Clostridium perfringens Epsilon-Toxin Increases Permeability of Single Perfused Microvessels of Rat Mesentery

Epsilon-toxin, the primary virulence factor of Clostridium perfringenstype D, causes mortality in livestock, particularly sheep andgoats, in which it induces an often-fatal enterotoxemia. Itis believed to compromise the intestinal barrier and then enterthe gut vasculature, from which it is carried systemically,causing widespread vascular endothelial damage and edema. Herewe used single perfused venular microvessels in rat mesentery,which enabled direct observation of permeability propertiesof the in situ vascular wall during exposure to toxin. We determinedthe hydraulic conductivity (L_p) of microvessels as a measureof the response to epsilon-toxin. We found that microvesselswere highly sensitive to toxin. At 10 µg ml^–1 theL_p increased irreversibly to more than 15 times the controlvalue by 10 min. At 0.3 µg ml^–1 no increase in L_pwas observed for up to 90 min. The toxin-induced increase inL_p was consistent with changes in ultrastructure of microvesselsexposed to the toxin. Those microvessels exhibited gaps eitherbetween or through endothelial cells where perfusate had directaccess to the basement membrane. Many endothelial cells appearednecrotic, highly attenuated, and with dense cytoplasm. We showedthat epsilon-toxin, in a time- and dose-dependent manner, rapidlyand irreversibly compromised the barrier function of venularmicrovessel endothelium. The results conformed to the hypothesisthat epsilon-toxin interacts with vascular endothelial cellsand increases the vessel wall permeability by direct damageof the endothelium.

INTRODUCTION

Top

Introduction

One of the most potent of the clostridial toxins, epsilon-toxinof Clostridium perfringens, has been named a class B selectagent of the Centers for Disease Control and Prevention anda category B priority agent of the National Institute of Allergyand Infectious Diseases. While newly regarded as a potentialbioterrorism tool, epsilon-toxin has long been recognized asan important cause of mortality in livestock, particularly sheepand goats, where it is the primary virulence factor of C. perfringenstype D and leads to an often-fatal enterotoxemia. C. perfringenstype D can be a normal inhabitant of the intestine of sheepand several other animal species, but when large amounts ofeasily fermentable carbohydrates pass into the intestine, generallydue to sudden changes in diet, this microorganism proliferatesand produces large quantities of epsilon-toxin. Although littleevidence is available in this regard, it is generally acceptedthat epsilon-toxin compromises the intestinal barrier and isthen taken up by the vasculature of the gut, from which it isspread systemically. The effects of injecting epsilon-toxinintravenously have been studied in sheep (7), mice (10), goats(38), rats (11), and cattle (39). The toxin thus administeredproduces increased vascular permeability in many tissues, themost significant effects of which are acute pulmonary and cerebraledema, which is associated with vascular endothelial injurydemonstrable with transmission electron microscopy (13). Thus,the toxemia has been assumed to directly cause widespread damageto the vascular endothelium, although no definitive proof hasbeen produced about the direct action of epsilon-toxin on endothelialcells.

Of the many cultured cells that have been tested for sensitivityto epsilon-toxin, only Madin-Darby canine kidney (MDCK) cellsand Caucasian renal leiomyoblastoma (G-402) cells are sensitiveenough for useful investigation of the toxin (4, 32, 40). TheMDCK cells are about 100-fold more sensitive than the G-402cells and have been used in most cultured cell studies of epsilon-toxin.In MDCK cells, epsilon-toxin was found in high-molecular-weightcomplexes and induced increased membrane permeability to ionsand small hydrophilic solutes (33). It forms an aqueous anion-selectivepore permeable to solutes up to 1 kDa (35). Analysis of labeledconstructs derived from epsilon-toxin clearly demonstrated thatit forms heptamers when incubated with rat brain synaptosomalmembranes and that cleavage of a C-terminal peptide is essentialfor oligomerization (26). Also in MDCK cells, epsilon-toxinassociates with apical membrane in preference to basolateralmembrane and forms a heptameric pore associated with detergent-resistantdomains (27, 34). While it has relatively low amino acid sequenceidentity (ca. 14%) to the heptameric pore-forming toxin aerolysinfrom Aeromonas hydrophila, it does have three-dimensional structuralsimilarity to aerolysin, one of a family of toxins known asß-pore-forming toxins (8). Epsilon-toxin also hassome amino acid sequence identity to a family of mosquitocidaltoxins typified by Mtx2 and Mtx3 of Bacillus sphaericus (21%and 24% identity, respectively) (9, 22, 37). Because only somecultured cells are highly sensitive to the toxin it has beensuggested that a specific host receptor is essential for bindingand pore formation, but no specific receptor molecule has beenfound. A recent study demonstrated that experimental developmentof tolerance to epsilon-toxin by MDCK cells was associated withloss of expression of a group of membrane proteins, lendingfurther support to the hypothesis that host proteins mediateepsilon-toxin binding and toxicity (4). Because of the assumedassociation of epsilon-toxin with vascular endothelium, developmentof an endothelial model is highly desirable. However, culturedaortic endothelial cells (goat, sheep, and cattle) failed toshow any morphological response to epsilon-toxin with up to48 h of exposure, even at concentrations much higher than weretoxic for MDCK cells (40). The latter observation brought intoquestion whether the in vivo toxicity of epsilon-toxin is dueto direct action of the toxin with vascular endothelium.

