The C-terminal p6 domain of the HIV-1 Pr55^Gag precursor is required for specific binding to the genomic RNA

ABSTRACT

The Pr55^Gag precursor specifically selects the HIV-1 genomic RNA (gRNA) from a large excess of cellular and partially or fully spliced viral RNAs and drives the virus assembly at the plasma membrane. During these processes, the NC domain of Pr55^Gag interacts with the gRNA, while its C-terminal p6 domain binds cellular and viral factors and orchestrates viral particle release. Gag∆p6 is a truncated form of Pr55^Gag lacking the p6 domain usually used as a default surrogate for wild type Pr55^Gag for in vitro analysis. With recent advance in production of full-length recombinant Pr55^Gag, here, we tested whether the p6 domain also contributes to the RNA binding specificity of Pr55^Gag by systematically comparing binding of Pr55^Gag and Gag∆p6 to a panel of viral and cellular RNAs. Unexpectedly, our fluorescence data reveal that the p6 domain is absolutely required for specific binding of Pr55^Gag to the HIV-1 gRNA. Its deletion resulted not only in a decreased affinity for gRNA, but also in an increased affinity for spliced viral and cellular RNAs. In contrast Gag∆p6 displayed a similar affinity for all tested RNAs. Removal of the C-terminal His-tag from Pr55^Gag and Gag∆p6 uniformly increased the Kd values of the RNA-protein complexes by ~ 2.5 fold but did not affect the binding specificities of these proteins. Altogether, our results demonstrate a novel role of the p6 domain in the specificity of Pr55^Gag-RNA interactions, and strongly suggest that the p6 domain contributes to the discrimination of HIV-1 gRNA from cellular and spliced viral mRNAs, which is necessary for its selective encapsidation.

1. Introduction

Retroviral genomes are specifically selected for packaging from an excess of cellular and spliced viral mRNAs (for reviews see [1–4]). The human immunodeficiency virus type 1 (HIV-1) Pr55^Gag precursor drives specific selection of the genomic RNA (gRNA) by interacting with packaging signals (Psi) located within the 5ʹ untranslated region (UTR) and the beginning of the gag gene of gRNA (Fig. 1A) [1,3–6]. The Psi is composed of four stem-loops (SL1 to SL4). SL1 constitutes the gRNA Dimerization Initiation Site (DIS) [7–11]; gRNA dimerization is mediated by a 6 nucleotides palindromic sequence located in the apical loop of SL1 [12]. Interestingly recent findings also showed that SL1 contains the main Pr55^Gag recognition signal [13,14]. SL2 contains the major splice donor (SD) site, and SL3 contributes to gRNA packaging [15–19]. Finally, the folding of SL4 is most likely in equilibrium with the so-called U5-AUG long-range interaction [20–23], which may regulate the last stages of the packaging process (Fig. 1A) [24,25]. Upstream of Psi, highly structured hairpins including the trans-activating responsive element (TAR), the poly(A) hairpin and the Primary Binding Site (PBS) domain have also been proposed to be involved in gRNA packaging [26–28]. Although packaging of gRNA is a highly selective process, spliced viral RNAs [5,29,30], and cellular RNAs [30,31] are also found in viral particles. For instance, 7SL RNA, a component of the signal recognition particle [32,33], tRNAs, U6 spliceosomal RNA [30,34] are enriched in retroviral particles. However, the encapsidation of spliced viral and cellular RNAs follow different mechanisms [30] (for review see [35]): indeed, while SL1 is crucial for packaging of gRNA, it does not seem to be involved in the packaging of spliced viral RNAs [30].

Figure 1.

Domains involved in the interaction between the HIV-1 RNA genome and the Gag polyprotein precursor. (A) Schematic representation of the secondary structure of the 5ʹ-end of HIV-1 genome. The red line delimits the packaging signal (Psi) region comprising SL1-SL4. SL1 contains in its apical loop the Dimerization Initiation Site (DIS). The major Splice Donor (SD) site and gag AUG initiation codon are represented. The secondary structure model of the dimer with the long-range U5-AUG base pairing is represented on the right side. (B) Schematic representation of the full-length Pr55^Gag and GagΔp6 proteins used in this study. Their different domains are indicated: the matrix (MA), capsid (CA), nucleocapsid (NC) flanked by the two spacer peptides SP1 and SP2, and finally the C-terminal p6 domain. Tryptophan residues W and p6 late domains are represented.

The 55-kDa Pr55^Gag precursor is composed of several domains, starting with the matrix (MA) at the N-terminus, capsid (CA), nucleocapsid (NC) flanked by the two small peptides SP1 and SP2, and p6 at its C-terminus (Fig. 1B) (for review see [36]). The MA domain mediates Pr55^Gag membrane binding through a bipartite signal consisting of the N-terminal myristoylated glycine and the highly basic region (HBR) [37,38]. The CA domain drives Pr55^Gag multimerization thus leading to the formation of the structural viral core, and the NC contains two zinc finger motifs, a major determinant for gRNA recognition and packaging [39–43]. Interestingly, the HBR has been shown to promote the interaction of the MA domain with RNA in vitro [43–45], as well as in the cytosol [46]. Finally, the C-terminal p6 domain regulates budding of nascent virions at the plasma membrane [47]. This process involves the interaction of its two late domains PTAP and YPXnL with two host factors associated to the ESCRT (Endosomal Sorting Complex Required for Transport) machinery, namely TSG101 (Tumor Susceptibility Gene 101) [48–50] and ALIX (ALG-2 interaction protein X) [51–54] (Fig. 1B). The LXXLF p6 domain also binds the viral Vpr protein, ensuring its encapsidation [55–57] and it is excluded from the capsid core during viral maturation [58,59] (Fig. 1B).

