In vitro identification and characterization of an early complex linking HIV-1 genomic RNA recognition and Pr55Gag multimerization.

The minimal protein requirements that drive virus-like particle formation of human immunodeficiency virus type 1 (HIV-1) have been established. The C-terminal domain of capsid (CTD-CA) and nucleocapsid (NC) are the most important domains in a so-called minimal Gag protein (mGag). The CTD is essential for Gag oligomerization. NC is known to bind and encapsidate HIV-1 genomic RNA. The spacer peptide, SP1, located between CA and NC is important for the multimerization process, viral maturation and recognition of HIV-1 genomic RNA by NC. In this study, we show that NC in the context of an mGag protein binds HIV-1 genomic RNA with almost 10-fold higher affinity. The protein region encompassing the 11th alpha-helix of CA and the proposed alpha-helix in the CA/SP1 boundary region play important roles in this increased binding capacity. Furthermore, sequences downstream from stem loop 4 of the HIV-1 genomic RNA are also important for this RNA-protein interaction. In gel shift assays using purified mGag and a model RNA spanning the region from +223 to +506 of HIV-1 genomic RNA, we have identified an early complex (EC) formation between 2 proteins and 1 RNA molecule. This EC was not present in experiments performed with a mutant mGag protein, which contains a CTD dimerization mutation (M318A). These data suggest that the dimerization interface of the CTD plays an important role in EC formation, and, as a consequence, in RNA-protein association and multimerization. We propose a model for the RNA-protein interaction, based on previous results and those presented in this study.

genomic RNA are structured. Several models have been proposed for the folding of this region, with alternative conformers, but the overall conformation of this 5Ј-leader region (5Ј-LR) remains to be fully defined (45). Several structures within the 5Ј-LR are well characterized, i.e. the Tat transactivation response element (TAR), poly(A), the primer binding site (PBS) (46 -48). The HIV-1 packaging signal () is multipartite containing four stem loops (SL1-4) (39,49). SL1 promotes the formation of the RNA dimer through a palindromic sequence and, accordingly, is also known as the dimerization initiation site (DIS) (50 -52). NC binds to the four stem loops of , but it has been shown that it binds much more strongly to SL2 and SL3 than to SL1 or SL4 (53)(54)(55). The high resolution structure of the complex formed between SL2/NC (56) and SL3/NC (57) has been resolved by NMR. SL4, with a weaker affinity for NC, has been proposed to stabilize the structure formed by SL2 and SL3 (58). Various authors have proposed long distance interactions between 5Ј-regions (between the poly(A) region and the PBS) with 3Ј-counterpart sequences (generally downstream of SL4) thought to promote a more stable conformer (59,60).
Taking into account the participation of SP1 in each of the multimerization process (26,30,61), viral maturation (24,25), and recognition of the HIV-1 ⌿ site by NC (31,32), we decided to study relevant RNA-protein interactions in vitro using a minimal Pr55 Gag (mGag) protein. This recombinant protein includes amino acids 2-7 of MA (the myristyl anchor), the CTD of CA, SP1, and NC (28). Using this protein, we have analyzed the binding and multimerization abilities of native NC compared with NC in the context of the precursor (mGag), using probes containing different extensions of the 5Ј-leader region (5Ј-LR). Here, we report that NC in the context of Pr55 Gag has a much higher affinity for the 5Ј-LR and that this higher affinity correlates with the formation/occurrence of an early complex (EC) that involves two proteins (mGag) that are bound to 5Ј-LR RNA. This complex was not present in experiments performed with mature NC alone. Dissection of this EC reveals that amino acids at the N terminus of NC are involved in the protein-RNA association and, furthermore, that N terminus extended versions of NC also acquire the ability to multimerize. Moreover, we have identified sequences downstream of SL4 that are involved in the recognition of the 5Ј-LR by mGag, as they increase the affinity of this interaction and enhance the formation of the EC. These results suggest that recognition of the 5Ј-LR and Gag dimerization (and eventually multimerization) are linked. Based on previous and our own data, we propose a model for the recognition of HIV-1 genomic RNA.

