Astrovirus Increases Epithelial Barrier Permeability Independently of Viral Replication

Abstract

Astrovirus infection in a variety of species results in an age-dependent diarrhea; however, the means by which astroviruses cause diarrhea remain unknown. Studies of astrovirus-infected humans and turkeys have demonstrated few histological changes and little inflammation during infection, suggesting that intestinal damage or an overzealous immune response is not the primary mediator of astrovirus diarrhea. An alternative contributor to diarrhea is increased intestinal barrier permeability. Here, we demonstrate that astrovirus increases barrier permeability in a Caco-2 cell culture model system following apical infection. Increased permeability correlated with disruption of the tight-junction protein occludin and decreased the number of actin stress fibers in the absence of cell death. Additionally, permeability was increased when monolayers were treated with UV-inactivated virus or purified recombinant human astrovirus serotype 1 capsid in the form of virus-like particles. Together, these results demonstrate that astrovirus-induced permeability occurs independently of viral replication and is modulated by the capsid protein, a property apparently unique to astroviruses. Based on these data, we propose that the capsid contributes to diarrhea in vivo.

Astroviruses are small enteric viruses belonging to the family Astroviridae (37). They possess single-stranded, positive-sense RNA genomes containing three open reading frames known as ORF1a, ORF1b, and ORF2. ORF1a encodes the viral nonstructural proteins, while ORF1b encodes the viral RNA-dependent RNA polymerase. ORF2 encodes the only capsid protein (33).

Astroviruses have been isolated from the young of a variety of species, including humans (32). In most animals, infection manifests primarily as a self-limiting diarrhea of 2 to 3 days in duration (17, 26). For humans, astroviruses are increasingly recognized as a leading cause of virally induced diarrhea in young children (9). Infection with human astrovirus serotype 1 (HAstV-1) is most commonly detected, although eight serotypes of HAstV (HAstV-1 to -8) have been isolated with various frequencies (18).

Despite its prevalence, little is known about the mechanism by which astrovirus causes diarrhea. Previous in vivo studies with animal models demonstrated that only mild histological changes occur during infection (24, 46). For humans, a recent study demonstrated that despite severe diarrhea, the morphological changes present in the biopsy of an astrovirus-infected child were relatively minor and nonspecific; in particular, the inflammatory response was only mild (42). The patient was immunocompromised, so the tissue reaction may have been modified; however, there was good evidence of engraftment and other acute and chronic inflammatory reactions. Furthermore, earlier studies with a turkey model demonstrated that astrovirus infection resulted in diarrhea in the absence of significant intestinal lesions, cell death, or inflammation (24). These results suggest that astrovirus does not employ two common mechanisms, destruction of the intestinal epithelium and stimulation of an inflammatory response, to cause diarrhea.

An alternative cause of diarrhea is an increase in intestinal barrier permeability, which could lead to increased fluid secretion into the intestinal lumen. The intestine is continuously subjected to an onslaught of potential pathogens. The principal physical defense against these pathogens is the formation of regulated epithelial cell-cell associations known as tight junctions (TJs). TJs are composed of transmembrane proteins, primarily occludin (15) and claudins (14), which form homotypic and heterotypic interactions with neighboring cells (25, 35). When properly formed, TJs form a dynamic barrier that is essentially impermeable to movement of fluids and solutes between the luminal and the serosal compartments. These transmembrane proteins interact with cytosolic adapter proteins such as zonula occludens-1 (ZO-1) (11, 16, 22). ZO-1 coordinates with various signaling proteins, such as symplekin, ZO-2, ZO-3, and cingulin (reviewed in reference 47), and links the cell membrane to the actin cytoskeleton (11). These interactions allow for strict regulation of TJs and intestinal barrier permeability. Disruption of TJs can result in the unregulated flux of pathogens and solutes into the body as well as the osmotic flux of water out of the body, which can manifest as diarrhea (29, 31).

To explore the effect of astrovirus on intestinal barrier permeability, we used an in vitro model of the intestinal barrier. When grown on semipermeable inserts, the intestinal carcinoma cell line Caco-2 spontaneously differentiates, forming a polarized monolayer reminiscent of the intestinal barrier (21). This model has been used in numerous studies to evaluate the effects of substances and pathogens on intestinal permeability (21), a potential indicator of diarrhea in vivo. Additionally, Caco-2 cells support human astrovirus replication (51). Using this system, we sought to evaluate the effect, if any, of astrovirus infection on barrier permeability. We found that astrovirus infection caused a polarized increase in barrier permeability. This increase in permeability temporally correlated with the movement of occludin away from the cellular periphery as well as a reduction in actin stress fibers in infected cells. These results could be reproduced with UV-inactivated virus, and purified recombinant capsid also increased barrier permeability. These data suggest that the astrovirus capsid protein mediates changes in permeability, likely triggered during early events in the virus life cycle.

