The rotavirus surface protein VP8 modulates the gate and fence function of tight junctions in epithelial cells

Rotaviruses are a major cause of diarrhea in young mammals. In developing countries rotavirus infections cause 600,000 childhood deaths per year and constitute an important source of morbidity in industrialized countries (Parashar et al., 2003). These viruses of the Reoviridae family have a genome composed of 11 double-stranded RNA segments, surrounded by three concentric layers of protein. The outermost layer is smooth and formed by a 37 kDa glycoprotein named VP7. From it, 60 spikes formed by an 88 kDa protein named VP4, project outwards (Estes, 2002). VP4 is essential for early virus-cell interactions, because it participates in receptor binding and cell penetration (Lopez and Arias, 2004). In fact, the infectivity of rotaviruses is dependent upon the specific cleavage by trypsin of VP4 into subunits VP5 and VP8 (Almeida et al., 1978; Espejo et al., 1981; Estes et al., 1981).

In vivo rotavirus infects primarily the mature enterocyte of the villi of the small intestine (Kapikian and Chanock, 2002). In vitro rotaviruses bind to a wide variety of cell lines, although only a subset of these, including cells of renal or intestinal origin, and transformed cell lines derived from breast, stomach, bone and lung, are efficiently infected (Ciarlet et al., 2002).

Some rotaviruses bind to a cell-surface receptor containing sialic acid (SA), whereas others (e.g. human rotavirus) apparently do not require SA for infection (Fukudome et al., 1989). Furthermore, association to a secondary SA-independent receptor can overcome the initial interaction of certain rotavirus with SA. SA-dependent rhesus rotaviruses (RRV) initially bind to the cell through VP8 (Fiore et al., 1991; Isa et al., 1997), whereas variants of RRV that no longer depend on the presence of SA (e.g. nar3), interact with the cell through VP5 (Zarate et al., 2000b).

The rotavirus surface proteins contain integrin ligand sequences (Coulson et al., 1997; Guerrero et al., 2000) and several integrins like α₂β₁ (Zarate et al., 2000a), α_νβ₃ (Guerrero et al., 2000), α_Xβ₂ and α₄β₁ (Coulson et al., 1997; Hewish et al., 2000) are known to function as rotavirus cell receptors (for a review, see Lopez and Arias, 2004).

In epithelial and endothelial cells integrins have a polarized distribution and localize at the basolateral plasma membrane (Gut et al., 1998; Hynes, 1992). Therefore, rotaviruses reaching the intestinal epithelium from the apical surface or being in contact with the lumenal surface of confluent epithelial cell lines would find their integrin receptors hidden beneath the tight junction (TJ) on the basolateral surface. How then might luminal rotaviruses with putative basolateral ligands interact in polarized epithelia? One possibility is that rotavirus outer capsid proteins might be capable of opening the paracellular space sealed by the TJ. In this study we have explored this issue, and have found that the VP8 protein of rotavirus is capable of modulating TJ sealing in epithelia.

Madin-Darby canine kidney (MDCK) cultures between 60th and 100th passage were grown as previously described (Perez-Moreno et al., 1998).

The RRV VP8 fragment (VP4 gene nucleotides 1 to 750) was cloned in pGEX-4T (Pharmacia) and in pET-28 (Novagen) plasmids as previously described (Dowling et al., 2000; Guerrero et al., 2000; Isa et al., 1997). The RRV VP5 fragment (VP4 gene nucleotides 749 to 2347) was cloned in pGEX-4T-2 (Pharmacia) as previously described (Zarate et al., 2000b). Expression and purification of fusion proteins was performed following previously described standard procedures (Frangioni and Neel, 1993). For their experimental use, fusion proteins VP5 and VP8 were dissolved in fresh Dulbecco's minimal essential medium (DMEM) and sterilized by passage through a 0.22 μm filtration unit.

The degree of sealing of the TJ was evaluated by measuring the transepithelial electrical resistance (TER) across the monolayers as previously described (Gonzalez-Mariscal et al., 1985). TER determinations were done on single monolayers that were then discarded. The number of monolayers employed for each time point is shown in parenthesis. The values of TER shown have been normalized as previously reported (Balda et al., 1991).

