|Home | About | Journals | Submit | Contact Us | Français|
The human immunodeficiency virus type 1 (HIV-1) matrix (MA) protein represents the N-terminal domain of the HIV-1 precursor Gag (PrGag) protein and carries an N-terminal myristate (Myr) group. HIV-1 MA fosters PrGag membrane binding, as well as assembly of envelope (Env) proteins into virus particles, and recent studies have shown that HIV-1 MA preferentially directs virus assembly at plasma membrane sites enriched in cholesterol and phosphatidylinositol-(4,5)-bisphosphate (PI[4,5]P2). To characterize the membrane binding of MA and PrGag proteins, we have examined how Myr-MA proteins, and proteins composed of Myr-MA and its neighbor Gag capsid (CA) protein associate on membranes containing cholesterol and PI[4,5]P2. Our results indicate that Myr-MA assembles as a hexamer of trimers on such membranes, and imply that MA trimers interconnect CA hexamer rings in immature virus particles. Our observations suggest a model for the organization of PrGag proteins, and for MA-Env protein interactions.
The HIV-1 matrix (MA) protein represents the N-terminal domain of the precursor Gag (PrGag) protein that directs the assembly of HIV particles (Swanstrom and Wills, 1997). Molecular biology experiments have demonstrated that MA plays essential roles in targeting PrGag proteins to membranes for assembly, and in incorporation of HIV SU/TM (gp120/gp41) envelope (Env) protein trimers into virus particles (Yu et al., 1992; Wang and Barklis, 1993; Wang et al., 1993; Facke et al., 1993; Dorfman et al., 1994; Zhou et al., 1994; Mammano et al., 1995; Freed and Martin, 1995, 1996; Zhou and Resh, 1996; Ono et al., 1997, 2000; Ono and Freed, 1999; Morikawa et al., 2000; Wyma et al., 2000; Tritel and Resh, 2001; Sanders et al., 2002; Zhu et al., 2003; Yang et al., 2005; Davis et al., 2006; Bhatia et al., 2007; Scholz et al., 2008). However, alternative membrane binding domains (MBDs) can replace MA in terms of directing the assembly of virus particles at membranes (Jouvenet et al., 2006; Scholz et al., 2008). Moreover, viruses carrying large deletions in the MA coding region can be infectious in cell culture systems, as long as the large cytoplasmic domain of SU is deleted (Wang et al., 1993; Reil et al., 1998). Despite these studies suggesting that MA is not absolutely essential for replication, the MA domains of natural HIV-1 isolates are highly conserved, indicating the importance of the protein for in vivo infections (Kuiken et al., 2002).
Among its structural features, HIV-1 MA is myristoylated at its N-terminus, and oligomerization of PrGag proteins appears to trigger a conformational switch, whereby the previously buried myristate (Myr) group becomes exposed, facilitating membrane binding (Bryant and Ratner, 1990; Zhou and Resh, 1996; Spearman et al., 1997; Hermida-Matsumoto and Resh, 1999; Paillart and Gottlinger, 1999; Bouamr et al., 2003; Resh, 2004; Tang et al., 2004). HIV-1 MA also has a positively charged face that permits it to bind to negatively charged phospholipids such as phosphatidylserine (PS; Zhou et al., 1994; Ehrlich et al., 1996; Zhou and Resh, 1996; Scarlata et al., 1998; Tritel and Resh, 2001; Murray et al., 2005; Alfadhli et al., 2007). However, in vivo and in vitro experiments have demonstrated that MA preferentially binds to another lipid, phosphatidylinositol-(4,5)-bisphosphate (PI[4,5]P2; Ono et al., 2004; Saad et al., 2006, 2007; Chukkapalli et al., 2008). These results are consistent with the observation that HIV-1 assembly occurs at PI[4,5]P2- and cholesterol-rich plasma membrane (PM) sites, and may be influenced by signaling pathways that generate PI[4,5]P2 (Nguyen and Hildreth, 2000; Lindwasser and Resh, 2001; Ono and Freed, 2001; Graham et al., 2003; Holm et al., 2003; Ono et al., 2004; Gomez and Hope, 2006; Saad et al., 2006, 2007; Chukkapalli et al., 2008).