Production of full-length recombinant Pr55^Gag has been hindered by the instability of the C-terminus of Pr55^Gag [60], hence a non-physiological C-terminus truncated version of this precursor named GagΔp6 (also named Pr50^GagΔp6) (Fig. 1B) is often used as a surrogate for biochemical analysis of Pr55^Gag function. It is indeed usually assumed that the p6 domain does not affect RNA binding, and hence the vast majority of in vitro studies using recombinant Gag were conducted with GagΔp6 [43,61–66]. However, recent studies comparing the RNA binding properties of NCp7 (i.e. the mature NC domain) and NCp15 (i.e. NC-SP2-p6), suggested that the p6 domain might indirectly affect RNA binding [67]. Similarly, the p6 domain was recently shown to affect binding of Gag proteins to short oligoribonucleotides [68]. We thus decided to test the role of the p6 domain in the specific recognition of Pr55^Gag to the HIV-1 gRNA. We performed a systematic comparison of Pr55^Gag and Gag∆p6 binding to a series of viral and cellular RNA species under strictly identical conditions using fluorescence spectroscopy. Unexpectedly, our results demonstrate that p6 deletion resulted not only in a decreased affinity for gRNA, but also in an increased affinity for spliced viral and cellular RNAs. Indeed, all tested RNAs bound Gag∆p6 with similar affinity, revealing a complete lack of binding specificity. Altogether our findings demonstrate a novel role for the p6 domain as a regulator of the binding specificity of Pr55^Gag to HIV-1 gRNA.

2. Results

2.1. DLS and SLS analysis of GagΔp6 and Pr55^Gag proteins

Our recombinant GagΔp6 and Pr55^Gag proteins were tested and compared using Diffusion Light Scattering (DLS) and Static Light Scattering (SLS) (Fig. 2). The DLS intensity distribution of GagΔp6 and Pr55^Gag in the storage buffer (50 mM Tris-HCl pH 8, 1 M NaCl, 5 mM DTT,) was unimodal and fairly monodisperse (P_d index ~ 18.3% and 11.8%, for GagΔp6 and Pr55^Gag, respectively). The higher molecular weight species observed in the DLS intensity distribution (Fig. 2A, left panel) correspond to very minor populations, and were not observed in the DLS distribution by number (Fig. 2A, right panel). These profiles correspond to mean hydrodynamic radii (Rh) of 4–5.1 nm and 6.5–7.3 nm for GagΔp6 and Pr55^Gag, respectively (Fig. 2A). Using the Stokes-Einstein equation and assuming spherical proteins (see Methods), our analysis suggests that in solution GagΔp6 corresponds to a dimer-trimer, while Pr55^Gag to tetramer-pentamer. On the other hand, SLS data on GagΔp6 and Pr55^Gag in the storage buffer provided molecular weight (MW) estimations of 46.9 ± 1.8 kDa and 114.0 ± 4.0 kDa, respectively, corresponding to a monomer state for GagΔp6 and a dimer for Pr55^Gag in a good agreement with previous analysis [69]. The discrepancy between the oligomeric states determined by DLS and SLS is probably due to the limiting hypothesis imposed by the Stokes-Einstein model, which assumes the proteins to be spherical. Nevertheless, DLS and SLS both indicated that the oligomeric state of the GagΔp6 and Pr55^Gag proteins is different. The DLS profiles and SLS analysis of GagΔp6 and Pr55^Gag were also performed in the binding buffer (30 mM Tris-HCl pH 7.5, 200 mM NaCl, 10 mM MgCl₂) and were shown to be similar to the storage conditions (Fig. 2B).

Figure 2.

DLS analysis of GagΔp6 and Pr55^Gag. DLS analysis of GagΔp6 (red continuous line) and Pr55^Gag (black discontinuous line) (A) in the storage buffer (1 M NaCl, 5 mM DTT, 50 mM Tris-HCl pH 8) (B) and in the binding buffer (30 mM Tris-HCl pH 7.5, 200 mM NaCl, 10 mM MgCl₂) are represented. The intensity (left) and number (right) distributions are represented. The polydispersity index (PdI) and hydrodynamic radius (Rh) of each protein are indicated close to its corresponding peak. Mean ± SD of ten measurements.

2.2. GagΔp6 and Pr55^Gag binding to the 5ʹ region of HIV-1 gRNA

We exploited the intrinsic fluorescent signal of GagΔp6 and Pr55^Gag, which harbor 9 Trp residues in their MA, CA, SP2 and NC domains [70] (Fig. 1B), to determine the GagΔp6 binding parameters (the dissociation constants K_d, and the binding stoichiometry n) for RNA fragments corresponding to the first 600/615 of the HIV-1 gRNA of NL4.3 (N1-600 WT) and MAL (M1-615 WT) isolates (Fig. 3, which were previously shown to bind Pr55^Gag with the same affinity [13,14]. In addition, RNA binding assays with GagΔp6 and Pr55^Gag were performed under the exact same conditions (buffer, temperature, refolding and incubation protocols) [14].

Figure 3.

Schematic representation of the RNA fragments used in this study. The fragments whose names start with M and N are derived from the MAL and NL4-3 strains, respectively [13,14].

The experimental binding curves for both RNAs to Gag∆p6 were fitted with one-binding site model (Fig. 4, Equation 7, see Material and Methods). Accordingly, the Scatchard plots of GagΔp6 to N1-600 WT and M1-615 WT RNAs confirmed the presence of only one class of binding sites of very similar affinity (K_d ~ 11–12 nM, Table 1, Fig. 4). This observation strongly contrasts with the binding of these RNAs to the full-length Pr55^Gag that displayed two classes of binding sites for those RNA fragments: a very high affinity binding site (K_d1 ~ 2–3 nM, Table 1, Fig. 4) and a lower one (K_d2 ~ 15 nM, Table 1, Fig. 4) [14] whose affinity was similar to the one observed for GagΔp6 (Table 1). Note that while the data in Tables 1–3 concerning Pr55^Gag are from our previous publication [14], all curves in Figs. 4 to 7 and in Supplementary Figures 1 and 2 were previously unpublished and allow direct comparison between Gag∆p6 and Pr55^Gag. Analysis of the stoichiometry showed about 6–7 Pr55^Gag proteins bound to the high affinity binding site, while only 3 to the lower one. By comparison, about 3–4 GagΔp6 proteins bound to N1-600 WT and M1-615 WT RNAs (Table 1). Altogether, our data show that GagΔp6 does not display a very high affinity binding site for the 5ʹ-end region of the HIV-1 gRNA, which is the hallmark of the specific binding of Pr55^Gag to this RNA.

Figure 4.

