ABSTRACT
The retroviral Gag protein is the main structural protein responsible for virus particle assembly and release. Like human immunodeficiency virus type 1 (HIV-1) Gag, human T-cell leukemia virus type 1 (HTLV-1) has a structurally conserved capsid (CA) domain, including a β-hairpin turn and a centralized coiled-coil-like structure of six α helices in the CA amino-terminal domain (NTD), as well as four α-helices in the CA carboxy-terminal domain (CTD). CA drives Gag oligomerization, which is critical for both immature Gag lattice formation and particle production. The HIV-1 CA CTD has previously been shown to be a primary determinant for CA-CA interactions, and while both the HTLV-1 CA NTD and CTD have been implicated in Gag-Gag interactions, our recent observations have implicated the HTLV-1 CA NTD as encoding key determinants that dictate particle morphology. Here, we have conducted alanine-scanning mutagenesis in the HTLV-1 CA NTD nucleotide-encoding sequences spanning the loop regions and amino acids at the beginning and ends of α-helices due to their structural dissimilarity from the HIV-1 CA NTD structure. We analyzed both Gag subcellular distribution and efficiency of particle production for these mutants. We discovered several important residues (i.e., M17, Q47/F48, and Y61). Modeling implicated that these residues reside at the dimer interface (i.e., M17 and Y61) or at the trimer interface (i.e., Q47/F48). Taken together, these observations highlight the critical role of the HTLV-1 CA NTD in Gag-Gag interactions and particle assembly, which is, to the best of our knowledge, in contrast to HIV-1 and other retroviruses.
IMPORTANCE Retrovirus particle assembly and release from infected cells is driven by the Gag structural protein. Gag-Gag interactions, which form an oligomeric lattice structure at a particle budding site, are essential to the biogenesis of an infectious virus particle. The CA domain of Gag is generally thought to possess the key determinants for Gag-Gag interactions, and the present study has discovered several critical amino acid residues in the CA domain of HTLV-1 Gag, an important cancer-causing human retrovirus, which are distinct from that of HIV-1 as well as other retroviruses studied to date. Altogether, our results provide important new insights into a poorly understood aspect of HTLV-1 replication that significantly enhances our understanding of the molecular nature of Gag-Gag interaction determinants crucial for virus particle assembly.
INTRODUCTION
The retroviral Gag protein is the critical structural protein that localizes to the inner leaflet of the plasma membrane, forms an oligomeric lattice, and orchestrates particle assembly, release, and maturation to create an infectious virus (1–3). Interfering with the role of Gag in virus assembly eliminates virus infectivity, making Gag an attractive antiviral target (4). Three Gag domains are found within the Gag polyprotein that have roles as discrete viral proteins after Gag is processed by the viral protease: matrix (MA), which is important for Gag-membrane binding; capsid (CA), which is important for Gag oligomerization during virus assembly and core formation during virus maturation; and nucleocapsid (NC), which interacts with the viral RNA packaging signal and encapsidates the viral genomic RNA (3–11). The CA-CA interactions that form the retroviral Gag oligomeric matrix structure can impact both viral particle size and morphology (12–14). Human T-cell leukemia virus type 1 (HTLV-1; a deltaretrovirus) produces mature particles with pleomorphic CA cores, while human immunodeficiency virus type 1 (HIV-1; a lentivirus) produces particles containing conically shaped cores (15–19). Rous sarcoma virus (RSV; an alpharetrovirus) particles also possess a pleomorphic core, although RSV particles are, on average, larger in size (20).
HIV-1 Gag and CA dimer interfaces have been extensively studied (21–28). These studies support the conclusion that the HIV-1 CA carboxy-terminal domain (CTD) acts as the primary driver of oligomerization of both full-length Gag as well as CA protein, with key CA residues being W184 and M185 (residues W316 and M317 in full-length Gag) (9, 21, 22, 29). In contrast, the molecular interactions of HTLV-1 CA involved in immature Gag lattice and mature core formation are poorly understood. Previous studies of the HTLV-1 CA subdomains suggest that each perform distinct functional roles, with the HTLV-1 CA N-terminal domain (NTD) functioning as the primary determinant of Gag oligomerization (30–33), although the HTLV-1 CA CTD may stabilize Gag-Gag interactions (13). Based upon these observations, we recently demonstrated that the HTLV-1 CA NTD was functionally interchangeable with the HIV-1 CA CTD, indicating disparate roles in assembly between the HTLV-1 and HIV-1 CA subdomains (13).
CA domain structure is highly conserved among orthoretroviruses despite substantial sequence variation. The NTD in CA is comprised of 6 to 7 α-helices, and the CTD is comprised of 4 helices. While structural conservation exists, only about 50% amino acid conservation is observed between the HTLV-1 CA NTD and the HIV-1 CA NTD (30). However, certain distinctions between the two CA domains could explain the differential roles in assembly function. First, the amino-terminal β-hairpin that forms following proteolytic processing of the CA is in a different orientation for HTLV-1 than that of HIV-1 (31). This difference is due to an aspartate residue (D54) in HTLV-1 CA that positions the β-hairpin away from the helical core, unlike HIV-1 CA NTD, which encodes an aspartate residue (D51) that positions the β-hairpin toward the helical core (34, 35). The RSV CA NTD has a β-hairpin orientation similar to that of HIV-1, which helps distinguish HTLV-1 from other orthoretroviruses (36, 37). Furthermore, the loop regions between the HTLV-1 CA NTD α-helices differ in both length and composition compared to that of the HIV-1 CA NTD (31).