HIV-1 MA has shown a predilection to form trimers in solution and in three-dimensional (3D) crystals (Hill et al., 1996; Morikawa et al., 1998; Tang et al., 2004). However, we previously showed that the protein appears to assemble as hexamers rather than trimers on PS-cholesterol membranes (Alfadhli et al., 2007). Nevertheless, these studies were performed at low ionic strength with high concentrations of PS, which is not a preferred ligand for MA, and which can form unusual lipid phases (Li and Schick, 2000; Fuller et al., 2003). Because of this, we have reinvestigated the membrane binding organization of HIV-1 MA proteins under physiological salt conditions, using membranes composed of phosphatidylcholine (PC), cholesterol, and PI[4,5]P2. Under these conditions we now show that myristoyated MA proteins (Myr-MA), as well as proteins composed of Myr-MA and its neighbor Gag capsid (CA) domain (Myr-MACA), assemble as hexamers of trimers. Our results support a model in which each MA trimer contributes to three separate hexamer rings, and MA proteins are positioned roughly above CA N-terminal domain (NTD) hexamers, that also are linked via CA C-terminal domain (CTD) contacts. This model suggests that the shells of immature HIV-1 virions are stabilized tightly by multiple Gag domain contacts, and has implications for how Env proteins assemble into virus particles.
Previous experiments showed that HIV-1 MA proteins bind to PS-rich membranes, and organize on them as hexamers rather than trimers (Alfadhli et al., 2007). However, evidence indicates that HIV-1 MA preferentially binds to membranes containing PI[4,5]P2 and to short chain PI[4,5]P2 analogues (Ono et al., 2004; Saad et al., 2006, 2007; Chukkapalli et al., 2008). Consequently, we decided to examine how myristoylated MA (Myr-MA) proteins organize on PI[4,5]P2-containing membranes, using a lipid monolayer system (Barklis et al., 1997, 1998; McDermott et al., 2000; Mayo et al., 2002, 2002b, 2003; Huseby et al., 2005). Interestingly, we found that Myr-MA proteins assembled readily onto PC membranes containing 20% PI[4,5]P2 (w/w) and 20% cholesterol (w/w), at higher salt concentrations (150 mM) than we observed previously (Figure 1A). Moreover, we found that calculated diffraction patterns (Fourier transforms) displayed as power spectra of Myr-MA proteins bound to PI[4,5]P2 membranes (Figure 1B) differed from patterns of proteins assembled on PS-cholesterol (Alfadhli et al., 2007) membranes. Notably, the unit cell sizes were smaller (1/9 nm versus 1/9.6 nm), and the 1,2 reflections were markedly brighter. When Fourier transform reflection amplitude and phase data were back-transformed assuming no symmetry constraints (p1) to obtain a Fourier filtered two-dimensional (2D) projection image of Myr-MA bound to PI[4,5]P2 membranes (Figure 1C), proteins (in white) appeared to organize as trimers, with six trimers surrounding protein-free hexamer holes.
To validate our initial observations, data collected from eleven separate PI[4,5]P2 membrane-bound Myr-MA 2D crystal images were examined. As detailed (Table 1), unit cell sizes were consistent (a = 90.3 ± 4.3Å, b = 90.1 ± 2.1Å, γ = 120.9 ± 1.2°). When phases of symmetry-related reflections were analyzed, the crystals were found to be compatible with trigonal (p3) or hexagonal (p6) symmetry, as evidenced by the low (10–20°) phase residuals of their space group fits (Table 1). Consequently, amplitude and phase data from the eleven transforms were merged to yield averaged images, assuming p3 or p6 symmetry. Out to 22Å resolution, the consistency of averaging was excellent, with a phase residual of 15–16°, and both merges were 100% complete (Table 1). Not surprisingly, the back-transformed images of the p3 (Figure 2A) and p6 (Figure 2B) merges appeared nearly identical, and similar to our initial results. As viewed perpendicularly from the membrane sides, the Myr-MA proteins appeared to organize as hexamer rings of trimers, with subunits of each trimer contributing to separate hexamer rings (Figure 2).