Representative experiments of GagΔp6 and Pr55^Gag binding to the first 600 nucleotides of gRNA. (A) Increasing concentrations of RNA were added to 50 nM of protein. The binding curves corresponding to GagΔp6 binding to N1-600 WT (red triangles) were best fitted according to a single binding site Scatchard model [82]. The interaction of Pr55^Gag with N1-600 WT RNA (black squares) was fitted with a Scatchard-like equation corresponding to a two binding sites-model as previously described [14]. (B) The residual plots for each curve fitted in A are represented. (C) On the left the Scatchard plot of GagΔp6 interaction with N1-600 WT RNA fragment (red triangles) yielded a single linear pattern. Conversely, on the right Pr55^Gag interaction with this same RNA yielded two linear patterns (black squares).

Table 1.

GagΔp6 and Pr55^Gag binding to the first 600/615 nts of the genomic RNA and SL1 mutant RNAs. On the left, the binding parameters derived from the single binding site model [82] and from the stoichiometry analysis [14] (see Methods) for GagΔp6 interacting with RNA fragments corresponding to the first 600/615 nts of gRNA and SL1 mutant RNAs are indicated. On the right, the binding parameters determined for Pr55^Gag in interaction with the same RNA fragments [14]. K_di (i = 1, 2, 3) correspond to the three different classes of binding affinity. n.d. stands for not determined values. Mean ± SD of at least three independent experiments.Gag.

GagΔp6			Pr55Gag
						Stoichiometry
RNA	Kd (nM)	Stoichiometry	Kd1 (nM)	Kd2 (nM)	Kd3 (nM)	Site 1	Site 2	Site 3
N1-600 WT	12.1 ± 0.8	3.7 ± 0.5	1.8 ± 0.4	14.6 ± 3.8		6.9 ± 0.7	3.1 ± 0.8
M1-615 WT	12.0 ± 1.9	3.5 ± 0.5	2.9 ± 0.6	15.4 ± 3.4		6.7 ± 1.1	3.0 ± 0.9
N1-600 sAL	12.1 ± 1.3	3.8 ± 0.3		9.5 ± 2.4	62.5 ± 2.4		2.1 ± 0.4	1.0 ± 0.3
M1-615 sAL	11.3 ± 1.6	4.0 ± 1.2		12.5 ± 2.7	65.5 ± 6.6		1.9 ± 0.2	1.1 ± 0.2
N1-600 srIL	11.1 ± 1.9	3.8 ± 0.2			53.6 ± 9.1			2.4 ± 0.2
N1-600 syIL	12.1 ± 1.9	3.6 ± 0.5	n.d.	n.d.	n.d.
N1-600 ΔIL	11.3 ± 1.8	3.8 ± 0.6	n.d.	n.d.	n.d.

Table 3.

GagΔp6 and Pr55^Gag binding to the spliced viral and cellular RNA species. On the left, binding parameters derived from the single binding site model [82] and from the stoichiometry analysis [14] (see Methods) for GagΔp6 in interaction with RNA fragments corresponding to spliced viral and cellular RNAs. On the right, binding parameters determined for Pr55^Gag in interaction with the same RNAs fragments [14]. K_di (i = 1, 2) correspond to the two different classes of binding affinity. Mean ± SD of at least three independent experiments.

		GagΔp6		Pr55Gag
						Stoichiometry
RNA	Length (nts)	Kd (nM)	Stoichiometry	Kd1 (nM)	Kd2 (nM)	Site 1	Site 2
N1-600 WT	600	12.1 ± 0.8	3.7 ± 0.5	1.8 ± 0.4	14.6 ± 3.8	6.9 ± 0.7	3.1 ± 0.8
N1-600 NEF	600	11.0 ± 0.5	2.8 ± 0.4		16.3 ± 4.2		2.7 ± 0.6
N1-600 REV	600	13.7 ± 1.7	3.4 ± 0.5		11.7 ± 2.0		3.5 ± 0.8
N1-600 VPR	600	12.4 ± 0.9	3.5 ± 0.1		18.1 ± 3.9		3.0 ± 0.2
A3G 3’UTR	325	10.6 ± 0.4	3.8 ± 0.4		18.5 ± 4.1		3.3 ± 0.6
7SL	300	14.5 ± 1.7	3.1 ± 0.2		20.5 ± 1.1		2.2 ± 0.1

Figure 7.

Representative experiments of GagΔp6 and Pr55^Gag binding to gRNA and non-viral RNA species. Increasing concentrations of RNA were added to 50 nM of protein. (A) On the left, the binding curves corresponding to GagΔp6 binding to N1-600 WT (black squares), APOBEC3G 3ʹUTR (red triangles) and 7SL (cyan circles) RNAs were fitted according to the single binding site model [82]. On the right, the data of Pr55^Gag in interaction with N1-600 WT RNA (black squares) were fitted with a two binding sites-model as previously described [14]. APOBEC3G 3ʹUTR (red triangles) and 7SL (cyan circles) RNAs were best fitted with the single binding site model. (B) The corresponding residual plots for each curve fitted in A are represented. (C) On the left, single linear patterns were observed for GagΔp6 in interaction with N1-600 WT (black squares), APOBEC3G 3ʹUTR (red triangles) and 7SL (cyan circles) RNAs. On the right, the Scatchard plots of Pr55^Gag interaction with N1-600 WT RNA yielded two linear patterns (black squares), while single linear patterns were observed for in interaction with APOBEC3G 3ʹUTR (red triangles) and 7SL (cyan circles) RNAs.

2.3. GagΔp6 and Pr55^Gag binding to SL1 RNA mutants

We previously identified two major determinants for the interaction between Pr55^Gag and gRNA located in SL1 [13,14,71]. They correspond to gRNA dimerization, which is ensured by a 6 nucleotides palindromic sequence in the apical loop [7–11], and to the purine-rich internal loop (Fig. 1A). To compare the role of gRNA dimerization on Gag∆p6- and Pr55^Gag-RNA binding, we analyzed RNA fragments in which the apical loop of SL1 was mutated to prevent gRNA dimerization (M1-615SL1sAL and N1-600SL1sAL, Fig. 3) [7,9]. Those RNA fragments are defective in RNA dimerization under the buffer conditions and at the RNA concentration used in this study [13,14]. Mutations in the SL1 apical loop had little or no effect on the binding of Gag∆p6, as our analysis displayed one class of binding sites, with binding affinity (K_d ~ 11–12 nM, Table 1) and stoichiometry (about 4 proteins, Table 1) very similar to gRNA. This is in contrast with the effect of these mutations on Pr55^Gag binding, which resulted in loss of the class of high-affinity binding sites and the appearance of a of low affinity binding class (K_d3 ~ 65 nM, Table 1) in addition to the moderate affinity binding class (K_d2 ~ 12 nM, Table 1) [14]. Furthermore, mutations in the SL1 apical loop did not significantly alter the Gag∆p6 binding stoichiometry (3–4 Gag∆p6/RNA, Table 1), while it reduced the number of Pr55^Gag molecules bound per RNA molecules from 9–10 to 3 (Table 1).

