Reovirus Outer Capsid Protein {micro}1 Induces Apoptosis and Associates with Lipid Droplets, Endoplasmic Reticulum, and Mitochondria

The mechanisms by which reoviruses induce apoptosis have notbeen fully elucidated. Earlier studies identified the mammalianreovirus S1 and M2 genes as determinants of apoptosis induction.However, no published results have demonstrated the capacitiesof the proteins encoded by these genes to induce apoptosis,either independently or in combination, in the absence of reovirusinfection. Here we report that the mammalian reovirus µ1protein, encoded by the M2 gene, was sufficient to induce apoptosisin transfected cells. We also found that µ1 localizedto lipid droplets, endoplasmic reticulum, and mitochondria inboth transfected cells and infected cells. Two small regionsencompassing amphipathic ${alpha}$ -helices within a carboxyl-terminalportion of µ1 were necessary for efficient induction ofapoptosis and association with lipid droplets, endoplasmic reticulum,and mitochondria in transfected cells. Induction of apoptosisby µ1 and its association with lipid droplets and intracellularmembranes in transfected cells were abrogated when µ1was coexpressed with ${sigma}$ 3, with which it is known to coassemble.We propose that µ1 plays a direct role in the inductionof apoptosis in infected cells and that this property may relateto the capacity of µ1 to associate with intracellular membranes.Moreover, during reovirus infection, association with ${sigma}$ 3 mayregulate apoptosis induction by µ1.

INTRODUCTION

Top

Introduction

The mammalian orthoreoviruses (reoviruses) are nonenveloped virusesthat carry 10 genomic segments of double-stranded RNA within twoconcentric capsid layers. Reovirus infection has become a well-studiedviral model of apoptosis in a variety of cultured cells as wellas in animals, where virus-induced apoptosis is a primary pathogeneticmechanism of tissue damage (reviewed in reference 17). Differencesamong reovirus strains in their capacities to induce apoptosishave allowed genetic determinants to be mapped to two viralgenes: S1, encoding both outer capsid protein ${sigma}$ 1 and nonstructuralprotein ${sigma}$ 1s, and M2, encoding outer capsid protein µ1 (51, 60, 61).

Much work has focused on the role of the S1-encoded ${sigma}$ 1 proteinin reovirus-induced apoptosis. ${sigma}$ 1 is a homotrimeric attachment protein(15, 56) that binds to junctional adhesion molecule A (JAM-A)and to ${alpha}$ -linked sialic acid residues on the surface of susceptiblecells (3, 4, 14, 46). Both of these binding events are requiredfor apoptosis induction by type 3 (T3) reovirus strains (4, 19),and these findings have led to the proposal that binding ofT3 reovirus ${sigma}$ 1 to cell surface JAM-A and ${alpha}$ -sialic acid transducessignals that are required for apoptosis induction in culturedcells (4). The role of sialic acid binding in animals may becomplex, however, in that it plays a role in apoptosis inductionin the central nervous system but not in myocardial tissuesof infected mice (32). Recently, the hypothesis that ${sigma}$ 1 bindingof JAM-A and sialic acid initiates signals required for apoptosisinduction has been seemingly disproven by findings that thesebinding activities of ${sigma}$ 1 are dispensable for reovirus-inducedapoptosis; specifically, virions with monoclonal antibodies(MAbs) bound to ${sigma}$ 1 can infect and induce apoptosis in CHO cellsthat express Fc receptor but not JAM-A or sialic acid; moreover,reassortant mapping shows that the M2-encoded µ1 proteinis the sole proapoptotic virus determinant in these cells (23).

The S1 gene is bicistronic and encodes nonstructural protein ${sigma}$ 1s in an overlapping reading frame relative to ${sigma}$ 1. ${sigma}$ 1s is responsiblefor reovirus-induced cell cycle arrest at G₂/M (47) but is notrequired for reovirus-induced apoptosis in cultured cells (52).Again, however, the situation in animals appears to be morecomplex, as ${sigma}$ 1s-null viruses are less virulent and induce lowerlevels of apoptosis in the heart and central nervous systemof infected mice (32).

At high multiplicity of infection (MOI) in cultured cells, reovirusescan induce apoptosis without the need for viral transcriptionor replication (61). Apoptosis induced in this way appears torequire both binding to JAM-A and ${alpha}$ -sialic acid followed by conversionof virions to intermediate subviral particles (ISVPs) by endo/lysosomalproteolysis (20). Little is known about the roles played byµ1 in reovirus-induced apoptosis. µ1 is a majorouter capsid protein (600 copies per virion) that undergoesproteolytic processing within endo/lysosomes following viral uptakefrom cell surfaces, and multiple lines of evidence suggest that oneor more of the µ1 cleavage products interact with cellular membranesto effect penetration of a partially disassembled, subviral particleinto the cytoplasm (7, 9, 10, 42, 44). Thus, a cleavage productof µ1 might provide the postbinding signal required for inductionof apoptosis during cell entry. Indeed, we have shown that a largefragment of virion-derived µ1 enters the cytosol and nucleusof infected cells early in infection (11).

Although high-multiplicity infection by reovirus in culturedcells can induce apoptosis without the need for viral transcriptionor replication, the latter processes appear to be required forefficient apoptosis induction in cells infected at lower multiplicity (61).In addition, as reovirus-induced apoptosis occurs relativelylate in the infectious cycle (17), it seems likely that replicativeevents are needed to amplify the proapoptotic signals that accompanycell entry by reovirus. Thus, de novo expression of the S1-encoded ${sigma}$ 1, S1-encoded ${sigma}$ 1s, and/or M2-encoded µ1 proteins mightplay a role in apoptosis induction during reovirus infection.

In this study, we tested the hypothesis that µ1 can induceapoptosis independently of ${sigma}$ 1 and ${sigma}$ 1s when expressed in culturedcells. We show that µ1 induced apoptosis and localizedto lipid droplets, endoplasmic reticulum (ER), and mitochondriain transfected cells and had similar distributions in infectedcells. We further show that regions encompassing two amphipathic ${alpha}$ -helices within a carboxyl (C)-terminal portion of µ1were key determinants of apoptosis induction and localizationto lipid droplets, ER, and mitochondria. Apoptosis inductionand localization to lipid droplets and intracellular membranesby µ1 in transfected cells were inhibited by coexpressionof its assembly partner ${sigma}$ 3, suggesting that in infected cellsthe levels of free µ1 may regulate apoptosis. Based onthese findings, we conclude that the µ1 protein and, infact, specific small regions of this protein play a major rolein apoptosis induction during reovirus infection.

MATERIALS AND METHODS

Top

Materials and Methods

Cells and viruses. CHO and CV-1 cells were grown in Ham's F-12 medium (CellGro)supplemented with 10% fetal bovine serum (HyClone), 100 U/mlpenicillin, 100 µg/ml streptomycin, 1 mM sodium pyruvate,and nonessential amino acids (CellGro). Reoviruses T1L and T3Dwere laboratory stocks of the isolates previously identifiedas T1/human/Ohio/Lang/1953 and T3/human/Ohio/Dearing/1955, respectively, (27).The superscript N in T3D^N differentiates a laboratory stockobtained from the Nibert laboratory from a T3D clone obtainedfrom L. W. Cashdollar (Medical College of Wisconsin), denotedT3D^C. The T3D^C clone differs from the T3D^N clone in viral factorymorphology and in the nucleotide sequence of its M1 genome segment(45). Viruses were plaque isolated and amplified in murine L929cells in Joklik's modified minimal essential medium (Gibco)supplemented with 2% fetal bovine serum, 2% bovine calf serum(HyClone), 2 mM glutamine, 100 U/ml penicillin, and 100 µg/ml streptomycin.

Antibodies and reagents. Mouse MAbs to µ1 (4A3) and ${sigma}$ 3 (5C3) and rabbit polyclonalantiserum to µNS and reovirus cores have been describedpreviously (8, 59, 62). MitoTracker Red CMXros (Molecular Probes),MAbs to human golgin-97 (Molecular Probes), adipose differentiation-relatedprotein (ADRP) (Progen), CD63 (LAMP1) (BD Pharmingen), calnexin(Affinity BioReagents), and protein disulfide isomerase (PDI)(Molecular Probes) were used as organelle markers for mitochondria,Golgi, lipid droplets, lysosomes, and ER, respectively. Rabbitpolyclonal antiserum to activated caspase-3 was obtained from CellSignaling Technologies. The FLAG epitope tag was detected with anti-FLAGM2 MAb (Stratagene or Sigma). Secondary antibodies for immunofluorescence(IF) microscopy were goat anti-mouse immunoglobulin G (IgG)and goat anti-rabbit IgG conjugated to Alexa 488 or Alexa 594 (MolecularProbes). All antibodies were titrated to optimize signal-to-noiseratios.

