Migration of Nucleocapsids in Vesicular Stomatitis Virus-Infected Cells Is Dependent on both Microtubules and Actin Filaments

ABSTRACT

The distribution of vesicular stomatitis virus (VSV) nucleocapsids in the cytoplasm of infected cells was analyzed by scanning confocal fluorescence microscopy using a newly developed quantitative approach called the border-to-border distribution method. Nucleocapsids were located near the cell nucleus at early times postinfection (2 h) but were redistributed during infection toward the edges of the cell. This redistribution was inhibited by treatment with nocodazole, colcemid, or cytochalasin D, indicating it is dependent on both microtubules and actin filaments. The role of actin filaments in nucleocapsid mobility was also confirmed by live-cell imaging of fluorescent nucleocapsids of a virus containing P protein fused to enhanced green fluorescent protein. However, in contrast to the overall redistribution in the cytoplasm, the incorporation of nucleocapsids into virions as determined in pulse-chase experiments was dependent on the activity of actin filaments with little if any effect on inhibition of microtubule function. These results indicate that the mechanisms by which nucleocapsids are transported to the farthest reaches of the cell differ from those required for incorporation into virions. This is likely due to the ability of nucleocapsids to follow shorter paths to the plasma membrane mediated by actin filaments.

IMPORTANCE Nucleocapsids of nonsegmented negative-strand viruses like VSV are assembled in the cytoplasm during genome RNA replication and must migrate to the plasma membrane for assembly into virions. Nucleocapsids are too large to diffuse in the cytoplasm in the time required for virus assembly and must be transported by cytoskeletal elements. Previous results suggested that microtubules were responsible for migration of VSV nucleocapsids to the plasma membrane for virus assembly. Data presented here show that both microtubules and actin filaments are responsible for mobility of nucleocapsids in the cytoplasm, but that actin filaments play a larger role than microtubules in incorporation of nucleocapsids into virions.

INTRODUCTION

Nucleocapsids of negative-strand RNA viruses must be transported from their sites of assembly in the cytoplasm to sites of virus budding from host membranes (1). For example, the nucleocapsids of vesicular stomatitis virus (VSV) behave as random coils with a hydrodynamic radius of approximately 90 nm (2), which is too large to diffuse through the cytoplasm in the time required for virus assembly (3). Transport of nucleocapsids to the membrane after assembly in the cytoplasm has been proposed to occur primarily along microtubules (4). The goal of the experiments presented here was to further test mechanisms of nucleocapsid transport by evaluating both microtubule-dependent and actin-dependent transport using recently developed analytical tools.

Actin filaments and microtubules have a general orientation in which the growing (plus) end is oriented toward the cell periphery and the minus end is oriented toward the center of the cell (5). Assembly of microtubules is usually nucleated at the microtubule organizing center near the nucleus, and they radiate long distances toward the cell periphery. In the case of actin filaments, there are both radially oriented and tangentially oriented fiber systems, especially at the cell periphery, with extensive connections between the two systems (6). These transport systems are given their sophistication by the wide variety of molecular motors, adapter proteins, and regulatory proteins with which their cargoes interact (5). In principle, any cellular element, such as viral nucleocapsids, can move in either direction on either actin filaments or microtubules. The distribution within the cytoplasm then depends on the relative affinity for the different molecular motors and adapter proteins, the relative abundance of these proteins in the cell, and the effects of regulatory proteins that govern the time of residence on any given path. Thus, there is probably no single transport mechanism responsible for distribution of nucleocapsids. As a result, it is likely that there is no single destination to which nucleocapsids are transported, but instead, they are distributed throughout the cell according to the relative activities of the different transport mechanisms with which they are associated.

We have developed new cellular imaging analyses to quantify the effects of experimental perturbations on the distribution of elements like viral nucleocapsids or cellular organelles, which we call the border-to-border distribution method (7). In the experiments described here, the borders are the nucleus and the plasma membrane at the edge of the cell, i.e., the borders that define the cytoplasm. The goal of this approach is to provide a quantitative description of the distribution of elements in individual cells using mathematical parameters used to describe the distribution of any population (i.e., mean, standard deviation, skew, and kurtosis). Statistical methods can then be used to analyze results from many cells to determine a “representative” distribution and to determine whether experimental perturbations have a statistically significant effect on any of the distribution parameters. This removes much of the subjectivity associated with determining which images are representative and has the power to reveal quantitative differences that may not be obvious by visual inspection alone due to heterogeneity in distributions among different individual cells.

As presented here, nucleocapsids in VSV-infected cells were analyzed by the border-to-border distribution method, which showed that nucleocapsids are originally located near the nucleus and redistribute during the first 2 to 6 h of infection toward the edges of the cell. This redistribution is dependent on both microtubules and actin filaments, as shown by the effects of cytoskeletal inhibitors on nucleocapsid distribution. The role of actin filaments in nucleocapsid mobility was also confirmed by live-cell imaging of fluorescent nucleocapsids. However, in contrast to the overall redistribution in the cytoplasm, the incorporation of nucleocapsids into virions was dependent on the activity of actin filaments with little if any effect of inhibition of microtubule function. These results indicate that the mechanisms by which nucleocapsids are transported to the farthest reaches of the cell differ from those required for incorporation into virions. This is likely due to the ability of nucleocapsids to follow shorter paths to the plasma membrane mediated by actin filaments.

MATERIALS AND METHODS

Cells and virus.