The immunofluorescence was performed according to standard procedures previously described (Jaramillo et al., 2004). The following antibodies were employed: rabbit polyclonals against ZO-1 (diluted 1:100 in 1% Ig-free BSA, Zymed 61-7300), claudin-3 (diluted 1:20 in 1% Ig-free BSA, Zymed 34-1700), occludin (diluted 1:100 in 1% Ig-free BSA, Zymed 71-1500) and α_νβ₃ integrin (diluted 1:75 in 1% Ig-free BSA, Chemicon International MAB1976); and mouse monoclonals against the apical cell surface glycoprotein GP135 (a generous gift of George Ojakian, New York State University) and the β Na⁺-K⁺-ATPase (a generous gift of Marcelino Cereijido, CINVESTAV, Mexico). As secondary antibodies we employed: FITC-conjugated goat anti-rabbit (catalogue no. 62-6111, diluted 1:100, Zymed), CY5-conjugated goat anti-rabbit (catalogue no. 62-6116, diluted 1:100, Zymed) or FITC-conjugated goat anti-mouse (catalogue no. 62-6511, diluted 1:100, Zymed). The fluorescence of the monolayers was examined using a confocal microscope (Leica SP2) with a Krypton-argon laser.

The ability of VP8 to inhibit the development of TJs was assessed utilizing the Ca²⁺ switch assay (Gonzalez-Mariscal et al., 1985).

This assay was done as has been described elsewhere (Cereijido et al., 2002). Briefly, 1.5 ml of the tracer solution (10 μg/ml FITC-Dextran of 4 or 70 kDa; Sigma-Aldrich, catalogue no. FD-4, FD-70S) were added to the apical side of confluent monolayers grown on Transwell filters (Costar, catalogue no. 3450). After one hour of incubation at 37°C the media from both the upper and bottom chambers was collected and the FITC Dextran was measured in a fluorometer (excitation 492 nm; emission 520 nm).

This assay was performed as has been previously described (Cereijido et al., 2002; Contreras et al., 2002). Briefly, MDCK cells grown on Transwell inserts (Costar, catalogue no. 3470) were labeled on the apical side for 10 minutes on ice with a 5 nmol/ml solution of the Bodipy® FL-C12-sphingomyelin-BSA complex (Molecular Probes, Eugene, OR, USA). Then they were washed thrice with cold P buffer (NaCl 145 mM, HEPES pH 7.4 10 mM, Na-pyruvate 1.0 mM, glucose 10 mM and CaCl₂ 3.0 mM) and incubated on ice for 30 minutes. The membranes were immediately observed by confocal microscopy.

Apical and basolateral plasma membrane proteins of confluent MDCK cells grown in Transwell cell culture inserts (Costar, catalogue no. 3470), were biotinylated by applying 500 μg/ml sulfo-NHS-SS biotin (Pierce, catalogue no. 21331), at either the apical (0.5 ml) or the basolateral (1.5 ml) chambers, as previously described (Gottardi et al., 1995). Filters were then excised and monolayers solubilized. β₁ integrin was immunoprecipitated with a specific antibody (Chemicon International, catalogue no. AB1952) as previously described (Perez-Moreno et al., 1998). The samples (50 μl each) were then separated on a 10% SDS-PAGE, transferred to PVDF membranes and blotted with HRP-streptavidin (Zymed 43-4323; diluted 1:2000). Proteins were then detected with the ECL Plus reagent.

Freeze-fracture replicas of TJs present in confluent monolayers of MDCK cells incubated for 1 hour with or without 4 μg/ml of GST-VP8 were performed according to standard procedures described before (Gonzalez-Mariscal et al., 1985) using a Balzers apparatus (BAF400T). Observations were done in an electron microscope JEOL 200EX.

To assess modifications in the pattern of TJ strands, we counted the number of filaments intercepting lines drawn perpendicular to the main axis of the junction, every 48 nm, as well as the number of loose ends found along the 62 μm and 60 μm of TJ networks analyzed respectively for control and VP8-treated monolayers.