Plasmid construction. The reovirus M2 and S4 genes from T1L were subcloned from plasmidspBKS-M2L and pcDNAI-S4L (12) into the mammalian expression vectorpCI-neo (Promega) to produce pCI-M2(T1L) and pCI-S4(T1L), respectively.

To generate in-frame truncations of µ1, as well as in-framefusions with enhanced green fluorescent protein (EGFP), fragmentsof the M2 gene were PCR amplified using pCI-M2(T1L) as the template.The primers used for amplification are shown in Tables 1 and 2.PCR products were purified and cut with the appropriate restrictionenzymes (Tables 1 and 2) and then ligated into the indicatedvectors cut with the same enzymes. The pEGFP-N1 and pEGFP-C1vectors were purchased from Clontech.

View this table:
[in this window]
[in a new window]
TABLE 1. Primers used to prepare plasmids expressing M2 truncations

View this table:
[in this window]
[in a new window]
TABLE 2. Primers used to prepare plasmids expressing EGFP-M2 truncations

To generate the FLAG-M2(582-708) plasmid, the pEGFP-C-M2(582-708) plasmidwas digested with XhoI and ApaI and the cut fragment was ligatedinto the pCMV-3Tag-1B (Stratagene) vector cut with the same enzymes.

To generate pEGFP-C-M2(582-675 ${Delta}$ H2), we used overlap extensionPCR to splice M2 fragments 582 to 611 and 637 to 675 (31). WePCR amplified the M2 fragments (582 to 611) and (637 to 675)from the pEGFP-C-M2(582-675) template using primer pair i andii and primer pair iii and iv, respectively (Table 2). The twoPCR products were then mixed in equimolar amounts and furtherPCR amplified with primers i and iv (Table 2). The resultingPCR product was digested with XhoI and HindIII and ligated intopEGFP-C1 cut with the same enzymes. All recombinant plasmidswere recovered from transformed Escherichia coli, and the insertswere sequenced for correctness (Cornell University BioResource Center).

Infections and transfections. Cells were seeded the day before transfection or infection ata density of 1 x 10⁴ to 2 x 10⁴ per cm² in six-well plates containing18-mm-diameter round coverslips. Infections were begun by adsorbingvirus stocks to cells at an MOI of 5 to 100 PFU/cell, as indicated,for 1 h at room temperature in phosphate-buffered saline (PBS)(137 mM NaCl, 3 mM KCl, 8 mM Na₂HPO₄, 1 mM KH₂PO₄ [pH 7.5])containing 2 mM MgCl₂. Cells were then overlaid with growthmedium and incubated at 37°C for 6 to 48 h, as indicated.All cells were transfected using FuGENE 6 transfection reagent(Roche) according to the manufacturer's instructions. In brief, thefollowing ratios of FuGENE reagent to DNA (µl of FuGENE/µgof DNA) were used: CV-1, 3:2; CHO, 6:1; HeLa and L929, 3:1.The complexes were incubated for 30 min then added to cells andincubated at 37°C for 18 to 48 h, as indicated.

IF microscopy. Cells on coverslips were fixed for 10 min at room temperaturein 2% paraformaldehyde in PBS, washed three times in PBS, andthen permeabilized for 5 min in PBS containing 1% bovine serumalbumin (BSA) and 0.1% Triton X-100 (PBSA-T) or 0.5% saponin.All antibody incubations were carried out for 25 to 40 min at roomtemperature in PBSA-T or PBS containing 1% BSA and 0.05% saponin. Coverslipswere washed three times in PBS between primary and secondary antibodyincubations. Cell nuclei were labeled by incubation of coverslipswith 300 nM DAPI (4',6'-diamidino-2-phenylindole) (MolecularProbes). Coverslips were mounted on glass slides with Prolongreagent (Molecular Probes). Fluorescence and phase images wereobtained with a Nikon TE2000 inverted microscope equipped withfluorescence and phase optics through a 60x 1.4 NA oil objectivewith 1.5x optical zoom. Images were collected digitally witha Coolsnap HQ charge-coupled-device camera (Roper) and Openlabsoftware (Improvision) and then were prepared for publicationwith Photoshop and Illustrator software (Adobe Systems).

Mitochondria or the ER was labeled with MitoTracker Red CMXros(Molecular Probes) or an anti-protein disulfide isomerase (PDI)MAb (SelectFX Alexa Fluor 488 ER labeling kit; Molecular Probes),respectively, according to the manufacturer's instructions.

Lipid droplets were detected by incubating coverslips with Bodipy493/503 (20 µg/ml) for 20 min prior to mounting or withOil Red O as described previously (55).

SDS-PAGE and immunoblot analysis of steady-state levels of µ1 and ${phi}$ truncation constructs. CHO cells in 60-mm dishes were transfected in duplicate witheach of the constructs shown in Fig. 6 by using FuGene6 (Roche) accordingto the manufacturer's instructions. Immediately after transfection,50 µM z-VAD-fmk (Biomol) or an equivalent volume of dimethylsulfoxide (DMSO) carrier was added to the transfected cells.At 24 h posttransfection (p.t.), cells were scraped into ice-coldbuffer (150 mM NaCl, 50 mM Tris-HCl [pH 7.2]) and pelleted at500 x g for 5 min at 4°C. The cell pellets were then lysedon ice in the preceding buffer that also included 1% NonidetP-40, 1% desoxycholate, 0.1% sodium dodecyl sulfate (SDS), proteaseinhibitor cocktail (Roche), and 1 mM phenylmethylsulfonyl fluoride(Sigma). DNA was sheared by passing the cell lysate througha 27-gauge needle five times. Protein concentrations were determinedby the DC Protein Assay (Bio-Rad Laboratories), and 50 µgprotein from each lysate was suspended in 1x sample buffer,boiled for 10 min, and subjected to 10 or 15% SDS-polyacrylamidegel electrophoresis (PAGE), as indicated. Proteins were transferredfrom gels to nitrocellulose membranes, and the expressed proteinswere detected with polyclonal rabbit anti-T1L virion serum (todetect µ1) or MAb anti-GFP (Clontech) in 0.1 M Tris-bufferedsaline (pH 7.4) containing 0.05% Tween 20, 1% BSA, and 2% powderedmilk. Binding of antibodies was detected with goat anti-rabbit oranti-mouse IgG conjugated to horseradish peroxidase (HRP) and developedwith SuperSignal West Pico Chemiluminescent Substrate (Pierce)followed by fluorography. As a loading control, the blots were alsoprobed with MAb anti-ß-actin (Sigma) either concurrently withthe anti-T1L serum or after the immunoblot had been strippedof previous antibodies with Restore Western Blot Stripping Buffer(Pierce) for 15 min at room temperature.

View larger version (32K):
[in this window]
[in a new window]
FIG. 6. Apoptosis induction and subcellular localization by ${phi}$ truncation mutants in transfected cells examined by fluorescence microscopy. (A) CHO cells were transfected with the indicated constructs, fixed at 48 h p.t., and then immunostained with rabbit anti-activated caspase-3 polyclonal antibody followed by goat anti-rabbit IgG conjugated to Alexa 594. The FLAG-tagged construct was detected by immunostaining with an anti-FLAG MAb followed by goat anti-mouse IgG conjugated to Alexa 488. Cells were scored for caspase-3 activation as in Fig. 1. The means and standard deviations of three determinations are shown; Kruskal-Wallis (P) values of the group for CHO and CV-1 cells are 0.0014 and 0.01, respectively. (B) Immunoblots showing expression of µ1 and ${phi}$ truncation constructs in CHO cells at 24 h p.t. in the absence or presence of the broad-spectrum caspase inhibitor z-VAD-fmk. Cells were transfected with 1 µg of each construct. Immediately after transfection, cells were treated with DMSO or 50 µM z-VAD-fmk. At 24 h p.t., cell lysates were collected, and samples were subjected to 10% or 15% SDS-PAGE, followed by protein transfer to nitrocellulose. Expression of µ1 constructs was detected with polyclonal rabbit anti-virion serum followed by goat anti-mouse IgG conjugated to HRP. Expression of EGFP-fused constructs was detected with MAb anti-GFP (Clontech) followed by goat anti-mouse IgG conjugated to HRP. In all cases, ß-actin was used as a loading control and detected (after the blot was stripped and reprobed) with MAb anti-ß-actin followed by HRP-conjugated goat anti-mouse IgG. Positions of molecular weight markers are indicated. (C) CHO cells were transfected with pEGFP-N-M2(582-611) and then at 48 h p.t. either were stained with MitoTracker CMXros to detect mitochondria and then fixed (top row) or were fixed and then immunostained with anticalnexin followed by goat anti-mouse IgG conjugated to Alexa 594 to detect ER (bottom row). Nuclei were stained with DAPI. Right panels show colored merges of the different staining patterns, with labels in matching colors. Scale bars, 5 µm. (D) CHO cells were transfected with pEGFP-C-M2(582-675) or CV-1 cells were transfected with pEGFP-C-M2(582-675 ${Delta}$ H2), fixed at 48 h p.t., and then immunostained with anti-ADRP followed by goat anti-mouse IgG conjugated to Alexa 594. Right panels show colored merges of the different staining patterns, with labels in matching colors. Scale bars, 5 µm.