Because the cultured endothelial cells were not morphologicallyaffected by epsilon-toxin, in this study we have used singleperfused microvessels in rat mesentery, which enable directobservation of permeability properties of the vascular wallduring exposure to toxin under controlled conditions. Whilethe effects of epsilon-toxin in tissue have been investigatedusing morphological and histochemical techniques to demonstrateaccumulations of fluid and labeled proteins, there are no reportsin the literature of biophysical permeability coefficients determinedfrom directly measured fluxes and controlled driving gradients.Moreover, by presenting the toxin directly to the endotheliumwithin an artificial perfusate, the technique enables investigationof the endothelial response in the absence of potential systemicmechanisms or interactions with blood plasma. Variations ofthe single perfused microvessel model have been used to investigatevascular permeability coefficients regarding convective waterflow and permeability of both small and large solutes (19, 24,25), regulation of inflammatory mechanisms (3, 16), ultrastructureof the endothelial cleft (2), and regulation of the adherensjunction complex (1, 42). Here we determine the hydraulic conductivitycoefficient (L_p) as a measure of the response to epsilon-toxin.We demonstrate that microvessels of the rat mesentery are highlysensitive to epsilon-toxin, and we describe an experimentalsystem in which to investigate the direct action of bacterialtoxins with vascular endothelium.

MATERIALS AND METHODS

Top

Materials and Methods

Animal preparation. Experiments were carried out on rats (male, Sprague-Dawley,350 to 450 g; Hilltop Laboratory Animals, Inc.) anesthetizedwith pentobarbital (65 mg kg^–1 body weight given subcutaneously).Anesthesia, assessed by toe-pinch reflex, was maintained bygiving additional pentobarbital (subcutaneous, 3 mg per dose)as needed. At the end of experiments, animals were euthanizedwith pentobarbital overdose. Animal protocols were approvedby the Institutional Animal Care and Use Committee of the Universityof California, Davis. Each rat was placed on a heating pad tomaintain normal body temperature. A midline surgical incisionwas made in the abdominal wall, and the gut gently was takenfrom the abdominal cavity to spread the mesentery over a quartzpillar for transillumination and observation on the fixed stageof an inverted microscope. The upper surface of the mesenterywas continuously superfused with Ringer's solution maintainedat 35°C to 37°C during preparation and experimentation.Because our present techniques are not suited for use of vesselssmaller than about 15 µm in diameter, we have not investigatedthe properties of true capillaries (diameter of ca. 6 µm).Therefore, experiments were performed on straight nonbranchedsegments of venular microvessels typically 25 to 40 µmin diameter and usually between 700 and 900 µm in length.Prior to cannulation, all vessels selected for experiments hadbrisk blood flow and were free of leukocytes sticking or rollingon the vessel wall.

Measurement of L_p of the microvessel wall. All measurements were based on the modified Landis microperfusiontechnique, which enables measurement of the volume flux of wateracross the wall of a microvessel due to a known pressure differencebetween the lumen and the interstitium. The technique and itsassumptions and limitations have been evaluated in detail (2,24). Briefly, each vessel was cannulated by the tip of a glassmicropipette filled with a physiological perfusate additionallycontaining flow marker erythrocytes (ca. 4%, vol/vol). The micropipettewas positioned in the vessel to achieve free downstream perfusionwith no leak to the upstream direction or to the tissue. Therear of the micropipette was in communication with a water manometerused to set the hydrostatic pressure and drive flow from themicropipette through the cannulated vessel. To measure the initialtranscapillary water flow per unit area of the capillary wall(J_v/A), a glass microoccluder was lowered onto the microvesselup to 900 µm from the cannulation site, with the hydraulicpressure set at predetermined values of 30 to 60 cm H₂O. Whenthe vessels were occluded, the intravascular hydraulic pressurewas equal to that of the manometer, and the motion of flow markererythrocytes was recorded on videotape for later analysis. MicrovesselL_p was calculated as the slope of the relation between J_v/Aand applied hydraulic pressure. For most experiments, J_v/A wasestimated from each of several occlusions using one hydraulicpressure, with the assumption that the net effective pressuredetermining fluid flow was equal to the applied hydraulic pressureminus 3.6 cm H₂O, the approximate osmotic pressure contributedby bovine serum albumin (BSA) at 10 mg ml^–1. Experimentalreagents (e.g., epsilon-toxin) were added to the perfusate inthe respective experiments and delivered to the vascular lumenvia the micropipette continuously during L_p measurement. Changesin perfusate were accomplished by withdrawing the initial micropipetteand replacing it with a second micropipette filled with newperfusate solution of the appropriate composition.

