (Figure 2B and Table 1). Surprisingly, the di-mers formed by the DISInv and DISInvA RNAsare very resistant to thermal denaturation: theyonly start to dissociate at 70�C. This result is unex-pected since on the basis of base-pair stacking en-ergies, inversion of the two central base-pairsshould destabilize the kissing loop complex (by�2.8 kcal/mol). In addition, different results wereobtained when these mutations were introduced inthe context of RNA 1-311 (see below).

Stability of the 1-311 RNA dimers

Since the dimers of RNAs encompassingnucleotides 1 to 615 of HIV-1 MAL are stabilizedby interactions involving the 30 DLS, in addition tothe main contact at the DIS (Marquet et al., 1994;Paillart et al., 1994; Skripkin et al., 1994), some ofthe DIS mutations were studied in the context ofRNAs spanning nucleotides 1 to 311, which do notcontain the 30 DLS. Due to the lower stability ofthe corresponding dimer, only 50 to 60% of the

wild-type 1-311 RNA is dimeric under our stan-dard conditions (Table 2), compared to >80% withwild-type 1-615 RNA (data not shown; Marquetet al., 1994; Paillart et al., 1994). Accordingly, themelting temperature of the wild-type 1-311 RNA is12�C lower than that of the 1-615 RNA dimer(Table 2 and Paillart et al., 1994), and its Kd value isincreased by ®vefold (Table 2). As already noticedwith the larger RNAs, the DISComp 1-311 RNAdimer has the same stability as the wild-type 1-311RNA dimer, while the dimers of DISA and DIS-CompA 1-311 RNAs, which have the same stab-ility, are substantially more stable than the wild-type RNA dimer (Table 2). Thus, it appears thatthe 30 DLS stabilizes the dimers of wild-type, DIS-Comp, DISA and DISCompA RNAs to a similarextent. These results also con®rm that the presenceof A273 stabilizes the RNA dimer independently ofthe nature of nucleotides 275 and 278.

The results obtained with DISInv and DISInvA1-311 RNAs were more surprising. At the standardconcentration of 400 nM, these RNAs yielded lessthan 10% dimer, suggesting that these dimers arequite unstable (Table 2). Given the low dimeriza-tion yield, it was not possible to determine theirthermal stabilities. However, determination of theirKd values con®rmed this hypothesis. Indeed, wecould only obtain a lower estimate of 2.5 mM forthe Kd values of DISInv and DISInvA 1-311 RNAs.At this concentration, the dimerization yields ofDISInv and DISInvA 1-311 RNAs were 38 and28%, respectively (Table 2). Thus, inversion of thecentral nucleotides of the self-complementary se-quence increases the stability of the dimer in thecontext of 1-615 RNA but dramatically decreases itin the context of 1-311 RNA. This result demon-

Table 1. Physico-chemical parameters of the dimers of wild-type and mutant 1-615 RNAs

1-615 RNA Loop sequence tm (�C)a Kd (nM) ka (mMÿ1 minÿ1)

Wild-type AGGUGCACA 53 � 2 21 � 10 3.75 � 0.5DIS�R ± ±GUGCAC± 42 � 2 120 � 20 1.25 � 0.25DISCU AGGUGCACCU 42 � 2 130 � 20 0.75 � 0.1DISComp AGGCGCGCA 54 � 2 10 � 5 3.25 � 0.5DISCompA AAGCGCGCA 62 � 2 4 � 2 8 � 1DISA AAGUGCACA 60 � 2 15 � 7 9 � 1DISInv AGGUCGACA >70 13 � 5 2.0 � 0.25DISInvA AAGUCGACA >70 20 � 10 3.5 � 0.5

Thermal stabilities (tm), equilibrium dissociation constants (Kd) and kinetics of dimerization (ka) are indicated.a All data were obtained at an RNA concentration of 400 nM.

Table 2. Dimerization yield and stability of the dimers of wild-type and mutant 1-311 RNAs

1-311 RNA Loop sequence Dimer (%)a tm (�C)a Kd (nM)

Wild-type AGGUGCACA 50±60 41 � 2 110 � 20DISA AAGUGCACA 60±70 50 � 2 60 � 15DISComp AGGCGCGCA 50±60 41 � 2 80 � 15DISCompA AAGCGCGCA 60±70 50 � 2 40 � 15DISInv AGGUCGACA <10 (38{) nd >2500DISInvA AAGUCGACA <10 (28{) nd >2500

a All data were obtained at an RNA concentration of 400 nM, except those indicated by { which were obtained 2.5 mM.nd, not determined.


strates that subtle modi®cations of the kissing loopcomplex can strongly in¯uence stabilization by the30 DLS. It is also interesting that, in the context of1-311 RNAs, substitution of G273 by an A destabi-lizes the dimer when the self-complementary se-quence is GUCGAC (according to the dimerizationyields of DISInv and DISInvA 1-311 RNAs at highconcentration (Table 2)), while it has a stabilizingeffect when the self-complementary sequence isGUGCAC or GCGCGC. This observation furthersupports the involvement of R273 in a non-canoni-cal interaction.