MATERIALS AND METHODS

Cell and virus propagation.

The human intestinal adenocarcinoma cell line Caco-2 was obtained from the ATCC (HTB-37; Manassas, VA). Cells were propagated in Dulbecco's modified Eagle's medium (DMEM; Invitrogen, Carlsbad, CA) supplemented with 20% fetal bovine serum (FBS; Invitrogen), 1× nonessential amino acids (NEAA), 0.15% sodium bicarbonate, 2 mM l-glutamine, and 1 mM sodium pyruvate (Fisher Scientific, Norcross, GA). Medium was replaced on alternating days, and cells were passaged when they reached ∼80% confluence. For differentiation studies, Caco-2 cells were cultured for 5 to 7 days (on glass coverslips) or 15 to 20 days on 0.3-cm² semipermeable tissue culture inserts (1-μm pore size; BD Biosciences, Bedford, MA), with medium changes on alternating days. LLC-MK₂ cells, a macaque kidney cell line, were a kind gift of David Watkins (University of Wisconsin—Madison). LLC-MK₂ cells were propagated in minimal essential medium (MEM) supplemented with 10% FBS (Harlan, Indianapolis, IN), 1× NEAA, 0.15% sodium bicarbonate, 2 mM l-glutamine, and 1 mM sodium pyruvate. These cells were grown at 37°C in 5% CO₂. The spodoptera cell line Sf9 (CRL-1711; ATCC) was grown in Sf900 medium (Invitrogen) at 27°C with shaking. Cultures were maintained at between 5 × 10⁵ and 5 × 10⁶ cells/ml.

HAstV-1 to -8 (kind gifts of Stephen Monroe, Centers for Disease Control and Prevention, Atlanta, GA; Dorsey Bass, Stanford University; and Neel Krishna, Eastern Virginia Medical School) were propagated as described previously (51). Briefly, LLC-MK₂ cells were serum starved in MEM for 1 h at 37°C. Astrovirus (10⁶ infectious particles as determined by cell culture reverse transcription-PCR [RT-PCR], discussed below) was pretreated with 5 μg/ml type IV porcine trypsin (Sigma-Aldrich, St. Louis, MO) in MEM at 37°C for 15 min and added to LLC-MK₂ cells. The inoculum and cells were incubated for 90 min at 37°C, after which the inoculum was removed and the cells were washed with MEM. Maintenance medium containing 10 μg/ml porcine trypsin in MEM was incubated on monolayers for 3 days at 37°C. Monolayers were subjected to three rounds of freezing and thawing, and supernatants were collected by centrifugation (3,000 rpm, 10 min). Mock-infected lysates were obtained in an identical manner, with MEM substituted for astrovirus inocula. Viral supernatants were aliquoted and stored at −70°C until use.

Viral titration by cell culture RT-PCR.

Viral titers were determined by cell culture RT-PCR as described previously (5, 38). Briefly, viral stocks were serially diluted in MEM plus 10 μg/ml trypsin onto LLC-MK₂ cells. Infections proceeded for 4 days at 37°C, after which total RNA was isolated using TRIzol reagent (Invitrogen) according to the manufacturer's instructions. RNA was subjected to RT-PCR using primers Mon340 and Mon348 (3). The reciprocal of the last positive well corresponded to the titer of the virus in RT-PCR units. Results were confirmed by immunofluorescence as described below.

Virus inactivation.

To determine the UV dose sufficient for viral inactivation, virus (approximately 10⁶ U in 100 μl) was subjected to 50, 100, 150, or 200 mJ/cm² UV with a UV cross-linker (Fisher Scientific). UV-treated virus was added to differentiated Caco-2 cells on glass coverslips. Twenty-four hours after infection, monolayers were fixed with ice-cold absolute methanol and were stained for HAstV-1 antigen and DNA as described below. Complete viral inactivation was defined as a lack of capsid expression. A UV dose of 100 mJ/cm² was the lowest dose demonstrating complete viral inactivation (see Fig. 4). Inactivation was confirmed by serial passage of the UV-treated virus on Caco-2 cells, followed by staining for HAstV-1 capsid protein as described below.