Statistical analyses. Comparisons of apoptosis levels induced by groups of three ormore different constructs or conditions were analyzed by theKruskal-Wallis rank sum test. Graphs were prepared with Kaleidagraph(Synergy Software).

RESULTS

Top

Results

Induction of apoptosis in transfected cells expressing reovirus µ1 protein. To test the hypothesis that µ1 can induce apoptosis incultured cells independently of ${sigma}$ 1 and ${sigma}$ 1s, we examined transfected cellsthat expressed this protein intracellularly. The µ1-encodingM2 gene derived from T1L reovirus was cloned into the expressionvector pCI-neo, under the control of a cytomegalovirus immediate-earlypromoter. Upon examining the M2-transfected CHO cells by phase-contrastand fluorescence microscopy, we saw that many of the cells expressingµ1 had morphological changes characteristic of apoptosis:the cells were rounded up, partially detached from the coverslip,and/or spiculated, and DAPI staining revealed that many nucleiwere small and had condensed, marginated chromatin (Fig. 1A).In addition, many of the cells expressing µ1 stained positive foractivated caspase-3 (Fig. 1B), an effector caspase whose activationdenotes commitment to apoptosis (48). At 48 h p.t., $~$ 30% and $~$ 20% of M2-transfected CHO and CV-1 cells, respectively, hadnuclear changes and/or activated caspase-3 (Fig. 1C; also seeFig. 5C). From these data, we conclude that expression of µ1induces apoptosis in a subset of transfected cells.

View larger version (11K):
[in this window]
[in a new window]
FIG. 1. Apoptosis induction by µ1 in transfected cells examined by fluorescence microscopy. CHO cells were transfected with pEGFP-C1 to express EGFP, as a control, or pCI-M2(T1L) to express µ1. Cells were fixed at 48 h p.t. and then immunostained with anti-µ1 (MAb 4A3) and rabbit anti-activated caspase-3 serum followed by goat anti-mouse IgG conjugated to Alexa 488 and goat anti-rabbit IgG conjugated to Alexa 594. Nuclei were stained with DAPI. (A and B) Marginated chromatin of an apoptotic nucleus (A) and activated caspase-3 (B) are present in a representative µ1-expressing cell. The inset in panel A shows a phase-contrast image. Scale bars, 5 µm. (C) Percentage of transfected cells showing apoptotic nuclear changes or activated caspase-3. At least 100 cells expressing either µ1 or EGFP were scored for the presence of apoptosis indicators. Means (± standard deviations) of three replicates are shown.

View larger version (44K):
[in this window]
[in a new window]
FIG. 5. Subcellular localization and apoptosis induction by µ1 truncation mutants in transfected cells examined by fluorescence microscopy. (A) CHO cells were transfected with pEGFP-C1 to express EGFP as control, pCI-M2(1-582) to express the equivalent of µ1 ${delta}$ , or pCI-M2(43-582) to express the equivalent of ${delta}$ . The cells were fixed at 48 h p.t. and then immunostained with anti-µ1 (MAb 4A3) followed—except for pEGFP-C1—by goat anti-mouse IgG conjugated to Alexa 488. The apparent concentration of µ1 ${delta}$ and ${delta}$ in the nuclear region is likely because the images are not confocal, and therefore the nuclear region is the thickest part of the cell that is imaged. Scale bars, 5 µm. (B) CHO cells (top rows) and CV-1 cells (bottom row) were transfected with pEGFP-C-M2(582-708) to express the equivalent of ${phi}$ tagged with EGFP. At 48 h p.t., the cells either were stained with MitoTracker CMXros to detect mitochondria (top row, middle panel) and then fixed or were fixed and then immunostained with either anti-ADRP to detect lipid droplets (middle row, middle panel) or anti-PDI to detect ER (bottom row, middle panel), followed in each of the last two cases by goat anti-mouse IgG conjugated to Alexa 594. Nuclei were stained with DAPI in each case. Left panels show EGFP-based fluorescence. Right panels show colored merges of the different staining patterns, with labels in matching colors. Arrowheads in the bottom panels indicate areas of colocalization between EGFP-M2(582-708) and ER. Scale bars, 5 µm. (C) CHO and CV-1 cells were transfected with the indicated constructs [including pCI-M2(1-708) to express full-length µ1] or infected with T1L (MOI = 100). The cells were fixed at 48 h p.t. or p.i. and then immunostained with anti-µ1 (MAb 4A3) and/or a rabbit anti-activated caspase-3 polyclonal antibody followed by goat anti-mouse IgG conjugated to Alexa 488 and goat anti-rabbit IgG conjugated to Alexa 594. Cells were scored for caspase-3 activation as in Fig. 1. The means and standard deviations of three determinations are shown; Kruskal-Wallis (P) values of the group for CHO and CV-1 cells are 0.0025 and 0.015, respectively.

Association of µ1 with annular rings and tubulovesicular structures in transfected cells. We next examined the distribution of µ1 in transfectedcells by immunostaining with µ1-specific MAbs. Despitethe effects in the nucleus described above, the distributionof µ1 in M2-transfected CHO cells (Fig. 1A and B) andCV-1 cells (Fig. 2A and C) was largely cytoplasmic. Moreover,although much of µ1 appeared diffuse in the cytoplasm,two distinctive patterns suggested that it might also be associatedwith cytoplasmic organelles. First, in many M2-transfected CV-1cells, µ1 localized to annular ring-like structures (Fig.2A) which surrounded phase-dense globules (Fig. 2B). These structuresmay correspond to the "discrete particles" seen at lower magnificationby Yue and Shatkin (65) following expression of µ1 intransfected HeLa cells. Second, though less frequently, µ1distributed in a tubulovesicular pattern in transfected CV-1cells (Fig. 2C). Similar patterns of µ1 distribution wereobserved in M2-transfected CHO, HeLa, and L929 cells (data notshown).

View larger version (26K):
[in this window]
[in a new window]
FIG. 2. Distributions of µ1 in transfected cells examined by fluorescence microscopy. CV-1 cells were transfected with pCI-M2(T1L) to express µ1, fixed at 24 h p.t., and then immunostained with anti-µ1 (MAb 4A3) followed by goat anti-mouse IgG conjugated to Alexa 488. Scale bars, 10 µm. (A) Annular ring distribution of µ1 (arrowheads; enlarged in inset). (B) Phase-dense globules in a phase-contrast image of the cells shown in panel A (arrowheads; enlarged in inset). (C) Tubulovesicular distribution of µ1 (arrowheads; enlarged in inset). This cell also contains µ1-staining annular rings, which are indistinct and brighter as a result of the longer exposure needed to image the tubulovesicular structures.

Association of µ1 with lipid droplets, ER, and mitochondria in transfected cells. To identify the cytoplasmic organelles with which µ1 mayassociate in transfected cells, we tested for µ1 colocalizationwith various cellular markers in M2-transfected CV-1 cells (Fig. 3)and HeLa cells (data not shown). We found no colocalizationbetween µ1 and antibody markers for the Golgi body orlysosomes (Fig. 3A and B). However, µ1 staining of annularring structures was often closely juxtaposed to ER, as shownby immunostaining for the integral ER membrane protein calnexin(Fig. 3C), and some µ1 staining appeared to colocalizedirectly with calnexin (Fig. 3C'). The tubulovesicular distributionof µ1 was faint in both CV-1 and HeLa cells but was oftenmore prominent in the HeLa cells. Although much of this tubulovesicularµ1 colocalized with the ER marker calnexin (Fig. 3C),some was also present on tubulovesicular structures that didnot appear to be ER and instead colocalized with MitoTracker,which stains mitochondria (Fig. 3D).