Given our recent observations that the HTLV-1 CA NTD encodes key determinants that dictate particle morphology, we conducted alanine-scanning mutagenesis in the HTLV-1 CA NTD sequences spanning the loop regions and amino acids at the beginning and ends of α-helices due to their structural dissimilarity from that of the HIV-1 CA NTD. A panel of 59 mutants was analyzed for both Gag subcellular distribution and for the efficiency of particle production in a tractable virus-like particle (VLP) model system that has been validated to closely mimic immature virus particle assembly and release (38–41). Several residues (i.e., M17, Q47/F48, and Y61) were identified as being important for Gag oligomerization and particle assembly, and modeling implicated that they reside at the dimer interface (i.e., M17 and Y61) or at the trimer interface (i.e., Q47/F48). These findings provide the first detailed insights into the critical role of the HTLV-1 CA NTD in Gag-Gag interactions and particle assembly. To the best of our knowledge, our observations are distinct from what has been observed with HIV-1 and other retroviruses studied to date and represent a significant advancement in knowledge regarding the nature of retroviral Gag-Gag interaction determinants.
RESULTS
The goal of this study was to conduct site-directed mutagenesis in order to identify key amino acid residues in the HTLV-1 CA NTD loop regions and short regions of α-helices. We sought to investigate these particular regions based upon the observations that (i) the 7 α-helices in the NTD are highly conserved among orthoretroviruses (42) and (ii) HTLV-1 CA NTD plays a functionally different role than that of the CA NTD from HIV-1 and RSV, which are not the primary determinants of full-length Gag-Gag interactions and particle assembly (9, 29, 43). Taken together, these observations led us to hypothesize that alanine mutations in the loop regions and the surrounding α-helices would be informative in determining the role of residues responsible for efficient Gag assembly. To do this, we created a panel of mutants in which we mutated two consecutive amino acids to alanine in overlapping pairs in a HTLV-1 Gag expression construct (HTLV-1 Gag-enhanced yellow fluorescent protein [eYFP]) used to study HTLV-1 immature particle assembly (44–47). For example, in P1-V2, the first proline and second valine were both mutated to alanines, and similarly, in V2-M3, the second valine and third methionine were mutated to alanines. For double mutants that initially had a non-wild-type (wt) phenotype, single mutants were made, and untagged, mCherry-tagged, and hemagglutinin (HA)-tagged versions were also created. In total, we engineered 57 double mutant and 2 single mutant gag genes in our eYFP-tagged construct.
Assessment of Gag oligomerization via subcellular Gag distribution.To qualitatively assess the ability of the mutant Gag proteins to oligomerize, we analyzed the subcellular distribution of Gag-eYFP in transiently transfected HeLa cells. Gag puncta biogenesis in HeLa cells is an indicator of Gag assembly (and a requisite for particle production from cells), and wt Gag-eYFP served as a positive control for wt levels of Gag puncta formation. In contrast, a lack of Gag puncta and the presence of diffuse Gag-eYFP fluorescence was interpreted as an indication of low or no Gag oligomerization, which would negatively impact particle release (13). HeLa cells were transiently transfected with wt or mutant Gag expression constructs, and then cells were fixed, permeabilized, and stained with 4′,6-diamidino-2-phenylindole (DAPI) and ActinRed555. Puncta formation was quantified using relative fluorescence area associated with Gag puncta based on eYFP area (Fig. 1A and B). In the present study, any mutant Gag protein that resulted in less than 50% of wt-level puncta formation was selected for further analysis.
Subcellular localization of HTLV-1 Gag for a panel of CA NTD alanine-scanning mutants. HeLa cells were transfected with a wt HTLV-1 Gag-eYFP or an alanine-scanning mutant. At least 5 cells were imaged per mutant, and the mean relative fluorescence area associated with fluorescent Gag puncta was determined by using ImageJ. (A) Confocal microscopy. Representative images for wt HTLV-1 Gag-eYFP and for the mutants P9-P10-eYFP, M17-eYFP, Q47-F48-eYFP, and Y61-eYFP are shown. (B) Efficiency of HTLV-1 Gag puncta formation. The mean relative fluorescence area associated with puncta, relative to that of HTLV-1 Gag wt-eYFP, was determined as described in Materials and Methods. Error bars represent standard deviations from 3 independent experiments. Scale bar, 10 μm. (C) Efficiency of HTLV-1 Gag puncta formation in the presence of untagged Gag. Shown is the mean relative fluorescence area associated with puncta in HeLa cells transfected with untagged Gag-eYFP and Gag (1:4 weight ratio). Error bars represent standard deviations from 3 independent experiments.
In this initial analysis of the mutant panel, the following double mutant Gag proteins resulted in less than 50% of wt Gag puncta formation: Q16-M17 (mean, 6.6%), M17-K18 (8.8%), Q47-F48 (32.7%), Q60-Y61 (5.3%), and Y61-L62 (32.5%) (Fig. 1B). Since these two double mutants involved the M17 and Y61 amino acid residues, single mutations of each were introduced individually. Both the M17 and Y61 single mutants resulted in significantly decreased puncta formation (mean of 18.4% and 9.8%, respectively) compared with that of the wt, indicating that the M17 and Y61 residues were directly associated with the reduced puncta formation observed with the Q16-M17, M17-K18, Q60-Y61, and Y61-L62 mutants. Both the Q46-Q47 and F48-D49 mutants had puncta formation at near-wt levels, so the mutation of both Q47 and F48 within the same construct was interpreted as being essential for the reduced puncta formation phenotype observed in the Q47-F48 double mutant.
To ensure that the YFP tag was not interfering with the subcellular localization of M17, Y61, or Q47-F48 Gag mutants, these mutations were introduced into a Gag expression construct without the carboxy-terminal eYFP tag (i.e., untagged Gag). We also cloned a mutant Gag protein that had near-wt levels of Gag puncta efficiency (i.e., P9 and P10) into the untagged Gag vector as a positive control. Experiments were then done with a 1:4 ratio of Gag-eYFP and untagged Gag, which has been previously shown to restore immature VLP morphology and size to be comparable to that of authentic immature particles (Fig. 1C) (47). The M17, Y61, and Q47-F48 mutants possessed the same phenotype in this experiment (i.e., resulted in significantly less puncta formation than the wt).