Structural analysis of wild-type and variant1-615 RNA dimers by chemical probing

In order to gain insight into the structure of thekissing loop complex, we untertook a detailedchemical probing analysis of the dimers formed bythe wild-type and mutant RNAs. Since the dimeri-zation yields are much higher with 1-615 RNAsthan with 1-311 RNAs, probing experiments wereconducted on these. The Watson±Crick positionsA-N1 and C-N3 were tested with dimethylsulfate(DMS) and positions G-N1 and U-N3 were testedwith 1-cyclohexyl 3-(morpholinoethyl)carbodiimidemetho-p-toluene sulfonate (CMCT). The N7 pos-itions of A and G were tested with diethylpyrocar-bonate (DEPC) and DMS followed by treatmentwith aniline. The RNAs were also subjected tolead-induced cleavage, which is very sensitive tosubtle structural alterations (Gornicki et al., 1989).Examples of probing experiments are shown inFigure 3 and the results are summarized inFigure 4.

With a few exceptions discussed in detail below,the probing data of the wild-type and mutantRNAs are very similar, indicating rather limitedstructural differences between the kissing loop di-mers. As expected from the kissing loop secondarystructure model (Figure 1C), the Watson±Crickpositions of the six nucleotide self-complementarysequence are non-reactive, with the exception ofU275, which displays a low reactivity towardsCMCT in wild-type, DISInv and DISInvA RNAs(Figure 4). The stability of the dimers formed bywild-type, DISComp and DISInv 1-311 RNAs com-pared to their counterparts with an A273 also corre-lates with the reactivity of the G274-N7 (Figures 3and 4). For each RNA pair, the one that forms theless stable dimer displays the higher reactivity ofG274-N7 towards DMS.

As expected from the kissing loop model, all A-N7 positions in the self-complementary sequenceare unreactive. Indeed, this position is usually notmodi®ed by DEPC when the adenine is involvedin an RNA double helix (Ehresmann et al., 1987).Except for one case, the G-N7 positions in the self-complementary sequence are not reactive, whichmay re¯ect the existence of tertiary interactions in-volving these nucleotides. Interestingly, the G277-N7 position of DISInvA RNA, which forms the

least stable kissing loop interaction in the absenceof the 30 DLS, is strongly modi®ed by DMS, whileposition N7 of the central G residues (G276 or G277)is unreactive in all other RNAs (Figures 3 and 4).This result supports the view that in all RNAs,except DISInvA, the two central base-pairs areinvolved in tertiary interactions. Since these inter-actions exist in DISInv but not in DISInvA, theymost likely involve R273.

The three purines surrounding the self-comp-lementary sequence display a characteristic patternof reactivities. The most striking feature concernsA280, which displays a hyper-reactivity of its N1position, while its N7 position is totally unreactive(Figure 4). In the wild-type 1-311 RNA, the reactiv-ity level of A280-N1 decreases from hyper-reactiveto strongly reactive, relative to the 1-615 RNA, re-¯ecting a structural in¯uence of the 30 DLS onthe local structure of the kissing loop complex(data not shown). The A272-N1 position is moder-ately reactive, while the reactivity of A272-N7 var-ies from moderate (DISComp, DISCompA,DISInv) to strong (wild-type, DISA, DISInvA).Reactivity of R273 is either weak or moderate,compatible with the existence of tertiary inter-actions involving this nucleotide. When thenucleotide at position 273 is a G (i.e. in wild-type, DISComp and DISInv) its N7 position isweakly reactive or unreactive. In contrast, whennucleotide 273 is an A (in DISA, DISCompA andDISInvA), its N7 position is moderately reactive(Figure 4). In all cases, the R273-N1 position iseither unreactive or weakly reactive (or not deter-mined in DISInv). This result suggests that A273

and G273 may be involved in slightly differentinteractions.

In general, two kinds of RNA lead-induced clea-vages can be observed: weak unspeci®c clevagesgenerally observed in ¯exible unpaired regions,due to free lead ions, and strong speci®c cleavagesdue to ions coordinated to the RNA (Gornicki et al.,1989). In the case of wild-type 1-615 RNA, onlyone strong lead-induced cleavage is observed. It islocalized in the DIS loop, 50 to G273 (Figures 3Cand 4). The lead-induced cleavage is also observedin DISComp 1-615 RNA, although it is weaker.Quanti®cation of the gel of Figure 3C using a FujiBAS 2000 BioImager reveals a threefold reductionof the cleavage intensity. Lead-induced cleavage ofthe wild-type and DISComp loops requires a G atposition 273, since no cleavage is observed in DISAand DISCompA RNAs. Inversion of the two cen-tral nucleotides of the DIS self-complementarysequence totally inhibits lead-induced cleavage,as shown with mutants DISInv and DISInvA(Figures 3C and 4). As dimerization is a dynamicand reversible process (Paillart et al., 1996a), onecannot totally exclude the possibility that themonomeric form of the RNAs, but not the dimer,is cleaved by lead. In order to test this possibility,we used DISC275 1-615 RNA, which has a pointmutation in the DIS self-complementary sequenceand hence is unable to dimerize (Paillart et al.,


Figure 3. Chemical probing of the wild-type and mutant 1-615 RNAs. Chemical reactions were carried out asdescribed in Materials and Methods and the modi®ed positions were identi®ed by primer extension. A, Modi®cationof RNA with DMS, followed by aniline-induced chain scission allowing probing of positions A-N1, C-N3 and G-N7.Lane 1, control without DMS ; lane 2, DMS treatment for four minutes; lane 3, DMS treatment for eight minutes.A, U, C, and G lanes correspond to a dideoxy sequencing reaction run in parallel on the denaturing polyacrylamidegel. B, Lead-induced cleavage. Lane 1, incubation without Pb(OAc)2; lane 1, incubation with 4 mM Pb(OAc)2 for ®veminutes; lane 3, incubation with 10 mM Pb(OAc)2 for ®ve minutes. A sequencing reaction generated with the sameprimer was run in parallel (lanes U, A, C, G). The position of the cleaved nucleotide is shown by an asterisk.