Expression Vectors and Purification of Recombinant
Proteins-Unless otherwise stated, all expression vectors were constructed by PCR amplification of the encompassing DNA sequence of relevant fragments of Pr55 Gag from an HBX2 plasmid clone. These fragments were cloned into the expression vector Topo 100-D (Invitrogen) (Fig. 1). This vector has a T7 expression cassette with a 36-amino acid N-terminal extension containing a 6-histidine motif, a FLAG motif, and an enterokinase site. Site-directed mutagenesis was performed to produce M318A from the wild-type (wt) mGag vector. For cloning procedures Top10 cells (Invitrogen) were used, while Bl21 (DE3) cells (Novagen) were employed for protein expression.
The different recombinant proteins were purified using Ni-NTA resin (Qiagen) in a batch purification protocol under denaturing conditions. Briefly, transformed bacteria were grown and induced for 3 h at 37°C. Cells were pelleted by centrifugation and lysed with buffer B (6 M GdnHCl, 0.1 M NaH 2 PO 4 , 0.01 M Tris-HCl, pH 8.0, 20 mM ␤-mercaptoethanol). Complete lysis was achieved by stirring for 1 h (h) at room temperature and the lysate was then clarified by centrifugation at 10,000 ϫ g for 30 min. The supernatant was mixed with previously equilibrated Ni-NTA resin in buffer B, and binding was performed for 90 min with gentle shaking. The lysate resin was loaded in empty 5-ml columns, allowed to drain, and washed twice with buffer C (same as B but pH 6.3). Finally, the protein was eluted in buffer E (same as B but pH 4.5). The buffer of the proteins was exchanged using centrifugal filter devices (Amicon) to 7 M urea, 50 mM Tris-HCl, pH 8.5 for further purification of the protein, and removal of nucleic acids by anion exchange chromatography using Q-Sepharose resin (Amersham Biosciences) in a batch procedure. Proteins were collected from the unbound material and redissolved to 0.5-2 mg/ml. Refolding was accomplished through protein dialysis against 0.5 M Tris-HCl, pH 8.0, 50 mM NaCl, 10 M ZnCl 2 , 5 mM/0.5 mM reduced/oxidized glutathione, 0.01% CHAPS at 4°C for 3 h. Two buffer exchanges were completed, and this was followed by a final dialysis step against storage buffer (same as dialysis but 50 mM Tris-HCl, pH 8.0 and no CHAPS). Removal of unfolded proteins was accomplished by ultracentrifugation at 100,000 ϫ g for 1 h.
Synthesis and Labeling of RNA-To produce probes for gel shift analysis, various DNA fragments encoding portions of the 5Ј-LR were amplified by PCR (Fig. 2). For sense probes, the T7 promoter was included at the 5Ј-end of the forward primer, whereas the same promoter was similarly added into the reverse primer in the case of antisense probes. Radiolabeled RNAs were transcribed using the Megashortscript kit (Ambion) following recommendations of the manufacturer, but in the presence of 96 M UTP and 50 Ci of [␣ 32 P]UTP (3000 Ci/mmol, ICN) for 2 h at 37°C. The DNA template was removed by the addition of 2 units of DNase I and further incubation at 37°C for 15 min. Radiolabeled probes were purified on denaturing polyacrylamide gels. Finally, the probe was dissolved in 20 mM Tris-HCl, 100 mM NaCl, and renatured by heating at 70°C for 5 min and cooling slowly to room temperature.
Gel Shift Assay-Gel shift assays were done as previously described with minor modifications (62). Binding reactions were performed in 10 l containing 0.2 nM of the radiolabeled probe, 10 mM HEPES pH 7.5, 100 mM KCl, 1 mM MgCl 2 , 0.5 mM EDTA, 1 mM dithiothreitol, 20 ng/l Poly I⅐Poly C, 7.5% Ficoll and 0 -3 M of recombinant protein. After incubation for 15 min on ice, the reaction mixtures were electrophoresed through nondenaturing 6% polyacrylamide gels in 0.5ϫ Tris borate/ EDTA at 120 V for 14 -16 hs at 4°C. Gels were dried before exposure to Bio-Max Kodak films at Ϫ80°C.
Cross-linking Experiments-Proteins were dialyzed for 2 h against a storage buffer containing 50 mM HEPES pH 7.5, and ultracentrifuged at 100,000 ϫ g for 1 h. Binding reactions were performed as for gel shifts and then irradiated by UV (UV Crosslinker 2400, Stratagene). Complexes were resolved by electrophoresis on denaturing 5% polyacrylamide gels, which were dried and exposed.