Preliminary experiments were designed to determine the effectiveosmotic pressure ( ${Delta}$ ${pi}$ _eff) of albumin (30 mg ml^–1) in theperfusate. Before each measurement the hydraulic pressure inthe vessel was held at 60 cm H₂O for 30 s by partial downstreamocclusion to establish a high-filtration steady state in whichthe BSA concentration on the downstream side of the selectivitybarrier was washed down to a minimal level, thus maximizingthe effective oncotic pressure difference (24, 25). J_v/A wasmeasured by monitoring the motion of flow marker erythrocytesduring brief (5 to 8 s) full occlusions. With 60 cm H₂O hydraulicpressure, erythrocytes moved toward the occlusion site as fluidmoved out of the vessel. Measurements of J_v/A at microvesselpressure of 10 cm H₂O were made after setting up steady-statefiltration at 60 cm H₂O, occluding, and rapidly (<0.5 s)switching the applied hydraulic pressure to 10 cm H₂O. Reabsorptivewater movement was seen as marker erythrocytes moved away fromthe occlusion site toward the cannulation site as fluid movedinto the vessel. Three to five measurements were made at eachpressure, yielding a control relationship between J_v/A and appliedhydraulic pressure. The slope of the relationship is the hydraulicconductivity (L_p), and the intercept on the pressure axis is ${Delta}$ ${pi}$ _eff of the serum albumin. Our test solutions of BSA (10 or 30mg ml^–1) had protein osmotic pressures of 3.6 and 14 cmH₂O at 37°C (calculated from published equations [23]).

Experimental protocols. Each vessel was initially perfused with a control solution containingBSA at 10 mg ml^–1 in Ringer's solution (except preliminaryexperiments as indicated). Usually between 5 and 10 occlusionsat 50 cm H₂O over 15 to 20 min were used to establish a controlL_p. Then, the first pipette was removed and a second pipettecontaining the test solution was introduced at the same cannulationsite. Multiple occlusions over the subsequent 15 to 120 minenabled estimation of L_p during exposure to the test solution.Occlusions were made every 20 to 30 s during the first 5 minof test perfusion to check for rapid initial change in J_v/Aand then made less frequently (three to five occlusions every5 to 10 min) for the remainder of the experiment.

Statistics. We tested for differences from basal L_p due to the perfusionwith epsilon-toxin by comparing the averaged L_p in each experimentalgroup with the corresponding time-matched value found in a control(no-toxin) group of vessels. Because basal L_p distribution isnon-Gaussian, we used the nonparametric Mann-Whitney statisticand assumed significance for a P value of <0.05. Throughout,descriptive statistics are reported as the mean ± thestandard error of the mean (SEM).

Ultrastructure. To investigate the structural nature of the response to toxin,four vessels exposed to epsilon-toxin (two using 3 µgml^–1 and two using 10 µg ml^–1) were fixedafter a substantial rise in L_p (25 to 40 min of toxin exposure).Two other vessels were perfused with vehicle solution (controlvessels) for a comparable time and also inspected using thefollowing methods. After the final determination of J_v/A, fixationfor electron microscopy was started by dripping ice-cold glutaraldehyde(3% in 0.1 mM cacodylate buffer) on the mesentery. This wasdone while vessels were still perfused and occluded to ensurethat the final perfusion conditions and geometry were maintainedduring fixation. After 1 to 2 min the perfusion pipette wasremoved while immersion fixation continued. A diagram of thearea enabled later identification of the perfused vessel segment.After 10 min in situ fixation, tissue was cut away from theanimal and glutaraldehyde fixation continued on ice for a totalof 1 h before changing to OsO₄ (1% in veronal acetate, pH 7.2)for 1 h. Tissue was treated with tannic acid (0.5% in maleatebuffer, pH 6) for 30 min to enhance membrane staining and placedin uranyl acetate (2% in maleate buffer, pH 6) overnight. Afterdehydration in a graded series of acetone, the tissues wereflat embedded in epoxy resin. Sections (50 nm thick) from transverselysectioned vessels were collected on Formvar-coated slots andwere stained with uranyl acetate and lead citrate. They wereexamined using transmission electron microscopy (TEM) with aPhilips CM120 at 80 kV.

Solutions and reagents. Mammalian Ringer's solution was composed (in mM) of 132 NaCl,4.6 KCl, 2 CaCl₂, 1.2 MgSO₄, 5.5 glucose, 5.0 NaHCO₃, and 20HEPES and Na-HEPES. The ratio of acid-HEPES to Na-HEPES wasadjusted to achieve pH 7.40 to 7.45 for Ringer's solutions.All perfusates were mammalian Ringer's solution additionallycontaining BSA (A4378; Sigma-Aldrich) at 10 or 30 mg ml^–1as indicated.