While the appearance of trimer units in 2D projection images (Figures 1–2) strongly suggested that the units corresponded to trimers observed in HIV-1 MA 3D crystals (Hill et al., 1996), we wished to compare the units more thoroughly. To do so, data were collected from 29 tilted and untilted PI[4,5]P2-bound Myr-MA 2D crystals so as obtain a 3D density map that could be compared with the atomic model. Data were merged as described in the Materials and Methods yielding a 3D map out to 22Å resolution parallel to the membrane, and 37.5Å resolution perpendicular to the membrane. Hand-fitting of an atomic model of a HIV-1 MA trimer unit from x-ray diffraction studies demonstrated a good fit (Figure 3). When viewed from the membrane side down to MA (Figure 3A), with the membrane tilted slightly toward the reader (Figures 3B–C), or parallel to the membrane at a contour of 2σ (Figure 3E), the model appears to match the density envelope well. When viewed parallel to the membrane at a contour of 1.5σ (Figure 3D), the volume appears stretched perpendicular to the membrane as a consequence of the lower resolution along this axis, and shows membrane-proximal knobs at the top of the figure that may represent exposed myristate groups. In any case, the trimer units in our 3D model of PI[4,5]P2 membrane-bound Myr-MA proteins appear compatible with trimers observed in x-ray studies (Hill et al., 1996).
Because HIV-1 MA domains are tethered to CA domains in the PrGag proteins that direct immature virus assembly, it was of interest to examine how Myr-MACA proteins assemble on PI[4,5]P2-containing membranes. To do so, we endeavored to examine the organization of PI[4,5]P2 membrane-bound proteins via the methods outlined above. The 2D crystals obtained with Myr-MACA proteins were smaller than those for Myr-MA, yielding power spectra of Fourier transforms with less intense reflections (data not shown). The unit cell dimensions (a = 87.2 ± 2.1Å; b = 86.4 ± 2.0Å) of Myr-MACA crystals also were slightly smaller than those for Myr-MA, but phase data were consistent with p3 or p6 space groups (Table 1). To generate projection images of Myr-MACA proteins assembled on PI[4,5]P2 membranes, data from seven images were merged to 22Å resolution, giving phase residuals of 12–14°. The results of our Myr-MACA p6 merge are shown in Figure 4A, and the back-transformation of the p3 merge appeared indistinguishable (data not shown). As illustrated (Figure 4A), the proteins (in white) organized as hexamer rings, with clear evidence of trimer units. Because our Myr-MACA crystals were not of sufficient quality for informative 3D reconstructions, we scaled our Myr-MA 2D projection image to the Myr-MACA image and calculated a coarse difference map (Figure 4B), emphasizing areas (in white) likely to be occupied by CA but not MA in the projection image. The difference map, as well as simple comparison of Myr-MACA (Figure 4A) versus Myr-MA (Figure 2B) projections suggest that MA largely aligns above CA units, but that CA may cover regions closer to the centers of hexamer holes than are occupied by MA trimers. Implications of our results are discussed below.
HIV-1 MA preferentially binds to PI[4,5]P2- and cholesterol-rich membranes (Nguyen and Hildreth, 2000; Lindwasser and Resh, 2001; Ono and Freed, 2001; Graham et al., 2003; Holm et al., 2003; Ono et al., 2004; Gomez and Hope, 2006; Saad et al., 2006, 2007; Chukkapalli et al., 2008), and forms trimer units in 3D crystals and in solution (Hill et al.,1997; Morikawa et al., 1998; Tang et al., 2004). Our results show that Myr-MA organizes on 60% PC, 20% PI[4,5]P2, 20% cholesterol membranes as hexamer rings of trimers (Figures 1–3). We believe that MA trimers did not form in previous work with PS-cholesterol membranes (Alfadhli et al., 2007) due to the high concentrations of PS used, the propensity of PS to form unusual lipid phases (Li and Schick, 2000; Fuller et al., 2003), the low ionic strength of the incubations, or a combination of these effects. As compared with the lipid composition of actual HIV-1 virions, the 20% w/w cholesterol concentration we have used in our current study corresponds to a lipid mole fraction of about 35%, within the range observed for virions (Brugger et al., 2006; Chan et al., 2008). Our experimental PI[4,5]P2 mole fraction was about 12%, but we also observed similar, albeit smaller Myr-MA crystals with PI[4,5]P2 mole fractions as low as 6%, which compares favorably to the 2.6–7.3% PIP2 mole fraction observed in purified retrovirus particles (Chan et al., 2008). Nevertheless, the lipid mix we have employed is not a perfect replica of the HIV-1 lipidome, and our results must be interpreted with that caveat.