Analysis of Gag mutant particle production efficiency.Reduced particle production is a predicted phenotypic outcome of inefficient Gag oligomerization and assembly. To assess whether the identified mutants resulted in reduced particle production, immunoblots of cell lysates and supernatants from 293T/17 cells transiently transfected with HA-tagged versions of each construct were collected and analyzed using an anti-HA antibody. HA-tagged Gag levels from cells were normalized relative to tubulin levels, and immunoblot signal levels were determined relative to wt Gag levels.
A density analysis revealed that the chemiluminescent signals for Gag were about 5- to 10-fold lower for M17 and Y61 than for the wt, indicating reduced particle production (Fig. 2). The chemiluminescent signal for Q47-F48 was approximately 3-fold lower than that for the wt Gag signal levels, indicating that the observed reductions in M17, Y61, and Q47-F48 particle production was not associated with low Gag expression levels in cells.
Particle production of selected CA-NTD alanine-scanning mutants. 293T/17 cells were transiently transfected with HA-tagged, HTLV-1 Gag expression constructs (wt or selected CA-NTD alanine-scanning mutants), and the cell culture supernatant was harvested 48 h posttransfection. (A) Immunoblot analysis. Immunoblot analysis was conducted to determine the relative amount of particle production for the selected CA-NTD mutants. Shown is one representative immunoblot of three replicates in which the Gag protein from cell lysates and the cell culture supernatant was detected by using an anti-HA antibody. (B) Analysis of relative particle production. A histogram showing the detection of wt HTLV-1 Gag compared to that of the mutants harvested from cell culture supernatants as determined by immunoblot analysis is indicated. (C) Analysis of relative Gag expression levels in cells. A histogram is shown which indicates the detection of wt HTLV-1 Gag relative to that of the mutants harvested from cell culture lysates as determined by immunoblot analysis.
Analysis of HTLV-1 wt and mutant Gag interactions.To test whether the diffuse cell fluorescence phenotype of Gag mutants could be altered by coexpression of HTLV-1 wt Gag, HeLa cells were cotransfected with mutant Gag-eYFP and wt Gag-mCherry expression constructs at a 1:4 tagged-to-untagged ratio. Cells were permeabilized, fixed, and stained with DAPI and Acti-stain 670 (not shown) to visualize the cell perimeter (Fig. 3A). The punctate formation phenotype was then determined for each mutant Gag as well as colocalization with wt Gag.
Colocalization analysis of wt parental HTLV-1 with selected CA-NTD alanine-scanning mutants. HeLa cells or 293T/17 cells were cotransfected with a selected HTLV-1 Gag-eYFP mutant and a wt parental HTLV-1 Gag-mCherry expression construct in order to evaluate the degree of Gag colocalization and Gag copackaging of particles, respectively. Untagged Gag expression constructs were cotransfected at a 1:4 weight ratio. Experiments were conducted in triplicate. (A) Confocal analysis of HTLV-1 Gag colocalization in cells. Representative images are shown of cotransfected HeLa cells with a mutant HTLV-1 Gag-eYFP and a wt Gag-mCherry expression construct. Scale bar, 10 μm. (B) Confocal analysis of HTLV-1 Gag copackaging. Representative images of particles collected from the cell culture supernatant of 293T/17 cells cotransfected with mutant Gag-eYFP and wt parental Gag-mCherry constructs are shown. Scale bar, 5 μm. (C) Gag puncta formation. Mean relative YFP fluorescence area associated with puncta in HeLa cells transfected with a mutant HTLV-1 Gag-eYFP and a wt parental Gag-mCherry expression construct. Error bars represent standard deviations. (D) Colocalization analysis. Colocalization of cells cotransfected with mutant Gag-eYFP and wt parental Gag-mCherry constructs was analyzed by using Pearson's coefficient.
Cotransfection of wt Gag did not rescue the M17, Q47-F48, or Y61 mutant phenotypes. In particular, both M17 and Y61 had less than 10% punctate formation compared with that of the wt, while Q47-Q48 had approximately 40% of the punctate formation observed with the wt (Fig. 3C and D). While the punctate phenotype was not rescued for any of the 3 mutant Gag proteins, a moderate level of colocalization with the wt was observed for M17 and Y61, and the Q47-F48 mutant colocalized with the wt to a degree equivalent to that of wt Gag-eYFP. The observed colocalization with M17, Q47-F48, and Y61 may have been the direct result of mutant Gag protein interactions with wt Gag or of trafficking patterns otherwise similar to those of wt Gag.
To further investigate this possibility, the ability of the mutant and wt Gag proteins to copackage into particles was analyzed by cotransfection into 293T/17 cells followed by harvesting cell culture supernatants 48 h posttransfection and analyzing in glass chamber slides. As predicted by the colocalization in cells, the mutant Gag proteins appeared to be packaged into particles (Fig. 3B). These observations provide support for the notion that each mutant Gag colocalizes with wt Gag at the plasma membrane and that the mutant Gag proteins retain the ability to traffic to particle budding sites despite these mutations, suggesting that the mutations do not lead to significant levels of protein misfolding that impact Gag trafficking.
Comparison of wt and mutant Gag packaging in VLPs by FLIM.Fluorescence lifetime imaging microscopy (FLIM) measurements of fluorescence resonant energy transfer (FRET) were performed on VLPs to investigate whether Y61 and M17 mutations were associated with changes in the molecular-scale organization of Gag. The FRET interaction has been described as a molecular ruler (48, 49) due to its sensitivity to nanometer-scale separations and orientations between donor and acceptor fluorophores. In the context of VLPs containing a putative immature lattice, observed differences in the efficiency of the FRET interaction can be used to infer differential Gag packaging. As shown in Fig. 4, dually labeled VLPs produced by cotransfecting wt Gag-eYFP, Gag-mCherry, and untagged Gag contained relatively high levels of FRET, with an average FRET efficiency of approximately 25%. VLPs produced by analogous cotransfection of Q47-F48 mutant Gags had an average FRET efficiency similar to that of the wt, while VLPs from Y61 and M17 mutant Gags showed a significant (P < 0.001) reduction of average FRET efficiency to approximately 17%. As a control for zero FRET, VLPs were produced by separate transfections of cells with wt Gag-eYFP and Gag-mCherry. The VLPs collected from each cell plate were mixed together before FLIM. As expected, the FRET efficiency when donor (eYFP) and acceptor (mCherry) fluorophores were segregated within different VLPs was statistically indistinguishable from zero (P > 0.05).