Western blots were made as previously described (Jaramillo et al., 2004) using polyclonal antibodies against ZO-1, claudin-3 and occludin.

Male Wistar rats (230-250 g) were maintained on PicoLab® Rodent Diet 20, sterilized by irradiation and water ad libitum, in the animal house (temperature 22-24°C, 50-55% humidity). Care and handling of the animals were in accordance with international recommended procedures.

Diabetes mellitus type I was induced with one intraperitoneal shot of streptozotocin (75 mg/kg weight; Sigma, catalogue no. S-0130), diluted in citrate buffer 0.1 M (Sigma, catalogue no. S-4641) pH 4.5. Streptozotocin solution was prepared immediately before use and protected from light exposure.

The animals were anesthetized by ethylic ether inhalation. Using heparinized capillary tubing (Chase Scientific Glass, catalogue no. 2501) a drop of blood was taken from the eye orbital sinus of the rats. Blood glucose concentration was determined using reactive strips (One Touch, LIFESCAN, Johnson & Johnson) and a commercial glucometer (One Touch Basic Plus, LIFESCAN, Johnson & Johnson). Glucose concentration was determined in healthy animals and in those that had received the streptozotocin intraperitoneal shot at least three days before being tested.

In order to study the changes in glucose concentration generated by the administration of insulin, the first blood sample was taken from diabetic animals that had fasted overnight. Immediately after, the animals received their food (Lab Diet 5053), and 30 minutes later they were treated according to the following protocols: (1) Human insulin of intermediate action (3-6 IU/rat; Humulin® N, HI-310, Lilly) was parenterally administered. (2) Human insulin of intermediate action (15-30 IU/rat; Humulin® N, HI-310, Lilly) was given orally through an esophagic rat cannula (Fine Science Tools; catalogue no. 18061-75), in a solution of 400 μl of NaHCO₃ (1.5 g/100 ml, pH 8.3-8.4), to neutralize gastric acidity. (3) Employing an esophagic rat cannula, 100 μg of GST-VP8 and human insulin of intermediate action (15-30 IU/rat; Humulin® N, HI-310, Lilly) were administered orally in a solution of 400 μl NaHCO₃ (1.5 g/100 ml, pH 8.3-8.4). (4) 100 μg of GST-VP8 was administered orally through an esophagic rat cannula. The level of blood glucose was determined at different times after the start of each of the above described procedures.

To investigate whether rotaviruses have the ability to open TJs, we proceeded to study the effect of the rotavirus outermost proteins on TER. We started by investigating whether VP5 and VP8 could modulate the TER of MDCK cells. For such aim we employed GST fusion proteins containing the RRV proteins VP5 and VP8. Fig. 1A shows how the addition of 4 μg/ml of GST-VP8 but not of GST-VP5 or of GST alone, significantly reduces the TER of epithelial monolayers after 30 minutes of treatment. Observe how the effect of VP8 upon the electrical resistance of the monolayer becomes reversible with time. To further illustrate the latter point we developed another assay in which GST-VP8 was withdrawn from the culture media (arrow). Observe how in these monolayers (open triangles) the TER recovers in a shorter time period. Fig. 1B shows how when 0.4 μg of GST-VP8 protein were used, the decline in TER obtained was smaller and lasted for a shorter period. In accordance, when a larger amount of GST-VP8 was employed, the TER was completely debased and the monolayers showed no recovery for the next two hours (Fig. 1B). We have also employed VP8 His-fusion protein and obtained similar results (data not shown). To analyze whether the effect of VP8 upon TER is dependent on the age of the monolayer, we proceeded to plate confluent monolayers and to determine their TER at different times of culture with and without VP8. As can be observed in Fig. 1C, MDCK monolayers exhibit an initial rapid increase in TER followed by a decrease to a stable level (gray bars). This variation may be associated, as previously reported (Cereijido et al., 1983; Rabito, 1986), with the number of cells in the monolayer. Thus, as the density increases, the length of the intercellular space per unit area of monolayer also augments. The amount of current that traverses the paracellular pathway depends not only on the resistance of this route, but also on how much pathway is available per epithelial surface [for a review on the structure-function relationship of TJs, see the following (Anderson and Cereijido, 2001; Gonzalez-Mariscal et al., 2001)]. Therefore the increased cell density of the monolayers might be responsible for the reduction in the value of TER. Treating the cells with 4 μg/ml of GST-VP8 for 30 minutes diminishes the TER values of all the monolayers (Fig. 1C, black bars). The white bars of Fig. 1C reveal how when a group of these monolayers is maintained for 2 hours with 4 μg/ml of GST-VP8 the effect disappears and the TER recovers.

