Measles Virus-Induced Disruption of the Glial-Fibrillary-Acidic Protein Cytoskeleton in an Astrocytoma Cell Line (U-251)

Abstract

A recombinant measles virus which expresses enhanced green fluorescent protein (MVeGFP) has been used to infect two astrocytoma cell lines (GCCM and U-251) to study the effect of virus infection on the cytoskeleton. Indirect immunocytochemistry was used to demonstrate the cellular localization of the cytoskeletal components. Enhanced green fluorescent protein autofluorescence was used to identify measles virus-infected cells. No alteration of the actin, tubulin, or vimentin components of the cytoskeleton was observed in either cell type, whereas a disruption of the glial-fibrillary-acidic protein filament (GFAP) network was noted in MVeGFP-infected U-251 cells. The relative amounts of GFAP present in infected and uninfected U-251 cells were quantified by image analysis of data sets obtained by confocal microscopy by using vimentin, another intermediate filament on which MVeGFP has no effect, as a control.

The advent of reverse genetics for negative-stranded RNA viruses provides new opportunities for the examination and reassessment of various aspects of the virus infection process. Measles virus (MV) is a Morbillivirus which belongs to the Paramyxoviridae. Like the other members of this family, MV has a single-stranded negative-sense RNA genome which is encapsidated by nucleoprotein (N). Six structural genes are encoded by the genome. The polymerase (L) and phosphoprotein (P) associate with the N protein to generate the helical ribonucleocapsid structure. Two glycoproteins, fusion (F) and hemagglutinin (H), are embedded in the pleomorphic virion envelope, and these mediate cell entry and fusion (9, 13, 39, 55). The matrix protein (M) associates with both the glycoproteins and the ribonucleocapsid structure and plays a key role in virion assembly (8, 41).

Many viruses have been shown to cause alterations to the cytoskeleton during in vitro infection (6, 10, 43, 47). For a review, see the work of Cudmore et al. (12). The Morbillivirus Canine distemper virus (CDV), has been reported to cause a total reorganization of the cytoskeleton, with the most notable alterations being in the microtubule and intermediate-filament networks (26). Vesicular stomatitis virus (VSV) infection, first, causes disassembly of the actin filaments and, second, alters the distribution of the microtubules and intermediate filaments (44, 47). Respiratory syncytial virus (RSV) also causes a disruption of the cytoskeleton (7, 21, 52). The effect of MV on the actin cytoskeleton is less clear. One group has reported a striking decrease in the overall number of actin bundles in human fibroblasts infected with MV. They also show a similar disruption upon infection with other Paramyxoviridae (16, 17). Contrary to this, a second group has not been able to demonstrate alterations to the actin cytoskeleton in MV-infected Vero cells (2).

Treatment of MV-infected cells with the actin-depolymerizing agent cytochalasin B (CB) results in the inhibition of virus maturation. This suggests that microfilaments play a role in the release of budding virions (2, 48, 51). Actin filaments have been shown to have a role in the movement of MV glycoproteins on the surfaces of infected cells (14). The involvement of actin filaments in the budding of MV has been examined by electron microscopy (4, 5). Again, a close association exists between actin filaments from the outer part of the cytoskeletal network and budding virus, with the filaments protruding into the particles. It has been suggested that budding is possibly the result of a vectorial growth of actin filaments (4). CB inhibits the production of infectious virus particles of other paramyxoviruses (7, 11, 24). Interestingly, CB has no effect on the maturation of VSV (23), which has been unequivocally shown to disrupt the actin cytoskeleton (44, 47). Recently the essential role of cellular actin in the gene expression and morphogenesis of RSV has been described. In this instance RSV infection causes a gross disruption of the actin cytoskeleton (7). Thus, there appears to be confusion in the literature. Additionally, it is not clear whether these alterations are active, i.e., induced to facilitate virus growth, or passive, i.e., simply caused as a result of infection but playing no formal role in virus replication.