FLIM measurements of FRET efficiency from VLPs. VLPs produced by cotransfecting 293T/17 cells with Gag-eYFP, Gag-mCherry, and unlabeled Gag were immobilized on poly-l-lysine-coated slides. FLIM was used to calculate the average FRET efficiency based on the fluorescence lifetime of eYFP in each sample. As indicated, the wt and Q47-F48 mutant VLPs were observed to have a FRET efficiency of ∼25%, while Y61 and M17 mutant VLPs exhibited a reduced FRET efficiency. As a FRET-negative control, wt VLPs with eYFP- and mCherry-labeled Gags were separately produced and then mixed immediately before FLIM. The FRET efficiency for this control was statistically indistinguishable from zero (P > 0.05). Experiments were done in triplicate. *, P < 0.001 versus HTLV-1 parental wt.
The differences in average FRET efficiency between wt Gag and mutant Y61 or M17 indicate differential packing of these mutant Gag proteins compared with that observed with the wt. Interestingly, average FRET efficiencies in Y61 and M17 mutant VLPs were similar, and FRET efficiencies in Q47-F48 and wt VLPs were also similar. Taken together, these data suggest that the Y61 and M17 mutations possess similar defects in Gag packaging within VLPs and that Q47-F48 mutant Gags package into VLPs in a manner similar to that of wt Gag.
Analysis of particle morphology by cryo-TEM.VLPs have been previously shown to represent good surrogates for analyzing immature virus particle morphology (18). Here, we sought to analyze whether the mutations in the CA-NTD had an impact on particle size and morphology. To do this, we transfected 293T/17 cells with wt or mutant Gag expression constructs, collected the cell culture supernatant after 48 h, and purified particles by gradient ultracentrifugation. This protocol was scaled up to produce enough particles for cryo-transmission electron microscopy (cryo-TEM) analysis for the M17 and Y61 mutants, as they produced lower particle levels in culture supernatants.
Cryo-TEM analysis of wt parental HTLV-1 particles revealed an average diameter of 115.6 nm (standard deviations [SD], 30.2 nm) and regions of flat Gag lattice electron density that did not follow the curvature of the spherical lipid bilayer (Fig. 5); these observations are similar to previous reports (13, 18, 38, 47). The flat regions of electron density attributed to the immature Gag lattice are unique and are a defining characteristic of HTLV-1-like particles (38). The M17 and Y61 mutants primarily produced spherical particles with undefined electron density that was evenly distributed within particles. Particles with defined electron density beneath the lipid bilayer (indicative of an immature Gag lattice) were not observed. These observations suggest that the M17 and Y61 mutants possess a defect in the ability of Gag to oligomerize and form a defined lattice structure, which is a requisite step for infectious particle formation. Intriguingly, both M17 and Y61 mutants had average diameters significantly smaller than those of wt particles (P value of <0.001 for both by Student's t test), which may be associated with a defect in the assembly process. It is interesting that particles produced from HTLV-1 Gag-eYFP also have been shown to contain no discernible Gag lattice and were smaller than particles produced from untagged wt HTLV-1 Gag (44, 47). Taken together, these data suggest that HTLV-1-like particle size is related to the nature of Gag oligomerization and formation of immature Gag lattice.
Cryo-TEM analysis of particles produced by CA-NTD alanine-scanning mutants. 293T/17 cells were transiently transfected with untagged, HTLV-1 Gag expression constructs (wt or selected CA-NTD alanine-scanning mutants), and the cell culture supernatant was harvested. Particles were concentrated and purified by gradient ultracentrifugation prior to microscopy analysis. (A) Cryo-TEM analysis of CA-NTD alanine-scanning mutants. Shown are representative images of CA-NTD alanine-scanning mutants. Scale bar, 100 nm. (B) Particle diameter analysis. The distribution of particle diameters produced by CA-NTD alanine-scanning mutants are shown. The number (n) of particles, means, and standard deviation (SD) are indicated.
The Q47-F48 mutant also produced particles that were spherical in shape, but these particles typically contained electron density characteristic of an immature HTLV-1 wt Gag lattice below the lipid bilayer. The Q47-F48 particles were morphologically comparable with wt particles and were not significantly different in overall diameter. However, these particles were more heterogeneous in size than the wt (Fig. 5B). Although the Q47-F48 mutation resulted in an increase in diffuse Gag localization in cells as well as reduced particle production levels, data from cryo-TEM provides evidence that this mutant was still capable of forming particles morphologically indistinguishable from those of the wt. This indicates that the reduction in the efficiency of particle production did not interfere with Gag lattice formation and particle budding.
Modeling of HTLV-1 Gag hexamers.To date, no molecular structures of HTLV-1 immature CA oligomers have been reported in the literature. Structures have been solved for both HIV-1 and RSV CA hexamers in the context of full-length Gag proteins. To help better understand the role of M17, Q47-F48, and Y61 amino acids in Gag-Gag interactions, we conducted comparative modeling of HTLV-1 CA hexamer structures using previously solved structures for RSV and HIV-1. The HTLV-1 CA structure was determined based on both the HIV-1 packing and RSV packing previously observed in high-resolution cryo-EM densities (Fig. 6).
Modeling of HTLV-1 CA-CA interactions. Comparative analysis of HTLV-1 CA-CA interactions. (A) Shown is a composite ribbon diagram of the HTLV-1 CA structure of the CA-NTD (yellow) and CA-CTD (blue) based upon previously published data, with the orientation adjusted based upon the known packing of HIV-1 CA or of RSV CA. (B) Model of HTLV-1 CA hexamer based upon HIV-1 CA packing. The locations of the HTLV-1 CA M17 (red), Y61 (cyan), and Q47-F48 (yellow) residues are indicated. (C) Model of HTLV-1 CA hexamer based upon RSV CA packing. The locations of the HTLV-1 CA M17 (red), Y61 (cyan), and Q47-F48 (yellow) residues are indicated.
