Restriction of HIV-1 by Rhesus TRIM5α Is Governed by Alpha Helices in the Linker2 Region

ABSTRACT

TRIM5α proteins are a potent barrier to the cross-species transmission of retroviruses. TRIM5α proteins exhibit an ability to self-associate at many levels, ultimately leading to the formation of protein assemblies with hexagonal symmetry in vitro and cytoplasmic assemblies when expressed in cells. However, the role of these assemblies in restriction, the determinants that mediate their formation, and the organization of TRIM5α molecules within these assemblies have remained unclear. Here we show that α-helical elements within the Linker2 region of rhesus macaque TRIM5α govern the ability to form cytoplasmic assemblies in cells and restrict HIV-1 infection. Mutations that reduce α-helix formation by the Linker2 region disrupt assembly and restriction. More importantly, mutations that enhance the α-helical content of the Linker2 region, relative to the wild-type protein, also exhibit an increased ability to form cytoplasmic assemblies and restrict HIV-1 infection. Molecular modeling of the TRIM5α dimer suggests a model in which α-helical elements within the Linker2 region dock to α-helices of the coiled-coil domain, likely establishing proper orientation and spacing of protein domains necessary for assembly and restriction. Collectively, these studies provide critical insight into the determinants governing TRIM5α assembly and restriction and demonstrate that the antiviral potency of TRIM5α proteins can be significantly increased without altering the affinity of SPRY/capsid binding.

IMPORTANCE Many members of the tripartite motif (TRIM) family of proteins act as restriction factors that directly inhibit viral infection and activate innate immune signaling pathways. Another common feature of TRIM proteins is the ability to form protein assemblies in the nucleus or the cytoplasm. However, the determinants in TRIM proteins required for assembly and the degree to which assembly affects TRIM protein function have been poorly understood. Here we show that alpha helices in the Linker2 (L2) region of rhesus TRIM5α govern assembly and restriction of HIV-1 infection. Helix-disrupting mutations disrupt the assembly and restriction of HIV-1, while helix-stabilizing mutations enhance assembly and restriction relative to the wild-type protein. Circular dichroism analysis suggests that that the formation of this helical structure is supported by intermolecular interactions with the coiled-coil (CC) domain in the CCL2 dimer. These studies reveal a novel mechanism by which the antiviral activity of TRIM5α proteins can be regulated and provide detailed insight into the assembly determinants of TRIM family proteins.

INTRODUCTION

Several cellular proteins, termed restriction factors, provide intrinsic immunity against a wide range of viruses by interfering with various stages of the viral life cycle (1, 2). One of the extensively studied components of this intrinsic immunity is the restriction factor TRIM5α, which is a member of the tripartite motif (TRIM) family of proteins (3). TRIM5α inhibits retroviral infection in a species-specific manner. For example, rhesus macaque TRIM5α (rhTRIM5α) potently restricts infection by HIV-1 (4), while human TRIM5α (huTRIM5α) restricts other retroviruses, such as N-tropic murine leukemia virus (N-MLV) and equine infectious anemia virus (EIAV), but exhibits minimal restriction of B-tropic murine leukemia virus (B-MLV) and HIV-1 (5–8). TRIM5 proteins block retroviral replication soon after the viral core enters the cell cytoplasm (4, 9–11). The viral determinants responsible for the susceptibility of retroviruses to TRIM5-mediated restriction have been mapped to the viral capsid (CA) protein (12, 13). In the case of TRIM5α, species-specific changes in the C-terminal B30.2/SPRY domain are known to modulate the spectrum of restriction observed for primate TRIM5α proteins (14–23).

One remarkable aspect of many TRIM family proteins, including TRIM5α, is the ability to self-associate into large protein assemblies in vitro and in cells (3, 4, 24, 25). In cells, TRIM5α forms cytoplasmic accumulations of protein, termed cytoplasmic bodies, and many other TRIM family proteins assemble into similar accumulations in the cytoplasm or nucleus (3, 4, 24). In vitro, TRIM5α has been demonstrated to self-associate to form a lattice with hexameric symmetry, the dimensions of which are superimposable on the organization of the hexameric assemblies of HIV-1 capsid in a mature assembled core (25). We have previously observed rhTRIM5α assemblies forming around individual viral complexes in live-cell imaging experiments (26), suggesting that these cytoplasmic assemblies observed microscopically are the in vivo manifestation of the assemblies previously observed in vitro by Ganser-Pornillos et al. (25).

Numerous domains and regions of rhTRIM5α contribute to its tendency to self-associate. The coiled-coil (CC) domain is known to be required for the formation of TRIM5α dimers (3, 16, 27–30). The BBox2 domain has been shown to mediate a higher-order multimerization (31, 32) that is required for retroviral CA binding and restriction (27, 29–35). We have previously demonstrated that the Linker2 (L2) region connecting the CC domain to the C-terminal SPRY domain facilitates the formation of cytoplasmic assemblies (36). As is the case with dimerization and BBox2-mediated self-association, the ability of rhTRIM5α to form cytoplasmic assemblies is also critical for its ability to restrict HIV-1 infection, as mutations that disrupt the ability to form cytoplasmic assemblies also abrogate restriction (36). This suggests that the L2 region of rhTRIM5α contains determinants that govern lattice assembly. Previous models of TRIM5α assembly proposed that the L2 regions of adjacent parallel dimers self-associate to organize TRIM5α assembly around a viral core (25, 37). However, Sanchez et al. recently reported that the crystal structure of the TRIM25 dimeric unit exists as an elongated antiparallel dimer in which the L2 region doubles back along the axis of the CC helix (38), suggesting that the L2 region may be promoting the formation of cytoplasmic assemblies and restriction via a previously unappreciated mechanism.

In this study, we show that the ability of rhTRIM5α to form cytoplasmic assemblies and restrict HIV-1 infection is governed by α-helices present in the L2 region. The helical content of the L2 region is reduced in a mutant that does not restrict HIV-1 infection or form cytoplasmic assemblies. Conversely, mutations that stabilize helices in the L2 region restrict HIV-1 and form cytoplasmic assemblies more efficiently than the wild-type (wt) rhTRIM5α protein. We show that the presence of the CC domain stabilizes helices in the L2 region. Homology modeling of the TRIM5 dimer and molecular dynamics simulations reveal that helix-destabilizing mutations prevent residues in the L2 region from stably docking with residues of the α-helical CC domain.