Next, we compared the interaction of GagΔp6 and Pr55^Gag with RNA mutants in which purines of the internal loop of SL1 were substituted (AGG -> GAA, N1-600 SL1srIL, or AGG-> UUU, N1-600 SL1syIL) or deleted (N1-600 SL1ΔIL) (Figs. 3 and 5A). Importantly, these mutations do not affect the RNA secondary structure, nor abrogate the RNA dimerization under our experimental conditions [13,14]. Mutations in the SL1 internal loop had little or no effect on Gag∆p6 binding, which displayed one class of binding sites (Fig. 5, left panels) with an affinity and a stoichiometry similar to gRNA (Fig. 5, left panels and Table 1). At the opposite, no binding of Pr55^Gag could be detected when the purines of the internal loop were deleted (N1-600 SL1ΔIL RNA) or substituted by pyrimidines (N1-600 SL1syIL RNA) (Fig. 5, right panels). Similarly, substituting GAA for AGG in this loop (N1-600 SL1srIL RNA) dramatically reduced binding of the full-length Pr55^Gag bound with moderate affinity (Fig. 5, right panels and Table 1) [14]. Thus, while Gag∆p6 has a lower affinity than Pr55^Gag for the wild type 5ʹ-end region of HIV-1 gRNA, it has a higher affinity than the full-length Gag precursor for the SL1 internal loop mutants.

Figure 5.

Representative experiments of GagΔp6 and Pr55^Gag binding to SL1 internal loop mutants. (A) The internal loop mutants of SL1 used in this study. The DIS sequence within SL1 apical loop is highlighted in magenta. Increasing concentrations of RNA were added to 50 nM of protein. (B) On the left hand side, the data corresponding to GagΔp6 binding to N1-600 WT (black squares), N1-600 SL1srIL (red circles), and N1-600 SL1ΔIL RNAs (blue triangles) were best fitted according to the single binding site model [82]. On the right hand side, the binding curves corresponding to Pr55^Gag interaction with N1-600 WT RNA (black squares) were fitted with a two binding sites-model as previously described [14]. N1-600 SL1srIL (red circles) was best fitted with the single binding site model, while data corresponding to N1-600 SL1DIL (blue triangles) could not be fitted. (C) The corresponding residual plots for each curve fitted in B are represented. (D) On the left, we observed for GagΔp6 interaction with N1-600 WT (black squares), N1-600 SL1srIL (red circles), and N1-600 SL1ΔIL (blue triangles) RNAs single linear patterns. On the right, the Scatchard plots of Pr55^Gag interaction with N1-600 WT RNA yielded two linear patterns (black squares), while a single linear pattern for N1-600 SL1srIL RNA (red circles) was observed.

In addition, to gain a better understanding of the interaction between GagΔp6 and Psi, we compared the interactions between GagΔp6 and Pr55^Gag with truncated viral RNA fragments partially including Psi (Supplementary Fig. 1A). In contrast with Pr55^Gag, truncations in the 5ʹ-end region of the HIV-1 gRNA did not impact the GagΔp6 binding affinity, implying that the interaction between Gag∆p6 and this gRNA portion is non-specific (Supplementary Table 1 and Supplementary Fig. 1).

2.4. GagΔp6 and Pr55^Gag binding to the Psi individual elements

We next compared binding of GagΔp6 and Pr55^Gag to each of the four stem-loops located in the Psi region (Fig. 1A). Our data with these short chemically synthesized RNA fragments (14 to 35 nucleotides) revealed that GagΔp6 bound these stem-loops with a very similar affinity (K_d ~ 21–29 nM) and stoichiometry (n = ~ 3) (Supplementary Fig. 2 and Table 2). The full-length precursor also displayed only one class of binding sites for these individual stem-loops (Table 2). However, Pr55^Gag displayed a preferential binding to SL1, since the affinity for this motif was found to be 3- to 8-fold higher compared to the affinity determined for SL2 to SL4 [14]. Altogether these results further confirm that while Pr55^Gag preferentially binds SL1, Gag∆p6 does not discriminate between the four hairpins located in Psi.

Table 2.

GagΔp6 and Pr55^Gag binding to the individual stem-loops of the Psi region. On the left, binding parameters derived from the single binding site model [82] and from the stoichiometry analysis [14] (see Methods) for GagΔp6 in interaction with individual stem-loops of Psi. On the right, the binding parameters determined for Pr55^Gag in interaction with the same RNA fragments [14]. Mean ± SD of at least three independent experiments.

		GagΔp6		Pr55Gag
RNA	Length (nts)	Kd (nM)	Stoichiometry	Kd (nM)	Stoichiometry
N35 SL1	35	21 ± 2	3.2 ± 0.4	5 ± 2	3.50 ± 0.65
NSL2	17	22 ± 4	3.1 ± 0.6	44 ± 4	2.17 ± 0.49
NSL3	14	29 ± 1	2.8 ± 0.5	21 ± 2	2.33 ± 0.67
NSL4	24	25 ± 3	2.9 ± 0.4	18 ± 4	2.35 ± 0.78

2.5. GagΔp6 and Pr55^Gag interaction with spliced viral RNAs and non-viral RNAs

Finally, we compared GagΔp6 and Pr55^Gag binding to several spliced viral RNAs (N1-600 NEF, N1-600 VPR, and N1-600 REV) (Fig. 3) [72]. GagΔp6 showed one class of binding sites of moderate affinity (K_d ~ 11–14 nM, Table 3, Fig. 6) for those RNAs with a stoichiometry of about 3 proteins per RNA molecule (Table 3). Of note, Pr55^Gag bound the same RNA fragments with similar affinity and stoichiometry (K_d ~ 12–20 nM, and n ~ 3, Table 3) [14].

