Analysis of Human T-Cell Leukemia Virus Type 1 Particles by Using Cryo-Electron Tomography

Abstract

The particle structure of human T-cell leukemia virus type 1 (HTLV-1) is poorly characterized. Here, we have used cryo-electron tomography to analyze HTLV-1 particle morphology. Particles produced from MT-2 cells were polymorphic, roughly spherical, and varied in size. Capsid cores, when present, were typically poorly defined polyhedral structures with at least one curved region contacting the inner face of the viral membrane. Most of the particles observed lacked a defined capsid core, which likely impacts HTLV-1 particle infectivity.

TEXT

Human T-cell leukemia virus type 1 (HTLV-1) is a human cancer-causing retrovirus that causes an adult T cell leukemia and HTLV-1-associated myelopathy/tropical spastic paraparesis (1, 2). It is generally accepted that cell-cell contact is indispensable for the infection of HTLV-1 (3); however, cell-free infection through dendritic cells also contributes to viral transmission and pathogenesis (4). HTLV-1 particle assembly and maturation involve the budding of immature virus particles from the plasma membrane, followed by the triggering of virally mediated proteolysis and particle maturation (5, 6). Similar to Gag of other retroviruses, the HTLV-1 Gag protein has three major functional domains, the matrix (MA), capsid (CA), and nucleocapsid (NC). The MA binds directly to the inner leaflet of the cell membrane, which is critical for particle assembly; the CA protein is necessary for Gag-Gag interactions that form the Gag lattice in immature particles and is also required in forming the viral capsid core in mature particles (7); and the NC protein interacts with and coats the viral RNA genome. During virus maturation, Gag is cleaved by the viral protease, after which CA proteins reassemble into a capsid core that encapsulates the NC-RNA complexes and the viral replication enzymes.

Studies using cryo-electron microscopy (cryo-EM) and cryo-electron tomography (cryo-ET) have demonstrated that retrovirus cores are highly polymorphic. The cores in human immunodeficiency virus type 1 (HIV-1) are predominately conical in shape (8, 9), whereas those in Rous sarcoma virus (RSV) are either irregularly polyhedral or spherical (10, 11). Tubular cores have also been observed for both HIV-1 and RSV. Despite the morphological diversity among retroviral cores, the fundamental principles by which the CA proteins are organized are similar. For example, the amino-terminal domain (NTD) of CA mediates the formation of hexameric or pentameric rings on the exterior of the capsid, while the carboxy-terminal domain (CTD) links the rings together through dimerization at the floor of the capsid. In addition to variation in the dihedral angles between neighboring CA hexamers (11, 12), closely packed hexamers are interspersed with CA pentamers to mediate larger changes in curvature (11, 13). The correct formation of cores is likely related to their functions in replication and the infectivity of the virus particles (14–16).

Despite the significant impact of HTLV-1 on human health, many of the details of its particle assembly and structure remain poorly understood. Previous structural analyses of HTLV-1 assembly have largely been limited to observing HTLV-1-infected cells by using thin-section transmission electron microscopy (TEM). In those studies, HTLV-1 particles produced from chronically infected MT-2 cells (17) appeared to be assembly intermediates, with the electron densities inside individual particles forming either a half-donut shape (18, 19) or a complete donut shape (17) along the inside of the viral membrane. Virus particles were observed that had a distinct capsid core or had the electron densities evenly distributed in the interior of the particle in the absence of a core structure. In the current study, we purified HTLV-1 particles from MT-2 cells and employed cryo-EM and cryo-ET methods to determine the morphology of HTLV-1 virions in a frozen-hydrated state.

A two-step procedure was used for the purification of HTLV-1 particles. First, MT-2 cells were cultured in 50 ml of RPMI 1640 medium supplemented with 10% FetalClone III for about 1 week or until the cells reached ∼90% confluence (as determined by the formation of cell clusters). Two new flasks were then each inoculated with ∼5% of the suspended cells from the first expansion cycle and filled with 25 ml of culture medium supernatant from the previous expansion cycle and 25 ml of fresh medium. The cultures were then grown for an additional week, after which HTLV-1 particles were purified from the culture medium supernatant and concentrated using a procedure modified from that of Grigsby et al. (18, 19). In particular, after removing large cellular debris by centrifugation and filtration, the sample was pelleted using an 8% OptiPrep (Sigma-Aldrich) cushion and applied to a 10-to-50% OptiPrep gradient for centrifugation. The virus-containing fraction was extracted, pelleted, and resuspended in STE buffer (10 mM Tris-HCl, pH 7.4, 100 mM NaCl, 1 mM EDTA). Previous immunoblot analyses of HTLV-1 particles produced from MT-2 cells found both processed and nonprocessed forms of HTLV-1 Gag (19, 20). HTLV-1 particles purified from MT-2 cells were found to be infectious but with an infectivity that was about 1,000-fold lower than that of HIV-1. Prior to cryo-EM sample preparation, 2,2′-dithiodipyridine (AT-2) was used to inactivate infectious HTLV-1 particles. Previous studies with HIV-1 have indicated that AT-2 covalently modifies the NC domain zinc fingers, which results in blocking of reverse transcription (21).