MATERIALS AND METHODS

Recombinant DNA constructs.

The wild-type rhTRIM5α plasmid was a kind gift from Joseph Sodroski (Harvard School of Public Health). The hemagglutinin (HA)-tagged rhTRIM5α and yellow fluorescent protein (YFP)-labeled rhTRIM5α constructs were generated as previously described (36). Single- and triple-alanine mutations were introduced into wt rhTRIM5α by using splicing by overlap extension (SOE) PCR, as previously described (36). To generate the 6×His-tagged CCL2 peptides, the CCL2 fragments of wt rhTRIM5α and its L2 mutants were cloned into the pET-15b vector by using the NdeI and BamHI restriction sites. Synthetic L2 peptides (about 98% pure) were obtained from Bio-Synthesis Inc.

Cell culture, virus production, stable cell lines, and infectivity assays.

HeLa and 293T cells were cultured in complete Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal bovine serum, penicillin (final concentration, 100 U/ml), and streptomycin (final concentration, 100 μg/ml). Vesicular stomatitis virus G protein (VSV-G)-pseudotyped HIV-1 reporter virus was produced by transfecting 293T cells as previously described (26).

Virus infectivity was assessed by infecting equivalent numbers of cells in a 24-well plate, and green fluorescent protein (GFP) expression was determined at 48 to 72 h postinfection by using a FACSCanto II flow cytometer (Becton, Dickinson). Vectors expressing YFP- or HA-tagged wt rhTRIM5α or the L2 mutants were made in a similar way, by transfecting 293T cells with the respective TRIM5 plasmids along with VSV-G and pCig-B.

HeLa cells stably expressing the respective epitope-tagged proteins were generated by G418 (400 μg/ml) selection at 48 h posttransduction. The stable cell lines were then analyzed by immunofluorescence and Western blotting, and cell lines expressing comparable amounts of protein were chosen for subsequent experiments.

Immunofluorescence.

HeLa cells stably expressing fluorescently labeled TRIM5α proteins were plated onto fibronectin-treated coverslips, allowed to adhere, and fixed for 5 min with 3.7% formaldehyde (Polysciences) in 0.1 M PIPES [piperazine-N,N′-bis(2-ethanesulfonic acid)] (pH 6.8) (Sigma). Cells were stained with 4′,6-diamidino-2-phenylindole (DAPI) and Texas Red phalloidin in 1× phosphate-buffered saline (PBS) containing 0.1% Triton X-100 (Sigma) and 0.01% NaN₃. Images were collected with a DeltaVision microscope (Applied Precision) equipped with a digital camera (CoolSNAP HQ; Photometrics), using a 1.4-numerical-aperture objective lens, and were deconvolved with SoftWoRx deconvolution software (Applied Precision).

Image analysis.

z-stack images of each cell line were acquired by using identical acquisition parameters. The coverslips were coded such that the individual acquiring the images did not know the identity of the cell lines. Deconvolved images were analyzed for YFP-rhTRIM5α cytoplasmic bodies and cortical actin, stained with Texas Red phalloidin, by using the Surface Finder function of the Imaris software package (Bitplane). Surfaces for cytoplasmic bodies in all samples analyzed were identified by using defined fluorescence intensity and size criteria.

Forty-five individual images of cells expressing YFP-rhTRIM5α (wt and L2 mutants) were analyzed. To calculate the relative expression levels of YFP-labeled rhTRIM5α proteins in each cell, a three-dimensional (3D) surface overlapping a single cell was created by using the surface finder function of Imaris. The sum intensity in the fluorescein isothiocyanate (FITC) channel (corresponding to the total YFP fluorescence in the cell) and the surface volume (corresponding to the cell volume) were determined. The relative YFP-rhTRIM5α protein expression level in each cell was then calculated by dividing the sum intensity in the FITC channel by the total surface volume. Background intensity, derived similarly from untransduced HeLa cells, was calculated and used to remove background fluorescence. The number of cytoplasmic bodies in each cell was determined and normalized by the relative protein expression level in the cell. The data were plotted by using Prism (Graphpad Software Inc.) for statistical analysis. Dunnett's multiple-comparison test was used to determine the statistical significance of the differences between cell lines.

Western blotting and Coomassie staining.

Whole-cell lysates were prepared by lysing cells with NP-40 lysis buffer (100 mM Tris [pH 8.0], 1% NP-40, 150 mM NaCl) containing a protease inhibitor cocktail (PIC) (Roche) for 15 min on ice. A Coomassie Plus Bradford assay (Thermo Scientific) or the absorbance at 280 nm was used to determine total protein concentrations. An equal amount of protein was loaded onto a 10% polyacrylamide gel for SDS-polyacrylamide gel electrophoresis (PAGE). Proteins were transferred onto nitrocellulose membranes and detected by incubation with anti-GFP (Covance) or anti-HA (clone 3F10) conjugated to horseradish peroxidase (HRP) (Roche). Secondary antibodies conjugated to HRP (Thermo Scientific) were used where necessary, and antibody complexes were detected by using SuperSignal West Femto chemiluminescent substrate (Thermo Scientific). Chemiluminescence was detected by using the UVP EC3 imaging system (UVP LLC) or the Bio-Rad Chemidoc system.

For Coomassie staining, protein samples from bacterial cell lysates or purified proteins were prepared by adding 2× or 6× SDS sample buffer and boiled for 10 min at 100°C. Protein samples were separated on polyacrylamide gels at 100 V. The gels were then fixed in a colloidal Coomassie fixative (45% methanol and 1% acetic acid in Milli Q water) for at least 1 h and stained with Coomassie stain (170 g ammonium sulfate, 1 g Coomassie G250, 0.5% acetic acid, and 34% methanol in Milli Q water) for at least 2 h. Stained gels were washed with deionized water to remove excess stain, and the gels were imaged by using the Bio-Rad Chemidoc system.

Glutaraldehyde cross-linking assay.

Glutaraldehyde cross-linking assays were performed as previously described (39). Briefly, cell lysates were incubated on ice for 30 min and centrifuged at 3,000 rpm for 1 min. Clarified lysates were incubated with 0, 1, 2, and 4 mM glutaraldehyde for 5 min at room temperature. The glutaraldehyde was saturated by the addition of 1 M glycine. Purified peptides were treated in a similar manner. The cross-linked proteins were then subjected to SDS-PAGE using 4%-to-15% gradient Tris-HCl gels (Ready Gels; Bio-Rad) and subsequent Western blot analysis or Coomassie staining.