Figure 6.

Representative experiments of GagΔp6 and Pr55^Gag binding to genomic and spliced viral RNA fragments. Increasing concentrations of RNA were added to 50 nM of protein. (A) On the left, the binding curves corresponding to GagΔp6 binding to N1-600 WT (black squares), N1-600 VPR (red circles) and N1-600 REV (cyan circles) RNAs were fitted according to the single binding site model. On the right, the data of Pr55^Gag in interaction with N1-600 WT RNA (black squares) were fitted with a two binding sites-model as previously described [14], while data corresponding to Pr55^Gag interaction with N1-600 VPR (red circles) and N1-600 TAT (cyan circles) RNAs were best fitted with a single binding site model [82]. (B) The corresponding residual plots for each curve fitted in A are represented. (C) On the left, single linear patterns were observed for GagΔp6 interaction with N1-600 WT (black squared), N1-600 VPR (red circles) and N1-600 REV (cyan circles) RNAs. On the right, the Scatchard plots of Pr55^Gag interaction with N1-600 WT RNA yielded two linear patterns (black squares), while single linear patterns were observed for N1-600 VPR (red circles) and N1-600 TAT (cyan circles) RNAs.

Finally, we analyzed the binding parameters of GagΔp6 towards non-viral RNAs such as 7SL RNA which is known to be packaged into HIV-1 viral particle [32,33], and the 3ʹUTR of APOBEC3G mRNA [73]. GagΔp6 displayed only one class of binding sites for those non-viral RNAs (Fig. 7) with a K_d of about 14 nM and 10.6 nM for the 7SL RNA and the 3ʹUTR of APOBEC3G mRNA, respectively (Table 3). Both RNAs were bound by 3 GagΔp6 proteins (Table 3). Pr55^Gag bound these RNAs similarly (K_d ~ 18–20 nM) with a stoichiometry of 2–3 proteins (Table 3) [14]. These data show that deletion of p6 leads to very similar binding parameters for viral and non-viral RNA fragments, thus resulting in non-specific RNA binding properties.

2.6. The C-terminal Hist-tag does not affect the binding specificity of Pr55^Gag and GagΔp6

To test whether the presence of a His-tag in the Gag proteins affects their binding specificity to RNAs, we compared binding of His tagged Pr55^Gag and GagΔp6 proteins with Pr55^Gag and GagΔp6 proteins from which the His-tag had been removed by TEV (Tobacco Etch Virus) protease (Supplementary Figs. 3 and 4) to a representative panel of RNA fragments, including N1-600WT RNA, SL1 mutants N1-600 sAL and N1-600 srIL RNAs, and spliced viral N1-600 VPR RNA (Table 4 and Fig. 8). The impact of cleavage of the C-terminal His-tag on GagΔp6 binding to the RNA fragments resulted in increased Kd values (~ 2.5-fold) compared to the ones obtained with the His-tagged GagΔp6. Similarly, the TEV-cleaved Pr55^Gag displayed ~ 2.5-fold increased Kd values for the two classes of binding sites to N1-600 WT RNA, compared to the His-tagged Pr55^Gag protein (Fig. 8). Binding to N1-600 sAL RNA showed a single class of binding sites corresponding to the lower affinity component previously observed for the His-tagged Pr55^Gag (Kd ~ 60 nM, Table 4). Finally, TEV-cleaved Pr55^Gag displayed rather similar values for N1-600 srIL and N1-600 VPR RNAs compared to the His-tagged Pr55^Gag (Table 4). Altogether these results demonstrate that even though removal of the His-tag results in a general increase of the Kd values, this tag has a significant effect neither on the specific binding of Pr55^Gag to N1-600 WT RNA nor on the complete lack of RNA binding specificity of Gag∆p6.

Figure 9.

Model of Pr55^Gag binding specificity to gRNA. (A) The C-terminal acidic p6 domain of Pr55^Gag folds over the highly basic NC domain [67], thus partially masking the NC positive charges. According to our model this would promote the specific binding of Pr55^Gag to gRNA. (B) Our data show that the Pr55^Gag p6 domain deletion results in a non-specific binding to all the tested viral and cellular RNAs, since GagΔp6 displayed a similar affinity for all of them.

Table 4.

His-tagged and TEV-cleaved GagΔp6 and Pr55^Gag proteins binding to the first 600/615 nts of the gRNA, SL1 mutant RNAs and N1-600 VPR spliced viral RNA. On the left side, binding parameters derived from the single binding site model[82] and from the stoichiometry analysis[14] (see Methods) for GagΔp6 His-tagged and TEV-cleaved in interaction with RNA fragments corresponding to the first 600/615 nts of the gRNA, N1-600 srIL and N1-600 sAL, SL1 mutant RNAs, and N1-600 VPR spliced viral RNA. On the right side, binding parameters determined for Pr55^Gag His-tagged[14], and for Pr55^Gag TEV-cleaved in interaction with the same RNA fragments. K_di (i = 1, 2) correspond to the two different classes of binding affinity. Mean ± SD of three independent experiments.

	GagΔp6				Pr55Gag
	His-tagged	TEV- cleaved			His-tagged	TEV-cleaved
RNA	Kd (nM)	Kd (nM)	Kd1 (nM)	Kd2 (nM)	Kd3 (nM)	Kd1 (nM)	Kd2 (nM)	Kd3 (nM)
N1-600 WT	12.1 ± 0.8	33.7 ± 2.4	1.8 ± 0.4	14.6 ± 3.8		4.2 ± 0.9		32.5 ± 5.8
N1-600 sAL	12.1 ± 1.3	22.7 ± 2.5		9.5 ± 2.4	62.5 ± 2.4			60.8 ± 4.2
N1-600 srIL	11.1 ± 1.9	31.2 ± 3.6			53.6 ± 9.1			40.1 ± 4.7
N1-600 VPR	11.3 ± 1.8	28.6 ± 4.1		18.1 ± 3.9			24.3 ± 2.1

Figure 8.