1994). Indeed, this monomeric RNA is not cleavedby lead (Figure 3C), strongly suggesting that onlythe dimeric form of wild-type and DISComp RNAsare cleaved.

Three-dimensional molecular modeling of thekissing loop complex

We used the biochemical and chemical probingdata to construct three-dimensional models of thewild-type and DISCompA kissing loop complexes,which correspond to the sequences of the naturalHIV-1 isolates (Figure 5). A ribbon representationof the wild-type DIS kissing loop complex isshown in Figure 5A. Only three bases are not in-

volved in Watson±Crick pairing: A272, G273 andA280. All bases in the intramolecular and intermole-cular (kissing) stems consist of Watson±Crick base-pairs. The three stems (intra, kissing and intra0)have parallel helical axes with stacking continuityand are topologically similar to the anticodon-anticodon dimer of crystallized tRNAAsp (Westhofet al., 1985). Nucleotides A272 and A280 (as well asA0272 and A0280) form a trans base-pair between theWatson±Crick positions of A272 (or A0272) and theHoogsteen sites of A280 (or A0280) (Figure 5B). Thischoice is favored both by the natural right-handedrotation of RNA and by the chemical probing data(see Figure 4, A272-N7 and A280-N1 highly reactive,with A280-N7 not reactive and A272-N1 midly reac-

Figure 4. Summary of the chemical probing data of wild-type and mutant 1-615 RNAs. Results are displayed on thesecondary structure model of the kissing loop complex. The color code is explained in the insert. n.d., not determined.The absence of a symbol corresponds to unreactive nucleotides. Mutated nucleotides are shown in red.


tive). It is also in keeping with the phylogeneticanalysis, which indicates that this A �A base-pairmay be replaced by a Watson±Crick base-pair insome exceptional cases (Figure 1B, isolates HIV-1ANT70 and SIVCPZ ANT).

The linking guanine residue G273 (G0273) forms abase triple in the deep groove with the purine ofthe kissing base-pair C277 �G0276 (C0277 �G276), with H-bonds linking the N1 and N2 sites of G273 (G0273)and the acceptor Hoogsteen sites of G0276 (G276).Thus, the triple interaction links the two loops ofthe wild-type complex, and guanine G273 and G0273

are stacked on each other. A sharp turn occurs inthe DIS loop between nucleotides 273 and 274.Essentially the same three-dimensional model ap-plies for the DISComp kissing loop complex(not shown).

It is interesting to note that the path made by re-sidues A272-G273 (A0272-G0273) is very similar to thatfollowed by the U8-R9 stretch which precedes theD-helix in tRNA structures (Saenger, 1984). IntRNAs, residue U8 leaving the acceptor stem formsa trans Hoogsteen pair with A14 in the dihydrouri-dine loop, while R9 forms a triple interaction withthe purine of the third base-pair of the dihydrouri-dine helix, Y12-R23, before entering the D-helix. Inthe kissing loop complex, the triple interactionsoccur at the fourth base-pair of the kissing stem.The polynucleotide backbone G271-G273 (G0271-G0273)forms a tight loop, analogous to the so-called P10-loop formed by residues eight to ten of tRNAswhich constitutes a strong metal ion binding site.Interestingly, lead cleavage is observed betweenA272 and G273 in wild-type RNA, and to a lesserextent in DISComp RNA (Figure 4).

In DISA and DISCompA mutants, theA273 �C277-G0276 (A0273.C0277-G276) triple interactionsare not favorable and the A273-N6 might interactpreferentially with the Hoogsteen sites (or the 50-phosphate) of G276 (G0276). Thus, unlike G273 andG0273 in wild-type and DISComp RNAs, A273 andA0273 make intramonomer interactions (Figure 5C),with A273 stacked below A0273 (Figure 5C), whileG273 is stacked above G0273 (Figure 5C) (G273 andA273 belong to the monomers at the top ofFigures 5B and 5C). With respect to wild-typeRNA, the turn in the loop of DISA and DISCompARNAs is looser and the binding of metal ion isdisfavored, as indicated by the absence of leadcleavage.

Similarly, the R273.C277-G0276 (R0273 C0277-G276) tripleinteractions are also destabilized in DISInv andDIS InvA mutants. Indeed, in order for R273 to

Figure 5. Three-dimensional models of the wild-typeand CompA DIS kissing loop complexes. A, Stereo viewof the ribbon three-dimensional structure of the wild-type kissing loop complex. Nucleotides A272 (A0272), G273

(G0273) and A280 (A0280) are represented. B, Detailed stereoviews of the wild-type kissing loop complex, and C, ofthe DISCompA kissing loop complex. Residues A272,A280, A0272 and A0280 are drawn in white and R273 andR0273 are represented in yellow.


form a base triple with the mutated guanine atposition G0277, it would collide with its symmetricalpartner, R0273. At the same time, mutants containinga G at position 277 cannot form appropriate hydro-gen bonds with G273. Thus, we suggest instead astacking between R273 and R0273 and a possible H-bond between G273-O6 (G0273-O6) and the aminogroup of C276 (C0276).