Because PrGag proteins rather than MA proteins assemble immature virus particles, it might be argued that the organization of MA proteins is strictly relevant to mature virions. However, we observed a similar pattern with membrane-bound Myr-MACA proteins (Figure 4), suggesting that hexamers of HIV-1 trimers occur in immature virus particles. Interestingly, the arrangements of Myr-MA and Myr-MACA proteins on PI[4,5]P2-containing membranes look remarkably similar to the mature-type pattern of HIV-1 CA proteins determined from 2D crystals of flattened spheres (Ganser-Pornillos et al., 2007). Although packing differences between immature and mature Gag lattices are expected (Nermut et al., 1994, 1998; Barklis et al., 1998; Li et al., 2000; Mayo et al., 2003; Briggs et al., 2003, 2004; Huseby et al., 2005; Ganser-Pornillos et al., 2007), for comparative purposes, it is useful to model MA trimers onto the CA lattice, based on our results (Figure 5). As illustrated and as described by Ganser-Pornillos et al. (2007), CA NTDs (green) form hexamer rings via helix 1–3 interactions, and their respective CTDs (blue) associate with neighbor NTDs that are positioned clockwise around hexameric holes, when viewed from perpendicularly from the membrane side. Consistent with previous results, the CTDs interconnect hexamer rings via dimer contacts at two-fold axes of symmetry (yellow ovals). MA trimer units (magenta) fit onto the CA map in an arrangement we have proposed previously (Huseby et al., 2005), forming a second set of links between rings at threefold symmetry axes (yellow triangles). At this point, we have no data to suggest that MA domains do not align directly above their respective (covalently attached) NTDs, but given the CA NTD-CTD precedent (Lanman et al., 2003, 2004; Ganser-Pornillos et al., 2007), it is important to emphasize that alternatives are possible.
The hexameric arrangement of HIV-1 MA trimers on PI[4,5]P2-containing membranes implies consistent contacts between trimers, and supports the notion that immature HIV-1 particles are very stable (Wang and Barklis, 1993; Swanstrom and Wills, 1997; Holm et al., 2003; Huseby et al., 2005), due to multiple Gag interactions thoughout the protein shells. Another implication is that MA interactions with the cytoplasmic tails of HIV-1 Env trimer units (Sanders et al., 2002; Zhu et al., 2003; Yang et al., 2005) occur at hexameric holes. Assuming this is the case, at least two degrees of freedom (offset by 60°) may be allowed for accommodation of Env trimers into Gag lattices. However, it is important to note that the availability of Gag hexameric holes is not the only determinant of Env incorporation into virions: the process may be dependent on the cellular protein TIP47; estimates indicate that not all cage holes can be occupied by Env trimers; and ratios of Env to Gag in HIV-1 particles are lower than might be predicted (Forster et al., 2000; Zhu et al., 2003; Lopez-Verges et al., 2006). Another caveat is that tomographic examination of immature and mature HIV-1 particles has yet to reveal an extensive alignment of the MA layer (Benjamin et al., 2005; Briggs et al., 2006, 2006b; Wright et al., 2007). This discrepancy could be due to lattice disassembly during virus sample preparation (Carlson et al., 2008), or to difficulties in identifying small ordered MA layers within individual virions via tomography. Despite these caveats, our studies demonstrate how Myr-MA and Myr-MACA proteins are ordered on PI[4,5]P2-rich membranes, and should facilitate further study on the matrix protein’s associations with membranes and other proteins.