Wild-type VSV (Indiana serotype, Orsay strain) or recombinant VSV containing a gene for enhanced green fluorescent protein (eGFP) fused in frame with the P gene [VSV-PeGFP (4)] was grown at 37°C in BHK cells in Dulbecco's modified Eagle medium containing 2% fetal bovine serum. HeLa cells were cultured in Dulbecco's modified Eagle medium containing 7% fetal bovine serum. Infections were carried out at a multiplicity of infection of 50 PFU per cell to rapidly synchronize infection.

Treatment with cytoskeletal inhibitors.

Cells were pretreated with nocodazole (33 μM [Acros Organics]) or colcemid (27 μM [Sigma Chemical Company]) beginning 1 h prior to infection and continuing for the period of infection described in the figure legends. Cells were treated with cytochalasin D (20 μM [Sigma Chemical Company]) or latrunculin A (1 μM) beginning at 1 h postinfection (hpi). The inhibitor concentrations to be used in these experiments were determined in pilot experiments as concentrations that did not inhibit viral protein synthesis by more than 50%, determined by incorporation of [³⁵S]methionine as described below. Effectiveness of cytoskeletal disruption was determined by labeling with antibody against α/β-tubulin (Cell Signaling Technology, catalog no. 2148) or Alexa 594-phalloidin (Molecular Probes) according to the manufacturers' protocols.

Confocal fluorescence microscopy.

HeLa cells were seeded at a density of 3 × 10⁵ on 8-well chamber slides (Millipore) and were infected with VSV for 2, 4, or 6 h. The cells were washed with phosphate-buffered saline and then labeled with PKH26 (1:500 in diluent C; Sigma-Aldrich). Samples were then fixed with 3.7% paraformaldehyde (Sigma-Aldrich) for 15 min and permeabilized with blocking buffer containing 1% bovine serum albumin (Sigma-Aldrich) and 0.1% saponin (Calbiochem) for 10 min. The samples were then labeled with anti-N protein antibody for 60 min and then with fluorescent secondary antibody for 60 min. The primary antibody used for these experiments was mouse anti-N protein monoclonal antibody 10G4 (8). The secondary antibody used was Alexa Fluor 488-conjugated rabbit anti-mouse IgG (Molecular Probes). ProLong Gold antifade mountant with DAPI (4′,6-diamidino-2-phenylindole [Life Technologies]) was added to label the nucleus. Confocal images for analysis by the border-to-border distribution method were collected on a Nikon C1Si confocal microscope system with a 60× multi-immersion objective at an x-y resolution of 3.5 pixels/μm and an optical slice size (z depth) of 1 μm. Higher-resolution images for publication were obtained at an x-y resolution of 13 pixels/μm. Images for Fig. 1 were collected on a Zeiss 510 LSCM confocal fluorescence system at an x-y resolution of 16.25 pixels/μm and a z depth of 0.45 μm.

FIG 1.

Distribution of VSV nucleocapsids changes during infection. HeLa cells were infected with VSV for 2 (a), 4 (b), or 6 (c) h and then were fixed, permeabilized, and labeled with antibody against N protein. The outlines of the cell periphery were determined from differential interference contrast images of the same fields. Scale bars, 10 μm.

Data collection.

Stacks containing images of the DAPI, PKH26, and N protein fluorescence of the same field were analyzed by ImageJ, v.1.48q (9), as described previously (7). Briefly, for each cell to be analyzed, the borders of the nuclei were defined from the DAPI image, and the edge of the membrane labeling was defined from the PKH26 image using the selection tools provided by ImageJ software. The centroid of the nucleus was determined from which segments radiate, and the cytoplasm was divided into 360 radial segments that start at the first border (the nucleus) and end at the second border (the edge of the cell). The segments were numbered beginning with the longest segment. Segments that crossed more than two borders were excluded from the analysis because there would be a gap in the data following the second border. The data collected for the pixels along each segment in the N protein image were segment number, angle to centroid, pixel number along the segment, distance in native units if any were associated with the image, normalized (fractional) distance between the nucleus and the edge of the cell, and fluorescence intensity.

Data analysis.

Data analysis was performed with the R statistical package (10). Histograms of pixel intensities were analyzed to determine the fluorescence of the population of low-intensity pixels that was considered to represent the background of diffuse labeling from soluble forms of N protein, which was subtracted from the data. Segments from each cell were sorted according to their lengths and selected for analysis as described in the figure legends. The distribution parameters calculated for nucleocapsid fluorescence in the selected segments were (i) mean fractional distance of the fluorescence, (ii) standard deviation in the distance distribution, (iii) skew in the distance distribution, and (iv) kurtosis in the distance distribution, calculated as described previously (7). Of these parameters, only the mean fractional distance and skew changed significantly during infection. Differences in distribution parameters derived from multiple cells were tested for statistical significance by analysis of variance with Tukey's post hoc test for multiple comparisons, for cases in which values were normally distributed. Data that were not normally distributed were analyzed by Kruskal-Wallis analysis of variance on ranks with Dunn's method of multiple pairwise comparisons. Analysis of variance and calculation of Pearson correlation coefficients (r) were performed using the R, SigmaStat, or Excel software packages. The fitted curves in Fig. 3 were determined by nonlinear least-squares analysis using the formula for a skewed normal distribution in Excel. It is important to point out that the formula needed to be fit to the same boundaries as the data, since the calculated skew of an infinite distribution would be different.

FIG 3.

Distribution of nucleocapsid fluorescence in the longest segments of representative cells. VSV-infected cells were analyzed as described in the legend to Fig. 2 at 2 (a), 4 (b), and 6 (c) hpi. Data from an individual cell are shown for the longest segment and surrounding segments that were ≥90% of its length. Blue dots are individual pixel intensities. The red lines are the average intensity at each distance. The green lines are fitted curves that have the same mean, standard deviation, and skew as the data.