Discussion

The DIS is the essential part of the DLS in vitro(Laughrea & JetteÂ, 1994; Paillart et al., 1994, 1996a;Skripkin et al., 1994; Muriaux et al., 1995), and iscrucial for ef®cient HIV-1 replication in cell culture(Berkhout & van Wamel, 1996; Paillart et al.,1996b). We have shown that, in vitro, two RNAmonomers form a kissing loop complex via inter-molecular interactions of the six nucleotide self-complementary sequence located in the DIS apicalloop (Paillart et al., 1994, 1996a; Skripkin et al.,1994). In addition, recent experimental evidencesuggests that the three purines ¯anking the self-complementary sequence might be essential fordimer formation (Skripkin et al., 1996, Clever et al.,1996).

Our results with 1-615 DIS�R and DISCU RNAsshow that the unpaired purines contribute to thestability of the dimer and the dimerization kinetics.The observation that these mutants form less stabledimers than the wild-type RNA is consistent withthe fact that the kissing-loop complex in HIV-1RNA dimer is not converted into an extended du-plex (Paillart et al., 1996a). Indeed, the extendedduplex of the mutant RNAs would form a longregular double helix, while that of the wild-typeRNA would be interrupted by two internal loops(Paillart et al., 1996a). The importance of the purineresidues was also demonstrated by Clever et al.(1996). Using an in vitro assay performed at 25�Cwith 200 nucleotide-long RNAs containing the DIS,but not the 30 DLS, these authors showed that del-etion or substitution of the purine residues by U re-sidues abolished dimerization. The mutant RNAslacking the purines in the DIS loop are reminiscentof the stem-loop structure involved in the dimeri-zation of murine leukemia viruses (MLV; Tounektiet al., 1992; Girard et al., 1995; Berkhout, 1996;Paillart et al., 1996c). Interestingly, in vitro dimeri-zation of MLV RNA is much slower than that ofHIV-1 RNA and has a higher temperature opti-mum (Roy et al., 1990; Girard et al., 1995).

Of the 64 possible six nucleotide self-comp-lementary sequences, only two are found in theDIS loop of natural HIV-1 isolates (Figure 1B). Thisobservation suggests that only a tiny fraction ofthe possible self-complementary sequences adoptsa loop-structure compatible with dimerizationor/and that other functional constraints on the DISsequence are present during viral replication. Ourresults with DISInv and DISInvA clearly indicatethat not all self-complementary sequences are able

to promote RNA dimerization (at least in the ab-sence of the 30 DLS). The importance of the centraldinucleotide has also been pointed out by Cleveret al. (1996) who found that two DIS loops contain-ing GCGUGC and GCACGC trans-complementarysequences are unable to form heterodimers. Simi-larly, two RNAs in which the DIS self-complemen-tary sequences have been replaced by GGGGGGand CCCCCC did not dimerize (Clever et al.,1996). Evidence for additional constraints on theDIS sequence not linked to the dimerization pro-cess also exists. Our data indicate that the stabilityof the dimers and the dimerization kinetics of DIS-Comp and DISA RNAs are indistinguishable fromthose of wild-type and DISCompA RNAs, respect-ively. However, only the latter sequences arefound in the natural HIV-1 isolates. In addition,deletion of the entire DIS stem-loop affects replica-tion of HIV-1 more dramatically than substitutionsin the DIS loop (Paillart et al., 1996b). Thus, the DISstem-loop might bind viral or cellullar factors.

Phylogenetic analysis of the natural DIS se-quences (Figure 1B) suggested ®rst that R273 mightform a base triple with the Y275-R0278 (or Y0275-R278)base-pair. However, the thermal stability of the di-mers formed by wild-type, DISComp, DISA andDISCompA RNAs does not support that hypoth-esis. Instead, as suggested by three-dimensionalmolecular modeling of the wild-type kissing loopcomplex, G273 forms an intermolecular interactionwith G0276-C277, while in the DISCompA kissingloop complex, A273 would form an intramolecularinteraction with G276-C0277. Indeed, the dramatic ef-fect of base inversion of the central dinucleotide ofthe self-complementary sequence supports the sug-gestion that R273 interacts with the G276-C0277 (orG0276-C277) base-pair. In both complexes, the tertiaryinteractions of R273 with the central base-pairs arestabilized by base-stacking of R273 and R0273. Thedif®culty of identifying base triples from phylo-genic data is a well-known problem. It is linked, inpart, to the fact that natural mutations in base tri-ples create structural changes that require compen-satory changes in neighboring base-pairs tomaintain the triple-helix conformation (Gautheretet al., 1995). In the kissing loop complex, the phylo-genetic covariation may re¯ect an additional stabil-ization between R273 and the base-pair Y275 �R0278

preceding the one with which it forms a triple. Thenature of this stabilization is presently dif®cult topin-point but could be stacking and/or bifurcatedH-bonds between R273 and Y275-R0278. A situationsimilar to that seen in the loop kissing complexwas found recently in the binding site of ribosomalprotein S8 on the 16 S ribosomal RNA (Moine et al.,1997). SELEX experiments showed that this bind-ing site contains a base triple, while phylogeneticanalysis suggested covariations of the third basewith a base-pair adjacent to the one highlighted bySELEX.

The intramolecular interaction between A273 andG276 that we propose in the kissing loop complexesformed by DISA and DISCompA RNAs might pre-


exist in the corresponding monomers. This inter-action would facilitate dimerization, and could ac-count for the faster dimerization kinetics and thehigher thermal stabilities observed with theseRNAs.

In the DIS kissing loop complex, chemical prob-ing experiments as well as molecular modelingstrongly suggest the existence of an A �A base pairformed between the ®rst and the last nucleotide inthe DIS loop. The same interaction apparently ex-ists in the dimer formed by the wild-type and DIS-CompA RNAs.