Once we had demonstrated that VP8, but not VP5, could decrease the TER of confluent epithelial monolayers with well-formed TJs, we proceeded to study whether it could also alter the development of resistance in monolayers that are in the process of establishing cell-cell contacts. Such experiments named Ca-switch assays are based on our previous reports demonstrating that monolayers incubated for 20 hours in Ca-free media (LC) are devoid of TJs, and develop them with a fast kinetics once they are transferred to media containing 1.8 mM Ca²⁺ (NC) (Gonzalez-Mariscal et al., 1985; Gonzalez-Mariscal et al., 1990). Fig. 1D shows how the addition of 4 μg/ml of GST-VP8 but not of GST-VP5, to monolayers transferred from LC to NC media, delays the development of TER.

TJs also function as selective barriers to the diffusion of non-ionic molecules through the paracellular route. Therefore we examined whether GST-VP8 could affect the passage of two non-ionic tracers: 4 kDa and 70 kDa FITC-Dextrans (Fig. 2). In confluent MDCK monolayers a small amount of these tracers diffuses across the monolayers. Treatment with 4 μg/ml of GST-VP5 induces a slight increase in the passage of the 4 kDa dextran. In contrast, a one-hour treatment with 4 μg/ml GST-VP8 induces a 25-fold increase in the paracellular permeability of 4 kDa FITC-Dextran and a 6-fold increase in the diffusion of the 70 kDa FITC-Dextran. We therefore conclude that the treatment with VP8 induces a leakage of small non-ionic tracers from the apical domain.

To examine whether VP8 could modify the fence function of the TJ, we analyzed the distribution of a lipid (Bodipy® FL-C12-sphingomyelin/BSA) and apical and basolateral membrane proteins in monolayers treated with VP8. For this set of experiments, confluent MDCK monolayers grown on Transwell® filters, remained untreated (control) or were incubated for 1 hour with 4 μg/ml of GST or GST-VP8. Another set of monolayers was incubated for 10 minutes with 1.8 mM EGTA, because previous reports demonstrated that incubation of epithelial cells with EGTA opens TJs and generates epithelial depolarization (Mandel et al., 1993; Martinez-Palomo et al., 1980).

The lipid complex was applied to the apical surface of confluent MDCK monolayers. After washing, the diffusion of the Bodipy® FL-C12-sphingomyelin-BSA complex was immediately analyzed. The confocal z-sections shown in Fig. 3A illustrate how in control and GST-treated cells, the fluorescent lipid is confined to the apical domain and no lateral staining is detected. However, in MDCK cells treated with EGTA the Bodipy® FL-C12-sphingomyelin complex diffuses across the TJ and stains the lateral membranes, leaving almost no detectable stain on the apical surface. The cultures incubated with GST-VP8 display a pattern, in which lateral membrane staining starts to be detected, thus indicating alterations in the TJ fence function.