FIG. 4.

HAstV-1 inactivation. Differentiated Caco-2 cells were mock infected (A) or were infected with infectious HAstV-1 (B) or UV-treated HAstV-1 at 50 mJ/cm² (C) or 100 mJ/cm² (D). At 24 hpi, monolayers were screened for HAstV-1 antigen by immunofluorescence. Inserts depict nuclei staining by DAPI. All images are at ×40 magnification.

Immunofluorescent microscopy.

Differentiated Caco-2 cells on glass coverslips were used to monitor HAstV-1 titers and inactivation. Caco-2 cells were rinsed once with phosphate-buffered saline (PBS) and infected with HAstV-1 (100-μl samples in a final volume of 250 μl MEM) or were mock infected for 1.5 h at 37°C. An equal volume of maintenance medium (4% bovine serum albumin [BSA] in MEM) was added to monolayers, and infection was allowed to continue for 24 h at 37°C. Supernatants were removed, monolayers were rinsed with PBS, and cells were fixed with ice-cold absolute methanol for 15 min at −20°C. Viral antigen then was detected by immunofluorescence as follows. Fixed monolayers were blocked with 5% normal goat serum (NGS; Fisher Scientific) for 15 min at room temperature. Supernatants from the astrovirus-specific hybridoma (20) cell line 8E7 (HB-11945; ATCC) were used undiluted to probe for HAstV-1 antigen (90 min at room temperature). Monolayers were washed and stained with secondary antibody (Alexa 594-labeled goat anti-mouse antibody; Invitrogen) at 1:200 and 4′,6′-diamidino-2-phenylindole (DAPI; 1:1,000; Invitrogen) in PBS plus 1% NGS. Coverslips were mounted with Prolong Gold (Invitrogen), and staining was visualized by epifluorescent microscopy using an Axiovert 100 TV microscope (Zeiss, Germany). Data were obtained and analyzed with OpenLab software (Improvision, Lexington, MA).

To evaluate the effect of astrovirus infection on cellular proteins, Caco-2 cells were plated on inserts as described below and were apically infected with HAstV (multiplicity of infection [MOI], 10) or an equal volume of mock lysate. At 24 h postinfection (hpi), monolayers were washed once with PBS and were fixed with 1% paraformaldehyde (for occludin and actin; 1 h at room temperature) or ice-cold absolute methanol (for claudin and ZO-1; 15 min at −20°C) and stained as follows. Briefly, paraformaldehyde-fixed monolayers were permeabilized with 0.5% NP-40 for 5 min, and all fixed cells were blocked with 5% NGS for 15 min (all steps were performed at room temperature). Monolayers were stained for occludin (OC-3F10; 1:100; Invitrogen), claudin (71-7800; 1:50; Invitrogen), ZO-1 (61-7300; 1:50; Invitrogen), or actin (Alexa 488-conjugated phalloidin; 1:40; Invitrogen) in PBS with 1% NGS or HAstV-1 (undiluted 8E7 hybridoma supernatants) for 90 min. Monolayers were washed and stained with secondary antibody (Alexa-594-labeled goat anti-mouse antibody; Invitrogen) at 1:200 in PBS with 1% NGS (48). Coverslips were mounted with Prolong Gold (Invitrogen), and proteins were visualized using a Zeiss LSM510 confocal microscope. Data were obtained and analyzed using LSM 5 Image software (Zeiss, Germany).

Measurement of transepithelial resistance of Caco-2 monolayers.

Differentiated Caco-2 monolayers restrict ion transport in vitro, developing an electrical gradient that can be measured as transepithelial electrical resistance (TER). The degree of barrier permeability can therefore be assessed by monitoring TER (8). Caco-2 cells (2 × 10⁵) were plated on 0.3-cm² semipermeable tissue culture inserts (1-μm pore size; BD Biosciences) and allowed to differentiate for 15 to 20 days, with medium changes on alternating days, until they achieved TER levels of at least 1,000 Ω*cm². Monolayers were rinsed with PBS and either apically or basally infected with HAstV (10⁵ U unless otherwise noted) or an equal volume of mock-infected lysate in MEM. In antibody neutralization assays, inocula were incubated alone, with an irrelevant immunoglobulin G antibody (1:20), or with polyclonal rabbit anti-HAstV-1 (1:20) (2) (kind gift of Dorsey Bass, Stanford University) for 1 h at 37°C prior to addition to monolayers. TER levels were measured in the presence of the inocula every 4 to 12 h throughout the course of the experiment using an EndOhm-6 chamber and an EVOM voltometer (World Precision Instruments, Sarasota, FL). Results are presented as percentages of the insert's initial (time zero) TER reading (34). Data are representative of at least seven experiments.