View larger version (56K):
[in this window]
[in a new window]
FIG.3. Subcellular localizations of µ1 in transfected cells examined by fluorescence microscopy. CV-1 cells transfected with pCI-M2(T1L) were fixed at 18 h p.t. and stained as described below for each row of panels. Right panels show colored merges of the different staining patterns, with labels in matching colors. Nuclei were stained with DAPI in each case. Scale bars, 10 µm. (A to C, C', and F) After fixation, cells were immunostained with the following MAbs as markers for organelles—anti-golgin-97 for Golgi complex (A), anti-LAMP1 for lysosomes (B), anticalnexin for ER (C and C'), and anti-ADRP for lipid droplets (F)—followed by goat anti-mouse IgG conjugated to Alexa 594. Cells were then fixed again and immunostained with anti-µ1 (MAb 4A3) conjugated to Cy2. The area outlined in panel C is enlarged in panel C' and shows colocalization between calnexin and µ1 (arrowheads). (D) Mitochondria were labeled with MitoTracker CMXros prior to fixation. After fixation, cells were immunostained with anti-µ1 (MAb 4A3) conjugated to Cy2. Arrowheads indicate areas of colocalization between µ1 and mitochondria. (E) After fixation, cells were immunostained with anti-µ1 (MAb 4A3) followed by goat anti-mouse IgG conjugated to Alexa 594. Neutral lipids were then stained with Bodipy 493/503.

View larger version (53K):
[in this window]
[in a new window]
FIG. 3— Continued.

The annular ring pattern was the most common and prominent distributionof µ1 observed in M2-transfected cells (present in $~$ 82%of transfected CV-1 cells). These rings surrounded lipid dropletsthat were labeled with Bodipy 493/503 (Fig. 3E) or Oil Red O(data not shown) dye, both of which stain for neutral fattyacids. Lipid droplets are storage organelles for cholesterolesters and triglycerides and are believed to be metabolic organellesinvolved in the synthesis and trafficking of cellular lipids(reviewed in reference 39). They are surrounded by a protein-encrustedmonolayer of phospholipid (57), and in most cells the majorprotein associated with this monolayer is ADRP (38). Strong colocalizationbetween µ1 and ADRP confirmed that µ1 was localizingto the periphery of lipid droplets (Fig. 3F).

The C-terminal, ${phi}$ region of µ1 determines both targeting to intracellular membranes and induction of apoptosis in transfected cells. Three major regions of the µ1 protein, divided by proteolyticcleavage sites, are commonly identified: the amino (N)-terminal,myristoylated fragment µ1N (residues 2 to 41); the centralfragment ${delta}$ (residues 42 to 582); and the C-terminal fragment ${phi}$ (residues 582 to 708) (40). In addition, pairwise combinationsof these fragments can yield two other species: µ1 ${delta}$ (residues2 to 582 plus the N-terminal, N-myristoyl group) and µ1C(residues 42 to 708) (40). To identify the region(s) of µ1responsible for its activities in the preceding Results sections,we constructed a series of M2 gene truncations that encode thenear-equivalents of each of these fragments (as some constructshave the addition of an initiating methionine residue at the Nterminus). In addition, we prepared a construct to encode full-length µ1lacking only residues 676 to 708 [µ1(1-675)], which aredisordered in the µ1: ${sigma}$ 3 crystal structure and geneticallyabsent from the µ1 homologs of avian reoviruses and aquareoviruses (1, 37, 43, 68).Because none of the available anti-µ1 MAbs (62) recognizethe µ1N or ${phi}$ fragment, we prepared constructs to express EGFP-taggedversions of those regions. The constructs and results are summarizedin Fig. 4A.

View larger version (30K):
[in this window]
[in a new window]
FIG. 4. Summary of µ1 constructs. (A) Amino acid residues of µ1 represented in each construct are indicated in the name. Full-length µ1 is indicated by a bar spanning residues 1 to 708, with colors corresponding to regions denoted in panel B, the structure of a µ1 monomer. Each truncation mutant is represented by a bar spanning the approximate portion of µ1. A black circle indicates EGFP fused to the N or C terminus of µ1 or ${phi}$ . An open triangle indicates three repeats of the FLAG epitope (3x FLAG) fused to the N terminus of ${phi}$ (582-708). The name of the proteolytic fragments represented by some of the constructs is shown in parentheses following the bar. Fine mapping of the ${phi}$ fragment of µ1 is represented similarly. A deletion mutant lacking the second (yellow) amphipathic ${alpha}$ -helix of ${phi}$ is denoted by ${Delta}$ H2. The capacity of each construct to induce apoptosis in CHO cells is indicated by +, –, or +/–. A summary of each mutant's subcellular localization is also indicated: diffuse (—), to mitochondria (MT), to lipid droplets (LD), and/or to ER. (B) Ribbon diagram of the µ1 monomer X-ray crystal structure. Green represents the µ1N (lighter) and ${delta}$ (darker) regions of µ1. The ${phi}$ region is colored red, yellow, and blue to represent subregions encompassing three consecutive amphipathic ${alpha}$ -helices and gray to represent the hydrophilic C-terminal tail. Amino acid residues at the boundaries of these subregions are numbered. The ribbon diagram was prepared with Pymol (Delano Software) from Protein Data Bank coordinates for the µ1: ${sigma}$ 3 heterohexamer, 1JMU (37).

When expressed in transfected CHO cells, the three truncations lackingthe ${phi}$ region (µ1 ${delta}$ , ${delta}$ , and µ1N-EGFP), as well as EGFPalone, were distributed diffusely through the cytosol and nucleus(Fig. 5A and data not shown). In contrast, the three truncations containingmost or all of the ${phi}$ region [µ1(1-675), µ1C, andEGFP/ ${phi}$ ] were targeted to mitochondria, lipid droplets, and ER(Fig. 5B and data not shown). The association of EGFP/ ${phi}$ withmitochondria was more prominent, and its association with lipiddroplets less so (Fig. 5B), than seen with full-length µ1(Fig. 3), µ1(1-675), or µ1C. EGFP/ ${phi}$ was associatedwith lipid droplets in $~$ 25% of transfected CHO cells comparedto $~$ 87% of CHO cells expressing full-length µ1. The reason(s)for these differences is not yet known. We also tested truncatedversions of µ1 for the capacity to induce apoptosis. At48 h p.t. in CHO and CV-1 cells, only the truncations containingmost or all of the ${phi}$ region [µ1(1-675), µ1C, andEGFP/ ${phi}$ ], and not EGFP alone, induced apoptosis in a substantialpercentage of cells, similarly to full-length µ1 (Fig. 5C).

Although the M2 gene we used for creating the µ1-expressingconstructs was derived from the T1L reovirus, T1L infectionof CHO cells at an MOI of 100 induced a low level of apoptosis( $~$ 10% of infected cells) (Fig. 5C). This finding is in agreementwith those of others showing that T1L is a poor inducer of apoptosis(51, 60). To address the possibility that virus strain differencesin the µ1 protein were responsible for different levelsof apoptosis induction, we examined the capacities of µ1derived from the T3D^N and T3D^C reoviruses (44) to induce apoptosis intransfected CHO cells. We found no differences in the capacitiesof µ1 derived from the T1L, T3D^N, or T3D^C strain to induceapoptosis; they induced apoptosis in $~$ 27%, $~$ 30%, and $~$ 29% of transfectedCHO cells, respectively, as assessed by nuclear changes. Thesefindings were somewhat surprising, as previous genetic studieshave shown that strain differences in the capacity to induceapoptosis are determined at least in part by the µ1-encodingM2 gene (23, 50, 59). One likely explanation is that in thecontext of viral infection, µ1-induced apoptosis is modulatedindirectly by how, or the extent to which, it interacts withother viral factors in a strain-dependent manner (see Discussion).