Next we proceeded to detect by immunofluorescence the distribution of polarized membrane proteins. In these assays occludin was simultaneously employed, because it allowed us to distinguish with precision the limit between the apical and the basolateral domains of the plasma membrane in control monolayers (arrows). Fig. 3B illustrates how in control and GST-treated monolayers, the apical cell surface glycoprotein GP135 (Ojakian and Schwimmer, 1988) is confined to the luminal surface, whereas in the cultures treated with GST-VP8 or EGTA the staining moves from the apical to the lateral membrane. In Fig. 3C we illustrate the effect of GST-VP8 on the distribution of the Na⁺K⁺-ATPase, a basolateral membrane marker. Observe how the clear basolateral distribution of the Na⁺K⁺-ATPase in control and GST-treated monolayers is altered when the monolayers are incubated with GST-VP8. This situation is more intensively observed in monolayers treated with EGTA.

As stated in the Introduction, because rotaviruses employ integrins as cell surface receptors, we proceeded to analyze whether VP8 altered their polarized distribution in epithelia. Fig. 3D reveals how in control and GST-treated monolayers, the rotavirus receptor, integrin α_νβ₃, is detected along the basolateral membrane. In contrast, treatment with VP8 or EGTA induces the appearance of the integrin on the apical surface. This effect was further confirmed in an assay in which either the apical or the basolateral membrane surfaces of monolayers were labeled with sulfo-NHS-SS-biotin. The results in Fig. 3E reveal a redistribution of β1 integrin from the basolateral to the apical surface. These observations thus demonstrate that VP8 has the capacity to alter the distribution of integrins in epithelial cells. This effect might be of importance for rotaviruses infectivity because it would allow enhanced exposure of integrins to the luminal environment containing the viral particles.

Next, we asked whether VP8 could affect the TJ structure. For such a purpose we examined by freeze-fracture EM the TJ strand organization (Fig. 4, upper panel). In control monolayers (A) fractured faces of the apical plasma membrane display the characteristic organization of TJs: a network of interconnecting fibrils in the P face of the plasma membrane complementary to grooves on the E face. Treatment with VP8 (B) induces the appearance of loose ends that depart from the main network of TJ fibrils (arrowheads) together with discontinuous series of stranded clusters (asterisks). The arrangement of the main TJ axis is also simpler, as filaments have lost their interconnected appearance.

The morphometric analysis (Fig. 4, lower panel) reveals no changes neither in the number of TJ strands nor in the total amount of junctions between control and VP8-treated monolayers. In contrast, in VP8-treated monolayers, the number of loose ends that run quasi perpendicular to the main junctional axis is increased (Table 1). This change in the junctional pattern could explain the altered fence and gate properties of the TJs observed in VP8-treated monolayers.

We then examined whether the observed TJ strand disorganization induced by VP8, correlated with changes in the localization of TJ proteins. For this purpose, confluent MDCK monolayers receiving no treatment or incubated for 1 hour with 4 μg/ml of GST or GST-VP8 were examined by confocal immunofluorescence. The distribution of ZO-1, a submembranous protein known for its scaffolding characteristics, as well as of two transmembrane proteins, claudin-3 and occludin [for a review on TJ proteins, see Gonzalez-Mariscal et al. (Gonzalez-Mariscal et al., 2002)], was analyzed in xy and xz sections. Fig. 5A shows conspicuous sharp ring-like structures of ZO-1, claudin-3 and occludin on the lateral membranes between neighboring cells in control monolayers and in those treated with GST. The xz sections of control and GST-treated cells show the typical TJ dotted staining of ZO-1, claudin-3 and occludin. In GST-VP8-treated monolayers, staining for these proteins is greatly diminished, thus indicating that VP8 alters the molecular architecture of TJs.

Next, we proceeded to study by western blot the presence of ZO-1, claudin-3 and occludin in monolayers treated with GST-VP8. We analyzed Triton X-100 soluble and insoluble fractions, because certain TJ altering treatments displace TJ proteins from the insoluble to the soluble fraction (Chen et al., 2000). After VP8 treatment, no significant change was detected in the amount or in the distribution of these TJ proteins in the soluble and insoluble fractions (Fig. 5B). Therefore the discontinuous staining of the TJ proteins observed by immunofluorescence suggests that the TJs can be disorganized by VP8 without decreasing the cellular content of these TJ proteins, as has been previously observed in epithelial cells treated with tumor necrosis factor-alpha (Poritz et al., 2004) and peptides from the extracellular loops of occludin (Tavelin et al., 2003).