A number of virus genomes, such as Simian virus 5, Mouse hepatitis virus, Human herpesvirus, and Simian varicella-zoster virus, have been engineered to express green fluorescent protein (GFP) (18, 19, 25b, 32). Recently the gene encoding enhanced GFP (EGFP) has been introduced into the MV genome, and a recombinant virus (MVeGFP) has been rescued (25a). We have demonstrated that EGFP is detectable in cells in the early stages of infection (13a). In all cases diffuse EGFP autofluorescence was detectable before viral antigen was detected by immunocytochemistry. Therefore, EGFP appears to be an ideal indicator of early MV cell infection and the recombinant virus appears to be very useful for in vitro studies and may also be beneficial for in vivo investigations. The diffuse nature of EGFP autofluorescence makes MVeGFP an ideal candidate for assessing the effects of virus replication on the cytoskeletons of MVeGFP-infected cells. As tubulin has been shown to stimulate MV RNA synthesis in vitro (36) and the fate of actin within MV-infected cells remains unclear (2, 16, 17), we decided to use MVeGFP to investigate the effects of MV infection on the cytoskeleton. Confocal scanning laser microscopy (CSLM) was used to gain maximal resolution in dually labeled specimens.

MVeGFP infects astrocytoma cells.

MV infection of the central nervous system (CNS) is a rare event (27, 30). Oligodendrocytes and neurons are the predominantly infected cell types, although infected astrocytes have also been described (1, 30, 31, 34). In this study we have used two astrocytoma cell lines, the first being GCCM cells, which have been used to study MV cell-to-cell spread (13a). This cell line was derived from an anaplastic astrocytoma (grade IV). The second cell line used was U-251 MG cells, which were established from a human glioma (3). These cells have been used to examine the induction of inflammatory cytokines upon MV infection (22, 45). Both cell lines were maintained in RPMI 1640 medium supplemented with 5% fetal calf serum. MVeGFP virus was rescued from a full-length antigenomic clone (25a) by using a cell line which expresses T7 RNA polymerase and the N and P proteins of MV (42). MVeGFP was propagated in African green monkey kidney cells (Vero). The gene encoding EGFP is present in an additional transcription unit (ATU) which is inserted before the N gene in the MV genome in the most promoter-proximal position. Due to this location, large amounts of EGFP are produced in infected cells. No major effects on the replication of the virus and the type of cell-pathogenic effect generated was observed (25a). Autofluorescence was readily observed during virus rescue and propagation by UV microscopy. Cells which showed none of the well-characterized signs of MV-induced cell-pathogenic effects were frequently observed. This demonstrates the strength of using MVeGFP for these experiments in that observations of the cytoskeletons of cells in the very early stages of virus infection can be made.

MVeGFP has no effect on the actin-, tubulin-, or vimentin-based cytoskeletons of astrocytoma cells.

GCCM and U-251 cells were grown to a confluence of 80% on glass coverslips. Cells were infected with MVeGFP at multiplicity of infection of 0.01 for 1 h at 37°C, after which time the inoculum was removed and maintenance medium, RPMI 1640 containing 2% fetal calf serum, was added. Infections were carried out for 50 h at 37°C. During this time the cells attained 95 to 100% confluence. Cells were permeabilized and fixed by using freshly prepared 4% paraformaldehyde. The cytoskeletons of the cells were visualized with a monoclonal antibody specific for either tubulin (Sigma) or vimentin (Dako). Antivimentin and antitubulin antibodies were diluted in phosphate-buffered saline (PBS) containing 0.5% Triton X-100 (1:100 and 1/1,500, respectively). The detergent was included to increase the permeability of the paraformaldehyde-fixed cells. Antibodies were incubated on the coverslips for 20 h at 4°C. Unbound antibodies were removed by three washes in PBS, each lasting 5 min. CY3-conjugated sheep anti-mouse immunoglobulin G (Sigma) was used as a secondary antibody. Dilutions (1:40) were made in PBS containing 0.5% Triton X-100, and the antibody was incubated on the coverslips for 3 h at 37°C. Unbound antibody was removed as described above. Tetramethyl rhodamine isothiocyanate (TRITC)-conjugated phalloidin (Sigma), a fluorescently conjugated phallotoxin from Amanita phalloides which specifically binds to F-actin, was used to directly stain the microfilaments. TRITC-conjugated phalloidin (200 ng/ml) in PBS was incubated on the coverslips for 2 h at 37°C. Excess phalloidin was removed by a single PBS wash. Coverslips were mounted with Citifluor (Amersham). A Leica TCS/NT confocal microscope equipped with a krypton-argon laser as the source for the ion beam was used to examine the samples for fluorescence. CY3-stained samples were imaged by excitation at 568 nm with a 564- to 596-band-pass emission filter. EGFP was visualized by virtue of its autofluorescence by excitation at 488 nm with a 506- to 538-band-pass emission filter. Data sets were collected by dual excitation, and image stacks were accumulated every 0.5 μm through an optical plane of 5 μm. Composite images were generated for the separate EGFP (green) and TRITC (red) channels in single-excitation mode to prevent spillover artifacts. Images were accumulated from regions of the monolayer which contained uninfected and infected cells and thereby permitted direct comparison of their cytoskeletal networks.