Epsilon-prototoxin was prepared from an overnight culture ofC. perfringens type D (strain NCTC 8346) in Trypticase-yeast-glucosemedium, under anaerobic conditions at 37°C. The overnightcultures were centrifuged at 10,000 rpm for 30 min at 4°C,and the supernatant containing epsilon-toxin was saved for toxinpurification. The toxin was then precipitated by ammonium sulfate.Two columns were prepared with DEAE and CM Sepharose (Pharmacia,Sweden), respectively, equilibrated in 10 mM Tris, pH 7.5. Thetoxin was applied to the DEAE column, and the effluent was monitoredat 220 nm. The initially eluted peak was saved and applied tothe CM column. Again the effluent was monitored at 220 nm, andthe first peak was collected, dialyzed against distilled water,and freeze-dried. This product was shown to contain pure epsilon-prototoxinwhen tested by sodium dodecyl sulfate-polyacrylamide gel electrophoresis.Prior to its use in these experiments, the prototoxin was reconstitutedin Ringer's and activated by incubation at 37°C during 30min with 0.1% trypsin (Sigma).

To further establish specificity of the L_p response to toxin,we tested the effectiveness of an epsilon-toxin neutralizingmouse monoclonal antibody (immunoglobulin G; 5B7) provided byPaul Hauer (Center of Veterinary Biologics, Ames, Iowa). Theantibody (final concentration, 0.13 mg ml^–1) was addedto perfusion solution containing epsilon-toxin (3 µg ml^–1)and incubated for 30 min at room temperature before use. Theeffect on microvascular L_p was measured and compared to thetime-matched set of measurements made with epsilon-toxin inthe absence of antibody.

RESULTS

Top

Results

Epsilon-toxin increases microvascular L_p. We perfused single microvessels of rat mesentery with solutionscontaining epsilon-toxin to test for increased L_p as an indicationof endothelial cell intoxication. In a few preliminary experimentsthe perfusate contained 30 mg ml^–1 BSA, enabling determinationof the effective protein osmotic pressure difference acrossthe vessel wall. The data from a representative experiment illustratethat during control measurement in the absence of toxin theL_p (slope of the relation between J_v/A and applied pressure)was low and the effective protein osmotic pressure (intercepton the pressure axis) was between 10 and 20 cm H₂O (Fig. 1).After removing the initial pipette and perfusing with a secondpipette containing toxin (10 µg ml^–1), the L_p increasedover time and the effective protein osmotic pressure fell tonear zero. Both changes are consistent with loss of endothelialbarrier function. This initial technique required multiple measurements,typically taking several minutes at various pressures. Whenpermeability changed rapidly, it was difficult to make estimatesof the intercept on the pressure axis. Therefore, in remainingexperiments the BSA concentration was 10 mg ml^–1 and estimatesof L_p were made using one applied pressure (usually 50 or 60cm H₂O) as described in Materials and Methods. The latter techniquebetter enabled our following of rapid L_p changes.

View larger version (17K):
[in this window]
[in a new window]
FIG. 1. Representative data from a single experiment showing transmural fluid flux (J_v/A) against applied hydraulic pressure. During control perfusion in the absence of toxin, the L_p was 2.5 x 10^–7 cm x (s x cm H₂O)^–1 and ${Delta}$ ${pi}$ _eff was near 18 cm H₂O. After 10 min of perfusion with toxin, the L_p increased to 5.5 x 10^–7 cm x (s x cm H₂O)^–1 and ${Delta}$ ${pi}$ _eff had fallen to near zero. By near 20 min of toxin exposure, the L_p had risen to 13.5 x 10^–7 cm x (s x cm H₂O)^–1 and ${Delta}$ ${pi}$ _eff remained negligible.

A representative experiment shows the increase in L_p with timeassociated with toxin exposure (Fig. 2A). Perfusion with vehiclecontrol solution for 30 min established the control L_p of 0.3x 10^–7 cm s^–1 cm H₂O^–1. The initial micropipettewas removed, and the vessel was cannulated a second time andperfused with toxin (3 µg ml^–1) for the next 40min (Fig. 2A). During the first 5 min of perfusion with toxin,there was no apparent change in L_p. However, within 10 min ofexposure to epsilon-toxin the L_p increased to about 2 x 10^–7cm s^–1 cm H₂O^–1. It continued to increase and wasabove 15 x 10^–7 cm s^–1 cm H₂O^–1 by 40 minof toxin exposure. Representative data from a vessel perfusedfor 90 min without toxin illustrate that the perfusion techniquealone does not increase L_p (Fig. 2B). Averaged data from groupsof experiments are described next.