In summary, we conclude that residues 582 to 675 in the ${phi}$ regionof µ1 contain determinants for both inducing apoptosisand targeting to lipid droplets, ER, and mitochondria in transfectedcells. Moreover, the determinants in the ${phi}$ region appear to benecessary and sufficient for the same activities exhibited byfull-length µ1. Possible differences in the levels ofapoptosis induced by the different ${phi}$ -containing proteins areaddressed in the Discussion.

Two regions encompassing amphipathic ${alpha}$ -helices in the ${phi}$ region of µ1 are major determinants for inducing apoptosis in transfected cells. To identify specific determinants within the ${phi}$ region for inducingapoptosis in transfected CHO cells, we constructed a furtherseries of truncation or deletion mutants for expressing thisregion fused to the N or C terminus of EGFP. The constructsand results are summarized in Fig. 4A. Considering that the weakcapacity of EGFP to dimerize (67) might influence the proapoptoticactivity of ${phi}$ , we also prepared a FLAG epitope-tagged versionof the ${phi}$ region (Fig. 4A). FLAG/ ${phi}$ behaved like EGFP/ ${phi}$ with regardto both induction of apoptosis and localization to intracellularmembranes (Fig.and data not shown), suggesting that EGFP wasnot involved in these activities.

As for further truncations in the ${phi}$ region, we first createda construct to express a version of ${phi}$ lacking µ1 residues676 to 708 [construct pEGFP-C-M2(582-675)], i.e., lacking theC-terminal region that is disordered in the µ1: ${sigma}$ 3 crystalstructure and genetically absent from the µ1 homologsof related viruses, as indicated above. Expression of this proteininduced higher levels of apoptosis than those of full-length ${phi}$ [construct pEGFP-C-M2(582-708)] (Fig. 6A). Thus, this C-terminal regionof ${phi}$ is dispensable for apoptosis induction and may even serveto downregulate it (see Discussion).

The truncation junctions of additional ${phi}$ mutants were designedto fall between three amphipathic ${alpha}$ -helices observed in the µ1: ${sigma}$ 3crystal structure (37) (Fig. 4B). Extended truncations fromthe C terminus were designed to remove either a region encompassinghelix 3 [construct pEGFP-C-M2(582-643)] or a region encompassingboth helix 3 and helix 2 [construct pEGFP-C-M2(582-611)]. Expressionof each of these mutants led to moderate levels of apoptosis,albeit somewhat lower than those of full-length ${phi}$ (Fig. 6A).Thus, the 582-to-611 region encompassing helix 1 alone retainedmost of the activity at inducing apoptosis, and this activitywas similar when the 582-to-611 region was fused to either theN or C terminus of EGFP (Fig. 6A). The reduction in levels ofapoptosis induced by the 582-to-643 region relative to those bythe 582-to-675 region and full-length ${phi}$ (Fig. 6A) suggests thathelix 3 may represent another determinant of apoptosis induction.However, a truncation designed to express a protein containingonly helix 2 and helix 3 [construct pEGFP-C-M2(610-675)] showedlimited activity at inducing apoptosis (Fig. 6A), suggestingthat helix 3 may exhibit its proapoptotic activity predominantlyin concert with helix 1. We obtained further evidence that theregion encompassing helix 3 contributes to proapoptotic activity,whereas the region encompassing helix 2 does not, by replacingthe helix 2 region with a short linker in a protein otherwisecontaining the regions encompassing helix 1 and helix 3 [constructpEGFP-C-M2(582-675 ${Delta}$ H2)]. Expression of this mutant induced highlevels of apoptosis, similar to those induced by the 582-to-675region containing all three helices (Fig. 6A). These results appearto identify the regions encompassing helix 1 and helix 3 asthe minimal determinants for inducing maximal levels of apoptosis.The importance of the region encompassing helix 1 was furthersuggested by reduced levels of apoptosis induced by a mutantin which only the N-terminal three residues of this region weremissing from full-length ${phi}$ [construct pEGFP-C-M2(585-675)] (Fig. 6A)(see Discussion).

In summary, we conclude that regions encompassing the firstand third amphipathic ${alpha}$ -helices of ${phi}$ mediate its full proapoptoticactivity. Moreover, including evidence from the previous section,these two helical regions of ${phi}$ are probably responsible for thefull proapoptotic activity of full-length µ1. We alsonote further evidence that residues 675 to 708 at the C terminusof µ1 and ${phi}$ may serve to downregulate this activity.

Steady-state levels of µ1 constructs and ${phi}$ truncation mutants in transfected CHO cells differ in the presence or absence of the broad-spectrum caspase inhibitor z-VAD-fmk. Although we were able to detect expression of all µ1 constructsand ${phi}$ truncation mutants in transfected cells by fluorescencemicroscopy, we had limited success at detecting those constructsthat induced apoptosis by immunoblotting. We also observed thatconstructs that induced apoptosis often appeared to have lowerlevels of relative fluorescence compared to constructs thatdid not induce apoptosis. As general translation is inhibitedin cells undergoing apoptosis (28), we hypothesized that theproapoptotic constructs downregulated their own translation. Totest this hypothesis, we compared expression of the different constructsin transfected CHO cells incubated with either the broad-spectrumcaspase inhibitor z-VAD-fmk (50 µM) or DMSO control (Fig.6B). We found that in the presence of z-VAD-fmk, those constructsthat induced apoptosis had notably increased expression levelscompared to untreated controls (e.g., compare expression ofµ1, µ1C, and EGFP/ ${phi}$ ), whereas there was little changein the relative expression levels of those constructs that didnot induce apoptosis. We conclude that in the presence of thebroad-spectrum caspase inhibitor, most of the constructs hadsimilar steady-state levels of expression. In addition, we wereable to confirm that each construct was of the appropriate size.

The regions of ${phi}$ that induce apoptosis also determine targeting to intracellular membranes. As described above, the full-length ${phi}$ region fused to eitherEGFP or FLAG localized to lipid droplets, ER, and mitochondria(Fig. 5B and data not shown). The localizations of the varioustruncation and deletion mutants in the ${phi}$ region are summarizedin Fig. 4A. All mutants containing residues 582 to 611 localizedto ER and mitochondria (Fig. 6C and data not shown), identifyingthis region encompassing helix 1 as a minimal determinant ofthese activities. Only mutants containing both the helix 1 andhelix 3 regions, however, localized to lipid droplets (Fig. 6D).The truncation containing residues 610 to 675 did not associatewith intracellular membranes, suggesting that the helix 3 regionis not sufficient for membrane targeting and identifying thehelix 1 and helix 3 regions together as minimal determinantsfor targeting to lipid droplets. In general, all mutants thattargeted to ER and mitochondria also induced apoptosis. Thiscorrelation appeared weakest in the case of the 585-to-708 protein,which showed clear membrane targeting (data not shown) but inducedlower levels of apoptosis. Targeting to lipid droplets, in contrast,did not correlate strongly with apoptosis induction. In summary,we conclude that similar regions of ${phi}$ determine both membranetargeting and apoptosis induction and that localization to ERand/or mitochondria may be part of the mechanism by which thesedeterminants induce apoptosis.

Membrane association and apoptosis induction by µ1, but not ${phi}$ , are abrogated by coexpression of ${sigma}$ 3. Previous reports have noted that coexpression of µ1 leadsto a redistribution of ${sigma}$ 3 in cells (58, 64). We therefore hypothesizedthat coexpression of ${sigma}$ 3 would reciprocally lead to a redistributionof µ1. To test this hypothesis, we cotransfected µ1-and ${sigma}$ 3-expressing plasmids into CV-1 or CHO cells at differentrelative molar ratios. As a control, we cotransfected the µ1-expressingplasmid with one encoding the reovirus ${sigma}$ 2 protein. As the ratioof ${sigma}$ 3-to-µ1 plasmid increased, the distribution of µ1became more diffuse throughout the cytosol and less associatedwith lipid droplets, ER, or mitochondria (Fig. 7A). At a molarratio of 14:1 (S4:M2), essentially all of µ1 was diffusein the cytosol and the nucleus in the vast majority of transfectedcells (data not shown). In contrast, coexpression of ${sigma}$ 2 withµ1 had no substantive effect on the subcellular distributionof µ1, i.e., µ1 remained strongly associated withintracellular membranes (Fig. 7B). We also found that as the ${sigma}$ 3-to-µ1 ratio increased, apoptosis levels in CHO cellsdecreased, whereas ${sigma}$ 2 coexpression with µ1 had little orno effect on apoptosis levels (Fig. 7D, left panel). In other words,increasing ${sigma}$ 3, but not ${sigma}$ 2, had an increasingly antiapoptotic effect.The ${sigma}$ 3 and µ1 proteins are known to coassemble into solubleheterohexameric oligomers when coexpressed (37); thus, one possible explanationfor these results is that progressive sequestration of µ1into µ1: ${sigma}$ 3 heterohexamers decreases the amount of freeµ1 able to associate with intracellular membranes andto induce apoptosis.