When HTLV-1 CA interactions were modeled based on HIV-1 packing (Fig. 6B), M17 and Y61 were found to be located at the Gag dimer interface. The residues are approximately 5.9 Å apart within a single Gag molecule, and they are approximately 6.6 Å apart across the dimer interface. The Q47-F48 amino acids are shown to be located within the trimer interface of the HTLV-1 hexamer based on HIV-1 packing, suggesting a role for these residues in stabilizing Gag trimers. When HTLV-1 CA interactions are modeled based on RSV packing, the M17 and Y61 residues still appear to interact within a single Gag molecule but are no longer at the dimer interface.
Based on our comparative modeling analysis, the site-directed mutagenesis analyses support that the HTLV-1 CA NTD oligomerizes in a manner similar to that of HIV-1 Gag. This is unexpected given that the CA NTD is the driving force for HTLV-1 immature CA-CA interactions (13, 30), whereas the HIV-1 CA CTD encodes the amino acid residues W184 and M185, which have been shown to be primary determinants of full-length Gag-Gag interactions (9, 21, 29, 50). Given these differences between HTLV-1 and HIV-1, it is likely that the HTLV-1 hexamer arrangement is unique and not fully consistent with the HIV-1 packing model. However, our comparative modeling analyses predict an HTLV-1 Gag packing model more similar to that of HIV-1 than to that of RSV.
Mutational analyses of M17 and Y61 based on modeling.Based on our comparative modeling of the HTLV-1 CA NTD interactions, we hypothesized that residues 17 and 61 were interacting in an inter- or intrahexameric manner at the dimer interface. To test the hypothesis that interactions occurring between amino acids 17 and/or 61 would be retained if the amino acids were interchanged (which would help differentiate between inter- and intrahexameric interactions), we created the single M17Y and Y61M mutants as well as a double mutant, M17Y/Y61M. Confocal microscopy (using eYFP-tagged Gag constructs) and immunoblot analysis (using HA-tagged Gag constructs) were performed to evaluate mutant phenotypes (Fig. 7). These analyses revealed that M17Y had a wt phenotype. In particular, puncta formation was present for M17Y (Fig. 7A and B), and this mutant produced particles at levels comparable to that of the wt (Fig. 7C). We interpret these observations as support of tyrosine-tyrosine interactions or “π-stacking” between the tyrosines at the 17 and 61 residues (51). In contrast, the Y61M mutation did not form puncta (Fig. 7A and B) and was associated with low levels of particle production (Fig. 7C). The phenotype of the Y61M mutation is indicative of a lack of efficient Gag-Gag interactions. The M17Y/Y61M double mutant primarily revealed a diffuse Gag distribution phenotype in cells, with Gag puncta formation significantly lower than that of wt Gag (Fig. 7A and B). Particle production was severely reduced and was comparable to that observed with the Y61M single mutant (Fig. 7C). These observations suggest that the potential interactions between the Y17 and the M61 amino acid residues in the Gag double mutant were limited by minor geometric perturbations not fully predicted by the model shown in Fig. 6. Taken together, these model-based mutational analyses of the M17 and Y61 residues provide support for the proposed model as a useful guide for CA-CA interactions.
Model-based mutational analyses of M17 and Y61 in the HTLV-1 CA-NTD. HeLa cells or 293T/17 cells were transfected with the indicated HTLV-1 Gag-eYFP mutant or a HTLV-1 Gag-HA construct, respectively. Transfections were performed in the presence of a 1:4 weight ratio of tagged to untagged HTLV-1 Gag constructs. (A) Confocal microscopy. Representative images for wt parental HTLV-1 Gag-eYFP and for the mutants M17Y-eYFP, Y61M-eYFP, and M17Y/Y61M-eYFP are shown. (B) HTLV-1 Gag puncta formation. The mean relative fluorescence area associated with puncta, relative to HTLV-1 Gag wt-eYFP, was determined as described in Materials and Methods. Error bars indicate the standard deviations from 3 independent replicates. Scale bar, 10 μm. (C) Analysis of relative Gag expression levels in cells as determined by immunoblot analysis as described in Materials and Methods.
DISCUSSION
The determinants involved in Gag-Gag interactions are generally known to reside in the CA domain for retroviruses, but the precise nature of these interactions among HTLV-1 domains is poorly understood. While HIV-1 and HTLV-1 are both important T-cell-tropic human retroviruses, the basic biology between these viruses is quite different, which emphasizes the importance of comparative analysis for understanding the detailed mechanisms that have evolved in the maintenance of human retroviral replication. The precise nature of HTLV-1 replication, including virus assembly, has been recalcitrant to detailed molecular analysis due to the difficulties in the handling of molecular clones and conducting studies in cell culture.
In this study, we used alanine-scanning mutagenesis to thoroughly interrogate the loop and neighboring helical regions of the HTLV-1 CA NTD. These specific regions were chosen for mutagenesis based on structural comparisons between the HTLV-1 and HIV-1 CA NTD, particularly where differences between the two CA protein structures were predicted. In our analysis of a panel composed of 59 mutants, we found that the majority of mutagenized residues had no overall effect of Gag subcellular localization, which provides evidence that the alteration to alanine had no discernible effect on Gag oligomerization. However, a potential limitation of this analysis is that it is formally possible that a Gag mutant that appears to form puncta does not produce particles. Such a mutant would not have been identified for subsequent analyses based upon the observations made by confocal microscopy. Given this, immunoblot analysis of cell culture supernatants would help identify such mutant phenotypes. Using the criteria that lack of Gag puncta was indicative of a lack of productive Gag-Gag interactions, we identified three mutants, M17, Q47-F48, and Y61, that significantly decreased the prevalence of punctate HTLV-1 Gag protein, suggesting a defect in Gag oligomerization and particle assembly. These Gag mutants produced particles at reduced levels compared with those of the wt, and the particles that were produced by M17 and Y61 lacked electron density beneath the lipid bilayer that would be expected for the immature Gag lattice. Moreover, M17 and Y61 were associated with significantly reduced average FRET efficiencies in dual-colored fluorescent particles, indicating perturbed Gag packing. This correspondence between FLIM and cryo-TEM particle phenotypes supports the use of FLIM to investigate perturbations in the immature Gag lattice. Taken together, these observations indicate that the M17 and Y61 residues are critical for Gag oligomerization and particle assembly.