For tomography data collection, the virus sample was mixed with 10-nm colloidal gold particles treated with bovine serum albumin solution (Sigma-Aldrich) and then vitrified on 200-mesh Quantifoil grids (Ted Pella, Inc.) in liquid ethane. An FEI F30 transmission electron microscope was used to record single-axis tilt series from −60° to +60° in 3° incremental steps. Data were collected automatically using SerialEM (22) on a Gatan Ultrascan 4000 charge-coupled device (CCD) camera at ×39,000 nominal magnification. The defocus level of the image at the 0° tilt angle in each tilt series was determined to be ∼13 μm based on the first zero of the contrast transfer function shown in RobEM (http://cryoem.ucsd.edu/programs.shtm). The tomograms, reconstructed in IMOD (23), were bin averaged so that the dimension of each voxel in the tomogram corresponded to 0.62 nm in the specimen. The final reconstruction maps were denoised using nonlinear anisotropic diffusion implemented in IMOD. We used 30 tomographic reconstructions for further analysis.

The HTLV-1 virions embedded in vitreous ice were roughly spherical and varied in size (Fig. 1; see also Movie S1 in the supplemental material). About 15% of the particles were found to contain an interior core-like structure that had a distinct ∼5-nm-thick density layer corresponding to the viral capsid shell. These particles appeared to best represent mature HTLV-1 particles with complete cores. Other density-filled particles (∼60%) were observed to contain evenly distributed electron densities inside the viral membrane but no density attributable to a capsid core. While the nature of these HTLV-1 particles is not entirely clear, they likely represent particles that lack viral infectivity, as they do not possess a complete (mature) capsid core. Assembly defects may have contributed to the morphology of some or perhaps most of these particles, although further analyses are needed to better characterize these particles. Approximately 5% of the particles observed contained partially mature capsid cores in which the shell of the capsid was not closed, and the electron densities in the open region appeared similar to those in the frequently observed particles that were density filled. Single HTLV-1 particles that contained either conically shaped cores or double cores were rare, and particles with tubular cores were not observed. Densities that resembled small knob-like structures projecting away from the viral membrane were observed (Fig. 2E). However, the limit of resolution in the tomographic data set did not allow the determination of whether these structures represent the HTLV-1 envelope proteins.

FIG 1.

Morphology of HTLV-1 particles as determined by cryo-ET. (A) A section in a tomogram showing authentic HTLV-1 particles produced and purified from MT-2 cells. White arrows point to particles with evenly distributed electron densities that do not have discernible cores. Black arrows point to virus particles with mature cores. White arrowheads point to particles that have partially complete cores. Particle 3 has a capsid core with roughly spherical morphology. (B) Surface-shaded models of the four particles indicated by number in panel A, showing the viral envelope (cyan), capsid shell (red), genome and/or mixture of proteins and nucleic acids derived from the host cell (beige), and densities between the capsid and viral membrane (beige). The image was created using Chimera software (36).

FIG 2.

Tomographic slices of core-containing HTLV-1 particles. (A, B) Small and large particles containing complete (mature) cores; note that there are curling edges in the capsid shell adjoining the viral membrane. (C) Particles with incomplete cores. (D) Particles with partially double-layered cores. (E) Particles with prominent spike structures (arrowheads) on the viral envelope. (F) Particles with external turret structure (arrows) on the viral envelope. The scale bar is 100 nm.

The average diameter of the mature and partially mature HTLV-1 virions, as measured from the outer margins of the viral membrane in 83 particles, was determined to be 113 ± 23 nm (mean ± standard deviation), which is smaller than the diameters of mature RSV (∼125 nm) (10) and HIV-1 (∼145 nm) (8, 24) particles. The HTLV-1 particles are larger than the virus-like particles previously produced from 293T cells by an HTLV-1 Gag expression construct with yellow fluorescent protein (YFP) fused to the carboxy terminus (18), suggesting that the YFP influenced the particle diameter. The capsid cores in HTLV-1 virions have a polyhedron-like structure. In the 2-dimensional (2-D) slices of the core-containing particles (Fig. 2), linear segments of the capsid shell were either jointed at a sharp angle or connected by a continuous density outline with a smooth curvature. In addition, we occasionally observed spherical cores (Fig. 1). The HTLV-1 cores tended to adjoin the boundary of the viral envelope, and there was at least one curved region or vertex in each capsid that touched the inner face of the viral membrane. This membrane-proximal region is situated ∼13 nm away from the viral envelope, which is similar to the case in HIV-1, where a 12-nm space was observed between the viral membrane and the broad end of the core (24). We observed that four particles had cores with regions of double-layered structure; that is, a second capsid sheet of the same thickness was stacked on top of the primary capsid shell (Fig. 2D). Capsids with double-layered regions have also been observed in HIV-1 (8, 24) and RSV (10, 12). In HTLV-1, the stacked shell sheets observed in the four particles appear to be flat (Fig. 2D). The average electron density inside the complete cores of HTLV-1 virions is very similar to that in the region between the capsid and viral membrane, suggesting that the packing densities of viral or cellular components are similar in these two compartments.