MVeGFP infection of GCCM and U-251 cells led to extensive fusion. Syncytia which are typical of MV-infected cells were observed. Nuclei clustered in the centers of the syncytia, and possibly due to a nonspecific accumulation of EGFP, these were brightly autofluorescent, as is shown for both cell types in Fig. 1. EGFP was present diffusely throughout the cytoplasm, and no overlap was observed between the green and red channels. This is particularly important as the most readily detectable MV antigens, N and P, produce a pronounced punctate staining pattern in extensively fused syncytia. Under these conditions, any overlap between channels by standard, dual-labeling indirect immunofluorescence may give the impression that alterations have occurred in the cytoskeleton. This effect is absent when MVeGFP infection is used in conjunction with CSLM, and thus this combined technology provides an excellent approach for examination of the effects of MV on the cytoskeleton. Using this approach, we investigated the effects of infection on microtubule, intermediate-filament, and microfilament components of the cytoskeletons of U-251 and GCCM cells.

FIG. 1.

Effect of MVeGFP infection on the cytoskeletons of GCCM and U-251 astrocytoma cells. Astrocytoma cells were infected with MVeGFP at a multiplicity of infection of 0.01 for 50 h. Cells were fixed and examined by CSLM for autofluorescence and immunoreactivity. Micrographs represent an 8- to 10-μm-deep composite optical section, and all images were obtained in double-excitation mode. EGFP was detected by virtue of its autofluorescence (green). (A) Actin microfilaments in GCCM and U-251 cells were visualized using TRITC-labeled phalloidin (red). The arrow indicates a single actin stress fiber in a GCCM cell. The arrowhead indicates a nonfibrillary aggregation of actin in a U-251 cell. (B) Tubulin was visualized using a monoclonal antibody and a CY3-conjugated secondary antibody (red). The arrow indicates a tubulin-rich astrocytic process from a GCCM cell. The arrowhead in the U-251 panel indicates tubulin accumulation around the nucleus of an unaffected cell. (C) Vimentin was visualized using a monoclonal antibody and a CY3-conjugated secondary antibody (red). The arrowhead indicates an astrocytic process from a U-251 cell. Magnification, ×400.

Actin.