Fluorescent flux assay.

Differentiated Caco-2 cells form tightly regulated cell-cell junctions that restrict the passage of fluids and solutes between luminal and serosal compartments (21). The permeability of the epithelial monolayer during HAstV-1 infection was assessed by measuring the diffusion of fluorescein isothiocyanate (FITC)-labeled dextran (FITC-dextran; 4,000 Da; Sigma-Aldrich) across the monolayer. FITC-dextran was prepared at 20 mg/ml; 1 μl was added with the experimental sample to the apical surface of polarized Caco-2 cells grown on semipermeable inserts. Every 12 hpi, 50-μl aliquots were removed from the basolateral chamber and were replaced with an equal amount of MEM for a period of up to 48 h. Fluorescence was measured using a SpectraMax Gemini EM spectrofluorometer (Molecular Devices, Sunnyvale, CA) at an excitation wavelength of 495 nm and an emission wavelength of 518 nm (34). Fluorescence was compared to that of the maximum migration of the probe across a cell-free insert; results are expressed as percentages of the maximum migration. Data are representative of at least four experiments.

HAstV-1 capsid purification.

The HAstV-1 capsid was cloned into the baculovirus transfer vector 1392, and recombinant baculovirus was produced as described previously (50). Briefly, Sf9 cells were infected at an MOI of 5 in Sf900 medium, and infection was allowed to continue for 4 days, which is the time of maximal protein expression as determined by immunofluorescent microscopy and Western blot analysis (data not shown). Infected Sf9 cells were pelleted (2,500 rpm, 5 min) and resuspended in 10% of the original volume. Cells were lysed by three rounds of freezing and thawing, and cellular debris was removed by centrifugation (2,500 rpm, 5 min). Capsid was purified by passing the resulting supernatant over a Sepharose size-exclusion column (CL-6B; Sigma) as described for turkey astrovirus type 2 (23). Purified capsid fractions corresponding to whole-virus fractions were verified by Western blot analysis (see Fig. 6A). Positive fractions were pooled and concentrated by lyophilization (2 h at 43°C in a Savant speedvac ISS 100 [GMI, Ramsey, MN]) to a final volume of approximately 300 μl. The concentrated fractions were dialyzed overnight against PBS at 4°C, and protein concentrations were determined by the Bradford colorimetric assay (Bio-Rad, Hercules, CA). Formation of virus-like particles (VLPs) by purified capsid protein was confirmed by transmission electron microscopy (4).

FIG. 6.

Purification of HAstV-1 VLPs. (A) HAstV-1 ORF2 was expressed in Sf9 cells by baculovirus, and capsid protein was purified by size-exclusion chromatography and examined by Western blot analysis. (B) VLPs then were verified by electron microscopy. Bar, 100 nm.

To determine if HAstV-1 capsid was responsible for increased permeability, assays for TER were performed as described above. Thirty-five micrograms of purified capsid or an equal volume of mock-infected fraction was incubated in the presence of porcine trypsin (10 μg/ml) for 15 min at 37°C. Trypsin activity was neutralized with 2% BSA (Sigma), an equal volume of MEM (50 μl) containing 2× penicillin-streptomycin (Fisher Scientific) was added, and the samples were added apically to inserts.

Cell viability.

Caco-2 cells were plated at 2 × 10⁵ on semipermeable tissue culture inserts and allowed to differentiate for 15 to 20 days, with medium changes on alternating days. Cells were rinsed with PBS and either were mock infected or were infected with HAstV-1 as described above. At various times postinfection, viability was determined by trypan blue exclusion.

Statistics.

All statistical analyses were conducted using the Student's t test in Microsoft Excel. Error bars represent standard deviations, and statistical significance was defined as P < 0.05.

RESULTS

Apical HAstV-1 infection increases barrier permeability.