Filter Binding Assay-The filter binding protocol was derived from previous work (Ref. 63 and references therein). Similar binding reactions as for gel shifts were performed in a volume of 20 l using 1 nM radiolabeled probe. After incubation for 15 min on ice, the samples were filtered through prewet 0.45 M nitrocellulose disks (Millipore). Disks were washed twice with storage buffer, dried, and the radioactivity retained by the filter was quantified by liquid scintillation counting. Data were corrected for background binding and are expressed as percentage of counts; 100% represents the saturation of the probe within each experiment.
Protein Ultracentrifugation-Preliminary binding experiments were done after centrifugation of the recombinant proteins at 10,000 ϫ g for 15 min. However, no interpretable data were obtained in either filter binding assays or in gel shift assays (data not shown). In contrast, ultracentrifugation of the protein at 100,000 ϫ g for 1 h at 4°C yielded interpretable data by both techniques. Analysis of the ultracentrifuged protein through gel filtration chromatography columns (Superdex 200, 30/10 HR; Amersham Biosciences) revealed a narrow peak at an elution volume corresponding to ϳ22 kDa, the expected size of the protein. Centrifugation at 10,000 ϫ g revealed the same peak, but a "shoulder" thought to be misfolded protein was also present. All subsequent experiments were performed after ultracentrifugation of the protein at 100,000 ϫ g for 1 h.

Differential Binding of NC and mGag Binding to HIV-1
Genomic RNA Probes-Our laboratory has generated in vivo evidence that SP1 is involved in the recognition of HIV-1 genomic RNA (32); therefore, we decided to also study this protein-RNA interaction in vitro. For this purpose, we designed and purified a minimal Gag protein (mGag), which possesses all the domains including NC, that are needed to generate VLPs (Fig. 1). The first gel shift assays were performed with these proteins, and a nucleic acid probe containing the fragment from nt ϩ223 to ϩ506 (1/6) (ϩ1 is capping site; strain HXB2) (Fig. 2), based on evidence that these sequences are important in RNA packaging (39,64).
As reported by others, a small percentage of the probe migrated very slowly, and this band was thought to represent RNA dimers (Fig. 3) (39, 51, 64 -67). In support of this idea, the use of probes lacking SL1 did not yield this band (see also Fig. 7, below and compare panel C with the other panels in this figure). We also observed multiple bands in our gel shift assays as reported by others (39,53,64).
The interactions of proteins with the probe differed in regard to both affinity and potential for multimerization. Gel shift assays performed in the presence of 50 mM NaCl, 20 ng/l of Poly I⅐Poly C and mGag revealed a 4 -6-fold stronger affinity for the 1/6 probe than that observed with NC (Fig. 3, compare lines 8 and 17). This was not expected, since it had been assumed that NC and Pr55 Gag have similar binding affinity for HIV-1 genomic RNA (39, 64).
Our gel shift assays also showed that mGag has a higher ability to multimerize even at low protein concentration, while NC binds to the probe in a more gradual and limited fashion (Fig. 3).
In order to further characterize the different affinities and the complexes observed, we studied this protein-RNA interaction in greater detail at the protein and RNA levels.
mGag Binds to the 5Ј-LR as a Dimer While NC Does So as a Monomer-In order to determine the identity of the different bands in the gel shift assays, the RNA-protein complexes were cross-linked and resolved in denaturing gels. As expected, NC bound to the probe as a monomer (Fig. 4B). At the lowest concentration of NC, a 1:1 RNA-protein association was observed. As the protein concentration was increased, a second and a third shifted band were seen, with molecular masses corresponding to 2 and 3 proteins bound to each probe, respectively. In contrast, in the same type of experiments performed with mGag, the first shifted band corresponded to the presence of two proteins that were bound to the RNA probe (probe, 90 kDa, mGag, 20 kDa) (Fig. 4A). It is important to note that this is not a RNA dimer, since it would have migrated at 180 kDa. A second and a third shifted band were also present, with molecular masses corresponding to the binding of four and six proteins to one molecule of RNA, respectively. These experiments suggest that NC binds to the RNA probe as a monomer while mGag does so as a dimer.