The average size of the particles with evenly distributed electron densities, as measured from the central sections of 156 particles in the tomograms, was determined to be 111 ± 20 nm. For these particles, we did not observe density features attributable to the capsid protein shell, as seen in the mature HTLV-1 virion. We also did not see density attributable to the immature Gag-Gag lattice structure, as seen in HIV-1 mutant samples subjected to partial Gag cleavage (25) or in samples treated with the maturation inhibitor bevirimat (26). Furthermore, we observed particles with incomplete formation of the capsid shell, which suggests that the Gag proteins in these particles may not have been properly assembled when budding out of the cell or that the CA protein failed to reorganize into a complete capsid core. It has been reported that MT-2 cells contain one complete provirus and seven defective proviral sequences (27). A 3.4-kb RNA transcript of the defective proviruses expresses a myristylated truncated Gag protein that is composed of MA, a truncated CA, a short pX region, and two long terminal repeats (28). The 3.4-kb RNA transcript and the truncated Gag proteins have previously been found packaged into virus particles produced from MT-2 cells (20). Since the truncated Gag has only about 30% CA, it is likely that after protease cleavage, the truncated CA does not form proper capsid cores. The assembly of defective Gag may contribute to particles with the evenly distributed internal density or to those particles with partially closed capsid cores. A long-held belief about HTLV-1 particles that is in contrast with data for the particles of HIV-1 and RSV is that cell-free particles are poorly infectious (3). For example, it has been previously reported that a number of T-cell lines (29), B-cell lines, (30), and nonlymphoid cells (31) could be infected with cell-free HTLV-1 particles produced from MT-2 cells but only at very low levels. We have found that the majority of virus particles analyzed from MT-2 cells had evenly distributed electron densities or partially mature particles instead of mature capsid cores. The low frequency of HTLV-1 particles with mature cores, along with the encapsulation of defective RNA transcripts, could help to explain the low infectivity associated with cell-free HTLV-1 particles produced from MT-2 cells.

Intriguingly, we observed in about 10% of the particles analyzed (both mature particles and those with evenly distributed electron densities) a turret structure associated with the viral membrane (Fig. 3A and B; see also Movie S2 in the supplemental material). In some instances, this organization was observed between two particles. In cross section, these turret structures are rectangular or trapezoidal in shape. The wider end of the structure is 30 to 60 nm in length and proximal to the associated viral membrane, while the height of the structure is 10 to 40 nm. We detected no correlation between the size of the turret structure and the size of the virus particle. The boundary of the turret has an electron density as strong as that of the viral envelope. In the tomograms, the turret structures have a filled interior with a density greater than that corresponding to the region inside a viral core. A 2-D cryo-EM image of the turret revealed juxtaposed density stacks with ∼3.5 nm between each density stack (Fig. 3C), which is similar to the distance between the cholesterol layers in the cryo-EM structure of human low-density lipoprotein (32). The turret structure does not resemble the previously reported in vitro-assembled endosomal-sorting complex required for transport III (ESCRT-III), which forms spiral (33) or hollow cylindrical (34) structures. A turret structure such as that observed here with HTLV-1 particles from MT-2 cells has not been previously reported in other HTLV-1 particle analyses with TEM, nor did we observe such a structure in our previous cryo-EM study with HTLV-like particles produced from 293T cells using an HTLV-1 Gag-only expression construct. Also, such a turret structure has not been observed with other retroviruses. It is formally possible that the appearance of the turret structure is cell-type dependent. The molecular composition, origin, and biological relevance of these turret structures in HTLV-1 replication are currently not known but represent a clear focus for further investigations.

FIG 3.

Turret structures on HTLV-1 particles. (A, B) Tomographic slices highlighting four particles with external structures (black arrowheads). (C) One particle with a turret structure on its surface. (D) Enlarged view of the turret structure showing layers of strong density separated by gaps of ∼3.5 nm.

This study has found that the capsid cores of HTLV-1 particles are neither conical, as are those in HIV-1 (8), nor polyhedral, as are the angular cores in RSV (10). Instead, the cores of HTLV-1 were found to be generally polyhedral in shape, with regions of smooth surface. It is possible that hexameric CA proteins are located in the flat facets, such as the double-layered regions of the cores, and that pentameric CAs are located at the corners of irregular polyhedrons. Both CA pentamers and hexamers tilted out of the plane (11, 12) may contribute to the curvature of the capsid shell, which has a continuously curving surface. The malleability of HTLV-1 particle cores may reflect the flexibility of the linkers that tether the NTD and CTD of the CA proteins. Indeed, the solution structure of the full-length CA protein from HTLV-1 indicates that these two domains behave independently, without a preference for a particular relative orientation (35), which suggests that the NTD and CTD of HTLV-1 CA molecules are able to dynamically adopt relative orientations within the assembled capsid core. Higher-resolution structures of the HTLV-1 Gag lattice in the immature particle and CA lattice in the mature particle will help to establish the molecular organization of immature and mature HTLV-1 particles and provide insight into HTLV-1 assembly and maturation.