Once we had demonstrated that the modulation of TJs by VP8 is reversible, time and dose dependent, we proceeded to study whether it could be used in an animal model to facilitate the passage of molecules through epithelia. We selected insulin as a prototype molecule to be tested for oral delivery, based on its relative small size, structure, biological activity and therapeutic relevance. We worked with streptozotocin-induced diabetic male rats in which we evaluated the bioactivity of insulin after enteral co-administration with 100 μg VP8. Fig. 6A illustrates how three days after a single streptozotocin intraperitoneal shot (75 mg/kg body weight) the rats display a much higher level of blood glucose (broken lines) after an overnight fasting period (time zero) and at different times after food is allowed, as compared with animals receiving a mock shot (continuous lines). When the streptozotocin-diabetic rats receive, via an esophagic cannula, oral insulin (30 IU), the blood glucose level does not decrease (Fig. 6B, broken lines), whereas as expected those animals parenterally receiving insulin (6 IU) display a sharp drop in their blood glucose concentration that is maintained for at least 3 hours after the animals are allowed to eat (Fig. 6B, continuous lines). When the streptozotocin-diabetic rats receive an oral administration of GST-VP8 (100 μg) no change in the animals' glucose level is detected (Fig. 6C, broken lines). Instead, when the diabetic rats receive a co-administration of GST-VP8 (100 μg) and insulin (30 IU), the decrease in their blood glucose level resembles that observed in rats receiving parenterally administered insulin (Fig. 6C, continuous lines). The same results were obtained when instead of GST-VP8, His-VP8 was used (data not shown). None of the animals treated with GST-VP8 experienced diarrhea, fever or other systemic symptoms. These results therefore suggest that VP8 could provide a new strategy for the oral delivery of drugs and proteins not normally absorbed through the intestine.

Our results demonstrate that the rotavirus protein VP8 is capable of disrupting the TJs of epithelial cells in a reversible, time and dose dependent manner. The effect of VP8 alters both the gate and fence function of intercellular TJs. The former is evidenced by the VP8-generated decrease in TER and by the observed increase in paracellular permeability to non-ionic tracers, thus indicating that the VP8-induced permeability changes apply to both charged and uncharged compounds. The structural organization of the TJ responsible for the intramembrane fence is impaired by VP8 treatment, as determined by the lateral diffusion of the apically administered lipid probe Bodipy® FL-C12-sphingomyelin-BSA complex and the apical marker GP135, and by the movement to the apical surface of the basolateral proteins Na⁺K⁺-ATPase and integrins α_νβ₃ and β₁ subunit.

In agreement with these results, we were able to detect an altered pattern of TJ strands in freeze-fracture replicas of VP8-treated monolayers, as well as a displaced distribution of the TJ constituents occludin, claudin-3 and ZO-1.

On analyzing the predicted amino acid sequence of RRV VP8 we found that several segments of this protein had a high similarity to domains present in the extracellular loops of claudins and occludin (Fig. 7), the transmembrane proteins of the TJ which are presumed to constitute the backbone of TJ strands (Tsukita et al., 2001). This observation poses the interesting possibility that rotavirus protein VP8 might open the TJ by perturbing the extracellular domains of TJ proteins in a manner similar to that exerted by synthetic peptides corresponding to the extracellular domains of occludin (Chung et al., 2001; Lacaz-Vieira et al., 1999; Van Itallie and Anderson, 1997; Wong and Gumbiner, 1997).

Besides humans, rotaviruses infect a wide variety of animal species such as monkeys, calves, pigs, horses, sheep, dogs, cats, mice, chicken, etc (Estes and Cohen, 1989; Kapikian and Chanock, 2002). Comparison of the deduced amino acid sequences of the VP4 proteins of rotavirus strains obtained from different animal species, has shown that although the VP4 sequence correlates with the species of origin of the virus strain, a high interspecies relatedness is present (Taniguchi et al., 1994). In addition, although VP8 is the less conserved region of VP4, a substantial degree of identity is still maintained among different rotavirus strains, and, in principle, the VP8 subunits from these strains could generate the same effects observed in this study.