To determine if HAstV-1 had any effect on the permeability of polarized epithelial cells, Caco-2 cells grown on semipermeable inserts were mock or HAstV-1 infected apically (Fig. 1A) or basally (Fig. 1B), and TER levels were measured every 4 to 12 h through 48 hpi. Apical HAstV-1 infection resulted in a significant decrease in TER (P < 0.001) by 24 hpi compared to the TER of mock-infected monolayers. By 32 hpi, TER levels were approximately 0% of their original levels (Fig. 1A). In contrast, basal infection led to no significant decrease in TER (P > 0.05) throughout the time course of the experiment (Fig. 1B).

FIG. 1.

HAstV-1 increases barrier permeability in apical, but not basal, infection. Differentiated Caco-2 monolayers grown on permeable supports were mock infected (diamonds) or were infected with HAstV-1 (MOI 10) (squares) from the apical (A) or the basal (B) surface. TER levels were measured from 0 to 48 hpi. (C and D) The degree of paracellular permeability was measured from 4 to 48 hpi by monitoring the migration of FD-4 to the basal chamber over mock-infected (gray) or HAstV-infected (black) Caco-2 monolayers infected either apically (C) or basally (D). Error bars represent standard deviations, and asterisks represent statistical significance (P < 0.05) compared to values for mock infection.

The effect of HAstV-1 infection on epithelial permeability was further examined by monitoring the ability of a fluorescently labeled, inert dye to migrate from the apical chamber to the basal chamber. Caco-2 cells grown on semipermeable inserts were mock or HAstV-1 infected from the apical or basal surface. At the time of infection, FITC-dextran (4 kDa) was added to the apical chamber. Flux was determined by monitoring fluorescence in the basal chamber at various times postinfection. By 36 hpi, a significant (P = 0.0061) increase in fluorescent flux was observed across apically infected monolayers (Fig. 1C). Consistent with the results of the TER studies, no significant (P > 0.05) increase in fluorescent flux was detected in basally infected monolayers at any time point (Fig. 1D). Together, these data suggest that HAstV-1 increases epithelial barrier permeability in a polarized, apical-infection-specific fashion. Additionally, all eight serotypes of HAstV increased barrier permeability, suggesting that this phenomenon is not specific to HAstV-1 (data not shown).

To investigate if the increase in permeability caused by HAstV-1 was dose dependent, differentiated Caco-2 monolayers were infected with increasing concentrations of HAstV-1 (MOI of 1, 5, or 10). Figure 2A shows a dose-dependent increase in permeability, with a significant drop in TER in monolayers infected at an MOI of 10 compared to the TER of mock-infected cells by 24 hpi (P = 0.0046). A similar drop in monolayers infected with 10-fold less virus does not occur until approximately 8 h later, at 32 hpi (P = 0.0169). Additionally, cells infected with the highest concentration of HAstV-1 obtained a TER level of zero around 32 hpi, while inserts infected at an MOI of 1 did not reach their lowest TER levels until approximately 48 hpi. Monolayers infected with the lowest concentration of virus did not reach a TER level of zero (Fig. 2A), and levels began to rebound by 52 hpi (data not shown), which was not observed at higher doses.

FIG. 2.

HAstV-1's effects are dose dependent and require virus-cell interactions. (A) Differentiated Caco-2 cells were mock infected (triangles) or were infected with HAstV-1 virus at an MOI of ∼10 (light gray squares), 5 (dark gray squares), or 1 (black squares), and TER levels were monitored from 0 to 48 hpi. Error bars represent standard errors of the means, and asterisks represent statistical significance (P < 0.05) compared to values for mock infection. (B) Differentiated Caco-2 cells were treated with mock-infected cell lysate (dark triangles), mock-infected lysate preincubated with anti-HAstV-1 antibody (open triangles), HAstV-1 alone (MOI 10) (dark squares), HAstV-1 preincubated with irrelevant antibody (open diamonds), or HAstV-1 preincubated with anti-HAstV-1 antibody (open squares), and TER levels were monitored from 0 to 48 hpi. Data are representative of at least two experiments.