To our knowledge, this work presents the ®rstexperimental evidence of non-canonical base-pairsand base triple in a kissing loop complex. Indeed,in the kissing loop structures formed by naturalantisense RNA with their sense RNA target, suchas RNA I/RNA II and copA/copT, which regulatethe copy number of plasmids ColE1 and R1, re-spectively, the interacting loops are fully comp-lementary, leaving no unpaired nucleotides(Simons, 1993). Complete base-pairing of the twoloops in the RNA I/RNA II complex has been de-monstrated by NMR experiments (Marino et al.,1995). Regarding the demonstrated or proposedDIS of other retroviruses, some have fully self-complementary loops, while others, includingHIV-2 and SIV (Simian immunode®ciency virus)contain unpaired purines in their loops thatmight be involved in non-canonical interactions(Berkhout, 1996; Paillart et al., 1996c). However, nostructural data concerning these RNA dimers areavailable.

Materials and Methods

Construction of the HIV-1 DIS variants

Plasmid construction and cleavage by restriction en-zymes were conducted according to published pro-cedures (Sambrook et al., 1989). Plasmid pJCB containsnucleotides 1 to 615 of the HIV-1 genome (MAL isolate;�1 corresponds to the ®rst nucleotide of the genomicRNA) under the control of the promoter of RNA poly-merase from phage T7, followed by a PvuII restrictionsite (Paillart et al., 1994). Plasmids pDISA, pDISCompA,pDISInv, pDISInvA, pDIS�R and pDISCU were ob-tained by inverse PCR on pJCB essentially as described(Paillart et al., 1996d). Plasmid pDISComp has been de-scribed (Paillart et al., 1994). Positions of the mutationsare indicated in Figure 1D. The PCR products werephosphorylated, puri®ed on 1% (w/v) agarose gels, li-gated and used to transform JM 109 Escherichia coli cells.

RNA synthesis and purification

Plasmids containing the wild-type and mutated HIV-1MAL sequences were linearized with PvuII and tran-scribed by bacteriophage T7 RNA polymerase. In vitrotranscription was for two hours, under previously de-scribed conditions (Marquet et al., 1991). Transcriptionwas followed by treatment for 30 minutes with RNase-free DNase I (Appligene), phenol extraction and ethanolprecipitation. RNAs were puri®ed by FPLC (Pharmacia)on a Bio-sil TSK 250 (BioRad) column, in a buffer con-

taining 200 mM sodium acetate (pH 6.5), 1% (v/v)methanol. Transcripts were concentrated in a Centricon50 (Amicon) unit, precipitated with ethanol and dis-solved in water prior to use. RNAs were labeled eitherduring transcription using [a-32P]ATP (Amersham)(37.5 mM of ATP, 50 mCi/mg of DNA template) or by 30end ligation using [50-32P]pCp (Amersham) (50 mCi/2 mgRNA) and RNA ligase from phage T4.

In vitro RNA dimerization

In a typical experiment, wild-type or mutant un-labeled RNAs were diluted in Milli-Q water (Millipore)at 400 nM ®nal concentration with the correspondingradioactive RNA (3 to 5 nCi, 0.01 to 0.04 mg), heated fortwo minutes at 90�C, and chilled on ice for two minutes.Dimerization was initiated by addition of 2 ml of ®vefoldconcentrated dimerization buffer (®nal concentration:50 mM sodium cacodylate (pH 7.5), 300 mM KCl, 5 mMMgCl2). The dimerization yield for wild-type and mutantRNAs was compared after a 30 minutes incubation at37�C. Samples were analyzed on 1.2% (w/v) agarosegels. Electrophoresis were performed in 45 mM Tris-bo-rate (pH 8.3), 0.1 mM MgCl2. Gels were ®xed for tenminutes in 10% (v/v) trichloroacetic acid and dried for40 minutes under vacuum at room temperature. Dimersand monomers were quanti®ed using a BAS 2000 BIO-Imager (Fuji) as described (Marquet et al., 1994; Paillartet al., 1994).

The fraction of RNA dimer is de®ned as the ratio ofthe dimer to the total RNA species (Marquet et al., 1994).Thermal stability of the dimers (tm), dimerization kinetics(ka) and equilibrium dissociation constants (Kd) were de-termined as described (Marquet et al., 1994; Paillart et al.,1994, 1996a).

Chemical probing of the 1-615 RNA dimers

Chemical probing was performed after the dimeriza-tion procedure on 1.6 mg of wild-type and variant 1-615RNAs in the presence of 2 mg of E. coli total tRNA. RNAswere modi®ed with DMS (Fluka), CMCT (Merck), orDEPC (Sigma) in high salt buffer (300 mM) as described(Baudin et al., 1993; Skripkin et al., 1994). For DMS modi-®cation, the reaction was carried out for four and eightminutes at 25�C in dimerization buffer in the presence of0.8 ml of DMS freshly diluted 1/20 (v/v) in ethanol.Chain scission at methylated G-N7 was performed byaniline treatment as described (Peattie & Herr, 1981).CMCT modi®cations were performed at 25�C in 20 ml ofD2 buffer (100 mM sodium borate, (pH 8.0), 300 mMKCl and 5 mM MgCl2) in the presence of 5 ml of CMCT(42 mg/ml in water) for 15, 30 and 45 minutes. Modi®-cation with DEPC was done at 25�C with 2.5 ml of DEPCfor ten and 20 minutes. For each reaction, a control with-out the chemical reagent was treated in parallel. All reac-tions were stopped by ethanol precipitation. Modi®edbases were detected by extension with AMV reversetranscriptase (Life Sciences) of a 50-[g32P]ATP-labeled pri-mer complementary to residues 400 to 415 of the HIV-1RNA as described (Baudin et al., 1993).