View larger version (26K):
[in this window]
[in a new window]
FIG. 7. Effect of coexpressing ${sigma}$ 3 on the subcellular localization and ability to induce apoptosis of µ1 or EGFP/ ${phi}$ in transfected cells examined by fluorescence microscopy. (A) CV-1 cells were transfected with pCI-S4(T1L) to express ${sigma}$ 3 plus pCI-M2(1-708) to express µ1 at plasmid DNA ratios of 1:2 and 2:1 (S4:M2). Cells were fixed at 48 h p.t. and then immunostained with Cy2-conjugated anti-µ1 (4A3) and Alexa 594-conjugated anti- ${sigma}$ 3 (MAb 5C3). Representative examples of the predominant distribution patterns are shown. Scale bars, 5 µm. (B) CV-1 cells were transfected with pCI-S2(T1L) to express ${sigma}$ 2 plus pCI-M2(1-708) to express µ1 at a plasmid DNA ratio of 2:1 (S2:M2). Cells were fixed at 48 h p.t. and then immunostained with Cy2-conjugated anti-µ1 (4A3) and rabbit anti-core serum (to detect ${sigma}$ 2) followed by goat anti-rabbit IgG conjugated to Alexa 594. Scale bar, 5 µm. (C) CV-1 cells were transfected with pEGFP-C-M2(582-708) to express the equivalent of ${phi}$ tagged with EGFP plus either pCI-S4(T1L) or pCI-S2(T1L). Again in this experiment, each plasmid pair was transfected at a ratio of 1:2, respectively. Cells were fixed at 48 h p.t. and then immunostained for ${sigma}$ 3 and ${sigma}$ 2 as in panel A. Representative examples of the predominant distribution patterns are shown. Scale bars, 5 µm. (D) CHO cells were transfected with the indicated plasmid pairs at ratios of 2:1, 1:1, or 1:2 and then immunostained with anti-activated caspase-3 followed by goat anti-rabbit IgG conjugated to Alexa 594. For samples expressing full-length µ1, cells were also immunostained with anti-µ1 (MAb 4A3) followed by goat anti-mouse IgG conjugated to Alexa 488. Cells were scored for caspase-3 activation as in Fig. 1. The means and standard deviations of three determinations are shown. Also shown are the Kruskal-Wallis (P) values for the differences within each group.

Another possible explanation for the preceding results for µ1- ${sigma}$ 3coexpression is that the antiapoptotic effect of ${sigma}$ 3 is independentof its interaction with µ1. ${sigma}$ 3 is known to interact with double-strandedRNA and to prevent activation of protein kinase R, an effectthat could be antiapoptotic (24, 64). To address the possibilitythat ${sigma}$ 3 inhibited apoptosis independently of its capacity tointeract with µ1, we examined whether coexpression of ${sigma}$ 3 with ${phi}$ [construct pEGFP-C-M2(582-708)] would abrogate the capacityof ${phi}$ to associate with intracellular membranes and/or induceapoptosis. We reasoned that since ${phi}$ lacks the vast majority ofresidues in µ1 that interact with ${sigma}$ 3 (37), ${sigma}$ 3 and ${phi}$ shouldnot interact upon coexpression and therefore any other antiapoptoticeffect of ${sigma}$ 3 would be revealed. As a control, we coexpressed ${phi}$ with the reovirus ${sigma}$ 2 protein. Coexpression of ${sigma}$ 3 or ${sigma}$ 2 with ${phi}$ neitheraltered the intracellular distribution of ${phi}$ (Fig. 7C) nor reducedthe induction of apoptosis (Fig. 7D, right panel).

In summary, we conclude that coexpression of ${sigma}$ 3 with full-length µ1,as a function of relative levels of the two proteins, progressivelyabrogates µ1 association with intracellular membranesand induction of apoptosis, most likely because of µ1 sequestrationinto µ1: ${sigma}$ 3 heterohexamers. Since the ${phi}$ region lacks thevast majority of µ1: ${sigma}$ 3 contacts apparent in the structureof the µ1: ${sigma}$ 3 heterohexamer, we conclude that ${sigma}$ 3 cannot sequester ${phi}$ when those two proteins are coexpressed; thus, ${phi}$ retains its activitiesat membrane association and apoptosis induction.

µ1 localizes to lipid droplets, ER, and mitochondria in infected cells. The µ1 protein is known to localize to viral factoriesin reovirus-infected cells (5, 53) but has previously not beenlocalized to membranous organelles. Upon examining the distributionof µ1 in T1L-infected CV-1 cells by IF microscopy, wefound that at 24 h postinfection (p.i.), in addition to localizingto viral factories, µ1 localized to ring-like and tubulovesicularstructures in a subset of infected cells (see Fig. 9). We discernedfour patterns of µ1 distribution at 24 h p.i.: (i) diffuse throughthe cytosol (Fig. 8A), (ii) colocalized with µNS in viralfactories (Fig. 8B), (iii) localized to ring-like structures(Fig. 8C), and (iv) localized to tubulovesicular structures(Fig. 8D). Many cells displayed more than one of these patterns(e.g., in Fig. 8B, diffuse and localized to viral factories).As with our findings in transfected cells, µ1 colocalizedwith markers for lipid droplets (Fig. 8E), ER (Fig. 8F), andmitochondria (Fig. 8G) in infected cells. We found similar distributionsof µ1 in infected HeLa, CHO, and L929 cells (data notshown, but see Fig. 9). We conclude that µ1 localizesto lipid droplets, ER, and mitochondria in T1L-infected cellsin addition to viral factories; thus, its distribution partiallymirrors that seen in transfected cells.

View larger version (24K):
[in this window]
[in a new window]
FIG. 9. Distribution of µ1 and apoptosis induction in T1L-, T3D^N-, and T3D^C-infected L929 cells. (A) Reovirus T1L-, T3D^N-, or T3D^C-infected cells (MOI = 10) were fixed at the indicated times p.i., permeabilized with Triton X-100 (left panel) or methanol (right panel), and then immunostained with anti-µ1 (MAb 4A3) and anti-µNS serum followed by goat anti-mouse IgG conjugated to Alexa 488 and goat anti-rabbit IgG conjugated to Alexa 594. The patterns of µ1 distribution in individual infected cells were scored for each time point as diffuse only, associated with intracellular membranes (tubulovesicular and ring-like structures) (membranes), or associated with viral factories (VF). At least 200 cells were scored per replicate. Each data point represents the average of two replicates. (B) Cells were infected as above, fixed at 48 h p.i., permeabilized with Triton X-100, and then immunostained with anti-µ1 (MAb 4A3) and rabbit anti-activated caspase-3 polyclonal antibody followed by goat anti-mouse IgG conjugated to Alexa 488 and goat anti-rabbit IgG conjugated to Alexa 594. Cells were scored for caspase-3 activation as in Fig. 1. The means and standard deviations of three determinations are shown.

View larger version (72K):
[in this window]
[in a new window]
FIG. 8. Distribution patterns and subcellular localizations of µ1 in infected cells examined by fluorescence microscopy. CV-1 cells infected with T1L reovirus were fixed at 24 h p.i., and the distribution patterns of µ1 were detected by immunostaining with anti-µ1 (MAb 4A3) followed by goat anti-mouse IgG conjugated to Alexa 488. Four patterns of µ1 staining were detected as follows. (A) Diffuse. (B) Associated with viral factories (VF). Factories were detected by immunostaining with a rabbit polyclonal serum to µNS followed by goat anti-rabbit IgG conjugated to Alexa 594. In this merged image, yellow indicates colocalization between µ1 (green) and µNS (red). (C) Associated with annular ring-like structures (Rings). (D) Associated with tubulovesicular structures (TV). (E, F, and G) To ascertain the subcellular localization of µ1, fixed cells were first immunostained with anti-ADRP to detect lipid droplets (E) and anti-PDI to detect ER (F) followed by goat anti-mouse IgG conjugated to Alexa 594. Cells were then fixed again and immunostained with anti-µ1 (MAb 4A3) conjugated to Cy2. Alternatively, cells were first stained with MitoTracker CMXros to detect mitochondria (G), after which they were fixed and immunostained with anti-µ1 followed by goat anti-mouse IgG conjugated to Alexa 488. Nuclei were stained with DAPI. Arrowheads indicate areas of colocalization between µ1 and ER (F) or mitochondria (G). Scale bars, 5 µm.