Comparative modeling suggested that the NTDs of HTLV-1 CA hexamers follow an HIV-1-like packing model, positioning the M17 and Y61 residues such that they interact with each other within the same Gag protein at or near the HTLV-1 CA dimer interface. Our model-based mutational analysis provides general support for the utility of modeling for investigating CA-CA interactions. Intriguingly, the residues that drive the HIV-1 CA dimer interface are also an aromatic residue and a methionine (W184 and M185) (9, 50). Interactions between methionines and aromatic residues are known to stabilize protein structures at a much higher rate than hydrophobic interactions, suggesting that M17 and Y61 are critical for proper HTLV-1 Gag packing and perhaps full-length Gag dimerization (52). In addition, oxidation increases the strength of methionine interactions with aromatic residues, and previous reports have indicated that HTLV-1 assembly and budding is dependent upon oxidative assembly (53, 54).
While comparative modeling of HTLV-1 CA hexamers is intriguing in that it supports an HIV-1-like packing model over an RSV-like packing model, it is likely that HTLV-1 CA hexamers assume a unique structure that more closely resembles HIV-1 than RSV. Given that our data indicate that the CA NTD plays a significant role in HTLV-1 CA-CA interactions, a more rigorous analysis of the structure-function relationship between the HTLV-1 CA NTD and HTLV-1 immature Gag-Gag interactions is warranted. Specifically, continued site-directed mutagenesis studies combined with cryo-electron tomography studies to elucidate HTLV-1 immature Gag lattice structure will be highly informative toward this end.
A particularly intriguing observation made in this study was that the distinct subcellular Gag localization and morphological particle phenotypes for the M17 and Y61 mutants did not preclude the production of spherical particles, albeit at low levels. It is of interest that mutation of W184 and M185 residues in HIV-1 Gag to alanine was shown to not entirely abrogate particle production due to functional redundancy within the Gag protein (55). The ability of Gag to use RNA as a scaffold is believed to rescue a low level of particle production in the HIV-1 Gag W184-M185 mutant. This could help explain our observations in this study with the HTLV-1 Gag M17 and Y61 mutants. While the M17 and Y61 Gag mutants do not appear to form wt oligomeric structures, the strong affinity of HTLV-1 MA for the plasma membrane along with the ability of the NC domain to bind RNA likely aid in allowing a low level of particle production to occur. Ongoing studies that are investigating the role of RNA binding in the production of spherical particles will be informative in providing new insights into whether HTLV-1 Gag has functional redundancies similar to those of HIV-1 Gag (55).
The key features of the M17 and Y61 mutant Gag proteins were (i) the absence of electron density below the lipid bilayer of the VLP that would be indicative of the immature Gag lattice, (ii) a significant reduction in VLP size, and (iii) a reduction in average FRET efficiency within dually labeled fluorescent particles. The first two phenotypes are similar to a previous report on the morphology of VLPs produced by the HTLV-1 Gag-eYFP expression construct (44, 45, 47). Given the similarities between our findings with VLPs produced from the M17 and Y61 mutants (both untagged and eYFP tagged), we hypothesize that the C-terminal eYFP tag interferes with and prevents proper Gag oligomerization, perhaps due in part to the inability of CA to interact and initiate wt oligomeric structures.
The findings presented here represent the first extensive alanine-scanning mutagenesis study conducted in the HTLV-1 CA domain of Gag. These studies have resulted in important new observations related to retroviral Gag-Gag interaction determinants and their impact on virus particle assembly. In particular, mutagenesis in the HTLV-1 CA NTD loop regions and neighboring regions of α-helices were performed based upon their structural dissimilarity from that of the HIV-1 CA NTD structure. Several residues (i.e., M17, Q47/F48, and Y61) important for Gag oligomerization and particle assembly were characterized, and modeling implicates that these residues reside at the dimer interface (i.e., M17 and Y61) or at the trimer interface (i.e., Q47/F48). Model-based mutational analyses provided further support for the M17 and Y61 residues being at the dimer interface. The observations from this study help to establish the critical role of the HTLV-1 CA NTD as a key determinant in Gag-Gag interactions and particle assembly, a surprising and somewhat unexpected finding. This clearly distinguishes HTLV-1 from HIV-1 as well as other retroviruses and argues that important differences exist in the HTLV-1 particle assembly pathway that were not predictable from previous observations. Further studies of the HTLV-1 particle assembly pathway will likely provide other surprising observations that enhance our understanding of the diversity of molecular interactions involved in virus assembly. Such differences have broad implications, which may be applicable to the discovery of novel therapeutic targets.
MATERIALS AND METHODS
Plasmids, cell lines, and reagents.The HTLV-1 codon-optimized gag-eYFP plasmid has been previously described (44), as have the HTLV-1 untagged gag, gag-mCherry, and gag-HA plasmids (13, 38). Both 293T/17 cells and HeLa cells were purchased from the ATCC (Manassas, VA) and were cultured in complete Dulbecco's modified Eagle medium supplemented with 10% fetal clone III (GE Healthcare Lifesciences, Logan, UT).