We next asked whether increased permeability was due to the virus itself or to a soluble factor present in the viral inoculum. To examine this, lysates from mock-infected or HAstV-1-infected cells were incubated in the presence or absence of an anti-HAstV-1 polyclonal antibody (previously shown to neutralize the virus [2]) or an irrelevant isotype control antibody and were added to monolayers. TER and flux then were measured (Fig. 2B and data not shown). Monolayers treated with mock-infected cell lysate in the presence or absence of antibody remained at TER levels of approximately 100%, while cells treated with HAstV-1 lysate alone or with the isotype control antibody demonstrated the typical drop in TER. When HAstV-1 was preincubated with an HastV-specific antibody, however, the drop in the TER level was delayed until 36 hpi and then fell to only 60% of initial values before rebounding. These data suggest that virus-cell interactions, not a soluble factor, are responsible for the wild-type increases in permeability.

Increased permeability is not due to increased cell death.

A possible cause of the measurable increase in permeability is cell death. To determine if HAstV-1 caused an increase in cell death, Caco-2 cells were mock infected or were infected with HAstV-1, and cell viability was measured by trypan blue exclusion. HAstV-1 did not significantly increase cell death compared to the levels of death of mock-infected cells by 24 hpi (P > 0.05), when barrier permeability begins to increase, or at 36 hpi, when there was a significant drop in TER (Fig. 3). These results imply that cell death is not the primary cause of HAstV-1-increased permeability.

FIG. 3.

HAstV-1 does not increase cell death. Cell viability was measured for mock- or HAstV-1-infected (MOI, 10) differentiated Caco-2 cells at 4, 12, 24, and 36 hpi by trypan blue exclusion. No significant difference between mock-infected cells and HAstV-1-infected cells was noted at any time point. Error bars represent standard deviations.

Increased permeability is independent of viral replication.

To determine if productive viral replication was required for increased permeability, polarized Caco-2 monolayers were apically treated with infectious HAstV-1 or an equivalent amount of UV-inactivated virus (Fig. 4). Monolayers treated with UV-inactivated HAstV-1 demonstrated a decrease in TER, with kinetics nearly identical to those of infectious virus (Fig. 5A). UV-inactivated virus no longer replicated, suggesting that increased permeability may be due to viral binding or entry.

FIG. 5.

HAstV-1-induced barrier permeability is independent of viral replication. (A) Differentiated Caco-2 cells were mock infected (diamonds), infected with HAstV-1 (MOI, 10) (squares), or infected with UV-inactivated HAstV-1 (circles). TER levels were measured from 0 to 24 hpi. Error bars represent standard errors of the means, and asterisks indicate P < 0.05 compared to values for mock infection. (B) At the indicated times postinfection, TER levels of mock-treated (n ≥ 3) (closed diamonds), HAstV-1-infected (n ≥ 3) (closed squares), PBS-treated (n = 3) (open diamonds), or purified HAstV-1 VLP-treated (35 μg; n ≥ 3) (open squares) monolayers were determined. Results are presented as percentages of the initial TER readings. Error bars represent standard deviations, and asterisks indicate significance between mock- and HAstV-1-infected samples; #, significance between PBS- and capsid-treated samples (P < 0.05).

To confirm that functional viral replication was not required for increased permeability, the HAstV-1 capsid protein was expressed in Sf9 cells by using a baculovirus expression system, which has been shown to produce VLPs (4). The capsid protein was purified by size-exclusion chromatography, and positive fractions corresponding to full-length capsid were identified by Western blot analysis, pooled, and concentrated (Fig. 6A). We confirmed formation of VLPs by electron microscopy (Fig. 6B). VLPs were further treated with trypsin, since HAstV-1 requires trypsin cleavage for optimal infection (1, 39); the cleaved virus is up to 10⁴ times more infectious than the uncleaved product (1). Differentiated Caco-2 monolayers then were treated with PBS pretreated with trypsin, HAstV-1 (MOI 10), or trypsin-treated VLPs (35 μg) and were monitored for increased permeability by TER. Monolayers treated with either infectious HAstV-1 or VLPs demonstrated decreased TER levels by 20 hpi, while monolayers treated with PBS in the presence of trypsin remained at TER levels of approximately 100% throughout the experiment (Fig. 5B). Combined with the results of the UV studies described above, these results suggest that the capsid protein is responsible for increased barrier permeability. Similar to infectious virus, UV-inactivated virus and purified viral capsid had no effect on cell death (data not shown), suggesting that the increase in permeability is not due to increased cell death.

HAstV-1 alters F-actin and occludin distribution.