Protein expression and induction.

Freshly transformed BL21(DE3) cells were grown in 0.5 to 1 liter of Luria broth containing 100 μg/ml carbenicillin (Invitrogen) until the optical density at 600 nm (OD₆₀₀) reached 0.6. The bacterial cultures were then induced to express wt or L2 mutant rhTRIM5α CCL2 peptides by adding 1 mM isopropyl β-d-1-thiogalactopyranoside (IPTG) (Invitrogen) and shaking the cultures for 5 h at 37°C. To purify 6×His-tagged CCL2 peptides, bacterial pellets were lysed in a solution containing 50 mM Na₂HPO₄, 500 mM NaCl, 10 mM imidazole, 1% Triton X-100, 0.5 mg/ml lysozyme (Sigma), and a protease inhibitor cocktail (PIC) (Roche), followed by sonication. The lysates were then centrifuged at 13,000 rpm at 4°C for 30 min. The supernatant was discarded, and the pellet was washed extensively with lysis buffer and once with a buffer containing 50 mM Na₂HPO₄, 500 mM NaCl, and 10 mM imidazole. The pellets were then resuspended in a buffer containing 40 mM Tris, 150 mM NaCl, 8 M urea, and 10 mM imidazole (pH 8.0). Talon metal affinity resin (Clontech) was added to the resuspended pellet, and the mixture was incubated at 4°C for 1 to 2 h with gentle mixing to facilitate binding of the His-tagged proteins to the resin. The mixture was passed through a 2-ml Talon disposable gravity column (Clontech) twice. The flowthrough was discarded, and the resin was washed with a buffer containing 40 mM Tris, 300 mM NaCl, and 8 M urea (pH 8.0). The 6×His-tagged proteins were eluted from the resin by using an elution buffer (40 mM Tris, 150 mM NaCl, 8 M urea, 300 mM imidazole). Protein concentrations were determined by measuring the absorbance at 280 nm. The protein fractions were analyzed by Coomassie staining of SDS-PAGE gels, and the fractions with the highest purity were pooled and used for further analysis.

Dialysis and concentration of purified proteins.

The purified proteins were diluted to a concentration of 100 μg/ml and dialyzed by using Slide-A-lyzer dialysis cassettes (10,000 molecular weight cutoff [MWCO] or 3,500 MWCO) (Thermo Scientific) in 2 liters of dialysis buffer overnight at 4°C with gentle mixing. The dialyzed peptides were then concentrated by using centrifugal filter units (10,000 MWCO or 3,500 MWCO) (Millipore) according to the manufacturer's instructions.

In vitro assembly of CA-NC complexes.

HIV-1 CA-NC particles were assembled in vitro by diluting the CA-NC protein to a concentration of 0.3 mM in a solution containing 50 mM Tris-HCl (pH 8.0), 0.5 M NaCl, and 2 mg of DNA oligo(TG)₅₀/ml. The mixture was incubated at 4°C overnight and centrifuged at 8,600 × g for 5 min. The pellet was resuspended in assembly buffer (50 mM Tris-HCl [pH 8.0], 0.5 M NaCl) at a final protein concentration of 0.15 mM and stored at 4°C until needed.

Binding of rhTRIM5α variants to HIV-1 capsid complexes.

293T cells were transfected with plasmids expressing wild-type or mutant rhTRIM5α proteins. At 48 h after transfection, cell lysates were prepared as follows. Previously washed cells were resuspended in hypotonic lysis buffer (10 mM Tris [pH 7.4], 1.5 mM MgCl₂, 10 mM KCl, 0.5 mM dithiothreitol [DTT]). The cell suspension was frozen, thawed, and incubated on ice for 10 min. The lysate was centrifuged at full speed in a refrigerated Eppendorf microcentrifuge (∼14,000 × g) for 5 min. The supernatant was supplemented with a 1/10 volume of 10× PBS and then used for the binding assay. In some cases, samples containing the rhTRIM5α variants were diluted with extracts prepared from untransfected cells in parallel. To test binding, 5 μl of CA-NC particles assembled in vitro was incubated with 200 μl of the cell lysate at room temperature for 1 h. A fraction of this mixture was stored (input). The mixture was spun through a 70% sucrose cushion (70% sucrose, 1× PBS, and 0.5 mM DTT) at 100,000 × g in an SW55 rotor (Beckman) for 1 h at 4°C. After centrifugation, the supernatant was carefully removed, and the pellet was resuspended in 1× SDS-PAGE loading buffer (pellet). The level of rhTRIM5α proteins was determined by Western blotting with an anti-HA antibody, as described above. The level of HIV-1 CA-NC protein in the pellet was assessed by Western blotting with an anti-p24 CA antibody.

CD spectroscopy.

UV spectra (190 to 260 nm) of the purified CCL2 or synthetic L2 peptides were measured by circular dichroism (CD) spectroscopy using a Jasco J-810 circular dichroism spectrometer. CD spectroscopy for the L2 peptides was performed in the presence and absence of 10% trifluoroethanol (TFE). Spectra were collected in triplicate in 1-mm-path-length cuvettes at 20°C, the resulting scans were averaged, and the buffer signal was subtracted. The specific ellipticities for each peptide spectrum were calculated, and computation of the relative amount of helix was performed by using ContinLL.

FRET analysis.