Live-cell imaging of fluorescent nucleocapsids.

HeLa cells were infected with VSV-PeGFP and analyzed by live-cell fluorescence imaging as described previously (11). Briefly, infected cells were untreated or treated with cytochalasin D and imaged at 4 h postinfection (hpi) by a Nikon Eclipse Ti inverted epifluorescence microscope using a 60× NA 1.4 oil-immersion objective. An automated shutter in the excitation light path illuminated the cells in 10 1-s bursts separated by 9-s dark intervals to reduce photobleaching. A high-speed complementary metal oxide semiconductor (CMOS) camera (PCO, Romulus, MI) synchronized with the shutter acquired 100 images during each 1-s burst, thus collecting data every 10 ms. The pixel size of the camera was 6.5 μm by 6.5 μm. This small pixel size ensured that the point spread function of light at the object plane, which has a Gaussian width of 0.080 μm for our objective, illuminated about 7 pixels of the CMOS detector. This allowed individual particles to be tracked to subpixel precision, approximately ±10 nm at the object plane, by Video Spot Tracker software (CISSM, University of North Carolina, Chapel Hill, NC). Note that this precision is much smaller than the width of the point spread function and also much less than the Rayleigh criterion for the resolution of two particles. The coordinates were then processed to remove stage drift and cell migration artifacts, and the mean square displacement (MSD) as a function of time interval (τ) for each particle was calculated using MATLAB software as described previously (11).

Pulse-chase analysis.

Cells were pretreated with nocodazole or colcemid for 1 h before infection or treated with cytochalasin D or latrunculin A beginning 1 h after infection. Medium containing [³⁵S]methionine (25 μCi/ml) ± cytoskeletal inhibitors was added 1 h postinfection for 3 or 5 h, followed by a 1-h chase in nonlabeled medium. The supernatant was collected and spun at 1,000 × g to remove cell debris and then placed on a 15% sucrose cushion and spun at 114,000 × g for 1 h. The virion pellet and cell extract were analyzed by SDS-PAGE and phosphorimaging by a Typhoon FLA 9500 imager (GE Healthcare) and quantified using ImageQuant software as described previously (12).

RESULTS

Border-to-border distribution method for analysis of VSV nucleocapsid distribution in the cytoplasm of infected cells.

The goal of the experiments presented here was to quantify changes in the distribution of VSV nucleocapsids in the cytoplasm of infected cells by fluorescence microscopy and to determine the relative contributions of microtubules versus actin filaments in the nucleocapsid distribution. VSV nucleocapsids are organized into small clusters that function much like inclusion bodies (13). During the course of VSV infection, viral nucleocapsids appear to be redistributed from locations near the nucleus to locations near the edge of the cell, as shown by the examples in Fig. 1. In these experiments, HeLa cells were infected with VSV, and at 2, 4, and 6 h postinfection (hpi), cells were fixed, permeabilized, and labeled for immunofluorescence microscopy with antibody against VSV N protein. At these time points, the distribution of viral components is not likely to be affected by viral cytopathic effects, which begin somewhat later at around 10 hpi (14). The VSV N protein is present in multiple forms in infected cells, including soluble forms as well as ribonucleoprotein particles, such as messenger RNPs (mRNPs) and nucleocapsids (15). The soluble N protein and mRNPs form a faint diffuse background labeling that is hardly visible in the images in Fig. 1. The majority of the bright fluorescence visible in Fig. 1 was due to nucleocapsids, since individual nucleocapsids contain approximately 1,200 copies of N protein (16) and form clusters of multiple nucleocapsids (13) with various sizes and intensities.

At 2 hpi, the nucleocapsid fluorescence appeared in a punctate pattern in the cytoplasm and was primarily localized near the nucleus. At 4 hpi, the clustering of nucleocapsids appeared to be reduced, and the fluorescence was more evenly distributed throughout the cytoplasm. At 6 hpi, the fluorescence appeared to be distributed toward the edges of the cell. These images suggest that there is a shift in localization of nucleocapsids as the infectious cycle progresses. However, this conclusion is limited by the subjective judgment of which images are “representative” and by the heterogeneity in the size and shape of HeLa cells. Both of these limitations are addressed by an approach we have called the border-to-border distribution method.

The border-to-border distribution method is a new method of analysis of cellular imaging data to quantify the distribution in the cytoplasm of particles such as viral nucleocapsids and cellular organelles (7). The method consists of four steps: (i) fluorescence labeling and imaging of infected cells, (ii) division of the image of the cytoplasm into segments, (iii) selection of segments of interest, and (iv) distribution analysis of pixel intensities as a function of distance along the selected segments. Steps 1 and 2 are illustrated in Fig. 2. In order to define the borders of the cytoplasm, VSV-infected cells were labeled with the fluorescent membrane probe PKH26 to label cellular membranes and then were fixed, permeabilized with saponin (which does not solubilize PKH26), and labeled with antibody against the VSV N protein and DAPI to label nuclei. Images were obtained by confocal microscopy using a slice size and focal plane that captured both the plasma membrane at the bottom of the cell and a portion of the nucleus. For each cell to be analyzed, the borders of the nuclei and the edge of the membrane labeling were delineated using the selection tools provided by ImageJ software (Fig. 2a and b). The centroid of the nucleus was determined, from which segments radiate (cross in Fig. 2), and the cytoplasm was divided into 360 radial segments that start at the first border (the nucleus) and end at the second border (the edge of the cell). For clarity, only 90 segments are shown in Fig. 2d. The segments were numbered beginning with the longest segment. Segments that crossed more than two borders were excluded from the analysis because there would be a gap in the data following the second border. The data set for each cell consisted of the fluorescence intensity and distance of each pixel along the approximately 360 segments included in the analysis.