Increased cell permeability often results from disruption of the actin cytoskeleton or TJ proteins (reviewed in reference 12); we therefore examined the effect of HAstV-1 infection on these structures. Differentiated Caco-2 cells grown on inserts were mock or HAstV-1 (infectious or UV inactivated) treated. At 24 hpi, when infected monolayers show significantly increased permeability, monolayers were stained for actin and visualized by immunofluorescent microscopy. Mock-infected monolayers had a well-formed cytoskeleton. Apically, actin appeared in a cobblestone pattern, representing the perijunctional actomyosin ring, and well-developed actin stress fibers were evident (Fig. 7A and D). No viral staining was observed (Fig. 7G). In contrast, by 24 h after HAstV-1 infection, there was a dramatic decrease in the number of stress fibers, and infected cells appeared to physically separate at the apical surface (Fig. 7B and G). The actin cytoskeleton was disrupted in cells not expressing viral antigen (Fig. 7H and E), suggesting either that infection in these cells was below the level of detection or that a factor released from neighboring infected cells was causing this effect. Monolayers that had been treated with UV-inactivated HAstV-1 showed similar changes in their actin cytoskeletons (Fig. 7C and F), despite the fact that no viral staining was detected (Fig. 7I).

FIG. 7.

HAstV-1 infection disrupts the actin cytoskeleton. Differentiated Caco-2 cells grown on inserts were treated with mock-infected lysates (A, D, and G), UV-inactivated HAstV-1 (C, F, and I), or HAstV-1 virus (B, E, and H). At 24 hpi, monolayers were fixed and stained for actin (A to F) and HAstV-1 capsid protein (G to I). Protein localization was visualized by epifluorescent microscopy. All images are at ×40 magnification; arrows identify regions of protein disruption.

Changes in the actin cytoskeleton can disrupt TJ components (30). To determine the effect of HAstV-1 on TJ proteins, monolayers were fixed; stained for occludin, claudin, and ZO-1; and visualized by confocal fluorescent microscopy. Mock-infected monolayers demonstrate peripherally restricted occludin, ZO-1, and claudin, as indicated by the cobblestone staining pattern, as well as considerable pools of cytoplasmic occludin (Fig. 8A and data not shown). In contrast, occludin was removed from TJ strands and breaks in the cobblestone staining pattern were evident in HAstV-1-infected monolayers by 24 hpi (Fig. 8B). The diffuse pools of cytoplasmic occludin staining observed in mock-infected cells was absent in HAstV-1-infected cells at this time. Furthermore, astrovirus-infected cells appeared distended compared to infected monolayers. Relocalization of ZO-1 and claudin during HAstV-1 infection was not observed until 36 hpi, suggesting that the actin and occludin changes precede the increase in permeability (data not shown). Comparable changes in occludin localization were observed in monolayers treated with UV-inactivated HAstV-1 (Fig. 8C), suggesting that UV-inactivated virus acts through the same mechanism as infectious virus to increase barrier permeability and further supporting the hypothesis that viral replication is not required for the observed effects.

FIG. 8.

HAstV-1 infection disrupts occludin. Differentiated Caco-2 cells grown on inserts were mock (A), UV-inactivated HAstV-1 (C), or HAstV-1 (B) infected. At 24 hpi, monolayers were fixed and stained for occludin. Protein localization was visualized by confocal microscopy. All images are at ×40 magnification; arrows identify regions of protein disruption between neighboring cells.

DISCUSSION

The mechanism by which astroviruses cause diarrhea remains unknown. Studies utilizing a turkey model of astrovirus pathogenesis have demonstrated that only mild histological changes occur during infection. No increase in cell death or inflammation was observed, nor was blood present in stools of astrovirus-infected turkeys (24). These observations, which have been supported in human studies (42), suggest that epithelial destruction or unregulated inflammation is not the cause of astrovirus-induced diarrhea. In the absence of other known causes of diarrhea, we hypothesized that increased intestinal barrier permeability played a role in astrovirus infection.

We have shown that the addition of HAstV-1 to the apical surface of a Caco-2 model intestinal barrier increases permeability. Studies with UV-inactivated virus, resulting in RNA breaks and uracil dimers (45), and with purified VLPs suggest that replication is not required and that the capsid alone can mediate the increase in permeability. The kinetics of increased permeability were similar between purified VLPs and infectious HAstV-1. Unfortunately, at this time we do not have the tools to determine how the concentration of the VLPs used in these studies correlates with the number of particles (both infectious and empty) in viral inoculum. Studies are under way to develop the methodologies to make these comparisons. Regardless, it is clear that events early in the virus life cycle, specifically binding or entry, mediate the increase in permeability.