View larger version (22K):
[in this window]
[in a new window]
FIG. 2. Representative L_p data from two individual microvessels. (A) After perfusion with control (no toxin) solution for 30 min, the vessel was recannulated with epsilon-toxin (3 µg ml^–1). After about 5 min of exposure to epsilon-toxin the L_p of this vessel increased steadily during 40 min of perfusion. (B) Data from a separate experiment illustrate that perfusion with control solution (no toxin) does not induce a change in L_p.

To test the concentration dependence of the sensitivity to epsilon-toxinin the microvasculature, the L_p response to toxin over timewas tested in four different groups of vessels (Fig. 3A). Aseparate group of vessels perfused without toxin for 90 mindemonstrated that L_p did not change in that group (n = 15) (Fig.3B). The L_p values from that group were used as time-matchedcontrol values for the toxin-perfused vessels. At the lowesttoxin concentration (0.3 µg ml^–1) no increase inL_p was observed for up to 90 min (n = 7). At 10 µg ml^–1the L_p increased rapidly, on average to more than 15 times thecontrol value within 10 min, and was significantly greater thanthe time-matched control group at all times (P < 0.05, Mann-Whitney).At intermediate concentrations the response to epsilon-toxinexposure took longer to develop and the mean values of L_p didnot reach as high. The threshold concentration for a measurableL_p response within 60 min was 1 µg ml^–1 (significantlygreater than time-matched control group at 50 min and longer).In two experiments, after the L_p had significantly increasedduring epsilon-toxin exposure, each vessel was recannulatedwith the control perfusate and L_p measurements were made for20 min to test for recovery from the epsilon-toxin. The L_p inthese vessels did not return toward control; the effect of epsilon-toxinto increase L_p was irreversible.

View larger version (28K):
[in this window]
[in a new window]
FIG. 3. (A) Dose-response relationship between L_p and toxin concentration determined in four different groups of venular microvessels. No response was found with up to 90 min of perfusion at 0.3 µg ml^–1 toxin (lowest dose tested). At 10 µg ml^–1 toxin some vessels responded within the first minute. The threshold concentration to measurably increase L_p within 60 to 90 min was 1 µg ml^–1. *, for both 10 and 3 µg ml^–1 the L_p values were greater than time-matched controls at all times (P < 0.05, Mann-Whitney test); ${dagger}$ , for 1 µg ml^–1 the L_p was greater than control at 50 min and longer (P < 0.05, Mann-Whitney test); ${ddagger}$ , the group treated with 0.3 µg ml^–1 was not different from control at any time (P > 0.05, Mann-Whitney test). (B) Average L_p in vessels perfused without toxin shows no change over 90 min. Error bars are smaller than symbols and do not appear for this group. Note the different vertical scale. The number of vessels in each group (mean ± SEM) is shown.

Antibody neutralization of the toxin effect on L_p. A separate group of vessels were perfused with solution containinga neutralizing monoclonal antibody (0.13 mg ml^–1) to epsilon-toxinin addition to epsilon-toxin (3 µg ml^–1). In perfusionslasting up to 40 min, the L_p in these vessels did not increase.L_p measured over time is shown in comparison with the mean responseto the same concentration of toxin alone in Fig. 4. Over thistime the antibody completely prevented the increase in L_p inducedby toxin in the absence of antibody.

View larger version (15K):
[in this window]
[in a new window]
FIG. 4. Demonstration of toxin neutralization with monoclonal antibody 5B7. In vessels perfused with toxin additionally treated with neutralizing antibody there was no change in L_p, showing that the L_p response is dependent on the toxin. The number of vessels in each group (mean ± SEM) is shown.