We then reanalyzed our gel shift assay results with this in mind and noted the absence of any intermediate 1:1 RNAprotein association. This, together with the stronger affinity of mGag than NC for the RNA probe suggested that an EC might have formed.
To examine this hypothesis, we asked whether an EC might be observed in the presence of M318A, a mutation known to disrupt the CTD dimerization interface. We used site-directed mutagenesis to generate M318A mGag and showed that presence of the EC was completely voided by this mutation, as was the pattern of multimerization (Fig. 5). These experiments confirm that wt mGag binds to the 5Ј-LR as a dimer, and that this association requires the dimerization interface of the CTD.
Binding Properties of N-terminal Extensions of NC to 5Ј-LR RNA Probes-To analyze which fragments were involved in the strongest binding of the protein to the RNA, we produced different clones expressing N-terminal-extended versions of NC. This led to the generation of SP1-NC, the complete putative CA-SP1 ␣-helix-NC (354-NC), and the 9th, 10th, and 11th ␣-helices of the CTD of CA-NC (308-NC, 327-NC, and 339-NC, respectively) (Fig. 1). Due to the pattern of multimerization, it was difficult to study binding affinities by gel shift. For this reason, the binding affinities of these proteins for HIV-1 genomic RNA probes were studied by filter binding assays under various conditions as described under "Materials and Methods." We first analyzed the binding properties of wt mGag, M318A mGag, SP1-NC, and NC to the 1/6 probe. As observed with the gel shift assays, NC showed a remarkably higher dissociation constant than mGag. The difference between proteins was 7.5fold in the presence of competitor and high salt concentration In the absence of competitor, M318A showed a K d 40 -60% higher than that observed for wt mGag (Fig. 6B). However, when a nonspecific RNA was added to the reaction (i.e. Poly I⅐Poly C), the K d observed almost doubled. These results suggest the possibility that the dimerization interface of wt mGag plays an important role in specific binding to the 5Ј-LR.
In contrast, gel shift analyses using SP1-NC and the 1/6 probe showed a different binding pattern than that observed for NC. SP1-NC protein showed evidence of more multimerization capabilities than NC. This is strongly supported by the fact that higher concentrations of SP1-NC resulted in large aggregates that could not enter the gel, which was never the case for NC. These results suggest that SP1 provides an interface that is required for protein-protein interactions.
When we included the other intermediate proteins in the analysis, we found that the complete putative ␣-helix CA-SP1 (354-NC) increased affinity by 20 -25%, while the 11th ␣-helix of CA (339-NC) had ϳ100% higher affinity than that observed for NC (Fig. 6A) (K d -M-; NC: 0.75, 354-NC: 0.57, 327-NC: 0.33). In the absence of a nonspecific RNA, the affinity of 339-NC was almost the same as that observed for M318A mGag, while in the presence of Poly I⅐Poly C, the K d of 339-NC was higher than that observed for M318A mGag. This suggests that 339-NC possesses all of the domains necessary for nucleic acid binding, and that the hydrophobic core of the CTD is important for binding when a RNA competitor is present, possibly because of a conformational requirement. SP1-NC does not increase this affinity whereas 354-NC does. This might be explained by the presence of the complete putative ␣-helix in 354-NC. These results, together with the gel shift SP1-NC data, suggest that the COOH half of the putative ␣-helix interacts with the same motif within the protein dimer of the EC, whereas the NH 2 half of the ␣-helix interacts with viral RNA.
We then analyzed the behavior of 339-NC in regard to salt concentration. We found that low salt concentrations resulted in an almost doubling of affinity of 339-NC than were attained at high salt concentrations (Fig. 6E). This suggests that the hydrophobic nature of the amino acids in this region might play an important role in RNA binding.
On the other hand, 327-NC and 308-NC possessed an affinity similar to or lower than NC. We have no clear explanation for this result, but it is possible that the presence of these ␣-helices in the absence of the complete CTD core might lead to a conformational disturbance in this region.