Representative experiments of TEV-cleaved GagΔp6 and Pr55^Gag proteins binding to gRNA, SL1 mutant RNAs and N1-600 VPR spliced viral RNA Increasing concentrations of RNA were added to 50 nM of protein. (A) On the left side, the binding curves corresponding to TEV-cleaved GagΔp6 binding to N1-600 WT (black squares), N1-600 sAL (red triangles), N1-600 srIL (blue circles), and N1-600 VPR (green triangles) RNAs were fitted according to the single binding site model [82]. On the right side, the data of TEV-cleaved Pr55^Gag in interaction with N1-600 WT RNA (black squares) were fitted with a two binding sites-model, as previously described [14]. N1-600 sAL (red triangles), N1-600 srIL (blue circles), and N1-600 VPR (green triangles) RNAs were best fitted with the single binding site model. (B) The corresponding residual plots for each curve fitted in A are represented. (C) On the left side, single linear patterns were observed for TEV-cleaved GagΔp6 in interaction with N1-600 WT (black squares), N1-600 sAL (red triangles), N1-600 srIL (blue circles), and N1-600 VPR (green triangles) RNAs. On the right side, the Scatchard plots of Pr55^Gag interaction with N1-600 WT RNA yielded two linear patterns (black squares), while single linear patterns were observed for in interaction with N1-600 sAL (red triangles), N1-600 srIL (blue circles), and N1-600 VPR (green triangles) RNAs.

Finally, to assess the impact of ionic strength on RNA binding, we also tested His-tagged and TEV-cleaved GagΔp6 and Pr55^Gag proteins binding to N1-600 WT under more physiological ionic conditions (30 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM MgCl₂ and 4.8 mM spermidine). The resulting binding parameters were found to be similar to the ones obtained in our standard binding buffer conditions (30 mM Tris-HCl pH 7.5, 200 mM NaCl, 10 mM MgCl₂) (Supplementary Table 2).

3. Discussion

It is usually assumed that the p6 domain of Pr55^Gag is not involved in its RNA binding properties, and Gag∆p6, which is easier to produce than the full-length precursor is frequently used as a surrogate for this protein in RNA binding studies [60–66]. However, an NMR study of NCp15 [67] and a recent ITC analysis of Pr55^Gag and Gag∆p6 binding to short oligoribonucleotides [68] suggested a different picture. To dissect the possible impact of the C-terminal p6 domain of Pr55^Gag on the specific recognition of HIV-1 gRNA, we systematically compared the binding of GagΔp6, and full-length Pr55^Gag with a panel of wild type and mutant RNAs, also including spliced viral and cellular RNAs, which Pr55^Gag must discriminate during the specific gRNA selection and packaging [27,30,32,33].

Unlike full-length Pr55^Gag, which bound HIV-1 gRNA with a ten-fold higher affinity than spliced viral and cellular RNAs (Tables 1 and 3), Gag∆p6 displayed a very narrow range of affinities (K_d ~ 11–14 nM; Tables 1 and 3) for all tested RNA fragments, therefore revealing a complete lack of binding specificity. In addition, while about 3–4 Gag∆p6 molecules bound to any of these RNAs, significantly more Pr55^Gag molecules bound to the 5ʹ-end region of the HIV-1 gRNA than to spliced viral or cellular RNAs (~ 10 versus ~ 2–3; Tables 1 and 3). Importantly, our present findings are in complete agreement with previous publications. On the one hand, preferential binding of Pr55^Gag to HIV-1 gRNA was reported using fluorescence spectroscopy (this work and Bernacchi et al. [14]), as well as filter binding, band-shift, and footprinting assays [13]. On the other hand, fluorescence anisotropy [43] and FCS [74] studies showed that binding of Gag∆p6 is non-specific, except in the presence of very high salt concentrations or competitors. The discrepancies between these studies could have various origins: the methods used, the experimental protocols, the RNA fragments, as well as the proteins used (Pr55^Gag vs. Gag∆p6, and His-tagged vs. non-tagged proteins); hence, the importance of conducting a comparative study under strictly identical conditions.

Altogether, our results indicate that Pr55^Gag and Gag∆p6 bind RNAs in very different ways, suggesting that Gag∆p6 is not an adequate surrogate for Pr55^Gag when looking at specific interaction with HIV-1 gRNA. Of note, the Pr55^Gag and Gag∆p6 proteins used in most of our experiments were His-tagged at their C-termini (Tables 1–3 and Figs. 2–7); however our experiments with TEV-cleaved Gag proteins showed that the His-tag does not impact the specificity of Pr55^Gag and Gag∆p6 binding to viral RNAs (Table 4 and Fig. 8); therefore, the differences observed cannot be attributed to this tag. In line with these results, recent reports showed that while a His-tag can affect the RNA binding properties of Pr55^Gag at low ionic strength, it has negligible effect under the ionic conditions used in our study [75] and the same pattern of Pr55^Gag binding to short nucleic acids was observed with untagged and His-tagged Pr55^Gag [68].

We previously showed that SL1 plays a key role in the specific binding of Pr55^Gag to HIV-1 gRNA [13,14,71], with the internal loop and the apical loop which mediates gRNA dimerization, both contributing to specificity. However, substituting or deleting the SL1 internal loop or preventing RNA dimerization by mutating the SL1 apical loop had no effect on Gag∆p6 binding, while the same mutations drastically reduced or even abolished binding of Pr55^Gag (Table 1). Indeed, Gag∆p6 bound the four hairpins present in Psi equally well (Table 2), whereas Pr55^Gag has a higher affinity for SL1 and requires the complete 5ʹ-end region of the HIV-1 gRNA for optimal binding. Interestingly, the p6 domain increased the specificity of Pr55^Gag not only by increasing its affinity for gRNA, but also by decreasing its affinity for mutant RNAs (especially those in the SL1 internal loop (Table 1) and cellular RNAs (Table 3).