Identification of viral receptors has yielded insight into mechanisms of viral entry. On several occasions these receptors have surprisingly turned out to be cell-cell adhesion proteins. For example, rhinoviruses bind to the intercellular adhesion molecule ICAM-1 (Greve et al., 1989; Staunton et al., 1989), α-herpes viruses to the poliovirus receptor related protein (Prr1) (Geraghty et al., 1998), coxsackie B and adenoviruses 2 and 5 bind to the CAR TJ protein (Bergelson et al., 1997), and reoviruses associate with the JAM protein located at the TJ (Barton et al., 2001). Furthermore, several diarrheogenic bacteria alter the junctional complexes in polarized epithelial cells. For example, infection with Yersinia pseudotuberculosis (Tafazoli et al., 2000), Salmonella enterica, Helicobacter pylori, enteropathogenic and enterohemorrhagic Escherichia coli results in alteration of the junctional complexes (for a review, see Sears, 2000). In the case of rotaviruses, entry into the cells appears to be a multi-step process involving the interaction with certain integrins (Lopez and Arias, 2004). Our results suggest that VP8 is capable of generating leaky TJs that allows the exposure of integrins to the luminal surface. After VP8 exposure, the epithelial TJs might still represent a diffusion barrier to large particles such as the rotavirus. Thus, a successful virus infection could alternatively be obtained through the appearance of apically positioned integrins like α₂β₁, α₄β₁, α_xβ₂ and α_νβ₃, which have been reported to function as rotavirus receptors in epithelial cells. This would allow rotavirus to bind and infect from the luminal domain without having to cross the TJ. Hyper-permeability of the intestine during rotavirus infection has been documented in vivo, with enteral markers such as lactulose and mannitol (Isolauri et al., 1989; Jalonen et al., 1991; Uhnoo et al., 1990), and rotavirus infection of epithelial cell lines has been shown to generate an increased transepithelial macromolecular paracellular permeability, disruption of TJs and loss of TER at late times of infection (Ciarlet et al., 2001; Dickman et al., 2000; Obert et al., 2000).

Finally, the results obtained in this work suggest that VP8 could be employed as a potential tool for the modulation of TJ permeability. In this respect it should be remembered that intestinal absorption of numerous compounds routinely used for the treatment of diseases is limited by their physicochemical characteristics. With the exception of molecules transported by passive diffusion across the cell membrane (lipophilic molecules) or by carrier-mediated transport, absorption of large and hydrophilic molecules is almost exclusively limited by the presence of TJs that control passage through the paracellular pathway. Therefore the design of agents that can effectively and safely increase paracellular permeability via modulation of TJs has become increasingly important (Ward et al., 2000). Here we have shown that VP8 can be used in rats to enhance the intestinal absorption of orally administered insulin through the paracellular pathway. VP8 displays multiple properties that make it a promising tool to promote the transport of drugs, proteins and peptides across epithelia, because: (1) When employed in low doses it partially abolishes the transepithelial resistance, (2) it induces a reversible increase of tissue permeability, (3) in the rat model analyzed, it does not induce diarrhea, and (4) its co-administration in rat with biologically active compounds such as insulin enhances intestinal absorption of the active molecule.

In summary, in this study we have demonstrated that a rotavirus protein of the outer capsid, VP8, is capable of transiently opening TJs. This knowledge can be employed as a tool to enhance enteral administration of macromolecules.

We thank our technician Miguel Rojas for his efficient management of the animals employed in the present study and Pedro Romero for the plasmid constructions. We are grateful to Marcia Hiriart from the Department of Cellular Physiology at UNAM, for her help in generating the streptozotocin-induced diabetic animals. This work was supported by grants 37846-N and G37621N from the Mexican National Council for Science and Technology (CONACYT), and by grants 55000613 and 55003662 from the Howard Hughes Medical Institute. Porfirio Nava is a recipient of a doctoral fellowship from CONACYT (143090).