293T cells were plated into DeltaT dishes at 50% confluence and transfected with a total of 375 ng of YFP- or cyan fluorescent protein (CFP)-tagged CCL2 rhTRIM5α or W117E-BCCL2 rhTRIM5α constructs in various fluorescence resonance energy transfer (FRET) combinations. At 18 h posttransfection, the FRET efficiency (E) was measured by using the following equation:

$$E=\frac{I_{\mathrm{D}\mathrm{A}}-a\hspace{0.17em}(I_{\mathrm{A}\mathrm{A}})-d(I_{\mathrm{D}\mathrm{D}})}{I_{\mathrm{D}\mathrm{A}}-a(I_{\mathrm{A}\mathrm{A}})+(G-d)(I_{\mathrm{D}\mathrm{D}})}$$

where I_DD is the intensity of the fluorescence emission detected in the donor channel (438-nm excitation wavelength [Ex]/470-nm emission wavelength [Em]), I_AA is the intensity of the fluorescence emission detected in the acceptor channel (513-nm Ex/559-nm Em), and I_DA is the intensity of the fluorescence emission detected in the FRET channel (438-nm Ex/559-nm Em). The “a” and “d” cross talk coefficients were determined from acceptor-only and donor-only samples, respectively. The cross talk coefficients were calculated for a by using I_DA/I_AA and for d by using I_DA/I_DD. We obtained a d value of 0.393 for CFP and an a value of 0.275 for YFP. G was calculated by using a construct provided by Seth Robia, which contains 5 glycines flanked by CFP and YFP. YFP was progressively photobleached while the increase in CFP intensity was measured. Once YFP was completely photobleached, we measured the amount of donor recovery and determined G as the ratio of sensitized emission to the corresponding amount of donor recovery, which was 3.02.

Homology modeling and molecular dynamics simulations.

The SWISS-MODEL toolkit (40) was utilized to generate a homology model of the CCL2 fragment of rhTRIM5α, using the recently reported structure of the TRIM25 CCL2 dimer (38) as a template.

Simulations were generated by using the GROMACS program (41, 42) with the CHARMM 27 force field (43–45) and TI3P water model (46). Energy minimization was performed on the structures by using the steepest-descent method for 1,000 steps, and each model was then solvated in a rectangular water box with a minimum of 20 Å from the surface of the protein to the edge of the solvent box. Na⁺ and Cl⁻ ions were added to the solution to neutralize the charge of the system and to produce an ion concentration of 150 mM. The particle mesh Ewald (PME) method (47, 48) was used to describe long-range electrostatic interactions. Molecular dynamics simulations were carried out with an integration time step of 2 fs. To reach the target temperature (300 K) and pressure (10⁵ Pa), the method of Berendsen et al. (49) was used, with a relaxation time of 0.1 ps. After a 100-ps equilibration, production simulations were performed with the isobaric-isothermal (NPT) ensemble by using a Nose-Hoover thermostat (50, 51) and a Parrinello-Rahman barostat (52, 53), with relaxation times of 1.0 ps. The production run was carried out for 10 ns.

RESULTS

Mutations in the L2 region of rhTRIM5α govern cytoplasmic assembly formation.

In cell lines stably expressing rhTRIM5α, the wt form of the protein localizes to both cytoplasmic assemblies and a more diffuse pool of cytoplasmic protein. We previously found that mutations of some residues in the L2 region of rhTRIM5α prevented the formation of rhTRIM5α cytoplasmic assemblies in cell lines stably expressing these mutants and that mutants which lost the ability to assemble into cytoplasmic assemblies also lost the ability to inhibit HIV-1 infection (36) (Fig. 1A). In subsequent analyses, we found that two of the mutants that we generated in our previous study, TFH269-271AAA and HKN271-273AAA, appear to form cytoplasmic assemblies very efficiently, forming more cytoplasmic assemblies per cell, with little or no diffuse cytoplasmic fraction being visible (Fig. 1B). To ensure that this difference in localization was not due to a difference in the expression levels of these proteins between cell lines, we developed an assay that allows us to measure the number of cytoplasmic puncta in a given cell while accounting for the total TRIM5α protein expression level in that cell. Cells stably expressing YFP-labeled wt or mutant versions of rhTRIM5α were stained for cortical actin by using phalloidin, which allowed us to define the cell periphery. We then generated 3D surfaces around individual cells using the cortical actin stain of these cells, which we used to calculate the volume of the cell as well as the total amount of YFP-TRIM5α fluorescence in the cell (Fig. 1C). Next, we calculated the relative protein concentration in the cell by dividing the total YFP fluorescence by the cell volume. The total number of cytoplasmic bodies per cell was determined as described previously (36) and normalized by the relative protein expression level in the cell (45 images per cell line were analyzed). This analysis revealed that the TFH-AAA and HKN-AAA mutants form more cytoplasmic bodies per cell than does wt rhTRIM5α when normalized for protein expression levels (Fig. 1D).

FIG 1.

Mutation of residues within the L2 region enhances formation of rhTRIM5α cytoplasmic assemblies. (A) Schematic representation of the rhTRIM5α protein showing the RING, BBox, coiled-coil (CC), and B30.2/SPRY domains. The abilities of the indicated YFP-rhTRIM5α proteins to form cytoplasmic bodies and restrict HIV-1 infection are listed. (B) Representative images of HeLa cells stably expressing YFP-labeled wt or L2 variant rhTRIM5α proteins (green) with DAPI stain (blue). Display conditions have been adjusted to reveal the localization of each protein. (C) Texas Red phalloidin was used to visualize cortical actin and create a three-dimensional boundary of the cell. (D) Three-dimensional surfaces were generated around individual cells as described above for panel C, allowing calculation of the intracellular YFP concentration in each cell. The number of cytoplasmic bodies (CBs) in each cell was calculated and normalized to the relative YFP-rhTRIM5α protein expression level in that cell. Both the TFH-AAA and HKN-AAA mutants formed significantly more cytoplasmic assemblies than the wt (P < 0.01). Forty-five z-stack images were collected for each cell line.

Mutations that increase the ability of rhTRIM5α to form cytoplasmic bodies increase the potency of HIV-1 restriction.

Since we previously observed that mutations which disrupt the ability of rhTRIM5α to form cytoplasmic assemblies also abrogate the ability to inhibit HIV-1 infection (36), we reasoned that the TFH-AAA and HKN-AAA mutants, which have an increased tendency to assemble, may exhibit increased HIV-1 restriction abilities. To test this hypothesis, we generated cell lines stably expressing wt rhTRIM5α or the TFH-AAA and HKN-AAA mutants, which were all expressed at similar levels in these cells (Fig. 2A and B). We infected these cells with increasing titers of HIV-1. As we hypothesized, the TFH-AAA and HKN-AAA mutants exhibited increased HIV-1 restriction ability compared to that of wt rhTRIM5α (Fig. 2C). To ensure that this enhanced ability to restrict HIV-1 was reproducible and that the differences that we observed were not owing to the stochastic selection of transduced cells that were more or less permissive to HIV-1 infection, we generated seven independent cell lines expressing wt rhTRIM5α or the TFH-AAA and HKN-AAA mutants. When these cell lines were infected with HIV-1, cells expressing the TFH-AAA and HKN-AAA mutants consistently and reproducibly inhibited HIV-1 more efficiently than did their wt rhTRIM5α counterpart (Fig. 2D). Collectively, these results demonstrate that the tendency of rhTRIM5α to assemble into cytoplasmic assemblies is directly correlated with its ability to restrict HIV-1 infection.