Binding Properties of mGag for Different Probes-In order to characterize the role of viral RNA in its interaction with protein, we analyzed both the multimerization and affinity properties of probes encompassing different domains of the 5Ј-LR (Fig. 2). The probe lacking SL1 (i.e. 2/6) had the same ability as 1/6 to generate the EC and to multimerize (Fig. 7, A and C) but, interestingly, the affinity was only about 40% of that observed for 1/6 (Fig. 6C). Another difference between these two probes is the absence of high mass complexes, seen with 1/6 but not with 2/6, suggesting the possibility that NC promotes probe dimerization only at a certain protein concentration.
The presence of EC was much more sensitive to reduction of the probe from the 3Ј-end. M-fold analysis of the nt ϩ 400 -500 region of the HBX2 strain of HIV-1 suggests the possibility of a similar conformation at the analogue position reported for HIV-1 Mal (59). Interestingly, when probes lacking this region were used (i.e. 1/4 or 1/5), the EC was generated but only at higher protein concentrations (Fig. 7, D and E) and, moreover, multimerization ability was dramatically reduced. Probes containing part of this structure (1/6i) restored both EC formation and multimerization ability to a certain extent (Fig. 7B), supporting the concept that this particular structure is involved in proper protein-RNA interactions. Furthermore, the K d for these "incomplete" probes was about 70% of that of 1/6c, while the 1/4 and 1/5 probes showed a marked reduction in affinity (ϳ26%). To rule out the possibility that the size of the probe was the crucial element in these interactions, we performed the same assays with a probe of a similar size but spanning from the PBS to downstream of SL4. This probe, PB/4, possessed an affinity similar to that observed for 2/6 ( Fig. 6C), but EC formation and the multimerization pattern were reduced to that observed for 1/4 and 1/5. As a control, we also verified the affinity of NC for the different probes. In contrast to our findings with mGag, no significant differences were observed, showing that NC alone is incapable of discerning between the probes (Fig. 6D). These results suggest an association between EC formation (by mGag) and the increased binding capabilities of this protein for the probe. Furthermore, it is possible that this interaction might facilitate the initial recognition and specific packaging of the 5Ј-LR of genomic RNA in vivo.

DISCUSSION
Much effort has gone into a better understanding of RNA packaging and the participation of Gag domains in virus assembly and release. We have designed an in vitro assay to study early events in viral protein-RNA interactions and the role of these interactions in viral assembly. Based on previous and our own data, we propose a model for the protein-genomic RNA interaction (Fig. 8).
Our first important finding is that NC, in the context of Pr55 Gag , binds with much higher affinity to the 5ЈLR than does NC alone (Fig. 6, A and B). Note that the former represents the immature form of the protein whereas the latter represents the mature version. This was not expected, since it had been assumed that NC and Pr55 Gag have similar binding affinity for HIV-1 genomic RNA (39,64). Several explanations include our preliminary data using Gag⌬p6 and CA-NC, showing a lower affinity for the 1/6 probe than mGag. Others have employed full-length MA sequences that do not acquire proper conformation in a prokaryotic environment (68,69). In addition, others have compared the binding activities of both proteins with a 178/383 probe comprising SL1, 2, 3, and 4, and we show here that sequences downstream of nucleotide 383 are crucial for the differential binding of mGag (39). Other investigators also used GST-tagged proteins, and a GST tag is larger than that which we employed (GST Х 28 kDa, His Х 3 kDa) (39,64).
Our data suggest that the 11th ␣-helix of CA and the putative CA/SP1 ␣-helix provide the surfaces for the increased affinity of mGag for the HIV-1 genomic RNA probe. As shown, the region encompassing the putative ␣-helix in CA/SP1 (354-NC) increases affinity of NC for HIV-1 genomic RNA, and 339-NC has 100% more affinity than NC alone. Furthermore, 339-NC in the absence of competition, has a similar K d to that of M318A, meaning that this extended NC already possesses all the required domains for RNA binding. In the presence of a nonspecific RNA, 339-NC shows a decrease in affinity in relation to M318A, suggesting a conformational requirement of the hydrophobic core of the CA CTD for the specific recognition of HIV-1 genomic RNA. Finally, the dimeric ability of wt mGag was responsible for the 2-fold difference in affinity in relation to M318A mGag in the presence of a nonspecific RNA.