Given the findings at 24 h p.i., we interpreted the differentµ1 distribution patterns as representing a continuum thatmay vary with time p.i. In preliminary experiments, we notedthat the distribution of µ1 was affected by the method usedto permeabilize cellular membranes: methanol reduced the staining ofµ1 with ring-like and tubulovesicular structures but increasedits staining within viral factories, whereas 0.1% Triton X-100had the opposite effects (data not shown). We therefore compared thedistribution of µ1 in T1L-, T3D^N-, and T3D^C-infected L929cells at 6, 12, 18, 24, and 48 h p.i. in cells permeabilizedwith either 0.1% Triton X-100 or 100% methanol (Fig. 9A). Withall three strains, we found that µ1 was predominantly diffusein the cytosol at 6 and 12 h p.i. regardless of the permeabilizationmethod. Thereafter, in methanol-permeabilized cells, µ1increasingly stained with viral factories and was seen associatingwith intracellular membranes (ring-like structures) only at thelatest time point (48 h). In contrast, in Triton X-100-permeabilizedcells, µ1 staining of intracellular membranes (tubulovesicularand ring-like structures) became visible at 12 to 18 h p.i.and increased to being seen in $~$ 30% of T1L-, $~$ 40% of T3D^N-, and $~$ 87% of T3D^C-infected cells by 48 h p.i. We moreover noted thatas µ1 became associated with viral factories and membrane structures,fewer cells had a diffuse distribution of µ1. From theseresults, we conclude that the pattern of µ1 distribution ininfected cells changes as infection progresses, being initially diffuseand then becoming localized to viral factories and associated withring-like and tubulovesicular structures from 12 to 18 h p.i.and beyond.

As we had found that coexpression of µ1 with ${sigma}$ 3 causedµ1 to redistribute from intracellular membranes to a predominantlydiffuse distribution in cells and abrogated µ1-inducedapoptosis, we speculated that the association of µ1 withintracellular membranes in infected cells was related to apoptosisinduction. In support of this hypothesis, we found that thatthe level of apoptosis induced by the T1L, T3D^N, and T3D^C strainsat 48 h p.i. in L929 cells (Fig. 9B) correlated with the percentageof infected cells in which µ1 associated with intracellularmembranes (Fig. 9A, left panels).

DISCUSSION

Top

Discussion

Consistent with genetic reassortant studies showing that thereovirus M2 gene is a determinant of virus strain differencesin the capacity to induce apoptosis during infection (51, 60, 61),we found that expression of the M2-encoded µ1 proteininduced apoptosis in uninfected, transfected cells. Both S1and M2 viral genes were previously identified as genetic determinantsof strain differences in reovirus-induced apoptosis (60); however,since these earlier studies, most attention has been paid tothe roles of the S1 gene products ${sigma}$ 1 and ${sigma}$ 1s, particularly toreceptor interactions by ${sigma}$ 1 with JAM-A and ${alpha}$ -sialic acid and theirpossible importance for proapoptotic signaling (reviewed in reference26). Recently, Danthi et al. showed that in CHO cells expressingthe Fc receptor, but not JAM-A or sialic acid, infection andreovirus-induced apoptosis can occur when virion-associated ${sigma}$ 1 is prebound with MAbs such that the Fc portion of the antibodymediates virus attachment (23). These authors also found thatunder such conditions the sole genetic determinant of apoptosisinduction was the M2 gene. The results presented here extend andsupport those conclusions, and we therefore propose that the µ1protein plays a more primary role in reovirus-induced apoptosisthan was previously appreciated.

If sufficient numbers (a high MOI) of reovirus particles areadded to cells, viral transcription or genome replication isnot required for induction of apoptosis (19, 20, 61). Nevertheless,an unidentified postattachment or disassembly step is required (20).In this study, we found that the ${phi}$ region of µ1 is necessaryand sufficient for inducing apoptosis in transfected cells.During infectious entry, proteolytic processing of virion-associated µ1within endo/lysosomes produces partially uncoated particles, ISVPs,which are primed for membrane penetration (reviewed in reference 10).The ${phi}$ fragment remains associated with ISVPs (41), and so itseems possible that particle-derived ${phi}$ could be released intothe cytosol after membrane penetration, where if present insufficient concentrations, it could induce apoptosis. Our previousfinding that the particle-derived ${delta}$ fragment of µ1 is presentin the cytosol and nucleus of the infected cell soon after penetrationis consistent with this hypothesis (11). At lower MOI, it is possiblethat fragments of µ1 could be released into the cytosolin smaller amounts that do not directly induce apoptosis but insteadprime the cells for apoptosis induction later in the infectious cycle.

In the current study, µ1 and all of its derived regionsthat induced apoptosis in transfected cells also associated withmitochondria and, to a lesser extent, with ER. It is temptingto speculate that association of these proteins with mitochondriaand/or ER is important for proapoptotic activity, as these organellesare intricately involved in the intrinsic apoptotic pathway (22, 50).However, one protein, EGFP/ ${phi}$ (585-708), associated with both mitochondriaand ER but induced only low levels of apoptosis, suggestingthat association with one or both of these organelles is not sufficientfor apoptosis induction, though perhaps it is still required.This construct reflects the ${phi}$ fragment that is generated by trypsindigestion of µ1 during generation of ISVPs (41). The relative differencein the proapoptotic abilities of the 582-to-708 versus the 585-to-708forms of ${phi}$ is intriguing, as it raises the possibility that levelsof apoptosis induction might be determined by cell-type-specificproteolytic processing of µ1, perhaps during cell entry.

Regions encompassing two amphipathic ${alpha}$ -helices (residues 582to 611 and 643 to 675) within the ${phi}$ region of µ1 were importantfor its proapoptotic activity. We also found that the disorderedregion in the crystal structure of µ1: ${sigma}$ 3 at the C terminusof µ1 (µ1 residues 676 to 708) was dispensable forthis activity and may even serve to downregulate it. The first30 residues of ${phi}$ (residues 582 to 611 of µ1), consistingof a short hydrophobic loop followed by an amphipathic helix,was the minimal identified region sufficient for induction ofapoptosis and was required for association with mitochondrialand ER membranes. Several other viral proteins associate withmitochondrial membranes and induce apoptosis. These includehuman immunodeficiency virus type 1 Vpr, influenza A virus PB1-F2,and the human T-cell leukemia virus type 1 p13^II accessory protein (16, 21, 33).All of these proteins have predicted amphipathic ${alpha}$ -helical regionsthat are required for interactions with mitochondrial membranes.However, the mechanism(s) of apoptosis induction is not completelyunderstood for any of these proteins. The Vpr and PB1-F2 proteinsare thought to promote apoptosis by directly interacting withcomponents of the mitochondrial permeability transition pore,thereby causing loss of the transmembrane potential with a resultantincrease in mitochondrial outer membrane permeability (33, 66).However, both of these proteins may induce and/or promote apoptosisin other ways. Vpr interacts with the antiapoptotic proteinHAX-1 on the outer mitochondrial membrane and may promote apoptosisby counteracting the HAX-1 antiapoptotic effect (63), and PB1-F2may directly permeabilize mitochondrial and/or other cellularmembranes by forming lipidic or proteolipidic pores (13). Previousauthors have shown that mitochondrial apoptotic pathways areactivated in reovirus-infected cells and have suggested thatthese pathways involve activation of caspase-8 and subsequentcleavage of the Bcl-2 family member Bid (35). However, it ispossible that µ1 or µ1 fragments directly activatemitochondrial apoptotic pathways, and we are currently investigatingthis possibility.