FIG 2.

rhTRIM5α L2 variants with an increased tendency to form cytoplasmic assemblies exhibit increased HIV-1 restriction. (A) Protein levels of rhTRIM5α were analyzed by Western blotting with a monoclonal anti-HA antibody. Actin was used as the loading control. (B) Densitometric analysis of HA-rhTRIM5α expression levels, as shown in panel A. Band intensities for HA and actin were calculated and expressed as percentages of wt HA-rhTRIM5α expression. Results are averages from 3 independent experiments. Error bars represent the standard deviations from 3 experiments. (C) Equivalent numbers of each cell line were infected with increasing titers of HIV-1 expressing GFP. The percentage of GFP-positive cells was determined at 48 h postinfec- tion. (D) HeLa cells were transduced with a vector expressing YFP-labeled wt, TFH-AAA, or HKN-AAA constructs to generate seven individual cell lines. These cells were infected with the indicated dilutions of HIV-1 expressing a GFP reporter, and the percentage of GFP-positive cells was determined as described above for panel C. Error bars represent the standard errors of the means. Differences between wt rhTRIM5α and the L2 variants were analyzed by using one-way analysis of variance (n = 7) (P < 0.01 for the wt versus the TFH mutant and P < 0.01 for the wt versus the HKN mutant). Data are representative of data from 3 independent experiments.

The α-helical content of the L2 region of rhTRIM5α is correlated with the ability to form cytoplasmic assemblies and restrict HIV-1 infection.

The observation that alanine mutagenesis can increase or decrease rhTRIM5α assembly and restriction suggested that we were altering secondary structural elements in the L2 region. To understand how these mutations might be altering secondary structural features present in the L2 region, we expressed and purified peptides comprising the CC domain and the L2 region of rhTRIM5α (amino acids [aa] 134 to 296). Peptides corresponding to wt rhTRIM5α and the RRV-AAA, TFH-AAA, and HKN-AAA mutants were utilized in this study. These peptides were then analyzed by circular dichroism (CD) spectroscopy to determine their secondary structures. We found that the CCL2 peptides possess a predominantly α-helical conformation, as seen by the minima at 222 nm and 208 nm (Fig. 3A). This is not surprising, since the CC domain present in this peptide is strongly predicted to form an α-helix. We also observed that our CCL2 peptides exhibited differences in the magnitude of the specific ellipticity minima at 222 nm and 208 nm, which correspond to differences in helical content in these peptides. Notably, the CCL2 peptide with the RRV-AAA mutation, which abrogates assembly and restriction of full-length rhTRIM5α, exhibited the lowest helical content of the peptides examined. In three experiments, the average helical content of this peptide was found to be 68%. The average helicity of wt CCL2 was observed to be 78%. Notably, CCL2 peptides containing the TFH-AAA and HKN-AAA mutations that exhibit enhanced abilities to assemble and restrict HIV-1 exhibited higher α-helical content than the wt CCL2 peptide, being 85% and 87% α-helical, respectively. As the CC domain comprises 62% of these peptides, these data demonstrate that the L2 region contains α-helical elements. Moreover, the propensity to form cytoplasmic assemblies and restrict HIV-1 infection is correlated with the α-helical content of the CCL2 peptide, as shown in Fig. 3B.

FIG 3.

The L2 region of rhTRIM5α has an α-helix, and the α-helical content is correlated with the cytoplasmic assembly and HIV-1 restriction potency of rhTRIM5α. Shown are data for CD spectroscopy of purified CCL2 and L2 peptides of wt rhTRIM5α and the L2 variants. (A) The far-UV spectra (200 to 240 nm) of the indicated CCL2 peptides were recorded and used to calculate the specific ellipticity curves for each peptide. The average helical content from three independent experiments is indicated below the specific ellipticity curves. (B) Relative α-helical content calculated for the indicated CCL2 peptide from the results of 3 independent experiments. *, P < 0.05, ***, P < 0.001 by one-way analysis of variance (ANOVA) multiple comparisons. Error bars represent the standard errors of the means (SEM) from 3 experiments. (C and D). The far-UV spectra (200 to 240 nm) of L2 peptides of wt rhTRIM5α and the indicated variants were recorded as described above for panel A in the absence (C) and presence (D) of 10% TFE. The specific ellipticity curves for each peptide are shown. Data are representative of data from 3 independent experiments. The average helical contents derived from three experiments are provided in the text.

We also utilized synthetic peptides comprising only the L2 region using the wt and the same panel of mutant peptides. CD analysis of these peptides did not reveal an α-helical secondary structure. Rather, these peptides generated CD spectra consistent with a random coil or unstructured proteins (Fig. 3C). The addition of 10% trifluoroethanol (TFE) to the peptide samples induced a coil-to-helix transition in the L2 peptides, inducing the characteristic minima at 222 nm and 208 nm indicative of α-helices (Fig. 3D). As seen in the case of the CCL2 peptides, the α-helical content of these peptides in the presence of 10% TFE correlated with their ability to assemble and restrict infection. These data demonstrate that the L2 region of rhTRIM5α is not intrinsically α-helical in the absence of the adjacent CC domain, suggesting that the presence of the CC domain somehow stabilizes the α-helices in the L2 region. However, in the presence of TFE, which can stabilize α-helical structures (54), these peptides assumed an α-helical conformation, and under these conditions, the helical content of these peptides corresponds to their ability to induce assembly and restriction in the context of full-length rhTRIM5α.

Mutations in the L2 region that enhance the formation of cytoplasmic assemblies and HIV-1 restriction do not affect TRIM5α self-association or capsid binding in vitro.