Myr-MA and Myr-MACA proteins were expressed in E. coli strain BL21 (DE3)/pLysS (Novagen) along with yeast NMT from pET-11a-based vectors kindly provided by Michael F. Summers (Univ. of Maryland Baltimore County; Tang et al., 2004). Proteins were expressed and purified under conditions which yield only traces of unmyristylated species (Tang et al., 2004; Alfadhli et al., 2007). The purified proteins were desalted by buffer exchange in Sephadex G25 spin columns in 10 mM sodium phosphate (pH 7.8), supplemented with β-mercaptoethanol (BME; 1 mM final), aliquoted and stored at −80°C under nitrogen gas. Protein purities were evaluated after fractionation by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE; Barklis et al., 1997; Mayo et al., 2003; Huseby et al., 2005) by Coomassie blue staining (Alfadhli et al., 2005; Huseby et al., 2005) and immunoblotting (Barklis et al., 1998; Huseby et al., 2005). Immunoblot detection of Myr-MACA proteins used a primary anti-HIV CA monclonal antibody (Hy183; Alfadhli et al., 2005), while for Myr-MA proteins, a sheep antiserum (#286; lot 1DV-010) to MA obtained from Dr. Micheal Phelan via the NIAID AIDS Research and Reference Reagent Program was employed. Myr-MACA proteins were purified and used at 0.3–1.0 mg/ml, while concentrations of Myr-MA proteins were 1.0–2.5 mg/ml.
Crystallizations were performed with 7.5 ul of protein plus 2.5 ul of 4x crystallization buffer (50 mM sodium phosphate, 10 mM sodium acetate [pH 7.6], 600 mM NaCl, 40% glycerol [vol/vvol], 5 mM BME): Myr-MA incubations used sodium phosphate at a pH of 7.8, while Myr-MACA incubations were at pH 8.3. For each protein mixture, ten ul drops were deposited onto glass slides and overlayed with 1 ul of lipid mix composed of 150 ug/ml egg PC (Avanti), 50 ug/ml brain PI(4,5)P2 (Avanti), and 50 ug/ml cholesterol (Sigma) in 1:1 chloroform:hexane. Incubations were performed for 16 h at 25°C in sealed 15 cm diameter plastic dishes, humidified with blotter paper wetted with 4 ml water. After incubations, samples were lifted onto lacey EM grids (Ted Pella), rinsed 30 sec on drops of distilled water, stained 45–60 sec with 1.33% (wt/vol) uranyl acetate, wicked and dried.
Samples were viewed on a Philips CM120 transmission electron microscope (TEM) under low dose conditions at defocus values between 200 nm and 1500 nm, in which the first zero of the contrast transfer function (CTF) is beyond 22 Å resolution. Images were recorded either on negatives (Kodak SO163) or as 14-bit grayscale images in Gatan Digital Micrograph 3 (DM3) format on a 1024 × 1024 pixel Gatan 794 CCD multi-scan (MSC) camera. For CCD images, Gatan Digital Micrograph 3.4.0 software was used to convert DM3 format raw images to eight-bit grayscale TIF images at 5.24 Å/pixel, while negatives were digitized using a Nikon Super Coolscan 8000 scanner to 2.04 Å/pixel.
Image analysis steps followed previously described procedures (Amos et al., 1992; Crowther et al., 1996; Barklis et al., 1997, 1998; McDermott et al., 2000; McDermott and Barklis 2002; Mayo et al., 2002a, 2002b, 2003). Tagged image file format (TIF) images were converted to MRC format and processed using the ICE suite of MRC programs (Crowther et al., 1996). Boxed images were Fourier transformed using the 2DFFT function of ICE, transform (TNF) files were examined using SPECTRA (Hardt et al., 1996), and reflections were indexed using the SPECTRA interface. Indexed lattices were unbent and APH format text files were produced by the LATREF, UNBENDA, and MNBOX functions of SPECTRA (Hardt et al., 1996).