Ultrastructure of microvessels exposed to toxin. Two vessels of the group perfused with 3 µg ml^–1epsilon-toxin and two of the group perfused with 10 µgml^–1 epsilon-toxin were examined with TEM. The two vesselsexposed to 3 µg ml^–1 toxin (25 and 40 min, respectively)reached an L_p about 5 times the control value just before fixation;in the two vessels exposed to 10 µg ml^–1 epsilon-toxin(25 and 35 min, respectively), L_p had increased to 13 and 24times the control L_p when fixed. Features consistent with directendothelial damage were observed in all epsilon-toxin-treatedvessels (Fig. 5). Most endothelial cells were very attenuated,showing loss of cytoplasm, although the cytoplasmic membraneappeared intact in most cases (Fig. 5A, B, C, and G). Differentdegrees of detachment from the basement membrane were also observed.A few cells were also fragmented, with some of the fragmentsremaining attached to the basement membrane. Thus, in placesthe endothelium was discontinuous, exhibiting gaps either betweenor through the endothelial cells (Fig. 5A, B, D, and E). Nearone such gap the subendothelial space between the basement membraneand pericyte was expanded and filled with flocculent material,possibly protein from disrupted cells (Fig. 5A and B). Sectionsof the basement membrane were exposed to the vascular lumen,although the basement membrane in these areas and elsewhereappeared intact. Some of the pericytes were also densely stainedand highly attenuated (Fig. 5A and B). Most endothelial cellshad angular shapes, and the cytoplasmic membrane presented anirregular outline, with numerous indentations and a few cytoplasmicblebs consisting of flocculent relatively electron-lucent materialdevoid of organelles and surrounded by cytoplasmic membrane.Most endothelial cells showed a diffusely electron-dense cytoplasmin which it was impossible to distinguish any cytoplasmic detail.A few endothelial cells, however, showed a clear cytoplasm inwhich dilated cisterns of endoplasmic reticulum were observed.Very few ribosomes were seen attached to the endoplasmic reticulummembranes. A few membrane-bound vacuoles with myelinic figuresin front of a clear background were also seen in the cytoplasmof some cells. The nuclei of some endothelial cells were shrunkenand showed large clumps of electron-dense chromatin. A few cellsshowed diffuse dense chromatin occupying the entire nucleus.Nuclear membranes were unaffected in most cases. Endothelialcells of control vessels perfused for comparable times withvehicle solutions in the absence of toxin did not show ultrastructuralabnormalities, such as gaps in the endothelium or necrotic characteristics(Fig. 6). Specifically, the endothelium was in contact withbasement membrane and formed a continuous intimal layer withintact junctional complexes. The peripheral cytoplasm was ofmoderate electron density and was relatively uniform in thickness(0.1 to 0.2 µm).

View larger version (152K):
[in this window]
[in a new window]
FIG. 5. TEM images of vessels perfused with epsilon-toxin, illustrating a range of damage to endothelial cells and pericytes. (A) Low-magnification overview shows gap in endothelium, enlarged vacuole (vac), and high-density nuclear chromatin in an endothelial cell (nuc). Pericyte (peri) is separated from basement membrane, and the space is filled with flocculent material (floc). (B) Detail of panel A shows irregular endothelial surface (is) with regions separated from basement membrane (bm). Extreme density of pericytes also suggests intoxication. Endothelial junctional complex (jc) appears intact. (C) Highly attenuated and electron-dense endothelial cell (ec1) contrasts with relatively normal appearance of second endothelial cell (ec2) and nearby mesothelial cell (mes). Collagen (col) and elastic fibers (el) appear normal. (D) Large region over which endothelium is discontinuous, with only cell debris remaining loosely adherent to exposed basement membrane. (E) Exposed basement membrane and nearby endothelium that appears irregularly bulbous and lacking typical fine structure. (F) Swollen segment of endothelial cell (luc) with vesicular structures on an unusually electron lucent background. Nearby junctional complexes appear intact. (G) Segment of an endothelial cell with two regions empty of cytoplasm such that the luminal cytoplasmic membrane has collapsed onto the abluminal membrane (arrows). er, endoplasmic reticulum. Inset at higher magnification clearly shows double membrane layers (small arrows) and an apparently intact endothelial tight junction (tj).

View larger version (126K):
[in this window]
[in a new window]
FIG. 6. TEM images of vessels perfused with control solutions illustrate endothelial appearance in the absence of epsilon-toxin exposure. (A) Luminal endothelial surface is smooth, and the nuclear chromatin (nuc) is moderately electron dense. (B) Detail of panel A shows endothelial cells are adherent to basement membrane (bm), and a pericyte (peri) with moderated electron density is seen within multiple layers of basement membrane closely associated with the endothelial cell. pv, plasmalemmal vesicle. (C) One further image shows the fine structure of an endothelial cell, including an intact junctional complex (jc), numerous vesicles, and a nucleus. The process of a fibroblast (fib) is present in the interstitial matrix.

DISCUSSION

Top

Discussion

We set out to test the specific hypothesis that epsilon-toxinwould increase the hydraulic permeability of microvessels inthe rat mesentery and to answer the more general question ofwhether the single perfused vessel model could be used to investigateepsilon-toxin. The data demonstrated a time- and dose-dependentirreversible increase in convective water flux across the vesselwalls in response to exposure to activated epsilon-toxin. Manyendothelial cells from toxin-treated vessels appeared necrotic,highly attenuated, and with dense cytoplasm, indicative of extremefluid loss associated with necrosis. Patent pathways throughthe endothelial layer were seen associated with fluid accumulationin the subendothelial spaces. Such pathways are never seen incontrol endothelium and represent a very high conductance pathfor both plasma water and protein. The results are consistentwith the hypothesis that epsilon-toxin interacts with endothelialcells of rat venular microvessels and increases the vessel wallpermeability by direct damage of the endothelium. The answerto the broad question of whether the rat mesentery model mightbe useful to study endothelial interaction with blood-bornetoxins such as epsilon-toxin is clear. The single perfused microvesselmodel does enable investigation of relatively small quantitiesof toxin with mammalian microvessel endothelium.