Nevertheless, it seems unlikely that the Pr55^Gag binding specificity is regulated by a direct interaction of p6 with gRNA, as this domain contains seven conserved Glu residues, conferring to p6 a high negative charge density. Interestingly, NMR studies on the first protease-induced maturation product, NCp15 (NC-sp2-p6) (Fig. 1B), suggested that the acidic p6 domain folds back and interacts with the basic zinc fingers motifs of the NC domain [67]. We propose that the same interaction may take place in the context of full-length Pr55^Gag and increase the RNA binding specificity by partially masking the NC positive charges, and likely increasing steric selection (Fig. 9). Thus, this comparative analysis of Pr55^Gag and GagΔp6 proteins performed under strictly identical conditions unveiled a new role for the p6 domain in the selective binding of Pr55^Gag to the 5ʹ-end region of the HIV-1 gRNA. Our study thus strongly suggests that the p6 domain contributes to the specific selection/discrimination of gRNA from spliced viral RNAs and cellular RNAs by Pr55^Gag required for its selective packaging into viral particles.

4. Material and methods

4.1. Pr55^gag and GagΔp6 protein expression and purification

The full-length Pr55^Gag and the GagΔp6 proteins were expressed and purified as described previously [76]. Protein samples were ultra-centrifuged for 1 h at 100,000 g at 4°C immediately prior use and the upper half of the samples was then carefully transferred into a protein low-binding microtube (Eppendorf) for further experiments.

The plasmid containing the full-length wild type Pr55^Gag gene with a non-cleavable hexa-histidine sequence at the C-terminus of the gene in an engineered pET28a vector was used as the template to generate the TEV cleavable GagΔp6-TEV-His and Pr55^Gag-TEV-His constructs. These proteins containing the TEV sequence were digested with 1:25 (w/w) TEV protease (His-tagged produced in-house) at 4°C for 14 hrs. The TEV digested proteins were applied to a 1 ml Ni-NTA column and the flow through was collected and passed over the column two more times. The cleaved His-Tag, His-TEV and uncut fusion proteins were eluted from the column using the elution buffer (1.0 M NaCl, 50 mM Tris-HCl pH 8.0, 5 mM MgCl₂, 250 mM imidazole, 1% (v/v) Tween-20, 10% (v/v) glycerol, 5 mM DTT). His-tag cleavage from Pr55^Gag and GagΔp6 proteins was checked by western blot and mass spectrometry analyses (Supplementary Figs. 3 and 4).

4.2. Western blots

One µg of His-tagged or TEV-cleaved Pr55^Gag and Gag∆p6 proteins was loaded on a 4–12% Criterion TGX 4–15% gels (Bio-Rad) and transferred to a 22 μm PVDF membranes using the Trans-Blot® Turbo™ Transfer System (Bio-Rad). Gag proteins were detected with a HIV-positive patient serum and anti-His monoclonal antibody (sc-8036, Santa Cruz Biotechnology,) followed by horseradish peroxidase conjugated anti-human (NA933, GE Healthcares) or anti-mouse antibodies (170–6516, Bio-Rad), respectively. Proteins were then visualized by chemiluminescence using the ECL Prime Western blotting detection reagent (GE Healthcares) and analyzed with the ChemiDoc^TM Touch Imaging System (Bio-Rad).

4.3. Mass spectrometry

Prior mass analysis, protein samples were dialyzed against 200 mM ammonium acetate in a Slide-A-Lyzer™ MINI Dialysis Device 7000 NMWL. Mass spectrometry was performed on an ESI-TOF mass spectrometer Synapt G2-s (Waters, MA, USA). Mass analysis was performed under denaturing conditions: the protein was diluted in 50/50 water acetonitrile (v/v) mixture acidified with 1% formic acid to achieve a final concentration of 0.3 µM. Data were acquired in the positive ionization mode from 500 to 2000 m/z. The ionization conditions were 30 V for the sample cone, 3.0 kV for the capillary and the source temperature was set to 130°C. Multiple charge spectrum were deconvoluted using Waters MassLynx MaxEnt1 software and protein molecular weights were calculated using the ProtParam tool (https://web.expasy.org/protparam/).

4.4. Dynamic light scattering (DLS) analysis

GagΔp6 and Pr55^Gag samples were characterized by DLS. Proteins were diluted to a final concentration of 5 µM in the storage buffer (50 mM Tris-HCl pH 8, 1 M NaCl, 5 mM DTT,) or in the binding buffer (30 mM Tris-HCl pH 7.5, 200 mM NaCl, 10 mM MgCl₂). Intensities of the scattered light and correlation times were measured using a Zetasizer^TM Nano S apparatus (4 mW He-Ne laser, λ₀ = 633 nm, scattering angle θ = 173°) (Malvern, UK). Measurements were performed at 20°C in a single 50 µl trUView cuvette (BioRad Laboratories, CA USA). Variations of the diffused light intensity were recorded at microsecond time intervals. An autocorrelation function was derived, allowing the determination of the translational diffusion coefficients (D). Assimilating the proteins in solution to spheres, the diffusion coefficients were related to the hydrodynamic radius (R_h) of the molecules populations present in solution, via the Stokes-Einstein equation:

$$D=\frac{kT}{R_{h}6\pi \mu }$$ (1)

in which k is the Boltzmann constant, T is the temperature and µ is the viscosity of the solvent. All experimental data were corrected for solvent viscosity and refractive index. In our experimental settings, solvent viscosity was 1.104 cP and 0.9891 cP for storage and binding buffer, respectively. Solvent refractive index was 1.341 and 1.333 for storage and binding buffer, respectively, as estimated using Malvern Zetasizer Software calculator (Malvern, UK).

4.5. Static light scattering (SLS) analysis

The molecular mass of GagΔp6 and Pr55^Gag in solution, in storage or in binding buffer, was determined by SLS. The intensities of scattered light were measured using a DynaPro Nanostar^TM (100 mW He-Ne laser) (Wyatt Technologies,) in a 1 µl quartz cuvette (JC-006, Wyatt Technologies). The intensity of the scattered light I_SL is a function of the particles size and is proportional to the particles concentration and their mass. Thus I_SL can be used to derive the mass of the particles in solution through Zimm’s development of Rayleigh equation [77,78]:

$$\frac{KC}{R\left(\theta ,C\right)}=\frac{1}{M_{w}P\left(\theta \right)}+2A_{2}C$$ (2)

in which R(θ,C) is the excess Rayleigh ratio of the solution as a function of scattering angle θ (with θ = 90° in our setup) and concentration C. C is the solute concentration, M_w is the weight-averaged solute molar mass, A₂ is the second virial coefficient, P(θ) is the angular dependence of sample scattering and K a constant defined as:

$$K=\frac{2{\pi }^2}{{{\lambda }_0}^4{N}_A}{\left(\frac{d_n}{d_c}{n}_0\right)}^2$$ (3)

where λ₀ is the laser wavelength (λ₀ = 633 nm), N_A is Avogadro’s number, n₀ is the refractive index of the solvent and dn/dc the increment refractive index with sample concentration. The refractive index of each solvent was set as mentioned above in the DLS section. Before sample acquisition the offset of the solvent was measured for subsequent sample data treatment. All buffers were filtered using 0.02 µm filters (Millex ®) before analysis or sample dilution.