We found that steady-state levels of µ1 and its constructsthat induced apoptosis were much lower than those of constructsthat did not induce apoptosis. Moreover, levels of the proapoptoticconstructs were markedly increased by incubation of transfectedcells with the broad-spectrum caspase inhibitor z-VAD-fmk (Fig. 6B).As apoptosis is reported to inhibit translation generally (28),one explanation for this finding is that µ1 inductionof apoptosis resulted in inhibition of its own translation.Apoptosis induction appeared to be downregulated by C-terminalresidues 676 to 708 within either µ1 or ${phi}$ . Interestingly,this polypeptide sequence contains a predicted PEST motif (PESTfind [http://emb1.bcc.univie.ac.at/content/view/21/45/]). PESTmotifs are short regions of polypeptide sequence that are often foundat the C terminus of proteins; are enriched in proline (P), glutamicacid (E), serine (S), and threonine (T) residues; and are usuallyflanked by basic residues. PEST sequences are thought to actas signals for rapid protein degradation (49). If this weretrue of µ1, it would explain the enhancement of apoptosisseen in constructs lacking this region. Avian reoviruses (ARVs)induce higher levels of apoptosis in infected cells at 24 hp.i. than do mammalian reoviruses (36). We have found that expressionof the ARV-176 µB protein (the homolog of µ1) alsorobustly induces apoptosis in transfected cells (C. M. Coffeyand J. S. L. Parker, unpublished data). We note that ARV µBlacks the C-terminal extension (68), including the putativePEST motif found in µ1, and we are currently investigatingthe role of this region of µ1 in protein stability.

EGFP fusions of the ${phi}$ region tended to localize to mitochondrialmembranes, weakly to ER membranes, and occasionally to lipiddroplets in transfected cells (Fig. 8), contrasting with the primarylocalization of full-length µ1 to lipid droplets and lessso to mitochondria and ER in both transfected and infected cells. Althoughboth µ1 and ${phi}$ induced apoptosis in transfected cells, theirdifferential localization patterns suggest that they may differin their proapoptotic functions during reovirus infection. Caspase-3and caspase-8 activation in cells infected with reovirus T3 Abneyis biphasic, supporting the concept that two sequential proapoptoticsignals may be present in infected cells (34). Kominsky et al. suggestedthat this biphasic pattern of caspase activation results from aninitial tumor necrosis factor-related apoptosis-inducing ligand (TRAIL)-dependentactivation of caspase-8 that leads to low-level activation ofcaspase-3 and cleavage of Bid, followed by more sustained activationof caspase-3 and caspase-8 that results from cleaved Bid-mediatedactivation of the intrinsic apoptotic pathway and release ofcytochrome c and Smac/DIABLO from mitochondria (34, 35). Thismodel proposes that during reovirus infection, TRAIL-dependentapoptotic pathways are activated before mitochondrial apoptoticpathways. Alternatively, it is possible that mitochondria becomesensitized to TRAIL-dependent apoptosis before TRAIL secretion.The influenza A virus PB1-F2 protein, which has a localizationpattern similar to that of ${phi}$ , is believed to sensitize transfectedcells to tumor necrosis factor alpha-induced apoptosis by modulatingmitochondrial membrane permeability (66). We speculate that ${phi}$ and/or a related fragment of µ1 released into the cytosolof infected cells during membrane penetration sensitizes cellsto apoptosis induction by TRAIL, perhaps by modulating mitochondrialmembrane permeability.

As noted above, full-length µ1 localized primarily tolipid droplets in both transfected cells and infected cells.The hepatitis C virus (HCV) core protein is similarly localized(2, 6, 54). Moreover, expression of the HCV core protein causesapoptosis induction in some, but not all, transfected cells(6, 25, 29). The determinants of lipid droplet localizationof the HCV core protein have been mapped to amphipathic ${alpha}$ -heliceswhose primary sequence is homologous to plant oleosins, whichassociate with lipid droplets (30). In addition, a 10-residuesequence at the C terminus of the HCV core protein mediates itslocalization to mitochondria (54). Our findings suggest thatregions encompassing two amphipathic helices (residues 582 to611 and 644 to 675) near the C terminus of µ1 are required forassociation with lipid droplets but that only the first of these regionsis strictly required for mitochondrial localization (Fig. 6).The biological significance of associations by the HCV coreprotein and reovirus µ1 protein with lipid droplets remainsuncertain. However, lipid droplets have recently been implicatedas potential intracellular signaling platforms that might functionanalogously to lipid rafts on the plasma membrane (39). If thisis the case, then it is possible that association of µ1 withlipid droplets may be important for activation of certain signalingpathways. In support of this idea, it is known that the M2 geneis the genetic determinant of strain differences in JNK activation duringreovirus infection (18). Association of µ1 with lipiddroplets may modulate its capacity to induce apoptosis. In supportof this idea, we found that proteins EGFP/ ${phi}$ (582-675) and EGFP/ ${phi}$ (582-675 ${Delta}$ H2),which associated with mitochondria, ER, and lipid droplets,appeared to induce substantially higher levels of apoptosisthan did EGFP/ ${phi}$ (582-643) and EGFP/ ${phi}$ (582-611), which associatedonly with mitochondria and ER.

It has been previously reported that coexpression of µ1with the reovirus ${sigma}$ 3 protein modulates the distribution of ${sigma}$ 3in transfected cells (58, 65). We have confirmed these findingsand shown that coexpression of ${sigma}$ 3 with µ1 abrogates boththe membrane association of µ1 and its capacity to induceapoptosis. We hypothesize that this is a result of coassemblyof µ1: ${sigma}$ 3 heterohexamers. In our model, coexpression ofµ1 with ${sigma}$ 3 leads to sequestration of µ1 from membranesas assembled µ1: ${sigma}$ 3 heterohexamers. Our finding that coexpressionof ${sigma}$ 3 with the ${phi}$ domain of µ1 does not abrogate the capacityof ${phi}$ to induce apoptosis supports this hypothesis. Late in infection,µ1 is more often associated with intracellular membranesin cells infected with T3 viruses such as T3D than in those infectedwith T1 viruses such as T1L (Fig. 9A). This observation correlateswith the increased capacity of the T3 viruses to induce apoptosis(60; also Fig. 9B). Schmechel et al. have proposed that differencesin the affinity of ${sigma}$ 3 for µ1 regulate the subcellular distributionof ${sigma}$ 3, which in turn determines its capacity to bind double-strandedRNA and prevent activation of PKR (53). Similarly, we speculatethat strain-dependent differences in the affinity of µ1for ${sigma}$ 3 or in the kinetics of µ1: ${sigma}$ 3 heterohexamer assemblymay in turn determine the levels of "free" µ1 and thus µ1-determinedstrain differences in proapoptotic activity. The capacity of ${sigma}$ 3 to interact with µ1 and to abrogate its proapoptoticactivity in transfected cells is reminiscent of the controlof proapoptotic Bcl-2 family members such as Bax and Bak by hetero-oligomerizationwith antiapoptotic members such as Bcl-X_L and Bcl-2 (reviewedin reference 22). Relative levels of ${sigma}$ 3-bound versus free µ1may thus be an important determinant of phenotypes and straindifferences relating to apoptosis induction, as well as to inhibitionof host translation, by reovirus.

. 6A . . . .

ACKNOWLEDGMENTS

We are grateful to Roy Duncan for the generous gift of reagents,to Hollis Erb for advice on statistical analyses, and to MelinaAgosto, Tijana Ivanovic, and Leslie Schiff for helpful commentson the manuscript.

This work was supported by a Burroughs Welcome Fund Investigatorsof Pathogenesis of Infectious Disease award (to J.S.L.P.) andby NIH grants K08 AI052209 (to J.S.L.P.), R01 AI063036 (to J.S.L.P.),R01 AI46440 (to M.L.N.), and T32 AI07618 (to C.M.C.).

FOOTNOTES

* Corresponding author. Mailing address for John S. L. Parker: Baker Institute for Animal Health, College of Veterinary Medicine, Cornell University, Hungerford Hill Road, Ithaca, NY 14853. Phone: (607) 256-5626. Fax: (607) 256-5608. E-mail: jsp7{at}cornell.edu. Mailing address for Max L. Nibert: Department of Microbiology and Molecular Genetics, Harvard Medical School, 200 Longwood Ave., Boston, MA 02115. Phone: (617) 645-3680. Fax: (617) 738-7664. E-mail: mnibert{at}hms.harvard.edu.

Present address: Biological Engineering Division, MassachusettsInstitute of Technology, Cambridge, MA 02139.

Present address: Training Program in Virology, Division of MedicalSciences, Harvard University, Boston, MA 02115.

Present address: Department of Medicine, Brigham and Women'sHospital and Harvard Medical School, Boston, MA 02115.

REFERENCES

Top