While this study is the first to measure biophysical permeabilitycoefficients in response to epsilon-toxin, there are numerousprevious reports indicating fluid and protein extravasationresulting from epsilon-toxin toxemia. Edema and excess proteinextravasation in various tissues have been noted at necropsyin experimentally and naturally intoxicated animals (12, 14,30, 38, 39, 43). Morphological and histochemical methods provideendpoint observations but do not enable finer temporal characterizationof the progress of intoxication. Moreover, because the toxinhas been shown to stimulate vascular smooth muscle contractionand elevate arterial pressure, it is very difficult to separatechanges in vessel wall permeability, tissue perfusion, and intravascularpressure in whole-animal models (28, 36). Under the conditionsof our experiments, the intravascular hydraulic pressure andprotein osmotic pressure are directly controlled, while extravascularhydraulic pressure and protein osmotic pressure are negligible(2, 20). The postcapillary microvessels used in this study arenot invested with smooth muscle cells and maintain a constantdiameter under the conditions of these experiments. Individualmeasurements in the single perfused vessel, taking between 5and 8 seconds, enable characterization of permeability changeswith temporal resolution comparable to cell culture methodsused to measure transepithelial resistance, small solute flux,or hydraulic conductivity (18, 34, 41).

The single vessel technique was used here to measure an increasein water flux crossing the microvessel wall resulting from epsilon-toxinexposure and was also used to infer loss of the barrier to serumalbumin (decrease of effective protein osmotic pressure shownin Fig. 1). Such changes in each of these parameters are directlyimportant in the development of edema. However, one cannot specifywhat level of L_p measured in a single vessel would be correlatedwith organ failure or death. We can say that the ultrastructuralexamination of all vessels perfused with epsilon-toxin revealedendothelial cells that were severely damaged and regions ofbasement membrane exposed directly to the perfusate. These endothelialgaps were not found in control vessels. The latter control dataagree with results of a much larger examination of normal endotheliumin a recent publication from our group in which nearly 700 micrographswere analyzed for tight junction structure using rat mesenterymicrovessels that were perfused for 15 to 30 min with vehiclesolutions similar to that in the present study (2). We alsonoted that once the L_p began to increase, removing perfusatecontaining toxin and reverting to control perfusate failed toreturn the L_p toward control. Indeed, in many instances whenthe perfusing pipette was removed at the end of an experimentand blood was allowed to return to the vessel, we observed hemostasisin vessels that had been perfused with toxin; vessels not perfusedwith toxin almost always resumed normal blood flow.

We also note that the effects of epsilon-toxin were not similarto an acute inflammatory response, such as that stimulated byplatelet activating factor or other acute inflammatory mediators.The typical response of mesenteric postcapillary venules toplatelet activating factor is an increase to a peak value within10 to 15 min followed by a recovery phase lasting another 10to 60 min (3). Similar patterns have been described for theinflammatory mediators histamine (21) and bradykinin (15) inthese vessels. Because the epsilon-toxin effect exhibited acontinuous L_p increase and did not return toward control, itdid not have the characteristic pattern associated with an acuteinflammatory response.

Previous studies of epsilon-toxin using MDCK cells have examinedparameters different from the water flux and loss of proteinosmotic pressure measured in the present study. Loss of intracellularK⁺ (29, 33, 35), decreased cell viability (27, 29, 33-35), increasedentry of propidium iodide (34, 35), and a fall in trans-epithelialresistance (34) all have been used to infer epsilon-toxin toxicitytoward MDCK cell monolayers. None of these measures are directlycomparable to increased L_p in microvessels. However, all thesemeasures are indicative of damaged cell membrane and were correlatedwith compromised epithelial monolayer barrier function. TheMDCK cell model is particularly sensitive to epsilon-toxin,and the concentrations of epsilon-toxin that elicited effectsin the cited MDCK studies ranged from near 10^–10 M to10^–8 M. In the in vivo mesentery microvessel, we founda threshold effect at 1 µg ml^–1 (approximately 3x 10^–8 M) when measuring increases in L_p. Endothelialgaps seen with TEM and correlated with increased L_p could haveformed as endothelial tight and adherens junctions releasedor could have resulted from cell disruption. Some clostridialtoxins inactivate small GTPases, leading to the breakdown ofadherens and tight junctions in both epithelial and endothelialcells (1, 5, 42). In a recent study using MDCK monolayers, ata time and dose of epsilon-toxin that elicited a complete lossof trans-epithelial resistance, the monolayer permeability toa solute (dextran 38500) restricted to the paracellular pathwaywas not increased and the immunofluorescent distribution ofperipheral band actin and components of the tight junction complexwas not different from control (34). These data indicated thatdirect cell disruption by pore formation on the apical membranepreceded any possible direct effect on the junctional complex.Thus, it is likely that the broad areas of basement membraneexposed during epsilon-toxin treatment of the mesentery microvesselsresult from the disruption of the endothelial cells rather thanfrom disassembly of cell-cell junctions.