Attempts to determine if entry was the key step were unsuccessful. Both monensin and NH₄Cl are known to inhibit viral entry by ≥99% at concentrations of 0.01 and 20 mM, respectively (10; data not shown); however, treatment of Caco-2 monolayers at these concentrations decreased TER levels to less than 15% of initial values by 12 hpi (data not shown), rendering such experiments ineffective. Thus, the specific trigger for HAstV-1-induced permeability, whether binding alone or binding and entry, requires further exploration.

We next investigated the cellular mechanisms of increased permeability during astrovirus infection. Reports of astrovirus-induced cell death are conflicting; for humans (42) and in a turkey animal model (24), no increase in cell death was observed during infection. However, porcine astrovirus is cytolytic in vitro (44), and HastV-4 and -8 have been shown to cause apoptosis in cell culture (19, 36). No significant difference in cell death between mock- and HAstV-1-infected cells was observed in our experiments, nor did UV-inactivated virus or VLPs increase cell death. These observations do not necessarily conflict with those of previous reports. We evaluated death through 36 hpi, a time at which a significant increase in permeability is observed. Guix et al. did not observe an increase in apoptosis until 48 hpi (19). It therefore seems clear that cell death is not the major cause of increased permeability in our system, although it may contribute later in infection.

We examined several cellular proteins responsible for controlling barrier permeability. The TJ complex is composed of transmembrane proteins that seal the intracellular spaces, the most well-characterized of which are occludin (13) and claudin (25). These proteins interact with cytosolic adapter and signaling molecules, which subsequently interact with the actin cytoskeleton (28). In this way, permeability is regulated by external or internal signals; disruptions at many levels of these interactions can result in increased barrier permeability (12). This phenomenon has been well described for infections by other enteric pathogens, such as rotavirus (20), enteropathogenic Escherichia coli (39), and Clostridium difficile (40). We therefore evaluated TJ protein and actin localization during HAstV-1 infection. In cells treated with infectious HAstV-1 or UV-inactivated virus, occludin was disrupted by 24 hpi, corresponding to a dramatic decrease in the number of actin stress fibers. Changes in ZO-1 and claudin were not observed until 36 hpi, suggesting that actin and occludin relocalization occurred first. This is in contrast to findings for mock-treated cells, which exhibited well-developed perijunctional actomyosin rings, stress fibers, and peripheral staining of TJ proteins. Most agents known to increase barrier permeability do so within minutes or a few hours (40). However, HAstV-1 takes ∼20 h. This is not without precedent. Nitric oxide treatment of Caco-2 cells increases barrier permeability by about 12 h (41), while tumor necrosis factor alpha-increased permeability requires at least 24 h (27). Studies are under way to determine if the binding/entry of HAstV-1 results in the synthesis of a cellular factor that increases permeability.

What is the benefit of increasing barrier permeability? One reason may be to increase the spread of the virus. Hypothetically, the flux of fluids driven by the increased permeability may increase viral dissemination. Alternatively, disruption of TJs may allow enteric viruses to move from the intestinal lumen into the serosa, where the virus gains access to the bloodstream and the rest of the body. Animal models have demonstrated systemic turkey astrovirus type 2 (24) and rotavirus (7) infections supporting this hypothesis. Finally, increased barrier permeability may expose a viral receptor previously sequestered at the basolateral surface or within junctional complexes themselves. This has been demonstrated nicely for type B coxsackieviruses, which utilize the junctional adhesion molecule CAR. Disruption of the TJ exposes CAR, allowing a productive infection to occur (6). If a similar scheme is required for astrovirus infection, it would be extremely beneficial for structural proteins to increase permeability, as they are produced in abundance and released during infection. HAstV-1 can infect Caco-2 cells from the basal surface (data not shown); however, the efficiency of infection and virus production relative to that of apical infection as well as the ability of the virus to spread basally are currently unknown. Investigations into the role of increased permeability during astrovirus infection and spread are currently under way.

In conclusion, we have demonstrated that astrovirus infection at the apical surface of model intestinal epithelia results in a time-dependent increase in barrier permeability. This increase in permeability is associated with disruption of the TJ protein occludin as well as the actin cytoskeleton and occurs independently of viral replication. This is the first study to demonstrate that astrovirus increases barrier permeability; future studies will focus on further understanding the cellular mechanisms that contribute to these processes both in vitro and in vivo.