Based on the data described above, we hypothesized that mutations in the L2 region that increase or decrease assembly would similarly affect the ability of rhTRIM5α dimers to self-associate. To understand how helices within the L2 region promote rhTRIM5α assembly formation and restriction, we utilized biochemical assays developed to measure the self-association of rhTRIM5α dimers (31, 32, 37). To avoid the complication of heterodimerization, self-association of TRIM5α was measured by mixing cell lysates following centrifugation to remove larger assemblies or aggregates. Lysates containing FLAG-TRIM5α were incubated with cell lysates containing HA-TRIM5α, and the amount of HA-TRIM5α that precipitated with FLAG-TRIM5α was determined. As shown in Fig. 4A, mutations that disrupted or enhanced assembly and restriction of rhTRIM5α did not grossly affect the ability of these proteins to self-associate biochemically.

FIG 4.

L2 mutations that enhance rhTRIM5α assembly and HIV-1 restriction do not affect the ability of rhTRIM5α to self-associate or bind in vitro-assembled HIV-1 capsids. (A) 293T cells were transfected with a FLAG- or HA-tagged version of the indicated rhTRIM5α variants. Clarified cell lysates were incubated, and self-association was measured as the amount of HA-rhTRIM5α that coimmunoprecipitated with FLAG-rhTRIM5α. IP, immunoprecipitation. (B) The indicated CCL2 peptides were cross-linked at room temperature by using increasing amounts of glutaraldehyde for 5 min. The glutaraldehyde was saturated by the addition of 1 M glycine, and cross-linked peptides were visualized by SDS-PAGE and Coomassie staining. (C) In vitro TRIM5α binding to HIV-1 CA. CA-NC complexes were assembled in vitro and mixed with lysates from 293T cells transfected with the relevant HA-TRIM5α-expressing plasmid constructs. CA-NC complexes were separated from soluble proteins by ultracentrifugation through a sucrose cushion and analyzed by Western blotting (WB) using CA and HA antibodies (pellet). A fraction of the precentrifugation mix was analyzed by Western blotting for CA content (input).

To confirm this result, we also utilized our CCL2 and L2 peptides in biochemical cross-linking experiments to determine if the L2 region can mediate the self-association of these peptides in vitro. In the case of the CCL2 peptide, a dimer was readily detectable at low concentrations of glutaraldehyde (1 mM) (Fig. 4B). At high concentrations of glutaraldehyde, we observed a laddering pattern consistent with nonspecific cross-linking of TRIM5α proteins. Even at these levels of glutaraldehyde, or using mutants that increase or decrease rhTRIM5α assembly and restriction, we did not detect any enrichment of the tetrameric and hexameric species that would be indicative of intrinsic self-association of CCL2 dimers (25, 37) (Fig. 4B). We also examined the ability of L2 to mediate self-association in the absence or the presence of the CC domain. Here we fused the CC domain, the CCL2 fragment, or the L2 region alone to YFP and performed glutaraldehyde cross-linking of these fusions (data not shown). YFP-CCL2, like the recombinant wt peptide, was readily cross-linked into a dimeric form in the presence of glutaraldehyde. However, glutaraldehyde cross-linking did not reveal any detectable self-association activity of the L2 region alone (YFP-L2). Notably, no dimerization of the YFP-CC construct was detected in the presence of glutaraldehyde, suggesting that the L2 region is important for CC-mediated dimerization, consistent with data from a previous report (28). Therefore, while the L2 region does not appear to be capable of intrinsic self-association, it does affect the ability of the CC region to dimerize. However, this ability is independent of its ability to affect assembly, as the mutations that we previously identified as affecting assembly of the full-length rhTRIM5α protein (Fig. 1A) (36) were still able to dimerize and form higher-order multimers in cells, as measured by glutaraldehyde cross-linking (36).

To further understand how the L2 region influences rhTRIM5α self-association, we next compared the influences of assembly-disrupting mutations in the L2 region and mutations in the BBox2 domain on the ability to bind in vitro-assembled CA tubes. It is known that the BBox2 domain mediates rhTRIM5α self-association, and mutations that disrupt BBox2-mediated self-association, such as W117E, abrogate the ability to bind CA tubes (25, 31, 32, 55). As previously established, wt rhTRIM5α effectively bound CA in this assay, while rhTRIM5α containing the W117E mutation did not pellet with capsid assemblies under these conditions (Fig. 4C). In contrast, the RRV-AAA mutant, which does not assemble into cytoplasmic bodies or restrict HIV-1 infection, bound efficiently to in vitro-assembled capsids, suggesting that the L2 region does not directly enhance the self-association of rhTRIM5α required for capsid binding. Collectively, these results demonstrate that the L2 region does not directly contribute to the self-association of rhTRIM5α dimers.

Residues in the N-terminal portion of the L2 region are also required for restriction and assembly in vivo.

To understand how secondary structural motifs on the N-terminal side of the L2 region may affect rhTRIM5α restriction, we also performed alanine mutagenesis of residues in this region (Fig. 5A). We observed that mutations of two additional stretches of residues substantially relieved the ability of rhTRIM5α to restrict HIV-1 infection (Fig. 5B). Specifically, the conversion of RLQ residues at positions 240 to 242 (RLQ240-242) or LQG249-251 to alanine attenuated HIV-1 restriction in these cells. As was the case with other mutations in the L2 region that affect retroviral restriction, these mutants also exhibited a more diffuse localization than did wt TRIM5α or mutants that could still restrict HIV-1 infection, which localized to cytoplasmic assemblies (Fig. 5C).

FIG 5.

Residues in the N-terminal region of the L2 region are also required for restriction and assembly in vivo. (A) Protein levels of HeLa cells stably expressing HA-tagged rhTRIM5α proteins were analyzed by Western blot analysis using an anti-HA antibody. Actin was used as a loading control. (B) Equivalent numbers of these cells were infected with a serial dilution of HIV-1 expressing a GFP reporter for 14 h. The percentage of GFP-positive cells was determined at 48 h postinfection by flow cytometry. (C) Representative images of HeLa cells stably expressing YFP-labeled wt or L2 variant rhTRIM5α proteins (green). Nuclear DAPI staining is shown in blue.

FRET analysis reveals that TRIM5α forms antiparallel dimers.