We have also shown that the stronger affinity of mGag for HIV-1 genomic RNA is dependent on structures downstream of SL2 and SL3. Indeed, the binding data suggest that SL6 is markedly more important than SL4 and SL5 (Fig. 6C). It has already been shown that sequences up to nt ϩ500 are important for HIV-1 packaging. Pr55Gag has been shown to have strong affinity for RNA sequences extending to the AccI restric- tion site (ϩ497) (39,64,70). Moreover, in vivo evidence for a role of genomic RNA in packaging comes from gene transfer experiments that used HIV-1 based vectors. A higher packaging rate was reported for a vector that included the coding region of Gag up to nt ϩ1000 compared with a vector that extended only to nt ϩ400 (42). These data were confirmed in studies showing that a region up to nt ϩ535 contributed to increased packaging of the vector used (41). Our data show that an RNA segment between nt ϩ400 to ϩ500 is involved in interactions between mGag and viral RNA.
Furthermore, we have shown that these stronger RNA-protein associations are related to the formation of a RNA-protein complex, including two molecules of protein and one of RNA (Figs. 4A and 5). The fact that two proteins bind to the 5Ј-LR is not surprising, since the encapsidation signal of the HIV-1 genomic RNA contains at least two strong affinity binding sites for NC (i.e. SL3 & SL2) and the C-terminal domain of CA has a dimer interface. The presence of a M318A mutation, which is known to impair each of CTD dimerization, proper Gag assembly, and viral infectivity (23,71), also abolished formation of the EC. The EC apparently has an additive impact on the affinity of protein for genomic RNA since M318A has approximately half of the binding affinity of wt mGag ([K d ] ϭ K d ⅐n) but was crucial for multimerization ability. Nevertheless, this complex was present at a much lower protein concentration (0.2-0.3 M) than the dimer equilibrium concentration for CTD (18 M) (23) suggesting that this association takes place very early in the assembly process. The EC links genomic RNA recognition, Pr55 Gag dimerization, and multimerization.
We were able to elucidate a new function for the 5Ј-LR other than its role in protein binding and RNA dimerization, i.e. promotion of the formation of the RNA-protein complex termed the EC. Probes that lacked SL1 showed a drastic reduction in protein affinity (Fig. 6C) but did not impair EC conformation (Fig. 7, C and F). In contrast, SL6 (or EC platform -ECP-) was crucial for the formation of this EC (Fig. 7, D-F), and this also had important impact on affinity (Fig. 6C). Probes lacking ECP still yielded complexes at higher protein concentrations but less efficiently (Fig. 7F). Interestingly, however, multimerization was drastically reduced. These results suggest that protein-RNA binding allows a particular conformation or lattice that is important for proper Gag dimerization. Multimerization by dimer blocks has already been reported (68,72). Furthermore, in vivo evidence for such early dimer formation has also been reported in two studies that employed cysteine specific cross-linking (73,74).
How does this protein-RNA interaction take place? According to structural studies, the CA/SP1 region is unstructured in solution (23). Nevertheless, formation of almost every RNAprotein complex that has been characterized involves confor-FIG. 6. Binding affinities studied by filter binding. Filter binding assay was performed as described under "Materials and Methods," which is a modification of the protocol introduced by Schmalzbauer et al. (63). In A and B the inverse of the concentration is plotted against the inverse of the saturation factor. The K d and R 2 are shown between parentheses. A, in the presence of 20 g/l of Poly I⅐C and 250 mM of NaCl; B, no competitor, 250 mM NaCl. In C and D is plotted the ratio between the K d of the different probes with that observed for 1/6c using mGag (C) and NC (D). Similarly, in E is plotted the ratio between the K d of NC (black), 339-NC (dark gray), and mGag (light gray) in relation to that observed for mGag in the presence of low salt concentration (50 mM NaCl, left panel) and high salt concentration (250 mM NaCl, right panel) mational changes in the protein, the RNA or both (75,76). It has also been shown that site-specific binding involves, at least for DNA-protein interactions, coupled changes in 2°, 3°, and/or 4°structure of the protein (77). Therefore, it is possible that the CA/SP1 region folds upon binding to viral RNA as suggested by the ability of the protein to form the EC in the presence of a specific structure in the probe (i.e. SL6-ECP). Computational modeling of this region predicts an ␣-helix conformation. Our results