Wild-type and mutant RNAs were also submitted tolead-induced cleavage (Gornicki et al., 1989). In a typicalexperiment, 1.6 mg of RNA was allowed to dimerize inan acetate buffer consisting of 50 mM Hepes (pH 7.0),5 mM magnesium acetate, 300 mM potassium acetate.Pb2� concentrations varied from 1 to 100 mM and the in-cubation reaction was for ®ve minutes at 37�C. Optimal


cleavage was observed between 2 and 10 mM Pb(OAc)2.Reactions were stopped by addition of EDTA to a ®nalconcentration of 50 mM; RNA was precipitated andlead-induced cleavages were revealed as describedabove.

Three-dimensional molecular modeling

Programs used for the construction of the RNA modelare described by Westhof, (1993). The assembly and con-nection of the secondary elements into a three-dimen-sional fold was realized on a graphic system PS300 fromEvans and Sutherland using FRODO (Jones, 1978),adapted for the PS300 (P¯ugrath et al., 1983). Finally, thegenerated models were subjected to restrained least-squares re®nement (Konnert & Hendrickson, 1980) withthe programs NUCLIN and NUCLSQ (Westhof et al.,1985) in order to ensure geometry and stereochemistrywith allowed distances between interacting atoms and toavoid steric con¯icts. The color views were generatedwith the program DRAWNA (Massire et al., 1994).

Acknowledgements

We thank E. Skripkin for preliminary lead cleavageexperiments, D. Mignot for technical assistance, J.S.Lodmell for critical reading of the manuscript and B.Masquida for help with color drawings. This work wassupported by grants from the Agence Nationale deRecherche sur le SIDA (ANRS). J.-C.P. was a fellowof the ANRS.

References

Awang, G. & Sen, D. (1993). Mode of dimerization ofHIV-1 genomic RNA. Biochemistry, 32, 11453±11457.

Baudin, F., Marquet, R., Isel, C., Darlix, J. L.,Ehresmann, B. & Ehresmann, C. (1993). The func-tional sites in the 50 region of human immunode®-ciency virus type 1 RNA form de®ned structuraldomains. J. Mol. Biol. 229, 382±397.

Bender, W. & Davidson, N. (1976). Mapping of poly(A)sequences in the electron microscope reveals unu-sual structure of type C oncornavirus RNAmolecules. Cell, 7, 595±607.

Bender, W., Chien, Y. H., Chattopadkyay, S., Vogt, P. K.,Gardner, M. B. & Davidson, N. (1978). High-mol-ecular weight RNAs of AKR, NZB and wild mouseviruses and avian reticuloendotheliosis virus allhave similar dimer structures. J. Virol. 25, 888±896.

Berkhout, B. (1996). Structure and function of thehuman immunode®ciency virus leader RNA. Prog.Nucl. Acids Res. Mol. Biol. 54, 1±34.

Berkhout, B. & van Wamel, J. L. B. (1996). Role of theDIS hairpin in replication of human immunode®-ciency virus type 1. J. Virol. 70, 6723±6732.

Bieth, E., Gabus, C. & Darlix, J. L. (1990). A study of thedimer formation of Rous sarcoma virus RNA andof its effect on viral protein synthesis in vitro. Nucl.Acids Res. 18, 119±127.

Clever, J. L., Wong, M. L. & Parslow, T. G. (1996).Requirements for kissing-loop-mediated dimeriza-tion of human immunode®ciency virus RNA.J. Virol. 70, 5902±5908.

Darlix, J. L., Lapadat-Tapolski, M., de Rocquigny, H. &Roques, B. P. (1995). First glimpses at structure-

function relationships of the nucleocapsid protein ofretroviruses. J. Mol. Biol. 254, 523±537.

Ehresmann, C., Baudin, F., Mougel, M., Romby, P., Ebel,J. P. & Ehresmann, B. (1987). Probing the structureof RNAs in solution. Nucl. Acids Res. 22, 9109±9128.

Fu, W., Gorelick, R. J. & Rein, A. (1994). Characteriz-ation of human immunode®ciency virus type 1dimeric RNA from wild-type and protease-defectivevirions. J. Virol. 68, 5013±5018.

Gautheret, D., Damberger, S. H. & Gutell, R. R. (1995).Identi®cation of base-triples in RNA using com-parative sequence analysis. J. Mol. Biol. 248, 27±43.

Girard, P.-M., Bonnet-MathonieÁre, B., Muriaux, D. &Paoletti, J. (1995). A short autocomplementarysequence in the 50 leader region is responsible fordimerization of MoMuLV genomic RNA. Biochemis-try, 34, 9785±9794.

Gorelick, R. J., Nigida, S. M., Jr, Arthur, L. O.,Henderson, L. E. & Rein, A. (1991). Roles of nucleo-capsid cysteine arrays in retroviral assembly andreplication: possible mechanisms in RNAencapsidation. In Advances in Molecular Biology andTargeted Treatment for AIDS (Kumar, A., ed.),pp. 257±272, Plenum Press, New York.