Site-directed mutagenesis of gag plasmids.For the alanine-scanning mutagenesis of portions of the HTLV-1 CA NTD, codons in the HTLV-1 gag-eYFP plasmid were changed to alanine-encoding codons using the QuikChange II XL site-directed mutagenesis kit (Agilent Technologies, Santa Clara, CA). For the M17Y and Y61M mutants, amino acids 17 and 61 were mutated to Y and M, respectively. The plasmids were confirmed by sequencing. Mutant Gag-eYFP expression constructs of interest were engineered into vectors containing no C-terminal tag, a C-terminal mCherry tag, or a C-terminal HA tag using the 5′ HindIII and 3′ BamHI restriction sites.
Screening of mutants with subcellular localization analysis.The subcellular localization of Gag-eYFP was evaluated as previously described (13). Briefly, HTLV-1 wt and mutant Gag-eYFP expression constructs were transiently transfected into HeLa cells with GenJet, version II (SignaGen, Gaithersburg, MD). Approximately 24 h posttransfection, cells were fixed, stained with DAPI, and stained with ActinRed 555. Cells were imaged using a Zeiss LSM 700 confocal laser scanning microscope with a Plan-Apochromat 63×/1.40-numerical-aperture (NA) oil objective at 1.2× zoom (Zeiss, Thornwood, NY). The degree of Gag protein assembly into puncta was quantified by evaluating the area associated with Gag puncta divided by the total area of Gag fluorescence, which resulted in a relative measurement of puncta formation. At least 15 individual cells were imaged across 3 independent replicates for a total of 15 cells. For the HTLV-1 parental wt, P9-P10, M17, Q47-F48, and Y61, the same analysis was done with a cotransfection of eYFP-tagged and untagged Gag expression constructs at a 1:4 weight ratio.
Colocalization of Gag proteins.HTLV-1 Gag-eYFP expression constructs were transiently cotransfected with HTLV-1 wt Gag-mCherry into HeLa cells to assess the colocalization of the Gag proteins, as previously described (13). The cotransfections were performed with a 1:4 weight ratio of tagged to untagged gag expression plasmids. The cells were stained with DAPI and Acti-stain 670 (Cytoskeleton, Inc., Denver, CO) before being fixed with paraformaldehyde. Colocalization was evaluated in at least 5 cells from 3 different replicates using ImageJ plugin Coloc 2 to determine Pearson's coefficient within the cell perimeter, as determined by Acti-stain 670.
Copackaging of Gag proteins.As previously described (13), HTLV-1 Gag-eYFP expression constructs were transiently cotransfected with HTLV-1 wt Gag-mCherry into 293T/17 cells along with untagged Gag expression constructs at a 1:4 weight ratio. The supernatant was collected and filtered approximately 48 h posttransfection, and the particles were imaged in an 8-well dish using a Zeiss LSM 700 with a Plan-Apochromat 100×/1.40-NA oil objective at 2.5× zoom (Carl Zeiss).
Immunoblot of Gag-HA proteins.293T/17 cells were transiently transfected with HA-tagged Gag expression constructs and analyzed via immunohistochemistry as previously described (13). Briefly, cell lysates and supernatant were collected 48 h posttransfection. Supernatants were ultracentrifuged to concentrate VLPs, which were resuspended in phosphate-buffered saline (PBS) with 1.0% Triton X-100. Cell lysates were clarified via centrifugation, and 30 μg of total protein was loaded into each well. Gag-HA proteins were detected with 1:1,000 anti-HA antibody (16B12; BioLegend, San Diego, CA). Gamma-tubulin was detected with 1:1,000 antitubulin antibody (GTU-88; Sigma-Aldrich). Horseradish peroxidase-conjugated goat anti-mouse secondary antibody was used for chemiluminescent detection (Jackson ImmunoResearch, West Grove, PA). Blots were imaged with a ChemiDoc Touch system (Bio-Rad, Hercules, CA) and analyzed using Image Lab (Bio-Rad). The development of chemiluminescent detection was done by real-time imaging, and analysis of band intensities was done in the linear detection range. Gag-HA expression levels were calculated relative to HTLV-1 wt, and blots were done in triplicate to ensure reproducibility of the observations.
Cryo-TEM analysis of particle morphology.VLP concentration and purification were previously described elsewhere (38). To ensure that we were not analyzing vesicles, the supernatant was filtered and run through an OptiPrep gradient. A visible band (not observed in cell culture supernatants from mock-transfected cells) was removed from the gradient for subsequent analyses. EM grids used were Lacey/Formvar 300 mesh, Lacey carbon 300 mesh (EMS, Hatfield, PA), R2/2 Holey carbon 200 mesh, and Multi A Holey carbon 200 mesh (Quantifoil, Germany). Grids were glow discharged and loaded on an FEI MarkIII Vitrobot system (Thermo Fisher Scientific, Waltham, MA). Purified and concentrated particle suspensions (3 to 4 μl) were loaded on the carbon side of a grid and manually blotted before being plunge-frozen in ultracooled liquid ethane. Grids were imaged on a Tecnai F30 FEG transmission microscope (Thermo Fisher Scientific, Waltham, MA) operating at 300 kV. Images were taken at ×49,300 magnification at an electron dose of ∼25 electrons/Å2 and 4 to 8 μm defocus values using a Gatan 4,000 by 4,000 charge-coupled device (CCD) camera (Gatan Inc., Pleasanton, CA). Particle diameters were measured using ImageJ software. Two perpendicular diameters were measured and averaged for each particle. Particle morphologies were qualitatively characterized.
Model building.The structure of monomeric HTLV-1 CA, PDB entry 1QRJ, was used as the starting point for the mutant simulations. Of the monomeric HTLV-1 Gag structure, residues 1 to 14 and residues 208 to 214 were disregarded for the present study. The packing conformations of HIV-1 and RSV immature CA, namely, 4USN and 5A9E, were used as a starting point to build the hexamer complex. Subsequently, the NTD and CTD of the HTLV monomer were rigid-body docked into the RSV and HIV-1 cryo-EM density using Chimera (56). Each domain was docked separately to the respective terminal for each monomer of HIV and RSV complex. The fitted NTD and CTD were connected by using MODELLER (57).