No disruption of the actin-based cytoskeleton was observed upon MVeGFP infection of either cell type (Fig. 1A). Actin is present in two forms within cells, a globular, monomeric form (G-actin) and a polymerized, filamentous form (F-actin). It is the latter form which contributes to the cytoskeleton and is detected by phalloidin (29). The distributions of F-actin in GCCM and U-251 cells were similar to that observed in a previous study (20). In uninfected cells F-actin was present in long parallel stress fibers. These ran along the long axes of the cells. Cortical filaments outlined the peripheries of both cell types. In MVeGFP-infected cells which formed syncytia, the microfilaments were integrated into a larger, but organizationally similar, network. Extended fibers, which were greater in length than those of the single cells, spanned the syncytia (Fig. 1A, GCCM), indicating that actin polymerization does not seem to be inhibited by virus infection. Clumping of actin was noted in the U-251 cells (Fig. 1A). Equivalent amounts of F-actin appear to be present in both uninfected and infected cells. Actin bundles were more prevalent in MVeGFP-infected GCCM cells (Fig. 1A) than in U-251 cells. These observations confirm what was previously shown for nonrecombinant MV-infected Vero cells (2) and contrast with the results of two studies (16, 17) which observed severe actin disruption during MV infection of a human lung cell line. We have confirmed that MVeGFP infection of Vero cells causes no disruption of the actin cytoskeleton (data not shown). In a recent report, (24) colocalization of human parainfluenza virus type 3 (HPIV3) ribonucleoprotein (RNP) and actin microfilaments was observed in infected CV-1 cells by confocal microscopy with an HPIV3 polyclonal antiserum. The extent of colocalization was striking and demonstrates the usefulness of CSLM in this type of investigation. In MV-infected cells we have never observed a close relationship between MV RNP and the actin cytoskeleton using either polyclonal or monoclonal antibodies (anti-P or anti-N antibodies) for staining. This result is in spite of the fact that actin is known to associate with MV nucleocapsid (36). Rather, a punctate perinuclear staining pattern in which antigen is detected in MV-induced cytoplasmic inclusion bodies is observed. The HPIV3 system appears to be unique in that the actin microfilaments seem to be directly involved in viral-RNA synthesis in vivo (24). Interestingly, no disruption of the actin cytoskeleton was observed upon HPIV3 infection.

Tubulin.

This component of the cytoskeleton has been implicated as having a role in the MV life cycle, possibly as a subunit of the viral RNA polymerase (35). The distribution of the microtubules in MV-infected cells has not been examined previously by immunocytochemistry. The tubulin-based cytoskeletons are quite similar in organization in both the GCCM and U-251 cells. The microtubule bundles are thinner and more filamentous than F-actin stress fibers which cross syncytia. Generally, tubulin was present throughout the cell at similar levels, although the processes seemed to be particularly rich in microtubules (Fig. 1B, GCCM). There may be a slight accumulation of tubulin around the nuclei of the U-251 cells (Fig. 1B). Tubulin distribution was examined in MVeGFP-infected GCCM cells. No disruption of the cytoskeleton was observed in infected cells. Filaments were longer in the syncytia, indicating that the dynamic process of microtubule assembly is not perturbed in infected cells engulfed in syncytia. One investigation has reported a thickening of the microtubules in Hep-2 cells infected with the closely related Morbillivirus CDV (26). Bright foci and thick, long bundles crossing near or among the multiple nuclei were also observed in these syncytia. We have not been able to detect any such accumulation in MV-infected astrocytoma cells. Disruption of the microtubules has been suggested to have a role in the bipolar budding of Sendai virus (50). A mutant virus which buds in a bipolarized manner has alterations in the M protein, and it has been suggested that this protein may cause the alteration of microtubules. Involvement of the VSV M protein in microtubule disruption has also been suggested (47). In that study major changes in tubulin distribution were detected soon after VSV infection. As is the case for MV, a role has also been suggested for tubulin in VSV viral transcription (35).

Vimentin.

The intermediate filament, vimentin, is a major component of the cytoskeleton. The fate of the vimentin network in MV-infected cells has not been examined previously. In uninfected GCCM and U-251 cells the overall structure of the vimentin component of the cytoskeleton was similar in organization to the fine structure of filamentous tubulin. It appears, however, that the filaments are less well organized in parallel arrays than either the microtubules or the microfilaments. Astrocytic processes were particularly detectable in the U-251 cells by vimentin staining (Fig. 1C). We have previously shown that these processes mediate cell-to-cell spread of MV in vitro (13a). Once again no disassembly of this intermediate filament was observed in MVeGFP-infected cells (Fig. 1C). Extended filaments which were longer than those present in single cells were visible in syncytia, again indicating that assembly is not noticeably impaired within infected cells, as was also the case for actin and tubulin. A number of viruses have been shown to cause alterations in the intermediate-filament-based cytoskeletons (37, 46). RSV infection of Hep-2 cells leads to morphological changes of vimentin and an overall reduction in abundance, possibly due to proteolytic degradation (21). CDV infection of epithelial cells has been shown to lead to a disruption of the intermediate-filament network (26).