4.6. Plasmids, in vitro RNA transcription and purification

All plasmids used for the synthesis of wild type and mutant HIV-1 RNA fragments were previously described [7,9,13,14,72,79], as well as plasmids used for in vitro transcription of 7SL RNA and of the 3ʹUTR of APOBEC3G mRNA [14,73]. Linearized plasmids were used as templates for the synthesis of RNA fragments by in vitro run-off transcription using T7 bacteriophage RNA polymerase, followed by purification on a size-exclusion chromatography column [80]. The purified transcripts were folded as described previously [14]. Briefly, RNA fragments were prepared in Milli-Q (Millipore), denatured for 2 min at 90°C and snap-cooled on ice for 2 min. Proper folding was achieved by the addition of the binding buffer (30 mM Tris-HCl pH 7.5, 200 mM NaCl, 10 mM MgCl₂) and incubation for 15 min at 37°C.

4.7. Synthetic HIV-1 RNA oligonucleotides

RNAs corresponding to the individual stem-loops of the Psi region were chemically synthesized and purified by reverse-phase HPLC and polyacrylamide gel electrophoresis (Integrated DNA Technologies Inc.). RNA oligonucleotides were then folded as described above.

4.8. Steady-state fluorescence spectroscopy

Prior to protein binding analysis, 1 μM RNA was folded as described above. Fluorescence measurements were performed in quartz cells at 20 ± 0.5°C on a Fluoromax-4 fluorimeter (HORIBA Jobin-Yvon Inc., NJ., USA). The excitation wavelength was set at 295 nm for selective excitation of tryptophan residues and the emission wavelength was scanned from 305 to 450 nm. The integration time was set on 0.1 s and the excitation and emission bandwidth on 5 nm. Increasing amounts of RNA were added to 50 or 100 nM GagΔp6 or Pr55^Gag in the binding buffer, so that the RNA/protein ratio varied from 0 to 5. To assess the impact of ionic strength in the binding experiments, we also tested GagΔp6 and Pr55^Gag binding to N1-600 WT under more standard physiological salt conditions (30 mM Tris-HCl pH 7.5, 1 mM MgCl₂, 150 mM NaCl, and 4.8 mM spermidine [81]) (see Results). After addition of each RNA aliquot, the quartz cell was rapidly homogenized and the fluorescence emission measured.

The emission spectra of each titration were integrated and the fluorescence intensities corrected for buffer fluorescence and dilution effects. In order to determine the binding parameters of GagΔp6 and Pr55^Gag to each RNA fragment, the corrected fluorescence intensity I measured for any added RNA concentration was converted into binding density ν which corresponds to the protein bound, P_b, to nucleic acid concentration, A_t, ratio:

$$=\frac{P_b}{A_t}=\frac{\left(I_0-I\right)}{\left(I_0-I_F\right)}\frac{P_t}{A_t}$$ (4)

With I₀ corresponding to the protein fluorescence intensity in absence of nucleic acids, I_F to the fluorescence intensity at the end of the titration, and P_t to the total protein concentration.

Since:

$$\frac{P_b}{P_t}=\frac{\left(I_0-I\right)}{\left(I_0-I_F\right)}$$ (5)

$${\mathrm{P}}_f=P_t-P_b=P_t-A_t$$ (6)

The concentration of bound and free protein (P_b and P_f, respectively) can be calculated as functions of I, I₀, I_f, P_t and A_t using Equations (5) and (6). The experimental observed affinity constant K_obs, and thus the dissociation constant K_d, mathematically corresponding to its inverse, was then computed by fitting the experimental data to equation [82]:

$$=\frac{\left(K_{obs}{P}_f\right)}{\left(1+K_{obs}{P}_f\right)}$$ (7)

Plots were fitted with equations corresponding to models with a single class or two classes of binding sites. To confirm the number of Pr55^Gag and GagΔp6 binding site classes for viral and cellular RNAs, we then plotted the fraction of bound protein vs. the unbound protein fraction (Scatchard plots). The presence of one or two linear pattern(s) in these plots indicated the presence of one or two classes of RNA binding sites, respectively.

The analysis of fluorescence binding curves also allowed determination of the GagΔp6 and Pr55^Gag binding stoichiometry (n), corresponding to the average number of proteins bound to one RNA molecule. The experimental data expressed as normalized fluorescence quenching were reported vs. the molar ratio of total [RNA] expressed in strands to [GagΔp6] or [Pr55^Gag]. The stoichiometry of the complexes could then be graphically recovered by the intersection of the initial slope at low [RNA]/[protein] ratio with the fluorescence plateau at the end of the titration, as previously described [14,83].

Biography

ND performed DLS, SLS and fluorescence spectroscopy experiments, analyzed results, and generated the figures. KK, SG, WM and JM produced Pr55^Gag and GagΔp6 proteins. TS performed western blots, and PW contributed with mass spectrometry analysis. JCP and RM contributed to the design of the study and the data analysis and revised the manuscript. JM contributed to revise the manuscript. SB designed and supervised the study, analyzed the data and drafted the manuscript.

Funding Statement

This work was supported by grants from: the Agence Nationale de la Recherche sur le SIDA et les Hepatites Virales (ANRS) to SB, and a fellowship of the French Ministry of Higher Education and Research to ND.

Acknowledgments

We thank Dr Bernard Lorber (UPR9002 CNRS, Université de Strasbourg) for assistance with DLS and SLS analyses. We thank Dr Han Tanwar (Deakin University) who has contributed to the production of recombinant Pr55^Gag and GagΔp6 proteins.

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplementary material

Supplemental data for this article can be accessed here.