Gornicki, P., Baudin, F., Romby, P., Wiewiorowski, M.,Krzyzosiak, J. W., Ebel, J.-P., Ehresmann, C. &Ehresmann, B. (1989). Use of lead(II) to probe thestructure of large RNAs. Conformation of the 30terminal domain of E. coli 16 S rRNA and its invol-vement in building the tRNA binding sites. J. Biomol.Struct. Dynam. 5, 971±984.

Haddrick, M., Lear, A. L., Cann, A. J. & Heaphy, S.(1996). Evidence that a kissing loop structure facili-tates genomic RNA dimerisation in HIV-1. J. Mol.Biol. 259, 58±68.

Hu, W. S. & Temin, H. M. (1990). Retroviral recombina-tion and reverse transcription. Science, 250, 1227±1233.

Jones, T. A. (1978). A graphic model building and re®ne-ment system for macromolecules. J. Appl. Crystallog.11, 268±272.

Konnert, J. H. & Hendrickson, W. A. (1980). Arestrained-parameter thermal-factor re®nementprocedure. Acta Crystallog. 36, 344±350.

Kung, H. J., Hu, S., Bender, W., Bailey, J. M., Davidson,N., Nicolson, M. O. & McAllister, R. M. (1976). RD-114, baboon, and wolly monkey viral RNAs com-pared in size and structure. Cell, 7, 609±620.

Laughrea, M. & JetteÂ, L. (1994). A 19-nucleotidesequence upstream of the 50 major splice donor ispart of the dimerization domain of human immuno-de®ciency virus 1 genomic RNA. Biochemistry, 33,13464±13474.

Laughrea, M. & JetteÂ, L. (1996a). HIV-1 genome dimeri-zation: formation kinetics and thermal stability ofdimeric HIV-1 LAI RNAs are not improved by the1-232 and 296-790 regions ¯anking the kissing-loopdomain. Biochemistry, 35, 9366±9374.

Laughrea, M. & JetteÂ, L. (1996b). Kissing-loop model ofHIV-1 genome dimerization: HIV-1 RNA canassume alternative dimeric forms, and all sequencesupstream or downstream of hairpin 248-271 are dis-pensable for dimer formation. Biochemistry, 35,1589±1598.

Linial, M. L. & Miller, A. D. (1990). Retroviral RNApackaging: sequence requirements and implications.Curr. Top. Microbiol. Immunol. 157, 125±152.

Marino, J. P., Gregorian, R. S. J., Csankovszki, G. &Crothers, D. M. (1995). Bent helix formation


between RNA hairpins with complementary loops.Science, 268, 1448±1454.

Marquet, R., Baudin, F., Gabus, C., Darlix, J. L., Mougel,M., Ehresmann, C. & Ehresmann, B. (1991). Dimeri-zation of human immunode®ciency virus (type 1)RNA: stimulation by cations and possiblemechanism. Nucl. Acids Res. 19, 2349±2357.

Marquet, R., Paillart, J.-C., Skripkin, E., Ehresmann, C. &Ehresmann, B. (1994). Dimerization of humanimmunode®ciency virus type 1 RNA involvessequences located upstream of the splice donor site.Nucl. Acids Res. 22, 145±151.

Marquet, R., Paillart, J.-C., Skripkin, E., Ehresmann, C. &Ehresmann, B. (1996). Localization of the dimeriza-tion initiation site of HIV-1 genomic RNA andmechanism of dimerization. In Biological Structureand Dynamics (Sarma, R. H. & Sarma, M. H., eds),vol. 2, pp. 61±72, Adenine Press, Albany, NY.

Massire, C., Gaspin, C. & Westhof, E. (1994). DRAWNA:a program for drawing schematic views of nucleicacids. J. Mol. Graph. 12, 201±207.

McBride, M. S. & Panganiban, A. T. (1996). The humanimmunode®ciency virus type 1 encapsidation site isa multipartite RNA element composed of functionalhairpin structures. J. Virol. 70, 2963±2973.

Mikkelsen, J. G., Lund, A. H., Kristensen, K. D., Duch,M., Sorensen, M. S., Jorgensen, P. & Pedersen, F. S.(1996). A preferred region for recombinationalpatch repair in the 50 untranslated region of primerbinding site-impaired murine leukemia virusvectors. J. Virol. 70, 1439±1447.

Moine, H., Cachia, C., Westhof, E., Ehresmann, B. &Ehresmann, C. (1997). The RNA binding site of S8ribosomal protein of Escherichia coli: SELEX and hy-droxyl radical probing studies. RNA, 3, 255±268.

Muriaux, D., Girard, P.-M., Bonnet-MathonieÁre, B. &Paoletti, J. (1995). Dimerization of HIV-1Lai RNA atlow ionic strength. An autocomplementarysequence in the 50 leader region is evidenced by anantisense oligonucleotide. J. Biol. Chem. 270, 8209±8216.

Muriaux, D., FosseÂ, P. & Paoletti, J. (1996). A kissingcomplex together with a stable dimer is involved inthe HIV-1Lai RNA dimerization process in vitro. Bio-chemistry, 35, 5075±5082.

Murti, K. G., Bondurant, M. & Tereba, A. (1981). Sec-ondary structural features in the 70 S RNAs ofMoloney murine leukemia and Rous sarcomaviruses as observed by electron microscopy. J. Virol.37, 411±419.

Myers, G., Korber, B., Hahn, B. H., Jeang, K.-T., Mellors,J. W., McCutchan, F. E., Henderson, L. E. &Pavlakis, G. N. (1995). A compilation and analysisof nucleic acid and amino acid sequences. In HumanRetroviruses and AIDS, Los Alamos National Labora-tory, Los Alamos, NM.