Unit cell size parameters were calculated from the positions of 1,0 and 0,1 reflections in power spectra, and were averaged from zero tilt images for Myr-MA and Myr-MACA crystals. Space group analysis to 22Å resolution was calculated using ALLSPACE (Crowther et al., 1996; Hardt et al., 1996) in 3° phase origin search steps, and average phase residuals for space group fits were derived from these data. For merging of untilted image files to obtain average 2D projection images, p3 or p6 symmetry was applied as indicated, and files with the lowest phase residuals and highest number of comparisons for each space group were chosen as references. Merging was performed using the TILT_ORG (McDermott and Barklis, 2002) interface running ORIGTILT_B (Crowther et al., 1996) on MyrMA or MyrMACA APH files. Merges to 22Å resolution used reflections of IQ = 6 (Crowther et al., 1996) or better and phase origin search steps of 3°. Merge completeness and phase residual values were determined as described previously (Mayo et al., 2003). For calculation of average projection maps, amplitudes and phases for each p3 or p6 reflection were vector-averaged using APH_EDIT (McDermott and Barklis, 2002). For image rendering, averaged p3 or p6 reflection values were used to generate corresponding p3- or p6-symmetrized p1 APH files, which were back-transformed using APH_EDIT running the MRC CREATE_TNF, FFTRANS, AND ICE_SKEW commands (Crowther et al., 1996; McDermott and Barklis, 2002). The same approach was utilized for generation of Fourier filtered projection images from single APH files, assuming no symmetry constraints (p1) as in Figure 2C. To obtain a coarse difference map between Myr-MACA and Myr-MA projection images, the images were scaled, and Myr-MA projection image brightness values were subtracted from MyrMACA values using the Adobe PhotoShop LAYER/DIFFERENCE operation.
For preparation of the Myr-MA 3D map, 29 files from images tilted between −45° and +45° were merged as above, assuming p6 symmetry, a z window of 0.002 Å−1, and a z thickness of 200.0Å. Phase origin search steps initially were 3.0°, and then refined to 0.5°. Tilt angles were taken from reported TEM goniometer values and confirmed using EMTILT (Shaw and Hills, 1981), while the magnification-dependent tilt axis to image x axis angle was determined to be 138° over the course of all images taken. For the 3D reconstruction, a phase residual of 17.5° was obtained, and merged results were exported as 3D data in the ORIG text file format, containing h and k indices, z* values, amplitudes, phases, film numbers, and resolutions for each reflection included in the merge. The ORIG file was converted to an HKL file by combining all h, k points with similar z* values into single hkl points (McDermott et al., 2000, 2002; Mayo et al., 2002a, 2002b, 2003), and reflections contributing to the reconstruction were of IQ ≤6 (Crowther et al., 1996). The HKL file was filtered to remove h, k (xy plane) reflections of better than 22Å resolution, and to remove l (z dimension) values of better than 37.5Å resolution. The filtered HKL file to these resolutions was 63% complete, and was converted to a XTALVIEW (McRee, 1992) PHS file and back-transformed to give a 3D FSFOUR MAP file, which was viewed using XFIT (McRee, 1992). For model fitting, a trimer unit from the HIV-1 MA PDB 1HIW (Hill et al., 1996) was input into XFIT and hand-fitted to the density map. The rotational matices for the volumes in Figure 3 were as follows: A, x[1, 0, 0], y[0, 1, 0], z[0, 0, 1]; B–C, x[0.9980, 0.0031, −0.0054], y[−0.0585, 0.9511, 0.3033], z[0.0243, −0.3024, 0.9520]; D–E, x[−0.0872, 0.9962, 0], y[0, 0, 1], z[0.9962, 0.872, 0]. For Figure 5, the HIV-1 MA trimer unit was aligned onto a CA NTD (PDB 1GWP; Gitti et al., 1996) plus CTD (PDB 1A43; Gamble et al., 1997) map that was prepared based on the observations of Ganser-Pornillos et al. (2007).
We are grateful to Michael F. Summers for the bacterial Myr-MA and Myr-MACA expression vectors. We also appreciate the help and support of Ben Kukull, Claudia Lopez, Hank McNett, Amelia Still, and Mike Webb. This research was supported by National Institutes of Health grants GM060170 and AI071798 to EB.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.