FIG 2.

Image analysis of VSV-infected cells for the border-to-border distribution method. VSV-infected cells were labeled with DAPI (a), PKH26 (b), and antibody against N protein (c and d). For individual cells, the selection tools in ImageJ software were used to define the border of the nucleus from the DAPI image (a) and the edges of the cell from the PKH26 image (b). The centroid of the nucleus was determined, from which segments radiate (cross), and the cytoplasm was divided into 360 radial segments that start at the first border (the nucleus) and end at the second border (the edge of the cell). For clarity, 90 segments are shown (d).

The next steps in the border-to-border distribution method are selection of the segments of interest and distribution analysis of pixel intensities as a function of distance along the selected segments. The segments of interest can be selected by any criterion that can be mathematically defined (length, intensity, etc.). In the following experiments, we chose to analyze the longest segments in each cell, because in the well-spread, relatively flat HeLa cells used here, both microtubules and actin filaments extend along the longest segments of the cell in this dimension, and these segments are most likely to require extensive motor activity for nucleocapsids to travel the longest distances. Figure 3 shows the distribution of pixel intensities as a function of distance for the longest segments of three individual cells at 2, 4, or 6 hpi (8 to 10 total segments for each cell). The y axis is the fluorescence intensity of VSV N protein, the x axis is the distance between the nucleus and the edge of the cell expressed as a percentage, and the dots are data for individual pixels in the longest segments. Background intensity contributed by soluble N protein (typically about 100 U) was subtracted from the data. The red lines are the mean pixel intensities for all of the selected segments. The green curves are theoretical normal distributions with the same mean distance, standard deviation, and skew as those calculated from the data. The mean distances from the nucleus to the edge of the cell for the nucleocapsids in Fig. 3 were 39%, 51%, and 58% at 2, 4, and 6 hpi, respectively. In the cell infected for 2 h, the data are skewed with a peak toward the nucleus. By convention, this skew is positive (0.8). The data from the cell infected for 4 h are not as skewed (skew is near zero), and data from the cell infected for 6 h are skewed with a peak toward the edge of the cell (skew is negative, −0.4).

Effects of cellular heterogeneity on distribution of nucleocapsids.

The utility of the border-to-border distribution method is that quantitative data from multiple cells can be combined to address issues of cellular heterogeneity by statistical analysis. Figure 4 addresses two types of cellular heterogeneity in HeLa cells—heterogeneity within the cytoplasm of individual cells and heterogeneity in morphology among different cells. Approximately 10 cells from each time point in three separate experiments (approximately 30 cells total for each time point) were analyzed for the mean distance and skew of the nucleocapsid distribution. To address the question of heterogeneity within the cytoplasm of individual cells, the segments in each cell were divided into deciles (90th to 100th percentiles, 80th to 90th percentiles, etc.) according to their lengths to determine whether the nucleocapsid distribution in longer segments differed from that in shorter segments. Figure 4a and b show the mean distance and skew of the nucleocapsid distribution from the 50th to 100th percentiles. Segments shorter than the 50th percentile often had too few pixels (<10) to accurately determine the distribution parameters.

FIG 4.

Effect of cellular heterogeneity on distribution of nucleocapsids. VSV-infected cells were analyzed as described in the legend to Fig. 2 at 2 hpi (n = 33 cells), 4 hpi (n = 33), or 6 hpi (n = 29). The mean fractional distance of nucleocapsids between the nucleus and the edge of the cell (a) and skew in the nucleocapsid distribution (b) were determined for segments in each decile for length between the 50th and 100th percentiles. The data shown are averages from all cells at each time point. For clarity, standard deviations are shown for 6 hpi only. *, P < 0.05 versus 2 hpi. The mean fractional distance (c) and skew (d) of nucleocapsid distribution of segments in the 90th to 100th percentiles of individual cells are shown as a function of mean segment length. Straight lines represent linear regression.

At 6 hpi, the mean distances of the nucleocapsid distribution were significantly greater than those at 2 hpi from the 60th to 100th percentile (Fig. 4a). Similarly the mean distances at 4 hpi were significantly greater than at 2 hpi for segments from the 70th to 100th percentile, confirming the impression given by individual cells (Fig. 1 and 3) that nucleocapsids were redistributed during infection toward the edges of the cell. Analysis of the skew in the nucleocapsid distribution showed similar trends, although only segments in the 90th to 100th percentile at 6 hpi were significantly more negative than those at 2 hpi, due primarily to large variation in the skew at 2 hpi. Also apparent in Fig. 4a and b was a significant correlation between segment length and both mean distance and skew at 4 and 6 hpi (r = 0.97, P < 0.05 for mean distance at both time points, r = −0.89 and −0.98, P < 0.05 for skew at 4 and 6 hpi, respectively), whereas there was no significant correlation at 2 hpi (r = −0.11 and 0.38, P > 0.05 for mean distance and skew, respectively). This analysis shows that the redistribution of nucleocapsids was more prominent in the longer segments of the cell. This is the opposite result from that predicted if nucleocapsids move at similar rates in all segments and likely reflects heterogeneity in the cytoplasm. For example, there may be greater motor activity in the longer segments, or nucleocapsids may differ in their degree of maturation and association with molecular motors during the time they spend traveling in longer segments.