The present results using in vivo endothelium and epsilon-toxinconcentrations up to 10 µg ml^–1 contrast with thoseof our previous study, in which the toxin at up to 120 µgml^–1 did not induce morphological changes in culturedaortic endothelial cells (tested from cattle, goats, and sheep)(40). One possibility for this difference was that activityof the toxin toward endothelium was dependent on a factor inblood not present in the culture medium. In our previous studywe addressed that possibility by testing for toxicity againstthe endothelial cells in the presence as well as the absenceof serum proteins, but we found that in either case there wasno morphological change induced by the toxin. While most ofthe present experiments were carried out in physiological saline(Ringer's) solution supplemented with bovine serum albumin,in several experiments in which toxin was diluted into Ringer'ssolution without BSA, epsilon-toxin also induced large increasesin L_p (not shown). Therefore, our results indicate that no othercomponent of circulating blood is required for toxicity againstthe mesenteric venular microvessel endothelium.

A specific receptor for epsilon-toxin has not been identified.However, the presence of a receptor, particularly in brain andkidney, has been suggested by uptake of labeled toxin (6, 30).Also, in vitro studies showed that treatment of rat brain homogenateswith neuraminidase or lipase reduced binding of epsilon-toxin,indicating that a receptor could be a sialoglycoprotein withina specific lipid environment (31). Following these earlier studies,recent work demonstrated that heptamers of epsilon-toxin werefound exclusively associated with the detergent-resistant membranefraction from MDCK cells, further demonstrating that epsilon-toxinbinding is dependent on specific cholesterol-containing microdomains,possibly lipid rafts or caveolae (27). Therefore, it is possiblethat the aortic endothelial cells which showed no morphologicalresponse to epsilon-toxin did not express the appropriate receptoror associated microdomain. We cannot say at present whetherthose cells express such a domain in vivo or could do so underdifferent culture conditions. The demonstration in the presentexperiments of the sensitivity of postcapillary microvesselendothelium to epsilon-toxin suggests that cultured endothelialmonolayers from other sources or under different culture conditionsshould be tested and that an appropriate endothelial culturemodel could be found.

In our experiments we also noted damage to pericytes of thetoxin-perfused microvessels. While some studies have suggestedthat pericytes promote vascular permeability through contractility,recent studies have focused on the role of pericytes in thematuration of microvessels (17). Loss of pericytes, in additionto endothelial cells, due to epsilon-toxin could potentiallyexacerbate local permeability increases and delay return tonormal vascular function in cases of sublethal intoxication.

The present study uses the mesenteric microvessel as a modelfor the direct vascular effects of epsilon-toxin. However, theprincipal lethal pathology in sheep is associated with brainand lungs. The traditionally accepted model to explain brainlesions due to epsilon-toxin (6) suggests that such lesionsare due to the action of this toxin on endothelial cells, whichgenerates perivascular edema followed by hypoxia and necrosisof the perivascular tissues. However, a recent study showedthat epsilon-toxin can also leak into the brain and act directlyon rat neurons, producing degeneration and necrosis of thesecells in cases of subacute or chronic epsilon-toxin intoxication(11). The results of the present experiments confirm that directdamage does occur in endothelial cells exposed to epsilon-toxin.However, it is possible that epsilon-toxin acts both on endothelialcells and neurons alike. Moreover, it is possible and as yetunknown whether differences exist between various vascular bedsin regard to toxin sensitivity and pathogenesis.

Summary. We have shown that epsilon-toxin, in a time- and dose-dependentmanner, rapidly and irreversibly compromises the barrier functionof venular microvessel endothelium in the rat mesentery. Thetoxin appeared to directly damage vascular endothelium, inducingformation of gaps in the endothelial cell layer correlated withloss of effective protein osmotic pressure and large increasesin measured L_p. These changes are consistent with those expectedto lead to rapid edema formation and would also promote accessof the toxin to cells in the surrounding tissue. The singleperfused vessel model will enable further investigation of epsilon-toxininteraction with the vascular endothelium in situ.

ACKNOWLEDGMENTS

This work was supported by Public Health Service grants fromthe National Heart, Lung, and Blood Institute (HL28607 and HL44485)and National Institute of Allergy and Infectious Diseases (AI56177).

We gratefully acknowledge the expert technical work of JoyceLenz.

FOOTNOTES

* Corresponding author. Mailing address: Department of Physiology and Membrane Biology, University of California, Davis, One Shields Avenue, Davis, CA 95616. Phone: (530) 752-2180. Fax: (530) 752-5423. E-mail: rhadamson{at}ucdavis.edu.

Editor: D. L. Burns

REFERENCES

Top