suggest that the 11th ␣-helix of CA and the putative CA/SP1 ␣-helix might bind to HIV-1 genomic RNA (mainly SL6-ECP) in a helix-loop-helix (HLH) fashion. These elements are comprised of a short ␣-helix (11th ␣-helix of CA) and a long ␣-helix (putative CA/SP1), connected by a flexible linker (78), that is usually followed by a highly basic domain involved in the protein: nucleic acid interaction (38,63). In this model, a Gag dimer would interact with the 5Ј-LR as shown in Fig. 8: one NC domain would interact with SL3 and the adjacent HLH domain would interact with one side of the ECP, while the other NC would interact with SL2 and the adjacent HLH domain with the other side of the ECP. The CTD core and its potential for dimerization would provide a conformational advantage for specific HIV-1 genomic RNA recognition and make this complex more stable. Interestingly, HLH elements are usually involved in DNA binding as well as in protein dimer formation. We have shown that the putative CA/SP1 affords both multimerization capabilities and increases affinity. Furthermore, this increase reached its maximum when the short ␣-helix (the complete HLH element) was present. In agreement with the formation of an HLH/RNA complex is the fact that mGag binds to the 5Ј-LR in a dimer fashion, as opposed to what is observed for NC, which binds as a monomer.
The virus could take advantage of this induced fit model in two ways: First, the energetic cost for protein folding would be "paid" only by binding with the 5Ј-LR of HIV-1 genomic RNA, assuring mutual recognition. As a consequence only mutually recognized counterparts could multimerize in the proper way, leading to the formation of infectious virus particles. The hydrophobic sensitivity of the protein-RNA interaction supports this model (77). Furthermore, given that the final conformation of the protein-RNA complexes would depend on sequences from both protein and RNA, the mutual recognition between the parts is likely to be much more flexible than would be the case between rigid surfaces. Finally, according to the evolutionary conservation of the CTD-SP1-NC region among retroviruses, it is possible that the HIV-1 genomic RNA is specifically recognized in a similar way by other viruses.
Others have identified a SL3 variant with greater affinity for NC than wt SL3 (79). They proposed that this SL3 variant is not present inside the virus, since a stronger affinity for the element could eventually interfere with other roles of NC in the viral life cycle. Here, we report that NC in the precursor context has a one order of magnitude increase in ability to recognize a SL2/SL3/ECP element, but this high affinity is lost upon Gag cleavage during maturation, leaving NC free to perform its other functions. It is also possible that the HLH/RNA interactions provide binding support to Pr55 Gag , whereas NC is able to bind and unbind a variety of important elements (tRNA Lys , RT, etc).
It is interesting to note that association forces become weaker in a C-terminal/N-terminal sense (NC-RNA Ϫ0.2 M 3 W317-M318 Ϫ18 M 3 NTD-NTD Ϫ100 M 3 MA-MA -mM range), giving strength to the idea that proper protein-RNA associations help drive the assembly process (14, 23, 80 -84). The process of multimerization and assembly can be seen as a protein folding process in which two alternatives are possible: proper folding and assembly versus aggregation. The M318A mutation strongly reduced HIV-1 infectivity with reduced particle production and defective assembly of capsid and Gag (71). Our data suggest that the early association of two Gag proteins with the 5Ј-LR provides a conformation that is crucial for assembly and proper Gag-Gag interactions. It has recently been reported that RNA-protein interactions do not abrogate viral assembly but play a major role in the stability of the particles formed (85). This is in agreement with our hypothesis of proper protein-protein interactions that occur upon binding to the HIV-1 5Ј-LR. SL6 (ECP) plays an important role in our in vitro system and it is possible that HIV-1 genomic RNA possesses SL6 (ECP) analogue elements throughout its extension that could be bound/recognized by Gag during the assembly process. The ECP might then represent the first of such structures.
In summary, we have shown that two proteins can bind to the 5Ј-LR to form an early complex (EC). This EC is composed of the CTD of CA, SP1 and NC bound as a dimer to an RNA probe stretching from the SL2 to the ECP. Given the high affinity of this association, the low protein concentration at which the EC is present, and its ability to multimerize in a particular pattern, we conclude that this EC has a crucial role in genomic recognition and proper Gag multimerization.