The morphology of the HeLa cells in these experiments was variable, as is typical, ranging from very flat, well-spread cells, whose longest segments were up to 60 pixels in length (11.7 μm), to more rounded cells, whose longest segments were as short as 15 pixels (2.9 μm). Figure 4c and d show the mean distance and skew of the nucleocapsid distributions at 6 hpi in individual cells for their segments in the 90th to 100th percentile, plotted as a function of mean segment length in the 90th to 100th percentile. There was no significant correlation between either mean distance or skew in the nucleocapsid distribution and length of the longest segments (r = −0.22 and 0.16, P > 0.05 for mean distance and skew, respectively). This result indicates that the distribution of nucleocapsids between the nucleus and edge of the cell was largely independent of cellular morphology as reflected in the length of the longest segments.

Role of microtubules and actin filaments in nucleocapsid distribution.

The role of microtubules in the distribution of nucleocapsids was tested by treating infected cells with nocodazole or colcemid beginning 1 h prior to infection. Typical images obtained at 6 hpi are shown in Fig. 5. The impression given by these images is that nucleocapsids remain clustered near the nucleus in cells treated with nocodazole or colcemid. This effect was quantified by the border-to-border distribution method. The mean distance and skew of nucleocapsid fluorescence in the longest segments of approximately 30 cells at each time point at 2, 4, and 6 hpi are shown in Fig. 6a and b. In contrast to control untreated cells, there were no significant differences in drug-treated cells in the distributions of nucleocapsids as a function of time postinfection. By 6 hpi, the differences between control and drug-treated cells were statistically significant. These data indicate that microtubules contribute to the redistribution of nucleocapsids from near the nucleus toward the most distant edges of the cell. In addition, the mean distance in nocodazole-treated cells was significantly greater than that in the control at 2 hpi, suggesting that microtubules play a role in localizing nucleocapsids near the nucleus at early times postinfection.

FIG 5.

Treatment of VSV-infected cells with nocodazole or colcemid. Cells were untreated (a and b) or treated with nocodazole (c and d) or colcemid (e and f) beginning 1 h prior to infection with VSV and labeled at 6 hpi as described in the legend to Fig. 2. The images shown are the merged images for DAPI, PKH26, and N protein labels (a, c, and e) or N protein label with the borders of the nuclei and edges of the cells defined as in the legend to Fig. 2 (b, d, and f).

FIG 6.

Effect of treatment with cytoskeletal inhibitors on nucleocapsid distribution. Control cells (con) or cells treated with nocodazole (NOC) or colcemid (COL) beginning 1 h prior to infection (a and b) or treated with cytochalasin D (CD) or latrunculin A (LatA) beginning 1 hpi (c and d) were analyzed by the border-to-border distribution method at 2, 4, or 6 hpi (a and b) or at 6 hpi (c and d). Data are mean fractional distance (a and c) and skew (b and d) in the nucleocapsid distribution in the longest segment and surrounding segments that were ≥90% of its length in approximately 30 cells at each time point. The data shown are means ± standard deviations. *, P < 0.05.

The role of actin filaments in nucleocapsid distribution was tested by treating infected cells with cytochalasin D or latrunculin A. Drugs were added 1 h after infection because of the requirement of actin filament activity for virus penetration (17, 18). Images obtained at 6 hpi are shown in Fig. 7, and quantification of multiple experiments is shown in Fig. 6c and d. The effects of treatment with cytochalasin D appeared similar to the effects of the microtubule inhibitors in Fig. 5. There was relatively little effect of drug treatment on the overall morphology of the cells (Fig. 7c and d), and the nucleocapsids primarily remained clustered near the nucleus, as reflected in the mean distance (Fig. 6c) and skew (Fig. 6d) of their distributions. These results indicate that actin filaments play a role in the distribution of nucleocapsids toward the edge of the cell.

FIG 7.

Treatment of VSV-infected cells with cytochalasin D or latrunculin A. Cells were untreated (a and b) or treated with cytochalasin D (c and d) or latrunculin A (e and f) beginning 1 hpi with VSV and labeled at 6 hpi as described in the legend to Fig. 2. The images shown are the merged images for DAPI, PKH26, and N protein labels (a, c, and d) or N protein label with the borders of the nuclei and edges of the cells defined as in the legend to Fig. 2.

The effects of treatment with latrunculin A were quite different from those of cytochalasin D. The cell morphology was altered such that the cells were rounded and the nuclei were misshapen and eccentrically located near one edge of the cell (Fig. 7e and f). Alterations in nuclear morphology in response to treatment with latrunculins have been observed previously (19, 20). However, the cell rounding appeared to be due to the combined cytopathic effects of drug treatment and virus infection, since neither treatment alone induced these cytopathic effects (not shown). The striking difference from the effects of cytochalasin D in terms of the purpose of the experiment was the nucleocapsid distribution, in which the nucleocapsids were primarily located only at the edges of the cells, as is clear in the images in Fig. 7 and as shown by the dramatic increase in the mean distance and negative skew in their distribution in Fig. 6c and d.

Latrunculin A causes disassembly of preexisting actin filaments, whereas cytochalasin D prevents polymerization of actin and depends on normal filament turnover for filament disassembly (21, 22). Thus, the difference in effects of cytochalasin D and latrunculin A could be due to activity of residual actin filaments that were not disassembled following cytochalasin D treatment. To test this hypothesis, VSV-infected cells were treated with cytochalasin D, latrunculin A, or the control treatment and labeled using fluorescent phalloidin, which binds to filamentous actin. In control cells, filamentous actin was distributed throughout the cytoplasm (Fig. 8a). In cytochalasin D-treated cells, filamentous actin was primarily in small clusters that were concentrated around the nucleus (Fig. 8b). In contrast, there was very little filamentous actin in cells treated with latrunculin A (Fig. 8c). These results support the idea that residual actin filaments in the presence of cytochalasin D are responsible for retention of nucleocapsids near the nucleus, whereas in the absence of filamentous actin, the distribution of nucleocapsids toward the edge of the cell is enhanced, with the caveat that the enhanced distribution could be an indirect result of the cytopathic effect of latrunculin A treatment.