A recently reported crystal structure of the CCL2 subunit of TRIM25 revealed that this protein forms an elongated antiparallel dimer in which the L2 regions double back along the CC helices of the dimer (38). This is consistent with our observation that the L2 region does not directly influence self-association of TRIM5α dimers. To determine if rhTRIM5α CCL2 fragments similarly form antiparallel dimers, we generated fluorescent fusions of CCL2 constructs tethered to CFP and YFP FRET pairs. Because FRET efficiency exhibits a sixth-order dependence on the separation distance between fluorophores, we utilized N- and C-terminal CFP and YFP fusions to CCL2 to probe the relative orientation of these fluorophores in CCL2 dimers using ensemble FRET measured in cells expressing both constructs. For example, if the CCL2 dimer is oriented in a parallel fashion, CFP and YFP fusions on the N termini of the CCL2 constructs (CFP-CCL2 and YFP-CCL2, respectively) would be expected to exhibit strong FRET activity. Conversely, an antiparallel dimer would be expected to exhibit low FRET activity in this orientation and higher FRET activity when N-terminal and C-terminal fluorescent fusions were paired (Fig. 6B). When the FRET efficiencies of all possible CCL2 FRET pairs were analyzed, they revealed a FRET pattern consistent with an antiparallel dimer of the type described previously by Sanchez et al. (38). N-terminal CFP and YFP fusions consistently exhibited the lowest FRET efficiency (Fig. 6C). Notably, C-terminal fusions to CFP and YFP consistently exhibited the highest FRET efficiency of the four possible combinations (Fig. 6C), significantly higher than the FRET efficiency observed with a single N-terminal and C-terminal FRET pair. Taken together, these data suggest that the CCL2 dimer is a dimer with antiparallel symmetry in which the L2 region traverses back down the CC α-helix to orient the FRET pairs closer to each other.

FIG 6.

FRET analysis reveals that TRIM5α forms an antiparallel dimer. (A) Schematic describing the constructs utilized and the relative FRET efficiencies expected based on the proposed models of TRIM5α dimerization and assembly. (B) CCL2 FRET pairs were transfected into 293T cells, and ensemble FRET efficiency was measured at 24 h posttransfection. (C) BCCL2 FRET pairs were transfected into 293T cells, and ensemble FRET efficiency was measured at 24 h posttransfection. Forty or more individual regions were analyzed for each FRET pair. *, P < 0.001. Results are indicative of data from three independent experiments.

To confirm this observation, we also made similar fluorescent fusions to constructs containing the BBox-CCL2 (BCCL2) fragment. Unlike the CCL2 fusions, all of which exhibited a diffuse cytoplasmic localization, the addition of the BBox to these constructs induced striking filamentous assemblies, which would not allow informative analysis of the dimerization state of the dimeric subunit (data not shown). We therefore introduced a mutation that is known to abrogate the self-association activity of the rhTRIM5α BBox2 domain (W117E) into these constructs. The introduction of this mutation reduced, but did not entirely eliminate, assembly formation following transient transfection of FRET pairs. More specifically, we could not generate a C-terminal BCCL2-YFP fusion that exhibited a diffuse localization, although the other three fusions yielded proteins that were predominantly diffuse and primarily formed dimers following glutaraldehyde cross-linking (data not shown). These constructs allowed us to measure the relative FRET efficiency of N-terminal CFP/YFP fusions (CFP-W117E-BCCL2 and YFP-W117E-BCCL2) and constructs with fluorophore tags on the opposite ends (YFP-W117E-BCCL2 and W117E-BCCL2-CFP). Consistent with the results observed for CCL2 fusions, the CFP and YFP N-terminal fusions to the BCCL2 region exhibited significantly lower FRET efficiencies than when the FRET pairs were placed on opposite termini (Fig. 6D). Collectively, these results demonstrate that rhTRIM5α forms an antiparallel dimer in which the L2 regions double back along the long axis of the dimer, similar to the recently reported structure of the TRIM25 CCL2 dimer (38).

Molecular modeling of the CCL2 region reveals intramolecular contacts critical for assembly.

Using the SWISS-MODEL toolkit (56), we next utilized the recently reported structure of the TRIM25 CCL2 dimer (38) as a template to develop a homology model of the CCL2 fragment of rhTRIM5α (Fig. 7). In keeping with the nomenclature used by Sanchez et al. to describe the TRIM25 dimer, there are three helices in the L2 region that appear to mediate critical contacts in the antiparallel dimer. Helix 1 comprises the entirety of the CC domain and the N-terminal portion of the L2 region. Helix 1 appears to form contacts with helix 2. This establishes a hairpin, which allows the L2 region to double back along the long axis of the dimer. The critical nature of this interaction is supported by our observation that mutation of RLQ240-242 or LQG249-251 substantially impairs or abrogates restriction (Fig. 5).

FIG 7.

Modeling of the rhTRIM5α CCL2 dimer reveals contacts critical for assembly and restriction. (A) View of the CCL2 region of the rhTRIM5α dimer derived from the crystal structure, The CC domains are shown in orange, and the L2 regions are shown in blue. For clarity, one monomer is presented in lighter colors. Restriction-disrupting mutations are shown in red, and restriction-enhancing mutations are shown in green. The model was generated as described in Materials and Methods. The top panel is rotated forward 90° in the bottom panel. Helices 1, 2, and 3 are indicated, as described in the text. (B) L2 helices 1 and 2 form a hairpin structure that is required for assembly and restriction. Restriction-disrupting mutations are shown in red. (C) The 275-RRV-277 motif in helix 3 forms close contacts with the DYD motif in the CC domain.

This model also provides mechanistic insight into how mutations that increase or decrease the helical content of the L2 region promote or disrupt assembly. In our rhTRIM5α homology model, the RRV motif, which is required for assembly and restriction, forms intermolecular interactions with the other CC helix in the dimer (Fig. 7). The contact point for the RRV motif, which includes two basic arginine residues, associates with putative contact points around the acidic DYD motif in the CC helix, which may interact favorably with the RRV motif in the cognate L2 helix (Fig. 7C), although our homology model may not reflect the precise atomic placement of these side chains.

To test the hypothesis that the RRV motif stabilizes the docking of L2 helix 3 to the CC helix, we performed molecular dynamics simulations using our rhTRIM5α homology model of the CCL2 dimer on wild-type rhTRIM5α and the RRV275-277AAA mutant. We performed three independent runs starting with a new set of velocities for each system (wt and RRV275-277AAA). All three runs for each system showed similar behaviors. Figure 8 shows starting configurations and a representative last configuration after a 10-ns molecular dynamics simulation. As shown in Fig. 8, the wt structure was stable during the 10 ns, while the RRV275-277AAA mutant lost tertiary contacts between monomers. In these simulations, the loss of tertiary contacts between mutant L2 and CC occurred quickly, leading to progressive displacement of both L2 helices during the 10-ns simulation (see Movies S1 and S2 in the supplemental material). These simulations support the hypothesis that the RRV motif establishes critical contacts between L2 and SPRY that stabilize the four-helix bundle in the center of the dimer.