Disruption of the cytoskeleton therefore seems to depend on the virus studied. Here we clearly demonstrate that MV infection does not perturb the microfilament-, intermediate-filament-, or microtubule-based cytoskeletons of the two astrocytoma cell lines even when cell-pathogenic effect has progressed to form large, but intact, syncytia.

MVeGFP disrupts the GFAP-based cytoskeleton of astrocytoma cells.

Glial-fibrillary-acidic protein (GFAP) is an intermediate filament of the astrocytic cytoskeleton. This protein is found almost exclusively in astrocytes and is therefore commonly used as a marker for these cells (33). Its initial expression marks the differentiation of precursor cells into astrocytes, and its up-regulation accompanies the reactive response to CNS injury (15). Due to the contribution of GFAP to the astrocyte cytoskeleton, we examined the effects of MVeGFP infection on the organization of this protein. Immunocytochemistry was carried out as described above. GFAP was detected using a rabbit polyclonal antiserum (Dako) at a dilution of 1:100 in PBS containing 0.5% Triton X-100. CY3-conjugated sheep anti-rabbit immunoglobulin G (Sigma) was used as the secondary antibody. Immunocytochemical detection of GFAP in paraformaldehyde-fixed GCCM cells proved problematic (data not shown). This was disappointing, as we wished to be consistent and continue to examine the cytoskeleton in MVeGFP-infected cells using EGFP autofluorescence as an indicator of infection. For this, paraformaldehyde fixation was a prerequisite, as EGFP autofluorescence is rapidly lost when all organic fixatives are used. Detection of GFAP expression in the U-251 cells was more satisfactory. Nevertheless, it was important to use the cells at a low passage number due to the overall decrease in the levels of expression of GFAP as cells were cultured. Not all U-251 cells stained equally for GFAP, and it proved essential to carry out incubations in the presence of detergent to improve the detection of the fibrillary network. However, greater than 99% of cells were GFAP positive, permitting a satisfactory examination of the effects of MVeGFP infection on this intermediate filament in these cells. Uninfected U-251 cells stain brightly for GFAP, and MVeGFP was observed to efficiently infect GFAP-positive cells and form syncytia (Fig. 2A). At this low magnification large numbers of infected and uninfected cells are shown. In the uninfected cells GFAP was present as a fibrillary network which was similar in organization to that of vimentin. Astrocytic processes, which connect the cells, also stained positive for GFAP. A severe disruption of the GFAP cytoskeletal network was seen in MVeGFP-infected cells, and the overall amount of GFAP staining was reduced compared to that of the uninfected cells. Intensity profiles were plotted, using quantification software installed on the Leica TCS/NT confocal microscope, to assess the overall levels of fluorescence derived from the presence of GFAP and EGFP. A line was drawn across the composite data sets through a region of uninfected and infected cells. The analysis software was used to determine the total intensity of red and green in each pixel present along the length of this line, and the results were obtained as graphs (Fig. 2C and D). A direct correlation between infection (EGFP positive) and decrease in GFAP levels was observed (Fig. 2C). The line from which this profile was obtained is shown in red in Fig. 2A. This type of analysis was repeated for complete data sets collected from 10 distinct syncytia present in different areas of the infected monolayer. Similar profiles were obtained in all cases. Duplicate coverslips, infected at the same time with the same virus pool, were stained, as described above, for the intermediate filament vimentin and are shown at the same magnification for comparison (Fig. 2B). No disruption of the vimentin cytoskeleton had been observed previously (Fig. 1C). Therefore, this intermediate-filament protein served as the best control for this type of analysis. In this case no decrease in the intensity of the vimentin staining was observed in the infected area (Fig. 2D). Again, the line chosen from which this profile was produced is shown in red in Fig. 2B. Ten control data sets showed no diminution of the amount of vimentin in infected cells. An extreme example of GFAP disruption is shown at a higher magnification (Fig. 2E to G). In this case GFAP was barely detectable within the main body of the syncytium. Positive staining was observed within the syncytium, albeit at a very low level. This residual GFAP seemed to aggregate around the nucleus. Cells on the periphery of a syncytium can be assumed to be more recently infected than those in the center. GFAP staining in these infected cells was diminished compared to that in the uninfected cells, and the cytoskeletal network showed a certain degree of reorganization. This seems, therefore, to represent an intermediate stage in the disintegration of the GFAP cytoskeletal network. Association of residual GFAP with the nucleus was also observed. Examination of the GFAP fibrils demonstrated that they were much shorter than those in uninfected cells. Therefore, a reorganization of the GFAP-based cytoskeleton occurs in many of the infected astrocytoma cells upon MVeGFP infection. This has not been previously reported for MV.