FIG 8.

Reduction in filamentous actin after treatment with cytochalasin D or latrunculin A. Cells were untreated (a) or treated with cytochalasin D (b) or latrunculin A (c) beginning 1 h postinfection with VSV and labeled with Alexa 594-phalloidin and DAPI at 6 hpi. Images were colored yellow (phalloidin) and magenta (DAPI) so that the phalloidin labeling is more readily visible to the human eye.

Live-cell imaging of actin-dependent mobility of nucleocapsids.

The role of actin filaments in nucleocapsid mobility was confirmed by live-cell imaging of fluorescent nucleocapsids. HeLa cells were infected with a recombinant VSV in which the gene for eGFP was fused in frame with the sequence encoding the hinge region of the P protein (VSV-PeGFP) (4). Cells were infected with VSV-PeGFP, and control and cytochalasin D-treated cells were imaged at 4 hpi. Cells were illuminated in 10 1-s bursts separated by 9-s dark intervals. A high-speed camera acquired 100 images during each 1-s burst, thus collecting data every 10 ms. Individual particles appeared to be clusters of nucleocapsids as described previously (13). Fluorescent particles were tracked to subpixel precision (11). During the 90-s time course of the data collection, few particles underwent the rapid directional movement characteristic of microtubule-dependent transport. These particles were not analyzed, since their movement has been well described previously (4). Figure 9a and b show the tracks of representative particles from untreated and cytochalasin D-treated cells. The seemingly random nature of the movement in untreated cells is apparent from both the time scale of each 1-s burst (indicated by different colors), as well as the overall migration during the 91-s time course. However, both the rapid movement and the overall migration were dramatically inhibited by treatment with cytochalasin D, indicating that they are dependent on actin filaments.

FIG 9.

Particle tracking of fluorescent nucleocapsid clusters in live-cell imaging experiments. Cells were infected with VSV-PeGFP and were untreated or treated with cytochalasin D beginning at 1 hpi and then imaged at 4 hpi. Cells were illuminated in 10 1-s bursts separated by 9-s dark intervals, and 100 images were collected during each 1-s burst. Individual particles were tracked to subpixel precision. Representative tracks of a control (a) or cytochalasin D-treated (b) cell are shown. Different 1-s time intervals are indicated by different colors. The scale for parts a and b is 0.108 μm/pixel. Mean square displacements (MSD) were determined as a function of time delay between measurements (τ) for each particle and are shown in panel c as the average log MSD + standard deviation from control cells (open symbols) and cytochalasin D-treated cells (solid symbols) in two independent experiments (circles and squares).

Particle tracking data were quantified by calculating the mean square displacement (MSD) of a particle as a function of time delay (τ) between two measurements. In a purely viscous medium, the MSD is proportional to time, with a slope proportional to the diffusion coefficient, or in a plot of log(MSD) versus log(τ), the slope is 1, and the y values are related to the diffusion coefficient. In a purely elastic medium, the MSD is independent of time, and its magnitude depends on the elasticity of the medium. In the log-log plot, the slope is zero and the y values are related to the elasticity. In a viscoelastic medium, the slope is between 0 and 1. If the direction of the motion is not random, the slope of the log-log plot is >1, and the motion is said to be directed.

Figure 9c shows log-log plots of MSD versus time delay (τ) averaged for individual particles from control and cytochalasin D-treated cells. The tracks from control cells were truncated at τ = 10 s, because some of these particles migrated out of the focal plane or encountered other particles, which the particle-tracking software could not distinguish. At times of <1 s, the slope of the curve was nearly 0, indicating that the motion was primarily elastic on this time scale, whereas at times of >1 s, the slope increased between 0 and 1, indicating both diffusion-like and elastic-like motions. It is important to point out that the lower limit of detection due to system noise is 0.00010 μm² independent of τ (11). If the noise were significantly larger, the flatness of the MSD curve at small values of τ could otherwise be an artifact of system noise. The key result is that both elastic-like and diffusion-like motions were dramatically inhibited by treatment with cytochalasin D, indicating that they are both actin dependent.

Effects of cytoskeletal inhibitors on virus release.

Reports have been inconsistent in describing the effect on virus release of drugs that disrupt the cytoskeleton (4, 23–25). One of the problems is that viral mRNAs require cytoskeletal function to be efficiently translated (13, 26). As a result, these drugs typically inhibit viral protein synthesis to some extent. A pulse-chase labeling protocol was used to control for this effect. In the experiments in Fig. 10, VSV-infected cells were untreated or were treated with nocodazole, colcemid, or cytochalasin D. The cells were labeled continuously with [³⁵S]methionine to uniformly label viral proteins until 4 or 6 hpi and then were chased in nonlabeled medium for 1 h. The virions released in the 1-h chase were harvested from the medium by ultracentrifugation, and the virions and cell extracts were analyzed by SDS-PAGE and phosphorimaging. Figure 10a and b show representative images and quantification of the N and P protein bands in these images. Treatment with nocodazole reduced viral protein accumulation in the cell extract to approximately 60% of the control, similar to previous data (13), whereas treatment with colcemid had relatively little effect. Similarly, treatment with cytochalasin D caused a modest decrease in the amount of N and P proteins present at 6 hpi in the cell extract. The effects of nocodazole and colcemid on N and P protein incorporation into virions were similar to their effects on protein accumulation in infected cells, whereas cytochalasin D had a much more dramatic effect on incorporation into virions. To quantify the effects on virus assembly from multiple experiments, the ratio of N and P protein present in virions compared to that in cell extracts was normalized to the ratio in the control samples (Fig. 10c). At 4 hpi, neither microtubule inhibitor had an effect on virus assembly that could not be accounted for by the decrease in protein accumulation. At 6 hpi, colcemid had a modest effect—around 75% compared to that of the control. Cytochalasin D, however, had a substantial effect at both 4 and 6 hpi. These data indicate that actin has a pivotal role during assembly of nucleocapsids into virions.