The resulting complex was solvated using the TIP3P water model (58). The water boxes were adjusted to fit the hexamer's molecular dimensions. For the HTLV-1–HIV packing, the simulation box dimensions were the following: a = 162.1 Å, b = 158.6 Å, and c = 81.1 Å, with unit cell angles of alpha = beta = gamma = 90°. For the HTLV-1–RSV packing, the simulation box dimensions were the following: a = 150.6 Å, b = 161.8 Å, and c = 87.8 Å, with unit cell angles of alpha = beta = gamma = 90°. The solvated systems were neutralized by adding chloride and sodium ions. The total concentration of NaCl was set to 150 mM for each system. The resulting simulation systems contained ∼190,000 atoms, including protein, water molecules, and ions.
MD simulations.Molecular dynamics (MD) simulations for each of the complex models were performed using NAMD v2.12 (59) and the CHARMM 36m force field (60, 61). Each simulated model was subjected to an energy minimization, followed by a thermalization of 10,000 steps. After heating, each system was subjected to 1 ns of equilibration. An integration time step of 2 fs was employed for all simulations. Equations of motion were integrated using a RESPA multi-time-step integrator. Long-range interactions were updated every 2 fs, while nonbonded interactions were recalculated every 1 fs. Electrostatics were calculated using the particle mesh Ewald algorithm with a grid spacing of 1 Å. Temperature was controlled by a Langevin piston at 310K. A Langevin barostat was employed to maintain the pressure of the simulation box at 1 atm.
Analysis of mutant residues.Analysis of the MD trajectories were performed using VMD (62). The positions of the residues of interest, M17A, Q47A, M47A, and Y61A, were highlighted on each monomer for both RSV- and HIV-based configurations. Intrahexameric and interhexameric contacts of the residues of interest were observed throughout the trajectory of both HIV-1 and RSV packing conformations. A contact map for each simulated complex was created exhibiting the distances between the potential residues.
Fluorescence lifetime imaging microscopy.FLIM measurements were performed on a Zeiss Axiovert 200 microscope modified for two-photon excitation. Briefly, 1,000-nm excitation light provided by a mode-locked Ti:S laser (Tsunami, Spectra Physics, Santa Clara, CA) is coupled into the microscope through a galvanometer-driven scan head (Yanus IV, FEI, Hillsboro, OR) and focused to a diffraction-limited spot by a 63×, 1.2-NA water immersion objective (C-Apochromat; Zeiss, Thornwood, NY). Fluorescence emission from eYFP and mCherry was separated by a dichroic mirror and band-pass filtered before being recorded by hybrid PMT detectors (HPM-100-40; Becker & Hickl GmbH, Berlin, Germany). Imaging control and data acquisition were provided by SimFCS software (Laboratory for Fluorescence Dynamics, UC at Irvine, CA), which controls a 3-axis voltage waveform generator and a FastFLIM data acquisition board (ISS, Inc., Champaign, IL). Frequency-domain FLIM data (63) were acquired as time stacks (∼30 frames, 64-μs pixel time, 256- by 256-pixel resolution) on 34-μm by 34-μm fields and saved in native (fbd) format for later analysis using routines written in IDL 8.5 (Harris Geospatial Solutions, Broomfield, CO). Data analysis was performed by calculating the average phase lifetime of the donor eYFP from the VLPs in each 34-μm by 34-μm image using the equation τϕ = ω−1 tan(S/G), where G and S are the real and imaginary components, respectively, of the fluorescence response at the angular frequency ω = 2π × 80 MHz (64). A measurement of fluorescein in 0.1 M NaOH (lifetime, 4.05 ns) was used for calibration. Phase lifetimes were converted into the apparent FRET efficiency by E = 1 − τϕ, sample/τϕ, eYFP, where τϕ, eYFP = 2.74 ± 0.03 ns (SEM) is the phase lifetime of VLPs produced by cotransfecting HTLV-1 Gag-eYFP and Gag at a 1:3 plasmid weight ratio and τϕ, sample is the phase lifetime of the sample under test.
Dually labeled VLPs for FLIM were produced by cotransfecting eYFP- and mCherry-labeled Gag proteins with unlabeled Gag proteins at 1:1:6 plasmid weight ratios. The eYFP- and mCherry-labeled VLPs were produced by cotransfecting labeled and unlabeled Gag proteins at a 1:3 plasmid weight ratio. VLPs from cell culture supernatant were concentrated 10- to 100-fold by centrifugation at 16,000 × g for 1 h and added to poly-l-lysine-coated 8-well chamber slides (Thermo Fisher Scientific, Pittsburgh, PA). After verifying that immobilized VLPs had accumulated to high density (∼500 to 1,000 particles per frame area), the well was washed twice with STE buffer and mounted on the microscope stage. FLIM images were acquired from multiple 34-μm by 34-μm areas, and the estimated FRET efficiencies represent an average from at least 2,000 VLPs per sample.
To verify the presence of FRET in each VLP sample, acceptor photobleaching of Gag-mCherry was performed by exposing the sample to 3 min of epifluorescence excitation provided by an Hg lamp with mCherry filter set (48), followed by a second FLIM measurement. FRET efficiencies after acceptor photobleaching were significantly reduced (E value of 3% to 5% after bleaching) in all VLP samples produced by cotransfection of eYFP- and mCherry-tagged Gags.
P values for comparisons of average FRET efficiencies were calculated from two-tailed tests using Welch's t statistic. P values were adjusted for multiple hypotheses using the Bonferroni correction, where claims of significance were made and unadjusted where a lack of statistical significance is noted.
ACKNOWLEDGMENTS
We gratefully acknowledge financial support from the National Institutes of Health (R01 GM098550 to L.M.M., T32 DA007097 to J.L.M., T32 AI083196 to I.A., and P50 GM082251 to J.R.P.) and the National Science Foundation (ACI-1548562 to J.R.P.).
FOOTNOTES
- Received 27 February 2018.
- Accepted 24 April 2018.
- Accepted manuscript posted online 25 April 2018.
- Copyright © 2018 American Society for Microbiology.