FIG. 2.

Effect of MVeGFP infection on the GFAP and vimentin cytoskeleton. U-251 cells were infected, fixed, and examined by CSLM for autofluorescence and immunoreactivity, as described for Fig. 1. GFAP was visualized using a polyclonal antiserum and CY3-conjugated secondary antibody (red). Vimentin was visualized using a monoclonal antibody and a CY3-conjugated secondary antibody (red). EGFP autofluorescence (green) indicates infected cells. (A) Disruption of GFAP within MVeGFP-infected syncytia. Magnification, ×160. The red line crossing the image indicates the region selected for quantification of green and red fluorescence using the TCS/NT software. (B) Lack of disruption of the vimentin cytoskeleton. The line used for subsequent quantification is shown in red. Magnification, ×160. (C) Intensity profile obtained from panel A showing a correlation between the decrease in GFAP (red) and the increase in EGFP (green) autofluorescence. (D) Intensity profile obtained from panel B showing no alteration in vimentin (red) staining within areas of infection (green). (E to G) Severe disruption of GFAP cytoskeleton in U-251 cells. (E) Infection of cells shown by EGFP autofluorescence. (F) Vimentin staining. (G) Overlaid image. Magnification, ×400. Strongly positive GFAP cells were readily observed (arrow c); these were not present within the syncytia. In more recently infected cells at the peripheries of the syncytia (arrow b), the intermediate-filament network was partially disrupted and the overall amount of GFAP staining was diminished. A residual amount of GFAP occasionally remained associated with the nuclei of cells in the central areas of the syncytia (arrow a).

The closely related Morbillivirus CDV causes a demyelinating disease (reviewed by Summers and Appel [49]). CDV primarily infects astrocytes producing intracytoplasmic and intranuclear inclusion bodies (40). A small percentage (5%) of macrophages are also thought to be infected in the acute demyelinating lesions (38). Double labeling-immunohistochemistry has been used to detect viral antigen and GFAP in astrocytes present in, or derived from, brain tissues of CDV-infected animals. In vitro-infected primary cultures have also been examined (25, 54, 56). An overall decrease has been observed, both in vivo and in vitro, in the numbers of GFAP-positive cells, the prevalence of astrocytic processes, and GFAP staining in some cells (53, 57). Neither of these studies, however, links this diminution in GFAP staining with a reorganization of the cytoskeleton. Infection of primary fetal astrocytes with a lytic varicella-zoster virus causes a down-regulation or modification of GFAP expression (28). Morphological changes in the GFAP cytoskeleton which were very similar to those observed for MV infection were observed (Fig. 2), demonstrating that an alteration of the GFAP organization is not without precedent, albeit, in the case of varicella-zoster virus, in a lytic virus.

In this study we set out to examine the effect of MV infection on the cytoskeleton. MVeGFP was used to facilitate the examination of the cytoskeleton by CSLM because of its ability to express EGFP, which produces a diffuse cytoplasmic autofluorescence. We observed no disruption of the actin-, vimentin-, or tubulin-based cytoskeletal networks. A disruption of the GFAP cytoskeleton was observed. A direct correlation between MVeGFP infection and a decrease in GFAP amount was confirmed using quantitative confocal fluorescence microscopy. This is the first time that this effect has been noted for a fusogenic virus. Whether this disruption is a passive or active phenomenon remains unclear. It also remains to be seen if this decrease in GFAP mirrors the in vivo situation. This is an in vitro study which has examined effects in transformed cell lines because they expressed appreciable levels of GFAP. Nevertheless, this finding is important as the effect of MV on GFAP may give rise to an underestimation of the numbers of infected astrocytes in MV infection of the CNS.