FIG 10.

Effect of cytoskeletal inhibitors on virus assembly. Cells were treated with nocodazole (N) or colcemid (C) for 1 h prior to infection with VSV (a), or with cytochalasin D (CD) beginning at 1 hpi (b). con, control. Cells were labeled with [³⁵S]methionine for 3 or 5 h beginning at 1 hpi and chased in nonlabeled medium for 1 h. Cell extracts and virions released during the chase were analyzed by SDS-PAGE and phosphorimaging. The N and P protein bands were quantified, and drug-treated samples were normalized to controls as illustrated in panels a and b. The ratio of N plus P protein in virions to N plus P protein in the cell extract is shown as the mean + standard deviation from 3 independent experiments (c). Infectious virus in the medium from the chase at 6 hpi was determined by plaque assay (d). The data shown are the mean + standard deviation from 3 independent experiments. *, P < 0.05 versus control.

These results were confirmed by analysis of virus yield by plaque assays. Treatment with nocodazole or colcemid reduced virus yield slightly by less than a log (likely a result of inhibition of protein synthesis), whereas treatment with cytochalasin D reduced virus yield by almost 2 logs (Fig. 10d). These data indicate that actin-based transport is more important than microtubule-based transport for incorporation of nucleocapsids into virions. The observation that nocodazole and colcemid inhibit the redistribution of nucleocapsids toward the edge of the cell (Fig. 6) but not the incorporation into virions (Fig. 10) shows that it is not necessary for nucleocapsids to be transported to the farthest reaches of the cell in order to be incorporated into virions. This is likely due to their ability to follow shorter paths to the plasma membrane mediated by actin filaments.

DISCUSSION

Transport along microtubules and actin filaments are the two main cellular mechanisms used in movement of cargo (5). Viruses use these pathways to transport proteins and genetic information within infected cells during their replication cycles. Earlier data showed that VSV nucleocapsids migrate in association with microtubules (4), but our data have shown that actin filaments are also important not only in nucleocapsid migration (Fig. 6, 7, and 9) but also virus assembly (Fig. 10). At early times postinfection (e.g., 2 hpi), VSV nucleocapsids are present in well-defined clusters that resemble inclusion bodies of other viruses, such as the related rhabdovirus rabies virus (13). These clusters are located primarily in the perinuclear region of the cell (4, 13) (Fig. 1, 3, and 4). As the infection progressed, the extent of clustering decreased (Fig. 1), and nucleocapsids were redistributed toward the edges of the cell (Fig. 1, 3, and 4). This movement was dependent on both microtubules (Fig. 5 and 6) and actin filaments (Fig. 6, 7, and 9). However, inhibition of microtubule function had relatively little effect on the efficiency of incorporation of nucleocapsids into virions, whereas inhibition of actin-dependent mobility had a more pronounced effect (Fig. 10). The observation that inhibition of microtubule function inhibited nucleocapsid redistribution but not incorporation into virions implies that redistribution of nucleocapsids toward the edges of the cell is not essential for virus assembly. This is likely due to the ability of nucleocapsids to migrate along shorter paths to the sites of assembly at the plasma membrane mediated by actin filaments.

The redistribution of nucleocapsids was quantified in these experiments by our recently developed method of analysis, which we have termed the border-to-border distribution method (7). This method was based on the realization that there is no single destination for cellular particles and organelles, and there is no single transport mechanism that governs their distribution (5). Instead, the distribution of intracellular particles such as nucleocapsids or organelles such as membranes of the secretory pathway depends on the activity of the large number of different molecular motors and adapter proteins that couple their transport to cytoskeletal elements (5). In particular, they are acted upon by both plus-end- and minus-end-directed motors of both microtubules and actin filaments. If plus-end-directed motors predominate, the distribution will be skewed with a peak toward the edges of the cell, and if minus-end-directed motors predominate, the distribution will be skewed with a peak toward the nucleus. For example, we validated the border-to-border distribution method by analyzing the distribution of VSV G protein in perinuclear membranes of the Golgi apparatus and its disruption by treatment with brefeldin A (7).