Paillart, J.-C., Marquet, R., Skripkin, E., Ehresmann, B. &Ehresmann, C. (1994). Mutational analysis of thebipartite dimer linkage structure of human immu-node®ciency virus type 1 genomic RNA. J. Biol.Chem. 269, 27486±27493.

Paillart, J.-C., Skripkin, E., Ehresmann, B., Ehresmann,C. & Marquet, R. (1996a). A loop-loop ``kissing''complex is the essential part of the dimer linkage ofgenomic HIV-1 RNA. Proc. Natl Acad. Sci. USA, 93,5572±5577.

Paillart, J.-C., Berthoux, L., Ottmann, M., Darlix, J.-L.,Marquet, R., Ehresmann, B. & Ehresmann, C.(1996b). A dual role of the putative RNA dimeriza-

tion initiation site of human immunode®ciencyvirus type 1 in genomic RNA packaging and pro-viral DNA synthesis. J. Virol. 70, 8348±8354.

Paillart, J.-C., Marquet, R., Skripkin, E., Ehresmann, C. &Ehresmann, B. (1996c). Dimerization of retroviralgenomic RNAs: structural and functionalimplications. Biochimie, 78, 639±653.

Paillart, J.-C., Skripkin, E., Ehresmann, B., Ehresmann,C. & Marquet, R. (1996d). The use of chemicalmodi®cation interference and inverse PCR muta-genesis to identify the dimerization initiation site ofHIV-1 genomic RNA. Pharma. Acta Helvetiae, 71,21±28.

Panganiban, A. & Fiore, D. (1988). Ordered interstrandand intrastrand DNA transfer during reversetranscription. Science, 241, 1064±1069.

Peattie, D. A. & Herr, W. (1981). Chemical probing oftRNA-ribosome complex. Proc. Natl Acad. Sci. USA,78, 2273±2277.

P¯ugrath, J. W., Saper, M. A. & Quiocho, F. A. (1983).Molecular modeling with the PS300: a new gener-ation graphics display system. J. Mol. Graph. 1, 53±54.

Rein, A. (1994). Retroviral RNA packaging: a review.Arch. Virol. 9, 513±522.

Roy, C., Tounekti, N., Mougel, M., Darlix, J. L.,Paoletti, C., Ehresmann, C., Ehresmann, B. &Paoletti, J. (1990). An analytical study of thedimerization of in vitro generated RNA of Moloneymurine leukemia virus MoMuLV. Nucl. Acids Res.18, 7287±7292.

Saenger, W. (1984). Principles of Nucleic Acid Structure,Springer-Verlag, New York.

Sambrook, J., Fritsch, E. F. & Maniatis, T. (1989). Molecu-lar Cloning: A Laboratory Manual, Cold Spring Har-bor Laboratory Press, Cold Spring Harbor, NY.

Simons, R. W. (1993). The control of prokaryotic andeukaryotic gene expression by naturally occurringantisense RNA. In Antisense Research and Appli-cations (Crooke, S. T. & Lebleu, B., eds), pp. 97±124,CRC Press, Inc., Boca Raton.

Skripkin, E., Paillart, J. C., Marquet, R., Ehresmann, B. &Ehresmann, C. (1994). Identi®cation of the primarysite of human immunode®ciency virus type 1 RNAdimerization in vitro. Proc. Natl Acad. Sci. USA, 91,4945±4949.

Skripkin, E., Paillart, J.-C., Marquet, R., Blumenfeld, M.,Ehresmann, B. & Ehresmann, C. (1996). Mechanismsof inhibition of in vitro dimerization of HIV-1 RNAby sense and antisense oligonucleotides. J. Biol.Chem. 271, 28812±28817.

Stuhlmann, H. & Berg, P. (1992). Homologous recombi-nation of copackaged retrovirus RNAs duringreverse transcription. J. Virol. 66, 2378±2388.

Sundquist, W. I. & Heaphy, S. (1993). Evidence for inter-strand quadruplex formation in the dimerization ofhuman immunode®ciency virus-1 genomic RNA.Proc. Natl Acad. Sci. USA, 90, 3393±3397.

Temin, H. M. (1991). Sex and recombination inretroviruses. Trends Genet. 7, 71±74.

Temin, H. M. (1993). Retrovirus variation and reversetranscription - abnormal strand transfers result inretrovirus genetic variation. Proc. Natl Acad. Sci.USA, 90, 6900±6903.

Tounekti, N., Mougel, M., Roy, C., Marquet, R., Darlix,J. L., Paoletti, J., Ehresmann, B. & Ehresmann, C.(1992). Effect of dimerization on the conformationof the encapsidation Psi domain of Moloney murineleukemia virus RNA. J. Mol. Biol. 223, 205±220.


Weiss, S., HaÈusl, G., Famulok, M. & KoÈnig, B. (1993).The multimerization state of retroviral RNA ismodulated by ammonium ions and affects HIV-1full-length cDNA synthesis in vitro. Nucl. Acids Res.21, 4879±4885.

Westhof, E. (1993). Modelling the three-dimensional struc-ture of ribonucleic acids. J. Mol. Struct. 286, 203±210.

Westhof, E., Dumas, P. & Moras, D. (1985). Crystallo-graphic re®nement of yeast aspartic acid transferRNA. J. Mol. Biol. 184, 119±145.

Edited by J. Karn

(Received 17 February 1997; accepted 4 April 1997)


business-driven enterprise information management

Documents