FIG 8.

Molecular dynamics simulations of CCL2 dimers. Molecular simulations of the CCL2 dimer were generated by using GROMACS software, as described in Materials and Methods. Simulations were run for CCL2 dimers of the wt sequence and the sequence harboring the RRV275-277AAA mutation. Simulations were initiated with coordinate positions derived from the CCL2 homology model (Fig. 7) (top). After 10 ns of simulation, helical docking of L2 helices to cognate CC helices was consistently destabilized in the RRV275-277AAA mutant. Residues 275 to 277 are shown as space-filling spheres.

DISCUSSION

The studies described above collectively demonstrate the role of the L2 region in TRIM5α restriction and assembly. We demonstrate that α-helices within the L2 region govern the formation of rhTRIM5α assemblies. Critically, the ability of the L2 region to facilitate the formation of cytoplasmic assemblies is directly correlated with the ability of rhTRIM5α to restrict HIV-1 infection (Fig. 1). In addition to revealing a novel determinant of rhTRIM5α assembly, these studies also demonstrate that the potency of TRIM5α antiviral activity can be enhanced independently of altering the ability of the B30.2 PRY/SPRY domain to bind the viral capsid. Alanine mutagenesis of the 269-TFHKN-273 motif increased HIV-1 restriction as well as the ability to form cytoplasmic bodies (Fig. 1 and 2). Mutations of these residues also increased the helical content of CCL2 dimers (Fig. 3). This supports the notion that L2 helix 3 docking to the CC domain to form a 4-helix bundle is critical for restriction and assembly. This is also supported by molecular dynamics simulations that showed that the RRV275-277AAA mutant, which does not restrict HIV-1 infection or form cytoplasmic bodies, does not form stable tertiary contacts with the CC helix, compared to the wt protein (Fig. 8). Moreover, our data collectively suggest that L2 helix 3 is intrinsically unstable and is stabilized by contacts between the L2 RRV motif and residues in the CC domain. In the context of the CCL2 dimer, the RRV-AAA mutation reduced the helical content of the dimer from 78% to 68% (Fig. 3A). This change suggests that mutation of these residues destabilizes most of helix 3, which comprises ∼11% of the CCL2 dimer, according to our homology model based on the structure described previously by Sanchez et al. (38). Additionally, CD analysis showed that in the absence of the CC domain, the L2 region itself is intrinsically disordered, although α-helices can be induced with TFE, which supports the formation of α-helices (Fig. 3B and C). Collectively, these results support the hypothesis that intermolecular docking of the RRV motif to the CC helix stabilizes the formation of an α-helix. Our observations suggest that mutation of the 269-TFHKN-273 motif may increase the ability for helix 3 formation. As alanines have a high tendency to form α-helices, these mutations may establish helix 3 more readily to more readily allow the critical RRV interactions with the CC domain required to induce the formation of the rest of helix 3. Alternatively, it may be that because of the smaller side chains, mutation of the TFHKN motif to alanine reduces steric hindrance that might otherwise make docking of the RRV motif to the CC domain inefficient. Regardless of the mechanism, these studies collectively demonstrate that altering the efficiency of helix 3 formation can significantly alter the efficacy of restriction, as increasing or decreasing the propensity to form helix 3 similarly affected the assembly and restriction observed in vivo.

The observation that TRIM proteins form antiparallel dimers also impacts the assembly models used to describe TRIM5 assembly formation in vitro and cytoplasmic body formation in cells. Previous models of TRIM5 assembly considered assembly in the context of parallel dimers (25, 37). However, in the context of an antiparallel dimer, the self-associative activities required for the assembly of a lattice with hexameric symmetry must be driven primarily by BBox2 positioned at either end of the antiparallel dimer. This is consistent with the observation that L2 mutations do not significantly alter self-association measured in vitro (Fig. 4) and our previously reported observation that alterations of the L2 region did not significantly alter the higher-order multimerization of rhTRIM5α observed following glutaraldehyde cross-linking (36). Superficially, this seems to be at odds with our observations that L2 mutations disrupt cytoplasmic body assembly. However, as we previously suggested, this is consistent with there being three components of TRIM5α self-association: (i) dimerization, mediated by the CC domain and some elements of the L2 region; (ii) higher-order multimerization, mediated by the BBox2 domain; and (iii) assembly, which is mediated by C-terminal elements of the L2 region (40). The role of the L2 region in assembly may therefore be to spatially orient the PRY/SPRY domains in a conformation conducive to assembly. Coordination of the PRY/SPRY domains has been hypothesized to be important for allowing multivalent recognition of capsid determinants presented in the context of an assembled capsid lattice (25, 38). However, the increased or decreased tendency of our mutants to assemble was observed in the absence of restriction-sensitive virus, suggesting that proper coordination of the PRY/SPRY domain, which represents ∼40% of the primary protein sequence, is a more general requirement for efficient assembly. This idea is also supported by the fact that we observed that restriction-perturbing or -enhancing mutations do not grossly affect capsid binding (Fig. 4C). We hypothesize that small changes in the ordering of domains in individual dimers may substantially alter the formation of an efficiently ordered lattice while not grossly affecting the ability of individual dimers to bind capsid or self-associate via BBox2 domain interactions. This suggests that L2-mediated assembly is required for a previously uncharacterized effector step of the restriction process, which is distinct from and occurs subsequent to capsid binding. However, it is also possible that the biochemical assays that we performed to measure self-association or capsid binding lack the resolution required to recapitulate our in vivo observations.

Finally, the observations that an antiparallel dimer is likely to be the basic unit of TRIM5α assemblies also has implications for studies of other TRIM family proteins. Given the conserved ability of many TRIM proteins to form cytoplasmic and nuclear assemblies (3), it seems likely that the antiparallel dimer observed here and the role of the L2 region in allowing assembly of these dimers are architectural features of many if not all TRIM family proteins. It will be interesting to see the degree to which functionally similar α-helices in the L2 region facilitate the assembly of other TRIM family proteins relevant to innate immune function and other human disease states. The studies described above provide a template for approaching these questions.