The effects on nucleocapsid distribution of drugs that inhibit the function of cytoskeletal elements can be explained by a model in which microtubule-associated transport is primarily due to plus-end-directed motors, i.e., members of the kinesin family, whereas actin-dependent transport involves a more balanced distribution of plus-end- and minus-end-directed motors, which would be different members of the myosin family. The predominance of plus-end-directed motors associated with microtubules is consistent with inhibition of redistribution of nucleocapsids toward the farthest edges of the cell by nocodazole and colcemid (Fig. 5 and 6) and also with the dramatic redistribution toward the edges of the cell when actin filaments were largely disrupted in the presence of latrunculin A (Fig. 6 and 7), with the caveat that the enhanced redistribution could be an indirect result of the cytopathic effect of latrunculin A treatment. The more balanced distribution mediated by actin filaments is consistent with the contrasting effects of treatment with cytochalasin D and latrunculin A. In the presence of residual actin filaments in cells treated with cytochalasin D (Fig. 8), mobility of nucleocapsids toward the edges of the cell was slowed but not prevented completely (Fig. 6, 7, and 9). This result indicates that actin-dependent mobility is involved in movement toward the edges of the cell, which is generally in the plus-end-oriented direction for actin filaments as well as microtubules. The enhanced distribution toward the edges of the cell when actin filaments were disrupted in the presence of latrunculin A suggests that actin-dependent mobility also acts to slow the plus-end-directed movement along microtubules. This opposition to microtubule-based motility would likely be mediated by minus-end-directed members of the myosin family. The more balanced distribution resulting from actin-dependent transport is also consistent with the observation that the skewed distribution of nucleocapsids toward the edges of the cell was more pronounced in the longer segments than the shorter segments of the cytoplasm (Fig. 4), since long-distance transport tends to be more dependent on microtubules, whereas short-distance transport tends to be actin dependent (5).

The importance of actin filaments in movement of nucleocapsids was also demonstrated by tracking of fluorescent nucleocapsids in live-cell imaging experiments (Fig. 9). Particle tracking data were quantified by modeling actin-dependent motion mathematically as random diffusion of particles in a viscoelastic medium, even though the movement is not due to diffusion (i.e., thermal motion), nor is it likely to be random on longer time scales. A similar approach has been used recently to describe the actin-dependent motion of polystyrene nanospheres microinjected into different cell types and the motion of fluorescent peroxisomes in breast cancer cells (11, 27). On short time scales (<1 s), the motion of nucleocapsids was primarily elastic, resembling movement of particles attached to a flexible tether or particles confined within a meshwork of filaments. On longer time scales (>1 s), the motion appeared to be viscoelastic, indicating both diffusion-like and elastic-like motions. The key result was that the elastic-like and diffusion-like motions were both actin dependent, as shown by their inhibition by treatment with cytochalasin D.

Both microtubules and actin filaments are involved in multiple steps of the VSV replication cycle. For example, actin filaments are involved in VSV penetration into host cells. Penetration is initiated by clathrin-dependent endocytosis but is completed by actin filaments, because the VSV virion is too large to be completely engulfed by clathrin-coated vesicles (17, 18). Although virus penetration does not appear to be dependent on microtubules, microtubules may play a role in the localization of endocytic vesicles involved in virus penetration (5). This is a possible explanation of why treatment with nocodazole affects the distribution of nucleocapsids at early times postinfection (e.g., 2 hpi in Fig. 6), although this effect was not statistically significant following treatment with colcemid (Fig. 6). These cytoskeletal elements also affect translation of viral mRNAs. Earlier data showed that VSV mRNAs are translated when associated with the host cytoskeleton (26). More recent data showed that efficient translation of viral proteins requires microtubule-dependent transport of VSV mRNPs away from their sites of synthesis in the inclusion body-like clusters of nucleocapsids (13). Our data show that VSV protein synthesis was reduced by similar amounts in cells treated with cytochalasin D as in cells treated with nocodazole, suggesting that actin-dependent as well as microtubule-dependent mechanisms are involved in viral protein expression (Fig. 10).

Controlling for the effect of cytoskeletal inhibitors on viral protein expression was important for demonstrating the differential effects of inhibition of actin- versus microtubule-based mechanisms on incorporation of nucleocapsids into virions (Fig. 10). The difference between the effects on incorporation into virions of treatment with cytochalasin D versus nocodazole or colcemid contrasts with their similar effects on migration of nucleocapsids toward the edges of the cell (Fig. 6). As mentioned previously, this is likely due to the ability of nucleocapsids to follow shorter paths to sites of virus assembly that are not dependent on microtubules. In addition, it is possible that actin filaments may play roles in virus assembly in addition to movement from the cytoplasm to the plasma membrane. For example, actin filaments are often involved in the clustering of membrane microdomains (28), which appears to be a key process in the formation of VSV budding sites (29, 30).

Finally, if microtubule-dependent mobility of nucleocapsids is not critical for virus assembly, then what is the role in virus infection for movement of nucleocapsids along microtubules? One possibility is that the dependence on microtubules for virus assembly is cell type dependent. For example, previous experiments demonstrating effects of microtubule inhibitors on virus release were performed in BHK cells (4), whereas those presented here were in HeLa cells. Another possibility is that microtubules are not critical for how many virus particles are assembled but are important for where virions are assembled. This is likely to be the case in cell types where nucleocapsids must migrate long distances for efficient transmission of virus infection to neighboring cells, such as in polarized epithelial cells and neurons. It has been known for many years that VSV preferentially buds from the basolateral surface of polarized epithelial cells (31) and from dendrites as well as the cell body of neurons (32). This preferential budding is likely to be important for spread of virus in vivo. Budding from the basolateral surface of epithelial cells may promote the systemic spread of virus, and budding from dendrites may promote the neuronal transmission of virus in the nervous system. In both cell types, the preferential budding reflects the preferential transport of G protein (33–35). In the case of neurons, the localization of G protein in dendrites is dependent on microtubules (36). There may be a similar dependence on microtubules or actin filaments for distribution of nucleocapsids into dendrites. In contrast to neurons, microtubules do not appear to be essential for the polarized distribution of G protein and basolateral budding of VSV in epithelial cells (37–39). In one report, inhibition of either microtubule- or actin-dependent transport had relatively little effect on VSV budding in polarized epithelial cells (38), raising the possibility that these two transport systems may be redundant in these cells. The quantitative approaches for analyzing nucleocapsid distribution presented here may shed light on these questions in the future.